//! /system/services/acpi — the ACPI discovery service: the x86 firmware
//! interpreter, moved out of ring 0 (docs/m19-m20-plan.md, M20). Claims the
//! `acpi-tables` node the kernel publishes (the AML blobs, the broad io_port
//! grant, a broad irq window, the SCI), and runs the **shared AML module** in
//! ring 3 — the same parser and interpreter the kernel uses.
//!
//! M20.2 (this increment): after parsing, walk the namespace and, for each
//! present Device with a hardware id (`_HID`), evaluate its current resource
//! settings (`_CRS`) through a ring-3 `Hal` (port I/O over the claimed node),
//! register it under the acpi-tables node (its I/O ports and IRQs contained by
//! the node's broad grants), and report it to the device manager with its
//! EISA-decoded hid as identity. Matching those reports to drivers (ps2-bus)
//! and retiring the kernel's own device build follow in M20.3.

const std = @import("std");
const runtime = @import("runtime");
const aml = @import("aml");
const device = runtime.device;
const protocol = runtime.device_manager_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 claimed acpi-tables node and the resource index of its broad io_port
// window — the Hal routes every port access through this one claim.
var node_id: u64 = 0;
var io_resource_index: u64 = 0;

// Pass-1 registration record (see main): what pass 2 reports.
const Registered = struct { hid: [8]u8 = .{0} ** 8, hid_len: usize = 0, device_id: u64 = 0, resource_count: u64 = 0 };
var registered: [64]Registered = undefined;
var registered_count: usize = 0;

// A scratch page returned for SystemMemory OperationRegion maps: the service
// cannot map arbitrary physical memory from ring 3, so such regions are
// unsupported and degrade to harmless zeros rather than faulting. The M20.2
// targets (ps2, the legacy devices) use SystemIO and static templates.
var mmio_scratch: [4096]u8 align(4096) = .{0} ** 4096;

fn halMapMmio(physical: u64, len: u64, writable: bool) u64 {
    _ = physical;
    _ = len;
    _ = writable;
    return @intFromPtr(&mmio_scratch);
}

fn halPioRead(width: u8, port: u16) u32 {
    return device.ioRead(node_id, io_resource_index, port, width) orelse 0;
}

fn halPioWrite(width: u8, port: u16, value: u32) void {
    _ = device.ioWrite(node_id, io_resource_index, port, width, value);
}

fn findTablesNode(buffer: []device.DeviceDescriptor) ?device.DeviceDescriptor {
    const total = device.enumerate(buffer);
    for (buffer[0..@min(total, buffer.len)]) |d| {
        if (d.class == @intFromEnum(device.DeviceClass.acpi_tables)) return d;
    }
    return null;
}

pub fn main(init: runtime.process.Init) void {
    // When the acpi-parse scenario spawns this directly, argv[1] is the kernel's
    // own device count to self-verify against — deterministic, no log-scraping.
    const expected: ?usize = if (init.arguments.get(1)) |a| (std.fmt.parseInt(usize, a, 10) catch null) else null;

    const buffer = runtime.allocator().alloc(device.DeviceDescriptor, 64) catch {
        _ = runtime.system.write("acpi: out of memory\n");
        return;
    };
    const node = findTablesNode(buffer) orelse {
        _ = runtime.system.write("acpi: no acpi-tables node to claim\n");
        return;
    };
    node_id = node.id;
    if (!device.claim(node_id)) {
        _ = runtime.system.write("acpi: unable to claim acpi-tables\n");
        return;
    }

    // Map each memory resource (an AML blob) and note the io_port resource.
    var blocks: [8][]const u8 = undefined;
    var block_count: usize = 0;
    var found_io = false;
    for (node.resources[0..@intCast(node.resource_count)], 0..) |resource, index| {
        if (resource.kind == @intFromEnum(device.ResourceKind.io_port) and !found_io) {
            io_resource_index = index;
            found_io = true;
            continue;
        }
        if (resource.kind != @intFromEnum(device.ResourceKind.memory)) continue;
        const base = device.mmioMap(node_id, index) orelse continue;
        const pointer: [*]const u8 = @ptrFromInt(base);
        blocks[block_count] = pointer[0..@intCast(resource.len)];
        block_count += 1;
        if (block_count == blocks.len) break;
    }
    if (block_count == 0) {
        _ = runtime.system.write("acpi: no AML blobs on the node\n");
        return;
    }

    const result = aml.parse(runtime.allocator(), blocks[0..block_count]) catch {
        _ = runtime.system.write("acpi: AML parse failed\n");
        return;
    };
    var namespace = result.namespace;
    const devices = aml.deviceCount(&namespace);
    writeLine("acpi: parsed {d} AML blob(s), {d} namespace devices\n", .{ block_count, devices });
    if (expected) |want| {
        if (devices == want) {
            _ = runtime.system.write("acpi-parse: ok\n");
        } else {
            writeLine("acpi-parse: mismatch (ring-3 {d} vs kernel {d})\n", .{ devices, want });
        }
        // Self-verify mode is standalone (no manager); stop before reporting.
        while (true) runtime.system.sleep(1000);
    }

    // Register + report the present _HID devices (M20.2).
    var arena = std.heap.ArenaAllocator.init(runtime.allocator());
    var interpreter = aml.Interpreter.init(&namespace, .{
        .mapMmio = halMapMmio,
        .pioRead = halPioRead,
        .pioWrite = halPioWrite,
    }, arena.allocator());

    // Pass 1: register every present _HID device under acpi-tables, remembering
    // each (hid, device id). Pass 2: report them all. Registering before any
    // report reaches the manager means a driver it spawns on the first report
    // already sees the whole set (no keyboard-before-mouse race for ps2-bus).
    registered_count = 0;
    walkDevices(namespace.root, &interpreter);

    const manager = runtime.ipc.lookup(.device_manager);
    var i: usize = 0;
    while (i < registered_count) : (i += 1) {
        const entry = registered[i];
        writeLine("acpi: reported {s} (device {d}, {d} resources)\n", .{ entry.hid[0..entry.hid_len], entry.device_id, entry.resource_count });
        if (manager) |h| {
            var report = protocol.ChildAdded{
                .parent = node_id,
                .bus_address = entry.device_id,
                .identity = 0,
                .device_id = entry.device_id,
            };
            @memcpy(report.hid[0..entry.hid_len], entry.hid[0..entry.hid_len]);
            var reply: [protocol.message_maximum]u8 = undefined;
            _ = runtime.ipc.call(h, std.mem.asBytes(&report), &reply) catch {};
        }
    }
    writeLine("acpi: reported {d} device(s) to the manager\n", .{registered_count});

    // Stay resident: the claim holds, and the service is here to grow into the
    // supervised discoverer (M20.3, then the M21 event side on the SCI).
    while (true) runtime.system.sleep(1000);
}

/// Depth-first walk: register + report each present device with a _HID, then
/// descend. Scopes (\_SB, \_GPE …) are descended without producing a node.
fn walkDevices(node: *aml.Node, interpreter: *aml.Interpreter) void {
    var child = node.first_child;
    while (child) |c| : (child = c.next_sibling) {
        if (c.kind != .device) {
            walkDevices(c, interpreter);
            continue;
        }
        if (!devicePresent(interpreter, c)) continue; // absent: skip it and its subtree

        if (readHid(c, interpreter)) |hid| {
            // Skip PCI roots — pci-bus already reports PCI functions; ACPI adds
            // only the non-PCI _HID devices (docs/m19-m20-plan.md M20.2).
            if (!std.mem.eql(u8, hid[0..7], "PNP0A03") and !std.mem.eql(u8, hid[0..7], "PNP0A08")) {
                registerDevice(c, hid, interpreter);
            }
        }
        walkDevices(c, interpreter);
    }
}

fn registerDevice(node: *aml.Node, hid: [8]u8, interpreter: *aml.Interpreter) void {
    if (registered_count >= registered.len) return;
    var descriptor = std.mem.zeroes(device.DeviceDescriptor);
    descriptor.class = @intFromEnum(device.DeviceClass.acpi_device);
    descriptor.pci_class = device.no_pci_class;
    const hid_len: u64 = std.mem.indexOfScalar(u8, &hid, 0) orelse hid.len;
    descriptor.hid_len = hid_len;
    @memcpy(descriptor.hid[0..@intCast(hid_len)], hid[0..@intCast(hid_len)]);
    applyCrs(&descriptor, node, interpreter);

    const id = device.register(node_id, &descriptor) orelse {
        writeLine("acpi: register refused for {s}\n", .{hid[0..@intCast(hid_len)]});
        return;
    };
    registered[registered_count] = .{ .hid = hid, .hid_len = @intCast(hid_len), .device_id = id, .resource_count = descriptor.resource_count };
    registered_count += 1;
}

/// _STA bit 0 (present); absent method or a failed evaluation is treated as
/// present, per the ACPI rules.
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;
}

/// The device's EISA-decoded _HID (e.g. "PNP0303"), or null.
fn readHid(node: *aml.Node, interpreter: *aml.Interpreter) ?[8]u8 {
    const hid = aml.Namespace.childOf(node, seg4("_HID")) orelse return null;
    var buffer: [8]u8 = .{0} ** 8;
    if (hid.kind == .method) {
        const obj = interpreter.evaluate(hid, &.{}) catch return null;
        switch (obj) {
            .integer => |n| {
                _ = eisaIdToStr(@truncate(n), &buffer);
                return buffer;
            },
            else => return null,
        }
    }
    if (hid.kind != .name or hid.value.len == 0) return null;
    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 null;
            _ = eisaIdToStr(@truncate(n), &buffer);
            return buffer;
        },
        else => return null,
    }
}

// --- _CRS resource-template decode (ported from the kernel's acpi.zig) --------

fn applyCrs(descriptor: *device.DeviceDescriptor, 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 bytes = switch (obj) {
        .buffer => |b| b,
        else => return,
    };
    var i: usize = 0;
    while (i < bytes.len) {
        const tag = bytes[i];
        if (tag & 0x80 == 0) {
            const len: usize = tag & 0x07;
            const body = i + 1;
            if (body + len > bytes.len) break;
            switch ((tag >> 3) & 0x0F) {
                0x04 => if (len >= 2) { // IRQ mask
                    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) addResource(descriptor, .irq, b, 1);
                    }
                },
                0x08 => if (len >= 7) addResource(descriptor, .io_port, rd16(bytes, body + 1), bytes[body + 6]),
                0x09 => if (len >= 3) addResource(descriptor, .io_port, rd16(bytes, body), bytes[body + 2]),
                0x0F => break,
                else => {},
            }
            i = body + len;
        } else {
            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) addResource(descriptor, .memory, rd32(bytes, body + 1), rd32(bytes, body + 13)),
                0x86 => if (len >= 9) addResource(descriptor, .memory, rd32(bytes, body + 1), rd32(bytes, body + 5)),
                0x89 => if (len >= 2) {
                    const count = bytes[body + 1];
                    var k: usize = 0;
                    while (k < count and body + 2 + k * 4 + 4 <= body + len) : (k += 1) {
                        addResource(descriptor, .irq, rd32(bytes, body + 2 + k * 4), 1);
                    }
                },
                else => {},
            }
            i = body + len;
        }
    }
}

fn addResource(descriptor: *device.DeviceDescriptor, kind: device.ResourceKind, start: u64, len: u64) void {
    if (descriptor.resource_count >= descriptor.resources.len) return;
    descriptor.resources[@intCast(descriptor.resource_count)] = .{ .kind = @intFromEnum(kind), .start = start, .len = len };
    descriptor.resource_count += 1;
}

// --- small helpers ported verbatim from the kernel's acpi.zig ----------------

fn seg4(comptime s: *const [4:0]u8) [4]u8 {
    return s[0..4].*;
}

fn hexDigit(n: u8) u8 {
    return if (n < 10) '0' + n else 'A' + (n - 10);
}

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);
    buffer[7] = 0;
    return buffer[0..7];
}

fn readIntObj(bytes: []const u8, p: *usize) ?u64 {
    if (p.* >= bytes.len) return null;
    const op = bytes[p.*];
    p.* += 1;
    switch (op) {
        0x00 => return 0,
        0x01 => return 1,
        0xFF => return 1,
        0x0A => {
            if (p.* >= bytes.len) return null;
            const v = bytes[p.*];
            p.* += 1;
            return v;
        },
        0x0B => {
            if (p.* + 2 > bytes.len) return null;
            const v = rd16(bytes, p.*);
            p.* += 2;
            return v;
        },
        0x0C => {
            if (p.* + 4 > bytes.len) return null;
            const v = rd32(bytes, p.*);
            p.* += 4;
            return v;
        },
        else => return null,
    }
}

fn rd16(bytes: []const u8, off: usize) u64 {
    return @as(u64, bytes[off]) | (@as(u64, bytes[off + 1]) << 8);
}

fn rd32(bytes: []const u8, off: usize) u64 {
    return rd16(bytes, off) | (rd16(bytes, off + 2) << 16);
}

pub const panic = runtime.panic;
comptime {
    _ = &runtime.start._start; // pull the runtime entry shim into the image
}