Moving the rest (except ACPI tables) to high mem

Also the debug messaging to verify it.

NOTES.md

@@ -11,6 +11,9 @@
 - Serial out based on circular/bip biffer and interrupts, not spinning on
   `write_ready()`
 - Split out more code into kutil for testing
+- AHCI / MSI interrupts on Vbox break?
+- FXSAVE to save XMM registers.
+  - optimization using #NM (0x7) to detect SSE usage
 
 
 - Device Tree
@@ -21,4 +24,3 @@
   - Multiprocessing
     - Processes in Ring 3
 - Stack tracer
-- build system upgrade (CMake / Waf / etc)

scripts/vmem_translate.py

@@ -0,0 +1,12 @@
+#!/usr/bin/env python
+
+def translate(i4 = 0, i3 = 0, i2 = 0, i1 = 0, offset = 0):
+    addr = (i4 << 39) + (i3 << 30) + (i2 << 21) + (i1 << 12) + offset
+    if addr & (1 << 47):
+        addr |= 0xffff000000000000
+    return addr
+
+if __name__ == "__main__":
+    import sys
+    print("{:016x}".format(translate(*map(int, sys.argv[1:]))))
+

src/arch/x86_64/kernel.ld

@@ -1,7 +1,7 @@
 ENTRY(_start)
 SECTIONS
 {
-	OFFSET = 0xFFFF800000000000;
+	OFFSET = 0xFFFFFF0000000000;
 	. = OFFSET + 0x100000;
 
 	.header : {

src/boot/loader.c

@@ -215,7 +215,7 @@ loader_load_kernel(
 		if (status == EFI_NOT_FOUND)
 			continue;
 
-		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_file: %s", kernel_name);
+		CHECK_EFI_STATUS_OR_RETURN(status, L"loader_load_elf: %s", kernel_name);
 
 		data->font = (void *)((uint64_t)data->kernel + data->kernel_length);
 		status = loader_load_font(bootsvc, root, data);

src/boot/loader.h

@@ -9,11 +9,7 @@
 #endif
 
 #ifndef KERNEL_VIRT_ADDRESS
-#define KERNEL_VIRT_ADDRESS 0xFFFF800000000000
-#endif
-
-#ifndef VIRTUAL_OFFSET
-#define VIRTUAL_OFFSET 0xf00000000
+#define KERNEL_VIRT_ADDRESS 0xFFFFFF0000000000
 #endif
 
 #ifndef KERNEL_MEMTYPE

src/boot/main.c

@@ -82,9 +82,9 @@ efi_main(EFI_HANDLE image_handle, EFI_SYSTEM_TABLE *system_table)
 	status = loader_load_kernel(bootsvc, &load);
 	CHECK_EFI_STATUS_OR_FAIL(status);
 
-	con_printf(L"    %u image bytes at 0x%x\r\n", load.kernel_length, load.kernel);
-	con_printf(L"    %u font bytes at 0x%x\r\n", load.font_length, load.font);
-	con_printf(L"    %u data bytes at 0x%x\r\n", load.data_length, load.data);
+	con_printf(L"    image bytes at 0x%x : %x\r\n", load.kernel, load.kernel_length);
+	con_printf(L"     font bytes at 0x%x : %x\r\n", load.font, load.font_length);
+	con_printf(L"     data bytes at 0x%x : %x\r\n", load.data, load.data_length);
 
 	struct kernel_header *version = (struct kernel_header *)load.kernel;
 	if (version->magic != KERNEL_HEADER_MAGIC) {

src/boot/memory.c

@@ -191,6 +191,7 @@ memory_virtualize(EFI_RUNTIME_SERVICES *runsvc, struct memory_map *map)
 		case KERNEL_FONT_MEMTYPE:
 		case KERNEL_DATA_MEMTYPE:
 			d->Attribute |= EFI_MEMORY_RUNTIME;
+			d->VirtualStart = d->PhysicalStart + KERNEL_VIRT_ADDRESS;
 
 		default:
 			if (d->Attribute & EFI_MEMORY_RUNTIME) {

src/kernel/ahci/driver.cpp

@@ -16,13 +16,3 @@ ahci_driver::register_device(pci_device *device)
 
 	ahci::hba &hba = m_devices.emplace(device);
 }
-
-ahci::port *
-ahci_driver::find_disk()
-{
-	for (auto &hba : m_devices) {
-		ahci::port *d = hba.find_disk();
-		if (d) return d;
-	}
-	return nullptr;
-}

src/kernel/ahci/driver.h

@@ -22,9 +22,6 @@ public:
 	/// \arg device  The PCI device to remove
 	void unregister_device(pci_device *device);
 
-	/// Debug: find the first disk
-	ahci::port * find_disk();
-
 private:
 	kutil::vector<ahci::hba> m_devices;
 };

src/kernel/ahci/hba.cpp

@@ -1,7 +1,9 @@
 #include <stdint.h>
 #include "ahci/ata.h"
 #include "ahci/hba.h"
+#include "console.h"
 #include "device_manager.h"
+#include "fs/gpt.h"
 #include "log.h"
 #include "page_manager.h"
 #include "pci.h"
@@ -57,10 +59,14 @@ void irq_cb(void *data)
 hba::hba(pci_device *device)
 {
 	page_manager *pm = page_manager::get();
+	device_manager &dm = device_manager::get();
 
 	uint32_t bar5 = device->get_bar(5);
-	m_data = reinterpret_cast<hba_data *>(bar5 & ~0xfffull);
-	pm->map_offset_pointer(reinterpret_cast<void **>(&m_data), 0x2000);
+	log::debug(logs::driver, "HBA raw BAR5 is %08lx", bar5);
+
+	void *data = reinterpret_cast<void *>(bar5 & ~0xfffull);
+	pm->map_offset_pointer(&data, 0x2000);
+	m_data = reinterpret_cast<hba_data volatile *>(data);
 
 	if (! bitfield_has(m_data->cap, hba_cap::ahci_only))
 		m_data->host_control |= 0x80000000; // Enable AHCI mode
@@ -70,17 +76,17 @@ hba::hba(pci_device *device)
 	unsigned ports = (icap & 0xf) + 1;
 	unsigned slots = ((icap >> 8) & 0x1f) + 1;
 
-	log::debug(logs::driver, "  %d ports", ports);
+	log::debug(logs::driver, "  %d ports: %08x", ports, m_data->port_impl);
 	log::debug(logs::driver, "  %d command slots", slots);
 
-	port_data *pd = reinterpret_cast<port_data *>(
+	auto *pd = reinterpret_cast<port_data volatile *>(
 			kutil::offset_pointer(m_data, 0x100));
 
 	bool needs_interrupt = false;
 	m_ports.ensure_capacity(ports);
 	for (unsigned i = 0; i < ports; ++i) {
 		bool impl = ((m_data->port_impl & (1 << i)) != 0);
-		port &p = m_ports.emplace(i, kutil::offset_pointer(pd, 0x80 * i), impl);
+		port &p = m_ports.emplace(this, i, kutil::offset_pointer(pd, 0x80 * i), impl);
 		if (p.get_state() == port::state::active)
 			needs_interrupt = true;
 	}
@@ -89,29 +95,48 @@ hba::hba(pci_device *device)
 		device_manager::get().allocate_msi("AHCI Device", *device, irq_cb, this);
 		m_data->host_control |= 0x02; // enable interrupts
 	}
-}
 
-port *
-hba::find_disk()
-{
-	for (auto &port : m_ports) {
-		if (port.get_state() == port::state::active &&
-			port.get_type() == sata_signature::sata_drive)
-			return &port;
+	for (auto &p : m_ports) {
+		if (!p.active()) continue;
+
+		if (p.get_type() == sata_signature::sata_drive) {
+			p.sata_reconnect();
+			/*
+			if (fs::partition::load(&p) == 0)
+				dm.register_block_device(&p);
+			*/
+		}
 	}
-
-	return nullptr;
 }
 
 void
 hba::handle_interrupt()
 {
+	uint32_t status = m_data->int_status;
 	for (auto &port : m_ports) {
-		if (m_data->int_status & (1 << port.index())) {
+		if (status & (1 << port.index())) {
 			port.handle_interrupt();
 		}
 	}
-	m_data->int_status = 0;
+	// Write 1 to the handled interrupts
+	m_data->int_status = status;
+}
+
+void
+hba::dump()
+{
+	console *cons = console::get();
+	static const char *regs[] = {
+		" CAP", " GHC", "  IS", "  PI", "  VS", " C3C",
+		" C3P", " EML", " EMC", "CAP2", "BOHC"
+	};
+
+	cons->printf("HBA Registers:\n");
+	auto *data = reinterpret_cast<uint32_t volatile *>(m_data);
+	for (int i = 0; i < 11; ++i) {
+		cons->printf("  %s: %08x\n", regs[i], data[i]);
+	}
+	cons->putc('\n');
 }
 
 } // namespace ahci

src/kernel/ahci/hba.h

@@ -25,12 +25,12 @@ public:
 	/// Interrupt handler.
 	void handle_interrupt();
 
-	/// Debug: find the first disk
-	port * find_disk();
+	/// Dump the HBA registers to the console
+	void dump();
 
 private:
 	pci_device *m_device;
-	hba_data *m_data;
+	hba_data volatile *m_data;
 	kutil::vector<port> m_ports;
 };
 

src/kernel/ahci/port.cpp

@@ -2,13 +2,21 @@
 #include "kutil/assert.h"
 #include "kutil/enum_bitfields.h"
 #include "ahci/ata.h"
+#include "ahci/hba.h"
 #include "ahci/fis.h"
 #include "ahci/port.h"
+#include "console.h"
 #include "io.h"
 #include "log.h"
 #include "page_manager.h"
 
+namespace ahci {
+	enum class cmd_list_flags : uint16_t;
+}
+
 IS_BITFIELD(ahci::port_cmd);
+IS_BITFIELD(volatile ahci::port_cmd);
+IS_BITFIELD(ahci::cmd_list_flags);
 
 namespace ahci {
 
@@ -110,10 +118,11 @@ struct port_data
 } __attribute__ ((packed));
 
 
-port::port(uint8_t index, port_data *data, bool impl) :
+port::port(hba *device, uint8_t index, port_data volatile *data, bool impl) :
 	m_index(index),
 	m_type(sata_signature::none),
 	m_state(state::unimpl),
+	m_hba(device),
 	m_data(data),
 	m_fis(nullptr),
 	m_cmd_list(nullptr),
@@ -154,7 +163,13 @@ port::update()
 
 		rebase();
 		m_pending.set_size(32);
-		m_data->interrupt_enable = 1;
+		for (auto &pend : m_pending) {
+			pend.type = command_type::none;
+		}
+
+		// Clear any old pending interrupts and enable interrupts
+		m_data->interrupt_status = m_data->interrupt_status;
+		m_data->interrupt_enable = 0xffffffff;
 	} else {
 		m_state = state::inactive;
 	}
@@ -190,19 +205,27 @@ port::stop_commands()
 }
 
 int
-port::make_command(size_t length)
+port::make_command(size_t length, fis_register_h2d **fis)
 {
 	int slot = -1;
-	uint32_t used_slots = (m_data->serial_active | m_data->cmd_issue);
+	uint32_t used_slots =
+		m_data->serial_active |
+		m_data->cmd_issue |
+		m_data->interrupt_status;
+
 	for (int i = 0; i < 32; ++i) {
-		if ((used_slots & (1 << i)) == 0) {
+		if (used_slots & (1 << i)) continue;
+
+		if (m_pending[i].type == command_type::none) {
 			slot = i;
 			break;
+		} else {
+			log::debug(logs::driver, "Type is %d", m_pending[i].type);
 		}
 	}
 
 	if (slot < 0) {
-		log::info(logs::driver, "AHCI could not get a free command slot.");
+		log::error(logs::driver, "AHCI could not get a free command slot.");
 		return -1;
 	}
 
@@ -214,6 +237,14 @@ port::make_command(size_t length)
 	kutil::memset(&cmdt, 0, sizeof(cmd_table) +
 			max_prd_count * sizeof(prdt_entry));
 
+	ent.flags = cmd_list_fis_size(sizeof(fis_register_h2d));
+
+	fis_register_h2d *cfis = reinterpret_cast<fis_register_h2d *>(&cmdt.cmd_fis);
+	kutil::memset(cfis, 0, sizeof(fis_register_h2d));
+	cfis->type = fis_type::register_h2d;
+	cfis->pm_port = 0x80; // set command register flag
+	*fis = cfis;
+
 	size_t remaining = length;
 	for (int i = 0; i < max_prd_count; ++i) {
 		size_t prd_len = std::min(remaining, 0x200000ul);
@@ -240,21 +271,20 @@ port::make_command(size_t length)
 	return slot;
 }
 
-bool
-port::read(uint64_t sector, size_t length, void *dest)
+int
+port::read_async(uint64_t offset, size_t length, void *dest)
 {
-	int slot = make_command(length);
+	fis_register_h2d *fis;
+	int slot = make_command(length, &fis);
 	if (slot < 0)
-		return false;
+		return 0;
 
 	cmd_table &cmdt = m_cmd_table[slot];
 
-	fis_register_h2d *fis = reinterpret_cast<fis_register_h2d *>(&cmdt.cmd_fis);
-	fis->type = fis_type::register_h2d;
-	fis->pm_port = 0x80; // set command register flag
 	fis->command = ata_cmd::read_dma_ext;
 	fis->device = 0x40; // ATA8-ACS p.175
 
+	uint64_t sector = offset >> 9;
 	fis->lba0 = (sector      ) & 0xff;
 	fis->lba1 = (sector >>  8) & 0xff;
 	fis->lba2 = (sector >> 16) & 0xff;
@@ -270,8 +300,64 @@ port::read(uint64_t sector, size_t length, void *dest)
 			count, sector, sector*512);
 
 	m_pending[slot].type = command_type::read;
+	m_pending[slot].offset = offset % 512;
+	m_pending[slot].count = 0;
 	m_pending[slot].data = dest;
-	return issue_command(slot);
+	if(issue_command(slot))
+		return slot;
+	else
+		return -1;
+}
+
+size_t
+port::read(uint64_t offset, size_t length, void *dest)
+{
+	int slot = read_async(offset, length, dest);
+
+	int timeout = 0;
+	while (m_pending[slot].type == command_type::read) {
+		if (timeout++ > 5) {
+			return 0;
+		}
+		asm("hlt");
+	}
+	kassert(m_pending[slot].type == command_type::finished,
+			"Read got unexpected command type");
+
+	size_t count = m_pending[slot].count;
+	m_pending[slot].type = command_type::none;
+	m_pending[slot].count = 0;
+
+	return count;
+}
+
+int
+port::identify_async()
+{
+	fis_register_h2d *fis;
+	int slot = make_command(512, &fis);
+	if (slot < 0)
+		return 0;
+
+	fis->command = ata_cmd::identify;
+
+	m_pending[slot].type = command_type::identify;
+	m_pending[slot].offset = 0;
+	m_pending[slot].count = 0;
+	m_pending[slot].data = 0;
+
+	if(issue_command(slot))
+		return slot;
+	else
+		return -1;
+}
+
+void
+port::sata_reconnect()
+{
+	m_data->serial_control |= 1;
+	io_wait(1000); // About 1ms
+	m_data->serial_control &= ~1;
 }
 
 bool
@@ -297,14 +383,23 @@ port::issue_command(int slot)
 void
 port::handle_interrupt()
 {
-	log::debug(logs::driver, "AHCI port %d got an interrupt");
+	log::debug(logs::driver, "AHCI port %d got an interrupt", m_index);
 
 	// TODO: handle other states in interrupt_status
 
-	if (m_data->interrupt_status & 0x40000000) {
+	uint32_t is = m_data->interrupt_status;
+
+	if (is & 0x00000040) {
+		// Port connect status change: For now clear the "exchange"
+		// bit in SERR, this should probably kick off diagnostics.
+		m_data->serial_error = 0x04000000;
+		identify_async();
+	}
+
+	if (is & 0x40000000) {
 		log::error(logs::driver, "AHCI task file error");
-		// TODO: clean up!
-		return;
+		dump();
+		kassert(0, "Task file error");
 	}
 
 	log::debug(logs::driver, "AHCI interrupt status: %08lx  %08lx",
@@ -312,19 +407,25 @@ port::handle_interrupt()
 
 	uint32_t ci = m_data->cmd_issue;
 	for (int i = 0; i < 32; ++i) {
+		// Skip commands still listed as "issued"
 		if (ci & (1 << i)) continue;
 
+		// Any commands not still listed as "issued" that are still pending for
+		// the driver are now finished, so handle them.
 		pending &p = m_pending[i];
 		switch (p.type) {
 		case command_type::read:
 			finish_read(i);
 			break;
+		case command_type::identify:
+			finish_identify(i);
+			break;
 		default:
 			break;
 		}
-		p.type = command_type::none;
-		p.data = nullptr;
 	}
+
+	// Clear the whole status register to mark it as handled
 	m_data->interrupt_status = m_data->interrupt_status;
 }
 
@@ -335,7 +436,9 @@ port::finish_read(int slot)
 	cmd_table &cmdt = m_cmd_table[slot];
 	cmd_list_entry &ent = m_cmd_list[slot];
 
+	size_t count = 0;
 	void *p = m_pending[slot].data;
+	uint8_t offset = m_pending[slot].offset;
 	for (int i = 0; i < ent.prd_table_length; ++i) {
 		size_t prd_len = (cmdt.entries[i].byte_count & 0x7fffffff) + 1;
 
@@ -343,13 +446,82 @@ port::finish_read(int slot)
 			static_cast<addr_t>(cmdt.entries[i].data_base_low) |
 			static_cast<addr_t>(cmdt.entries[i].data_base_high) << 32;
 
-		void *mem = pm->offset_virt(phys);
+		void *mem = kutil::offset_pointer(pm->offset_virt(phys), offset);
+
+		log::debug(logs::driver, "Reading PRD %2d: %016lx->%016lx [%lxb]", i, mem, p, prd_len);
+
 		kutil::memcpy(p, mem, prd_len);
-		p = kutil::offset_pointer(p, prd_len);
+		p = kutil::offset_pointer(p, prd_len - offset);
+		count += (prd_len - offset);
+		offset = 0;
 
 		pm->unmap_pages(mem, page_count(prd_len));
 	}
 
+	m_pending[slot].count = count;
+	m_pending[slot].type = command_type::finished;
+	m_pending[slot].data = nullptr;
+}
+
+static void
+ident_strcpy(uint16_t *from, int words, char *dest)
+{
+	for (int i = 0; i < words; ++i) {
+		*dest++ = *from >> 8;
+		*dest++ = *from & 0xff;
+		from++;
+	}
+	*dest = 0;
+}
+
+void
+port::finish_identify(int slot)
+{
+	page_manager *pm = page_manager::get();
+	cmd_table &cmdt = m_cmd_table[slot];
+	cmd_list_entry &ent = m_cmd_list[slot];
+
+	kassert(ent.prd_table_length == 1, "AHCI identify used multiple PRDs");
+
+	size_t prd_len = (cmdt.entries[0].byte_count & 0x7fffffff) + 1;
+
+	addr_t phys =
+		static_cast<addr_t>(cmdt.entries[0].data_base_low) |
+		static_cast<addr_t>(cmdt.entries[0].data_base_high) << 32;
+
+	log::debug(logs::driver, "Reading ident PRD:");
+
+	uint16_t *mem = reinterpret_cast<uint16_t *>(pm->offset_virt(phys));
+	char string[41];
+
+	ident_strcpy(&mem[10], 10, &string[0]);
+	log::debug(logs::driver, "    Device serial: %s", string);
+
+	ident_strcpy(&mem[23], 4, &string[0]);
+	log::debug(logs::driver, "   Device version: %s", string);
+
+	ident_strcpy(&mem[27], 20, &string[0]);
+	log::debug(logs::driver, "     Device model: %s", string);
+
+	uint32_t sectors = mem[60] | (mem[61] << 16);
+	log::debug(logs::driver, "      Max sectors: %xh", sectors);
+
+	uint16_t lb_size = mem[106];
+	log::debug(logs::driver, " lsects per psect: %d %s %s", 1 << (lb_size & 0xf),
+			lb_size & 0x20 ? "multiple logical per physical" : "",
+			lb_size & 0x10 ? "physical > 512b" : "");
+
+	uint32_t b_per_ls = 2 * (mem[117] | (mem[118] << 16));
+	log::debug(logs::driver, "      b per lsect: %d", b_per_ls);
+
+	/*
+	for (int i=0; i<256; i += 4)
+		log::debug(logs::driver, "  %3d: %04x  %3d: %04x  %3d: %04x  %3d: %04x",
+				i, mem[i], i+1, mem[i+1], i+2, mem[i+2], i+3, mem[i+3]);
+	*/
+
+	pm->unmap_pages(mem, page_count(prd_len));
+	m_pending[slot].type = command_type::none;
 }
 
 void
@@ -430,4 +602,22 @@ port::rebase()
 	start_commands();
 }
 
+void
+port::dump()
+{
+	console *cons = console::get();
+	static const char *regs[] = {
+		" CLB", "+CLB", "  FB", " +FB", "  IS", "  IE",
+		" CMD", nullptr, " TFD", " SIG", "SSTS", "SCTL", "SERR",
+		"SACT", "  CI", "SNTF", " FBS", "DEVS"
+	};
+
+	cons->printf("Port Registers:\n");
+	auto *data = reinterpret_cast<volatile uint32_t *>(m_data);
+	for (int i = 0; i < 18; ++i) {
+		if (regs[i]) cons->printf("  %s: %08x\n", regs[i], data[i]);
+	}
+	cons->putc('\n');
+}
+
 } // namespace ahci

src/kernel/ahci/port.h

@@ -4,25 +4,30 @@
 #include <stddef.h>
 #include <stdint.h>
 #include "kutil/vector.h"
+#include "block_device.h"
 
 namespace ahci {
 
 struct cmd_list_entry;
 struct cmd_table;
+struct fis_register_h2d;
+class hba;
 enum class sata_signature : uint32_t;
 enum class port_cmd : uint32_t;
 struct port_data;
 
 
 /// A port on an AHCI HBA
-class port
+class port :
+	public block_device
 {
 public:
 	/// Constructor.
-	/// \arg index Index of the port on its HBA
-	/// \arg data  Pointer to the device's registers for this port
-	/// \arg impl  Whether this port is marked as implemented in the HBA
-	port(uint8_t index, port_data *data, bool impl);
+	/// \arg device The HBA device this port belongs to
+	/// \arg index  Index of the port on its HBA
+	/// \arg data   Pointer to the device's registers for this port
+	/// \arg impl   Whether this port is marked as implemented in the HBA
+	port(hba *device, uint8_t index, port_data volatile *data, bool impl);
 
 	/// Destructor
 	~port();
@@ -37,6 +42,10 @@ public:
 	/// \returns  An enum representing the state
 	inline state get_state() const { return m_state; }
 
+	/// Check if this device is active
+	/// \returns  True if the device state is active
+	inline bool active() const { return m_state == state::active; }
+
 	/// Get the type signature of this device
 	/// \returns  An enum representing the type of device
 	inline sata_signature get_type() const { return m_type; }
@@ -53,16 +62,34 @@ public:
 	/// Stop command processing from this port
 	void stop_commands();
 
-	/// Read data from the drive.
-	/// \arg sector  Starting sector to read
+	/// Start a read operation from the drive.
+	/// \arg offset  Offset to start from
 	/// \arg length  Number of bytes to read
 	/// \arg dest    A buffer where the data will be placed
-	/// \returns     True if the command succeeded
-	bool read(uint64_t sector, size_t length, void *dest);
+	/// \returns     A handle to the read operation, or -1 on error
+	int read_async(uint64_t offset, size_t length, void *dest);
+
+	/// Read from the drive, blocking until finished.
+	/// \arg offset  Offset to start from
+	/// \arg length  Number of bytes to read
+	/// \arg dest    A buffer where the data will be placed
+	/// \returns     The number of bytes read
+	virtual size_t read(uint64_t offset, size_t length, void *dest);
+
+	/// Start an identify operation for the drive.
+	/// \returns     A handle to the read operation, or -1 on error
+	int identify_async();
+
+	/// Tell the HBA to reconnect to the SATA device. A successful
+	/// reconnect will kick off an identify command.
+	void sata_reconnect();
 
 	/// Handle an incoming interrupt
 	void handle_interrupt();
 
+	/// Dump the port registers to the console
+	void dump();
+
 private:
 	/// Rebase the port command structures to a new location in system
 	/// memory, to be allocated from the page manager.
@@ -70,8 +97,9 @@ private:
 
 	/// Initialize a command structure
 	/// \arg length  The number of bytes of data needed in the PRDs
+	/// \arg fis     [out] The FIS for this command
 	/// \returns     The index of the command slot, or -1 if none available
-	int make_command(size_t length);
+	int make_command(size_t length, fis_register_h2d **fis);
 
 	/// Send a constructed command to the hardware
 	/// \arg slot   The index of the command slot used
@@ -88,20 +116,29 @@ private:
 	/// \arg slot  The command slot that the read command used
 	void finish_read(int slot);
 
+	/// Finish an identify command started by `identify_async()`.
+	/// This will free the structures allocated, so `free_command()` is
+	/// not necessary.
+	/// \arg slot  The command slot that the read command used
+	void finish_identify(int slot);
+
 	sata_signature m_type;
 	uint8_t m_index;
 	state m_state;
 
-	port_data *m_data;
+	hba *m_hba;
+	port_data volatile *m_data;
 	void *m_fis;
 	cmd_list_entry *m_cmd_list;
 	cmd_table *m_cmd_table;
 
-	enum class command_type : uint8_t { none, read, write };
+	enum class command_type : uint8_t { none, read, write, identify, finished };
 
 	struct pending
 	{
 		command_type type;
+		uint8_t offset;
+		size_t count;
 		void *data;
 	};
 

src/kernel/allocator.cpp

@@ -1,9 +1,18 @@
 #include "kutil/memory_manager.h"
+#include "log.h"
 
 kutil::memory_manager g_kernel_memory_manager;
 
 // kutil malloc/free implementation
 namespace kutil {
-	void * malloc(size_t n) { return g_kernel_memory_manager.allocate(n); }
-	void free(void *p) { g_kernel_memory_manager.free(p); }
+
+void *
+malloc(size_t n) {
+	void *p = g_kernel_memory_manager.allocate(n);
+	log::debug(logs::memory, "Heap allocated %ld bytes: %016lx", n, p);
+	return p;
+}
+
+void free(void *p) { g_kernel_memory_manager.free(p); }
+
 }

src/kernel/apic.cpp

@@ -52,28 +52,38 @@ lapic::lapic(uint32_t *base, isr spurious) :
 void
 lapic::enable_timer(isr vector, uint8_t divisor, uint32_t count, bool repeat)
 {
+	uint32_t divbits = 0;
+
 	switch (divisor) {
-	case   1: divisor = 11; break;
-	case   2: divisor = 0; break;
-	case   4: divisor = 1; break;
-	case   8: divisor = 2; break;
-	case  16: divisor = 3; break;
-	case  32: divisor = 8; break;
-	case  64: divisor = 9; break;
-	case 128: divisor = 10; break;
+	case   1: divbits = 0xb; break;
+	case   2: divbits = 0x0; break;
+	case   4: divbits = 0x1; break;
+	case   8: divbits = 0x2; break;
+	case  16: divbits = 0x3; break;
+	case  32: divbits = 0x8; break;
+	case  64: divbits = 0x9; break;
+	case 128: divbits = 0xa; break;
 	default:
 		kassert(0, "Invalid divisor passed to lapic::enable_timer");
 	}
 
-	apic_write(m_base, 0x3e0, divisor);
-	apic_write(m_base, 0x380, count);
-
 	uint32_t lvte = static_cast<uint8_t>(vector);
 	if (repeat)
 		lvte |= 0x20000;
 
 	log::debug(logs::apic, "Enabling APIC timer with isr %d.", vector);
 	apic_write(m_base, 0x320, lvte);
+	apic_write(m_base, 0x3e0, divbits);
+
+	reset_timer(count);
+}
+
+uint32_t
+lapic::reset_timer(uint32_t count)
+{
+	uint32_t remaining = apic_read(m_base, 0x380);
+	apic_write(m_base, 0x380, count);
+	return remaining;
 }
 
 void

src/kernel/apic.h

@@ -37,6 +37,15 @@ public:
 	/// \arg repeat   If false, this timer is one-off, otherwise repeating
 	void enable_timer(isr vector, uint8_t divisor, uint32_t count, bool repeat = true);
 
+	/// Reset the timer countdown.
+	/// \arg count  The count of ticks before an interrupt, or 0 to stop the timer
+	/// \returns    The count of ticks that were remaining before reset
+	uint32_t reset_timer(uint32_t count);
+
+	/// Stop the timer.
+	/// \returns  The count of ticks remaining before an interrupt was to happen
+	inline uint32_t stop_timer() { return reset_timer(0); }
+
 	/// Enable interrupts for the LAPIC LINT0 pin.
 	/// \arg num      Local interrupt number (0 or 1)
 	/// \arg vector   Interrupt vector LINT0 should use

src/kernel/assert.cpp

@@ -7,7 +7,7 @@ __kernel_assert(const char *file, unsigned line, const char *message)
 	console *cons = console::get();
 	if (cons) {
 		cons->set_color(9 , 0);
-		cons->puts("\n\n  ERROR: ");
+		cons->puts("\n\n       ERROR: ");
 		cons->puts(file);
 		cons->puts(":");
 		cons->put_dec(line);
@@ -15,15 +15,6 @@ __kernel_assert(const char *file, unsigned line, const char *message)
 		cons->puts(message);
 	}
 
-	__asm__ __volatile__( 
-		"movq %0, %%r8;"
-		"movq %1, %%r9;"
-		"movq %2, %%r10;"
-		"movq $0, %%rdx;"
-		"divq %%rdx;"
-		: // no outputs
-		: "r"((uint64_t)line), "r"(file), "r"(message)
-		: "rax", "rdx", "r8", "r9", "r10");
-
-	while (1);
+	__asm__ ( "int $0e7h" );
+	while (1) __asm__ ("hlt");
 }

src/kernel/block_device.h

@@ -0,0 +1,11 @@
+#pragma once
+/// \file block_device.h
+/// Interface definition for block devices
+#include <stddef.h>
+
+/// Interface for block devices
+class block_device
+{
+public:
+	virtual size_t read(size_t offset, size_t length, void *buffer) = 0;
+};

src/kernel/boot.s

@@ -28,8 +28,8 @@ _start:
 	extern kernel_main
 	call kernel_main
 
-	cli
-
+	; Kernel init is over, wait for the scheduler to
+	; take over
 .hang:
 	hlt
 	jmp .hang

src/kernel/console.cpp

@@ -1,4 +1,5 @@
 #include "kutil/coord.h"
+#include "kutil/guid.h"
 #include "kutil/memory.h"
 #include "console.h"
 #include "font.h"
@@ -334,6 +335,23 @@ void console::vprintf(const char *fmt, va_list args)
 					done = true;
 					break;
 
+				case 'G': {
+						// Special: GUID type
+						kutil::guid g = va_arg(args, kutil::guid);
+						put_hex<uint32_t>(g.a, 8, '0');
+						putc('-');
+						put_hex<uint16_t>((g.a >> 32) & 0xffff, 4, '0');
+						putc('-');
+						put_hex<uint16_t>((g.a >> 48) & 0xffff, 4, '0');
+						putc('-');
+						put_hex<uint16_t>((kutil::byteswap(g.b) >> 16) & 0xffff, 4, '0');
+						putc('-');
+						put_hex<uint16_t>(kutil::byteswap(g.b) & 0xffff, 4, '0');
+						put_hex<uint32_t>(kutil::byteswap(g.b >> 32), 8, '0');
+					}
+					done = true;
+					break;
+
 				case 'l':
 					switch (*r++) {
 						case 'x': put_hex<uint64_t>(va_arg(args, uint64_t), right ? width : -width, pad); done = true; break;

src/kernel/cpu.h

@@ -1,5 +1,13 @@
 #pragma once
 
+struct cpu_state
+{
+	uint64_t ds;
+	uint64_t r15, r14, r13, r12, r11, r10, r9,  r8;
+	uint64_t rdi, rsi, rbp, rbx, rdx, rcx, rax;
+	uint64_t interrupt, errorcode;
+	uint64_t rip, cs, rflags, user_rsp, ss;
+};
 
 class cpu_id
 {

src/kernel/debug.s

@@ -1,10 +1,12 @@
+global get_rsp
+get_rsp:
+	mov rax, rsp
+	ret
 
-section .text
-global do_the_set_registers
-do_the_set_registers:
-	mov rax, 0xdeadbeef0badc0de
-	mov r8, rcx
-	mov r9, rdi
+global get_rip
+get_rip:
+	pop rax ; do the same thing as 'ret', except with 'jmp'
+	jmp rax ; with the return address still in rax
 
 global _halt
 _halt:

src/kernel/device_manager.cpp

@@ -320,3 +320,9 @@ device_manager::allocate_msi(const char *name, pci_device &device, irq_callback
 			static_cast<uint16_t>(vector));
 	return true;
 }
+
+void
+device_manager::register_block_device(block_device *blockdev)
+{
+	m_blockdevs.append(blockdev);
+}

src/kernel/device_manager.h

@@ -7,6 +7,7 @@
 struct acpi_xsdt;
 struct acpi_apic;
 struct acpi_mcfg;
+class block_device;
 class lapic;
 class ioapic;
 
@@ -65,6 +66,23 @@ public:
 		return false;
 	}
 
+	/// Register the existance of a block device.
+	/// \arg blockdev  Pointer to the block device
+	void register_block_device(block_device *blockdev);
+
+	/// Get the number of block devices in the system
+	/// \returns  A count of devices
+	inline unsigned get_num_block_devices() const { return m_blockdevs.count(); }
+
+	/// Get a block device
+	/// \arg i    Index of the device to get
+	/// \returns  A pointer to the requested device, or nullptr
+	inline block_device * get_block_device(unsigned i)
+	{
+		return i < m_blockdevs.count() ?
+			m_blockdevs[i] : nullptr;
+	}
+
 private:
 	/// Parse the ACPI XSDT and load relevant sub-tables.
 	/// \arg xsdt  Pointer to the XSDT from the firmware
@@ -100,6 +118,8 @@ private:
 	};
 	kutil::vector<irq_allocation> m_irqs;
 
+	kutil::vector<block_device *> m_blockdevs;
+
 	static device_manager s_instance;
 
 	device_manager() = delete;

src/kernel/fs/gpt.cpp

@@ -0,0 +1,114 @@
+#include "kutil/assert.h"
+#include "kutil/guid.h"
+#include "kutil/memory.h"
+#include "device_manager.h"
+#include "fs/gpt.h"
+#include "log.h"
+
+namespace fs {
+
+const kutil::guid efi_system_part = kutil::make_guid(0xC12A7328, 0xF81F, 0x11D2, 0xBA4B, 0x00A0C93EC93B);
+const kutil::guid efi_unused_part = kutil::make_guid(0, 0, 0, 0, 0);
+
+const uint64_t gpt_signature = 0x5452415020494645; // "EFI PART"
+const size_t block_size = 512;
+
+struct gpt_header
+{
+	uint64_t signature;
+	uint32_t revision;
+	uint32_t headersize;
+	uint32_t crc32;
+	uint32_t reserved;
+	uint64_t my_lba;
+	uint64_t alt_lba;
+	uint64_t first_usable_lba;
+	uint64_t last_usable_lba;
+	kutil::guid disk_guid;
+	uint64_t table_lba;
+	uint32_t entry_count;
+	uint32_t entry_length;
+	uint32_t table_crc32;
+} __attribute__ ((packed));
+
+struct gpt_entry
+{
+	kutil::guid type;
+	kutil::guid part_guid;
+	uint64_t start_lba;
+	uint64_t end_lba;
+	uint64_t attributes;
+	uint16_t name_wide[36];
+} __attribute__ ((packed));
+
+
+partition::partition(block_device *parent, size_t start, size_t length) :
+	m_parent(parent),
+	m_start(start),
+	m_length(length)
+{
+}
+
+size_t
+partition::read(size_t offset, size_t length, void *buffer)
+{
+	if (offset + length > m_length)
+		offset = m_length - offset;
+	return m_parent->read(m_start + offset, length, buffer);
+}
+
+
+unsigned
+partition::load(block_device *device)
+{
+	// Read LBA 1
+	uint8_t block[block_size];
+	size_t count = device->read(block_size, block_size, &block);
+	kassert(count == block_size, "Short read for GPT header.");
+
+	gpt_header *header = reinterpret_cast<gpt_header *>(&block);
+	if (header->signature != gpt_signature)
+		return 0;
+
+	size_t arraysize = header->entry_length * header->entry_count;
+	log::debug(logs::fs, "Found GPT header: %d paritions, size 0x%lx",
+			header->entry_count, arraysize);
+
+	uint8_t *array = new uint8_t[arraysize];
+	count = device->read(block_size * header->table_lba, arraysize, array);
+	kassert(count == arraysize, "Short read for GPT entry array.");
+
+	auto &dm = device_manager::get();
+
+	unsigned found = 0;
+	gpt_entry *entry0 = reinterpret_cast<gpt_entry *>(array);
+	for (uint32_t i = 0; i < header->entry_count; ++i) {
+		gpt_entry *entry = kutil::offset_pointer(entry0, i * header->entry_length);
+		if (entry->type == efi_unused_part) continue;
+
+		// TODO: real UTF16->UTF8
+		char name[sizeof(gpt_entry::name_wide) / 2];
+		for (int i = 0; i < sizeof(name); ++i)
+			name[i] = entry->name_wide[i];
+
+		log::debug(logs::fs, "Found partition %02x at %lx-%lx", i, entry->start_lba, entry->end_lba);
+		if (entry->type == efi_system_part)
+			log::debug(logs::fs, "   type EFI SYSTEM PARTITION");
+		else
+			log::debug(logs::fs, "   type %G", entry->type);
+		log::debug(logs::fs, "   name %s", name);
+		log::debug(logs::fs, "   attr %016lx", entry->attributes);
+
+		found += 1;
+		partition *part = new partition(
+				device,
+				entry->start_lba * block_size,
+				(entry->end_lba - entry->start_lba) * block_size);
+		dm.register_block_device(part);
+	}
+
+	return found;
+}
+
+
+} // namespace fs

src/kernel/fs/gpt.h

@@ -0,0 +1,37 @@
+#pragma once
+/// \file gpt.h
+/// Definitions for dealing with GUID Partition Tables
+#include "block_device.h"
+
+namespace fs {
+
+
+class partition :
+	public block_device
+{
+public:
+	/// Constructor.
+	/// \arg parent  The block device this partition is a part of
+	/// \arg start   The starting offset in bytes from the start of the parent
+	/// \arg lenght  The length in bytes of this partition
+	partition(block_device *parent, size_t start, size_t length);
+
+	/// Read bytes from the partition.
+	/// \arg offset  The offset in bytes at which to start reading
+	/// \arg length  The number of bytes to read
+	/// \arg buffer  [out] Data is read into this buffer
+	/// \returns     The number of bytes read
+	virtual size_t read(size_t offset, size_t length, void *buffer);
+
+	/// Find partitions on a block device and add them to the device manager
+	/// \arg device  The device to search for partitions
+	/// \returns     The number of partitions found
+	static unsigned load(block_device *device);
+
+private:
+	block_device *m_parent;
+	size_t m_start;
+	size_t m_length;
+};
+
+} // namespace fs

src/kernel/gdt.cpp

@@ -0,0 +1,273 @@
+#include <stdint.h>
+
+#include "kutil/assert.h"
+#include "kutil/enum_bitfields.h"
+#include "kutil/memory.h"
+#include "console.h"
+#include "log.h"
+
+
+enum class gdt_type : uint8_t
+{
+	accessed	= 0x01,
+	read_write	= 0x02,
+	conforming	= 0x04,
+	execute		= 0x08,
+	system		= 0x10,
+	ring1		= 0x20,
+	ring2		= 0x40,
+	ring3		= 0x60,
+	present		= 0x80
+};
+IS_BITFIELD(gdt_type);
+
+struct gdt_descriptor
+{
+	uint16_t limit_low;
+	uint16_t base_low;
+	uint8_t base_mid;
+	gdt_type type;
+	uint8_t size;
+	uint8_t base_high;
+} __attribute__ ((packed));
+
+struct tss_descriptor
+{
+	uint16_t limit_low;
+	uint16_t base_00;
+	uint8_t base_16;
+	gdt_type type;
+	uint8_t size;
+	uint8_t base_24;
+	uint32_t base_32;
+	uint32_t reserved;
+} __attribute__ ((packed));
+
+struct tss_entry
+{
+	uint32_t reserved0;
+
+	uint64_t rsp[3]; // stack pointers for CPL 0-2
+	uint64_t ist[8]; // ist[0] is reserved
+
+	uint64_t reserved1;
+	uint16_t reserved2;
+	uint16_t iomap_offset;
+} __attribute__ ((packed));
+
+struct idt_descriptor
+{
+	uint16_t base_low;
+	uint16_t selector;
+	uint8_t ist;
+	uint8_t flags;
+	uint16_t base_mid;
+	uint32_t base_high;
+	uint32_t reserved;		// must be zero
+} __attribute__ ((packed));
+
+struct table_ptr
+{
+	uint16_t limit;
+	uint64_t base;
+} __attribute__ ((packed));
+
+
+gdt_descriptor g_gdt_table[10];
+idt_descriptor g_idt_table[256];
+table_ptr g_gdtr;
+table_ptr g_idtr;
+tss_entry g_tss;
+
+
+extern "C" {
+	void idt_write();
+	void idt_load();
+
+	void gdt_write(uint16_t cs, uint16_t ds, uint16_t tr);
+	void gdt_load();
+}
+
+void
+gdt_set_entry(uint8_t i, uint32_t base, uint64_t limit, bool is64, gdt_type type)
+{
+	g_gdt_table[i].limit_low = limit & 0xffff;
+	g_gdt_table[i].size = (limit >> 16) & 0xf;
+	g_gdt_table[i].size |= (is64 ? 0xa0 : 0xc0);
+
+	g_gdt_table[i].base_low = base & 0xffff;
+	g_gdt_table[i].base_mid = (base >> 16) & 0xff;
+	g_gdt_table[i].base_high = (base >> 24) & 0xff;
+
+	g_gdt_table[i].type = type | gdt_type::system | gdt_type::present;
+}
+
+void
+tss_set_entry(uint8_t i, uint64_t base, uint64_t limit)
+{
+	tss_descriptor tssd;
+	tssd.limit_low = limit & 0xffff;
+	tssd.size = (limit >> 16) & 0xf;
+
+	tssd.base_00 = base & 0xffff;
+	tssd.base_16 = (base >> 16) & 0xff;
+	tssd.base_24 = (base >> 24) & 0xff;
+	tssd.base_32 = (base >> 32) & 0xffffffff;
+	tssd.reserved = 0;
+
+	tssd.type =
+		gdt_type::accessed |
+		gdt_type::execute |
+		gdt_type::ring3 |
+		gdt_type::present;
+	kutil::memcpy(&g_gdt_table[i], &tssd, sizeof(tss_descriptor));
+}
+
+void
+idt_set_entry(uint8_t i, uint64_t addr, uint16_t selector, uint8_t flags)
+{
+	g_idt_table[i].base_low = addr & 0xffff;
+	g_idt_table[i].base_mid = (addr >> 16) & 0xffff;
+	g_idt_table[i].base_high = (addr >> 32) & 0xffffffff;
+	g_idt_table[i].selector = selector;
+	g_idt_table[i].flags = flags;
+	g_idt_table[i].ist = 0;
+	g_idt_table[i].reserved = 0;
+}
+
+void
+tss_set_stack(int ring, addr_t rsp)
+{
+	kassert(ring < 3, "Bad ring passed to tss_set_stack.");
+	g_tss.rsp[ring] = rsp;
+}
+
+addr_t
+tss_get_stack(int ring)
+{
+	kassert(ring < 3, "Bad ring passed to tss_get_stack.");
+	return g_tss.rsp[ring];
+}
+
+void
+gdt_init()
+{
+	kutil::memset(&g_gdt_table, 0, sizeof(g_gdt_table));
+	kutil::memset(&g_idt_table, 0, sizeof(g_idt_table));
+
+	g_gdtr.limit = sizeof(g_gdt_table) - 1;
+	g_gdtr.base = reinterpret_cast<uint64_t>(&g_gdt_table);
+
+	// Kernel CS/SS - always 64bit
+	gdt_set_entry(1, 0, 0xfffff,  true, gdt_type::read_write | gdt_type::execute);
+	gdt_set_entry(2, 0, 0xfffff,  true, gdt_type::read_write);
+
+	// User CS32/SS/CS64 - layout expected by SYSRET
+	gdt_set_entry(3, 0, 0xfffff,  false, gdt_type::ring3 | gdt_type::read_write | gdt_type::execute);
+	gdt_set_entry(4, 0, 0xfffff,  true, gdt_type::ring3 | gdt_type::read_write);
+	gdt_set_entry(5, 0, 0xfffff,  true, gdt_type::ring3 | gdt_type::read_write | gdt_type::execute);
+
+	kutil::memset(&g_tss, 0, sizeof(tss_entry));
+	g_tss.iomap_offset = sizeof(tss_entry);
+
+	addr_t tss_base = reinterpret_cast<addr_t>(&g_tss);
+
+	// Note that this takes TWO GDT entries
+	tss_set_entry(6, tss_base, sizeof(tss_entry));
+
+	gdt_write(1 << 3, 2 << 3, 6 << 3);
+
+	g_idtr.limit = sizeof(g_idt_table) - 1;
+	g_idtr.base = reinterpret_cast<uint64_t>(&g_idt_table);
+
+	idt_write();
+}
+
+void
+gdt_dump()
+{
+	const table_ptr &table = g_gdtr;
+
+	console *cons = console::get();
+	cons->printf("         GDT: loc:%lx size:%d\n", table.base, table.limit+1);
+
+	int count = (table.limit + 1) / sizeof(gdt_descriptor);
+	const gdt_descriptor *gdt =
+		reinterpret_cast<const gdt_descriptor *>(table.base);
+
+	for (int i = 0; i < count; ++i) {
+		uint32_t base =
+			(gdt[i].base_high << 24) |
+			(gdt[i].base_mid << 16) |
+			gdt[i].base_low;
+
+		uint32_t limit =
+			static_cast<uint32_t>(gdt[i].size & 0x0f) << 16 |
+			gdt[i].limit_low;
+
+		cons->printf("          %02d:", i);
+		if (! bitfield_has(gdt[i].type, gdt_type::present)) {
+			cons->puts(" Not Present\n");
+			continue;
+		}
+
+		cons->printf(" Base %08x  limit %05x ", base, limit);
+
+		switch (gdt[i].type & gdt_type::ring3) {
+			case gdt_type::ring3: cons->puts("ring3"); break;
+			case gdt_type::ring2: cons->puts("ring2"); break;
+			case gdt_type::ring1: cons->puts("ring1"); break;
+			default: cons->puts("ring0"); break;
+		}
+
+		cons->printf(" %s %s %s %s %s %s %s\n",
+			bitfield_has(gdt[i].type, gdt_type::accessed) ? "A" : " ",
+			bitfield_has(gdt[i].type, gdt_type::read_write) ? "RW" : "  ",
+			bitfield_has(gdt[i].type, gdt_type::conforming) ? "C" : " ",
+			bitfield_has(gdt[i].type, gdt_type::execute) ? "EX" : "  ",
+			bitfield_has(gdt[i].type, gdt_type::system) ? "S" : " ",
+			(gdt[i].size & 0x80) ? "KB" : " B",
+			(gdt[i].size & 0x60) == 0x20 ? "64" :
+				(gdt[i].size & 0x60) == 0x40 ? "32" : "16");
+	}
+}
+
+void
+idt_dump()
+{
+	const table_ptr &table = g_idtr;
+
+	log::info(logs::boot, "Loaded IDT at: %lx size: %d bytes", table.base, table.limit+1);
+
+	int count = (table.limit + 1) / sizeof(idt_descriptor);
+	const idt_descriptor *idt =
+		reinterpret_cast<const idt_descriptor *>(table.base);
+
+	for (int i = 0; i < count; ++i) {
+		uint64_t base =
+			(static_cast<uint64_t>(idt[i].base_high) << 32) |
+			(static_cast<uint64_t>(idt[i].base_mid)  << 16) |
+			idt[i].base_low;
+
+		char const *type;
+		switch (idt[i].flags & 0xf) {
+			case 0x5: type = " 32tsk "; break;
+			case 0x6: type = " 16int "; break;
+			case 0x7: type = " 16trp "; break;
+			case 0xe: type = " 32int "; break;
+			case 0xf: type = " 32trp "; break;
+			default:  type = " ????? "; break;
+		}
+
+		if (idt[i].flags & 0x80) {
+			log::debug(logs::boot,
+					"   Entry %3d: Base:%lx Sel(rpl %d, ti %d, %3d) IST:%d %s DPL:%d", i, base,
+					(idt[i].selector & 0x3),
+					((idt[i].selector & 0x4) >> 2),
+					(idt[i].selector >> 3),
+					idt[i].ist,
+					type,
+					((idt[i].flags >> 5) & 0x3));
+		}
+	}
+}

src/kernel/gdt.h

@@ -0,0 +1,32 @@
+#pragma once
+/// \file gdt.h
+/// Definitions relating to system descriptor tables: GDT, IDT, TSS
+#include <stdint.h>
+#include "kutil/memory.h"
+
+/// Set up the GDT and TSS, and switch segment registers to point
+/// to them.
+void gdt_init();
+
+/// Set an entry in the IDT
+/// \arg i         Index in the IDT (vector of the interrupt this handles)
+/// \arg addr      Address of the handler
+/// \arg selector  GDT selector to set when invoking this handler
+/// \arg flags     Descriptor flags to set
+void idt_set_entry(uint8_t i, uint64_t addr, uint16_t selector, uint8_t flags);
+
+/// Set the stack pointer for a given ring in the TSS
+/// \arg ring  Ring to set for (0-2)
+/// \arg rsp   Stack pointer to set
+void tss_set_stack(int ring, addr_t rsp);
+
+/// Get the stack pointer for a given ring in the TSS
+/// \arg ring  Ring to get (0-2)
+/// \returns   Stack pointers for that ring
+addr_t tss_get_stack(int ring);
+
+/// Dump information about the current GDT to the screen
+void gdt_dump();
+
+/// Dump information about the current IDT to the screen
+void idt_dump();

src/kernel/interrupt_isrs.inc

@@ -134,6 +134,8 @@ IRQ (0x7d, 0x5d, irq5D)
 IRQ (0x7e, 0x5e, irq5E)
 IRQ (0x7f, 0x5f, irq5F)
 
+ISR (0xe7, isrAssert)
+
 ISR (0xec, isrTimer)
 ISR (0xed, isrLINT0)
 ISR (0xee, isrLINT1)

src/kernel/interrupts.cpp

@@ -1,72 +1,18 @@
-#include <stddef.h>
 #include <stdint.h>
 
-#include "kutil/enum_bitfields.h"
-#include "kutil/memory.h"
 #include "console.h"
+#include "cpu.h"
 #include "device_manager.h"
+#include "gdt.h"
 #include "interrupts.h"
 #include "io.h"
 #include "log.h"
-
-enum class gdt_flags : uint8_t
-{
-	ac = 0x01,
-	rw = 0x02,
-	dc = 0x04,
-	ex = 0x08,
-	r1 = 0x20,
-	r2 = 0x40,
-	r3 = 0x60,
-	pr = 0x80,
-
-	res_1 = 0x10
-};
-IS_BITFIELD(gdt_flags);
-
-struct gdt_descriptor
-{
-	uint16_t limit_low;
-	uint16_t base_low;
-	uint8_t base_mid;
-	uint8_t flags;
-	uint8_t granularity;
-	uint8_t base_high;
-} __attribute__ ((packed));
-
-struct idt_descriptor
-{
-	uint16_t base_low;
-	uint16_t selector;
-	uint8_t ist;
-	uint8_t flags;
-	uint16_t base_mid;
-	uint32_t base_high;
-	uint32_t reserved;		// must be zero
-} __attribute__ ((packed));
-
-struct table_ptr
-{
-	uint16_t limit;
-	uint64_t base;
-} __attribute__ ((packed));
-
-gdt_descriptor g_gdt_table[10];
-idt_descriptor g_idt_table[256];
-table_ptr g_gdtr;
-table_ptr g_idtr;
-
-struct registers;
+#include "scheduler.h"
 
 extern "C" {
-	void idt_write();
-	void idt_load();
-
-	void gdt_write();
-	void gdt_load();
-
-	void isr_handler(registers);
-	void irq_handler(registers);
+	addr_t isr_handler(addr_t, cpu_state);
+	addr_t irq_handler(addr_t, cpu_state);
+	addr_t syscall_handler(addr_t, cpu_state);
 
 #define ISR(i, name)     extern void name ();
 #define EISR(i, name)    extern void name ();
@@ -77,9 +23,6 @@ extern "C" {
 #undef ISR
 }
 
-void idt_dump(const table_ptr &table);
-void gdt_dump(const table_ptr &table);
-
 isr
 operator+(const isr &lhs, int rhs)
 {
@@ -103,33 +46,6 @@ get_irq(unsigned vector)
 	}
 }
 
-void
-set_gdt_entry(uint8_t i, uint32_t base, uint32_t limit, bool is64, gdt_flags flags)
-{
-	g_gdt_table[i].limit_low = limit & 0xffff;
-
-	g_gdt_table[i].base_low = base & 0xffff;
-	g_gdt_table[i].base_mid = (base >> 16) & 0xff;
-	g_gdt_table[i].base_high = (base >> 24) & 0xff;
-
-	g_gdt_table[i].granularity =
-			((limit >> 16) & 0xf) | (is64 ? 0xa0 : 0xc0);
-
-	g_gdt_table[i].flags = static_cast<uint8_t>(flags | gdt_flags::res_1 | gdt_flags::pr);
-}
-
-void
-set_idt_entry(uint8_t i, uint64_t addr, uint16_t selector, uint8_t flags)
-{
-	g_idt_table[i].base_low = addr & 0xffff;
-	g_idt_table[i].base_mid = (addr >> 16) & 0xffff;
-	g_idt_table[i].base_high = (addr >> 32) & 0xffffffff;
-	g_idt_table[i].selector = selector;
-	g_idt_table[i].flags = flags;
-	g_idt_table[i].ist = 0;
-	g_idt_table[i].reserved = 0;
-}
-
 static void
 disable_legacy_pic()
 {
@@ -164,48 +80,20 @@ enable_serial_interrupts()
 void
 interrupts_init()
 {
-	kutil::memset(&g_gdt_table, 0, sizeof(g_gdt_table));
-	kutil::memset(&g_idt_table, 0, sizeof(g_idt_table));
-
-	g_gdtr.limit = sizeof(g_gdt_table) - 1;
-	g_gdtr.base = reinterpret_cast<uint64_t>(&g_gdt_table);
-
-	set_gdt_entry(1, 0, 0xfffff, false, gdt_flags::rw);
-	set_gdt_entry(2, 0, 0xfffff, false, gdt_flags::rw | gdt_flags::ex | gdt_flags::dc);
-	set_gdt_entry(3, 0, 0xfffff, false, gdt_flags::rw);
-	set_gdt_entry(4, 0, 0xfffff, false, gdt_flags::rw | gdt_flags::ex);
-
-	set_gdt_entry(6, 0, 0xfffff, false, gdt_flags::rw);
-	set_gdt_entry(7, 0, 0xfffff,  true, gdt_flags::rw | gdt_flags::ex);
-
-	gdt_write();
-
-	g_idtr.limit = sizeof(g_idt_table) - 1;
-	g_idtr.base = reinterpret_cast<uint64_t>(&g_idt_table);
-
-#define ISR(i, name)     set_idt_entry(i, reinterpret_cast<uint64_t>(& name), 0x38, 0x8e);
-#define EISR(i, name)    set_idt_entry(i, reinterpret_cast<uint64_t>(& name), 0x38, 0x8e);
-#define IRQ(i, q, name)  set_idt_entry(i, reinterpret_cast<uint64_t>(& name), 0x38, 0x8e);
+#define ISR(i, name)     idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0x8e);
+#define EISR(i, name)    idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0x8e);
+#define IRQ(i, q, name)  idt_set_entry(i, reinterpret_cast<uint64_t>(& name), 0x08, 0x8e);
 #include "interrupt_isrs.inc"
 #undef IRQ
 #undef EISR
 #undef ISR
 
-	idt_write();
 	disable_legacy_pic();
 	enable_serial_interrupts();
 
 	log::info(logs::boot, "Interrupts enabled.");
 }
 
-struct registers
-{
-	uint64_t ds;
-	uint64_t rdi, rsi, rbp, rsp, rbx, rdx, rcx, rax;
-	uint64_t interrupt, errorcode;
-	uint64_t rip, cs, eflags, user_esp, ss;
-};
-
 #define print_reg(name, value) cons->printf("         %s: %016lx\n", name, (value));
 
 extern "C" uint64_t get_frame(int frame);
@@ -222,14 +110,16 @@ print_stacktrace(int skip = 0)
 	}
 }
 
-void
-isr_handler(registers regs)
+addr_t
+isr_handler(addr_t return_rsp, cpu_state regs)
 {
 	console *cons = console::get();
 
 	switch (static_cast<isr>(regs.interrupt & 0xff)) {
-	case isr::isrTimer:
-		cons->puts("\nTICK\n");
+	case isr::isrTimer: {
+			scheduler &s = scheduler::get();
+			return_rsp = s.tick(return_rsp);
+		}
 		break;
 
 	case isr::isrLINT0:
@@ -248,11 +138,67 @@ isr_handler(registers regs)
 	case isr::isrIgnore5:
 	case isr::isrIgnore6:
 	case isr::isrIgnore7:
+		break;
 
-		/*
-		cons->printf("\nIGNORED PIC INTERRUPT %d\n",
-				(regs.interrupt % 0xff) - 0xf0);
-		*/
+	case isr::isrGPFault: {
+			cons->set_color(9);
+			cons->puts("\nGeneral Protection Fault:\n");
+			cons->set_color();
+
+			cons->puts("       flags:");
+			if (regs.errorcode & 0x01) cons->puts(" external");
+
+			int index = (regs.errorcode & 0xf8) >> 3;
+			if (index) {
+				switch (regs.errorcode & 0x06) {
+				case 0:
+					cons->printf(" GDT[%d]\n", index);
+					gdt_dump();
+					break;
+
+				case 1:
+				case 3:
+					cons->printf(" IDT[%d]\n", index);
+					idt_dump();
+					break;
+
+				default:
+					cons->printf(" LDT[%d]??\n", index);
+					break;
+				}
+			} else {
+				cons->putc('\n');
+			}
+
+			cons->puts("\n");
+			print_reg(" ds", regs.ds);
+			print_reg(" cs", regs.cs);
+			print_reg(" ss", regs.ss);
+
+
+			cons->puts("\n");
+			print_reg("rax", regs.rax);
+			print_reg("rbx", regs.rbx);
+			print_reg("rcx", regs.rcx);
+			print_reg("rdx", regs.rdx);
+			print_reg("rdi", regs.rdi);
+			print_reg("rsi", regs.rsi);
+
+			cons->puts("\n");
+			print_reg("rbp", regs.rbp);
+
+			cons->puts("\n");
+			print_reg("rip", regs.rip);
+			print_stacktrace(2);
+
+			cons->puts("\nStack:\n");
+			uint64_t sp = regs.user_rsp;
+			while (sp <= regs.rbp) {
+				cons->printf("%016x: %016x\n", sp, *reinterpret_cast<uint64_t *>(sp));
+				sp += sizeof(uint64_t);
+			}
+		}
+		while(1) asm("hlt");
 		break;
 
 	case isr::isrPageFault: {
@@ -272,6 +218,7 @@ isr_handler(registers regs)
 			__asm__ __volatile__ ("mov %%cr2, %0" : "=r"(cr2));
 			print_reg("cr2", cr2);
 
+			print_reg("rsp", regs.user_rsp);
 			print_reg("rip", regs.rip);
 
 			cons->puts("\n");
@@ -280,6 +227,28 @@ isr_handler(registers regs)
 		while(1) asm("hlt");
 		break;
 
+	case isr::isrAssert: {
+			cons->set_color();
+
+			cons->puts("\n");
+			print_reg("rax", regs.rax);
+			print_reg("rbx", regs.rbx);
+			print_reg("rcx", regs.rcx);
+			print_reg("rdx", regs.rdx);
+			print_reg("rdi", regs.rdi);
+			print_reg("rsi", regs.rsi);
+
+			cons->puts("\n");
+			print_reg("rbp", regs.rbp);
+			print_reg("rsp", regs.user_rsp);
+
+			cons->puts("\n");
+			print_reg("rip", regs.rip);
+			print_stacktrace(2);
+		}
+		while(1) asm("hlt");
+		break;
+
 	default:
 		cons->set_color(9);
 		cons->puts("\nReceived ISR interrupt:\n");
@@ -293,7 +262,6 @@ isr_handler(registers regs)
 		print_reg("rdi", regs.rdi);
 		print_reg("rsi", regs.rsi);
 		print_reg("rbp", regs.rbp);
-		print_reg("rsp", regs.rsp);
 		print_reg("rbx", regs.rbx);
 		print_reg("rdx", regs.rdx);
 		print_reg("rcx", regs.rcx);
@@ -302,20 +270,21 @@ isr_handler(registers regs)
 
 		print_reg("rip", regs.rip);
 		print_reg(" cs", regs.cs);
-		print_reg(" ef", regs.eflags);
-		print_reg("esp", regs.user_esp);
+		print_reg(" ef", regs.rflags);
+		print_reg("rsp", regs.user_rsp);
 		print_reg(" ss", regs.ss);
 
 		cons->puts("\n");
 		print_stacktrace(2);
 		while(1) asm("hlt");
 	}
-
 	*reinterpret_cast<uint32_t *>(0xffffff80fee000b0) = 0;
+
+	return return_rsp;
 }
 
-void
-irq_handler(registers regs)
+addr_t
+irq_handler(addr_t return_rsp, cpu_state regs)
 {
 	console *cons = console::get();
 	uint8_t irq = get_irq(regs.interrupt);
@@ -329,7 +298,6 @@ irq_handler(registers regs)
 		print_reg("rdi", regs.rdi);
 		print_reg("rsi", regs.rsi);
 		print_reg("rbp", regs.rbp);
-		print_reg("rsp", regs.rsp);
 		print_reg("rbx", regs.rbx);
 		print_reg("rdx", regs.rdx);
 		print_reg("rcx", regs.rcx);
@@ -338,87 +306,46 @@ irq_handler(registers regs)
 
 		print_reg("rip", regs.rip);
 		print_reg(" cs", regs.cs);
-		print_reg(" ef", regs.eflags);
-		print_reg("esp", regs.user_esp);
+		print_reg(" ef", regs.rflags);
+		print_reg("rsp", regs.user_rsp);
 		print_reg(" ss", regs.ss);
 		while(1) asm("hlt");
 	}
 
 	*reinterpret_cast<uint32_t *>(0xffffff80fee000b0) = 0;
+	return return_rsp;
 }
 
-
-void
-gdt_dump(const table_ptr &table)
+addr_t
+syscall_handler(addr_t return_rsp, cpu_state regs)
 {
-	log::info(logs::boot, "Loaded GDT at: %lx size: %d bytes", table.base, table.limit+1);
+	console *cons = console::get();
+	cons->printf("SYSCALL\n");
 
-	int count = (table.limit + 1) / sizeof(gdt_descriptor);
-	const gdt_descriptor *gdt =
-		reinterpret_cast<const gdt_descriptor *>(table.base);
+	cons->puts("\n");
+	print_reg("rax", regs.rax);
+	print_reg("rbx", regs.rbx);
+	print_reg("rcx", regs.rcx);
+	print_reg("rdx", regs.rdx);
+	print_reg("rdi", regs.rdi);
+	print_reg("rsi", regs.rsi);
 
-	for (int i = 0; i < count; ++i) {
-		uint32_t base =
-			(gdt[i].base_high << 24) |
-			(gdt[i].base_mid << 16) |
-			gdt[i].base_low;
+	cons->puts("\n");
+	print_reg(" r8", regs.r8);
+	print_reg(" r9", regs.r9);
+	print_reg("r10", regs.r10);
+	print_reg("r11", regs.r11);
+	print_reg("r12", regs.r12);
+	print_reg("r13", regs.r13);
+	print_reg("r14", regs.r14);
+	print_reg("r15", regs.r15);
 
-		uint32_t limit =
-			static_cast<uint32_t>(gdt[i].granularity & 0x0f) << 16 |
-			gdt[i].limit_low;
+	cons->puts("\n");
+	print_reg("rbp", regs.rbp);
+	print_reg("rsp", regs.user_rsp);
+	print_reg("sp0", tss_get_stack(0));
 
-		if (gdt[i].flags & 0x80) {
-			log::debug(logs::boot,
-					"   Entry %3d: Base %x  limit %x  privs %d  flags %s%s%s%s%s%s",
-					i, base, limit, ((gdt[i].flags >> 5) & 0x03),
-
-					(gdt[i].flags & 0x80) ? "P " : "  ",
-					(gdt[i].flags & 0x08) ? "ex " : "   ",
-					(gdt[i].flags & 0x04) ? "dc " : "   ",
-					(gdt[i].flags & 0x02) ? "rw " : "   ",
-
-					(gdt[i].granularity & 0x80) ? "kb " : "b  ",
-					(gdt[i].granularity & 0x60) == 0x60 ? "64" :
-						(gdt[i].granularity & 0x60) == 0x40 ? "32" : "16"
-					);
-		}
-	}
-}
-
-void
-idt_dump(const table_ptr &table)
-{
-	log::info(logs::boot, "Loaded IDT at: %lx size: %d bytes", table.base, table.limit+1);
-
-	int count = (table.limit + 1) / sizeof(idt_descriptor);
-	const idt_descriptor *idt =
-		reinterpret_cast<const idt_descriptor *>(table.base);
-
-	for (int i = 0; i < count; ++i) {
-		uint64_t base =
-			(static_cast<uint64_t>(idt[i].base_high) << 32) |
-			(static_cast<uint64_t>(idt[i].base_mid)  << 16) |
-			idt[i].base_low;
-
-		char const *type;
-		switch (idt[i].flags & 0xf) {
-			case 0x5: type = " 32tsk "; break;
-			case 0x6: type = " 16int "; break;
-			case 0x7: type = " 16trp "; break;
-			case 0xe: type = " 32int "; break;
-			case 0xf: type = " 32trp "; break;
-			default:  type = " ????? "; break;
-		}
-
-		if (idt[i].flags & 0x80) {
-			log::debug(logs::boot,
-					"   Entry %3d: Base:%lx Sel(rpl %d, ti %d, %3d) IST:%d %s DPL:%d", i, base,
-					(idt[i].selector & 0x3),
-					((idt[i].selector & 0x4) >> 2),
-					(idt[i].selector >> 3),
-					idt[i].ist,
-					type,
-					((idt[i].flags >> 5) & 0x3));
-		}
-	}
+	cons->puts("\n");
+	print_reg("rip", regs.rip);
+	return return_rsp;
 }

src/kernel/interrupts.h

@@ -1,6 +1,7 @@
 #pragma once
 /// \file interrupts.h
 /// Free functions and definitions related to interrupt service vectors
+#include <stdint.h>
 
 
 /// Enum of all defined ISR/IRQ vectors
@@ -17,11 +18,17 @@ enum class isr : uint8_t
 	_zero = 0
 };
 
+/// Helper operator to add an offset to an isr vector
 isr operator+(const isr &lhs, int rhs);
 
 extern "C" {
+	/// Set the CPU interrupt enable flag (sti)
 	void interrupts_enable();
+
+	/// Set the CPU interrupt disable flag (cli)
 	void interrupts_disable();
 }
 
+/// Fill the IDT with our ISRs, and disable the legacy
+/// PIC interrupts.
 void interrupts_init();

src/kernel/interrupts.s

@@ -14,6 +14,18 @@ idt_load:
 global gdt_write
 gdt_write:
 	lgdt [rel g_gdtr]
+	mov ax, si ; second arg is data segment
+	mov ds, ax
+	mov es, ax
+	mov fs, ax
+	mov gs, ax
+	mov ss, ax
+	push qword rdi ; first arg is code segment
+	lea rax, [rel .next]
+	push rax
+	o64 retf
+.next:
+	ltr dx
 	ret
 
 global gdt_load
@@ -26,11 +38,19 @@ gdt_load:
 	push rcx
 	push rdx
 	push rbx
-	push rsp
 	push rbp
 	push rsi
 	push rdi
 
+	push r8
+	push r9
+	push r10
+	push r11
+	push r12
+	push r13
+	push r14
+	push r15
+
 	mov ax, ds
 	push rax
 %endmacro
@@ -42,31 +62,32 @@ gdt_load:
 	mov fs, ax
 	mov gs, ax
 
+	pop r15
+	pop r14
+	pop r13
+	pop r12
+	pop r11
+	pop r10
+	pop r9
+	pop r8
+
 	pop rdi
 	pop rsi
 	pop rbp
-	pop rsp
 	pop rbx
 	pop rdx
 	pop rcx
 	pop rax
 %endmacro
 
-%macro load_kernel_segments 0
-	mov ax, 0x10	; load the kernel data segment
-	mov ds, ax
-	mov es, ax
-	mov fs, ax
-	mov gs, ax
-%endmacro
-
 extern isr_handler
 global isr_handler_prelude
 isr_handler_prelude:
 	push_all_and_segments
-	load_kernel_segments
 
+	mov rdi, rsp
 	call isr_handler
+	mov rsp, rax
 
 	pop_all_and_segments
 
@@ -78,9 +99,10 @@ extern irq_handler
 global irq_handler_prelude
 irq_handler_prelude:
 	push_all_and_segments
-	load_kernel_segments
 
+	mov rdi, rsp
 	call irq_handler
+	mov rsp, rax
 
 	pop_all_and_segments
 
@@ -120,3 +142,75 @@ irq_handler_prelude:
 
 section .isrs
 %include "interrupt_isrs.inc"
+
+extern syscall_handler
+syscall_handler_prelude:
+	push 0		; ss, doesn't matter here
+	push rsp
+	pushf
+	push 0		; cs, doesn't matter here
+	push rcx	; user rip
+	push 0		; bogus interrupt
+	push 0		; bogus errorcode
+	push_all_and_segments
+
+	mov rdi, rsp
+	call syscall_handler
+	mov rsp, rax
+
+	pop_all_and_segments
+	add rsp, 16	; ignore bogus interrupt / error
+	pop rcx		; user rip
+	add rsp, 32 ; ignore cs, flags, rsp, ss
+
+	o64 sysret
+
+global syscall_enable
+syscall_enable:
+	; IA32_EFER - set bit 0, syscall enable
+	mov rcx, 0xc0000080
+	rdmsr
+	or rax, 0x1
+	wrmsr
+
+	; IA32_STAR - cs for syscall
+	mov rcx, 0xc0000081
+	mov rax, 0			; not used
+	mov rdx, 0x00180008	; GDT:3 (user code), GDT:1 (kernel code)
+	wrmsr
+
+	; IA32_LSTAR - RIP for syscall
+	mov rcx, 0xc0000082
+	lea rax, [rel syscall_handler_prelude]
+	mov rdx, rax
+	shr rdx, 32
+	wrmsr
+
+	; IA32_FMASK - FLAGS mask inside syscall
+	mov rcx, 0xc0000084
+	mov rax, 0x200
+	mov rdx, 0
+	wrmsr
+
+
+	ret
+
+global taskA
+taskA:
+	push rbp
+	mov rbp, rsp
+	mov rax, 0xaaaaaaaaaaaaaaaa
+
+.loop:
+	syscall
+	jmp .loop
+
+global taskB
+taskB:
+	push rbp
+	mov rbp, rsp
+	mov rax, 0xbbbbbbbbbbbbbbbb
+
+.loop:
+	syscall
+	jmp .loop

src/kernel/io.cpp

@@ -31,7 +31,8 @@ wrmsr(uint64_t addr, uint64_t value)
 }
 
 void
-io_wait()
+io_wait(unsigned times)
 {
-	outb(0x80, 0);
+	for (unsigned i = 0; i < times; ++i)
+		outb(0x80, 0);
 }

src/kernel/io.h

@@ -25,7 +25,8 @@ uint64_t rdmsr(uint64_t addr);
 void wrmsr(uint64_t addr, uint64_t value);
 
 /// Pause briefly by doing IO to port 0x80
-void io_wait();
+/// \arg times  Number of times to delay by writing
+void io_wait(unsigned times = 1);
 
 }
 

src/kernel/log.cpp

@@ -10,11 +10,13 @@ static const uint8_t level_colors[] = {0x07, 0x0f, 0x0b, 0x09};
 
 static const char *levels[] = {"debug", " info", " warn", "error", "fatal"};
 static const char *areas[] = {
-	"boot",
-	"mem ",
-	"apic",
-	"dev ",
-	"driv",
+	"boot  ",
+	"memory",
+	"apic  ",
+	"device",
+	"driver",
+	"file  ",
+	"task  ",
 
 	nullptr
 };

src/kernel/log.h

@@ -12,6 +12,8 @@ enum class logs
 	apic,
 	device,
 	driver,
+	fs,
+	task,
 
 	max
 };

src/kernel/main.cpp

@@ -3,32 +3,29 @@
 
 #include "kutil/assert.h"
 #include "kutil/memory.h"
-
-#include "ahci/driver.h"
-#include "ahci/port.h"
-
+#include "block_device.h"
 #include "console.h"
 #include "cpu.h"
 #include "device_manager.h"
 #include "font.h"
+#include "gdt.h"
 #include "interrupts.h"
 #include "io.h"
 #include "kernel_data.h"
 #include "log.h"
 #include "memory.h"
 #include "page_manager.h"
+#include "scheduler.h"
 #include "screen.h"
 #include "serial.h"
 
 extern "C" {
-	void do_the_set_registers(popcorn_data *header);
 	void kernel_main(popcorn_data *header);
+	void syscall_enable();
 
 	void *__bss_start, *__bss_end;
 }
 
-extern ahci_driver ahcid;
-
 extern void __kernel_assert(const char *, unsigned, const char *);
 
 void
@@ -58,7 +55,9 @@ init_console(const popcorn_data *header)
 	log::enable(logs::apic, log::level::info);
 	log::enable(logs::device, log::level::debug);
 	log::enable(logs::driver, log::level::debug);
-	log::enable(logs::memory, log::level::debug);
+	log::enable(logs::memory, log::level::info);
+	log::enable(logs::fs, log::level::debug);
+	log::enable(logs::task, log::level::debug);
 }
 
 void do_error_3() { volatile int x = 1; volatile int y = 0; volatile int z = x / y; }
@@ -70,6 +69,9 @@ kernel_main(popcorn_data *header)
 {
 	kutil::assert_set_callback(__kernel_assert);
 
+	gdt_init();
+	interrupts_init();
+
 	page_manager *pager = new (&g_page_manager) page_manager;
 
 	memory_initialize(
@@ -82,43 +84,73 @@ kernel_main(popcorn_data *header)
 			header->frame_buffer_length);
 
 	init_console(header);
-	// pager->dump_blocks();
 
-	interrupts_init();
+	log::debug(logs::boot, " Popcorn header is at: %016lx", header);
+	log::debug(logs::boot, "    Framebuffer is at: %016lx", header->frame_buffer);
+	log::debug(logs::boot, "      Font data is at: %016lx", header->font);
+	log::debug(logs::boot, "    Kernel data is at: %016lx", header->data);
+	log::debug(logs::boot, "     Memory map is at: %016lx", header->memory_map);
+	log::debug(logs::boot, "ACPI root table is at: %016lx", header->acpi_table);
+	log::debug(logs::boot, "Runtime service is at: %016lx", header->runtime);
+
+	/*
+	pager->dump_pml4(nullptr, 0);
+	pager->dump_blocks(true);
+	*/
 
 	device_manager *devices =
 		new (&device_manager::get()) device_manager(header->acpi_table);
 
 	interrupts_enable();
 
+	/*
 	cpu_id cpu;
 	log::info(logs::boot, "CPU Vendor: %s", cpu.vendor_id());
 	log::info(logs::boot, "CPU Family %x Model %x Stepping %x",
 			cpu.family(), cpu.model(), cpu.stepping());
 
+	auto r = cpu.get(0x15);
+	log::info(logs::boot, "CPU Crystal: %dHz", r.ecx);
+
+	addr_t cr4 = 0;
+	__asm__ __volatile__ ( "mov %%cr4, %0" : "=r" (cr4) );
+	log::info(logs::boot, "cr4: %016x", cr4);
+	*/
+
 	devices->init_drivers();
 
-	ahci::port *disk = ahcid.find_disk();
+	/*
+	block_device *disk = devices->get_block_device(0);
 	if (disk) {
-		uint8_t buf[512];
-		kutil::memset(buf, 0, 512);
+		for (int i=0; i<1; ++i) {
+			uint8_t buf[512];
+			kutil::memset(buf, 0, 512);
 
-		disk->read(1, sizeof(buf), buf);
-		while (buf[0] == 0) io_wait();
+			kassert(disk->read(0x200, sizeof(buf), buf),
+					"Disk read returned 0");
 
-		console *cons = console::get();
-		uint8_t *p = &buf[0];
-		for (int i = 0; i < 8; ++i) {
-			for (int j = 0; j < 16; ++j) {
-				cons->printf(" %02x", *p++);
+			console *cons = console::get();
+			uint8_t *p = &buf[0];
+			for (int i = 0; i < 8; ++i) {
+				for (int j = 0; j < 16; ++j) {
+					cons->printf(" %02x", *p++);
+				}
+				cons->putc('\n');
 			}
-			cons->putc('\n');
 		}
+	} else {
+		log::warn(logs::boot, "No block devices present.");
 	}
+	*/
 
 	// do_error_1();
 	// __asm__ __volatile__("int $15");
 
-	g_console.puts("boogity!");
-	do_the_set_registers(header);
+	// pager->dump_pml4();
+
+	syscall_enable();
+	scheduler *sched = new (&scheduler::get()) scheduler(devices->get_lapic());
+	sched->start();
+
+	g_console.puts("boogity!\n");
 }

src/kernel/memory_bootstrap.cpp

@@ -4,6 +4,7 @@
 #include "page_manager.h"
 
 const unsigned efi_page_size = 0x1000;
+const unsigned ident_page_flags = 0xb;
 
 enum class efi_memory_type : uint32_t
 {
@@ -306,7 +307,7 @@ copy_new_table(page_table *base, unsigned index, page_table *new_table)
 		for (int i = 0; i < 512; ++i) new_table->entries[i] = 0;
 	}
 
-	base->entries[index] = reinterpret_cast<uint64_t>(new_table) | 0xb;
+	base->entries[index] = reinterpret_cast<uint64_t>(new_table) | ident_page_flags;
 }
 
 static uint64_t
@@ -363,7 +364,7 @@ check_needs_page_ident(page_table *table, unsigned index, page_table **free_page
 
 	page_table *new_table = (*free_pages)++;
 	for (int i=0; i<512; ++i) new_table->entries[i] = 0;
-	table->entries[index] = reinterpret_cast<uint64_t>(new_table) | 0xb;
+	table->entries[index] = reinterpret_cast<uint64_t>(new_table) | ident_page_flags;
 	return 1;
 }
 
@@ -390,12 +391,24 @@ page_in_ident(
 					tables[1]->entries[idx[1]] & ~0xfffull);
 
 			for (; idx[2] < 512; idx[2] += 1, idx[3] = 0) {
+				if (idx[3] == 0 &&
+					count >= 512 &&
+					tables[2]->get(idx[2]) == nullptr) {
+					// Do a 2MiB page instead
+					tables[2]->entries[idx[2]] = phys_addr | 0x80 | ident_page_flags;
+
+					phys_addr += page_manager::page_size * 512;
+					count -= 512;
+					if (count == 0) return pages_consumed;
+					continue;
+				}
+
 				pages_consumed += check_needs_page_ident(tables[2], idx[2], &free_pages);
 				tables[3] = reinterpret_cast<page_table *>(
 						tables[2]->entries[idx[2]] & ~0xfffull);
 
 				for (; idx[3] < 512; idx[3] += 1) {
-					tables[3]->entries[idx[3]] = phys_addr | 0xb;
+					tables[3]->entries[idx[3]] = phys_addr | ident_page_flags;
 					phys_addr += page_manager::page_size;
 					if (--count == 0) return pages_consumed;
 				}
@@ -423,7 +436,7 @@ memory_initialize(const void *memory_map, size_t map_length, size_t desc_length)
 
 	// Offset-map this region into the higher half.
 	uint64_t free_region_start_virt =
-		free_region_start_phys + page_manager::high_offset;
+		free_region_start_phys + page_manager::page_offset;
 
 	uint64_t free_next = free_region_start_virt;
 

src/kernel/page_manager.cpp

@@ -2,6 +2,7 @@
 
 #include "kutil/assert.h"
 #include "kutil/memory_manager.h"
+#include "console.h"
 #include "log.h"
 #include "page_manager.h"
 
@@ -140,7 +141,7 @@ page_block::dump(page_block *list, const char *name, bool show_unmapped)
 
 		if (cur->virtual_address) {
 			page_table_indices start{cur->virtual_address};
-			log::info(logs::memory, "  %lx %x [%6d] %lx (%d,%d,%d,%d)",
+			log::info(logs::memory, "  %016lx %08x [%6d] %016lx (%d,%d,%d,%d)",
 					cur->physical_address,
 					cur->flags,
 					cur->count,
@@ -148,7 +149,7 @@ page_block::dump(page_block *list, const char *name, bool show_unmapped)
 					start[0], start[1], start[2], start[3]);
 		} else {
 			page_table_indices start{cur->virtual_address};
-			log::info(logs::memory, "  %lx %x [%6d]",
+			log::info(logs::memory, "  %016lx %08x [%6d]",
 					cur->physical_address,
 					cur->flags,
 					cur->count);
@@ -218,6 +219,34 @@ page_manager::init(
 	new (&g_kernel_memory_manager) kutil::memory_manager(
 			reinterpret_cast<void *>(end),
 			mm_grow_callback);
+
+	m_kernel_pml4 = get_pml4();
+}
+
+page_table *
+page_manager::create_process_map()
+{
+	page_table *table = get_table_page();
+
+	kutil::memset(table, 0, page_size);
+	table->entries[510] = m_kernel_pml4->entries[510];
+	table->entries[511] = m_kernel_pml4->entries[511];
+
+	// Create the initial user stack
+	map_pages(
+		initial_stack - (initial_stack_pages * page_size),
+		initial_stack_pages,
+		true, // This is the ring3 stack, user = true
+		table);
+
+	return table;
+}
+
+void
+page_manager::delete_process_map(page_table *table)
+{
+	// TODO: recurse table entries and mark them free
+	unmap_pages(table, 1);
 }
 
 void
@@ -247,10 +276,19 @@ page_manager::map_offset_pointer(void **pointer, size_t length)
 }
 
 void
-page_manager::dump_blocks()
+page_manager::dump_blocks(bool used_only)
 {
 	page_block::dump(m_used, "used", true);
-	page_block::dump(m_free, "free", true);
+	if (!used_only)
+		page_block::dump(m_free, "free", true);
+}
+
+void
+page_manager::dump_pml4(page_table *pml4, int max_index)
+{
+	if (pml4 == nullptr)
+		pml4 = get_pml4();
+	pml4->dump(4, max_index);
 }
 
 page_block *
@@ -338,10 +376,11 @@ page_manager::consolidate_blocks()
 }
 
 void *
-page_manager::map_pages(addr_t address, size_t count)
+page_manager::map_pages(addr_t address, size_t count, bool user, page_table *pml4)
 {
 	void *ret = reinterpret_cast<void *>(address);
-	page_table *pml4 = get_pml4();
+	if (pml4 == nullptr)
+		pml4 = get_pml4();
 
 	while (count) {
 		kassert(m_free, "page_manager::map_pages ran out of free pages!");
@@ -358,7 +397,10 @@ page_manager::map_pages(addr_t address, size_t count)
 				page_block_flags::mapped;
 		page_block::insert(m_used, block);
 
-		page_in(pml4, phys, address, n);
+		log::debug(logs::memory, "Paging in %d pages at p:%016lx to v:%016lx into %016lx table",
+				n, phys, address, pml4);
+
+		page_in(pml4, phys, address, n, user);
 
 		address += n * page_size;
 		count -= n;
@@ -374,7 +416,6 @@ page_manager::map_offset_pages(size_t count)
 	page_block *free = m_free;
 	page_block *prev = nullptr;
 
-	log::debug(logs::memory, "Got request to offset map %d pages", count);
 
 	while (free) {
 		if (free->count < count) {
@@ -405,6 +446,7 @@ page_manager::map_offset_pages(size_t count)
 			m_block_cache = free;
 		}
 
+		log::debug(logs::memory, "Got request for offset map %016lx [%d]", used->virtual_address, count);
 		page_in(pml4, used->physical_address, used->virtual_address, count);
 		return reinterpret_cast<void *>(used->virtual_address);
 	}
@@ -480,36 +522,47 @@ page_manager::unmap_pages(void* address, size_t count)
 }
 
 void
-page_manager::check_needs_page(page_table *table, unsigned index)
+page_manager::check_needs_page(page_table *table, unsigned index, bool user)
 {
 	if ((table->entries[index] & 0x1) == 1) return;
 
 	page_table *new_table = get_table_page();
 	for (int i=0; i<512; ++i) new_table->entries[i] = 0;
-	table->entries[index] = pt_to_phys(new_table) | 0xb;
+	table->entries[index] = pt_to_phys(new_table) | (user ? 0xf : 0xb);
 }
 
 void
-page_manager::page_in(page_table *pml4, addr_t phys_addr, addr_t virt_addr, size_t count)
+page_manager::page_in(page_table *pml4, addr_t phys_addr, addr_t virt_addr, size_t count, bool user)
 {
 	page_table_indices idx{virt_addr};
 	page_table *tables[4] = {pml4, nullptr, nullptr, nullptr};
 
 	for (; idx[0] < 512; idx[0] += 1) {
-		check_needs_page(tables[0], idx[0]);
+		check_needs_page(tables[0], idx[0], user);
 		tables[1] = tables[0]->get(idx[0]);
 
 		for (; idx[1] < 512; idx[1] += 1, idx[2] = 0, idx[3] = 0) {
-			check_needs_page(tables[1], idx[1]);
+			check_needs_page(tables[1], idx[1], user);
 			tables[2] = tables[1]->get(idx[1]);
 
 			for (; idx[2] < 512; idx[2] += 1, idx[3] = 0) {
-				check_needs_page(tables[2], idx[2]);
+				if (idx[3] == 0 &&
+					count >= 512 &&
+					tables[2]->get(idx[2]) == nullptr) {
+					// Do a 2MiB page instead
+					tables[2]->entries[idx[2]] = phys_addr | (user ? 0x8f : 0x8b);
+					phys_addr += page_size * 512;
+					count -= 512;
+					if (count == 0) return;
+					continue;
+				}
+
+				check_needs_page(tables[2], idx[2], user);
 				tables[3] = tables[2]->get(idx[2]);
 
 				for (; idx[3] < 512; idx[3] += 1) {
-					tables[3]->entries[idx[3]] = phys_addr | 0xb;
-					phys_addr += page_manager::page_size;
+					tables[3]->entries[idx[3]] = phys_addr | (user ? 0xf : 0xb);
+					phys_addr += page_size;
 					if (--count == 0) return;
 				}
 			}
@@ -571,40 +624,40 @@ page_manager::pop_pages(size_t count, addr_t *address)
 
 
 void
-page_table::dump(int level, uint64_t offset)
+page_table::dump(int level, int max_index, uint64_t offset)
 {
-	log::info(logs::memory, "Level %d page table @ %lx (off %lx):", level, this, offset);
+	console *cons = console::get();
+
+	cons->printf("\nLevel %d page table @ %lx (off %lx):\n", level, this, offset);
 	for (int i=0; i<512; ++i) {
 		uint64_t ent = entries[i];
 		if (ent == 0) continue;
 
 		if ((ent & 0x1) == 0) {
-			log::info(logs::memory, "  %3d: %lx   NOT PRESENT", i, ent);
+			cons->printf("  %3d: %016lx   NOT PRESENT\n", i, ent);
 			continue;
 		}
 
 		if ((level == 2 || level == 3) && (ent & 0x80) == 0x80) {
-			log::info(logs::memory, "  %3d: %lx   -> Large page at     %lx",
-					i, ent, ent & ~0xfffull);
+			cons->printf("  %3d: %016lx -> Large page at    %016lx\n", i, ent, ent & ~0xfffull);
 			continue;
 		} else if (level == 1) {
-			log::info(logs::memory, "  %3d: %lx   -> Page at           %lx",
-					i, ent, ent & ~0xfffull);
+			cons->printf("  %3d: %016lx -> Page at          %016lx\n", i, ent, ent & ~0xfffull);
 		} else {
-			log::info(logs::memory, "  %3d: %lx   -> Level %d table at %lx",
+			cons->printf("  %3d: %016lx -> Level %d table at %016lx\n",
 					i, ent, level - 1, (ent & ~0xfffull) + offset);
 			continue;
 		}
 	}
 
 	if (--level > 0) {
-		for (int i=0; i<512; ++i) {
+		for (int i=0; i<=max_index; ++i) {
 			uint64_t ent = entries[i];
 			if ((ent & 0x1) == 0) continue;
 			if ((ent & 0x80)) continue;
 
 			page_table *next = reinterpret_cast<page_table *>((ent & ~0xffful) + offset); 
-			next->dump(level, offset);
+			next->dump(level, 511, offset);
 		}
 	}
 }

src/kernel/page_manager.h

@@ -21,18 +21,51 @@ public:
 	static const size_t page_size = 0x1000;
 
 	/// Start of the higher half.
-	static const addr_t high_offset = 0xffff800000000000;
+	static const addr_t high_offset = 0xffffff0000000000;
 
 	/// Offset from physical where page tables are mapped.
 	static const addr_t page_offset = 0xffffff8000000000;
 
+	/// Initial process thread's stack address
+	static const addr_t initial_stack = 0x0000800000000000;
+
+	/// Initial process thread's stack size, in pages
+	static const unsigned initial_stack_pages = 1;
+
 	page_manager();
 
+	/// Helper to read the PML4 table from CR3.
+	/// \returns  A pointer to the current PML4 table.
+	static inline page_table * get_pml4()
+	{
+		addr_t pml4 = 0;
+		__asm__ __volatile__ ( "mov %%cr3, %0" : "=r" (pml4) );
+		return reinterpret_cast<page_table *>((pml4 & ~0xfffull) + page_offset);
+	}
+
+	/// Helper to set the PML4 table pointer in CR3.
+	/// \arg pml4  A pointer to the PML4 table to install.
+	static inline void set_pml4(page_table *pml4)
+	{
+		addr_t p = reinterpret_cast<addr_t>(pml4) - page_offset;
+		__asm__ __volatile__ ( "mov %0, %%cr3" :: "r" (p & ~0xfffull) );
+	}
+
+	/// Allocate but don't switch to a new PML4 table. This table
+	/// should only have global kernel pages mapped.
+	/// \returns      A pointer to the PML4 table
+	page_table * create_process_map();
+
+	/// Deallocate a process' PML4 table.
+	void delete_process_map(page_table *table);
+
 	/// Allocate and map pages into virtual memory.
 	/// \arg address  The virtual address at which to map the pages
 	/// \arg count    The number of pages to map
+	/// \arg user     True is this memory is user-accessible
+	/// \arg pml4     The pml4 to map into - null for the current one
 	/// \returns      A pointer to the start of the mapped region
-	void * map_pages(addr_t address, size_t count);
+	void * map_pages(addr_t address, size_t count, bool user = false, page_table *pml4 = nullptr);
 
 	/// Allocate and map contiguous pages into virtual memory, with
 	/// a constant offset from their physical address.
@@ -68,7 +101,13 @@ public:
 	}
 
 	/// Log the current free/used block lists.
-	void dump_blocks();
+	/// \arg used_only  If true, skip printing free list. Default false.
+	void dump_blocks(bool used_only = false);
+
+	/// Dump the given or current PML4 to the console
+	/// \arg pml4       The page table to use, null for the current one
+	/// \arg max_index  The max index of pml4 to print
+	void dump_pml4(page_table *pml4 = nullptr, int max_index = 511);
 
 	/// Get the system page manager.
 	/// \returns  A pointer to the system page manager
@@ -105,40 +144,26 @@ private:
 	/// to the cache.
 	void consolidate_blocks();
 
-	/// Helper to read the PML4 table from CR3.
-	/// \returns  A pointer to the current PML4 table.
-	static inline page_table * get_pml4()
-	{
-		addr_t pml4 = 0;
-		__asm__ __volatile__ ( "mov %%cr3, %0" : "=r" (pml4) );
-		return reinterpret_cast<page_table *>((pml4 & ~0xfffull) + page_offset);
-	}
-
-	/// Helper to set the PML4 table pointer in CR3.
-	/// \arg pml4  A pointer to the PML4 table to install.
-	static inline void set_pml4(page_table *pml4)
-	{
-		addr_t p = reinterpret_cast<addr_t>(pml4) - page_offset;
-		__asm__ __volatile__ ( "mov %0, %%cr3" :: "r" (p & ~0xfffull) );
-	}
-
 	/// Helper function to allocate a new page table. If table entry `i` in
 	/// table `base` is empty, allocate a new page table and point `base[i]` at
 	/// it.
 	/// \arg base  Existing page table being indexed into
 	/// \arg i     Index into the existing table to check
-	void check_needs_page(page_table *base, unsigned i);
+	/// \art user  True if this is a userspace mapping
+	void check_needs_page(page_table *base, unsigned i, bool user);
 
 	/// Low-level routine for mapping a number of pages into the given page table.
 	/// \arg pml4       The root page table to map into
 	/// \arg phys_addr  The starting physical address of the pages to be mapped
 	/// \arg virt_addr  The starting virtual address ot the memory to be mapped
 	/// \arg count      The number of pages to map
+	/// \art user  True if this is a userspace mapping
 	void page_in(
 			page_table *pml4,
 			addr_t phys_addr,
 			addr_t virt_addr,
-			size_t count);
+			size_t count,
+			bool user = true);
 
 	/// Low-level routine for unmapping a number of pages from the given page table.
 	/// \arg pml4       The root page table for this mapping
@@ -157,6 +182,8 @@ private:
 	/// \returns      The number of pages retrieved
 	size_t pop_pages(size_t count, addr_t *address);
 
+	page_table *m_kernel_pml4; ///< The PML4 of just kernel pages
+
 	page_block *m_free; ///< Free pages list
 	page_block *m_used; ///< In-use pages list
 
@@ -278,7 +305,7 @@ struct page_table
 		entries[i] = (reinterpret_cast<uint64_t>(p) - pm::page_offset) | (flags & 0xfff);
 	}
 
-	void dump(int level = 4, uint64_t offset = page_manager::page_offset);
+	void dump(int level = 4, int max_index = 511, uint64_t offset = page_manager::page_offset);
 };
 
 

src/kernel/pci.cpp

@@ -1,11 +1,32 @@
 #include "kutil/assert.h"
+#include "console.h"
 #include "log.h"
 #include "interrupts.h"
 #include "pci.h"
 
-struct pci_cap_msi :
-	public pci_cap
+struct pci_cap_msi
 {
+	pci_cap::type id;
+	uint8_t next;
+	uint16_t control;
+} __attribute__ ((packed));
+
+struct pci_cap_msi32
+{
+	pci_cap::type id;
+	uint8_t next;
+	uint16_t control;
+	uint32_t address;
+	uint16_t data;
+	uint16_t reserved;
+	uint32_t mask;
+	uint32_t pending;
+} __attribute__ ((packed));
+
+struct pci_cap_msi64
+{
+	pci_cap::type id;
+	uint8_t next;
 	uint16_t control;
 	uint64_t address;
 	uint16_t data;
@@ -14,17 +35,33 @@ struct pci_cap_msi :
 	uint32_t pending;
 } __attribute__ ((packed));
 
-struct pci_cap_msix :
-	public pci_cap
-{
-	uint16_t control;
-	uint64_t address;
-	uint16_t data;
-	uint16_t reserved;
-	uint32_t mask;
-	uint32_t pending;
-} __attribute__ ((packed));
 
+void dump_msi(pci_cap_msi *cap)
+{
+	auto cons = console::get();
+	cons->printf("MSI Cap:\n");
+	cons->printf("     id: %02x\n", cap->id);
+	cons->printf("   next: %02x\n", cap->next);
+	cons->printf("control: %04x\n", cap->control);
+	if (cap->control & 0x0080) {
+		pci_cap_msi64 *cap64 = reinterpret_cast<pci_cap_msi64 *>(cap);
+		cons->printf("address: %016x\n", cap64->address);
+		cons->printf("   data: %04x\n", cap64->data);
+		if (cap->control & 0x100) {
+			cons->printf("   mask: %08x\n", cap64->mask);
+			cons->printf("pending: %08x\n", cap64->pending);
+		}
+	} else {
+		pci_cap_msi32 *cap32 = reinterpret_cast<pci_cap_msi32 *>(cap);
+		cons->printf("address: %08x\n", cap32->address);
+		cons->printf("   data: %04x\n", cap32->data);
+		if (cap->control & 0x100) {
+			cons->printf("   mask: %08x\n", cap32->mask);
+			cons->printf("pending: %08x\n", cap32->pending);
+		}
+	}
+	cons->putc('\n');
+};
 
 pci_device::pci_device() :
 	m_base(nullptr),
@@ -60,19 +97,25 @@ pci_device::pci_device(pci_group &group, uint8_t bus, uint8_t device, uint8_t fu
 	uint16_t *command = reinterpret_cast<uint16_t *>(&m_base[1]);
 	*command |= 0x400; // Mask old INTx style interrupts
 
-	log::info(logs::device, "Found PCIe device at %02d:%02d:%d of type %d.%d id %04x:%04x",
-			bus, device, func, m_class, m_subclass, m_vendor, m_device);
+	uint16_t *status = command + 1;
 
-	// Walk the extended capabilities list
-	uint8_t next = m_base[13] & 0xff;
-	while (next) {
-		pci_cap *cap = reinterpret_cast<pci_cap *>(kutil::offset_pointer(m_base, next));
-		next = cap->next;
+	log::info(logs::device, "Found PCIe device at %02d:%02d:%d of type %x.%x.%x id %04x:%04x",
+			bus, device, func, m_class, m_subclass, m_progif, m_vendor, m_device);
 
-		if (cap->id == pci_cap::type::msi) {
-			m_msi = cap;
-			pci_cap_msi *mcap = reinterpret_cast<pci_cap_msi *>(cap);
-			mcap->control |= ~0x1; // Mask interrupts
+	if (*status & 0x0010) {
+		// Walk the extended capabilities list
+		uint8_t next = m_base[13] & 0xff;
+		while (next) {
+			pci_cap *cap = reinterpret_cast<pci_cap *>(kutil::offset_pointer(m_base, next));
+			next = cap->next;
+			log::debug(logs::device, "  - found PCI cap type %02x", cap->id);
+
+			if (cap->id == pci_cap::type::msi) {
+				m_msi = cap;
+				pci_cap_msi *mcap = reinterpret_cast<pci_cap_msi *>(cap);
+				mcap->control &= ~0x70; // at most 1 vector allocated
+				mcap->control |= 0x01; // Enable interrupts, at most 1 vector allocated
+			}
 		}
 	}
 }
@@ -111,9 +154,19 @@ pci_device::write_msi_regs(addr_t address, uint16_t data)
 	kassert(m_msi, "Tried to write MSI for a device without that cap");
 	if (m_msi->id == pci_cap::type::msi) {
 		pci_cap_msi *mcap = reinterpret_cast<pci_cap_msi *>(m_msi);
-		mcap->address = address;
-		mcap->data = data;
-		mcap->control |= 1;
+		if (mcap->control & 0x0080) {
+			pci_cap_msi64 *mcap64 = reinterpret_cast<pci_cap_msi64 *>(m_msi);
+			mcap64->address = address;
+			mcap64->data = data;
+		} else {
+			pci_cap_msi32 *mcap32 = reinterpret_cast<pci_cap_msi32 *>(m_msi);
+			mcap32->address = address;
+			mcap32->data = data;
+		}
+		uint16_t control = mcap->control;
+		control &= 0xff8f; // We're allocating one vector, clear 6::4
+		control |= 0x0001; // Enable MSI
+		mcap->control = control;
 	} else {
 		kassert(0, "MIS-X is NYI");
 	}

src/kernel/scheduler.cpp

@@ -0,0 +1,90 @@
+#include "apic.h"
+#include "console.h"
+#include "cpu.h"
+#include "gdt.h"
+#include "interrupts.h"
+#include "log.h"
+#include "page_manager.h"
+#include "scheduler.h"
+
+scheduler scheduler::s_instance(nullptr);
+//static const uint32_t quantum = 2000000;
+static const uint32_t quantum =  20000000;
+
+const int stack_size = 0x1000;
+
+extern "C" void taskA();
+extern "C" void taskB();
+
+
+scheduler::scheduler(lapic *apic) :
+	m_apic(apic),
+	m_current(0)
+{
+	m_processes.ensure_capacity(50);
+}
+
+static process
+create_process(uint16_t pid, void (*rip)())
+{
+	uint64_t flags;
+	__asm__ __volatile__ ( "pushf; pop %0" : "=r" (flags) );
+
+	uint16_t kcs = (1 << 3) | 0; // Kernel CS is GDT entry 1, ring 0
+	uint16_t cs = (5 << 3) | 3;  // User CS is GDT entry 5, ring 3
+
+	uint16_t kss = (2 << 3) | 0; // Kernel SS is GDT entry 2, ring 0
+	uint16_t ss = (4 << 3) | 3;  // User SS is GDT entry 4, ring 3
+
+	void *stack0 = kutil::malloc(stack_size);
+	void *sp0 = kutil::offset_pointer(stack0, stack_size);
+	cpu_state *state = reinterpret_cast<cpu_state *>(sp0) - 1;
+	kutil::memset(state, 0, sizeof(cpu_state));
+
+	state->ds = state->ss = ss;
+	state->cs = cs;
+	state->rflags = 0x202;
+	state->rip = reinterpret_cast<uint64_t>(rip);
+
+	page_table *pml4 = page_manager::get()->create_process_map();
+	state->user_rsp = page_manager::initial_stack;
+
+	log::debug(logs::task, "Creating PID %d:", pid);
+	log::debug(logs::task, "  RSP0 %016lx", state);
+	log::debug(logs::task, "  RSP3 %016lx", state->user_rsp);
+	log::debug(logs::task, "  PML4 %016lx", pml4);
+
+	return {pid, reinterpret_cast<addr_t>(state), pml4};
+}
+
+void
+scheduler::start()
+{
+	log::info(logs::task, "Starting scheduler.");
+
+	m_apic->enable_timer(isr::isrTimer, 128, quantum, false);
+
+	m_processes.append({0, 0, page_manager::get_pml4()}); // The kernel idle task, also the thread we're in now
+	m_processes.append(create_process(1, &taskA));
+	m_processes.append(create_process(2, &taskB));
+}
+
+addr_t
+scheduler::tick(addr_t rsp0)
+{
+	log::debug(logs::task, "Scheduler tick.");
+
+	m_processes[m_current].rsp = rsp0;
+	m_current = (m_current + 1) % m_processes.count();
+	rsp0 = m_processes[m_current].rsp;
+
+	// Set rsp0 to after the end of the about-to-be-popped cpu state
+	tss_set_stack(0, rsp0 + sizeof(cpu_state));
+
+	// Swap page tables
+	page_table *pml4 = m_processes[m_current].pml4;
+	page_manager::set_pml4(pml4);
+
+	m_apic->reset_timer(quantum);
+	return rsp0;
+}

src/kernel/scheduler.h

@@ -0,0 +1,46 @@
+#pragma once
+/// \file scheduler.h
+/// The task scheduler and related definitions
+#include "kutil/memory.h"
+#include "kutil/vector.h"
+
+class lapic;
+struct page_table;
+
+
+struct process
+{
+	uint16_t pid;
+	addr_t rsp;
+	page_table *pml4;
+};
+
+
+/// The task scheduler
+class scheduler
+{
+public:
+	/// Constructor.
+	/// \arg apic  Pointer to the local APIC object
+	scheduler(lapic *apic);
+
+	/// Start the scheduler working. This may involve starting
+	/// timer interrupts or other preemption methods.
+	void start();
+
+	/// Handle a timer tick
+	/// \arg rsp0  The stack pointer of the current interrupt handler
+	/// \returns   The stack pointer to handler to switch to
+	addr_t tick(addr_t rsp0);
+
+	/// Get a reference to the system scheduler
+	/// \returns  A reference to the global system scheduler
+	static scheduler & get() { return s_instance; }
+
+private:
+	lapic *m_apic;
+	kutil::vector<process> m_processes;
+	uint16_t m_current;
+
+	static scheduler s_instance;
+};

src/kernel/wscript

@@ -29,7 +29,11 @@ def build(bld):
             return node.path_from(node.ctx.launch_node())
         def run(self):
             from subprocess import check_output
-            args = self.env.objdump + ["--source", "-D", self.inputs[0].abspath()]
+            args = self.env.objdump + [
+                    "--source",
+                    "-D",
+                    "-M", "intel",
+                    self.inputs[0].abspath()]
             with file(self.outputs[0].abspath(), 'w') as output:
                 output.write(check_output(args))
 

src/modules/kutil/enum_bitfields.h

@@ -8,27 +8,27 @@ struct is_enum_bitfield { static constexpr bool value = false; };
 #define IS_BITFIELD(name) \
 	template<> struct ::is_enum_bitfield<name> {static constexpr bool value=true;}
 
-template <typename E>
+template <typename E, typename F>
 typename std::enable_if<is_enum_bitfield<E>::value,E>::type
-operator & (E lhs, E rhs)
+operator & (E lhs, F rhs)
 {
 	return static_cast<E> (
 		static_cast<typename std::underlying_type<E>::type>(lhs) &
 		static_cast<typename std::underlying_type<E>::type>(rhs));
 }
 
-template <typename E>
+template <typename E, typename F>
 typename std::enable_if<is_enum_bitfield<E>::value,E>::type
-operator | (E lhs, E rhs)
+operator | (E lhs, F rhs)
 {
 	return static_cast<E> (
 		static_cast<typename std::underlying_type<E>::type>(lhs) |
 		static_cast<typename std::underlying_type<E>::type>(rhs));
 }
 
-template <typename E>
+template <typename E, typename F>
 typename std::enable_if<is_enum_bitfield<E>::value,E>::type
-operator ^ (E lhs, E rhs)
+operator ^ (E lhs, F rhs)
 {
 	return static_cast<E> (
 		static_cast<typename std::underlying_type<E>::type>(lhs) ^
@@ -42,9 +42,9 @@ operator ~ (E rhs)
 	return static_cast<E>(~static_cast<typename std::underlying_type<E>::type>(rhs));
 }
 
-template <typename E>
+template <typename E, typename F>
 typename std::enable_if<is_enum_bitfield<E>::value,E>::type&
-operator |= (E &lhs, E rhs)
+operator |= (E &lhs, F rhs)
 {
 	lhs = static_cast<E>(
 		static_cast<typename std::underlying_type<E>::type>(lhs) |
@@ -53,9 +53,9 @@ operator |= (E &lhs, E rhs)
 	return lhs;
 }
 
-template <typename E>
+template <typename E, typename F>
 typename std::enable_if<is_enum_bitfield<E>::value,E>::type&
-operator &= (E &lhs, E rhs)
+operator &= (E &lhs, F rhs)
 {
 	lhs = static_cast<E>(
 		static_cast<typename std::underlying_type<E>::type>(lhs) &
@@ -64,9 +64,9 @@ operator &= (E &lhs, E rhs)
 	return lhs;
 }
 
-template <typename E>
+template <typename E, typename F>
 typename std::enable_if<is_enum_bitfield<E>::value,E>::type&
-operator ^= (E &lhs, E rhs)
+operator ^= (E &lhs, F rhs)
 {
 	lhs = static_cast<E>(
 		static_cast<typename std::underlying_type<E>::type>(lhs) ^

src/modules/kutil/guid.h

@@ -0,0 +1,38 @@
+#pragma once
+/// \file guid.h
+/// Definition of the guid type and related functions
+#include <stdint.h>
+#include "kutil/misc.h"
+
+namespace kutil {
+
+
+/// A GUID
+struct guid
+{
+	uint64_t a, b;
+};
+
+/// Make a GUID by writing it naturally-ordered in code:
+///  AAAAAAAA-BBBB-CCCC-DDDD-EEEEEEEEEEEE becomes:
+///  make_guid(0xAAAAAAAA, 0xBBBB, 0xCCCC, 0xDDDD, 0xEEEEEEEEEEEE);
+/// \returns  The guid object
+inline constexpr guid make_guid(uint32_t a, uint16_t b, uint16_t c, uint16_t d, uint64_t e)
+{
+	const uint64_t h =
+		static_cast<uint64_t>(c) << 48 |
+		static_cast<uint64_t>(b) << 32 |
+		a;
+
+	const uint64_t l =
+		static_cast<uint64_t>(byteswap(e & 0xffffffff)) << 32 |
+		(byteswap(e >> 32) & 0xffff0000) |
+		((d << 8) & 0xff00) | ((d >> 8) & 0xff);
+
+	return {h, l};
+}
+
+} // namespace kutil
+
+
+inline bool operator==(const kutil::guid &a, const kutil::guid &b) { return a.a == b.a && a.b == b.b; }

wscript

@@ -80,6 +80,7 @@ def configure(ctx):
         '-ffreestanding',
         '-nodefaultlibs',
         '-fno-builtin',
+        '-mno-sse',
         '-fno-omit-frame-pointer',
         '-mno-red-zone',
     ]
@@ -141,6 +142,20 @@ def configure(ctx):
         '-nostartfiles',
     ])
 
+    ctx.env.append_value('QEMUOPTS', [
+        '-smp', '1',
+        '-m', '512',
+        '-d', 'mmu,int,guest_errors',
+        '-D', 'popcorn.log',
+        '-cpu', 'Broadwell',
+        '-M', 'q35',
+        '-no-reboot',
+        '-nographic',
+    ])
+
+    if os.path.exists('/dev/kvm'):
+        ctx.env.append_value('QEMUOPTS', ['--enable-kvm'])
+
     env = ctx.env
     ctx.setenv('boot', env=env)
     ctx.recurse(join("src", "boot"))
@@ -268,6 +283,14 @@ def build(bld):
         bld.recurse(join("src", "tests"))
 
 
+class QemuContext(BuildContext):
+    cmd = 'qemu'
+    fun = 'qemu'
+
+class DebugQemuContext(QemuContext):
+    cmd = 'debug'
+    fun = 'qemu'
+
 def qemu(ctx):
     import subprocess
     subprocess.call("rm popcorn.log", shell=True)
@@ -275,24 +298,34 @@ def qemu(ctx):
         'qemu-system-x86_64',
         '-drive', 'if=pflash,format=raw,file={}/flash.img'.format(out),
         '-drive', 'format=raw,file={}/popcorn.img'.format(out),
-        '-smp', '1',
-        '-m', '512',
-        '-d', 'mmu,int,guest_errors',
-        '-D', 'popcorn.log',
-        '-cpu', 'Broadwell',
-        '-M', 'q35',
-        '-no-reboot',
-        '-nographic',
-    ])
+        ] + ctx.env.QEMUOPTS)
+
+def debug(ctx):
+    import subprocess
+    subprocess.call("rm popcorn.log", shell=True)
+    subprocess.call([
+        'qemu-system-x86_64',
+        '-drive', 'if=pflash,format=raw,file={}/flash.img'.format(out),
+        '-drive', 'format=raw,file={}/popcorn.img'.format(out),
+        '-s',
+        ] + ctx.env.QEMUOPTS)
 
 
 def vbox(ctx):
     import os
+    from os.path import join
     from shutil import copy
     from subprocess import call
 
-    dest = os.getenv("VBOX_DEST")
-    copy("{}/popcorn.img".format(out), "{}/popcorn.img".format(dest))
+    vbox_dest = os.getenv("VBOX_DEST")
+
+    ext = 'qed'
+    dest = join(vbox_dest, "popcorn.{}".format(ext))
+    call("qemu-img convert -f raw -O {} build/popcorn.img {}".format(ext, dest), shell=True)
+    copy(join("build", "src", "kernel", "popcorn.elf"), vbox_dest);
+
+def listen(ctx):
+    from subprocess import call
     call("nc -l -p 5555", shell=True)