Documentation Index
Fetch the complete documentation index at: https://mintlify.com/CRISTOP-bot/cris-os-v2/llms.txt
Use this file to discover all available pages before exploring further.
kmain is the C entry point of CrisOS v2. It is called by the boot assembly stub in boot/boot.S with a single argument — the physical address of the Multiboot information structure populated by GRUB — and orchestrates the initialization of every kernel subsystem in a deterministic, sequential order. Control never returns from kmain; it ends with an infinite call to shell_run(), and the fallback hlt loop below it is never reached under normal operation.
Full kmain Source
void kmain(unsigned long mbi_addr)
{
console_clear();
console_print("================================\n");
console_print(" CrisOS Kernel\n");
console_print("================================\n\n");
console_print("[ OK ] Console initialized\n");
gdt_init();
idt_init();
pic_init();
console_print("[ OK ] PIC initialized (IRQs 0-15 remapped)\n");
pic_mask(0xFD, 0xFF); /* enable only IRQ1 (keyboard) -- polling mode */
timer_init(100);
console_print("[ OK ] PIT timer initialized (100 Hz)\n");
keyboard_init();
console_print("[ OK ] Keyboard initialized\n");
mouse_init();
pic_mask(0xFC, 0xEF); /* enable IRQ1 + IRQ12 */
bool rootfs_loaded = false;
if (mbi_addr) {
struct multiboot_info *mbi = (struct multiboot_info *)mbi_addr;
if (mbi->flags & 0x8) {
console_print("[ OK ] Multiboot modules detected\n");
struct multiboot_module *mods =
(struct multiboot_module *)mbi->mods_addr;
for (unsigned long i = 0; i < mbi->mods_count; ++i) {
const char *name = (const char *)mods[i].cmdline;
if (name && kstrstr(name, "rootfs")) {
console_print("[ INFO ] Mounting rootfs...\n");
if (fs_init((const void *)mods[i].mod_start,
mods[i].mod_end - mods[i].mod_start)) {
console_print("[ OK ] Rootfs mounted\n");
if (vfs_init()) {
console_print("[ OK ] VFS initialized\n");
rootfs_loaded = true;
} else {
console_print("[FAIL] VFS initialization failed\n");
}
} else {
console_print("[FAIL] Rootfs mount failed\n");
}
break;
}
}
if (!rootfs_loaded)
console_print("[WARN] No rootfs module found or failed to mount\n");
} else {
console_print("[FAIL] No multiboot modules found\n");
}
} else {
console_print("[FAIL] Invalid multiboot info\n");
}
boot_init();
console_print("[ OK ] Boot manager initialized\n");
systemd_init();
console_print("[ OK ] Service manager initialized\n");
lcp_init();
console_print("[ OK ] LCP package manager initialized\n\n");
int sum = add_asm(7, 5);
char buf[32];
kitoa(sum, buf, sizeof(buf));
console_print("[ASM] 7 + 5 = ");
console_print(buf);
console_print("\n\n");
console_print("Launching shell...\n\n");
shell_run();
while (1)
asm volatile("hlt");
}
Initialization Order
The table below maps each subsystem to the exact function called during startup. Order matters — for example, the IDT must be installed before the PIC is unmasked, and the PIC must be initialized before any IRQ can fire.
| Subsystem | Function(s) Called | Notes |
|---|
| Console | console_clear() | Clears the VGA buffer; all subsequent [ OK ] messages are visible from this point |
| GDT | gdt_init() | Installs a 5-entry flat GDT (null, ring-0 code/data, ring-3 code/data) via lgdt |
| IDT | idt_init() | Registers 32 CPU exception handlers and 16 IRQ handlers via lidt |
| PIC | pic_init() then pic_mask(0xFD, 0xFF) | Remaps IRQs 0–15 to INT 0x20–0x2F; initially enables IRQ1 (keyboard) only |
| Timer | timer_init(100) | Programs the 8253 PIT to fire at 100 Hz (10 ms per tick) |
| Keyboard | keyboard_init() | Initializes the PS/2 keyboard driver |
| Mouse | mouse_init() then pic_mask(0xFC, 0xEF) | Initializes the PS/2 mouse driver; unmasks IRQ12 alongside IRQ1 |
| Rootfs | fs_init() from Multiboot module, then vfs_init() | Mounts the flat-file rootfs image loaded by GRUB as a Multiboot module |
| Boot manager | boot_init() | Parses .boot unit files from the rootfs and runs boot.target |
| Service manager | systemd_init() | Registers system services |
| LCP | lcp_init() | Initializes the LCP package manager |
| Shell | shell_run() | Enters the interactive command loop — does not return |
Multiboot Module Loading
kmain uses two structs to parse the Multiboot information block passed by GRUB:
struct multiboot_info {
unsigned long flags; /* bitmask of valid fields */
unsigned long mem_lower; /* lower memory in KB */
unsigned long mem_upper; /* upper memory in KB */
unsigned long boot_device;
unsigned long cmdline;
unsigned long mods_count; /* number of loaded modules */
unsigned long mods_addr; /* physical address of module list */
};
struct multiboot_module {
unsigned long mod_start; /* physical start address of module data */
unsigned long mod_end; /* physical end address (exclusive) */
unsigned long cmdline; /* pointer to module command-line string */
unsigned long reserved;
};
mbi->flags & 0x8 to confirm that the module list fields are valid. It then iterates over the mods_count entries at mods_addr and calls kstrstr(name, "rootfs") on each module’s command-line string. The first match is passed to fs_init() as a raw byte pointer plus length (mod_end - mod_start).
Kernel Panic
When an unrecoverable error occurs — including any CPU exception caught by the IDT — the kernel calls kernel_panic():
void kernel_panic(const char *message) {
console_clear_color(0x4F); /* red background, white text (attr 0x4F) */
console_print("KERNEL PANIC\n\n");
if (message && message[0]) {
console_print(message);
console_print("\n\n");
}
console_print("El sistema se ha detenido. Presiona una tecla para bloquear.");
keyboard_read_char(); /* wait for a keypress */
halt_cpu(); /* cli; hlt loop — never returns */
}
0x4F encodes white-on-red (0x4 = red background, 0xF = bright white foreground), making the panic screen immediately recognizable. After displaying the message the system waits for one keypress, then calls the assembly halt_cpu routine which disables interrupts and halts forever.
To add a new kernel subsystem, implement an init() function and any command handler, then call init() from kmain before the shell_run() call. Register a command keyword in the shell dispatch table so the subsystem can be exercised interactively without rebooting.