Add an example user space process

build.sh

@@ -15,3 +15,4 @@ dd if=build/bootloader.bin of=build/os.img bs="${sector_size}" seek=1 count="$((
 mkfs.fat -F 16 --offset "${partition_offset}" build/os.img
 mcopy -i build/os.img@@"$((partition_offset * sector_size))" src ::src
 mcopy -i build/os.img@@"$((partition_offset * sector_size))" build/kernel.bin ::kernel.bin
+mcopy -i build/os.img@@"$((partition_offset * sector_size))" build/hello.bin ::hello.bin

meson.build

@@ -31,6 +31,14 @@ custom_target(
   build_by_default: true,
 )
 
+custom_target(
+  'hello',
+  input: 'src/app/hello.asm',
+  output: 'hello.bin',
+  command: [nasm, '-f', 'bin', '-l', 'hello.lst', '@INPUT@', '-o', '@OUTPUT@'],
+  build_by_default: true,
+)
+
 kernel_entry_o = custom_target(
   'kernel_entry',
   input: 'src/kernel_entry.asm',
@@ -38,6 +46,12 @@ kernel_entry_o = custom_target(
   command: [nasm, '-f', 'elf64', '@INPUT@', '-o', '@OUTPUT@'],
 )
 
+syscall_o = custom_target('syscall',
+  input:  'src/syscall.asm',
+  output: 'syscall.o',
+  command: [nasm, '-f', 'elf64', '@INPUT@', '-o', '@OUTPUT@'],
+)
+
 cc = meson.get_compiler('c')
 
 kernel_sources = files(
@@ -47,6 +61,7 @@ kernel_sources = files(
   'src/ata.c',
   'src/fat16.c',
   'src/fs.c',
+  'src/gdt.c',
   'src/idt.c',
   'src/kernel.c',
   'src/keyboard.c',
@@ -54,19 +69,23 @@ kernel_sources = files(
   'src/panic.c',
   'src/path.c',
   'src/pic.c',
+  'src/process.c',
   'src/string.c',
+  'src/syscall.c',
+  'src/tss.c',
   'src/vga.c',
 )
 
 kernel_elf = executable(
   'kernel.elf',
-  sources: [kernel_entry_o, kernel_sources],
+  sources: [kernel_entry_o, syscall_o, kernel_sources],
   c_args: [
     '-ffreestanding',
     '-nostdlib',
     '-nostartfiles',
     '-mno-red-zone',
     '-mgeneral-regs-only',
+    '-mcmodel=kernel',
     '-Wno-unused-command-line-argument',
     '-Wconversion',
     '-DVERSION="' + meson.project_version() + '"',

src/app/hello.asm

@@ -0,0 +1,16 @@
+bits 64
+org 0x400000
+
+section .text
+    mov rax, 1
+    mov rdi, 1
+    lea rsi, [rel msg]
+    mov rdx, 14
+    syscall
+
+    mov rax, 60
+    xor rdi, rdi
+    syscall
+
+section .data
+msg: db "Hello, ring 3!", 10

src/app/terminal.c

@@ -270,9 +270,7 @@ void terminal(process_t *p, __attribute__((unused)) uint8_t argc, __attribute__(
 
   proc = p;
 
-  char *greeting = string_format("Welcome to FreywarOS v%s!\n\n", VERSION);
-  WRITE_D(greeting);
-  memory_free(greeting);
+  WRITE_D("Welcome to FreywarOS v" VERSION "!\n\n");
 
   print_prompt();
 

src/bootloader.asm

@@ -224,30 +224,30 @@ jmp load_gdt
 
 gdt_start:
 gdt_0: dq 0x0000000000000000
-gdt_1:
+gdt_32_code:
 .limit_low:     dw 0xFFFF
 .base_low:      dw 0x0000
 .base_middle:   db 0x00
 .access:        db 10011010b ; present, ring 0, code, executable, readable
 .flags:         db 11001111b ; 32-bit, 4KB granularity
 .base_high:     db 0x00
-gdt_1_end:
-gdt_2:
+gdt_32_code_end:
+gdt_64_code:
 .limit_low:     dw 0xFFFF
 .base_low:      dw 0x0000
 .base_middle:   db 0x00
 .access:        db 10011010b ; present, ring 0, code, executable, readable
 .flags:         db 10101111b ; long mode (L bit)
 .base_high:     db 0x00
-gdt_2_end:
-gdt_3:
+gdt_64_code_end:
+gdt_data:
 .limit_low:     dw 0xFFFF
 .base_low:      dw 0x0000
 .base_middle:   db 0x00
 .access:        db 10010010b ; present, ring 0, data, writable
 .flags:         db 00000000b
 .base_high:     db 0x00
-gdt_3_end:
+gdt_data_end:
 gdt_end:
 
 gdt_descriptor:
@@ -276,23 +276,30 @@ protected_mode:
     mov ss, ax
     mov esp, 0x00090000
 
-    ; zero page table memory at 0x00010000 (3 pages = 768 dwords)
+    ; zero page table memory
     mov edi, 0x00010000
-    mov ecx, 768
+    mov ecx, 256 * 5
     xor eax, eax
     rep stosd
 
     ; PML4[0] -> PDPT at 0x00011000
     mov dword [0x00010000], 0x00011003
-    ; PDPT[0] -> PD at 0x00012000
-    mov dword [0x00011000], 0x00012003
+    ; PML4[511] -> PDPT at 0x00012000
+    mov dword [0x00010FF8], 0x00012003
+    ; PML4[0].PDPT[0] -> PD at 0x00013000
+    mov dword [0x00011000], 0x00013003
+    ; PML4[511].PDPT[510] -> PD at 0x00014000
+    mov dword [0x00012FF0], 0x00014003
     ; PD: fill 64 entries, each mapping 2MB
-    mov edi, 0x00012000
+    mov edi, 0x00013000
     mov eax, 0x00000083   ; present, writable, huge page, base 0
     mov ecx, 64
 
 .fill_pd:
+    mov ebx, edi
+    add ebx, 0x1000
     mov dword [edi], eax
+    mov dword [ebx], eax
     add edi, 8
     add eax, 0x00200000   ; next 2MB
     loop .fill_pd
@@ -327,11 +334,11 @@ long_mode:
     mov ds, ax
     mov es, ax
     mov ss, ax
-    mov rsp, 0x0000000000090000
+    mov rsp, 0xFFFFFFFF80090000
 
     xor rax, rax
     xor rbx, rbx
     xor rcx, rcx
     xor rdx, rdx
 
-    jmp 0x00020000
+    jmp 0xFFFFFFFF80020000

src/fs.h

@@ -11,7 +11,6 @@ typedef struct fs_node {
   uint8_t type;
 } fs_node_t;
 
-
 fs_node_t *fs_mount();
 
 fs_node_t *fs_open_by(const fs_node_t *directory, const char *name);

src/gdt.c

@@ -0,0 +1,68 @@
+#include "src/gdt.h"
+#include "src/tss.h"
+#include <stdint.h>
+
+typedef struct {
+  uint16_t limit_low;
+  uint16_t base_low;
+  uint8_t base_mid;
+  uint8_t access;
+  uint8_t flags_limit_high;
+  uint8_t base_high;
+} __attribute__((packed)) gdt_entry_t;
+
+typedef struct {
+  gdt_entry_t base;
+  uint32_t base_upper;
+  uint32_t reserved;
+} __attribute__((packed)) gdt_system_entry_t;
+
+typedef struct {
+  gdt_entry_t gdt[6];
+  gdt_system_entry_t tss_entry;
+} __attribute__((packed)) gdt_table_t;
+
+typedef struct {
+  uint16_t limit;
+  uint64_t offset;
+} __attribute__((packed)) gdt_descriptor_t;
+
+static gdt_table_t gdt;
+static gdt_descriptor_t desc;
+
+static void set_entry(gdt_entry_t *e, uint8_t access, uint8_t flags) {
+  e->limit_low = 0xFFFF;
+  e->base_low = 0;
+  e->base_mid = 0;
+  e->access = access;
+  e->flags_limit_high = flags | 0x0F;
+  e->base_high = 0;
+}
+
+static void set_tss_entry(void *base) {
+  uint64_t ibase = (uint64_t)base;
+  gdt.tss_entry.base.limit_low = sizeof(tss_t) - 1;
+  gdt.tss_entry.base.base_low = ibase & 0xFFFF;
+  gdt.tss_entry.base.base_mid = (ibase >> 16) & 0xFF;
+  gdt.tss_entry.base.access = 0x89; // present, type=TSS available
+  gdt.tss_entry.base.flags_limit_high = 0x00;
+  gdt.tss_entry.base.base_high = (ibase >> 24) & 0xFF;
+  gdt.tss_entry.base_upper = ibase >> 32;
+  gdt.tss_entry.reserved = 0;
+}
+
+void gdt_init() {
+  gdt.gdt[0] = (gdt_entry_t){0};
+  set_entry(&gdt.gdt[1], 0x9A, 0xCF); // present, ring 0, code, executable, readable, 32-bit, 4KB granularity
+  set_entry(&gdt.gdt[2], 0x9A, 0xA0); // present, ring 0, code, executable, readable, 64-bit
+  set_entry(&gdt.gdt[3], 0x92, 0x00); // present, ring 0, data, writable
+  set_entry(&gdt.gdt[4], 0xF2, 0x00); // present, ring 3, data, writable
+  set_entry(&gdt.gdt[5], 0xFA, 0xA0); // present, ring 3, code, executable, readable, 64-bit
+  set_tss_entry(&tss);
+
+  desc.limit = sizeof(gdt) - 1;
+  desc.offset = (uint64_t)&gdt;
+
+  __asm__ volatile("lgdt %0" : : "m"(desc));
+  __asm__ volatile("ltr %0" : : "r"((uint16_t)TSS_SEL));
+}

src/gdt.h

@@ -0,0 +1,12 @@
+#pragma once
+
+#define OBSOLETE_CS 0x08
+#define KERNEL_CS (OBSOLETE_CS + 0x08)
+#define KERNEL_DS (KERNEL_CS + 0x08)
+#define USER_DS_BASE (KERNEL_DS + 0x08)
+#define USER_DS (USER_DS_BASE | 3)
+#define USER_CS_BASE (USER_DS_BASE + 0x08)
+#define USER_CS (USER_CS_BASE | 3)
+#define TSS_SEL (USER_CS_BASE + 0x08)
+
+void gdt_init();

src/kernel.c

@@ -1,11 +1,13 @@
-#include "src/app/terminal.h"
 #include "src/fs.h"
+#include "src/gdt.h"
 #include "src/idt.h"
 #include "src/keyboard.h"
 #include "src/memory.h"
+#include "src/panic.h"
 #include "src/path.h"
 #include "src/pic.h"
 #include "src/process.h"
+#include "src/syscall.h"
 #include "src/util.h"
 #include "src/vga.h"
 
@@ -42,15 +44,34 @@ void kernel_main() {
   idt_set_entry(46, isr_ata_primary, 0x8E);
   __asm__ volatile("sti");
 
+  gdt_init();
+  syscall_init();
+
   memory_init();
 
-  process_t *proc = memory_allocate(sizeof(process_t));
-  proc->cwd = memory_allocate(sizeof(path_t));
-  proc->root = fs_mount();
-  proc->stdin = keyboard_init();
-  proc->stdout = vga_init();
+  process_t *kp = memory_allocate(sizeof(process_t));
+  kp->root = fs_mount();
+  kp->cwd = memory_allocate(sizeof(path_t));
+  kp->stdin = keyboard_init();
+  kp->stdout = vga_init();
 
-  terminal(proc, 0, NUL);
+  fs_node_t *hello = fs_open_by(kp->root, "HELLO.BIN");
+
+  ASSERT(hello, "kernel: hello.bin not found");
+
+  uint8_t *hello_code = memory_allocate(hello->size);
+  fs_read(hello, 0, hello->size, hello_code);
+
+  process_t *proc = process_create(kp, hello_code, hello->size);
+
+  memory_free(hello_code);
+  fs_close(hello);
+
+  process_run(proc);
+
+  while (1);
+
+  fs_unmount(kp->root);
 
   outw(0x604, 0x2000); // TODO: parse ACPI tables for real hardware
   __asm__ volatile("cli; hlt");

src/linker.ld

@@ -1,7 +1,7 @@
 ENTRY(kernel_main)
 
 SECTIONS {
-    . = 0x0000000000020000;
+    . = 0xFFFFFFFF80020000;
 
     .text : {
         *(.text)

src/memory.c

@@ -1,11 +1,7 @@
 #include "src/memory.h"
 #include "src/panic.h"
 #include "src/util.h"
-#include <stdint.h>
 
-#define MEMORY_SIZE 134217728
-#define PAGE_SIZE 4096
-#define PAGE_COUNT (MEMORY_SIZE / PAGE_SIZE)
 #define MAP_UNIT 64
 #define FULL_UNIT 0xFFFFFFFFFFFFFFFF
 
@@ -26,7 +22,7 @@ static void set_allocated(uint16_t page, uint8_t allocated) {
   }
 }
 
-static void *memory_page_allocate() {
+void *memory_page_allocate() {
   ASSERT(free_page < PAGE_COUNT, "memory_page_allocate: out of memory");
   const uint16_t page = free_page;
   set_allocated(page, 1);
@@ -40,7 +36,7 @@ static void *memory_page_allocate() {
   return (void *)((uint64_t)page * PAGE_SIZE);
 }
 
-__attribute__((unused)) static void memory_page_free(void *address) {
+void memory_page_free(void *address) {
   uint16_t page = (uint16_t)((uint64_t)address / PAGE_SIZE);
   ASSERT(get_allocated(page), "memory_page_free: page not allocated")
   set_allocated(page, 0);
@@ -49,7 +45,65 @@ __attribute__((unused)) static void memory_page_free(void *address) {
   }
 }
 
-#define HEAP_PAGE_COUNT 256
+void memory_page_map(uint64_t *pml4, void *virt, void *phys, uint64_t flags) {
+  const uint64_t ivirt = (uint64_t)virt;
+  const uint64_t pml4_index = (ivirt >> 39) & 0x1FF;
+  const uint64_t pdpt_index = (ivirt >> 30) & 0x1FF;
+  const uint64_t pd_index = (ivirt >> 21) & 0x1FF;
+  const uint64_t pt_index = (ivirt >> 12) & 0x1FF;
+
+  if (!(pml4[pml4_index] & PAGE_PRESENT)) {
+    uint64_t *pdpt = memory_page_allocate();
+    memory_set(0, PAGE_SIZE, PHYS_TO_VIRT(pdpt));
+    pml4[pml4_index] = (uint64_t)pdpt | PAGE_PRESENT | PAGE_WRITABLE | PAGE_USER;
+  }
+
+  uint64_t *pdpt = PHYS_TO_VIRT(pml4[pml4_index] & ~(uint64_t)0xFFF);
+  if (!(pdpt[pdpt_index] & PAGE_PRESENT)) {
+    uint64_t *pd = memory_page_allocate();
+    memory_set(0, PAGE_SIZE, PHYS_TO_VIRT(pd));
+    pdpt[pdpt_index] = (uint64_t)pd | PAGE_PRESENT | PAGE_WRITABLE | PAGE_USER;
+  }
+
+  uint64_t *pd = PHYS_TO_VIRT(pdpt[pdpt_index] & ~(uint64_t)0xFFF);
+  ASSERT(!(pd[pd_index] & 0x80), "memory_page_map: huge page in PD");
+  if (!(pd[pd_index] & PAGE_PRESENT)) {
+    uint64_t *pt = memory_page_allocate();
+    memory_set(0, PAGE_SIZE, PHYS_TO_VIRT(pt));
+    pd[pd_index] = (uint64_t)pt | PAGE_PRESENT | PAGE_WRITABLE | PAGE_USER;
+  }
+
+  uint64_t *pt = PHYS_TO_VIRT(pd[pd_index] & ~(uint64_t)0xFFF);
+  pt[pt_index] = (uint64_t)phys | flags | PAGE_PRESENT;
+}
+
+void memory_page_unmap(uint64_t *pml4, void *virt) {
+  const uint64_t ivirt = (uint64_t)virt;
+  const uint64_t pml4_index = (ivirt >> 39) & 0x1FF;
+  const uint64_t pdpt_index = (ivirt >> 30) & 0x1FF;
+  const uint64_t pd_index = (ivirt >> 21) & 0x1FF;
+  const uint64_t pt_index = (ivirt >> 12) & 0x1FF;
+
+  if (!(pml4[pml4_index] & PAGE_PRESENT)) {
+    return;
+  }
+
+  uint64_t *pdpt = PHYS_TO_VIRT(pml4[pml4_index] & ~(uint64_t)0xFFF);
+  if (!(pdpt[pdpt_index] & PAGE_PRESENT)) {
+    return;
+  }
+
+  uint64_t *pd = PHYS_TO_VIRT(pdpt[pdpt_index] & ~(uint64_t)0xFFF);
+  if (!(pd[pd_index] & PAGE_PRESENT)) {
+    return;
+  }
+
+  uint64_t *pt = PHYS_TO_VIRT(pd[pd_index] & ~(uint64_t)0xFFF);
+  memory_page_free((void *)(pt[pt_index] & ~(uint64_t)0xFFF));
+  pt[pt_index] = 0;
+
+  __asm__ volatile("invlpg (%0)" : : "r"(virt) : "memory");
+}
 
 typedef struct free_chunk {
   uint64_t size; // header + data
@@ -59,9 +113,9 @@ typedef struct free_chunk {
 static free_chunk_t *free_list;
 
 void memory_init() {
-  free_list = memory_page_allocate();
+  free_list = PHYS_TO_VIRT(memory_page_allocate());
   for (uint64_t i = 1; i < HEAP_PAGE_COUNT; i++) {
-    memory_page_allocate(); // TODO Map the pages.
+    memory_page_allocate();
   }
   free_list->size = HEAP_PAGE_COUNT * PAGE_SIZE;
   free_list->next = NUL;

src/memory.h

@@ -3,8 +3,44 @@
 #include "src/panic.h"
 #include <stdint.h>
 
+#define PAGE_PRESENT 0x01
+#define PAGE_WRITABLE 0x02
+#define PAGE_USER 0x04
+#define PAGE_PWT 0x08
+#define PAGE_PCD 0x10
+#define PAGE_ACCESSED 0x20
+#define PAGE_DIRTY 0x40
+#define PAGE_HUGE 0x80
+#define PAGE_GLOBAL 0x100
+#define PAGE_NX (1ULL << 63)
+
+#define MEMORY_SIZE 0x08000000
+#define PAGE_SIZE 4096
+#define PAGE_COUNT (MEMORY_SIZE / PAGE_SIZE)
+
+#define HEAP_PAGE_COUNT 256
+
+#define KERNEL_VIRTUAL_BASE 0xFFFFFFFF80000000ULL
+#define PHYS_TO_VIRT(phys) ((void *)((uint64_t)(phys) + KERNEL_VIRTUAL_BASE))
+#define VIRT_TO_PHYS(virt) ((uint64_t)(virt) - KERNEL_VIRTUAL_BASE)
+
+#define KERNEL_PML4 0xFFFFFFFF80010000ULL
+
+#define USER_VIRTUAL_BASE 0x0000000000000000ULL
+#define USER_VIRTUAL_CODE (USER_VIRTUAL_BASE + 0x400000)
+#define USER_VIRTUAL_HEAP (USER_VIRTUAL_CODE + 0x400000)
+#define USER_VIRTUAL_STACK (0x0000700000000000ULL - PAGE_SIZE)
+
 void memory_init();
 
+void *memory_page_allocate();
+
+void memory_page_free(void *page);
+
+void memory_page_map(uint64_t *pml4, void *virt, void *phys, uint64_t flags);
+
+void memory_page_unmap(uint64_t *pml4, void *virt);
+
 void *memory_allocate(uint64_t size);
 
 void memory_free(void *pointer);

src/path.c

@@ -3,7 +3,6 @@
 #include "src/memory.h"
 #include "src/string.h"
 #include "src/util.h"
-#include <stdint.h>
 
 path_t *path_open(const fs_node_t *root, const path_t *source, char *path) {
   path_t *result = memory_allocate(sizeof(path_t));

src/process.c

@@ -0,0 +1,112 @@
+#include "src/process.h"
+#include "src/gdt.h"
+#include "src/memory.h"
+#include "src/panic.h"
+#include "src/path.h"
+#include "src/tss.h"
+
+process_t *current_process;
+
+process_t *process_create(const process_t *parent, const uint8_t *code, uint64_t size) {
+  ASSERT(size <= USER_VIRTUAL_HEAP - USER_VIRTUAL_CODE, "process_create: binary too big");
+
+  process_t *proc = memory_allocate(sizeof(process_t));
+
+  proc->kernel_stack = (uint8_t *)memory_allocate(PAGE_SIZE) + PAGE_SIZE;
+
+  proc->pml4 = PHYS_TO_VIRT(memory_page_allocate());
+  memory_set(0, PAGE_SIZE / 2, proc->pml4);
+  memory_copy((uint64_t *)KERNEL_PML4 + 256, PAGE_SIZE / 2, proc->pml4 + 256);
+
+  memory_page_map(proc->pml4, (void *)USER_VIRTUAL_STACK, memory_page_allocate(), PAGE_PRESENT | PAGE_WRITABLE | PAGE_USER);
+
+  for (uint64_t page = 0; size > 0; page++) {
+    void *p = memory_page_allocate();
+    uint64_t s = size < PAGE_SIZE ? size : PAGE_SIZE;
+    memory_page_map(proc->pml4, (void *)(USER_VIRTUAL_CODE + page * PAGE_SIZE), p, PAGE_PRESENT | PAGE_USER);
+    memory_copy(code, s, PHYS_TO_VIRT(p));
+    code += s;
+    size -= s;
+  }
+
+  for (uint64_t page = 0; page < HEAP_PAGE_COUNT; page++) {
+    memory_page_map(proc->pml4, (void *)(USER_VIRTUAL_HEAP + page * PAGE_SIZE), memory_page_allocate(),
+                    PAGE_PRESENT | PAGE_WRITABLE | PAGE_USER);
+  }
+
+  proc->cwd = memory_allocate(sizeof(path_t));
+  proc->cwd->depth = parent->cwd->depth;
+  for (uint8_t i = 0; i < parent->cwd->depth; i++) {
+    proc->cwd->stack[i] = fs_open_again(parent->cwd->stack[i]);
+  }
+
+  proc->root = parent->root;
+  proc->stdin = parent->stdin;
+  proc->stdout = parent->stdout;
+
+  return proc;
+}
+
+void process_run(process_t *proc) {
+  current_process = proc;
+  tss.rsp0 = (uint64_t)proc->kernel_stack;
+
+  __asm__ volatile("mov %0, %%cr3" : : "r"(VIRT_TO_PHYS(proc->pml4)) : "memory");
+
+  __asm__ volatile("push %0\n" // ss
+                   "push %1\n" // rsp
+                   "push %2\n" // rflags
+                   "push %3\n" // cs
+                   "push %4\n" // rip
+                   "iretq"
+                   :
+                   : "r"((uint64_t)USER_DS), "r"(USER_VIRTUAL_STACK + PAGE_SIZE), "r"((uint64_t)USER_RFLAGS), "r"((uint64_t)USER_CS),
+                     "r"(USER_VIRTUAL_CODE)
+                   : "memory");
+}
+
+void process_destroy(process_t *proc) {
+  path_close(proc->cwd);
+
+  for (uint16_t pml4i = 0; pml4i < 256; pml4i++) {
+    if (!(proc->pml4[pml4i] & PAGE_PRESENT)) {
+      continue;
+    }
+
+    uint64_t *pdpt = PHYS_TO_VIRT(proc->pml4[pml4i] & ~(uint64_t)0xFFF);
+    for (uint16_t pdpti = 0; pdpti < 512; pdpti++) {
+      if (!(pdpt[pdpti] & PAGE_PRESENT)) {
+        continue;
+      }
+
+      uint64_t *pd = PHYS_TO_VIRT(pdpt[pdpti] & ~(uint64_t)0xFFF);
+      for (uint16_t pdi = 0; pdi < 512; pdi++) {
+        if (!(pd[pdi] & PAGE_PRESENT)) {
+          continue;
+        }
+
+        if (pd[pdi] & PAGE_HUGE) {
+          memory_page_free((void *)(pd[pdi] & ~(uint64_t)0xFFF));
+          continue;
+        }
+
+        uint64_t *pt = PHYS_TO_VIRT(pd[pdi] & ~(uint64_t)0xFFF);
+        for (uint16_t pti = 0; pti < 512; pti++) {
+          if (!(pt[pti] & PAGE_PRESENT)) {
+            continue;
+          }
+          memory_page_free((void *)(pt[pti] & ~(uint64_t)0xFFF));
+        }
+        memory_page_free((void *)(pd[pdi] & ~(uint64_t)0xFFF));
+      }
+      memory_page_free((void *)(pdpt[pdpti] & ~(uint64_t)0xFFF));
+    }
+    memory_page_free((void *)(proc->pml4[pml4i] & ~(uint64_t)0xFFF));
+  }
+
+  memory_page_free((void *)VIRT_TO_PHYS(proc->pml4));
+
+  memory_free(proc->kernel_stack);
+
+  memory_free(proc);
+}

src/process.h

@@ -5,14 +5,26 @@
 #include "src/stream.h"
 #include "src/string.h"
 
-typedef struct {
-  path_t *cwd;
+#define USER_RFLAGS 0x202
+
+typedef struct process {
+  uint64_t *pml4;
+  void *kernel_stack;
   fs_node_t *root;
+  path_t *cwd;
   stream_t *stdin;
   stream_t *stdout;
 } process_t;
 
 typedef void (*app_t)(process_t *proc, uint8_t argc, char **argv);
 
+extern process_t *current_process;
+
+process_t *process_create(const process_t *parent, const uint8_t *code, uint64_t size);
+
+void process_run(process_t *proc);
+
+void process_destroy(process_t *proc);
+
 #define WRITE_S(s) proc->stdout->write(proc->stdout, s, sizeof(s) - 1);
 #define WRITE_D(s) proc->stdout->write(proc->stdout, s, string_length(s));

src/string.c

@@ -1,7 +1,5 @@
 #include "src/string.h"
-#include "src/memory.h"
 #include <stdarg.h>
-#include <stdint.h>
 
 uint8_t bytes_equal(const char *l, const char *r, uint64_t count) {
   while (count--) {
@@ -164,41 +162,39 @@ uint64_t string_uint_to_hex(uint64_t value, uint64_t max, char *result) {
   return i;
 }
 
-char *string_format(const char *format, ...) {
+uint64_t string_format(const char *format, uint64_t max, char *output, ...) {
   va_list args;
-  va_start(args, format);
+  va_start(args, output);
 
-  uint64_t limit = 1023;
-
-  char *result = memory_allocate(limit + 1);
+  uint64_t limit = max - 1;
 
   uint64_t fi = 0, ri = 0;
   while (format[fi] && ri < limit) {
     if (format[fi] != '%') {
-      result[ri++] = format[fi++];
+      output[ri++] = format[fi++];
     } else {
       switch (format[fi + 1]) {
       case 'c':
-        result[ri++] = (char)va_arg(args, int);
+        output[ri++] = (char)va_arg(args, int);
         fi += 2;
         break;
       case 'd':
-        string_int_to_decimal(va_arg(args, int64_t), limit - ri, result + ri);
-        while (result[ri]) {
+        string_int_to_decimal(va_arg(args, int64_t), limit - ri, output + ri);
+        while (output[ri]) {
           ri++;
         }
         fi += 2;
         break;
       case 'u':
-        string_uint_to_decimal(va_arg(args, uint64_t), limit - ri, result + ri);
-        while (result[ri]) {
+        string_uint_to_decimal(va_arg(args, uint64_t), limit - ri, output + ri);
+        while (output[ri]) {
           ri++;
         }
         fi += 2;
         break;
       case 'x':
-        string_uint_to_hex(va_arg(args, uint64_t), limit - ri, result + ri);
-        while (result[ri]) {
+        string_uint_to_hex(va_arg(args, uint64_t), limit - ri, output + ri);
+        while (output[ri]) {
           ri++;
         }
         fi += 2;
@@ -206,25 +202,26 @@ char *string_format(const char *format, ...) {
       case 's': {
         char *s = va_arg(args, char *);
         while (*s && ri < limit) {
-          result[ri++] = *s;
+          output[ri++] = *s;
           s++;
         }
         fi += 2;
         break;
       }
       case '%': {
-        result[ri++] = '%';
+        output[ri++] = '%';
         fi += 2;
         break;
       }
       default:
-        result[ri++] = format[fi++];
-        result[ri++] = format[fi++];
+        output[ri++] = format[fi++];
+        output[ri++] = format[fi++];
         break;
       }
     }
   }
-  result[ri] = '\0';
+  output[ri] = '\0';
   va_end(args);
-  return result;
+
+  return ri;
 }

src/string.h

@@ -20,4 +20,4 @@ uint64_t string_int_to_decimal(int64_t value, uint64_t max, char *result);
 
 uint64_t string_uint_to_hex(uint64_t value, uint64_t max, char *result);
 
-char *string_format(const char *format, ...);
+uint64_t string_format(const char *format, uint64_t max, char *output, ...);

src/syscall.asm

@@ -0,0 +1,39 @@
+bits 64
+
+extern tss
+extern syscall_dispatch
+
+global syscall_entry
+
+syscall_entry:
+    mov [user_rsp_tmp], rsp
+    mov rsp, [tss + 4]
+
+    push rcx
+    push r11
+    push rbp
+    push rbx
+    push r12
+    push r13
+    push r14
+    push r15
+
+    mov rcx, rdx    ; arg3
+    mov rdx, rsi    ; arg2
+    mov rsi, rdi    ; arg1
+    mov rdi, rax    ; syscall number
+    call syscall_dispatch
+
+    pop r15
+    pop r14
+    pop r13
+    pop r12
+    pop rbx
+    pop rbp
+    pop r11
+    pop rcx
+
+    mov rsp, [rel user_rsp_tmp]
+    o64 sysret
+
+user_rsp_tmp: dq 0

src/syscall.c

@@ -0,0 +1,62 @@
+#include "src/syscall.h"
+#include "src/gdt.h"
+#include "src/memory.h"
+#include "src/process.h"
+#include "src/util.h"
+
+#define MSR_EFER 0xC0000080
+#define MSR_STAR 0xC0000081
+#define MSR_LSTAR 0xC0000082
+#define MSR_FMASK 0xC0000084
+
+static inline uint64_t rdmsr(uint32_t msr) {
+  uint32_t lo, hi;
+  __asm__ volatile("rdmsr" : "=a"(lo), "=d"(hi) : "c"(msr));
+  return ((uint64_t)hi << 32) | lo;
+}
+
+static inline void wrmsr(uint32_t msr, uint64_t value) {
+  __asm__ volatile("wrmsr" : : "c"(msr), "a"((uint32_t)value), "d"((uint32_t)(value >> 32)));
+}
+
+extern void syscall_entry();
+
+void syscall_init() {
+  wrmsr(MSR_EFER, rdmsr(MSR_EFER) | 0x01);
+  wrmsr(MSR_STAR, ((uint64_t)KERNEL_DS << 48) | ((uint64_t)KERNEL_CS << 32));
+  wrmsr(MSR_LSTAR, (uint64_t)syscall_entry);
+  wrmsr(MSR_FMASK, 0x200);
+}
+
+static void write(uint64_t fd, uint64_t buf, uint64_t size) {
+  (void)fd;
+  current_process->stdout->write(current_process->stdout, (const char *)buf, size);
+}
+
+static void exit() {
+   __asm__ volatile("mov %0, %%cr3" : : "r"(KERNEL_PML4) : "memory");
+
+
+  process_destroy(current_process);
+  current_process = NUL;
+
+  while (1)
+    ;
+
+  // shutdown for now
+  // outw(0x604, 0x2000);
+  // __asm__ volatile("cli; hlt");
+}
+
+void syscall_dispatch(uint64_t func, uint64_t arg1, uint64_t arg2, uint64_t arg3) {
+  switch (func) {
+  case SYSCALL_WRITE:
+    write(arg1, arg2, arg3);
+    break;
+  case SYSCALL_EXIT:
+    exit();
+    break;
+  default:
+    break;
+  }
+}

src/syscall.h

@@ -0,0 +1,10 @@
+#pragma once
+
+#include <stdint.h>
+
+#define SYSCALL_WRITE 1
+#define SYSCALL_EXIT 60
+
+void syscall_init();
+
+void syscall_dispatch(uint64_t func, uint64_t arg1, uint64_t arg2, uint64_t arg3);

src/tss.c

@@ -0,0 +1,9 @@
+#include "src/tss.h"
+
+tss_t tss = {
+    .iopb_offset = sizeof(tss_t) // points past end of TSS = no I/O permissions
+};
+
+void tss_set_kernel_stack(void *rsp0) {
+  tss.rsp0 = (uint64_t)rsp0;
+}

src/tss.h

@@ -0,0 +1,20 @@
+// tss.h
+#pragma once
+
+#include <stdint.h>
+
+typedef struct __attribute__((packed)) {
+  uint32_t reserved0;
+  uint64_t rsp0;
+  uint64_t rsp1;
+  uint64_t rsp2;
+  uint64_t reserved1;
+  uint64_t ist[7];
+  uint64_t reserved2;
+  uint16_t reserved3;
+  uint16_t iopb_offset;
+} tss_t;
+
+extern tss_t tss;
+
+void tss_set_kernel_stack(void *rsp0);

src/vga.c

@@ -1,9 +1,8 @@
 #include "src/vga.h"
 #include "src/memory.h"
 #include "src/util.h"
-#include <stdint.h>
 
-static uint16_t *vga = (uint16_t *)0x000B8000;
+static uint16_t *vga = (uint16_t *)0xFFFFFFFF800B8000;
 static uint16_t color = (uint16_t)0x0F << 8;
 static uint16_t offset = 0;