kernel/ATA: Pass extra information to DeviceRegistry

This is needed since merging e7d482e from main.

apps/CMakeLists.txt

@@ -26,3 +26,4 @@ luna_app(login.cpp login)
 luna_app(ipc-test.cpp ipc-test)
 luna_app(mount.cpp mount)
 luna_app(umount.cpp umount)
+luna_app(ps.cpp ps)

apps/chown.cpp

@@ -40,6 +40,8 @@ Result<int> luna_main(int argc, char** argv)
                 }
 
                 gid = grp->gr_gid;
+
+                endgrent();
             }
         }
     }
@@ -73,6 +75,8 @@ Result<int> luna_main(int argc, char** argv)
 
                 gid = pw->pw_gid;
             }
+
+            endpwent();
         }
     }
 

apps/ps.cpp

@@ -0,0 +1,44 @@
+#include <errno.h>
+#include <os/ArgumentParser.h>
+#include <os/File.h>
+#include <pwd.h>
+#include <sys/pstat.h>
+#include <time.h>
+
+Result<int> luna_main(int argc, char** argv)
+{
+    os::ArgumentParser parser;
+    parser.add_description("Show a list of processes on the system.");
+    parser.add_system_program_info("ps"_sv);
+    parser.parse(argc, argv);
+
+    os::println("UID         PID   PPID       TIME CMD");
+
+    pid_t last = pstat(-1, nullptr);
+    for (pid_t pid = 1; pid <= last; pid++)
+    {
+        static process ps;
+        if (pstat(pid, &ps) < 0)
+        {
+            if (errno == ESRCH) continue;
+            return err(errno);
+        }
+
+        struct tm tm;
+        gmtime_r(&ps.ps_time.tv_sec, &tm);
+
+        char timebuf[256];
+        strftime(timebuf, sizeof(timebuf), "%H:%M:%S", &tm);
+
+        const char* user = "???";
+
+        passwd* pw = getpwuid(ps.ps_uid);
+        if (pw) user = pw->pw_name;
+
+        os::println("%-8s %6d %6d %10s %s", user, ps.ps_pid, ps.ps_ppid, timebuf, ps.ps_name);
+    }
+
+    endpwent();
+
+    return 0;
+}

apps/sh.cpp

@@ -74,6 +74,7 @@ Result<int> luna_main(int argc, char** argv)
         struct passwd* pw = getpwuid(getuid());
         if (pw) { username = pw->pw_name; }
         else { username = getenv("USER"); }
+        endpwent();
     }
 
     while (1)

apps/su.cpp

@@ -78,6 +78,8 @@ Result<int> luna_main(int argc, char** argv)
         return 1;
     }
 
+    endpwent();
+
     if ((prompt_password || getuid() != geteuid()) && *entry->pw_passwd)
     {
         char* pass = getpass();

kernel/CMakeLists.txt

@@ -31,7 +31,7 @@ set(SOURCES
     src/sys/file.cpp
     src/sys/id.cpp
     src/sys/mkdir.cpp
-    src/sys/mknod.cpp
+    src/sys/pstat.cpp
     src/sys/waitpid.cpp
     src/sys/getdents.cpp
     src/sys/stat.cpp
@@ -63,6 +63,7 @@ if("${LUNA_ARCH}" MATCHES "x86_64")
         src/arch/x86_64/Thread.cpp
         src/arch/x86_64/PCI.cpp
         src/arch/x86_64/Keyboard.cpp
+        src/arch/x86_64/disk/ATA.cpp
         src/arch/x86_64/init/GDT.cpp
         src/arch/x86_64/init/IDT.cpp
         src/arch/x86_64/init/PIC.cpp

kernel/src/arch/CPU.h

@@ -22,6 +22,9 @@ namespace CPU
     void disable_interrupts();
     void wait_for_interrupt();
 
+    bool save_interrupts();
+    void restore_interrupts(bool saved);
+
     void get_stack_trace(void (*callback)(u64, void*), void* arg);
     void print_stack_trace();
     void get_stack_trace_at(Registers* regs, void (*callback)(u64, void*), void* arg);
@@ -29,5 +32,8 @@ namespace CPU
 
     u16 get_processor_id();
 
+    bool register_interrupt(u8 interrupt, void (*handler)(Registers*, void*), void* context);
+    void sync_interrupts();
+
     void pause();
 }

kernel/src/arch/PCI.cpp

@@ -9,6 +9,10 @@ struct ScanInfo
 
 namespace PCI
 {
+    BAR::BAR(u32 raw) : m_raw(raw)
+    {
+    }
+
     Device::ID read_id(const Device::Address& address)
     {
         const u16 vendor = read16(address, Field::VendorID);
@@ -108,4 +112,13 @@ namespace PCI
             }
         }
     }
+
+    BAR Device::getBAR(u8 index) const
+    {
+        check(index < 6);
+
+        u32 raw = read32(address, 0x10 + (index * 4));
+
+        return { raw };
+    }
 }

kernel/src/arch/PCI.h

@@ -17,7 +17,67 @@ namespace PCI
 
         HeaderType = 0x0e,
 
-        SecondaryBus = 0x19
+        BAR0 = 0x10,
+        BAR1 = 0x14,
+        BAR2 = 0x18,
+        BAR3 = 0x1c,
+
+        SecondaryBus = 0x19,
+
+        InterruptLine = 0x3c,
+    };
+
+    enum CommandField : u16
+    {
+        CMD_IO_SPACE = 1 << 0,
+        CMD_MEMORY_SPACE = 1 << 1,
+        CMD_BUS_MASTER = 1 << 2,
+        CMD_SPECIAL_CYCLES = 1 << 3,
+        CMD_MEMORY_WRITE_AND_INVALIDATE = 1 << 4,
+        CMD_VGA_PALETTE_SNOOP = 1 << 5,
+        CMD_PARITY_ERROR_RESPONSE = 1 << 6,
+        CMD_SERR = 1 << 8,
+        CMD_FAST_BACK_TO_BACK = 1 << 9,
+        CMD_INTERRUPT_DISABLE = 1 << 10,
+    };
+
+    struct BAR
+    {
+      public:
+        BAR(u32 raw);
+
+        bool is_iospace()
+        {
+            return m_raw & 0x01;
+        }
+
+        bool is_memory_space()
+        {
+            return !is_iospace();
+        }
+
+        u16 port()
+        {
+            return (u16)(m_raw & 0xfffffffc);
+        }
+
+        u8 type()
+        {
+            return (m_raw >> 1) & 0x03;
+        }
+
+        bool is_prefetchable()
+        {
+            return m_raw & (1 << 3);
+        }
+
+        u32 address_32bit()
+        {
+            return m_raw & 0xFFFFFFF0;
+        }
+
+      private:
+        u32 m_raw;
     };
 
     struct Device
@@ -42,6 +102,8 @@ namespace PCI
             u32 function;
         };
 
+        BAR getBAR(u8 index) const;
+
         ID id;
         Type type;
         Address address;

kernel/src/arch/x86_64/CPU.asm

@@ -188,4 +188,18 @@ ISR_ERROR 21 ; control-protection exception (#CP)
 ; ISR 22-31 reserved
 IRQ 32, 0 ; timer interrupt
 IRQ 33, 1 ; keyboard interrupt
+IRQ 34, 2
+IRQ 35, 3
+IRQ 36, 4
+IRQ 37, 5
+IRQ 38, 6
+IRQ 39, 7
+IRQ 40, 8
+IRQ 41, 9
+IRQ 42, 10
+IRQ 43, 11
+IRQ 44, 12
+IRQ 45, 13
+IRQ 46, 14
+IRQ 47, 15
 ISR 66 ; system call

kernel/src/arch/x86_64/CPU.cpp

@@ -22,12 +22,23 @@ extern "C" void enable_nx();
 
 extern void setup_gdt();
 extern void remap_pic();
+extern void change_pic_masks(u8 pic1_mask, u8 pic2_mask);
 extern void pic_eoi(unsigned char irq);
 extern void pic_eoi(Registers* regs);
 extern void setup_idt();
 
 static Thread* g_io_thread;
 
+typedef void (*interrupt_handler_t)(Registers*, void*);
+
+struct InterruptHandler
+{
+    interrupt_handler_t function;
+    void* context;
+};
+
+static InterruptHandler irq_handlers[16];
+
 void FPData::save()
 {
     asm volatile("fxsave (%0)" : : "r"(m_data));
@@ -135,21 +146,34 @@ void io_thread()
     }
 }
 
-// Called from _asm_interrupt_entry
+static void timer_interrupt(Registers* regs, void*)
-extern "C" void arch_interrupt_entry(Registers* regs)
-{
-    if (regs->isr < 32) handle_x86_exception(regs);
-    else if (regs->isr == 32) // Timer interrupt
 {
     Timer::tick();
     if (should_invoke_scheduler()) Scheduler::invoke(regs);
-        pic_eoi(regs);
 }
-    else if (regs->isr == 33) // Keyboard interrupt
+
+static void keyboard_interrupt(Registers*, void*)
 {
     u8 scancode = IO::inb(0x60);
     scancode_queue.try_push(scancode);
     g_io_thread->wake_up();
+}
+
+// Called from _asm_interrupt_entry
+extern "C" void arch_interrupt_entry(Registers* regs)
+{
+    if (regs->isr < 32) handle_x86_exception(regs);
+    else if (regs->isr >= 32 && regs->isr < 48) // IRQ from the PIC
+    {
+        u64 irq = regs->error;
+        auto handler = irq_handlers[irq];
+        if (!handler.function)
+        {
+            kwarnln("Unhandled IRQ catched! Halting.");
+            CPU::efficient_halt();
+        }
+
+        handler.function(regs, handler.context);
         pic_eoi(regs);
     }
     else if (regs->isr == 66) // System call
@@ -159,7 +183,7 @@ extern "C" void arch_interrupt_entry(Registers* regs)
     }
     else
     {
-        kwarnln("IRQ catched! Halting.");
+        kwarnln("Unhandled interrupt catched! Halting.");
         CPU::efficient_halt();
     }
 }
@@ -218,6 +242,10 @@ namespace CPU
             kwarnln("not setting the NX bit as it is unsupported");
         setup_gdt();
         setup_idt();
+
+        memset(irq_handlers, 0, sizeof(irq_handlers));
+        register_interrupt(0, timer_interrupt, nullptr);
+        register_interrupt(1, keyboard_interrupt, nullptr);
     }
 
     void platform_finish_init()
@@ -226,6 +254,8 @@ namespace CPU
                           .expect_value("Could not create the IO background thread!");
 
         remap_pic();
+
+        sync_interrupts();
     }
 
     void enable_interrupts()
@@ -238,6 +268,20 @@ namespace CPU
         asm volatile("cli");
     }
 
+    bool save_interrupts()
+    {
+        u64 flags;
+        asm volatile("pushfq; pop %0" : "=r"(flags));
+        return flags & 0x200;
+    }
+
+    void restore_interrupts(bool saved)
+    {
+        if (saved) enable_interrupts();
+        else
+            disable_interrupts();
+    }
+
     void wait_for_interrupt()
     {
         asm volatile("hlt");
@@ -338,6 +382,35 @@ namespace CPU
         __get_cpuid(1, &unused, &ebx, &unused, &unused);
         return (u16)(ebx >> 24);
     }
+
+    bool register_interrupt(u8 interrupt, interrupt_handler_t handler, void* context)
+    {
+        if (irq_handlers[interrupt].function) return false;
+
+        irq_handlers[interrupt] = { handler, context };
+
+        sync_interrupts();
+
+        return true;
+    }
+
+    void sync_interrupts()
+    {
+        u8 pic1_mask, pic2_mask;
+        pic1_mask = pic2_mask = 0b11111111;
+        for (int i = 0; i < 8; i++)
+        {
+            if (irq_handlers[i].function) pic1_mask &= (u8)(~(1 << i));
+            if (irq_handlers[i + 8].function) pic2_mask &= (u8)(~(1 << i));
+        }
+
+        if (pic2_mask != 0b11111111) pic1_mask &= 0b11111011;
+
+        auto val = CPU::save_interrupts();
+        CPU::disable_interrupts();
+        change_pic_masks(pic1_mask, pic2_mask);
+        CPU::restore_interrupts(val);
+    }
 }
 
 // called by kernel_yield

kernel/src/arch/x86_64/PCI.cpp

@@ -33,19 +33,32 @@ namespace PCI
 
     void write8(const Device::Address& address, u32 field, u8 value)
     {
-        ignore(address, field, value);
+        u8 offset = (u8)(field & ~0x3);
-        todo();
+        union {
+            u8 split[4];
+            u32 full;
+        };
+        full = read32(address, offset);
+        split[(field & 0x3)] = value;
+        write32(address, offset, full);
     }
 
-    void write16(const Device::Address& address, u32 field, u8 value)
+    void write16(const Device::Address& address, u32 field, u16 value)
     {
-        ignore(address, field, value);
+        u8 offset = (u8)(field & ~0x3);
-        todo();
+        union {
+            u8 split[4];
+            u32 full;
+        };
+        full = read32(address, offset);
+        split[(field & 0x3)] = (u8)(value >> 8);
+        split[(field & 0x3) + 1] = (u8)(value & 0xff);
+        write32(address, offset, full);
     }
 
-    void write32(const Device::Address& address, u32 field, u8 value)
+    void write32(const Device::Address& address, u32 field, u32 value)
     {
-        ignore(address, field, value);
+        IO::outl(PCI_ADDRESS_PORT, make_pci_address(address, field));
-        todo();
+        IO::outl(PCI_VALUE_PORT, value);
     }
 }

kernel/src/arch/x86_64/disk/ATA.cpp

@@ -0,0 +1,780 @@
+#include "arch/x86_64/disk/ATA.h"
+#include "Log.h"
+#include "arch/Serial.h"
+#include "arch/Timer.h"
+#include "arch/x86_64/IO.h"
+#include "memory/MemoryManager.h"
+#include <luna/Alignment.h>
+#include <luna/CType.h>
+#include <luna/SafeArithmetic.h>
+#include <luna/Vector.h>
+
+SharedPtr<ATA::Controller> g_controller;
+
+static void irq_handler(Registers* regs, void* ctx)
+{
+    ((ATA::Channel*)ctx)->irq_handler(regs);
+}
+
+static usize copy_ata_string(char* out, u16* in, usize size)
+{
+    for (usize i = 0; i < size; i += 2)
+    {
+        u16 val = in[i / 2];
+        out[i] = (u8)(val >> 8);
+        out[i + 1] = (u8)(val & 0xff);
+    }
+
+    out[size + 1] = '\0';
+
+    return size;
+}
+
+namespace ATA
+{
+    Result<void> Controller::scan()
+    {
+        // FIXME: Propagate errors.
+        PCI::scan(
+            [](const PCI::Device& device) {
+                if (!g_controller)
+                {
+                    auto controller = adopt_shared_if_nonnull(new (std::nothrow) Controller(device)).release_value();
+                    kinfoln("ata: Found ATA controller on PCI bus (%x:%x:%x)", device.address.bus,
+                            device.address.function, device.address.slot);
+
+                    if (controller->initialize()) g_controller = controller;
+                }
+            },
+            { .klass = 1, .subclass = 1 });
+
+        if (!g_controller) kwarnln("ata: No ATA controller found.");
+
+        return {};
+    }
+
+    bool Controller::initialize()
+    {
+        u16 command_old = PCI::read16(m_device.address, PCI::Command);
+        u16 command_new = command_old;
+
+        command_new &= ~PCI::CMD_INTERRUPT_DISABLE;
+        command_new |= PCI::CMD_IO_SPACE;
+        command_new |= PCI::CMD_BUS_MASTER;
+
+        if (command_new != command_old) PCI::write16(m_device.address, PCI::Command, command_new);
+
+        if (!m_primary_channel.initialize()) return false;
+
+        return m_secondary_channel.initialize();
+    }
+
+    Controller::Controller(const PCI::Device& device)
+        : m_device(device), m_primary_channel(this, 0, {}), m_secondary_channel(this, 1, {})
+    {
+    }
+
+    Channel::Channel(Controller* controller, u8 channel_index, Badge<Controller>)
+        : m_controller(controller), m_channel_index(channel_index)
+    {
+    }
+
+    u8 Channel::read_register(Register reg)
+    {
+        return IO::inb(m_io_base + (u16)reg);
+    }
+
+    u16 Channel::read_data()
+    {
+        return IO::inw(m_io_base + (u16)Register::Data);
+    }
+
+    void Channel::write_data(u16 value)
+    {
+        IO::outw(m_io_base + (u16)Register::Data, value);
+    }
+
+    void Channel::write_register(Register reg, u8 value)
+    {
+        IO::outb(m_io_base + (u16)reg, value);
+    }
+
+    u8 Channel::read_control(ControlRegister reg)
+    {
+        return IO::inb(m_control_base + (u16)reg);
+    }
+
+    void Channel::write_control(ControlRegister reg, u8 value)
+    {
+        IO::outb(m_control_base + (u16)reg, value);
+    }
+
+    u8 Channel::read_bm(BusmasterRegister reg)
+    {
+        return IO::inb(m_busmaster_base + (u16)reg);
+    }
+
+    void Channel::write_bm(BusmasterRegister reg, u8 value)
+    {
+        IO::outb(m_busmaster_base + (u16)reg, value);
+    }
+
+    u32 Channel::read_prdt_address()
+    {
+        return IO::inl(m_busmaster_base + (u16)BusmasterRegister::PRDTAddress);
+    }
+
+    void Channel::write_prdt_address(u32 value)
+    {
+        IO::outl(m_busmaster_base + (u16)BusmasterRegister::PRDTAddress, value);
+    }
+
+    void Channel::delay_400ns()
+    {
+        // FIXME: We should use kernel_sleep(), but it doesn't support nanosecond granularity.
+        for (int i = 0; i < 14; i++) { [[maybe_unused]] volatile u8 val = read_control(ControlRegister::AltStatus); }
+    }
+
+    void Channel::select(u8 drive)
+    {
+        if (drive == m_current_drive) return;
+
+        u8 value = (drive << 4) | 0xa0;
+        write_register(Register::DriveSelect, value);
+
+        delay_400ns();
+
+        m_current_drive = drive;
+    }
+
+    void Channel::irq_handler(Registers*)
+    {
+        // FIXME: Read the Busmaster register to make sure the IRQ is for this channel.
+
+        if (m_current_drive < 2 && m_drives[m_current_drive]) m_drives[m_current_drive]->irq_handler();
+
+        m_irq_called = true;
+
+        if (m_thread) m_thread->wake_up();
+    }
+
+    void Channel::prepare_for_irq()
+    {
+        m_thread = Scheduler::current();
+        m_irq_called = false;
+    }
+
+    void Channel::wait_for_irq()
+    {
+        if (!m_irq_called) kernel_wait_for_event();
+
+        m_irq_called = false;
+    }
+
+    bool Channel::wait_for_irq_or_timeout(u64 timeout)
+    {
+        if (!m_irq_called)
+        {
+            kernel_sleep(timeout);
+            m_irq_called = false;
+            return m_thread->sleep_ticks_left;
+        }
+
+        m_irq_called = false;
+        return true;
+    }
+
+    bool Channel::wait_for_reg_set(Register reg, u8 value, u64 timeout)
+    {
+        u64 begin = Timer::ticks_ms();
+        while (true)
+        {
+            u8 reg_value = reg == Register::Status ? read_control(ControlRegister::AltStatus) : read_register(reg);
+            if (reg_value & value) return true;
+            if ((Timer::ticks_ms() - begin) >= timeout) return false;
+            kernel_sleep(1);
+        }
+    }
+
+    bool Channel::wait_for_reg_clear(Register reg, u8 value, u64 timeout)
+    {
+        u64 begin = Timer::ticks_ms();
+        while (true)
+        {
+            u8 reg_value = reg == Register::Status ? read_control(ControlRegister::AltStatus) : read_register(reg);
+            if ((reg_value & value) == 0) return true;
+            if ((Timer::ticks_ms() - begin) >= timeout) return false;
+            kernel_sleep(1);
+        }
+    }
+
+    Result<void> Channel::wait_until_ready()
+    {
+        if (!wait_for_reg_clear(Register::Status, SR_Busy, 1000))
+        {
+            kwarnln("ata: Drive %d:%d timed out (BSY)", m_channel_index, m_current_drive);
+            return err(EIO);
+        }
+
+        if (!wait_for_reg_set(Register::Status, SR_DataRequestReady | SR_Error, 1000))
+        {
+            kwarnln("ata: Drive %d:%d timed out (DRQ)", m_channel_index, m_current_drive);
+            return err(EIO);
+        }
+
+        u8 status = read_control(ControlRegister::AltStatus);
+        if (status & SR_Error)
+        {
+            kwarnln("ata: An error occurred in drive %d:%d while waiting for data to become available", m_channel_index,
+                    m_current_drive);
+            return err(EIO);
+        }
+
+        return {};
+    }
+
+    bool Channel::initialize()
+    {
+        int offset = m_channel_index ? 2 : 0;
+        m_is_pci_native_mode = m_controller->device().type.prog_if & (1 << offset);
+
+        u16 control_port_base_address;
+        u16 io_base_address;
+
+        if (m_is_pci_native_mode)
+        {
+            auto io_base = m_controller->device().getBAR(m_channel_index ? 2 : 0);
+            if (!io_base.is_iospace())
+            {
+                kwarnln("ata: Channel %d's IO base BAR is not in IO space", m_channel_index);
+                return false;
+            }
+            io_base_address = io_base.port();
+
+            auto io_control = m_controller->device().getBAR(m_channel_index ? 3 : 1);
+            if (!io_control.is_iospace())
+            {
+                kwarnln("ata: Channel %d's control base BAR is not in IO space", m_channel_index);
+                return false;
+            }
+
+            control_port_base_address = io_control.port() + 2;
+        }
+        else
+        {
+            io_base_address = m_channel_index ? 0x170 : 0x1f0;
+            control_port_base_address = m_channel_index ? 0x376 : 0x3f6;
+        }
+
+        m_io_base = io_base_address;
+        m_control_base = control_port_base_address;
+
+        auto io_busmaster = m_controller->device().getBAR(4);
+        if (!io_busmaster.is_iospace())
+        {
+            kwarnln("ata: Channel %d's busmaster base BAR is not in IO space", m_channel_index);
+            return false;
+        }
+        m_busmaster_base = io_busmaster.port() + (u16)(m_channel_index * 8u);
+
+        if (m_is_pci_native_mode) m_interrupt_line = PCI::read8(m_controller->device().address, PCI::InterruptLine);
+        else
+            m_interrupt_line = m_channel_index ? 15 : 14;
+
+        write_control(ControlRegister::DeviceControl, 0);
+
+        for (u8 drive = 0; drive < 2; drive++)
+        {
+            ScopedKMutexLock<100> lock(m_lock);
+
+            select(drive);
+
+            if (read_register(Register::Status) == 0)
+            {
+                // No drive on this slot.
+                continue;
+            }
+
+            kinfoln("ata: Channel %d has a drive on slot %d!", m_channel_index, drive);
+
+            auto rc = adopt_shared_if_nonnull(new (std::nothrow) Drive(this, drive, {}));
+            if (rc.has_error())
+            {
+                kinfoln("ata: Failed to create drive object: %s", rc.error_string());
+                return false;
+            }
+
+            m_drives[drive] = rc.release_value();
+
+            if (!m_drives[drive]->initialize())
+            {
+                m_drives[drive] = {};
+                return false;
+            }
+        }
+
+        bool ok = CPU::register_interrupt(m_interrupt_line, ::irq_handler, this);
+        if (!ok)
+        {
+            kerrorln("ata: Failed to register IRQ handler for ATA channel %d (IRQ %d)", m_channel_index,
+                     m_interrupt_line);
+            return false;
+        }
+
+        for (u8 drive = 0; drive < 2; drive++)
+        {
+            if (m_drives[drive])
+            {
+                if (!m_drives[drive]->post_initialize())
+                {
+                    m_drives[drive] = {};
+                    return false;
+                }
+
+                if (ATADevice::create(m_drives[drive]).has_error())
+                {
+                    kwarnln("ata: Failed to register ATA drive %d:%d in DeviceRegistry", m_channel_index, drive);
+                    continue;
+                }
+            }
+        }
+
+        return true;
+    }
+
+    Drive::Drive(Channel* channel, u8 drive_index, Badge<Channel>) : m_channel(channel), m_drive_index(drive_index)
+    {
+    }
+
+    bool Drive::identify_ata()
+    {
+        m_channel->write_register(Register::Command, m_is_atapi ? CMD_Identify_Packet : CMD_Identify);
+
+        m_channel->delay_400ns();
+
+        if (!m_channel->wait_for_reg_clear(Register::Status, SR_Busy, 1000))
+        {
+            kwarnln("ata: Drive %d timed out clearing SR_Busy (waited for 1000 ms)", m_drive_index);
+            return false;
+        }
+
+        if (m_channel->read_register(Register::Status) & SR_Error)
+        {
+            u8 lbam = m_channel->read_register(Register::LBAMiddle);
+            u8 lbah = m_channel->read_register(Register::LBAHigh);
+
+            if ((lbam == 0x14 && lbah == 0xeb) || (lbam == 0x69 && lbah == 0x96))
+            {
+                if (!m_is_atapi)
+                {
+                    kinfoln("ata: Drive %d is ATAPI, sending IDENTIFY_PACKET command", m_drive_index);
+                    m_is_atapi = true;
+                    return identify_ata();
+                }
+            }
+
+            kwarnln("ata: IDENTIFY command for drive %d returned error", m_drive_index);
+
+            return false;
+        }
+
+        if (!m_channel->wait_for_reg_set(Register::Status, SR_DataRequestReady | SR_Error, 1000))
+        {
+            kwarnln("ata: Drive %d timed out setting SR_DataRequestReady (waited for 1000 ms)", m_drive_index);
+            return false;
+        }
+
+        u8 status = m_channel->read_register(Register::Status);
+        if (status & SR_Error)
+        {
+            kwarnln("ata: IDENTIFY command for drive %d returned error", m_drive_index);
+            return false;
+        }
+
+        for (usize i = 0; i < 256; i++)
+        {
+            u16 data = m_channel->read_data();
+            m_identify_words[i] = data;
+        }
+
+        return true;
+    }
+
+    bool Drive::initialize()
+    {
+        m_channel->select(m_drive_index);
+
+        m_channel->write_register(Register::SectorCount, 0);
+        m_channel->write_register(Register::LBALow, 0);
+        m_channel->write_register(Register::LBAMiddle, 0);
+        m_channel->write_register(Register::LBAHigh, 0);
+
+        if (!identify_ata()) return false;
+
+        m_serial.set_length(copy_ata_string(m_serial.data(), &m_identify_words[10], SERIAL_LEN));
+        m_revision.set_length(copy_ata_string(m_revision.data(), &m_identify_words[23], REVISION_LEN));
+        m_model.set_length(copy_ata_string(m_model.data(), &m_identify_words[27], MODEL_LEN));
+
+        m_serial.trim(" ");
+        m_revision.trim(" ");
+        m_model.trim(" ");
+
+        kinfoln("ata: Drive IDENTIFY returned serial='%s', revision='%s' and model='%s'", m_serial.chars(),
+                m_revision.chars(), m_model.chars());
+
+        auto status = m_channel->read_bm(BusmasterRegister::Status);
+        if (status & BMS_SimplexOnly)
+        {
+            kwarnln("ata: Drive %d will not use DMA because of simplex shenanigans", m_drive_index);
+            m_uses_dma = false;
+        }
+
+        auto frame = MemoryManager::alloc_frame();
+        if (frame.has_error() || frame.value() > 0xffffffff)
+        {
+            kwarnln("ata: Failed to allocate memory below the 32-bit limit for the PRDT");
+            return false;
+        }
+        m_dma_prdt_phys = frame.release_value();
+        m_dma_prdt = (prdt_entry*)MMU::translate_physical_address(m_dma_prdt_phys);
+
+        memset(m_dma_prdt, 0, ARCH_PAGE_SIZE);
+
+        frame = MemoryManager::alloc_frame();
+        if (frame.has_error() || frame.value() > 0xffffffff)
+        {
+            kwarnln("ata: Failed to allocate memory below the 32-bit limit for DMA memory");
+            return false;
+        }
+        m_dma_mem_phys = frame.release_value();
+        m_dma_mem = (void*)MMU::translate_physical_address(m_dma_mem_phys);
+
+        memset(const_cast<void*>(m_dma_mem), 0, ARCH_PAGE_SIZE);
+
+        if (m_uses_dma)
+        {
+            auto cmd = m_channel->read_bm(BusmasterRegister::Command);
+            cmd &= ~BMC_StartStop;
+            m_channel->write_bm(BusmasterRegister::Command, cmd);
+        }
+
+        return true;
+    }
+
+    bool Drive::post_initialize()
+    {
+        if (m_is_atapi)
+        {
+            atapi_packet packet;
+            memset(&packet, 0, sizeof(packet));
+            packet.command_bytes[0] = ATAPI_ReadCapacity;
+
+            atapi_read_capacity_reply reply;
+
+            if (send_packet_atapi_pio(&packet, &reply, sizeof(reply)).has_error())
+            {
+                kwarnln("ata: Failed to send Read Capacity command to ATAPI drive");
+                return false;
+            }
+
+            m_is_lba48 = true;
+
+            // FIXME: This assumes the host machine is little-endian.
+            u32 last_lba = __builtin_bswap32(reply.last_lba);
+            u32 sector_size = __builtin_bswap32(reply.sector_size);
+
+            m_block_count = last_lba + 1;
+            m_block_size = sector_size;
+        }
+        else
+        {
+            if (m_identify_data.big_lba) m_is_lba48 = true;
+
+            if (m_is_lba48) m_block_count = m_identify_data.sectors_48;
+            else
+                m_block_count = m_identify_data.sectors_28;
+
+            // FIXME: Should we check for CHS?
+
+            // FIXME: Maybe a different block size is in use? Detect that.
+            m_block_size = 512;
+        }
+
+        u64 total_capacity;
+        if (!safe_mul(m_block_count, m_block_size).try_set_value(total_capacity))
+        {
+            kwarnln("ata: Drive %d's total capacity is too large", m_drive_index);
+            return false;
+        }
+
+        kinfoln("ata: Drive %d capacity information: Block Count=%lu, Block Size=%lu, Total Capacity=%lu",
+                m_drive_index, m_block_count, m_block_size, total_capacity);
+
+        char buf[2048];
+        if (atapi_read_pio(0, buf, sizeof(buf)).has_error()) return false;
+
+        char readable[2049];
+        for (int i = 0; i < 2048; i++) { readable[i] = _iscntrl(buf[i]) ? '.' : buf[i]; }
+        readable[2048] = 0;
+        kinfoln("ata: Read first sector from CD-ROM: %s", readable);
+
+        return true;
+    }
+
+    Result<void> Drive::send_packet_atapi_pio(const atapi_packet* packet, void* out, u16 response_size)
+    {
+        u8* ptr = (u8*)out;
+
+        m_channel->select(m_drive_index);
+
+        // We use PIO here.
+        m_channel->write_register(Register::Features, 0x00);
+
+        m_channel->write_register(Register::LBAMiddle, (u8)(response_size & 0xff));
+        m_channel->write_register(Register::LBAHigh, (u8)(response_size >> 8));
+
+        m_channel->write_register(Register::Command, CMD_Packet);
+
+        m_channel->delay_400ns();
+
+        usize i = 0;
+
+        TRY(m_channel->wait_until_ready());
+
+        for (int j = 0; j < 6; j++) m_channel->write_data(packet->command_words[j]);
+
+        while (i < response_size)
+        {
+            TRY(m_channel->wait_until_ready());
+
+            usize byte_count =
+                m_channel->read_register(Register::LBAHigh) << 8 | m_channel->read_register(Register::LBAMiddle);
+            usize word_count = byte_count / 2;
+
+            while (word_count--)
+            {
+                u16 value = m_channel->read_data();
+                ptr[0] = (u8)(value & 0xff);
+                ptr[1] = (u8)(value >> 8);
+                ptr += 2;
+            }
+
+            i += byte_count;
+        }
+
+        return {};
+    }
+
+#if 0
+
+    Result<void> Drive::send_packet_atapi_dma(const atapi_packet* packet, void* out, u16 response_size)
+    {
+        check(m_uses_dma);
+
+        m_channel->select(m_drive_index);
+
+        kdbgln("have selected");
+
+        // We use DMA here.
+        m_channel->write_register(Register::Features, 0x01);
+
+        m_channel->write_register(Register::LBAMiddle, 0);
+        m_channel->write_register(Register::LBAHigh, 0);
+
+        kdbgln("will do_dma_command");
+
+        TRY(do_dma_command(CMD_Packet, response_size, false));
+
+        TRY(m_channel->wait_until_ready());
+
+        kdbgln("send atapi packet data");
+
+        for (int j = 0; j < 6; j++) m_channel->write_data(packet->command_words[j]);
+
+        kdbgln("do dma transfer");
+
+        TRY(do_dma_transfer());
+
+        memcpy(out, const_cast<void*>(m_dma_mem), response_size);
+
+        return {};
+    }
+
+    Result<void> Drive::do_dma_command(u8 command, u16 count, bool write)
+    {
+        m_dma_prdt->address = (u32)m_dma_mem_phys;
+        m_dma_prdt->count = count;
+        m_dma_prdt->flags = END_OF_PRDT;
+
+        kdbgln("ata: do_dma_command: phys=%x, command=%x, count=%u, write=%d", m_dma_prdt->address, command, count,
+               write);
+
+        m_channel->write_prdt_address((u32)m_dma_prdt_phys);
+
+        auto status = m_channel->read_bm(BusmasterRegister::Status);
+        status &= ~(BMS_DMAFailure | BMS_IRQPending);
+        m_channel->write_bm(BusmasterRegister::Status, status);
+
+        auto cmd = m_channel->read_bm(BusmasterRegister::Command);
+        if (!write) cmd |= BMC_ReadWrite;
+        else
+            cmd &= ~BMC_ReadWrite;
+        m_channel->write_bm(BusmasterRegister::Command, cmd);
+
+        m_channel->prepare_for_irq();
+
+        m_channel->write_register(Register::Command, command);
+
+        cmd = m_channel->read_bm(BusmasterRegister::Command);
+        cmd |= BMC_StartStop;
+        m_channel->write_bm(BusmasterRegister::Command, cmd);
+
+        m_channel->delay_400ns();
+
+        return {};
+    }
+
+    Result<void> Drive::do_dma_transfer()
+    {
+        if (!m_channel->wait_for_irq_or_timeout(2000))
+        {
+            kwarnln("ata: Drive %d timed out (DMA)", m_drive_index);
+            return err(EIO);
+        }
+
+        u8 status = m_channel->read_control(ControlRegister::AltStatus);
+        kdbgln("ata: status after irq: %#x", status);
+
+        m_channel->delay_400ns();
+
+        auto cmd = m_channel->read_bm(BusmasterRegister::Command);
+        cmd &= ~BMC_StartStop;
+        m_channel->write_bm(BusmasterRegister::Command, cmd);
+
+        status = m_channel->read_bm(BusmasterRegister::Status);
+        m_channel->write_bm(BusmasterRegister::Status, status & ~(BMS_DMAFailure | BMS_IRQPending));
+
+        if (status & BMS_DMAFailure)
+        {
+            kwarnln("ata: DMA failure while trying to read drive %d", m_drive_index);
+            return err(EIO);
+        }
+
+        return {};
+    }
+
+#endif
+
+    Result<void> Drive::atapi_read_pio(u64 lba, void* out, usize size)
+    {
+        check(lba < m_block_count);
+        check(size <= ARCH_PAGE_SIZE);
+
+        atapi_packet read_packet;
+        memset(&read_packet, 0, sizeof(read_packet));
+        read_packet.command_bytes[0] = ATAPI_Read;
+        read_packet.command_bytes[2] = (lba >> 0x18) & 0xff;
+        read_packet.command_bytes[3] = (lba >> 0x10) & 0xff;
+        read_packet.command_bytes[4] = (lba >> 0x08) & 0xff;
+        read_packet.command_bytes[5] = (lba >> 0x00) & 0xff;
+        read_packet.command_bytes[9] = (u8)(size / m_block_size);
+
+        return send_packet_atapi_pio(&read_packet, out, (u16)size);
+    }
+
+    Result<void> Drive::read_lba(u64 lba, void* out, usize nblocks)
+    {
+        const usize blocks_per_page = ARCH_PAGE_SIZE / m_block_size;
+        if (m_is_atapi)
+        {
+            kdbgln("ata: Reading %zu blocks from ATAPI drive using PIO, at LBA %ld", nblocks, lba);
+            while (nblocks > blocks_per_page)
+            {
+                TRY(atapi_read_pio(lba, out, ARCH_PAGE_SIZE));
+                lba += blocks_per_page;
+                nblocks -= blocks_per_page;
+                out = offset_ptr(out, ARCH_PAGE_SIZE);
+            }
+            return atapi_read_pio(lba, out, nblocks * m_block_size);
+        }
+        else
+            todo();
+    }
+
+    void Drive::irq_handler()
+    {
+        // Clear the IRQ flag.
+        u8 status = m_channel->read_register(Register::Status);
+
+        if (status & SR_Error)
+        {
+            u8 error = m_channel->read_register(Register::Error);
+            (void)error;
+        }
+
+        if (m_uses_dma)
+        {
+            status = m_channel->read_bm(BusmasterRegister::Status);
+            if (status & BMS_DMAFailure) { kwarnln("ata: DMA failure in irq"); }
+            m_channel->write_bm(BusmasterRegister::Status, 4);
+        }
+    }
+}
+
+static u32 next_minor = 0;
+
+Result<void> ATADevice::create(SharedPtr<ATA::Drive> drive)
+{
+    auto device = TRY(adopt_shared_if_nonnull(new (std::nothrow) ATADevice()));
+    device->m_drive = drive;
+    return DeviceRegistry::register_special_device(DeviceRegistry::Disk, next_minor++, device, "cdrom", 0400);
+}
+
+Result<u64> ATADevice::read(u8* buf, usize offset, usize size) const
+{
+    if (size == 0) return 0;
+
+    if (offset > m_drive->capacity()) return 0;
+    if (offset + size > m_drive->capacity()) size = m_drive->capacity() - offset;
+
+    usize length = size;
+
+    auto block_size = m_drive->block_size();
+
+    auto* temp = TRY(make_array<u8>(block_size));
+    auto guard = make_scope_guard([temp] { delete[] temp; });
+
+    if (offset % block_size)
+    {
+        usize extra_size = block_size - (offset % block_size);
+        TRY(m_drive->read_lba(offset / block_size, temp, 1));
+        memcpy(buf, temp + (offset % block_size), extra_size);
+        offset += extra_size;
+        size -= extra_size;
+        buf += extra_size;
+    }
+
+    while (size >= ARCH_PAGE_SIZE)
+    {
+        TRY(m_drive->read_lba(offset / block_size, buf, ARCH_PAGE_SIZE / block_size));
+        offset += ARCH_PAGE_SIZE;
+        size -= ARCH_PAGE_SIZE;
+        buf += ARCH_PAGE_SIZE;
+    }
+
+    while (size >= block_size)
+    {
+        TRY(m_drive->read_lba(offset / block_size, buf, 1));
+        offset += block_size;
+        size -= block_size;
+        buf += block_size;
+    }
+
+    if (size)
+    {
+        TRY(m_drive->read_lba(offset / block_size, temp, 1));
+        memcpy(buf, temp, size);
+    }
+
+    return length;
+}

kernel/src/arch/x86_64/disk/ATA.h

@@ -0,0 +1,309 @@
+#pragma once
+#include "arch/PCI.h"
+#include "fs/devices/DeviceRegistry.h"
+#include "lib/KMutex.h"
+#include <luna/Atomic.h>
+#include <luna/SharedPtr.h>
+#include <luna/StaticString.h>
+
+namespace ATA
+{
+    enum class Register : u16
+    {
+        Data = 0,
+        Error = 1,
+        Features = 1,
+        SectorCount = 2,
+        SectorNumber = 3,
+        LBALow = 3,
+        CylinderLow = 4,
+        LBAMiddle = 4,
+        CylinderHigh = 5,
+        LBAHigh = 5,
+        DriveSelect = 6,
+        Status = 7,
+        Command = 7,
+    };
+
+    enum class ControlRegister : u16
+    {
+        AltStatus = 0,
+        DeviceControl = 0,
+        DriveAddress = 1,
+    };
+
+    enum class BusmasterRegister : u16
+    {
+        Command = 0,
+        Status = 2,
+        PRDTAddress = 4,
+    };
+
+    enum StatusRegister : u8
+    {
+        SR_Busy = 0x80,
+        SR_DriveReady = 0x40,
+        SR_WriteFault = 0x20,
+        SR_SeekComplete = 0x10,
+        SR_DataRequestReady = 0x08,
+        SR_CorrectedData = 0x04,
+        SR_Index = 0x02,
+        SR_Error = 0x01
+    };
+
+    enum CommandRegister : u8
+    {
+        CMD_Identify = 0xec,
+        CMD_Packet = 0xa0,
+        CMD_Identify_Packet = 0xa1
+    };
+
+    enum BusMasterStatus : u8
+    {
+        BMS_SimplexOnly = 0x80,
+        BMS_SlaveInit = 0x40,
+        BMS_MasterInit = 0x20,
+        BMS_IRQPending = 0x4,
+        BMS_DMAFailure = 0x2,
+        BMS_DMAMode = 0x1
+    };
+
+    enum BusMasterCommand : u8
+    {
+        BMC_ReadWrite = 0x8,
+        BMC_StartStop = 0x1,
+    };
+
+    struct ATAIdentify
+    {
+        u16 flags;
+        u16 unused1[9];
+        char serial[20];
+        u16 unused2[3];
+        char firmware[8];
+        char model[40];
+        u16 sectors_per_int;
+        u16 unused3;
+        u16 capabilities[2];
+        u16 unused4[2];
+        u16 valid_ext_data;
+        u16 unused5[5];
+        u16 size_of_rw_mult;
+        u32 sectors_28;
+        u16 unused6[21];
+        u16 unused7 : 10;
+        u16 big_lba : 1;
+        u16 unused8 : 5;
+        u16 unused9[17];
+        u64 sectors_48;
+        u16 unused10[152];
+    };
+
+    enum ATAPICommand : u8
+    {
+        ATAPI_ReadCapacity = 0x25,
+        ATAPI_Read = 0xa8,
+    };
+
+    class Controller;
+    class Channel;
+
+    struct prdt_entry
+    {
+        u32 address;
+        u16 count;
+        u16 flags;
+    };
+
+    struct atapi_packet
+    {
+        union {
+            u16 command_words[6];
+            u8 command_bytes[12];
+        };
+    };
+
+    struct atapi_read_capacity_reply
+    {
+        u32 last_lba;
+        u32 sector_size;
+    };
+
+    static constexpr u16 END_OF_PRDT = (1 << 15);
+
+    class Drive
+    {
+      public:
+        Drive(Channel* channel, u8 drive_index, Badge<Channel>);
+
+        bool initialize();
+
+        bool post_initialize();
+
+        void irq_handler();
+
+        usize block_size() const
+        {
+            return m_block_size;
+        }
+
+        usize block_count() const
+        {
+            return m_block_count;
+        }
+
+        usize capacity() const
+        {
+            return m_block_count * m_block_size;
+        }
+
+        Result<void> read_lba(u64 lba, void* out, usize nblocks);
+
+      private:
+        bool identify_ata();
+
+        Result<void> send_packet_atapi_pio(const atapi_packet* packet, void* out, u16 response_size);
+#if 0
+        Result<void> send_packet_atapi_dma(const atapi_packet* packet, void* out, u16 response_size);
+
+        Result<void> do_dma_command(u8 command, u16 count, bool write);
+        Result<void> do_dma_transfer();
+#endif
+
+        Result<void> atapi_read_pio(u64 lba, void* out, usize size);
+
+        Channel* m_channel;
+
+        u8 m_drive_index;
+        union {
+            u16 m_identify_words[256];
+            ATAIdentify m_identify_data;
+        };
+
+        bool m_is_atapi { false };
+        bool m_uses_dma { true };
+
+        bool m_is_lba48;
+        u64 m_block_count;
+        u64 m_block_size;
+
+        prdt_entry* m_dma_prdt;
+        u64 m_dma_prdt_phys;
+        volatile void* m_dma_mem;
+        u64 m_dma_mem_phys;
+
+        constexpr static usize SERIAL_LEN = 20;
+        constexpr static usize REVISION_LEN = 8;
+        constexpr static usize MODEL_LEN = 40;
+
+        StaticString<SERIAL_LEN> m_serial;
+        StaticString<REVISION_LEN> m_revision;
+        StaticString<MODEL_LEN> m_model;
+    };
+
+    class Channel
+    {
+      public:
+        Channel(Controller* controller, u8 channel_index, Badge<Controller>);
+
+        u8 read_register(Register reg);
+        u16 read_data();
+        void write_data(u16 value);
+        void write_register(Register reg, u8 value);
+        u8 read_control(ControlRegister reg);
+        void write_control(ControlRegister reg, u8 value);
+
+        u8 read_bm(BusmasterRegister reg);
+        void write_bm(BusmasterRegister reg, u8 value);
+        u32 read_prdt_address();
+        void write_prdt_address(u32 value);
+
+        bool wait_for_reg_set(Register reg, u8 value, u64 timeout);
+        bool wait_for_reg_clear(Register reg, u8 value, u64 timeout);
+
+        Result<void> wait_until_ready();
+
+        void delay_400ns();
+
+        void prepare_for_irq();
+
+        void wait_for_irq();
+        bool wait_for_irq_or_timeout(u64 timeout);
+        void irq_handler(Registers*);
+
+        void select(u8 drive);
+
+        bool initialize();
+
+      private:
+        Controller* m_controller;
+        u8 m_channel_index;
+        bool m_is_pci_native_mode;
+
+        u8 m_interrupt_line;
+
+        KMutex<100> m_lock {};
+
+        Thread* m_thread { nullptr };
+
+        u16 m_io_base;
+        u16 m_control_base;
+        u16 m_busmaster_base;
+
+        bool m_irq_called { false };
+
+        u8 m_current_drive = (u8)-1;
+
+        SharedPtr<Drive> m_drives[2];
+    };
+
+    class Controller
+    {
+      public:
+        static Result<void> scan();
+
+        const PCI::Device& device() const
+        {
+            return m_device;
+        }
+
+        bool initialize();
+
+      private:
+        Controller(const PCI::Device& device);
+        PCI::Device m_device;
+        Channel m_primary_channel;
+        Channel m_secondary_channel;
+    };
+
+}
+
+class ATADevice : public Device
+{
+  public:
+    // Initializer for DeviceRegistry.
+    static Result<void> create(SharedPtr<ATA::Drive> drive);
+
+    Result<usize> read(u8*, usize, usize) const override;
+
+    Result<usize> write(const u8*, usize, usize) override
+    {
+        return err(ENOTSUP);
+    }
+
+    bool blocking() const override
+    {
+        return false;
+    }
+
+    usize size() const override
+    {
+        return m_drive->capacity();
+    }
+
+    virtual ~ATADevice() = default;
+
+  private:
+    ATADevice() = default;
+    SharedPtr<ATA::Drive> m_drive;
+};

kernel/src/arch/x86_64/init/IDT.cpp

@@ -84,6 +84,20 @@ INT(20);
 INT(21);
 INT(32);
 INT(33);
+INT(34);
+INT(35);
+INT(36);
+INT(37);
+INT(38);
+INT(39);
+INT(40);
+INT(41);
+INT(42);
+INT(43);
+INT(44);
+INT(45);
+INT(46);
+INT(47);
 INT(66);
 
 void setup_idt()
@@ -112,6 +126,20 @@ void setup_idt()
     TRAP(21);
     IRQ(32);
     IRQ(33);
+    IRQ(34);
+    IRQ(35);
+    IRQ(36);
+    IRQ(37);
+    IRQ(38);
+    IRQ(39);
+    IRQ(40);
+    IRQ(41);
+    IRQ(42);
+    IRQ(43);
+    IRQ(44);
+    IRQ(45);
+    IRQ(46);
+    IRQ(47);
     SYS(66);
 
     static IDTR idtr;

kernel/src/arch/x86_64/init/PIC.cpp

@@ -36,9 +36,18 @@ void remap_pic()
     IO::outb(PIC2_DATA, ICW4_8086);
     io_delay();
 
-    IO::outb(PIC1_DATA, 0b11111100);
+    IO::outb(PIC1_DATA, 0b11111111);
     io_delay();
     IO::outb(PIC2_DATA, 0b11111111);
+    io_delay();
+}
+
+void change_pic_masks(u8 pic1_mask, u8 pic2_mask)
+{
+    IO::outb(PIC1_DATA, pic1_mask);
+    io_delay();
+    IO::outb(PIC2_DATA, pic2_mask);
+    io_delay();
 }
 
 void pic_eoi(unsigned char irq)

kernel/src/fs/devices/DeviceRegistry.h

@@ -12,6 +12,7 @@ namespace DeviceRegistry
         Console = 1,
         Memory = 2,
         Framebuffer = 3,
+        Disk = 4,
     };
 
     Result<SharedPtr<Device>> fetch_special_device(u32 major, u32 minor);

kernel/src/lib/KMutex.h

@@ -0,0 +1,77 @@
+#pragma once
+#include "Log.h"
+#include "arch/CPU.h"
+#include "thread/Scheduler.h"
+#include "thread/Thread.h"
+#include <luna/CircularQueue.h>
+
+template <usize ConcurrentThreads> class KMutex
+{
+  public:
+    void lock()
+    {
+        int expected = 0;
+        while (!m_lock.compare_exchange_strong(expected, 1))
+        {
+            expected = 0;
+            auto* current = Scheduler::current();
+
+            // We cannot be interrupted between these functions, otherwise we might never exit the loop
+            CPU::disable_interrupts();
+            bool ok = m_blocked_threads.try_push(current);
+            if (!ok) kernel_sleep(10);
+            else
+                kernel_wait_for_event();
+            CPU::enable_interrupts();
+        }
+    };
+
+    void unlock()
+    {
+        int expected = 1;
+        if (!m_lock.compare_exchange_strong(expected, 0))
+        {
+            kwarnln("KMutex::unlock() called on an unlocked lock with value %d", expected);
+        }
+
+        Thread* blocked;
+        if (m_blocked_threads.try_pop(blocked)) blocked->wake_up();
+    }
+
+    bool try_lock()
+    {
+        int expected = 0;
+        return m_lock.compare_exchange_strong(expected, 1);
+    }
+
+  private:
+    CircularQueue<Thread*, ConcurrentThreads> m_blocked_threads;
+    Atomic<int> m_lock;
+};
+
+template <usize ConcurrentThreads> class ScopedKMutexLock
+{
+  public:
+    ScopedKMutexLock(KMutex<ConcurrentThreads>& lock) : m_lock(lock)
+    {
+        m_lock.lock();
+    }
+
+    ~ScopedKMutexLock()
+    {
+        if (!m_taken_over) m_lock.unlock();
+    }
+
+    ScopedKMutexLock(const ScopedKMutexLock&) = delete;
+    ScopedKMutexLock(ScopedKMutexLock&&) = delete;
+
+    KMutex<ConcurrentThreads>& take_over()
+    {
+        m_taken_over = true;
+        return m_lock;
+    }
+
+  private:
+    KMutex<ConcurrentThreads>& m_lock;
+    bool m_taken_over { false };
+};

kernel/src/main.cpp

@@ -1,6 +1,5 @@
 #include "Log.h"
 #include "arch/CPU.h"
-#include "arch/PCI.h"
 #include "arch/Timer.h"
 #include "boot/Init.h"
 #include "config.h"
@@ -11,6 +10,10 @@
 #include "thread/Scheduler.h"
 #include <luna/Units.h>
 
+#ifdef ARCH_X86_64
+#include "arch/x86_64/disk/ATA.h"
+#endif
+
 extern void set_host_name(StringView);
 
 void reap_thread()
@@ -53,12 +56,9 @@ Result<void> init()
     auto reap = Scheduler::new_kernel_thread(reap_thread, "[reap]").release_value();
     Scheduler::set_reap_thread(reap);
 
-    PCI::scan(
+#ifdef ARCH_X86_64
-        [](const PCI::Device& device) {
+    ATA::Controller::scan();
-            kinfoln("Found PCI mass storage device %.4x:%.4x, at address %u:%u:%u", device.id.vendor, device.id.device,
+#endif
-                    device.address.bus, device.address.slot, device.address.function);
-        },
-        { .klass = 1 });
 
     // Disable console logging before transferring control to userspace.
     setup_log(log_debug_enabled(), log_serial_enabled(), false);

kernel/src/sys/mknod.cpp

@@ -1,36 +0,0 @@
-#include "Log.h"
-#include "fs/VFS.h"
-#include "memory/MemoryManager.h"
-#include "sys/Syscall.h"
-#include "thread/Scheduler.h"
-#include <bits/makedev.h>
-#include <luna/PathParser.h>
-
-Result<u64> sys_mknod(Registers*, SyscallArgs args)
-{
-    auto path = TRY(MemoryManager::strdup_from_user(args[0]));
-    mode_t mode = (mode_t)args[1];
-    dev_t dev = (dev_t)args[2];
-
-    Thread* current = Scheduler::current();
-
-    u32 maj = luna_dev_major(dev);
-    u32 min = luna_dev_minor(dev);
-
-    auto parser = TRY(PathParser::create(path.chars()));
-
-    auto dirname = TRY(parser.dirname());
-    auto basename = TRY(parser.basename());
-
-    TRY(VFS::validate_filename(basename.view()));
-
-    auto parent = TRY(VFS::resolve_path(dirname.chars(), current->auth, current->current_directory));
-    if (!VFS::can_write(parent, current->auth)) return err(EACCES);
-
-    auto inode = TRY(parent->fs()->create_device_inode(maj, min));
-    TRY(parent->add_entry(inode, basename.chars()));
-
-    TRY(inode->chmod(mode));
-
-    return 0;
-}

kernel/src/sys/pstat.cpp

@@ -0,0 +1,41 @@
+#include "memory/MemoryManager.h"
+#include "sys/Syscall.h"
+#include "thread/Scheduler.h"
+#include <bits/pstat.h>
+
+static void set_timespec(struct timespec& ts, u64 ticks)
+{
+    ts.tv_sec = ticks / 1000;
+    ts.tv_nsec = (ticks % 1000) * 1000 * 1000;
+}
+
+Result<u64> sys_pstat(Registers*, SyscallArgs args)
+{
+    pid_t pid = (pid_t)args[0];
+    struct process* ps = (struct process*)args[1];
+
+    // If pid == -1, return the PID of the last spawned thread.
+    if (pid == -1) return g_threads.expect_last()->id;
+
+    auto* thread = TRY(Result<Thread*>::from_option(Scheduler::find_by_pid(pid), ESRCH));
+
+    struct process proc;
+    proc.ps_pid = (pid_t)thread->id;
+    proc.ps_ppid = thread->parent ? (pid_t)thread->parent->id : 0;
+    proc.ps_uid = thread->auth.uid;
+    proc.ps_gid = thread->auth.gid;
+    proc.ps_euid = thread->auth.euid;
+    proc.ps_egid = thread->auth.egid;
+    proc.ps_state = (int)thread->state;
+    proc.ps_flags = thread->is_kernel ? PS_FLAG_KRNL : 0;
+    set_timespec(proc.ps_time, thread->ticks);
+    set_timespec(proc.ps_ktime, thread->ticks_in_kernel);
+    set_timespec(proc.ps_utime, thread->ticks_in_user);
+    strlcpy(proc.ps_name, thread->name.chars(), sizeof(proc.ps_name));
+    strlcpy(proc.ps_cwd, thread->current_directory_path.is_empty() ? "/" : thread->current_directory_path.chars(),
+            sizeof(proc.ps_cwd));
+
+    if (!MemoryManager::copy_to_user_typed(ps, &proc)) return err(EFAULT);
+
+    return (u64)pid;
+}

libc/CMakeLists.txt

@@ -25,6 +25,7 @@ set(SOURCES
     src/sys/ioctl.cpp
     src/sys/utsname.cpp
     src/sys/mount.cpp
+    src/sys/pstat.cpp
 )
 
 if(${LUNA_ARCH} STREQUAL "x86_64")

libc/include/bits/pstat.h

@@ -0,0 +1,34 @@
+/* bits/pstat.h: The process structure and flags for the pstat() system call. */
+
+#ifndef _BITS_PSTAT_H
+#define _BITS_PSTAT_H
+
+#include <bits/timespec.h>
+#include <sys/types.h>
+
+#define PS_IDLE 1
+#define PS_RUNNABLE 2
+#define PS_SLEEPING 3
+#define PS_WAITING 4
+#define PS_ENDED 5
+
+#define PS_FLAG_KRNL 1
+
+struct process
+{
+    pid_t ps_pid;
+    pid_t ps_ppid;
+    uid_t ps_uid;
+    uid_t ps_euid;
+    gid_t ps_gid;
+    gid_t ps_egid;
+    int ps_state;
+    int ps_flags;
+    struct timespec ps_time;
+    struct timespec ps_ktime;
+    struct timespec ps_utime;
+    char ps_name[129];
+    char ps_cwd[256];
+};
+
+#endif

libc/include/sys/pstat.h

@@ -0,0 +1,23 @@
+/* sys/pstat.h: The pstat() system call. */
+
+#ifndef _SYS_PSTAT_H
+#define _SYS_PSTAT_H
+
+#include <bits/pstat.h>
+
+#ifdef __cplusplus
+extern "C"
+{
+#endif
+
+    /* Retrieve information about a process. */
+    pid_t pstat(pid_t pid, struct process* ps);
+
+    /* Get a string representing a process' state. */
+    const char* ps_state_name(int state);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

libc/include/sys/stat.h

@@ -15,9 +15,6 @@ extern "C"
     /* Create a directory. */
     int mkdir(const char* path, mode_t mode);
 
-    /* Create a special file. */
-    int mknod(const char* path, mode_t mode, dev_t dev);
-
     /* Change the mode bits of a file. */
     int chmod(const char* path, mode_t mode);
 

libc/src/grp.cpp

@@ -1,4 +1,5 @@
 #include <errno.h>
+#include <fcntl.h>
 #include <grp.h>
 #include <luna/Vector.h>
 #include <stdio.h>
@@ -20,6 +21,7 @@ extern "C"
         {
             f = fopen("/etc/group", "r");
             if (!f) return nullptr;
+            fcntl(fileno(f), F_SETFD, FD_CLOEXEC);
         }
 
         while (true)
@@ -82,14 +84,9 @@ extern "C"
 
         while ((entry = getgrent()))
         {
-            if (!strcmp(entry->gr_name, name))
+            if (!strcmp(entry->gr_name, name)) { return entry; }
-            {
-                endgrent();
-                return entry;
-            }
         }
 
-        endgrent();
         return entry;
     }
 
@@ -101,14 +98,9 @@ extern "C"
 
         while ((entry = getgrent()))
         {
-            if (entry->gr_gid == gid)
+            if (entry->gr_gid == gid) { return entry; }
-            {
-                endgrent();
-                return entry;
-            }
         }
 
-        endgrent();
         return entry;
     }
 

libc/src/pwd.cpp

@@ -1,3 +1,4 @@
+#include <fcntl.h>
 #include <pwd.h>
 #include <stdio.h>
 #include <stdlib.h>
@@ -45,6 +46,7 @@ extern "C"
         {
             f = fopen("/etc/passwd", "r");
             if (!f) return nullptr;
+            fcntl(fileno(f), F_SETFD, FD_CLOEXEC);
         }
 
         while (true)
@@ -93,14 +95,9 @@ extern "C"
 
         while ((entry = getpwent()))
         {
-            if (!strcmp(entry->pw_name, name))
+            if (!strcmp(entry->pw_name, name)) { return entry; }
-            {
-                endpwent();
-                return entry;
-            }
         }
 
-        endpwent();
         return entry;
     }
 
@@ -112,14 +109,9 @@ extern "C"
 
         while ((entry = getpwent()))
         {
-            if (entry->pw_uid == uid)
+            if (entry->pw_uid == uid) { return entry; }
-            {
-                endpwent();
-                return entry;
-            }
         }
 
-        endpwent();
         return entry;
     }
 

libc/src/sys/pstat.cpp

@@ -0,0 +1,26 @@
+#include <bits/errno-return.h>
+#include <sys/pstat.h>
+#include <sys/syscall.h>
+#include <unistd.h>
+
+extern "C"
+{
+    pid_t pstat(pid_t pid, struct process* ps)
+    {
+        long rc = syscall(SYS_pstat, pid, ps);
+        __errno_return(rc, pid_t);
+    }
+
+    const char* ps_state_name(int state)
+    {
+        switch (state)
+        {
+        case PS_IDLE: return "Idle";
+        case PS_RUNNABLE: return "Running";
+        case PS_SLEEPING: return "Sleeping";
+        case PS_WAITING: return "Waiting";
+        case PS_ENDED: return "Zombie";
+        default: return "???";
+        }
+    }
+}

libc/src/sys/stat.cpp

@@ -12,12 +12,6 @@ extern "C"
         __errno_return(rc, int);
     }
 
-    int mknod(const char* path, mode_t mode, dev_t dev)
-    {
-        long rc = syscall(SYS_mknod, path, mode, dev);
-        __errno_return(rc, int);
-    }
-
     int chmod(const char* path, mode_t mode)
     {
         long rc = syscall(SYS_fchmodat, AT_FDCWD, path, mode, 0);

libluna/include/luna/StaticString.h

@@ -1,5 +1,6 @@
 #pragma once
 #include <luna/CString.h>
+#include <luna/StringView.h>
 #include <luna/Types.h>
 
 template <usize Size> class StaticString
@@ -24,12 +25,27 @@ template <usize Size> class StaticString
             m_length = length;
     }
 
+    void adopt(StringView string)
+    {
+        usize length = strlcpy(m_buffer, string.chars(),
+                               string.length() > sizeof(m_buffer) ? sizeof(m_buffer) : string.length() + 1);
+        if (length > Size) { m_length = Size; }
+        else
+            m_length = length;
+    }
+
     StaticString<Size>& operator=(const char* string)
     {
         adopt(string);
         return *this;
     }
 
+    StaticString<Size>& operator=(StringView string)
+    {
+        adopt(string);
+        return *this;
+    }
+
     template <usize OtherSize> StaticString<Size>& operator=(const StaticString<OtherSize>& string)
     {
         if constexpr (OtherSize == Size)
@@ -46,11 +62,38 @@ template <usize Size> class StaticString
         return m_buffer;
     }
 
+    char* data()
+    {
+        return m_buffer;
+    }
+
+    void set_length(usize len)
+    {
+        m_length = len;
+    }
+
     usize length() const
     {
         return m_length;
     }
 
+    void trim(StringView delim)
+    {
+        isize i = (isize)m_length;
+
+        while (i--)
+        {
+            char c = m_buffer[i];
+            if (!strchr(delim.chars(), c)) break;
+        }
+
+        i++;
+
+        m_buffer[i] = '\0';
+
+        m_length = (usize)i;
+    }
+
   private:
     char m_buffer[Size + 1];
     usize m_length { 0 };

libluna/include/luna/Syscall.h

@@ -2,10 +2,10 @@
 
 #define enumerate_syscalls(_e)                                                                                         \
     _e(exit) _e(clock_gettime) _e(mmap) _e(munmap) _e(usleep) _e(openat) _e(close) _e(read) _e(getpid) _e(write)       \
-        _e(lseek) _e(mkdir) _e(execve) _e(mknod) _e(fork) _e(waitpid) _e(getppid) _e(fcntl) _e(getdents) _e(getuid)    \
+        _e(lseek) _e(mkdir) _e(execve) _e(fork) _e(waitpid) _e(getppid) _e(fcntl) _e(getdents) _e(getuid) _e(geteuid)  \
-            _e(geteuid) _e(getgid) _e(getegid) _e(setuid) _e(setgid) _e(seteuid) _e(setegid) _e(fchmodat) _e(fchownat) \
+            _e(getgid) _e(getegid) _e(setuid) _e(setgid) _e(seteuid) _e(setegid) _e(fchmodat) _e(fchownat) _e(ioctl)   \
-                _e(ioctl) _e(fstatat) _e(chdir) _e(getcwd) _e(unlinkat) _e(uname) _e(sethostname) _e(dup2) _e(pipe)    \
+                _e(fstatat) _e(chdir) _e(getcwd) _e(unlinkat) _e(uname) _e(sethostname) _e(dup2) _e(pipe) _e(mount)    \
-                    _e(mount) _e(umount)
+                    _e(umount) _e(pstat)
 
 enum Syscalls
 {