Phase 2d (ii): filesystem modification time (mtime)

Completes Phase 2: the FAT filesystem now stamps and reports a real modification
time, built on the Phase 2d(i) kernel wall-clock. This is the last stat field the
compiler's build cache needs to reason about (source vs cached output).

- on-disk.zig: fatToEpoch / epochToFatDateTime convert between the two 16-bit DOS
  date/time fields and Unix epoch seconds (UTC — FAT has no timezone). Host-tested
  round-trip + an absolute check (1577836800 == 2020-01-01).
- engine: a settable current_time_epoch that create/write stamp into the entry's
  write (and creation) date/time; Node/Listing gained an mtime decoded from those
  fields on read. Host test: a create stamps the mtime, read back through resolve
  and listEntry.
- vfs protocol FileStatus + runtime.fs.Attributes gained an mtime field; the fat
  server sets current_time_epoch from runtime.system.wallClock() per request and
  returns mtime from stat. The flat ramfs reports 0 (it has no timestamps).
- fat-test reads the created file's mtime through stat and checks it is a real
  current time, behind a new `fat-mtime` QEMU case.

Verified against the host: the guest stamped mtime 1783971676 while the host clock
was 1783971680 (boot+test lag) — the file's mtime is real current time. zig build,
zig build test (the epoch<->DOS conversions + the engine mtime test),
zig build check-fat-image, and a sequential QEMU sweep — fat-mount, fat-mutations,
fat-rename, fat-mtime, vfs, vfs-client-death, log-flush, orderly-shutdown,
initial-ramdisk, smoke, wall-clock, usb-storage — all green. mode/inode remain.

.editorconfig

@@ -0,0 +1,16 @@
+# EditorConfig: https://editorconfig.org/
+# Follows the Zig style guide: https://ziglang.org/documentation/0.16.0/#Style-Guide
+
+root = true
+
+[*]
+charset = utf-8
+end_of_line = lf
+indent_style = space
+indent_size = 4
+trim_trailing_whitespace = true
+insert_final_newline = true
+
+[*.zig]
+# "Line length: aim for 100; use common sense."
+max_line_length = 100

.gitattributes

@@ -0,0 +1 @@
+*.zig text eol=lf

README.md

@@ -2,12 +2,31 @@
 Codename: Shodan
 Version: 1
 
-A small operating system, written from scratch in Zig — a bootloader (`boot/`)
-and a microkernel (`system/kernel/`), sharing a neutral handoff contract (`system/boot-handoff.zig`).
-It boots x86-64 via UEFI, and so far has a framebuffer console, a physical frame
-allocator, its own paging with W^X permissions, interrupt/exception handling, a
-LAPIC timer, a kernel heap, a fixed-priority preemptive scheduler, and in-kernel IPC
-channels. See [`docs/`](docs/README.md) for how each piece works.
+A small resilient operating system, written from scratch in Zig.
+
+## Zen of DanOS:
+
+- Resilient Micro-Kernel Architecture.
+    - Every process run in an isolated user space not kernel space.
+    - Processes cannot take down the entire OS with it when they die or is killed
+- Stable public runtime library, private OS ABI.
+    - Keeps a stable runtime for user space processes between OS versions (great for backwards compatibility)
+    - Allows the underlying OS to be changed without effecting applications
+    - Provides a boundary to enable compatibility between OS's e.g. POSIX, MUSL etc
+- Drivers are just isolated processes in user space.
+    - Thin binaries that can be restarted like applications.
+    - Useful during driver development.
+    - Drivers can claim MMIO / ports
+    - Driver resources (e.g. IRQ/Port/MMIO) claims are automatically cleaned up if the driver dies or is killed
+    - Drivers can also hook into the process lifecyle to clean up or reset hardware
+- No legacy to deal with
+    - Zig code uses a clean coding style (Zen of Zig)
+    - Favor reading code over writing code.
+    - No magic numbers.
+    - No shortend names unless its for ABI compatibility or acronyms
+- Inter-Process Communication (IPC)
+  - Publish and subscribe to Asynchronous Messages
+  - Talk to services and processes synchronously
 
 ## Prerequisites
 
@@ -60,9 +79,13 @@ straight into CI.
 
 ## Documentation
 
-Design notes explaining the *why* behind the code live in
+Design notes explaining *why* behind the code live in
 [`docs/`](docs/README.md) — start with [`docs/README.md`](docs/README.md).
 
+For the hardware needed to run DanOS — minimum specs plus a plain-language guide
+matching Intel/AMD CPU generations by name — see
+[`docs/system-requirements.md`](docs/system-requirements.md).
+
 ## Logo
 
 San Serif Text "Dan OS" with a black karate belt around it.

boot/efi.zig

@@ -29,7 +29,7 @@ pub fn main() uefi.Status {
     // report the reason (boot services are still up) and park the machine so the
     // message stays on screen.
     boot() catch |err| {
-        log("\r\ndanos: boot failed: ");
+        log("\r\nEFI: boot failed: ");
         logBytes(@errorName(err));
         log("\r\n");
         while (true) asm volatile ("hlt");
@@ -65,14 +65,14 @@ fn boot() !noreturn {
 
     // Best effort: a volume without /system/services/init still boots (kernel-only).
     loadInit(bs, &boot_information) catch |err| {
-        log("danos: no /system/services/init (");
+        log("EFI: no /system/services/init (");
         logBytes(@errorName(err));
         log(") - booting without user space\r\n");
     };
 
     // Best effort: the initial_ramdisk (VFS server + drivers) is optional too.
     loadInitialRamdisk(bs, &boot_information) catch |err| {
-        log("danos: no initial_ramdisk (");
+        log("EFI: no initial_ramdisk (");
         logBytes(@errorName(err));
         log(")\r\n");
     };
@@ -84,7 +84,7 @@ fn boot() !noreturn {
     // the map and exiting would invalidate the map key.
     const cr3 = try buildBootstrapTables(bs, &boot_information);
 
-    log("danos: kernel loaded, exiting boot services\r\n");
+    log("EFI: kernel loaded, exiting boot services\r\n");
     boot_information.memory_map = try exitBootServices(bs);
 
     // Switch onto our tables and jump to the kernel in one uninterruptible step.
@@ -395,7 +395,7 @@ fn loadInit(bs: *uefi.tables.BootServices, boot_information: *BootInformation) !
     const image = try loadFile(bs, init_file_name);
     boot_information.init_base = @intFromPtr(image.ptr);
     boot_information.init_len = image.len;
-    log("danos: /system/services/init loaded\r\n");
+    log("EFI: /system/services/init loaded\r\n");
 }
 
 /// Ferry the initial_ramdisk (the VFS server + drivers) to the kernel, same as init.
@@ -403,7 +403,7 @@ fn loadInitialRamdisk(bs: *uefi.tables.BootServices, boot_information: *BootInfo
     const image = try loadFile(bs, initial_ramdisk_file_name);
     boot_information.initial_ramdisk_base = @intFromPtr(image.ptr);
     boot_information.initial_ramdisk_len = image.len;
-    log("danos: initial_ramdisk loaded\r\n");
+    log("EFI: initial_ramdisk loaded\r\n");
 }
 
 /// Validate the ELF, copy every PT_LOAD segment to its physical address, and

build.zig

@@ -58,8 +58,9 @@ fn addUserBinary(
     b: *std.Build,
     target: std.Build.ResolvedTarget,
     runtime_module: *std.Build.Module,
-    posix_module: *std.Build.Module,
     mmio_module: *std.Build.Module,
+    xkeyboard_config_module: *std.Build.Module,
+    acpi_ids_module: *std.Build.Module,
     name: []const u8,
     root: []const u8,
 ) *std.Build.Step.Compile {
@@ -76,11 +77,14 @@ fn addUserBinary(
             .stack_protector = false,
             .imports = &.{
                 .{ .name = "runtime", .module = runtime_module },
-                // POSIX/C compatibility layer, available to any program that wants it
-                // (danos-native code uses `runtime` directly). See library/posix/.
-                .{ .name = "posix", .module = posix_module },
                 // Typed volatile MMIO + memory barriers, for drivers. See library/mmio/.
                 .{ .name = "mmio", .module = mmio_module },
+                // Keyboard layouts (keycode + modifiers -> keysym/character), available
+                // to any program that wants it. See library/xkeyboard-config/.
+                .{ .name = "xkeyboard-config", .module = xkeyboard_config_module },
+                // ACPI/PnP hardware-ID registry, so drivers name devices
+                // (HardwareId.ps2_keyboard) instead of magic "_HID" strings.
+                .{ .name = "acpi-ids", .module = acpi_ids_module },
             },
         }),
     });
@@ -123,10 +127,39 @@ pub fn build(b: *std.Build) void {
     });
     // ACPI/PnP hardware-ID (_HID) names — the flat analog of pci-class for acpi_device
     // nodes. Also shared reference data.
+    // The AML interpreter, a build module so the ring-3 acpi service can run the
+    // same parser the kernel does (docs/discovery.md — the shared AML module).
+    // Pure Zig, no kernel imports — one source, two builds.
+    const aml_module = b.addModule("aml", .{
+        .root_source_file = b.path("system/devices/aml/aml.zig"),
+    });
+
     const acpi_ids_module = b.addModule("acpi-ids", .{
         .root_source_file = b.path("system/devices/acpi-ids.zig"),
     });
 
+    // The USB device-framework wire ABI (chapter-9 set-up packets, standard +
+    // class requests, descriptors) and the USB class-code taxonomy — the flat
+    // reference the xHCI bus driver, the USB class drivers, and the device
+    // manager's identity matcher all share. Pure data, like pci-class/acpi-ids.
+    const usb_abi_module = b.addModule("usb-abi", .{
+        .root_source_file = b.path("system/devices/usb-abi.zig"),
+    });
+    const usb_ids_module = b.addModule("usb-ids", .{
+        .root_source_file = b.path("system/devices/usb-ids.zig"),
+    });
+    // The USB transfer protocol: what a USB class driver says to the xHCI bus
+    // driver to drive its device (open / control / interrupt / bulk). A protocol
+    // module like vfs-protocol, shared by the bus driver and every class driver.
+    const usb_transfer_protocol_module = b.addModule("usb-transfer-protocol", .{
+        .root_source_file = b.path("system/drivers/usb-xhci-bus/usb-transfer-protocol.zig"),
+    });
+    // The block-device protocol: read/write of fixed-size blocks, spoken between a
+    // filesystem and a block driver (usb-storage). A protocol module like the rest.
+    const block_protocol_module = b.addModule("block-protocol", .{
+        .root_source_file = b.path("system/services/block/protocol.zig"),
+    });
+
     // Kernel tunables (maximum_cpus, stack sizes, tick rate). A dependency-free module of
     // compile-time constants, imported wherever a knob is read; keeps the trade-offs
     // in one place instead of scattered across the tree. See system/parameters.zig.
@@ -178,6 +211,13 @@ pub fn build(b: *std.Build) void {
         .root_source_file = b.path("system/services/vfs/protocol.zig"),
     });
 
+    // The input wire protocol: the input service's public interface, exposed as its own
+    // module the same way vfs-protocol is. Shared by the input service, the runtime's
+    // `input` helper (subscribe/publish), and every source and subscriber.
+    const input_protocol_module = b.addModule("input-protocol", .{
+        .root_source_file = b.path("system/services/input/protocol.zig"),
+    });
+
     // The danos-native user-space runtime: system_call wrappers, the C-convention
     // heap, IPC helpers, the process start shim, device access. This is the stable
     // application ABI; POSIX compatibility is a separate library on top (see below).
@@ -193,9 +233,28 @@ pub fn build(b: *std.Build) void {
             .{ .name = "abi", .module = abi_module },
             .{ .name = "device-abi", .module = device_abi_module },
             .{ .name = "vfs-protocol", .module = vfs_protocol_module },
+            .{ .name = "input-protocol", .module = input_protocol_module },
         },
     });
 
+    // The device-manager protocol: hello + (M18.2) tree reports, exposed as its
+    // own module like the other protocol modules. Imported through the runtime.
+    const device_manager_protocol_module = b.addModule("device-manager-protocol", .{
+        .root_source_file = b.path("system/services/device-manager/device-manager-protocol.zig"),
+    });
+    runtime_module.addImport("device-manager-protocol", device_manager_protocol_module);
+    // The USB transfer protocol, so runtime.usb (the class-driver client) can speak
+    // it, the way runtime.input speaks the input protocol.
+    runtime_module.addImport("usb-transfer-protocol", usb_transfer_protocol_module);
+    // The block protocol, so runtime.block (the block-device client) can speak it.
+    runtime_module.addImport("block-protocol", block_protocol_module);
+
+    // The power protocol: system power's domain-named surface (docs/power.md).
+    const power_protocol_module = b.addModule("power-protocol", .{
+        .root_source_file = b.path("system/services/power/protocol.zig"),
+    });
+    runtime_module.addImport("power-protocol", power_protocol_module);
+
     // Typed volatile MMIO register access + memory-ordering barriers, for drivers on
     // top of an mmio_map grant. Depends only on `builtin` (arch-conditional barriers);
     // no target set, so it inherits each driver's. See library/mmio/mmio.zig.
@@ -203,15 +262,17 @@ pub fn build(b: *std.Build) void {
         .root_source_file = b.path("library/mmio/mmio.zig"),
     });
 
-    // The POSIX / C compatibility layer, a separate library layered strictly over the
-    // runtime (it calls the runtime's IPC/heap, never system calls directly). This is
-    // the one place POSIX/C spellings are allowed verbatim — see docs/coding-standards.md
-    // and library/posix/posix.zig.
-    const posix_module = b.addModule("posix", .{
-        .root_source_file = b.path("library/posix/posix.zig"),
+    // Keyboard layouts compiled from the X11 xkeyboard-config database into native Zig
+    // (keycode + modifiers -> keysym/character). The `layouts` tables are generated by
+    // tools/make-xkeyboard-config.py; `xkeyboard-config` is the hand-written API over them.
+    // No target set, so each inherits its importer's. See library/xkeyboard-config/.
+    const xkb_layouts_module = b.addModule("layouts", .{
+        .root_source_file = b.path("library/xkeyboard-config/generated/layouts.zig"),
+    });
+    const xkeyboard_config_module = b.addModule("xkeyboard-config", .{
+        .root_source_file = b.path("library/xkeyboard-config/xkeyboard-config.zig"),
         .imports = &.{
-            .{ .name = "runtime", .module = runtime_module },
-            .{ .name = "vfs-protocol", .module = vfs_protocol_module },
+            .{ .name = "layouts", .module = xkb_layouts_module },
         },
     });
 
@@ -286,7 +347,7 @@ pub fn build(b: *std.Build) void {
     // Built by the shared user-binary recipe (see addUserBinary): freestanding,
     // linked into the kernel's user region against the `runtime` runtime library, and
     // started in ring 3 by the kernel's user-ELF loader.
-    const init_exe = addUserBinary(b, kernel_target, runtime_module, posix_module, mmio_module, "init", "system/services/init/init.zig");
+    const init_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "init", "system/services/init/init.zig");
     const init_install = b.addInstallArtifact(init_exe, .{ .dest_dir = .{ .override = .{ .custom = "system/services" } } });
     b.getInstallStep().dependOn(&init_install.step);
 
@@ -294,11 +355,70 @@ pub fn build(b: *std.Build) void {
     // Each is built by the same user-binary recipe, then packed into one image by
     // the host-side make-initial-ramdisk tool. The bootloader ferries the image to the kernel,
     // which unpacks it and spawns each program (system/initial-ramdisk.zig).
-    const vfs_exe = addUserBinary(b, kernel_target, runtime_module, posix_module, mmio_module, "vfs", "system/services/vfs/vfs.zig");
-    const vfstest_exe = addUserBinary(b, kernel_target, runtime_module, posix_module, mmio_module, "vfs-test", "system/services/vfs/vfs-test.zig");
-    const hpet_exe = addUserBinary(b, kernel_target, runtime_module, posix_module, mmio_module, "hpet", "system/drivers/hpet/hpet.zig");
-    const bus_exe = addUserBinary(b, kernel_target, runtime_module, posix_module, mmio_module, "bus", "system/drivers/bus/bus.zig");
-    const device_manager_exe = addUserBinary(b, kernel_target, runtime_module, posix_module, mmio_module, "device-manager", "system/services/device-manager/device-manager.zig");
+    const vfs_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "vfs", "system/services/vfs/vfs.zig");
+    const vfstest_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "vfs-test", "system/services/vfs/vfs-test.zig");
+    const ps2_bus_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "ps2-bus", "system/drivers/ps2-bus/ps2-bus.zig");
+    const ps2_keyboard_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "ps2-keyboard", "system/drivers/ps2-bus/keyboard.zig");
+    const ps2_mouse_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "ps2-mouse", "system/drivers/ps2-bus/mouse.zig");
+    const usb_xhci_bus_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "usb-xhci-bus", "system/drivers/usb-xhci-bus/usb-xhci-bus.zig");
+    // The xHCI bus driver builds chapter-9 requests and decodes descriptors from
+    // usb-abi, and reports each interface's (class,subclass,protocol) identity via
+    // usb-ids.packTriple.
+    usb_xhci_bus_exe.root_module.addImport("usb-abi", usb_abi_module);
+    usb_xhci_bus_exe.root_module.addImport("usb-ids", usb_ids_module);
+    usb_xhci_bus_exe.root_module.addImport("usb-transfer-protocol", usb_transfer_protocol_module);
+    // The USB HID class drivers: keyboard and mouse. They own no hardware — each
+    // opens its device through runtime.usb (the transfer protocol) and publishes to
+    // the input service. They build chapter-9 class requests from usb-abi.
+    const usb_hid_keyboard_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "usb-hid-keyboard", "system/drivers/usb-hid/keyboard.zig");
+    usb_hid_keyboard_exe.root_module.addImport("usb-abi", usb_abi_module);
+    const usb_hid_mouse_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "usb-hid-mouse", "system/drivers/usb-hid/mouse.zig");
+    usb_hid_mouse_exe.root_module.addImport("usb-abi", usb_abi_module);
+    // The USB mass-storage class driver: opens its device via runtime.usb, drives it
+    // with Bulk-Only Transport + SCSI, and serves the block protocol under `.block`.
+    const usb_storage_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "usb-storage", "system/drivers/usb-storage/usb-storage.zig");
+    usb_storage_exe.root_module.addImport("block-protocol", block_protocol_module);
+    // The FAT filesystem server: mounts the block device and serves it into the VFS
+    // at /mnt/usb. Its engine (engine.zig / on-disk.zig) is imported relatively.
+    const fat_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "fat", "system/services/fat/fat.zig");
+    const fat_test_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "fat-test", "system/services/fat/fat-test.zig");
+    const pci_bus_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "pci-bus", "system/drivers/pci-bus/pci-bus.zig");
+    // The PCI bus driver decodes each function's class triple to human names in its
+    // boot log (class/subclass/prog-IF), so pull in the shared pci-class reference.
+    pci_bus_exe.root_module.addImport("pci-class", pci_class_module);
+    // A test fixture, not a real driver: hellos to the device manager, then faults —
+    // what the driver-restart scenario drives the crash-loop cap with.
+    const crash_test_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "crash-test", "system/services/crash-test/crash-test.zig");
+    const device_list_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "device-list", "system/services/device-list/device-list.zig");
+    // The discovery service: one swappable process per firmware
+    // (docs/discovery.md), bundled under the neutral ramdisk name
+    // "discovery" so the device manager never learns which firmware it is on.
+    // x86 boots describe hardware with ACPI; the Raspberry Pis hand over a
+    // flattened device tree — the aarch64 target flips the default when it
+    // lands (docs/arm.md). Both are placeholders until M20.1 (acpi) and the
+    // ARM bring-up (fdt).
+    const Discovery = enum { acpi, fdt };
+    const discovery = b.option(Discovery, "discovery", "Which discovery service fills the ramdisk's 'discovery' slot (default: acpi)") orelse Discovery.acpi;
+    const discovery_source: []const u8 = switch (discovery) {
+        .acpi => "system/services/acpi/acpi.zig",
+        .fdt => "system/services/fdt/fdt.zig",
+    };
+    const discovery_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "discovery", discovery_source);
+    if (discovery == .acpi) discovery_exe.root_module.addImport("aml", aml_module);
+    const device_manager_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "device-manager", "system/services/device-manager/device-manager.zig");
+    // Names the xHCI PCI class triple from the shared taxonomy instead of a bare 0x0C0330.
+    device_manager_exe.root_module.addImport("pci-class", pci_class_module);
+    // The manager matches reported USB interfaces by their (class,subclass,protocol)
+    // triple (usbDriverForIdentity), built from the named usb-ids codes.
+    device_manager_exe.root_module.addImport("usb-ids", usb_ids_module);
+    // The input service and its exercisers: the fan-out server, a hardware-free synthetic
+    // source, and a subscriber that doubles as the `input` test's oracle. See docs/input.md.
+    const input_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "input", "system/services/input/input.zig");
+    const input_source_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "input-source", "system/services/input-source/input-source.zig");
+    const input_test_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "input-test", "system/services/input-test/input-test.zig");
+    const args_echo_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "args-echo", "system/services/args-echo/args-echo.zig");
+    const process_test_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "process-test", "system/services/process-test/process-test.zig");
+    const log_flush_exe = addUserBinary(b, kernel_target, runtime_module, mmio_module, xkeyboard_config_module, acpi_ids_module, "log-flush", "system/services/log-flush/log-flush.zig");
 
     // Pack the user binaries into the initial_ramdisk image with the host-side Python tool
     // (the container format is trivial, and Python sidesteps std API churn). Args:
@@ -310,20 +430,62 @@ pub fn build(b: *std.Build) void {
     mk_run.addFileArg(vfs_exe.getEmittedBin());
     mk_run.addArg("vfs-test");
     mk_run.addFileArg(vfstest_exe.getEmittedBin());
-    mk_run.addArg("hpet");
-    mk_run.addFileArg(hpet_exe.getEmittedBin());
-    mk_run.addArg("bus");
-    mk_run.addFileArg(bus_exe.getEmittedBin());
+    mk_run.addArg("ps2-bus");
+    mk_run.addFileArg(ps2_bus_exe.getEmittedBin());
+    mk_run.addArg("ps2-keyboard");
+    mk_run.addFileArg(ps2_keyboard_exe.getEmittedBin());
+    mk_run.addArg("ps2-mouse");
+    mk_run.addFileArg(ps2_mouse_exe.getEmittedBin());
+    mk_run.addArg("usb-xhci-bus");
+    mk_run.addFileArg(usb_xhci_bus_exe.getEmittedBin());
+    mk_run.addArg("usb-hid-keyboard");
+    mk_run.addFileArg(usb_hid_keyboard_exe.getEmittedBin());
+    mk_run.addArg("usb-hid-mouse");
+    mk_run.addFileArg(usb_hid_mouse_exe.getEmittedBin());
+    mk_run.addArg("usb-storage");
+    mk_run.addFileArg(usb_storage_exe.getEmittedBin());
+    mk_run.addArg("fat");
+    mk_run.addFileArg(fat_exe.getEmittedBin());
+    mk_run.addArg("fat-test");
+    mk_run.addFileArg(fat_test_exe.getEmittedBin());
+    mk_run.addArg("pci-bus");
+    mk_run.addFileArg(pci_bus_exe.getEmittedBin());
+    mk_run.addArg("crash-test");
+    mk_run.addFileArg(crash_test_exe.getEmittedBin());
+    mk_run.addArg("device-list");
+    mk_run.addFileArg(device_list_exe.getEmittedBin());
+    mk_run.addArg("discovery");
+    mk_run.addFileArg(discovery_exe.getEmittedBin());
     mk_run.addArg("device-manager");
     mk_run.addFileArg(device_manager_exe.getEmittedBin());
+    mk_run.addArg("input");
+    mk_run.addFileArg(input_exe.getEmittedBin());
+    mk_run.addArg("input-source");
+    mk_run.addFileArg(input_source_exe.getEmittedBin());
+    mk_run.addArg("input-test");
+    mk_run.addFileArg(input_test_exe.getEmittedBin());
+    mk_run.addArg("args-echo");
+    mk_run.addFileArg(args_echo_exe.getEmittedBin());
+    mk_run.addArg("process-test");
+    mk_run.addFileArg(process_test_exe.getEmittedBin());
+    mk_run.addArg("log-flush");
+    mk_run.addFileArg(log_flush_exe.getEmittedBin());
 
     // Also install the packed binaries to their FHS homes, so zig-out is a true image
     // of the filesystem — even though at boot they arrive inside the initial-ramdisk.
     for ([_]struct { *std.Build.Step.Compile, []const u8 }{
         .{ vfs_exe, "system/services" },
         .{ device_manager_exe, "system/services" },
-        .{ hpet_exe, "system/drivers" },
-        .{ bus_exe, "system/drivers" },
+        .{ input_exe, "system/services" },
+        .{ ps2_bus_exe, "system/drivers" },
+        .{ ps2_keyboard_exe, "system/drivers" },
+        .{ ps2_mouse_exe, "system/drivers" },
+        .{ usb_xhci_bus_exe, "system/drivers" },
+        .{ usb_hid_keyboard_exe, "system/drivers" },
+        .{ usb_hid_mouse_exe, "system/drivers" },
+        .{ usb_storage_exe, "system/drivers" },
+        .{ fat_exe, "system/services" },
+        .{ log_flush_exe, "system/services" },
     }) |entry| {
         const step = b.addInstallArtifact(entry[0], .{ .dest_dir = .{ .override = .{ .custom = entry[1] } } });
         b.getInstallStep().dependOn(&step.step);
@@ -357,6 +519,36 @@ pub fn build(b: *std.Build) void {
     const efi_install = b.addInstallArtifact(efiexe, .{ .dest_dir = .{ .override = .{ .custom = "EFI/BOOT" } } });
     b.getInstallStep().dependOn(&efi_install.step);
 
+    // --- danos-usb.img: the bootable FAT32 USB image ---
+    // Format a real FAT32 image (the in-repo Python builder, no external tools)
+    // holding exactly what the firmware and bootloader need off the ESP: the EFI
+    // stub, the kernel, init, and the initial-ramdisk. QEMU presents this image as
+    // a USB mass-storage device the guest boots from (see run-x86-64 and the test
+    // harness), and the danos fat driver mounts the same image at /mnt/usb.
+    const mk_fat = b.addSystemCommand(&.{"python3"});
+    mk_fat.addFileArg(b.path("tools/make-fat-image.py"));
+    const fat_image = mk_fat.addOutputFileArg("danos-usb.img");
+    mk_fat.addArg("64"); // MiB
+    mk_fat.addArg("EFI/BOOT/BOOTX64.efi");
+    mk_fat.addFileArg(efiexe.getEmittedBin());
+    mk_fat.addArg("system/kernel");
+    mk_fat.addFileArg(exe.getEmittedBin());
+    mk_fat.addArg("system/services/init");
+    mk_fat.addFileArg(init_exe.getEmittedBin());
+    mk_fat.addArg("boot/initial-ramdisk.img");
+    mk_fat.addFileArg(initial_ramdisk_img);
+    const fat_image_install = b.addInstallFile(fat_image, "danos-usb.img");
+    b.getInstallStep().dependOn(&fat_image_install.step);
+
+    // `zig build check-fat-image` — validate the produced image is a real FAT32
+    // with the EFI stub present (the builder's own --verify, no external tools).
+    const check_fat = b.addSystemCommand(&.{"python3"});
+    check_fat.addFileArg(b.path("tools/make-fat-image.py"));
+    check_fat.addArg("--verify");
+    check_fat.addFileArg(fat_image);
+    const check_fat_step = b.step("check-fat-image", "Verify the FAT32 USB image is valid and bootable");
+    check_fat_step.dependOn(&check_fat.step);
+
     // --- run-x86-64: boot the x86-64 kernel in QEMU via UEFI/OVMF ---
     // Firmware lives in different places per OS/distro, so probe the known
     // layouts (Architecture, Debian/Ubuntu, Fedora, macOS Homebrew) and use the first
@@ -395,6 +587,19 @@ pub fn build(b: *std.Build) void {
 
     const run_efi = b.addSystemCommand(&.{
         "qemu-system-x86_64",
+        "-device",
+        "qemu-xhci,id=xhci",
+        "-device",
+        "usb-mouse,bus=xhci.0",
+        "-device",
+        "usb-kbd,bus=xhci.0",
+        // "-usb",
+        // "-device",
+        // "usb-ehci,id=ehci",
+        // "-device",
+        // "usb-tablet,bus=usb-bus.0",
+        // "-device",
+        // "usb-mouse,bus=ehci.0",
         "-machine",
         "q35",
         "-m",
@@ -404,10 +609,13 @@ pub fn build(b: *std.Build) void {
     });
     run_efi.addArg("-drive");
     run_efi.addPrefixedFileArg("if=pflash,format=raw,file=", vars_out);
-    // Present the FHS zig-out to the guest as a FAT drive — it is the boot volume.
+    // Boot off the FAT32 USB image: a mass-storage device on the same xHCI bus as
+    // the keyboard and mouse. OVMF finds \EFI\BOOT\BOOTX64.efi on it and boots.
+    run_efi.addArg("-drive");
+    run_efi.addPrefixedFileArg("if=none,id=bootusb,format=raw,file=", fat_image);
     run_efi.addArgs(&.{
-        "-drive",
-        b.fmt("format=raw,file=fat:rw:{s}", .{b.install_path}),
+        "-device",
+        "usb-storage,bus=xhci.0,drive=bootusb,removable=on,bootindex=0",
         "-net",
         "none",
         // Emulated display advertising 1280x720 as its native (EDID preferred)
@@ -453,7 +661,19 @@ pub fn build(b: *std.Build) void {
         "system/devices/device-abi.zig",
         "system/devices/pci-class.zig", // class/subclass/prog-IF name decoding
         "system/devices/acpi-ids.zig", // _HID name decoding
+        "system/devices/aml/aml.zig", // AML parse + interpret, incl. Notify dispatch (M21)
+        "system/devices/usb-abi.zig", // wire sizes + bit packings + set-up packet encodings
+        "system/devices/usb-ids.zig", // class/subclass/protocol code assignments
         "library/mmio/mmio.zig", // barriers assemble + registers round-trip
+        "system/drivers/ps2-bus/scancode.zig", // set-2 decode + keyboard state machine
+        "system/drivers/ps2-bus/mouse-packet.zig", // 3-byte mouse packet assembly
+        "system/drivers/usb-hid/hid-report.zig", // HID boot-report keyboard/mouse decode
+        "system/drivers/usb-storage/bulk-only-transport.zig", // CBW/CSW wrapper sizes
+        "system/drivers/usb-storage/scsi.zig", // SCSI CDB encodings (big-endian)
+        "system/services/vfs/path.zig", // mount-prefix path matching
+        "system/services/vfs/protocol.zig", // NodeKind / DirectoryEntry sizes + op values
+        "system/services/fat/on-disk.zig", // FAT on-disk struct sizes + type detection
+        "system/services/fat/engine.zig", // FAT read/write over a RAM-backed image
     }) |root| {
         const mod_tests = b.addTest(.{
             .root_module = b.createModule(.{
@@ -464,4 +684,40 @@ pub fn build(b: *std.Build) void {
         });
         test_step.dependOn(&b.addRunArtifact(mod_tests).step);
     }
+
+    // The xkeyboard-config keymap tests need its generated `layouts` import wired, so they
+    // don't fit the plain loop above. Its keycode->character assertions are the end-to-end
+    // proof that the xkb-data -> generator -> Zig-lookup pipeline is correct.
+    const xkb_tests = b.addTest(.{
+        .root_module = b.createModule(.{
+            .root_source_file = b.path("library/xkeyboard-config/xkeyboard-config.zig"),
+            .target = target,
+            .optimize = optimize,
+            .imports = &.{
+                .{ .name = "layouts", .module = xkb_layouts_module },
+            },
+        }),
+    });
+    test_step.dependOn(&b.addRunArtifact(xkb_tests).step);
+
+    // runtime.time's Instant/Duration arithmetic. time.zig pulls in system.zig (the
+    // syscall wrappers), which needs the `abi` module, so it doesn't fit the plain
+    // loop above.
+    const time_tests = b.addTest(.{
+        .root_module = b.createModule(.{
+            .root_source_file = b.path("library/runtime/time.zig"),
+            .target = target,
+            .optimize = optimize,
+            .imports = &.{
+                .{ .name = "abi", .module = abi_module },
+            },
+        }),
+    });
+    test_step.dependOn(&b.addRunArtifact(time_tests).step);
+
+    // Convenience: `zig build gen-xkeyboard-config` regenerates the layout tables from the
+    // vendored data (offline). `fetch` (the network step) stays a manual script run.
+    const gen_xkb = b.addSystemCommand(&.{ "python3", "tools/make-xkeyboard-config.py", "generate" });
+    const gen_xkb_step = b.step("gen-xkeyboard-config", "Regenerate library/xkeyboard-config/generated from the vendored data");
+    gen_xkb_step.dependOn(&gen_xkb.step);
 }

docs/README.md

@@ -45,10 +45,31 @@ rather than restate it. Roughly in the order things happen at runtime:
     until its hardware interrupts it**. The claim is the capability; `irq_ack` is the
     unmask.
 14. **[driver-model.md](driver-model.md) — buses, classes and host controllers.** How
-    real driver stacks factor into three shapes, how families share code, and the
-    proposed ABI for the three primitives still missing (capability passing, DMA +
-    memory barriers, MSI).
-15. **[halting.md](halting.md) — halting.** Why a kernel can't just "exit", and
+    real driver stacks factor into three shapes and how families share code. The
+    three primitives it proposed are long since built (M13 capability passing,
+    M14 DMA + barriers, M15 MSI), and the driver *contract* on top of them —
+    hello, supervision, restart — is built too (device-manager.md, M18).
+15. **[process-management.md](process-management.md) — process management.** The
+    microkernel's `ps`/`kill`/SIGCHLD: enumerate as a table snapshot, the
+    supervision link as the kill authority, and child-exit notifications over the
+    same endpoints IRQs arrive on.
+16. **[process-lifecycle.md](process-lifecycle.md) — the process lifecycle.** Built
+    (M17): signals over IPC as the one lifecycle vocabulary every process speaks — the
+    POSIX.1-1990 words with message delivery instead of stack hijack, the stable
+    `runtime.process` interface, exit reasons, published exit events any stateful
+    service can subscribe to (the VFS releasing dead clients' handles), and the two
+    iron rules (cleanup is the kernel's job; kill is not a signal).
+17. **[device-manager.md](device-manager.md) — the device manager.** Built (M18,
+    through the app surface): the
+    tree, the matcher, and the supervisor. Tree structure lives in the manager,
+    authority stays in the kernel; bus drivers report what they see; drivers are
+    restarted through the lifecycle vocabulary — the plan that turns
+    [resilience.md](resilience.md)'s restart goal into increments.
+18. **[input.md](input.md) — the input module.** Broadcasting input events (keyboard,
+    mouse, joystick): why a synchronous rendezvous can't fan out to many listeners, the
+    asynchronous `ipc_send` primitive built to fix it, and the per-device subscribe/publish
+    service layered on top.
+19. **[halting.md](halting.md) — halting.** Why a kernel can't just "exit", and
     how `while (true) hlt` parks the CPU safely once there's nothing left to do.
 
 Start with the north star:
@@ -61,9 +82,19 @@ Start with the north star:
 - **[resilience.md](resilience.md) — resilience.** A design note (not built yet) on
   fault isolation + live restart — the reincarnation-server + capability model that
   makes "if I break it, I can restart it" real. danos's core motivation.
+- **[zig-self-hosting.md](zig-self-hosting.md) — running Zig on danos.** A design note
+  (not built yet) on making danos a real Zig target (`-target x86_64-danos`) and
+  eventually running the compiler on it. The key realisation: Zig 0.16 reduces an OS
+  port to **one seam** (`std.os.danos`), so we build `runtime.os` (→ that seam) plus a
+  thin `runtime.fs`, retire the `posix` shim, and follow a phased path to
+  `zig build-exe hello.zig` running on danos — **not** Linux-ABI emulation.
 
 Cutting across all of these:
 
+- **[system-requirements.md](system-requirements.md) — system requirements.** The
+  hardware needed to run danos: minimum specs (UEFI x86-64, ACPI, PCIe ECAM,
+  xHCI, ~128 MiB RAM) grounded in what the boot path actually assumes, plus a
+  plain-language guide matching Intel/AMD CPU generations by name.
 - **[arch.md](arch.md) — the architecture split.** How CPU-specific code is kept
   behind a build-time `arch` module so the generic kernel never names x86_64,
   leaving room for other systems (e.g. an AArch64 Raspberry Pi) later.
@@ -75,7 +106,16 @@ Cutting across all of these:
   when to build it, and how to keep it architecture-agnostic.
 - **[acpi.md](acpi.md) — finding the ACPI tables.** The concrete x86 locator chain:
   how the loader captures the **RSDP**, hands its physical address across in `BootInfo`,
-  and how the platform derives the **RSDT/XSDT** from it and walks the SDTs.
+  and how the platform derives the **RSDT/XSDT** from it and walks the SDTs — plus the
+  live event side (the SCI, the power button, GPE/Notify) the ring-3 acpi service runs.
+- **[power.md](power.md) — the power service.** System power as a domain-named
+  service: button/lid/battery events published to subscribers, and init's orderly
+  shutdown composing the [lifecycle](process-lifecycle.md) stop sequence with an ACPI
+  S5 write. Firmware-neutral — a PSCI backend drops in on ARM.
+- **[timers.md](timers.md) — timers and time.** The ring-3 surface for reading the
+  clock and waiting: why `now()` is a syscall rather than a service, and the one-shot
+  timer notification (`timer_bind`) that gives supervisors a timed wait — built on the
+  LAPIC heartbeat and calibrated TSC of [device-interrupts.md](device-interrupts.md).
 - **[smp.md](smp.md) — multiple cores.** A design/research note on how microkernels
   (L4, seL4) handle SMP — big kernel lock vs per-CPU vs multikernel — and how the
   right choice depends on whether danos is chasing real-time or resilience.
@@ -132,7 +172,7 @@ addressed as **`system/services/init`** — the repeated leaf resolves away:
 | Source (root file)                     | Addressed as (module / binary / FHS path) |
 |----------------------------------------|--------------------------------------------|
 | `system/services/init/init.zig`        | `system/services/init` → `/system/services/init` |
-| `system/drivers/hpet/hpet.zig`         | `system/drivers/hpet` → `/system/drivers/hpet`   |
+| `system/drivers/ps2-bus/ps2-bus.zig`   | `system/drivers/ps2-bus` → `/system/drivers/ps2-bus` |
 | `library/runtime/runtime.zig`          | `library/runtime` (the `runtime` module)         |
 
 In **source**, a sub-project is a directory so it can hold many files — the entry is
@@ -157,21 +197,22 @@ system/          → /system   danos's own internals (the self-representation)
   services/    init/ vfs/ device-manager/   system servers → /system/services (vfs/ holds
                               vfs.zig, vfs-test.zig, protocol.zig)
 library/         → /lib       libraries, one sub-directory each
-  runtime/                    the danos-native runtime — the stable application ABI
-  posix/                      POSIX/C compatibility, layered over runtime
+  runtime/                    the danos-native runtime + file API (fs) — the stable application ABI
 boot/            → /boot      the loaders
 tools/  test/                 host-side build + QEMU test harness
 ```
 
 A sub-project exposes its **public interface as a module**: `system/services/vfs/` owns
-the VFS wire protocol (`protocol.zig`, the `vfs-protocol` module), which the POSIX
-layer imports by name. `usb`/`block` drivers will expose their protocols the same way.
+the VFS wire protocol (`protocol.zig`, the `vfs-protocol` module), which the runtime's
+file API (`runtime.fs`) imports by name. `usb`/`block` drivers expose their protocols the
+same way.
 
-`library/posix/` is special: it is the **one place** POSIX/C spellings are allowed
-verbatim (`stat`, `O_CREAT`, `fopen`, `errno`). Everywhere else follows the danos
-naming rule with no exception — see [coding-standards.md](coding-standards.md). The
-POSIX layer calls the runtime, never the kernel's system calls directly, so it never
-appears in the private-ABI path.
+There is **no POSIX/C compatibility layer today**: danos programs do file I/O through the
+danos-native `runtime.fs` (open/read/write/list over the VFS). A hand-rolled POSIX shim
+(`library/posix/`) was retired as premature — the real POSIX/C surface will come later
+from the `std.os.danos` seam (and, eventually, musl) when danos becomes a Zig target (see
+[zig-self-hosting.md](zig-self-hosting.md)). When it does, the foreign-ABI naming
+exception in [coding-standards.md](coding-standards.md) applies to that seam.
 
 ## Source map
 
@@ -195,9 +236,8 @@ appears in the private-ABI path.
 | Framebuffer text console (mirrors to serial) | `system/kernel/console.zig` |
 | In-kernel test cases | `system/kernel/tests.zig` |
 | Arch-specific kernel code (`halt`, GDT/IDT/TSS, exception + interrupt stubs, page tables, APIC/IO-APIC/timer, serial, linker script) | `system/kernel/architecture/x86_64/` |
-| danos-native runtime (`runtime`): syscall wrappers, heap, IPC, device access — the stable application ABI | `library/runtime/` |
-| POSIX/C compatibility (`posix`): unistd, stdio — the one place POSIX names are allowed | `library/posix/` |
+| danos-native runtime (`runtime`): syscall wrappers, heap, IPC, device access, the file API (`fs`) — the stable application ABI | `library/runtime/` |
 | System services (init, the VFS server + `protocol`, the device-manager) | `system/services/` |
-| Device drivers, one sub-project each (`hpet` leaf driver, `bus` bus driver) | `system/drivers/` |
+| Device drivers, one sub-project each (`pci-bus`, `ps2-bus`, `usb-xhci-bus` bus drivers) | `system/drivers/` |
 | Build + `run-x86-64` (QEMU/OVMF) | `build.zig` |
 | QEMU integration test harness | `test/qemu_test.py` |

docs/acpi.md

@@ -107,12 +107,66 @@ firmware-agnostic [device model](discovery.md) gets populated; this note stops a
 part that answers "where are the tables?" — everything past the RSDP is just following
 more pointers the tables themselves provide.
 
+## ACPI events: the SCI, the power button, and GPEs (M21)
+
+The tables above are static description; ACPI is also a *live* channel. Hardware
+raises the **SCI** (System Control Interrupt) — one shared, level-triggered line
+whose vector the FADT names — and the OS reads status registers to learn what
+happened: a fixed event like the power button, or a **General-Purpose Event**
+(GPE) whose handler is an AML method. Since [discovery](discovery.md) moved AML
+to ring 3, the event side lives there too, in the same **acpi service** — the
+device discoverer and the event source are one process, because both need the
+namespace and the port grant.
+
+**The kernel hands the service what it needs and no more.** Reading PM1 event
+blocks and GPE blocks requires the FADT, which the kernel already parses for its
+own `\_S5` poweroff. Rather than re-parse, the kernel appends the **FADT as one
+more memory resource** on the `acpi-tables` node; the service tells it apart
+from the AML blob resources by signature — the FADT keeps its intact `"FACP"`
+header, while the blob resources are header-stripped bytecode that starts with
+no signature. The kernel's own FADT parse is untouched; the service reads the
+PM1 *event* blocks (which the kernel never parsed — it only needs PM1 *control*
+for `\_S5`) and the GPE0/GPE1 blocks straight from its copy. The **SCI itself**
+arrives as the node's one `len == 1` irq resource (distinct from the broad
+`[0, 256)` window that covers children's legacy lines), which is how the service
+finds the line to `irq_bind`.
+
+With those in hand the service enables ACPI mode (only if `SCI_EN` is clear —
+some firmwares boot with it already set), sets `PWRBTN_EN`, and on each SCI:
+
+- **The power button** is a *fixed* event: a set `PWRBTN_STS` bit in PM1 status.
+  The handler clears it (write-1-to-clear), logs the press, and publishes a
+  [`power`](power.md) `power_button` event to subscribers.
+- **GPEs** are the general path: for each set-and-enabled GPE bit `n`, the
+  service evaluates its `\_GPE._L%02X` (level) or `_E%02X` (edge) handler
+  method, drains the **Notify** queue that method produced, maps each notified
+  device to an event (battery, AC, lid, or a generic `notify` with its code),
+  and clears the status bit. A missing handler method is clear-and-log, not an
+  error. Making GPEs work required teaching the interpreter one opcode it never
+  handled — `Notify` (`0x86`) — which it now folds into a bounded queue drained
+  per evaluation; everything else a handler needs (field access, control flow,
+  method calls) was already proven by the ring-3 `_STA`/`_CRS` work.
+
+**How this is tested.** QEMU cannot raise GPEs deterministically on this config,
+so GPE/Notify correctness is proven by **host unit tests** — hand-encoded AML
+with a `Notify` inside a method body, run under `zig build test`. The QEMU
+`power-button` scenario proves the fixed-event path end to end: a QMP
+`system_powerdown` injects a real ACPI power-button press, and the service's SCI
+handler must log it. Battery/AC/lid and the embedded controller's `_Qxx` queries
+are interface-complete but validated on real hardware later.
+
+The service surface these events are *published on* — subscription, the event
+vocabulary, and orderly shutdown — is the power service, [power.md](power.md).
+
 ## Related
 
 - [efi.md](efi.md) — the loader that captures the RSDP before `ExitBootServices`.
 - [memory-map.md](memory-map.md) — the same loader-captures / kernel-consumes seam, and
   the ACPI-reclaim memory the RSDP lives in.
 - [discovery.md](discovery.md) — the broader (still-evolving) plan for turning these
-  tables into one neutral device model shared with the ARM device-tree path.
+  tables into one neutral device model shared with the ARM device-tree path, and how
+  ACPI enumeration and events moved to the ring-3 acpi service.
+- [power.md](power.md) — the domain-named power service the ACPI event side publishes
+  to (button, lid, battery) and its orderly-shutdown path into S5.
 - [arch.md](arch.md) — why the kernel reaches the device code through a `platform`
   module and never names ACPI directly.

docs/coding-standards.md

@@ -65,15 +65,18 @@ Three, and only three.
    `errno`, `O_CREAT`. We don't get to rename `fwrite` to `fileWrite` — it wouldn't be
    `fwrite` any more.
 
-   **This exception is scoped to one place: `library/posix/`.** A file under
-   `library/posix/` *is* the foreign ABI, so it keeps the ABI's spellings — that is the
-   whole rule for that directory. **Everywhere else, Zig/danos naming applies with no
-   POSIX exception**, so there is nothing to get wrong: if you're not in
-   `library/posix/`, expand it. A concept POSIX also has gets a danos name outside that
-   layer — the VFS wire protocol carries a `FileStatus`, not a `Stat`, and a `create`
-   flag, not `O_CREAT`; `library/posix/` is what maps `stat`→`status` and
-   `O_CREAT`→`create` at the boundary. (The `syscall` *wrappers* elsewhere are not an
-   exception to this — they wrap the private danos ABI, so they use danos names.)
+   **This exception is scoped to a file that *is* a foreign ABI, and nothing else.**
+   danos has no such file today: the old `library/posix/` compatibility shim was retired
+   once its callers moved to the danos-native `runtime.fs`, since a hand-rolled POSIX
+   layer is premature until danos actually needs it (see
+   [zig-self-hosting.md](zig-self-hosting.md)). The exception will apply again to the
+   `std.os.danos` seam when danos becomes a real Zig target — that module *is* the C-ABI
+   `system` interface, so it keeps `open`/`read`/`errno`/`O_CREAT`. **Everywhere else,
+   Zig/danos naming applies with no exception**: a concept POSIX also has gets a danos
+   name — the VFS wire protocol carries a `FileStatus`, not a `Stat`, and a `create`
+   flag, not `O_CREAT`; the boundary is where `stat`→`status` and `O_CREAT`→`create` get
+   mapped. (The `syscall` *wrappers* elsewhere are not an exception — they wrap the
+   private danos ABI, so they use danos names.)
 
 2. **Zig idioms are spelled the way Zig spells them.** Three names are the language's,
    not ours, and are left alone:
@@ -96,7 +99,7 @@ That's all — no Unix-abbreviation exception. The source directories are full w
 (`system`, `library`, not `src`/`lib`), and there is no daemon `d` suffix: a driver
 lives in `system/drivers/` and a service in `system/services/`, so the *location*
 already says what it is. Encoding the role in the name too (`busd`, `vfsd`) is
-redundant — the program is just `bus`, `vfs`. Don't put in a name what its directory
+redundant — the program is just `ps2-bus`, `vfs`. Don't put in a name what its directory
 already tells you.
 
 ## A note on collisions
@@ -138,10 +141,36 @@ single word or acronym needs no hyphen: `scheduler.zig`, `paging.zig`, `apic.zig
 conventions above — `snake_case` — because it's an identifier, not a filename.)
 
 **A sub-project's entry point repeats its directory's name** — `init/init.zig`,
-`runtime/runtime.zig`, `hpet/hpet.zig` — and the sub-project is addressed by the
+`runtime/runtime.zig`, `ps2-bus/ps2-bus.zig` — and the sub-project is addressed by the
 *directory* (`system/services/init`, `library/runtime`), with the repeated leaf
 resolving away. See the repository-layout section of [README.md](README.md).
 
+## Named values, not magic numbers
+
+The naming rule has a twin: **a value with meaning gets a name, too.** The same
+principle drives both — a reader should never have to leave the code to understand it.
+An abbreviated *name* forces a reader to guess; a bare *number* forces them worse, out
+to a spec or a header or a comment three files away, to learn what the value even *is*.
+If `0x0C` is the PCI serial-bus class, the code says `BaseClass.serial_bus`, not `0x0C`;
+if `0x04` is the ACPI IRQ resource descriptor, it says `SmallResourceType.irq`, not
+`0x04`. The number is an implementation detail of the name — recorded once, where the
+name is defined, and never spelled again at a use site.
+
+**Prefer an `enum`** when the values form a set (device classes, AML opcodes, resource
+descriptor types, states): the type then also says *which* set a value belongs to, and
+the compiler rejects a value from the wrong one. A lone `pub const` with a descriptive
+name suffices for a one-off (`const large_descriptor_bit = 0x80`). Reach for the enum
+the moment code elsewhere compares against, packs, or produces the value — a packed PCI
+class triple is written from named parts (`.serial_bus`, `.usb`, `.xhci`), never as
+`0x0C_03_30` under a comment that decodes the bytes.
+
+The exceptions are the numbers that carry no hidden meaning: `0` and `1` as plain zero
+and one, an index step, a field width, a bit shift. `x + 1`, `buffer[0]`, and `<< 8`
+need no christening — there is nothing to look up. The test is exactly the naming test:
+*would a reader have to look this up to know what it means?* If yes, name it. This is
+what `opcodes.zig`'s `*_opcode` constants, `acpi-ids`'s `HardwareId`, and `pci-class`'s
+class enums already are — reference data defined once and named everywhere it is used.
+
 ## Why acronyms are the line
 
 Because an acronym has no letters to restore. `MMIO` doesn't become "memory mapped
@@ -150,3 +179,19 @@ input output" in code — that expansion is what the acronym *is for*. But `msg`
 test for "is this an abbreviation I must expand" is simply: *is there a longer word this
 is a clipped form of?* If yes, write the word. If it's an initialism standing in for a
 phrase, leave it.
+
+## Zen of Zig
+
+* Communicate intent precisely.
+* Edge cases matter.
+* Favor reading code over writing code.
+* Only one obvious way to do things.
+* Runtime crashes are better than bugs.
+* Compile errors are better than runtime crashes.
+* Incremental improvements.
+* Avoid local maximums.
+* Reduce the amount one must remember.
+* Focus on code rather than style.
+* Resource allocation may fail; resource deallocation must succeed.
+* Memory is a resource.
+* Together we serve the users.

docs/danos-file-system-hierarchy-FSH.md

@@ -17,7 +17,7 @@ Most modern Unix and Unix-like operating systems follow the FHS. DanOS has its o
 | /srv             | Site-specific data served by this system, such as data and scripts for web servers, data offered by FTP servers, and repositories for version control systems       |
 | /system          | DanOS operating system files (similar idea to C:\Windows). A true representation of danos — its layout mirrors the source tree, so `/system` is what danos *is*.    |
 | /system/devices  | danos virtual device tree e.g. similar to /sys on linux but with danos device tree conventions (the structures in the devices module)                               |
-| /system/drivers  | driver binaries, one sub-project each (e.g. /system/drivers/hpet)                                                                                                   |
+| /system/drivers  | driver binaries, one sub-project each (e.g. /system/drivers/pci-bus, /system/drivers/ps2-bus)                                                                       |
 | /system/services | system-service binaries — the VFS server, init, and other user-mode servers (e.g. /system/services/vfs, /system/services/init)                                      |
 | /system/kernel   | the kernel image                                                                                                                                                    |
 | /tmp             | Directory for temporary files (see also /var/tmp). Often not preserved between system reboots and may be severely size-restricted.                                  |
@@ -61,8 +61,8 @@ to the driver in the order written, and a read consumes what is there. Terminals
 serial lines, keyboards and mice are all of this shape. These are the natural first
 device nodes in danos, because a character driver needs nothing the kernel doesn't
 already provide — it claims its device, maps its registers with `mmio_map`, and blocks
-on `replyWait` for either an interrupt or a client request. `system/drivers/hpet/hpet.zig` is already
-that program, minus the client half.
+on `replyWait` for either an interrupt or a client request. `system/drivers/ps2-bus/ps2-bus.zig`
+is already that program, minus the file-node client half.
 
 The obstacle was never the file type; it is which hardware a ring-3 driver can reach.
 Direct `in`/`out` from user space is still a #GP (no TSS I/O bitmap, IOPL never raised),

docs/device-interrupts.md

@@ -78,6 +78,40 @@ preemption and wakeups (1 ms granularity); the **TSC** is the resolution you rea
 time at. Making `sleep` itself sub-millisecond would take a tickless one-shot
 timer — a later step.
 
+### Is the TSC trustworthy? Invariant, and synchronized
+
+A cycle counter is only a valid *clock* if two things hold, and danos checks both,
+because they decide whether we read time with a cheap `rdtsc` or fall back to the HPET.
+
+**Invariant.** An old TSC counted core clock cycles, so it sped up and slowed down with
+frequency scaling — useless as wall time. Modern CPUs (all of danos's targets) provide an
+**invariant TSC**: a constant rate across P/C-states that never stops. The guarantee is a
+CPUID bit — leaf `0x80000007`, EDX bit 8 — on both Intel *and* AMD. danos reads it in
+`calibrate`, and a TSC that doesn't advertise it is not used as the clocksource. AMD is
+why this matters in practice: it doesn't populate the Intel leaf `0x15` that enumerates
+the TSC *frequency*, so danos already measures AMD's rate against the HPET — but a
+measured frequency without the invariance guarantee is not enough.
+
+**Synchronized.** Each core has its own TSC. Even invariant ones can start at different
+values (a second socket, some firmware), so a thread migrating from a core reading
+`1_000_000` to one reading `999_000` would see time jump *backward*. danos runs a **warp
+check** as each application processor comes online (`checkWarpSource`, adapted from
+Linux's): the waking core and the BSP hammer a shared "highest seen" TSC under a lock,
+and if either ever reads below it, the cores' TSCs are skewed. It's pairwise because APs
+come up one at a time ([smp.md](smp.md)).
+
+**The fallback.** When the TSC fails either test — non-invariant (a bare VM such as the
+default qemu64), or warped between cores — danos moves the monotonic clock onto the
+**HPET** main counter: one fixed-rate counter, so it can neither skew between cores nor
+drift with frequency. It costs a memory-mapped read instead of a register read, but it
+keeps time *accurate*, which is the whole point. The switch preserves the current value,
+so the clock never jumps. The boot log names the outcome:
+
+```
+/system/kernel: clocksource tsc  (TSC invariant: yes, synchronized: yes)   # real Intel/AMD
+/system/kernel: clocksource hpet (TSC invariant: no,  synchronized: yes)   # a bare VM (TCG)
+```
+
 ## Two kinds of vector, one dispatch
 
 The IDT now installs gates `0-47`: the 32 exceptions plus the device range. Every

docs/device-manager.md

@@ -0,0 +1,172 @@
+# The device manager
+
+**Status: the protocol and supervision are built** (M18.1, 2026-07-13): `hello`
+with its deadline, supervised spawn, restart with backoff, and the crash-loop
+cap are in — usb-xhci-bus is the first conforming driver, and the
+`driver-restart` scenario proves fault → backoff → re-claim → cap end to end.
+Tree reports are built too (M18.2, 2026-07-13): the xHCI driver scans its
+root-hub ports and reports each connected device (`child_added`); the manager
+mirrors them and prunes a dead reporter's children, and the `usb-report`
+scenario proves report → prune → respawn → re-report. The application surface is built (M18.3, 2026-07-13):
+`enumerate` and `subscribe` over IPC, with `device-list` as the first client —
+the manager is now the one answer to "what devices exist" for applications.
+The primitives underneath are real ([process-management.md](process-management.md):
+spawn/supervise/kill/exit-notification; [driver-model.md](driver-model.md): the device
+table as a capability system; [drivers.md](drivers.md): claim/map/IRQ), and the first
+per-device driver spawn works (the device manager matches the xHCI controller by PCI
+class and spawns `usb-xhci-bus` with the device id as argv[1]). This document designs
+the rest: the device manager as **the tree, the matcher, and the supervisor** — the
+policy process that turns [resilience.md](resilience.md)'s restart goal into practice
+for drivers.
+
+How processes stop, reload, and report their deaths is deliberately **not** in this
+document: that is the universal lifecycle every danos process speaks —
+[process-lifecycle.md](process-lifecycle.md), signals over IPC and the stable
+`runtime.process` interface. The device manager is that design's first serious
+customer, not its owner. Its own protocol contains nothing lifecycle-shaped; a
+driver is stopped, health-checked, and buried exactly like any other process.
+
+## The tree: structure in the manager, authority in the kernel
+
+The device tree is two things fused: *information* (what exists, how it nests) and
+*authority* (a descriptor is a licence to map physical memory). They separate:
+
+- The **kernel keeps the capability system** — device, I/O-port, and interrupt
+  claims, resource containment on `device_register`, the
+  `mmio_map`/`irq_bind`/`msi_bind` gates — and **cleans all of it up when a process
+  dies** (settled; it is increment 1 of
+  [process-lifecycle.md](process-lifecycle.md)). The three invariants in
+  [driver-model.md](driver-model.md) stay exactly where they are. A device manager
+  that could mint MMIO mappings by its own say-so would be a second kernel, and a
+  buggy one would un-earn everything the microkernel bought.
+- The **device manager owns the tree as data** — identity, topology, naming, driver
+  matching, hotplug events, and being the one process everything else asks about
+  devices. Firmware discovery seeds it (today via the kernel's snapshot); **bus
+  drivers grow it** by reporting what they see; applications query and watch it.
+  `device_enumerate` fades to a manager-internal (then deleted) seam.
+
+Long-term, discovery itself leaves the kernel — but not *into* the manager. PCI
+enumeration is a **pci-bus driver**: the manager spawns it against the host bridge
+(already a device with the ECAM window as a resource), it scans, it reports functions
+like any bus reports children. ACPI becomes an **acpi service** that interprets the
+tables and reports the namespace. The manager only orchestrates and merges. Moving
+AML interpretation out of ring 0 is its own project on its own track; nothing here
+depends on when it lands. (It landed: [discovery.md](discovery.md), M19–M20.)
+
+`device_register` is **idempotent on exact match**: a re-registration with an
+identical (parent, class, identity, resources) tuple returns the existing id
+instead of appending a duplicate. The kernel table has no unregister, so without
+this a restarted registering bus would re-report its children as fresh nodes on
+every respawn. Idempotence is what makes restart-and-re-report sound for *every*
+reporting bus — pci-bus, the acpi service, a future fdt service — not just one,
+and it is why supervision (below) can prune a dead bus's subtree and trust the
+restarted instance to rebuild exactly the same ids.
+
+## The protocol
+
+A `device-manager-protocol` module (the vfs-protocol pattern): extern-struct
+messages, a version in the handshake, reserved fields everywhere. The manager is a
+well-known endpoint (`ipc.register(.device_manager)`); the badge tells it who is
+talking; the same endpoint receives its children's exit notifications — one loop,
+one world.
+
+| Direction | Message | Purpose |
+|---|---|---|
+| driver → manager | `hello { version, role, device_id }` | confirms the argv assignment, starts the deadline clock |
+| bus → manager | `child_added { parent, identity, resources }` | one node the bus discovered |
+| bus → manager | `child_removed { id }` | unplug, or the bus lost it |
+| app → manager | `enumerate` | snapshot of the tree (read-only) |
+| app → manager | `subscribe` | receive published add/remove events |
+
+`hello` is the one deadline the manager enforces itself: spawned and silent past the
+deadline means wrong binary, wrong protocol version, or wedged before main — apply
+the stop sequence and the restart policy. Everything else lifecycle-shaped
+(terminate, the common `ping` liveness call, exit reasons) arrives through
+[process-lifecycle.md](process-lifecycle.md)'s vocabulary, not this protocol.
+
+Assignment stays argv (`usb-xhci-bus <device id>`) for now — simple, and it works.
+The step after `hello` exists is delegation: the manager claims (or is granted) the
+devices and passes the claim to the driver over IPC (the M13 capability-transfer
+mechanism), replacing first-come-first-served `device_claim` with policy. Identity in
+`child_added` is per-bus: PCI children carry the class triple (`pci_class`, as the
+xHCI match already uses); USB children carry the (class, subclass, protocol) triple
+from usb-ids.zig — each bus's native language, decoded by the shared ids modules.
+
+## Supervision and restart
+
+Every driver is spawned with the manager's exit endpoint (`spawnSupervised` — built).
+On a death notification:
+
+1. **Read the reason** ([process-lifecycle.md](process-lifecycle.md) increment 2).
+   Clean exit → it meant to; don't restart. Fault or missed `hello` deadline →
+   restart with **backoff**, and a crash-loop cap (three fast deaths → mark failed,
+   stop respawning, log loudly; a later `reload` to the manager can retry).
+2. **Prune the subtree** the dead bus driver reported. Its children describe
+   protocol state (xHCI slot ids, transfer rings) that died with the process;
+   keeping the nodes would be keeping a lie. Watchers receive `child_removed` — the
+   input service losing, then regaining, a keyboard is the *honest* description of
+   what happened. The restarted instance rediscovers and re-reports.
+3. **The claim is already free** because the kernel released it at death — the
+   restarted instance claims the same controller and comes up.
+
+Who supervises the supervisor: **init** (PID 1), which already supervises the
+services it starts. If the manager dies, drivers keep running (they hold their
+claims; the kernel doesn't care who their supervisor was — though their exit
+notifications now dangle harmlessly). The restarted manager re-learns the world:
+kernel snapshot, then a re-`hello` round — drivers answer a broadcast or are stopped
+and respawned. Full state handoff is deliberately not attempted.
+
+## Thin drivers, class protocols
+
+The [driver-model.md](driver-model.md) three-shape split, restated as processes:
+
+- A **bus driver** (usb-xhci-bus) owns its controller — claim, MMIO, IRQ/MSI, DMA
+  rings — and offers a *transfer* protocol ("submit a control transfer to device N",
+  built from the usb-abi request constructors) plus tree reports to the manager.
+- A **class driver** (usb-hid, usb-storage) owns nothing: it is matched to a reported
+  child by its identity triple, speaks the bus's transfer protocol downward and its
+  service's protocol upward — HID reports to the input service, blocks to the block
+  service. It works unchanged over any controller.
+- **Services** (input, display, block) aggregate class drivers and face applications.
+
+Each arrow is a protocol module. The manager routes none of the data plane — it
+introduces the parties (matching), supervises them (lifecycle), and gets out of the
+way.
+
+## Increments
+
+Increments 1–4 are the lifecycle prerequisites and live in
+[process-lifecycle.md](process-lifecycle.md) (claim cleanup on death, exit reasons,
+published exit events, signals + `runtime.process`). On top of those:
+
+5. **device-manager-protocol**: `hello`, supervised spawn with restart policy;
+   usb-xhci-bus becomes the first conforming driver.
+6. **Tree reports**: `child_added`/`child_removed`; the manager mirrors; xHCI reports
+   the mouse and keyboard QEMU already hangs off it.
+7. **App surface**: `enumerate`/`subscribe` over IPC; `device_enumerate` retreats
+   to a manager-internal seam.
+8. **Discovery migration** — DONE (M19–M20, 2026-07-13): enumeration moved to
+   ring 3 as swappable per-firmware discoverers — the pci-bus driver (M19) then
+   the acpi service (M20), see [discovery.md](discovery.md); the kernel seeds
+   only the host bridge and the acpi-tables node. Matching moved with it:
+   `child_added` grew a `device_id` (the kernel-registered id, `no_device` for
+   unregistered leaves like USB ports) and a firmware `hid`, and the manager now
+   matches drivers from those **reports** rather than its boot-time snapshot. The
+   PCI arm flipped in M19.3, the ACPI arm (ps2-bus matched from `_HID`) in M20.3
+   — each in a single phase so no device is ever matched from both sources at
+   once. The acpi service reports only the non-PCI `_HID` devices, since pci-bus
+   already reports PCI functions (M20.2).
+
+## Settled questions (2026-07-12)
+
+- **Stateful buses**: pruning the subtree on bus-driver death is right for USB. A
+  future storage bus with in-flight writes wants drain-before-terminate — which is
+  exactly the `deadline_ms` parameter `stop()` already has; a per-driver deadline
+  is one value in the manager's policy table when such a bus arrives. No design
+  change.
+- **Manager death**: drivers survive the manager; the restarted manager re-learns
+  the world (above). Checkpointing driver state with the manager is deferred until
+  something demonstrates the need.
+- **Matching stays code until the third bus.** `driverFor`/`pciDriverFor` are
+  honest at two bus types; the third triggers the manifest (a driver declares what
+  it binds: a PCI class triple, a USB class triple, an ACPI `_HID`).

docs/discovery.md

@@ -167,3 +167,81 @@ free; discovery on x86 is partly about *finding* what ARM just tells you.
 - [ipc.md](ipc.md) — the channels that interrupts-as-messages and the device manager
   will ride on.
 - [vision.md](vision.md) — why drivers belong in isolated user space at all.
+
+## Update (M19.3, 2026-07-13): PCI enumeration left the kernel
+
+The kernel now seeds only the `pci_host_bridge` node (ECAM window, MMIO
+apertures derived from the memory map's holes, bus range, and the 16-bit I/O
+window). The per-function walk moved to the ring-3 `pci-bus` driver
+([device-manager.md](device-manager.md)): it claims the bridge, repeats the
+ECAM scan through its mmio grant, and `device_register`s what it finds, which
+the device manager mirrors and matches. The ACPI namespace walk follows in M20;
+the static tables (MADT, HPET, MCFG, FADT + `\\_S5`) stay kernel-side.
+
+## Update (M20.3, 2026-07-13): ACPI enumeration left the kernel too
+
+The kernel no longer folds the AML namespace's Device objects into the device
+tree. It still parses the *static* tables (MADT for SMP, HPET for the tick, MCFG
+for the host bridge, FADT) and still builds the AML namespace — but only to read
+the `\\_S5` sleep type for poweroff. Device discovery is the ring-3 **acpi
+service** ([device-manager.md](device-manager.md)): it claims the `acpi-tables`
+node the kernel publishes (the AML blobs, a broad io_port grant, the SCI),
+re-parses the same blobs with the shared AML module, evaluates `_STA`/`_CRS`,
+and registers + reports each `_HID` device — the device manager matches drivers
+(ps2-bus) from those reports. With M19's pci-bus driver, discovery now runs
+entirely in user space; the kernel seeds only the host bridge and the
+acpi-tables node.
+
+## Discovery is a swappable process per firmware (M19–M20)
+
+Moving PCI and ACPI enumeration out of ring 0 was not just a relocation — it
+made discovery **firmware-neutral by construction**, which is the whole reason
+to do it before the second architecture rather than after. Everything at and
+above the [device-manager](device-manager.md) protocol — descriptors,
+containment, reports, matching, supervision — is generic and may never become
+x86-specific. Discovery is the single firmware-specific piece, and it is
+isolated as **one swappable process per firmware**:
+
+- **x86** boots describe hardware with ACPI, so the discoverer is the **acpi
+  service** ([acpi.md](acpi.md)): it claims the `acpi-tables` node and runs AML.
+- **The Raspberry Pis** hand over a flattened device tree, so the discoverer is
+  an **fdt service**: it claims a `devicetree-blob` node and walks the tree —
+  pure data, no bytecode, so it needs neither a port grant nor an interpreter,
+  strictly simpler than ACPI. (A placeholder until the [aarch64](arm.md)
+  bring-up fills it in.)
+
+The device manager spawns the discoverer under the **neutral ramdisk name
+`discovery`** and never learns which firmware it is on; the build's
+`-Ddiscovery=acpi|fdt` option fills that slot (x86 defaults to `acpi`, the
+aarch64 target flips the default when it lands). The manager owns the device
+tree as *data* and touches no hardware, ever — firmware bytecode runs only
+inside the crashable, supervised discoverer, so an AML fault can never take
+down the supervisor.
+
+Two consequences of neutrality bind on later work:
+
+- **Cross-firmware surfaces are named by domain, not firmware.** System power is
+  a [`power`](power.md) protocol, not an "ACPI events" protocol: on x86 the acpi
+  service registers it, on ARM a PSCI/mailbox service registers the same
+  `ServiceId.power`, and subscribers never learn the difference.
+- **Identity must widen before the fdt service exists.** `DeviceDescriptor`'s
+  8-byte `hid` holds an EISA id but cannot hold an FDT `compatible` string
+  (`"brcm,bcm2835-aux-uart"`); the identity field grows before the ARM path can
+  report a real node.
+
+Two supporting decisions keep the kernel's remaining slice honest:
+
+- **The AML interpreter is a shared build module**, compiled into both the
+  kernel and the acpi service — one source, two builds, no fork. The kernel
+  links it for the `\_S5` poweroff evaluation, the service links it for
+  everything else, and the `acpi-parse` test asserts the two produce the same
+  device count across the ring-3 move.
+- **Bridge apertures come from the firmware memory map, not AML.** Registered
+  PCI functions carry BAR resources, and `device_register` containment demands
+  the bridge own windows that cover them. Those apertures are derived
+  kernel-side from the boot memory map's MMIO holes (regions that are neither
+  RAM nor tables) — mechanical, AML-free, and available at boot regardless of
+  what later moved to user space. The acpi service's authority is likewise
+  exactly one node: the `acpi-tables` node, whose broad io_port grant is the
+  documented trust boundary for the one process allowed to run firmware
+  bytecode.

docs/driver-model.md

@@ -57,8 +57,9 @@ is not an address window. Discovery is trusted; user space is not.
 
 ### What a bus driver looks like
 
-`system/drivers/bus/bus.zig` is the smallest honest one. Its "bus" is the HPET's register block and
-its "devices" are the block's comparators:
+danos ships no demo bus driver — the real ones are `pci-bus`, `ps2-bus`, and
+`usb-xhci-bus`. The smallest *honest* shape, illustrated here with an HPET register block
+as the "bus" and its comparators as the "devices", is:
 
 ```zig
 _ = dev.claim(bus.id);                       // 1. own the bus
@@ -78,8 +79,8 @@ for (0..n) |i| {                             // 3. publish each child
 
 Each child is left **unclaimed**, which is the handoff: a comparator driver can now
 `device_claim` one and `mmio_map` it, and will see only its own 0x20-byte window. A child
-whose window escapes the bus is refused — `bus` asserts that, and the `bus` test
-asserts the kernel's table upholds it.
+whose window escapes the bus is refused; the in-kernel `containment` test asserts the
+kernel's table upholds that ([drivers.md](drivers.md)).
 
 A USB device has *no* resources at all: `resource_count = 0`, because it's addressed
 through its controller, not by MMIO. That case is allowed and is the common one.
@@ -143,7 +144,7 @@ If a class driver needs `mmio`, it has become an HCD and should be one.
   physically-contiguous, pinned, uncacheable, reclaim-on-teardown buffers with the
   physical address exposed (`pmm.allocContiguous`, a DMA arena, `mapUserDmaInto`).
   `dma_below_4g` caps the address for legacy engines; `dma_write_combining` is accepted
-  but falls back to coherent until PAT is programmed. hpet is refactored onto `/lib/mmio`;
+  but falls back to coherent until PAT is programmed. The bus drivers use `/lib/mmio`;
   no DMA driver consumes `dma_alloc` yet.
 - **M15** — interrupts for PCI devices, the MSI half. Discovery now gives every PCI
   function its 4 KiB ECAM config space as resource 0 (unblocking the capability walk
@@ -164,8 +165,11 @@ If a class driver needs `mmio`, it has become an HCD and should be one.
   DMA is still unprotected** (the caveat below). Enforcement lands with the first DMA
   driver, which is what there is to protect and test against. Proven in the `iommu` test,
   booted with an emulated `intel-iommu`.
-- **`system_spawn`** — a user-space supervisor starts a driver: `system_spawn(name)`
-  loads a binary bundled in the initial-ramdisk as a fresh ring-3 process. This is what
+- **`system_spawn`** — a user-space supervisor starts a driver:
+  `system_spawn(name, arguments)` loads a binary bundled in the initial-ramdisk as a
+  fresh ring-3 process; `name` becomes the child's argv[0] and the optional
+  NUL-separated `arguments` blob its argv[1..], delivered on a SysV entry stack
+  ([sysv.md](sysv.md)). This is what
   turned the device manager from "log the match" into "run the driver": the kernel now
   spawns only `init`, `init` spawns the services, and the **device-manager** discovers
   the hardware and spawns each driver ([drivers.md](drivers.md)). Ungated for now — a
@@ -299,8 +303,8 @@ rather than an out-struct. The rest of this section is the original design note.
 
 **The blocker, and it's a hard one.** No PCI device can take an interrupt today.
 [`addBars`](system/devices/acpi.zig) records `.memory` and `.io_port` BARs and never an
-`.irq`; there is no `_PRT` parsing anywhere in the tree. `hpet` only works because the
-HPET advertises its own routing options in its own registers — a privilege no ordinary
+`.irq`; there is no `_PRT` parsing anywhere in the tree. The HPET is the one exception —
+it advertises its own interrupt routing in its own registers, a privilege no ordinary
 device has.
 
 **The fix, in two halves.**
@@ -323,7 +327,7 @@ which means **discovery should give each `pci_device` a `.memory` resource for i
 4 KiB ECAM slot**. That's a small change to `parseMcfg` and it unblocks the whole
 capability walk (MSI, MSI-X, PCIe extended caps) without any new syscall.
 
-Note QEMU's HPET reports `Tn_FSB_INT_DEL_CAP = 0` — no MSI — so `hpet` can never
+Note QEMU's HPET reports `Tn_FSB_INT_DEL_CAP = 0` — no MSI — so an HPET timer could never
 exercise this path. The first MSI driver will be the first PCI driver.
 
 ## M16 — the IOMMU, and the honest caveat  ◑ detection done, enforcement pending
@@ -350,7 +354,7 @@ gap should be named rather than implied.
 
 `M13` (capability passing) is independent of `M14`/`M15` and is the cheapest. It
 unlocks class drivers, which are the shape with no hardware requirements at all — you
-could write a real one against `bus`'s comparators tomorrow.
+could write a real one against any device a bus driver publishes tomorrow.
 
 `M14` and `M15` together unlock the first HCD. `M14`'s barrier layer is worth landing
 on its own regardless: it's small, obviously correct, and stops every future driver

docs/drivers.md

@@ -22,12 +22,12 @@ say.*
 
 ## How a driver gets started: discover, match, spawn
 
-Nothing in the kernel decides that the HPET needs the `hpet` driver — that is policy,
-and policy lives in user space. Boot brings user space up as a three-level supervision
-hierarchy, each level owning one job:
+Nothing in the kernel decides that the PCI host bridge needs the `pci-bus` driver — that
+is policy, and policy lives in user space. Boot brings user space up as a three-level
+supervision hierarchy, each level owning one job:
 
 ```
-kernel  ──spawns──►  init (PID 1)  ──spawns──►  device-manager  ──spawns──►  hpet
+kernel  ──spawns──►  init (PID 1)  ──spawns──►  device-manager  ──spawns──►  pci-bus
   |                     |                          |
   spawns only init,     the service supervisor:    the driver supervisor: enumerates
   publishes the         starts the system          /system/devices, matches each device
@@ -37,8 +37,9 @@ kernel  ──spawns──►  init (PID 1)  ──spawns──►  device-manag
 ```
 
 The kernel launches exactly one process — `init` — and hands it nothing but the raw
-ability to start more (`system_spawn(name)`, which loads a binary bundled in the
-initial-ramdisk as a fresh ring-3 process). Everything else is a user-space decision:
+ability to start more (`system_spawn(name, arguments)`, which loads a binary bundled
+in the initial-ramdisk as a fresh ring-3 process — `name` becoming its argv[0],
+the optional arguments its argv[1..], on a SysV entry stack, see sysv.md). Everything else is a user-space decision:
 
 - **init** ([system/services/init](system/services/init/init.zig)) is the **service
   supervisor**. It spawns the system services danos brings up at boot — today `vfs` and
@@ -52,7 +53,8 @@ initial-ramdisk as a fresh ring-3 process). Everything else is a user-space deci
      is a table (`driverFor`): today a static `timer → hpet` map; a fuller system reads
      what each driver *binds* (a manifest under `/system/drivers`, or the driver
      describing its own match).
-  3. **Spawn** — `system_spawn(driver_name)` starts the matched driver, which then claims
+  3. **Spawn** — `system_spawn(driver_name, arguments)` starts the matched driver (the
+     arguments can carry *which* device it matched), which then claims
      its device and runs the event loop below.
 
 So "how is a driver discovered and configured" has two halves: **discovery** is the
@@ -182,7 +184,11 @@ Two properties worth knowing:
 
 ## A whole driver
 
-`system/drivers/hpet/hpet.zig` is ~150 lines and does all of it. The shape:
+A minimal leaf driver is only ~150 lines and does all of it. danos ships **no such
+example binary** — the driver model is proven by the real drivers (`pci-bus`, `ps2-bus`,
+`usb-xhci-bus`), and a teaching example belongs here, in the docs, rather than as a
+compiled program nobody runs. Illustrated with a hypothetical HPET timer driver, the
+shape is:
 
 ```zig
 const hpet = findHpet(buf) orelse return;      // device_enumerate, look for
@@ -207,8 +213,8 @@ while (...) {
 }
 ```
 
-The HPET is a good first driver for a reason that isn't obvious. Its *counter* is a
-clocksource — the only way to use it is to read it, so it proved `mmio_map` without
+The HPET makes a good illustration for a reason that isn't obvious. Its *counter* is a
+clocksource — the only way to use it is to read it, so it exercises `mmio_map` without
 needing interrupts at all. Its *comparators* are a clockevent, and can be configured
 **level-triggered** (`Tn_INT_TYPE_CNF`), which asserts a bit in `GENERAL_INT_STATUS`
 that the driver must write-1-to-clear. That's a genuine deassert step, so the full
@@ -250,9 +256,10 @@ bus driver may only ever subdivide what it already owns.
 A device with **no resources** is legal and common. A USB device is reached through its
 controller, not by MMIO, so it gets `resource_count = 0`.
 
-See [`system/drivers/bus/bus.zig`](../system/drivers/bus/bus.zig) for a complete one, and
-[driver-model.md](driver-model.md) for how bus drivers, class drivers and host
-controller drivers fit together.
+See [`system/drivers/pci-bus/pci-bus.zig`](../system/drivers/pci-bus/pci-bus.zig) for a
+real one — it claims a PCI host bridge, maps its ECAM window, and publishes each function
+it finds as a child — and [driver-model.md](driver-model.md) for how bus drivers, class
+drivers and host controller drivers fit together.
 
 ## What the kernel does not do for you
 
@@ -311,32 +318,38 @@ uncacheable, physical address exposed), and **memory barriers** (`/lib/mmio`'s
 
 ## Verifying it
 
-The `hpet` test spawns `hpet` from the initial ramdisk and watches the serial log. The driver
-prints `hpet: ok` only after being woken five times, and its loop's only exit is
-through `replyWait` returning a notification — it cannot reach that line by polling.
+No demo driver ships to prove this end to end; the *real* drivers do, so the tests
+target them and the kernel primitives directly:
 
-The last check doesn't trust the driver's self-report at all: the kernel reads the I/O
-APIC redirection entry back and asserts the line really is routed to a device vector,
-really is level-triggered, and really was left unmasked by the driver's final
-`irq_ack`.
+- **`device-manager`** — boots only the device manager, which discovers the PCI host
+  bridge, matches `pci-bus`, and `system_spawn`s it. The test reads kernel state — the
+  process table and the device tree — to confirm pci-bus came up and registered the
+  functions it enumerated: the whole discover → match → spawn → driver-up chain.
+- **`acpi-ps2`** — a user-space driver (`ps2-bus`) is woken by its device's IRQ,
+  delivered as an IPC notification, and attaches the keyboard: IRQ-as-IPC, end to end.
+- **`pci-scan`** — a user-space driver (`pci-bus`) maps its device's MMIO (the ECAM
+  window) and walks it: `mmio_map`, end to end.
+- **`containment`** — the kernel refuses a `device_register` whose child window escapes
+  the parent's grant (else it would be a syscall for mapping arbitrary memory), while an
+  identical re-register stays idempotent. Asserted in-kernel, straight against the broker.
+- **`irqfree`** — the teardown path. Binds two owners to one shared endpoint, releases
+  one, and reads the I/O APIC back: the departing owner's line is masked, the sibling's
+  is not. That second half is why bindings are keyed on the owning *task* and not on the
+  endpoint pointer — endpoints are shared, so releasing "everything pointing at this
+  endpoint" would silently mask a live driver's device.
+- **`iopass`** — the `device_grant` teardown rule, so destroying a driver's address
+  space never returns MMIO frames to the RAM pool.
 
 ```
-$ python3 test/qemu_test.py hpet irqfree iopass
-  hpet         ... PASS  (matched 'DANOS-TEST-RESULT: PASS')
+$ python3 test/qemu_test.py device-manager acpi-ps2 pci-scan containment irqfree iopass
+  device-manager ... PASS  (matched 'DANOS-TEST-RESULT: PASS')
+  acpi-ps2     ... PASS
+  pci-scan     ... PASS  (matched 'DANOS-TEST-RESULT: PASS')
+  containment  ... PASS  (matched 'DANOS-TEST-RESULT: PASS')
   irqfree      ... PASS  (matched 'DANOS-TEST-RESULT: PASS')
   iopass       ... PASS  (matched 'DANOS-TEST-RESULT: PASS')
 ```
 
-Two companions cover what `hpet` can't, because it never exits:
-
-- **`irqfree`** — the teardown path. Binds two owners to one shared endpoint, releases
-  one, and reads the I/O APIC back: the departing owner's line is masked, the sibling's
-  is not. That second half is why bindings are keyed on the owning *task* and not on
-  the endpoint pointer — endpoints are shared, so releasing "everything pointing at
-  this endpoint" would silently mask a live driver's device.
-- **`iopass`** — the `device_grant` teardown rule, so destroying a driver's address
-  space never returns MMIO frames to the RAM pool.
-
 ## What's next (not done here)
 
 The big driver-model pieces — capability passing (class drivers), DMA + barriers, MSI,
@@ -360,3 +373,23 @@ the first DMA driver to protect and test against) and these smaller items:
 - **Interrupt priority / threaded IRQ latency.** `notifyFromIsr` enqueues the woken
   driver but doesn't preempt (`wakeLocked` deliberately leaves that to the caller), so
   a woken driver waits for the next scheduling point.
+
+## The driver contract (M17–M18)
+
+Claiming and mapping is half of being a danos driver; the other half is the
+**lifecycle and protocol contract**, and the runtime makes it nearly free:
+
+- Build on `runtime.service.run` — one replyWait loop folding protocol
+  requests, signals, and notifications into callbacks. The harness answers the
+  universal zero-length ping and turns `terminate` into a clean exit for you
+  ([process-lifecycle.md](process-lifecycle.md)).
+- A driver spawned with an assignment (its device id as argv[1]) sends the
+  versioned `hello` to the device manager inside the deadline, and a **bus**
+  driver reports what it discovers with `child_added`
+  ([device-manager.md](device-manager.md); usb-xhci-bus is the reference
+  implementation).
+- Crash freely — that is the design. The kernel releases your claims, IRQ
+  bindings, and MSI vectors at death; the manager reads your exit reason,
+  prunes what you reported, restarts you with backoff, and your fresh instance
+  re-claims and re-reports. Never depend on your own cleanup running
+  (iron rule 1).

docs/input.md

@@ -0,0 +1,161 @@
+# The input module: broadcasting input events
+
+A keyboard driver has one keystroke and *many* programs that might want it — a shell, a
+window server, a logger. None of them owns the hardware, and the driver should not know
+who is listening. So between the drivers and the listeners sits the **input service**
+(`system/services/input/`): drivers **publish** events to it, programs **subscribe**, and
+it fans each event out to every interested subscriber. It is an ordinary ring-3 process
+reached over IPC, like the [VFS server](../system/services/vfs/vfs.zig) — no kernel knows
+what a key is.
+
+## One service, several device classes
+
+The service carries three device classes today — **keyboard**, **mouse**, and
+**joystick/gamepad** — and is built to take more
+([protocol.zig](../system/services/input/protocol.zig)). Each class has its own typed
+event:
+
+- `KeyEvent` — `key_down`/`key_up` (physical make/break) and `key_press` (a character was
+  produced, carrying the Unicode scalar); plus a layout-independent `keycode` and a
+  `modifiers` bitmask.
+- `MouseEvent` — relative `motion` (`dx`/`dy`), `button_down`/`button_up`, and `scroll`.
+- `JoystickEvent` — `axis` moves (a signed value on a `control` index) and
+  `button_down`/`button_up`.
+
+All three travel in one **`InputEvent` envelope** tagged with a `DeviceKind`, so the
+fan-out is a single code path and a subscriber can take a mix of classes on one stream.
+Decode an envelope with `asKeyboard()` / `asMouse()` / `asJoystick()` (each returns null
+unless the tag matches). A subscriber names the classes it wants with a **`device_mask`**,
+and the service routes each event only to subscribers whose mask includes its class — so a
+mouse-only listener never wakes for keystrokes.
+
+## Why this needed a new kernel primitive
+
+The interesting part is delivery, and it runs straight into the shape of danos IPC.
+[ipc.md](ipc.md) describes a **synchronous rendezvous**: a server holds exactly one
+pending reply (`Task.ipc_client`) and *must* answer it on its next `replyWait`. Two
+consequences decide the whole design:
+
+1. **You cannot block N subscribers waiting for "the next event".** A server can hold only
+   one caller at a time, so the natural "subscriber calls `next_event()` and blocks" API
+   is impossible for more than one subscriber. Delivery therefore has to be **push** — the
+   service reaching out to subscribers — not pull.
+
+2. **A synchronous push can hang the whole service.** If the service delivered with
+   `ipc_call`, it would block until each subscriber replied. `ipc_call` has no timeout, and
+   the kernel does **not** wake a caller parked on a *dead* peer's endpoint (it only fails a
+   peer that was mid-reply — see [process.zig](../system/kernel/process.zig)
+   `releaseTaskResourcesLocked`). One subscriber that exits mid-delivery would wedge input
+   for everyone. That is the opposite of the resilience the microkernel is for.
+
+The fix is the asynchronous send that [ipc.md](ipc.md) had already earmarked as future
+work ("asynchronous / buffered send … for notifications between servers"):
+
+```
+ipc_send(handle, message_ptr, message_len) -> 0 / -errno
+```
+
+`ipc_send` copies a small payload into the endpoint's **bounded queue** and wakes a
+receiver, then returns immediately — it never blocks and so can never hang on a dead or
+slow subscriber. The receiver picks it up through the same `replyWait` it already runs:
+the wake arrives as a **buffered message** — `notify_badge_bit | notify_message_bit` set in
+the badge (distinguishing it from a bare IRQ/child-exit notification), the sender's task id
+in the low bits, and the payload in the receive buffer, with no reply owed. The queue holds
+16 messages per endpoint; a full queue **drops the oldest**, because a buffered message is
+discrete data, not a coalescing "level" like an interrupt. See
+[ipc-synchronous.zig](../system/kernel/ipc-synchronous.zig) (`sendLocked`, `popPost`, and
+the `replyWait` receive loop).
+
+This is the async counterpart of `ipc_call`, and the input service is its first consumer.
+
+## How the pieces fit
+
+```
+  keyboard/mouse driver, input-source       input service               subscriber(s)
+  -----------------------------------       -------------               -------------
+  connectSource(); loop:                    replyWait:                  subscribeKeyboard()/…All:
+    publishKeyboardEvent(k) ─ ipc_call ─▶  publish → broadcast:            createIpcEndpoint()
+    publishMouseEvent(m)                     for each sub whose            callCap(subscribe,
+    publishJoystickEvent(j)                   mask matches event.device:     send_cap = ep,
+                                               ipc_send(sub_ep) ──────▶       device_mask)
+                                       reply ok                          loop: next()
+                                     subscribe → store {ep cap,          └─ replyWait(ep)
+                                       task id, device_mask}                → InputEvent
+```
+
+- A **subscriber** calls `input.subscribe(mask)` — or a typed helper: `subscribeKeyboard()`,
+  `subscribeMouse()`, `subscribeJoystick()` (one class, `next()` returns the decoded event),
+  or `subscribeAll()` (every class, `next()` returns a tagged `InputEvent`)
+  ([library/runtime/input.zig](../library/runtime/input.zig)). It creates its own endpoint
+  and hands it to the service as a **capability** (M13 capability passing — the input
+  service is that feature's first real user), along with its `device_mask`. Then it loops on
+  `next()`, a `replyWait` on that endpoint returning each pushed event.
+- A **source** (a keyboard, mouse, or joystick driver) calls `input.connectSource()` and the
+  method for its class: `publishKeyboardEvent`, `publishMouseEvent`, or
+  `publishJoystickEvent`. Publishing is a short synchronous `ipc_call` the service answers at
+  once; the service's own fan-out is asynchronous, so publishing never blocks on a slow
+  subscriber.
+- The **service** ([input.zig](../system/services/input/input.zig)) keeps a small subscriber
+  table (endpoint handle + owning task id + `device_mask`). On `publish` it `ipc_send`s the
+  event to every subscriber whose mask includes the event's device class. On `subscribe` it
+  stores the passed capability and mask and, as housekeeping, prunes any slot whose owning
+  process has exited (checked against `process_enumerate`) — not for correctness (an async
+  send to an orphaned endpoint is harmless) but to reclaim the slot.
+
+Publisher and subscriber must be **separate processes**: a single thread that both
+published and serviced its own subscription would deadlock (its `publish` call blocks until
+the service delivers to its endpoint, which only the same thread could receive).
+
+## Status and follow-ups
+
+- **The keyboard is real.** The `ps2-bus` driver owns PNP0303, which carries *both* the
+  0x60/0x64 ports and IRQ1, so reading the hardware lives in the bus, not in
+  [keyboard.zig](../system/drivers/ps2-bus/keyboard.zig): the bus binds IRQ1 and, on each
+  interrupt, drains port 0x60, routing every byte by the status register's
+  auxiliary-output bit to whichever child driver **attached** for that device (an
+  `AttachRequest` to the well-known `ps2_bus` service, carrying the child's endpoint as a
+  capability; the bytes then arrive as asynchronous `ForwardedByte` messages, so the IRQ
+  path never blocks on a child). The keyboard driver decodes the stream — scancode **set 2**,
+  what the keyboard sends with the 8042's legacy translation off, decoded by
+  [scancode.zig](../system/drivers/ps2-bus/scancode.zig) into USB HID usage keycodes with
+  make/break, typematic-repeat, and modifier tracking (host-tested under `zig build test`) —
+  and publishes real `key_down`/`key_press`/`key_up` events.
+- **Keycode → character** is wired in: the keyboard driver fills a `key_press` event's
+  `character` through [`library/xkeyboard-config`](../library/xkeyboard-config/README.md)
+  (`xkb.map(layout, keycode, mods)` → keysym + Unicode character), synthesizing the ASCII
+  control characters for Enter/Tab/Backspace/Escape, whose keysyms map to no Unicode. The
+  layout defaults to `us`; the bus can pass another as the driver's argv[2] — the seam for
+  a future settings source.
+- **The mouse is real too.** IRQ12 is enumerated on the auxiliary device's own ACPI node
+  (PNP0F13), so the bus claims that node alongside the controller and routes both IRQs to
+  its one endpoint, acking whichever line the notification's badge names.
+  [mouse.zig](../system/drivers/ps2-bus/mouse.zig) attaches the way the keyboard does and
+  assembles the forwarded bytes with
+  [mouse-packet.zig](../system/drivers/ps2-bus/mouse-packet.zig) (three-byte stream-mode
+  packets: sync/overflow handling, nine-bit movement, screen-convention `dy` — host-tested
+  under `zig build test`) into `button_down`/`button_up` transitions and `motion` events.
+  **Follow-up:** the IntelliMouse magic-knock for a scroll wheel (four-byte packets) and
+  `scroll` events. The hardware-free `input-source` still rotates through all three classes
+  synthetically (including a joystick, which has no driver yet) via the
+  `input.synthetic*Event` helpers.
+- **Drop-oldest under overflow** is a defined loss; the 16-slot ring absorbs normal bursts.
+  Real backpressure/flow-control is future work.
+- **`publish` is unauthenticated** — any process may publish, consistent with the current
+  bring-up trust model (see [driver-model.md](driver-model.md)). A source capability is
+  future work.
+
+## Verifying it
+
+The `input` case (`python3 test/qemu_test.py input`, in
+[tests.zig](../system/kernel/tests.zig) `inputTest`) boots the real kernel and spawns the
+service, the synthetic source (which cycles keyboard, mouse, and joystick events), and a
+subscriber that took all three classes. It passes only when the subscriber heartbeats
+`input-test: ok` — proof that an event travelled source → service → subscriber over IPC,
+exercising `ipc_send`, capability-passing subscription, and per-device routing. Each
+serial line names the class received, so the log shows all three arriving on one stream.
+
+## See also
+
+- [ipc.md](ipc.md) — the synchronous rendezvous and the notification path `ipc_send` extends.
+- [syscall.md](syscall.md) — the system-call surface, including `ipc_send`.
+- [driver-model.md](driver-model.md) — class drivers, capability passing (M13), the trust model.

docs/interrupts.md

@@ -80,11 +80,22 @@ inline). `build.zig` adds `isr.s` to the arch module.
 ## Reporting a fault
 
 `isr_common` calls `exceptionHandler`, which forwards to a swappable `on_fault`
-hook. The generic kernel installs a reporter (`onException` in `main.zig`) that
-prints, in red, the exception name and vector, the error code, the faulting RIP
+hook. The generic kernel installs a reporter (`onException` in `kernel.zig`) that
+prints the exception name and vector, the error code, the faulting RIP
 and RSP, and — for a page fault (#PF, vector 14) — the faulting address from
-**CR2**. Then it halts. There's no fault *recovery* yet, so every exception is
-terminal; the point is that it's now **visible** instead of a silent reset.
+**CR2**. What happens next depends on where the fault came from:
+
+- **User mode (CPL 3): kill the process, keep the machine.** The kernel is intact
+  (the CPU trapped onto the task's kernel stack), so the faulting process is
+  killed — address space, IRQ bindings, and IPC handles reclaimed; a client it
+  owed a reply to is failed with `-EPEER` — and the core reschedules. A crashing
+  driver takes itself down, never the OS. This is fault recovery step 2 of
+  [resilience.md](resilience.md). NMI, double fault, and machine check are
+  excluded: they report machine trouble regardless of what was running.
+- **Kernel mode: halt this core.** The trusted base itself is broken, so there is
+  nothing safe to kill; the fault is still *contained* to the core (an
+  application-processor fault leaves the rest of the system running), and the
+  report makes it **visible** instead of a silent reset.
 
 The hook is set before `arch.init()` in `kmain`, so a fault during setup is still
 caught.

docs/ipc.md

@@ -95,6 +95,30 @@ This is what makes a user-space driver possible at all, and it's the subject of
   every capability is either well-known (the registry) or inherited — there's no way
   to delegate one.
 - **Asynchronous / buffered send** for the cases where a rendezvous is the wrong
-  shape (logging, notifications between servers).
+  shape (logging, notifications between servers). *Landed as `ipc_send`* — a
+  non-blocking post to an endpoint's bounded payload queue, delivered through
+  `reply_wait` as a buffered message (badge bit `notify_message_bit`). Built for, and
+  first used by, the [input service](input.md)'s keyboard-event broadcast, where a
+  synchronous push would let one dead subscriber hang the fan-out. A full queue drops
+  the oldest (discrete messages, not a coalescing level like the notification ring).
 - **A bounded reply.** `MSG_MAX` is 256 bytes and the copy runs under the big kernel
   lock; a bulk transfer wants shared pages, not a copy.
+
+## Lifecycle conventions over IPC (M17)
+
+Three conventions from [process-lifecycle.md](process-lifecycle.md) ride the
+notification mechanism:
+
+- **Signals** arrive as notifications on the endpoint a process nominated with
+  `signal_bind` (`runtime.process.bindSignals`): badge = the signal bit plus the
+  coalesced pending mask (`runtime.process.signalsFrom` decodes). Statements,
+  never questions; no payload, no reply.
+- **One-shot timers** (`timer_bind`, `runtime.system.timerOnce`) land as a
+  timer-bit notification — the timed wait: a service arms a deadline and keeps
+  serving, instead of blocking in sleep.
+- **The universal ping**: a **zero-length request is the liveness probe**,
+  answered with a zero-length reply by the service harness itself
+  (`runtime.service.run`). No protocol's requests start at length zero, so the
+  encoding cannot collide, and a wedged service simply fails to answer — which
+  is the diagnosis. Deep health ("can I reach my hardware?") stays a per-service
+  protocol message.

docs/power.md

@@ -0,0 +1,128 @@
+# The power service: events and shutdown
+
+A laptop lid closes, a battery drains, someone presses the power button — and
+several parts of the system might care: a session manager dims the screen, a
+logger notes it, and ultimately *something* has to turn the machine off. None of
+them owns the hardware that reported the event, and the reporter should not know
+who is listening. So system power is a **service**: an event source **publishes**
+button/lid/battery/AC events, interested processes **subscribe**, and one
+privileged caller — init — can ask it to power the machine off. It is the same
+publish/subscribe shape as the [input service](input.md), applied to power.
+
+## Why a service, and why it is named for the domain, not the firmware
+
+Where the events come from is firmware-specific — on x86 they ride the ACPI SCI
+([acpi.md](acpi.md)); on a Raspberry Pi they would come from PSCI or a mailbox.
+What subscribers want is not: *the lid closed* means the same thing regardless of
+who noticed. So the surface is **domain-named**. There is a `power-protocol`
+module and a well-known `ServiceId.power = 5`; on x86 the **acpi service**
+registers it, and on ARM a PSCI/mailbox service will register the *same* id.
+Subscribers call `runtime.ipc.lookup(.power)` and never learn which firmware they
+are on — the neutrality the whole [discovery](discovery.md) migration exists to
+preserve, carried one layer up into a running-system surface.
+
+This is why the protocol is `power`, not "ACPI events": naming a cross-firmware
+surface after one firmware would leak x86 into code the ARM port must reuse
+unchanged.
+
+## The protocol
+
+The `power-protocol` module ([system/services/power/protocol.zig](../system/services/power/protocol.zig))
+follows the vfs-protocol pattern — extern-struct messages, a version, reserved
+fields. Three operations:
+
+| Direction | Operation | Purpose |
+|---|---|---|
+| subscriber → service | `subscribe` | receive published events; the subscriber's endpoint rides as the call's **capability** (the input/device-manager pattern) |
+| init → service | `shutdown` | orderly shutdown's last step: enter S5 (soft off) |
+| service → subscriber | `event` | a published `EventMessage`, delivered as a buffered message (never sent *to* the service) |
+
+Events are published, not polled: like the input service, the service holds
+subscriber endpoints as capabilities and `ipc_send`s each event as a buffered
+message, so a slow or dead subscriber can never wedge the source. The event
+vocabulary is hardware-neutral:
+
+- `power_button` — the button was pressed (a fixed ACPI event on x86).
+- `lid`, `ac`, `battery` — the named GPE-driven events.
+- `notify` — a device notification that maps to none of the above; its `code`
+  (the ACPI `Notify` argument) and the notifying device's `hid` say which device
+  and what happened.
+
+An `EventMessage` carries the `event` tag plus `code` and an 8-byte `hid`, so a
+generic `notify` is fully described without a second round trip.
+
+**`shutdown` is authority, not information.** It is the only operation that
+*does* something irreversible, so it is gated: the contract is that only init
+(PID 1) may request it, because init is the process that has already run the stop
+sequence over everything else. The acpi service implements this as a **soft
+gate** — it honors `shutdown` only from a process that is a *subscriber*, and
+init is the one subscriber. That stands in for "only the system supervisor may
+power off" without hard-coding a pid, so it still holds under tests where PID 1
+is not init.
+
+## Orderly shutdown
+
+Powering off cleanly is where the power service, the [process
+lifecycle](process-lifecycle.md), and [ACPI events](acpi.md) compose. init
+already supervises the services it starts; for shutdown it runs **one event loop
+over one endpoint** that carries three things at once: its children's exit
+notifications, the lifecycle **signals** it can receive (`terminate`), and the
+**power events** it subscribes to — plus a re-arming heartbeat timer proving PID
+1 is alive. (init subscribes with retries, because the power service registers
+`.power` well after init starts; a missing power service is not fatal — a
+`terminate` signal drives the same path.)
+
+On a `power_button` event or a `terminate` signal, init:
+
+1. logs that it is shutting down,
+2. runs the standard stop sequence — `runtime.process.stop(child, deadline,
+   endpoint)` — over its children **in reverse spawn order**, so the VFS stops
+   last (other services may flush through it), each child getting the
+   *terminate → deadline → kill* escalation from
+   [process-lifecycle.md](process-lifecycle.md), and
+3. requests `.power` `shutdown`.
+
+The service then enters **S5** (soft off) by writing `SLP_TYP | SLP_EN` to the
+PM1 control register(s) from ring 3, mirroring the kernel's own
+`system/devices/power.zig` `sleepValue`. If the write returns instead of powering
+the machine off, it logs loudly so a test fails rather than hangs.
+
+**No new system call was needed for S5.** The broad io_port grant on the
+`acpi-tables` node ([discovery.md](discovery.md)) already put the PM1 control
+ports in the acpi service's hands, so writing S5 from ring 3 is something it
+could physically already do; formalizing it as a protocol operation added a
+contract, not authority. The kernel keeps `power.zig` for its own test paths and
+panic-time poweroff, where no user space is available to ask.
+
+## Verifying it
+
+Two QEMU scenarios exercise the path, both injecting a real ACPI power-button
+press via QMP `system_powerdown` (there is no other deterministic power event on
+this config):
+
+- `power-button` proves the source: the acpi service's SCI handler logs the
+  press and publishes `power_button` (the ACPI half is in [acpi.md](acpi.md)).
+- `orderly-shutdown` proves the whole composition: button → init logs shutting
+  down → children stopped → the service enters S5 → QEMU exits. The ordered
+  regex is the proof, and QEMU's self-exit through S5 is the pass.
+
+## Scope
+
+Interface-complete but validated on real hardware (the author's laptop) later,
+because QEMU does not emulate them: battery `_BST`/`_BIF` evaluation beyond the
+interface stubs, lid and AC events, and the embedded controller's `_Qxx`
+queries. Deliberately out of scope for now: reboot over the power protocol, S3
+sleep, per-device D-states (a future lifecycle-vocabulary extension, since
+"suspend" has the shape of a signal every driver must answer and has no consumer
+until laptop sleep), and thermal zones.
+
+## See also
+
+- [acpi.md](acpi.md) — where the events come from on x86: the SCI, the power
+  button fixed event, and GPE/Notify dispatch in the acpi service.
+- [discovery.md](discovery.md) — why the surface is domain-named, and the
+  firmware neutrality that makes a PSCI backend drop-in on ARM.
+- [process-lifecycle.md](process-lifecycle.md) — the stop sequence
+  (`terminate → deadline → kill`) and signals init composes into shutdown.
+- [device-manager.md](device-manager.md) — the supervision model init mirrors for
+  its own children.

docs/process-lifecycle.md

@@ -0,0 +1,327 @@
+# Process lifecycle: signals over IPC
+
+**Status: increments 1–4 built** (2026-07-12): claim release on death, exit
+reasons, published exit events, and signals + one-shot timers + the service
+harness are all in — the interface below is as-built. The primitives underneath
+predate this design ([process-management.md](process-management.md):
+spawn, the supervision link, kill, child-exit notifications); this document designs
+the layer above them — the standard vocabulary a danos process speaks about its own
+life, and the stable `runtime.process` interface that carries it. Nothing here is
+device- or driver-specific: a driver, the VFS, and a user application all stop,
+reload, and die the same way. The device manager is simply this design's first
+serious customer ([device-manager.md](device-manager.md)).
+
+**"POSIX" in this document means the concepts, never the letter of the standard.**
+danos borrows the ideas and the hard-won lessons (what SIGTERM *means*, why SIGPIPE
+was a mistake) without inheriting the mechanism, the API, or the names. The naming
+rule is danos's own and it is strict: plain words that communicate intent
+(`terminate`, `reload`, `exited`) and the IPC vocabulary the system already speaks
+(`bind`, `subscribe`, `publish`, `endpoint`) — never `SIG*`, never a second word for
+a concept that already has one. Literal POSIX arrives later and lives elsewhere: the
+`std.os.danos` seam that makes danos a Zig target, and eventually a **musl-based C
+layer** on the same native surface (see [zig-self-hosting.md](zig-self-hosting.md)) —
+musl's syscall surface retargeted at danos system calls and IPC protocols (files onto
+the VFS protocol, `sigaction`/`wait` onto this lifecycle, sockets onto whatever
+networking becomes). Ported programs see POSIX; the system underneath never does.
+
+## Why a standard vocabulary
+
+A supervisor can only manage processes it has never heard of if "please exit" means
+the same thing to all of them. That is the one thing POSIX signals got deeply right:
+`SIGTERM` means the same thing to nginx and to a five-line script, which is why
+process supervision on Unix (init systems, container runtimes) is possible at all.
+danos wants that property from day one, because supervision-and-restart is the
+system's core motivation ([resilience.md](resilience.md)).
+
+What POSIX got wrong — for a system like this — is the **delivery mechanism**:
+asynchronous control-flow hijack. A Unix handler runs on a stolen stack at an
+arbitrary instruction boundary, which is why the async-signal-safe function list
+exists, why `errno` must be saved, and why the canonical signal bug is a SIGTERM
+handler innocently calling `printf` mid-`malloc`. That entire bug class comes from
+the mechanism, not the vocabulary, and none of it is worth importing.
+
+A microkernel already has the right channel: **a signal is a message.** QNX delivers
+POSIX signals over its message passing; seL4 has notification objects; Erlang turned
+"death is a message to whoever linked" into a reliability philosophy. danos has
+already done it once without naming it: a child's death arrives as a notification
+badge on the supervisor's endpoint — the microkernel's SIGCHLD, the IRQ-as-IPC
+pattern reused. Signals are the same pattern reused a third time.
+
+## The mechanism
+
+- **`signal_bind(endpoint)`** — a process nominates the endpoint its signals arrive
+  on, exactly as `irq_bind` nominates where a device's interrupts land. The runtime
+  does this at startup for any program that opts in.
+- **`process_signal(id, signal)`** — posts the signal as an asynchronous
+  notification to the target's bound endpoint: badge = `notify_badge_bit |
+  notify_signal_bit | pending signals`. Non-blocking for the sender, always.
+- **Pending signals coalesce** in a per-process bitmask until the target next waits
+  — exactly like interrupt notifications, and exactly POSIX's own semantics for
+  non-realtime signals (two pending SIGTERMs are one SIGTERM). The bitmask *is* the
+  design: signals carry no payload. Anything with a payload is a protocol message.
+- **Authority**: the supervisor may signal its children — the same link that is
+  already the kill authority. A process may signal itself. Anything broader waits
+  for transferable process handles.
+- **No binding, no problem**: a process that never calls `signal_bind` is not
+  broken — its signals pend unread and only `process_kill` works on it. Simple
+  programs stay simple; the vocabulary is opt-in, the kill authority is not.
+
+Because delivery is a message into the process's own event loop, there is no
+async-signal-safe list in danos: a handler is ordinary code running at a point the
+process chose. The bug class is gone by construction, not by discipline.
+
+## The vocabulary: POSIX.1-1990, sorted honestly
+
+The full 1990 set, and what each becomes. Two intrinsically problematic cases get a
+defense below the table.
+
+| POSIX.1-1990 | danos disposition | Notes |
+|---|---|---|
+| SIGTERM | signal `terminate` | finish up and exit; the supervisor's polite half |
+| SIGHUP | signal `reload` | re-read configuration / re-scan |
+| SIGINT | signal `interrupt` | interactive interrupt; meaningful once a console can send it, in the vocabulary now so numbering is stable |
+| SIGQUIT | signal `quit` | as SIGINT, without the core-dump baggage |
+| SIGALRM | signal `alarm` | timer expiry as a message; the Unix SIGALRM+`longjmp` timeout hacks are impossible here. In the vocabulary, unbuilt: no consumer yet, and when one appears it is runtime sugar over the existing timer — zero kernel work |
+| SIGUSR1, SIGUSR2 | signals `user_1`, `user_2` | service-defined |
+| SIGCHLD | **already exists** — the exit notification | the badge carries the child id, dodging the classic coalescing bug (Unix code must loop `waitpid`) |
+| SIGKILL | `process_kill` — kernel mechanism | its definition is "cannot be handled"; it was never really a signal |
+| SIGABRT | exit reason `abort` | `abort()` is synchronous self-termination, not an event |
+| SIGSEGV, SIGILL, SIGFPE | exit reasons, **never delivered** | see below |
+| SIGPIPE | **an error return**, not a signal | see below |
+| SIGSTOP, SIGTSTP, SIGTTIN, SIGTTOU, SIGCONT | deferred | job control needs terminals, sessions, and process groups; stop/continue is scheduler territory |
+
+**The fault signals (SIGSEGV, SIGILL, SIGFPE) are intrinsically wrong for messages.**
+They are *synchronous* — raised at a specific faulting instruction, not "sometime
+soon". A message cannot be delivered to a process whose next instruction re-faults;
+it never reaches its event loop to read it. POSIX only makes fault handlers "work"
+via the async hijack (run the handler *instead of* the instruction), and even there,
+returning from a SIGSEGV handler without curing the cause is undefined behavior.
+danos's architecture already has the better answer: fault → the kernel kills the
+process ([resilience.md](resilience.md) step 2, built) → the supervisor reads the
+reason → restart. Recovery is restart, not a handler. This is also truer to the 1990
+standard than handling is: the standard's default action for all three was
+"terminate the process".
+
+**SIGPIPE deserves special contempt.** Its default kills a process that writes to a
+closed pipe — which is why "the whole server died because one client disconnected"
+is roughly every network daemon's first production bug, and why every mature codebase
+contains the same fix: ignore SIGPIPE, handle the `EPIPE` error return. danos made
+the right choice natively already — a reply owed to a dead peer fails with `-EPEER`.
+Errors from operations are error returns from those operations. The posix layer can
+synthesize SIGPIPE for ported code that expects it.
+
+### Statements, not questions
+
+A signal and a protocol message both travel over IPC — the difference is the
+**contract**, not the transport. danos IPC has two primitives, both already in
+daily use: the **asynchronous notification** (a badge — bits that coalesce into a
+pending mask; the sender never blocks; no payload, *no reply path*; how IRQs and
+exit events arrive) and the **synchronous call** (a rendezvous — payload both
+ways, the caller waits for the reply; how VFS requests work). A signal is the
+first kind: a *statement*. `terminate` wants no reply — the exit notification is
+its acknowledgement.
+
+A health probe is the second kind: a *question*, worthless without its answer —
+and the answer's absence within a deadline is the very thing being measured.
+Asked as a signal it has no reply channel (a coalescing bit can't carry an answer,
+and the authority rule forbids a child signalling its supervisor back); asked as a
+call, the timeout-is-the-diagnosis semantics come free. So there is no `health`
+signal. Liveness is the common **`ping`**: a reserved request every harness-run
+service answers automatically on its main endpoint — still free for the service
+author, still one obvious way — and a supervisor's probe is a `ping` call with a
+deadline.
+
+## The two iron rules
+
+1. **Cleanup is the kernel's job.** A process can die with no warning — fault,
+   kill, power. Correctness must never depend on a `terminate` handler running. On
+   any death the kernel releases the address space, IPC handles, IRQ bindings, and
+   owed replies (built), and must also release **device, I/O-port, and interrupt
+   claims and MSI vectors** (the known gap in
+   [process-management.md](process-management.md); increment 1). A signal handler is
+   for *graceful* work — flushing, deregistering, saving — never for *necessary*
+   work.
+2. **Kill is not a signal, and exit reasons are load-bearing.** The standard stop
+   sequence is *terminate → deadline → `process_kill`*; the unhandleable kill stays
+   a kernel mechanism. And a supervisor deciding whether to restart must know *how*
+   the child died: clean exit (meant to — don't restart), fault (restart with
+   backoff), killed (the supervisor did it). The exit notification today carries
+   only the id; it grows a reason. Restart policy cannot be written without it.
+
+## Who learns of a death
+
+A death has three audiences, and conflating them is how systems end up with either
+zombie state or privileged snooping:
+
+1. **The supervisor** — gets the exit notification on the endpoint it gave at spawn
+   (built), which grows the `ExitReason` (increment 2). The supervisor is the only
+   audience that needs the *reason*, because it is the only one deciding whether to
+   restart.
+2. **The peer owed a reply** — already built: a client that dies mid-request fails
+   the server's reply with `-EPEER`; a server that dies fails its waiting clients
+   the same way. This covers the *synchronous* case only.
+3. **The subscribers** — the new piece, and it is the input service's
+   publish/subscribe shape ([input.md](input.md)) applied to exits. A stateful
+   service accumulates per-client state across many requests: the VFS holds a dead
+   client's open file handles, the input service holds its subscriptions, a future
+   network stack holds its sockets. None of these are the client's supervisor, and
+   none learn anything from a failed reply if the client simply never calls again.
+   So the kernel **publishes every exit** to whoever subscribed:
+   `process_subscribe(endpoint)` adds a subscriber, and each death posts a
+   notification to every subscriber (badge = `notify_exit_bit | process id` — the
+   same encoding supervisors already decode, the IRQ-as-IPC pattern once more). The
+   subscriber filters for ids it holds state for and releases what the dead client
+   held. Correlating is free of bookkeeping: an IPC sender's badge already *is* its
+   task id (`runtime.ipc.Received`), so the id a service has been keying client
+   state by all along is the id the exit event carries.
+
+   Subscription, not broadcast-to-everyone: only processes that asked receive
+   events, the kernel keeps a bounded subscriber table, and delivery is the same
+   non-blocking coalescing notification as everything else — a dying process never
+   waits on its mourners. Subscribing is ungated, like `process_enumerate`: what is
+   running (and dying) is not a secret between cooperating processes. Subscribers
+   do not receive the exit reason — the VFS does not care *why* the client died.
+
+This is the service-side mirror of iron rule 1: **a service must never depend on
+its clients cleaning up after themselves.** Handle release on client death is the
+service's job, triggered by the published exit event — never by a courtesy
+"closing now" message that a crashed client will never send.
+
+## The stable interface: `runtime.process`
+
+`runtime.process` already owns what a process receives at birth (`Init`, the
+argv contract). It grows to own the other end of life.
+
+**The runtime is the stable interface; the numbers are not.** danos applications do
+not make system calls — they call the runtime library, and the system-call numbers,
+notification bits, and signal bit positions beneath it are a **private kernel ↔
+runtime contract** that may change at any time (settled 2026-07-12). This is why
+the runtime exists. Today kernel and runtime ship from one tree in one image, so
+"stability" is simply building them together. When driver binaries start shipping
+as separately-versioned applications — the whole point of the restart design — the
+binary's embedded runtime version becomes compatibility metadata (the same idea as
+the protocol version in the device manager's `hello`), and the kernel refuses what
+it cannot serve. Signals therefore need no reserved numbering scheme: the enum
+below is vocabulary, not ABI.
+
+```zig
+/// The signal vocabulary. The value is the bit position in the pending mask — a
+/// private kernel/runtime detail, free to change while they ship together.
+pub const Signal = enum(u5) {
+    terminate = 0, // SIGTERM: finish up and exit
+    reload = 1, // SIGHUP: re-read configuration
+    interrupt = 2, // SIGINT
+    quit = 3, // SIGQUIT
+    alarm = 4, // SIGALRM
+    user_1 = 5, // SIGUSR1
+    user_2 = 6, // SIGUSR2
+};
+
+/// A decoded pending mask: the coalesced set of signals a notification delivered.
+pub const SignalSet = struct {
+    pending: u32,
+    pub fn has(set: SignalSet, signal: Signal) bool { ... }
+    pub fn iterate(set: SignalSet) Iterator { ... }
+};
+
+/// Nominate `endpoint` as this process's signal endpoint (signal_bind). The
+/// runtime's service harness calls this; a bare program may call it directly and
+/// fold signals into its own replyWait loop.
+pub fn bindSignals(endpoint: usize) bool { ... }
+
+/// Decode a received badge into signals, or null if the badge is not a signal
+/// notification (mirrors ipc.Received.isChildExit).
+pub fn signalsFrom(badge: usize) ?SignalSet { ... }
+
+/// Send `signal` to process `id`. Supervisor-gated, like kill; non-blocking.
+pub fn sendSignal(id: u32, signal: Signal) bool { ... }
+
+/// The standard stop sequence: terminate, wait up to `deadline_ms` for the exit
+/// notification, then process_kill. The one call a supervisor needs.
+pub fn stop(id: u32, deadline_ms: u64) void { ... }
+
+/// Subscribe `endpoint` to published exit events (process_subscribe). Every
+/// process death posts an asynchronous notification: badge = notify_exit_bit |
+/// process id — the same encoding a supervisor's exit notification uses, decoded
+/// by the same ipc.Received helpers. For stateful services: release what the dead
+/// client held (file handles, subscriptions, sockets). Ungated, like
+/// process_enumerate.
+pub fn subscribeExits(endpoint: usize) bool { ... }
+
+/// How a process ended — queried after the exit notification (the kernel records
+/// it first, so the two never race). What restart policy reads. (Built in M17.2.)
+pub const ExitReason = enum(u8) {
+    exited, // returned from main / clean exit
+    aborted, // abort() — deliberate self-termination (SIGABRT's ghost; reserved)
+    segmentation_fault, // SIGSEGV's ghost
+    illegal_instruction, // SIGILL's ghost
+    arithmetic_fault, // SIGFPE's ghost
+    protection_fault, // general protection fault
+    fault, // any other CPU exception
+    killed, // process_kill
+};
+```
+
+Two deliberate absences. There is no `mask`/`block` API — a process that is not
+ready for a signal simply has not waited on its endpoint yet; the pending mask *is*
+the blocked set. And there is no per-signal handler registration at this layer —
+dispatch is the process's own `switch` over `SignalSet`, or the service harness's
+callbacks (`on_terminate`, `on_reload`) for programs that want defaults.
+
+### The service harness
+
+`runtime.service` owns the `replyWait` loop and folds every event source — signals,
+child exits, protocol messages — into callbacks, with the vocabulary's defaults:
+`terminate` returns from the loop (clean exit), the common `ping` is answered automatically,
+`reload` is ignored unless overridden. One loop, no locking, nothing reentrant. A
+service author writes domain logic; the lifecycle contract is satisfied by the
+harness. A process that bypasses the harness and ignores its signals meets the
+deadline-then-kill escalation — you cannot force a process to implement an
+interface, but you can make compliance free and non-compliance fatal.
+
+### The musl layer later
+
+The POSIX C layer is a **musl port**: musl's arch/syscall layer retargeted so that
+what musl believes are kernel syscalls become danos runtime calls and IPC — `open`
+and `read` onto the VFS protocol, `kill`/`sigaction`/`waitpid` onto this document's
+vocabulary, `exit` onto the runtime's exit path. `sigaction` handlers registered
+through it are invoked by the runtime's loop when the signal message arrives —
+synchronous underneath, async-looking to ported code, delivered at wait boundaries
+the way most Unix programs already experience signals (at syscalls). No stack hijack
+ever happens, `SA_RESTART` semantics come free because nothing was interrupted, and
+SIGPIPE can be synthesized from `-EPEER` for the programs that expect it. C programs
+get POSIX; danos-native programs never pay for it.
+
+## Increments
+
+1. **Kernel: release device/port/IRQ claims and MSI vectors on death** — the
+   cleanup half of iron rule 1, and the prerequisite for any restart story. Test:
+   kill a claiming driver, spawn it again, the claim succeeds.
+2. **Exit reason in the death notification** (`ExitReason` above).
+3. **Exit events**: `process_subscribe` in the kernel (bounded subscriber table,
+   publishes on every death), `runtime.process.subscribeExits`; the VFS becomes the
+   first subscriber — releasing a dead client's handles is its proof test.
+4. **Signals**: `signal_bind` + `process_signal` + the pending mask in the kernel;
+   `runtime.process` grows the interface above; the service harness handles
+   `terminate` and answers the common `ping`; `stop()` for supervisors.
+
+[device-manager.md](device-manager.md) builds directly on all four.
+
+## Settled questions (2026-07-12)
+
+- **Signal numbering is not ABI**: the runtime is the stable interface; the numbers
+  beneath it are a private kernel ↔ runtime contract (see "The stable interface").
+- **Liveness is a `ping` call, not a signal**: signals are statements, questions
+  are synchronous calls (see "Statements, not questions"). A service wanting *deep*
+  health ("can I reach my hardware?") defines its own protocol message on top.
+- **Process handles: deferred.** Pids + the supervisor gate cover everything
+  planned; transferable handles (Fuchsia-style, delegating signalling without
+  delegating kill) wait for the capability table to grow types beyond endpoints.
+- **`alarm`: in the vocabulary, unbuilt.** No consumer yet; when one appears it is
+  runtime sugar over the existing timer (arm a timer that posts your own signal) —
+  zero kernel work, so deferring costs nothing.
+- **Subscription granularity: all exits**, subscriber-side filtering — one
+  subscription per service, a bounded kernel table. Per-id subscriptions only if
+  event volume ever matters (hundreds of processes, not before).
+- **Client identity across the exit boundary: no convention needed** — an IPC
+  sender's badge already is its task id (see "Who learns of a death").

docs/process-management.md

@@ -0,0 +1,120 @@
+# Process Management
+
+How danos lists, supervises, and kills processes — the microkernel answer to
+`ps`, `kill`, and `SIGCHLD`/`wait`.
+
+## Why system calls, not `/proc`
+
+Unix systems sit on a spectrum. Classic BSD/macOS list processes through
+syscalls (`sysctl(KERN_PROC)`) and kill through `kill(2)`; Linux renders the
+process table as `/proc` for *reading* but still kills through a syscall; Plan 9
+made the file tree the whole interface (`echo kill > /proc/n/ctl`). Microkernels
+mostly abandon ambient PIDs: Minix and QNX route everything through a user-space
+process-manager server, and Fuchsia/seL4 control processes only through handles.
+
+danos rules out `/proc` **as the primitive**: here a `/proc` would be served by
+the VFS server — a user process — which would put the VFS in the path of process
+control. If the VFS (or anything under it) hangs, nothing could be listed or
+killed, *including the hung VFS*. The control plane for processes must not
+depend on a process. So the primitives are kernel system calls; a read-only
+`/proc` rendering can be layered on later, and a POSIX-style process-manager
+server can be built *from* these primitives when one is needed.
+
+## The three primitives
+
+### `process_enumerate(buffer, maximum) -> total`
+
+A snapshot of the task table into a caller buffer of `abi.ProcessDescriptor`
+(id, supervisor, state, priority, name) — the exact shape of
+`device_enumerate`, so `ps` is a user program over a snapshot, not a kernel
+service. The total may exceed what fit; call again with a larger buffer. Kernel
+tasks are included with an empty name — an honest listing shows the idle tasks
+too. Ungated and read-only: what is running is not a secret between cooperating
+bring-up processes.
+
+### `system_spawn(..., exit_endpoint) -> child id`, and the supervision link
+
+`system_spawn` records the caller as the child's **supervisor** and returns the
+child's process id (ids are monotonic, never reused — a stale id can only miss).
+That link is the kill authority: it answers "who may kill process 7?" without
+inventing users or permissions, the same way a device *claim* is the capability
+for `mmio_map`. It composes with the supervision hierarchy the device manager
+already forms: init supervises the services it starts, the device manager
+supervises the drivers it matches. (A transferable process *handle* — Fuchsia
+style — can replace the id once the handle table grows types beyond endpoints.)
+
+`exit_endpoint` (a handle, or `abi.no_cap`) is the supervisor's death-watch: when
+the child ends — clean exit, CPU fault, or `process_kill` — the kernel posts an
+asynchronous notification to that endpoint, exactly like a bound IRQ. The badge
+carries `abi.notify_badge_bit | abi.notify_exit_bit | child_id`, so one endpoint
+supervises many children and can even share with IRQ notifications. This is the
+microkernel's SIGCHLD: no new mechanism, just the IRQ-as-IPC pattern reused, and
+a supervisor's event loop (`ipc.replyWait`) already knows how to receive it. The
+child holds a reference to the endpoint from birth, so the notification cannot
+dangle even if the supervisor dies first.
+
+### `process_kill(id) -> 0 / -ESRCH / -EPERM`
+
+Only the supervisor may kill; kernel tasks are not killable processes. Like a
+signal, delivery is prompt but asynchronous — 0 means the kill is accepted and
+irrevocable; the exit notification confirms completion.
+
+## How a kill lands (the kernel mechanics)
+
+Everything below runs under the big kernel lock, where task states cannot move.
+
+- **Target ready or blocked** (not on any core): reaped on the killer's own
+  call. The reap releases what death always releases (IRQ bindings first, then
+  a client the target still owed a reply to is failed with `-EPEER`, IPC handles
+  closed, the exit notification posted last) — plus the unlinking only a
+  *remote* death needs: out of the ready queue, out of an endpoint's sender FIFO
+  (`Task.ipc_wait_endpoint`), out of a receive wait queue (`Task.wait_queue`),
+  and out of any server's owed-reply slot, so nothing ever dequeues a dangling
+  pointer. Destroying the address space is safe because no core can have it
+  loaded: every switch away from a task loads the next task's tables.
+- **Target running on another core**: it cannot be torn down mid-instruction,
+  so it is condemned (`Task.kill_pending`) and dies at whichever comes first:
+  - its next **system_call entry** — checked before dispatch, so a condemned
+    process cannot spawn, claim, or message anything on its way out;
+  - its core's next **timer tick** — but only when the task is not inside one
+    of its own system calls (`Task.in_system_call`): the tick may have
+    interrupted kernel code mid-operation, where teardown would leak whatever
+    the operation held. User-mode execution is always a safe kill point. The
+    tick-time terminate abandons the interrupt frame exactly like the fault
+    path (the LAPIC is acknowledged before the tick hook runs);
+  - any core's tick finding it **blocked or ready** (it entered a syscall and
+    parked after being condemned) — reaped by the same remote-reap path.
+
+  A pure user-mode spin loop that never makes a system call therefore dies
+  within one tick; nothing a process does can outrun the kill.
+
+The scheduler stays below the process layer: finishing a kill (IRQ bindings,
+handles, the notification) is called *up* through two hooks process.zig
+registers at boot (`terminate_current_hook`, `reap_task_hook`), mirroring how
+the architecture layer calls up into `tick`.
+
+## Known gaps (bring-up honesty)
+
+- ~~Device claims are not released on death~~ Closed (M17.1): every path out of a
+  process releases its device claims alongside its IRQ and MSI bindings
+  (`releaseTaskResourcesLocked`), so a restarted driver can claim its hardware
+  again — the cleanup half of [process-lifecycle.md](process-lifecycle.md)'s iron
+  rule 1. The `claim-release` test proves the kill → release → re-claim cycle.
+- Kernel stacks of dead tasks are leaked, as on every exit path (no reaper yet).
+- ~~There is no exit status in the notification~~ Closed (M17.2): the kernel
+  records how every process ends — exited, a fault class, or killed — before it
+  posts the exit notification, and the supervisor reads it with
+  `process_exit_reason` (`runtime.process.exitReason`). This is the input to
+  restart policy ([process-lifecycle.md](process-lifecycle.md)); an exit *code*
+  for the clean case can still ride alongside later.
+- Enumerate writes through the caller's raw pointer under the bring-up trust
+  model, like `device_enumerate` (an unmapped page is a self-DoS, not an
+  isolation break).
+
+## Tests
+
+`process-list` (enumerate), `process-kill` (kernel-level kill paths, refusals,
+notifications), `supervision` (the whole user-side surface via the process-test
+service: spawn supervised → enumerate → kill blocked and spinning children →
+notifications → gone), `claim-release` (a killed claim-holder's device is
+claimable again). See test/qemu_test.py.

docs/resilience.md

@@ -1,6 +1,17 @@
 # Resilience: fault isolation and live restart
 
-A design/research note, not built yet. This is the property danos is really chasing:
+Steps 1–4 of the ordering below are **built** (M17–M18, 2026-07-13): user-mode
+isolation; fault → kill the process → keep the core (`onException`; the
+`fault-recovery` test); the supervisor notification **with exit reasons**
+([process-lifecycle.md](process-lifecycle.md) — clean exit, fault class, or
+killed, recorded before the notice posts); and the **restart policy itself**
+([device-manager.md](device-manager.md)): the device manager supervises every
+driver, restarts crashes with backoff, caps crash loops, and re-claims work
+because the kernel releases a dead process's claims. The `driver-restart` and
+`usb-report` scenarios prove kill → release → respawn → re-claim → re-report
+end to end. What remains of this document's ladder is scope, not mechanism:
+more of the system moved into restartable processes (the discovery migration,
+[discovery.md](discovery.md), is the next rung). This is the property danos is really chasing:
 **if a part of the OS breaks, isolate it, and re-initialise it — without rebooting.**
 A crashed driver gets restarted; a wedged service gets killed and brought back. It's
 the reason the [microkernel](vision.md) shape was chosen, and it's a *separate* goal
@@ -111,9 +122,11 @@ Honest boundaries:
 ## Suggested ordering
 
 1. **User mode + address-space isolation** — the shared prerequisite (also on the
-   path for everything else).
+   path for everything else). **Done.**
 2. **Kernel: fault → kill process → notify.** Turn today's "halt on fault" into
-   "confine to the process and report it."
+   "confine to the process and report it." **Done** (the kill and reclaim; the
+   supervisor notification waits for step 3's supervisor). A killed server's
+   pending client is unblocked with `-EPEER` rather than hung.
 3. **A minimal supervisor server** that can (re)start a process.
 4. **Resource cleanup on death** — reclaim memory/MMIO/IPC/IRQ, via caps or a grant
    table.

docs/syscall.md

@@ -47,6 +47,8 @@ Everything else---including`read()`,`write()`,`malloc()`, and`fork()`---will run
     -   **What it does:**Used strictly by your background user-space servers (like your disk driver or filesystem). It sends a reply to the last client that called it, and immediately puts the server to sleep until the next request arrives.[[1](https://news.ycombinator.com/item?id=33078441)]
 3.  **`Yield()`/`Thread_Ctrl()`**
     -   **What it does:**Allows a thread to voluntarily give up its CPU time slice, or allows a root task to spawn/kill threads.
+4.  **`ipc_send(endpoint, message_buffer)`(Asynchronous Send)**
+    -   **What it does:**Posts a small payload to an endpoint's bounded queue and returns *without* blocking — no rendezvous, no reply. The receiver picks it up through the same `IPC_ReplyWait`, as a buffered message. It is the async counterpart of `IPC_Call`, for one-to-many broadcasts where a synchronous rendezvous would let one dead or slow receiver hang the sender. The [input service](input.md) — keyboard-event fan-out — is its first user. A full queue drops the oldest message (a buffered message is discrete data, unlike a coalescing interrupt notification).
 
 * * * * *
 

docs/system-requirements.md

@@ -0,0 +1,227 @@
+# System Requirements
+
+Minimum and recommended hardware for running danos. Every requirement below is
+grounded in what the current code actually assumes at boot — this is a
+description of the real target, not an aspirational one.
+
+## Summary
+
+danos targets a **modern UEFI x86-64 PC with ACPI and PCIe**. The practical
+minimum is:
+
+- 64-bit x86-64 CPU with SSE2, APIC, and `syscall`/`sysret`
+- UEFI firmware (no BIOS / legacy boot)
+- ACPI tables: MADT, MCFG, FADT
+- PCIe with an ECAM (MMConfig) window
+- **128 MiB RAM** (target); see [Memory](#memory) for the breakdown
+- USB via **xHCI only**
+
+There is no support for legacy BIOS boot, x2APIC, port-IO PCI configuration, or
+any USB host controller other than xHCI.
+
+## Plain-language hardware guide
+
+If you don't want to cross-reference chipset datasheets, here's roughly what era
+of PC works. These are **guidance based on when the required features became
+standard**, not a list of tested machines — the authoritative rules are in the
+technical sections below.
+
+The feature that sets the floor is **built-in xHCI USB** (danos supports no other
+USB controller) combined with **UEFI firmware**. Both became standard on
+mainstream desktops and laptops around **2012**.
+
+| | Known-good baseline | Comfortable recommendation |
+|---|---|---|
+| **Intel** | 3rd-gen Core "Ivy Bridge" (2012) with a 7-series "Panther Point" chipset — Intel's first chipset with xHCI built in | 6th-gen Core "Skylake" (2015) or newer |
+| **AMD** | A-series "Llano" APU with an A75 FCH (2011) — the industry's first chipset with built-in xHCI | Any AM4 platform, i.e. Ryzen (2017) or newer |
+
+**AMD is not behind Intel here — it was first.** AMD's A75 FCH shipped with
+native xHCI in April 2011, about a year *ahead* of Intel's 7-series (2012); AMD
+was the first vendor to earn USB-IF certification for chipset-level USB 3.0. The
+two "comfortable recommendation" dates differ only because they name convenient,
+long-supported product lines (Skylake, Ryzen) — not because of any USB
+capability gap. Every AMD desktop platform from the A75 FCH (2011) and FM2/AM3+
+era onward has built-in xHCI, and any of them qualifies as a baseline.
+
+Older 64-bit machines (e.g. Intel Core 2, Nehalem, Sandy Bridge) meet the CPU
+requirements but typically **lack built-in xHCI and/or ship with BIOS instead of
+UEFI**, so they are not supported.
+
+### Matching your CPU by name
+
+If you know your chip's marketing name or codename, find it here. Everything from
+the **Supported** rows down works; the **Too old** row does not.
+
+**Intel Core** (the "-lake"/"-bridge"/"-well" codenames):
+
+| Status | Generation | Codename(s) | Year |
+|---|---|---|---|
+| Too old | 2nd gen | Sandy Bridge | 2011 |
+| Supported (baseline) | 3rd gen | Ivy Bridge | 2012 |
+| Supported | 4th–5th gen | Haswell, Broadwell | 2013–2014 |
+| **Recommended** | 6th–9th gen | **Skylake**, Kaby Lake, Coffee Lake | 2015–2018 |
+| Recommended | 10th–11th gen | Comet Lake, Ice Lake, Tiger Lake, Rocket Lake | 2019–2021 |
+| Recommended | 12th gen+ | Alder Lake, Raptor Lake | 2021–2023 |
+| Recommended | Core Ultra | Meteor Lake, Arrow Lake, Lunar Lake | 2023+ |
+
+**AMD:**
+
+| Status | Family | Codename(s) | Year |
+|---|---|---|---|
+| Supported (baseline) | A-series APU (A75/A85 FCH) | Llano, Trinity, Richland, Kaveri | 2011–2014 |
+| Supported | FX (AM3+) | Bulldozer, Piledriver | 2011–2012 |
+| **Recommended** | **Ryzen** 1000–5000 (AM4) | Summit/Pinnacle Ridge, Matisse, Vermeer (Zen–Zen 3) | 2017–2020 |
+| Recommended | Ryzen 7000+ (AM5) | Raphael, Granite Ridge (Zen 4 / Zen 5) | 2022+ |
+| Recommended | Threadripper / EPYC | Zen and later | 2017+ |
+
+(These map generations to the era their platforms shipped built-in xHCI + UEFI;
+they are guidance, not a tested-hardware list.)
+
+**Two caveats that matter regardless of CPU:**
+
+- **Firmware must be UEFI.** Many 2011-era machines could do either UEFI or
+  legacy BIOS — danos needs it set to UEFI. There is no BIOS boot path.
+- **Input is PS/2 only, for now.** danos does not yet support USB
+  keyboards/mice. This is fine on most **laptops** (their built-in keyboards are
+  wired to a PS/2-style i8042 controller) but means a **desktop with only USB
+  ports** currently has no usable keyboard. USB HID input is planned.
+
+Virtual machines are the easiest way to meet every requirement: QEMU (with OVMF/
+UEFI, a `qemu-xhci` controller, and the default Q35 machine type), or any
+hypervisor configured for UEFI firmware and an xHCI USB controller.
+
+## CPU / architecture
+
+| Requirement | Detail | Source |
+|---|---|---|
+| **x86-64, 64-bit only** | Kernel and loader are built exclusively for `x86_64`; the loader rejects any non-x86-64 kernel ELF (`error.WrongArchitecture`). | `build.zig:285`, `boot/efi.zig:418` |
+| **Long mode + PAE + NX** | AP trampoline sets `CR4.PAE`, `EFER.LME`, `EFER.NXE`; NX is used in kernel page-table entries. | `system/kernel/architecture/x86_64/trampoline.s:62` |
+| **SSE / SSE2** | Baseline: the compiler emits SSE for ordinary struct copies. Trampoline enables `CR4.OSFXSR` + `OSXMMEXCPT` and clears `CR0.EM`. | `build.zig:282`, `trampoline.s:62` |
+| **`syscall` / `sysret`** | Primary user↔kernel entry path. `EFER.SCE` enabled; `STAR`/`LSTAR`/`SFMASK` programmed per core. (`int 0x80` exists as a parallel gate.) | `architecture/x86_64/per-cpu.zig:59`, `isr.s:169` |
+| **Local APIC (xAPIC)** | LAPIC accessed via MMIO at `0xFEE00000`. LAPIC ID read as a `u8` — classic xAPIC. **x2APIC is not supported** (no MSR path). | `apic.zig:62`, `apic.zig:414` |
+| **CPUID + RDTSC** | CPUID leaf `0x15` for TSC frequency; RDTSC is the monotonic clock. | `apic.zig:279`, `apic.zig:84` |
+| **SMP (optional)** | Multi-core supported via INIT–SIPI–SIPI; ceiling `maximum_cpus = 128`. Single core is fine. Cores beyond the ceiling are parked. | `system/parameters.zig:16`, `apic.zig:144` |
+
+## Firmware / boot
+
+- **UEFI only.** A custom UEFI application loader is installed to
+  `\EFI\BOOT\BOOTX64.efi`. There is **no BIOS, multiboot, or limine** path. The
+  loader tolerates UEFI Class-3 machines with no legacy PIC/PIT.
+  (`build.zig:464`, `boot/efi.zig`)
+- **ACPI is the hardware-discovery mechanism.** The RSDP is taken from the UEFI
+  configuration table (ACPI 2.0 GUID preferred, 1.0 fallback). Without a valid
+  RSDP there is **no device discovery** — no SMP, no IOAPIC routing, no PCI/USB.
+  (`efi.zig:578`, `boot-handoff.zig:144`)
+- **Required ACPI tables:** MADT (interrupt topology), MCFG (PCIe ECAM base),
+  FADT (power / PM timer). Optionally consumed: HPET, DMAR, SPCR.
+  (`system/devices/acpi.zig:3`)
+- The loader reads `/system/kernel`, `/system/services/init`, and
+  `/boot/initial-ramdisk.img` off the FAT boot volume. The kernel can boot
+  "kernel-only" without init or the ramdisk. (`efi.zig:14`, `efi.zig:66`)
+
+## Interrupt controller
+
+- **Local APIC + I/O APIC required.** I/O APIC base, GSI base, and MADT
+  interrupt-source overrides come from ACPI. (`cpu.zig:365`, `apic.zig:119`)
+- **MSI supported** — edge-triggered, keyed by vector, no I/O APIC mask cycle.
+  Vector window 33–46, timer on 32, spurious on 47. (`system/kernel/irq.zig:70`,
+  `cpu.zig:397`)
+- The legacy 8259 PIC is remapped and masked **only if present** (MADT
+  `PCAT_COMPAT`); it is not required. (`apic.zig:103`)
+
+## PCI / PCIe
+
+- **PCIe with ECAM (MMConfig) required.** The PCI bus driver maps the host
+  bridge's ECAM window (1 MiB config space per bus) and computes config
+  addresses directly. **There is no legacy CF8/CFC port-IO config path** — the
+  driver bails if the bridge exposes no ECAM window. The ECAM base comes from
+  the ACPI MCFG table. (`system/drivers/pci-bus/pci-bus.zig:41`, `acpi.zig:6`)
+
+## USB
+
+- **xHCI only.** The sole USB driver is `usb-xhci-bus`, and the device manager
+  binds it strictly to PCI prog-IF `0x30` (xHCI). UHCI / OHCI / EHCI exist only
+  as report strings with no driver behind them — **USB 1.x/2.0-only controllers
+  are not supported.** (`system/drivers/usb-xhci-bus/`,
+  `system/services/device-manager/device-manager.zig:34`)
+- USB input (keyboard/mouse over HID) is future work; the current input stack is
+  PS/2. See [Buses & devices](#buses--devices).
+
+## Timers
+
+Calibration prefers, in order: (1) CPUID leaf `0x15` TSC frequency, (2) HPET,
+(3) ACPI PM timer (3.579545 MHz, from FADT), (4) legacy PIT. Any one suffices —
+HPET/PM-timer/PIT are optional fallbacks when CPUID `0x15` is absent.
+(`apic.zig:180`)
+
+- **TSC** — monotonic high-resolution clock.
+- **LAPIC timer** — scheduler heartbeat, periodic at `timer_hz = 1000 Hz`.
+  (`parameters.zig:39`)
+
+## Memory
+
+**Target: 128 MiB RAM.** The system uses 4 KiB pages and a bitmap physical-frame
+allocator built from the firmware memory map. There is no hardcoded minimum-RAM
+constant — the allocator only panics if there is no usable region, or none large
+enough to hold its own bitmap. (`system/kernel/pmm.zig:13`, `pmm.zig:77`)
+
+Where the budget goes:
+
+| Consumer | Size | Source |
+|---|---|---|
+| Kernel heap (cap, grown one page at a time) | up to **64 MiB** | `system/kernel/heap.zig:26` |
+| Kernel stack, per CPU | 16 KiB | `parameters.zig:26` |
+| IST stack, per CPU | 16 KiB | `parameters.zig:36` |
+| User stack, per task | 8 pages / 32 KiB | `parameters.zig:32` |
+| Max concurrent tasks | 32 | `parameters.zig:23` |
+| Boot page-table pool | 64 frames / 256 KiB | `efi.zig:299` |
+
+The 64 MiB heap cap plus kernel image, per-CPU stacks, task stacks, the frame
+bitmap, and DMA-contiguous allocations fit comfortably within 128 MiB on a
+single- or low-core-count machine. Very high core counts (toward the 128-CPU
+ceiling) add per-CPU stack overhead and push toward more RAM.
+
+**Note on the 4 GiB physmap:** the loader identity-maps and physmaps the low
+4 GiB of address space with 2 MiB leaves. This is *virtual address* reach, not a
+RAM requirement — RAM above 4 GiB simply needs an extra mapping window and is not
+needed to boot. (`efi.zig:305`)
+
+Virtual-memory layout (`boot-handoff.zig:47`):
+
+| Region | Base |
+|---|---|
+| User space | `0x0000_7000_0000_0000` |
+| Kernel heap | `0xFFFF_8000_0000_0000` |
+| Physmap | `0xFFFF_8800_0000_0000` |
+| Kernel image | `0xFFFF_FFFF_8000_0000` |
+
+## Buses & devices
+
+Buses with real drivers today:
+
+- **PCIe** via ECAM (`pci-bus`)
+- **xHCI USB** (`usb-xhci-bus`)
+- **PS/2** keyboard + mouse (`ps2-bus`) — the current input stack
+- **Serial UART** (16550/16450), configured from the ACPI SPCR table
+
+**No storage driver exists yet.** AHCI / NVMe / IDE are named for reporting only;
+there is no block-device driver. Persistent storage is future work.
+
+## IOMMU
+
+**Detection only; enforcement deferred.** The ACPI DMAR table is parsed for the
+first VT-d DRHD unit and its capabilities are exposed via `PlatformInfo`
+(`iommu_present`, `iommu_base`, `iommu_version`). No DMA-remapping tables are
+programmed and no translation is enforced. An IOMMU is therefore **not required**
+and does not currently constrain devices. (`system/devices/acpi.zig:96`)
+
+## What is explicitly NOT supported
+
+- Legacy BIOS / multiboot / limine boot
+- 32-bit x86
+- x2APIC
+- Legacy port-IO (CF8/CFC) PCI configuration
+- Non-xHCI USB (UHCI / OHCI / EHCI)
+- Machines without ACPI (no device discovery)
+- Persistent storage (no AHCI / NVMe / IDE driver yet)
+- USB HID input (PS/2 only for now)

docs/sysv.md

@@ -65,6 +65,38 @@ function-pointer type and the kernel's `_start` both carry
 whole reason `kernel_abi` lives in the shared contract — see [efi.md](efi.md) for
 the handoff it governs.
 
+## The process-entry stack (argc/argv)
+
+The SysV ABI also fixes what a *fresh process* finds on its stack — and danos
+follows it, so its own runtime and any future C libc read arguments the same way.
+At the first user instruction, `rsp` is 16-byte aligned and points at (addresses
+growing upward):
+
+```
+rsp →  argc                       u64
+       argv[0] … argv[argc-1]     pointers into the strings area below
+       NULL                       argv terminator
+       NULL                       envp terminator (no environment yet)
+       {AT_PAGESZ, page size}     auxiliary vector
+       {AT_NULL, 0}               auxiliary-vector terminator
+       argv string bytes          NUL-terminated
+       ─────────────────────────  stack top (stack_top_virtual)
+```
+
+The kernel builds this block at the top of the process's stack — 8 pages (32 KiB,
+`parameters.user_stack_pages`) mapped RW+NX below a fixed top, with the page below
+them left unmapped as a **guard**, so a stack overflow faults (killing only that
+process) instead of silently corrupting the image
+(`buildEntryStack` in `system/kernel/process.zig`); `argv[0]` is always the path
+or initial-ramdisk name the process was spawned as, and `system_spawn`'s optional
+argument blob becomes `argv[1..]`. The runtime's `_start`
+(`library/runtime/start.zig`) hands the block to `rt_start`, which builds a
+`runtime.process.Init` from it and passes that to the program's `main`
+(`pub fn main(init: runtime.process.Init)`; a parameterless `main()` is also
+accepted). A C runtime's `crt0` would walk
+the identical layout unmodified — that's the compatibility being bought. The
+`args` test proves the round trip.
+
 ## Where else it surfaces
 
 - **The red zone → `red_zone = false`.** `build.zig` disables the red zone for the

docs/timers.md

@@ -0,0 +1,117 @@
+# Timers and time
+
+Two different needs hide under the word "timer", and danos keeps them apart:
+
+- **Reading the clock** — *what time is it?* A read of a free-running counter.
+- **Waiting** — *wake me in N milliseconds*, or *notify me when a deadline passes.*
+
+Both are answered by the **kernel**, because the kernel already owns a timer: it has
+to, to preempt tasks. The LAPIC heartbeat and the calibrated TSC that back all of this
+are built in [device-interrupts.md](device-interrupts.md); the scheduler's blocking and
+wait queues are in [scheduling.md](scheduling.md). This page is about the surface a
+ring-3 program actually uses, and one deliberate absence: **there is no user-space time
+service.**
+
+## Why time is a syscall, not a service
+
+The tempting microkernel move is to put a timer *driver* in user space and have
+applications ask it for the time over IPC. For a **monotonic clock that is wrong** —
+reading `now()` should never cost an IPC round trip. The kernel is already holding the
+answer: it computes the current time every time it schedules, from the TSC, in a couple
+of instructions. Surfacing that as a system call is pure mechanism; routing it through a
+message to another process would be slower *and* redundant, and a device like the HPET
+(uncacheable MMIO reads) is a particularly bad thing to read on every `now()`.
+
+This is the same conclusion every serious system reaches: Linux and Zircon read the
+counter in the vDSO, L4 exposes a clock field in a shared kernel page, seL4 reads the
+cycle counter directly. None of them make a clock read an IPC. danos makes it a syscall.
+
+That "from the TSC" hides a portability question, because the TSC is only a valid clock
+when the CPU guarantees it is *invariant* and when every core's TSC is *synchronized*.
+danos checks both — the invariant-TSC CPUID bit (`0x80000007` EDX[8], set on Intel and
+AMD), and a cross-core "warp" check as the cores come up — and falls back to the HPET
+counter when either fails. So `now()` stays accurate on a real Intel box, a real AMD box,
+and inside a VM alike; only the source behind it differs. The mechanism is in
+[device-interrupts.md](device-interrupts.md).
+
+So the timer hardware lives in the kernel, and there is **no `hpet` driver and no time
+server** to consume. (An earlier HPET driver existed only to *demonstrate* the driver
+model; that role now lives in [drivers.md](drivers.md), as documentation.) The one place
+a user-space time service *is* justified — **wall-clock / calendar time** — is discussed
+at the end; it is deliberately not built yet.
+
+## The three system calls
+
+Time and waiting are three entries in the small syscall table ([syscall.md](syscall.md)):
+
+- **`clock` (#23)** → monotonic nanoseconds since boot. It only moves forward. Not
+  wall-clock: no date, no timezone. Backed by `architecture.nanos()` (TSC, scaled with a
+  128-bit intermediate so a long uptime can't overflow) — a few nanoseconds of
+  resolution, and just an `rdtsc` plus a multiply.
+- **`sleep` (#3)** → block the caller for N milliseconds. The scheduler records a wake
+  deadline and the tick sweep wakes it (`scheduler.sleep`).
+- **`timer_bind` (#31)** → arm a one-shot timer that, after N milliseconds, posts a
+  **timer notification** to an IPC endpoint. Unlike `sleep` it does **not** block: a
+  service can keep answering messages on the same endpoint while a deadline is pending.
+  This is the timed wait that stop-sequence escalation, hello deadlines, and restart
+  backoff are built from ([process-lifecycle.md](process-lifecycle.md),
+  [device-manager.md](device-manager.md)).
+
+The kernel's own scheduling timer (the LAPIC, vector 32) is never exposed to user space;
+programs read the TSC through `clock` and get timed wakeups through `sleep`/`timer_bind`,
+both riding the scheduler tick.
+
+## `runtime.time` — the generic interface
+
+Applications don't call the syscalls directly; they use `runtime.time`
+(`library/runtime/time.zig`), a thin `Instant`/`Duration` layer over them — an ergonomic
+front door, not new mechanism.
+
+```zig
+const time = @import("runtime").time;
+
+const start = time.now();                 // Instant — monotonic
+doWork();
+const took = start.elapsed();             // Duration
+time.sleep(time.Duration.fromMillis(5));  // block ~5 ms
+
+// A deadline delivered as a notification, so a service keeps serving meanwhile:
+_ = time.after(endpoint, time.Duration.fromMillis(200));
+```
+
+- `Duration` is nanoseconds under the hood, with `fromNanos/fromMicros/fromMillis/
+  fromSeconds` and `asNanos/asMillis`. `ceilMillis` rounds *up* to the kernel's
+  millisecond granularity, so a sub-millisecond `sleep` never rounds down to zero and
+  returns early. All arithmetic saturates rather than wraps.
+- `Instant` is a point on the monotonic clock: `since`, `elapsed`, `plus`, `reached` —
+  built for deadline loops (`while (!deadline.reached()) …`).
+- `now()` / `monotonicNanos()` wrap `clock`. `available()` reports whether the clock is
+  calibrated at all (the kernel returns 0 until the TSC frequency is known, so a caller
+  that needs real time can treat 0 as "unavailable" rather than assume it advances).
+- `sleep(d)` wraps `sleep`; `spin(d)` busy-polls `now()` for the sub-millisecond delays
+  the millisecond tick can't express; `after(endpoint, d)` wraps `timer_bind`.
+
+The raw wrappers (`system.clock`, `system.sleep`, `system.timerOnce`) stay in
+`library/runtime/system.zig`; `runtime.time` is the layer meant for everyday use.
+
+## Wall-clock time (not built)
+
+Everything above is **monotonic**: elapsed time since boot, perfect for timeouts and
+measurement, useless for "what is the date?" Calendar time — a real-time clock, time
+zones, leap seconds — is genuinely a **user-space** concern, and it *is* the case a time
+service is for. It would be backed by an **RTC** driver (the CMOS real-time clock), not
+the HPET, and exposed as a `CLOCK_REALTIME`-style service alongside the monotonic
+syscall. It is deferred until something needs it; the monotonic clock the kernel already
+owns covers every current use.
+
+## Verifying it
+
+`runtime.time`'s `Instant`/`Duration` arithmetic has unit tests that run on the host:
+
+```
+$ zig build test        # includes library/runtime/time.zig
+```
+
+End to end, the proof the clock is real is that it *advances*: read `now()`, `sleep` a
+`Duration`, read `now()` again, and the second reading is later — the kernel's timer
+driving a ring-3 program with no service in between.

docs/zig-self-hosting.md

@@ -0,0 +1,349 @@
+# Running Zig on danos: the self-hosting roadmap
+
+A design note (not built yet) on the path to making danos a **real Zig target** — a
+target you can name (`-target x86_64-danos`) and, eventually, run the Zig compiler
+itself on. It is forward-looking, like [vision.md](vision.md): it sets a direction
+and the decisions that follow from it, so the code we write now bends toward it
+instead of away.
+
+This note deliberately does **not** cover a text editor or terminal. Those are
+easier (single-process, I/O-bound) and fall out of the early phases here almost for
+free; the hard, shaping problem is the standard-library surface, so that is what
+this roadmap is about.
+
+The analysis behind it was done against **Zig 0.16** (the pinned toolchain). Zig's
+standard library moves between releases — especially the parts described here — so
+treat upstream references as "the shape in 0.16.x," and expect to re-check them on a
+toolchain bump.
+
+## The win condition
+
+danos runs the Zig compiler when a bare
+
+```
+zig build-exe hello.zig
+```
+
+completes **on danos** and produces a runnable danos binary. Note the milestone is
+`build-exe`, not `zig build`: the `zig build` runner spawns child processes (the
+build steps), which needs a whole process-control surface danos does not have yet.
+A single `build-exe` needs none of that (see Phase 3). Reaching `build-exe` is
+"self-hosting"; reaching `zig build` is a later, separate lift.
+
+### Non-goals
+
+- **No Linux syscall/ABI emulation.** danos will not implement the Linux `syscall`
+  interface so that stock `x86_64-linux` binaries run. That is a permanent
+  compatibility treadmill and it inverts the microkernel design — explicitly out.
+- **No musl port yet.** A musl libc port is a reasonable *later* effort (it unlocks
+  the C ecosystem), but it is not on the critical path to Zig-on-danos, and it is
+  deferred. The roadmap below is arranged so the work still pays off if musl ever
+  happens (see "The same surface, twice").
+- **Editor/terminal are out of scope for this note** (they are downstream of Phase 1).
+
+**On FFI.** Foreign-function interop splits the same way as the doors below. Zig-level
+and C-ABI-*exposing* FFI (`extern`, `callconv(.c)`, C-ABI structs) work on a real target
+immediately — and the `std.os.danos` seam is C-ABI-shaped by construction, so it is
+FFI-friendly from the start. *Consuming* C libraries (`@cImport`, linking archives) is
+the part that needs a libc + headers, i.e. the deferred musl door. So an eventual FFI
+need reinforces keeping that door open; it does not change the plan.
+
+## The realization that shapes everything: 0.16 gives us *one* seam
+
+The instinct "to target Zig we'd have to reimplement all the `std` namespaces" was
+how older Zig worked. Zig 0.16 (post-"writergate") is far kinder:
+
+- **`std.fs` is essentially gone.** It is now path helpers plus deprecated aliases;
+  there is no `std.fs.File`, `std.fs.Dir`, or `std.fs.cwd()`. File and directory
+  work goes through **`std.Io`** — a single runtime **vtable** (`Io.zig`) of
+  function pointers handed to `main` as `std.process.Init.io`. `std.Io.File` and
+  `std.Io.Dir` are thin forwarders to that vtable. `Io.zig` and the `fs` shim carry
+  **zero** per-OS branches.
+- **`std.posix` is one generic body** parameterised over a single `system` module.
+  With no libc, `system` resolves **per target OS**: `.linux => std.os.linux`,
+  `.plan9 => std.os.plan9`, and so on. The generic `std.posix.read`/`write`/`open`
+  bodies are just `system.read(...)` plus an errno switch — *identical for every
+  OS*. The only variable is what `system` binds to.
+- **`std.os.<tag>`** (e.g. `std/os/linux.zig`) is therefore the real porting seam: a
+  low-level, C-ABI-shaped module of `read/write/open/close/lseek/mmap/clock/exit/…`
+  plus an `errno` enum and the constant tables (`O_*`, `CLOCK_*`, `S_*`).
+
+Put together: **to port danos we write `std.os.danos` once** — the ~30-operation
+seam — and the whole `std.posix` / `std.fs` / `std.Io` tower above it lights up
+generically, because none of it branches on the OS. That is a dramatically smaller
+and more contained target than "reimplement the namespaces."
+
+## Three doors, and why we take the first
+
+| Door | What it is | Verdict |
+|------|-----------|---------|
+| **1. Implement the std seam** (`std.os.danos`) | Write the ~30-op `system` module over danos's native ABI + VFS; the generic std tower lights up. | **Take this.** The only door that touches neither C nor the Linux ABI. |
+| **2. Port musl** | Port musl libc to danos, link Zig against it. | Defer. Good later for the *C* ecosystem; barely helps *Zig* (std only uses libc on the libc-linked path). |
+| **3. Emulate the Linux ABI** | Implement Linux syscalls so stock linux binaries run. | Reject. Bottomless compatibility treadmill; against the design. |
+
+### The same surface, twice
+
+Doors 1 and 2 are the **same native surface at different layers**. `std.posix.read`
+is `system.read(...)` + an errno switch *regardless of OS* — the only question is
+whether `system` is **`std.os.danos` (Zig)** or **musl (C)**. Either way, the set of
+danos-facing operations you must implement is the *same* ~30 ops, all bottoming out
+in danos's native syscalls + the VFS/FAT server.
+
+So the runtime work below is **not throwaway** if musl ever happens: you are building
+the danos-native implementations of that surface either way. Door 1 just packages
+them as Zig; a future musl re-uses the identical kernel/VFS operations underneath. The
+two symmetries worth keeping in mind: doors 1 and 2 converge at the **top** (identical
+POSIX surface); doors 2 and 3 converge at the **bottom** (unmodified musl needs the
+Linux syscall ABI). Door 1 is the only one that avoids both C and Linux.
+
+### A fork is table stakes — for any door
+
+`std.Target.Os.Tag` is a **closed enum** baked into the compiler binary *and* into
+the `std` linked with every program; `-target x86_64-danos` resolves through it. So
+adding `danos` as a name requires patching and rebuilding the compiler — even the
+musl door needs this. "Fork Zig" is therefore not an extra cost unique to door 1; it
+is the price of admission for *any* real target. What door 1 adds on top is small and
+localised (below).
+
+## The architecture decision: `runtime.os` + `runtime.fs`, and retire `posix`
+
+danos already has the right split ([the private-ABI boundary](../README.md)): the
+kernel exposes a minimal syscall ABI ([syscall.md](syscall.md)); the **`runtime`**
+library is the stable, danos-native application ABI. What this roadmap adds:
+
+- **`runtime.os` — the seam.** A C-ABI-shaped module of the ~30 operations
+  (`read/write/open/close/lseek/mmap/munmap/clock/exit/…`) + an errno enum + the
+  constant tables, each backed by danos's native syscalls and the VFS. **Structure it
+  to mirror `std/os/linux.zig`.** This is the load-bearing, *non-throwaway* artifact:
+  when we fork Zig, `runtime.os` is copy-pasted (near-verbatim) into `std.os.danos`.
+- **`runtime.fs` — the thin native file API** danos programs use *today*, layered
+  over `runtime.os`. It is also the concrete backing for the `std.Io` vtable's
+  file-write entry once we're a real target, which is why program stdout, diagnostics,
+  and file writes should all be *decided once at that seam* rather than as bespoke
+  per-call helpers (see "How this informs decisions now").
+
+**Do not hand-mirror the high-level std namespaces.** `std.fs`/`std.Io`/`std.process`
+are generic and OS-agnostic; once `std.os.danos` exists and we fork, upstream *gives*
+them to danos for free. Hand-writing `runtime.std.fs` to imitate them would be
+redundant the day the fork works, and it would chase a moving target (0.16's `std.Io`
+is large and still shifting). Build the seam well; take the tower for free.
+
+**Why not a library called `std`?** Because `@import("std")` resolves to the
+compiler-provided standard library; a user module named `std` would *shadow* it for
+anything that imports it that way. That is the real reason the seam lives *inside* a
+forked std as `std/os/danos.zig`, not as a `runtime.std` library — and why danos's end
+state (`@import("std")` just working, and knowing danos) is the most natively Zig it can
+be. `runtime.os` is only the interim staging ground: developed against the stock
+toolchain so Phase 1 need not wait on the fork, then promoted near-verbatim into the
+fork's `std/os/danos.zig`.
+
+### Retire `library/posix`
+
+The `posix` compatibility layer (`unistd`, `stdio`) was the right instinct too early.
+Its whole value is POSIX *spellings* for POSIX software — and danos has no POSIX
+software; every current caller is danos-native code that could use `runtime.fs`
+directly. The real POSIX story arrives later and from elsewhere (musl, or upstream
+`std`'s own posix over `std.os.danos`), which supersedes a hand-rolled shim. So it is
+premature abstraction that adds a "which layer do I use?" fork with no payoff yet.
+
+Its footprint is tiny: **five** call sites, all `unistd` file operations —
+`system/services/fat/fat.zig` (`mount`), the `vfs-test` and `fat-test` clients, and
+(from the boot-log work) `init.zig` and `log-flush.zig`. `stdio.zig` is dead — nothing
+imports it. The plan: build `runtime.fs`, migrate those five to it, delete
+`library/posix/`, and drop the `posix` module from `build.zig`'s `addUserBinary`.
+
+## Where danos stands: coverage vs. the gaps
+
+What the seam needs, and what danos already provides:
+
+| std need | danos today | Gap |
+|----------|-------------|-----|
+| open / read / write / close / lseek | VFS (via the current `unistd`, → `runtime.fs`) | none — repackage |
+| directory read (`getdents`) | VFS `readdir` | none — repackage |
+| mmap / munmap | native syscalls ([abi.zig](../system/abi.zig)) | none |
+| page allocator | over `mmap`, via `root.os.heap.page_allocator` override | ~30-line hook |
+| monotonic clock | `clock` syscall | none |
+| args / argv | SysV entry stack ([sysv.md](sysv.md)), `runtime.process.Init` | none |
+| stdout / stderr | `debug_write` today | wire fd 1/2 to a console **byte** stream |
+| mkdir / unlink / rename / truncate | done — engine + VFS + `runtime.fs` (Phase 2) | — |
+| stat fields | `{size, kind, mtime}` | **mode / inode** still missing (cache validity) |
+| wall-clock / realtime | done — `wall_clock` syscall (CMOS RTC, Phase 2d) | — |
+| **environment variables** | `Init` has no env field | missing (can start empty) |
+| **cwd / chdir** | paths are absolute or bare | missing (no cwd anchor) |
+| **entropy / random** | — | missing (needed behind `vtable.random`) |
+| process spawn + exit status | `system_spawn` starts a *named ramdisk binary*; `ExitReason` is a *category* | no exec-of-path, no numeric `WEXITSTATUS` |
+| threads | one thread per process | avoided via `-fsingle-threaded` (below) |
+| symlinks | `NodeKind` has the tag; unimplemented | low priority |
+
+The clustering is clear: reads and memory are basically done; the real work is
+**filesystem mutation + richer stat + wall-clock**, and a few small seam pieces
+(page-allocator hook, stdio bytes, entropy). Process spawning and threads are
+side-stepped entirely for a single `build-exe`.
+
+## The roadmap
+
+### Phase 0 — Make `danos` a real target
+
+**Host, target, self-host — keep the three roles straight.** The *host* is where the
+compiler runs (your mac + linux dev machines); the *target* is what it emits (`danos`);
+and eventually danos becomes a host too (self-hosting — the win condition). So the move
+is: fork the compiler, build it **for** your dev hosts, and teach it to **cross-compile
+to** danos. You already do this — danos is cross-compiled `freestanding` from your dev
+host today; Phase 0 swaps that `freestanding` target for a real `x86_64-danos` one, which
+is what unlocks the native `std`.
+
+**Why a compiler fork, not just a `--zig-lib-dir` override.** `std.Target.Os.Tag` is a
+*closed enum compiled into the compiler binary*, so `-target x86_64-danos` will not even
+parse unless the compiler itself knows the tag. Overriding the std lib directory alone
+cannot add a target — and there is no libc-only shortcut (a future musl needs the same
+patch). The only alternative, staying on `freestanding` + hand-shims, is exactly the
+non-native feel we are leaving: `@import("std")` there is stubbed, not real.
+
+**The fork.** Clone `ziglang/zig` at the pinned 0.16 tag; build it with a stock
+same-version `zig` (`zig build` in the tree — a standard, LLVM-pulling, roughly one-time
+build); point danos's `build.zig`/CI at the resulting binary. Four localised patches:
+
+- add `danos` to `std.Target.Os.Tag`, in the "no version range" group alongside
+  plan9/serenity;
+- add `danos` to the freestanding/other **no-op `_start` list** in `std`'s `start.zig`,
+  so std does *not* emit its own System-V `_start` — danos keeps owning the entry shim
+  and `Init`/argv construction it already builds ([sysv.md](sysv.md));
+- wire the `system` selector `.danos => std.os.danos` in `std.posix`;
+- add `std/os/danos.zig` — **the seam itself**, promoted near-verbatim from the
+  `runtime.os` developed first in Phase 1 (against the stock toolchain, so the fork is
+  not a prerequisite for starting).
+
+This is the fork treadmill we accept once. Keep the patch set tiny and `else`-friendly,
+pin to one 0.16.x, and rebase on point releases.
+
+### Phase 1 — `runtime.os` read-side + allocator + stdio + cwd; retire `posix`
+
+Author `runtime.os` (→ `std.os.danos`): the `errno` enum, the constant tables, and
+the C-convention `read / write / open / openat / close / lseek / mmap / munmap /
+exit`, each returning result-or-`-errno`. Most backing already exists (VFS + native
+mmap + clock).
+
+- Provide `page_allocator` via `root.os.heap.page_allocator` (a thin override over
+  danos `mmap`). This sits **outside** the `std.Io` vtable, so it is wired separately.
+- Wire fd 0/1/2 to a console **byte** stream (today output only reaches `debug_write`;
+  input is structured `InputEvent` IPC — a byte tty is a new, small thing in both
+  directions).
+- Add a `getcwd`/`chdir` anchor so `std.fs.cwd()`-style resolution has something to
+  resolve against.
+- Build `runtime.fs` over `runtime.os`; migrate the five `posix` callers to it; delete
+  `library/posix/` and drop its build module.
+
+After Phase 1, the surface an editor or terminal needs (open/read/write/close/lseek/
+readdir/isatty/args/exit) exists. Those are downstream and out of scope here.
+
+### Phase 2 — Filesystem mutation + real stat (the compiler's cache tower)
+
+danos's biggest genuine gap, and the correctness-critical one:
+
+- Add **mkdir / unlink / rename / truncate** to *both* the VFS wire protocol
+  ([protocol.zig](../system/services/vfs/protocol.zig)) and the FAT engine
+  ([engine.zig](../system/services/fat/engine.zig)), then expose them via `runtime.os`.
+- Extend `stat` beyond `{size, kind}` to carry **mtime + inode + mode** — `std`'s file
+  stat needs them for build-cache validity — which in turn needs **wall-clock** time
+  (danos is monotonic-only today; an RTC/time service is the dependency).
+
+Because `std.fs`/`std.Io` have no per-OS branches, finishing this in `runtime.os`
+lights up the whole file tower for the compiler at once. Environment can stay an empty
+map until the kernel populates a non-empty `envp`.
+
+**Status — Phase 2 complete.** `truncate` (O_TRUNC, closing the boot-log stale-tail
+bug), `mkdir`, `unlink`, and `rename` are all wired through the FAT engine, the VFS
+protocol + router, and `runtime.fs` (`makeDirectory` / `remove` / `rename`) —
+host-tested and QEMU-tested (`fat-mutations` + `fat-rename` make a directory, write+read
+a file in it, rename it, then remove it through the mount). `removeFile` and `rename`
+are LFN-aware; `rename` is same-directory + 8.3 (cross-directory and long-name-
+preserving rename are noted limitations). Wall-clock is now a kernel syscall
+(`wall_clock`, a CMOS-RTC read anchored to the monotonic clock), and the FAT engine
+stamps and reports **mtime** — `stat` / `runtime.fs.Attributes` carry a real
+modification time (the `fat-mtime` case reads it back within seconds of the host clock).
+The remaining `stat` fields, `mode`/`inode`, are deferred (not needed until the
+compiler's cache layer wants them). **Everything past here is gated on Phase 0 (the
+fork):** the `runtime.os` seam, `cwd`, stdio-as-fds, and the compiler bring-up.
+
+### Phase 3 — Single-threaded, self-linked compiler bring-up
+
+Build the compiler with **two load-bearing flags**:
+
+- **`-fsingle-threaded`** removes `std.Thread` entirely — `Thread.spawn` is a hard
+  compile error under it, and `std.Io`'s threaded backend runs inline. danos being
+  one-thread-per-process is therefore **not** a blocker. Parallel codegen is a
+  throughput optimisation, not a correctness requirement.
+- **`-fno-llvm -fno-lld`** keeps codegen and linking **in-process** (the self-hosted
+  x86-64 backend + self-linker), so a single `build-exe` **never forks a child**. That
+  is what lets us defer the entire spawn/exec/wait surface.
+
+Then supply the few remaining seam pieces: `now` (wrap the danos clock), an entropy
+source behind `vtable.random` (`randomSecure` can alias it initially — low volume, for
+temp-file names and hashmap seeds), and the Phase-2 mkdir/rename/unlink for cache dir
+trees and atomic temp-then-rename output.
+
+**Explicitly deferred** (not on the `build-exe` path): child-process spawn/exec (only
+`zig build` and external tools need it), `std.Thread`, `fsync` (FAT is write-through
+today), symlinks, and musl.
+
+## Risks and gotchas
+
+- **The std-fork rebase treadmill is the main ongoing cost.** A new OS tag touches the
+  same broad file set plan9/serenity touch (hundreds of `native_os` sites, plus
+  "unsupported OS" `@compileError` dead-ends a new tag must be routed around), and the
+  entire `std.Io` layer is new in 0.16 and still moving. Stay pinned to one 0.16.x,
+  keep additions localised and `else`-friendly. Watch the closed-enum gotcha: adding
+  `danos` to `Os.Tag` can break existing *exhaustive* switches that lack an `else`, so
+  expect to touch switch sites beyond the ones you implement.
+- **Single-threaded is load-bearing.** The "no `std.Thread`" simplification rests
+  entirely on `-fsingle-threaded`. If a dependency or flag flips threading back on, you
+  inherit an unescapable compile error (no root-hook exists) — the only outs are a full
+  thread-impl fork or linking libc for pthreads. Keep `single_threaded` asserted end to
+  end.
+- **In-process linking is load-bearing.** Reaching the compiler without fork/exec
+  depends on `-fno-llvm -fno-lld`. The moment you shell out to LLD/`ld`, you need the
+  full `spawn`/`wait` surface — the hardest microkernel piece — and danos's
+  `system_spawn` only starts a *named ramdisk binary*, not exec of an arbitrary path.
+  Verify the self-hosted backend covers the target output before assuming child
+  processes are optional.
+- **The shim cannot host the compiler.** danos's current `runtime`/`posix` is fine for
+  danos's *own* native programs, but the compiler `import`s *upstream* `std`, which on
+  a non-target hits the void `system` stub. So the compiler forces the real target
+  (Phase 0's fork). Do not over-invest in extending the hand-shim for compiler
+  purposes; put that effort into `runtime.os` + the VFS/FAT operations, which both the
+  fork *and* a future musl consume.
+- **`"w"`/`O_CREAT` does not truncate — a silent-corruption bug on this road.** The FAT
+  engine's `writeFile` only *grows* `node.size`, so overwriting a shorter file leaves
+  trailing garbage. Harmless for the boot log today, but for a compiler it means
+  **corrupt `.o`/cache files that look like nondeterministic compiler bugs.** Land
+  `truncate` (Phase 2) before the compiler ever writes cache.
+- **Exit status is categorical, not numeric.** `process_exit_reason` returns an
+  `ExitReason` *category*, not a numeric code (`WEXITSTATUS`). Fine while spawn is
+  stubbed; the day `zig build` or external tools arrive, plan a kernel exit-record
+  extension — do not let it surprise you.
+
+## How this informs decisions now
+
+Two current decisions fall out of this roadmap:
+
+1. **The `runtime.fs` / `std.Io` question resolves at the vtable seam.** Because 0.16
+   routes *all* output through the `std.Io` vtable's file-write entry, and stdout/stderr
+   are just `File`s with well-known handles, build `runtime.fs` (and the console stdout)
+   as the concrete backing for that entry — not as a bespoke `std.Io.Writer`-only shim.
+   Decide it once, at the seam, and program stdout, diagnostics, and file writes all
+   flow through the same danos VFS/console path.
+2. **The boot-log `truncate` caveat is now fixed** (Phase 2a). It was the same
+   `writeFile`-only-grows gap that on the self-hosting road would corrupt build output;
+   `engine.truncate` + an O_TRUNC open flag now free the old chain so a shorter rewrite
+   leaves no stale tail, and the boot-log flush opens with it.
+
+## Related
+
+- [vision.md](vision.md) — the north star this serves.
+- [syscall.md](syscall.md) — the kernel↔runtime ABI `runtime.os` is built on.
+- [sysv.md](sysv.md) — the entry stack (`argc/argv/envp/auxv`) danos already constructs.
+- [ipc.md](ipc.md) — the IPC the VFS/FAT operations travel over.
+- [danos-file-system-hierarchy-FSH.md](danos-file-system-hierarchy-FSH.md) — the
+  filesystem layout the file surface serves.
+- [coding-standards.md](coding-standards.md) — danos naming (why the compat spellings
+  are confined, and now retired).

library/posix/posix.zig

@@ -1,13 +0,0 @@
-//! DanOS's POSIX / C compatibility layer — `unistd`, `stdio`, and (later) the C
-//! `errno` / `struct stat` / `extern "C"` surface. This is the *one* place POSIX and
-//! C spellings are allowed to appear verbatim (see docs/coding-standards.md): a file
-//! under library/posix/ *is* the foreign ABI, so it keeps the ABI's names. Everything
-//! it touches on the danos side (the VFS protocol, the runtime) uses danos names,
-//! which this layer translates to at the boundary.
-//!
-//! It is layered strictly *over* the runtime: it calls the runtime's IPC and heap,
-//! never the kernel's system calls directly. danos-native applications use the
-//! runtime; this exists so *POSIX* software can too.
-
-pub const unistd = @import("unistd.zig");
-pub const stdio = @import("stdio.zig");

library/posix/stdio.zig

@@ -1,115 +0,0 @@
-//! A small C stdio layer over the POSIX-style file API (unistd.zig). Unbuffered
-//! for now — each fread/fwrite is one VFS round trip; an internal buffer (fewer
-//! IPC calls) is a later optimisation. Both a Zig-callable API and `extern "C"`
-//! symbols are provided, so Zig and future C programs share it.
-
-const std = @import("std");
-const unistd = @import("unistd.zig");
-const heap = @import("runtime").heap;
-
-pub const SEEK_SET = unistd.SEEK_SET;
-pub const SEEK_CURRENT = unistd.SEEK_CURRENT;
-pub const SEEK_END = unistd.SEEK_END;
-
-/// A C `FILE`: an fd plus sticky end-of-file / error flags. Allocated on the
-/// heap; `fclose` frees it.
-pub const FILE = extern struct {
-    fd: i32,
-    eof: c_int = 0,
-    err: c_int = 0,
-};
-
-fn flagsFor(mode: []const u8) u32 {
-    if (mode.len == 0) return 0;
-    return switch (mode[0]) {
-        'w', 'a' => unistd.O_CREAT,
-        else => 0,
-    };
-}
-
-/// Open `path` in `mode` ("r"/"w"/"a", '+' ignored for now). Returns null on error.
-pub fn fopen(path: []const u8, mode: []const u8) ?*FILE {
-    const fd = unistd.open(path, flagsFor(mode));
-    if (fd < 0) return null;
-    const f = heap.allocator().create(FILE) catch {
-        unistd.close(fd);
-        return null;
-    };
-    f.* = .{ .fd = fd };
-    if (mode.len > 0 and mode[0] == 'a') _ = unistd.lseek(fd, 0, unistd.SEEK_END);
-    return f;
-}
-
-pub fn fclose(f: *FILE) c_int {
-    unistd.close(f.fd);
-    heap.allocator().destroy(f);
-    return 0;
-}
-
-/// Read `size*nmemb` bytes; returns the number of whole items read.
-pub fn fread(buffer: []u8, size: usize, nmemb: usize, f: *FILE) usize {
-    const total = size * nmemb;
-    if (total == 0) return 0;
-    const n = unistd.read(f.fd, buffer[0..@min(buffer.len, total)]);
-    if (n <= 0) {
-        f.eof = 1;
-        return 0;
-    }
-    return @as(usize, @intCast(n)) / size;
-}
-
-/// Write `size*nmemb` bytes; returns the number of whole items written.
-pub fn fwrite(data: []const u8, size: usize, nmemb: usize, f: *FILE) usize {
-    const total = @min(data.len, size * nmemb);
-    if (total == 0) return 0;
-    const n = unistd.write(f.fd, data[0..total]);
-    if (n <= 0) {
-        f.err = 1;
-        return 0;
-    }
-    return @as(usize, @intCast(n)) / size;
-}
-
-pub fn fseek(f: *FILE, off: i64, whence: u32) c_int {
-    f.eof = 0;
-    return if (unistd.lseek(f.fd, off, whence) < 0) -1 else 0;
-}
-
-pub fn ftell(f: *FILE) i64 {
-    return unistd.lseek(f.fd, 0, unistd.SEEK_CURRENT);
-}
-
-pub fn rewind(f: *FILE) void {
-    _ = fseek(f, 0, SEEK_SET);
-}
-
-pub fn feof(f: *FILE) c_int {
-    return f.eof;
-}
-
-pub fn ferror(f: *FILE) c_int {
-    return f.err;
-}
-
-pub fn fputs(s: []const u8, f: *FILE) c_int {
-    return if (unistd.write(f.fd, s) < 0) -1 else 0;
-}
-
-pub fn fputc(c: u8, f: *FILE) c_int {
-    const b = [_]u8{c};
-    return if (unistd.write(f.fd, &b) == 1) c else -1;
-}
-
-pub fn fgetc(f: *FILE) c_int {
-    var b: [1]u8 = undefined;
-    const n = unistd.read(f.fd, &b);
-    if (n <= 0) {
-        f.eof = 1;
-        return -1; // EOF
-    }
-    return b[0];
-}
-
-// Real `extern "C"` symbols (fopen/fread/fseek/...) — with a C-string signature
-// distinct from the Zig slice API above — land with the first C program, wired
-// via @export so they don't collide with these Zig names.

library/posix/unistd.zig

@@ -1,148 +0,0 @@
-//! POSIX-style file API for user programs — the low level under C stdio. Files
-//! are named objects served by the user-space VFS server (system/services/vfs/vfs.zig); each
-//! call marshals a request, IPC_Calls the VFS, and unmarshals the reply. The
-//! kernel knows nothing of files or fds — the fd table lives here, per process.
-
-const std = @import("std");
-const protocol = @import("vfs-protocol");
-const ipc = @import("runtime").ipc;
-
-pub const O_CREAT = protocol.create;
-pub const SEEK_SET: u32 = 0;
-pub const SEEK_CURRENT: u32 = 1;
-pub const SEEK_END: u32 = 2;
-
-// Resolve (and cache) the VFS server endpoint, looked up by well-known id.
-var vfs_handle: usize = 0;
-var vfs_resolved = false;
-fn vfs() ?usize {
-    if (!vfs_resolved) {
-        vfs_handle = ipc.lookup(.vfs) orelse return null;
-        vfs_resolved = true;
-    }
-    return vfs_handle;
-}
-
-const maximum_fds = 32;
-const Fd = struct { used: bool = false, node: u64 = 0, offset: u64 = 0 };
-var fds = [_]Fd{.{}} ** maximum_fds;
-
-fn allocFd() ?usize {
-    for (&fds, 0..) |*f, i| {
-        if (!f.used) {
-            f.* = .{ .used = true };
-            return i;
-        }
-    }
-    return null;
-}
-
-const Result = struct { reply: protocol.Reply, payload: []u8 };
-
-/// One request/reply round trip: [Request header][send payload] -> VFS ->
-/// [Reply header][receive payload]. The receive payload is written into `out`.
-fn transact(request: protocol.Request, send: []const u8, out: []u8) ?Result {
-    const h = vfs() orelse return null;
-    var message: [protocol.message_maximum]u8 = undefined;
-    @memcpy(message[0..protocol.request_size], std.mem.asBytes(&request));
-    const slen = @min(send.len, protocol.maximum_payload);
-    @memcpy(message[protocol.request_size..][0..slen], send[0..slen]);
-
-    var rbuf: [protocol.message_maximum]u8 = undefined;
-    const n = ipc.call(h, message[0 .. protocol.request_size + slen], &rbuf) catch return null;
-    if (n < protocol.reply_size) return null;
-    const reply = std.mem.bytesToValue(protocol.Reply, rbuf[0..protocol.reply_size]);
-    const rpl = @min(n - protocol.reply_size, out.len);
-    @memcpy(out[0..rpl], rbuf[protocol.reply_size..][0..rpl]);
-    return .{ .reply = reply, .payload = out[0..rpl] };
-}
-
-/// Open (or create, with O_CREAT) `path`; returns an fd or -1.
-pub fn open(path: []const u8, flags: u32) i32 {
-    const fd = allocFd() orelse return -1;
-    const request = protocol.Request{ .operation = .open, .node = 0, .offset = 0, .len = @intCast(path.len), .flags = flags };
-    const r = transact(request, path, &.{}) orelse {
-        fds[fd].used = false;
-        return -1;
-    };
-    if (r.reply.status != 0) {
-        fds[fd].used = false;
-        return -1;
-    }
-    fds[fd] = .{ .used = true, .node = r.reply.node, .offset = 0 };
-    return @intCast(fd);
-}
-
-fn fdPtr(fd: i32) ?*Fd {
-    if (fd < 0 or fd >= maximum_fds) return null;
-    const f = &fds[@intCast(fd)];
-    return if (f.used) f else null;
-}
-
-/// Read up to `buffer.len` bytes at the current offset; returns the count or -1.
-pub fn read(fd: i32, buffer: []u8) isize {
-    const f = fdPtr(fd) orelse return -1;
-    const want: u32 = @intCast(@min(buffer.len, protocol.maximum_payload));
-    const request = protocol.Request{ .operation = .read, .node = f.node, .offset = f.offset, .len = want, .flags = 0 };
-    const r = transact(request, &.{}, buffer) orelse return -1;
-    if (r.reply.status != 0) return -1;
-    f.offset += r.reply.len;
-    return @intCast(r.reply.len);
-}
-
-/// Write `data` at the current offset; returns the count or -1.
-pub fn write(fd: i32, data: []const u8) isize {
-    const f = fdPtr(fd) orelse return -1;
-    const want: u32 = @intCast(@min(data.len, protocol.maximum_payload));
-    const request = protocol.Request{ .operation = .write, .node = f.node, .offset = f.offset, .len = want, .flags = 0 };
-    const r = transact(request, data[0..want], &.{}) orelse return -1;
-    if (r.reply.status != 0) return -1;
-    f.offset += r.reply.len;
-    return @intCast(r.reply.len);
-}
-
-/// Reposition the fd's offset. Returns the new offset or -1. (SEEK_END needs the
-/// file size, which `stat` provides; handled by fetching it here.)
-pub fn lseek(fd: i32, off: i64, whence: u32) i64 {
-    const f = fdPtr(fd) orelse return -1;
-    const base: i64 = switch (whence) {
-        SEEK_SET => 0,
-        SEEK_CURRENT => @intCast(f.offset),
-        SEEK_END => blk: {
-            const request = protocol.Request{ .operation = .status, .node = f.node, .offset = 0, .len = 0, .flags = 0 };
-            var sbuf: [@sizeOf(protocol.FileStatus)]u8 = undefined;
-            const r = transact(request, &.{}, &sbuf) orelse return -1;
-            if (r.reply.status != 0 or r.payload.len < @sizeOf(protocol.FileStatus)) return -1;
-            const st = std.mem.bytesToValue(protocol.FileStatus, sbuf[0..@sizeOf(protocol.FileStatus)]);
-            break :blk @intCast(st.size);
-        },
-        else => return -1,
-    };
-    const pos = base + off;
-    if (pos < 0) return -1;
-    f.offset = @intCast(pos);
-    return pos;
-}
-
-/// Stat `path`. Returns 0 or -1.
-pub fn stat(path: []const u8, out: *protocol.FileStatus) i32 {
-    // Open, stat by node, close — simple and enough for now.
-    const fd = open(path, 0);
-    if (fd < 0) return -1;
-    defer close(fd);
-    const f = fdPtr(fd).?;
-    const request = protocol.Request{ .operation = .status, .node = f.node, .offset = 0, .len = 0, .flags = 0 };
-    var sbuf: [@sizeOf(protocol.FileStatus)]u8 = undefined;
-    const r = transact(request, &.{}, &sbuf) orelse return -1;
-    if (r.reply.status != 0 or r.payload.len < @sizeOf(protocol.FileStatus)) return -1;
-    out.* = std.mem.bytesToValue(protocol.FileStatus, sbuf[0..@sizeOf(protocol.FileStatus)]);
-    return 0;
-}
-
-/// Close an fd (best effort — tells the VFS to release the open file).
-pub fn close(fd: i32) void {
-    const f = fdPtr(fd) orelse return;
-    const request = protocol.Request{ .operation = .close, .node = f.node, .offset = 0, .len = 0, .flags = 0 };
-    _ = transact(request, &.{}, &.{});
-    f.used = false;
-}

library/runtime/block.zig

@@ -0,0 +1,62 @@
+//! Block-device client: the helper a filesystem uses to read and write a block
+//! device (a USB stick, via usb-storage) without hand-rolling the block-protocol
+//! IPC. Layered over `ipc` and the shared `block-protocol` wire format, like
+//! `runtime.usb` over the transfer protocol.
+//!
+//! Transfers name a caller-owned DMA buffer by physical address (from
+//! `runtime.dma.alloc`), so whole sectors move without crossing the IPC size
+//! limit — the same handoff usb-storage uses toward the controller.
+
+const std = @import("std");
+const ipc = @import("ipc.zig");
+const system = @import("system.zig");
+const protocol = @import("block-protocol");
+
+pub const Geometry = struct { block_size: u32, block_count: u64 };
+
+pub const Device = struct {
+    endpoint: ipc.Handle,
+
+    /// The device's block size and total block count.
+    pub fn geometry(self: Device) ?Geometry {
+        var request = protocol.Request{ .operation = @intFromEnum(protocol.Operation.geometry), .lba = 0, .count = 0, .physical = 0 };
+        var reply: [protocol.reply_size]u8 = undefined;
+        const n = ipc.call(self.endpoint, std.mem.asBytes(&request), &reply) catch return null;
+        if (n < protocol.reply_size) return null;
+        const result = std.mem.bytesToValue(protocol.Reply, reply[0..protocol.reply_size]);
+        if (result.status != 0) return null;
+        return .{ .block_size = result.block_size, .block_count = result.block_count };
+    }
+
+    /// Read `count` blocks starting at `lba` into the DMA buffer at `physical`.
+    pub fn read(self: Device, lba: u64, count: u32, physical: u64) bool {
+        return self.transfer(.read, lba, count, physical);
+    }
+
+    /// Write `count` blocks starting at `lba` from the DMA buffer at `physical`.
+    pub fn write(self: Device, lba: u64, count: u32, physical: u64) bool {
+        return self.transfer(.write, lba, count, physical);
+    }
+
+    fn transfer(self: Device, operation: protocol.Operation, lba: u64, count: u32, physical: u64) bool {
+        var request = protocol.Request{ .operation = @intFromEnum(operation), .lba = lba, .count = count, .physical = physical };
+        var reply: [protocol.reply_size]u8 = undefined;
+        const n = ipc.call(self.endpoint, std.mem.asBytes(&request), &reply) catch return false;
+        if (n < protocol.reply_size) return false;
+        return std.mem.bytesToValue(protocol.Reply, reply[0..protocol.reply_size]).status == 0;
+    }
+};
+
+/// Look up the block device, retrying generously while the USB storage chain
+/// (controller reset, enumeration, mass-storage bring-up) comes up.
+pub fn open() ?Device {
+    // Patient: the whole USB storage chain (firmware discovery, xHCI reset and
+    // enumeration, mass-storage bring-up) must complete first, which can take
+    // tens of seconds under emulation.
+    var attempts: usize = 0;
+    while (attempts < 1200) : (attempts += 1) {
+        if (ipc.lookup(.block)) |handle| return .{ .endpoint = handle };
+        system.sleep(50);
+    }
+    return null;
+}

library/runtime/device.zig

@@ -3,6 +3,7 @@
 //! ownership of its hardware; the claim is the capability the kernel checks before
 //! mapping registers or routing an IRQ.
 
+const std = @import("std");
 const abi = @import("abi");
 const device_abi = @import("device-abi");
 const sc = @import("system-call.zig");
@@ -36,6 +37,10 @@ pub fn mmioMap(device_id: u64, resource_index: u64) ?usize {
 /// `DeviceDescriptor.parent` for a device with no parent.
 pub const no_parent = device_abi.no_parent;
 
+/// `DeviceDescriptor.pci_class` for a device that is not a PCI function. Set this on
+/// descriptors passed to `register` unless the child really is one.
+pub const no_pci_class = device_abi.no_pci_class;
+
 /// Publish `descriptor` as a child of `parent_id`, which this process must have claimed.
 /// Returns the new device id. The child is left unclaimed, so whichever driver owns
 /// that class of device can `claim` it — that is how a bus hands off a device.
@@ -108,3 +113,19 @@ pub fn ioRead(device_id: u64, resource_index: u64, offset: u64, width: u8) ?u32
 pub fn ioWrite(device_id: u64, resource_index: u64, offset: u64, width: u8, value: u32) bool {
     return !failed(sc.systemCall5(.io_write, device_id, resource_index, offset, width, value));
 }
+
+/// Find DeviceDescription by hid
+///
+/// Utility function for driver development
+pub fn findDeviceDescriptorByHid(buffer: []DeviceDescriptor, hid_needle: []const u8) ?DeviceDescriptor {
+    const total = enumerate(buffer);
+    const n = @min(total, buffer.len);
+    for (@as([]DeviceDescriptor, buffer[0..n])) |d| {
+        const hid_haystack = d.hid[0..@intCast(d.hid_len)];
+        if (std.mem.eql(u8, hid_haystack, hid_needle)) {
+            return d;
+        }
+    }
+
+    return null;
+}

library/runtime/fs.zig

@@ -0,0 +1,274 @@
+//! runtime.fs — the danos-native file API. A program opens, reads, writes, and
+//! lists files served by the user-space VFS (system/services/vfs), each call
+//! marshalling a vfs-protocol request over IPC. This is the danos-native layer
+//! danos programs use directly; it is also where the file operations that later
+//! become `std.os.danos` are staged (see docs/zig-self-hosting.md). It replaces
+//! the old POSIX `unistd` shim — a compatibility spelling danos does not need yet.
+//!
+//! Handles are *values*, not entries in a global descriptor table: a `File` /
+//! `Directory` owns its VFS node id and (for files) a byte offset. So there is no
+//! per-process fd limit and no shared table to synchronise — the danos-native
+//! shape, unlike the POSIX fd model the old shim emulated.
+
+const std = @import("std");
+const ipc = @import("ipc.zig");
+const protocol = @import("vfs-protocol");
+
+/// The kind of a filesystem node — re-exported so a caller need not import the
+/// wire protocol.
+pub const Kind = protocol.NodeKind;
+
+/// A node's metadata (the answer to a status request).
+pub const Attributes = struct {
+    size: u64,
+    kind: Kind,
+    /// Modification time — Unix epoch seconds, UTC. 0 if the filesystem has none.
+    mtime: u64 = 0,
+};
+
+// Map a wire `NodeKind` value to the enum, defaulting anything unrecognised to
+// `.regular` (the server is trusted, but a value outside the enum would be
+// illegal to `@enumFromInt` directly).
+fn kindFromWire(value: u32) Kind {
+    return switch (value) {
+        @intFromEnum(Kind.directory) => .directory,
+        @intFromEnum(Kind.character_device) => .character_device,
+        @intFromEnum(Kind.block_device) => .block_device,
+        @intFromEnum(Kind.symbolic_link) => .symbolic_link,
+        @intFromEnum(Kind.fifo) => .fifo,
+        @intFromEnum(Kind.socket) => .socket,
+        else => .regular,
+    };
+}
+
+/// How to open a path.
+pub const OpenOptions = struct {
+    /// Create the file if it does not exist.
+    create: bool = false,
+    /// Open a directory node (for listing) rather than a file.
+    directory: bool = false,
+    /// Truncate an existing file to zero length on open (O_TRUNC) — replace its
+    /// contents rather than overwriting in place.
+    truncate: bool = false,
+
+    fn wireFlags(self: OpenOptions) u32 {
+        var f: u32 = 0;
+        if (self.create) f |= protocol.create;
+        if (self.directory) f |= protocol.directory;
+        if (self.truncate) f |= protocol.truncate;
+        return f;
+    }
+};
+
+// The VFS server endpoint, looked up once by well-known id and cached.
+var vfs_handle: ipc.Handle = 0;
+var vfs_resolved = false;
+fn vfs() ?ipc.Handle {
+    if (!vfs_resolved) {
+        vfs_handle = ipc.lookup(.vfs) orelse return null;
+        vfs_resolved = true;
+    }
+    return vfs_handle;
+}
+
+const Result = struct { reply: protocol.Reply, payload: []u8 };
+
+// One request/reply round trip: [Request header][send payload] -> VFS ->
+// [Reply header][receive payload]. The receive payload lands in `out`.
+fn transact(request: protocol.Request, send: []const u8, out: []u8) ?Result {
+    const h = vfs() orelse return null;
+    var message: [protocol.message_maximum]u8 = undefined;
+    @memcpy(message[0..protocol.request_size], std.mem.asBytes(&request));
+    const slen = @min(send.len, protocol.maximum_payload);
+    @memcpy(message[protocol.request_size..][0..slen], send[0..slen]);
+
+    var rbuf: [protocol.message_maximum]u8 = undefined;
+    const n = ipc.call(h, message[0 .. protocol.request_size + slen], &rbuf) catch return null;
+    if (n < protocol.reply_size) return null;
+    const reply = std.mem.bytesToValue(protocol.Reply, rbuf[0..protocol.reply_size]);
+    const rpl = @min(n - protocol.reply_size, out.len);
+    @memcpy(out[0..rpl], rbuf[protocol.reply_size..][0..rpl]);
+    return .{ .reply = reply, .payload = out[0..rpl] };
+}
+
+/// An open file: a VFS node plus a byte cursor. Read and write advance the cursor.
+pub const File = struct {
+    node: u64,
+    offset: u64 = 0,
+
+    /// Read up to `buffer.len` bytes at the current offset; returns the count, or
+    /// null on error.
+    pub fn read(self: *File, buffer: []u8) ?usize {
+        const want: u32 = @intCast(@min(buffer.len, protocol.maximum_payload));
+        const request = protocol.Request{ .operation = .read, .node = self.node, .offset = self.offset, .len = want, .flags = 0 };
+        const r = transact(request, &.{}, buffer) orelse return null;
+        if (r.reply.status != 0) return null;
+        self.offset += r.reply.len;
+        return r.reply.len;
+    }
+
+    /// Write `data` at the current offset; returns the count written. A single
+    /// call is capped at the VFS payload size, so the return may be short — use
+    /// `writeAll` to write the whole slice. Null on error.
+    pub fn write(self: *File, data: []const u8) ?usize {
+        const want: u32 = @intCast(@min(data.len, protocol.maximum_payload));
+        const request = protocol.Request{ .operation = .write, .node = self.node, .offset = self.offset, .len = want, .flags = 0 };
+        const r = transact(request, data[0..want], &.{}) orelse return null;
+        if (r.reply.status != 0) return null;
+        self.offset += r.reply.len;
+        return r.reply.len;
+    }
+
+    /// Write all of `data`, looping past the per-call payload cap. Returns the
+    /// total written, or null if a write failed before any progress.
+    pub fn writeAll(self: *File, data: []const u8) ?usize {
+        var written: usize = 0;
+        while (written < data.len) {
+            const n = self.write(data[written..]) orelse return if (written == 0) null else written;
+            if (n == 0) return written; // no forward progress; stop rather than spin
+            written += n;
+        }
+        return written;
+    }
+
+    /// Move the read/write cursor to an absolute byte position.
+    pub fn seekTo(self: *File, position: u64) void {
+        self.offset = position;
+    }
+
+    /// This file's metadata.
+    pub fn attributes(self: *File) ?Attributes {
+        const request = protocol.Request{ .operation = .status, .node = self.node, .offset = 0, .len = 0, .flags = 0 };
+        var buffer: [@sizeOf(protocol.FileStatus)]u8 = undefined;
+        const r = transact(request, &.{}, &buffer) orelse return null;
+        if (r.reply.status != 0 or r.payload.len < @sizeOf(protocol.FileStatus)) return null;
+        const status = std.mem.bytesToValue(protocol.FileStatus, buffer[0..@sizeOf(protocol.FileStatus)]);
+        return .{ .size = status.size, .kind = kindFromWire(status.kind), .mtime = status.mtime };
+    }
+
+    /// Release the VFS's open handle for this file.
+    pub fn close(self: *File) void {
+        const request = protocol.Request{ .operation = .close, .node = self.node, .offset = 0, .len = 0, .flags = 0 };
+        _ = transact(request, &.{}, &.{});
+    }
+};
+
+/// Open (or create, with `.create`) `path`. Returns the open file, or null.
+pub fn open(path: []const u8, options: OpenOptions) ?File {
+    const request = protocol.Request{ .operation = .open, .node = 0, .offset = 0, .len = @intCast(path.len), .flags = options.wireFlags() };
+    const r = transact(request, path, &.{}) orelse return null;
+    if (r.reply.status != 0) return null;
+    return .{ .node = r.reply.node };
+}
+
+/// A path's metadata without keeping it open (open -> status -> close).
+pub fn attributes(path: []const u8) ?Attributes {
+    var file = open(path, .{}) orelse return null;
+    defer file.close();
+    return file.attributes();
+}
+
+/// Whether `path` resolves — handy as a readiness check (e.g. waiting for a mount
+/// to come up before writing to it).
+pub fn exists(path: []const u8) bool {
+    return attributes(path) != null;
+}
+
+/// One entry returned by `Directory.next`.
+pub const Entry = struct {
+    kind: Kind = .regular,
+    size: u64 = 0,
+    name_buffer: [64]u8 = undefined,
+    name_len: usize = 0,
+
+    pub fn name(self: *const Entry) []const u8 {
+        return self.name_buffer[0..self.name_len];
+    }
+};
+
+/// An open directory being listed, cursor-advanced by `next`.
+pub const Directory = struct {
+    node: u64,
+    cursor: u64 = 0,
+
+    /// Fill `entry` with the next directory entry; false at end of directory or
+    /// on error.
+    pub fn next(self: *Directory, entry: *Entry) bool {
+        const request = protocol.Request{ .operation = .readdir, .node = self.node, .offset = self.cursor, .len = 0, .flags = 0 };
+        var buffer: [protocol.message_maximum]u8 = undefined;
+        const r = transact(request, &.{}, &buffer) orelse return false;
+        if (r.reply.status != 0 or r.reply.len == 0) return false; // error or EOF
+        if (r.payload.len < protocol.directory_entry_size) return false;
+        const header = std.mem.bytesToValue(protocol.DirectoryEntry, r.payload[0..protocol.directory_entry_size]);
+        entry.kind = kindFromWire(header.kind);
+        entry.size = header.size;
+        const source = r.payload[protocol.directory_entry_size..];
+        const nlen = @min(@min(@as(usize, header.name_len), source.len), entry.name_buffer.len);
+        @memcpy(entry.name_buffer[0..nlen], source[0..nlen]);
+        entry.name_len = nlen;
+        self.cursor += 1;
+        return true;
+    }
+
+    /// Release the VFS's open handle for this directory.
+    pub fn close(self: *Directory) void {
+        var f = File{ .node = self.node };
+        f.close();
+    }
+};
+
+/// Open `path` as a directory for listing. Returns null if it isn't one / on error.
+pub fn openDirectory(path: []const u8) ?Directory {
+    const file = open(path, .{ .directory = true }) orelse return null;
+    return .{ .node = file.node };
+}
+
+// A path-based request that returns only a status (mkdir, unlink).
+fn pathOperation(operation: protocol.Operation, path: []const u8) bool {
+    const request = protocol.Request{ .operation = operation, .node = 0, .offset = 0, .len = @intCast(path.len), .flags = 0 };
+    const r = transact(request, path, &.{}) orelse return false;
+    return r.reply.status == 0;
+}
+
+/// Create a directory at `path` (its parent must already exist). Returns true on
+/// success. Only works under a mounted filesystem that supports directories.
+pub fn makeDirectory(path: []const u8) bool {
+    return pathOperation(.mkdir, path);
+}
+
+/// Remove the file at `path`. Returns true on success. Directories are refused
+/// (a separate directory-removal would have to check emptiness).
+pub fn remove(path: []const u8) bool {
+    return pathOperation(.unlink, path);
+}
+
+/// Rename `old_path` to `new_path`. Both must be in the same directory (same-
+/// directory, 8.3-name rename only for now). Returns true on success.
+pub fn rename(old_path: []const u8, new_path: []const u8) bool {
+    const total = old_path.len + 1 + new_path.len;
+    if (total > protocol.maximum_payload) return false;
+    var payload: [protocol.maximum_payload]u8 = undefined;
+    @memcpy(payload[0..old_path.len], old_path);
+    payload[old_path.len] = 0;
+    @memcpy(payload[old_path.len + 1 ..][0..new_path.len], new_path);
+    const request = protocol.Request{ .operation = .rename, .node = 0, .offset = 0, .len = @intCast(total), .flags = 0 };
+    const r = transact(request, payload[0..total], &.{}) orelse return false;
+    return r.reply.status == 0;
+}
+
+/// Mount a filesystem backend (its server endpoint) at absolute path `target`;
+/// the VFS then routes everything under `target` to that backend. This is the one
+/// call that hands the VFS a capability (the backend endpoint). Returns true on
+/// success.
+pub fn mount(target: []const u8, backend: ipc.Handle) bool {
+    const h = vfs() orelse return false;
+    const request = protocol.Request{ .operation = .mount, .node = 0, .offset = 0, .len = @intCast(target.len), .flags = 0 };
+    var message: [protocol.message_maximum]u8 = undefined;
+    @memcpy(message[0..protocol.request_size], std.mem.asBytes(&request));
+    const tlen = @min(target.len, protocol.maximum_payload);
+    @memcpy(message[protocol.request_size..][0..tlen], target[0..tlen]);
+    var rbuf: [protocol.message_maximum]u8 = undefined;
+    const result = ipc.callCap(h, message[0 .. protocol.request_size + tlen], &rbuf, backend) catch return false;
+    if (result.len < protocol.reply_size) return false;
+    return std.mem.bytesToValue(protocol.Reply, rbuf[0..protocol.reply_size]).status == 0;
+}

library/runtime/input.zig

@@ -0,0 +1,220 @@
+//! User-space input helpers: the client and publisher sides of the input service, so a
+//! program listening for input events — or a driver broadcasting them — doesn't hand-roll
+//! the IPC. Layered over `ipc` (endpoints, capability passing, `send`) and the shared
+//! `input-protocol` wire format, the way `device.zig` layers over the raw `device_*` calls.
+//! See system/services/input/input.zig.
+//!
+//! The service carries several device classes (keyboard, mouse, joystick/gamepad). A
+//! **source** publishes its class with the matching method:
+//!     var source = input.connectSource() orelse return;
+//!     _ = source.publishKeyboardEvent(.{ .kind = ..., .keycode = ..., ... });
+//!     _ = source.publishMouseEvent(.{ ... });
+//!     _ = source.publishJoystickEvent(.{ ... });
+//!
+//! A **subscriber** either takes one class with a typed helper —
+//!     var keys = input.subscribeKeyboard() orelse return;
+//!     while (true) { const key = keys.next() orelse continue; ... }
+//! — or takes several at once and inspects the tagged envelope:
+//!     var listener = input.subscribeAll() orelse return;
+//!     while (true) {
+//!         const event = listener.next() orelse continue;
+//!         if (event.asKeyboard()) |k| { ... } else if (event.asMouse()) |m| { ... }
+//!     }
+
+const std = @import("std");
+const abi = @import("abi");
+const ipc = @import("ipc.zig");
+const system = @import("system.zig");
+const protocol = @import("input-protocol");
+
+pub const DeviceKind = protocol.DeviceKind;
+pub const InputEvent = protocol.InputEvent;
+pub const KeyEvent = protocol.KeyEvent;
+pub const MouseEvent = protocol.MouseEvent;
+pub const JoystickEvent = protocol.JoystickEvent;
+pub const EventKind = protocol.EventKind;
+pub const MouseEventKind = protocol.MouseEventKind;
+pub const JoystickEventKind = protocol.JoystickEventKind;
+pub const Keycode = protocol.Keycode;
+
+/// Interest masks re-exported so a caller can `subscribe(input.device_keyboard |
+/// input.device_mouse)`.
+pub const device_keyboard = protocol.device_keyboard;
+pub const device_mouse = protocol.device_mouse;
+pub const device_joystick = protocol.device_joystick;
+pub const device_all = protocol.device_all;
+
+/// Look up the input service, retrying while it is still coming up. Both a subscriber and
+/// a source race the service's registration at boot, so both wait for it here rather than
+/// failing. Returns the service endpoint handle, or null if it never appears.
+fn lookupService() ?ipc.Handle {
+    var attempts: usize = 0;
+    while (attempts < 100) : (attempts += 1) {
+        if (ipc.lookup(.input)) |handle| return handle;
+        system.sleep(50);
+    }
+    return null;
+}
+
+// --- subscribing ------------------------------------------------------------
+
+/// A subscription to the input service: our own endpoint, which the service pushes events
+/// to. `next` returns each event as a tagged `InputEvent`; use `asKeyboard`/`asMouse`/
+/// `asJoystick` to decode. Created with `subscribe`/`subscribeAll`; for a single device
+/// class prefer the typed helpers (`subscribeKeyboard`, ...), which return decoded events.
+pub const Subscriber = struct {
+    /// The endpoint the service delivers events to (created and owned by us; its handle
+    /// was handed to the service as a capability at subscribe time).
+    endpoint: ipc.Handle,
+    receive: [protocol.event_size]u8 = undefined,
+
+    /// Block until the next event is pushed, and return it. Events arrive as asynchronous
+    /// buffered messages (`ipc_send` from the service), so nothing is owed in reply — the
+    /// empty reply this issues is a harmless no-op. Returns null for any non-event wake-up
+    /// (there should be none), so callers can loop.
+    pub fn next(self: *Subscriber) ?InputEvent {
+        const got = ipc.replyWait(self.endpoint, &.{}, &self.receive, null);
+        if (!got.isMessage() or got.len < protocol.event_size) return null;
+        return std.mem.bytesToValue(InputEvent, self.receive[0..protocol.event_size]);
+    }
+};
+
+/// Subscribe to the input classes named in `device_mask` (an OR of `device_*`, or
+/// `device_all`). Creates an endpoint for the service to push to and hands it over as a
+/// capability. Returns a `Subscriber` to loop `next` on, or null on failure.
+pub fn subscribe(device_mask: u32) ?Subscriber {
+    const service = lookupService() orelse return null;
+    const endpoint = ipc.createIpcEndpoint() orelse return null;
+
+    var request = protocol.Request{ .operation = @intFromEnum(protocol.Operation.subscribe), .device_mask = device_mask };
+    var reply: [protocol.reply_size]u8 = undefined;
+    const result = ipc.callCap(service, std.mem.asBytes(&request), &reply, endpoint) catch return null;
+    if (result.len < protocol.reply_size) return null;
+    if (std.mem.bytesToValue(protocol.Reply, reply[0..protocol.reply_size]).status != 0) return null;
+    return .{ .endpoint = endpoint };
+}
+
+/// Subscribe to every input class (keyboard, mouse, joystick) on one stream.
+pub fn subscribeAll() ?Subscriber {
+    return subscribe(device_all);
+}
+
+/// A subscriber filtered to keyboard events, whose `next` returns a decoded `KeyEvent`.
+pub const KeyboardSubscriber = struct {
+    inner: Subscriber,
+    pub fn next(self: *KeyboardSubscriber) ?KeyEvent {
+        return (self.inner.next() orelse return null).asKeyboard();
+    }
+};
+
+/// A subscriber filtered to mouse events, whose `next` returns a decoded `MouseEvent`.
+pub const MouseSubscriber = struct {
+    inner: Subscriber,
+    pub fn next(self: *MouseSubscriber) ?MouseEvent {
+        return (self.inner.next() orelse return null).asMouse();
+    }
+};
+
+/// A subscriber filtered to joystick/gamepad events, whose `next` returns a decoded
+/// `JoystickEvent`.
+pub const JoystickSubscriber = struct {
+    inner: Subscriber,
+    pub fn next(self: *JoystickSubscriber) ?JoystickEvent {
+        return (self.inner.next() orelse return null).asJoystick();
+    }
+};
+
+/// Subscribe to keyboard events only; `next` returns decoded `KeyEvent`s.
+pub fn subscribeKeyboard() ?KeyboardSubscriber {
+    return .{ .inner = subscribe(device_keyboard) orelse return null };
+}
+
+/// Subscribe to mouse events only; `next` returns decoded `MouseEvent`s.
+pub fn subscribeMouse() ?MouseSubscriber {
+    return .{ .inner = subscribe(device_mouse) orelse return null };
+}
+
+/// Subscribe to joystick/gamepad events only; `next` returns decoded `JoystickEvent`s.
+pub fn subscribeJoystick() ?JoystickSubscriber {
+    return .{ .inner = subscribe(device_joystick) orelse return null };
+}
+
+// --- publishing -------------------------------------------------------------
+
+/// A connection to the input service for a source (a keyboard/mouse/joystick driver) that
+/// publishes events. Each `publish*Event` is a short synchronous call the service answers
+/// at once; its own fan-out to subscribers is asynchronous, so publishing never blocks on
+/// a slow subscriber.
+pub const Publisher = struct {
+    service: ipc.Handle,
+
+    fn publish(self: Publisher, event: InputEvent) bool {
+        var request = protocol.Request{ .operation = @intFromEnum(protocol.Operation.publish), .event = event };
+        var reply: [protocol.reply_size]u8 = undefined;
+        const len = ipc.call(self.service, std.mem.asBytes(&request), &reply) catch return false;
+        if (len < protocol.reply_size) return false;
+        return std.mem.bytesToValue(protocol.Reply, reply[0..protocol.reply_size]).status == 0;
+    }
+
+    /// Broadcast a keyboard event to every subscriber that took keyboard events.
+    pub fn publishKeyboardEvent(self: Publisher, event: KeyEvent) bool {
+        return self.publish(InputEvent.fromKeyboard(event));
+    }
+    /// Broadcast a mouse event to every subscriber that took mouse events.
+    pub fn publishMouseEvent(self: Publisher, event: MouseEvent) bool {
+        return self.publish(InputEvent.fromMouse(event));
+    }
+    /// Broadcast a joystick/gamepad event to every subscriber that took joystick events.
+    pub fn publishJoystickEvent(self: Publisher, event: JoystickEvent) bool {
+        return self.publish(InputEvent.fromJoystick(event));
+    }
+};
+
+/// Connect to the input service as an event source, waiting for it to come up. Returns a
+/// `Publisher`, or null if the service never registered.
+pub fn connectSource() ?Publisher {
+    return .{ .service = lookupService() orelse return null };
+}
+
+// --- synthetic scaffolding --------------------------------------------------
+
+/// Synthetic key events, shared by the demo source and the keyboard driver's placeholder
+/// stream while real scancode decoding is still a follow-up. `step` rolls through A..E,
+/// emitting for each key a `key_down`, then a `key_press` carrying the character, then a
+/// `key_up`. Scaffolding, not wire protocol — hence it lives with the helpers.
+pub fn syntheticKeyEvent(step: usize) KeyEvent {
+    const Key = struct { code: Keycode, character: u32 };
+    const keys = [_]Key{
+        .{ .code = .a, .character = 'A' },
+        .{ .code = .b, .character = 'B' },
+        .{ .code = .c, .character = 'C' },
+        .{ .code = .d, .character = 'D' },
+        .{ .code = .e, .character = 'E' },
+    };
+    const key = keys[(step / 3) % keys.len];
+    return switch (step % 3) {
+        0 => .{ .kind = @intFromEnum(EventKind.key_down), .keycode = @intFromEnum(key.code), .character = 0, .modifiers = 0 },
+        1 => .{ .kind = @intFromEnum(EventKind.key_press), .keycode = @intFromEnum(key.code), .character = key.character, .modifiers = 0 },
+        else => .{ .kind = @intFromEnum(EventKind.key_up), .keycode = @intFromEnum(key.code), .character = 0, .modifiers = 0 },
+    };
+}
+
+/// Synthetic mouse events (placeholder until real PS/2 packet decoding). `step` alternates
+/// a small diagonal motion with a left-button click.
+pub fn syntheticMouseEvent(step: usize) MouseEvent {
+    return switch (step % 3) {
+        0 => .{ .kind = @intFromEnum(MouseEventKind.motion), .button = 0, .dx = 1, .dy = 1, .scroll_x = 0, .scroll_y = 0, .buttons = 0 },
+        1 => .{ .kind = @intFromEnum(MouseEventKind.button_down), .button = protocol.mouse_button_left, .dx = 0, .dy = 0, .scroll_x = 0, .scroll_y = 0, .buttons = protocol.mouse_button_left },
+        else => .{ .kind = @intFromEnum(MouseEventKind.button_up), .button = protocol.mouse_button_left, .dx = 0, .dy = 0, .scroll_x = 0, .scroll_y = 0, .buttons = 0 },
+    };
+}
+
+/// Synthetic joystick/gamepad events (placeholder until a real controller driver). `step`
+/// sweeps axis 0 and toggles button 0.
+pub fn syntheticJoystickEvent(step: usize) JoystickEvent {
+    return switch (step % 3) {
+        0 => .{ .kind = @intFromEnum(JoystickEventKind.axis), .control = 0, .value = 16384, .buttons = 0 },
+        1 => .{ .kind = @intFromEnum(JoystickEventKind.button_down), .control = 0, .value = 0, .buttons = 1 },
+        else => .{ .kind = @intFromEnum(JoystickEventKind.button_up), .control = 0, .value = 0, .buttons = 0 },
+    };
+}

library/runtime/ipc.zig

@@ -79,11 +79,41 @@ pub fn call(h: Handle, message: []const u8, reply: []u8) CallError!usize {
     return (try callCap(h, message, reply, null)).len;
 }
 
+/// Post `message` to endpoint `h`'s asynchronous queue and return immediately — no
+/// rendezvous, no reply, no blocking. The receiver picks it up through `replyWait` as a
+/// buffered message (`Received.isMessage`). Unlike `call`, this **cannot hang on a dead
+/// or slow peer**, which is why a broadcaster (the input service) delivers events this
+/// way. The payload must fit an endpoint slot (64 bytes); a full queue drops the oldest
+/// message. Returns false on failure (bad handle, oversized payload, bad buffer).
+pub fn send(h: Handle, message: []const u8) bool {
+    return !failed(sc.systemCall3(.ipc_send, h, @intFromPtr(message.ptr), message.len));
+}
+
 /// Set in `Received.badge` when what arrived is an asynchronous notification — a
 /// bound device interrupt — rather than a client's message. The low bits carry the
 /// GSI. See `isNotification`.
 pub const notify_badge_bit: u64 = abi.notify_badge_bit;
 
+/// Set alongside `notify_badge_bit` when the notification is a **signal** — the
+/// lifecycle vocabulary of docs/process-lifecycle.md, delivered to the endpoint
+/// nominated with `process.bindSignals`. Decode with `process.signalsFrom`.
+pub const notify_signal_bit: u64 = abi.notify_signal_bit;
+
+/// Set alongside `notify_badge_bit` when the notification is a **one-shot timer**
+/// landing (`system.timerOnce`).
+pub const notify_timer_bit: u64 = abi.notify_timer_bit;
+
+/// Set alongside `notify_badge_bit` when the notification is a **child-exit
+/// notice** — a process this one spawned (with an exit endpoint) has ended —
+/// rather than a device interrupt. The low bits carry the child's process id.
+pub const notify_exit_bit: u64 = abi.notify_exit_bit;
+
+/// Set alongside `notify_badge_bit` when the wake-up is a **buffered message** — a payload
+/// posted with `send` (`ipc_send`) — rather than a bare device interrupt or child-exit
+/// notice. The payload is in the `replyWait` receive buffer (`Received.len` bytes); the
+/// low bits of the badge carry the sender's task id. See `Received.isMessage`.
+pub const notify_message_bit: u64 = abi.notify_message_bit;
+
 /// The result of a `replyWait`: the request length, the sender's badge (a task id, or
 /// an IRQ notification if the high bit is set), and any capability the request carried.
 pub const Received = struct {
@@ -91,16 +121,53 @@ pub const Received = struct {
     badge: u64,
     cap: ?Handle,
 
-    /// True if this wake-up was a device interrupt, not a client request. A driver's
-    /// event loop branches on this; there is no reply owed on the notification path.
+    /// True if this wake-up was an asynchronous notification (a device interrupt
+    /// or a child-exit notice), not a client request. An event loop branches on
+    /// this; there is no reply owed on the notification path.
     pub fn isNotification(self: Received) bool {
         return self.badge & notify_badge_bit != 0;
     }
 
-    /// The interrupt source (a GSI), meaningful only when `isNotification`.
+    /// True if this wake-up tells of a supervised child's end — the notification
+    /// requested by passing an exit endpoint to `system.spawnSupervised`.
+    pub fn isChildExit(self: Received) bool {
+        return self.isNotification() and self.badge & notify_exit_bit != 0;
+    }
+
+    /// True if this wake-up is a **buffered message** posted with `send` (`ipc_send`):
+    /// there is a payload in the receive buffer (`self.len` bytes) and no reply is owed.
+    /// The subscriber side of a broadcast branches on this.
+    pub fn isMessage(self: Received) bool {
+        return self.isNotification() and self.badge & notify_message_bit != 0;
+    }
+
+    /// The task id of whoever posted a buffered message, meaningful only when
+    /// Whether this arrival is a signal notification — decode the set with
+    /// `process.signalsFrom(badge)`.
+    pub fn isSignal(self: Received) bool {
+        return self.isNotification() and self.badge & notify_signal_bit != 0;
+    }
+
+    /// Whether this arrival is a one-shot timer landing (`system.timerOnce`).
+    pub fn isTimer(self: Received) bool {
+        return self.isNotification() and self.badge & notify_timer_bit != 0;
+    }
+
+    /// `isMessage`. (The badge's low bits, with the three high marker bits masked off.)
+    pub fn senderTaskId(self: Received) u32 {
+        return @intCast(self.badge & ~(notify_badge_bit | notify_exit_bit | notify_message_bit));
+    }
+
+    /// The interrupt source (a GSI), meaningful only when `isNotification` and
+    /// not `isChildExit`.
     pub fn source(self: Received) u64 {
         return self.badge & ~notify_badge_bit;
     }
+
+    /// The ended child's process id, meaningful only when `isChildExit`.
+    pub fn childProcessId(self: Received) u32 {
+        return @intCast(self.badge & ~(notify_badge_bit | notify_exit_bit));
+    }
 };
 
 /// Server side of IPC_ReplyWait: deliver `reply` to the client last received (if any,

library/runtime/process.zig

@@ -0,0 +1,137 @@
+//! Process-level runtime types: what a user program receives at entry (`Init`,
+//! the argv contract) and the process end of the lifecycle
+//! (docs/process-lifecycle.md) — today the exit reason a supervisor reads to
+//! decide restart; signals and the stop sequence land here with M17.4. Mirrors
+//! the spirit of `std.process.Init.Minimal` in danos terms — std's `Args` holds
+//! no data on freestanding targets, so the type is danos's own.
+
+const std = @import("std");
+const abi = @import("abi");
+const sc = @import("system-call.zig");
+const ipc = @import("ipc.zig");
+const system = @import("system.zig");
+
+/// Everything a program receives at entry. Passed to
+/// `pub fn main(init: runtime.process.Init)`; programs that need nothing keep
+/// `pub fn main() void`. An `environment` field is added here once the kernel
+/// passes a non-empty envp (today it is always empty — see docs/sysv.md).
+pub const Init = struct {
+    arguments: Arguments,
+};
+
+/// The process arguments (argc/argv), parsed from the kernel-built System V
+/// entry block. The bytes live in the entry block at the top of the stack page,
+/// NUL-terminated, valid for the process's lifetime.
+pub const Arguments = struct {
+    /// argc — at least 1: argument 0 is the path or name this binary was
+    /// spawned as.
+    count: usize,
+    /// The argv pointers in the entry block (NULL-terminated after `count`
+    /// entries).
+    vector: [*]const [*:0]const u8,
+
+    /// Argument `index` (0 = the program's own path/name), or null if out of
+    /// range.
+    pub fn get(arguments: Arguments, index: usize) ?[:0]const u8 {
+        if (index >= arguments.count) return null;
+        return std.mem.span(arguments.vector[index]);
+    }
+
+    pub fn iterate(arguments: Arguments) Iterator {
+        return .{ .arguments = arguments };
+    }
+
+    pub const Iterator = struct {
+        arguments: Arguments,
+        index: usize = 0,
+
+        pub fn next(iterator: *Iterator) ?[:0]const u8 {
+            const argument = iterator.arguments.get(iterator.index) orelse return null;
+            iterator.index += 1;
+            return argument;
+        }
+    };
+};
+
+/// How a process ended — what a supervisor's restart policy reads: a clean exit
+/// meant to stop, a fault wants a restart with backoff, killed means the
+/// supervisor did it itself (docs/process-lifecycle.md).
+pub const ExitReason = abi.ExitReason;
+
+/// How dead child `id` ended. Ask after the exit notification arrives — the
+/// kernel records the reason before it posts the notification, so this never
+/// races it. Returns null for an id that never lived, is still alive, was
+/// evicted from the kernel's bounded record, or is not this process's child
+/// (the same authority gate as `kill`).
+pub fn exitReason(id: u32) ?ExitReason {
+    const r = sc.systemCall1(.process_exit_reason, id);
+    if (r > ~@as(usize, 0) - 4095) return null; // a wrapped -errno
+    return @enumFromInt(r);
+}
+
+/// The signal vocabulary (docs/process-lifecycle.md): POSIX's concepts, danos's
+/// names, message delivery. A signal is a one-way coalescing statement — never a
+/// question (liveness is the zero-length ping call) and never kill (that is
+/// `system.kill`, unhandleable by definition).
+pub const Signal = abi.Signal;
+
+/// The coalesced set of signals one notification delivered: two pending
+/// terminates arrive as one. Decode a received badge with `signalsFrom`.
+pub const SignalSet = struct {
+    pending: u32,
+
+    pub fn has(set: SignalSet, signal: Signal) bool {
+        return set.pending & (@as(u32, 1) << @intFromEnum(signal)) != 0;
+    }
+};
+
+/// Nominate `endpoint` as this process's signal endpoint. Signals posted while
+/// unbound have pended; they are delivered immediately on bind, coalesced.
+pub fn bindSignals(endpoint: usize) bool {
+    return sc.systemCall1(.signal_bind, endpoint) == 0;
+}
+
+/// Decode a received badge into the signals it delivered, or null if it is not
+/// a signal notification.
+pub fn signalsFrom(badge: u64) ?SignalSet {
+    if (badge & abi.notify_badge_bit == 0 or badge & abi.notify_signal_bit == 0) return null;
+    return .{ .pending = @truncate(badge & ~(abi.notify_badge_bit | abi.notify_signal_bit)) };
+}
+
+/// Post `signal` to child `id` (or to yourself). Supervisor-gated, like kill;
+/// non-blocking, always — a statement, not a conversation.
+pub fn sendSignal(id: u32, signal: Signal) bool {
+    return sc.systemCall2(.process_signal, id, @intFromEnum(signal)) == 0;
+}
+
+/// The standard stop sequence (docs/process-lifecycle.md): terminate, wait up to
+/// `deadline_ms` for the exit notification on `exit_endpoint` (the endpoint the
+/// child was spawned with), then kill. Any *other* notifications arriving on
+/// that endpoint while stopping are consumed and dropped — a supervisor with
+/// concurrent traffic implements the same sequence inside its own event loop
+/// (arm `system.timerOnce`, keep serving) instead of calling this.
+pub fn stop(id: u32, deadline_ms: u64, exit_endpoint: usize) void {
+    _ = sendSignal(id, .terminate);
+    _ = system.timerOnce(exit_endpoint, deadline_ms);
+    var receive: [8]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(exit_endpoint, &.{}, &receive, null);
+        if (got.isChildExit() and got.childProcessId() == id) return;
+        if (got.isTimer()) break; // the deadline passed first — escalate
+    }
+    _ = system.kill(id);
+    while (true) {
+        const got = ipc.replyWait(exit_endpoint, &.{}, &receive, null);
+        if (got.isChildExit() and got.childProcessId() == id) return;
+    }
+}
+
+/// Subscribe `endpoint` to published exit events: every process death posts an
+/// asynchronous notification with the same badge encoding as a supervisor's exit
+/// notice (decode with `ipc.Received.isChildExit`/`childProcessId`). For stateful
+/// services: release what the dead client held — file handles, subscriptions —
+/// because a service must never depend on clients cleaning up after themselves
+/// (docs/process-lifecycle.md). Ungated, like `system.processes`.
+pub fn subscribeExits(endpoint: usize) bool {
+    return sc.systemCall1(.process_subscribe, endpoint) == 0;
+}

library/runtime/runtime.zig

@@ -8,14 +8,29 @@
 //!   const runtime = @import("runtime");
 //!   pub const panic = runtime.panic;
 //!   comptime { _ = &runtime.start._start; }   // pull the entry shim in
-//! and a `pub fn main() void`.
+//! and a `pub fn main() void` or `pub fn main(init: runtime.process.Init) void`
+//! (arguments arrive via `init`).
 
 pub const system = @import("system.zig");
+/// Monotonic time, delays, and deadlines over the kernel clock/sleep/timer syscalls
+/// — an `Instant`/`Duration` front door, no time service (docs/timers.md).
+pub const time = @import("time.zig");
 pub const heap = @import("heap.zig");
 pub const ipc = @import("ipc.zig");
 pub const start = @import("start.zig");
 /// The VFS wire protocol (shared with the VFS server).
 pub const vfs_protocol = @import("vfs-protocol");
+
+/// The device-manager protocol: hello + tree reports (docs/device-manager.md).
+pub const device_manager_protocol = @import("device-manager-protocol");
+
+/// The power protocol: events (button, lid, battery) + shutdown (docs/power.md).
+pub const power_protocol = @import("power-protocol");
+/// Keyboard-event listening (subscribe/next) and broadcasting (publish), over the input
+/// service. See library/runtime/input.zig and system/services/input/.
+pub const input = @import("input.zig");
+/// The input wire protocol (shared with the input service and its clients).
+pub const input_protocol = @import("input-protocol");
 /// POSIX-style file API: open/read/write/lseek/stat/close.
 /// C stdio: fopen/fread/fwrite/fseek/ftell/fclose over unistd.
 /// Device access for drivers: enumerate/claim/mmioMap.
@@ -23,8 +38,28 @@ pub const device = @import("device.zig");
 /// DMA-capable memory for drivers: contiguous, pinned, uncacheable buffers.
 pub const dma = @import("dma.zig");
 
+/// USB class-driver client: open a device on the xHCI bus and drive it
+/// (control / interrupt / bulk transfers). See library/runtime/usb.zig.
+pub const usb = @import("usb.zig");
+
+/// Block-device client: read/write a block device (a USB stick, via
+/// usb-storage). See library/runtime/block.zig.
+pub const block = @import("block.zig");
+
+/// The danos-native file API (open/read/write/list over the user-space VFS) — the
+/// layer danos programs use directly, and where the operations that later become
+/// `std.os.danos` are staged. See docs/zig-self-hosting.md.
+pub const fs = @import("fs.zig");
+
 /// Re-exported so a user binary can `pub const panic = runtime.panic;`.
 pub const panic = start.panic;
 
+/// Process entry types: the `Init` handed to `main`, and its `Arguments`.
+pub const process = @import("process.zig");
+
+/// The service harness: one replyWait loop folding requests, signals, and
+/// notifications into callbacks (docs/process-lifecycle.md).
+pub const service = @import("service.zig");
+
 /// The heap as a `std.mem.Allocator`, for Zig `std` containers in user code.
 pub const allocator = heap.allocator;

library/runtime/service.zig

@@ -0,0 +1,83 @@
+//! The service harness (docs/process-lifecycle.md): one replyWait loop that
+//! folds protocol requests, signals, and subscribed notifications into
+//! callbacks — so the lifecycle contract ("answers ping, exits on terminate")
+//! is satisfied by construction and a service author writes domain logic only.
+//! Nothing is asynchronous inside the process: a callback runs at a point the
+//! loop chose, never on a hijacked stack — the whole reason signals are
+//! messages.
+//!
+//! The liveness probe: a **zero-length request is the universal ping**, answered
+//! with a zero-length reply by the harness itself. No protocol's requests start
+//! at length zero, so the encoding cannot collide, and there is nothing for a
+//! service author to implement — a wedged service simply fails to answer, which
+//! is the diagnosis (see docs/ipc.md).
+
+const abi = @import("abi");
+const ipc = @import("ipc.zig");
+const process = @import("process.zig");
+
+pub const Callbacks = struct {
+    /// Called once with the service's endpoint before the loop starts — the
+    /// place to subscribe to exit events, bind IRQs, or announce readiness.
+    /// Return false to abort startup (the process exits).
+    init: ?*const fn (endpoint: ipc.Handle) bool = null,
+    /// One protocol request from `sender` (a task id): write the reply into
+    /// `reply`, return its length. `capability` is the handle the request
+    /// carried, if any (M13 cap passing — how a subscriber hands over its
+    /// endpoint). The zero-length ping never reaches this.
+    on_message: *const fn (message: []const u8, reply: []u8, sender: u32, capability: ?ipc.Handle) usize,
+    /// A notification that is not a signal — a subscribed exit event, a bound
+    /// IRQ, a timer landing. The raw badge; decode with the ipc helpers.
+    on_notification: ?*const fn (badge: u64) void = null,
+    /// The reload signal. Default: ignored.
+    on_reload: ?*const fn () void = null,
+    /// The terminate signal, called before the loop returns. The clean exit is
+    /// the return itself — never put *necessary* work here (iron rule 1: a kill
+    /// arrives with no warning; this is for graceful extras only).
+    on_terminate: ?*const fn () void = null,
+    /// Publish the endpoint under a well-known service id at startup.
+    service: ?abi.ServiceId = null,
+};
+
+/// Run the service: create and (optionally) register the endpoint, bind signals
+/// to it, call `init`, then serve until `terminate` arrives — at which point the
+/// loop returns and main's return is the clean exit the supervisor reads as
+/// `ExitReason.exited`. `maximum_message` sizes the receive and reply buffers
+/// (a service passes its protocol's message maximum).
+pub fn run(comptime maximum_message: usize, callbacks: Callbacks) void {
+    const endpoint = ipc.createIpcEndpoint() orelse return;
+    if (callbacks.service) |id| {
+        if (!ipc.register(id, endpoint)) return;
+    }
+    _ = process.bindSignals(endpoint);
+    if (callbacks.init) |initialise| {
+        if (!initialise(endpoint)) return;
+    }
+
+    var reply_buffer: [maximum_message]u8 = undefined;
+    var reply_len: usize = 0;
+    var receive: [maximum_message]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(endpoint, reply_buffer[0..reply_len], &receive, null);
+        if (got.isNotification()) {
+            reply_len = 0; // nothing owed for a notification
+            if (process.signalsFrom(got.badge)) |signals| {
+                if (signals.has(.reload)) {
+                    if (callbacks.on_reload) |onReload| onReload();
+                }
+                if (signals.has(.terminate)) {
+                    if (callbacks.on_terminate) |onTerminate| onTerminate();
+                    return; // the loop's return IS the clean exit
+                }
+                continue;
+            }
+            if (callbacks.on_notification) |onNotification| onNotification(got.badge);
+            continue;
+        }
+        if (got.len == 0) {
+            reply_len = 0; // the universal ping: a zero-length reply, from the harness
+            continue;
+        }
+        reply_len = callbacks.on_message(receive[0..got.len], &reply_buffer, got.senderTaskId(), got.cap);
+    }
+}

library/runtime/start.zig

@@ -4,24 +4,78 @@
 
 const std = @import("std");
 const system = @import("system.zig");
+const process = @import("process.zig");
 
-/// The kernel enters at `_start` with rsp 16-aligned, but a SystemV function expects
-/// rsp ≡ 8 (mod 16) on entry (as if reached by `call`). The `call` below pushes
-/// the 8-byte return address, satisfying the ABI before any Zig frame runs; the
-/// `ud2` is a safety net if `rt_start` ever returns.
+/// The kernel enters at `_start` with rsp 16-aligned, pointing at the System V
+/// process-entry block it built: argc, argv pointers, NULL, envp terminator, the
+/// auxiliary vector, then the strings (see system/kernel/process.zig,
+/// `buildEntryStack`). Capture that address in rdi — the first SysV argument —
+/// before `call` disturbs the stack; the call's pushed return address also puts
+/// rsp ≡ 8 (mod 16), satisfying the ABI before any Zig frame runs. The `ud2` is a
+/// safety net if `rt_start` ever returns.
 pub export fn _start() callconv(.naked) noreturn {
     asm volatile (
+        \\mov %%rsp, %%rdi
         \\call rt_start
         \\ud2
     );
 }
 
-/// The first Zig frame. The heap is lazy (first alloc grows it), so there is no
-/// runtime init to order here — just hand control to the program's `main`.
-export fn rt_start() callconv(.c) noreturn {
+/// The first Zig frame, entered with `stack` pointing at the kernel-built entry
+/// block. Build the `process.Init` from it and dispatch to the program's `main`,
+/// whose signature is inspected at comptime. The heap is lazy (first alloc grows
+/// it), so there is no other runtime init to order here.
+export fn rt_start(stack: [*]const u64) callconv(.c) noreturn {
+    const init: process.Init = .{ .arguments = .{
+        .count = stack[0],
+        .vector = @ptrCast(stack + 1),
+    } };
+    system.exit(callMain(init));
+}
+
+/// Comptime-dispatch on root.main's signature, in the spirit of std's start.zig:
+/// zero parameters or one `process.Init`; returns void, noreturn, u8, !void, or !u8.
+fn callMain(init: process.Init) u8 {
     const root = @import("root"); // the user binary's root source file
-    root.main();
-    system.exit(0);
+    const main_information = @typeInfo(@TypeOf(root.main)).@"fn";
+
+    const call_arguments = switch (main_information.params.len) {
+        0 => .{},
+        1 => arguments: {
+            const Parameter = main_information.params[0].type orelse
+                @compileError("main's parameter must be runtime.process.Init (not anytype)");
+            if (Parameter != process.Init)
+                @compileError("main's parameter must be runtime.process.Init, found " ++ @typeName(Parameter));
+            break :arguments .{init};
+        },
+        else => @compileError("main takes no parameters or a single runtime.process.Init"),
+    };
+
+    const ReturnType = main_information.return_type.?;
+    switch (@typeInfo(ReturnType)) {
+        .noreturn => @call(.auto, root.main, call_arguments),
+        .void => {
+            @call(.auto, root.main, call_arguments);
+            return 0;
+        },
+        .int => {
+            if (ReturnType != u8)
+                @compileError("main's integer return type must be u8, found " ++ @typeName(ReturnType));
+            return @call(.auto, root.main, call_arguments);
+        },
+        .error_union => {
+            const payload = @call(.auto, root.main, call_arguments) catch |err| {
+                var buffer: [128]u8 = undefined;
+                const line = std.fmt.bufPrint(&buffer, "main returned error: {s}\n", .{@errorName(err)}) catch "main returned an error\n";
+                _ = system.write(line);
+                return 1; // distinct from panic's 127
+            };
+            if (@TypeOf(payload) == void) return 0;
+            if (@TypeOf(payload) == u8) return payload;
+            @compileError("main's error-union payload must be void or u8, found " ++ @typeName(@TypeOf(payload)));
+        },
+        else => @compileError("main must return void, noreturn, u8, !void, or !u8, found " ++ @typeName(ReturnType)),
+    }
 }
 
 /// No runtime to unwind into — report a panic as a nonzero exit code.

library/runtime/system-call.zig

@@ -19,34 +19,49 @@ pub inline fn systemCall0(n: SystemCall) usize {
 pub inline fn systemCall1(n: SystemCall, a0: usize) usize {
     return asm volatile ("syscall"
         : [ret] "={rax}" (-> usize),
-        : [n] "{rax}" (@intFromEnum(n)), [a0] "{rdi}" (a0),
+        : [n] "{rax}" (@intFromEnum(n)),
+          [a0] "{rdi}" (a0),
         : .{ .rcx = true, .r11 = true, .memory = true });
 }
 
 pub inline fn systemCall2(n: SystemCall, a0: usize, a1: usize) usize {
     return asm volatile ("syscall"
         : [ret] "={rax}" (-> usize),
-        : [n] "{rax}" (@intFromEnum(n)), [a0] "{rdi}" (a0), [a1] "{rsi}" (a1),
+        : [n] "{rax}" (@intFromEnum(n)),
+          [a0] "{rdi}" (a0),
+          [a1] "{rsi}" (a1),
         : .{ .rcx = true, .r11 = true, .memory = true });
 }
 
 pub inline fn systemCall3(n: SystemCall, a0: usize, a1: usize, a2: usize) usize {
     return asm volatile ("syscall"
         : [ret] "={rax}" (-> usize),
-        : [n] "{rax}" (@intFromEnum(n)), [a0] "{rdi}" (a0), [a1] "{rsi}" (a1), [a2] "{rdx}" (a2),
+        : [n] "{rax}" (@intFromEnum(n)),
+          [a0] "{rdi}" (a0),
+          [a1] "{rsi}" (a1),
+          [a2] "{rdx}" (a2),
         : .{ .rcx = true, .r11 = true, .memory = true });
 }
 
 pub inline fn systemCall4(n: SystemCall, a0: usize, a1: usize, a2: usize, a3: usize) usize {
     return asm volatile ("syscall"
         : [ret] "={rax}" (-> usize),
-        : [n] "{rax}" (@intFromEnum(n)), [a0] "{rdi}" (a0), [a1] "{rsi}" (a1), [a2] "{rdx}" (a2), [a3] "{r10}" (a3),
+        : [n] "{rax}" (@intFromEnum(n)),
+          [a0] "{rdi}" (a0),
+          [a1] "{rsi}" (a1),
+          [a2] "{rdx}" (a2),
+          [a3] "{r10}" (a3),
         : .{ .rcx = true, .r11 = true, .memory = true });
 }
 
 pub inline fn systemCall5(n: SystemCall, a0: usize, a1: usize, a2: usize, a3: usize, a4: usize) usize {
     return asm volatile ("syscall"
         : [ret] "={rax}" (-> usize),
-        : [n] "{rax}" (@intFromEnum(n)), [a0] "{rdi}" (a0), [a1] "{rsi}" (a1), [a2] "{rdx}" (a2), [a3] "{r10}" (a3), [a4] "{r8}" (a4),
+        : [n] "{rax}" (@intFromEnum(n)),
+          [a0] "{rdi}" (a0),
+          [a1] "{rsi}" (a1),
+          [a2] "{rdx}" (a2),
+          [a3] "{r10}" (a3),
+          [a4] "{r8}" (a4),
         : .{ .rcx = true, .r11 = true, .memory = true });
 }

library/runtime/system.zig

@@ -2,6 +2,7 @@
 //! stubs, one per kernel call. Numbers come from `abi.SystemCall`, the single
 //! source of truth shared with the kernel dispatcher.
 
+const std = @import("std");
 const abi = @import("abi");
 const sc = @import("system-call.zig");
 
@@ -11,6 +12,10 @@ pub const PROT_READ: usize = abi.prot_read;
 pub const PROT_WRITE: usize = abi.prot_write;
 pub const PROT_EXEC: usize = abi.prot_exec;
 
+/// One `processes` entry — re-exported from the shared ABI so a user program can
+/// declare its snapshot buffer without importing `abi` itself.
+pub const ProcessDescriptor = abi.ProcessDescriptor;
+
 /// Give up the rest of this quantum.
 pub fn yield() void {
     _ = sc.systemCall0(.yield);
@@ -27,6 +32,15 @@ pub fn sleep(ms: usize) void {
     _ = sc.systemCall1(.sleep, ms);
 }
 
+/// Arm a one-shot timer: after `ms` milliseconds the kernel posts a timer
+/// notification (`ipc.Received.isTimer`) to `endpoint`. The timed wait of
+/// docs/process-lifecycle.md — a service arms a deadline and keeps serving,
+/// instead of blocking in sleep; what stop-sequence escalation, hello deadlines,
+/// and restart backoff are built from.
+pub fn timerOnce(endpoint: usize, ms: u64) bool {
+    return sc.systemCall2(.timer_bind, endpoint, ms) == 0;
+}
+
 /// Monotonic nanoseconds since boot — a time source for timeouts and short delays. It
 /// only ever moves forward. This is *not* wall-clock time (no date, no timezone — that
 /// is a user-space service layered on top). Deadline pattern for a bounded poll loop:
@@ -37,6 +51,24 @@ pub fn clock() u64 {
     return @intCast(sc.systemCall0(.clock));
 }
 
+/// Wall-clock time in Unix epoch seconds (UTC) — the real date/time, from the RTC.
+/// Unlike `clock` (monotonic since boot), this tracks calendar time, so it is what a
+/// filesystem stamps as a file's modification time. Formatting it into a calendar
+/// date/timezone is user-space policy layered on top.
+pub fn wallClock() u64 {
+    return @intCast(sc.systemCall0(.wall_clock));
+}
+
+/// Copy bytes out of the kernel's in-memory diagnostic log — the accumulated
+/// stream of everything `write` (and the kernel itself) has emitted — starting at
+/// `offset`, into `out`. Returns the number of bytes copied (0 at end of buffer).
+/// A program reads the whole log by looping from offset 0, advancing by the return
+/// value, until it gets 0. This is how the boot log is persisted to disk on a
+/// headless/real machine where serial output is otherwise lost.
+pub fn klogRead(offset: usize, out: []u8) usize {
+    return sc.systemCall3(.klog_read, offset, @intFromPtr(out.ptr), out.len);
+}
+
 /// End the process. Never returns.
 pub fn exit(code: usize) noreturn {
     _ = sc.systemCall1(.exit, code);
@@ -44,11 +76,74 @@ pub fn exit(code: usize) noreturn {
 }
 
 /// Start the binary bundled in the initial-ramdisk under `name` as a new ring-3
-/// process, returning true on success. This is how a supervisor (the device manager)
-/// launches a driver it matched — danos-native, not POSIX (a spawn/exec family comes
-/// with the process work later).
-pub fn spawn(name: []const u8) bool {
-    return sc.systemCall2(.system_spawn, @intFromPtr(name.ptr), name.len) == 0;
+/// process, returning the child's process id (or null on failure). The child's
+/// argv[0] is `name`, and the caller becomes its **supervisor** — the only process
+/// allowed to `kill` it. This is how a supervisor (the device manager) launches a
+/// driver it matched — danos-native, not POSIX (a spawn/exec family comes with the
+/// POSIX layer later).
+pub fn spawn(name: []const u8) ?u32 {
+    return spawnSupervised(name, &.{}, null);
+}
+
+/// Like `spawn`, but hands the child command-line arguments: they arrive as
+/// argv[1..] on its System V entry stack (argv[0] is still `name`).
+pub fn spawnWithArguments(name: []const u8, arguments: []const []const u8) ?u32 {
+    return spawnSupervised(name, arguments, null);
+}
+
+/// The full spawn: command-line arguments for the child, and an optional endpoint
+/// (a handle from `ipc.createIpcEndpoint`) the kernel notifies when the child ends
+/// — any way it ends: clean exit, fault, or `kill`. The notification arrives via
+/// `ipc.replyWait` as a badge with the child-exit bit set and the child's id in
+/// the low bits (`ipc.Received.isChildExit`/`childProcessId`), so one endpoint can
+/// supervise many children. Arguments are marshalled to the kernel as one
+/// NUL-separated blob; the combined arguments must fit `blob` (the kernel caps the
+/// blob at 256 bytes and argc at 8 anyway). Returns the child's process id, or
+/// null on failure.
+pub fn spawnSupervised(name: []const u8, arguments: []const []const u8, exit_endpoint: ?usize) ?u32 {
+    var blob: [256]u8 = undefined;
+    var len: usize = 0;
+    for (arguments, 0..) |argument, i| {
+        if (i != 0) {
+            if (len >= blob.len) return null;
+            blob[len] = 0;
+            len += 1;
+        }
+        if (len + argument.len > blob.len) return null;
+        @memcpy(blob[len..][0..argument.len], argument);
+        len += argument.len;
+    }
+    const r = sc.systemCall5(.system_spawn, @intFromPtr(name.ptr), name.len, if (len == 0) 0 else @intFromPtr(&blob), len, exit_endpoint orelse abi.no_cap);
+    if (r > ~@as(usize, 0) - 4095) return null; // a wrapped -errno
+    return @intCast(r);
+}
+
+/// Snapshot the process table into `out` (up to its length) and return the total
+/// number of live processes — which may exceed `out.len`; call again with a larger
+/// buffer for the full listing. Kernel tasks are included, with an empty name.
+/// The primitive `ps` is built on.
+pub fn processes(out: []abi.ProcessDescriptor) usize {
+    return sc.systemCall2(.process_enumerate, @intFromPtr(out.ptr), out.len);
+}
+
+/// Whether a process spawned under `name` (its argv[0]) is currently alive.
+pub fn isProcessRunning(name: []const u8) bool {
+    var table: [32]ProcessDescriptor = undefined;
+    const total = processes(&table);
+    for (table[0..@min(total, table.len)]) |descriptor| {
+        if (std.mem.eql(u8, descriptor.name[0..descriptor.name_length], name)) return true;
+    }
+    return false;
+}
+
+/// End process `id`. Only its supervisor — the process that spawned it — may;
+/// anyone else gets false, as does a stale or unknown id (ids are never reused).
+/// Delivery is prompt but asynchronous, like a signal: a target caught running on
+/// another core dies at its next system call or timer tick. True means the kill
+/// is accepted and irrevocable; the exit notification (if an endpoint was given
+/// at spawn) confirms completion.
+pub fn kill(id: u32) bool {
+    return sc.systemCall1(.process_kill, id) == 0;
 }
 
 /// Grant `len` bytes (rounded up to whole pages) of fresh, zeroed, writable

library/runtime/time.zig

@@ -0,0 +1,169 @@
+//! The danos time interface — monotonic time, delays, and deadlines for user space.
+//!
+//! There is no time *service*: the kernel already owns the scheduling timer and
+//! surfaces it directly, so reading the clock is one system call (an `rdtsc` and a
+//! scale), never an IPC round trip (docs/timers.md explains why). This module is a
+//! thin, generic layer over the `clock`/`sleep`/`timer_bind` wrappers in `system.zig`
+//! — an ergonomic `Instant`/`Duration` front door, not new mechanism.
+//!
+//! It is **monotonic** time only: nanoseconds since boot, moving forward, no date or
+//! timezone. Wall-clock/calendar time is a separate user-space service (an RTC-backed
+//! CLOCK_REALTIME) layered on top later.
+
+const std = @import("std");
+const system = @import("system.zig");
+
+const nanos_per_micro: u64 = 1_000;
+const nanos_per_milli: u64 = 1_000_000;
+const nanos_per_second: u64 = 1_000_000_000;
+
+/// A span of time, held as nanoseconds. Constructors name their unit; accessors
+/// truncate toward zero. `ceilMillis` rounds *up*, since `sleep`/`after` land on the
+/// kernel's millisecond granularity and rounding down could return early.
+pub const Duration = struct {
+    ns: u64,
+
+    pub fn fromNanos(n: u64) Duration {
+        return .{ .ns = n };
+    }
+    pub fn fromMicros(n: u64) Duration {
+        return .{ .ns = n *| nanos_per_micro };
+    }
+    pub fn fromMillis(n: u64) Duration {
+        return .{ .ns = n *| nanos_per_milli };
+    }
+    pub fn fromSeconds(n: u64) Duration {
+        return .{ .ns = n *| nanos_per_second };
+    }
+
+    pub fn asNanos(d: Duration) u64 {
+        return d.ns;
+    }
+    pub fn asMicros(d: Duration) u64 {
+        return d.ns / nanos_per_micro;
+    }
+    pub fn asMillis(d: Duration) u64 {
+        return d.ns / nanos_per_milli;
+    }
+    pub fn asSeconds(d: Duration) u64 {
+        return d.ns / nanos_per_second;
+    }
+
+    /// Whole milliseconds, rounded up — the argument `sleep`/`after` pass the kernel.
+    /// A non-zero sub-millisecond duration becomes 1 ms rather than 0.
+    pub fn ceilMillis(d: Duration) u64 {
+        return (d.ns +| (nanos_per_milli - 1)) / nanos_per_milli;
+    }
+
+    pub fn plus(a: Duration, b: Duration) Duration {
+        return .{ .ns = a.ns +| b.ns };
+    }
+};
+
+/// A point on the monotonic clock — nanoseconds since boot. Compare and subtract
+/// instants to measure elapsed time; it never runs backward, so `since` is safe to
+/// saturate at zero rather than wrap.
+pub const Instant = struct {
+    ns: u64,
+
+    /// The span from `earlier` to `self`, saturating at zero if `earlier` is later
+    /// (which the monotonic clock should never produce, but callers may pass any pair).
+    pub fn since(self: Instant, earlier: Instant) Duration {
+        return .{ .ns = self.ns -| earlier.ns };
+    }
+
+    /// How long since this instant, sampled now.
+    pub fn elapsed(self: Instant) Duration {
+        return now().since(self);
+    }
+
+    /// This instant advanced by `d` (a deadline, `d` from here).
+    pub fn plus(self: Instant, d: Duration) Instant {
+        return .{ .ns = self.ns +| d.ns };
+    }
+
+    /// Whether the monotonic clock has reached this instant (used as a deadline).
+    pub fn reached(deadline: Instant) bool {
+        return now().ns >= deadline.ns;
+    }
+};
+
+/// The current monotonic time.
+pub fn now() Instant {
+    return .{ .ns = system.clock() };
+}
+
+/// Monotonic nanoseconds since boot — the raw `clock()` reading, for callers that
+/// want a plain integer instead of an `Instant`.
+pub fn monotonicNanos() u64 {
+    return system.clock();
+}
+
+/// Whether the monotonic clock is usable. The kernel returns 0 until the TSC is
+/// calibrated (`tsc_hz == 0`); a caller that needs real time can treat that as
+/// "unavailable" instead of assuming the clock advances.
+pub fn available() bool {
+    return system.clock() != 0;
+}
+
+/// Block the caller for at least `d`, rounded up to the kernel's millisecond
+/// granularity. For sub-millisecond precision the scheduler cannot express, use
+/// `spin`.
+pub fn sleep(d: Duration) void {
+    system.sleep(d.ceilMillis());
+}
+
+/// Block the caller for `ms` milliseconds — the coarse, allocation-free form.
+pub fn sleepMillis(ms: u64) void {
+    system.sleep(ms);
+}
+
+/// Busy-wait until `d` has elapsed, polling the monotonic clock. This burns the CPU
+/// on purpose, to hit sub-millisecond delays the scheduler's millisecond tick cannot.
+/// Prefer `sleep` for anything at or above a millisecond.
+pub fn spin(d: Duration) void {
+    const deadline = now().plus(d);
+    while (!deadline.reached()) {}
+}
+
+/// Arm a one-shot timer against `endpoint` (a handle from `ipc.createIpcEndpoint`):
+/// after `d` the kernel posts a timer notification (`ipc.Received.isTimer`) there.
+/// Unlike `sleep`, this does not block — a service can keep serving IPC on the same
+/// endpoint while the deadline is pending. Rounds `d` up to milliseconds; returns
+/// false if the timer could not be armed. See `system.timerOnce`.
+pub fn after(endpoint: usize, d: Duration) bool {
+    return system.timerOnce(endpoint, d.ceilMillis());
+}
+
+test "Duration unit conversions round toward zero" {
+    try std.testing.expectEqual(@as(u64, 1_000_000_000), Duration.fromSeconds(1).asNanos());
+    try std.testing.expectEqual(@as(u64, 1_500), Duration.fromNanos(1_500).asNanos());
+    try std.testing.expectEqual(@as(u64, 2), Duration.fromMillis(2).asMillis());
+    try std.testing.expectEqual(@as(u64, 1), Duration.fromNanos(1_999_999).asMillis());
+    try std.testing.expectEqual(@as(u64, 250), Duration.fromMicros(250).asMicros());
+}
+
+test "ceilMillis rounds up, and never turns a nonzero span into zero" {
+    try std.testing.expectEqual(@as(u64, 0), Duration.fromNanos(0).ceilMillis());
+    try std.testing.expectEqual(@as(u64, 1), Duration.fromNanos(1).ceilMillis());
+    try std.testing.expectEqual(@as(u64, 1), Duration.fromMillis(1).ceilMillis());
+    try std.testing.expectEqual(@as(u64, 2), Duration.fromNanos(nanos_per_milli + 1).ceilMillis());
+    try std.testing.expectEqual(@as(u64, 5), Duration.fromMillis(5).ceilMillis());
+}
+
+test "Instant arithmetic: since saturates, plus/reached form deadlines" {
+    const t0 = Instant{ .ns = 1_000 };
+    const t1 = Instant{ .ns = 4_000 };
+    try std.testing.expectEqual(@as(u64, 3_000), t1.since(t0).asNanos());
+    // earlier-than-self can't happen on a monotonic clock; saturate rather than wrap.
+    try std.testing.expectEqual(@as(u64, 0), t0.since(t1).asNanos());
+    const deadline = t0.plus(Duration.fromNanos(2_500));
+    try std.testing.expectEqual(@as(u64, 3_500), deadline.ns);
+}
+
+test "saturating arithmetic does not overflow at the u64 ceiling" {
+    const big = Duration.fromSeconds(std.math.maxInt(u64));
+    try std.testing.expectEqual(@as(u64, std.math.maxInt(u64)), big.asNanos());
+    const late = Instant{ .ns = std.math.maxInt(u64) };
+    try std.testing.expectEqual(@as(u64, std.math.maxInt(u64)), late.plus(Duration.fromSeconds(10)).ns);
+}

library/runtime/usb.zig

@@ -0,0 +1,162 @@
+//! USB class-driver client: the helper a keyboard, mouse, or mass-storage driver
+//! uses to reach its device through the xHCI bus driver, so it never hand-rolls
+//! the transfer-protocol IPC. Layered over `ipc` and the shared
+//! `usb-transfer-protocol` wire format, the way `input.zig` layers over the input
+//! service and `device.zig` over the raw device calls.
+//!
+//! A class driver, spawned with its interface's assigned device id as argv[1]:
+//!     if (!usb.helloManager(id)) return;              // meet the spawn deadline
+//!     var device = usb.open(id) orelse return;        // open + get its endpoints
+//!     _ = device.controlOut(usb_abi.setProtocol(...));// class requests, descriptors
+//!     _ = device.subscribeInterrupt(address, length); // reports arrive asynchronously
+//!     while (true) { ... ipc.replyWait(device.endpoint, ...) ... }  // its own loop
+//!
+//! Reports are delivered to `device.endpoint` as asynchronous `InterruptReport`
+//! messages (the class driver runs a bare `replyWait` loop to read them, because
+//! the service harness drops buffered-message payloads — see service.zig).
+
+const std = @import("std");
+const ipc = @import("ipc.zig");
+const system = @import("system.zig");
+const protocol = @import("usb-transfer-protocol");
+const device_manager = @import("device-manager-protocol");
+
+pub const Endpoint = protocol.Endpoint;
+pub const InterruptReport = protocol.InterruptReport;
+pub const max_report_data = protocol.max_report_data;
+
+// Endpoint transfer types (EndpointDescriptor attributes), for `findEndpoint`.
+pub const transfer_type_bulk: u8 = 2;
+pub const transfer_type_interrupt: u8 = 3;
+
+/// An opened USB device: the bus endpoint to send requests to, this driver's own
+/// endpoint that reports arrive on, the device token, and the interface's
+/// endpoints (so a driver need not re-read the configuration descriptor).
+pub const Device = struct {
+    bus: ipc.Handle,
+    endpoint: ipc.Handle,
+    token: u64,
+    class: u8,
+    subclass: u8,
+    protocol_code: u8,
+    interface_number: u8,
+    endpoint_count: usize = 0,
+    endpoints: [protocol.max_reported_endpoints]Endpoint = undefined,
+
+    /// The interface's first endpoint of the given transfer type and direction
+    /// (`transfer_type_bulk` / `transfer_type_interrupt`), or null.
+    pub fn findEndpoint(self: *const Device, transfer_type: u8, direction_in: bool) ?Endpoint {
+        for (self.endpoints[0..self.endpoint_count]) |endpoint| {
+            if (endpoint.transfer_type == transfer_type and (endpoint.address & 0x80 != 0) == direction_in) return endpoint;
+        }
+        return null;
+    }
+
+    fn controlTransfer(self: *Device, setup: [8]u8, direction_in: bool, data: []u8) ?usize {
+        var request = protocol.ControlRequest{
+            .device_token = self.token,
+            .setup = setup,
+            .direction_in = @intFromBool(direction_in),
+            .data_length = @intCast(data.len),
+        };
+        if (!direction_in and data.len > 0) @memcpy(request.data[0..data.len], data);
+        var reply: [@sizeOf(protocol.ControlReply)]u8 = undefined;
+        const length = ipc.call(self.bus, std.mem.asBytes(&request), &reply) catch return null;
+        if (length < @sizeOf(protocol.ControlReply)) return null;
+        const control_reply = std.mem.bytesToValue(protocol.ControlReply, reply[0..@sizeOf(protocol.ControlReply)]);
+        if (control_reply.status != 0) return null;
+        const actual = @min(control_reply.actual_length, data.len);
+        if (direction_in and actual > 0) @memcpy(data[0..actual], control_reply.data[0..actual]);
+        return actual;
+    }
+
+    /// A control transfer with no data stage (SET_PROTOCOL, SET_IDLE, ...). The
+    /// `setup` is a bit-cast `usb_abi.Request`.
+    pub fn controlOut(self: *Device, setup: [8]u8) bool {
+        return self.controlTransfer(setup, false, &.{}) != null;
+    }
+
+    /// A device-to-host control transfer, returning the bytes read into `out`.
+    pub fn controlIn(self: *Device, setup: [8]u8, out: []u8) ?usize {
+        return self.controlTransfer(setup, true, out);
+    }
+
+    /// Begin periodic IN polling of an interrupt endpoint; reports flow back to
+    /// `self.endpoint` as asynchronous `InterruptReport` messages.
+    pub fn subscribeInterrupt(self: *Device, endpoint_address: u8, max_length: u16) bool {
+        var request = protocol.InterruptSubscribeRequest{
+            .device_token = self.token,
+            .endpoint_address = endpoint_address,
+            .max_length = max_length,
+        };
+        var reply: [@sizeOf(protocol.InterruptSubscribeReply)]u8 = undefined;
+        const length = ipc.call(self.bus, std.mem.asBytes(&request), &reply) catch return false;
+        if (length < @sizeOf(protocol.InterruptSubscribeReply)) return false;
+        return std.mem.bytesToValue(protocol.InterruptSubscribeReply, reply[0..@sizeOf(protocol.InterruptSubscribeReply)]).status == 0;
+    }
+
+    /// One bulk transfer (IN or OUT per `endpoint_address`'s direction bit) to or
+    /// from the caller's own DMA buffer at `physical`. Returns the bytes moved.
+    pub fn bulk(self: *Device, endpoint_address: u8, physical: u64, length: u32) ?u32 {
+        var request = protocol.BulkRequest{
+            .device_token = self.token,
+            .physical_address = physical,
+            .length = length,
+            .endpoint_address = endpoint_address,
+        };
+        var reply: [@sizeOf(protocol.BulkReply)]u8 = undefined;
+        const replied = ipc.call(self.bus, std.mem.asBytes(&request), &reply) catch return null;
+        if (replied < @sizeOf(protocol.BulkReply)) return null;
+        const bulk_reply = std.mem.bytesToValue(protocol.BulkReply, reply[0..@sizeOf(protocol.BulkReply)]);
+        if (bulk_reply.status != 0) return null;
+        return bulk_reply.actual_length;
+    }
+};
+
+/// Look up the USB bus and open the device with the assigned id, handing over a
+/// freshly created endpoint for asynchronous interrupt reports. Retries while the
+/// bus is still coming up (a class driver races the bus driver at boot).
+pub fn open(device_id: u64) ?Device {
+    var attempts: usize = 0;
+    const bus = while (attempts < 100) : (attempts += 1) {
+        if (ipc.lookup(.usb_bus)) |handle| break handle;
+        system.sleep(20);
+    } else return null;
+
+    const endpoint = ipc.createIpcEndpoint() orelse return null;
+    var request = protocol.OpenRequest{ .device_id = device_id };
+    var reply: [@sizeOf(protocol.OpenReply)]u8 = undefined;
+    const result = ipc.callCap(bus, std.mem.asBytes(&request), &reply, endpoint) catch return null;
+    if (result.len < @sizeOf(protocol.OpenReply)) return null;
+    const open_reply = std.mem.bytesToValue(protocol.OpenReply, reply[0..@sizeOf(protocol.OpenReply)]);
+    if (open_reply.status != 0) return null;
+
+    var device = Device{
+        .bus = bus,
+        .endpoint = endpoint,
+        .token = open_reply.device_token,
+        .class = open_reply.interface_class,
+        .subclass = open_reply.interface_subclass,
+        .protocol_code = open_reply.interface_protocol,
+        .interface_number = open_reply.interface_number,
+        .endpoint_count = @min(open_reply.endpoint_count, protocol.max_reported_endpoints),
+    };
+    for (0..device.endpoint_count) |index| device.endpoints[index] = open_reply.endpoints[index];
+    return device;
+}
+
+/// Hello the device manager as a class driver (Role.device) so a supervised
+/// spawn meets its hello deadline. Retries while the manager comes up.
+pub fn helloManager(device_id: u64) bool {
+    var attempts: usize = 0;
+    const manager = while (attempts < 100) : (attempts += 1) {
+        if (ipc.lookup(.device_manager)) |handle| break handle;
+        system.sleep(20);
+    } else return false;
+
+    const hello = device_manager.Hello{ .role = @intFromEnum(device_manager.Role.device), .device_id = device_id };
+    var reply: [device_manager.message_maximum]u8 = undefined;
+    const length = ipc.call(manager, std.mem.asBytes(&hello), &reply) catch return false;
+    if (length < device_manager.reply_size) return false;
+    return std.mem.bytesToValue(device_manager.HelloReply, reply[0..device_manager.reply_size]).status == 0;
+}

library/xkeyboard-config/README.md

@@ -0,0 +1,65 @@
+# xkeyboard-config — X11 keyboard layouts, compiled to Zig
+
+This module turns a physical key (a **USB HID usage**, as the [input module](../../docs/input.md)
+delivers in `KeyEvent.keycode`) plus a modifier state into a **keysym** and, when the key
+produces one, a **character** (a Unicode scalar). It is what lets a `keycode` become a
+`character` — a keymap — without danos shipping an X11 runtime.
+
+The layout data comes from the X11 [xkeyboard-config](https://gitlab.freedesktop.org/xkeyboard-config/xkeyboard-config)
+database, but it is **compiled to native Zig at build time** rather than parsed at runtime.
+`tools/make-xkeyboard-config.py` reads the vendored xkb data and emits pure-data tables into
+`generated/layouts.zig`; `xkeyboard-config.zig` is the hand-written API over them. This is
+the same build-time-codegen pattern as `tools/make-initial-ramdisk.py`.
+
+## Using it
+
+```zig
+const xkb = @import("xkeyboard-config");
+
+const m = xkb.map(xkb.us, key_event.keycode, .{ .shift = shift_held, .caps_lock = caps });
+if (m.character) |ch| { /* a printable Unicode scalar */ }
+// m.keysym is always set (e.g. an X11 keysym for Return / F1 / a dead key).
+
+const layout = xkb.byName("gb") orelse xkb.us;   // choose a layout by name
+for (xkb.all) |l| { /* enumerate available layouts */ }
+```
+
+`Modifiers` carries `shift`, `caps_lock`, `level3` (AltGr), and `control`. `map` selects the
+level from the key's XKB *type* (the generated data) and those modifiers (the policy, in
+`xkeyboard-config.zig`), so data and semantics stay separable.
+
+Layouts: **us, gb, de, fr, es, dvorak**.
+
+## Regenerating
+
+```sh
+python3 tools/make-xkeyboard-config.py fetch      # network: download + vendor the data subset
+python3 tools/make-xkeyboard-config.py generate   # offline: emit generated/layouts.zig
+# or, from the build:
+zig build gen-xkeyboard-config
+```
+
+- **`fetch`** downloads the pinned xkeyboard-config release (version + sha256 in the script),
+  resolves the `include` graph for the configured layouts, and vendors *only* the symbols
+  files actually reached (plus `keysymdef.h`, `COPYING`, and `PROVENANCE.md`) into `vendor/`.
+  Run it when bumping the version or adding a layout.
+- **`generate`** is deterministic and offline — same vendored input produces byte-identical
+  output. To add a layout, extend `TARGETS` (and `HID_TO_NAME` if a new physical key is
+  involved), then re-run `fetch` (to vendor any new includes) and `generate`.
+
+## Scope
+
+A pragmatic subset, enough for real Latin-script typing:
+
+- **Group 1 only** — no multi-layout group switching.
+- **No dead-key / compose composition** — a dead key returns its keysym with no `character`
+  (composing `´` + `e` → `é` is a higher layer's job).
+- **Curated key types** — the common XKB types (one/two-level, alphabetic, four-level, …);
+  unmapped keys and unknown types fall back to level-by-shift.
+- **6 layouts** — extend via `TARGETS` as above.
+
+## Licensing
+
+xkeyboard-config and `keysymdef.h` (xorgproto) are MIT/X11 licensed. The vendored data
+subset carries the upstream `vendor/COPYING`, and `vendor/PROVENANCE.md` records the exact
+version, source URL, and sha256. The generated tables are a derived work under the same terms.

library/xkeyboard-config/vendor/COPYING

@@ -0,0 +1,190 @@
+Copyright 1996 by Joseph Moss
+Copyright (C) 2002-2007 Free Software Foundation, Inc.
+Copyright (C) Dmitry Golubev <lastguru@mail.ru>, 2003-2004
+Copyright (C) 2004, Gregory Mokhin <mokhin@bog.msu.ru>
+Copyright (C) 2006 Erdal Ronahî
+
+Permission to use, copy, modify, distribute, and sell this software and its
+documentation for any purpose is hereby granted without fee, provided that
+the above copyright notice appear in all copies and that both that
+copyright notice and this permission notice appear in supporting
+documentation, and that the name of the copyright holder(s) not be used in
+advertising or publicity pertaining to distribution of the software without
+specific, written prior permission.  The copyright holder(s) makes no
+representations about the suitability of this software for any purpose.  It
+is provided "as is" without express or implied warranty.
+
+THE COPYRIGHT HOLDER(S) DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
+INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO
+EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY SPECIAL, INDIRECT OR
+CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
+TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
+PERFORMANCE OF THIS SOFTWARE.
+
+
+Copyright (c) 1996  Digital Equipment Corporation
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be included
+in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL DIGITAL EQUIPMENT CORPORATION BE LIABLE FOR ANY CLAIM,
+DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
+THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of the Digital Equipment
+Corporation shall not be used in advertising or otherwise to promote
+the sale, use or other dealings in this Software without prior written
+authorization from Digital Equipment Corporation.
+
+
+Copyright 1996, 1998  The Open Group
+
+Permission to use, copy, modify, distribute, and sell this software and its
+documentation for any purpose is hereby granted without fee, provided that
+the above copyright notice appear in all copies and that both that
+copyright notice and this permission notice appear in supporting
+documentation.
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE OPEN GROUP BE LIABLE FOR ANY CLAIM, DAMAGES OR
+OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of The Open Group shall
+not be used in advertising or otherwise to promote the sale, use or
+other dealings in this Software without prior written authorization
+from The Open Group.
+
+
+Copyright 2004-2005 Sun Microsystems, Inc.  All rights reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a
+copy of this software and associated documentation files (the "Software"),
+to deal in the Software without restriction, including without limitation
+the rights to use, copy, modify, merge, publish, distribute, sublicense,
+and/or sell copies of the Software, and to permit persons to whom the
+Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice (including the next
+paragraph) shall be included in all copies or substantial portions of the
+Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+DEALINGS IN THE SOFTWARE.
+
+
+Copyright (c) 1996 by Silicon Graphics Computer Systems, Inc.
+
+Permission to use, copy, modify, and distribute this
+software and its documentation for any purpose and without
+fee is hereby granted, provided that the above copyright
+notice appear in all copies and that both that copyright
+notice and this permission notice appear in supporting
+documentation, and that the name of Silicon Graphics not be
+used in advertising or publicity pertaining to distribution
+of the software without specific prior written permission.
+Silicon Graphics makes no representation about the suitability
+of this software for any purpose. It is provided "as is"
+without any express or implied warranty.
+
+SILICON GRAPHICS DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS
+SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
+AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT SHALL SILICON
+GRAPHICS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL
+DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE,
+DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
+OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION  WITH
+THE USE OR PERFORMANCE OF THIS SOFTWARE.
+
+
+Copyright (c) 1996  X Consortium
+
+Permission is hereby granted, free of charge, to any person obtaining
+a copy of this software and associated documentation files (the
+"Software"), to deal in the Software without restriction, including
+without limitation the rights to use, copy, modify, merge, publish,
+distribute, sublicense, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR
+OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of the X Consortium shall
+not be used in advertising or otherwise to promote the sale, use or
+other dealings in this Software without prior written authorization
+from the X Consortium.
+
+
+Copyright (C) 2004, 2006 Ævar Arnfjörð Bjarmason <avarab@gmail.com>
+
+Permission to use, copy, modify, distribute, and sell this software and its
+documentation for any purpose is hereby granted without fee, provided that
+the above copyright notice appear in all copies and that both that
+copyright notice and this permission notice appear in supporting
+documentation.
+
+The above copyright notice and this permission notice shall be
+included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+IN NO EVENT SHALL THE OPEN GROUP BE LIABLE FOR ANY CLAIM, DAMAGES OR
+OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+OTHER DEALINGS IN THE SOFTWARE.
+
+Except as contained in this notice, the name of a copyright holder shall
+not be used in advertising or otherwise to promote the sale, use or
+other dealings in this Software without prior written authorization of
+the copyright holder.
+
+
+Copyright (C) 1999, 2000 by Anton Zinoviev <anton@lml.bas.bg>
+
+This software may be used, modified, copied, distributed, and sold,
+in both source and binary form provided that the above copyright
+and these terms are retained. Under no circumstances is the author
+responsible for the proper functioning of this software, nor does
+the author assume any responsibility for damages incurred with its
+use.
+
+Permission is granted to anyone to use, distribute and modify
+this file in any way, provided that the above copyright notice
+is left intact and the author of the modification summarizes
+the changes in this header.
+
+This file is distributed without any expressed or implied warranty.

library/xkeyboard-config/vendor/PROVENANCE.md

@@ -0,0 +1,21 @@
+# Vendored xkeyboard-config subset
+
+- **Package**: xkeyboard-config 2.44
+- **Source**: https://gitlab.freedesktop.org/xkeyboard-config/xkeyboard-config/-/archive/xkeyboard-config-2.44/xkeyboard-config-2.44.tar.gz
+- **sha256**: `35e34edeaf4e8da8d0696ff6b241ee11ddb1b8c6730bac7252d4d0a88ea5f05b`
+- **keysymdef.h**: xorgproto, copied from `/opt/homebrew/include/X11/keysymdef.h`
+- **License**: MIT/X11 (see COPYING)
+
+Only the symbols files reachable from the generated layouts (tools/make-xkeyboard-config.py `TARGETS`) are vendored; regenerate with
+`python3 tools/make-xkeyboard-config.py fetch` then `... generate`.
+
+Vendored symbols files:
+
+- `symbols/de`
+- `symbols/es`
+- `symbols/fr`
+- `symbols/gb`
+- `symbols/kpdl`
+- `symbols/latin`
+- `symbols/level3`
+- `symbols/us`

library/xkeyboard-config/vendor/symbols/es

@@ -0,0 +1,250 @@
+// Keyboard layouts for Spain.
+
+// Modified for a real Spanish keyboard by Jon Tombs.
+default partial alphanumeric_keys
+xkb_symbols "basic" {
+
+    include "latin(type4)"
+
+    name[Group1]="Spanish";
+
+    key <TLDE>	{ [ masculine, ordfeminine,    backslash,   backslash ]	};
+    key <AE01>	{ [         1,     exclam,          bar,   exclamdown ]	};
+    key <AE03>	{ [         3, periodcentered, numbersign,   sterling ]	};
+    key <AE04>	{ [         4,     dollar,   asciitilde,       dollar ]	};
+    key <AE11>	{ [apostrophe,   question,    backslash, questiondown ]	};
+    key <AE12>	{ [exclamdown, questiondown, dead_cedilla, dead_ogonek]	};
+
+    key <AD11>	{ [dead_grave, dead_circumflex, bracketleft, dead_abovering ] };
+    key <AD12>	{ [      plus,   asterisk, bracketright,  dead_macron ]	};
+
+    key <AC10>	{ [    ntilde, Ntilde,   dead_tilde, dead_doubleacute ]	};
+    key <AC11>	{ [dead_acute, dead_diaeresis, braceleft,  dead_caron ]	};
+    key <BKSL>	{ [  ccedilla,   Ccedilla,   braceright,   dead_breve ]	};
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "winkeys" {
+
+    include "es(basic)"
+    name[Group1]="Spanish (Windows)";
+    include "eurosign(5)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "nodeadkeys" {
+
+    include "es(basic)"
+
+    name[Group1]="Spanish (no dead keys)";
+
+    key <AE12>	{ [exclamdown, questiondown,     cedilla,      ogonek ]	};
+    key <AD11>	{ [     grave, asciicircum,  bracketleft,      degree ]	};
+    key <AD12>	{ [      plus,   asterisk, bracketright,       macron ]	};
+    key <AC07>	{ [         j,          J,          ezh,          EZH ] };
+    key <AC10>	{ [    ntilde,     Ntilde,   asciitilde,  doubleacute ]	};
+    key <AC11>	{ [     acute,  diaeresis,    braceleft,        caron ]	};
+    key <BKSL>	{ [  ccedilla,   Ccedilla,   braceright,        breve ]	};
+    key <AB10>	{ [     minus, underscore,     ellipsis,     abovedot ]	};
+};
+
+// Spanish Dvorak mapping (note R-H exchange)
+partial alphanumeric_keys
+xkb_symbols "dvorak" {
+
+    name[Group1]="Spanish (Dvorak)";
+
+    key <TLDE> {[  masculine, ordfeminine, backslash, degree		]};
+    key <AE01> {[          1, exclam, bar, onesuperior			]};
+    key <AE02> {[          2, quotedbl, at, twosuperior			]};
+    key <AE03> {[          3, periodcentered, numbersign, threesuperior	]};
+    key <AE04> {[          4, dollar, asciitilde, onequarter		]};
+    key <AE05> {[          5, percent, brokenbar, fiveeighths		]};
+    key <AE06> {[          6, ampersand, notsign, threequarters		]};
+    key <AE07> {[          7, slash, onehalf, seveneighths		]};
+    key <AE08> {[          8, parenleft, oneeighth, threeeighths	]};
+    key <AE09> {[          9, parenright, asciicircum			]};
+    key <AE10> {[          0, equal, grave, dead_doubleacute		]};
+    key <AE11> {[ apostrophe, question, dead_macron, dead_ogonek	]};
+    key <AE12> {[ exclamdown, questiondown, dead_breve, dead_abovedot	]};
+
+    key <AD01> {[     period, colon, less, guillemotleft                ]};
+    key <AD02> {[      comma, semicolon, greater, guillemotright        ]};
+    key <AD03> {[     ntilde, Ntilde, lstroke, Lstroke			]};
+    key <AD04> {[          p, P, paragraph				]};
+    key <AD05> {[          y, Y, yen					]};
+    key <AD06> {[          f, F, tslash, Tslash				]};
+    key <AD07> {[          g, G, dstroke, Dstroke			]};
+    key <AD08> {[          c, C, cent, copyright			]};
+    key <AD09> {[          h, H, hstroke, Hstroke			]};
+    key <AD10> {[          l, L, sterling				]};
+    key <AD11> {[ dead_grave, dead_circumflex, bracketleft, dead_caron	]};
+    key <AD12> {[       plus, asterisk, bracketright, plusminus		]};
+
+    key <AC01> {[          a, A, ae, AE					]};
+    key <AC02> {[          o, O, oslash, Oslash				]};
+    key <AC03> {[          e, E, EuroSign				]};
+    key <AC04> {[          u, U, aring, Aring				]};
+    key <AC05> {[          i, I, oe, OE					]};
+    key <AC06> {[          d, D, eth, ETH				]};
+    key <AC07> {[          r, R, registered, trademark			]};
+    key <AC08> {[          t, T, thorn, THORN				]};
+    key <AC09> {[          n, N, eng, ENG				]};
+    key <AC10> {[          s, S, ssharp, section			]};
+    key <AC11> {[ dead_acute, dead_diaeresis, braceleft, dead_tilde	]};
+    key <BKSL> {[   ccedilla, Ccedilla, braceright, dead_cedilla	]};
+
+    key <LSGT> {[       less, greater, guillemotleft, guillemotright	]};
+    key <AB01> {[      minus, underscore, hyphen, macron		]};
+    key <AB02> {[          q, Q, currency				]};
+    key <AB03> {[          j, J						]};
+    key <AB04> {[          k, K, kra					]};
+    key <AB05> {[          x, X, multiply, division			]};
+    key <AB06> {[          b, B						]};
+    key <AB07> {[          m, M, mu					]};
+    key <AB08> {[          w, W						]};
+    key <AB09> {[          v, V						]};
+    key <AB10> {[          z, Z						]};
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "cat" {
+
+    include "es(basic)"
+
+    name[Group1]="Catalan (Spain, with middle-dot L)";
+
+    key <AC09>  { [	    l,      L,		0x1000140,  0x100013F ] };
+};
+
+partial alphanumeric_keys
+xkb_symbols "ast" {
+
+    include "es(basic)"
+
+    name[Group1]="Asturian (Spain, with bottom-dot H and L)";
+
+    key <AC06>  { [	    h,      H,		0x1001E25,  0x1001E24 ] };
+    key <AC09>  { [	    l,      L,		0x1001E37,  0x1001E36 ] };
+};
+
+
+partial alphanumeric_keys
+xkb_symbols "olpc" {
+
+    // #HW-SPECIFIC
+
+    // http://wiki.laptop.org/go/OLPC_Spanish_Keyboard
+
+    include "us(basic)"
+    name[Group1]="Spanish";
+
+    key <AE00> { [    masculine,    ordfeminine      ] };
+    key <AE01> { [            1,         exclam,             bar ] };
+    key <AE02> { [            2,       quotedbl,              at ] };
+    key <AE03> { [            3,     dead_grave,      numbersign,   grave           ] };
+    key <AE05> { [            5,        percent,     asciicircum,   dead_circumflex ] };
+    key <AE06> { [            6,      ampersand,         notsign ] };
+    key <AE07> { [            7,          slash,       backslash ] };
+    key <AE08> { [            8,      parenleft                         ] };
+    key <AE09> { [            9,     parenright                         ] };
+    key <AE10> { [            0,          equal                  ] };
+    key <AE11> { [   apostrophe,       question                  ] };
+    key <AE12> { [   exclamdown,   questiondown                  ] };
+
+    key <AD03> { [            e,              E,        EuroSign ] };
+    key <AD11> { [   dead_acute, dead_diaeresis,           acute,   dead_abovering  ] };
+    key <AD12> { [  bracketleft,      braceleft                  ] };
+
+    key <AC10> { [       ntilde,        Ntilde                  ] };
+    key <AC11> { [         plus,       asterisk,      dead_tilde ] };
+    key <AC12> { [ bracketright,     braceright,      section    ] };
+
+    key <AB08> { [        comma,      semicolon                  ] };
+    key <AB09> { [       period,          colon                  ] };
+    key <AB10> { [        minus,     underscore                  ] };
+
+    key <I219> { [         less,        greater,  ISO_Next_Group ] };
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "olpcm" {
+
+    // #HW-SPECIFIC
+
+    // Mechanical (non-membrane) OLPC Spanish keyboard layout.
+    // See: http://wiki.laptop.org/go/OLPC_Spanish_Non-membrane_Keyboard
+
+    include "us(basic)"
+    name[Group1]="Spanish";
+
+    key <AE00> { [    questiondown,    exclamdown, backslash      ] };
+    key <AE01> { [            1,         exclam,             bar ] };
+    key <AE02> { [            2,       quotedbl,              at ] };
+    key <AE03> { [            3,     dead_grave,      numbersign,   grave           ] };
+    key <AE04> { [            4,     dollar,     asciitilde,     dead_tilde ] };
+    key <AE05> { [            5,        percent,     asciicircum,   dead_circumflex ] };
+    key <AE06> { [            6,      ampersand,         notsign ] };
+    key <AE07> { [            7,          slash,       backslash ] };   // no '\' label on olpcm, leave for compatibility
+    key <AE08> { [            8,      parenleft,    masculine                ] };
+    key <AE09> { [            9,     parenright,    ordfeminine                 ] };
+    key <AE10> { [            0,          equal                  ] };
+    key <AE11> { [   apostrophe,       question                  ] };
+
+    key <AD03> { [            e,              E,        EuroSign ] };
+    key <AD11> { [   dead_acute, dead_diaeresis,     dead_abovering, acute  ] };
+    key <AD12> { [  plus,      asterisk                  ] };
+
+    key <AC10> { [       ntilde,        Ntilde                  ] };
+    // no AC11 or AC12 on olpcm
+
+    key <AB08> { [        comma,      semicolon                  ] };
+    key <AB09> { [       period,          colon                  ] };
+    key <AB10> { [        minus,     underscore                  ] };
+
+    key <AA02>  { [        less,        greater ] };
+    key <AA06>  { [  bracketleft,      braceleft,   ccedilla,         Ccedilla  ] };
+    key <AA07>  { [  bracketright,     braceright                ] };
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "deadtilde" {
+
+    include "es(basic)"
+
+    name[Group1]="Spanish (dead tilde)";
+
+    key <AE04>	{ [         4,     dollar,   dead_tilde,       dollar ]	};
+    key <AC10>	{ [    ntilde, Ntilde,   asciitilde, dead_doubleacute ]	};
+};
+
+partial alphanumeric_keys
+xkb_symbols "olpc2" {
+    // #HW-SPECIFIC
+
+    // Modified variant of US International layout, specifically for Peru
+    // Contact: Sayamindu Dasgupta <sayamindu@laptop.org>
+
+    include "us(olpc)"
+    name[Group1]="Spanish";
+
+    key <AE03> { [  3, numbersign,   dead_grave,  dead_grave] }; // combining grave
+    key <I236>  { [ XF86Start ] };
+
+    include "level3(ralt_switch)"
+};
+
+// EXTRAS:
+
+partial alphanumeric_keys
+	xkb_symbols "sun_type6" {
+	include "sun_vndr/es(sun_type6)"
+};

library/xkeyboard-config/vendor/symbols/gb

@@ -0,0 +1,249 @@
+// Keyboard layouts for Great Britain.
+
+default partial alphanumeric_keys
+xkb_symbols "basic" {
+
+    // The basic UK layout, also known as the IBM 166 layout,
+    // but with the useless brokenbar pushed two levels up.
+
+    include "latin"
+
+    name[Group1]="English (UK)";
+
+    key <TLDE>	{ [     grave,    notsign,          bar,          bar ]	};
+    key <AE02>	{ [         2,   quotedbl,  twosuperior,    oneeighth ]	};
+    key <AE03>	{ [         3,   sterling, threesuperior,    sterling ]	};
+    key <AE04>	{ [         4,     dollar,     EuroSign,   onequarter ]	};
+
+    key <AC11>	{ [apostrophe,         at, dead_circumflex, dead_caron]	};
+    key <BKSL>	{ [numbersign, asciitilde,   dead_grave,   dead_breve ]	};
+
+    key <LSGT>	{ [ backslash,        bar,          bar,    brokenbar ]	};
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "intl" {
+
+    // A UK layout but with five accents made into dead keys:
+    // grave, diaeresis, circumflex, acute, and tilde.
+    // By Phil Jones <philjones1 at blueyonder.co.uk>.
+
+    include "latin"
+
+    name[Group1]="English (UK, intl., with dead keys)";
+
+    key <TLDE>  { [ dead_grave,    notsign,            bar,         bar ] };
+    key <AE02>  { [   2,    dead_diaeresis,    twosuperior,     onehalf ] };
+    key <AE03>  { [   3,          sterling,  threesuperior,    onethird ] };
+    key <AE04>  { [   4,            dollar,       EuroSign,  onequarter ] };
+    key <AE06>  { [   6,   dead_circumflex,  threequarters,    onesixth ] };
+
+    key <AC11>  { [ dead_acute,         at,     apostrophe,         bar ] };
+    key <BKSL>  { [ numbersign, dead_tilde,            bar,         bar ] };
+
+    key <LSGT>  { [  backslash,        bar,            bar,         bar ] };
+    key <AB08>  { [      comma,       less,       ccedilla,    Ccedilla ] };
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "extd" {
+    // Clone of the Microsoft "United Kingdom Extended" layout, which
+    // includes dead keys for: grave; diaeresis; circumflex; tilde; and
+    // accute. It also enables direct access to accute characters using
+    // the Multi_key (Alt Gr).
+    //
+    // Taken from...
+    //    "Windows Keyboard Layouts"
+    //    https://docs.microsoft.com/en-gb/globalization/windows-keyboard-layouts#U
+    //
+    // -- Jonathan Miles <jon@cybah.co.uk>
+
+    include "latin"
+
+    name[Group1]="English (UK, extended, Windows)";
+
+    key <TLDE> { [ dead_grave,    notsign,       brokenbar,    NoSymbol ] };
+    key <AE02> { [     2,        quotedbl,  dead_diaeresis,     onehalf ] };
+    key <AE03> { [     3,        sterling,   threesuperior,    onethird ] };
+    key <AE04> { [     4,          dollar,        EuroSign,  onequarter ] };
+    key <AE06> { [     6,     asciicircum, dead_circumflex,    NoSymbol ] };
+
+    key <AD02> { [   w,                 W,          wacute,      Wacute ] };
+    key <AD03> { [   e,                 E,          eacute,      Eacute ] };
+    key <AD06> { [   y,                 Y,          yacute,      Yacute ] };
+    key <AD07> { [   u,                 U,          uacute,      Uacute ] };
+    key <AD08> { [   i,                 I,          iacute,      Iacute ] };
+    key <AD09> { [   o,                 O,          oacute,      Oacute ] };
+    key <AD12> { [ bracketright, braceright,      NoSymbol,         bar ] };
+
+    key <AC01> { [   a,                 A,          aacute,      Aacute ] };
+    key <AC11> { [  apostrophe,         at,     dead_acute,      grave  ] };
+    key <BKSL>  { [ numbersign, asciitilde,     dead_tilde,    backslash ] };
+
+    key <LSGT>  { [  backslash,        bar,       NoSymbol,    NoSymbol ] };
+    key <AB03> { [   c,                 C,        ccedilla,    Ccedilla ] };
+
+    include "level3(ralt_switch)"
+};
+
+// Describe the differences between the US Colemak layout
+// and a UK variant. By Andy Buckley (andy@insectnation.org)
+
+partial alphanumeric_keys
+xkb_symbols "colemak" {
+    include "us(colemak)"
+
+    name[Group1]="English (UK, Colemak)";
+
+    key <TLDE>	{ [     grave,    notsign,          bar,   asciitilde ]	};
+    key <AE02>	{ [         2,   quotedbl,  twosuperior,    oneeighth ]	};
+    key <AE03>	{ [         3,   sterling, threesuperior,    sterling ]	};
+    key <AE04>	{ [         4,     dollar,     EuroSign,   onequarter ]	};
+
+    key <AC11>	{ [apostrophe,         at, dead_circumflex, dead_caron]	};
+    key <BKSL>	{ [numbersign, asciitilde,   dead_grave,   dead_breve ]	};
+
+    key <LSGT>	{ [ backslash,        bar,   asciitilde,    brokenbar ]	};
+};
+
+// Colemak-DH (ISO) layout, UK Variant, https://colemakmods.github.io/mod-dh/
+
+partial alphanumeric_keys
+xkb_symbols "colemak_dh" {
+    include "us(colemak_dh)"
+
+    name[Group1]="English (UK, Colemak-DH)";
+
+    key <TLDE>	{ [     grave,    notsign,          bar,   asciitilde ]	};
+    key <AE02>	{ [         2,   quotedbl,  twosuperior,    oneeighth ]	};
+    key <AE03>	{ [         3,   sterling, threesuperior,    sterling ]	};
+    key <AE04>	{ [         4,     dollar,     EuroSign,   onequarter ]	};
+
+    key <AC11>	{ [apostrophe,         at, dead_circumflex, dead_caron]	};
+    key <BKSL>	{ [numbersign, asciitilde,   dead_grave,   dead_breve ]	};
+
+    key <AB05>	{ [ backslash,        bar,   asciitilde,    brokenbar ]	};
+};
+
+
+// Dvorak (UK) keymap (by odaen) allowing the usage of
+// the £ and ? key and swapping the @ and " keys.
+
+partial alphanumeric_keys
+xkb_symbols "dvorak" {
+    include "us(dvorak-alt-intl)"
+
+    name[Group1]="English (UK, Dvorak)";
+
+    key <TLDE> { [	grave,	notsign,   bar,           bar		] };
+    key <AE02> { [	    2,	quotedbl,  twosuperior,   NoSymbol	] };
+    key <AE03> { [	    3,	sterling,  threesuperior, NoSymbol	] };
+    key <AD01> { [ apostrophe,	at		] };
+    key <BKSL> { [ numbersign,	asciitilde	] };
+    key <LSGT> { [  backslash,	bar		] };
+};
+
+// Dvorak letter positions, but punctuation all in the normal UK positions.
+
+partial alphanumeric_keys
+xkb_symbols "dvorakukp" {
+    include "gb(dvorak)"
+
+    name[Group1]="English (UK, Dvorak, with UK punctuation)";
+
+    key <AE11> {        [     minus,    underscore      ]       };
+    key <AE12> {        [     equal,    plus            ]       };
+    key <AD11> {        [ bracketleft,  braceleft       ]       };
+    key <AD12> {        [ bracketright, braceright      ]       };
+    key <AD01> {        [ slash,        question        ]       };
+    key <AC11> {        [apostrophe,    at,         dead_circumflex, dead_caron] };
+};
+
+partial alphanumeric_keys
+xkb_symbols "mac" {
+
+    include "latin"
+
+    name[Group1]= "English (UK, Macintosh)";
+
+    key <TLDE> {	[         section,       plusminus ]	};
+    key <AE02> {	[               2,              at,         EuroSign	]	};
+    key <AE03> {	[               3,        sterling,       numbersign	]	};
+    key <LSGT> {	[           grave,      asciitilde ]	};
+
+    include "level3(ralt_switch)"
+    include "level3(enter_switch)"
+};
+
+
+partial alphanumeric_keys
+xkb_symbols "mac_intl" {
+
+    include "latin"
+
+    name[Group1]="English (UK, Macintosh, intl.)";
+
+    key <TLDE>  { [ section,            plusminus,            notsign,    notsign ] }; //dead_grave
+    key <AE02>  { [   2,                       at,           EuroSign,    onehalf ] };
+    key <AE03>  { [   3,                 sterling,        twosuperior,   onethird ] };
+    key <AE04>  { [   4,                   dollar,      threesuperior, onequarter ] };
+    key <AE06>  { [   6,          dead_circumflex,           NoSymbol,   onesixth ] };
+    key <AD09>    { [   o,                        O,                 oe,       OE ] };
+
+    key <AC11>  { [ dead_acute,    dead_diaeresis,     dead_diaeresis,        bar ] }; //dead_doubleacute
+    key <BKSL>  { [ backslash,                bar,         numbersign,        bar ] };
+
+    key <LSGT>  { [ dead_grave,        dead_tilde,          brokenbar,        bar ] };
+
+    include "level3(ralt_switch)"
+};
+
+partial alphanumeric_keys
+xkb_symbols "pl" {
+
+    // Polish accented letters on upper levels of corresponding base letters.
+    // Idea from Wawrzyniec Niewodniczański, adapted by Aleksander Kowalski.
+
+    include "gb(basic)"
+
+    name[Group1]="Polish (British keyboard)";
+
+    key <AD03>  { [         e,          E,      eogonek,      Eogonek ] };
+    key <AD09>  { [         o,          O,       oacute,       Oacute ] };
+
+    key <AC01>  { [         a,          A,      aogonek,      Aogonek ] };
+    key <AC02>  { [         s,          S,       sacute,       Sacute ] };
+
+    key <AB01>  { [         z,          Z,    zabovedot,    Zabovedot ] };
+    key <AB02>  { [         x,          X,       zacute,       Zacute ] };
+    key <AB03>  { [         c,          C,       cacute,       Cacute ] };
+    key <AB06>  { [         n,          N,       nacute,       Nacute ] };
+};
+
+partial alphanumeric_keys
+xkb_symbols "gla" {
+
+    // Grave-accented letters on the upper levels of the relevant vowels.
+
+    include "gb(basic)"
+
+    name[Group1]="Scottish Gaelic";
+
+    key <AD03> { [   e,                 E,          egrave,      Egrave ] };
+    key <AD07> { [   u,                 U,          ugrave,      Ugrave ] };
+    key <AD08> { [   i,                 I,          igrave,      Igrave ] };
+    key <AD09> { [   o,                 O,          ograve,      Ograve ] };
+
+    key <AC01> { [   a,                 A,          agrave,      Agrave ] };
+};
+
+// EXTRAS:
+
+partial alphanumeric_keys
+	xkb_symbols "sun_type6" {
+	include "sun_vndr/gb(sun_type6)"
+};

library/xkeyboard-config/vendor/symbols/kpdl

@@ -0,0 +1,102 @@
+// The <KPDL> key is a mess.
+// It was probably originally meant to be a decimal separator.
+// Except since it was declared by USA people it didn't use the original
+// SI separator "," but a "." (since then the USA managed to f-up the SI
+// by making "." an accepted alternative, but standards still use "," as
+// default)
+// As a result users of SI-abiding countries expect either a "." or a ","
+// or a "decimal_separator" which may or may not be translated in one of the
+// above depending on applications.
+// It's not possible to define a default per-country since user expectations
+// depend on the conflicting choices of their most-used applications,
+// operating system, etc. Therefore it needs to be a configuration setting
+// Copyright © 2007 Nicolas Mailhot <nicolas.mailhot @ laposte.net>
+
+
+// Legacy <KPDL> #1
+// This assumes KP_Decimal will be translated in a dot
+partial keypad_keys
+xkb_symbols "dot" {
+
+    key.type[Group1]="KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, KP_Decimal ] }; // <delete> <separator>
+};
+
+
+// Legacy <KPDL> #2
+// This assumes KP_Separator will be translated in a comma
+partial keypad_keys
+xkb_symbols "comma" {
+
+    key.type[Group1]="KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, KP_Separator ] }; // <delete> <separator>
+};
+
+
+// Period <KPDL>, usual keyboard serigraphy in most countries
+partial keypad_keys
+xkb_symbols "dotoss" {
+
+    key.type[Group1]="FOUR_LEVEL_MIXED_KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, period, comma, 0x100202F ] }; // <delete> . , ⍽ (narrow no-break space)
+};
+
+
+// Period <KPDL>, usual keyboard serigraphy in most countries, latin-9 restriction
+partial keypad_keys
+xkb_symbols "dotoss_latin9" {
+
+    key.type[Group1]="FOUR_LEVEL_MIXED_KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, period, comma, nobreakspace ] }; // <delete> . , ⍽ (no-break space)
+};
+
+
+// Comma <KPDL>, what most non anglo-saxon people consider the real separator
+partial keypad_keys
+xkb_symbols "commaoss" {
+
+    key.type[Group1]="FOUR_LEVEL_MIXED_KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, comma, period, 0x100202F ] }; // <delete> , . ⍽ (narrow no-break space)
+};
+
+
+// Momayyez <KPDL>: Bahrain, Iran, Iraq, Kuwait, Oman, Qatar, Saudi Arabia, Syria, UAE
+partial keypad_keys
+xkb_symbols "momayyezoss" {
+
+    key.type[Group1]="FOUR_LEVEL_MIXED_KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, 0x100066B, comma, 0x100202F ] }; // <delete> ? , ⍽ (narrow no-break space)
+};
+
+
+// Abstracted <KPDL>, pray everything will work out (it usually does not)
+partial keypad_keys
+xkb_symbols "kposs" {
+
+    key.type[Group1]="FOUR_LEVEL_MIXED_KEYPAD" ;
+
+    key <KPDL> { [ KP_Delete, KP_Decimal, KP_Separator, 0x100202F ] }; // <delete> ? ? ⍽ (narrow no-break space)
+};
+
+// Spreadsheets may be configured to use the dot as decimal
+// punctuation, comma as a thousands separator and then semi-colon as
+// the list separator. Of these, dot and semi-colon is most important
+// when entering data by the keyboard; the comma can then be inferred
+// and added to the presentation afterwards. Using semi-colon as a
+// general separator may in fact be preferred to avoid ambiguities
+// in data files. Most times a decimal separator is hard-coded, it
+// seems to be period, probably since this is the syntax used in
+// (most) programming languages.
+partial keypad_keys
+xkb_symbols "semi" {
+
+    key.type[Group1]="FOUR_LEVEL_MIXED_KEYPAD" ;
+
+    key <KPDL> { [ NoSymbol, NoSymbol, semicolon ] };
+};

library/xkeyboard-config/vendor/symbols/latin

@@ -0,0 +1,255 @@
+// Common Latin alphabet layout
+
+default partial
+xkb_symbols "basic" {
+
+    key <AE01>	{ [         1,     exclam,  onesuperior,   exclamdown ]	};
+    key <AE02>	{ [         2,         at,  twosuperior,    oneeighth ]	};
+    key <AE03>	{ [         3, numbersign, threesuperior,    sterling ]	};
+    key <AE04>	{ [         4,     dollar,   onequarter,       dollar ]	};
+    key <AE05>	{ [         5,    percent,      onehalf, threeeighths ]	};
+    key <AE06>	{ [         6, asciicircum, threequarters, fiveeighths ] };
+    key <AE07>	{ [         7,  ampersand,    braceleft, seveneighths ]	};
+    key <AE08>	{ [         8,   asterisk,  bracketleft,    trademark ]	};
+    key <AE09>	{ [         9,  parenleft, bracketright,    plusminus ]	};
+    key <AE10>	{ [         0, parenright,   braceright,       degree ]	};
+    key <AE11>	{ [     minus, underscore,    backslash, questiondown ]	};
+    key <AE12>	{ [     equal,       plus, dead_cedilla,  dead_ogonek ]	};
+
+    key <AD01>	{ [         q,          Q,           at,  Greek_OMEGA ]	};
+    key <AD02>	{ [         w,          W,        U017F,      section ]	};
+    key <AD03>	{ [         e,          E,            e,            E ]	};
+    key <AD04>	{ [         r,          R,    paragraph,   registered ]	};
+    key <AD05>	{ [         t,          T,       tslash,       Tslash ]	};
+    key <AD06>	{ [         y,          Y,    leftarrow,          yen ]	};
+    key <AD07>	{ [         u,          U,    downarrow,      uparrow ]	};
+    key <AD08>	{ [         i,          I,   rightarrow,     idotless ]	};
+    key <AD09>	{ [         o,          O,       oslash,       Oslash ]	};
+    key <AD10>	{ [         p,          P,        thorn,        THORN ]	};
+    key <AD11>	{ [bracketleft,  braceleft, dead_diaeresis, dead_abovering ] };
+    key <AD12>	{ [bracketright, braceright, dead_tilde,  dead_macron ]	};
+
+    key <AC01>	{ [         a,          A,           ae,           AE ]	};
+    key <AC02>	{ [         s,          S,       ssharp,        U1E9E ]	};
+    key <AC03>	{ [         d,          D,          eth,          ETH ]	};
+    key <AC04>	{ [         f,          F,      dstroke,  ordfeminine ]	};
+    key <AC05>	{ [         g,          G,          eng,          ENG ]	};
+    key <AC06>	{ [         h,          H,      hstroke,      Hstroke ]	};
+    key <AC07>	{ [         j,          J,    dead_hook,    dead_horn ] };
+    key <AC08>	{ [         k,          K,          kra,    ampersand ]	};
+    key <AC09>	{ [         l,          L,      lstroke,      Lstroke ]	};
+    key <AC10>	{ [ semicolon,    colon, dead_acute, dead_doubleacute ]	};
+    key <AC11>	{ [apostrophe, quotedbl, dead_circumflex,  dead_caron ]	};
+    key <TLDE>	{ [     grave, asciitilde,      notsign,      notsign ]	};
+
+    key <BKSL>	{ [ backslash,        bar,   dead_grave,   dead_breve ]	};
+    key <AB01>	{ [         z,          Z, guillemotleft,        less ]	};
+    key <AB02>	{ [         x,          X, guillemotright,    greater ]	};
+    key <AB03>	{ [         c,          C,         cent,    copyright ]	};
+    key <AB04>	{ [         v,          V,   doublelowquotemark, singlelowquotemark ]	};
+    key <AB05>	{ [         b,          B,  leftdoublequotemark, leftsinglequotemark ] };
+    key <AB06>	{ [         n,          N, rightdoublequotemark, rightsinglequotemark ]	};
+    key <AB07>	{ [         m,          M,           mu,    masculine ]	};
+    key <AB08>	{ [     comma,       less,        U2022,     multiply ]	}; // bullet
+    key <AB09>	{ [    period,    greater, periodcentered,   division ]	};
+    key <AB10>	{ [     slash,   question, dead_belowdot, dead_abovedot ] };
+};
+
+// Northern Europe ( Danish, Finnish, Norwegian, Swedish) common layout
+
+partial
+xkb_symbols "type2" {
+
+    include "latin"
+
+    key <AE01>	{ [         1,     exclam,   exclamdown,  onesuperior ]	};
+    key <AE02>	{ [         2,   quotedbl,           at,  twosuperior ]	};
+    key <AE03>	{ [         3, numbersign,     sterling, threesuperior]	};
+    key <AE04>	{ [         4,   currency,       dollar,   onequarter ]	};
+    key <AE05>	{ [         5,    percent,      onehalf,         cent ]	};
+    key <AE06>	{ [         6,  ampersand,          yen,  fiveeighths ]	};
+    key <AE07>	{ [         7,      slash,    braceleft,     division ]	};
+    key <AE08>	{ [         8,  parenleft,  bracketleft, guillemotleft]	};
+    key <AE09>	{ [         9, parenright, bracketright, guillemotright] };
+    key <AE10>	{ [         0,      equal,   braceright,       degree ]	};
+
+    key <AD03>	{ [         e,          E,     EuroSign,         cent ]	};
+    key <AD04>	{ [         r,          R,   registered,   registered ]	};
+    key <AD05>	{ [         t,          T,        thorn,        THORN ]	};
+    key <AD09>	{ [         o,          O,           oe,           OE ]	};
+    key <AD11>	{ [     aring,  Aring, dead_diaeresis, dead_abovering ]	};
+    key <AD12>	{ [dead_diaeresis, dead_circumflex, dead_tilde, dead_caron ] };
+
+    key <AC01>	{ [         a,          A,  ordfeminine,    masculine ]	};
+
+    key <AB03>	{ [         c,          C,    copyright,    copyright ]	};
+    key <AB08>	{ [     comma,  semicolon, dead_cedilla,  dead_ogonek ]	};
+    key <AB09>	{ [    period,   colon, periodcentered, dead_abovedot ]	};
+    key <AB10>	{ [     minus, underscore, dead_belowdot, dead_abovedot ] };
+};
+
+// Slavic Latin ( Albanian, Croatian, Polish, Slovene, Yugoslav)
+// common layout
+
+partial
+xkb_symbols "type3" {
+
+    include "latin"
+
+    key <AD01>	{ [         q,          Q,    backslash,  Greek_OMEGA ]	};
+    key <AD02>	{ [         w,          W,          bar,      section ]	};
+    key <AD06>	{ [         z,          Z,    leftarrow,          yen ]	};
+
+    key <AC04>	{ [         f,          F,  bracketleft,  ordfeminine ]	};
+    key <AC05>	{ [         g,          G, bracketright,          ENG ]	};
+    key <AC08>	{ [         k,          K,      lstroke,    ampersand ]	};
+
+    key <AB01>	{ [         y,          Y, guillemotleft,        less ]	};
+    key <AB04>	{ [         v,          V,           at,        grave ]	};
+    key <AB05>	{ [         b,          B,    braceleft,   apostrophe ]	};
+    key <AB06>	{ [         n,          N,   braceright,        acute ]	};
+    key <AB07>	{ [         m,          M,      section,    masculine ]	};
+    key <AB08>	{ [     comma,  semicolon,         less,     multiply ]	};
+    key <AB09>	{ [    period,      colon,      greater,     division ]	};
+};
+
+// Another common Latin layout
+// (German, Estonian, Spanish, Icelandic, Italian, Latin American, Portuguese)
+
+partial
+xkb_symbols "type4" {
+
+    include "latin"
+
+    key <AE02>	{ [         2,   quotedbl,           at,    oneeighth ]	};
+    key <AE06>	{ [         6,  ampersand,      notsign,  fiveeighths ]	};
+    key <AE07>	{ [         7,      slash,    braceleft, seveneighths ]	};
+    key <AE08>	{ [         8,  parenleft,  bracketleft,    trademark ]	};
+    key <AE09>	{ [         9, parenright, bracketright,    plusminus ]	};
+    key <AE10>	{ [         0,      equal,   braceright,       degree ]	};
+
+    key <AD03>	{ [         e,          E,     EuroSign,         cent ]	};
+
+    key <AB08>	{ [   comma,  semicolon,          U2022,     multiply ]	}; // bullet
+    key <AB09>	{ [  period,      colon, periodcentered,     division ]	};
+    key <AB10>	{ [   minus, underscore, dead_belowdot, dead_abovedot ]	};
+};
+
+partial
+xkb_symbols "nodeadkeys" {
+
+    key <AE12>	{ [     equal,       plus,     cedilla,        ogonek ]	};
+    key <AD11>	{ [bracketleft,  braceleft,  diaeresis,        degree ]	};
+    key <AD12>	{ [bracketright, braceright, asciitilde,       macron ]	};
+    key <AC07>	{ [         j,          J,         ezh,           EZH ] };
+    key <AC10>	{ [ semicolon,      colon,       acute,   doubleacute ]	};
+    key <AC11>	{ [apostrophe,   quotedbl, asciicircum,         caron ]	};
+    key <BKSL>	{ [ backslash,        bar,       grave,         breve ]	};
+    key <AB10>	{ [     slash,   question,    ellipsis,      abovedot ]	};
+};
+
+partial
+xkb_symbols "type2_nodeadkeys" {
+
+    include "latin(nodeadkeys)"
+
+    key <AD11>	{ [     aring,      Aring,   diaeresis,        degree ]	};
+    key <AD12>	{ [ diaeresis, asciicircum, asciitilde,         caron ]	};
+    key <AB08>	{ [     comma,  semicolon,     cedilla,        ogonek ]	};
+    key <AB09>	{ [    period,   colon, periodcentered,      abovedot ]	};
+    key <AB10>	{ [   minus, underscore,      ellipsis,      abovedot ]	};
+};
+
+partial
+xkb_symbols "type3_nodeadkeys" {
+
+    include "latin(nodeadkeys)"
+};
+
+partial
+xkb_symbols "type4_nodeadkeys" {
+
+    include "latin(nodeadkeys)"
+
+    key <AB10>	{ [   minus, underscore,      ellipsis,      abovedot ]	};
+};
+
+// Added 2008.03.05 by Marcin Woliński
+// See http://marcinwolinski.pl/keyboard/ for a description.
+// Used by pl(intl)
+//
+// ┌─────┐
+// │ 2 4 │   2 = Shift,  4 = Level3 + Shift
+// │ 1 3 │   1 = Normal, 3 = Level3
+// └─────┘
+// ┌─────┬─────┬─────┬─────┬─────┬─────┬─────┬─────┬─────┬─────┬─────┬─────┬─────┲━━━━━━━━━┓
+// │ ~ ~ │ ! ' │ @ " │ # ˝ │ $ ¸ │ % ˇ │ ^ ^ │ & ˘ │ * ̇  │ ( ̣  │ ) ° │ _ ¯ │ + ˛ ┃ ⌫ Back- ┃
+// │ ` ` │ 1 ¡ │ 2 © │ 3 • │ 4 § │ 5 € │ 6 ¢ │ 7 − │ 8 × │ 9 ÷ │ 0 ° │ - – │ = — ┃  space  ┃
+// ┢━━━━━┷━┱───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┴─┬───┺━┳━━━━━━━┫
+// ┃       ┃ Q   │ W   │ E   │ R   │ T   │ Y   │ U   │ I   │ O   │ P   │ { « │ } » ┃ Enter ┃
+// ┃Tab ↹  ┃ q   │ w   │ e   │ r   │ t   │ y   │ u   │ i   │ o   │ p   │ [ ‹ │ ] › ┃   ⏎   ┃
+// ┣━━━━━━━┻┱────┴┬────┴┬────┴┬────┴┬────┴┬────┴┬────┴┬────┴┬────┴┬────┴┬────┴┬────┺┓      ┃
+// ┃        ┃ A   │ S   │ D   │ F   │ G   │ H   │ J   │ K   │ L   │ : “ │ " ” │ | ¶ ┃      ┃
+// ┃Caps ⇬  ┃ a   │ s   │ d   │ f   │ g   │ h   │ j   │ k   │ l   │ ; ‘ │ ' ’ │ \   ┃      ┃
+// ┣━━━━━━━━┹────┬┴────┬┴────┬┴────┬┴────┬┴────┬┴────┬┴────┬┴────┬┴────┬┴────┲┷━━━━━┻━━━━━━┫
+// ┃             │ Z   │ X   │ C   │ V   │ B   │ N   │ M   │ < „ │ > · │ ? ¿ ┃             ┃
+// ┃Shift ⇧      │ z   │ x   │ c   │ v   │ b   │ n   │ m   │ , ‚ │ . … │ / ⁄ ┃Shift ⇧      ┃
+// ┣━━━━━━━┳━━━━━┷━┳━━━┷━━━┱─┴─────┴─────┴─────┴─────┴─────┴───┲━┷━━━━━╈━━━━━┻━┳━━━━━━━┳━━━┛
+// ┃       ┃       ┃       ┃ ␣                               ⍽ ┃       ┃       ┃       ┃
+// ┃Ctrl   ┃Meta   ┃Alt    ┃ ␣           Space               ⍽ ┃AltGr ⇮┃Menu   ┃Ctrl   ┃
+// ┗━━━━━━━┻━━━━━━━┻━━━━━━━┹───────────────────────────────────┺━━━━━━━┻━━━━━━━┻━━━━━━━┛
+
+partial
+xkb_symbols "intl" {
+
+    key <TLDE>	{ [     grave,  asciitilde, dead_grave,       dead_tilde ]	};
+    key <AE01>	{ [         1,      exclam, exclamdown,       dead_acute ]	};
+    key <AE02>	{ [         2,          at,  copyright,   dead_diaeresis ]	};
+    key <AE03>	{ [         3,  numbersign,      U2022, dead_doubleacute ]	}; // U+2022 is bullet (the name bullet does not work)
+    key <AE04>	{ [         4,      dollar,    section,     dead_cedilla ]	};
+    key <AE05>	{ [         5,     percent,   EuroSign,       dead_caron ]	};
+    key <AE06>	{ [         6, asciicircum,       cent,  dead_circumflex ]	};
+    key <AE07>	{ [         7,   ampersand,      U2212,       dead_breve ]	}; // U+2212 is MINUS SIGN
+    key <AE08>	{ [         8,    asterisk,   multiply,    dead_abovedot ]	};
+    key <AE09>	{ [         9,   parenleft,   division,    dead_belowdot ]	};
+    key <AE10>	{ [         0,  parenright,     degree,   dead_abovering ]	};
+    key <AE11>	{ [     minus,  underscore,     endash,      dead_macron ]	};
+    key <AE12>	{ [     equal,        plus,     emdash,      dead_ogonek ]	};
+
+    key <AD01>	{ [         q,          Q ]	};
+    key <AD02>	{ [         w,          W ]	};
+    key <AD03>	{ [         e,          E ]	};
+    key <AD04>	{ [         r,          R ]	};
+    key <AD05>	{ [         t,          T ]	};
+    key <AD06>	{ [         y,          Y ]	};
+    key <AD07>	{ [         u,          U ]	};
+    key <AD08>	{ [         i,          I ]	};
+    key <AD09>	{ [         o,          O ]	};
+    key <AD10>	{ [         p,          P ]	};
+    key <AD11>	{ [bracketleft,   braceleft,  U2039, guillemotleft ]    };
+    key <AD12>	{ [bracketright, braceright, U203A, guillemotright ]	};
+
+    key <AC01>	{ [         a,          A ]	};
+    key <AC02>	{ [         s,          S ]	};
+    key <AC03>	{ [         d,          D ]	};
+    key <AC04>	{ [         f,          F ]	};
+    key <AC05>	{ [         g,          G ]	};
+    key <AC06>	{ [         h,          H ]	};
+    key <AC07>	{ [         j,          J ]	};
+    key <AC08>	{ [         k,          K ]	};
+    key <AC09>	{ [         l,          L ]	};
+    key <AC10>	{ [ semicolon,      colon,  leftsinglequotemark,  leftdoublequotemark  ]  };
+    key <AC11>	{ [apostrophe,   quotedbl, rightsinglequotemark, rightdoublequotemark  ]  };
+
+    key <BKSL>	{ [ backslash,        bar, 	NoSymbol, paragraph ] };
+    key <AB01>	{ [         z,          Z ]	};
+    key <AB02>	{ [         x,          X ]	};
+    key <AB03>	{ [         c,          C ]	};
+    key <AB04>	{ [         v,          V ]	};
+    key <AB05>	{ [         b,          B ]     };
+    key <AB06>	{ [         n,          N ]	};
+    key <AB07>	{ [         m,          M ]	};
+    key <AB08>	{ [     comma,       less, singlelowquotemark, doublelowquotemark ]	};
+    key <AB09>	{ [    period,    greater, ellipsis, periodcentered ]	};
+    key <AB10>	{ [     slash,   question, U2044,  questiondown ]     };  // U+2044 is FRACTION SLASH
+};

library/xkeyboard-config/vendor/symbols/level3

@@ -0,0 +1,156 @@
+// These variants assign ISO_Level3_Shift to various keys
+// so that levels 3 and 4 can be reached.
+
+// The default behaviour:
+// the right Alt key (AltGr) chooses the third symbol engraved on a key.
+default partial modifier_keys
+xkb_symbols "ralt_switch" {
+  key <RALT> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The right Alt key never chooses the third level.
+// This option attempts to undo the effect of a layout's inclusion of
+// 'ralt_switch'.  You may want to also select another level3 option
+// to map the level3 shift to some other key.
+partial modifier_keys
+xkb_symbols "ralt_alt" {
+  key <RALT> {[  Alt_R,  Meta_R  ], type[group1]="TWO_LEVEL" };
+  modifier_map Mod1 { <RALT> };
+};
+
+// The right Alt key (while pressed) chooses the third shift level,
+// and Compose is mapped to its second level.
+partial modifier_keys
+xkb_symbols "ralt_switch_multikey" {
+  key <RALT> {[  ISO_Level3_Shift,  Multi_key  ], type[group1]="TWO_LEVEL" };
+};
+
+// Either Alt key (while pressed) chooses the third shift level.
+// (To be used mostly to imitate Mac OS functionality.)
+partial modifier_keys
+xkb_symbols "alt_switch" {
+  include "level3(lalt_switch)"
+  include "level3(ralt_switch)"
+};
+
+// The left Alt key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "lalt_switch" {
+  key <LALT> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The right Ctrl key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "switch" {
+  key <RCTL> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The Menu key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "menu_switch" {
+  key <MENU> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// Either Win key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "win_switch" {
+  include "level3(lwin_switch)"
+  include "level3(rwin_switch)"
+};
+
+// The left Win key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "lwin_switch" {
+  key <LWIN> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The right Win key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "rwin_switch" {
+  key <RWIN> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The Enter key on the kepypad (while pressed) chooses the third shift level.
+// (This is especially useful for Mac laptops which miss the right Alt key.)
+partial modifier_keys
+xkb_symbols "enter_switch" {
+  key <KPEN> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The CapsLock key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "caps_switch" {
+  key <CAPS> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The CapsLock key (while pressed) chooses the third shift level and
+// Ctrl + CapsLock has the original CapsLock function.
+// The 2023 DIN standard for German keyboards recommends it as an option:
+// - https://de.wikipedia.org/wiki/E1_(Tastaturbelegung)#Feststelltaste/Umschaltsperre
+// - https://en.wikipedia.org/wiki/Caps_Lock#Abolition
+partial modifier_keys
+xkb_symbols "caps_switch_capslock_with_ctrl" {
+    virtual_modifiers  LevelThree;
+
+    key <CAPS> {
+        type[Group1] = "PC_CONTROL_LEVEL2",
+        symbols[Group1] = [ ISO_Level3_Shift, Caps_Lock ],
+        // Explicit actions are preferred over modMap None/Mod5 { Caps_Lock }
+        // because they have no side effect
+        actions[Group1] = [ SetMods(modifiers = LevelThree), LockMods(modifiers = Lock) ]
+    };
+};
+
+// The Backslash key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "bksl_switch" {
+  key <BKSL> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The AC11 key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "ac11_switch" {
+  key <AC11> {[  ISO_Level3_Shift  ], type[Group1]="ONE_LEVEL" };
+};
+
+// The Less/Greater key (while pressed) chooses the third shift level.
+partial modifier_keys
+xkb_symbols "lsgt_switch" {
+  key <LSGT> {[  ISO_Level3_Shift  ], type[group1]="ONE_LEVEL" };
+};
+
+// The CapsLock key (while pressed) chooses the third shift level,
+// and latches when pressed together with another third-level chooser.
+partial modifier_keys
+xkb_symbols "caps_switch_latch" {
+  key <CAPS> {[  ISO_Level3_Shift,  ISO_Level3_Shift,  ISO_Level3_Latch  ],
+                 type[group1]="THREE_LEVEL" };
+};
+
+// The Backslash key (while pressed) chooses the third shift level,
+// and latches when pressed together with another third-level chooser.
+partial modifier_keys
+xkb_symbols "bksl_switch_latch" {
+  key <BKSL> {[  ISO_Level3_Shift,  ISO_Level3_Shift,  ISO_Level3_Latch  ],
+                 type[group1]="THREE_LEVEL" };
+};
+
+// The Less/Greater key (while pressed) chooses the third shift level,
+// and latches when pressed together with another third-level chooser.
+partial modifier_keys
+xkb_symbols "lsgt_switch_latch" {
+  key <LSGT> {[  ISO_Level3_Shift,  ISO_Level3_Shift,  ISO_Level3_Latch  ],
+                 type[group1]="THREE_LEVEL" };
+};
+
+// Top-row digit key 4 chooses third shift level when pressed alone.
+partial modifier_keys
+xkb_symbols "4_switch_isolated" {
+  override key <AE04> {[  ISO_Level3_Shift  ]};
+};
+
+// Top-row digit key 9 chooses third shift level when pressed alone.
+partial modifier_keys
+xkb_symbols "9_switch_isolated" {
+  override key <AE09> {[  ISO_Level3_Shift  ]};
+};

library/xkeyboard-config/xkeyboard-config.zig

@@ -0,0 +1,153 @@
+//! xkeyboard-config — keyboard layouts, compiled from the X11 xkeyboard-config database
+//! into native Zig. It turns a physical key (a USB HID usage, as the input module delivers)
+//! plus a modifier state into a **keysym** and, when the key produces one, a **character**
+//! (a Unicode scalar). This is the piece that lets a `KeyEvent.keycode` become a
+//! `KeyEvent.character`, without shipping an X11 runtime.
+//!
+//! The layout tables in `generated/layouts.zig` are produced by
+//! `tools/make-xkeyboard-config.py` (see ./README.md to regenerate). Those tables are
+//! deliberately pure data — each key carries its up-to-four levels and an XKB *type*. The
+//! type -> level selection semantics (which modifier picks which level) live here, so the
+//! data and the policy are separable.
+//!
+//! Scope (documented in README.md): group 1 only, no dead-key/compose composition (a dead
+//! key returns its keysym with no character), and a curated set of key types. Layouts:
+//! us, gb, de, fr, es, dvorak.
+//!
+//! Upstream xkeyboard-config and keysymdef.h are MIT/X11 licensed; see vendor/COPYING and
+//! vendor/PROVENANCE.md.
+
+const std = @import("std");
+const generated = @import("layouts");
+
+pub const Level = generated.Level;
+pub const KeyType = generated.KeyType;
+pub const Key = generated.Key;
+pub const Layout = generated.Layout;
+
+/// The generated layouts, by name — as pointers, so they share identity with `all` and
+/// `byName` (and match the `*const Layout` that `map` takes).
+pub const us: *const Layout = &generated.us;
+pub const gb: *const Layout = &generated.gb;
+pub const de: *const Layout = &generated.de;
+pub const fr: *const Layout = &generated.fr;
+pub const es: *const Layout = &generated.es;
+pub const dvorak: *const Layout = &generated.dvorak;
+
+/// Every generated layout, for enumeration (e.g. a settings UI).
+pub const all = generated.all;
+
+/// The modifier state that selects a key's level. `level3` is AltGr (ISO Level3 Shift);
+/// `control` is accepted for completeness but does not affect level selection here.
+pub const Modifiers = struct {
+    shift: bool = false,
+    caps_lock: bool = false,
+    level3: bool = false,
+    control: bool = false,
+};
+
+/// The result of a lookup: the X11 `keysym`, and the `character` it produces (a Unicode
+/// scalar) when it is a printable key — null for keys that produce none (Return, F1, a
+/// bare dead key, an unmapped key).
+pub const Mapping = struct {
+    keysym: u32,
+    character: ?u21,
+};
+
+/// Which level (0..3) a key of `kind` selects under `mods`. XKB's canonical semantics:
+/// Shift picks the odd level, AltGr (level3) adds 2, and Caps acts like Shift for the
+/// alphabetic types. See the XKB "key types" — this covers the ones the vendored layouts
+/// use; anything else falls back to shift-or-not.
+fn selectLevel(kind: KeyType, mods: Modifiers) usize {
+    const shift_or_caps = mods.shift != mods.caps_lock; // XOR: Caps behaves like Shift
+    const low: usize = if (mods.shift) 1 else 0;
+    const high: usize = if (mods.level3) 2 else 0;
+    return switch (kind) {
+        .one_level => 0,
+        .two_level, .keypad, .other => low,
+        .alphabetic => if (shift_or_caps) 1 else 0,
+        .four_level => low + high,
+        .four_level_alphabetic => (if (shift_or_caps) @as(usize, 1) else 0) + high,
+        // Caps affects only the base pair, not the AltGr pair.
+        .four_level_semialphabetic => if (mods.level3) 2 + low else (if (shift_or_caps) @as(usize, 1) else 0),
+    };
+}
+
+/// Map a physical key (`hid_usage`, a USB HID keyboard-page usage) under `mods` on
+/// `layout` to its keysym and character. Falls back gracefully when the selected level is
+/// undefined for the key: it drops the AltGr component, then the shift component, so a key
+/// with only a base/shift pair still yields something sensible under AltGr.
+pub fn map(layout: *const Layout, hid_usage: u8, mods: Modifiers) Mapping {
+    const key = &layout.keys[hid_usage];
+    var level = selectLevel(key.kind, mods);
+    // Fall back to a defined level: full -> without AltGr -> base.
+    if (key.levels[level].keysym == 0 and key.levels[level].unicode == 0) {
+        const candidates = [_]usize{ level & 1, 0 };
+        for (candidates) |candidate| {
+            if (key.levels[candidate].keysym != 0 or key.levels[candidate].unicode != 0) {
+                level = candidate;
+                break;
+            }
+        }
+    }
+    const chosen = key.levels[level];
+    return .{
+        .keysym = chosen.keysym,
+        .character = if (chosen.unicode != 0) @intCast(chosen.unicode) else null,
+    };
+}
+
+/// Look up a layout by its name (`"us"`, `"gb"`, ...), or null if unknown.
+pub fn byName(name: []const u8) ?*const Layout {
+    for (all) |layout| {
+        if (std.mem.eql(u8, layout.name, name)) return layout;
+    }
+    return null;
+}
+
+// --- tests (host-run via `zig build test`) ---------------------------------
+
+const testing = std.testing;
+
+// USB HID usages used in the tests (keyboard page 0x07).
+const hid_a: u8 = 0x04;
+const hid_1: u8 = 0x1e;
+const hid_3: u8 = 0x20;
+
+test "us: letters obey shift and caps" {
+    try testing.expectEqual(@as(?u21, 'a'), map(us, hid_a, .{}).character);
+    try testing.expectEqual(@as(?u21, 'A'), map(us, hid_a, .{ .shift = true }).character);
+    try testing.expectEqual(@as(?u21, 'A'), map(us, hid_a, .{ .caps_lock = true }).character);
+    // Shift + Caps cancels for an alphabetic key.
+    try testing.expectEqual(@as(?u21, 'a'), map(us, hid_a, .{ .shift = true, .caps_lock = true }).character);
+}
+
+test "us: digits and their shifted symbols" {
+    try testing.expectEqual(@as(?u21, '1'), map(us, hid_1, .{}).character);
+    try testing.expectEqual(@as(?u21, '!'), map(us, hid_1, .{ .shift = true }).character);
+    try testing.expectEqual(@as(?u21, '3'), map(us, hid_3, .{}).character);
+    try testing.expectEqual(@as(?u21, '#'), map(us, hid_3, .{ .shift = true }).character);
+    // A digit is not alphabetic: Caps alone must not shift it.
+    try testing.expectEqual(@as(?u21, '3'), map(us, hid_3, .{ .caps_lock = true }).character);
+}
+
+test "layouts differ: GB pound vs US hash on shift+3" {
+    try testing.expectEqual(@as(?u21, '#'), map(us, hid_3, .{ .shift = true }).character);
+    try testing.expectEqual(@as(?u21, '£'), map(gb, hid_3, .{ .shift = true }).character);
+}
+
+test "french azerty places q where us has a" {
+    try testing.expectEqual(@as(?u21, 'q'), map(fr, hid_a, .{}).character);
+    try testing.expectEqual(@as(?u21, 'Q'), map(fr, hid_a, .{ .shift = true }).character);
+}
+
+test "byName resolves and rejects" {
+    try testing.expect(byName("us") == us);
+    try testing.expect(byName("gb") == gb);
+    try testing.expect(byName("nonsense") == null);
+}
+
+test "unmapped key yields no character" {
+    // HID 0x00 is not a key; every level is empty.
+    try testing.expectEqual(@as(?u21, null), map(us, 0x00, .{}).character);
+}

system/abi.zig

@@ -43,16 +43,43 @@ pub const SystemCall = enum(u64) {
     irq_bind = 14, // irq_bind(id, resource_index, endpoint): deliver a device IRQ as an IPC notification
     irq_ack = 15, // irq_ack(id, resource_index): re-arm a bound IRQ after servicing it
     device_register = 16, // device_register(parent_id, descriptor) -> id: publish a child of a device you claimed
-    system_spawn = 17, // system_spawn(name_ptr, name_len) -> 0: start a named initial-ramdisk binary as a new ring-3 process
+    system_spawn = 17, // system_spawn(name_ptr, name_len, arguments_ptr, arguments_len, exit_endpoint) -> child process id: start a named initial-ramdisk binary as a new ring-3 process
     dma_alloc = 18, // dma_alloc(len, flags) -> vaddr (rax), paddr (rdx): contiguous, pinned, uncacheable DMA memory
     dma_free = 19, // dma_free(vaddr, len) -> 0: release a prior dma_alloc
     msi_bind = 20, // msi_bind(device_id, endpoint) -> address (rax), data (rdx): a per-device MSI vector for a claimed device
     io_read = 21, // io_read(device_id, resource_index, offset, width) -> value: read a port in a claimed device's io_port resource
     io_write = 22, // io_write(device_id, resource_index, offset, width, value) -> 0: write a port in a claimed device's io_port resource
     clock = 23, // clock() -> nanoseconds since boot: a monotonic time source (for timeouts/delays)
+    process_enumerate = 24, // process_enumerate(buffer, maximum) -> total: snapshot the task table
+    process_kill = 25, // process_kill(id) -> 0/-errno: end a process this process spawned
+    ipc_send = 26, // ipc_send(handle, message_ptr, message_len) -> 0/-errno: post a payload to an endpoint's async queue without blocking
+    process_exit_reason = 27, // process_exit_reason(id) -> ExitReason/-errno: how a dead child ended (its supervisor only)
+    process_subscribe = 28, // process_subscribe(endpoint) -> 0/-errno: subscribe to published exit events — every death posts a notification
+    signal_bind = 29, // signal_bind(endpoint) -> 0/-errno: nominate the endpoint this process's signals arrive on
+    process_signal = 30, // process_signal(id, signal) -> 0/-errno: post a signal to a child (or to yourself)
+    timer_bind = 31, // timer_bind(endpoint, ms) -> 0/-errno: one-shot timer — posts a notification when ms elapse
+    klog_read = 32, // klog_read(offset, ptr, len) -> bytes copied: copy the kernel RAM log buffer out to a user buffer (for persisting the boot log to disk)
+    wall_clock = 33, // wall_clock() -> Unix epoch seconds (UTC): the RTC wall-clock time, for filesystem timestamps (mtime). Monotonic time is `clock`.
     _,
 };
 
+/// How a process ended — recorded by the kernel at death, queried by the
+/// supervisor with `process_exit_reason`, and the input to its restart decision
+/// (docs/process-lifecycle.md): a clean exit meant to stop, a fault wants a
+/// restart with backoff, killed means the supervisor did it itself. The faults
+/// mirror the CPU exceptions a ring-3 process can die of; they are exit reasons,
+/// never delivered to the faulting process (recovery is restart, not a handler).
+pub const ExitReason = enum(u8) {
+    exited = 0, // returned from main / called exit
+    aborted = 1, // deliberate self-termination (reserved: no abort path yet)
+    segmentation_fault = 2, // page fault
+    illegal_instruction = 3, // invalid opcode
+    arithmetic_fault = 4, // divide error, x87 or SIMD fault
+    protection_fault = 5, // general protection fault
+    fault = 6, // any other CPU exception
+    killed = 7, // process_kill
+};
+
 /// The x86 MSI message address base (`0xFEE0_0000`): a device raises an MSI by writing
 /// `data` to this address, which the Local APIC turns into an interrupt at the vector
 /// in `data`. The kernel returns the concrete (address, data) from `msi_bind`; this is
@@ -67,17 +94,95 @@ pub const dma_write_combining: u64 = 2; // write-combining (framebuffers); needs
 pub const dma_below_4g: u64 = 4; // physical address must fit 32 bits (legacy DMA engines)
 
 /// Set in the badge returned by `ipc_reply_wait` when what arrived is an
-/// **asynchronous notification** (today: a device interrupt bound with `irq_bind`)
-/// rather than a message from a client. There is no payload and no reply owed; the
-/// low bits carry the source, a GSI. Shared so the kernel's ISR and the driver's
-/// event loop can't disagree about which bit means "the hardware spoke".
+/// **asynchronous notification** (a device interrupt bound with `irq_bind`, or a
+/// child-exit notice — see `notify_exit_bit`) rather than a message from a client.
+/// There is no payload and no reply owed; the low bits carry the source. Shared so
+/// the kernel's ISR and the driver's event loop can't disagree about which bit
+/// means "the hardware spoke".
 pub const notify_badge_bit: u64 = 1 << 63;
 
+/// Set (alongside `notify_badge_bit`) in the badge of a **child-exit notification**:
+/// posted to the endpoint a supervisor passed to `system_spawn` when that child ends
+/// — by clean exit, by a fault, or by `process_kill`. The low bits carry the child's
+/// process id, so one endpoint can supervise many children (and even share with IRQ
+/// notifications, which never set this bit). The microkernel's SIGCHLD.
+pub const notify_exit_bit: u64 = 1 << 62;
+
+/// Set (alongside `notify_badge_bit`) in the badge of a **buffered message** — a payload
+/// posted to an endpoint's async queue by `ipc_send`, delivered through `ipc_reply_wait`
+/// like a notification (no reply owed) but carrying bytes in the receive buffer, not just
+/// a badge. This is what distinguishes a payload-bearing async message from a bare IRQ /
+/// child-exit notification (which sets neither this nor `notify_exit_bit`). The low bits
+/// carry the sender's task id. The async counterpart of the synchronous `ipc_call`, for
+/// broadcasts where a rendezvous is the wrong shape (the input service is the first user).
+pub const notify_message_bit: u64 = 1 << 61;
+
+/// Set (alongside `notify_badge_bit`) in the badge of a **signal notification** —
+/// the process-lifecycle vocabulary of docs/process-lifecycle.md, delivered to the
+/// endpoint the process nominated with `signal_bind`. The low bits carry the
+/// coalesced pending mask (bit positions = `Signal` values): signals are
+/// statements, not questions, and two pending terminates are one terminate.
+pub const notify_signal_bit: u64 = 1 << 60;
+
+/// Set (alongside `notify_badge_bit`) in the badge of a **timer notification** —
+/// a one-shot `timer_bind` deadline landing. No payload bits: what to do when the
+/// deadline fires is whatever the receiver armed it for (a stop-sequence
+/// escalation, a restart backoff, an alarm).
+pub const notify_timer_bit: u64 = 1 << 59;
+
+/// The signal vocabulary (docs/process-lifecycle.md): POSIX's concepts, danos's
+/// names, message delivery. The value is the bit position in the pending mask — a
+/// private kernel/runtime detail, free to change while they ship together. Kill
+/// is not here (it is `process_kill`, unhandleable by definition); faults are not
+/// here (they are `ExitReason`s — recovery is restart, not a handler); liveness is
+/// not here (a question, asked as the zero-length ping call, not a statement).
+pub const Signal = enum(u5) {
+    terminate = 0, // finish up and exit (the polite half of the stop sequence)
+    reload = 1, // re-read configuration / re-scan
+    interrupt = 2, // interactive interrupt (no sender until a console exists)
+    quit = 3, // as interrupt, by convention more final
+    alarm = 4, // a timer the process armed for itself (unbuilt: no consumer yet)
+    user_1 = 5, // service-defined
+    user_2 = 6, // service-defined
+};
+
+/// Capacity of `ProcessDescriptor.name` — matches the longest name `system_spawn`
+/// accepts, so a process's recorded name (its argv[0]) is never truncated.
+pub const maximum_process_name = 64;
+
+/// What a process is doing right now, as reported by `process_enumerate`. Crosses
+/// the system_call boundary as `ProcessDescriptor.state`.
+pub const ProcessState = enum(u32) {
+    ready = 0, // runnable, waiting for a core
+    running = 1, // executing on a core right now
+    blocked = 2, // waiting (sleeping, or blocked in IPC)
+};
+
+/// One `process_enumerate` entry — the kernel's view of a live task, kernel tasks
+/// included (they carry an empty name and id 0 is the boot task). Fixed layout
+/// (extern) because it crosses the kernel↔user boundary by memory copy, like
+/// `DeviceDescriptor` in the device ABI.
+pub const ProcessDescriptor = extern struct {
+    id: u32, // kernel-assigned process id; never reused (monotonic)
+    supervisor: u32, // id of the process that spawned it (0 = the kernel)
+    state: u32, // a ProcessState value
+    priority: u32,
+    name_length: u32,
+    name: [maximum_process_name]u8, // argv[0] at spawn; empty for kernel tasks
+};
+
 /// Well-known IPC service ids for the bootstrap name registry (create_ipc_endpoint +
 /// ipc_register/ipc_lookup). Small integers, so no string interning is needed
 /// during bring-up. The VFS server registers under `vfs`; clients look it up.
 pub const ServiceId = enum(u32) {
     vfs = 1,
+    input = 2,
+    ps2_bus = 3, // the 8042 owner; child device drivers attach here for raw bytes
+    device_manager = 4, // the tree, the matcher, the supervisor (docs/device-manager.md)
+    power = 5, // system power: events (button, lid, battery) + shutdown (docs/power.md; domain-named per docs/discovery.md — the acpi service registers it on x86, a PSCI service will on ARM)
+    usb_bus = 6, // the xHCI host-controller driver's transfer endpoint; USB class drivers look it up and `callCap`-open their device to get a private per-device transfer channel (docs/driver-model.md)
+    block = 7, // a block-device driver (USB mass storage today): read/write of fixed-size blocks, the storage a filesystem sits on
+    fat = 8, // the FAT filesystem server; the VFS mounts it and forwards paths under its mount point (/mnt/usb) to it
     _,
 };
 

system/devices/acpi-ids.zig

@@ -1,60 +1,120 @@
 //! ACPI / PnP hardware-ID (`_HID`) names: the flat analog of pci-class.zig for
 //! `acpi_device` nodes. Unlike PCI, ACPI has no class/subclass/prog-IF taxonomy — a
 //! device's identity *is* its `_HID` string (`PNP0303` simply means "PS/2 keyboard"),
-//! so this is a plain id -> description registry rather than a hierarchical decoder.
+//! so this is a plain id <-> name registry rather than a hierarchical decoder.
 //! The well-known PnP/ACPI IDs; vendor-specific ids (e.g. `QEMU0002`, `INTC1234`) have
-//! no standard name and return "". Pure reference data, so it is shared by kernel
-//! discovery (the device-tree dump) and any user-space tool.
+//! no standard name and decode to nothing. Pure reference data, so it is shared by
+//! kernel discovery (the device-tree dump) and any user-space driver or tool.
+//!
+//! Code that means a specific device names the `HardwareId` variant instead of its
+//! `_HID` string — `HardwareId.ps2_keyboard.hid()` reads without a registry lookup,
+//! where a bare `"PNP0303"` does not.
 
 const std = @import("std");
 
-const Entry = struct { hid: []const u8, name: []const u8 };
+/// The common standard PnP/ACPI hardware IDs, as named values. Prefix ranges hint at
+/// the grouping (PNP03xx keyboards, PNP0Fxx pointing devices, PNP0Cxx ACPI
+/// power/thermal, PNP0Axx buses), but there is no formal hierarchy — hence a flat
+/// enum over a flat registry.
+pub const HardwareId = enum {
+    programmable_interrupt_controller,
+    system_timer,
+    high_precision_event_timer,
+    dma_controller,
+    ps2_keyboard,
+    parallel_port,
+    ecp_parallel_port,
+    serial_port,
+    floppy_disk_controller,
+    system_speaker,
+    pci_bus,
+    generic_container,
+    /// The second id the ACPI spec assigns the same "Generic Container Device" name.
+    generic_container_extended,
+    pci_express_root_bridge,
+    real_time_clock,
+    system_board,
+    motherboard_reserved_resources,
+    math_coprocessor,
+    acpi_system_board,
+    embedded_controller,
+    control_method_battery,
+    fan,
+    power_button,
+    lid,
+    sleep_button,
+    pci_interrupt_link,
+    microsoft_ps2_mouse,
+    ps2_mouse,
+    ac_adapter,
+    processor_device,
+    processor_aggregator,
+    processor_container,
 
-/// The common standard PnP/ACPI hardware IDs. Prefix ranges hint at the grouping
-/// (PNP03xx keyboards, PNP0Fxx pointing devices, PNP0Cxx ACPI power/thermal,
-/// PNP0Axx buses), but there is no formal hierarchy — hence a flat table.
-const table = [_]Entry{
-    .{ .hid = "PNP0000", .name = "Programmable Interrupt Controller (PIC)" },
-    .{ .hid = "PNP0100", .name = "System Timer (PIT)" },
-    .{ .hid = "PNP0103", .name = "High Precision Event Timer (HPET)" },
-    .{ .hid = "PNP0200", .name = "DMA Controller" },
-    .{ .hid = "PNP0303", .name = "PS/2 Keyboard" },
-    .{ .hid = "PNP0400", .name = "Standard LPT Parallel Port" },
-    .{ .hid = "PNP0401", .name = "ECP Parallel Port" },
-    .{ .hid = "PNP0501", .name = "16550A-compatible Serial Port" },
-    .{ .hid = "PNP0700", .name = "PC Floppy Disk Controller" },
-    .{ .hid = "PNP0800", .name = "System Speaker" },
-    .{ .hid = "PNP0A03", .name = "PCI Bus" },
-    .{ .hid = "PNP0A05", .name = "Generic Container Device" },
-    .{ .hid = "PNP0A06", .name = "Generic Container Device" },
-    .{ .hid = "PNP0A08", .name = "PCI Express Root Bridge" },
-    .{ .hid = "PNP0B00", .name = "Real-Time Clock (RTC)" },
-    .{ .hid = "PNP0C01", .name = "System Board" },
-    .{ .hid = "PNP0C02", .name = "Motherboard Reserved Resources" },
-    .{ .hid = "PNP0C04", .name = "Math Coprocessor" },
-    .{ .hid = "PNP0C08", .name = "ACPI System Board" },
-    .{ .hid = "PNP0C09", .name = "ACPI Embedded Controller" },
-    .{ .hid = "PNP0C0A", .name = "ACPI Control Method Battery" },
-    .{ .hid = "PNP0C0B", .name = "ACPI Fan" },
-    .{ .hid = "PNP0C0C", .name = "ACPI Power Button" },
-    .{ .hid = "PNP0C0D", .name = "ACPI Lid" },
-    .{ .hid = "PNP0C0E", .name = "ACPI Sleep Button" },
-    .{ .hid = "PNP0C0F", .name = "PCI Interrupt Link Device" },
-    .{ .hid = "PNP0F03", .name = "Microsoft PS/2 Mouse" },
-    .{ .hid = "PNP0F13", .name = "PS/2 Mouse" },
-    .{ .hid = "ACPI0003", .name = "AC Adapter" },
-    .{ .hid = "ACPI0007", .name = "Processor Device" },
-    .{ .hid = "ACPI000C", .name = "Processor Aggregator" },
-    .{ .hid = "ACPI0010", .name = "Processor Container" },
+    const Entry = struct { hid: []const u8, name: []const u8 };
+
+    /// The registry row for this id: its `_HID` string and human-readable name.
+    fn entry(self: HardwareId) Entry {
+        return switch (self) {
+            .programmable_interrupt_controller => .{ .hid = "PNP0000", .name = "Programmable Interrupt Controller (PIC)" },
+            .system_timer => .{ .hid = "PNP0100", .name = "System Timer (PIT)" },
+            .high_precision_event_timer => .{ .hid = "PNP0103", .name = "High Precision Event Timer (HPET)" },
+            .dma_controller => .{ .hid = "PNP0200", .name = "DMA Controller" },
+            .ps2_keyboard => .{ .hid = "PNP0303", .name = "PS/2 Keyboard" },
+            .parallel_port => .{ .hid = "PNP0400", .name = "Standard LPT Parallel Port" },
+            .ecp_parallel_port => .{ .hid = "PNP0401", .name = "ECP Parallel Port" },
+            .serial_port => .{ .hid = "PNP0501", .name = "16550A-compatible Serial Port" },
+            .floppy_disk_controller => .{ .hid = "PNP0700", .name = "PC Floppy Disk Controller" },
+            .system_speaker => .{ .hid = "PNP0800", .name = "System Speaker" },
+            .pci_bus => .{ .hid = "PNP0A03", .name = "PCI Bus" },
+            .generic_container => .{ .hid = "PNP0A05", .name = "Generic Container Device" },
+            .generic_container_extended => .{ .hid = "PNP0A06", .name = "Generic Container Device" },
+            .pci_express_root_bridge => .{ .hid = "PNP0A08", .name = "PCI Express Root Bridge" },
+            .real_time_clock => .{ .hid = "PNP0B00", .name = "Real-Time Clock (RTC)" },
+            .system_board => .{ .hid = "PNP0C01", .name = "System Board" },
+            .motherboard_reserved_resources => .{ .hid = "PNP0C02", .name = "Motherboard Reserved Resources" },
+            .math_coprocessor => .{ .hid = "PNP0C04", .name = "Math Coprocessor" },
+            .acpi_system_board => .{ .hid = "PNP0C08", .name = "ACPI System Board" },
+            .embedded_controller => .{ .hid = "PNP0C09", .name = "ACPI Embedded Controller" },
+            .control_method_battery => .{ .hid = "PNP0C0A", .name = "ACPI Control Method Battery" },
+            .fan => .{ .hid = "PNP0C0B", .name = "ACPI Fan" },
+            .power_button => .{ .hid = "PNP0C0C", .name = "ACPI Power Button" },
+            .lid => .{ .hid = "PNP0C0D", .name = "ACPI Lid" },
+            .sleep_button => .{ .hid = "PNP0C0E", .name = "ACPI Sleep Button" },
+            .pci_interrupt_link => .{ .hid = "PNP0C0F", .name = "PCI Interrupt Link Device" },
+            .microsoft_ps2_mouse => .{ .hid = "PNP0F03", .name = "Microsoft PS/2 Mouse" },
+            .ps2_mouse => .{ .hid = "PNP0F13", .name = "PS/2 Mouse" },
+            .ac_adapter => .{ .hid = "ACPI0003", .name = "AC Adapter" },
+            .processor_device => .{ .hid = "ACPI0007", .name = "Processor Device" },
+            .processor_aggregator => .{ .hid = "ACPI000C", .name = "Processor Aggregator" },
+            .processor_container => .{ .hid = "ACPI0010", .name = "Processor Container" },
+        };
+    }
+
+    /// This id's `_HID` string (e.g. `.ps2_keyboard` -> "PNP0303").
+    pub fn hid(self: HardwareId) []const u8 {
+        return self.entry().hid;
+    }
+
+    /// This id's human-readable name (e.g. `.ps2_keyboard` -> "PS/2 Keyboard").
+    pub fn description(self: HardwareId) []const u8 {
+        return self.entry().name;
+    }
+
+    /// The named value for a `_HID` string, or null if it is not a known standard
+    /// id (vendor-specific ids are not in the registry).
+    pub fn fromHid(hid_string: []const u8) ?HardwareId {
+        for (std.enums.values(HardwareId)) |id| {
+            if (std.mem.eql(u8, id.hid(), hid_string)) return id;
+        }
+        return null;
+    }
 };
 
-/// The human-readable name for a `_HID`, or "" if it is not a known standard id
-/// (vendor-specific ids have no registry name — callers just print the raw HID).
+/// The human-readable name for a `_HID` string, or "" if it is not a known standard
+/// id (vendor-specific ids have no registry name — callers just print the raw HID).
 pub fn description(hid: []const u8) []const u8 {
-    for (table) |entry| {
-        if (std.mem.eql(u8, entry.hid, hid)) return entry.name;
-    }
-    return "";
+    return (HardwareId.fromHid(hid) orelse return "").description();
 }
 
 test "decodes standard PnP/ACPI ids and leaves the rest alone" {
@@ -66,3 +126,13 @@ test "decodes standard PnP/ACPI ids and leaves the rest alone" {
     try eq("", description("QEMU0002")); // vendor-specific: no standard name
     try eq("", description("")); // no HID at all
 }
+
+test "named values round-trip through their _HID strings" {
+    const testing = std.testing;
+    try testing.expectEqualStrings("PNP0303", HardwareId.ps2_keyboard.hid());
+    try testing.expectEqual(@as(?HardwareId, .ps2_mouse), HardwareId.fromHid("PNP0F13"));
+    try testing.expectEqual(@as(?HardwareId, null), HardwareId.fromHid("QEMU0002"));
+    for (std.enums.values(HardwareId)) |id| {
+        try testing.expectEqual(@as(?HardwareId, id), HardwareId.fromHid(id.hid()));
+    }
+}

system/devices/acpi.zig

@@ -40,6 +40,9 @@ pub const RegisterAccess = struct {
 /// Everything the power subsystem needs, extracted from the FADT and the AML
 /// sleep packages during discovery. Populated by `discover`, read by `power`.
 pub const PowerInformation = struct {
+    /// The System Control Interrupt's GSI (FADT SCI_INT) — the line ACPI events
+    /// (power button, GPEs) arrive on. Published to the acpi service for M21.
+    sci_interrupt: u16 = 0,
     /// The SMM command port and the value that switches the platform into ACPI mode.
     smi_cmd: u16 = 0,
     acpi_enable: u8 = 0,
@@ -51,7 +54,8 @@ pub const PowerInformation = struct {
     reset: RegisterAccess = .{},
     reset_value: u8 = 0,
     reset_supported: bool = false,
-    /// SLP_TYP values for S5 (soft off) and S3 (suspend), from the AML sleep-state (`_Sx`) packages.
+    /// SLP_TYP values for S5 (soft off) and S3 (suspend), from the AML sleep-state (`_Sx`)
+    /// packages.
     s5: ?aml.SleepType = null,
     s3: ?aml.SleepType = null,
 };
@@ -153,6 +157,13 @@ pub var namespace: ?aml.Namespace = null;
 /// Physical address of the DSDT the FADT points at, or 0.
 pub var dsdt_physical: u64 = 0;
 
+/// The FADT itself (physical + length), published on the acpi-tables node so
+/// the ring-3 acpi service can read the PM1 event and GPE blocks it needs for
+/// the event side (docs/acpi.md — ACPI events). Distinguished from the AML
+/// blob resources by its intact "FACP" header — the blobs are header-stripped.
+var fadt_physical: u64 = 0;
+var fadt_length: u64 = 0;
+
 // AML blocks (DSDT + any SSDTs) collected during the table walk, as physical
 // address + length of each table's post-header bytecode. Scanned after the walk
 // for the sleep-state (`_Sx`) packages.
@@ -197,7 +208,8 @@ const ExtendedSystemDescriptorPointer = extern struct {
     root_system_description_table_address: u32 align(1),
     /// The size of the RSDP.
     length: u32 align(1),
-    /// A 64-bit physical address pointing to the XSDT. If the revision is at least 2, the XSDT should be used regardless of architecture, as the RSDT was deprecated.
+    /// A 64-bit physical address pointing to the XSDT. If the revision is at least 2, the XSDT
+    /// should be used regardless of architecture, as the RSDT was deprecated.
     extended_system_descriptor_table_address: u64 align(1),
     /// A checksum used for the entire table.
     extended_checksum: u8,
@@ -357,35 +369,19 @@ const Hpet = extern struct {
     page_protection: u8,
 };
 
-// --- PCI configuration-space header (first 64 bytes, common fields) ---------
-
-const PciHeader = extern struct {
-    vendor_id: u16 align(1),
-    device_id: u16 align(1),
-    command: u16 align(1),
-    status: u16 align(1),
-    revision_id: u8,
-    prog_if: u8,
-    subclass: u8,
-    class_code: u8,
-    cache_line_size: u8,
-    latency_timer: u8,
-    /// bit 7 set => multi-function device.
-    header_type: u8,
-    bist: u8,
-    // 0x10 onward (BARs, etc.) depends on header_type; read separately.
-};
-
 // --- Entry point ------------------------------------------------------------
 
 /// Discover hardware from the ACPI tables rooted at `rsdp_physical` and populate
 /// `device_tree`. `hal` provides MMIO mapping (for PCIe ECAM) and port I/O. Also parses the
 /// FADT and the AML sleep-state (`_Sx`) packages into `power_information` for the power service.
-pub fn discover(rsdp_physical: u64, device_tree: *DeviceTree, hal: Hal) !void {
+pub fn discover(rsdp_physical: u64, memory_regions: []const boot_handoff.MemoryRegion, device_tree: *DeviceTree, hal: Hal) !void {
     if (rsdp_physical == 0) return error.NoRsdp;
+    boot_memory_regions = memory_regions;
 
     // Start clean so a re-run doesn't accumulate stale state.
     power_information = .{};
+    fadt_physical = 0;
+    fadt_length = 0;
     platform_information = .{};
     aml_stats = .{};
     namespace = null;
@@ -417,12 +413,58 @@ pub fn discover(rsdp_physical: u64, device_tree: *DeviceTree, hal: Hal) !void {
         aml_stats = .{ .nodes = namespace.?.nodeCount(), .consumed = pr.consumed, .total = pr.total };
         power_information.s5 = aml.sleepState(&namespace.?, 5);
         power_information.s3 = aml.sleepState(&namespace.?, 3);
-        // Fold the namespace's Device objects into the generic tree.
-        wireAcpiDevices(device_tree, &namespace.?, hal) catch {};
+        // The namespace's Device objects are no longer folded into the kernel
+        // tree (M20.3): the ring-3 acpi service claims the acpi-tables node
+        // (published below), re-parses the same blobs, and registers + reports
+        // the _HID devices itself. The kernel keeps the namespace only for the
+        // \_S5 sleep type above.
     } else |_| {
         // AML parse failed (e.g. out of memory); power stays best-effort with
         // whatever the FADT alone provided.
     }
+
+    // Publish the acpi-tables node (docs/discovery.md): the AML blobs as
+    // memory resources for the acpi service to map and parse in ring 3, a broad
+    // io_port grant for the OperationRegion access its interpreter needs, and
+    // the SCI for the events track (M21). Exactly one node, one trusted
+    // claimant. Kept even when the kernel-side device building (above) retires
+    // in M20.3 — the kernel still owns the *static* tables and \_S5.
+    publishAcpiTablesNode(device_tree) catch {};
+}
+
+/// Build the acpi-tables node (see the call site in discover). Best-effort: a
+/// failure here leaves the kernel-seeded tree working, only the ring-3 service
+/// finds nothing to claim.
+fn publishAcpiTablesNode(device_tree: *DeviceTree) !void {
+    const node = try device_tree.addChild(device_tree.root, .acpi_tables, "acpi-tables");
+    // One memory resource per AML block — page-aligned base down, length padded
+    // up to cover the bytecode, so mmio_map hands the service a pointer into it.
+    var i: usize = 0;
+    while (i < aml_block_count and i < device_model.maximum_resources - 2) : (i += 1) {
+        // mmio_map preserves the sub-page offset, so the service maps this and
+        // gets a pointer straight to the bytecode.
+        _ = node.addResource(.memory, aml_block_physical[i], aml_block_len[i]);
+    }
+    // The broad I/O grant: OperationRegions name whatever ports the firmware
+    // chose (EC, PM1, GPE, SMBus); which ports cannot be known before the AML
+    // that names them is parsed, so the grant is the whole space — the honest
+    // trust boundary of docs/discovery.md (the acpi service's one trusted node).
+    _ = node.addResource(.io_port, 0, 1 << 16);
+    // A broad interrupt window: ACPI _CRS names legacy ISA IRQs (the PS/2 lines
+    // 1 and 12, the RTC, …), and the service registers those devices under this
+    // node, so it must own a superset. The range [0, 256) covers every GSI; the
+    // SCI (recorded first, len 1) stays distinct so M21 can pick it out.
+    if (power_information.sci_interrupt != 0) _ = node.addResource(.irq, power_information.sci_interrupt, 1);
+    _ = node.addResource(.irq, 0, 256);
+    // The FADT rides along (M21): the service reads the PM1 event / GPE blocks
+    // from its own copy, telling it apart from the AML blobs by signature.
+    if (fadt_physical != 0) _ = node.addResource(.memory, fadt_physical, fadt_length);
+}
+
+/// The number of Device objects in the namespace built during discovery, or 0.
+pub fn amlDeviceCount() usize {
+    if (namespace) |*ns| return aml.deviceCount(ns);
+    return 0;
 }
 
 /// Walk the RSDT (Entry = u32) or XSDT (Entry = u64): validate it, then dispatch
@@ -448,10 +490,12 @@ fn handleTable(device_tree: *DeviceTree, hal: Hal, sdt_physical: u64) !void {
     if (std.mem.eql(u8, &sig, &APIC)) {
         try parseMadt(device_tree, header);
     } else if (std.mem.eql(u8, &sig, &MCFG)) {
-        try parseMcfg(device_tree, hal, header);
+        try parseMcfg(device_tree, header);
     } else if (std.mem.eql(u8, &sig, &HPET)) {
         try parseHpet(device_tree, hal, header);
     } else if (std.mem.eql(u8, &sig, &FACP)) {
+        fadt_physical = sdt_physical;
+        fadt_length = header.length;
         parseFadt(header);
     } else if (std.mem.eql(u8, &sig, &SPCR)) {
         parseSpcr(header);
@@ -533,7 +577,7 @@ fn parseMadt(device_tree: *DeviceTree, header: *const SystemDescriptorTableHeade
 }
 
 /// MCFG -> a pci_host_bridge per ECAM segment, then a PCI enumeration underneath.
-fn parseMcfg(device_tree: *DeviceTree, hal: Hal, header: *const SystemDescriptorTableHeader) !void {
+fn parseMcfg(device_tree: *DeviceTree, header: *const SystemDescriptorTableHeader) !void {
     const total: usize = header.length;
     const base: [*]const u8 = @ptrCast(header);
 
@@ -548,109 +592,85 @@ fn parseMcfg(device_tree: *DeviceTree, hal: Hal, header: *const SystemDescriptor
         // ECAM window: 1 MiB of configuration space per bus.
         _ = bridge.addResource(.memory, alloc.base_address, bus_count << 20);
         _ = bridge.addResource(.bus_range, alloc.start_bus, bus_count);
+        addBridgeApertures(bridge);
+        // The bridge decodes the whole 16-bit I/O space toward its bus — the
+        // window functions' I/O BARs must register-contain within (M19.2).
+        _ = bridge.addResource(.io_port, 0, 1 << 16);
 
-        try enumeratePci(device_tree, bridge, hal, alloc.*);
+        // The function walk itself retired to ring 3 (M19.3): the pci-bus
+        // driver claims this bridge, repeats the scan through its ECAM grant,
+        // and device_registers what it finds — the kernel seeds only the
+        // bridge. The scan's equivalence was proven before the hand-off
+        // (pci-scan), and the walk's history is in git if archaeology calls.
     }
 }
 
-/// Brute-force scan the ECAM window's bus range for present PCI functions. No
-/// bridge recursion yet: on the ECAM path the host bridge decodes every bus in
-/// the window, so scanning the declared range finds everything QEMU exposes.
-fn enumeratePci(
-    device_tree: *DeviceTree,
-    bridge: *device_model.Device,
-    hal: Hal,
-    alloc: McfgAllocation,
-) !void {
-    var bus: u16 = alloc.start_bus;
-    while (bus <= alloc.end_bus) : (bus += 1) {
-        var device: u8 = 0;
-        while (device < 32) : (device += 1) {
-            const h0: *align(1) const PciHeader = @ptrCast(pciConfigurationPtr(alloc, hal, @intCast(bus), device, 0));
-            if (h0.vendor_id == 0xFFFF) continue; // no function 0 => slot empty
+/// The boot memory map, stored at discover() entry for the aperture derivation
+/// below (and, in M20, for the acpi-tables node's containment windows).
+var boot_memory_regions: []const boot_handoff.MemoryRegion = &.{};
 
-            const funcs: u8 = if (h0.header_type & 0x80 != 0) 8 else 1;
-            var function: u8 = 0;
-            while (function < funcs) : (function += 1) {
-                const configuration = pciConfigurationPtr(alloc, hal, @intCast(bus), device, function);
-                const h: *align(1) const PciHeader = @ptrCast(configuration);
-                if (h.vendor_id == 0xFFFF) continue;
-
-                var nb: [24]u8 = undefined;
-                const nm = std.fmt.bufPrint(&nb, "{s}:{x:0>2}:{x:0>2}.{d}", .{
-                    bridge.name(), bus, device, function,
-                }) catch "pcidev";
-                const node = try device_tree.addChild(bridge, .pci_device, nm);
-                // Resource 0 is the function's own 4 KiB ECAM configuration space. A
-                // claimed PCI driver mmio_maps this to reach its command register,
-                // BARs, and — the point — its capability list (MSI/MSI-X, PCIe
-                // extended caps), without any new syscall. Physical address per the
-                // ECAM formula (same as pciConfigurationPtr).
-                const config_physical = alloc.base_address +
-                    (@as(u64, @as(u8, @intCast(bus)) - alloc.start_bus) << 20) +
-                    (@as(u64, device) << 15) + (@as(u64, function) << 12);
-                _ = node.addResource(.memory, config_physical, abi.page_size);
-                node.ids.pci_vendor = h.vendor_id;
-                node.ids.pci_device = h.device_id;
-                node.ids.pci_class = (@as(u24, h.class_code) << 16) |
-                    (@as(u24, h.subclass) << 8) | h.prog_if;
-                node.ids.pci_bdf = (@as(u16, @intCast(bus)) << 8) | (@as(u16, device) << 3) | function;
-
-                // BARs only exist in header type 0 (normal devices), not bridges.
-                if (h.header_type & 0x7F == 0) addBars(node, configuration);
+/// The bridge's MMIO apertures, derived from the boot memory map's holes
+/// (docs/discovery.md — apertures from the memory map): registered PCI functions carry BAR
+/// resources, and `device_register` containment demands the bridge own windows
+/// that cover them. Everything the firmware described is "not hole"; the low
+/// aperture runs from the end of the described space below 4 GiB up to the
+/// I/O-APIC region, the high one from 4 GiB (or the end of RAM above it) to
+/// the 46-bit line. Coarse, mechanical, and AML-free — available at boot no
+/// matter what later moved to user space.
+fn addBridgeApertures(bridge: *device_model.Device) void {
+    // Below 4 GiB the described regions are sparse (RAM low, firmware flash
+    // and tables high), so the holes are the *gaps between* them — a single
+    // "after the last region" rule dies on OVMF's flash at the very top.
+    // Sort-merge the described ranges, then keep the three largest gaps
+    // (resource slots are bounded at 8 per device; ECAM + bus range + 3 + the
+    // high aperture fits). Above 4 GiB one aperture runs from the end of the
+    // described space to the 46-bit line.
+    const Range = struct { base: u64, end: u64 };
+    var below: [64]Range = undefined;
+    var below_count: usize = 0;
+    var high_end: u64 = 1 << 32;
+    for (boot_memory_regions) |region| {
+        const end = region.base + region.pages * 4096;
+        // Above 4 GiB only *usable RAM* blocks the aperture: OVMF describes
+        // its own 64-bit PCI window as a reserved region and then programs
+        // BARs inside it — honoring reserved there would exclude the very
+        // space BARs live in. Below 4 GiB every described region blocks (the
+        // kernel image, the tables, the ramdisk all live there). Bring-up
+        // trust: only the bridge's claimant can register into the aperture.
+        if (region.kind == .usable and end > high_end) high_end = end;
+        if (region.base >= (1 << 32) or below_count == below.len) continue;
+        below[below_count] = .{ .base = region.base, .end = @min(end, 1 << 32) };
+        below_count += 1;
+    }
+    // Insertion sort by base (the map is small and this runs once at boot).
+    for (1..below_count) |i| {
+        const key = below[i];
+        var j = i;
+        while (j > 0 and below[j - 1].base > key.base) : (j -= 1) below[j] = below[j - 1];
+        below[j] = key;
+    }
+    // Walk the sorted ranges, collecting inter-region gaps of at least 1 MiB.
+    var gaps: [3]Range = .{Range{ .base = 0, .end = 0 }} ** 3;
+    var cursor: u64 = 0;
+    var index: usize = 0;
+    while (index <= below_count) : (index += 1) {
+        const gap_end = if (index == below_count) (1 << 32) else below[index].base;
+        if (gap_end > cursor and gap_end - cursor >= (1 << 20)) {
+            // Keep the three largest, replacing the smallest kept so far.
+            var smallest: usize = 0;
+            for (gaps, 0..) |gap, gi| {
+                if (gap.end - gap.base < gaps[smallest].end - gaps[smallest].base) smallest = gi;
+            }
+            if (gap_end - cursor > gaps[smallest].end - gaps[smallest].base) {
+                gaps[smallest] = .{ .base = cursor, .end = gap_end };
             }
         }
+        if (index < below_count and below[index].end > cursor) cursor = below[index].end;
     }
-}
-
-/// Record and size the memory/IO windows named by a device's Base Address
-/// Registers. Sizing is the standard probe: disable decode, write all-ones, read
-/// back the writable (address) bits, restore. `size = ~mask + 1`.
-fn addBars(node: *device_model.Device, configuration: [*]align(1) u8) void {
-    // Stop the device decoding its BARs while we transiently write all-ones.
-    const command = rd(u16, configuration, 0x04);
-    wr(u16, configuration, 0x04, command & ~@as(u16, 0b11));
-
-    var i: usize = 0;
-    while (i < 6) : (i += 1) {
-        const off = 0x10 + i * 4;
-        const orig = rd(u32, configuration, off);
-        if (orig == 0) continue;
-
-        if (orig & 1 != 0) {
-            // I/O-space BAR (16-bit address space on x86).
-            wr(u32, configuration, off, 0xFFFF_FFFF);
-            const readback = rd(u32, configuration, off);
-            wr(u32, configuration, off, orig);
-            const mask = readback & 0xFFFF_FFFC;
-            const size: u32 = if (mask == 0) 0 else (~mask +% 1) & 0xFFFF;
-            _ = node.addResource(.io_port, orig & 0xFFFF_FFFC, size);
-        } else if ((orig >> 1) & 0x3 == 2) {
-            // 64-bit memory BAR: this BAR pair spans two configuration slots.
-            const orig_hi = rd(u32, configuration, off + 4);
-            wr(u32, configuration, off, 0xFFFF_FFFF);
-            wr(u32, configuration, off + 4, 0xFFFF_FFFF);
-            const lo = rd(u32, configuration, off);
-            const hi = rd(u32, configuration, off + 4);
-            wr(u32, configuration, off, orig);
-            wr(u32, configuration, off + 4, orig_hi);
-            const readback = (@as(u64, hi) << 32) | (lo & 0xFFFF_FFF0);
-            const size: u64 = if (readback == 0) 0 else ~readback +% 1;
-            const address = (@as(u64, orig_hi) << 32) | (orig & 0xFFFF_FFF0);
-            _ = node.addResource(.memory, address, size);
-            i += 1; // consumed the high half
-        } else {
-            // 32-bit memory BAR.
-            wr(u32, configuration, off, 0xFFFF_FFFF);
-            const readback = rd(u32, configuration, off);
-            wr(u32, configuration, off, orig);
-            const mask = readback & 0xFFFF_FFF0;
-            const size: u32 = if (mask == 0) 0 else ~mask +% 1;
-            _ = node.addResource(.memory, orig & 0xFFFF_FFF0, size);
+    for (gaps) |gap| {
+        if (gap.end > gap.base) _ = bridge.addResource(.memory, gap.base, gap.end - gap.base);
     }
-    }
-
-    wr(u16, configuration, 0x04, command); // restore decode
+    _ = bridge.addResource(.memory, high_end, (@as(u64, 1) << 46) - high_end);
 }
 
 /// HPET -> a timer node with its register block as an MMIO resource, plus the GSI
@@ -709,6 +729,7 @@ const fadt_pm1a_cnt_blk = 64; // u32 (I/O port)
 const fadt_pm1b_cnt_blk = 68; // u32 (I/O port)
 const fadt_pm_tmr_blk = 76; // u32 (I/O port) — the PM timer counter
 const fadt_pm1_cnt_len = 89; // u8 (bytes)
+const fadt_sci_int = 46; // u16 (the SCI's GSI)
 const fadt_flags = 112; // u32
 const fadt_reset_register = 116; // GAS (12 bytes)
 const fadt_reset_value = 128; // u8
@@ -726,6 +747,7 @@ fn parseFadt(header: *const SystemDescriptorTableHeader) void {
     const len: usize = header.length;
     const pi = &power_information;
 
+    pi.sci_interrupt = @truncate(fadt(u16, base, len, fadt_sci_int) orelse 0);
     pi.smi_cmd = @truncate(fadt(u32, base, len, fadt_smi_cmd) orelse 0);
     pi.acpi_enable = fadt(u8, base, len, fadt_acpi_enable) orelse 0;
     pi.acpi_disable = fadt(u8, base, len, fadt_acpi_disable) orelse 0;
@@ -805,337 +827,6 @@ fn parseDmar(hal: Hal, header: *const SystemDescriptorTableHeader) void {
     }
 }
 
-// --- AML namespace -> generic device tree -----------------------------------
-
-/// The PCI bus context while descending the ACPI namespace: the generic host
-/// bridge whose children ACPI address (`_ADR`) devices resolve against, and the bus number.
-const PciContext = struct { bridge: *device_model.Device, bus: u8 };
-
-/// Mirror the ACPI namespace's Device objects into the generic tree, *merging*
-/// them with the PCI-enumerated nodes: a PCI root bridge (`PNP0A03`/`PNP0A08`)
-/// folds onto the existing `pci_host_bridge`, and each addressed (`_ADR`) device folds onto
-/// the matching PCI function (annotating it with the ACPI hardware ID (`_HID`) and nesting the
-/// ACPI-only children — keyboard, RTC, … — beneath it). Namespace devices with no
-/// PCI match land under a synthetic `acpi` node.
-fn wireAcpiDevices(device_tree: *DeviceTree, aml_namespace: *aml.Namespace, hal: Hal) !void {
-    var arena = std.heap.ArenaAllocator.init(device_tree.allocator);
-    defer arena.deinit();
-    var interpreter = aml.Interpreter.init(aml_namespace, .{
-        .mapMmio = hal.mapMmio,
-        .pioRead = hal.pioRead,
-        .pioWrite = hal.pioWrite,
-    }, arena.allocator());
-
-    const acpi_root = try device_tree.addChild(device_tree.root, .unknown, "acpi");
-    try mirrorDevices(device_tree, aml_namespace.root, acpi_root, null, &interpreter);
-}
-
-fn mirrorDevices(device_tree: *DeviceTree, node: *aml.Node, parent_device: *device_model.Device, context: ?PciContext, interpreter: *aml.Interpreter) (error{OutOfMemory})!void {
-    var child = node.first_child;
-    while (child) |c| : (child = c.next_sibling) {
-        if (c.kind != .device) {
-            // A scope — the System Bus (\_SB), General Purpose Events (\_GPE), … —
-            // descend without adding a node.
-            try mirrorDevices(device_tree, c, parent_device, context, interpreter);
-            continue;
-        }
-
-        // Skip devices the firmware reports as not present (via a device-status (`_STA`) method),
-        // along with their whole subtree — per the ACPI rules.
-        if (!devicePresent(interpreter, c)) continue;
-
-        var mirrored_device: *device_model.Device = undefined;
-        var child_context = context;
-
-        if (isPciRootNode(c)) {
-            // The PCI root bridge folds onto the generic host bridge.
-            mirrored_device = matchHostBridge(device_tree) orelse
-                try device_tree.addChild(parent_device, .acpi_device, &c.segment);
-            child_context = .{ .bridge = mirrored_device, .bus = 0 };
-        } else {
-            // An addressed device folds onto its matching PCI function; anything
-            // else becomes a fresh node under the current parent.
-            mirrored_device = pick: {
-                if (context) |pc| {
-                    if (readAdr(c)) |adr| {
-                        if (findPciNode(pc.bridge, pc.bus, adr)) |pnode| break :pick pnode;
-                    }
-                }
-                break :pick try device_tree.addChild(parent_device, .acpi_device, &c.segment);
-            };
-        }
-
-        applyHid(mirrored_device, c, interpreter);
-        applyCrs(mirrored_device, c, interpreter);
-        try mirrorDevices(device_tree, c, mirrored_device, child_context, interpreter);
-    }
-}
-
-/// Evaluate a device's status (`_STA`) to decide if it is present. An absent status
-/// (`_STA`) means present by default; an evaluation failure is treated as present too (we'd
-/// rather over-report than hide a device we couldn't introspect).
-fn devicePresent(interpreter: *aml.Interpreter, node: *aml.Node) bool {
-    const sta = aml.Namespace.childOf(node, seg4("_STA")) orelse return true;
-    const obj = interpreter.evaluate(sta, &.{}) catch return true;
-    const status = obj.asInteger() catch return true;
-    return (status & 0x01) != 0; // bit 0 = present
-}
-
-/// The first PCI host bridge in the generic tree (segment 0).
-fn matchHostBridge(device_tree: *DeviceTree) ?*device_model.Device {
-    var c = device_tree.root.first_child;
-    while (c) |ch| : (c = ch.next_sibling) {
-        if (ch.class == .pci_host_bridge) return ch;
-    }
-    return null;
-}
-
-/// The PCI function node under `bridge` at the address the device's address object
-/// (`_ADR`) names (device/function on
-/// `bus`), or null.
-fn findPciNode(bridge: *device_model.Device, bus: u8, adr: u32) ?*device_model.Device {
-    const device: u16 = @truncate((adr >> 16) & 0x1F);
-    const function: u16 = @truncate(adr & 0x7);
-    const target: u16 = (@as(u16, bus) << 8) | (device << 3) | function;
-    var c = bridge.first_child;
-    while (c) |ch| : (c = ch.next_sibling) {
-        if (ch.ids.pci_bdf) |bdf| {
-            if (bdf == target) return ch;
-        }
-    }
-    return null;
-}
-
-/// A device's address (`_ADR`) — a static integer Name — or null.
-fn readAdr(node: *aml.Node) ?u32 {
-    const n = aml.Namespace.childOf(node, seg4("_ADR")) orelse return null;
-    if (n.kind != .name) return null;
-    var p: usize = 0;
-    return @truncate(readIntObj(n.value, &p) orelse return null);
-}
-
-/// Whether a namespace device is a PCI(e) host bridge (`PNP0A03` / `PNP0A08`).
-fn isPciRootNode(node: *aml.Node) bool {
-    const hid = aml.Namespace.childOf(node, seg4("_HID")) orelse return false;
-    if (hid.kind != .name or hid.value.len == 0) return false;
-    switch (hid.value[0]) {
-        0x00, 0x01, 0xFF, 0x0A, 0x0B, 0x0C, 0x0E => {
-            var p: usize = 0;
-            const n = readIntObj(hid.value, &p) orelse return false;
-            return n == 0x030AD041 or n == 0x080AD041; // PNP0A03 / PNP0A08
-        },
-        0x0D => {
-            const s = cstr(hid.value[1..]);
-            return std.mem.eql(u8, s, "PNP0A03") or std.mem.eql(u8, s, "PNP0A08");
-        },
-        else => return false,
-    }
-}
-
-/// Read a device's hardware ID (`_HID`) into the generic device: an integer decodes as an EISA
-/// id ("PNP0A03"), a string is taken verbatim. Handles both the common static
-/// Name form and a Method form (evaluated).
-fn applyHid(device: *device_model.Device, node: *aml.Node, interpreter: *aml.Interpreter) void {
-    const hid = aml.Namespace.childOf(node, seg4("_HID")) orelse return;
-    if (hid.kind == .method) {
-        const obj = interpreter.evaluate(hid, &.{}) catch return;
-        switch (obj) {
-            .integer => |n| setEisaHid(device, @truncate(n)),
-            .string => |s| device.setHid(s),
-            else => {},
-        }
-        return;
-    }
-    if (hid.kind != .name or hid.value.len == 0) return;
-    const v = hid.value;
-    switch (v[0]) {
-        0x00, 0x01, 0xFF, 0x0A, 0x0B, 0x0C, 0x0E => {
-            var p: usize = 0;
-            const n = readIntObj(v, &p) orelse return;
-            setEisaHid(device, @truncate(n));
-        },
-        0x0D => device.setHid(cstr(v[1..])), // StringPrefix
-        else => {},
-    }
-}
-
-fn setEisaHid(device: *device_model.Device, id: u32) void {
-    device.ids.acpi_hid = id;
-    var buffer: [8]u8 = undefined;
-    device.setHid(eisaIdToStr(id, &buffer));
-}
-
-/// Parse a device's current resource settings (`_CRS`). The evaluator handles both the static
-/// `Buffer` form (a `Name`) and the method form uniformly, yielding the
-/// ResourceTemplate bytes we then decode.
-fn applyCrs(device: *device_model.Device, node: *aml.Node, interpreter: *aml.Interpreter) void {
-    const crs = aml.Namespace.childOf(node, seg4("_CRS")) orelse return;
-    const obj = interpreter.evaluate(crs, &.{}) catch return;
-    const buffer = switch (obj) {
-        .buffer => |b| b,
-        else => return,
-    };
-    parseResourceTemplate(device, buffer);
-}
-
-/// Walk a ResourceTemplate byte list, adding recognised descriptors as resources.
-fn parseResourceTemplate(device: *device_model.Device, bytes: []const u8) void {
-    var i: usize = 0;
-    while (i < bytes.len) {
-        const tag = bytes[i];
-        if (tag & 0x80 == 0) {
-            // Small descriptor: length in low 3 bits, type in bits [6:3].
-            const len: usize = tag & 0x07;
-            const body = i + 1;
-            if (body + len > bytes.len) break;
-            switch ((tag >> 3) & 0x0F) {
-                0x04 => if (len >= 2) { // IRQ: a 16-bit mask, one resource per set bit
-                    const mask = @as(u16, bytes[body]) | (@as(u16, bytes[body + 1]) << 8);
-                    var b: usize = 0;
-                    while (b < 16) : (b += 1) {
-                        if (mask & (@as(u16, 1) << @intCast(b)) != 0) _ = device.addResource(.irq, b, 1);
-                    }
-                },
-                0x08 => if (len >= 7) { // IO port: minimum at +1, length at +6
-                    _ = device.addResource(.io_port, rd16(bytes, body + 1), bytes[body + 6]);
-                },
-                0x09 => if (len >= 3) { // Fixed IO: base at +0, length at +2
-                    _ = device.addResource(.io_port, rd16(bytes, body), bytes[body + 2]);
-                },
-                0x0F => break, // EndTag
-                else => {},
-            }
-            i = body + len;
-        } else {
-            // Large descriptor: 16-bit length follows the tag.
-            if (i + 3 > bytes.len) break;
-            const len: usize = @intCast(rd16(bytes, i + 1));
-            const body = i + 3;
-            if (body + len > bytes.len) break;
-            switch (tag) {
-                0x85 => if (len >= 17) { // Memory32: minimum at +1, length at +13
-                    _ = device.addResource(.memory, rd32(bytes, body + 1), rd32(bytes, body + 13));
-                },
-                0x86 => if (len >= 9) { // Memory32Fixed: base at +1, length at +5
-                    _ = device.addResource(.memory, rd32(bytes, body + 1), rd32(bytes, body + 5));
-                },
-                0x89 => if (len >= 2) { // Extended IRQ: count at +1, then count u32s
-                    const count = bytes[body + 1];
-                    var k: usize = 0;
-                    while (k < count and body + 2 + k * 4 + 4 <= body + len) : (k += 1) {
-                        _ = device.addResource(.irq, rd32(bytes, body + 2 + k * 4), 1);
-                    }
-                },
-                0x87, 0x88, 0x8A => parseAddressSpace(device, tag, bytes[body .. body + len]),
-                else => {},
-            }
-            i = body + len;
-        }
-    }
-}
-
-/// Word/DWord/QWord address-space descriptors: resource type at [0], then
-/// granularity/minimum/maximum/translation/length, each of width `w`.
-fn parseAddressSpace(device: *device_model.Device, tag: u8, body: []const u8) void {
-    const w: usize = switch (tag) {
-        0x88 => 2, // Word
-        0x87 => 4, // DWord
-        else => 8, // QWord (0x8A)
-    };
-    if (body.len < 3 + 5 * w) return;
-    const minimum = readN(body, 3 + w, w);
-    const length = readN(body, 3 + 4 * w, w);
-    const kind: device_model.ResourceKind = switch (body[0]) {
-        0 => .memory,
-        1 => .io_port,
-        else => .bus_range,
-    };
-    _ = device.addResource(kind, minimum, length);
-}
-
-/// Decode a packed EISA id into its 7-char string (e.g. 0x030AD041 -> "PNP0A03").
-fn eisaIdToStr(id: u32, buffer: *[8]u8) []const u8 {
-    const b0: u16 = @intCast(id & 0xFF);
-    const b1: u16 = @intCast((id >> 8) & 0xFF);
-    const b2: u8 = @truncate(id >> 16);
-    const b3: u8 = @truncate(id >> 24);
-    const mfg = (b0 << 8) | b1;
-    buffer[0] = '@' + @as(u8, @intCast((mfg >> 10) & 0x1F));
-    buffer[1] = '@' + @as(u8, @intCast((mfg >> 5) & 0x1F));
-    buffer[2] = '@' + @as(u8, @intCast(mfg & 0x1F));
-    buffer[3] = hexDigit((b2 >> 4) & 0xF);
-    buffer[4] = hexDigit(b2 & 0xF);
-    buffer[5] = hexDigit((b3 >> 4) & 0xF);
-    buffer[6] = hexDigit(b3 & 0xF);
-    return buffer[0..7];
-}
-
-fn hexDigit(n: u8) u8 {
-    return if (n < 10) '0' + n else 'A' + (n - 10);
-}
-
-fn seg4(comptime s: *const [4:0]u8) [4]u8 {
-    return s[0..4].*;
-}
-
-fn cstr(bytes: []const u8) []const u8 {
-    const index = std.mem.indexOfScalar(u8, bytes, 0) orelse bytes.len;
-    return bytes[0..index];
-}
-
-const PkgLen = struct { value: usize, size: usize };
-
-fn packageLength(bytes: []const u8, p: usize) ?PkgLen {
-    if (p >= bytes.len) return null;
-    const lead = bytes[p];
-    const follow: usize = lead >> 6;
-    if (p + 1 + follow > bytes.len) return null;
-    if (follow == 0) return .{ .value = lead & 0x3F, .size = 1 };
-    var value: usize = lead & 0x0F;
-    var i: usize = 0;
-    while (i < follow) : (i += 1) value |= @as(usize, bytes[p + 1 + i]) << @intCast(4 + i * 8);
-    return .{ .value = value, .size = 1 + follow };
-}
-
-/// Read an AML integer object at `p`, advancing `p` past it.
-fn readIntObj(bytes: []const u8, p: *usize) ?u64 {
-    if (p.* >= bytes.len) return null;
-    const opcode = bytes[p.*];
-    p.* += 1;
-    return switch (opcode) {
-        0x00 => 0,
-        0x01 => 1,
-        0xFF => 0xFF,
-        0x0A => readLE(bytes, p, 1),
-        0x0B => readLE(bytes, p, 2),
-        0x0C => readLE(bytes, p, 4),
-        0x0E => readLE(bytes, p, 8),
-        else => null,
-    };
-}
-
-fn readLE(bytes: []const u8, p: *usize, n: usize) ?u64 {
-    if (p.* + n > bytes.len) return null;
-    const v = readN(bytes, p.*, n);
-    p.* += n;
-    return v;
-}
-
-fn readN(bytes: []const u8, off: usize, n: usize) u64 {
-    var v: u64 = 0;
-    var k: usize = 0;
-    while (k < n and off + k < bytes.len) : (k += 1) v |= @as(u64, bytes[off + k]) << @intCast(k * 8);
-    return v;
-}
-
-fn rd16(bytes: []const u8, off: usize) u64 {
-    return readN(bytes, off, 2);
-}
-
-fn rd32(bytes: []const u8, off: usize) u64 {
-    return readN(bytes, off, 4);
-}
-
 // --- helpers ----------------------------------------------------------------
 
 /// Sum `len` bytes; an ACPI table/pointer is valid when the low 8 bits are zero.
@@ -1176,58 +867,9 @@ fn readCntRegister(base: [*]align(1) const u8, len: usize, xoff: usize, legacy_o
     return .{ .mmio = false, .address = port, .width = width };
 }
 
-/// The mapped configuration space of one PCI function (its 4 KiB ECAM page). Mapped
-/// writable so BAR sizing can probe it; reads and writes both go through here.
-fn pciConfigurationPtr(alloc: McfgAllocation, hal: Hal, bus: u8, device: u8, function: u8) [*]align(1) u8 {
-    const physical = alloc.base_address +
-        (@as(u64, bus - alloc.start_bus) << 20) +
-        (@as(u64, device) << 15) +
-        (@as(u64, function) << 12);
-    // Map the configuration page (writable, for BAR sizing) and use the virtual
-    // address the HAL hands back.
-    return @ptrFromInt(hal.mapMmio(physical, abi.page_size, true));
-}
-
 /// Read a little-endian integer at `off` from a (possibly unaligned) byte pointer.
 /// x86 is little-endian and native, so an unaligned load suffices.
 fn rd(comptime T: type, bytes: [*]align(1) const u8, off: usize) T {
     const p: *align(1) const T = @ptrCast(bytes + off);
     return p.*;
 }
-
-/// Write a little-endian integer at `off` through a (possibly unaligned) pointer.
-fn wr(comptime T: type, bytes: [*]align(1) u8, off: usize, value: T) void {
-    const p: *align(1) T = @ptrCast(bytes + off);
-    p.* = value;
-}
-
-// --- tests ------------------------------------------------------------------
-
-test "eisaIdToStr decodes a packed EISA id" {
-    var buffer: [8]u8 = undefined;
-    // 0x030AD041 is the well-known encoding of "PNP0A03" (PCI root bridge).
-    try std.testing.expectEqualStrings("PNP0A03", eisaIdToStr(0x030AD041, &buffer));
-}
-
-test "parseResourceTemplate extracts IO, IRQ, and fixed memory" {
-    // ResourceTemplate { IO(minimum 0x60, len 8), IRQ(4), Memory32Fixed(0xFED00000, 0x1000) }
-    const runtime = [_]u8{
-        0x47, 0x01, 0x60, 0x00, 0x60, 0x00, 0x01, 0x08, // small IO descriptor
-        0x22, 0x10, 0x00, // small IRQ descriptor (mask bit 4 -> IRQ 4)
-        0x86, 0x09, 0x00, 0x01, 0x00, 0x00, 0xD0, 0xFE, 0x00, 0x10, 0x00, 0x00, // Memory32Fixed
-        0x79, 0x00, // EndTag
-    };
-    var device = device_model.Device{};
-    parseResourceTemplate(&device, &runtime);
-
-    try std.testing.expectEqual(@as(u8, 3), device.resource_count);
-    const rs = device.resources[0..device.resource_count];
-    try std.testing.expectEqual(device_model.ResourceKind.io_port, rs[0].kind);
-    try std.testing.expectEqual(@as(u64, 0x60), rs[0].start);
-    try std.testing.expectEqual(@as(u64, 8), rs[0].len);
-    try std.testing.expectEqual(device_model.ResourceKind.irq, rs[1].kind);
-    try std.testing.expectEqual(@as(u64, 4), rs[1].start);
-    try std.testing.expectEqual(device_model.ResourceKind.memory, rs[2].kind);
-    try std.testing.expectEqual(@as(u64, 0xFED00000), rs[2].start);
-    try std.testing.expectEqual(@as(u64, 0x1000), rs[2].len);
-}

system/devices/aml/aml.zig

@@ -12,6 +12,11 @@ const std = @import("std");
 const opcode = @import("opcodes.zig");
 const parser = @import("parser.zig");
 
+/// The named AML opcode/prefix bytes (`zero_opcode`, `byte_prefix`, …). Re-exported so
+/// callers that decode raw AML bytes — e.g. the acpi service reading a `_HID` integer —
+/// name the opcodes instead of writing bare 0x0A/0x0B/… literals (docs/coding-standards.md).
+pub const opcodes = @import("opcodes.zig");
+
 pub const Namespace = @import("namespace.zig").Namespace;
 pub const Node = @import("namespace.zig").Node;
 pub const NodeKind = @import("namespace.zig").NodeKind;
@@ -50,6 +55,20 @@ pub fn parse(allocator: std.mem.Allocator, blocks: []const []const u8) !ParseRes
     return .{ .namespace = namespace, .consumed = consumed, .total = total };
 }
 
+/// Count the Device objects in a parsed namespace — what the acpi service
+/// (docs/discovery.md) reports, and what the kernel's own parse counts
+/// so the two can be checked equal across the ring-3 move.
+pub fn deviceCount(namespace: *const Namespace) usize {
+    return countKind(namespace.root, .device);
+}
+
+fn countKind(node: *const Node, kind: NodeKind) usize {
+    var n: usize = if (node.kind == kind) 1 else 0;
+    var c = node.first_child;
+    while (c) |child| : (c = child.next_sibling) n += countKind(child, kind);
+    return n;
+}
+
 /// Look up the `\_S{state}` sleep package in a parsed namespace and return its
 /// first two integer elements (SLP_TYP for PM1a / PM1b), or null if absent.
 pub fn sleepState(namespace: *Namespace, state: u8) ?SleepType {
@@ -126,17 +145,22 @@ test "parses a nested namespace and finds the sleep package" {
         // Scope(\_SB) packagelen=0x27
         0x10, 0x27, 0x5C, 0x5F, 0x53, 0x42, 0x5F,
         // Device(PCI0) packagelen=0x1F
-        0x5B, 0x82, 0x1F, 0x50, 0x43, 0x49, 0x30,
+        0x5B, 0x82, 0x1F, 0x50, 0x43,
+        0x49, 0x30,
         // Name(_HID, 0x11)
         0x08, 0x5F, 0x48, 0x49, 0x44, 0x0A, 0x11,
         // Method(MTHD, flags=1) empty, packagelen=0x06
-        0x14, 0x06, 0x4D, 0x54, 0x48, 0x44, 0x01,
+        0x14, 0x06, 0x4D,
+        0x54, 0x48, 0x44, 0x01,
         // Method(CALL, flags=0) { MTHD(Zero) }, packagelen=0x0B
-        0x14, 0x0B, 0x43, 0x41, 0x4C, 0x4C, 0x00, 0x4D, 0x54, 0x48, 0x44, 0x00,
+        0x14, 0x0B, 0x43, 0x41, 0x4C, 0x4C, 0x00, 0x4D,
+        0x54, 0x48, 0x44, 0x00,
         // OperationRegion(DBG0, SystemIO, Word 0x0402, Byte 1)
-        0x5B, 0x80, 0x44, 0x42, 0x47, 0x30, 0x01, 0x0B, 0x02, 0x04, 0x0A, 0x01,
+        0x5B, 0x80, 0x44, 0x42, 0x47, 0x30, 0x01, 0x0B,
+        0x02, 0x04, 0x0A, 0x01,
         // Field(DBG0, flags=1) { DBGB, 8 }, packagelen=0x0B
-        0x5B, 0x81, 0x0B, 0x44, 0x42, 0x47, 0x30, 0x01, 0x44, 0x42, 0x47, 0x42, 0x08,
+        0x5B, 0x81, 0x0B, 0x44, 0x42, 0x47, 0x30, 0x01,
+        0x44, 0x42, 0x47, 0x42, 0x08,
     };
 
     var arena = std.heap.ArenaAllocator.init(std.testing.allocator);
@@ -206,3 +230,31 @@ test "interpreter runs a method with args, arithmetic, and control flow" {
     const lo = try interpreter.evaluate(tst, &.{.{ .integer = 2 }}); // 2+5=7 !> 10 -> 0
     try std.testing.expectEqual(@as(u64, 0), try lo.asInteger());
 }
+
+test "interpreter records Notify(device, code)" {
+    // Device(DEV_) { Name(_HID, 0x030AD041) }   // PNP0A03-ish placeholder
+    // Method(TST_, 0) { Notify(DEV_, 0x80); Return(Zero) }
+    // Encoded: a Device holding a Name, then a Method issuing Notify on it.
+    const blob = [_]u8{
+        0x5B, 0x82, 0x0F, 0x44, 0x45, 0x56, 0x5F, // Device(DEV_) len=0x0F (pkglen + DEV_ + Name)
+        0x08, 0x5F, 0x48, 0x49, 0x44, 0x0C, 0x41, 0xD0, 0x0A, 0x03, // Name(_HID, DWord 0x030AD041)
+        0x14, 0x0F, 0x54, 0x53, 0x54, 0x5F, 0x00, // Method(TST_, 0) len=0x0F (pkglen + TST_ + flags + body)
+        0x86, 0x44, 0x45, 0x56, 0x5F, 0x0A, 0x80, // Notify(DEV_, 0x80)
+        0xA4, 0x00, // Return(Zero)
+    };
+
+    var arena = std.heap.ArenaAllocator.init(std.testing.allocator);
+    defer arena.deinit();
+    var result = try parse(arena.allocator(), &.{&blob});
+    const namespace = &result.namespace;
+    const tst = namespace.resolve(namespace.root, false, 0, &.{.{ 'T', 'S', 'T', '_' }}) orelse return error.NoMethod;
+    const dev = namespace.resolve(namespace.root, false, 0, &.{.{ 'D', 'E', 'V', '_' }}) orelse return error.NoDevice;
+
+    var interpreter = Interpreter.init(namespace, .{ .mapMmio = noMap, .pioRead = noRead, .pioWrite = noWrite }, arena.allocator());
+    _ = try interpreter.evaluate(tst, &.{});
+
+    const events = interpreter.takeNotifications();
+    try std.testing.expectEqual(@as(usize, 1), events.len);
+    try std.testing.expectEqual(dev, events[0].node);
+    try std.testing.expectEqual(@as(u64, 0x80), events[0].code);
+}

system/devices/aml/interpreter.zig

@@ -141,6 +141,9 @@ const Frame = struct {
 /// A CreateField binding: a name that indexes into a buffer object.
 const BufferField = struct { buffer: *Node, byte_off: usize, bit_width: u32 };
 
+/// One Notify(device, code) the interpreter executed.
+pub const NotifyEvent = struct { node: *Node, code: u64 };
+
 pub const Interpreter = struct {
     namespace: *Namespace,
     hal: Hal,
@@ -149,6 +152,11 @@ pub const Interpreter = struct {
     dynamic_overrides: std.AutoHashMapUnmanaged(*Node, Object) = .{},
     /// CreateField bindings active for the current evaluation.
     fields: std.AutoHashMapUnmanaged(*Node, BufferField) = .{},
+    /// Notify(device, code) operations the last evaluation executed — a GPE or
+    /// EC handler tells the OS "look at this device" this way. Bounded; the
+    /// caller drains it with `takeNotifications` after `evaluate` (M21).
+    notify_queue: [16]NotifyEvent = undefined,
+    notify_count: usize = 0,
 
     pub fn init(namespace: *Namespace, hal: Hal, arena: std.mem.Allocator) Interpreter {
         return .{ .namespace = namespace, .hal = hal, .arena = arena };
@@ -157,6 +165,7 @@ pub const Interpreter = struct {
     /// Evaluate a namespace object: invoke a Method, read a Name's value, or read a
     /// Field. Resets per-evaluation runtime state first.
     pub fn evaluate(self: *Interpreter, node: *Node, args: []const Object) Error!Object {
+        self.notify_count = 0;
         self.dynamic_overrides.clearRetainingCapacity();
         self.fields.clearRetainingCapacity();
         return self.invoke(node, args);
@@ -267,6 +276,8 @@ pub const Interpreter = struct {
             },
             opcode.to_buffer_opcode => try self.passThroughUnary(current, frame),
 
+            opcode.notify_opcode => try self.notify(current, frame),
+
             opcode.extended_opcode_prefix => try self.ext(current, frame),
 
             // CreateXField: source, index, name  (bit widths differ by op)
@@ -542,6 +553,36 @@ pub const Interpreter = struct {
         try self.storeInto(current, frame, value);
     }
 
+    /// Notify(SuperName, NotifyValue): resolve the named device, evaluate the
+    /// code, and record the pair for the caller to dispatch. AML control flow
+    /// continues (Notify returns nothing).
+    fn notify(self: *Interpreter, current: *Cursor, frame: *Frame) Error!Object {
+        const lead = current.peek() orelse return error.Truncated;
+        var target: ?*Node = null;
+        if (isNameStart(lead)) {
+            const name_path = try current.nameString();
+            target = self.namespace.resolve(frame.scope, name_path.rooted, name_path.parents, name_path.slice());
+        } else {
+            // A non-name SuperName (Local/Arg holding a reference).
+            const obj = try self.term(current, frame);
+            if (obj == .reference) target = obj.reference;
+        }
+        const code = try self.evaluateInteger(current, frame);
+        if (target) |node| {
+            if (self.notify_count < self.notify_queue.len) {
+                self.notify_queue[self.notify_count] = .{ .node = node, .code = code };
+                self.notify_count += 1;
+            }
+        }
+        return .uninitialized;
+    }
+
+    /// The Notify events the last `evaluate` produced. Valid until the next
+    /// `evaluate` clears the queue.
+    pub fn takeNotifications(self: *Interpreter) []const NotifyEvent {
+        return self.notify_queue[0..self.notify_count];
+    }
+
     fn storeInto(self: *Interpreter, current: *Cursor, frame: *Frame, value: Object) Error!void {
         const lead = current.peek() orelse return error.Truncated;
         if (isNameStart(lead)) {

system/devices/device-abi.zig

@@ -28,6 +28,16 @@ pub const DeviceClass = enum(u32) {
     /// A device named in the ACPI namespace (from the DSDT/SSDT), carrying a
     /// hardware ID (`_HID`) and, where static, current resource settings (`_CRS`).
     acpi_device,
+    /// The ACPI tables themselves, published as one node for the user-space acpi
+    /// service (docs/discovery.md): memory resources over the AML blobs,
+    /// a broad io_port grant for OperationRegion access, and the SCI interrupt.
+    /// The one node whose claimant is trusted to run firmware bytecode.
+    acpi_tables,
+    /// One interface of a USB device, registered by the xHCI bus driver. It owns
+    /// no MMIO — it is reached through its controller — so it carries no
+    /// resources; the (class, subclass, protocol) triple that says what it is
+    /// travels in the bus report's identity, not here.
+    usb_device,
     unknown,
 };
 
@@ -56,6 +66,10 @@ pub const maximum_device_resources = 8;
 /// `DeviceDescriptor.parent` for a device with no parent — a root of the device tree.
 pub const no_parent: u64 = ~@as(u64, 0);
 
+/// `DeviceDescriptor.pci_class` for a device that is not a PCI function. (Zero would be
+/// ambiguous: 0x000000 is a real class code, "unclassified device".)
+pub const no_pci_class: u64 = ~@as(u64, 0);
+
 /// A device, as snapshotted for user space by `device_enumerate`. A driver scans
 /// these to find the hardware it owns, claims it, and maps its MMIO.
 ///
@@ -69,6 +83,11 @@ pub const DeviceDescriptor = extern struct {
     id: u64,
     parent: u64, // a device id, or `no_parent`
     class: u64, // a DeviceClass value
+    // The PCI class/subclass/prog-IF triple packed as 0xCCSSPP when this device is a PCI
+    // function, or `no_pci_class` otherwise. This is how a manager tells *what* a
+    // `pci_device` is (an xHCI controller, an AHCI controller) — decode the triple into
+    // names with the pci-class module.
+    pci_class: u64,
     hid_len: u64,
     resource_count: u64,
     hid: [8]u8,

system/devices/device-model.zig

@@ -3,7 +3,7 @@
 //! Discovery backends (ACPI today, device-tree later) translate their native
 //! hardware description into this one shape, so the rest of the kernel walks a
 //! plain `Device` tree without knowing which firmware described the machine —
-//! the same discipline `root.zig`'s `MemoryKind` applies to memory and `architecture`
+//! the same discipline `ps2-library.zig`'s `MemoryKind` applies to memory and `architecture`
 //! applies to the CPU.
 //!
 //! This is deliberately minimal: enough to *describe* what was discovered (a
@@ -198,8 +198,7 @@ fn dumpNode(device: *const Device, depth: usize, emit: *const fn ([]const u8) vo
             std.fmt.bufPrint(buffer[indent..], "{s} [{s}] hid={s} ({s})\n", .{ device.name(), @tagName(device.class), device.hid(), desc }) catch return
         else
             std.fmt.bufPrint(buffer[indent..], "{s} [{s}] hid={s}\n", .{ device.name(), @tagName(device.class), device.hid() }) catch return;
-    } else
-        std.fmt.bufPrint(buffer[indent..], "{s} [{s}]\n", .{ device.name(), @tagName(device.class) }) catch return;
+    } else std.fmt.bufPrint(buffer[indent..], "{s} [{s}]\n", .{ device.name(), @tagName(device.class) }) catch return;
     emit(buffer[0 .. indent + body.len]);
 
     // For a PCI function, decode its class code — the (class / subclass / prog-IF)

system/devices/pci-class.zig

@@ -7,6 +7,16 @@
 //! apart. Pure reference data (from the PCI spec; see https://wiki.osdev.org/PCI) — no
 //! hardware access — so it is shared by kernel discovery (the device-tree dump) and any
 //! user-space tool (a future lspci, driver matching).
+//!
+//! The taxonomy is named, not numbered (docs/coding-standards.md, "Named values"): the
+//! base class is a `BaseClass` enum, and each class with defined subclasses gets a
+//! namespace holding its `SubClass` enum (and, where the spec defines them, per-subclass
+//! `ProgIf` enums) — the same shape as `usb-ids.zig`. Code that *means* a specific class
+//! names it (`BaseClass.serial_bus`, `serial_bus.usb.ProgIf.xhci`) rather than writing a
+//! bare 0x0C/0x03/0x30. The `className`/`subclassName`/`progIfName` functions still take
+//! the raw bytes a function reports in its header, because that is what hardware hands us.
+
+const std = @import("std");
 
 /// The three bytes of a PCI class code, unpacked from the `0xCCSSPP` value discovery
 /// records in `Device.ids.pci_class` (CC = base class, SS = subclass, PP = prog-IF).
@@ -22,148 +32,465 @@ pub const ClassCode = struct {
             .prog_if = @intCast(packed_code & 0xFF),
         };
     }
+
+    /// Re-pack the triple into the `0xCCSSPP` form. Lets code name a whole class code
+    /// from its parts — `pack(.{ .base = @intFromEnum(BaseClass.serial_bus), … })` —
+    /// instead of writing the literal 0x0C0330.
+    pub fn pack(self: ClassCode) u24 {
+        return (@as(u24, self.base) << 16) | (@as(u24, self.subclass) << 8) | self.prog_if;
+    }
 };
 
+/// Base class (config byte 0x0B). Non-exhaustive: an unlisted code is a real but
+/// unnamed class, decoded as "Unknown" rather than rejected.
+pub const BaseClass = enum(u8) {
+    unclassified = 0x00,
+    mass_storage = 0x01,
+    network = 0x02,
+    display = 0x03,
+    multimedia = 0x04,
+    memory = 0x05,
+    bridge = 0x06,
+    simple_communication = 0x07,
+    base_system_peripheral = 0x08,
+    input_device = 0x09,
+    docking_station = 0x0A,
+    processor = 0x0B,
+    serial_bus = 0x0C,
+    wireless = 0x0D,
+    intelligent = 0x0E,
+    satellite_communication = 0x0F,
+    encryption = 0x10,
+    signal_processing = 0x11,
+    processing_accelerator = 0x12,
+    non_essential_instrumentation = 0x13,
+    co_processor = 0x40,
+    unassigned = 0xFF,
+    _,
+
+    pub fn name(self: BaseClass) []const u8 {
+        return switch (self) {
+            .unclassified => "Unclassified",
+            .mass_storage => "Mass Storage Controller",
+            .network => "Network Controller",
+            .display => "Display Controller",
+            .multimedia => "Multimedia Controller",
+            .memory => "Memory Controller",
+            .bridge => "Bridge",
+            .simple_communication => "Simple Communication Controller",
+            .base_system_peripheral => "Base System Peripheral",
+            .input_device => "Input Device Controller",
+            .docking_station => "Docking Station",
+            .processor => "Processor",
+            .serial_bus => "Serial Bus Controller",
+            .wireless => "Wireless Controller",
+            .intelligent => "Intelligent Controller",
+            .satellite_communication => "Satellite Communication Controller",
+            .encryption => "Encryption Controller",
+            .signal_processing => "Signal Processing Controller",
+            .processing_accelerator => "Processing Accelerator",
+            .non_essential_instrumentation => "Non-Essential Instrumentation",
+            .co_processor => "Co-Processor",
+            .unassigned => "Unassigned Class (Vendor specific)",
+            _ => "Unknown",
+        };
+    }
+};
+
+// --- Per-class subclass (and prog-IF) taxonomies --------------------------------------
+// One namespace per base class that has defined subclasses, named after the class. Each
+// holds an exhaustive `SubClass` enum (so an unlisted code decodes to the class default,
+// not a wrong name), and, where the spec assigns them, per-subclass `ProgIf` enums.
+
+pub const mass_storage = struct {
+    pub const SubClass = enum(u8) {
+        scsi_bus = 0x00,
+        ide = 0x01,
+        floppy = 0x02,
+        ipi_bus = 0x03,
+        raid = 0x04,
+        ata = 0x05,
+        serial_ata = 0x06,
+        serial_attached_scsi = 0x07,
+        non_volatile_memory = 0x08,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .scsi_bus => "SCSI Bus Controller",
+                .ide => "IDE Controller",
+                .floppy => "Floppy Disk Controller",
+                .ipi_bus => "IPI Bus Controller",
+                .raid => "RAID Controller",
+                .ata => "ATA Controller",
+                .serial_ata => "Serial ATA Controller",
+                .serial_attached_scsi => "Serial Attached SCSI Controller",
+                .non_volatile_memory => "Non-Volatile Memory Controller",
+            };
+        }
+    };
+
+    pub const serial_ata = struct {
+        pub const ProgIf = enum(u8) {
+            vendor_specific = 0x00,
+            ahci = 0x01,
+            serial_storage_bus = 0x02,
+
+            pub fn name(self: ProgIf) []const u8 {
+                return switch (self) {
+                    .vendor_specific => "Vendor Specific Interface",
+                    .ahci => "AHCI 1.0",
+                    .serial_storage_bus => "Serial Storage Bus",
+                };
+            }
+        };
+    };
+
+    pub const non_volatile_memory = struct {
+        pub const ProgIf = enum(u8) {
+            nvmhci = 0x01,
+            nvm_express = 0x02,
+
+            pub fn name(self: ProgIf) []const u8 {
+                return switch (self) {
+                    .nvmhci => "NVMHCI",
+                    .nvm_express => "NVM Express",
+                };
+            }
+        };
+    };
+};
+
+pub const network = struct {
+    pub const SubClass = enum(u8) {
+        ethernet = 0x00,
+        token_ring = 0x01,
+        fddi = 0x02,
+        atm = 0x03,
+        isdn = 0x04,
+        picmg_multi_computing = 0x06,
+        infiniband = 0x07,
+        fabric = 0x08,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .ethernet => "Ethernet Controller",
+                .token_ring => "Token Ring Controller",
+                .fddi => "FDDI Controller",
+                .atm => "ATM Controller",
+                .isdn => "ISDN Controller",
+                .picmg_multi_computing => "PICMG 2.14 Multi Computing Controller",
+                .infiniband => "Infiniband Controller",
+                .fabric => "Fabric Controller",
+            };
+        }
+    };
+};
+
+pub const display = struct {
+    pub const SubClass = enum(u8) {
+        vga_compatible = 0x00,
+        xga = 0x01,
+        three_dimensional = 0x02,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .vga_compatible => "VGA Compatible Controller",
+                .xga => "XGA Controller",
+                .three_dimensional => "3D Controller (Not VGA-Compatible)",
+            };
+        }
+    };
+
+    pub const vga_compatible = struct {
+        pub const ProgIf = enum(u8) {
+            vga = 0x00,
+            compatible_8514 = 0x01,
+
+            pub fn name(self: ProgIf) []const u8 {
+                return switch (self) {
+                    .vga => "VGA Controller",
+                    .compatible_8514 => "8514-Compatible Controller",
+                };
+            }
+        };
+    };
+};
+
+pub const multimedia = struct {
+    pub const SubClass = enum(u8) {
+        video = 0x00,
+        audio = 0x01,
+        telephony = 0x02,
+        audio_device = 0x03,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .video => "Multimedia Video Controller",
+                .audio => "Multimedia Audio Controller",
+                .telephony => "Computer Telephony Device",
+                .audio_device => "Audio Device",
+            };
+        }
+    };
+};
+
+pub const memory = struct {
+    pub const SubClass = enum(u8) {
+        ram = 0x00,
+        flash = 0x01,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .ram => "RAM Controller",
+                .flash => "Flash Controller",
+            };
+        }
+    };
+};
+
+pub const bridge = struct {
+    pub const SubClass = enum(u8) {
+        host = 0x00,
+        isa = 0x01,
+        eisa = 0x02,
+        mca = 0x03,
+        pci_to_pci = 0x04,
+        pcmcia = 0x05,
+        nubus = 0x06,
+        cardbus = 0x07,
+        raceway = 0x08,
+        pci_to_pci_semi_transparent = 0x09,
+        infiniband_to_pci = 0x0A,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .host => "Host Bridge",
+                .isa => "ISA Bridge",
+                .eisa => "EISA Bridge",
+                .mca => "MCA Bridge",
+                .pci_to_pci => "PCI-to-PCI Bridge",
+                .pcmcia => "PCMCIA Bridge",
+                .nubus => "NuBus Bridge",
+                .cardbus => "CardBus Bridge",
+                .raceway => "RACEway Bridge",
+                .pci_to_pci_semi_transparent => "PCI-to-PCI Bridge (Semi-Transparent)",
+                .infiniband_to_pci => "InfiniBand-to-PCI Host Bridge",
+            };
+        }
+    };
+
+    pub const pci_to_pci = struct {
+        pub const ProgIf = enum(u8) {
+            normal_decode = 0x00,
+            subtractive_decode = 0x01,
+
+            pub fn name(self: ProgIf) []const u8 {
+                return switch (self) {
+                    .normal_decode => "Normal Decode",
+                    .subtractive_decode => "Subtractive Decode",
+                };
+            }
+        };
+    };
+};
+
+pub const simple_communication = struct {
+    pub const SubClass = enum(u8) {
+        serial = 0x00,
+        parallel = 0x01,
+        multiport_serial = 0x02,
+        modem = 0x03,
+        gpib = 0x04,
+        smart_card = 0x05,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .serial => "Serial Controller",
+                .parallel => "Parallel Controller",
+                .multiport_serial => "Multiport Serial Controller",
+                .modem => "Modem",
+                .gpib => "IEEE 488.1/2 (GPIB) Controller",
+                .smart_card => "Smart Card Controller",
+            };
+        }
+    };
+
+    pub const serial = struct {
+        pub const ProgIf = enum(u8) {
+            compatible_8250 = 0x00,
+            compatible_16450 = 0x01,
+            compatible_16550 = 0x02,
+            compatible_16650 = 0x03,
+            compatible_16750 = 0x04,
+            compatible_16850 = 0x05,
+            compatible_16950 = 0x06,
+
+            pub fn name(self: ProgIf) []const u8 {
+                return switch (self) {
+                    .compatible_8250 => "8250-Compatible (Generic XT)",
+                    .compatible_16450 => "16450-Compatible",
+                    .compatible_16550 => "16550-Compatible",
+                    .compatible_16650 => "16650-Compatible",
+                    .compatible_16750 => "16750-Compatible",
+                    .compatible_16850 => "16850-Compatible",
+                    .compatible_16950 => "16950-Compatible",
+                };
+            }
+        };
+    };
+};
+
+pub const base_system_peripheral = struct {
+    pub const SubClass = enum(u8) {
+        pic = 0x00,
+        dma = 0x01,
+        timer = 0x02,
+        rtc = 0x03,
+        pci_hot_plug = 0x04,
+        sd_host = 0x05,
+        iommu = 0x06,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .pic => "PIC",
+                .dma => "DMA Controller",
+                .timer => "Timer",
+                .rtc => "RTC Controller",
+                .pci_hot_plug => "PCI Hot-Plug Controller",
+                .sd_host => "SD Host Controller",
+                .iommu => "IOMMU",
+            };
+        }
+    };
+};
+
+pub const input_device = struct {
+    pub const SubClass = enum(u8) {
+        keyboard = 0x00,
+        digitizer_pen = 0x01,
+        mouse = 0x02,
+        scanner = 0x03,
+        gameport = 0x04,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .keyboard => "Keyboard Controller",
+                .digitizer_pen => "Digitizer Pen",
+                .mouse => "Mouse Controller",
+                .scanner => "Scanner Controller",
+                .gameport => "Gameport Controller",
+            };
+        }
+    };
+};
+
+pub const serial_bus = struct {
+    pub const SubClass = enum(u8) {
+        firewire = 0x00,
+        access_bus = 0x01,
+        ssa = 0x02,
+        usb = 0x03,
+        fibre_channel = 0x04,
+        smbus = 0x05,
+        infiniband = 0x06,
+        ipmi = 0x07,
+        sercos = 0x08,
+        canbus = 0x09,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .firewire => "FireWire (IEEE 1394) Controller",
+                .access_bus => "ACCESS Bus Controller",
+                .ssa => "SSA",
+                .usb => "USB Controller",
+                .fibre_channel => "Fibre Channel",
+                .smbus => "SMBus Controller",
+                .infiniband => "InfiniBand Controller",
+                .ipmi => "IPMI Interface",
+                .sercos => "SERCOS Interface (IEC 61491)",
+                .canbus => "CANbus Controller",
+            };
+        }
+    };
+
+    pub const usb = struct {
+        pub const ProgIf = enum(u8) {
+            uhci = 0x00,
+            ohci = 0x10,
+            ehci = 0x20,
+            xhci = 0x30,
+            unspecified = 0x80,
+            device = 0xFE,
+
+            pub fn name(self: ProgIf) []const u8 {
+                return switch (self) {
+                    .uhci => "UHCI Controller",
+                    .ohci => "OHCI Controller",
+                    .ehci => "EHCI (USB2) Controller",
+                    .xhci => "XHCI (USB3) Controller",
+                    .unspecified => "Unspecified",
+                    .device => "USB Device (not a host controller)",
+                };
+            }
+        };
+    };
+};
+
+pub const wireless = struct {
+    pub const SubClass = enum(u8) {
+        irda = 0x00,
+        consumer_ir = 0x01,
+        rf = 0x10,
+        bluetooth = 0x11,
+        broadband = 0x12,
+        ethernet_802_1a = 0x20,
+        ethernet_802_1b = 0x21,
+
+        pub fn name(self: SubClass) []const u8 {
+            return switch (self) {
+                .irda => "iRDA Compatible Controller",
+                .consumer_ir => "Consumer IR Controller",
+                .rf => "RF Controller",
+                .bluetooth => "Bluetooth Controller",
+                .broadband => "Broadband Controller",
+                .ethernet_802_1a => "Ethernet Controller (802.1a)",
+                .ethernet_802_1b => "Ethernet Controller (802.1b)",
+            };
+        }
+    };
+};
+
+// --- Raw-byte decoding (what a function reports in its header) -------------------------
+
+/// The name of an exhaustive class-code enum member, or null if `value` is not one — the
+/// bridge from a raw config byte to a named taxonomy above.
+fn enumName(comptime Enum: type, value: u8) ?[]const u8 {
+    return (std.enums.fromInt(Enum, value) orelse return null).name();
+}
+
 /// Name of the base class (byte 0x0B), e.g. `0x06` -> "Bridge".
 pub fn className(base: u8) []const u8 {
-    return switch (base) {
-        0x00 => "Unclassified",
-        0x01 => "Mass Storage Controller",
-        0x02 => "Network Controller",
-        0x03 => "Display Controller",
-        0x04 => "Multimedia Controller",
-        0x05 => "Memory Controller",
-        0x06 => "Bridge",
-        0x07 => "Simple Communication Controller",
-        0x08 => "Base System Peripheral",
-        0x09 => "Input Device Controller",
-        0x0A => "Docking Station",
-        0x0B => "Processor",
-        0x0C => "Serial Bus Controller",
-        0x0D => "Wireless Controller",
-        0x0E => "Intelligent Controller",
-        0x0F => "Satellite Communication Controller",
-        0x10 => "Encryption Controller",
-        0x11 => "Signal Processing Controller",
-        0x12 => "Processing Accelerator",
-        0x13 => "Non-Essential Instrumentation",
-        0x40 => "Co-Processor",
-        0xFF => "Unassigned Class (Vendor specific)",
-        else => "Unknown",
-    };
+    return @as(BaseClass, @enumFromInt(base)).name();
 }
 
 /// Name of the subclass within its base class, e.g. `(0x06, 0x01)` -> "ISA Bridge".
 /// Subclass `0x80` is "Other" by PCI convention; anything unlisted is "Unknown".
 pub fn subclassName(base: u8, subclass: u8) []const u8 {
-    return switch (base) {
-        0x01 => switch (subclass) {
-            0x00 => "SCSI Bus Controller",
-            0x01 => "IDE Controller",
-            0x02 => "Floppy Disk Controller",
-            0x03 => "IPI Bus Controller",
-            0x04 => "RAID Controller",
-            0x05 => "ATA Controller",
-            0x06 => "Serial ATA Controller",
-            0x07 => "Serial Attached SCSI Controller",
-            0x08 => "Non-Volatile Memory Controller",
-            else => defaultSubclass(subclass),
-        },
-        0x02 => switch (subclass) {
-            0x00 => "Ethernet Controller",
-            0x01 => "Token Ring Controller",
-            0x02 => "FDDI Controller",
-            0x03 => "ATM Controller",
-            0x04 => "ISDN Controller",
-            0x06 => "PICMG 2.14 Multi Computing Controller",
-            0x07 => "Infiniband Controller",
-            0x08 => "Fabric Controller",
-            else => defaultSubclass(subclass),
-        },
-        0x03 => switch (subclass) {
-            0x00 => "VGA Compatible Controller",
-            0x01 => "XGA Controller",
-            0x02 => "3D Controller (Not VGA-Compatible)",
-            else => defaultSubclass(subclass),
-        },
-        0x04 => switch (subclass) {
-            0x00 => "Multimedia Video Controller",
-            0x01 => "Multimedia Audio Controller",
-            0x02 => "Computer Telephony Device",
-            0x03 => "Audio Device",
-            else => defaultSubclass(subclass),
-        },
-        0x05 => switch (subclass) {
-            0x00 => "RAM Controller",
-            0x01 => "Flash Controller",
-            else => defaultSubclass(subclass),
-        },
-        0x06 => switch (subclass) {
-            0x00 => "Host Bridge",
-            0x01 => "ISA Bridge",
-            0x02 => "EISA Bridge",
-            0x03 => "MCA Bridge",
-            0x04 => "PCI-to-PCI Bridge",
-            0x05 => "PCMCIA Bridge",
-            0x06 => "NuBus Bridge",
-            0x07 => "CardBus Bridge",
-            0x08 => "RACEway Bridge",
-            0x09 => "PCI-to-PCI Bridge (Semi-Transparent)",
-            0x0A => "InfiniBand-to-PCI Host Bridge",
-            else => defaultSubclass(subclass),
-        },
-        0x07 => switch (subclass) {
-            0x00 => "Serial Controller",
-            0x01 => "Parallel Controller",
-            0x02 => "Multiport Serial Controller",
-            0x03 => "Modem",
-            0x04 => "IEEE 488.1/2 (GPIB) Controller",
-            0x05 => "Smart Card Controller",
-            else => defaultSubclass(subclass),
-        },
-        0x08 => switch (subclass) {
-            0x00 => "PIC",
-            0x01 => "DMA Controller",
-            0x02 => "Timer",
-            0x03 => "RTC Controller",
-            0x04 => "PCI Hot-Plug Controller",
-            0x05 => "SD Host Controller",
-            0x06 => "IOMMU",
-            else => defaultSubclass(subclass),
-        },
-        0x09 => switch (subclass) {
-            0x00 => "Keyboard Controller",
-            0x01 => "Digitizer Pen",
-            0x02 => "Mouse Controller",
-            0x03 => "Scanner Controller",
-            0x04 => "Gameport Controller",
-            else => defaultSubclass(subclass),
-        },
-        0x0C => switch (subclass) {
-            0x00 => "FireWire (IEEE 1394) Controller",
-            0x01 => "ACCESS Bus Controller",
-            0x02 => "SSA",
-            0x03 => "USB Controller",
-            0x04 => "Fibre Channel",
-            0x05 => "SMBus Controller",
-            0x06 => "InfiniBand Controller",
-            0x07 => "IPMI Interface",
-            0x08 => "SERCOS Interface (IEC 61491)",
-            0x09 => "CANbus Controller",
-            else => defaultSubclass(subclass),
-        },
-        0x0D => switch (subclass) {
-            0x00 => "iRDA Compatible Controller",
-            0x01 => "Consumer IR Controller",
-            0x10 => "RF Controller",
-            0x11 => "Bluetooth Controller",
-            0x12 => "Broadband Controller",
-            0x20 => "Ethernet Controller (802.1a)",
-            0x21 => "Ethernet Controller (802.1b)",
-            else => defaultSubclass(subclass),
-        },
-        else => defaultSubclass(subclass),
+    const named: ?[]const u8 = switch (@as(BaseClass, @enumFromInt(base))) {
+        .mass_storage => enumName(mass_storage.SubClass, subclass),
+        .network => enumName(network.SubClass, subclass),
+        .display => enumName(display.SubClass, subclass),
+        .multimedia => enumName(multimedia.SubClass, subclass),
+        .memory => enumName(memory.SubClass, subclass),
+        .bridge => enumName(bridge.SubClass, subclass),
+        .simple_communication => enumName(simple_communication.SubClass, subclass),
+        .base_system_peripheral => enumName(base_system_peripheral.SubClass, subclass),
+        .input_device => enumName(input_device.SubClass, subclass),
+        .serial_bus => enumName(serial_bus.SubClass, subclass),
+        .wireless => enumName(wireless.SubClass, subclass),
+        else => null,
     };
+    return named orelse defaultSubclass(subclass);
 }
 
 fn defaultSubclass(subclass: u8) []const u8 {
@@ -175,68 +502,34 @@ fn defaultSubclass(subclass: u8) []const u8 {
 /// Returns "" when the prog-IF carries no standard meaning for this class/subclass —
 /// callers just print the hex byte in that case.
 pub fn progIfName(base: u8, subclass: u8, prog_if: u8) []const u8 {
-    return switch (base) {
-        0x01 => switch (subclass) {
-            0x06 => switch (prog_if) { // Serial ATA
-                0x00 => "Vendor Specific Interface",
-                0x01 => "AHCI 1.0",
-                0x02 => "Serial Storage Bus",
-                else => "",
+    const named: ?[]const u8 = switch (@as(BaseClass, @enumFromInt(base))) {
+        .mass_storage => switch (std.enums.fromInt(mass_storage.SubClass, subclass) orelse return "") {
+            .serial_ata => enumName(mass_storage.serial_ata.ProgIf, prog_if),
+            .non_volatile_memory => enumName(mass_storage.non_volatile_memory.ProgIf, prog_if),
+            else => null,
         },
-            0x08 => switch (prog_if) { // Non-Volatile Memory
-                0x01 => "NVMHCI",
-                0x02 => "NVM Express",
-                else => "",
+        .display => switch (std.enums.fromInt(display.SubClass, subclass) orelse return "") {
+            .vga_compatible => enumName(display.vga_compatible.ProgIf, prog_if),
+            else => null,
         },
-            else => "",
+        .bridge => switch (std.enums.fromInt(bridge.SubClass, subclass) orelse return "") {
+            .pci_to_pci => enumName(bridge.pci_to_pci.ProgIf, prog_if),
+            else => null,
         },
-        0x03 => switch (subclass) {
-            0x00 => switch (prog_if) { // VGA Compatible
-                0x00 => "VGA Controller",
-                0x01 => "8514-Compatible Controller",
-                else => "",
+        .simple_communication => switch (std.enums.fromInt(simple_communication.SubClass, subclass) orelse return "") {
+            .serial => enumName(simple_communication.serial.ProgIf, prog_if),
+            else => null,
         },
-            else => "",
+        .serial_bus => switch (std.enums.fromInt(serial_bus.SubClass, subclass) orelse return "") {
+            .usb => enumName(serial_bus.usb.ProgIf, prog_if),
+            else => null,
         },
-        0x06 => switch (subclass) {
-            0x04 => switch (prog_if) { // PCI-to-PCI Bridge
-                0x00 => "Normal Decode",
-                0x01 => "Subtractive Decode",
-                else => "",
-            },
-            else => "",
-        },
-        0x07 => switch (subclass) {
-            0x00 => switch (prog_if) { // Serial Controller
-                0x00 => "8250-Compatible (Generic XT)",
-                0x01 => "16450-Compatible",
-                0x02 => "16550-Compatible",
-                0x03 => "16650-Compatible",
-                0x04 => "16750-Compatible",
-                0x05 => "16850-Compatible",
-                0x06 => "16950-Compatible",
-                else => "",
-            },
-            else => "",
-        },
-        0x0C => switch (subclass) {
-            0x03 => switch (prog_if) { // USB Controller
-                0x00 => "UHCI Controller",
-                0x10 => "OHCI Controller",
-                0x20 => "EHCI (USB2) Controller",
-                0x30 => "XHCI (USB3) Controller",
-                0x80 => "Unspecified",
-                0xFE => "USB Device (not a host controller)",
-                else => "",
-            },
-            else => "",
-        },
-        else => "",
+        else => null,
     };
+    return named orelse "";
 }
 
 test "decodes the common class codes" {
-    const std = @import("std");
     const eq = std.testing.expectEqualStrings;
 
     const isa = ClassCode.unpack(0x06_01_00);
@@ -251,11 +544,25 @@ test "decodes the common class codes" {
     try eq("AHCI 1.0", progIfName(ahci.base, ahci.subclass, ahci.prog_if));
 
     const xhci = ClassCode.unpack(0x0C_03_30);
+    try eq("Serial Bus Controller", className(xhci.base));
     try eq("USB Controller", subclassName(xhci.base, xhci.subclass));
     try eq("XHCI (USB3) Controller", progIfName(xhci.base, xhci.subclass, xhci.prog_if));
+}
 
-    // Unknowns and the "Other" convention.
-    try eq("Other", subclassName(0x02, 0x80));
-    try eq("Unknown", subclassName(0x06, 0x7E));
-    try eq("", progIfName(0x06, 0x00, 0x00)); // host bridge: prog-IF has no standard name
+test "unlisted codes fall back without a wrong name" {
+    const eq = std.testing.expectEqualStrings;
+    try eq("Unknown", className(0x77)); // no such base class
+    try eq("Other", subclassName(0x01, 0x80)); // 0x80 is the PCI "Other" convention
+    try eq("Unknown", subclassName(0x01, 0x7A)); // unlisted mass-storage subclass
+    try eq("", progIfName(0x01, 0x06, 0x7F)); // no standard SATA prog-IF for 0x7F
+    try eq("", progIfName(0x02, 0x00, 0x00)); // class with no prog-IF taxonomy at all
+}
+
+test "named parts pack to the raw triple" {
+    const xhci = ClassCode{
+        .base = @intFromEnum(BaseClass.serial_bus),
+        .subclass = @intFromEnum(serial_bus.SubClass.usb),
+        .prog_if = @intFromEnum(serial_bus.usb.ProgIf.xhci),
+    };
+    try std.testing.expectEqual(@as(u24, 0x0C_03_30), xhci.pack());
 }

system/devices/platform.zig

@@ -40,6 +40,13 @@ pub fn platformInformation() PlatformInformation {
 }
 
 /// AML parse integrity/diagnostics (namespace node count, bytes consumed).
+/// The number of Device objects in the kernel's own AML namespace, or 0 if the
+/// parse produced none — the `acpi-parse` test compares the ring-3 service's
+/// count against this.
+pub fn amlDeviceCount() usize {
+    return acpi.amlDeviceCount();
+}
+
 pub fn amlStats() AmlStats {
     return acpi.aml_stats;
 }
@@ -72,7 +79,8 @@ pub fn discover(
     var device_tree = try DeviceTree.init(allocator);
 
     if (boot_information.acpi_rsdp != 0) {
-        try acpi.discover(boot_information.acpi_rsdp, &device_tree, hal);
+        const memory_regions = @as([*]const boot_handoff.MemoryRegion, @ptrFromInt(boot_handoff.physicalToVirtual(boot_information.memory_map.regions)))[0..boot_information.memory_map.len];
+        try acpi.discover(boot_information.acpi_rsdp, memory_regions, &device_tree, hal);
     } else {
         // No ACPI RSDP. A device-tree boot would parse its blob here; today that
         // path is a stub, so this reports the machine described itself no way we

system/devices/usb-abi.zig

@@ -0,0 +1,945 @@
+//! USB device-framework wire ABI: the set-up packets, standard requests, and standard
+//! descriptors every USB device speaks over its default control pipe, as defined by chapter 9
+//! of the USB 2.0 specification (see https://wiki.osdev.org/Universal_Serial_Bus). Pure data
+//! definitions — no hardware access — shared by the host-controller bus drivers (which build
+//! the requests) and anything that parses what devices return (device naming, driver
+//! matching, configuration). The structs mirror the wire byte-for-byte: multi-byte fields are
+//! little-endian and align(1), so a descriptor can be bit-cast straight out of a transfer
+//! buffer at any offset, and bitmap bytes are packed structs so no caller ever needs a magic
+//! mask. Class, subclass, and protocol code tables live in usb-ids.zig.
+
+pub const DeviceState = enum(u8) {
+    // Immediately after the USB device is attached to the USB system, it is in this state.
+    // The USB specifications do not define the state of a USB device that is detached from
+    // a USB system.
+    attached,
+    // A device is in this state after it has both been attached to the bus, and the VBUS line is
+    // applied to the device (the host controller drives the VBUS at +5V, however this is only
+    // particularly important for hardware developers). In this state, the device must not respond
+    // to any bus transactions. The USB specification recognizes three potential scenarios with
+    // respect to how a device draws power:
+    //    - Self-Powered Devices draw power from an external power source (e.g, a USB printer plugs
+    //      into the wall as well as a USB port). Although the device may be considered
+    //      technically "powered" even before attachment to the USB, it is still only considered
+    //      powered after the VBUS line is applied to the device.
+    //    - Bus-Powered Devices draw power solely from the USB up to 100mA.
+    //    - Self- or Bus-Powered Devices may draw power from either the bus or an external power
+    //      source, depending on the configuration. These devices may change power source at any
+    //      time. If a device is currently self-powered and requires more than 100mA of power, but
+    //      switches to being bus-powered, then the device must return to the Address state.
+    powered,
+    // A device in the powered state enters the default state after receiving a bus reset. In this
+    // state, the device is addressable at the default, reserved address of 0. At this point, the
+    // device is operating at the correct speed. The host is expected to allow 10 milliseconds
+    // before expecting the device to respond to data transfers after reset.
+    default,
+    // A device enters this state after the host assigns it an address via the default control pipe,
+    // which is always accessible whether the device's address has been set or not.
+    address,
+    // A device is in this state after the host examines its possible configurations and selects
+    // one. All endpoint's data toggle bits are initialized to zero when a device enters this state.
+    configured,
+    // When no traffic is observed on the bus for a period of 1 millisecond, a USB device enters
+    // this state, characterized by its low power consumption. The device's address and
+    // configuration settings are maintained while suspended. A device exits the suspended state as
+    // soon as it begins seeing bus activity again. The host is expected to allow 10 milliseconds
+    // before expecting the device to respond to data transfers after resume.
+    suspended,
+};
+
+pub const RequestCode = enum(u8) {
+    get_status = 0,
+    clear_feature = 1,
+    set_feature = 3,
+    set_address = 5,
+    get_descriptor = 6,
+    set_descriptor = 7,
+    get_configuration = 8,
+    set_configuration = 9,
+    get_interface = 10,
+    set_interface = 11,
+    sync_frame = 12,
+    // Non-exhaustive: class-specific requests (HID, mass storage) reuse this byte
+    // field with codes from their own class's namespace — see the class-request
+    // constructors below. Some class codes numerically coincide with a standard
+    // one; the wire byte is what matters, and the constructors set it explicitly.
+    _,
+};
+
+// Direction of an endpoint, from the host's point of view
+pub const EndpointDirection = enum(u1) {
+    out = 0,
+    in = 1,
+};
+
+// Identifier newtypes: distinct wire-sized types for values that identify something on the
+// device rather than count something. Each is a non-exhaustive enum whose values originate
+// in the descriptors below and flow, still typed, into the standard request constructors —
+// so an interface number can never be passed where a configuration value is expected.
+
+// The bus address of a device, assigned by the host with SET_ADDRESS. Addresses are 7 bits
+// wide.
+pub const DeviceAddress = enum(u7) {
+    // The default address every device answers at after a reset, until SET_ADDRESS
+    // completes
+    default = 0,
+    _,
+};
+
+// Identifies a configuration; from ConfigurationDescriptor.configuration_value.
+pub const ConfigurationValue = enum(u8) {
+    // Not configured: returned by GET_CONFIGURATION while the device is in the address
+    // state, and passed to SET_CONFIGURATION to return a configured device to the address
+    // state
+    none = 0,
+    _,
+};
+
+// Identifies an interface within a configuration; from
+// InterfaceDescriptor.interface_number.
+pub const InterfaceNumber = enum(u8) { _ };
+
+// Selects between the alternate settings of one interface; from
+// InterfaceDescriptor.alternate_setting.
+pub const AlternateSetting = enum(u8) {
+    // The default setting of an interface
+    default = 0,
+    _,
+};
+
+// The number of an endpoint within a device, 4 bits wide. The direction bit carried
+// alongside it tells the two endpoints sharing a number apart.
+pub const EndpointNumber = enum(u4) {
+    // Endpoint zero: the default control pipe every device provides
+    default_control = 0,
+    _,
+};
+
+// Index of a STRING descriptor, stored in descriptors that reference a string and passed to
+// GET_DESCRIPTOR to read it.
+pub const StringIndex = enum(u8) {
+    // The device has no string descriptor for this field
+    none = 0,
+    _,
+};
+
+// Characteristics of a device request (the bmRequestType field of a set-up packet). Fields are
+// declared least-significant first: recipient occupies bits 4...0, kind bits 6...5, and
+// direction bit 7.
+pub const RequestType = packed struct(u8) {
+    // The recipient of the request (values 4...31 are reserved)
+    recipient: Recipient,
+    // The type of the request
+    kind: Kind,
+    // Data transfer direction. The value of this bit is ignored when length is zero.
+    direction: Direction,
+
+    pub const Recipient = enum(u5) {
+        device = 0,
+        interface = 1,
+        endpoint = 2,
+        other = 3,
+    };
+
+    pub const Kind = enum(u2) {
+        standard = 0,
+        class = 1,
+        vendor = 2,
+        reserved = 3,
+    };
+
+    pub const Direction = enum(u1) {
+        host_to_device = 0,
+        device_to_host = 1,
+    };
+};
+
+pub const Request = extern struct {
+    // Characteristics of the request
+    request_type: RequestType,
+    // Specific request
+    request_code: RequestCode,
+    // Word-sized field that may (or may not) serve as a parameter to the request, depending
+    // on the specific request. For GET_DESCRIPTOR and SET_DESCRIPTOR, bit-cast a
+    // DescriptorValue into this field.
+    value: u16 align(1),
+    // Word-sized field that may (or may not) serve as a parameter to the request, depending
+    // on the specific request. Typically this field holds an index or an offset value. When
+    // request_type specifies an endpoint or an interface as the recipient, bit-cast an
+    // EndpointIndex or an InterfaceIndex into this field.
+    index: u16 align(1),
+    // Number of bytes to transfer if there is a DATA stage.
+    //   - If this field is non-zero, and request_type indicates a transfer from
+    //     device-to-host, then the device must never return more than length bytes of data.
+    //     However, a device may return less.
+    //   - If this field is non-zero, and request_type indicates a transfer from
+    //     host-to-device, then the host must send exactly length bytes of data. If the host
+    //     sends more than length bytes, the behavior of the device is undefined.
+    length: u16 align(1),
+
+    // The format of the index field when request_type specifies an endpoint as the
+    // recipient. The host should always set the direction bit to zero (but the device
+    // should accept either value) when the endpoint is part of a control pipe.
+    pub const EndpointIndex = packed struct(u16) {
+        // Endpoint number
+        number: EndpointNumber,
+        // Reserved (reset to zero)
+        reserved: u3 = 0,
+        // Selects the OUT or the IN endpoint with the specified endpoint number
+        direction: EndpointDirection,
+        // Reserved (reset to zero)
+        reserved_high: u8 = 0,
+    };
+
+    // The format of the index field when request_type specifies an interface as the
+    // recipient.
+    pub const InterfaceIndex = packed struct(u16) {
+        // Interface number
+        number: u8,
+        // Reserved (reset to zero)
+        reserved: u8 = 0,
+    };
+
+    // The format of the value field of GET_DESCRIPTOR and SET_DESCRIPTOR requests: the
+    // descriptor type in the high byte, and the descriptor index in the low byte. The index
+    // is used to select a specific descriptor (only for CONFIGURATION and STRING
+    // descriptors) when several descriptors of that type are implemented by a device.
+    pub const DescriptorValue = packed struct(u16) {
+        // Descriptor index
+        index: u8 = 0,
+        // Descriptor type
+        kind: DescriptorType,
+    };
+};
+
+// Feature selectors, used as the value field of CLEAR_FEATURE and SET_FEATURE requests. The
+// comment on each value notes the recipient the selector applies to.
+pub const FeatureSelector = enum(u16) {
+    // Halts an endpoint (recipient: endpoint)
+    endpoint_halt = 0,
+    // Enables or disables the device's remote wakeup capability (recipient: device)
+    device_remote_wakeup = 1,
+    // Puts a hi-speed device into a test mode, selected by a TestMode value in the high
+    // byte of the index field (recipient: device)
+    test_mode = 2,
+};
+
+// Test mode selectors, passed in the high byte of the index field of a SET_FEATURE request
+// with the test_mode feature selector. Values 06h...3Fh are reserved for standard test
+// selectors and C0h...FFh for vendor-specific test modes; all other unlisted values are
+// reserved.
+pub const TestMode = enum(u8) {
+    test_j = 0x01,
+    test_k = 0x02,
+    test_se0_nak = 0x03,
+    test_packet = 0x04,
+    test_force_enable = 0x05,
+    _,
+};
+
+// The two bytes returned by a GET_STATUS request directed at a device. Fields are declared
+// least-significant first.
+pub const DeviceStatus = packed struct(u16) {
+    // Whether the device is currently self-powered (as opposed to bus-powered). This bit
+    // cannot be changed with the SET_FEATURE or CLEAR_FEATURE requests.
+    self_powered: bool,
+    // Whether the device is currently enabled to request remote wakeup. Changed with the
+    // SET_FEATURE and CLEAR_FEATURE requests using the device_remote_wakeup feature
+    // selector.
+    remote_wakeup: bool,
+    // Reserved (reset to zero)
+    reserved: u14,
+};
+
+// The two bytes returned by a GET_STATUS request directed at an endpoint. (A GET_STATUS
+// request directed at an interface returns two bytes that are entirely reserved.)
+pub const EndpointStatus = packed struct(u16) {
+    // Whether the endpoint is currently halted. Set with the SET_FEATURE request using the
+    // endpoint_halt feature selector, and cleared with CLEAR_FEATURE.
+    halted: bool,
+    // Reserved (reset to zero)
+    reserved: u15,
+};
+
+// A target for the standard requests that may be directed at the device, an interface, or
+// an endpoint.
+pub const Target = union(enum) {
+    device,
+    interface: InterfaceNumber,
+    endpoint: Request.EndpointIndex,
+
+    fn recipient(target: Target) RequestType.Recipient {
+        return switch (target) {
+            .device => .device,
+            .interface => .interface,
+            .endpoint => .endpoint,
+        };
+    }
+
+    fn index(target: Target) u16 {
+        return switch (target) {
+            .device => 0,
+            .interface => |number| @intFromEnum(number),
+            .endpoint => |endpoint| @bitCast(endpoint),
+        };
+    }
+};
+
+// Constructors for the standard device requests, one per RequestCode. Each returns a
+// ready-to-send set-up packet with the request_type, value, index, and length fields the
+// specification prescribes for that request.
+
+// Reads the status of the given target: bit-cast the two bytes the device returns into a
+// DeviceStatus or an EndpointStatus. (The two bytes returned for an interface are entirely
+// reserved.)
+pub fn getStatus(target: Target) Request {
+    return .{
+        .request_type = .{
+            .recipient = target.recipient(),
+            .kind = .standard,
+            .direction = .device_to_host,
+        },
+        .request_code = .get_status,
+        .value = 0,
+        .index = target.index(),
+        .length = 2,
+    };
+}
+
+// Clears or disables the given feature. A device cannot be taken out of a test mode with
+// this request; test_mode is only cleared by cycling power.
+pub fn clearFeature(feature: FeatureSelector, target: Target) Request {
+    return .{
+        .request_type = .{
+            .recipient = target.recipient(),
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .clear_feature,
+        .value = @intFromEnum(feature),
+        .index = target.index(),
+        .length = 0,
+    };
+}
+
+// Sets or enables the given feature. For the test_mode feature selector, use setTestMode
+// instead: the test selector rides in the high byte of the index field.
+pub fn setFeature(feature: FeatureSelector, target: Target) Request {
+    return .{
+        .request_type = .{
+            .recipient = target.recipient(),
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .set_feature,
+        .value = @intFromEnum(feature),
+        .index = target.index(),
+        .length = 0,
+    };
+}
+
+// Puts a hi-speed device into the given test mode: a SET_FEATURE request with the test_mode
+// feature selector and the test selector in the high byte of the index field.
+pub fn setTestMode(mode: TestMode) Request {
+    return .{
+        .request_type = .{
+            .recipient = .device,
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .set_feature,
+        .value = @intFromEnum(FeatureSelector.test_mode),
+        .index = @as(u16, @intFromEnum(mode)) << 8,
+        .length = 0,
+    };
+}
+
+// Assigns the device its bus address, moving it from the default state to the address
+// state. The device does not answer at the new address until the status stage of this
+// request completes.
+pub fn setAddress(address: DeviceAddress) Request {
+    return .{
+        .request_type = .{
+            .recipient = .device,
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .set_address,
+        .value = @intFromEnum(address),
+        .index = 0,
+        .length = 0,
+    };
+}
+
+// Reads a descriptor from the device.
+//   - descriptor_index selects among descriptors of the same type, and is only used for
+//     configuration and string descriptors.
+//   - language_id selects the language of a string descriptor, and is zero otherwise.
+//   - length is the number of bytes to read; a device never returns more than length bytes,
+//     but may return less if the descriptor is shorter.
+pub fn getDescriptor(kind: DescriptorType, descriptor_index: u8, language_id: u16, length: u16) Request {
+    return .{
+        .request_type = .{
+            .recipient = .device,
+            .kind = .standard,
+            .direction = .device_to_host,
+        },
+        .request_code = .get_descriptor,
+        .value = @bitCast(Request.DescriptorValue{ .index = descriptor_index, .kind = kind }),
+        .index = language_id,
+        .length = length,
+    };
+}
+
+// Updates an existing descriptor or adds a new one (optional; many devices do not support
+// this request). The parameters mirror getDescriptor; the descriptor itself is sent in the
+// DATA stage.
+pub fn setDescriptor(kind: DescriptorType, descriptor_index: u8, language_id: u16, length: u16) Request {
+    return .{
+        .request_type = .{
+            .recipient = .device,
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .set_descriptor,
+        .value = @bitCast(Request.DescriptorValue{ .index = descriptor_index, .kind = kind }),
+        .index = language_id,
+        .length = length,
+    };
+}
+
+// Reads the currently active configuration: @enumFromInt the byte the device returns into a
+// ConfigurationValue, which is none while the device is not configured.
+pub fn getConfiguration() Request {
+    return .{
+        .request_type = .{
+            .recipient = .device,
+            .kind = .standard,
+            .direction = .device_to_host,
+        },
+        .request_code = .get_configuration,
+        .value = 0,
+        .index = 0,
+        .length = 1,
+    };
+}
+
+// Selects the configuration with the given configuration_value (from
+// ConfigurationDescriptor.configuration_value), moving the device from the address state to
+// the configured state. Selecting none returns the device to the address state.
+pub fn setConfiguration(configuration_value: ConfigurationValue) Request {
+    return .{
+        .request_type = .{
+            .recipient = .device,
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .set_configuration,
+        .value = @intFromEnum(configuration_value),
+        .index = 0,
+        .length = 0,
+    };
+}
+
+// Reads the alternate setting currently selected for the given interface: @enumFromInt the
+// byte the device returns into an AlternateSetting.
+pub fn getInterface(interface: InterfaceNumber) Request {
+    return .{
+        .request_type = .{
+            .recipient = .interface,
+            .kind = .standard,
+            .direction = .device_to_host,
+        },
+        .request_code = .get_interface,
+        .value = 0,
+        .index = @intFromEnum(interface),
+        .length = 1,
+    };
+}
+
+// Selects an alternate setting (from InterfaceDescriptor.alternate_setting) for the given
+// interface.
+pub fn setInterface(interface: InterfaceNumber, alternate_setting: AlternateSetting) Request {
+    return .{
+        .request_type = .{
+            .recipient = .interface,
+            .kind = .standard,
+            .direction = .host_to_device,
+        },
+        .request_code = .set_interface,
+        .value = @intFromEnum(alternate_setting),
+        .index = @intFromEnum(interface),
+        .length = 0,
+    };
+}
+
+// Reads the two-byte number of the frame in which the given isochronous endpoint's
+// repeating pattern of transfers begins.
+pub fn syncFrame(endpoint: Request.EndpointIndex) Request {
+    return .{
+        .request_type = .{
+            .recipient = .endpoint,
+            .kind = .standard,
+            .direction = .device_to_host,
+        },
+        .request_code = .sync_frame,
+        .value = 0,
+        .index = @bitCast(endpoint),
+        .length = 2,
+    };
+}
+
+// Class-specific requests. These carry a `kind = .class` request_type and a
+// request_code from the interface's class namespace (not the standard
+// RequestCode set above); the code is written into the same byte field, which
+// is why RequestCode is non-exhaustive. Each is directed at an interface, whose
+// number rides in the index field.
+
+// The HID class request codes (USB HID 1.11 §7.2). Only the ones danos issues
+// are named; the field on the wire is the raw byte.
+pub const HidRequestCode = enum(u8) {
+    get_report = 0x01,
+    get_idle = 0x02,
+    get_protocol = 0x03,
+    set_report = 0x09,
+    set_idle = 0x0A,
+    set_protocol = 0x0B,
+};
+
+// The two protocols a boot-capable HID device can run (USB HID 1.11 §7.2.5).
+// A driver selects `boot` for the simplified fixed-format boot report, usable
+// before a full report-descriptor parser exists.
+pub const HidProtocol = enum(u8) {
+    boot = 0,
+    report = 1,
+};
+
+// SET_PROTOCOL: choose the boot or report protocol on a HID interface.
+pub fn setProtocol(interface: InterfaceNumber, protocol: HidProtocol) Request {
+    return .{
+        .request_type = .{ .recipient = .interface, .kind = .class, .direction = .host_to_device },
+        .request_code = @enumFromInt(@intFromEnum(HidRequestCode.set_protocol)),
+        .value = @intFromEnum(protocol),
+        .index = @intFromEnum(interface),
+        .length = 0,
+    };
+}
+
+// SET_IDLE: bound a HID interface's report rate. `duration` is in 4 ms units
+// (0 means report only on change); `report_id` selects a report (0 = all).
+pub fn setIdle(interface: InterfaceNumber, duration: u8, report_id: u8) Request {
+    return .{
+        .request_type = .{ .recipient = .interface, .kind = .class, .direction = .host_to_device },
+        .request_code = @enumFromInt(@intFromEnum(HidRequestCode.set_idle)),
+        .value = (@as(u16, duration) << 8) | report_id,
+        .index = @intFromEnum(interface),
+        .length = 0,
+    };
+}
+
+// Bulk-Only Mass Storage Reset (USB MSC BOT §3.1): ready a mass-storage
+// interface for the next Command Block Wrapper after a protocol error.
+pub fn bulkOnlyMassStorageReset(interface: InterfaceNumber) Request {
+    return .{
+        .request_type = .{ .recipient = .interface, .kind = .class, .direction = .host_to_device },
+        .request_code = @enumFromInt(0xFF),
+        .value = 0,
+        .index = @intFromEnum(interface),
+        .length = 0,
+    };
+}
+
+// Get Max LUN (USB MSC BOT §3.2): read the highest logical unit number the
+// device supports (0 for a single-LUN flash drive). One byte is returned.
+pub fn getMaxLun(interface: InterfaceNumber) Request {
+    return .{
+        .request_type = .{ .recipient = .interface, .kind = .class, .direction = .device_to_host },
+        .request_code = @enumFromInt(0xFE),
+        .value = 0,
+        .index = @intFromEnum(interface),
+        .length = 1,
+    };
+}
+
+pub const DescriptorType = enum(u8) {
+    device = 1,
+    configuration = 2,
+    string = 3,
+    interface = 4,
+    endpoint = 5,
+    device_qualifier = 6,
+    other_speed_configuration = 7,
+    interface_power = 8,
+    _,
+};
+
+pub const DeviceDescriptor = extern struct {
+    // Size of this descriptor in bytes
+    length: u8,
+    // DEVICE Descriptor Type
+    descriptor_type: DescriptorType,
+    // USB Specification Release Number in Binary-Coded Decimal (i.e, 2.10 is expressed as 210h).
+    // Identifies the release of the USB Specification with with the device and its
+    // descriptors are compliant.
+    bcd_usb: u16 align(1),
+    // Class code (assigned by the USB-IF)
+    //   - This field is reset to zero if each interface within a configuration specifies its own
+    //     class information and the various interfaces operate independently.
+    //   - A value of FFh in this field indicates the device class is vendor-specific.
+    device_class: u8,
+    // Subclass Code (assigned by the USB-IF)
+    //   - The subclass code of a device is qualified by the class code of that device.
+    //   - If device_class is reset to zero, then this field must also be reset to zero.
+    //   - When device_class is not set to FFh, then all values for this field are reserved for
+    //     assignment by the USB-IF.
+    device_subclass: u8,
+    // Protocol code (assigned by the USB-IF)
+    //   - The protocol code of a device is qualified by both the class and subclass codes of
+    //     that device.
+    //   - A value of 00h in this field means that the device may specify class-specific
+    //     protocols on an interface basis, though this is not a requirement.
+    //   - If this field is set to FFh, then the device uses a vendor-specific protocol.
+    device_protocol: u8,
+    // Maximum packet size for endpoint zero (8, 16, 32, or 64 are the only valid options)
+    max_packet_size_0: u8,
+    // Vendor ID (assigned by the USB-IF)
+    vendor_id: u16 align(1),
+    // Product ID (assigned by the USB-IF)
+    product_id: u16 align(1),
+    // Device release number in binary-coded decimal
+    bcd_device: u16 align(1),
+    // Index of STRING descriptor describing manufacturer
+    manufacturer_index: StringIndex,
+    // Index of STRING descriptor describing product
+    product_index: StringIndex,
+    // Index of STRING descriptor describing the device's serial number
+    serial_number_index: StringIndex,
+    // Number of possible configurations
+    configuration_count: u8,
+};
+
+pub const DeviceQualifierDescriptor = extern struct {
+    // Size of this descriptor in bytes
+    length: u8,
+    // DEVICE_QUALIFIER Descriptor Type
+    descriptor_type: DescriptorType,
+    // USB Specification Release Number in Binary-Coded Decimal (i.e, 2.00 is expressed as 200h).
+    // Identifies the release of the USB Specification with with the device and its
+    // descriptors are compliant. This field must be at least 0200h.
+    bcd_usb: u16 align(1),
+    // Class code (assigned by the USB-IF)
+    device_class: u8,
+    // Subclass Code (assigned by the USB-IF)
+    device_subclass: u8,
+    // Protocol code (assigned by the USB-IF)
+    device_protocol: u8,
+    // Maximum packet size for endpoint zero (8, 16, 32, or 64 are the only valid options)
+    max_packet_size_0: u8,
+    // Number of possible configurations
+    configuration_count: u8,
+    // Reserved for future uses, must be zero.
+    reserved: u8,
+};
+
+pub const ConfigurationDescriptor = extern struct {
+    // Size of this descriptor in bytes
+    length: u8,
+    // CONFIGURATION Descriptor Type
+    descriptor_type: DescriptorType,
+    // The total combined length in bytes of all the descriptors returned with the request for
+    // this CONFIGURATION descriptor (including CONFIGURATION, INTERFACE, ENDPOINT, class- and
+    // vendor-specific descriptors).
+    total_length: u16 align(1),
+    // Number of interfaces supported by this configuration
+    interface_count: u8,
+    // Value which when used as an argument in the SET_CONFIGURATION request, causes the device
+    // to assume the configuration described by this descriptor.
+    configuration_value: ConfigurationValue,
+    // Index of STRING descriptor describing this configuration.
+    configuration_index: StringIndex,
+    // Configuration Characteristics
+    attributes: Attributes,
+    // Maximum power consumption of this device from the bus when fully operational and using
+    // this configuration. Expressed in units of 2mA (i.e., a value of 50 in this field
+    // indicates 100mA).
+    //   - A device reports with the attributes field whether the configuration is bus- or
+    //     self-powered, but the device status (retrieved with a GET_STATUS request) reports
+    //     whether the device is currently self-powered.
+    //   - If a device is disconnected from an external power source, it may not draw more
+    //     power from the bus than specified in this field.
+    max_power: u8,
+
+    // Configuration characteristics. Fields are declared least-significant first.
+    pub const Attributes = packed struct(u8) {
+        // Reserved, reset to zero (D4...0)
+        reserved: u5,
+        // Whether Remote Wakeup is supported by this configuration (D5)
+        remote_wakeup: bool,
+        // Self-Powered (D6)
+        //   - false: Device runs on power supplied by the bus
+        //   - true: Device provides a local power source; if max_power is non-zero, the
+        //     device also may use bus power.
+        self_powered: bool,
+        // Reserved, must be set to one for historical reasons (D7)
+        reserved_one: u1,
+    };
+};
+
+// This descriptor describes the configuration of a high-speed device if it were operating at
+// its alternative speed. The structure of the OTHER_SPEED_CONFIGURATION is identical to that
+// of the CONFIGURATION descriptor; the only difference is that the descriptor_type field
+// reflects that the descriptor is an OTHER_SPEED_CONFIGURATION descriptor.
+pub const OtherSpeedConfigurationDescriptor = ConfigurationDescriptor;
+
+pub const InterfaceDescriptor = extern struct {
+    // Size of this descriptor in bytes
+    length: u8,
+    // INTERFACE Descriptor Type
+    descriptor_type: DescriptorType,
+    // Number of this interface. Zero-based value which identifies the index of this interface
+    // in the array of interfaces supported within a configuration.
+    interface_number: InterfaceNumber,
+    // Value used to select the alternate settings described by this INTERFACE descriptor for
+    // the interface with the interface_number in the previous field. This value is zero if
+    // this descriptor describes the default settings for a particular interface.
+    alternate_setting: AlternateSetting,
+    // Number of endpoints used by this interface, not including endpoint zero.
+    endpoint_count: u8,
+    // Class code (assigned by the USB-IF)
+    //   - A value of zero here is reserved for future standardization.
+    //   - If this value is FFh, the interface class is vendor-specific.
+    //   - All other values are reserved for assignment by the USB-IF.
+    interface_class: u8,
+    // Subclass code (assigned by the USB-IF)
+    //   - The subclass code in this field is qualified by the value of the interface_class
+    //     field.
+    //   - If interface_class is reset to zero, then this field must also be reset to zero.
+    //   - If interface_class is not set to the value of FFh, then all values of this field are
+    //     reserved for assignment by the USB-IF.
+    interface_subclass: u8,
+    // Protocol code (assigned by the USB-IF)
+    //   - The protocol code in this field is qualified by the values of the interface_class
+    //     and interface_subclass fields.
+    //   - If an interface supports class-specific requests, then this field identifies the
+    //     protocols that the device uses as defined by the specifications of the device class.
+    //   - If this field is reset to zero, then the device does not use a class-specific
+    //     protocol on this interface.
+    //   - If this field is set to FFh, then the device uses a vendor-specific protocol on
+    //     this interface.
+    interface_protocol: u8,
+    // Index of STRING descriptor describing this interface
+    interface_index: StringIndex,
+};
+
+pub const EndpointDescriptor = extern struct {
+    // Size of this descriptor in bytes
+    length: u8,
+    // ENDPOINT Descriptor Type
+    descriptor_type: DescriptorType,
+    // The address of the endpoint on the USB device described by this descriptor
+    endpoint_address: Address,
+    // The endpoint's attributes
+    attributes: Attributes,
+    // Maximum packet size that this endpoint is capable of sending or receiving. For
+    // isochronous endpoints, this value is used to reserve bus time; the pipe, however, may
+    // not always use all of the reserved bus time.
+    max_packet_size: MaxPacketSize align(1),
+    // Interval for polling a device during a data transfer, expressed in units of microframes
+    // for high-speed devices, and frames for low- and full-speed devices. The exact meaning of
+    // the value in this field depends on the endpoint type and the operating speed of the
+    // device:
+    //   - Full- and High-speed isochronous endpoints, and high-speed interrupt endpoints:
+    //     This field must be in the range from 1 to 16, and is used to calculate the period
+    //     as 2^(interval - 1). That is, a value of 4 calculates to 2^(4 - 1) = 2^3 = 8.
+    //   - Full- and Low-speed interrupt endpoints: This field must be in the range from
+    //     1 to 255.
+    //   - High-speed bulk and control OUT endpoints: This field must be in the range from
+    //     0 to 255, and specifies the maximum NAK rate of the endpoint. A value of zero
+    //     indicates that the endpoint never NAKs; other values indicate at most 1 NAK each
+    //     interval number of microframes.
+    interval: u8,
+
+    // The address of an endpoint. Fields are declared least-significant first.
+    pub const Address = packed struct(u8) {
+        // Endpoint Number (D3...0)
+        number: EndpointNumber,
+        // Reserved, reset to zero (D6...4)
+        reserved: u3,
+        // Direction, ignored for control endpoints (D7)
+        direction: EndpointDirection,
+    };
+
+    // An endpoint's attributes. Fields are declared least-significant first.
+    pub const Attributes = packed struct(u8) {
+        // Transfer Type (D1...0)
+        transfer_type: TransferType,
+        // Synchronization Type; isochronous endpoints only, reserved and reset to zero for
+        // other endpoint types (D3...2)
+        synchronization: Synchronization,
+        // Usage Type; isochronous endpoints only, reserved and reset to zero for other
+        // endpoints (D5...4)
+        usage: Usage,
+        // Reserved, reset to zero (D7...6)
+        reserved: u2,
+    };
+
+    pub const TransferType = enum(u2) {
+        control = 0,
+        isochronous = 1,
+        bulk = 2,
+        interrupt = 3,
+    };
+
+    pub const Synchronization = enum(u2) {
+        none = 0,
+        asynchronous = 1,
+        adaptive = 2,
+        synchronous = 3,
+    };
+
+    pub const Usage = enum(u2) {
+        data = 0,
+        feedback = 1,
+        implicit_feedback_data = 2,
+        _,
+    };
+
+    // The maximum packet size of an endpoint. Fields are declared least-significant first.
+    pub const MaxPacketSize = packed struct(u16) {
+        // Maximum packet size in bytes (bits 10...0)
+        size: u11,
+        // Number of additional transaction opportunities per microframe, for high-speed
+        // isochronous and interrupt endpoints; reserved and reset to zero for other
+        // endpoints (bits 12...11)
+        additional_transactions: AdditionalTransactions,
+        // Reserved, must be reset to zero (bits 15...13)
+        reserved: u3,
+    };
+
+    pub const AdditionalTransactions = enum(u2) {
+        // None (1 transaction per microframe)
+        none = 0,
+        // 1 additional (2 transactions per microframe)
+        one = 1,
+        // 2 additional (3 transactions per microframe)
+        two = 2,
+        _,
+    };
+};
+
+// A STRING descriptor at index zero returns the list of LANGID codes supported by the
+// device; all other indices return a Unicode string. Both forms start with this two-byte
+// header, followed by the variable-length payload:
+//   - index 0: an array of two-byte LANGID codes (wLangID[0] through wLangID[x])
+//   - other indices: a Unicode string of N bytes
+pub const StringDescriptor = extern struct {
+    // Size of this descriptor in bytes
+    length: u8,
+    // STRING Descriptor Type
+    descriptor_type: DescriptorType,
+};
+
+const std = @import("std");
+
+test "wire sizes and offsets match the specification" {
+    const expectEqual = std.testing.expectEqual;
+
+    try expectEqual(8, @sizeOf(Request));
+    try expectEqual(18, @sizeOf(DeviceDescriptor));
+    try expectEqual(10, @sizeOf(DeviceQualifierDescriptor));
+    try expectEqual(9, @sizeOf(ConfigurationDescriptor));
+    try expectEqual(9, @sizeOf(InterfaceDescriptor));
+    try expectEqual(7, @sizeOf(EndpointDescriptor));
+    try expectEqual(2, @sizeOf(StringDescriptor));
+
+    try expectEqual(2, @offsetOf(DeviceDescriptor, "bcd_usb"));
+    try expectEqual(8, @offsetOf(DeviceDescriptor, "vendor_id"));
+    try expectEqual(17, @offsetOf(DeviceDescriptor, "configuration_count"));
+    try expectEqual(2, @offsetOf(ConfigurationDescriptor, "total_length"));
+    try expectEqual(4, @offsetOf(EndpointDescriptor, "max_packet_size"));
+}
+
+test "bitmap packings match the specification" {
+    const expectEqual = std.testing.expectEqual;
+    const expect = std.testing.expect;
+
+    // bmRequestType for GET_DESCRIPTOR: device-to-host | standard | device = 80h
+    const request_type = RequestType{
+        .recipient = .device,
+        .kind = .standard,
+        .direction = .device_to_host,
+    };
+    try expectEqual(0x80, @as(u8, @bitCast(request_type)));
+
+    // wValue for GET_DESCRIPTOR(CONFIGURATION, index 0) = 0200h
+    const descriptor_value = Request.DescriptorValue{ .kind = .configuration };
+    try expectEqual(0x0200, @as(u16, @bitCast(descriptor_value)));
+
+    // wIndex for the IN endpoint 1 = 0081h
+    const endpoint_index = Request.EndpointIndex{ .number = @enumFromInt(1), .direction = .in };
+    try expectEqual(0x0081, @as(u16, @bitCast(endpoint_index)));
+
+    // Endpoint address 81h = IN endpoint 1
+    const address: EndpointDescriptor.Address = @bitCast(@as(u8, 0x81));
+    try expectEqual(1, @intFromEnum(address.number));
+    try expectEqual(.in, address.direction);
+
+    // Endpoint attributes 03h = interrupt transfer
+    const attributes: EndpointDescriptor.Attributes = @bitCast(@as(u8, 0x03));
+    try expectEqual(.interrupt, attributes.transfer_type);
+
+    // wMaxPacketSize 0008h = 8 bytes, no additional transactions
+    const max_packet_size: EndpointDescriptor.MaxPacketSize = @bitCast(@as(u16, 0x0008));
+    try expectEqual(8, max_packet_size.size);
+    try expectEqual(.none, max_packet_size.additional_transactions);
+
+    // Configuration attributes C0h = self-powered, with the historical D7 bit set
+    const configuration_attributes: ConfigurationDescriptor.Attributes = @bitCast(@as(u8, 0xC0));
+    try expect(configuration_attributes.self_powered);
+    try expect(!configuration_attributes.remote_wakeup);
+    try expectEqual(1, configuration_attributes.reserved_one);
+
+    // GET_STATUS words: device 0001h = self-powered; endpoint 0001h = halted
+    const device_status: DeviceStatus = @bitCast(@as(u16, 0x0001));
+    try expect(device_status.self_powered and !device_status.remote_wakeup);
+    const endpoint_status: EndpointStatus = @bitCast(@as(u16, 0x0001));
+    try expect(endpoint_status.halted);
+
+    // DescriptorType is non-exhaustive: class-specific values (HID = 21h) pass through
+    const hid_type: DescriptorType = @enumFromInt(0x21);
+    try expectEqual(0x21, @intFromEnum(hid_type));
+    try expect(hid_type != .device);
+}
+
+pub fn expectRequestBytes(request: Request, expected: [8]u8) !void {
+    try std.testing.expectEqualSlices(u8, &expected, std.mem.asBytes(&request));
+}
+
+test "standard request constructors encode the specification's set-up packets" {
+    try expectRequestBytes(getStatus(.device), .{ 0x80, 0, 0, 0, 0, 0, 2, 0 });
+    try expectRequestBytes(getStatus(.{ .interface = @enumFromInt(3) }), .{ 0x81, 0, 0, 0, 3, 0, 2, 0 });
+    try expectRequestBytes(getStatus(.{ .endpoint = .{ .number = @enumFromInt(2), .direction = .in } }), .{ 0x82, 0, 0, 0, 0x82, 0, 2, 0 });
+    try expectRequestBytes(clearFeature(.endpoint_halt, .{ .endpoint = .{ .number = @enumFromInt(1), .direction = .out } }), .{ 0x02, 1, 0, 0, 0x01, 0, 0, 0 });
+    try expectRequestBytes(setFeature(.device_remote_wakeup, .device), .{ 0x00, 3, 1, 0, 0, 0, 0, 0 });
+    try expectRequestBytes(setTestMode(.test_packet), .{ 0x00, 3, 2, 0, 0, 0x04, 0, 0 });
+    try expectRequestBytes(setAddress(@enumFromInt(5)), .{ 0x00, 5, 5, 0, 0, 0, 0, 0 });
+    try expectRequestBytes(getDescriptor(.device, 0, 0, 18), .{ 0x80, 6, 0, 1, 0, 0, 18, 0 });
+    try expectRequestBytes(getDescriptor(.string, 2, 0x0409, 255), .{ 0x80, 6, 2, 3, 0x09, 0x04, 255, 0 });
+    try expectRequestBytes(setDescriptor(.string, 2, 0x0409, 16), .{ 0x00, 7, 2, 3, 0x09, 0x04, 16, 0 });
+    try expectRequestBytes(getConfiguration(), .{ 0x80, 8, 0, 0, 0, 0, 1, 0 });
+    try expectRequestBytes(setConfiguration(@enumFromInt(1)), .{ 0x00, 9, 1, 0, 0, 0, 0, 0 });
+    try expectRequestBytes(getInterface(@enumFromInt(2)), .{ 0x81, 10, 0, 0, 2, 0, 1, 0 });
+    try expectRequestBytes(setInterface(@enumFromInt(2), @enumFromInt(1)), .{ 0x01, 11, 1, 0, 2, 0, 0, 0 });
+    try expectRequestBytes(syncFrame(.{ .number = @enumFromInt(3), .direction = .in }), .{ 0x82, 12, 0, 0, 0x83, 0, 2, 0 });
+}
+
+test "class request constructors encode the specification's set-up packets" {
+    // bmRequestType for a host-to-device class request to an interface = 0x21;
+    // device-to-host = 0xA1. The request_code byte is the class code, not a
+    // standard one — SET_PROTOCOL 0x0B, SET_IDLE 0x0A, BOT reset 0xFF, Max LUN 0xFE.
+    try expectRequestBytes(setProtocol(@enumFromInt(0), .boot), .{ 0x21, 0x0B, 0, 0, 0, 0, 0, 0 });
+    try expectRequestBytes(setProtocol(@enumFromInt(1), .report), .{ 0x21, 0x0B, 1, 0, 1, 0, 0, 0 });
+    try expectRequestBytes(setIdle(@enumFromInt(1), 0, 0), .{ 0x21, 0x0A, 0, 0, 1, 0, 0, 0 });
+    try expectRequestBytes(bulkOnlyMassStorageReset(@enumFromInt(0)), .{ 0x21, 0xFF, 0, 0, 0, 0, 0, 0 });
+    try expectRequestBytes(getMaxLun(@enumFromInt(0)), .{ 0xA1, 0xFE, 0, 0, 0, 0, 1, 0 });
+}

system/devices/usb-ids.zig

@@ -0,0 +1,307 @@
+//! USB class-code decoding: turn the (class, subclass, protocol) triple a USB device or
+//! interface reports in its descriptors into typed values. The device descriptor carries one
+//! triple for the whole device, and each interface descriptor carries its own; a class code
+//! of zero at the device level defers entirely to the interfaces. Subclass and protocol
+//! codes are qualified by the class code — the same value means different things under
+//! different classes — so there is no single SubClass or Protocol enum: each class with
+//! spec-defined codes gets its own namespace below. Pure reference data (from the USB-IF
+//! defined class codes; see https://www.usb.org/defined-class-codes) — no hardware access —
+//! so it is shared by kernel discovery and any user-space tool (device naming, driver
+//! matching).
+
+// Base class codes (assigned by the USB-IF). The comment on each value notes where the code
+// may legally appear: in the device descriptor, in interface descriptors, or both.
+pub const Class = enum(u8) {
+    // Use class information in the interface descriptors (device descriptor only). Each
+    // interface within a configuration specifies its own class information and the various
+    // interfaces operate independently.
+    per_interface = 0x00,
+    // Audio: speakers, microphones, sound cards (interface)
+    audio = 0x01,
+    // Communications and CDC control: modems, network adapters (both)
+    communications = 0x02,
+    // Human Interface Device: keyboards, mice, game controllers (interface)
+    hid = 0x03,
+    // Physical: force-feedback devices (interface)
+    physical = 0x05,
+    // Image: still-imaging cameras, scanners (interface)
+    image = 0x06,
+    // Printer (interface)
+    printer = 0x07,
+    // Mass storage: flash drives, external disks, card readers (interface)
+    mass_storage = 0x08,
+    // Hub (device descriptor only)
+    hub = 0x09,
+    // CDC-Data: the data interfaces paired with a communications control interface
+    // (interface)
+    cdc_data = 0x0A,
+    // Smart card readers (interface)
+    smart_card = 0x0B,
+    // Content security (interface)
+    content_security = 0x0D,
+    // Video: webcams (interface)
+    video = 0x0E,
+    // Personal healthcare devices (interface)
+    personal_healthcare = 0x0F,
+    // Audio/Video devices (interface)
+    audio_video = 0x10,
+    // Billboard: describes alternate modes a USB Type-C device supports (device descriptor
+    // only)
+    billboard = 0x11,
+    // USB Type-C bridge (interface)
+    type_c_bridge = 0x12,
+    // USB Bulk Display Protocol devices (interface)
+    bulk_display = 0x13,
+    // MCTP over USB protocol endpoint devices (interface)
+    mctp = 0x14,
+    // I3C devices (interface)
+    i3c = 0x3C,
+    // Diagnostic devices (both)
+    diagnostic = 0xDC,
+    // Wireless controllers: Bluetooth adapters (interface)
+    wireless_controller = 0xE0,
+    // Miscellaneous (both)
+    miscellaneous = 0xEF,
+    // Application-specific: firmware upgrade, IrDA bridges, test and measurement
+    // (interface)
+    application_specific = 0xFE,
+    // Vendor-specific (both)
+    vendor_specific = 0xFF,
+    _,
+};
+
+// Subclass and protocol codes qualified by Class.hub. Hubs have no subclass codes; the
+// protocol distinguishes the hub's transaction-translator arrangement.
+pub const hub = struct {
+    pub const Protocol = enum(u8) {
+        // Full-speed hub
+        full_speed = 0x00,
+        // Hi-speed hub with a single transaction translator
+        hi_speed_single_tt = 0x01,
+        // Hi-speed hub with multiple transaction translators
+        hi_speed_multi_tt = 0x02,
+        // SuperSpeed hub (USB 3)
+        super_speed = 0x03,
+        _,
+    };
+};
+
+// Subclass and protocol codes qualified by Class.hid.
+pub const hid = struct {
+    pub const SubClass = enum(u8) {
+        // No subclass
+        none = 0x00,
+        // Boot interface: the device also supports the simplified boot protocol, usable by
+        // firmware before a full HID report-descriptor parser is available
+        boot = 0x01,
+        _,
+    };
+
+    // Only meaningful when the subclass is boot
+    pub const Protocol = enum(u8) {
+        none = 0x00,
+        keyboard = 0x01,
+        mouse = 0x02,
+        _,
+    };
+};
+
+// Subclass and protocol codes qualified by Class.mass_storage. The subclass identifies the
+// command set the device understands; the protocol identifies the transport used to carry
+// commands, data, and status over the bus.
+pub const mass_storage = struct {
+    pub const SubClass = enum(u8) {
+        // SCSI command set not reported; de facto, treat as scsi
+        not_reported = 0x00,
+        // Reduced Block Commands: typically flash devices
+        rbc = 0x01,
+        // MMC-5 (ATAPI): CD and DVD drives
+        atapi = 0x02,
+        // QIC-157 tape drives (obsolete)
+        qic_157 = 0x03,
+        // UFI: floppy disk drives
+        ufi = 0x04,
+        // SFF-8070i (obsolete)
+        sff_8070i = 0x05,
+        // Transparent SCSI command set: the common case for flash drives and disks
+        scsi = 0x06,
+        // LSD FS: negotiated access to large storage devices
+        lsd_fs = 0x07,
+        // IEEE 1667
+        ieee_1667 = 0x08,
+        // Vendor-specific
+        vendor_specific = 0xFF,
+        _,
+    };
+
+    pub const Protocol = enum(u8) {
+        // Control/Bulk/Interrupt with command completion interrupt
+        cbi_completion_interrupt = 0x00,
+        // Control/Bulk/Interrupt without command completion interrupt
+        cbi = 0x01,
+        // Bulk-only transport: the common case for flash drives and disks
+        bulk_only = 0x50,
+        // USB attached SCSI
+        uas = 0x62,
+        // Vendor-specific
+        vendor_specific = 0xFF,
+        _,
+    };
+};
+
+// Subclass and protocol codes qualified by Class.communications (CDC). The protocol codes
+// are model-specific; the useful invariant is the subclass, which selects the control model
+// the interface implements.
+pub const communications = struct {
+    pub const SubClass = enum(u8) {
+        // Direct line control model
+        direct_line = 0x01,
+        // Abstract control model: USB modems and serial adapters
+        abstract_control = 0x02,
+        // Telephone control model
+        telephone = 0x03,
+        // Multi-channel control model
+        multi_channel = 0x04,
+        // CAPI control model
+        capi = 0x05,
+        // Ethernet networking control model
+        ethernet = 0x06,
+        // ATM networking control model
+        atm = 0x07,
+        // Wireless handset control model
+        wireless_handset = 0x08,
+        // Device management
+        device_management = 0x09,
+        // Mobile direct line model
+        mobile_direct_line = 0x0A,
+        // OBEX
+        obex = 0x0B,
+        // Ethernet emulation model
+        ethernet_emulation = 0x0C,
+        // Network control model
+        network_control = 0x0D,
+        _,
+    };
+};
+
+// Subclass and protocol codes qualified by Class.wireless_controller.
+pub const wireless_controller = struct {
+    pub const SubClass = enum(u8) {
+        // Radio frequency controllers
+        radio_frequency = 0x01,
+        _,
+    };
+
+    // Only meaningful when the subclass is radio_frequency
+    pub const Protocol = enum(u8) {
+        // Bluetooth programming interface
+        bluetooth = 0x01,
+        // Ultra-wideband radio control
+        ultra_wideband = 0x02,
+        // Remote NDIS
+        remote_ndis = 0x03,
+        // Bluetooth AMP controller
+        bluetooth_amp = 0x04,
+        _,
+    };
+};
+
+// Subclass and protocol codes qualified by Class.miscellaneous.
+pub const miscellaneous = struct {
+    pub const SubClass = enum(u8) {
+        // Common class
+        common = 0x02,
+        _,
+    };
+
+    // Only meaningful when the subclass is common
+    pub const Protocol = enum(u8) {
+        // Interface association descriptor: at the device level, announces that the
+        // configuration groups interfaces into functions with IADs
+        interface_association = 0x01,
+        _,
+    };
+};
+
+// Subclass and protocol codes qualified by Class.application_specific.
+pub const application_specific = struct {
+    pub const SubClass = enum(u8) {
+        // Device firmware upgrade
+        firmware_upgrade = 0x01,
+        // IrDA bridge
+        irda_bridge = 0x02,
+        // Test and measurement
+        test_and_measurement = 0x03,
+        _,
+    };
+};
+
+/// Pack a (class, subclass, protocol) triple into one 0xCCSSPP value — the
+/// bus-native identity a USB bus driver reports in `ChildAdded.identity` and the
+/// device manager matches on (the USB analog of a packed PCI class code). Mirrors
+/// `pci_class.ClassCode.pack`, so both sides build/decode the identical u64.
+pub fn packTriple(class: u8, subclass: u8, protocol: u8) u64 {
+    return (@as(u64, class) << 16) | (@as(u64, subclass) << 8) | protocol;
+}
+
+/// The inverse of `packTriple`.
+pub fn unpackTriple(triple: u64) struct { class: u8, subclass: u8, protocol: u8 } {
+    return .{
+        .class = @truncate(triple >> 16),
+        .subclass = @truncate(triple >> 8),
+        .protocol = @truncate(triple),
+    };
+}
+
+test "class codes match the USB-IF assignments" {
+    const std = @import("std");
+    const expectEqual = std.testing.expectEqual;
+
+    try expectEqual(0x03, @intFromEnum(Class.hid));
+    try expectEqual(0x09, @intFromEnum(Class.hub));
+    try expectEqual(0xFF, @intFromEnum(Class.vendor_specific));
+
+    // A typical flash drive: mass storage, transparent SCSI, bulk-only transport.
+    try expectEqual(0x06, @intFromEnum(mass_storage.SubClass.scsi));
+    try expectEqual(0x50, @intFromEnum(mass_storage.Protocol.bulk_only));
+
+    // A boot keyboard: HID, boot subclass, keyboard protocol.
+    try expectEqual(0x01, @intFromEnum(hid.SubClass.boot));
+    try expectEqual(0x01, @intFromEnum(hid.Protocol.keyboard));
+
+    // Class codes are non-exhaustive: unlisted values pass through undamaged.
+    const unknown: Class = @enumFromInt(0x42);
+    try expectEqual(0x42, @intFromEnum(unknown));
+
+    _ = hub.Protocol.hi_speed_multi_tt;
+    _ = communications.SubClass.abstract_control;
+    _ = wireless_controller.Protocol.bluetooth;
+    _ = miscellaneous.Protocol.interface_association;
+    _ = application_specific.SubClass.firmware_upgrade;
+}
+
+test "packTriple / unpackTriple round-trip the identity a bus driver reports" {
+    const std = @import("std");
+    const expectEqual = std.testing.expectEqual;
+
+    // A boot keyboard interface: HID / boot / keyboard.
+    const keyboard = packTriple(
+        @intFromEnum(Class.hid),
+        @intFromEnum(hid.SubClass.boot),
+        @intFromEnum(hid.Protocol.keyboard),
+    );
+    try expectEqual(@as(u64, 0x03_01_01), keyboard);
+
+    // A flash drive interface: mass storage / SCSI / bulk-only.
+    const storage = packTriple(
+        @intFromEnum(Class.mass_storage),
+        @intFromEnum(mass_storage.SubClass.scsi),
+        @intFromEnum(mass_storage.Protocol.bulk_only),
+    );
+    try expectEqual(@as(u64, 0x08_06_50), storage);
+
+    const parts = unpackTriple(storage);
+    try expectEqual(@as(u8, 0x08), parts.class);
+    try expectEqual(@as(u8, 0x06), parts.subclass);
+    try expectEqual(@as(u8, 0x50), parts.protocol);
+}

system/drivers/bus/bus.zig

@@ -1,210 +0,0 @@
-//! /system/drivers/bus — a user-space **bus driver**, and the smallest honest example of one.
-//!
-//! A bus driver owns a device that *contains other devices*, enumerates them by some
-//! bus-specific protocol, and publishes each one into the kernel's device table so a
-//! class driver can claim it. PCI walks configuration space; USB walks hub descriptors. Here
-//! the "bus" is the HPET's register block and the "devices" are its comparators, each
-//! a 0x20-byte window at 0x100 + 0x20*n that can be driven independently.
-//!
-//! It's a toy bus, but nothing about the mechanism is: `bus` reads how many children
-//! exist from the hardware (GENERAL_CAP bits [12:8]), publishes one `DeviceDescriptor` per
-//! child with a sub-window of its own MMIO plus the shared IRQ, and the kernel checks
-//! every one of those resources is contained in what `bus` was granted. A comparator
-//! driver then claims a child and maps only *its* registers — not the whole block.
-//!
-//! It also proves the negative: registering a child whose window escapes the parent's
-//! is refused. Without that check, `device_register` would be a system_call for mapping
-//! arbitrary physical memory.
-
-const std = @import("std");
-const runtime = @import("runtime");
-const device = runtime.device;
-
-const register_general_cap = 0x000;
-
-/// Comparator n's registers: configuration+comparator+FSB route, 0x20 bytes.
-fn timerWindow(hpet_base: u64, n: u64) device.ResourceDescriptor {
-    return .{
-        .kind = @intFromEnum(device.ResourceKind.memory),
-        .start = hpet_base + 0x100 + 0x20 * n,
-        .len = 0x20,
-    };
-}
-
-fn findHpet(buffer: []device.DeviceDescriptor) ?device.DeviceDescriptor {
-    const total = device.enumerate(buffer);
-    const n = @min(total, buffer.len);
-    for (buffer[0..n]) |d| {
-        if (d.class != @intFromEnum(device.DeviceClass.timer)) continue;
-        if (d.parent != device.no_parent) continue; // the block, not a comparator child
-        for (0..d.resource_count) |j| {
-            if (d.resources[j].kind == @intFromEnum(device.ResourceKind.memory)) return d;
-        }
-    }
-    return null;
-}
-
-/// The parent's MMIO resource, and its IRQ if it has one.
-fn resourcesOf(d: device.DeviceDescriptor) struct { mmio: device.ResourceDescriptor, irq: ?device.ResourceDescriptor } {
-    var mmio: device.ResourceDescriptor = undefined;
-    var irq: ?device.ResourceDescriptor = null;
-    for (0..d.resource_count) |j| {
-        const r = d.resources[j];
-        if (r.kind == @intFromEnum(device.ResourceKind.memory)) mmio = r;
-        if (r.kind == @intFromEnum(device.ResourceKind.irq)) irq = r;
-    }
-    return .{ .mmio = mmio, .irq = irq };
-}
-
-fn firstChildOf(buffer: []device.DeviceDescriptor, total: usize, parent_id: u64) ?u64 {
-    for (buffer[0..@min(total, buffer.len)]) |d| {
-        if (d.parent == parent_id) return d.id;
-    }
-    return null;
-}
-
-pub fn main() void {
-    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
-        _ = runtime.system.write("bus: out of memory\n");
-        return;
-    };
-
-    const parent = findHpet(buffer) orelse {
-        _ = runtime.system.write("bus: no HPET\n");
-        return;
-    };
-    const resource = resourcesOf(parent);
-
-    // Claim the bus. Everything below is subdivision of what this claim granted.
-    //
-    // Claims are exclusive, and at a normal boot the kernel spawns every initial_ramdisk
-    // binary — so hpet may own the HPET already. That's not an error, it's the
-    // capability model working: exit quietly and leave the device to its owner. The
-    // `bus` test spawns bus alone, so there it wins the claim.
-    if (!device.claim(parent.id)) {
-        _ = runtime.system.write("bus: HPET already claimed by another driver, nothing to do\n");
-        return;
-    }
-
-    // Enumerate the bus: ask the hardware how many children it has.
-    const base = device.mmioMap(parent.id, 0) orelse {
-        _ = runtime.system.write("bus: mmio_map failed\n");
-        return;
-    };
-    const cap: *volatile u64 = @ptrFromInt(base + register_general_cap);
-    const n_children = ((cap.* >> 8) & 0x1F) + 1;
-
-    // Publish one child per comparator, each owning only its own window.
-    var published: u64 = 0;
-    var n: u64 = 0;
-    while (n < n_children) : (n += 1) {
-        var child = std.mem.zeroes(device.DeviceDescriptor);
-        child.class = @intFromEnum(device.DeviceClass.timer);
-        child.hid_len = 6;
-        child.hid[0..6].* = "hpet-t".*;
-        child.resource_count = 1;
-        child.resources[0] = timerWindow(resource.mmio.start, n);
-        // Comparators share the block's interrupt line; only one child can bind it,
-        // but all of them may legitimately name it.
-        if (resource.irq) |i| {
-            child.resources[child.resource_count] = i;
-            child.resource_count += 1;
-        }
-
-        if (device.register(parent.id, &child) == null) {
-            _ = runtime.system.write("bus: register failed\n");
-            return;
-        }
-        published += 1;
-    }
-
-    // The negative case. A window one byte past the end of the parent's must be
-    // refused — otherwise device_register would be "map any physical page you like".
-    // Confirm the table did not grow, not merely that the call returned null: null
-    // also means NoSpace/BadParent, so a size check is what actually proves the
-    // *containment* rule fired.
-    const before = device.enumerate(buffer);
-    var rogue = std.mem.zeroes(device.DeviceDescriptor);
-    rogue.class = @intFromEnum(device.DeviceClass.unknown);
-    rogue.resource_count = 1;
-    rogue.resources[0] = .{
-        .kind = @intFromEnum(device.ResourceKind.memory),
-        .start = resource.mmio.start + resource.mmio.len,
-        .len = 0x1000,
-    };
-    if (device.register(parent.id, &rogue) != null) {
-        _ = runtime.system.write("bus: FAIL out-of-window child was accepted\n");
-        return;
-    }
-    if (device.enumerate(buffer) != before) {
-        _ = runtime.system.write("bus: FAIL rogue child leaked into the table\n");
-        return;
-    }
-
-    // And confirm the children came back with the right parent and a *narrower*
-    // window than the bus — read from the table, not from our own memory.
-    const total = device.enumerate(buffer);
-    var seen: u64 = 0;
-    for (buffer[0..@min(total, buffer.len)]) |d| {
-        if (d.parent != parent.id) continue;
-        const w = d.resources[0];
-        if (w.start < resource.mmio.start or w.len >= resource.mmio.len) {
-            _ = runtime.system.write("bus: FAIL child window is not inside the bus\n");
-            return;
-        }
-        seen += 1;
-    }
-    if (seen != published) {
-        _ = runtime.system.write("bus: FAIL child count mismatch\n");
-        return;
-    }
-
-    // Delegation, end to end: claim a child and map *it*. A real class driver would be
-    // a different process; here bus plays both parts, which exercises the same path.
-    // The child's window is 0x20 bytes at parent+0x100, so the register it sees at
-    // offset 0 must be the same timer-0 configuration register the bus sees at 0x100.
-    //
-    // (mmio_map rounds to a page, so the child's mapping physically covers the whole
-    // 4 KiB the HPET lives in — the granularity limit documented in docs/drivers.md.
-    // The *resource* is narrow even though the page isn't.)
-    const child_id = firstChildOf(buffer, device.enumerate(buffer), parent.id) orelse {
-        _ = runtime.system.write("bus: FAIL no child to claim\n");
-        return;
-    };
-    if (!device.claim(child_id)) {
-        _ = runtime.system.write("bus: FAIL could not claim own child\n");
-        return;
-    }
-    const child_base = device.mmioMap(child_id, 0) orelse {
-        _ = runtime.system.write("bus: FAIL child mmio_map refused\n");
-        return;
-    };
-    const via_child: *volatile u64 = @ptrFromInt(child_base);
-    const via_bus: *volatile u64 = @ptrFromInt(base + 0x100);
-    if (via_child.* != via_bus.*) {
-        _ = runtime.system.write("bus: FAIL child window does not alias the bus register\n");
-        return;
-    }
-
-    // A descriptor pointer into an unmapped page must fail the call, not fault the
-    // kernel. Grab a page, free it, and register through the stale address: if the
-    // kernel dereferenced it raw (rather than copying in through the page tables) this
-    // would triple-fault QEMU and the test would time out instead of printing ok.
-    const scratch = runtime.system.mmap(0x1000, runtime.system.PROT_READ | runtime.system.PROT_WRITE);
-    if (!runtime.system.mmapFailed(scratch)) {
-        _ = runtime.system.munmap(scratch, 0x1000);
-        const descriptor: *const device.DeviceDescriptor = @ptrFromInt(scratch);
-        if (device.register(parent.id, descriptor) != null) {
-            _ = runtime.system.write("bus: FAIL register accepted an unmapped descriptor\n");
-            return;
-        }
-    }
-
-    _ = runtime.system.write("bus: ok\n");
-    while (true) runtime.system.sleep(1000);
-}
-
-pub const panic = runtime.panic;
-comptime {
-    _ = &runtime.start._start;
-}

system/drivers/hpet/hpet.zig

@@ -1,195 +0,0 @@
-//! /system/drivers/hpet — a user-space HPET driver. It proves the whole driver model end to
-//! end: enumerate the device table, find the HPET, claim it, map its registers into
-//! this ring-3 address space (strong-uncacheable), **bind its interrupt to an IPC
-//! endpoint**, then sit blocked in `replyWait` until the hardware wakes it.
-//!
-//! Nothing here polls. Between interrupts the process is `.blocked` and off every
-//! scheduler queue; the core runs other work or idles. That is the point of the
-//! exercise — a driver is a process that sleeps until its device has something to
-//! say (see docs/drivers.md).
-//!
-//! The comparator is configured **level-triggered** on purpose. Edge would be
-//! simpler, but level is the discipline every real device line needs, and it forces
-//! the full cycle to be correct:
-//!
-//!   kernel ISR   mask the GSI -> EOI -> notify this endpoint
-//!   hpet        wake, clear GENERAL_INT_STATUS (deasserts the line), re-arm
-//!   hpet        irq_ack -> kernel unmasks the GSI
-//!
-//! Clear the status bit *before* acking, or the line is still asserted when the
-//! kernel unmasks and the I/O APIC redelivers forever.
-//!
-//! Register map (HPET spec 1.0a):
-//!   0x000 GENERAL_CAP     [63:32] fs per tick, [12:8] number timers - 1
-//!   0x010 GENERAL_CONFIGURATION  bit0 ENABLE_CNF, bit1 LEG_RT_CNF
-//!   0x020 GENERAL_INT_STATUS  bit n = timer n asserted (write 1 to clear)
-//!   0x0F0 MAIN_COUNTER
-//!   0x100 TIMER0_CONFIGURATION   bit1 INT_TYPE(1=level) bit2 INT_ENB bit3 TYPE(periodic)
-//!                         bits[13:9] INT_ROUTE, [63:32] INT_ROUTE_CAP
-//!   0x108 TIMER0_COMPARATOR
-
-const runtime = @import("runtime");
-const mmio = @import("mmio");
-const device = runtime.device;
-const ipc = runtime.ipc;
-
-const register_general_cap = 0x000;
-const register_general_configuration = 0x010;
-const register_int_status = 0x020;
-const register_main_counter = 0x0F0;
-const register_timer0_configuration = 0x100;
-const register_timer0_comparator = 0x108;
-
-const configuration_enable: u64 = 1 << 0; // GENERAL_CONFIGURATION.ENABLE_CNF
-const configuration_leg_rt: u64 = 1 << 1; // GENERAL_CONFIGURATION.LEG_RT_CNF
-const tn_int_type_level: u64 = 1 << 1;
-const tn_int_enb: u64 = 1 << 2;
-const tn_type_periodic: u64 = 1 << 3;
-const tn_route_shift = 9;
-const tn_route_mask: u64 = 0x1F << tn_route_shift;
-
-/// Interrupts to observe before declaring victory.
-const target_ticks = 5;
-
-/// Read/write a 64-bit HPET register through the typed volatile MMIO layer (/lib/mmio).
-/// The HPET is pure MMIO with no DMA, and on x86 its grant is strong-uncacheable (so
-/// UC writes are already ordered) — no barriers are needed here; the point is the
-/// typed, arch-portable access every driver should use.
-inline fn rd(base: usize, off: usize) u64 {
-    return mmio.read(u64, base + off);
-}
-inline fn wr(base: usize, off: usize, value: u64) void {
-    mmio.write(u64, base + off, value);
-}
-
-/// A timer-class device exposing both an MMIO window and an IRQ: its id, the two
-/// resource indices, and the GSI discovery chose out of `Tn_INT_ROUTE_CAP`.
-const Found = struct { device_id: u64, mmio: u64, irq: u64, gsi: u64 };
-
-fn findHpet(buffer: []device.DeviceDescriptor) ?Found {
-    const total = device.enumerate(buffer);
-    const n = @min(total, buffer.len);
-    for (buffer[0..n]) |d| {
-        if (d.class != @intFromEnum(device.DeviceClass.timer)) continue;
-        // Skip comparator children a bus driver may have published below the block
-        // (see system/drivers/bus/bus.zig) — we want the register block itself.
-        if (d.parent != device.no_parent) continue;
-        var mmio_index: ?u64 = null;
-        var irq: ?u64 = null;
-        for (0..d.resource_count) |j| {
-            switch (d.resources[j].kind) {
-                @intFromEnum(device.ResourceKind.memory) => mmio_index = mmio_index orelse j,
-                @intFromEnum(device.ResourceKind.irq) => irq = irq orelse j,
-                else => {},
-            }
-        }
-        if (mmio_index) |m| if (irq) |i| {
-            return .{ .device_id = d.id, .mmio = m, .irq = i, .gsi = d.resources[i].start };
-        };
-    }
-    return null;
-}
-
-pub fn main() void {
-    // Enumerate into a heap buffer (too big for the one-page user stack).
-    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 32) catch {
-        _ = runtime.system.write("hpet: out of memory\n");
-        return;
-    };
-
-    const hpet = findHpet(buffer) orelse {
-        _ = runtime.system.write("hpet: no HPET with an IRQ\n");
-        return;
-    };
-
-    if (!device.claim(hpet.device_id)) {
-        _ = runtime.system.write("hpet: claim failed\n");
-        return;
-    }
-    const base = device.mmioMap(hpet.device_id, hpet.mmio) orelse {
-        _ = runtime.system.write("hpet: mmio_map failed\n");
-        return;
-    };
-
-    // The GSI discovery picked for us out of Tn_INT_ROUTE_CAP. Program the comparator
-    // to raise exactly this line — the kernel will only bind the one it recorded.
-    const gsi = hpet.gsi;
-
-    const endpoint = ipc.createIpcEndpoint() orelse {
-        _ = runtime.system.write("hpet: create_ipc_endpoint failed\n");
-        return;
-    };
-
-    // --- program the hardware ------------------------------------------------
-    // Counter period, so we can arm the comparator a fixed wall-clock distance out.
-    const femtos_per_tick = rd(base, register_general_cap) >> 32;
-    if (femtos_per_tick == 0) {
-        _ = runtime.system.write("hpet: bad HPET period\n");
-        return;
-    }
-    const ticks_per_ms = 1_000_000_000_000 / femtos_per_tick;
-
-    // Stop the counter and take the legacy route off while we reconfigure.
-    wr(base, register_general_configuration, rd(base, register_general_configuration) & ~(configuration_enable | configuration_leg_rt));
-
-    // Timer 0: one-shot, level-triggered, routed to our GSI, interrupt enabled.
-    // One-shot (not periodic) sidesteps the HPET's Tn_value_SET accumulator quirk —
-    // we simply re-arm from the driver on each interrupt, which is what a tickless
-    // timer driver does anyway.
-    var t0 = rd(base, register_timer0_configuration);
-    t0 &= ~(tn_route_mask | tn_type_periodic);
-    t0 |= tn_int_type_level | tn_int_enb | (gsi << tn_route_shift);
-    wr(base, register_timer0_configuration, t0);
-
-    // Clear any stale assertion, then arm ~100 ms out and start the counter.
-    wr(base, register_int_status, 1);
-    wr(base, register_timer0_comparator, rd(base, register_main_counter) + ticks_per_ms * 100);
-    wr(base, register_general_configuration, rd(base, register_general_configuration) | configuration_enable);
-
-    if (!device.irqBind(hpet.device_id, hpet.irq, endpoint)) {
-        _ = runtime.system.write("hpet: irq_bind failed\n");
-        return;
-    }
-    _ = runtime.system.write("hpet: bound, sleeping until the hardware speaks\n");
-
-    // --- the driver loop -----------------------------------------------------
-    // Blocked in replyWait. No polling, no spinning: the next line of this function
-    // runs only because an interrupt fired.
-    var receive: [64]u8 = undefined;
-    var count: usize = 0;
-    while (count < target_ticks) {
-        // Blocked here. The task is `.blocked` and off every scheduler queue; the
-        // next line runs only because the HPET raised its line.
-        const r = ipc.replyWait(endpoint, &.{}, &receive, null);
-        if (!r.isNotification()) continue; // a client request, not our IRQ
-
-        // Quiet the device: write 1 to timer 0's status bit. Until this lands, the
-        // line is still asserted and unmasking would refire immediately.
-        wr(base, register_int_status, 1);
-        count += 1;
-
-        if (count < target_ticks) {
-            wr(base, register_timer0_comparator, rd(base, register_main_counter) + ticks_per_ms * 100);
-        } else {
-            // Last one: stop the source rather than re-arming, so the line is left
-            // both quiet *and* unmasked by the ack below. Re-arming here would leave
-            // a pending interrupt that nobody is waiting for, and the ISR would mask
-            // the line again a moment later.
-            wr(base, register_timer0_configuration, rd(base, register_timer0_configuration) & ~tn_int_enb);
-        }
-
-        _ = runtime.system.write("hpet: irq\n");
-        if (!device.irqAck(hpet.device_id, hpet.irq)) {
-            _ = runtime.system.write("hpet: irq_ack failed\n");
-            return;
-        }
-    }
-
-    _ = runtime.system.write("hpet: ok\n");
-    while (true) runtime.system.sleep(1000);
-}
-
-pub const panic = runtime.panic;
-comptime {
-    _ = &runtime.start._start;
-}

system/drivers/pci-bus/pci-bus.zig

@@ -0,0 +1,270 @@
+//! /system/drivers/pci-bus — the PCI bus driver: enumeration moved out of ring 0
+//! (docs/discovery.md). The device manager matches the `pci_host_bridge`
+//! node and spawns one instance per bridge, the bridge's device id as argv[1] —
+//! the same per-device contract as usb-xhci-bus.
+//!
+//! M19.1 (this increment): claim the bridge, map its ECAM window (resource 0;
+//! the bus range and the MMIO apertures follow it), walk every
+//! bus/device/function config header, and log what the walk finds — ending
+//! with "/system/drivers/pci-bus: N functions found", which the `pci-scan` scenario compares
+//! against the kernel's own enumeration. Registration and reports (M19.2), and
+//! the kernel walk's retirement (M19.3), build on this proven-equivalent scan.
+
+const std = @import("std");
+const runtime = @import("runtime");
+const protocol = runtime.device_manager_protocol;
+const device = runtime.device;
+const pci_class = @import("pci-class");
+
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+/// Log a discovered function with its (class / subclass / prog-IF) triple decoded
+/// to human names — the boot-log breadcrumb that says *what* the hardware is, so
+/// "class 0x01 (Mass Storage Controller) subclass 0x06 (Serial ATA Controller)
+/// progif 0x01 (AHCI 1.0)" reads straight off the log when writing a new driver.
+/// A dedicated wider buffer than `writeLine`'s, since the decoded names are long.
+fn logFunction(bus: u64, dev: u64, function: u64, class_triple: u32) void {
+    const cc = pci_class.ClassCode.unpack(@truncate(class_triple));
+    const pif = pci_class.progIfName(cc.base, cc.subclass, cc.prog_if);
+    var line: [200]u8 = undefined;
+    const text = if (pif.len != 0)
+        std.fmt.bufPrint(&line, "/system/drivers/pci-bus: {d}:{d}.{d} class 0x{x:0>2} ({s})  subclass 0x{x:0>2} ({s})  progif 0x{x:0>2} ({s})\n", .{ bus, dev, function, cc.base, pci_class.className(cc.base), cc.subclass, pci_class.subclassName(cc.base, cc.subclass), cc.prog_if, pif }) catch return
+    else
+        std.fmt.bufPrint(&line, "/system/drivers/pci-bus: {d}:{d}.{d} class 0x{x:0>2} ({s})  subclass 0x{x:0>2} ({s})  progif 0x{x:0>2}\n", .{ bus, dev, function, cc.base, pci_class.className(cc.base), cc.subclass, pci_class.subclassName(cc.base, cc.subclass), cc.prog_if }) catch return;
+    _ = runtime.system.write(text);
+}
+
+var bridge_id: u64 = protocol.no_device;
+var ecam_base: usize = 0;
+var ecam_physical: u64 = 0;
+var start_bus: u64 = 0;
+var bus_count: u64 = 0;
+var manager_handle: runtime.ipc.Handle = 0;
+
+/// One aligned 32-bit read from a function's configuration space.
+fn configRead(bus: u64, dev: u64, function: u64, offset: u64) u32 {
+    const address = ecam_base + (((bus - start_bus) << 20) | (dev << 15) | (function << 12) | offset);
+    const register: *volatile u32 = @ptrFromInt(address);
+    return register.*;
+}
+
+fn configWrite(bus: u64, dev: u64, function: u64, offset: u64, value: u32) void {
+    const address = ecam_base + (((bus - start_bus) << 20) | (dev << 15) | (function << 12) | offset);
+    const register: *volatile u32 = @ptrFromInt(address);
+    register.* = value;
+}
+
+fn configRead16(bus: u64, dev: u64, function: u64, offset: u64) u16 {
+    const word = configRead(bus, dev, function, offset & ~@as(u64, 3));
+    return @truncate(word >> @intCast((offset & 3) * 8));
+}
+
+fn configWrite16(bus: u64, dev: u64, function: u64, offset: u64, value: u16) void {
+    const aligned = offset & ~@as(u64, 3);
+    const shift: u5 = @intCast((offset & 3) * 8);
+    const word = configRead(bus, dev, function, aligned);
+    const mask = @as(u32, 0xFFFF) << shift;
+    configWrite(bus, dev, function, aligned, (word & ~mask) | (@as(u32, value) << shift));
+}
+
+/// Claim the bridge, map the ECAM, hello the manager, then scan.
+fn initialise(endpoint: runtime.ipc.Handle) bool {
+    _ = endpoint;
+    if (!device.claim(bridge_id)) {
+        writeLine("/system/drivers/pci-bus: unable to claim bridge device {d}\n", .{bridge_id});
+        return false;
+    }
+    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
+        _ = runtime.system.write("/system/drivers/pci-bus: out of memory\n");
+        return false;
+    };
+    const total = device.enumerate(buffer);
+    const descriptor = for (buffer[0..@min(total, buffer.len)]) |d| {
+        if (d.id == bridge_id) break d;
+    } else {
+        writeLine("/system/drivers/pci-bus: device {d} not in the device tree\n", .{bridge_id});
+        return false;
+    };
+    // Resource 0 is the ECAM window (1 MiB of config space per bus); the bus
+    // range rides beside it. The MMIO apertures (M19.0) come after both.
+    if (descriptor.resource_count < 2 or descriptor.resources[0].kind != @intFromEnum(device.ResourceKind.memory)) {
+        _ = runtime.system.write("/system/drivers/pci-bus: bridge has no ECAM window\n");
+        return false;
+    }
+    const bus_range = for (descriptor.resources[0..@intCast(descriptor.resource_count)]) |resource| {
+        if (resource.kind == @intFromEnum(device.ResourceKind.bus_range)) break resource;
+    } else {
+        _ = runtime.system.write("/system/drivers/pci-bus: bridge has no bus range\n");
+        return false;
+    };
+    start_bus = bus_range.start;
+    bus_count = bus_range.len;
+    ecam_physical = descriptor.resources[0].start;
+    ecam_base = device.mmioMap(bridge_id, 0) orelse {
+        _ = runtime.system.write("/system/drivers/pci-bus: ECAM mmio_map failed\n");
+        return false;
+    };
+
+    // The handshake, then the scan (reports join in M19.2).
+    var manager: ?runtime.ipc.Handle = null;
+    var tries: u32 = 0;
+    while (manager == null and tries < 100) : (tries += 1) {
+        manager = runtime.ipc.lookup(.device_manager);
+        if (manager == null) runtime.system.sleep(20);
+    }
+    const h = manager orelse {
+        _ = runtime.system.write("/system/drivers/pci-bus: no device manager to hello\n");
+        return false;
+    };
+    const hello = protocol.Hello{ .role = @intFromEnum(protocol.Role.bus), .device_id = bridge_id };
+    var reply: [protocol.message_maximum]u8 = undefined;
+    const n = runtime.ipc.call(h, std.mem.asBytes(&hello), &reply) catch {
+        _ = runtime.system.write("/system/drivers/pci-bus: hello call failed\n");
+        return false;
+    };
+    if (n < protocol.reply_size or std.mem.bytesToValue(protocol.HelloReply, reply[0..protocol.reply_size]).status != 0) {
+        _ = runtime.system.write("/system/drivers/pci-bus: hello refused\n");
+        return false;
+    }
+    manager_handle = h;
+
+    scan();
+    return true;
+}
+
+/// The brute-force walk the kernel does today, from ring 3: every bus in the
+/// range, 32 devices, 8 functions; vendor id FFFFh means nothing decodes there,
+/// and only multifunction devices get their functions 1..7 probed.
+fn scan() void {
+    var found: u32 = 0;
+    var bus: u64 = start_bus;
+    while (bus < start_bus + bus_count) : (bus += 1) {
+        var dev: u64 = 0;
+        while (dev < 32) : (dev += 1) {
+            const first = configRead(bus, dev, 0, 0);
+            if (first & 0xFFFF == 0xFFFF) continue;
+            const multifunction = (configRead(bus, dev, 0, 0x0C) >> 16) & 0x80 != 0;
+            var function: u64 = 0;
+            while (function < 8) : (function += 1) {
+                if (function != 0 and !multifunction) break;
+                const vendor_device = configRead(bus, dev, function, 0);
+                if (vendor_device & 0xFFFF == 0xFFFF) continue;
+                const class_revision = configRead(bus, dev, function, 0x08);
+                found += 1;
+                logFunction(bus, dev, function, class_revision >> 8);
+                registerAndReport(bus, dev, function, class_revision >> 8);
+            }
+        }
+    }
+    writeLine("/system/drivers/pci-bus: {d} functions found\n", .{found});
+}
+
+/// Register one function under the bridge and report it to the manager. The
+/// descriptor mirrors the kernel's own recording byte for byte — config slice
+/// as resource 0, then the sized BARs — so during coexistence the idempotent
+/// device_register (M19.0) returns the kernel's existing node id rather than
+/// growing a duplicate, and the report carries the id drivers already use.
+fn registerAndReport(bus: u64, dev: u64, function: u64, class_triple: u32) void {
+    var descriptor = std.mem.zeroes(device.DeviceDescriptor);
+    descriptor.class = @intFromEnum(device.DeviceClass.pci_device);
+    descriptor.pci_class = class_triple;
+    descriptor.resources[0] = .{
+        .kind = @intFromEnum(device.ResourceKind.memory),
+        .start = ecam_physical + (((bus - start_bus) << 20) | (dev << 15) | (function << 12)),
+        .len = 4096,
+    };
+    descriptor.resource_count = 1;
+
+    // The standard BAR-sizing probe, exactly as the kernel does it: decode off,
+    // write all-ones, read the writable mask back, restore. Header type 0 only.
+    const header_type = (configRead(bus, dev, function, 0x0C) >> 16) & 0x7F;
+    if (header_type == 0) {
+        const command = configRead16(bus, dev, function, 0x04);
+        configWrite16(bus, dev, function, 0x04, command & ~@as(u16, 0b11));
+        var i: u64 = 0;
+        while (i < 6) : (i += 1) {
+            if (descriptor.resource_count >= 8) break;
+            const off = 0x10 + i * 4;
+            const original = configRead(bus, dev, function, off);
+            if (original == 0) continue;
+            const slot: usize = @intCast(descriptor.resource_count);
+            if (original & 1 != 0) {
+                configWrite(bus, dev, function, off, 0xFFFF_FFFF);
+                const readback = configRead(bus, dev, function, off);
+                configWrite(bus, dev, function, off, original);
+                const mask = readback & 0xFFFF_FFFC;
+                const size: u32 = if (mask == 0) 0 else (~mask +% 1) & 0xFFFF;
+                if (size == 0) continue; // unimplemented BAR — nothing to register
+                descriptor.resources[slot] = .{ .kind = @intFromEnum(device.ResourceKind.io_port), .start = original & 0xFFFF_FFFC, .len = size };
+                descriptor.resource_count += 1;
+            } else if ((original >> 1) & 0x3 == 2) {
+                const original_high = configRead(bus, dev, function, off + 4);
+                configWrite(bus, dev, function, off, 0xFFFF_FFFF);
+                configWrite(bus, dev, function, off + 4, 0xFFFF_FFFF);
+                const lo = configRead(bus, dev, function, off);
+                const hi = configRead(bus, dev, function, off + 4);
+                configWrite(bus, dev, function, off, original);
+                configWrite(bus, dev, function, off + 4, original_high);
+                const readback = (@as(u64, hi) << 32) | (lo & 0xFFFF_FFF0);
+                const size: u64 = if (readback == 0) 0 else ~readback +% 1;
+                i += 1; // consumed the high half regardless
+                if (size == 0) continue;
+                descriptor.resources[slot] = .{ .kind = @intFromEnum(device.ResourceKind.memory), .start = (@as(u64, original_high) << 32) | (original & 0xFFFF_FFF0), .len = size };
+                descriptor.resource_count += 1;
+            } else {
+                configWrite(bus, dev, function, off, 0xFFFF_FFFF);
+                const readback = configRead(bus, dev, function, off);
+                configWrite(bus, dev, function, off, original);
+                const mask = readback & 0xFFFF_FFF0;
+                const size: u32 = if (mask == 0) 0 else ~mask +% 1;
+                if (size == 0) continue;
+                descriptor.resources[slot] = .{ .kind = @intFromEnum(device.ResourceKind.memory), .start = original & 0xFFFF_FFF0, .len = size };
+                descriptor.resource_count += 1;
+            }
+        }
+        configWrite16(bus, dev, function, 0x04, command);
+    }
+
+    const registered = device.register(bridge_id, &descriptor) orelse {
+        writeLine("/system/drivers/pci-bus: register refused for {d}:{d}.{d}\n", .{ bus, dev, function });
+        return;
+    };
+    const report = protocol.ChildAdded{
+        .parent = bridge_id,
+        .bus_address = (bus << 8) | (dev << 3) | function,
+        .identity = class_triple,
+        .device_id = registered,
+    };
+    var reply: [protocol.message_maximum]u8 = undefined;
+    _ = runtime.ipc.call(manager_handle, std.mem.asBytes(&report), &reply) catch {
+        writeLine("/system/drivers/pci-bus: child report for {d}:{d}.{d} failed\n", .{ bus, dev, function });
+    };
+}
+
+fn onMessage(message: []const u8, reply: []u8, sender: u32, capability: ?runtime.ipc.Handle) usize {
+    _ = message;
+    _ = reply;
+    _ = sender;
+    _ = capability;
+    return 0;
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const argument = init.arguments.get(1) orelse return; // bare (ramdisk sweep): stay silent
+    bridge_id = std.fmt.parseInt(u64, argument, 10) catch {
+        writeLine("/system/drivers/pci-bus: malformed bridge device id '{s}'\n", .{argument});
+        return;
+    };
+    runtime.service.run(protocol.message_maximum, .{
+        .init = initialise,
+        .on_message = onMessage,
+    });
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start; // pull the runtime entry shim into the image
+}

system/drivers/ps2-bus/keyboard.zig

@@ -0,0 +1,184 @@
+//! PS/2 Keyboard Driver
+//!
+//! Spawned by the ps2-bus driver once the controller is initialized and the port
+//! has passed its interface test and device reset. The bus driver hands us our
+//! device HID as argv[1] and, optionally, a layout name (`"us"`, `"gb"`, ...) as
+//! argv[2].
+//!
+//! The 8042's ports (0x60/0x64) and IRQ1 live on the PNP0303 node, which the
+//! ps2-bus driver exclusively owns — so this driver never touches the hardware.
+//! Instead it **attaches** to the bus (handing over its endpoint as a capability)
+//! and receives every scancode byte as a forwarded asynchronous message. Each byte
+//! feeds the set-2 decoder; a decoded key becomes input-protocol events:
+//!
+//!   scancode byte -> HID usage keycode -> key_down / key_up
+//!                                      -> xkeyboard-config -> character -> key_press
+
+const std = @import("std");
+const runtime = @import("runtime");
+const xkb = @import("xkeyboard-config");
+const ps2 = @import("ps2-library.zig");
+const scancode = @import("scancode.zig");
+const device = runtime.device;
+const ipc = runtime.ipc;
+const protocol = runtime.input_protocol;
+
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+/// Look up the ps2-bus service, retrying while the bus (which spawned us before
+/// registering) is still coming up.
+fn lookupBus() ?ipc.Handle {
+    var attempts: usize = 0;
+    while (attempts < 100) : (attempts += 1) {
+        if (ipc.lookup(.ps2_bus)) |handle| return handle;
+        runtime.system.sleep(50);
+    }
+    return null;
+}
+
+/// The character a pressed key produces under `modifiers`, or 0 for none. The
+/// layout lookup answers for printable keys; the keys whose keysym has no Unicode
+/// mapping but that every consumer still expects as a character (Enter, Tab,
+/// Backspace, Escape) are given their ASCII control characters here.
+fn characterFor(layout: *const xkb.Layout, usage: u8, modifiers: scancode.ModifierSnapshot) u32 {
+    const mapping = xkb.map(layout, usage, .{
+        .shift = modifiers.shift,
+        .caps_lock = modifiers.caps_lock,
+        .level3 = modifiers.right_alt,
+        .control = modifiers.control,
+    });
+    if (mapping.character) |character| return character;
+    return switch (@as(protocol.Keycode, @enumFromInt(usage))) {
+        .enter, .keypad_enter => '\n',
+        .tab => '\t',
+        .backspace => 0x08,
+        .escape => 0x1B,
+        else => 0,
+    };
+}
+
+/// The input protocol's modifier word for a snapshot.
+fn modifierWord(modifiers: scancode.ModifierSnapshot) u32 {
+    var word: u32 = 0;
+    if (modifiers.shift) word |= protocol.modifier_shift;
+    if (modifiers.control) word |= protocol.modifier_control;
+    if (modifiers.alt) word |= protocol.modifier_alt;
+    return word;
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const hid = init.arguments.get(1).?;
+    if (hid.len == 0) {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: no HID argument\n");
+        return;
+    }
+    writeLine("/system/drivers/ps2-bus/keyboard: starting for hid {s}\n", .{hid});
+
+    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: out of memory\n");
+        return;
+    };
+    if (device.findDeviceDescriptorByHid(buffer, hid) == null) {
+        writeLine("/system/drivers/ps2-bus/keyboard: no device for hid {s}\n", .{hid});
+        return;
+    }
+
+    // The layout is a spawn argument so a later settings source can choose it;
+    // absent (as today) it defaults to us.
+    const layout_name = init.arguments.get(2) orelse "us";
+    const layout = xkb.byName(layout_name) orelse xkb.us;
+    writeLine("/system/drivers/ps2-bus/keyboard: layout {s}\n", .{layout.name});
+
+    // Attach to the bus: hand it our endpoint, and it forwards every byte the
+    // keyboard sends (it owns the controller; we own the decoding).
+    const bus = lookupBus() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: ps2-bus service unavailable\n");
+        return;
+    };
+    const endpoint = ipc.createIpcEndpoint() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: no endpoint\n");
+        return;
+    };
+    var attach = ps2.AttachRequest{ .device_type = @intFromEnum(ps2.DeviceType.keyboard) };
+    var attach_reply: [@sizeOf(ps2.AttachReply)]u8 = undefined;
+    const attached = ipc.callCap(bus, std.mem.asBytes(&attach), &attach_reply, endpoint) catch {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: attach call failed\n");
+        return;
+    };
+    if (attached.len < @sizeOf(ps2.AttachReply) or
+        std.mem.bytesToValue(ps2.AttachReply, attach_reply[0..@sizeOf(ps2.AttachReply)]).status != @intFromEnum(ps2.AttachStatus.ok))
+    {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: attach refused\n");
+        return;
+    }
+
+    // Broadcast keyboard events through the input service so programs can listen
+    // for them (docs/input.md).
+    var source = runtime.input.connectSource() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: input service unavailable\n");
+        return;
+    };
+    _ = runtime.system.write("/system/drivers/ps2-bus/keyboard: ok\n");
+
+    var decoder = scancode.Decoder{};
+    var state = scancode.KeyboardState{};
+    var receive: [@sizeOf(ps2.ForwardedByte)]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(endpoint, &.{}, &receive, null);
+        if (!got.isMessage() or got.len < @sizeOf(ps2.ForwardedByte)) continue;
+        const forwarded = std.mem.bytesToValue(ps2.ForwardedByte, receive[0..@sizeOf(ps2.ForwardedByte)]);
+
+        const key = decoder.feed(@intCast(forwarded.byte & 0xFF)) orelse continue;
+        const transition = state.apply(key);
+        const modifiers = modifierWord(transition.modifiers);
+
+        switch (transition.action) {
+            .pressed => {
+                _ = source.publishKeyboardEvent(.{
+                    .kind = @intFromEnum(protocol.EventKind.key_down),
+                    .keycode = key.usage,
+                    .character = 0,
+                    .modifiers = modifiers,
+                });
+                const character = characterFor(layout, key.usage, transition.modifiers);
+                if (character != 0) {
+                    _ = source.publishKeyboardEvent(.{
+                        .kind = @intFromEnum(protocol.EventKind.key_press),
+                        .keycode = key.usage,
+                        .character = character,
+                        .modifiers = modifiers,
+                    });
+                }
+            },
+            // Typematic repeat: the key did not physically go down again, so no
+            // key_down — but it keeps producing its character.
+            .repeated => {
+                const character = characterFor(layout, key.usage, transition.modifiers);
+                if (character != 0) {
+                    _ = source.publishKeyboardEvent(.{
+                        .kind = @intFromEnum(protocol.EventKind.key_press),
+                        .keycode = key.usage,
+                        .character = character,
+                        .modifiers = modifiers,
+                    });
+                }
+            },
+            .released => {
+                _ = source.publishKeyboardEvent(.{
+                    .kind = @intFromEnum(protocol.EventKind.key_up),
+                    .keycode = key.usage,
+                    .character = 0,
+                    .modifiers = modifiers,
+                });
+            },
+        }
+    }
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start;
+}

system/drivers/ps2-bus/mouse-packet.zig

@@ -0,0 +1,143 @@
+//! PS/2 mouse packet assembly — the byte stream a streaming mouse sends, turned
+//! into decoded movement/button reports.
+//!
+//! A standard PS/2 mouse in stream mode sends three-byte packets:
+//!
+//!   byte 0: | Y ovf | X ovf | Y sign | X sign | 1 | middle | right | left |
+//!   byte 1: X movement (low eight bits; the sign bit lives in byte 0)
+//!   byte 2: Y movement (likewise)
+//!
+//! Movement is nine-bit two's complement, PS/2 convention: positive X right,
+//! positive Y **up**. The decoded packet converts Y to the screen convention
+//! (positive down), matching what every consumer of relative motion expects.
+//! Bit 3 of byte 0 is always set — the resynchronization anchor: a byte at
+//! packet start with bit 3 clear cannot be a packet header and is dropped.
+//!
+//! Everything here is pure (no imports beyond `std`, no IO), so it is
+//! host-testable: the tests at the bottom run under `zig build test`.
+
+const std = @import("std");
+
+/// One decoded movement/button report, in screen convention (positive dy down).
+pub const Packet = struct {
+    left: bool,
+    right: bool,
+    middle: bool,
+    dx: i16,
+    dy: i16,
+};
+
+const header_always_set: u8 = 1 << 3;
+const header_left: u8 = 1 << 0;
+const header_right: u8 = 1 << 1;
+const header_middle: u8 = 1 << 2;
+const header_x_sign: u8 = 1 << 4;
+const header_y_sign: u8 = 1 << 5;
+const header_x_overflow: u8 = 1 << 6;
+const header_y_overflow: u8 = 1 << 7;
+
+/// Device protocol bytes that can reach the packet stream around bring-up (the
+/// acknowledge to enable-reporting, a reset's self-test result). Both have bit 3
+/// set, so the header check alone cannot reject them; they are recognized only
+/// at packet start, where a real header cannot be one of them in practice.
+const response_acknowledge: u8 = 0xFA;
+const response_self_test_passed: u8 = 0xAA;
+
+/// Accumulates the byte stream into `Packet`s. Feed it every byte the mouse
+/// sends; the third byte of each well-formed packet returns one.
+pub const Assembler = struct {
+    bytes: [3]u8 = undefined,
+    count: u8 = 0,
+
+    pub fn feed(self: *Assembler, byte: u8) ?Packet {
+        if (self.count == 0) {
+            // Resynchronize: a packet must start with a plausible header.
+            if (byte & header_always_set == 0) return null;
+            if (byte == response_acknowledge or byte == response_self_test_passed) return null;
+        }
+        self.bytes[self.count] = byte;
+        self.count += 1;
+        if (self.count < 3) return null;
+        self.count = 0;
+
+        const header = self.bytes[0];
+        // An overflowed count is garbage by definition; discard the packet.
+        if (header & (header_x_overflow | header_y_overflow) != 0) return null;
+
+        return .{
+            .left = header & header_left != 0,
+            .right = header & header_right != 0,
+            .middle = header & header_middle != 0,
+            .dx = movement(self.bytes[1], header & header_x_sign != 0),
+            // PS/2 positive Y is up; screen positive Y is down.
+            .dy = -movement(self.bytes[2], header & header_y_sign != 0),
+        };
+    }
+
+    /// Nine-bit two's complement: the eight movement bits plus the header's sign.
+    fn movement(low: u8, negative: bool) i16 {
+        const value: i16 = low;
+        return if (negative) value - 256 else value;
+    }
+};
+
+// --- tests (host-run via `zig build test`) ------------------------------------
+
+const testing = std.testing;
+
+fn feedAll(assembler: *Assembler, bytes: []const u8) ?Packet {
+    var result: ?Packet = null;
+    for (bytes) |byte| {
+        if (assembler.feed(byte)) |packet| result = packet;
+    }
+    return result;
+}
+
+test "plain motion decodes with screen-convention y" {
+    var assembler = Assembler{};
+    const packet = feedAll(&assembler, &.{ 0x08, 5, 3 }).?;
+    try testing.expectEqual(@as(i16, 5), packet.dx);
+    try testing.expectEqual(@as(i16, -3), packet.dy); // PS/2 up 3 -> screen -3
+    try testing.expect(!packet.left and !packet.right and !packet.middle);
+}
+
+test "negative movement sign-extends through the header bits" {
+    var assembler = Assembler{};
+    // X sign and Y sign set: dx = 0xFB - 256 = -5, dy raw = 0xFE - 256 = -2 -> screen +2.
+    const packet = feedAll(&assembler, &.{ 0x08 | 0x10 | 0x20, 0xFB, 0xFE }).?;
+    try testing.expectEqual(@as(i16, -5), packet.dx);
+    try testing.expectEqual(@as(i16, 2), packet.dy);
+}
+
+test "buttons decode from the header" {
+    var assembler = Assembler{};
+    const packet = feedAll(&assembler, &.{ 0x08 | 0x01 | 0x02, 0, 0 }).?;
+    try testing.expect(packet.left);
+    try testing.expect(packet.right);
+    try testing.expect(!packet.middle);
+}
+
+test "a byte with bit 3 clear at packet start is dropped" {
+    var assembler = Assembler{};
+    // The stray 0x02 cannot be a header; the following packet still decodes.
+    try testing.expectEqual(@as(?Packet, null), assembler.feed(0x02));
+    const packet = feedAll(&assembler, &.{ 0x09, 1, 0 }).?;
+    try testing.expect(packet.left);
+    try testing.expectEqual(@as(i16, 1), packet.dx);
+}
+
+test "protocol bytes at packet start are dropped" {
+    var assembler = Assembler{};
+    try testing.expectEqual(@as(?Packet, null), assembler.feed(0xFA)); // enable-reporting ACK
+    try testing.expectEqual(@as(?Packet, null), assembler.feed(0xAA)); // self-test passed
+    const packet = feedAll(&assembler, &.{ 0x08, 7, 0 }).?;
+    try testing.expectEqual(@as(i16, 7), packet.dx);
+}
+
+test "an overflowed packet is discarded whole" {
+    var assembler = Assembler{};
+    try testing.expectEqual(@as(?Packet, null), feedAll(&assembler, &.{ 0x08 | 0x40, 0xFF, 0xFF }));
+    // The assembler is back at packet start.
+    const packet = feedAll(&assembler, &.{ 0x08, 1, 1 }).?;
+    try testing.expectEqual(@as(i16, 1), packet.dx);
+}

system/drivers/ps2-bus/mouse.zig

@@ -0,0 +1,144 @@
+//! PS/2 Mouse Driver
+//!
+//! Spawned by the ps2-bus driver once the controller is initialized and the port
+//! has passed its interface test and device reset. The bus driver hands us our
+//! device HID as argv[1].
+//!
+//! Like the keyboard, this driver never touches the hardware: the 8042's ports
+//! and both port IRQs are owned by the ps2-bus driver (the auxiliary port's
+//! IRQ12 lives on the PNP0F13 node, which the bus claims alongside the
+//! controller). The driver **attaches** to the bus and receives every byte the
+//! mouse sends as a forwarded asynchronous message. The bytes assemble into
+//! three-byte packets, and each packet becomes input-protocol events:
+//!
+//!   packet -> button transitions -> button_down / button_up
+//!          -> movement            -> motion (dx/dy, screen convention)
+
+const std = @import("std");
+const runtime = @import("runtime");
+const ps2 = @import("ps2-library.zig");
+const mouse_packet = @import("mouse-packet.zig");
+const device = runtime.device;
+const ipc = runtime.ipc;
+const protocol = runtime.input_protocol;
+
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+/// Look up the ps2-bus service, retrying while the bus (which spawned us before
+/// registering) is still coming up.
+fn lookupBus() ?ipc.Handle {
+    var attempts: usize = 0;
+    while (attempts < 100) : (attempts += 1) {
+        if (ipc.lookup(.ps2_bus)) |handle| return handle;
+        runtime.system.sleep(50);
+    }
+    return null;
+}
+
+/// The protocol's pressed-button bitmask for a packet.
+fn buttonMask(packet: mouse_packet.Packet) u32 {
+    var mask: u32 = 0;
+    if (packet.left) mask |= protocol.mouse_button_left;
+    if (packet.right) mask |= protocol.mouse_button_right;
+    if (packet.middle) mask |= protocol.mouse_button_middle;
+    return mask;
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const hid = init.arguments.get(1).?;
+
+    if (hid.len == 0) {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: no HID argument\n");
+        return;
+    }
+    writeLine("/system/drivers/ps2-bus/mouse: starting for hid {s}\n", .{hid});
+
+    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: out of memory\n");
+        return;
+    };
+    if (device.findDeviceDescriptorByHid(buffer, hid) == null) {
+        writeLine("/system/drivers/ps2-bus/mouse: no device for hid {s}\n", .{hid});
+        return;
+    }
+
+    // Attach to the bus: hand it our endpoint, and it forwards every byte the
+    // mouse sends (it owns the controller; we own the decoding).
+    const bus = lookupBus() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: ps2-bus service unavailable\n");
+        return;
+    };
+    const endpoint = ipc.createIpcEndpoint() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: no endpoint\n");
+        return;
+    };
+    var attach = ps2.AttachRequest{ .device_type = @intFromEnum(ps2.DeviceType.mouse) };
+    var attach_reply: [@sizeOf(ps2.AttachReply)]u8 = undefined;
+    const attached = ipc.callCap(bus, std.mem.asBytes(&attach), &attach_reply, endpoint) catch {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: attach call failed\n");
+        return;
+    };
+    if (attached.len < @sizeOf(ps2.AttachReply) or
+        std.mem.bytesToValue(ps2.AttachReply, attach_reply[0..@sizeOf(ps2.AttachReply)]).status != @intFromEnum(ps2.AttachStatus.ok))
+    {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: attach refused\n");
+        return;
+    }
+
+    // Broadcast mouse events through the input service so programs can listen
+    // for them (docs/input.md).
+    var source = runtime.input.connectSource() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus/mouse: input service unavailable\n");
+        return;
+    };
+    _ = runtime.system.write("/system/drivers/ps2-bus/mouse: ok\n");
+
+    var assembler = mouse_packet.Assembler{};
+    var buttons: u32 = 0;
+    var receive: [@sizeOf(ps2.ForwardedByte)]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(endpoint, &.{}, &receive, null);
+        if (!got.isMessage() or got.len < @sizeOf(ps2.ForwardedByte)) continue;
+        const forwarded = std.mem.bytesToValue(ps2.ForwardedByte, receive[0..@sizeOf(ps2.ForwardedByte)]);
+
+        const packet = assembler.feed(@intCast(forwarded.byte & 0xFF)) orelse continue;
+        const new_buttons = buttonMask(packet);
+
+        // A button transition per changed button, carrying the new whole mask.
+        const changed = buttons ^ new_buttons;
+        for ([_]u32{ protocol.mouse_button_left, protocol.mouse_button_right, protocol.mouse_button_middle }) |button| {
+            if (changed & button == 0) continue;
+            const kind: protocol.MouseEventKind = if (new_buttons & button != 0) .button_down else .button_up;
+            _ = source.publishMouseEvent(.{
+                .kind = @intFromEnum(kind),
+                .button = button,
+                .dx = 0,
+                .dy = 0,
+                .scroll_x = 0,
+                .scroll_y = 0,
+                .buttons = new_buttons,
+            });
+        }
+        buttons = new_buttons;
+
+        if (packet.dx != 0 or packet.dy != 0) {
+            _ = source.publishMouseEvent(.{
+                .kind = @intFromEnum(protocol.MouseEventKind.motion),
+                .button = 0,
+                .dx = packet.dx,
+                .dy = packet.dy,
+                .scroll_x = 0,
+                .scroll_y = 0,
+                .buttons = new_buttons,
+            });
+        }
+    }
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start;
+}

system/drivers/ps2-bus/ps2-bus.zig

@@ -0,0 +1,315 @@
+//! The PS/2 Controller is located on the mainboard.
+//! In the early days the controller was a single chip (Intel 8042).
+//! As of today it is part of the Advanced Integrated Peripheral.
+//!
+//! It shows up in the device discovery as:
+//! KBD_ [acpi_device] hid=PNP0303 (PS/2 Keyboard)
+//!       - io_port 0x60 len 0x1
+//!        - io_port 0x64 len 0x1
+//!        - irq 0x1 len 0x1
+//!  MOU_ [acpi_device] hid=PNP0F13 (PS/2 Mouse)
+//!        - irq 0xc len 0x1
+const std = @import("std");
+const runtime = @import("runtime");
+const acpi_ids = @import("acpi-ids");
+const ps2 = @import("ps2-library.zig");
+const device = runtime.device;
+const ipc = runtime.ipc;
+
+/// Format one whole log line and emit it in a single `debug_write`, so output
+/// from the child drivers (which run concurrently) can never interleave with it.
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+/// Ask the device on `port` what it is, then spawn the matching driver from the
+/// initial-ramdisk, handing it the device's HID as argv[1]. The driver is chosen
+/// from what the device reports, not from the port number. Returns the identified
+/// type so the forwarding loop can route that port's bytes to the driver once it
+/// attaches, or null if nothing was spawned.
+fn spawnIdentifiedDriver(controller: ps2.Controller, port: ps2.Port) ?ps2.DeviceType {
+    const device_type = controller.identifyDevice(port) orelse {
+        writeLine("/system/drivers/ps2-bus: identify timed out on port {s}\n", .{@tagName(port)});
+        return null;
+    };
+    const driver_name = device_type.driverName() orelse {
+        writeLine("/system/drivers/ps2-bus: unrecognized device on port {s}\n", .{@tagName(port)});
+        return null;
+    };
+    const hid = device_type.hid() orelse "";
+    if (runtime.system.spawnWithArguments(driver_name, &.{hid}) != null) {
+        writeLine("/system/drivers/ps2-bus: port {s} is a {s}, spawned {s}\n", .{ @tagName(port), hid, driver_name });
+        return device_type;
+    }
+    writeLine("/system/drivers/ps2-bus: failed to spawn {s}\n", .{driver_name});
+    return null;
+}
+
+/// Resource index of the controller's IRQ (IRQ1) on the PNP0303 descriptor, found
+/// the way the ports are found in `Controller.init`.
+fn findInterruptResourceIndex(descriptor: device.DeviceDescriptor) ?u64 {
+    for (0..descriptor.resource_count) |index| {
+        if (descriptor.resources[index].kind == @intFromEnum(device.ResourceKind.irq)) return index;
+    }
+    return null;
+}
+
+/// Forwarding endpoints of the attached child drivers, indexed by `ps2.Port`.
+/// Written when a child's `AttachRequest` arrives, read on every forwarded byte.
+var port_endpoints = [_]?ipc.Handle{ null, null };
+
+/// Which device type each port identified as, so an attaching child (which knows
+/// its type, not its port) can be matched to the right port's byte stream.
+var port_device_types = [_]?ps2.DeviceType{ null, null };
+
+/// Handle a child driver's `AttachRequest`: record the endpoint capability it
+/// passed as the forwarding target for the port whose device matches its type.
+/// Writes an `AttachReply` into `out` and returns its length.
+fn handleAttach(message: []const u8, got: ipc.Received, out: []u8) usize {
+    const reply = struct {
+        fn write(buffer: []u8, status: ps2.AttachStatus) usize {
+            const header = ps2.AttachReply{ .status = @intFromEnum(status) };
+            @memcpy(buffer[0..@sizeOf(ps2.AttachReply)], std.mem.asBytes(&header));
+            return @sizeOf(ps2.AttachReply);
+        }
+    };
+
+    if (message.len < @sizeOf(ps2.AttachRequest)) return reply.write(out, .invalid_request);
+    const request = std.mem.bytesToValue(ps2.AttachRequest, message[0..@sizeOf(ps2.AttachRequest)]);
+    const endpoint = got.cap orelse return reply.write(out, .missing_endpoint);
+
+    for (&port_device_types, 0..) |maybe_type, port_index| {
+        const device_type = maybe_type orelse continue;
+        if (@intFromEnum(device_type) != request.device_type) continue;
+        port_endpoints[port_index] = endpoint;
+        writeLine("/system/drivers/ps2-bus: {s} driver attached\n", .{@tagName(device_type)});
+        return reply.write(out, .ok);
+    }
+    return reply.write(out, .no_such_device);
+}
+
+pub fn main() void {
+    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
+        _ = runtime.system.write("/system/drivers/ps2-bus: out of memory\n");
+        return;
+    };
+
+    var has_two_channels = false;
+    var maybe_controller: ?ps2.Controller = null;
+    var maybe_interrupt_index: ?u64 = null;
+    // The 8042's IO ports (0x60/0x64) are enumerated under the keyboard ACPI node
+    // (PNP0303), so we init the controller from that descriptor — but which device
+    // is on which port is decided later by identify, not by this HID.
+    const maybe_controller_device_descriptor = device.findDeviceDescriptorByHid(buffer, acpi_ids.HardwareId.ps2_keyboard.hid());
+    if (maybe_controller_device_descriptor) |controller_device_descriptor| {
+        _ = runtime.system.write("/system/drivers/ps2-bus: found PS/2 controller\n");
+        _ = runtime.system.write("/system/drivers/ps2-bus: initializing controller\n");
+
+        if (!device.claim(controller_device_descriptor.id)) {
+            _ = runtime.system.write("/system/drivers/ps2-bus: unable to claim controller \n");
+            return;
+        }
+
+        const controller = ps2.Controller.init(controller_device_descriptor) orelse {
+            _ = runtime.system.write("/system/drivers/ps2-bus: controller is missing its IO ports\n");
+            return;
+        };
+        maybe_controller = controller;
+        maybe_interrupt_index = findInterruptResourceIndex(controller_device_descriptor);
+
+        controller.disablePort(.one);
+        controller.disablePort(.two);
+        controller.flushOutputBuffer();
+
+        const current = controller.readConfigurationByte() orelse {
+            _ = runtime.system.write("/system/drivers/ps2-bus: controller configuration timed out\n");
+            return;
+        };
+
+        const update = current & ~(ps2.configuration_first_port_interrupt |
+            ps2.configuration_second_port_interrupt |
+            ps2.configuration_first_port_translation);
+
+        if (controller.writeConfigurationByte(update) == null) {
+            _ = runtime.system.write("/system/drivers/ps2-bus: controller configuration timed out\n");
+            return;
+        }
+
+        if (controller.performSelfTest()) |reply| {
+            if (reply != ps2.response_controller_test_passed) {
+                _ = runtime.system.write("/system/drivers/ps2-bus: perform controller self test failed\n");
+                return;
+            }
+        } else {
+            _ = runtime.system.write("/system/drivers/ps2-bus: controller self test timed out\n");
+            return;
+        }
+
+        has_two_channels = controller.hasTwoChannels() orelse {
+            _ = runtime.system.write("/system/drivers/ps2-bus: controller channels timed out\n");
+            return;
+        };
+
+        if (has_two_channels) {
+            _ = runtime.system.write("/system/drivers/ps2-bus: has two channels\n");
+            // keep the bus quiet until we have tested the ports and are ready to use them
+            controller.disablePort(.two);
+        } else {
+            _ = runtime.system.write("/system/drivers/ps2-bus: has one channel\n");
+        }
+
+        // interface tests: always test port 1, test port 2 only if it exists
+        const port_one_works = (controller.testPort(.one) orelse {
+            _ = runtime.system.write("/system/drivers/ps2-bus: port 1 test timed out\n");
+            return;
+        }) == ps2.response_port_test_passed;
+
+        var port_two_works = false;
+        if (has_two_channels) {
+            port_two_works = (controller.testPort(.two) orelse {
+                _ = runtime.system.write("/system/drivers/ps2-bus: port 2 test timed out\n");
+                return;
+            }) == ps2.response_port_test_passed;
+        }
+
+        if (!port_one_works and !port_two_works) {
+            _ = runtime.system.write("/system/drivers/ps2-bus: no usable ports\n");
+            return;
+        }
+
+        // Enable the working ports. Their interrupts stay off until IRQ1 is bound
+        // below — reset and identify use polled reads, which must never race the
+        // interrupt-driven drain loop for bytes.
+        controller.enablePort(.one);
+        if (port_two_works) controller.enablePort(.two);
+
+        // reset each working device; a failing device is logged but does not
+        // abort bring-up of the other one
+        if (port_one_works) {
+            if (controller.resetDevice(.one)) |passed| {
+                if (!passed) _ = runtime.system.write("/system/drivers/ps2-bus: port 1 device reset failed\n");
+            } else {
+                _ = runtime.system.write("/system/drivers/ps2-bus: port 1 device reset timed out\n");
+            }
+        }
+        if (port_two_works) {
+            if (controller.resetDevice(.two)) |passed| {
+                if (!passed) _ = runtime.system.write("/system/drivers/ps2-bus: port 2 device reset failed\n");
+            } else {
+                _ = runtime.system.write("/system/drivers/ps2-bus: port 2 device reset timed out\n");
+            }
+        }
+
+        // Identify the device on each working port and hand it off to the driver
+        // that matches what it reported — a port is not assumed to be a keyboard
+        // or a mouse by its number.
+        if (port_one_works) port_device_types[@intFromEnum(ps2.Port.one)] = spawnIdentifiedDriver(controller, .one);
+        if (port_two_works) port_device_types[@intFromEnum(ps2.Port.two)] = spawnIdentifiedDriver(controller, .two);
+    } else {
+        _ = runtime.system.write("/system/drivers/ps2-bus: no PS/2 controller found\n");
+        return;
+    }
+
+    const controller = maybe_controller.?;
+    const interrupt_index = maybe_interrupt_index orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus: controller is missing its IRQ\n");
+        return;
+    };
+
+    // The endpoint the child drivers attach to and IRQ1 wakes. Registered under a
+    // well-known id so the children can find it, the way input subscribers find
+    // the input service.
+    const endpoint = ipc.createIpcEndpoint() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus: no endpoint\n");
+        return;
+    };
+    if (!ipc.register(.ps2_bus, endpoint)) {
+        _ = runtime.system.write("/system/drivers/ps2-bus: register failed\n");
+        return;
+    }
+
+    // From here on, only the interrupt path reads the data port. Drop anything a
+    // device sent between enable-scanning and now, bind the IRQs, and only then
+    // let the controller raise them — an interrupt with nobody bound is lost.
+    controller.drainOutputBuffer();
+    if (!device.irqBind(controller.device_id, interrupt_index, endpoint)) {
+        _ = runtime.system.write("/system/drivers/ps2-bus: irq_bind failed\n");
+        return;
+    }
+
+    // Port 2's interrupt (IRQ12) is enumerated on the auxiliary device's own ACPI
+    // node (PNP0F13), not on the controller's — so if port 2 carries a device,
+    // claim that node too and route its IRQ to the same endpoint. The IRQ belongs
+    // to the *port*, whatever device identify found on it.
+    var maybe_auxiliary_interrupt: ?struct { device_id: u64, interrupt_index: u64, gsi: u64 } = null;
+    if (port_device_types[@intFromEnum(ps2.Port.two)] != null) {
+        if (device.findDeviceDescriptorByHid(buffer, acpi_ids.HardwareId.ps2_mouse.hid())) |descriptor| {
+            if (findInterruptResourceIndex(descriptor)) |auxiliary_index| {
+                if (device.claim(descriptor.id) and device.irqBind(descriptor.id, auxiliary_index, endpoint)) {
+                    maybe_auxiliary_interrupt = .{
+                        .device_id = descriptor.id,
+                        .interrupt_index = auxiliary_index,
+                        .gsi = descriptor.resources[auxiliary_index].start,
+                    };
+                } else {
+                    _ = runtime.system.write("/system/drivers/ps2-bus: auxiliary irq_bind failed\n");
+                }
+            }
+        }
+    }
+
+    var configuration = controller.readConfigurationByte() orelse {
+        _ = runtime.system.write("/system/drivers/ps2-bus: controller configuration timed out\n");
+        return;
+    };
+    if (port_device_types[@intFromEnum(ps2.Port.one)] != null) configuration |= ps2.Port.one.interruptBit();
+    if (maybe_auxiliary_interrupt != null) configuration |= ps2.Port.two.interruptBit();
+    _ = controller.writeConfigurationByte(configuration);
+
+    _ = runtime.system.write("/system/drivers/ps2-bus: ok\n");
+
+    // The forwarding loop: an IRQ1 notification drains the output buffer, routing
+    // each byte to the attached driver of the port it came from; a client message
+    // is a child driver's AttachRequest.
+    var reply_buffer: [@sizeOf(ps2.AttachReply)]u8 = undefined;
+    var reply_len: usize = 0;
+    var receive: [@sizeOf(ps2.AttachRequest)]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(endpoint, reply_buffer[0..reply_len], &receive, null);
+        if (got.isNotification()) {
+            reply_len = 0;
+            if (got.isMessage() or got.isChildExit()) continue; // nothing sends us these
+            while (true) {
+                const current_status = ps2.status(controller.device_id, controller.status_index);
+                if (current_status & ps2.status_output_buffer_full == 0) break;
+                const byte = device.ioRead(controller.device_id, controller.data_index, 0, 1) orelse break;
+                const port: ps2.Port = if (current_status & ps2.status_auxiliary_output != 0) .two else .one;
+                if (port_endpoints[@intFromEnum(port)]) |child| {
+                    const forwarded = ps2.ForwardedByte{ .port = @intFromEnum(port), .byte = byte };
+                    _ = ipc.send(child, std.mem.asBytes(&forwarded));
+                }
+                // An unattached port's byte is dropped — e.g. a keystroke before
+                // the keyboard driver has attached.
+            }
+            // Re-arm the line that woke us: the notification badge carries the
+            // GSI, and IRQ1 and IRQ12 are acked through different device claims.
+            if (maybe_auxiliary_interrupt) |auxiliary| {
+                if (got.source() == auxiliary.gsi) {
+                    _ = device.irqAck(auxiliary.device_id, auxiliary.interrupt_index);
+                } else {
+                    _ = device.irqAck(controller.device_id, interrupt_index);
+                }
+            } else {
+                _ = device.irqAck(controller.device_id, interrupt_index);
+            }
+            continue;
+        }
+        reply_len = handleAttach(receive[0..got.len], got, &reply_buffer);
+    }
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start;
+}

system/drivers/ps2-bus/ps2-library.zig

@@ -0,0 +1,485 @@
+//! shared definitions between the different PS/2 drivers
+const std = @import("std");
+const runtime = @import("runtime");
+const acpi_ids = @import("acpi-ids");
+const device = runtime.device;
+const system = runtime.system;
+
+/// PS-2 io ports:
+/// The PS/2 Controller itself uses 2 IO ports (usually, IO ports 0x60 and 0x64). Like many IO
+///  ports, reads and writes may access different internal registers.
+///
+/// Historical note: The PC-XT PPI had used port 0x61 to reset the keyboard interrupt request
+/// signal (among other unrelated functions). Port 0x61 has no keyboard related functions on AT and
+/// PS/2 compatibles.
+///
+/// The Data Port (typically IO Port 0x60) is used for reading data that was received from a PS/2
+/// device or from the PS/2 controller itself and writing data to a PS/2 device or to the PS/2
+/// controller itself.
+// Access type: Read/Write
+pub const dataPort = 0x60;
+// Access type: Read
+pub const statusRegisterPort = 0x64;
+// Access type: Write
+pub const CommandRegisterPort = 0x64;
+
+/// How long to poll the status register before giving up. PS/2 controller
+/// responses normally arrive within a few milliseconds.
+pub const default_wait_timeout_nanoseconds: u64 = 10_000_000; // 10 ms
+
+/// A PS/2 device reset (0xFF) runs the device's self-test (BAT), whose reply can
+/// take far longer than an ordinary controller response.
+pub const device_reset_timeout_nanoseconds: u64 = 750_000_000; // 750 ms
+
+/// PS/2 controller commands, written to the command register (port 0x64).
+pub const cmd_read_configuration_byte: u8 = 0x20; // read controller configuration byte (internal RAM byte 0)
+pub const cmd_write_configuration_byte: u8 = 0x60; // write controller configuration byte (internal RAM byte 0)
+pub const cmd_disable_second_port: u8 = 0xA7; // disable second PS/2 port (dual-channel controllers only)
+pub const cmd_enable_second_port: u8 = 0xA8; // enable second PS/2 port (dual-channel controllers only)
+pub const cmd_test_second_port: u8 = 0xA9; // test second PS/2 port
+pub const cmd_test_controller: u8 = 0xAA; // controller self-test
+pub const cmd_test_first_port: u8 = 0xAB; // test first PS/2 port
+pub const cmd_diagnostic_dump: u8 = 0xAC; // read all bytes of internal RAM
+pub const cmd_disable_first_port: u8 = 0xAD; // disable first PS/2 port
+pub const cmd_enable_first_port: u8 = 0xAE; // enable first PS/2 port
+pub const cmd_read_controller_input_port: u8 = 0xC0; // read controller input port
+pub const cmd_read_controller_output_port: u8 = 0xD0; // read controller output port
+pub const cmd_write_controller_output_port: u8 = 0xD1; // write next data byte to the controller output port
+pub const cmd_write_first_port_output: u8 = 0xD2; // write next data byte to the first port output buffer
+pub const cmd_write_second_port_output: u8 = 0xD3; // write next data byte to the second port output buffer
+pub const cmd_write_second_port_input: u8 = 0xD4; // write next data byte to the second port input buffer (to the mouse)
+pub const cmd_pulse_system_reset: u8 = 0xFE; // pulse output line 0 low: resets the CPU
+
+/// PS/2 status register bits (read from the status port, 0x64). Bit 4 is
+/// chipset-specific and intentionally omitted.
+pub const status_output_buffer_full: u8 = 1 << 0; // 1 = a byte is waiting to be read from the data port
+pub const status_input_buffer_full: u8 = 1 << 1; // 1 = the controller has not yet consumed the last write
+pub const status_system_flag: u8 = 1 << 2; // set once the controller passes POST
+pub const status_command_or_data: u8 = 1 << 3; // 1 = last write was a command, 0 = data
+/// Chipset-specific in the original spec, universal in practice on dual-channel
+/// controllers: set = the waiting byte came from the second port (the mouse).
+pub const status_auxiliary_output: u8 = 1 << 5;
+pub const status_timeout_error: u8 = 1 << 6; // 1 = time-out error
+pub const status_parity_error: u8 = 1 << 7; // 1 = parity error
+
+/// Controller configuration byte bits (internal RAM byte 0; read/written via 0x20/0x60).
+pub const configuration_first_port_interrupt: u8 = 1 << 0; // 1 = first port IRQ (IRQ1) enabled
+pub const configuration_second_port_interrupt: u8 = 1 << 1; // 1 = second port IRQ (IRQ12) enabled
+pub const configuration_system_flag: u8 = 1 << 2; // 1 = system passed POST
+pub const configuration_first_port_clock_disabled: u8 = 1 << 4; // 1 = first port clock disabled
+pub const configuration_second_port_clock_disabled: u8 = 1 << 5; // 1 = second port clock disabled
+pub const configuration_first_port_translation: u8 = 1 << 6; // 1 = first port scancode translation enabled
+
+/// Controller output port bits (read/written via 0xD0/0xD1).
+pub const output_port_system_reset: u8 = 1 << 0; // WARNING: keep this 1; writing 0 can lock the machine
+pub const output_port_a20_gate: u8 = 1 << 1; // A20 gate
+pub const output_port_second_port_clock: u8 = 1 << 2; // dual-channel controllers only
+pub const output_port_second_port_data: u8 = 1 << 3; // dual-channel controllers only
+pub const output_port_first_port_output_full: u8 = 1 << 4; // output buffer full from first port (IRQ1)
+pub const output_port_second_port_output_full: u8 = 1 << 5; // output buffer full from second port (IRQ12)
+pub const output_port_first_port_clock: u8 = 1 << 6; // first port clock
+pub const output_port_first_port_data: u8 = 1 << 7; // first port data
+
+/// Controller self-test (0xAA) result codes.
+pub const response_controller_test_passed: u8 = 0x55;
+pub const response_controller_test_failed: u8 = 0xFC;
+
+/// Port test (0xAB / 0xA9) result codes.
+pub const response_port_test_passed: u8 = 0x00;
+pub const response_port_test_clock_stuck_low: u8 = 0x01;
+pub const response_port_test_clock_stuck_high: u8 = 0x02;
+pub const response_port_test_data_stuck_low: u8 = 0x03;
+pub const response_port_test_data_stuck_high: u8 = 0x04;
+
+/// PS/2 device commands, written to the data port (0x60) to reach the attached device.
+pub const device_cmd_identify: u8 = 0xF2; // identify device
+pub const device_cmd_enable_scanning: u8 = 0xF4;
+pub const device_cmd_disable_scanning: u8 = 0xF5;
+pub const device_cmd_reset: u8 = 0xFF; // reset and run the device self-test (BAT)
+
+/// PS/2 device response bytes, read from the data port (0x60).
+pub const device_response_self_test_passed: u8 = 0xAA; // BAT succeeded after a reset
+pub const device_response_echo: u8 = 0xEE;
+pub const device_response_acknowledge: u8 = 0xFA; // ACK
+pub const device_response_self_test_failed_1: u8 = 0xFC; // BAT failure
+pub const device_response_self_test_failed_2: u8 = 0xFD; // BAT failure
+pub const device_response_resend: u8 = 0xFE; // ask the host to resend the last byte
+
+/// PS/2 device identify (0xF2) reply bytes. A keyboard returns a two-byte id
+/// beginning with 0xAB; a mouse returns a single-byte id (0x00/0x03/0x04); an
+/// ancient AT keyboard returns nothing at all.
+pub const identify_keyboard_mf2: u8 = 0xAB; // first byte of a MF2 keyboard id (a subtype byte follows)
+pub const identify_mouse_standard: u8 = 0x00;
+pub const identify_mouse_scroll: u8 = 0x03; // mouse with scroll wheel
+pub const identify_mouse_five_button: u8 = 0x04; // 5-button mouse
+
+fn waitReadable(id: u64, cmd_index: u64, wait_timeout_nanoseconds: u64) bool {
+    const deadline = system.clock() + wait_timeout_nanoseconds;
+    while (system.clock() < deadline) {
+        if (status(id, cmd_index) & status_output_buffer_full != 0) return true; // OBF set -> data ready
+    }
+    return false;
+}
+
+fn waitWritable(id: u64, cmd_index: u64, wait_timeout_nanoseconds: u64) bool {
+    const deadline = system.clock() + wait_timeout_nanoseconds;
+    while (system.clock() < deadline) {
+        if (status(id, cmd_index) & status_input_buffer_full == 0) return true; // IBF clear -> ok to write
+    }
+    return false; // timed out
+}
+
+pub fn status(id: u64, cmd_index: u64) u8 {
+    return @intCast(device.ioRead(id, cmd_index, 0, 1) orelse 0);
+}
+
+pub fn sendCommand(id: u64, cmd_index: u64, byte: u8, timeout_nanoseconds: u64) bool {
+    // wait IBF clear
+    if (!waitWritable(id, cmd_index, timeout_nanoseconds)) return false;
+    return device.ioWrite(id, cmd_index, 0, 1, byte);
+}
+
+pub fn readData(id: u64, status_index: u64, data_index: u64, timeout_nanoseconds: u64) ?u8 {
+    // OBF lives in the status register (0x64); wait for it there, then read the data port (0x60)
+    if (!waitReadable(id, status_index, timeout_nanoseconds)) return null;
+    return @intCast(device.ioRead(id, data_index, 0, 1) orelse 0);
+}
+
+pub fn writeData(id: u64, status_index: u64, data_index: u64, byte: u8, timeout_nanoseconds: u64) bool {
+    // IBF lives in the status register (0x64); wait for it to clear there, then write the data port
+    // (0x60)
+    if (!waitWritable(id, status_index, timeout_nanoseconds)) return false;
+    return device.ioWrite(id, data_index, 0, 1, byte);
+}
+
+pub const Port = enum(u2) {
+    one,
+    two,
+
+    /// Command register byte that disables this port.
+    fn disableCommand(self: Port) u8 {
+        return switch (self) {
+            .one => cmd_disable_first_port,
+            .two => cmd_disable_second_port,
+        };
+    }
+
+    /// Command register byte that enables this port (and its clock).
+    fn enableCommand(self: Port) u8 {
+        return switch (self) {
+            .one => cmd_enable_first_port,
+            .two => cmd_enable_second_port,
+        };
+    }
+
+    /// Command register byte that runs this port's interface test.
+    fn testCommand(self: Port) u8 {
+        return switch (self) {
+            .one => cmd_test_first_port,
+            .two => cmd_test_second_port,
+        };
+    }
+
+    /// Configuration-byte bit that, when set, disables this port's clock.
+    pub fn clockDisabledBit(self: Port) u8 {
+        return switch (self) {
+            .one => configuration_first_port_clock_disabled,
+            .two => configuration_second_port_clock_disabled,
+        };
+    }
+
+    /// Configuration-byte bit that, when set, enables this port's interrupt.
+    pub fn interruptBit(self: Port) u8 {
+        return switch (self) {
+            .one => configuration_first_port_interrupt,
+            .two => configuration_second_port_interrupt,
+        };
+    }
+
+    /// Command register byte that writes the next data byte into this port's
+    /// output buffer (makes a byte appear as if it came from the device).
+    pub fn writeOutputBufferCommand(self: Port) u8 {
+        return switch (self) {
+            .one => cmd_write_first_port_output,
+            .two => cmd_write_second_port_output,
+        };
+    }
+
+    /// Controller command that must prefix a byte destined for this port's
+    /// device. Port 1 is the default target of the data port, so it needs no
+    /// prefix (null); port 2 requires the "write second port input" command.
+    pub fn deviceInputCommand(self: Port) ?u8 {
+        return switch (self) {
+            .one => null,
+            .two => cmd_write_second_port_input,
+        };
+    }
+
+    /// Controller output-port bit driving this port's clock line.
+    pub fn outputPortClockBit(self: Port) u8 {
+        return switch (self) {
+            .one => output_port_first_port_clock,
+            .two => output_port_second_port_clock,
+        };
+    }
+
+    /// Controller output-port bit driving this port's data line.
+    pub fn outputPortDataBit(self: Port) u8 {
+        return switch (self) {
+            .one => output_port_first_port_data,
+            .two => output_port_second_port_data,
+        };
+    }
+
+    /// Controller output-port bit set when this port's output buffer is full
+    /// (wired to the port's IRQ line).
+    pub fn outputPortBufferFullBit(self: Port) u8 {
+        return switch (self) {
+            .one => output_port_first_port_output_full,
+            .two => output_port_second_port_output_full,
+        };
+    }
+};
+
+/// The kind of device attached to a port, as reported by the device itself in
+/// response to the identify command — not assumed from the port number. Fixed
+/// `u32` values because the type also travels in an `AttachRequest`.
+pub const DeviceType = enum(u32) {
+    keyboard = 0,
+    mouse = 1,
+    unknown = 2,
+
+    /// Initial-ramdisk name of the driver that serves this device type, or null
+    /// if we could not classify it.
+    pub fn driverName(self: DeviceType) ?[]const u8 {
+        return switch (self) {
+            .keyboard => "ps2-keyboard",
+            .mouse => "ps2-mouse",
+            .unknown => null,
+        };
+    }
+
+    /// Canonical ACPI HID for this device type, handed to the spawned driver as
+    /// its command-line argument, or null if we could not classify it.
+    pub fn hid(self: DeviceType) ?[]const u8 {
+        return switch (self) {
+            .keyboard => acpi_ids.HardwareId.ps2_keyboard.hid(),
+            .mouse => acpi_ids.HardwareId.ps2_mouse.hid(),
+            .unknown => null,
+        };
+    }
+};
+
+// --- the bus <-> child-driver forwarding protocol -----------------------------
+//
+// The 8042's ports and IRQ1 live on the PNP0303 node that only the ps2-bus driver
+// claims, so the child device drivers (ps2-keyboard, ps2-mouse) cannot read port
+// 0x60 themselves. Instead each child **attaches**: it calls the bus's well-known
+// `ps2_bus` endpoint with an `AttachRequest`, handing over its own endpoint as the
+// call's capability. From then on the bus forwards every byte the device sends as
+// a `ForwardedByte` via the asynchronous `ipc.send` — the IRQ path in the bus can
+// never block on a slow child, and the child never touches the controller.
+
+/// A child driver registering for its device's bytes. `device_type` is a
+/// `DeviceType` value; the child's receive endpoint travels as the call's
+/// capability (`send_cap`).
+pub const AttachRequest = extern struct {
+    device_type: u32,
+};
+
+/// How the bus answered an `AttachRequest` (`AttachReply.status`).
+pub const AttachStatus = enum(i32) {
+    ok = 0,
+    /// The request was malformed (too short to be an `AttachRequest`).
+    invalid_request = -1,
+    /// The call carried no endpoint capability to forward to.
+    missing_endpoint = -2,
+    /// No port identified a device of the requested type.
+    no_such_device = -3,
+};
+
+/// Reply to an `AttachRequest`. `status` is an `AttachStatus` value.
+pub const AttachReply = extern struct {
+    status: i32,
+    _padding: u32 = 0,
+};
+
+/// One raw byte read from the data port, forwarded to the attached child whose
+/// port it came from (routed by the status register's auxiliary-output bit).
+pub const ForwardedByte = extern struct {
+    /// The `Port` the byte came from, as `@intFromEnum`.
+    port: u32,
+    byte: u32,
+};
+
+/// A single PS/2 (8042) controller. Construct one with `Controller.init` and
+/// drive the controller through its methods; there is only ever one 8042 per
+/// machine, but holding the resolved resource indices in an instance keeps the
+/// call sites free of global state.
+pub const Controller = struct {
+    device_id: u64,
+    /// Resource index of the command/status port (0x64).
+    status_index: u64,
+    /// Resource index of the data port (0x60).
+    data_index: u64,
+
+    /// Resolve the controller's IO-port resource indices from its device
+    /// descriptor. Returns null if either the data or command/status port is
+    /// missing from the descriptor.
+    pub fn init(device_descriptor: device.DeviceDescriptor) ?Controller {
+        var data_index: ?u64 = null;
+        var status_index: ?u64 = null;
+
+        for (device_descriptor.resources, 0..device_descriptor.resource_count) |resource, resource_index| {
+            if (resource.kind != @intFromEnum(device.ResourceKind.io_port)) continue;
+            if (resource.start == dataPort) {
+                data_index = @intCast(resource_index);
+            } else if (resource.start == statusRegisterPort) {
+                status_index = @intCast(resource_index);
+            }
+        }
+
+        return .{
+            .device_id = device_descriptor.id,
+            .data_index = data_index orelse return null,
+            .status_index = status_index orelse return null,
+        };
+    }
+
+    pub fn disablePort(self: Controller, port: Port) void {
+        // port enable/disable are controller commands and go to the command register (0x64)
+        _ = sendCommand(self.device_id, self.status_index, port.disableCommand(), default_wait_timeout_nanoseconds);
+    }
+
+    pub fn enablePort(self: Controller, port: Port) void {
+        // enabling a port also starts its clock
+        _ = sendCommand(self.device_id, self.status_index, port.enableCommand(), default_wait_timeout_nanoseconds);
+    }
+
+    /// Run a port's interface test. Returns the controller's reply — compare it
+    /// to `response_port_test_passed` (0x00) — or null on timeout.
+    pub fn testPort(self: Controller, port: Port) ?u8 {
+        if (!sendCommand(self.device_id, self.status_index, port.testCommand(), default_wait_timeout_nanoseconds)) return null;
+        return readData(self.device_id, self.status_index, self.data_index, default_wait_timeout_nanoseconds);
+    }
+
+    pub fn flushOutputBuffer(self: Controller) void {
+        // flush any stale byte the controller buffered
+        _ = device.ioRead(self.device_id, self.data_index, 0, 1);
+    }
+
+    pub fn readConfigurationByte(self: Controller) ?u8 {
+        // ask the controller to place its configuration byte in the output buffer, then read it
+        if (!sendCommand(self.device_id, self.status_index, cmd_read_configuration_byte, default_wait_timeout_nanoseconds)) return null;
+        return readData(self.device_id, self.status_index, self.data_index, default_wait_timeout_nanoseconds);
+    }
+
+    pub fn writeConfigurationByte(self: Controller, update_byte: u8) ?u8 {
+        // command 0x60 makes the controller store the next data-port byte as its configuration byte
+        if (!sendCommand(self.device_id, self.status_index, cmd_write_configuration_byte, default_wait_timeout_nanoseconds)) return null;
+        if (!writeData(self.device_id, self.status_index, self.data_index, update_byte, default_wait_timeout_nanoseconds)) return null;
+        return update_byte;
+    }
+
+    /// Run the controller self-test. Returns the reply — compare it to
+    /// `response_controller_test_passed` (0x55) — or null on timeout.
+    pub fn performSelfTest(self: Controller) ?u8 {
+        if (!sendCommand(self.device_id, self.status_index, cmd_test_controller, default_wait_timeout_nanoseconds)) return null;
+        return readData(self.device_id, self.status_index, self.data_index, default_wait_timeout_nanoseconds);
+    }
+
+    /// Detect whether this is a dual-channel controller by temporarily enabling
+    /// port 2 and checking whether its clock turned on. Note: this leaves port 2
+    /// enabled; the caller should disable it again to keep the bus quiet until
+    /// device bring-up.
+    pub fn hasTwoChannels(self: Controller) ?bool {
+        self.enablePort(.two);
+        const configuration = self.readConfigurationByte() orelse return null;
+        return (configuration & Port.two.clockDisabledBit()) == 0;
+    }
+
+    /// Reset the device attached to `port` (device command 0xFF) and wait for
+    /// its power-on self-test (BAT) result. Returns true if the device both
+    /// acknowledged and passed, false if it reported a self-test failure, or
+    /// null on timeout. The BAT reply can be slow, so the response reads use
+    /// `device_reset_timeout_nanoseconds`.
+    pub fn resetDevice(self: Controller, port: Port) ?bool {
+        // A byte destined for port 2 must be prefixed with the "write to second
+        // port input buffer" controller command (0xD4); port 1 is the default.
+        if (port.deviceInputCommand()) |prefix| {
+            if (!sendCommand(self.device_id, self.status_index, prefix, default_wait_timeout_nanoseconds)) return null;
+        }
+        if (!writeData(self.device_id, self.status_index, self.data_index, device_cmd_reset, default_wait_timeout_nanoseconds)) return null;
+
+        // A successful reset yields both an ACK (0xFA) and a self-test-passed
+        // byte (0xAA). Their order is not guaranteed, so accept either ordering.
+        var saw_acknowledge = false;
+        var saw_self_test_passed = false;
+        var reads: u8 = 0;
+        while (reads < 2) : (reads += 1) {
+            const reply = readData(self.device_id, self.status_index, self.data_index, device_reset_timeout_nanoseconds) orelse return null;
+            switch (reply) {
+                device_response_acknowledge => saw_acknowledge = true,
+                device_response_self_test_passed => saw_self_test_passed = true,
+                device_response_self_test_failed_1, device_response_self_test_failed_2 => return false,
+                else => {},
+            }
+        }
+        return saw_acknowledge and saw_self_test_passed;
+    }
+
+    /// Send one command byte to the device on `port` (applying the port-2 prefix
+    /// as needed) and consume its acknowledgement. Returns true on ACK (0xFA),
+    /// false on any other reply, or null on timeout.
+    pub fn sendToDevice(self: Controller, port: Port, byte: u8) ?bool {
+        if (port.deviceInputCommand()) |prefix| {
+            if (!sendCommand(self.device_id, self.status_index, prefix, default_wait_timeout_nanoseconds)) return null;
+        }
+        if (!writeData(self.device_id, self.status_index, self.data_index, byte, default_wait_timeout_nanoseconds)) return null;
+        const reply = readData(self.device_id, self.status_index, self.data_index, default_wait_timeout_nanoseconds) orelse return null;
+        return reply == device_response_acknowledge;
+    }
+
+    /// Discard any bytes sitting in the output buffer (for example the device-id
+    /// byte a mouse emits after a reset) so they cannot be mistaken for the reply
+    /// to a subsequent command.
+    pub fn drainOutputBuffer(self: Controller) void {
+        var guard: u8 = 0;
+        while (guard < 16) : (guard += 1) {
+            if (status(self.device_id, self.status_index) & status_output_buffer_full == 0) return;
+            _ = device.ioRead(self.device_id, self.data_index, 0, 1);
+        }
+    }
+
+    /// Ask the device on `port` what it is (command 0xF2) and classify the reply.
+    /// Scanning is disabled around the query so a streaming device cannot inject
+    /// data bytes that look like the identifier. Returns the device type, or null
+    /// if the identify command itself timed out.
+    pub fn identifyDevice(self: Controller, port: Port) ?DeviceType {
+        // Clear any leftover bytes (e.g. a post-reset mouse id) before we start.
+        self.drainOutputBuffer();
+
+        // Stop the device reporting so its data can't be mistaken for the reply.
+        if (self.sendToDevice(port, device_cmd_disable_scanning) == null) return null;
+
+        if (self.sendToDevice(port, device_cmd_identify) == null) return null;
+
+        // After the ACK, the device sends 0, 1, or 2 identifier bytes.
+        const first = readData(self.device_id, self.status_index, self.data_index, default_wait_timeout_nanoseconds);
+        const device_type: DeviceType = if (first) |id| switch (id) {
+            identify_keyboard_mf2 => blk: {
+                // A MF2 keyboard sends a second subtype byte; consume and ignore it.
+                _ = readData(self.device_id, self.status_index, self.data_index, default_wait_timeout_nanoseconds);
+                break :blk .keyboard;
+            },
+            identify_mouse_standard, identify_mouse_scroll, identify_mouse_five_button => .mouse,
+            else => .unknown,
+        } else
+            // No identifier bytes at all is a legacy AT keyboard.
+            .keyboard;
+
+        // Resume scanning so the device works once its driver takes over.
+        _ = self.sendToDevice(port, device_cmd_enable_scanning);
+        return device_type;
+    }
+};

system/drivers/ps2-bus/scancode.zig

@@ -0,0 +1,389 @@
+//! PS/2 scancode set 2 → USB HID usage decoding, plus the keyboard state a driver
+//! needs on top of it (pressed keys, modifier tracking, caps-lock toggle).
+//!
+//! Set 2 is what a keyboard sends when the 8042's legacy set-1 translation is off —
+//! which is how ps2-bus.zig deliberately configures the controller. A key's **make**
+//! code is one byte (two with an `E0` prefix for the "extended" keys added after the
+//! original AT layout); its **break** code is the same code behind an `F0` prefix.
+//! Pause alone is an eight-byte `E1` sequence with no break.
+//!
+//! The output vocabulary is USB HID keyboard-page usages (a=4, enter=40, ...), the
+//! same numbering the input protocol's `Keycode` and the xkeyboard-config layout
+//! tables use — so a decoded usage indexes a layout directly.
+//!
+//! Everything here is pure (no imports beyond `std`, no IO), so it is host-testable:
+//! the tests at the bottom run under `zig build test`.
+
+const std = @import("std");
+
+// --- USB HID usages the state machine itself needs to recognize --------------
+
+pub const usage_caps_lock: u8 = 0x39;
+pub const usage_left_control: u8 = 0xE0;
+pub const usage_left_shift: u8 = 0xE1;
+pub const usage_left_alt: u8 = 0xE2;
+pub const usage_right_control: u8 = 0xE4;
+pub const usage_right_shift: u8 = 0xE5;
+pub const usage_right_alt: u8 = 0xE6; // AltGr — selects XKB level 3
+
+// --- scancode set 2 → HID usage tables ---------------------------------------
+
+/// Single-byte (non-`E0`) make codes. Zero means "no key" — protocol bytes (ACK,
+/// BAT results) and reserved codes land there and decode to nothing.
+pub const set2_base: [256]u8 = blk: {
+    var table = [_]u8{0} ** 256;
+    // function row
+    table[0x01] = 0x42; // F9
+    table[0x03] = 0x3E; // F5
+    table[0x04] = 0x3C; // F3
+    table[0x05] = 0x3A; // F1
+    table[0x06] = 0x3B; // F2
+    table[0x07] = 0x45; // F12
+    table[0x09] = 0x43; // F10
+    table[0x0A] = 0x41; // F8
+    table[0x0B] = 0x3F; // F6
+    table[0x0C] = 0x3D; // F4
+    table[0x78] = 0x44; // F11
+    table[0x83] = 0x40; // F7
+    // letters
+    table[0x1C] = 0x04; // A
+    table[0x32] = 0x05; // B
+    table[0x21] = 0x06; // C
+    table[0x23] = 0x07; // D
+    table[0x24] = 0x08; // E
+    table[0x2B] = 0x09; // F
+    table[0x34] = 0x0A; // G
+    table[0x33] = 0x0B; // H
+    table[0x43] = 0x0C; // I
+    table[0x3B] = 0x0D; // J
+    table[0x42] = 0x0E; // K
+    table[0x4B] = 0x0F; // L
+    table[0x3A] = 0x10; // M
+    table[0x31] = 0x11; // N
+    table[0x44] = 0x12; // O
+    table[0x4D] = 0x13; // P
+    table[0x15] = 0x14; // Q
+    table[0x2D] = 0x15; // R
+    table[0x1B] = 0x16; // S
+    table[0x2C] = 0x17; // T
+    table[0x3C] = 0x18; // U
+    table[0x2A] = 0x19; // V
+    table[0x1D] = 0x1A; // W
+    table[0x22] = 0x1B; // X
+    table[0x35] = 0x1C; // Y
+    table[0x1A] = 0x1D; // Z
+    // digit row
+    table[0x16] = 0x1E; // 1
+    table[0x1E] = 0x1F; // 2
+    table[0x26] = 0x20; // 3
+    table[0x25] = 0x21; // 4
+    table[0x2E] = 0x22; // 5
+    table[0x36] = 0x23; // 6
+    table[0x3D] = 0x24; // 7
+    table[0x3E] = 0x25; // 8
+    table[0x46] = 0x26; // 9
+    table[0x45] = 0x27; // 0
+    // control and whitespace
+    table[0x5A] = 0x28; // Enter
+    table[0x76] = 0x29; // Escape
+    table[0x66] = 0x2A; // Backspace
+    table[0x0D] = 0x2B; // Tab
+    table[0x29] = 0x2C; // Space
+    // punctuation
+    table[0x4E] = 0x2D; // - _
+    table[0x55] = 0x2E; // = +
+    table[0x54] = 0x2F; // [ {
+    table[0x5B] = 0x30; // ] }
+    table[0x5D] = 0x31; // \ | (non-US hash on ISO boards, same position)
+    table[0x4C] = 0x33; // ; :
+    table[0x52] = 0x34; // ' "
+    table[0x0E] = 0x35; // ` ~
+    table[0x41] = 0x36; // , <
+    table[0x49] = 0x37; // . >
+    table[0x4A] = 0x38; // / ?
+    table[0x61] = 0x64; // non-US backslash (the extra ISO key between shift and Z)
+    // locks
+    table[0x58] = usage_caps_lock;
+    table[0x77] = 0x53; // Num Lock
+    table[0x7E] = 0x47; // Scroll Lock
+    // keypad
+    table[0x7C] = 0x55; // keypad *
+    table[0x7B] = 0x56; // keypad -
+    table[0x79] = 0x57; // keypad +
+    table[0x69] = 0x59; // keypad 1
+    table[0x72] = 0x5A; // keypad 2
+    table[0x7A] = 0x5B; // keypad 3
+    table[0x6B] = 0x5C; // keypad 4
+    table[0x73] = 0x5D; // keypad 5
+    table[0x74] = 0x5E; // keypad 6
+    table[0x6C] = 0x5F; // keypad 7
+    table[0x75] = 0x60; // keypad 8
+    table[0x7D] = 0x61; // keypad 9
+    table[0x70] = 0x62; // keypad 0
+    table[0x71] = 0x63; // keypad .
+    // modifiers
+    table[0x14] = usage_left_control;
+    table[0x12] = usage_left_shift;
+    table[0x11] = usage_left_alt;
+    table[0x59] = usage_right_shift;
+    break :blk table;
+};
+
+/// `E0`-prefixed make codes. `E0 12` is the "fake shift" the keyboard wraps around
+/// Print Screen and navigation keys when a real shift is involved; it maps to zero
+/// here, so it decodes to nothing and only the real key comes through.
+pub const set2_extended: [256]u8 = blk: {
+    var table = [_]u8{0} ** 256;
+    table[0x11] = usage_right_alt;
+    table[0x14] = usage_right_control;
+    table[0x1F] = 0xE3; // left GUI
+    table[0x27] = 0xE7; // right GUI
+    table[0x2F] = 0x65; // application (menu)
+    table[0x7C] = 0x46; // Print Screen (arrives as E0 12 E0 7C; the E0 12 decodes to nothing)
+    table[0x4A] = 0x54; // keypad /
+    table[0x5A] = 0x58; // keypad Enter
+    table[0x70] = 0x49; // Insert
+    table[0x6C] = 0x4A; // Home
+    table[0x7D] = 0x4B; // Page Up
+    table[0x71] = 0x4C; // Delete
+    table[0x69] = 0x4D; // End
+    table[0x7A] = 0x4E; // Page Down
+    table[0x74] = 0x4F; // right arrow
+    table[0x6B] = 0x50; // left arrow
+    table[0x72] = 0x51; // down arrow
+    table[0x75] = 0x52; // up arrow
+    break :blk table;
+};
+
+// --- the byte-stream decoder --------------------------------------------------
+
+/// One decoded key transition: which key (as a USB HID usage) and whether this is
+/// a make (press or typematic repeat) or a break (release).
+pub const DecodedKey = struct {
+    usage: u8,
+    make: bool,
+};
+
+/// Turns the raw set-2 byte stream into `DecodedKey`s. Feed it every byte the
+/// keyboard sends; most bytes complete a key and return one, prefix bytes return
+/// null and arm the state machine for the next byte.
+pub const Decoder = struct {
+    const State = enum {
+        idle,
+        extended, // saw E0
+        break_prefix, // saw F0
+        extended_break, // saw E0 F0
+        pause_skip, // inside the 8-byte E1 Pause sequence
+    };
+
+    state: State = .idle,
+    /// Bytes still to swallow in `pause_skip`.
+    skip: u8 = 0,
+
+    /// The whole Pause make sequence is `E1 14 77 E1 F0 14 F0 77` — seven bytes
+    /// after the leading `E1`, and no break sequence ever follows.
+    const pause_bytes_after_e1: u8 = 7;
+
+    pub fn feed(self: *Decoder, byte: u8) ?DecodedKey {
+        switch (self.state) {
+            .idle => switch (byte) {
+                0xE0 => self.state = .extended,
+                0xF0 => self.state = .break_prefix,
+                0xE1 => {
+                    self.state = .pause_skip;
+                    self.skip = pause_bytes_after_e1;
+                },
+                // Anything else is a make code — or a protocol byte (0xFA ACK,
+                // 0xAA BAT-passed, 0xEE echo, ...), which the tables map to zero.
+                else => return decoded(set2_base[byte], true),
+            },
+            .extended => switch (byte) {
+                0xF0 => self.state = .extended_break,
+                else => {
+                    self.state = .idle;
+                    return decoded(set2_extended[byte], true);
+                },
+            },
+            .break_prefix => {
+                self.state = .idle;
+                return decoded(set2_base[byte], false);
+            },
+            .extended_break => {
+                self.state = .idle;
+                return decoded(set2_extended[byte], false);
+            },
+            .pause_skip => {
+                self.skip -= 1;
+                if (self.skip == 0) self.state = .idle;
+            },
+        }
+        return null;
+    }
+
+    fn decoded(usage: u8, make: bool) ?DecodedKey {
+        if (usage == 0) return null; // unmapped or a protocol byte
+        return .{ .usage = usage, .make = make };
+    }
+};
+
+// --- driver-side keyboard state -----------------------------------------------
+
+/// What a key transition did, plus the modifier state to stamp on the resulting
+/// events (snapshotted after the transition was applied).
+pub const Transition = struct {
+    pub const Action = enum {
+        pressed, // physical make of a key that was up
+        repeated, // typematic make of a key already down — no new key_down
+        released, // physical break
+    };
+    action: Action,
+    modifiers: ModifierSnapshot,
+};
+
+/// The modifier state at one instant, in both vocabularies a driver needs: the
+/// input protocol's coarse bits (shift/control/alt) and the level-selection
+/// inputs xkeyboard-config takes (shift, caps_lock, AltGr as level3).
+pub const ModifierSnapshot = struct {
+    shift: bool, // either shift held
+    control: bool, // either control held
+    alt: bool, // either alt held (including AltGr)
+    right_alt: bool, // AltGr specifically — the XKB level-3 selector
+    caps_lock: bool, // the toggle, not the key
+};
+
+/// Tracks which keys are physically down and the caps-lock toggle, and classifies
+/// each decoded transition. Pure state — no IO — so repeat detection and modifier
+/// snapshots are host-testable.
+pub const KeyboardState = struct {
+    /// One bit per HID usage: set while the key is physically down.
+    pressed: [32]u8 = [_]u8{0} ** 32,
+    caps_lock: bool = false,
+
+    pub fn apply(self: *KeyboardState, key: DecodedKey) Transition {
+        const already_down = self.isPressed(key.usage);
+        if (key.make) {
+            if (!already_down) {
+                self.setPressed(key.usage, true);
+                if (key.usage == usage_caps_lock) self.caps_lock = !self.caps_lock;
+            }
+            return .{
+                .action = if (already_down) .repeated else .pressed,
+                .modifiers = self.snapshot(),
+            };
+        }
+        self.setPressed(key.usage, false);
+        return .{ .action = .released, .modifiers = self.snapshot() };
+    }
+
+    pub fn isPressed(self: *const KeyboardState, usage: u8) bool {
+        return self.pressed[usage / 8] & (@as(u8, 1) << @intCast(usage % 8)) != 0;
+    }
+
+    fn setPressed(self: *KeyboardState, usage: u8, down: bool) void {
+        const bit = @as(u8, 1) << @intCast(usage % 8);
+        if (down) {
+            self.pressed[usage / 8] |= bit;
+        } else {
+            self.pressed[usage / 8] &= ~bit;
+        }
+    }
+
+    fn snapshot(self: *const KeyboardState) ModifierSnapshot {
+        const right_alt = self.isPressed(usage_right_alt);
+        return .{
+            .shift = self.isPressed(usage_left_shift) or self.isPressed(usage_right_shift),
+            .control = self.isPressed(usage_left_control) or self.isPressed(usage_right_control),
+            .alt = self.isPressed(usage_left_alt) or right_alt,
+            .right_alt = right_alt,
+            .caps_lock = self.caps_lock,
+        };
+    }
+};
+
+// --- tests (host-run via `zig build test`) ------------------------------------
+
+const testing = std.testing;
+
+/// Feed `bytes` and return the single DecodedKey they should produce (fails the
+/// test if they produce none or more than one).
+fn feedOne(decoder: *Decoder, bytes: []const u8) !DecodedKey {
+    var result: ?DecodedKey = null;
+    for (bytes) |byte| {
+        if (decoder.feed(byte)) |key| {
+            try testing.expect(result == null);
+            result = key;
+        }
+    }
+    return result orelse error.TestExpectedResult;
+}
+
+fn feedNone(decoder: *Decoder, bytes: []const u8) !void {
+    for (bytes) |byte| try testing.expectEqual(@as(?DecodedKey, null), decoder.feed(byte));
+}
+
+test "base make and break: A" {
+    var decoder = Decoder{};
+    try testing.expectEqual(DecodedKey{ .usage = 0x04, .make = true }, try feedOne(&decoder, &.{0x1C}));
+    try testing.expectEqual(DecodedKey{ .usage = 0x04, .make = false }, try feedOne(&decoder, &.{ 0xF0, 0x1C }));
+}
+
+test "extended make and break: right arrow" {
+    var decoder = Decoder{};
+    try testing.expectEqual(DecodedKey{ .usage = 0x4F, .make = true }, try feedOne(&decoder, &.{ 0xE0, 0x74 }));
+    try testing.expectEqual(DecodedKey{ .usage = 0x4F, .make = false }, try feedOne(&decoder, &.{ 0xE0, 0xF0, 0x74 }));
+}
+
+test "pause: the E1 sequence is consumed silently" {
+    var decoder = Decoder{};
+    try feedNone(&decoder, &.{ 0xE1, 0x14, 0x77, 0xE1, 0xF0, 0x14, 0xF0, 0x77 });
+    // The decoder is back in idle: an ordinary key still decodes.
+    try testing.expectEqual(DecodedKey{ .usage = 0x04, .make = true }, try feedOne(&decoder, &.{0x1C}));
+}
+
+test "protocol bytes decode to nothing" {
+    var decoder = Decoder{};
+    try feedNone(&decoder, &.{ 0xFA, 0xAA, 0xEE }); // ACK, BAT-passed, echo
+}
+
+test "print screen: the fake-shift E0 12 decodes to nothing" {
+    var decoder = Decoder{};
+    try feedNone(&decoder, &.{ 0xE0, 0x12 });
+    try testing.expectEqual(DecodedKey{ .usage = 0x46, .make = true }, try feedOne(&decoder, &.{ 0xE0, 0x7C }));
+}
+
+test "typematic repeat is classified, not re-pressed" {
+    var state = KeyboardState{};
+    const a = DecodedKey{ .usage = 0x04, .make = true };
+    try testing.expectEqual(Transition.Action.pressed, state.apply(a).action);
+    try testing.expectEqual(Transition.Action.repeated, state.apply(a).action);
+    try testing.expectEqual(Transition.Action.repeated, state.apply(a).action);
+    try testing.expectEqual(Transition.Action.released, state.apply(.{ .usage = 0x04, .make = false }).action);
+    try testing.expectEqual(Transition.Action.pressed, state.apply(a).action);
+}
+
+test "shift held shows in the snapshot of other keys" {
+    var state = KeyboardState{};
+    _ = state.apply(.{ .usage = usage_left_shift, .make = true });
+    const transition = state.apply(.{ .usage = 0x04, .make = true });
+    try testing.expect(transition.modifiers.shift);
+    try testing.expect(!transition.modifiers.control);
+    _ = state.apply(.{ .usage = usage_left_shift, .make = false });
+    _ = state.apply(.{ .usage = 0x04, .make = false });
+    try testing.expect(!state.apply(.{ .usage = 0x04, .make = true }).modifiers.shift);
+}
+
+test "right alt reports both alt and the level-3 selector" {
+    var state = KeyboardState{};
+    _ = state.apply(.{ .usage = usage_right_alt, .make = true });
+    const transition = state.apply(.{ .usage = 0x04, .make = true });
+    try testing.expect(transition.modifiers.alt);
+    try testing.expect(transition.modifiers.right_alt);
+}
+
+test "caps lock toggles on make, not on repeat or break" {
+    var state = KeyboardState{};
+    try testing.expect(state.apply(.{ .usage = usage_caps_lock, .make = true }).modifiers.caps_lock);
+    try testing.expect(state.apply(.{ .usage = usage_caps_lock, .make = true }).modifiers.caps_lock); // repeat
+    try testing.expect(state.apply(.{ .usage = usage_caps_lock, .make = false }).modifiers.caps_lock);
+    try testing.expect(!state.apply(.{ .usage = usage_caps_lock, .make = true }).modifiers.caps_lock); // second press: off
+}

system/drivers/usb-hid/hid-report.zig

@@ -0,0 +1,191 @@
+//! Pure decoders for USB HID **boot-protocol** reports — the simplified,
+//! fixed-format reports a boot keyboard and boot mouse send, the USB analog of
+//! the PS/2 scancode and mouse-packet decoders. No I/O: these turn report bytes
+//! into make/break transitions and motion, which the usb-hid drivers publish to
+//! the input service. Host-testable in isolation (like mouse-packet.zig).
+//!
+//! "Boot protocol" is a USB HID term (USB HID 1.11 §B) — the device reports in
+//! this fixed layout after SET_PROTOCOL(boot); it has nothing to do with system
+//! boot.
+
+const std = @import("std");
+
+// --- keyboard ---------------------------------------------------------------
+
+/// The 8-byte boot keyboard report: a modifier bitmap, a reserved byte, and up
+/// to six concurrently-pressed key usages.
+pub const KeyboardReport = extern struct {
+    modifiers: u8 = 0,
+    reserved: u8 = 0,
+    keys: [6]u8 = .{ 0, 0, 0, 0, 0, 0 },
+};
+
+// The modifier byte's bits (HID keyboard boot report).
+pub const modifier_left_control: u8 = 1 << 0;
+pub const modifier_left_shift: u8 = 1 << 1;
+pub const modifier_left_alt: u8 = 1 << 2;
+pub const modifier_left_gui: u8 = 1 << 3;
+pub const modifier_right_control: u8 = 1 << 4;
+pub const modifier_right_shift: u8 = 1 << 5;
+pub const modifier_right_alt: u8 = 1 << 6;
+pub const modifier_right_gui: u8 = 1 << 7;
+
+pub const TransitionKind = enum { pressed, released };
+
+/// One key going down or up. `usage` is a HID keyboard-page usage — modifier keys
+/// map to usages 224..231 — which is exactly the input protocol's `Keycode`.
+pub const Transition = struct { kind: TransitionKind, usage: u8 };
+
+// A report can change at most all 8 modifiers and all 6 keys at once.
+pub const max_transitions = 8 + 6;
+
+pub const Transitions = struct {
+    items: [max_transitions]Transition = undefined,
+    count: usize = 0,
+
+    fn add(self: *Transitions, transition: Transition) void {
+        if (self.count < self.items.len) {
+            self.items[self.count] = transition;
+            self.count += 1;
+        }
+    }
+
+    pub fn slice(self: *const Transitions) []const Transition {
+        return self.items[0..self.count];
+    }
+};
+
+/// Turns a stream of boot keyboard reports into make/break transitions by diffing
+/// each report against the last.
+pub const KeyboardDecoder = struct {
+    previous: KeyboardReport = .{},
+
+    pub fn feed(self: *KeyboardDecoder, current: KeyboardReport) Transitions {
+        var out = Transitions{};
+
+        // Rollover: 0x01 (ErrorRollOver) means more keys are held than the report
+        // can carry, so the key array is invalid. Emit nothing and keep the prior
+        // state (so the eventual releases still resolve against real keys).
+        for (current.keys) |key| {
+            if (key == 0x01) return out;
+        }
+
+        // Modifiers: one make/break per changed bit; modifier usages are 224..231.
+        const changed = current.modifiers ^ self.previous.modifiers;
+        var bit: u3 = 0;
+        while (true) : (bit += 1) {
+            const mask = @as(u8, 1) << bit;
+            if (changed & mask != 0) {
+                out.add(.{
+                    .kind = if (current.modifiers & mask != 0) .pressed else .released,
+                    .usage = 224 + @as(u8, bit),
+                });
+            }
+            if (bit == 7) break;
+        }
+
+        // Keys made: present now, absent before.
+        for (current.keys) |key| {
+            if (key != 0 and !contains(&self.previous.keys, key)) out.add(.{ .kind = .pressed, .usage = key });
+        }
+        // Keys broken: present before, absent now.
+        for (self.previous.keys) |key| {
+            if (key != 0 and !contains(&current.keys, key)) out.add(.{ .kind = .released, .usage = key });
+        }
+
+        self.previous = current;
+        return out;
+    }
+};
+
+fn contains(keys: *const [6]u8, value: u8) bool {
+    for (keys) |key| {
+        if (key == value) return true;
+    }
+    return false;
+}
+
+// --- mouse ------------------------------------------------------------------
+
+/// A decoded boot mouse report: the button bitmap and relative motion. The wheel
+/// byte is present only on 4-byte reports (QEMU's usb-mouse sends one).
+pub const MouseReport = struct {
+    buttons: u8 = 0,
+    dx: i8 = 0,
+    dy: i8 = 0,
+    wheel: i8 = 0,
+    has_wheel: bool = false,
+};
+
+pub const mouse_button_left: u8 = 1 << 0;
+pub const mouse_button_right: u8 = 1 << 1;
+pub const mouse_button_middle: u8 = 1 << 2;
+
+/// Parse a 3- or 4-byte boot mouse report. Note HID reports Y in screen
+/// convention (positive = down), so — unlike PS/2 — `dy` is NOT negated.
+pub fn parseMouse(bytes: []const u8) ?MouseReport {
+    if (bytes.len < 3) return null;
+    return .{
+        .buttons = bytes[0],
+        .dx = @bitCast(bytes[1]),
+        .dy = @bitCast(bytes[2]),
+        .wheel = if (bytes.len >= 4) @bitCast(bytes[3]) else 0,
+        .has_wheel = bytes.len >= 4,
+    };
+}
+
+// --- tests ------------------------------------------------------------------
+
+test "keyboard diff produces make and break transitions" {
+    var decoder = KeyboardDecoder{};
+
+    // Press 'a' (usage 4).
+    var t = decoder.feed(.{ .keys = .{ 4, 0, 0, 0, 0, 0 } });
+    try std.testing.expectEqual(@as(usize, 1), t.count);
+    try std.testing.expectEqual(TransitionKind.pressed, t.items[0].kind);
+    try std.testing.expectEqual(@as(u8, 4), t.items[0].usage);
+
+    // Hold 'a', press 'b' (usage 5): only 'b' is new.
+    t = decoder.feed(.{ .keys = .{ 4, 5, 0, 0, 0, 0 } });
+    try std.testing.expectEqual(@as(usize, 1), t.count);
+    try std.testing.expectEqual(@as(u8, 5), t.items[0].usage);
+
+    // Release everything: 'a' and 'b' both break.
+    t = decoder.feed(.{ .keys = .{ 0, 0, 0, 0, 0, 0 } });
+    try std.testing.expectEqual(@as(usize, 2), t.count);
+    try std.testing.expectEqual(TransitionKind.released, t.items[0].kind);
+
+    // Press Left Shift (modifier bit 1 -> usage 225).
+    t = decoder.feed(.{ .modifiers = modifier_left_shift });
+    try std.testing.expectEqual(@as(usize, 1), t.count);
+    try std.testing.expectEqual(@as(u8, 225), t.items[0].usage);
+    try std.testing.expectEqual(TransitionKind.pressed, t.items[0].kind);
+}
+
+test "rollover report is ignored but state is preserved" {
+    var decoder = KeyboardDecoder{};
+    _ = decoder.feed(.{ .keys = .{ 4, 0, 0, 0, 0, 0 } }); // press 'a'
+
+    const rollover = decoder.feed(.{ .keys = .{ 0x01, 0x01, 0x01, 0x01, 0x01, 0x01 } });
+    try std.testing.expectEqual(@as(usize, 0), rollover.count);
+
+    // 'a' is still considered down, so releasing all keys now breaks it.
+    const release = decoder.feed(.{ .keys = .{ 0, 0, 0, 0, 0, 0 } });
+    try std.testing.expectEqual(@as(usize, 1), release.count);
+    try std.testing.expectEqual(@as(u8, 4), release.items[0].usage);
+    try std.testing.expectEqual(TransitionKind.released, release.items[0].kind);
+}
+
+test "mouse report parses motion without inverting Y" {
+    const three = parseMouse(&.{ mouse_button_left, 5, 0xFB }).?; // dy = -5
+    try std.testing.expectEqual(mouse_button_left, three.buttons);
+    try std.testing.expectEqual(@as(i8, 5), three.dx);
+    try std.testing.expectEqual(@as(i8, -5), three.dy);
+    try std.testing.expect(!three.has_wheel);
+
+    const four = parseMouse(&.{ 0, 0, 0, 0xFF }).?; // wheel = -1
+    try std.testing.expect(four.has_wheel);
+    try std.testing.expectEqual(@as(i8, -1), four.wheel);
+
+    try std.testing.expect(parseMouse(&.{ 0, 0 }) == null); // too short
+}

system/drivers/usb-hid/keyboard.zig

@@ -0,0 +1,174 @@
+//! USB HID boot keyboard driver.
+//!
+//! Spawned by the device manager when the xHCI bus driver reports a HID / boot /
+//! keyboard interface (class 3, subclass 1, protocol 1); its assigned device id
+//! arrives as argv[1] and an optional layout name ("us", "gb", ...) as argv[2].
+//! It owns no hardware: it opens its device through the USB transfer protocol
+//! (`runtime.usb`), asks the device for the boot protocol, subscribes to its
+//! interrupt-IN endpoint, and turns each 8-byte boot report into input-protocol
+//! events, published to the input service — the USB analogue of ps2-bus/keyboard.
+//!
+//!   interrupt report -> hid-report diff -> key_down / key_up
+//!                                       -> xkeyboard-config -> character -> key_press
+//!
+//! Because a USB keyboard's usages ARE the input protocol's keycodes (both are
+//! HID keyboard page 0x07), the decode is nearly 1:1 — no scancode translation.
+
+const std = @import("std");
+const runtime = @import("runtime");
+const usb_abi = @import("usb-abi");
+const xkb = @import("xkeyboard-config");
+const hid = @import("hid-report.zig");
+const ipc = runtime.ipc;
+const process = runtime.process;
+const input_protocol = runtime.input_protocol;
+
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+// The modifier state a character lookup needs — derived from the report's
+// modifier byte, plus the driver-tracked caps-lock toggle.
+const ModifierSnapshot = struct {
+    shift: bool,
+    control: bool,
+    right_alt: bool,
+    caps_lock: bool,
+};
+
+/// The character a key produces under `modifiers`, or 0 for none — the layout
+/// lookup for printable keys, with ASCII control characters for the keys every
+/// consumer expects (Enter, Tab, Backspace, Escape), exactly as ps2-bus/keyboard.
+fn characterFor(layout: *const xkb.Layout, usage: u8, modifiers: ModifierSnapshot) u32 {
+    const mapping = xkb.map(layout, usage, .{
+        .shift = modifiers.shift,
+        .caps_lock = modifiers.caps_lock,
+        .level3 = modifiers.right_alt,
+        .control = modifiers.control,
+    });
+    if (mapping.character) |character| return character;
+    return switch (@as(input_protocol.Keycode, @enumFromInt(usage))) {
+        .enter, .keypad_enter => '\n',
+        .tab => '\t',
+        .backspace => 0x08,
+        .escape => 0x1B,
+        else => 0,
+    };
+}
+
+fn modifierWord(modifiers: u8) u32 {
+    var word: u32 = 0;
+    if (modifiers & (hid.modifier_left_shift | hid.modifier_right_shift) != 0) word |= input_protocol.modifier_shift;
+    if (modifiers & (hid.modifier_left_control | hid.modifier_right_control) != 0) word |= input_protocol.modifier_control;
+    if (modifiers & (hid.modifier_left_alt | hid.modifier_right_alt) != 0) word |= input_protocol.modifier_alt;
+    return word;
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const argument = init.arguments.get(1) orelse {
+        _ = runtime.system.write("/system/drivers/usb-hid/keyboard: missing device id (argv[1])\n");
+        return;
+    };
+    const device_id = std.fmt.parseInt(u64, argument, 10) catch {
+        writeLine("/system/drivers/usb-hid/keyboard: malformed device id '{s}'\n", .{argument});
+        return;
+    };
+    const layout = xkb.byName(init.arguments.get(2) orelse "us") orelse xkb.us;
+
+    // Hello the manager first (meet the spawn deadline), then open the device.
+    if (!runtime.usb.helloManager(device_id)) {
+        _ = runtime.system.write("/system/drivers/usb-hid/keyboard: hello to device manager failed\n");
+        return;
+    }
+    var device = runtime.usb.open(device_id) orelse {
+        writeLine("/system/drivers/usb-hid/keyboard: could not open device {d}\n", .{device_id});
+        return;
+    };
+    const endpoint = device.findEndpoint(runtime.usb.transfer_type_interrupt, true) orelse {
+        _ = runtime.system.write("/system/drivers/usb-hid/keyboard: no interrupt-IN endpoint\n");
+        return;
+    };
+
+    // Ask for the boot protocol and an indefinite idle (report only on change).
+    _ = device.controlOut(@bitCast(usb_abi.setProtocol(@enumFromInt(device.interface_number), .boot)));
+    _ = device.controlOut(@bitCast(usb_abi.setIdle(@enumFromInt(device.interface_number), 0, 0)));
+
+    if (!device.subscribeInterrupt(endpoint.address, endpoint.max_packet_size)) {
+        _ = runtime.system.write("/system/drivers/usb-hid/keyboard: interrupt subscribe failed\n");
+        return;
+    }
+
+    var source = runtime.input.connectSource() orelse {
+        _ = runtime.system.write("/system/drivers/usb-hid/keyboard: input service unavailable\n");
+        return;
+    };
+    _ = process.bindSignals(device.endpoint);
+    writeLine("/system/drivers/usb-hid/keyboard: ok (device {d}, interface {d}, layout {s})\n", .{ device_id, device.interface_number, layout.name });
+
+    var decoder = hid.KeyboardDecoder{};
+    var caps_lock = false;
+    var receive: [64]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(device.endpoint, &.{}, &receive, null);
+        if (!got.isNotification()) continue;
+        if (process.signalsFrom(got.badge)) |signals| {
+            if (signals.has(.terminate)) return;
+            continue;
+        }
+        if (!got.isMessage() or got.len < @sizeOf(runtime.usb.InterruptReport)) continue;
+
+        const message = std.mem.bytesToValue(runtime.usb.InterruptReport, receive[0..@sizeOf(runtime.usb.InterruptReport)]);
+        if (message.length < @sizeOf(hid.KeyboardReport)) continue;
+        const report = std.mem.bytesToValue(hid.KeyboardReport, message.data[0..@sizeOf(hid.KeyboardReport)]);
+        const transitions = decoder.feed(report);
+
+        // Caps Lock toggles on its own key-down (a stateful lock, not a modifier).
+        for (transitions.slice()) |transition| {
+            if (transition.kind == .pressed and @as(input_protocol.Keycode, @enumFromInt(transition.usage)) == .caps_lock) caps_lock = !caps_lock;
+        }
+
+        const modifiers = ModifierSnapshot{
+            .shift = report.modifiers & (hid.modifier_left_shift | hid.modifier_right_shift) != 0,
+            .control = report.modifiers & (hid.modifier_left_control | hid.modifier_right_control) != 0,
+            .right_alt = report.modifiers & hid.modifier_right_alt != 0,
+            .caps_lock = caps_lock,
+        };
+        const modifier_word = modifierWord(report.modifiers);
+
+        for (transitions.slice()) |transition| {
+            switch (transition.kind) {
+                .pressed => {
+                    _ = source.publishKeyboardEvent(.{
+                        .kind = @intFromEnum(input_protocol.EventKind.key_down),
+                        .keycode = transition.usage,
+                        .character = 0,
+                        .modifiers = modifier_word,
+                    });
+                    const character = characterFor(layout, transition.usage, modifiers);
+                    if (character != 0) {
+                        _ = source.publishKeyboardEvent(.{
+                            .kind = @intFromEnum(input_protocol.EventKind.key_press),
+                            .keycode = transition.usage,
+                            .character = character,
+                            .modifiers = modifier_word,
+                        });
+                    }
+                },
+                .released => {
+                    _ = source.publishKeyboardEvent(.{
+                        .kind = @intFromEnum(input_protocol.EventKind.key_up),
+                        .keycode = transition.usage,
+                        .character = 0,
+                        .modifiers = modifier_word,
+                    });
+                },
+            }
+        }
+    }
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start;
+}

system/drivers/usb-hid/mouse.zig

@@ -0,0 +1,140 @@
+//! USB HID boot mouse driver.
+//!
+//! Spawned by the device manager when the xHCI bus driver reports a HID / boot /
+//! mouse interface (class 3, subclass 1, protocol 2); its assigned device id
+//! arrives as argv[1]. Like the keyboard driver it owns no hardware: it opens its
+//! device through the USB transfer protocol (`runtime.usb`), asks for the boot
+//! protocol, subscribes to its interrupt-IN endpoint, and turns each 3- or 4-byte
+//! boot report into input-protocol mouse events published to the input service.
+//!
+//! Unlike PS/2, HID reports Y in screen convention (positive = down), so motion
+//! is passed straight through (the decode in hid-report.zig does not negate it).
+
+const std = @import("std");
+const runtime = @import("runtime");
+const usb_abi = @import("usb-abi");
+const hid = @import("hid-report.zig");
+const ipc = runtime.ipc;
+const process = runtime.process;
+const input_protocol = runtime.input_protocol;
+
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+// The current pressed-button bitmask in input-protocol terms.
+fn buttonMask(buttons: u8) u32 {
+    var mask: u32 = 0;
+    if (buttons & hid.mouse_button_left != 0) mask |= input_protocol.mouse_button_left;
+    if (buttons & hid.mouse_button_right != 0) mask |= input_protocol.mouse_button_right;
+    if (buttons & hid.mouse_button_middle != 0) mask |= input_protocol.mouse_button_middle;
+    return mask;
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const argument = init.arguments.get(1) orelse {
+        _ = runtime.system.write("/system/drivers/usb-hid/mouse: missing device id (argv[1])\n");
+        return;
+    };
+    const device_id = std.fmt.parseInt(u64, argument, 10) catch {
+        writeLine("/system/drivers/usb-hid/mouse: malformed device id '{s}'\n", .{argument});
+        return;
+    };
+
+    if (!runtime.usb.helloManager(device_id)) {
+        _ = runtime.system.write("/system/drivers/usb-hid/mouse: hello to device manager failed\n");
+        return;
+    }
+    var device = runtime.usb.open(device_id) orelse {
+        writeLine("/system/drivers/usb-hid/mouse: could not open device {d}\n", .{device_id});
+        return;
+    };
+    const endpoint = device.findEndpoint(runtime.usb.transfer_type_interrupt, true) orelse {
+        _ = runtime.system.write("/system/drivers/usb-hid/mouse: no interrupt-IN endpoint\n");
+        return;
+    };
+
+    _ = device.controlOut(@bitCast(usb_abi.setProtocol(@enumFromInt(device.interface_number), .boot)));
+
+    if (!device.subscribeInterrupt(endpoint.address, endpoint.max_packet_size)) {
+        _ = runtime.system.write("/system/drivers/usb-hid/mouse: interrupt subscribe failed\n");
+        return;
+    }
+
+    var source = runtime.input.connectSource() orelse {
+        _ = runtime.system.write("/system/drivers/usb-hid/mouse: input service unavailable\n");
+        return;
+    };
+    _ = process.bindSignals(device.endpoint);
+    writeLine("/system/drivers/usb-hid/mouse: ok (device {d}, interface {d})\n", .{ device_id, device.interface_number });
+
+    var previous_buttons: u8 = 0;
+    var receive: [64]u8 = undefined;
+    while (true) {
+        const got = ipc.replyWait(device.endpoint, &.{}, &receive, null);
+        if (!got.isNotification()) continue;
+        if (process.signalsFrom(got.badge)) |signals| {
+            if (signals.has(.terminate)) return;
+            continue;
+        }
+        if (!got.isMessage() or got.len < @sizeOf(runtime.usb.InterruptReport)) continue;
+
+        const message = std.mem.bytesToValue(runtime.usb.InterruptReport, receive[0..@sizeOf(runtime.usb.InterruptReport)]);
+        const length = @min(message.length, message.data.len);
+        const report = hid.parseMouse(message.data[0..length]) orelse continue;
+        const mask = buttonMask(report.buttons);
+
+        // Button transitions: one event per changed button bit.
+        const changed = report.buttons ^ previous_buttons;
+        inline for (.{
+            .{ hid.mouse_button_left, input_protocol.mouse_button_left },
+            .{ hid.mouse_button_right, input_protocol.mouse_button_right },
+            .{ hid.mouse_button_middle, input_protocol.mouse_button_middle },
+        }) |pair| {
+            if (changed & pair[0] != 0) {
+                _ = source.publishMouseEvent(.{
+                    .kind = @intFromEnum(if (report.buttons & pair[0] != 0) input_protocol.MouseEventKind.button_down else input_protocol.MouseEventKind.button_up),
+                    .button = pair[1],
+                    .dx = 0,
+                    .dy = 0,
+                    .scroll_x = 0,
+                    .scroll_y = 0,
+                    .buttons = mask,
+                });
+            }
+        }
+        previous_buttons = report.buttons;
+
+        // Relative motion (dy straight through — HID Y is already screen convention).
+        if (report.dx != 0 or report.dy != 0) {
+            _ = source.publishMouseEvent(.{
+                .kind = @intFromEnum(input_protocol.MouseEventKind.motion),
+                .button = 0,
+                .dx = report.dx,
+                .dy = report.dy,
+                .scroll_x = 0,
+                .scroll_y = 0,
+                .buttons = mask,
+            });
+        }
+
+        // Wheel (4-byte reports only): positive = scroll up.
+        if (report.has_wheel and report.wheel != 0) {
+            _ = source.publishMouseEvent(.{
+                .kind = @intFromEnum(input_protocol.MouseEventKind.scroll),
+                .button = 0,
+                .dx = 0,
+                .dy = 0,
+                .scroll_x = 0,
+                .scroll_y = report.wheel,
+                .buttons = mask,
+            });
+        }
+    }
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start;
+}

system/drivers/usb-storage/bulk-only-transport.zig

@@ -0,0 +1,73 @@
+//! USB Mass Storage Bulk-Only Transport (BOT) wire structures — the Command and
+//! Command Status Wrappers that bracket every command (USB MSC BOT §5). Pure data
+//! definitions, host-testable in isolation. The command inside the CBW is a SCSI
+//! CDB (see scsi.zig); the transport here just carries it and reports status.
+//!
+//! One command is three bulk transfers: CBW out, an optional data stage, CSW in.
+
+const std = @import("std");
+
+/// "USBC" — the signature at the head of every Command Block Wrapper.
+pub const cbw_signature: u32 = 0x43425355;
+/// "USBS" — the signature at the head of every Command Status Wrapper.
+pub const csw_signature: u32 = 0x53425355;
+
+/// CBW `flags`: set for a device-to-host (IN) data stage, clear for OUT.
+pub const flag_data_in: u8 = 0x80;
+
+/// The 31-byte Command Block Wrapper, sent on the bulk-OUT endpoint.
+pub const CommandBlockWrapper = extern struct {
+    signature: u32 align(1) = cbw_signature,
+    tag: u32 align(1),
+    data_transfer_length: u32 align(1),
+    flags: u8,
+    lun: u8,
+    cdb_length: u8,
+    cdb: [16]u8 = [_]u8{0} ** 16,
+};
+
+/// A device's answer to a command (the CSW `status` byte).
+pub const CommandStatus = enum(u8) {
+    passed = 0,
+    failed = 1,
+    phase_error = 2,
+    _,
+};
+
+/// The 13-byte Command Status Wrapper, read from the bulk-IN endpoint.
+pub const CommandStatusWrapper = extern struct {
+    signature: u32 align(1) = csw_signature,
+    tag: u32 align(1),
+    data_residue: u32 align(1),
+    status: u8,
+};
+
+comptime {
+    std.debug.assert(@sizeOf(CommandBlockWrapper) == 31);
+    std.debug.assert(@sizeOf(CommandStatusWrapper) == 13);
+}
+
+test "wrapper sizes and signatures match the specification" {
+    const cbw = CommandBlockWrapper{
+        .tag = 0x11223344,
+        .data_transfer_length = 512,
+        .flags = flag_data_in,
+        .lun = 0,
+        .cdb_length = 10,
+    };
+    const bytes = std.mem.asBytes(&cbw);
+    try std.testing.expectEqual(@as(usize, 31), bytes.len);
+    // "USBC" little-endian.
+    try std.testing.expectEqualSlices(u8, "USBC", bytes[0..4]);
+    try std.testing.expectEqual(flag_data_in, bytes[12]);
+
+    const csw = std.mem.bytesToValue(CommandStatusWrapper, &[_]u8{
+        0x55, 0x53, 0x42, 0x53, // "USBS"
+        0x44, 0x33, 0x22, 0x11, // tag
+        0x00, 0x00, 0x00, 0x00, // residue
+        0x00, // passed
+    });
+    try std.testing.expectEqual(csw_signature, csw.signature);
+    try std.testing.expectEqual(@as(u32, 0x11223344), csw.tag);
+    try std.testing.expectEqual(@as(u8, @intFromEnum(CommandStatus.passed)), csw.status);
+}

system/drivers/usb-storage/scsi.zig

@@ -0,0 +1,86 @@
+//! The SCSI command descriptor blocks a transparent-SCSI (subclass 0x06) mass
+//! storage device understands, and the parsers for what they return. Pure data —
+//! host-testable. These CDBs go inside a Bulk-Only-Transport CBW (see
+//! bulk-only-transport.zig).
+//!
+//! Every multi-byte SCSI field is **big-endian** — the opposite of the USB wire
+//! ABI — so the LBA and transfer-length encodings are the load-bearing detail.
+
+const std = @import("std");
+
+// SCSI operation codes.
+const op_test_unit_ready: u8 = 0x00;
+const op_request_sense: u8 = 0x03;
+const op_inquiry: u8 = 0x12;
+const op_read_capacity_10: u8 = 0x25;
+const op_read_10: u8 = 0x28;
+const op_write_10: u8 = 0x2A;
+
+/// INQUIRY: standard device data (36 bytes: peripheral type, removable, vendor
+/// and product strings).
+pub fn inquiry(allocation_length: u8) [6]u8 {
+    return .{ op_inquiry, 0, 0, 0, allocation_length, 0 };
+}
+
+/// TEST UNIT READY: no data; success (CSW passed) means the unit is ready.
+pub fn testUnitReady() [6]u8 {
+    return .{ op_test_unit_ready, 0, 0, 0, 0, 0 };
+}
+
+/// REQUEST SENSE: 18 bytes of sense data (sense key + ASC/ASCQ) explaining the
+/// previous failure.
+pub fn requestSense(allocation_length: u8) [6]u8 {
+    return .{ op_request_sense, 0, 0, 0, allocation_length, 0 };
+}
+
+/// READ CAPACITY(10): 8 bytes back — the last LBA and the block size, both u32
+/// big-endian. Block count is last_lba + 1.
+pub fn readCapacity10() [10]u8 {
+    return .{ op_read_capacity_10, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
+}
+
+/// READ(10): read `blocks` logical blocks starting at `lba` into the data stage.
+pub fn read10(lba: u32, blocks: u16) [10]u8 {
+    var cdb = [_]u8{0} ** 10;
+    cdb[0] = op_read_10;
+    std.mem.writeInt(u32, cdb[2..6], lba, .big);
+    std.mem.writeInt(u16, cdb[7..9], blocks, .big);
+    return cdb;
+}
+
+/// WRITE(10): write `blocks` logical blocks starting at `lba` from the data stage.
+pub fn write10(lba: u32, blocks: u16) [10]u8 {
+    var cdb = [_]u8{0} ** 10;
+    cdb[0] = op_write_10;
+    std.mem.writeInt(u32, cdb[2..6], lba, .big);
+    std.mem.writeInt(u16, cdb[7..9], blocks, .big);
+    return cdb;
+}
+
+/// Decode an 8-byte READ CAPACITY(10) reply.
+pub fn parseCapacity(bytes: [8]u8) struct { last_lba: u32, block_size: u32 } {
+    return .{
+        .last_lba = std.mem.readInt(u32, bytes[0..4], .big),
+        .block_size = std.mem.readInt(u32, bytes[4..8], .big),
+    };
+}
+
+test "read/write CDBs encode the LBA and length big-endian" {
+    const read = read10(0x01020304, 8);
+    try std.testing.expectEqualSlices(u8, &.{ 0x28, 0x00, 0x01, 0x02, 0x03, 0x04, 0x00, 0x00, 0x08, 0x00 }, &read);
+
+    const write = write10(0xAABBCCDD, 1);
+    try std.testing.expectEqualSlices(u8, &.{ 0x2A, 0x00, 0xAA, 0xBB, 0xCC, 0xDD, 0x00, 0x00, 0x01, 0x00 }, &write);
+
+    try std.testing.expectEqual(@as(u8, 0x25), readCapacity10()[0]);
+    try std.testing.expectEqual(@as(u8, 0x12), inquiry(36)[0]);
+    try std.testing.expectEqual(@as(u8, 36), inquiry(36)[4]);
+    try std.testing.expectEqual(@as(u8, 0x00), testUnitReady()[0]);
+}
+
+test "read capacity parses last LBA and block size" {
+    // last_lba = 0x0003FFFF (262144 blocks), block_size = 512.
+    const capacity = parseCapacity(.{ 0x00, 0x03, 0xFF, 0xFF, 0x00, 0x00, 0x02, 0x00 });
+    try std.testing.expectEqual(@as(u32, 0x0003FFFF), capacity.last_lba);
+    try std.testing.expectEqual(@as(u32, 512), capacity.block_size);
+}

system/drivers/usb-storage/usb-storage.zig

@@ -0,0 +1,179 @@
+//! USB mass-storage class driver (Bulk-Only Transport + transparent SCSI).
+//!
+//! Spawned by the device manager when the xHCI bus driver reports a mass-storage
+//! / SCSI / bulk-only interface (class 8, subclass 6, protocol 0x50); its device
+//! id arrives as argv[1]. It owns no hardware: it opens its device through the
+//! USB transfer protocol (`runtime.usb`), then drives it with the BOT command
+//! cycle — CBW out, an optional data stage, CSW in — carrying SCSI commands
+//! (READ CAPACITY, READ(10), WRITE(10)). Upward it is a block device: it serves
+//! the block protocol under `.block`, the storage a FAT filesystem sits on.
+//!
+//! Block data never crosses IPC: read/write name a caller-owned DMA buffer by
+//! physical address, which the data stage DMAs straight to/from.
+
+const std = @import("std");
+const runtime = @import("runtime");
+const scsi = @import("scsi.zig");
+const bot = @import("bulk-only-transport.zig");
+const block_protocol = @import("block-protocol");
+const dma = runtime.dma;
+
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+var device_id: u64 = 0;
+var device: runtime.usb.Device = undefined;
+var bulk_in: runtime.usb.Endpoint = undefined;
+var bulk_out: runtime.usb.Endpoint = undefined;
+
+// DMA buffers for the transport: the 31-byte CBW, the 13-byte CSW, and a page
+// for the small command data (INQUIRY / READ CAPACITY / the self-check sector).
+var command_wrapper: dma.Region = undefined;
+var status_wrapper: dma.Region = undefined;
+var command_data: dma.Region = undefined;
+
+var next_tag: u32 = 1;
+var block_size: u32 = 512;
+var block_count: u64 = 0;
+
+/// One Bulk-Only-Transport command: send the CBW, run the data stage (to/from
+/// `data_physical`), read and validate the CSW. Returns true on a passed status.
+fn transact(cdb: []const u8, direction_in: bool, data_physical: u64, data_length: u32) bool {
+    const tag = next_tag;
+    next_tag +%= 1;
+
+    const wrapper: *bot.CommandBlockWrapper = @ptrFromInt(command_wrapper.virtual);
+    wrapper.* = .{
+        .tag = tag,
+        .data_transfer_length = data_length,
+        .flags = if (direction_in) bot.flag_data_in else 0,
+        .lun = 0,
+        .cdb_length = @intCast(cdb.len),
+    };
+    @memcpy(wrapper.cdb[0..cdb.len], cdb);
+
+    if (device.bulk(bulk_out.address, command_wrapper.physical, @sizeOf(bot.CommandBlockWrapper)) == null) return false;
+    if (data_length > 0) {
+        const endpoint = if (direction_in) bulk_in.address else bulk_out.address;
+        if (device.bulk(endpoint, data_physical, data_length) == null) return false;
+    }
+    if (device.bulk(bulk_in.address, status_wrapper.physical, @sizeOf(bot.CommandStatusWrapper)) == null) return false;
+
+    const status: *const bot.CommandStatusWrapper = @ptrFromInt(status_wrapper.virtual);
+    if (status.signature != bot.csw_signature or status.tag != tag) return false;
+    return status.status == @intFromEnum(bot.CommandStatus.passed);
+}
+
+fn initialise(endpoint: runtime.ipc.Handle) bool {
+    _ = endpoint;
+    if (!runtime.usb.helloManager(device_id)) {
+        _ = runtime.system.write("/system/drivers/usb-storage: hello to device manager failed\n");
+        return false;
+    }
+    device = runtime.usb.open(device_id) orelse {
+        writeLine("/system/drivers/usb-storage: could not open device {d}\n", .{device_id});
+        return false;
+    };
+    bulk_in = device.findEndpoint(runtime.usb.transfer_type_bulk, true) orelse {
+        _ = runtime.system.write("/system/drivers/usb-storage: no bulk-IN endpoint\n");
+        return false;
+    };
+    bulk_out = device.findEndpoint(runtime.usb.transfer_type_bulk, false) orelse {
+        _ = runtime.system.write("/system/drivers/usb-storage: no bulk-OUT endpoint\n");
+        return false;
+    };
+    command_wrapper = dma.alloc(4096, dma.coherent) orelse return false;
+    status_wrapper = dma.alloc(4096, dma.coherent) orelse return false;
+    command_data = dma.alloc(4096, dma.coherent) orelse return false;
+
+    // Bring the LUN up: wait for it to be ready (clearing the initial unit-attention
+    // with REQUEST SENSE), identify it, and read its capacity.
+    var tries: u32 = 0;
+    while (tries < 10) : (tries += 1) {
+        const ready = scsi.testUnitReady();
+        if (transact(&ready, false, 0, 0)) break;
+        const sense = scsi.requestSense(18);
+        _ = transact(&sense, true, command_data.physical, 18);
+        runtime.system.sleep(50);
+    }
+    const inquiry = scsi.inquiry(36);
+    _ = transact(&inquiry, true, command_data.physical, 36);
+
+    const capacity_command = scsi.readCapacity10();
+    if (!transact(&capacity_command, true, command_data.physical, 8)) {
+        _ = runtime.system.write("/system/drivers/usb-storage: READ CAPACITY failed\n");
+        return false;
+    }
+    var capacity_bytes: [8]u8 = undefined;
+    const capacity_source: [*]const u8 = @ptrFromInt(command_data.virtual);
+    @memcpy(&capacity_bytes, capacity_source[0..8]);
+    const capacity = scsi.parseCapacity(capacity_bytes);
+    block_size = capacity.block_size;
+    block_count = @as(u64, capacity.last_lba) + 1;
+    writeLine("/system/drivers/usb-storage: ready ({d} blocks x {d} bytes)\n", .{ block_count, block_size });
+
+    // Self-check: read block 0 and log its trailing signature (0x55AA for a boot
+    // sector) — proof READ(10) works end to end over the bulk path.
+    const read0 = scsi.read10(0, 1);
+    if (block_size <= 4096 and transact(&read0, true, command_data.physical, block_size)) {
+        const sector: [*]const u8 = @ptrFromInt(command_data.virtual);
+        writeLine("/system/drivers/usb-storage: block 0 signature 0x{x:0>2}{x:0>2}\n", .{ sector[510], sector[511] });
+    }
+    return true;
+}
+
+/// Serve the block protocol: geometry, and whole-block read/write to/from the
+/// caller's DMA buffer (named by physical address).
+fn onMessage(message: []const u8, reply: []u8, sender: u32, capability: ?runtime.ipc.Handle) usize {
+    _ = sender;
+    _ = capability;
+    if (message.len < block_protocol.request_size) return 0;
+    const request = std.mem.bytesToValue(block_protocol.Request, message[0..block_protocol.request_size]);
+    switch (request.operation) {
+        @intFromEnum(block_protocol.Operation.geometry) => {
+            return writeReply(reply, .{ .status = 0, .block_size = block_size, .block_count = block_count });
+        },
+        @intFromEnum(block_protocol.Operation.read) => {
+            const count: u16 = @intCast(request.count);
+            const cdb = scsi.read10(@intCast(request.lba), count);
+            const ok = transact(&cdb, true, request.physical, request.count * block_size);
+            return writeReply(reply, .{ .status = if (ok) 0 else -1, .block_size = block_size, .block_count = if (ok) request.count else 0 });
+        },
+        @intFromEnum(block_protocol.Operation.write) => {
+            const count: u16 = @intCast(request.count);
+            const cdb = scsi.write10(@intCast(request.lba), count);
+            const ok = transact(&cdb, false, request.physical, request.count * block_size);
+            return writeReply(reply, .{ .status = if (ok) 0 else -1, .block_size = block_size, .block_count = if (ok) request.count else 0 });
+        },
+        else => return 0,
+    }
+}
+
+fn writeReply(reply: []u8, value: block_protocol.Reply) usize {
+    const bytes = std.mem.asBytes(&value);
+    @memcpy(reply[0..bytes.len], bytes);
+    return bytes.len;
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const argument = init.arguments.get(1) orelse {
+        _ = runtime.system.write("/system/drivers/usb-storage: missing device id (argv[1])\n");
+        return;
+    };
+    device_id = std.fmt.parseInt(u64, argument, 10) catch {
+        writeLine("/system/drivers/usb-storage: malformed device id '{s}'\n", .{argument});
+        return;
+    };
+    runtime.service.run(block_protocol.message_maximum, .{
+        .service = .block,
+        .init = initialise,
+        .on_message = onMessage,
+    });
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start;
+}

system/drivers/usb-xhci-bus/usb-transfer-protocol.zig

@@ -0,0 +1,159 @@
+//! The USB transfer protocol: what a USB class driver (a keyboard, mouse, or
+//! mass-storage driver) says to the xHCI bus driver over its well-known
+//! `.usb_bus` endpoint to drive its device. The class driver owns no hardware —
+//! it reaches its device entirely through these messages, the way a PS/2 keyboard
+//! driver reaches the 8042 through the ps2-bus. Extern-struct messages tagged by
+//! `Operation`, the vfs-protocol / device-manager-protocol pattern.
+//!
+//! The shape:
+//!   - **open** (a capability-passing `ipc.callCap`): the class driver hands over
+//!     its own endpoint (for asynchronous interrupt reports) and its assigned
+//!     device id, and receives a `device_token` plus its interface's endpoints.
+//!   - **control / bulk** (synchronous `ipc.call`): one transfer, answered when
+//!     it completes. Control data travels inline (descriptors, HID/MSC class
+//!     requests are all small); bulk data travels by **physical address** — the
+//!     class driver's own `dma_alloc`'d buffer — so a 512-byte sector never has
+//!     to cross the 256-byte IPC boundary.
+//!   - **interrupt_subscribe** (synchronous): arm periodic IN polling of an
+//!     interrupt endpoint; each report the device produces is then pushed to the
+//!     class driver's endpoint as an asynchronous `InterruptReport` (`ipc.send`),
+//!     exactly how the input service delivers events.
+//!
+//! Single controller assumption: one `.usb_bus` singleton serves QEMU's one xHCI.
+//! A multi-controller machine would need a per-controller endpoint (the device
+//! manager handing each class driver the right one); noted, not built.
+
+/// Fits one synchronous IPC message (kernel MESSAGE_MAXIMUM).
+pub const message_maximum: usize = 256;
+
+/// The largest inline control-transfer payload. Sized so a whole message
+/// (header + data) stays under `message_maximum`: descriptors and HID/MSC class
+/// requests are all far smaller.
+pub const max_inline_data: usize = 200;
+
+/// The largest interrupt report pushed asynchronously. Sized so `InterruptReport`
+/// fits an `ipc_send` payload slot (POST_MAXIMUM = 64): boot keyboard reports are
+/// 8 bytes, boot mouse reports 3–4.
+pub const max_report_data: usize = 48;
+
+/// Endpoints per interface reported back in an open reply (a boot HID interface
+/// has one interrupt endpoint, a mass-storage interface two bulk endpoints).
+pub const max_reported_endpoints: usize = 4;
+
+pub const Operation = enum(u32) {
+    open = 0,
+    control = 1,
+    interrupt_subscribe = 2,
+    bulk = 3,
+};
+
+/// The endpoint facts a class driver needs, lifted from the endpoint descriptor
+/// the bus driver already parsed during enumeration.
+pub const Endpoint = extern struct {
+    /// EndpointDescriptor address: direction in bit 7, number in bits 3:0.
+    address: u8,
+    /// 0 control, 1 isochronous, 2 bulk, 3 interrupt.
+    transfer_type: u8,
+    max_packet_size: u16,
+    interval: u8,
+    reserved: [3]u8 = .{ 0, 0, 0 },
+};
+
+/// open: the class driver's receive endpoint rides as the call's capability, and
+/// `device_id` is the interface's assigned id (its argv[1]).
+pub const OpenRequest = extern struct {
+    operation: u32 = @intFromEnum(Operation.open),
+    reserved: u32 = 0,
+    device_id: u64,
+};
+
+/// The answer to open: a token scoping every later request to this device, the
+/// interface's class triple (a sanity check), and its endpoints.
+pub const OpenReply = extern struct {
+    status: i32,
+    endpoint_count: u32,
+    device_token: u64,
+    interface_class: u8,
+    interface_subclass: u8,
+    interface_protocol: u8,
+    interface_number: u8,
+    reserved2: u32 = 0,
+    endpoints: [max_reported_endpoints]Endpoint = [_]Endpoint{.{ .address = 0, .transfer_type = 0, .max_packet_size = 0, .interval = 0 }} ** max_reported_endpoints,
+};
+
+/// control: one EP0 control transfer. `setup` is a bit-cast `usb_abi.Request`.
+/// For an OUT transfer `data[0..data_length]` is sent; for an IN transfer the
+/// reply carries up to `data_length` bytes back.
+pub const ControlRequest = extern struct {
+    operation: u32 = @intFromEnum(Operation.control),
+    reserved: u32 = 0,
+    device_token: u64,
+    setup: [8]u8,
+    direction_in: u8, // 1 = device-to-host (IN), 0 = host-to-device (OUT)
+    reserved2: u8 = 0,
+    data_length: u16,
+    reserved3: u32 = 0,
+    data: [max_inline_data]u8 = [_]u8{0} ** max_inline_data,
+};
+
+pub const ControlReply = extern struct {
+    status: i32, // 0 success, negative on failure/stall
+    actual_length: u32,
+    data: [max_inline_data]u8 = [_]u8{0} ** max_inline_data,
+};
+
+/// interrupt_subscribe: begin periodic IN polling of an interrupt endpoint. Each
+/// report the device returns is pushed to the caller's endpoint (handed over at
+/// open) as an asynchronous `InterruptReport`.
+pub const InterruptSubscribeRequest = extern struct {
+    operation: u32 = @intFromEnum(Operation.interrupt_subscribe),
+    reserved: u32 = 0,
+    device_token: u64,
+    endpoint_address: u8,
+    reserved2: u8 = 0,
+    max_length: u16, // bytes to request per poll (the endpoint's max packet size)
+};
+
+pub const InterruptSubscribeReply = extern struct {
+    status: i32,
+    reserved: u32 = 0,
+};
+
+/// bulk: one bulk IN or OUT transfer. `physical_address` is the class driver's own
+/// `dma_alloc`'d buffer — the controller DMAs straight to/from it, so the bulk
+/// data never crosses IPC. `endpoint_address`'s bit 7 selects IN vs OUT.
+pub const BulkRequest = extern struct {
+    operation: u32 = @intFromEnum(Operation.bulk),
+    reserved: u32 = 0,
+    device_token: u64,
+    physical_address: u64,
+    length: u32,
+    endpoint_address: u8,
+    reserved2: u8 = 0,
+    reserved3: u16 = 0,
+};
+
+pub const BulkReply = extern struct {
+    status: i32,
+    actual_length: u32,
+};
+
+/// An asynchronous interrupt report, pushed with `ipc.send` to a subscriber's
+/// endpoint. `Received.isMessage()` is set; there is no reply owed.
+pub const InterruptReport = extern struct {
+    device_token: u64,
+    endpoint_address: u8,
+    length: u8,
+    reserved: u16 = 0,
+    data: [max_report_data]u8 = [_]u8{0} ** max_report_data,
+};
+
+comptime {
+    const std = @import("std");
+    // Every synchronous message must fit one IPC message; the async report must
+    // fit an ipc_send payload slot.
+    std.debug.assert(@sizeOf(ControlRequest) <= message_maximum);
+    std.debug.assert(@sizeOf(ControlReply) <= message_maximum);
+    std.debug.assert(@sizeOf(OpenReply) <= message_maximum);
+    std.debug.assert(@sizeOf(InterruptReport) <= 64);
+}

system/drivers/usb-xhci-bus/usb-xhci-bus.zig

@@ -0,0 +1,424 @@
+//! /system/drivers/usb-xhci-bus — the xHCI (USB 3) host-controller bus driver.
+//! The device manager spawns **one instance per controller** it discovers (a
+//! machine can carry several), passing the controller's device-tree id as
+//! argv[1]; this instance claims that device and no other, so multiple
+//! instances never fight over hardware.
+//!
+//! M18.2 (this increment): after the hello, real hardware — map the xHC's
+//! register window (the first memory BAR; resource 0 is the ECAM config
+//! space), read the capability registers, and walk the root-hub ports: one
+//! `child_added` report to the manager per connected port, carrying the port
+//! number and the PORTSC speed class as identity. No transfer rings yet —
+//! descriptors and USB class matching are the USB track; the connect bit and
+//! speed come straight from PORTSC, which reflects hardware state whether or
+//! not the controller is running.
+
+const std = @import("std");
+const runtime = @import("runtime");
+const protocol = runtime.device_manager_protocol;
+const device = runtime.device;
+const usb_ids = @import("usb-ids");
+const usb_abi = @import("usb-abi");
+const transfer = @import("usb-transfer-protocol");
+const library = @import("usb-xhci-library.zig");
+
+/// The controller engine (reset, rings, transfers), stood up in `initialise`.
+var controller: ?library.Controller = null;
+
+/// This driver's service endpoint (registered as `.usb_bus`), where class-driver
+/// requests, signals, and the interrupt-poll timer all arrive.
+var service_endpoint: runtime.ipc.Handle = 0;
+
+/// How often the driver drains the event ring for interrupt reports (~125 Hz),
+/// re-armed each tick. Frequent enough for responsive input.
+const poll_interval_ms: u64 = 8;
+
+/// The class driver endpoints that opened each device, so interrupt reports can
+/// be pushed back to them. Keyed by the device token (the interface's device id).
+const Open = struct {
+    used: bool = false,
+    device_token: u64 = 0,
+    report_endpoint: usize = 0,
+};
+var opens = [_]Open{.{}} ** 16;
+
+fn recordOpen(device_token: u64, report_endpoint: usize) void {
+    for (&opens) |*open| {
+        if (open.used and open.device_token == device_token) {
+            open.report_endpoint = report_endpoint;
+            return;
+        }
+    }
+    for (&opens) |*open| {
+        if (!open.used) {
+            open.* = .{ .used = true, .device_token = device_token, .report_endpoint = report_endpoint };
+            return;
+        }
+    }
+}
+
+fn reportEndpointFor(device_token: u64) ?usize {
+    for (&opens) |*open| {
+        if (open.used and open.device_token == device_token) return open.report_endpoint;
+    }
+    return null;
+}
+
+/// Format one whole log line and emit it in a single `debug_write`, so
+/// concurrent instances (one per controller) can never interleave mid-line.
+fn writeLine(comptime fmt: []const u8, arguments: anytype) void {
+    var line: [128]u8 = undefined;
+    _ = runtime.system.write(std.fmt.bufPrint(&line, fmt, arguments) catch return);
+}
+
+var controller_id: u64 = protocol.no_device;
+
+/// Claim the assigned controller, find its register window, and hello the
+/// manager. Any failure returns false: the process exits cleanly, which the
+/// manager reads as "meant to stop" — a missing assignment is not a crash loop.
+fn initialise(endpoint: runtime.ipc.Handle) bool {
+    service_endpoint = endpoint;
+    if (!device.claim(controller_id)) {
+        writeLine("/system/drivers/usb-xhci-bus: unable to claim controller device {d}\n", .{controller_id});
+        return false;
+    }
+
+    // Fetch our own descriptor back for the controller's resources.
+    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: out of memory\n");
+        return false;
+    };
+    const total = device.enumerate(buffer);
+    const descriptor = for (buffer[0..@min(total, buffer.len)]) |d| {
+        if (d.id == controller_id) break d;
+    } else {
+        writeLine("/system/drivers/usb-xhci-bus: device {d} not in the device tree\n", .{controller_id});
+        return false;
+    };
+
+    // The xHC's registers live behind the first memory BAR. Resource 0 is the
+    // function's ECAM configuration space (M15), so the walk starts at 1.
+    var register_index: u64 = 0;
+    const register_window = for (descriptor.resources[1..@intCast(descriptor.resource_count)], 1..) |resource, index| {
+        if (resource.kind == @intFromEnum(device.ResourceKind.memory)) {
+            register_index = index;
+            break resource;
+        }
+    } else {
+        writeLine("/system/drivers/usb-xhci-bus: controller device {d} has no register BAR\n", .{controller_id});
+        return false;
+    };
+    writeLine("/system/drivers/usb-xhci-bus: claimed controller device {d} (registers at 0x{x}, {d} bytes)\n", .{
+        controller_id,
+        register_window.start,
+        register_window.len,
+    });
+    register_base = device.mmioMap(controller_id, register_index) orelse {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: mmio_map failed\n");
+        return false;
+    };
+
+    // Bring the controller up: reset it, stand up the command and event rings,
+    // and start it running (the hardware half lives in usb-xhci-library.zig).
+    controller = library.Controller.init(register_base) orelse {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: controller reset/bring-up failed\n");
+        return false;
+    };
+    writeLine("/system/drivers/usb-xhci-bus: controller running ({d} slots, {d}-byte contexts)\n", .{
+        controller.?.max_slots,
+        controller.?.context_size,
+    });
+    // The proof of life: a No-Op command round-trips the command ring, the event
+    // ring, the doorbell, and the cycle-bit bookkeeping. If this completes, the
+    // engine is sound; transfers build on exactly this machinery.
+    if (controller.?.noOpCommand()) {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: command ring running (no-op ok)\n");
+    } else {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: no-op command did not complete\n");
+        return false;
+    }
+
+    // The handshake: role, protocol version, assignment — inside the manager's
+    // deadline (the lookup retries cover the manager still registering).
+    var manager: ?runtime.ipc.Handle = null;
+    var tries: u32 = 0;
+    while (manager == null and tries < 100) : (tries += 1) {
+        manager = runtime.ipc.lookup(.device_manager);
+        if (manager == null) runtime.system.sleep(20);
+    }
+    const h = manager orelse {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: no device manager to hello\n");
+        return false;
+    };
+    const hello = protocol.Hello{ .role = @intFromEnum(protocol.Role.bus), .device_id = controller_id };
+    var reply: [protocol.message_maximum]u8 = undefined;
+    const n = runtime.ipc.call(h, std.mem.asBytes(&hello), &reply) catch {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: hello call failed\n");
+        return false;
+    };
+    if (n < protocol.reply_size or std.mem.bytesToValue(protocol.HelloReply, reply[0..protocol.reply_size]).status != 0) {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: hello refused\n");
+        return false;
+    }
+    _ = runtime.system.write("/system/drivers/usb-xhci-bus: hello acknowledged\n");
+
+    scanPorts(h);
+
+    // Arm the poll timer that drains interrupt reports from the event ring. It is
+    // re-armed on each tick in onNotification; class drivers subscribe later.
+    _ = runtime.system.timerOnce(service_endpoint, poll_interval_ms);
+    return true;
+}
+
+var register_base: usize = 0;
+
+/// The xHCI default Protocol Speed IDs (the PORTSC port-speed field, bits 13:10)
+/// decoded to human names — the boot-log breadcrumb for what actually enumerated on
+/// a port, the USB analog of the pci-bus class-code line. A controller may redefine
+/// these through its Supported Protocol capability, but the defaults cover every
+/// speed QEMU and real hardware report at this (pre-descriptor) stage.
+fn speedName(speed: u32) []const u8 {
+    return switch (speed) {
+        1 => "Full-speed (USB 2.0, 12 Mb/s)",
+        2 => "Low-speed (USB 2.0, 1.5 Mb/s)",
+        3 => "High-speed (USB 2.0, 480 Mb/s)",
+        4 => "SuperSpeed (USB 3.0, 5 Gb/s)",
+        5 => "SuperSpeedPlus (USB 3.1, 10 Gb/s)",
+        else => "unknown speed",
+    };
+}
+
+/// The root-hub scan and enumeration: for each connected port, bring the device
+/// up (reset → enable slot → address), read its descriptors, and register +
+/// report one child per interface — carrying the interface's (class, subclass,
+/// protocol) triple as identity, which is what the device manager matches a
+/// class driver against.
+fn scanPorts(manager: runtime.ipc.Handle) void {
+    const engine = if (controller) |*c| c else {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: controller not initialised\n");
+        return;
+    };
+    writeLine("/system/drivers/usb-xhci-bus: {d} root-hub ports\n", .{engine.max_ports});
+
+    var port: u32 = 1;
+    var connected: u32 = 0;
+    while (port <= engine.max_ports) : (port += 1) {
+        const port_status = engine.portStatus(port);
+        if (port_status & 1 == 0) continue; // CCS: nothing connected
+        connected += 1;
+        const speed = (port_status >> 10) & 0xF; // the PORTSC port-speed class
+        writeLine("/system/drivers/usb-xhci-bus: port {d} connected — {s} (speed class {d})\n", .{ port, speedName(speed), speed });
+
+        const usb_device = engine.setupDevice(port, speed) orelse {
+            writeLine("/system/drivers/usb-xhci-bus: port {d} device setup failed\n", .{port});
+            continue;
+        };
+        if (!engine.enumerate(usb_device)) {
+            writeLine("/system/drivers/usb-xhci-bus: port {d} enumeration failed\n", .{port});
+            continue;
+        }
+        writeLine("/system/drivers/usb-xhci-bus: port {d} device vendor 0x{x:0>4} product 0x{x:0>4}, {d} interface(s)\n", .{
+            port,
+            usb_device.device_descriptor.vendor_id,
+            usb_device.device_descriptor.product_id,
+            usb_device.interface_count,
+        });
+
+        for (usb_device.interfaces[0..usb_device.interface_count]) |*interface| {
+            // Record the id each interface was registered as, so a class driver
+            // opening the interface (by that id) resolves to it.
+            if (reportInterface(manager, port, interface.*)) |registered| {
+                interface.registered_device_id = registered;
+            }
+        }
+    }
+    if (connected == 0) _ = runtime.system.write("/system/drivers/usb-xhci-bus: no devices connected\n");
+}
+
+/// Register one interface as a resource-less child of the controller and report
+/// it to the device manager. The identity is the packed USB class triple, so the
+/// manager can match a class driver (HID keyboard, mouse, mass storage); the
+/// registered device id becomes that driver's argv[1] assignment. Returns the
+/// registered device id, or null if registration or the report failed.
+fn reportInterface(manager: runtime.ipc.Handle, port: u32, interface: library.InterfaceInfo) ?u64 {
+    const identity = usb_ids.packTriple(interface.class, interface.subclass, interface.protocol);
+
+    // A USB device is reached through its controller, not by MMIO, so the child
+    // carries no resources; register() allows that. Its bus-local identity — the
+    // (port, interface) address, written as a short "P<port>I<interface>" tag in
+    // the hid field — makes each interface a distinct kernel node (the register
+    // dedup keys on class/pci_class/hid/resources, all otherwise identical here)
+    // and keeps re-registration idempotent across a bus restart: the same port
+    // and interface always map back to the same device id.
+    var descriptor = std.mem.zeroes(device.DeviceDescriptor);
+    descriptor.class = @intFromEnum(device.DeviceClass.usb_device);
+    descriptor.pci_class = device.no_pci_class;
+    descriptor.resource_count = 0;
+    var hid_buffer: [8]u8 = undefined;
+    const hid_text = std.fmt.bufPrint(&hid_buffer, "P{d}I{d}", .{ port, interface.number }) catch "";
+    descriptor.hid_len = hid_text.len;
+    @memcpy(descriptor.hid[0..hid_text.len], hid_text);
+    const registered = device.register(controller_id, &descriptor) orelse {
+        writeLine("/system/drivers/usb-xhci-bus: register refused for port {d} interface {d}\n", .{ port, interface.number });
+        return null;
+    };
+
+    const report = protocol.ChildAdded{
+        .parent = controller_id,
+        .bus_address = (@as(u64, port) << 8) | interface.number,
+        .identity = identity,
+        .device_id = registered,
+    };
+    var reply: [protocol.message_maximum]u8 = undefined;
+    _ = runtime.ipc.call(manager, std.mem.asBytes(&report), &reply) catch {
+        writeLine("/system/drivers/usb-xhci-bus: child report for port {d} interface {d} failed\n", .{ port, interface.number });
+        return null;
+    };
+    writeLine("/system/drivers/usb-xhci-bus: port {d} interface {d} class {d}/{d}/{d} registered as device {d}\n", .{
+        port,
+        interface.number,
+        interface.class,
+        interface.subclass,
+        interface.protocol,
+        registered,
+    });
+    return registered;
+}
+
+/// Serve the USB transfer protocol: a class driver opens its device, then issues
+/// control / interrupt-subscribe / bulk requests against it.
+fn onMessage(message: []const u8, reply: []u8, sender: u32, capability: ?runtime.ipc.Handle) usize {
+    _ = sender;
+    if (message.len < 4) return 0;
+    const operation = std.mem.readInt(u32, message[0..4], .little);
+    return switch (operation) {
+        @intFromEnum(transfer.Operation.open) => handleOpen(message, reply, capability),
+        @intFromEnum(transfer.Operation.control) => handleControl(message, reply),
+        @intFromEnum(transfer.Operation.interrupt_subscribe) => handleSubscribe(message, reply),
+        @intFromEnum(transfer.Operation.bulk) => handleBulk(message, reply),
+        else => 0,
+    };
+}
+
+fn writeReply(reply: []u8, value: anytype) usize {
+    const bytes = std.mem.asBytes(&value);
+    @memcpy(reply[0..bytes.len], bytes);
+    return bytes.len;
+}
+
+/// open: resolve the assigned device id to an interface, remember the caller's
+/// endpoint (for interrupt reports), and answer with a device token + the
+/// interface's endpoints so the class driver need not re-read the config.
+fn handleOpen(message: []const u8, reply: []u8, capability: ?runtime.ipc.Handle) usize {
+    if (message.len < @sizeOf(transfer.OpenRequest)) return writeReply(reply, transfer.OpenReply{ .status = -1, .endpoint_count = 0, .device_token = 0, .interface_class = 0, .interface_subclass = 0, .interface_protocol = 0, .interface_number = 0 });
+    const request = std.mem.bytesToValue(transfer.OpenRequest, message[0..@sizeOf(transfer.OpenRequest)]);
+    const engine = if (controller) |*c| c else return writeReply(reply, transfer.OpenReply{ .status = -1, .endpoint_count = 0, .device_token = 0, .interface_class = 0, .interface_subclass = 0, .interface_protocol = 0, .interface_number = 0 });
+    const found = engine.findInterface(request.device_id) orelse return writeReply(reply, transfer.OpenReply{ .status = -1, .endpoint_count = 0, .device_token = 0, .interface_class = 0, .interface_subclass = 0, .interface_protocol = 0, .interface_number = 0 });
+
+    if (capability) |endpoint| recordOpen(request.device_id, endpoint);
+
+    var open_reply = transfer.OpenReply{
+        .status = 0,
+        .endpoint_count = found.interface.endpoint_count,
+        .device_token = request.device_id,
+        .interface_class = found.interface.class,
+        .interface_subclass = found.interface.subclass,
+        .interface_protocol = found.interface.protocol,
+        .interface_number = found.interface.number,
+    };
+    const count = @min(found.interface.endpoint_count, transfer.max_reported_endpoints);
+    for (found.interface.endpoints[0..count], 0..) |endpoint, index| {
+        open_reply.endpoints[index] = .{
+            .address = endpoint.address,
+            .transfer_type = endpoint.transfer_type,
+            .max_packet_size = endpoint.max_packet_size,
+            .interval = endpoint.interval,
+        };
+    }
+    return writeReply(reply, open_reply);
+}
+
+/// control: one EP0 control transfer, small data inline both ways.
+fn handleControl(message: []const u8, reply: []u8) usize {
+    if (message.len < @sizeOf(transfer.ControlRequest)) return writeReply(reply, transfer.ControlReply{ .status = -1, .actual_length = 0 });
+    const request = std.mem.bytesToValue(transfer.ControlRequest, message[0..@sizeOf(transfer.ControlRequest)]);
+    const engine = if (controller) |*c| c else return writeReply(reply, transfer.ControlReply{ .status = -1, .actual_length = 0 });
+    const found = engine.findInterface(request.device_token) orelse return writeReply(reply, transfer.ControlReply{ .status = -1, .actual_length = 0 });
+
+    const setup = std.mem.bytesToValue(usb_abi.Request, &request.setup);
+    const direction_in = request.direction_in != 0;
+    const data_length = @min(request.data_length, transfer.max_inline_data);
+    var data: [transfer.max_inline_data]u8 = undefined;
+    if (!direction_in) @memcpy(data[0..data_length], request.data[0..data_length]);
+
+    const ok = engine.controlTransfer(found.device, setup, data[0..data_length], direction_in);
+    var control_reply = transfer.ControlReply{ .status = if (ok) 0 else -1, .actual_length = if (ok) data_length else 0 };
+    if (ok and direction_in) @memcpy(control_reply.data[0..data_length], data[0..data_length]);
+    return writeReply(reply, control_reply);
+}
+
+/// interrupt_subscribe: arm periodic IN polling; reports flow back asynchronously.
+fn handleSubscribe(message: []const u8, reply: []u8) usize {
+    if (message.len < @sizeOf(transfer.InterruptSubscribeRequest)) return writeReply(reply, transfer.InterruptSubscribeReply{ .status = -1 });
+    const request = std.mem.bytesToValue(transfer.InterruptSubscribeRequest, message[0..@sizeOf(transfer.InterruptSubscribeRequest)]);
+    const engine = if (controller) |*c| c else return writeReply(reply, transfer.InterruptSubscribeReply{ .status = -1 });
+    const found = engine.findInterface(request.device_token) orelse return writeReply(reply, transfer.InterruptSubscribeReply{ .status = -1 });
+    const endpoint = library.Controller.endpointForAddress(found.interface, request.endpoint_address) orelse return writeReply(reply, transfer.InterruptSubscribeReply{ .status = -1 });
+    const report_endpoint = reportEndpointFor(request.device_token) orelse return writeReply(reply, transfer.InterruptSubscribeReply{ .status = -1 });
+    const ok = engine.subscribeInterrupt(found.device, endpoint, request.device_token, report_endpoint);
+    return writeReply(reply, transfer.InterruptSubscribeReply{ .status = if (ok) 0 else -1 });
+}
+
+/// bulk: one bulk transfer to/from the class driver's own DMA buffer (by physical
+/// address), so sector-sized data never crosses IPC.
+fn handleBulk(message: []const u8, reply: []u8) usize {
+    if (message.len < @sizeOf(transfer.BulkRequest)) return writeReply(reply, transfer.BulkReply{ .status = -1, .actual_length = 0 });
+    const request = std.mem.bytesToValue(transfer.BulkRequest, message[0..@sizeOf(transfer.BulkRequest)]);
+    const engine = if (controller) |*c| c else return writeReply(reply, transfer.BulkReply{ .status = -1, .actual_length = 0 });
+    const found = engine.findInterface(request.device_token) orelse return writeReply(reply, transfer.BulkReply{ .status = -1, .actual_length = 0 });
+    const endpoint = library.Controller.endpointForAddress(found.interface, request.endpoint_address) orelse return writeReply(reply, transfer.BulkReply{ .status = -1, .actual_length = 0 });
+    const transferred = engine.bulkTransfer(found.device, endpoint, request.physical_address, request.length);
+    return writeReply(reply, transfer.BulkReply{ .status = if (transferred != null) 0 else -1, .actual_length = transferred orelse 0 });
+}
+
+/// The poll timer landed: drain any interrupt reports off the event ring and push
+/// each to the class driver that subscribed, then re-arm the timer.
+fn onNotification(badge: u64) void {
+    if (badge & runtime.ipc.notify_timer_bit == 0) return;
+    if (controller) |*engine| {
+        engine.pump();
+        while (engine.takeReport()) |report| {
+            var message = transfer.InterruptReport{
+                .device_token = report.device_token,
+                .endpoint_address = report.endpoint_address,
+                .length = @intCast(@min(report.length, transfer.max_report_data)),
+            };
+            const n = @min(report.length, transfer.max_report_data);
+            @memcpy(message.data[0..n], report.data[0..n]);
+            _ = runtime.ipc.send(report.report_endpoint, std.mem.asBytes(&message));
+        }
+    }
+    _ = runtime.system.timerOnce(service_endpoint, poll_interval_ms);
+}
+
+pub fn main(init: runtime.process.Init) void {
+    const argument = init.arguments.get(1) orelse {
+        _ = runtime.system.write("/system/drivers/usb-xhci-bus: missing controller device id (argv[1])\n");
+        return;
+    };
+    controller_id = std.fmt.parseInt(u64, argument, 10) catch {
+        writeLine("/system/drivers/usb-xhci-bus: malformed controller device id '{s}'\n", .{argument});
+        return;
+    };
+    runtime.service.run(transfer.message_maximum, .{
+        .service = .usb_bus,
+        .init = initialise,
+        .on_message = onMessage,
+        .on_notification = onNotification,
+    });
+}
+
+pub const panic = runtime.panic;
+comptime {
+    _ = &runtime.start._start; // pull the runtime entry shim into the image
+}

system/kernel/architecture/x86_64/apic.zig

@@ -80,6 +80,35 @@ var timer_hz: u32 = 0;
 var tsc_hz: u64 = 0;
 var tsc_base: u64 = 0;
 
+/// Whether the TSC is architecturally **invariant** — a constant rate regardless of
+/// P/C-state transitions, and thus valid as a clocksource (CPUID leaf 0x80000007,
+/// EDX bit 8). AMD and modern Intel set it; the bare qemu64 model does not. Measured
+/// frequency alone is not enough: a non-invariant TSC speeds up and slows down with
+/// the core clock, so reading it as wall time would drift.
+var tsc_invariant: bool = false;
+/// Cleared if the cross-core warp check (checkWarpSource) ever sees the TSC read
+/// lower on one core than the max another core has already published — i.e. the
+/// per-core TSCs are not synchronized, and a task migrating cores could see time go
+/// backward. Starts true (assume synchronized until proven otherwise).
+var tsc_synced: bool = true;
+/// The worst backward skew the warp check observed, in TSC cycles (0 = none).
+var tsc_warp_cycles: u64 = 0;
+
+/// The monotonic clock's source. The TSC when it is invariant *and* synchronized —
+/// the fast `rdtsc` path taken on real Intel/AMD and modern VMs. Otherwise the HPET
+/// main counter: a single fixed-rate counter, immune to both per-core skew and
+/// frequency scaling, so it stays accurate on a bare VM or a warped machine.
+const ClockSource = enum { tsc, hpet };
+var clock_source: ClockSource = .tsc;
+
+/// HPET standby clocksource, set up in calibrate() whenever an HPET exists (whether
+/// or not calibration itself measured against it): its frequency, the counter value
+/// chosen as the zero point, and its width mask. Only a 64-bit HPET is used as a
+/// clocksource — a 32-bit one wraps too fast to be monotonic without accumulation.
+var hpet_clock_hz: u64 = 0;
+var hpet_clock_base: u64 = 0;
+var hpet_clock_mask: u64 = ~@as(u64, 0);
+
 /// Read the 64-bit Time Stamp Counter.
 fn rdtsc() u64 {
     var low: u32 = undefined;
@@ -220,6 +249,31 @@ pub fn calibrate() void {
     }
 
     tsc_base = rdtsc(); // the clock's zero point (boot)
+
+    // Decide whether the TSC is trustworthy as a clocksource. Frequency (measured
+    // above, possibly against the HPET/PIT) is necessary but not sufficient: the TSC
+    // must also be *invariant* (CPUID 0x80000007 EDX[8]). AMD and modern Intel set
+    // this; the bare qemu64 model does not.
+    tsc_invariant = tscIsInvariant();
+
+    // Bring up the HPET as a standby clocksource whenever one exists — even on the
+    // CPUID-0x15 path where calibration never touched it — so a non-invariant TSC
+    // (here) or an unsynchronized one (checkWarpSource, during SMP bring-up) can fall
+    // back to a source that is immune to both. hpetHz() maps + enables the counter
+    // and is idempotent if calibration already used it.
+    if (configuration_hpet_base != 0) {
+        if (hpetHz()) |hz| {
+            hpet_clock_mask = hpetMask();
+            if (hpet_clock_mask == ~@as(u64, 0)) { // only a 64-bit HPET is monotonic enough
+                hpet_clock_hz = hz;
+                hpet_clock_base = readHpet();
+            }
+        }
+    }
+
+    // Select the source: the fast TSC when invariant, else the HPET if we have one.
+    // (checkWarpSource may still demote TSC -> HPET later if the cores' TSCs skew.)
+    if (!tsc_invariant and hpet_clock_hz != 0) clock_source = .hpet;
 }
 
 /// Run the LAPIC timer one-shot from its maximum count while a monotonic reference
@@ -275,7 +329,7 @@ fn calibratePit() void {
 // --- reference clocks ------------------------------------------------------
 
 /// TSC frequency from CPUID leaf 0x15 (crystal_hz * numerator / denominator), or
-/// null if the CPU doesn't enumerate it (common under QEMU).
+/// null if the CPU doesn't enumerate it (common under QEMU, and on AMD).
 fn cpuidTscHz() ?u64 {
     if (cpuid(0).eax < 0x15) return null;
     const r = cpuid(0x15);
@@ -283,6 +337,15 @@ fn cpuidTscHz() ?u64 {
     return @as(u64, r.ecx) * r.ebx / r.eax;
 }
 
+/// Whether the CPU advertises an **invariant** TSC (CPUID leaf 0x80000007, EDX
+/// bit 8) — the architectural guarantee, on both Intel and AMD, that the TSC ticks
+/// at a constant rate across P/C-states and never stops. Requires the extended-leaf
+/// range to reach 0x80000007 first.
+fn tscIsInvariant() bool {
+    if (cpuid(0x80000000).eax < 0x80000007) return false;
+    return (cpuid(0x80000007).edx & (1 << 8)) != 0;
+}
+
 const CpuidRegs = struct { eax: u32, ebx: u32, ecx: u32, edx: u32 };
 
 fn cpuid(leaf: u32) CpuidRegs {
@@ -310,11 +373,20 @@ fn hpetWrite64(off: usize, value: u64) void {
     @as(*volatile u64, @ptrFromInt(configuration_hpet_base + off)).* = value;
 }
 
+/// Whether the HPET has been mapped into the physmap yet, so `configuration_hpet_base`
+/// already holds the virtual address. `hpetHz` is called more than once (calibration
+/// may use the HPET, and the standby-clocksource setup asks for it again), and mapping
+/// an already-mapped base a second time would double-offset it into an overflow.
+var hpet_mapped: bool = false;
+
 /// Map + enable the HPET and return its tick frequency, or null if unusable.
 /// Maps the HPET into the physmap and switches configuration_hpet_base to that virtual
-/// address, so the register accessors reach it without the identity map.
+/// address, so the register accessors reach it without the identity map. Idempotent.
 fn hpetHz() ?u64 {
+    if (!hpet_mapped) {
         configuration_hpet_base = paging.mapMmio(configuration_hpet_base, 0x400, true);
+        hpet_mapped = true;
+    }
     const caps = hpetRead64(0x00);
     const period_fs = caps >> 32; // femtoseconds per tick
     if (period_fs == 0) return null;
@@ -364,24 +436,169 @@ pub fn tscHz() u64 {
     return tsc_hz;
 }
 
-// Monotonic high-resolution clock, from the TSC. A function per resolution, each
-// scaling the cycle delta directly at its unit (the 128-bit intermediate avoids
-// overflow across a long uptime). nanos() resolves to a few ns; millis() is what
-// the scheduler uses for sleep deadlines.
+// Monotonic high-resolution clock. A function per resolution, each scaling the
+// counter delta directly at its unit (the 128-bit intermediate avoids overflow
+// across a long uptime). nanos() resolves to a few ns on the TSC; millis() is what
+// the scheduler uses for sleep deadlines. The source is the TSC when it is invariant
+// and synchronized, else the HPET counter (see clock_source) — the branch is one
+// global load and the TSC path is unchanged from before.
+
+/// The selected source's counter delta since its zero point.
+fn clockCount() u64 {
+    return switch (clock_source) {
+        .tsc => rdtsc() -% tsc_base,
+        // A 64-bit HPET (the only kind we select) never wraps in any realistic
+        // uptime, so the wrapping subtraction is exact.
+        .hpet => readHpet() -% hpet_clock_base,
+    };
+}
+
+/// The selected source's frequency (0 if the clock is unavailable/uncalibrated).
+fn clockHertz() u64 {
+    return switch (clock_source) {
+        .tsc => tsc_hz,
+        .hpet => hpet_clock_hz,
+    };
+}
 
 pub fn nanos() u64 {
-    if (tsc_hz == 0) return 0;
-    return @intCast(@as(u128, rdtsc() -% tsc_base) * 1_000_000_000 / tsc_hz);
+    const hz = clockHertz();
+    if (hz == 0) return 0;
+    return @intCast(@as(u128, clockCount()) * 1_000_000_000 / hz);
 }
 
 pub fn micros() u64 {
-    if (tsc_hz == 0) return 0;
-    return @intCast(@as(u128, rdtsc() -% tsc_base) * 1_000_000 / tsc_hz);
+    const hz = clockHertz();
+    if (hz == 0) return 0;
+    return @intCast(@as(u128, clockCount()) * 1_000_000 / hz);
 }
 
 pub fn millis() u64 {
-    if (tsc_hz == 0) return 0;
-    return @intCast(@as(u128, rdtsc() -% tsc_base) * 1_000 / tsc_hz);
+    const hz = clockHertz();
+    if (hz == 0) return 0;
+    return @intCast(@as(u128, clockCount()) * 1_000 / hz);
+}
+
+/// Whether the CPU advertises an invariant TSC (CPUID 0x80000007 EDX[8]).
+pub fn tscInvariant() bool {
+    return tsc_invariant;
+}
+
+/// Test hook: force the TSC clocksource on, as if the CPU had advertised an invariant
+/// TSC. QEMU's TCG accelerator (the only one for an x86 guest on an Apple-Silicon
+/// host) does not expose the invariant-TSC bit — its emulated TSC isn't invariant — so
+/// the tsc-sync test can't reach the real-Intel/AMD/KVM path through CPUID. This lets
+/// that test exercise the TSC clocksource and the cross-core warp check anyway. tsc_base
+/// is left as-is so the switch from the HPET is continuous.
+pub fn forceTscClocksourceForTest() void {
+    tsc_invariant = true;
+    clock_source = .tsc;
+}
+
+/// How many per-AP warp checks actually ran (a rendezvous completed) — lets a test
+/// confirm the cross-core check executed rather than being skipped.
+pub fn warpChecksRun() u32 {
+    return warp_checks;
+}
+
+/// Whether the per-core TSCs are synchronized (no backward warp seen at bring-up).
+pub fn tscSynced() bool {
+    return tsc_synced;
+}
+
+/// The active monotonic clocksource, for the boot log and tests.
+pub fn clockSourceName() []const u8 {
+    return switch (clock_source) {
+        .tsc => "tsc",
+        .hpet => "hpet",
+    };
+}
+
+// --- cross-core TSC synchronization ("warp") check -------------------------
+// Two cores hammer a shared "max seen" TSC value under a lock; if either reads a
+// value below that max, its TSC lags the other's, and time would run backward for a
+// task migrating between them (Linux calls this a warp). danos brings APs up one at a
+// time, so this runs pairwise: the BSP (source) against each AP (target) as it comes
+// online. It only matters — and only runs — while the TSC is the clocksource; on a
+// machine already on the HPET (a bare VM) the whole rendezvous is skipped.
+
+var warp_lock: u32 = 0;
+var warp_last: u64 = 0;
+var warp_bsp_ready: u32 = 0;
+var warp_ap_ready: u32 = 0;
+var warp_stop: u32 = 0;
+var warp_checks: u32 = 0; // completed per-AP rendezvous count (for the tsc-sync test)
+
+const warp_rounds: u32 = 1 << 20; // locked reads on the BSP: ~1 ms at GHz rates
+const warp_spin_limit: u64 = 1 << 32; // bound every rendezvous wait so a lost core can't hang boot
+
+fn warpTick() void {
+    while (@cmpxchgWeak(u32, &warp_lock, 0, 1, .acquire, .monotonic) != null) asm volatile ("pause");
+    const t = rdtsc();
+    if (t < warp_last) {
+        const delta = warp_last - t;
+        if (delta > tsc_warp_cycles) tsc_warp_cycles = delta;
+        tsc_synced = false;
+    } else {
+        warp_last = t;
+    }
+    @atomicStore(u32, &warp_lock, 0, .release);
+}
+
+/// Spin (bounded) until `flag` is nonzero; false on timeout.
+fn warpAwait(flag: *u32) bool {
+    var spins: u64 = 0;
+    while (@atomicLoad(u32, flag, .acquire) == 0) : (spins += 1) {
+        if (spins >= warp_spin_limit) return false;
+        asm volatile ("pause");
+    }
+    return true;
+}
+
+/// BSP side of the pairwise TSC warp check, run once per AP as it reports in. No-op
+/// unless the TSC is the active clocksource. If the AP's TSC proves to lag, demote
+/// the monotonic clock to the HPET without a discontinuity.
+pub fn checkWarpSource() void {
+    if (clock_source != .tsc) return;
+    warp_last = 0;
+    @atomicStore(u32, &warp_stop, 0, .release);
+    @atomicStore(u32, &warp_ap_ready, 0, .release);
+    @atomicStore(u32, &warp_bsp_ready, 1, .release);
+    if (!warpAwait(&warp_ap_ready)) { // AP never joined the rendezvous; skip, don't hang
+        @atomicStore(u32, &warp_bsp_ready, 0, .release);
+        return;
+    }
+    var i: u32 = 0;
+    while (i < warp_rounds) : (i += 1) warpTick();
+    @atomicStore(u32, &warp_stop, 1, .release);
+    @atomicStore(u32, &warp_bsp_ready, 0, .release);
+    warp_checks += 1;
+
+    if (!tsc_synced and hpet_clock_hz != 0) demoteToHpet();
+}
+
+/// AP side: join the BSP's warp check, then return so the core can enter the
+/// scheduler. Bounded so a missing BSP can't strand the core.
+pub fn checkWarpTarget() void {
+    if (clock_source != .tsc) return;
+    if (!warpAwait(&warp_bsp_ready)) return;
+    @atomicStore(u32, &warp_ap_ready, 1, .release);
+    var spins: u64 = 0;
+    while (@atomicLoad(u32, &warp_stop, .acquire) == 0) : (spins += 1) {
+        if (spins >= warp_spin_limit) return;
+        warpTick();
+    }
+}
+
+/// Switch the clocksource from the TSC to the HPET without a discontinuity: choose
+/// the HPET zero point so it reads the same nanosecond value the TSC does right now,
+/// so time neither jumps nor runs backward across the switch. Called when the warp
+/// check proves the per-core TSCs unsynchronized.
+fn demoteToHpet() void {
+    const now_ns = @as(u128, rdtsc() -% tsc_base) * 1_000_000_000 / tsc_hz;
+    const equivalent_ticks: u64 = @intCast(now_ns * hpet_clock_hz / 1_000_000_000);
+    hpet_clock_base = readHpet() -% equivalent_ticks;
+    clock_source = .hpet;
 }
 
 /// Acknowledge the current interrupt so the LAPIC will deliver the next one.

system/kernel/architecture/x86_64/cpu.zig

@@ -469,6 +469,124 @@ pub fn clockHz() u64 {
     return apic.tscHz();
 }
 
+// --- real-time clock (CMOS) --------------------------------------------------
+//
+// The battery-backed CMOS clock, read once at boot and thereafter anchored to the
+// monotonic clock (see kernel/wall-clock.zig) — so this is never on a hot path and
+// needs no lock. Wall-clock *seconds* are mechanism the kernel owns (the hardware's
+// value), like the monotonic clock; calendars/timezones are policy layered on top.
+
+fn cmosRead(register: u8) u8 {
+    io.outb(0x70, register);
+    return io.inb(0x71);
+}
+
+const RtcFields = struct { second: u8, minute: u8, hour: u8, day: u8, month: u8, year: u8 };
+
+fn rtcRaw() RtcFields {
+    while (cmosRead(0x0A) & 0x80 != 0) {} // wait out any update in progress (status A bit 7)
+    return .{
+        .second = cmosRead(0x00),
+        .minute = cmosRead(0x02),
+        .hour = cmosRead(0x04),
+        .day = cmosRead(0x07),
+        .month = cmosRead(0x08),
+        .year = cmosRead(0x09),
+    };
+}
+
+fn bcdToBinary(v: u8) u8 {
+    return (v & 0x0F) + ((v >> 4) * 10);
+}
+
+fn isLeapYear(y: u32) bool {
+    return (y % 4 == 0 and y % 100 != 0) or (y % 400 == 0);
+}
+
+/// Read the CMOS real-time clock and convert it to Unix epoch seconds (UTC).
+pub fn readRtcUnixSeconds() u64 {
+    // Read until two consecutive reads agree, so we never latch a half-updated time.
+    var a = rtcRaw();
+    while (true) {
+        const b = rtcRaw();
+        if (a.second == b.second and a.minute == b.minute and a.hour == b.hour and
+            a.day == b.day and a.month == b.month and a.year == b.year) break;
+        a = b;
+    }
+
+    const status_b = cmosRead(0x0B);
+    const binary_mode = status_b & 0x04 != 0; // else BCD
+    const hour_24 = status_b & 0x02 != 0; // else 12-hour with a PM bit
+
+    var second = a.second;
+    var minute = a.minute;
+    var hour_field = a.hour;
+    var day = a.day;
+    var month = a.month;
+    var year = a.year;
+    if (!binary_mode) {
+        second = bcdToBinary(second);
+        minute = bcdToBinary(minute);
+        hour_field = bcdToBinary(hour_field & 0x7F) | (hour_field & 0x80); // preserve the PM bit
+        day = bcdToBinary(day);
+        month = bcdToBinary(month);
+        year = bcdToBinary(year);
+    }
+
+    var hour: u32 = hour_field & 0x7F;
+    if (!hour_24) {
+        const pm = hour_field & 0x80 != 0;
+        hour %= 12; // 12 AM/PM -> 0
+        if (pm) hour += 12;
+    }
+
+    // The CMOS year is 0..99; QEMU and modern hardware mean 20xx (there is no
+    // reliable century register on QEMU). Treat < 70 as 20xx, else 19xx.
+    const full_year: u32 = if (year < 70) 2000 + @as(u32, year) else 1900 + @as(u32, year);
+
+    var days: u64 = 0;
+    var y: u32 = 1970;
+    while (y < full_year) : (y += 1) days += if (isLeapYear(y)) 366 else 365;
+    const month_lengths = [_]u8{ 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 };
+    var m: u8 = 1;
+    while (m < month) : (m += 1) {
+        days += month_lengths[m - 1];
+        if (m == 2 and isLeapYear(full_year)) days += 1;
+    }
+    days += @as(u64, day) - 1;
+
+    return ((days * 24 + hour) * 60 + minute) * 60 + second;
+}
+
+/// Whether the CPU guarantees an **invariant** TSC (CPUID 0x80000007 EDX[8] on
+/// x86; the analogous architectural guarantee elsewhere). When false the TSC is not
+/// used as the clocksource.
+pub fn clockInvariant() bool {
+    return apic.tscInvariant();
+}
+
+/// Whether the per-core clock counters are synchronized (no backward warp observed
+/// at SMP bring-up). When false the clock falls back off the TSC.
+pub fn clockSynchronized() bool {
+    return apic.tscSynced();
+}
+
+/// The active monotonic clocksource, for the boot log ("tsc" or "hpet" on x86).
+pub fn clockSourceName() []const u8 {
+    return apic.clockSourceName();
+}
+
+/// Test hook: force the TSC clocksource on, to exercise the TSC + warp-check path on
+/// a hypervisor that won't advertise an invariant TSC (see apic.forceTscClocksourceForTest).
+pub fn forceTscClocksourceForTest() void {
+    apic.forceTscClocksourceForTest();
+}
+
+/// How many per-AP TSC warp checks completed (for the tsc-sync test).
+pub fn warpChecksRun() u32 {
+    return apic.warpChecksRun();
+}
+
 /// Unmask maskable interrupts (`sti`) so device interrupts get delivered.
 pub fn enableInterrupts() void {
     asm volatile ("sti");
@@ -490,8 +608,7 @@ pub fn saveInterrupts() u64 {
         \\cli
         : [f] "=r" (flags),
         :
-        : .{ .memory = true }
-    );
+        : .{ .memory = true });
     return flags;
 }
 

system/kernel/architecture/x86_64/idt.zig

@@ -3,9 +3,11 @@
 //! triple-faults and silently resets the machine. With it, the CPU vectors into
 //! our stubs, which capture the register state and hand it to a dispatcher.
 //!
-//! Vectors split in two: 0-31 are CPU exceptions (terminal — reported and
-//! halted); 32+ are device interrupts (a registered handler runs, the APIC is
-//! acknowledged, and we return to the interrupted code).
+//! Vectors split in two: 0-31 are CPU exceptions, handed to the `on_fault` hook
+//! and never returned from (the kernel's handler kills a faulting user process
+//! and reschedules, or halts the core for a kernel-mode fault); 32+ are device
+//! interrupts (a registered handler runs, the APIC is acknowledged, and we
+//! return to the interrupted code).
 
 const gdt = @import("gdt.zig");
 const tss = @import("tss.zig");
@@ -162,8 +164,9 @@ pub fn loadOnThisCpu() void {
 }
 
 /// Called by isr_common (isr.s) with a pointer to the trap frame. Exported so the
-/// assembly stubs can `call` it by name. Exceptions are terminal; device
-/// interrupts run their handler, get acknowledged, and return.
+/// assembly stubs can `call` it by name. Exceptions never return here (on_fault
+/// kills the faulting process or halts the core); device interrupts run their
+/// handler, get acknowledged, and return.
 export fn interruptDispatch(state: *CpuState) callconv(.c) void {
     if (state.vector < 32) {
         on_fault(state); // CPU exception — never returns

system/kernel/architecture/x86_64/paging.zig

@@ -166,8 +166,7 @@ pub fn init(allocFrame: *const fn () ?u64, freeFrame: *const fn (u64) void, boot
     asm volatile ("mov %[pml4], %%cr3"
         :
         : [pml4] "r" (pml4),
-        : .{ .memory = true }
-    );
+        : .{ .memory = true });
     on_own_tables = true; // now on the kernel's physmap (covers all RAM)
     init_done = true; // the kernel half is fixed from here
 }
@@ -409,6 +408,5 @@ fn invalidate(virtual: u64) void {
         \\invlpg (%%rax)
         :
         : [v] "r" (virtual),
-        : .{ .rax = true, .memory = true }
-    );
+        : .{ .rax = true, .memory = true });
 }

system/kernel/architecture/x86_64/smp.zig

@@ -148,7 +148,14 @@ pub fn startAp(apic_id: u32, stack_top: usize, percpu: usize, index: usize, cr3:
     // Wait up to 100 ms for the AP to reach apEntry and set the flag.
     const deadline = apic.millis() + 100;
     while (apic.millis() < deadline) {
-        if (@atomicLoad(u32, &ap_alive, .acquire) != 0) return true;
+        if (@atomicLoad(u32, &ap_alive, .acquire) != 0) {
+            // Cross-check this core's TSC against the BSP's before it joins the run
+            // loop: an unsynchronized TSC must be caught before any task can migrate
+            // onto this core and observe time going backward. No-op unless the TSC is
+            // the clocksource (apic.checkWarpSource).
+            apic.checkWarpSource();
+            return true;
+        }
         asm volatile ("pause");
     }
     return false;
@@ -177,6 +184,11 @@ fn apEntry(percpu: usize) callconv(.c) noreturn {
 
     @atomicStore(u32, &ap_alive, 1, .release); // "architecture state up" — BSP is polling this
 
+    // Rendezvous with the BSP for the TSC warp check (no-op unless the TSC is the
+    // clocksource) before joining the run loop, so this core's clock is vetted before
+    // it can run any task.
+    apic.checkWarpTarget();
+
     if (secondary_entry) |enterScheduler| enterScheduler(); // joins the run loop
     while (true) asm volatile ("hlt"); // (only if no entry was registered)
 }

system/kernel/console.zig

@@ -154,5 +154,3 @@ pub const Console = struct {
         while (x < self.fb.width) : (x += 1) destination[x] = source[x];
     }
 };
-
-

system/kernel/devices-broker.zig

@@ -66,6 +66,7 @@ fn record(node: *platform.Device, parent_id: u64) u64 {
     d.id = count;
     d.parent = parent_id;
     d.class = @intFromEnum(node.class);
+    d.pci_class = if (node.ids.pci_class) |code| code else device_abi.no_pci_class;
     const h = node.hid();
     d.hid_len = @min(h.len, d.hid.len);
     @memcpy(d.hid[0..d.hid_len], h[0..d.hid_len]);
@@ -103,6 +104,19 @@ pub fn ownerOf(id: u64) ?u32 {
     return claimed[@intCast(id)];
 }
 
+/// Release every claim held by `owner` — called by the process layer on every
+/// path out of a process (exit, fault, kill), so a restarted driver can claim its
+/// hardware again (docs/process-lifecycle.md iron rule 1: cleanup is the kernel's
+/// job). The devices stay in the table — they describe hardware, which did not go
+/// away — only their ownership clears.
+pub fn releaseAllOwnedBy(owner: u32) void {
+    for (claimed[0..count]) |*slot| {
+        if (slot.*) |o| {
+            if (o == owner) slot.* = null;
+        }
+    }
+}
+
 /// Resource `index` of device `id`, or null if out of range.
 pub fn resourceOf(id: u64, index: u64) ?device_abi.ResourceDescriptor {
     if (id >= count) return null;
@@ -117,7 +131,16 @@ pub fn resourceOf(id: u64, index: u64) ?device_abi.ResourceDescriptor {
 /// and would otherwise vacuously "fit" anywhere.
 fn contains(parent: device_abi.ResourceDescriptor, child: device_abi.ResourceDescriptor) bool {
     if (parent.kind != child.kind) return false;
-    if (child.kind == @intFromEnum(device_abi.ResourceKind.irq)) return parent.start == child.start;
+    if (child.kind == @intFromEnum(device_abi.ResourceKind.irq)) {
+        // Range containment: an interrupt line is still indivisible (a child owns
+        // exactly one GSI), but a parent may own a *range* of lines so a broad
+        // owner — the acpi-tables node, whose firmware names any legacy IRQ —
+        // can contain its children's specific lines. A length-1 parent range is
+        // exactly the old equality rule, so existing single-IRQ parents are
+        // unaffected.
+        const span = if (parent.len == 0) 1 else parent.len;
+        return child.start >= parent.start and child.start < parent.start + span;
+    }
     if (child.len == 0 or parent.len == 0) return false;
     // No overflow: a resource that wraps the address space is not containable.
     const child_end = std.math.add(u64, child.start, child.len) catch return false;
@@ -166,10 +189,31 @@ pub fn register(parent_id: u64, owner: u32, descriptor: *const device_abi.Device
         if (!ok) return error.NotContained;
     }
 
+    // Idempotent on exact match (docs/device-manager.md): a restarted
+    // registering bus re-registers what it rediscovers, and the table has no
+    // unregister — an identical (class, identity, resources) child under the
+    // same parent returns the existing id instead of appending a duplicate.
+    for (devices[0..count]) |*existing| {
+        if (existing.parent != parent_id) continue;
+        if (existing.class != descriptor.class) continue;
+        if (existing.pci_class != descriptor.pci_class) continue;
+        if (existing.hid_len != descriptor.hid_len) continue;
+        if (!std.mem.eql(u8, existing.hid[0..@intCast(existing.hid_len)], descriptor.hid[0..@intCast(descriptor.hid_len)])) continue;
+        if (existing.resource_count != descriptor.resource_count) continue;
+        var same = true;
+        for (0..@intCast(descriptor.resource_count)) |i| {
+            const a = existing.resources[i];
+            const b = descriptor.resources[i];
+            if (a.kind != b.kind or a.start != b.start or a.len != b.len) same = false;
+        }
+        if (same) return existing.id;
+    }
+
     var d = std.mem.zeroes(device_abi.DeviceDescriptor);
     d.id = count;
     d.parent = parent_id;
     d.class = descriptor.class;
+    d.pci_class = descriptor.pci_class;
     d.hid_len = @min(descriptor.hid_len, d.hid.len);
     @memcpy(d.hid[0..@intCast(d.hid_len)], descriptor.hid[0..@intCast(d.hid_len)]);
     d.resource_count = descriptor.resource_count;

system/kernel/ipc-synchronous.zig

@@ -37,7 +37,9 @@ const Task = scheduler.Task;
 pub const MESSAGE_MAXIMUM: usize = 256;
 
 pub const maximum_handles = scheduler.ipc_maximum_handles;
-pub const maximum_services = 8;
+// The name registry is indexed directly by ServiceId, so this must exceed the
+// largest id (currently fat = 8). Sized with headroom for new services.
+pub const maximum_services = 16;
 
 /// Errno-style failures, returned as `-value` in the system_call result register.
 pub const EBADF: i64 = 1; // bad handle
@@ -46,6 +48,9 @@ pub const EFAULT: i64 = 3; // buffer unmapped / out of the user half
 pub const ENOENT: i64 = 4; // no such registered service
 pub const ENOSPC: i64 = 5; // handle table or registry full
 pub const ENOMEM: i64 = 6; // out of memory
+pub const EPEER: i64 = 7; // peer died before replying (its process exited or was killed)
+pub const ESRCH: i64 = 8; // no such process (process_kill of an unknown/dead id)
+pub const EPERM: i64 = 9; // not permitted (process_kill by anyone but the supervisor)
 
 /// A badge with this bit set is an asynchronous notification (e.g. an IRQ), not a
 /// message from a client — there is no reply owed. The low bits carry the source
@@ -54,6 +59,29 @@ pub const ENOMEM: i64 = 6; // out of memory
 /// shared kernel↔user ABI (system/abi.zig), because ring 3 has to test the same bit.
 pub const notify_badge_bit: u64 = abi.notify_badge_bit;
 
+/// Set (with `notify_badge_bit`) when a `replyWait` wake carries a buffered payload
+/// posted by `send` (`ipc_send`), rather than a bare IRQ/exit notification. Shared with
+/// ring 3 through the ABI so the receiver can tell "a message arrived" from "the hardware
+/// spoke".
+pub const notify_message_bit: u64 = abi.notify_message_bit;
+
+/// Largest payload a single `send` (`ipc_send`) may post. Kept small — the payload rides
+/// inline in every `Endpoint`, and the async path is for events (a `KeyEvent` is 16
+/// bytes), not bulk transfer, which is what `call` and future shared pages are for.
+pub const POST_MAXIMUM: usize = 64;
+
+/// Depth of an endpoint's async payload ring. Absorbs a burst while a receiver is briefly
+/// busy; a full ring drops the *oldest* message (see `send`).
+const post_capacity: usize = 16;
+
+/// One buffered message: a length-prefixed payload plus the sender's task id (delivered
+/// in the low bits of the receiver's badge).
+const PostSlot = struct {
+    length: u16 = 0,
+    sender_id: u64 = 0,
+    bytes: [POST_MAXIMUM]u8 = undefined,
+};
+
 /// End of the user (low) canonical half — user buffers must lie below it.
 const user_half_end: u64 = 0x0000_8000_0000_0000;
 
@@ -71,6 +99,12 @@ pub const Endpoint = struct {
     notify_buffer: [8]u64 = undefined,
     notify_head: u8 = 0,
     notify_tail: u8 = 0,
+    // Pending buffered messages (payloads posted by `send`), a small FIFO ring. Unlike
+    // notifications — which are a level and coalesce — these are discrete messages, so a
+    // full ring drops the oldest rather than merging.
+    post_buffer: [post_capacity]PostSlot = undefined,
+    post_head: u16 = 0,
+    post_tail: u16 = 0,
 };
 
 pub fn createIpcEndpoint() ?*Endpoint {
@@ -92,6 +126,7 @@ pub fn dropRef(endpoint: *Endpoint) void {
 // --- sender FIFO (endpoint-local, via Task.next) ----------------------------
 
 fn enqueueSender(endpoint: *Endpoint, t: *Task) void {
+    t.ipc_wait_endpoint = @ptrCast(endpoint); // so a kill can unlink a parked caller
     t.next = null;
     if (endpoint.sender_tail) |tail| tail.next = t else endpoint.sender_head = t;
     endpoint.sender_tail = t;
@@ -101,10 +136,34 @@ fn dequeueSender(endpoint: *Endpoint) ?*Task {
     const t = endpoint.sender_head orelse return null;
     endpoint.sender_head = t.next;
     if (endpoint.sender_head == null) endpoint.sender_tail = null;
+    t.ipc_wait_endpoint = null;
     t.next = null;
     return t;
 }
 
+/// Unlink `t` from the sender FIFO it queues in, if any — the kill path for a
+/// client parked in `call` that no server has received yet. Without this, a dead
+/// caller would later be dequeued as a dangling pointer. The endpoint is still
+/// alive here: `t`'s own handle table holds a reference until closeHandles runs
+/// (which the kill path does *after* this). Precondition: the big kernel lock is
+/// held.
+pub fn abandonSenderLocked(t: *Task) void {
+    const endpoint: *Endpoint = @ptrCast(@alignCast(t.ipc_wait_endpoint orelse return));
+    t.ipc_wait_endpoint = null;
+    var previous: ?*Task = null;
+    var node = endpoint.sender_head;
+    while (node) |n| : ({
+        previous = n;
+        node = n.next;
+    }) {
+        if (n != t) continue;
+        if (previous) |p| p.next = t.next else endpoint.sender_head = t.next;
+        if (endpoint.sender_tail == t) endpoint.sender_tail = previous;
+        t.next = null;
+        return;
+    }
+}
+
 // --- cross-address-space copy ----------------------------------------------
 
 /// Copy `len` bytes from `source_va` in address space `source_as` to `destination_va` in
@@ -237,12 +296,23 @@ pub fn replyWait(endpoint: *Endpoint, reply_ptr: u64, reply_len: u64, receive_pt
         scheduler.readyLocked(client); // its `call` now returns
     }
 
-    // (2) Receive the next request (or notification), blocking until one is ready.
+    // (2) Receive the next request (or notification / buffered message), blocking until
+    // one is ready. Bare notifications (IRQ/exit) come first — they're latency-sensitive
+    // and carry no payload — then buffered messages, then synchronous client requests.
     while (true) {
         if (popNotify(endpoint)) |badge| {
             out_badge.* = badge | notify_badge_bit;
             return 0; // notification: no payload, no reply owed, no cap
         }
+        if (popPost(endpoint)) |slot| {
+            const n = @min(@as(usize, slot.length), receive_cap);
+            // Copy from the kernel-resident ring slot (source aspace 0) into the receiver.
+            if (!copyAcross(0, @intFromPtr(&slot.bytes), me.aspace, receive_ptr, n)) {
+                continue; // bad receive buffer: drop this message, keep serving
+            }
+            out_badge.* = slot.sender_id | notify_badge_bit | notify_message_bit;
+            return @intCast(n); // async message: payload delivered, no reply owed, no cap
+        }
         if (dequeueSender(endpoint)) |caller| {
             const n = @min(caller.ipc_send_len, receive_cap);
             if (!copyAcross(caller.aspace, caller.ipc_send_ptr, me.aspace, receive_ptr, n)) {
@@ -278,6 +348,44 @@ fn popNotify(endpoint: *Endpoint) ?u64 {
     return badge;
 }
 
+/// Take the oldest buffered message from the post ring, or null if empty. Returns a
+/// pointer into the endpoint's own storage — valid until the next `send`/`popPost` under
+/// the same lock region, which is all the copy-out in `replyWait` needs.
+fn popPost(endpoint: *Endpoint) ?*const PostSlot {
+    if (endpoint.post_head == endpoint.post_tail) return null;
+    const slot = &endpoint.post_buffer[endpoint.post_head % post_capacity];
+    endpoint.post_head +%= 1;
+    return slot;
+}
+
+/// Client-free side of async IPC (`ipc_send`): copy `[source_va, len)` from address space
+/// `source_as` into `endpoint`'s post ring and wake a waiting receiver — **without
+/// blocking the sender** and with no reply owed. `sender_id` rides along, delivered in the
+/// low bits of the receiver's badge. Returns 0, or a negative errno (`-E2BIG` if the
+/// payload exceeds `POST_MAXIMUM`, `-EFAULT` if the source buffer is unmapped / out of the
+/// user half). A full ring drops the *oldest* message (advancing `post_head`), because a
+/// buffered message is discrete, not a level: keeping the newest keeps input responsive.
+/// Precondition: the big kernel lock is held.
+pub fn sendLocked(endpoint: *Endpoint, source_as: u64, source_va: u64, len: u64, sender_id: u64) i64 {
+    if (len > POST_MAXIMUM) return -E2BIG;
+    // Drop the oldest if the ring is full, so this newest message always lands.
+    if (endpoint.post_tail -% endpoint.post_head >= post_capacity) endpoint.post_head +%= 1;
+    const slot = &endpoint.post_buffer[endpoint.post_tail % post_capacity];
+    if (!copyFromUser(source_as, source_va, slot.bytes[0..@intCast(len)])) return -EFAULT;
+    slot.length = @intCast(len);
+    slot.sender_id = sender_id;
+    endpoint.post_tail +%= 1;
+    scheduler.wakeLocked(&endpoint.receive_wait_queue);
+    return 0;
+}
+
+/// `sendLocked` wrapped in its own critical section, for the `ipc_send` syscall path.
+pub fn send(endpoint: *Endpoint, source_as: u64, source_va: u64, len: u64, sender_id: u64) i64 {
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    return sendLocked(endpoint, source_as, source_va, len, sender_id);
+}
+
 /// Post an asynchronous notification carrying `badge` to `endpoint` and wake a waiting
 /// receiver. Precondition: the big kernel lock is held.
 ///

system/kernel/kernel.zig

@@ -5,6 +5,7 @@ const parameters = @import("parameters");
 const architecture = @import("architecture");
 const console = @import("console.zig");
 const log = @import("log.zig");
+const wall_clock = @import("wall-clock.zig");
 const pmm = @import("pmm.zig");
 const heap = @import("heap.zig");
 const scheduler = @import("scheduler.zig");
@@ -62,6 +63,10 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     architecture.serialInit();
     log.addSink(architecture.serialWrite);
     if (architecture.debugconPresent()) log.addSink(architecture.debugconWrite);
+    // Retain the whole stream in a RAM buffer too, so a user program can later
+    // read it back (klog_read) and persist the boot log to disk — the only way to
+    // see it on a headless/real machine with no host capturing serial.
+    log.addSink(log.ramSink);
 
     // The **framebuffer** is deliberately *not* a log sink. It's a separate output
     // surface — a bootstrap text console today, a graphics device driver later — so
@@ -77,17 +82,17 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     architecture.setFaultHandler(onException);
     architecture.init();
 
-    status("danos: initialising kernel...\n");
+    status("/system/kernel: initialising kernel...\n");
     log.write(if (console.present())
-        "danos: framebuffer console online (bootstrap; graphics driver later)\n"
+        "/system/kernel: framebuffer console online (bootstrap; graphics driver later)\n"
     else
-        "danos: no framebuffer (headless) -> logging to serial/debugcon only\n");
-    log.write("danos: cpu tables online (GDT, IDT, TSS)\n");
+        "/system/kernel: no framebuffer (headless) -> logging to serial/debugcon only\n");
+    log.write("/system/kernel: cpu tables online (GDT, IDT, TSS)\n");
     log.print("  resolution : {d}x{d}\n", .{ fb.width, fb.height });
     log.print("  pitch      : {d} bytes\n", .{fb.pitch});
     log.print("  format     : {s}\n", .{@tagName(fb.format)});
     log.print("  framebuffer: 0x{x:0>16}\n", .{fb.base});
-    log.print (" footprint  : {d} MiB\n", .{(fb.pitch * fb.height) / (1024 * 1024)});
+    log.print(" footprint  : {d} MiB\n", .{(fb.pitch * fb.height) / (1024 * 1024)});
 
     // Summarise the physical memory the loader handed us. The array is danos's
     // own MemoryRegion, so this is a plain slice — no firmware layout in sight.
@@ -105,7 +110,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     const total_bytes = total_pages * abi.page_size;
     const gib = 1 << 30;
 
-    log.write("\ndanos: physical memory\n");
+    log.write("\n/system/kernel: physical memory\n");
     log.print("  total RAM  : {d}.{d:0>2} GiB ({d} MiB) - RAM the firmware reported\n", .{ total_bytes / gib, (total_bytes % gib) * 100 / gib, mib(total_pages) });
     log.print("  usable     : {d} MiB - free RAM (incl. reclaimed boot-services memory)\n", .{mib(usable_pages)});
     log.print("  reserved   : {d} MiB - kernel image, boot stack, ACPI, runtime services\n", .{mib(reserved_pages)});
@@ -119,7 +124,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     // until SMP bring-up; 0 means none was available (we stay uniprocessor).
     ap_trampoline_page = pmm.allocBelow(0x100000) orelse 0;
     const s1 = pmm.stats();
-    log.print("\ndanos: frame allocator online\n", .{});
+    log.print("\n/system/kernel: frame allocator online\n", .{});
     log.print("  free frames: {d} ({d} MiB)\n", .{ s1.free_frames, mib(s1.free_frames) });
     const f0 = pmm.alloc();
     const f1 = pmm.alloc();
@@ -133,14 +138,14 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     // Switch off the firmware's page tables onto our own (with real permissions).
     architecture.enablePaging(pmm.alloc, pmm.free, boot_information);
     log.checkpoint(cp_paging);
-    log.print("\ndanos: paging enabled\n", .{});
+    log.print("\n/system/kernel: paging enabled\n", .{});
     log.print("  page tables: root = 0x{x:0>16}\n", .{architecture.activePageTable()});
     log.print("  kernel segs: {d} (mapped with W^X permissions)\n", .{boot_information.kernel_segment_count});
 
     // Bring up the kernel heap (dynamic allocation), built on the VMM.
     heap.init();
     log.checkpoint(cp_heap);
-    log.write("\ndanos: kernel heap online\n");
+    log.write("\n/system/kernel: kernel heap online\n");
     // Measure the amount of resources the kernel is actually using
     const s2 = pmm.stats();
     log.print("  Kernel footprint: {d} KiB\n", .{kib(s1.free_frames - s2.free_frames)});
@@ -156,7 +161,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     };
     if (platform.discover(boot_information, heap.allocator(), hal)) |devtree| {
         var device_tree = devtree;
-        log.write("\ndanos: device discovery online\n");
+        log.write("\n/system/kernel: device discovery online\n");
         device_tree.dump(log.write);
 
         // Snapshot the device tree for user-space drivers (device_enumerate/claim/
@@ -164,7 +169,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
         devices_broker.init(&device_tree);
         if (devices_broker.dropped > 0) {
             // Otherwise entirely silent: drivers would just never see that hardware.
-            log.print("danos: WARNING {d} device(s) dropped — table full\n", .{devices_broker.dropped});
+            log.print("/system/kernel: WARNING {d} device(s) dropped — table full\n", .{devices_broker.dropped});
         }
 
         // Install the device-IRQ trampolines, so a driver's irq_bind has vectors to
@@ -173,7 +178,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
 
         // Power register map extracted from the FADT + AML, for confidence it parsed.
         const pw = platform.powerInformation();
-        log.write("danos: power\n");
+        log.write("/system/kernel: power\n");
         log.print("  pm1a_cnt   : {s} 0x{x} (width {d})\n", .{ if (pw.pm1a_cnt.mmio) "mmio" else "io", pw.pm1a_cnt.address, pw.pm1a_cnt.width });
         if (pw.s5) |s| {
             log.print("  S5 slp_typ : a={d} b={d}\n", .{ s.slp_typ_a, s.slp_typ_b });
@@ -221,7 +226,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
         });
         if (pinfo.spcr_uart) |u| architecture.serialReconfigure(u.mmio, u.address);
 
-        log.write("danos: platform\n");
+        log.write("/system/kernel: platform\n");
         log.print("  8259 PIC   : {s}\n", .{if (pinfo.pic_present) "present" else "absent"});
         log.print("  lapic base : 0x{x}\n", .{pinfo.lapic_base});
         log.print("  hpet base  : 0x{x}\n", .{hpet_base});
@@ -237,7 +242,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
         if (platform.cpusDropped() > 0)
             log.print("  cpus       : WARNING {d} core(s) beyond pool cap dropped\n", .{platform.cpusDropped()});
     } else |err| {
-        log.print("\ndanos: device discovery failed: {s}\n", .{@errorName(err)});
+        log.print("\n/system/kernel: device discovery failed: {s}\n", .{@errorName(err)});
     }
     log.checkpoint(cp_discovery);
 
@@ -248,19 +253,43 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     // Register the current context as the first task before enabling preemption.
     scheduler.init(4);
     log.checkpoint(cp_scheduler);
-    log.write("\ndanos: scheduler online\n");
+    log.write("\n/system/kernel: scheduler online\n");
 
     // Start the timer and unmask interrupts — the kernel now has a heartbeat, and
     // the timer preempts among tasks.
     architecture.startTimer();
     architecture.enableInterrupts();
     log.checkpoint(cp_timer);
-    log.print("danos: timer online ({d} Hz tick; timer clock {d} MHz, clock {d} MHz; calibrated via {s})\n", .{ architecture.timer_hz, architecture.timerClockHz() / 1_000_000, architecture.clockHz() / 1_000_000, architecture.timerCalibrationSource() });
+    log.print("/system/kernel: timer online ({d} Hz tick; timer clock {d} MHz, clock {d} MHz; calibrated via {s})\n", .{ architecture.timer_hz, architecture.timerClockHz() / 1_000_000, architecture.clockHz() / 1_000_000, architecture.timerCalibrationSource() });
+
+    // The tsc-sync test forces the TSC clocksource on before the cores come up, so the
+    // TSC + warp-check path is exercised even under TCG (which won't advertise an
+    // invariant TSC). Inert in a normal build (docs/timers.md).
+    if (build_options.test_case) |tc| {
+        if (std.mem.eql(u8, tc, "tsc-sync")) architecture.forceTscClocksourceForTest();
+    }
 
     // Wake the other cores (application processors). A no-op on a single-core
-    // machine; on SMP each AP climbs to long mode and reports in (docs/smp.md).
+    // machine; on SMP each AP climbs to long mode and reports in (docs/smp.md). The
+    // per-core TSC warp check rides this: each AP is vetted before it joins the run
+    // loop (docs/timers.md).
     bringUpSecondaries();
 
+    // Report the monotonic clock's final reliability, now the warp check has run on
+    // every core. On real Intel/AMD this is the invariant, synchronized TSC; a bare
+    // VM (no invariant bit) or a machine whose cores' TSCs skew uses the HPET instead.
+    log.print("/system/kernel: clocksource {s} (TSC invariant: {s}, synchronized: {s})\n", .{
+        architecture.clockSourceName(),
+        if (architecture.clockInvariant()) "yes" else "no",
+        if (architecture.clockSynchronized()) "yes" else "no",
+    });
+    if (!architecture.clockSynchronized())
+        log.write("/system/kernel: WARNING: per-core TSCs are not synchronized; monotonic clock moved off the TSC\n");
+
+    // Anchor wall-clock time: read the RTC once, now the monotonic clock is final.
+    wall_clock.init();
+    log.print("/system/kernel: wall clock {d} (Unix epoch seconds, UTC, from the RTC)\n", .{wall_clock.nowSeconds()});
+
     // In a test build (`zig build -Dtest-case=<name>`), run that case and stop.
     // Normal builds fall through to the idle halt.
     if (build_options.test_case) |case| {
@@ -269,7 +298,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     }
 
     log.checkpoint(cp_running);
-    status("kernel initialised.\n");
+    status("/system/kernel: initialised.\n");
 
     // Publish the initial-ramdisk so user space can `system_spawn` its bundled
     // binaries by name. The kernel no longer launches them itself: init is the
@@ -282,10 +311,10 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     // manager then discovers the hardware and spawns each driver. init runs on its own
     // address space, preemptively — this boot context becomes the BSP's idle loop.
     if (boot_information.init_len != 0) {
-        status("starting /system/services/init...\n");
+        status("/system/kernel: starting /system/services/init...\n");
         const image = @as([*]const u8, @ptrFromInt(boot_handoff.physicalToVirtual(boot_information.init_base)))[0..boot_information.init_len];
-        process.spawnProcess(image, 4) catch |err| {
-            statusPrint("/system/services/init failed to load: {s}\n", .{@errorName(err)});
+        process.spawnProcess(image, 4, &.{"/system/services/init"}) catch |err| {
+            statusPrint("/system/kernel: /system/services/init failed to load: {s}\n", .{@errorName(err)});
         };
     } else {
         status("no /system/services/init on the boot volume.\n");
@@ -294,7 +323,7 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     // Become the idle task: drop below every real task and halt until an
     // interrupt. The timer keeps preempting into init and any other work.
     scheduler.setPriority(0);
-    status("\nkernel idle; user space is running.\n");
+    status("\n/system/kernel: kernel idle; user space is running.\n");
     architecture.halt();
 }
 
@@ -321,7 +350,7 @@ fn bringUpSecondaries() void {
     // vector addresses it). It's kept for the system's life — armed only during a
     // wake, inert (zeroed, non-executable) otherwise — so cores can be re-woken later.
     if (ap_trampoline_page == 0) {
-        log.write("danos: smp: no low page for the AP trampoline; staying uniprocessor\n");
+        log.write("/system/kernel: smp: no low page for the AP trampoline; staying uniprocessor\n");
         return;
     }
     architecture.setTrampolinePage(ap_trampoline_page);
@@ -333,7 +362,7 @@ fn bringUpSecondaries() void {
         if (std.mem.eql(u8, tc, "smp-retry")) architecture.testFailNextWakes(1);
     }
 
-    log.print("\ndanos: bringing up {d} application processor(s)\n", .{cores.len - 1});
+    log.print("\n/system/kernel: bringing up {d} application processor(s)\n", .{cores.len - 1});
     const maximum_wake_attempts = 3; // a core that misses the first INIT-SIPI-SIPI gets retried
     for (cores[1..], 1..) |core, index| {
         const stack = heap.allocator().alloc(u8, parameters.kernel_stack_size) catch {
@@ -344,7 +373,7 @@ fn bringUpSecondaries() void {
         // This core's dedicated fault stack — allocated only now that the core is
         // real, rather than reserved statically for every possible core.
         const fault_stack = heap.allocator().alloc(u8, architecture.fault_stack_size) catch {
-            log.print("  cpu apic_id {d}: no fault stack; skipped\n", .{core.apic_id});
+            log.print("/system/kernel:  cpu apic_id {d}: no fault stack; skipped\n", .{core.apic_id});
             continue;
         };
         architecture.setFaultStack(index, (@intFromPtr(fault_stack.ptr) + fault_stack.len) & ~@as(usize, 15));
@@ -353,14 +382,14 @@ fn bringUpSecondaries() void {
         while (attempt <= maximum_wake_attempts) : (attempt += 1) {
             if (architecture.startSecondary(core.apic_id, stack_top, @intFromPtr(pc), index)) {
                 pc.online = true;
-                log.print("  cpu apic_id {d}: online (attempt {d})\n", .{ core.apic_id, attempt });
+                log.print("/system/kernel:  cpu apic_id {d}: online (attempt {d})\n", .{ core.apic_id, attempt });
                 break;
             }
             if (attempt == maximum_wake_attempts)
-                log.print("  cpu apic_id {d}: no response after {d} attempts (parked)\n", .{ core.apic_id, maximum_wake_attempts });
+                log.print("/system/kernel:  cpu apic_id {d}: no response after {d} attempts (parked)\n", .{ core.apic_id, maximum_wake_attempts });
         }
     }
-    log.print("danos: {d}/{d} cores online\n", .{ scheduler.onlineCount(), cores.len });
+    log.print("/system/kernel: {d}/{d} cores online\n", .{ scheduler.onlineCount(), cores.len });
 }
 
 /// A user-facing status line: to the diagnostic `log` *and* the on-screen console
@@ -385,13 +414,55 @@ fn kib(frames: u64) u64 {
     return frames * abi.page_size / (1024);
 }
 
-/// Report a CPU exception and halt **this core**. There's no fault recovery yet, so
-/// the faulting core is terminal — but the fault is *contained* to it: on an
-/// application processor only that core stops, and the rest of the system keeps
-/// running (full recovery — kill the task, keep the core — is the resilience track,
-/// see docs/resilience.md). The report names the core so an AP fault is attributed,
-/// and goes to every output sink plus a POST code and a persistent breadcrumb.
+/// Whether a ring-3 exception is attributable to the process that raised it — and
+/// therefore recoverable by killing that process. NMI (2), double fault (8), and
+/// machine check (18) report machine or kernel trouble even when they arrive with a
+/// user CS (an NMI interrupts whatever happens to be running), so they stay terminal.
+fn recoverableFault(vector: u64) bool {
+    return switch (vector) {
+        2, 8, 18 => false,
+        else => true,
+    };
+}
+
+/// Report a CPU exception. Two outcomes (docs/resilience.md):
+///
+/// **A fault taken in user mode kills the faulting process, not the machine.** The
+/// kernel is intact — the CPU trapped onto the task's kernel stack — so the process
+/// is killed, everything it held (address space, IRQ bindings, IPC handles, device
+/// grants' frames) is reclaimed, and the core reschedules. A crashing driver takes
+/// itself down, never the OS.
+///
+/// **Everything else halts this core.** A kernel-mode fault means the trusted base
+/// itself is broken — there is nothing safe to kill — and NMI/#DF/#MC report machine
+/// trouble regardless of CS (`recoverableFault`). Even then the fault is *contained*:
+/// on an application processor only that core stops and the rest keep running. The
+/// report names the core so an AP fault is attributed, and goes to every output sink
+/// plus a POST code and a persistent breadcrumb. (A ring-3 fault on a *borrowed*
+/// kernel thread — process.run, the user-pf isolation probe — also lands here: there
+/// is no scheduled process to kill.)
+/// Classify a CPU exception vector as the ExitReason a supervisor reads — the
+/// fault classes of docs/process-lifecycle.md. Faults are exit reasons, never
+/// signals delivered to the faulting process: recovery is restart, not a handler.
+fn exitReasonForVector(vector: u64) abi.ExitReason {
+    return switch (vector) {
+        14 => .segmentation_fault, // page fault
+        6 => .illegal_instruction, // invalid opcode
+        0, 16, 19 => .arithmetic_fault, // divide error, x87, SIMD
+        13 => .protection_fault, // general protection
+        else => .fault,
+    };
+}
+
 fn onException(state: *const architecture.CpuState) noreturn {
+    if (architecture.fromUser(state) and scheduler.currentIsUserProcess() and recoverableFault(state.vector)) {
+        statusPrint("\n/system/kernel: process {d} ({s}) killed by {s} (vector {d}) on core {d}\n", .{ scheduler.currentId(), scheduler.current().name(), architecture.exceptionName(state.vector), state.vector, scheduler.currentCpuIndex() });
+        statusPrint("  error code : 0x{x}\n", .{state.error_code});
+        statusPrint("  IP         : 0x{x:0>16}\n", .{architecture.instructionPointer(state)});
+        if (architecture.faultAddress(state)) |address| statusPrint("  fault addr : 0x{x:0>16}\n", .{address});
+        process.killCurrentProcess(exitReasonForVector(state.vector)); // reclaims everything, reschedules; never returns
+    }
+
     log.checkpoint(cp_exception);
     const core = scheduler.currentCpuIndex();
     // A fault is user-facing enough to paint on screen too (via statusPrint), on

system/kernel/log.zig

@@ -41,6 +41,39 @@ pub fn write(bytes: []const u8) void {
     for (sinks[0..sink_count]) |sink| sink(bytes);
 }
 
+// --- the RAM sink: a retained copy of the whole diagnostic stream ------------
+//
+// A fixed in-image buffer that accumulates every logged byte, so a user program
+// (`log-flush`, and init at shutdown) can read it back through `klog_read` and
+// persist it to a file — the boot log survives on a headless/real machine that
+// has no host capturing serial. It is a *sink like any other*: register it with
+// `addSink(ramSink)` at boot. No allocation (works pre-heap and in a panic).
+//
+// It fills linearly and stops when full: the earliest output — the most valuable
+// for diagnosing a boot — is kept, and the tail is still on the live serial sink.
+// 256 KiB comfortably holds a full boot plus a long run (a boot is ~15 KiB).
+
+const ram_capacity = 256 * 1024;
+var ram_buffer: [ram_capacity]u8 = undefined;
+var ram_len: usize = 0;
+
+/// The RAM sink. Best-effort and self-guarding like every sink: appends what fits
+/// and silently drops the rest once full. (Concurrency matches the other sinks —
+/// the dominant writer, debug_write, already holds the kernel lock; a rare torn
+/// append on a kernel-internal line is an accepted diagnostic imperfection.)
+pub fn ramSink(bytes: []const u8) void {
+    const n = @min(ram_buffer.len - ram_len, bytes.len);
+    if (n != 0) {
+        @memcpy(ram_buffer[ram_len..][0..n], bytes[0..n]);
+        ram_len += n;
+    }
+}
+
+/// The accumulated log so far — what `klog_read` copies out.
+pub fn ramSnapshot() []const u8 {
+    return ram_buffer[0..ram_len];
+}
+
 /// A formatted log line. Truncates past 256 bytes; the buffer is on the stack, so
 /// this is safe to call from interrupt context and from a panic.
 pub fn print(comptime fmt: []const u8, args: anytype) void {

system/kernel/process.zig

@@ -24,6 +24,7 @@ const elf = std.elf;
 const boot_handoff = @import("boot-handoff");
 const abi = @import("abi");
 const device_abi = @import("device-abi");
+const parameters = @import("parameters");
 const architecture = @import("architecture");
 const pmm = @import("pmm.zig");
 const scheduler = @import("scheduler.zig");
@@ -33,6 +34,7 @@ const devices_broker = @import("devices-broker.zig");
 const irq = @import("irq.zig");
 const initial_ramdisk = @import("initial-ramdisk");
 const log = @import("log.zig");
+const wall_clock = @import("wall-clock.zig");
 
 const page_size = abi.page_size;
 const SystemCall = abi.SystemCall;
@@ -40,9 +42,18 @@ const SystemCall = abi.SystemCall;
 /// User virtual addresses. PML4 index 224 — a user-exclusive region, far from
 /// the identity map (low indices) and the vmm test address (index 128), so
 /// setting the U/S bit on its intermediate tables widens no kernel mapping.
-/// An ELF image may occupy [code_virtual, stack_virtual); the stack page sits above.
+/// An ELF image may occupy [code_virtual, stack_virtual); the stack sits above.
 pub const code_virtual: u64 = 0x0000_7000_0000_0000;
-pub const stack_virtual: u64 = 0x0000_7000_0020_0000;
+
+/// The stack region, above the image. The page at `stack_virtual` is **never
+/// mapped** — it is the guard page: a process that overflows its stack walks into
+/// it and faults (killing only that process) rather than silently corrupting the
+/// top of its own image. The stack proper is `parameters.user_stack_pages` pages
+/// at [stack_base_virtual, stack_top_virtual), RW + NX, with the System V entry
+/// block (argc/argv) at the very top.
+pub const stack_virtual: u64 = 0x0000_7000_0020_0000; // guard page (unmapped)
+pub const stack_base_virtual: u64 = stack_virtual + page_size;
+pub const stack_top_virtual: u64 = stack_base_virtual + parameters.user_stack_pages * page_size;
 
 /// The mmap grant arena: where `mmap` hands out fresh user pages, above the image
 /// and stack but still inside PML4[224] (so no kernel mapping is widened). Each
@@ -73,6 +84,19 @@ pub const dma_arena_end: u64 = dma_arena_base + (256 << 20); // 256 MiB per proc
 /// chunks, so this bound is generous; it also caps the frame scratch array below.
 const maximum_mmap_pages = 256;
 
+/// Ceiling on a process's argv entries, including argv[0]. Arguments are spawn
+/// parameters ("you are the driver for device 12"), not bulk data — IPC carries
+/// that — so the bound is small and everything fits the single stack page.
+pub const maximum_arguments = 8;
+
+/// Ceiling on the `system_spawn` extra-arguments blob (argv[1..], NUL-separated).
+pub const maximum_argument_bytes = 256;
+
+/// Auxiliary-vector entry types (System V AMD64 process entry). Only what the
+/// kernel emits today; a C runtime scans the vector until the null terminator.
+const auxiliary_vector_null: u64 = 0; // AT_NULL — end of the vector
+const auxiliary_vector_page_size: u64 = 6; // AT_PAGESZ
+
 // The hand-assembled user program blob (isr.s, .rodata) — the isolation probe.
 const pf_start = @extern([*]const u8, .{ .name = "user_pf_start" });
 const pf_end = @extern([*]const u8, .{ .name = "user_pf_end" });
@@ -106,9 +130,15 @@ pub fn setInitialRamdisk(image: []const u8) void {
 /// written back into the trap frame, since the entry paths restore user registers
 /// from it. One handler serves both the system_call/sysret and int-0x80 entry paths.
 ///
-/// Install it once at boot (before any user code runs) via `init`.
+/// Install it once at boot (before any user code runs) via `init`. Also registers
+/// the scheduler's kill hooks: the scheduler sits below this layer, so finishing a
+/// deferred process_kill (IRQ bindings, IPC handles, the exit notification) is
+/// called back up into here from the tick (see scheduler.reapKillPendingLocked).
 pub fn init() void {
     architecture.setSystemCallHandler(system_call);
+    scheduler.terminate_current_hook = terminateCurrentLocked;
+    scheduler.reap_task_hook = reapTaskLocked;
+    scheduler.timer_tick_hook = timerSweepLocked;
 }
 
 /// Return -1 (as an unsigned bit pattern) in the system_call result register.
@@ -117,21 +147,27 @@ fn fail(state: *architecture.CpuState) void {
 }
 
 fn system_call(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    const user = t.aspace != 0;
+    if (user) {
+        // A condemned process (process_kill caught it running) dies at its next
+        // kernel entry — before it can spawn, claim, or message anything else.
+        if (t.kill_pending) terminateCurrent();
+        // Mark the span of this call so the timer tick never tears the task down
+        // in the middle of a kernel operation (scheduler.reapKillPendingLocked).
+        t.in_system_call = true;
+    }
+    defer if (user) {
+        t.in_system_call = false;
+    };
     switch (@as(SystemCall, @enumFromInt(architecture.systemCallNumber(state)))) {
         .exit => {
             exit_code = architecture.systemCallArg(state, 0);
-            // A scheduled process drops its endpoint references, frees its address
-            // space, and reschedules; a borrowed test thread unwinds back to the
-            // kernel that entered it.
+            // A scheduled process tears down fully (terminateCurrent); a borrowed
+            // test thread unwinds back to the kernel that entered it.
             if (scheduler.currentIsUserProcess()) {
-                // Unbind before closeHandles: dropping the last reference destroys the
-                // Endpoint, and a still-bound GSI would have an ISR call
-                // notifyFromIsr on freed memory the next time the device fired.
-                // unbindAll also leaves the line masked, so a dead driver's device
-                // goes quiet rather than storming.
-                releaseIrqs(scheduler.current());
-                ipc.closeHandles(scheduler.current());
-                scheduler.exitUser();
+                scheduler.current().exit_reason = .exited;
+                terminateCurrent();
             } else architecture.userExit();
         },
         .yield => {
@@ -150,6 +186,7 @@ fn system_call(state: *architecture.CpuState) void {
         .ipc_lookup => systemIpcLookup(state),
         .ipc_call => systemIpcCall(state),
         .ipc_reply_wait => systemIpcReplyWait(state),
+        .ipc_send => systemIpcSend(state),
         .device_enumerate => systemDeviceEnumerate(state),
         .device_claim => systemDeviceClaim(state),
         .mmio_map => systemMmioMap(state),
@@ -163,6 +200,15 @@ fn system_call(state: *architecture.CpuState) void {
         .io_read => systemIoRead(state),
         .io_write => systemIoWrite(state),
         .clock => systemClock(state),
+        .process_enumerate => systemProcessEnumerate(state),
+        .process_kill => systemProcessKill(state),
+        .process_exit_reason => systemProcessExitReason(state),
+        .process_subscribe => systemProcessSubscribe(state),
+        .signal_bind => systemSignalBind(state),
+        .process_signal => systemProcessSignal(state),
+        .timer_bind => systemTimerBind(state),
+        .klog_read => systemKlogRead(state),
+        .wall_clock => systemWallClock(state),
         _ => fail(state),
     }
 }
@@ -228,6 +274,18 @@ fn systemIpcReplyWait(state: *architecture.CpuState) void {
     architecture.setSystemCallResult3(state, received_cap);
 }
 
+/// ipc_send(handle, message_ptr, message_len) -> 0/-errno: post a payload to an
+/// endpoint's async queue and wake a receiver, without blocking the caller. The async
+/// counterpart of ipc_call — for broadcasts (the input service) where a rendezvous would
+/// let one dead subscriber hang the sender. Delivered through ipc_reply_wait as a
+/// buffered message (badge carries notify_message_bit and the caller's task id).
+fn systemIpcSend(state: *architecture.CpuState) void {
+    const me = scheduler.current();
+    const endpoint = ipc.resolveHandle(me, architecture.systemCallArg(state, 0)) orelse return failErr(state, ipc.EBADF);
+    const r = ipc.send(endpoint, me.aspace, architecture.systemCallArg(state, 1), architecture.systemCallArg(state, 2), me.id);
+    architecture.setSystemCallResult(state, @bitCast(r));
+}
+
 /// device_enumerate(buffer, maximum) -> total: snapshot the device table into the caller's
 /// buffer (up to `maximum` entries), returning the total device count.
 fn systemDeviceEnumerate(state: *architecture.CpuState) void {
@@ -243,6 +301,8 @@ fn systemDeviceEnumerate(state: *architecture.CpuState) void {
 
 /// device_claim(id) -> 0/-1: take exclusive ownership of a device for this process.
 fn systemDeviceClaim(state: *architecture.CpuState) void {
+    const claim_flags = sync.enter();
+    defer sync.leave(claim_flags);
     if (devices_broker.claim(architecture.systemCallArg(state, 0), scheduler.current().id))
         architecture.setSystemCallResult(state, 0)
     else
@@ -257,10 +317,22 @@ fn systemMmioMap(state: *architecture.CpuState) void {
     const resource_index = architecture.systemCallArg(state, 1);
     const t = scheduler.current();
     if (t.aspace == 0) return fail(state);
+    // Read the broker table under the lock: ring-3 device_register (M19) now
+    // mutates it concurrently on other cores, so a lock-free read here could
+    // see a torn resource (and a torn length used to panic the arithmetic
+    // below on integer overflow).
+    const r = blk: {
+        const flags = sync.enter();
+        defer sync.leave(flags);
         const owner = devices_broker.ownerOf(device_id) orelse return fail(state);
         if (owner != t.id) return fail(state); // not claimed by this process
-    const r = devices_broker.resourceOf(device_id, resource_index) orelse return fail(state);
+        break :blk devices_broker.resourceOf(device_id, resource_index) orelse return fail(state);
+    };
     if (r.kind != @intFromEnum(device_abi.ResourceKind.memory)) return fail(state);
+    // A zero-length or wrapping window is not mappable — fail cleanly rather
+    // than underflow `r.len - 1`.
+    if (r.len == 0) return fail(state);
+    if (@addWithOverflow(r.start, r.len)[1] != 0) return fail(state);
 
     if (t.device_map_next == 0) t.device_map_next = device_arena_base;
     const first = r.start & ~@as(u64, page_size - 1);
@@ -396,44 +468,406 @@ fn systemDeviceRegister(state: *architecture.CpuState) void {
     var descriptor: device_abi.DeviceDescriptor = undefined;
     if (!ipc.copyFromUser(t.aspace, descriptor_ptr, std.mem.asBytes(&descriptor))) return fail(state);
 
+    // Under the big kernel lock: the broker's table is also mutated by the
+    // death sweep (releaseAllOwnedBy) and read by enumerate on other cores —
+    // ring-3 registration (M19) made those genuinely concurrent.
+    const flags = sync.enter();
+    defer sync.leave(flags);
     const id = devices_broker.register(parent_id, t.id, &descriptor) catch return fail(state);
     architecture.setSystemCallResult(state, id);
 }
 
-/// system_spawn(name_ptr, name_len) -> 0 on success, -1 on failure. Load the binary
-/// bundled in the initial-ramdisk under `name` as a fresh ring-3 process. This is the
-/// mechanism a user-space supervisor (the device manager) uses to start a driver it
-/// matched: discovery and policy stay in user space, the kernel only spawns.
+/// system_spawn(name_ptr, name_len, arguments_ptr, arguments_len, exit_endpoint)
+/// -> the child's process id on success, -1 on failure. Load the binary bundled in
+/// the initial-ramdisk under `name` as a fresh ring-3 process. `name` becomes the
+/// child's argv[0] (and its task name, so a fault report can say which binary
+/// died); `arguments` is an optional NUL-separated blob that becomes argv[1..] —
+/// how a supervisor parameterises what it starts ("you are the driver for device
+/// 12"). 0/0 means no extra arguments. This is the mechanism a user-space
+/// supervisor (the device manager) uses to start a driver it matched: discovery
+/// and policy stay in user space, the kernel only spawns.
 ///
-/// Ungated for now — any process may spawn any bundled binary. A capability (only a
-/// supervisor holds the right to spawn) belongs here once the model grows one; see
-/// docs/driver-model.md. The name is bounds-checked into the user half exactly like
-/// `debug_write`, and an unknown name or a load failure returns -1.
+/// The caller is recorded as the child's **supervisor** — the sole holder of the
+/// right to `process_kill` it (docs/process-management.md). `exit_endpoint` (a
+/// handle, or `abi.no_cap` for none) names an endpoint of the caller's to notify
+/// when the child ends, any way it ends — the IRQ-as-IPC pattern reused as the
+/// microkernel's SIGCHLD.
+///
+/// Spawning itself is still ungated — any process may spawn any bundled binary; a
+/// spawn capability belongs here once the model grows one (docs/driver-model.md).
+/// Both buffers are bounds-checked into the user half exactly like `debug_write`,
+/// and an unknown name or a load failure returns -1.
 fn systemSpawn(state: *architecture.CpuState) void {
     const ptr = architecture.systemCallArg(state, 0);
     const len = architecture.systemCallArg(state, 1);
-    if (len == 0 or len > 64 or ptr >= user_half_end or ptr + len > user_half_end) return fail(state);
+    const arguments_ptr = architecture.systemCallArg(state, 2);
+    const arguments_len = architecture.systemCallArg(state, 3);
+    const exit_handle = architecture.systemCallArg(state, 4);
+    const t = scheduler.current();
+    if (len == 0 or len > scheduler.maximum_task_name or ptr >= user_half_end or ptr + len > user_half_end) return fail(state);
+    if (arguments_len > maximum_argument_bytes) return fail(state);
+    if (arguments_len != 0 and (arguments_ptr >= user_half_end or arguments_ptr + arguments_len > user_half_end)) return fail(state);
+    const exit_endpoint: ?*ipc.Endpoint = if (exit_handle == abi.no_cap)
+        null
+    else
+        ipc.resolveHandle(t, exit_handle) orelse return failErr(state, ipc.EBADF);
     const image = ramdisk_image orelse return fail(state);
     const rd = initial_ramdisk.Reader.init(image) orelse return fail(state);
 
     const name = @as([*]const u8, @ptrFromInt(ptr))[0..len];
+    var argv: [maximum_arguments][]const u8 = undefined;
+    argv[0] = name;
+    var argc: usize = 1;
+    if (arguments_len != 0) {
+        const blob = @as([*]const u8, @ptrFromInt(arguments_ptr))[0..arguments_len];
+        var pieces = std.mem.tokenizeScalar(u8, blob, 0);
+        while (pieces.next()) |piece| {
+            if (argc == maximum_arguments) return fail(state);
+            argv[argc] = piece;
+            argc += 1;
+        }
+    }
+
     var i: u32 = 0;
     while (i < rd.count) : (i += 1) {
         const item = rd.entry(i) orelse continue;
         if (!std.mem.eql(u8, item.name, name)) continue;
-        spawnProcess(item.blob, 4) catch return fail(state);
-        architecture.setSystemCallResult(state, 0);
+        const child = spawnProcessSupervised(item.blob, 4, argv[0..argc], t.id, exit_endpoint) catch return fail(state);
+        architecture.setSystemCallResult(state, child);
         return;
     }
     fail(state); // no bundled binary by that name
 }
 
-/// Drop every IRQ binding `t` made. Called on exit, before the handle table is closed
-/// (which is what frees the endpoints an ISR would otherwise notify into).
-fn releaseIrqs(t: *scheduler.Task) void {
+/// process_enumerate(buffer, maximum) -> total: snapshot the task table into the
+/// caller's buffer (up to `maximum` `abi.ProcessDescriptor` entries), returning
+/// the total live-task count — the exact shape of `device_enumerate`, so a `ps`
+/// is a user program over a snapshot, not a kernel service. Read-only and
+/// ungated: what is running is not a secret between cooperating bring-up
+/// processes.
+fn systemProcessEnumerate(state: *architecture.CpuState) void {
+    const buffer_ptr = architecture.systemCallArg(state, 0);
+    const maximum = architecture.systemCallArg(state, 1);
+    const t = scheduler.current();
+    if (t.aspace == 0 or buffer_ptr >= user_half_end) return fail(state);
+    const sz = @sizeOf(abi.ProcessDescriptor);
+    const cap = @min(maximum, (user_half_end - buffer_ptr) / sz); // clamp to the user half
+    const out: [*]abi.ProcessDescriptor = @ptrFromInt(buffer_ptr);
+    architecture.setSystemCallResult(state, scheduler.enumerate(out[0..@intCast(cap)]));
+}
+
+/// process_kill(id) -> 0 / -ESRCH / -EPERM: end the process `id`. Only its
+/// supervisor — the process that spawned it — may do so; the supervision link is
+/// the kill capability, so no user/permission model is needed and a stray id
+/// cannot be a weapon (ids are never reused, so a stale one just misses).
+fn systemProcessKill(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    if (t.aspace == 0) return fail(state);
+    const id = architecture.systemCallArg(state, 0);
+    if (id > std.math.maxInt(u32)) return failErr(state, ipc.ESRCH);
+    const r = killProcess(t.id, @intCast(id));
+    architecture.setSystemCallResult(state, @bitCast(r));
+}
+
+/// Processes killed by a CPU fault rather than a clean exit. Evidence for the
+/// fault-recovery test, and a health signal a supervisor can consult later.
+pub var fault_kill_count: u64 = 0;
+
+/// Release everything a dying task holds and tell its supervisor — the shared
+/// half of every path out of a process: clean exit, fault kill, and process_kill
+/// (both the immediate reap and the deferred tick-time terminate). The order
+/// matters:
+/// - IRQ bindings are dropped before the handle table closes: dropping the last
+///   endpoint reference destroys the Endpoint, and a still-bound GSI would have an
+///   ISR call notifyFromIsr on freed memory the next time the device fired.
+///   `releaseOwner` also leaves the line masked, so a dead driver's device goes
+///   quiet rather than storming. (It drops MSI vectors by the same owner sweep.)
+/// - Device claims are released with the IRQ bindings, so a restarted driver can
+///   claim the same hardware again — the cleanup half of process-lifecycle.md's
+///   iron rule 1. Claims hold no pointers, so ordering is free; they go here so
+///   the exit notification (below, last) observes a fully-released child.
+/// - A client this task still owes a reply to (it died between receive and reply)
+///   is failed with -EPEER rather than left blocked forever — a dead server must
+///   not hang its callers.
+/// - The task is unlinked from wherever IPC parked it (an endpoint's sender FIFO,
+///   a receive wait queue, or a server's owed-reply slot) *before* the handles
+///   close, so nothing ever dequeues a dangling pointer. These are no-ops for a
+///   running task ending itself; they matter when process_kill reaps a blocked one.
+/// - The exit notification is posted last, once the process can no longer act, so
+///   a supervisor that receives it observes a fully-released child. The endpoint
+///   reference taken at spawn is dropped with it.
+/// Precondition: the big kernel lock is held.
+fn releaseTaskResourcesLocked(t: *scheduler.Task) void {
+    recordExitLocked(t);
+    irq.releaseOwner(t.id);
+    devices_broker.releaseAllOwnedBy(t.id);
+    // The dying task's signal endpoint and one-shot timers go with it.
+    if (t.signal_endpoint) |raw| {
+        ipc.dropRef(@ptrCast(@alignCast(raw)));
+        t.signal_endpoint = null;
+    }
+    t.pending_signals = 0;
+    for (&one_shot_timers) |*slot| {
+        if (slot.*) |timer| {
+            if (timer.owner == t.id) {
+                ipc.dropRef(timer.endpoint);
+                slot.* = null;
+            }
+        }
+    }
+    // A dead subscriber's own subscriptions go first: it must not hear about
+    // itself, and the slots' endpoint references drop with it.
+    for (&exit_subscribers) |*slot| {
+        if (slot.*) |subscriber| {
+            if (subscriber.owner == t.id) {
+                ipc.dropRef(subscriber.endpoint);
+                slot.* = null;
+            }
+        }
+    }
+    if (t.ipc_client) |client| {
+        t.ipc_client = null;
+        client.ipc_status = -ipc.EPEER;
+        scheduler.readyLocked(client); // its blocked `call` now returns the error
+    }
+    ipc.abandonSenderLocked(t);
+    scheduler.removeFromWaitQueueLocked(t);
+    scheduler.forgetIpcClientLocked(t);
+    ipc.closeHandles(t);
+    // Publish the exit to every subscriber (docs/process-lifecycle.md): the same
+    // badge encoding as the supervisor's notification, and equally late, so a
+    // subscriber also observes a fully-released child.
+    for (&exit_subscribers) |*slot| {
+        if (slot.*) |subscriber| ipc.notifyLocked(subscriber.endpoint, abi.notify_exit_bit | t.id);
+    }
+    if (t.exit_endpoint) |raw| {
+        const endpoint: *ipc.Endpoint = @ptrCast(@alignCast(raw));
+        t.exit_endpoint = null;
+        ipc.notifyLocked(endpoint, abi.notify_exit_bit | t.id);
+        ipc.dropRef(endpoint);
+    }
+}
+
+/// Tear down the current user process and reschedule; never returns. Shared by the
+/// exit system call and the fault path (`killCurrentProcess`). See
+/// `releaseTaskResourcesLocked` for what is released, and in what order.
+pub fn terminateCurrent() noreturn {
+    _ = sync.enter(); // handed off through the exit switch, released by the resumed task
+    terminateCurrentLocked();
+}
+
+/// The body of `terminateCurrent` for a caller that already holds the big kernel
+/// lock — the scheduler's tick calls this (via `terminate_current_hook`) to finish
+/// a deferred process_kill on its own core's current task. Never returns; the
+/// tick's abandoned interrupt frame is fine (the LAPIC was acknowledged before the
+/// tick hook ran), exactly as on the fault path.
+fn terminateCurrentLocked() noreturn {
+    releaseTaskResourcesLocked(scheduler.current());
+    scheduler.exitUserLocked();
+}
+
+/// Reap a condemned task that is NOT running on any core (ready or blocked — and
+/// it cannot start running: state changes need the lock we hold). The other half
+/// of a deferred process_kill, called by the scheduler's tick (via
+/// `reap_task_hook`) and directly by `killProcess` for targets caught off-CPU.
+/// Precondition: the big kernel lock is held.
+fn reapTaskLocked(t: *scheduler.Task) void {
+    releaseTaskResourcesLocked(t);
+    scheduler.removeFromReadyQueueLocked(t); // no-op unless it was ready in a queue
+    scheduler.destroyTaskLocked(t);
+}
+
+/// Kill process `target_id` on behalf of `caller_id` — the kernel half of the
+/// process_kill system call. Returns 0, -ESRCH (no such live process — kernel
+/// tasks are not killable processes and stale ids miss, since ids are never
+/// reused), or -EPERM (the caller is not the target's supervisor).
+///
+/// A target that is ready or blocked is reaped on the spot. One that is running
+/// on another core cannot be torn down mid-instruction, so it is condemned
+/// (`kill_pending`) and dies at its next system_call entry, block, or timer tick
+/// — like a Unix signal, delivery is prompt but asynchronous. Either way the
+/// call returns 0: the kill is accepted and irrevocable.
+pub fn killProcess(caller_id: u32, target_id: u32) i64 {
     const flags = sync.enter();
     defer sync.leave(flags);
-    irq.releaseOwner(t.id);
+    const target = scheduler.taskByIdLocked(target_id) orelse return -ipc.ESRCH;
+    if (target.aspace == 0) return -ipc.ESRCH; // kernel tasks are not processes
+    if (target.supervisor != caller_id) return -ipc.EPERM;
+    target.exit_reason = .killed;
+    if (target.state == .running) {
+        target.kill_pending = true;
+    } else {
+        reapTaskLocked(target);
+    }
+    return 0;
+}
+
+/// Kill the current user process in response to a CPU fault it raised in ring 3.
+/// The fault is confined to the process — the kernel trapped it on the task's own
+/// kernel stack and is intact — so everything the process held is reclaimed and the
+/// core reschedules. The system keeps running; only the faulting process dies
+/// (docs/resilience.md: fault -> kill -> continue). `reason` is the fault class
+/// (from the vector), recorded for the supervisor's `process_exit_reason`.
+pub fn killCurrentProcess(reason: abi.ExitReason) noreturn {
+    scheduler.current().exit_reason = reason;
+    fault_kill_count += 1;
+    terminateCurrent();
+}
+
+/// The bounded record of recent deaths, for `process_exit_reason`: ids are never
+/// reused, so a ring keyed by id is enough — a record evicted by wraparound reads
+/// as -ESRCH, the same as an id that never lived, which a supervisor treats as
+/// "too late to ask". Written under the big kernel lock by the reap.
+const exit_record_capacity = 64;
+const ExitRecord = struct { id: u32 = 0, supervisor: u32 = 0, reason: abi.ExitReason = .exited, valid: bool = false };
+var exit_records: [exit_record_capacity]ExitRecord = .{ExitRecord{}} ** exit_record_capacity;
+var exit_record_next: usize = 0;
+
+/// Record a dying task's (id, supervisor, reason) — called by the reap before the
+/// exit notification is posted, so a supervisor that hears the notification can
+/// always still query the reason. Precondition: the big kernel lock is held.
+fn recordExitLocked(t: *scheduler.Task) void {
+    exit_records[exit_record_next] = .{ .id = t.id, .supervisor = t.supervisor, .reason = t.exit_reason, .valid = true };
+    exit_record_next = (exit_record_next + 1) % exit_record_capacity;
+}
+
+/// How dead process `id` ended, for `caller` — the kernel half of the
+/// process_exit_reason system call. Returns the ExitReason value, -ESRCH (never
+/// lived, still alive, or evicted from the ring), or -EPERM (the caller was not
+/// its supervisor — the same authority gate as process_kill).
+pub fn exitReasonOf(caller_id: u32, target_id: u32) i64 {
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    for (&exit_records) |*record| {
+        if (record.valid and record.id == target_id) {
+            if (record.supervisor != caller_id) return -ipc.EPERM;
+            return @intFromEnum(record.reason);
+        }
+    }
+    return -ipc.ESRCH;
+}
+
+/// The published exit events' subscribers (docs/process-lifecycle.md "Who learns
+/// of a death"): stateful services — the VFS's file handles, input's
+/// subscriptions — that must release what a dead client held and cannot learn it
+/// any other way (a client that simply never calls again looks like silence).
+/// Bounded like every kernel table; each entry holds its own endpoint reference.
+const exit_subscriber_capacity = 8;
+const ExitSubscriber = struct { endpoint: *ipc.Endpoint, owner: u32 };
+var exit_subscribers: [exit_subscriber_capacity]?ExitSubscriber = .{null} ** exit_subscriber_capacity;
+
+/// process_subscribe(endpoint): subscribe the caller's endpoint to published exit
+/// events. Ungated, like process_enumerate — what is running (and dying) is not a
+/// secret between cooperating processes. -ENOSPC when the table is full.
+fn systemProcessSubscribe(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    if (t.aspace == 0) return fail(state);
+    const endpoint = ipc.resolveHandle(t, architecture.systemCallArg(state, 0)) orelse return failErr(state, ipc.EBADF);
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    for (&exit_subscribers) |*slot| {
+        if (slot.* == null) {
+            endpoint.refcount += 1; // the slot's own reference, dropped on unsubscribe-by-death
+            slot.* = .{ .endpoint = endpoint, .owner = t.id };
+            return architecture.setSystemCallResult(state, 0);
+        }
+    }
+    failErr(state, ipc.ENOSPC);
+}
+
+/// signal_bind(endpoint): nominate where this process's signals arrive — the
+/// IRQ-as-IPC pattern a fourth time (docs/process-lifecycle.md). Replacing a
+/// binding drops the old reference; signals that pended while unbound are
+/// delivered immediately on bind, coalesced into one notification.
+fn systemSignalBind(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    if (t.aspace == 0) return fail(state);
+    const endpoint = ipc.resolveHandle(t, architecture.systemCallArg(state, 0)) orelse return failErr(state, ipc.EBADF);
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    if (t.signal_endpoint) |raw| ipc.dropRef(@ptrCast(@alignCast(raw)));
+    endpoint.refcount += 1;
+    t.signal_endpoint = @ptrCast(endpoint);
+    if (t.pending_signals != 0) {
+        ipc.notifyLocked(endpoint, abi.notify_signal_bit | t.pending_signals);
+        t.pending_signals = 0;
+    }
+    architecture.setSystemCallResult(state, 0);
+}
+
+/// process_signal(id, signal): post a signal — a one-way, coalescing statement,
+/// never a question (docs/process-lifecycle.md). The authority gate is the
+/// supervision link, like kill; a process may also signal itself. Unbound
+/// targets accumulate the signal in their pending mask.
+fn systemProcessSignal(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    if (t.aspace == 0) return fail(state);
+    const id = architecture.systemCallArg(state, 0);
+    const signal = architecture.systemCallArg(state, 1);
+    if (id > std.math.maxInt(u32)) return failErr(state, ipc.ESRCH);
+    if (signal > 31) return failErr(state, ipc.EBADF); // not a Signal bit position
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    const target = scheduler.taskByIdLocked(@intCast(id)) orelse return failErr(state, ipc.ESRCH);
+    if (target.aspace == 0) return failErr(state, ipc.ESRCH);
+    if (target.supervisor != t.id and target.id != t.id) return failErr(state, ipc.EPERM);
+    target.pending_signals |= @as(u32, 1) << @intCast(signal);
+    if (target.signal_endpoint) |raw| {
+        const endpoint: *ipc.Endpoint = @ptrCast(@alignCast(raw));
+        ipc.notifyLocked(endpoint, abi.notify_signal_bit | target.pending_signals);
+        target.pending_signals = 0;
+    }
+    architecture.setSystemCallResult(state, 0);
+}
+
+/// The one-shot timers of timer_bind: the missing timed wait. A service arms a
+/// deadline and keeps serving; the expiry arrives in the same replyWait as
+/// everything else (notify_timer_bit). What stop-sequence escalation, hello
+/// deadlines, and restart backoff are built from — and later, `alarm`.
+const timer_capacity = 16;
+const OneShotTimer = struct { deadline: u64, endpoint: *ipc.Endpoint, owner: u32 };
+var one_shot_timers: [timer_capacity]?OneShotTimer = .{null} ** timer_capacity;
+
+/// Sweep expired timers — hung on scheduler.timer_tick_hook, so it runs on every
+/// tick with the big kernel lock held, like the sleeper wake it rides beside.
+fn timerSweepLocked() void {
+    const now = architecture.millis();
+    for (&one_shot_timers) |*slot| {
+        if (slot.*) |timer| {
+            if (now >= timer.deadline) {
+                ipc.notifyLocked(timer.endpoint, abi.notify_timer_bit);
+                ipc.dropRef(timer.endpoint);
+                slot.* = null;
+            }
+        }
+    }
+}
+
+/// timer_bind(endpoint, ms): arm a one-shot timer. -ENOSPC when the table is full.
+fn systemTimerBind(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    if (t.aspace == 0) return fail(state);
+    const endpoint = ipc.resolveHandle(t, architecture.systemCallArg(state, 0)) orelse return failErr(state, ipc.EBADF);
+    const ms = architecture.systemCallArg(state, 1);
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    for (&one_shot_timers) |*slot| {
+        if (slot.* == null) {
+            endpoint.refcount += 1;
+            slot.* = .{ .deadline = architecture.millis() + ms, .endpoint = endpoint, .owner = t.id };
+            return architecture.setSystemCallResult(state, 0);
+        }
+    }
+    failErr(state, ipc.ENOSPC);
+}
+
+fn systemProcessExitReason(state: *architecture.CpuState) void {
+    const t = scheduler.current();
+    if (t.aspace == 0) return fail(state);
+    const id = architecture.systemCallArg(state, 0);
+    if (id > std.math.maxInt(u32)) return failErr(state, ipc.ESRCH);
+    const r = exitReasonOf(t.id, @intCast(id));
+    architecture.setSystemCallResult(state, @bitCast(r));
 }
 
 /// Resolve `(device_id, resource_index)` to a GSI this process is entitled to bind, or null.
@@ -503,6 +937,11 @@ fn systemIrqAck(state: *architecture.CpuState) void {
     if (irq.ack(gsi)) architecture.setSystemCallResult(state, 0) else fail(state);
 }
 
+/// Whether the debug_write stream sits at the start of a line — the last emitted
+/// byte was a newline (true at boot: nothing emitted yet). Guarded by the kernel
+/// lock in `systemDebugWrite`, like the stream it describes.
+var write_at_line_start: bool = true;
+
 /// debug_write(ptr, len): copy bytes from user memory into the kernel log.
 /// A bring-up diagnostic — real output goes through the VFS/console later.
 ///
@@ -512,23 +951,63 @@ fn systemIrqAck(state: *architecture.CpuState) void {
 /// Known gap (fine for trusted user code): a pointer into an *unmapped* hole in
 /// the user half passes the check and the read #PFs -> on_fault halts — a
 /// self-DoS, not an isolation break. Fault-recovering copy-in is a later item.
+///
+/// The emit runs under the kernel lock, so a message is atomic on the wire — two
+/// processes writing from different cores can interleave *messages*, never bytes.
+/// The "DANOS-INIT: " marker is emitted only at the start of a line (not per
+/// call), so a process may assemble a line from several writes without the marker
+/// (or, with the lock, another byte) landing in the middle. Cleanly-terminated
+/// lines from concurrent writers stay whole either way.
 fn systemDebugWrite(state: *architecture.CpuState) void {
     const ptr = architecture.systemCallArg(state, 0);
     const len = architecture.systemCallArg(state, 1);
     if (len <= write_buffer.len and ptr < user_half_end and ptr + len <= user_half_end) {
         const source: [*]const u8 = @ptrFromInt(ptr);
+        const flags = sync.enter();
+        defer sync.leave(flags);
         @memcpy(write_buffer[0..len], source[0..len]); // keep the latest message
         write_len = len;
         write_from_user = architecture.fromUser(state);
         write_count += 1;
-        log.write("DANOS-INIT: ");
         log.write(source[0..len]);
+        if (len != 0) write_at_line_start = source[len - 1] == '\n';
         architecture.setSystemCallResult(state, len);
     } else {
         fail(state);
     }
 }
 
+/// klog_read(offset, ptr, len) -> bytes copied: copy the kernel's in-memory
+/// diagnostic log (the RAM sink in log.zig) out to the user buffer at `ptr`,
+/// starting at `offset`. Returns the count copied — 0 once `offset` reaches the
+/// end — so a program reads the whole log by looping from 0 until it gets 0.
+///
+/// The mirror of `debug_write`: the same overflow-safe user-half bounds check,
+/// but the copy runs kernel -> user. Written under the kernel lock so the source
+/// snapshot can't grow underneath the copy. A read-only diagnostic — it exposes
+/// only the log the kernel already broadcasts to serial, nothing else.
+fn systemKlogRead(state: *architecture.CpuState) void {
+    const offset = architecture.systemCallArg(state, 0);
+    const ptr = architecture.systemCallArg(state, 1);
+    const len = architecture.systemCallArg(state, 2);
+    // Confine the whole destination span to the user (low) half. `len <=
+    // user_half_end - ptr` bounds the length without an overflowing add.
+    if (ptr < user_half_end and len <= user_half_end - ptr) {
+        const flags = sync.enter();
+        defer sync.leave(flags);
+        const snapshot = log.ramSnapshot();
+        var n: usize = 0;
+        if (offset < snapshot.len) {
+            n = @min(len, snapshot.len - offset);
+            const dest: [*]u8 = @ptrFromInt(ptr);
+            @memcpy(dest[0..n], snapshot[offset..][0..n]);
+        }
+        architecture.setSystemCallResult(state, n);
+    } else {
+        fail(state);
+    }
+}
+
 /// mmap(len, prot) -> base: grant `len` bytes (rounded up to whole pages) of
 /// fresh, zeroed, writable+NX memory in the caller's mmap arena, and return the
 /// base virtual address. `prot` is accepted but not yet honoured (grants are
@@ -617,15 +1096,15 @@ pub fn run(blob: []const u8) RunError!void {
     @memset(code[blob.len..page_size], 0xCC);
 
     architecture.mapUserPage(code_virtual, code_frame, false, true); // RO + X
-    architecture.mapUserPage(stack_virtual, stack_frame, true, false); // RW + NX
+    architecture.mapUserPage(stack_base_virtual, stack_frame, true, false); // RW + NX (one page; the probe barely stacks)
     resetRecords();
 
-    architecture.enterUser(scheduler.currentCpuIndex(), code_virtual, stack_virtual + page_size);
+    architecture.enterUser(scheduler.currentCpuIndex(), code_virtual, stack_base_virtual + page_size);
 
     // Back via the exit system_call; the interrupt gate left IF clear.
     architecture.enableInterrupts();
     architecture.unmapPage(code_virtual);
-    architecture.unmapPage(stack_virtual);
+    architecture.unmapPage(stack_base_virtual);
     pmm.free(code_frame);
     pmm.free(stack_frame);
 }
@@ -636,6 +1115,7 @@ pub const InitError = error{
     BadElf, // malformed/inapplicable image (magic, class, machine, type, bounds)
     BadSegment, // PT_LOAD unaligned, out of the user region, W&X, or overlapping
     BadEntry, // e_entry not inside an executable segment
+    BadArguments, // no argv[0], too many entries, or too many bytes for the entry stack
     ProgramTooBig, // more pages than the loader's budget
     OutOfMemory,
 };
@@ -736,13 +1216,84 @@ fn loadPageInto(aspace: u64, image: []const u8, seg: Segment, page_index: u64) I
     architecture.mapUserPageInto(aspace, seg.vaddr + page_off, frame, seg.writable, seg.executable);
 }
 
+/// Build the System V AMD64 process-entry block at the top of a process's stack
+/// page and return the initial user stack pointer. At the first user instruction,
+/// rsp is 16-byte aligned and points at (addresses growing upward):
+///
+///   argc, argv[0..argc-1], NULL, NULL (empty envp), auxiliary vector, strings
+///
+/// — the layout every C runtime's startup code walks, so danos's own runtime and a
+/// future libc port read arguments identically (docs/sysv.md). `page` is the kernel
+/// (physmap) view of the stack's **top** frame and `page_user_base` that frame's
+/// user address (stack_top_virtual - page_size); the pointers written into it are
+/// user addresses inside that page. The caller has validated the sizes
+/// (`entryStackBytes`), so this cannot overrun.
+fn buildEntryStack(page: [*]u8, page_user_base: u64, argv: []const []const u8) u64 {
+    // The strings live at the very top of the page, packed from the end downward.
+    var string_offset: usize = page_size;
+    var pointers: [maximum_arguments]u64 = undefined;
+    var i: usize = argv.len;
+    while (i > 0) {
+        i -= 1;
+        string_offset -= argv[i].len + 1;
+        @memcpy(page[string_offset..][0..argv[i].len], argv[i]);
+        page[string_offset + argv[i].len] = 0; // NUL-terminated, as C expects
+        pointers[i] = page_user_base + string_offset;
+    }
+
+    // The vector sits below the strings: argc, the argv pointers, the argv
+    // terminator, an empty envp (terminator only), then the auxiliary vector.
+    const word_count = 1 + argv.len + 1 + 1 + 4;
+    const vector_offset = (string_offset - word_count * 8) & ~@as(usize, 15); // entry rsp % 16 == 0
+    const words: [*]u64 = @ptrCast(@alignCast(page + vector_offset));
+    var w: usize = 0;
+    words[w] = argv.len; // argc
+    w += 1;
+    for (pointers[0..argv.len]) |pointer| {
+        words[w] = pointer;
+        w += 1;
+    }
+    words[w] = 0; // argv terminator
+    words[w + 1] = 0; // envp: no environment yet, just the terminator
+    words[w + 2] = auxiliary_vector_page_size;
+    words[w + 3] = page_size;
+    words[w + 4] = auxiliary_vector_null; // end of the auxiliary vector
+    words[w + 5] = 0;
+    return page_user_base + vector_offset;
+}
+
+/// Bytes the entry block for `argv` occupies at the top of the stack page:
+/// strings (each NUL-terminated), vector words, and the alignment slack.
+fn entryStackBytes(argv: []const []const u8) usize {
+    var string_bytes: usize = 0;
+    for (argv) |argument| string_bytes += argument.len + 1;
+    return string_bytes + (1 + argv.len + 1 + 1 + 4) * 8 + 16;
+}
+
 /// Load a user ELF image into a fresh address space and spawn it as a scheduled
-/// ring-3 process at `priority`. Returns immediately — the process runs
-/// preemptively on its own page tables alongside everything else, and its exit
-/// is handled by the system_call layer. The whole build (address space + ELF load +
-/// task) runs under the kernel lock so it appears atomically and can't race
-/// pmm/heap on another core.
-pub fn spawnProcess(image: []const u8, priority: u3) InitError!void {
+/// ring-3 process at `priority`, entered with `argv` on its stack per the System V
+/// convention (`buildEntryStack`). `argv[0]` is required — it names the process:
+/// the path or initial-ramdisk name it was spawned as. It is also recorded on the
+/// task, so a fault report can say *which* binary died, not just its id.
+/// The kernel-internal spawn (init at boot, tests): supervisor 0, no exit
+/// notification. `spawnProcessSupervised` is the full form.
+pub fn spawnProcess(image: []const u8, priority: u3, argv: []const []const u8) InitError!void {
+    _ = try spawnProcessSupervised(image, priority, argv, 0, null);
+}
+
+/// `spawnProcess`, recording `supervisor` (the id of the process that asked — the
+/// kill authority) and, if given, `exit_endpoint` to notify when the child ends
+/// (a reference is taken here and dropped when the notification posts).
+/// Returns the child's process id.
+/// Returns immediately — the process runs preemptively on its own page tables
+/// alongside everything else, and its exit is handled by the system_call layer.
+/// The whole build (address space + ELF load + task) runs under the kernel lock so
+/// it appears atomically and can't race pmm/heap on another core.
+pub fn spawnProcessSupervised(image: []const u8, priority: u3, argv: []const []const u8, supervisor: u32, exit_endpoint: ?*ipc.Endpoint) InitError!u32 {
+    if (argv.len == 0 or argv.len > maximum_arguments) return error.BadArguments;
+    // The entry block must leave most of the page as actual stack.
+    if (entryStackBytes(argv) > page_size / 2) return error.BadArguments;
+
     var segs: [maximum_segments]Segment = undefined;
     const parsed = try parseSegments(image, &segs);
 
@@ -755,11 +1306,28 @@ pub fn spawnProcess(image: []const u8, priority: u3) InitError!void {
     for (segs[0..parsed.count]) |seg| {
         for (0..seg.pages()) |i| try loadPageInto(aspace, image, seg, i);
     }
-    const stack_frame = pmm.alloc() orelse return error.OutOfMemory;
-    architecture.mapUserPageInto(aspace, stack_virtual, stack_frame, true, false); // RW + NX
 
-    if (!scheduler.spawnUserLocked(aspace, parsed.entry, stack_virtual + page_size, priority))
+    // The stack: `user_stack_pages` zeroed pages below stack_top_virtual, RW + NX.
+    // The page below them (`stack_virtual`) stays unmapped as the overflow guard.
+    // The entry block goes at the top of the highest page.
+    var user_sp: u64 = 0;
+    for (0..parameters.user_stack_pages) |i| {
+        const stack_frame = pmm.alloc() orelse return error.OutOfMemory;
+        const stack_page: [*]u8 = @ptrFromInt(boot_handoff.physicalToVirtual(stack_frame));
+        @memset(stack_page[0..page_size], 0); // no stale frame contents leak into user space
+        const page_virtual = stack_base_virtual + i * page_size;
+        if (i == parameters.user_stack_pages - 1)
+            user_sp = buildEntryStack(stack_page, page_virtual, argv);
+        architecture.mapUserPageInto(aspace, page_virtual, stack_frame, true, false); // RW + NX
+    }
+
+    const child = scheduler.spawnUserLocked(aspace, parsed.entry, user_sp, priority, argv[0], supervisor, if (exit_endpoint) |endpoint| @ptrCast(endpoint) else null) orelse
         return error.OutOfMemory;
+    // The child holds a reference to its exit endpoint from birth to death. Taken
+    // only now, after nothing can fail; the lock is still held, so the child
+    // cannot run (let alone die) before the reference exists.
+    if (exit_endpoint) |endpoint| endpoint.refcount += 1;
+    return child;
 }
 
 /// clock() -> nanoseconds since boot: a monotonic time source. The kernel already owns
@@ -770,3 +1338,10 @@ pub fn spawnProcess(image: []const u8, priority: u3) InitError!void {
 fn systemClock(state: *architecture.CpuState) void {
     architecture.setSystemCallResult(state, architecture.nanos());
 }
+
+/// wall_clock() -> Unix epoch seconds (UTC). The RTC value, read at boot and offset
+/// by the monotonic clock (wall-clock.zig) — mechanism, not policy: calendars and
+/// timezones layer on top in user space. Needed for filesystem timestamps (mtime).
+fn systemWallClock(state: *architecture.CpuState) void {
+    architecture.setSystemCallResult(state, wall_clock.nowSeconds());
+}

system/kernel/scheduler.zig

@@ -18,6 +18,7 @@
 //! shared queues.
 
 const std = @import("std");
+const abi = @import("abi");
 const parameters = @import("parameters");
 const architecture = @import("architecture");
 const heap = @import("heap.zig");
@@ -41,6 +42,37 @@ pub const Task = struct {
     kstack_top: usize = 0, // top of `stack` (== TSS.rsp0 for a user task); 0 = none
     wake_at: u64 = 0, // uptime (ms) to wake a sleeping task; 0 = not sleeping
     affinity: ?u32 = null, // null = runs on any core; else the index of its pinned core
+    // --- process management (process.zig) ---
+    // Id of the process that spawned this one (0 = the kernel). The supervision
+    // link is the kill authority: only the supervisor may process_kill a child.
+    supervisor: u32 = 0,
+    // Endpoint to notify when this process ends (any way: exit, fault, kill), or
+    // null. Holds its own reference, dropped when the notification is posted.
+    // Opaque here for the same reason as `handles` below.
+    exit_endpoint: ?*anyopaque = null,
+    // How this process ended — set by the death paths (exit, fault, kill) just
+    // before the reap records it for `process_exit_reason`. Meaningless while
+    // the task lives.
+    exit_reason: abi.ExitReason = .exited,
+    // Endpoint this process's signals arrive on (signal_bind), or null — same
+    // ownership rules as exit_endpoint (holds a reference; opaque here).
+    signal_endpoint: ?*anyopaque = null,
+    // Signals posted but not yet delivered: the coalescing pending mask
+    // (docs/process-lifecycle.md). Bits are abi.Signal values. Signals pend here
+    // until an endpoint is bound; two pending terminates are one terminate.
+    pending_signals: u32 = 0,
+    // Set by process_kill on a task that is running on another core; the kernel
+    // finishes the kill at that task's next system call or timer tick.
+    kill_pending: bool = false,
+    // True while this task executes its own system call — the timer tick must not
+    // tear a task down in the middle of a kernel operation, only while it runs
+    // user code (or sits at a block point, where teardown is safe).
+    in_system_call: bool = false,
+    // Where this task is parked while blocked, so a kill can unlink it: the
+    // WaitQueue it waits on (maintained by waitLocked/wakeLocked), or the endpoint
+    // whose sender FIFO it queues in (maintained by the IPC layer; opaque here).
+    wait_queue: ?*WaitQueue = null,
+    ipc_wait_endpoint: ?*anyopaque = null,
     // Physical root of this task's address space, or 0 for a kernel task (which
     // runs on the shared kernel page tables). A user task carries its own.
     aspace: u64 = 0,
@@ -70,8 +102,25 @@ pub const Task = struct {
     ipc_send_cap: u64 = ~@as(u64, 0), // handle to transfer with this message (abi.no_cap = none)
     ipc_received_cap: u64 = ~@as(u64, 0), // client: handle the reply's transferred cap landed at (abi.no_cap = none)
     next: ?*Task = null, // ready-queue link (also the endpoint sender-FIFO link)
+    // The process's name — argv[0] as it was spawned (a boot-volume path for init,
+    // an initial-ramdisk name for everything else); empty for kernel tasks. Fixed
+    // storage, so the fault path can name the dead without touching the heap.
+    // Zero-initialised (not `undefined`): an undefined default is materialised as
+    // a 0xAA fill, which would move the whole static task pool out of .bss.
+    name_buffer: [maximum_task_name]u8 = .{0} ** maximum_task_name,
+    name_length: u8 = 0,
+
+    /// The task's name (argv[0] at spawn), or empty for a kernel task.
+    pub fn name(self: *const Task) []const u8 {
+        return self.name_buffer[0..self.name_length];
+    }
 };
 
+/// Capacity of `Task.name_buffer` — matches the longest name `system_spawn`
+/// accepts, so a spawned name is never truncated. Shared with the ABI's
+/// ProcessDescriptor, so `enumerate` copies names without clipping.
+pub const maximum_task_name = abi.maximum_process_name;
+
 /// Size of each task's IPC handle table. Kept here (not in ipc_sync.zig) because
 /// it dimensions a field of `Task`; ipc_sync.zig re-exports it.
 pub const ipc_maximum_handles = 16;
@@ -258,14 +307,19 @@ pub fn spawnOn(entry: *const fn () void, priority: Priority, cpu: u32) bool {
 }
 
 /// Spawn a **user** task: a task with its own address space (`aspace`) that starts
-/// in user mode at `entry` on `user_sp`. It gets a fresh kernel stack for
-/// syscalls/interrupts, and its first switch-in lands in `user_task_trampoline`.
-/// Returns false (creating nothing) if the table is full or out of memory.
+/// in user mode at `entry` on `user_sp`, recorded under `name` (its argv[0]).
+/// `supervisor` is the id of the spawning process (0 = the kernel) — the kill
+/// authority — and `exit_endpoint` (an *ipc.Endpoint whose reference the caller
+/// has already taken, or null) is notified when this process ends.
+/// It gets a fresh kernel stack for syscalls/interrupts, and its first switch-in
+/// lands in `user_task_trampoline`.
+/// Returns the new process id, or null (creating nothing) if the table is full or
+/// out of memory.
 /// **Caller must hold the kernel lock** (the loader that builds `aspace` holds it
 /// across the whole spawn, so the address space and the task appear atomically).
-pub fn spawnUserLocked(aspace: u64, entry: u64, user_sp: u64, priority: Priority) bool {
-    const t = freeSlot() orelse return false;
-    const stack = heap.allocator().alloc(u8, stack_size) catch return false;
+pub fn spawnUserLocked(aspace: u64, entry: u64, user_sp: u64, priority: Priority, task_name: []const u8, supervisor: u32, exit_endpoint: ?*anyopaque) ?u32 {
+    const t = freeSlot() orelse return null;
+    const stack = heap.allocator().alloc(u8, stack_size) catch return null;
     t.* = .{
         .id = next_id,
         .state = .ready,
@@ -274,7 +328,12 @@ pub fn spawnUserLocked(aspace: u64, entry: u64, user_sp: u64, priority: Priority
         .aspace = aspace,
         .user_ip = entry,
         .user_sp = user_sp,
+        .supervisor = supervisor,
+        .exit_endpoint = exit_endpoint,
     };
+    const name_length = @min(task_name.len, maximum_task_name);
+    @memcpy(t.name_buffer[0..name_length], task_name[0..name_length]);
+    t.name_length = @intCast(name_length);
     next_id += 1;
     const top = @intFromPtr(stack.ptr) + stack.len;
     t.kstack_top = top;
@@ -282,7 +341,7 @@ pub fn spawnUserLocked(aspace: u64, entry: u64, user_sp: u64, priority: Priority
     // the user entry/stack from the Task itself).
     t.sp = architecture.initTaskStack(top, @intFromPtr(&startUserTask));
     enqueue(t);
-    return true;
+    return t.id;
 }
 
 /// The first thing a fresh user task runs (in ring 0, via task_trampoline). It
@@ -291,8 +350,9 @@ pub fn spawnUserLocked(aspace: u64, entry: u64, user_sp: u64, priority: Priority
 /// context switch and lock release.
 fn startUserTask() void {
     const t = current();
-    var buffer: [96]u8 = undefined;
-    architecture.serialWrite(std.fmt.bufPrint(&buffer, "DBG startUserTask ip=0x{x} sp=0x{x} aspace=0x{x} kstack=0x{x}\n", .{ t.user_ip, t.user_sp, t.aspace, t.kstack_top }) catch "");
+    // No serial chatter here: this runs on every spawn, unserialized against
+    // user-space writes, and its output used to shear concurrent log lines in
+    // half — the largest source of corrupted markers in the QEMU scenarios.
     architecture.jumpToUser(t.user_ip, t.user_sp); // noreturn
 }
 
@@ -394,6 +454,7 @@ pub const WaitQueue = struct {
 pub fn waitLocked(wait_queue: *WaitQueue) void {
     const t = current();
     t.state = .blocked;
+    t.wait_queue = wait_queue; // so a kill can unlink a parked waiter
     t.next = wait_queue.head;
     wait_queue.head = t;
     schedule();
@@ -418,10 +479,78 @@ pub fn wakeLocked(wait_queue: *WaitQueue) void {
     }
     const t = best orelse return;
     if (best_previous) |p| p.next = t.next else wait_queue.head = t.next;
+    t.wait_queue = null;
     t.state = .ready;
     enqueue(t);
 }
 
+/// Unlink `t` from the wait queue it is parked on, if any (the kill path — a
+/// killed waiter must not be woken later as a dangling pointer). Precondition:
+/// the big kernel lock is held.
+pub fn removeFromWaitQueueLocked(t: *Task) void {
+    const wait_queue = t.wait_queue orelse return;
+    t.wait_queue = null;
+    var previous: ?*Task = null;
+    var node = wait_queue.head;
+    while (node) |n| : ({
+        previous = n;
+        node = n.next;
+    }) {
+        if (n != t) continue;
+        if (previous) |p| p.next = t.next else wait_queue.head = t.next;
+        t.next = null;
+        return;
+    }
+}
+
+/// Unlink `t` from the ready queue it sits in (global, or its affinity core's
+/// pinned queue) — the kill path for a task that is runnable but not running.
+/// Precondition: the big kernel lock is held.
+pub fn removeFromReadyQueueLocked(t: *Task) void {
+    if (t.affinity) |cpu| {
+        const pc = &cpus[cpu];
+        removeFrom(&pc.pinned_head, &pc.pinned_tail, &pc.pinned_bitmap, t);
+    } else {
+        removeFrom(&ready_head, &ready_tail, &ready_bitmap, t);
+    }
+}
+
+fn removeFrom(head: *[number_priorities]?*Task, tail: *[number_priorities]?*Task, bitmap: *u8, t: *Task) void {
+    const level: usize = t.priority;
+    var previous: ?*Task = null;
+    var node = head[level];
+    while (node) |n| : ({
+        previous = n;
+        node = n.next;
+    }) {
+        if (n != t) continue;
+        if (previous) |p| p.next = t.next else head[level] = t.next;
+        if (tail[level] == t) tail[level] = previous;
+        if (head[level] == null) bitmap.* &= ~(@as(u8, 1) << @intCast(level));
+        t.next = null;
+        return;
+    }
+}
+
+/// Find a live task by process id, or null. Ids are monotonic and never reused,
+/// so a stale id misses cleanly rather than naming a recycled slot.
+/// Precondition: the big kernel lock is held.
+pub fn taskByIdLocked(id: u32) ?*Task {
+    for (&tasks) |*t| {
+        if (t.state != .free and t.id == id) return t;
+    }
+    return null;
+}
+
+/// Make every server that still holds `t` as the client it owes a reply to forget
+/// it — the reply of a dead client is dropped, not delivered into freed state.
+/// Precondition: the big kernel lock is held.
+pub fn forgetIpcClientLocked(t: *Task) void {
+    for (&tasks) |*other| {
+        if (other.state != .free and other.ipc_client == t) other.ipc_client = null;
+    }
+}
+
 /// Block the current task and switch away, without putting it on any wait queue —
 /// the caller has already linked it wherever it belongs (e.g. an endpoint's sender
 /// FIFO). Precondition: the big kernel lock is held; still held on return (when the
@@ -481,14 +610,61 @@ fn wakeExpired() void {
     }
 }
 
+// Process-teardown hooks, registered by process.zig at init — the scheduler sits
+// below the process layer, so finishing a kill (IRQ bindings, IPC handles, exit
+// notification) is called *up* through these, mirroring how the architecture
+// layer calls up into `tick`.
+//
+// `terminate_current_hook` ends the task running on THIS core (lock held, never
+// returns — it switches away like `exitUserLocked`). `reap_task_hook` tears down
+// a task that is NOT running on any core (lock held).
+pub var terminate_current_hook: ?*const fn () noreturn = null;
+pub var reap_task_hook: ?*const fn (*Task) void = null;
+
+/// Finish any pending kills this core can see (the deferred half of process_kill;
+/// the immediate half runs in the killer's own call). Precondition: the big kernel
+/// lock is held, from `tick`.
+///
+/// - This core's *current* task, if condemned, is terminated here — but only when
+///   it is not inside one of its own system calls (`in_system_call`): the tick may
+///   have interrupted kernel code mid-operation, where teardown would leak or
+///   corrupt what that operation holds. User-mode execution (and the system_call
+///   entry/exit stubs, which hold nothing) are safe termination points. A task
+///   that *is* mid-call dies at its next block, tick, or system_call entry instead.
+///   The hook never returns; abandoning the interrupt frame is fine — the LAPIC
+///   was acknowledged before the tick hook ran (see apic.timerTick), exactly as on
+///   the fault-kill path.
+/// - Condemned tasks that are ready or blocked are not running anywhere (state
+///   changes need the lock we hold), so they are reaped in place.
+fn reapKillPendingLocked() void {
+    const pc = thisCpu();
+    const cur = pc.current;
+    if (cur.kill_pending and cur.aspace != 0 and !cur.in_system_call) {
+        if (terminate_current_hook) |hook| hook(); // noreturn
+    }
+    if (reap_task_hook) |hook| {
+        for (&tasks) |*t| {
+            if (!t.kill_pending) continue;
+            if (t.state == .ready or t.state == .blocked) hook(t);
+        }
+    }
+}
+
 /// Called from the timer interrupt (interrupts already disabled): wake due
-/// sleepers, then preempt. Takes the kernel lock like any other critical section,
-/// but releases it *without* touching the interrupt flag — the handler's `iretq`
-/// restores the interrupted context's flags, so re-enabling here would open a
-/// nested-interrupt window before the return.
+/// sleepers, finish pending kills, then preempt. Takes the kernel lock like any
+/// other critical section, but releases it *without* touching the interrupt flag
+/// — the handler's `iretq` restores the interrupted context's flags, so
+/// re-enabling here would open a nested-interrupt window before the return.
+/// Called from the tick with the big kernel lock held — process.zig hangs the
+/// one-shot timer sweep here (timer_bind), the same call-up pattern as the
+/// teardown hooks below.
+pub var timer_tick_hook: ?*const fn () void = null;
+
 pub fn tick() void {
     _ = sync.enter();
     wakeExpired();
+    if (timer_tick_hook) |hook| hook();
+    reapKillPendingLocked();
     if (preemption_enabled) schedule();
     sync.leaveIsr();
 }
@@ -520,6 +696,13 @@ pub fn exit() noreturn {
 /// itself is leaked, as in `exit` (no reaper yet). Never returns.
 pub fn exitUser() noreturn {
     _ = sync.enter();
+    exitUserLocked();
+}
+
+/// The body of `exitUser` for callers that already hold the big kernel lock (the
+/// tick-time terminate path, which enters with the lock held). The lock is handed
+/// off through the switch and released by the task that resumes. Never returns.
+pub fn exitUserLocked() noreturn {
     const pc = thisCpu();
     const dying = pc.current;
     const as = dying.aspace;
@@ -531,6 +714,8 @@ pub fn exitUser() noreturn {
     }
     dying.state = .free;
     dying.aspace = 0;
+    dying.kill_pending = false;
+    dying.in_system_call = false;
     const next = dequeueHighest(pc) orelse @panic("sched: no task left to run");
     next.state = .running;
     pc.current = next;
@@ -539,6 +724,54 @@ pub fn exitUser() noreturn {
     unreachable;
 }
 
+/// Free a task that is NOT running on any core (it is ready or blocked, and the
+/// caller — the kill path — has already unlinked it from every queue and released
+/// what it held). Destroys its address space: safe here because no core can have
+/// it loaded (every switch away from a task loads the next task's tables, and the
+/// task isn't running). The kernel stack is leaked, as in `exitUser` (no reaper
+/// yet). Precondition: the big kernel lock is held.
+pub fn destroyTaskLocked(t: *Task) void {
+    if (t.aspace != 0) architecture.destroyAddressSpace(t.aspace);
+    t.aspace = 0;
+    t.kill_pending = false;
+    t.in_system_call = false;
+    t.wake_at = 0;
+    t.state = .free;
+}
+
+/// Snapshot the task table into `out` (up to its length), returning the total
+/// number of live tasks — the kernel half of `process_enumerate`, mirroring
+/// devices_broker.enumerate. Kernel tasks are included (empty name, supervisor 0):
+/// an honest `ps` shows the idle tasks too. `out` may be user memory: the caller's
+/// address space is loaded during its system call, and the same bring-up trust
+/// applies as for device_enumerate (an unmapped user page faults the kernel).
+pub fn enumerate(out: []abi.ProcessDescriptor) u64 {
+    const flags = sync.enter();
+    defer sync.leave(flags);
+    var total: u64 = 0;
+    for (&tasks) |*t| {
+        if (t.state == .free) continue;
+        if (total < out.len) {
+            const d = &out[total];
+            d.* = .{
+                .id = t.id,
+                .supervisor = t.supervisor,
+                .state = @intFromEnum(@as(abi.ProcessState, switch (t.state) {
+                    .ready => .ready,
+                    .running => .running,
+                    .blocked => .blocked,
+                    .free => unreachable,
+                })),
+                .priority = t.priority,
+                .name_length = t.name_length,
+                .name = t.name_buffer,
+            };
+        }
+        total += 1;
+    }
+    return total;
+}
+
 /// Whether the running task is a user process (has its own address space).
 pub fn currentIsUserProcess() bool {
     return current().aspace != 0;

system/kernel/wall-clock.zig

@@ -0,0 +1,26 @@
+//! Wall-clock time: the CMOS real-time clock read once at boot and anchored to the
+//! monotonic clock, so a query is a cheap arithmetic offset — no per-call CMOS poll,
+//! no lock, no SMP hazard on the shared 0x70/0x71 ports.
+//!
+//! Wall-clock *seconds* are mechanism the kernel owns, exactly like the monotonic
+//! clock ([[time-architecture]]): reading the hardware's value is not policy.
+//! Calendars, timezones, and formatting layer on top in user space. It exists so the
+//! filesystem can stamp real timestamps (mtime) — see docs/zig-self-hosting.md.
+
+const architecture = @import("architecture");
+
+var boot_unix_seconds: u64 = 0;
+var boot_nanos: u64 = 0;
+
+/// Read the RTC once and anchor it to the monotonic clock. Call at boot, after the
+/// monotonic clock is calibrated.
+pub fn init() void {
+    boot_unix_seconds = architecture.readRtcUnixSeconds();
+    boot_nanos = architecture.nanos();
+}
+
+/// The current wall-clock time in Unix epoch seconds (UTC): the boot RTC value plus
+/// the monotonic time elapsed since. Zero until `init` runs.
+pub fn nowSeconds() u64 {
+    return boot_unix_seconds + (architecture.nanos() -% boot_nanos) / 1_000_000_000;
+}

system/parameters.zig

@@ -4,7 +4,7 @@
 //! hiding the trade-offs. Keeping them here makes them visible at a glance and gives
 //! one spot to change them. They're plain `comptime` constants (zero runtime cost);
 //! any one can later be promoted to a `-D` build option if a target needs to vary it
-//! (see build.zig's `-Dtest-case` for the pattern). This keeps root.zig to what it
+//! (see build.zig's `-Dtest-case` for the pattern). This keeps ps2-library.zig to what it
 //! actually is — the bootloader↔kernel handoff *contract* — with tunables living here.
 
 /// Ceiling on logical CPUs the kernel tracks — the size of the per-CPU bookkeeping
@@ -16,12 +16,24 @@
 pub const maximum_cpus = 128;
 
 /// Maximum tasks (kernel threads) alive at once — the static task-table size. Each
-/// online core consumes one slot for its idle task, plus task 0 on the BSP.
-pub const maximum_tasks = 16;
+/// online core consumes one slot for its idle task, plus task 0 on the BSP. Sized
+/// for the initial-ramdisk sweep (the bundled binaries spawned at once) plus the
+/// device manager's supervised children with room to grow — at 16 the sweep
+/// started failing spawns once the bundle passed a dozen binaries. Raised to 48
+/// for the USB stack: the xHCI bus driver spawns a supervised class-driver instance
+/// per matched interface (keyboard, mouse, mass storage), on top of the FAT and
+/// block servers and the growing ramdisk bundle.
+pub const maximum_tasks = 48;
 
 /// Each task's kernel stack (also each AP's bring-up stack), in bytes.
 pub const kernel_stack_size = 16 * 1024;
 
+/// Each user process's stack, in pages (32 KiB). Mapped just below a fixed top;
+/// the System V entry block (argc/argv) occupies the top of the highest page, and
+/// the page below the mapping is left unmapped as a guard, so an overflow faults
+/// (killing only that process) instead of silently corrupting the image.
+pub const user_stack_pages = 8;
+
 /// Each core's IST (double-fault) stack, in bytes. The BSP's is static; an AP's is
 /// heap-allocated at bring-up.
 pub const ist_stack_size = 16 * 1024;