gitignore: /var/log — real-hardware log pulls stay out of the tree

(An earlier append landed on a line missing its newline, mangling
'.github/' and silently un-ignoring var/ — which let a stray add sweep
pulled logs into the tree. Repaired, and the logs untracked again.)

.gitignore

@@ -7,3 +7,4 @@ zig-out/
 
 .claude/
 .github/
+/var/log/

boot/efi.zig

@@ -380,11 +380,19 @@ const Bundled = struct {
     data: []align(8) u8,
 };
 
-/// Walk the boot volume's /system tree and pack every regular file (except the
-/// kernel image itself — the only top-level file) into an in-RAM v2
-/// initial_ramdisk image, entries named by full FHS path. This is what makes the
-/// volume's file structure the single source of truth: there is no packed
-/// ramdisk artifact on disk, and init travels in the table like everything else.
+/// Gather the boot volume's user binaries into an in-RAM v2 initial_ramdisk
+/// image, entries named by full FHS path — the volume's file structure is the
+/// single source of truth (no packed ramdisk artifact; init travels in the
+/// table like everything else).
+///
+/// Two strategies, most portable first:
+///   1. /system/manifest (written by the build): each listed path is opened BY
+///      NAME — the case-insensitive lookup every firmware FAT driver gets
+///      right, and the only file access the pre-tree loader ever used.
+///   2. No manifest: ENUMERATE the /system tree. Portable in principle, but
+///      firmware differs in what names enumeration returns (bare 8.3 entries
+///      come back uppercase on some drivers), so this is the fallback for
+///      hand-assembled sticks, not the primary path.
 fn loadSystemTree(bs: *uefi.tables.BootServices, boot_information: *BootInformation) !void {
     const loaded = (try bs.handleProtocol(uefi.protocol.LoadedImage, uefi.handle)) orelse
         return error.NoLoadedImage;
@@ -395,12 +403,30 @@ fn loadSystemTree(bs: *uefi.tables.BootServices, boot_information: *BootInformat
     const root = try fs.openVolume();
     defer _ = root.close() catch {};
 
-    const system_directory = try root.open(system_directory_name, .read, .{});
-    defer _ = system_directory.close() catch {};
+    // Unconditional breadcrumb (con_out, independent of -Dserial): this phase
+    // is where a slow firmware stalls, and a silent black screen here already
+    // cost a real-hardware debugging session.
+    log("EFI: loading the system...\r\n");
+
+    // The capsule (boot\system.img) first: one open + one sequential read is
+    // the only firmware file I/O shape that is fast everywhere. It is already
+    // the kernel's wire format — hand it over as-is.
+    if (loadCapsule(bs, root, boot_information)) {
+        log("EFI: system image loaded, starting the kernel\r\n");
+        return;
+    }
 
     var list: [maximum_bundled]Bundled = undefined;
     var count: usize = 0;
+
+    loadByManifest(bs, root, &list, &count) catch {
+        count = 0; // a torn manifest read leaves partial entries; start over
+    };
+    if (count == 0) {
+        const system_directory = try root.open(system_directory_name, .read, .{});
+        defer _ = system_directory.close() catch {};
         try walkDirectory(bs, system_directory, "/system", 0, &list, &count);
+    }
     if (count == 0) return error.NoBinaries;
 
     // Assemble the v2 image: header, entry table, then the blobs.
@@ -426,7 +452,72 @@ fn loadSystemTree(bs: *uefi.tables.BootServices, boot_information: *BootInformat
 
     boot_information.initial_ramdisk_base = @intFromPtr(image.ptr);
     boot_information.initial_ramdisk_len = total;
-    progress("EFI: /system tree loaded\r\n");
+    log("EFI: /system tree loaded, starting the kernel\r\n");
+}
+
+/// The boot capsule: the bundled binaries as one v2 initial_ramdisk image.
+const capsule_file_name = std.unicode.utf8ToUtf16LeStringLiteral("boot\\system.img");
+
+/// Load boot\system.img whole and hand it to the kernel unmodified — it is
+/// already the initial_ramdisk wire format. Returns false (capsule absent or
+/// unreadable or wrong magic) to let the caller fall back to per-file loading.
+fn loadCapsule(bs: *uefi.tables.BootServices, root: *uefi.protocol.File, boot_information: *BootInformation) bool {
+    const file = root.open(capsule_file_name, .read, .{}) catch return false;
+    defer _ = file.close() catch {};
+    const image = readWholeFile(bs, file) catch return false;
+    if (image.len < @sizeOf(initial_ramdisk.Header) or
+        std.mem.bytesToValue(initial_ramdisk.Header, image[0..@sizeOf(initial_ramdisk.Header)]).magic != initial_ramdisk.magic)
+    {
+        _ = bs.freePool(image.ptr) catch {};
+        return false;
+    }
+    boot_information.initial_ramdisk_base = @intFromPtr(image.ptr);
+    boot_information.initial_ramdisk_len = image.len;
+    return true;
+}
+
+/// The manifest path, and a scratch limit for its UTF-16 conversion.
+const manifest_file_name = std.unicode.utf8ToUtf16LeStringLiteral("system\\manifest");
+
+/// Load every binary the manifest lists, opening each path by name from the
+/// volume root. A listed-but-unopenable file is skipped (the kernel reports the
+/// absence); a missing manifest errors so the caller falls back to the walk.
+fn loadByManifest(bs: *uefi.tables.BootServices, root: *uefi.protocol.File, list: *[maximum_bundled]Bundled, count: *usize) !void {
+    const manifest_handle = try root.open(manifest_file_name, .read, .{});
+    var manifest_open = true;
+    defer if (manifest_open) {
+        _ = manifest_handle.close() catch {};
+    };
+    const manifest = try readWholeFile(bs, manifest_handle);
+    _ = manifest_handle.close() catch {};
+    manifest_open = false;
+    defer _ = bs.freePool(manifest.ptr) catch {};
+
+    var lines = std.mem.tokenizeAny(u8, manifest, "\r\n");
+    while (lines.next()) |line| {
+        if (line.len < 2 or line[0] != '/') continue;
+        if (line.len >= initial_ramdisk.maximum_name) continue;
+        if (count.* == maximum_bundled) return;
+
+        // "/system/services/init" -> UTF-16 "system\services\init".
+        var name16: [initial_ramdisk.maximum_name]u16 = undefined;
+        var i: usize = 0;
+        for (line[1..]) |c| {
+            name16[i] = if (c == '/') '\\' else c;
+            i += 1;
+        }
+        name16[i] = 0;
+
+        const file = root.open(@ptrCast(name16[0..i :0]), .read, .{}) catch continue;
+        defer _ = file.close() catch {};
+        const data = readWholeFile(bs, file) catch continue;
+
+        var entry: *Bundled = &list[count.*];
+        @memcpy(entry.path[0..line.len], line);
+        entry.path_len = line.len;
+        entry.data = data;
+        count.* += 1;
+    }
 }
 
 /// Recursively collect the regular files below `directory` into `list`. Top-level
@@ -448,17 +539,34 @@ fn walkDirectory(
         const info: *const uefi.protocol.File.Info.File = @ptrCast(@alignCast(&info_buffer));
         const name16 = info.getFileName();
 
-        // Convert the (ASCII in practice) UTF-16 name; skip "." and "..".
+        // Convert the (ASCII in practice) UTF-16 name. A hostile-shaped entry
+        // (too long, non-ASCII) is SKIPPED, never fatal — one odd file on a
+        // hand-written stick must not cost the whole boot. Names are lowered:
+        // the danos tree is canonically lowercase and FAT lookups are
+        // case-insensitive, but firmware ENUMERATION returns whatever the
+        // directory stores — an 8.3 short entry comes back uppercase ("INIT"),
+        // which would otherwise poison every path comparison downstream.
         var name_buffer: [initial_ramdisk.maximum_name]u8 = undefined;
         var name_length: usize = 0;
+        var name_ok = true;
         while (name16[name_length] != 0) : (name_length += 1) {
-            if (name_length == name_buffer.len) return error.NameTooLong;
-            const c = name16[name_length];
-            if (c > 0x7F) return error.UnsupportedName;
-            name_buffer[name_length] = @intCast(c);
+            if (name_length == name_buffer.len) {
+                name_ok = false;
+                break;
             }
+            const c = name16[name_length];
+            if (c > 0x7F) {
+                name_ok = false;
+                break;
+            }
+            name_buffer[name_length] = std.ascii.toLower(@intCast(c));
+        }
+        if (!name_ok) continue;
         const name = name_buffer[0..name_length];
-        if (std.mem.eql(u8, name, ".") or std.mem.eql(u8, name, "..")) continue;
+        // Skip dot entries: "." / ".." and host-OS litter (macOS "._*" AppleDouble
+        // resource forks, ".fseventsd", ".Spotlight-V100") a copied-onto stick
+        // accumulates — none of it is a danos binary.
+        if (name.len == 0 or name[0] == '.') continue;
 
         if (info.attribute.directory) {
             if (depth == maximum_tree_depth) continue;
@@ -474,12 +582,12 @@ fn walkDirectory(
         if (count.* == maximum_bundled) return error.TooManyBinaries;
 
         var entry: *Bundled = &list[count.*];
-        const path = try std.fmt.bufPrint(&entry.path, "{s}/{s}", .{ prefix, name });
+        const path = std.fmt.bufPrint(&entry.path, "{s}/{s}", .{ prefix, name }) catch continue; // path too long: skip the file, keep the boot
         entry.path_len = path.len;
 
-        const file = try directory.open(name16, .read, .{});
+        const file = directory.open(name16, .read, .{}) catch continue;
         defer _ = file.close() catch {};
-        entry.data = try readWholeFile(bs, file);
+        entry.data = readWholeFile(bs, file) catch continue; // unreadable/empty: skip
         count.* += 1;
     }
 }

build.zig

@@ -235,6 +235,8 @@ fn addBootImage(
     b: *std.Build,
     kernel_bin: std.Build.LazyPath,
     efi_bin: std.Build.LazyPath,
+    manifest: std.Build.LazyPath,
+    capsule: std.Build.LazyPath,
     bundled: []const BundledBinary,
 ) std.Build.LazyPath {
     const mk_fat = b.addSystemCommand(&.{"python3"});
@@ -245,6 +247,10 @@ fn addBootImage(
     mk_fat.addFileArg(efi_bin);
     mk_fat.addArg("system/kernel");
     mk_fat.addFileArg(kernel_bin);
+    mk_fat.addArg("system/manifest");
+    mk_fat.addFileArg(manifest);
+    mk_fat.addArg("boot/system.img");
+    mk_fat.addFileArg(capsule);
     for (bundled) |item| {
         mk_fat.addArg(item.path);
         mk_fat.addFileArg(item.binary);
@@ -462,6 +468,10 @@ pub fn build(b: *std.Build) void {
     // QEMU test harness (test/qemu_test.py, which asserts on serial markers) turn
     // it on; a flashable `zig build` image leaves it out. See serial.zig.
     const serial = b.option(bool, "serial", "Compile the serial-console log sink into the kernel (default: off; run-x86-64 and the test harness enable it)") orelse false;
+    // The diagnose boot: init skips the display service (and demo), so the
+    // on-screen boot transcript is never suppressed — the full timestamped
+    // timeline stays on the screen for real-hardware debugging by eye.
+    const diagnose = b.option(bool, "diagnose", "Boot without the display service so the timestamped boot transcript stays on screen (real-hardware debugging)") orelse false;
 
     // --- Kernel: freestanding x86_64 ELF, jumped to by the bootloader ---
     // SSE2 is part of the x86_64 baseline and UEFI leaves it enabled at handoff,
@@ -505,6 +515,7 @@ pub fn build(b: *std.Build) void {
     // the heartbeat stays present under test.
     const init_options = b.addOptions();
     init_options.addOption(bool, "serial", serial);
+    init_options.addOption(bool, "diagnose", diagnose);
     programModule(init_exe).addImport("build_options", init_options.createModule());
 
     // --- the rest of the /system tree: services, drivers, test fixtures ---
@@ -622,6 +633,40 @@ pub fn build(b: *std.Build) void {
         .{ .path = "system/tests/thread-test", .binary = thread_test_exe.getEmittedBin() },
     };
 
+    // The boot manifest: the FHS path of every bundled binary, one per line. The
+    // EFI loader reads THIS by name and opens each listed path by name — FAT
+    // name lookup is case-insensitive and firmware-portable, unlike directory
+    // ENUMERATION, whose returned names vary by firmware (bare 8.3 entries come
+    // back uppercase on some FAT drivers). The tree walk remains only as the
+    // loader's fallback for hand-assembled sticks without a manifest.
+    var manifest_text: std.ArrayListUnmanaged(u8) = .empty;
+    for (bundled) |item| {
+        manifest_text.append(b.allocator, '/') catch @panic("OOM");
+        manifest_text.appendSlice(b.allocator, item.path) catch @panic("OOM");
+        manifest_text.append(b.allocator, '\n') catch @panic("OOM");
+    }
+    const manifest_files = b.addWriteFiles();
+    const manifest_file = manifest_files.add("manifest", manifest_text.items);
+    const manifest_install = b.addInstallFileWithDir(manifest_file, .prefix, "system/manifest");
+    b.getInstallStep().dependOn(&manifest_install.step);
+
+    // The boot capsule: the same bundled list packed into ONE file (v2
+    // initial_ramdisk format), because a single open + sequential read is the
+    // only firmware file I/O shape that is fast everywhere — a per-file tree
+    // walk measured MINUTES on real firmware. The loader tries this first,
+    // then the manifest, then the walk; the running system cannot tell the
+    // difference (it always receives the same in-RAM table). Derived from the
+    // tree in the same build graph, so the two cannot drift.
+    const mk_capsule = b.addSystemCommand(&.{"python3"});
+    mk_capsule.addFileArg(b.path("tools/pack-system-image.py"));
+    const capsule_img = mk_capsule.addOutputFileArg("system.img");
+    for (bundled) |item| {
+        mk_capsule.addArg(item.path);
+        mk_capsule.addFileArg(item.binary);
+    }
+    const capsule_install = b.addInstallFile(capsule_img, "boot/system.img");
+    b.getInstallStep().dependOn(&capsule_install.step);
+
     // Install every bundled binary to its FHS home, so zig-out is a true image of
     // the filesystem — the same tree make-fat-image.py lays out on the boot volume.
     for (bundled) |item| {
@@ -669,7 +714,7 @@ pub fn build(b: *std.Build) void {
     // binaries at their FHS paths. 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 fat_image = addBootImage(b, exe.getEmittedBin(), efiexe.getEmittedBin(), &bundled);
+    const fat_image = addBootImage(b, exe.getEmittedBin(), efiexe.getEmittedBin(), manifest_file, capsule_img, &bundled);
     const fat_image_install = b.addInstallFile(fat_image, "danos-usb.img");
     b.getInstallStep().dependOn(&fat_image_install.step);
 
@@ -678,7 +723,7 @@ pub fn build(b: *std.Build) void {
     // log captured to serial0 — without baking serial into the image users flash.
     // Built lazily (only when `run-x86-64` is requested), and never installed.
     const exe_serial = addKernel(b, kernel_target, optimize, kernel_modules, test_case, true);
-    const fat_image_serial = addBootImage(b, exe_serial.getEmittedBin(), efiexe.getEmittedBin(), &bundled);
+    const fat_image_serial = addBootImage(b, exe_serial.getEmittedBin(), efiexe.getEmittedBin(), manifest_file, capsule_img, &bundled);
 
     // `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).

library/runtime/block.zig

@@ -61,7 +61,10 @@ pub fn open() ?Device {
     // 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) {
+    // 30 s covers the slowest observed healthy chain (a flaky QEMU enumeration
+    // completed at ~24 s); a machine whose stick genuinely failed setup should
+    // not sit a further minute pretending otherwise.
+    while (attempts < 600) : (attempts += 1) {
         if (ipc.lookup(.block)) |handle| return .{ .endpoint = handle };
         system.sleep(50);
     }

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

@@ -620,8 +620,15 @@ pub const Controller = struct {
             .parameter = device.input_context.physical,
             .control = trbControl(.address_device, @as(u32, device.slot_id) << 24),
         });
-        const code = self.awaitCommand(physical) orelse return false;
-        return code == @intFromEnum(CompletionCode.success);
+        const code = self.awaitCommand(physical) orelse {
+            std.log.info("port {d} setup: Address Device timed out", .{device.port});
+            return false;
+        };
+        if (code != @intFromEnum(CompletionCode.success)) {
+            std.log.info("port {d} setup: Address Device completion code {d}", .{ device.port, code });
+            return false;
+        }
+        return true;
     }
 
     /// Reset the port, enable a slot, and address the device on it: after this the
@@ -629,9 +636,25 @@ pub const Controller = struct {
     /// or null on any failure. The EP0 MPS is taken from the speed default and
     /// corrected from the device descriptor by `refreshMaxPacketSize0` if needed.
     pub fn setupDevice(self: *Controller, port: u32, speed: u32) ?*Device {
-        if (!self.resetPort(port)) return null;
-        const slot_id = self.enableSlot() orelse return null;
-        const device = self.allocateDevice() orelse return null;
+        // A SuperSpeed port that has trained its link is ALREADY enabled — the
+        // xHCI advances USB3 ports to Enabled with no reset (spec 4.3). Driving
+        // a hot reset into a live SS link drops PED mid-reset on real silicon
+        // (observed: "setup failed" in the same millisecond as "connected").
+        // Only a not-yet-enabled port — every USB2 device, or a stuck SS link —
+        // needs the reset to enable.
+        const already_enabled = speed >= 4 and self.portStatus(port) & portsc_enabled != 0;
+        if (!already_enabled and !self.resetPort(port)) {
+            std.log.info("port {d} setup: port reset failed (PORTSC 0x{x:0>8})", .{ port, self.portStatus(port) });
+            return null;
+        }
+        const slot_id = self.enableSlot() orelse {
+            std.log.info("port {d} setup: Enable Slot failed", .{port});
+            return null;
+        };
+        const device = self.allocateDevice() orelse {
+            std.log.info("port {d} setup: no free device slot", .{port});
+            return null;
+        };
         device.* = .{
             .used = true,
             .slot_id = slot_id,
@@ -652,6 +675,7 @@ pub const Controller = struct {
         return device;
     }
 
+
     fn abandon(self: *Controller, device: *Device) ?*Device {
         _ = self;
         device.used = false;

system/initial-ramdisk.zig

@@ -76,17 +76,20 @@ pub const Reader = struct {
 
     /// Look a binary up by name: an exact path match wins; otherwise a unique
     /// basename match ("fat" finds "/system/services/fat") keeps pre-path callers
-    /// working. The returned Item's name is always the stored full path.
+    /// working. Comparisons are ASCII case-insensitive — the entries come from a
+    /// FAT volume, whose name lookups are case-insensitive by definition (and
+    /// whose short entries store uppercase). The returned Item's name is always
+    /// the stored full path.
     pub fn find(self: Reader, name: []const u8) ?Item {
         var i: u32 = 0;
         while (i < self.count) : (i += 1) {
             const item = self.entry(i) orelse continue;
-            if (std.mem.eql(u8, item.name, name)) return item;
+            if (std.ascii.eqlIgnoreCase(item.name, name)) return item;
         }
         i = 0;
         while (i < self.count) : (i += 1) {
             const item = self.entry(i) orelse continue;
-            if (std.mem.eql(u8, basename(item.name), name)) return item;
+            if (std.ascii.eqlIgnoreCase(basename(item.name), name)) return item;
         }
         return null;
     }

system/kernel/architecture/x86_64/apic.zig

@@ -529,8 +529,22 @@ 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
+// The warp check is bounded by TIME, not iterations: a warp tick is a locked
+// read-modify-write on a cacheline two cores are fighting over — microseconds
+// under real contention, not the nanosecond an uncontended count assumes (a
+// 1<<20-round budget measured 2-21 SECONDS per core on a 16-core machine), and
+// a PAUSE costs ~140 cycles on modern Intel, so an iteration-counted await
+// mis-measures by two orders of magnitude too. ~5 ms of pairwise hammering per
+// core is plenty to catch a lagging TSC (Linux's check_tsc_warp budget), and
+// ~100 ms is a generous rendezvous window for a healthy core.
+const warp_check_ns: u64 = 5_000_000; // per-AP pairwise check duration
+const warp_await_ns: u64 = 100_000_000; // rendezvous wait before giving up
+
+/// TSC ticks for `ns` nanoseconds (valid whenever the warp check runs: the TSC
+/// is the clocksource, so tsc_hz is calibrated).
+fn warpTicksFor(ns: u64) u64 {
+    return @intCast(@as(u128, ns) * tsc_hz / 1_000_000_000);
+}
 
 fn warpTick() void {
     while (@cmpxchgWeak(u32, &warp_lock, 0, 1, .acquire, .monotonic) != null) asm volatile ("pause");
@@ -545,11 +559,11 @@ fn warpTick() void {
     @atomicStore(u32, &warp_lock, 0, .release);
 }
 
-/// Spin (bounded) until `flag` is nonzero; false on timeout.
+/// Spin (time-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;
+    const deadline = rdtsc() +% warpTicksFor(warp_await_ns);
+    while (@atomicLoad(u32, flag, .acquire) == 0) {
+        if (rdtsc() -% deadline < (1 << 62)) return false; // past the deadline
         asm volatile ("pause");
     }
     return true;
@@ -568,8 +582,8 @@ pub fn checkWarpSource() void {
         @atomicStore(u32, &warp_bsp_ready, 0, .release);
         return;
     }
-    var i: u32 = 0;
-    while (i < warp_rounds) : (i += 1) warpTick();
+    const deadline = rdtsc() +% warpTicksFor(warp_check_ns);
+    while (rdtsc() -% deadline >= (1 << 62)) warpTick(); // until the time budget is spent
     @atomicStore(u32, &warp_stop, 1, .release);
     @atomicStore(u32, &warp_bsp_ready, 0, .release);
     warp_checks += 1;
@@ -583,9 +597,10 @@ 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;
+    // The BSP owns the budget; this bound only protects against a lost BSP.
+    const deadline = rdtsc() +% warpTicksFor(2 * warp_check_ns + warp_await_ns);
+    while (@atomicLoad(u32, &warp_stop, .acquire) == 0) {
+        if (rdtsc() -% deadline < (1 << 62)) return; // past the deadline
         warpTick();
     }
 }

system/kernel/console.zig

@@ -138,14 +138,16 @@ pub const Console = struct {
         }
     }
 
-    /// Shift the visible text up one glyph row and clear the freed bottom row,
-    /// leaving the cursor on that now-blank last line.
+    /// The screen is full: start a fresh page at the top. NEVER scroll by
+    /// copying pixel rows — that READS the framebuffer, and VRAM reads are
+    /// uncached-slow on real hardware (measured: 16-core bring-up took ~90 s
+    /// purely from boot lines each paying a whole-screen scroll copy). A page
+    /// clear is writes only, and only once per screenful.
     fn scroll(self: *Console) void {
         const visible = self.rows * glyph_h;
         var y: u32 = 0;
-        while (y + glyph_h < visible) : (y += 1) self.copyRow(y, y + glyph_h);
         while (y < visible) : (y += 1) self.fillRow(y, self.bg);
-        self.row = self.rows - 1;
+        self.row = 0;
     }
 
     inline fn rowPtr(self: *Console, y: u32) [*]volatile u32 {
@@ -163,10 +165,4 @@ pub const Console = struct {
         while (x < self.fb.width) : (x += 1) row[x] = color;
     }
 
-    fn copyRow(self: *Console, destination_y: u32, source_y: u32) void {
-        const destination = self.rowPtr(destination_y);
-        const source = self.rowPtr(source_y);
-        var x: u32 = 0;
-        while (x < self.fb.width) : (x += 1) destination[x] = source[x];
-    }
 };

system/kernel/kernel.zig

@@ -162,6 +162,13 @@ fn kmain(boot_information: *const BootInformation) noreturn {
     // uncached crawl). Routine boot output goes only to the log; this console now exists for
     // early-boot and fatal (`fatal`/panic) output, until the display service takes over.
     console.init(fb);
+    // The console joins the log sinks: the boot transcript — kernel AND
+    // userspace lines, each timestamped by the renderer — shows on screen
+    // until the display service claims the framebuffer (which flips the
+    // console's `suppressed` and silences this sink). On a machine with no
+    // serial this is the only live view of the boot, and a slow boot becomes
+    // diagnosable by eye: the timeline is right there.
+    if (console.present()) log.addSink(console.write);
     log.write(if (console.present())
         "/system/kernel: framebuffer ready (early-boot + fatal fallback; the display service drives it in normal operation)\n"
     else

system/kernel/log.zig

@@ -126,7 +126,7 @@ fn appendLocked(pid: u32, name: []const u8, level: abi.KlogLevel, now: u64, byte
         const line_complete = newline != null or level != .raw;
         if (line.len != 0 or line_complete)
             _ = ring.append(pid, name, level, now, line, line.len > abi.klog_maximum_message);
-        render(pid, name, level, line, line_complete);
+        render(pid, name, level, now, line, line_complete);
         rest = if (newline) |i| rest[i + 1 ..] else rest[rest.len..];
     }
 }
@@ -136,7 +136,7 @@ fn appendLocked(pid: u32, name: []const u8, level: abi.KlogLevel, now: u64, byte
 /// own "name: " prefixes until the std.log migration). Leveled (std.log)
 /// records get a kernel-rendered "<name>: " prefix at line start — err/warn/
 /// debug also get their level spelled out.
-fn render(pid: u32, name: []const u8, level: abi.KlogLevel, line: []const u8, line_complete: bool) void {
+fn render(pid: u32, name: []const u8, level: abi.KlogLevel, now: u64, line: []const u8, line_complete: bool) void {
     if (sink_count == 0) return;
     if (line.len == 0 and !line_complete) return;
     // Compose the whole rendered piece first and emit it in ONE sink call per
@@ -149,6 +149,14 @@ fn render(pid: u32, name: []const u8, level: abi.KlogLevel, line: []const u8, li
         used += 1;
         at_line_start = true;
     }
+    if (at_line_start) {
+        // Every line starts with its boot-relative time: the live transcript
+        // (serial AND the on-screen boot console) is a readable timeline —
+        // which is how a slow real-hardware boot gets diagnosed by eye.
+        const seconds = now / 1_000_000_000;
+        const millis = (now / 1_000_000) % 1000;
+        used += (std.fmt.bufPrint(buffer[used..], "[{d:>4}.{d:0>3}] ", .{ seconds, millis }) catch buffer[used..used]).len;
+    }
     if (at_line_start and level != .raw) {
         used += place(buffer[used..], name);
         used += place(buffer[used..], ": ");
@@ -169,8 +177,8 @@ fn render(pid: u32, name: []const u8, level: abi.KlogLevel, line: []const u8, li
     open_line_pid = pid;
 }
 
-/// newline + name + ": warning: " + a full payload line + newline.
-const render_buffer_size = 1 + abi.maximum_process_name + 11 + abi.klog_maximum_message + 1;
+/// newline + timestamp + name + ": warning: " + a full payload line + newline.
+const render_buffer_size = 1 + 16 + abi.maximum_process_name + 11 + abi.klog_maximum_message + 1;
 
 fn place(destination: []u8, bytes: []const u8) usize {
     const n = @min(destination.len, bytes.len);

system/services/init/init.zig

@@ -27,7 +27,15 @@ const build_options = @import("build_options");
 /// kernel. Drivers are absent on purpose: the device manager owns those. (A
 /// future init reads this from a manifest under /system/services instead of a
 /// hardcoded list.)
-const boot_services = [_][]const u8{
+const boot_services = if (build_options.diagnose) [_][]const u8{
+    // The diagnose boot: no display service, so the kernel's on-screen boot
+    // transcript is never suppressed — the timestamped timeline (USB bring-up,
+    // storage, logger) stays readable on real hardware with no serial.
+    "/system/services/input",
+    "/system/services/device-manager",
+    "/system/services/fat",
+    "/system/services/logger",
+} else [_][]const u8{
     "/system/services/input",
     "/system/services/device-manager",
     "/system/services/fat",

system/services/logger/logger.zig

@@ -112,9 +112,14 @@ fn onNotification(badge: u64) void {
 }
 
 fn onTerminate() void {
-    // Final drain: everything still in the ring, then close (= flush) all files.
+    // The completeness receipt FIRST: this record enters the ring before the
+    // final drain, so the drain carries it into logger.log — a directory whose
+    // logger.log ends with this marker is complete through shutdown; one that
+    // doesn't was cut early and may be missing tails.
+    _ = system.write("logger: shutting down; final flush\n");
     drain();
     closeAll();
+    // Serial-only epilogue (after the drain, so it reaches no file — by design).
     var line: [96]u8 = undefined;
     _ = system.write(std.fmt.bufPrint(&line, "logger: flushed through sequence {d}\n", .{next_expected_sequence}) catch return);
 }

tools/pack-system-image.py

@@ -0,0 +1,53 @@
+#!/usr/bin/env python3
+"""Pack the bundled binaries into the boot capsule (boot/system.img) — the
+single file the EFI loader reads in one sequential pass, which is the only
+shape firmware file I/O is fast at (a per-file tree walk measured minutes on
+real firmware). The format is the v2 initial_ramdisk (system/initial-ramdisk.zig):
+entries named by full FHS path, so the running system is identical whether the
+loader read the capsule or walked the tree.
+
+Usage: pack-system-image.py <out.img> [<path> <file>]...
+Layout (little-endian): Header{magic "DNR2", count}, Entry{name[64], offset, len}*N, blobs.
+"""
+import struct
+import sys
+
+MAGIC = 0x32524E44  # "DNR2"
+HEADER = struct.Struct("<II")
+ENTRY = struct.Struct("<64sQQ")
+
+
+def main() -> int:
+    out_path = sys.argv[1]
+    rest = sys.argv[2:]
+    if len(rest) % 2 != 0:
+        sys.stderr.write("usage: pack-system-image.py <out.img> [<path> <file>]...\n")
+        return 2
+    items = [(rest[i], rest[i + 1]) for i in range(0, len(rest), 2)]
+
+    table_end = HEADER.size + len(items) * ENTRY.size
+    entries = b""
+    blobs = []
+    offset = table_end
+    for path, source in items:
+        name = path if path.startswith("/") else "/" + path
+        encoded = name.encode("ascii")
+        if len(encoded) > 63:
+            sys.stderr.write(f"pack-system-image: path too long (>63): {name}\n")
+            return 2
+        with open(source, "rb") as f:
+            data = f.read()
+        entries += ENTRY.pack(encoded, offset, len(data))
+        blobs.append(data)
+        offset += len(data)
+
+    with open(out_path, "wb") as f:
+        f.write(HEADER.pack(MAGIC, len(items)))
+        f.write(entries)
+        for blob in blobs:
+            f.write(blob)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())