Run first app from ramdisk!

kernel/proc/.gitignore

@@ -0,0 +1 @@
+*.o

kernel/proc/proc.c

@@ -0,0 +1,245 @@
+#include <aux/compiler.h>
+#include <aux/elf.h>
+#include <irq/irq.h>
+#include <libk/align.h>
+#include <libk/list.h>
+#include <libk/std.h>
+#include <libk/string.h>
+#include <limine/requests.h>
+#include <mm/liballoc.h>
+#include <mm/pmm.h>
+#include <proc/proc.h>
+#include <rd/rd.h>
+#include <sync/spin_lock.h>
+#include <sys/debug.h>
+#include <sys/mm.h>
+#include <sys/sched.h>
+#include <sys/smp.h>
+
+#if defined(__x86_64__)
+#include <amd64/intr_defs.h>
+#endif
+
+struct elf_aux {
+  uint64_t entry;
+  uint64_t phdr;
+  uint64_t phent;
+  uint64_t phnum;
+};
+
+static struct procw* procs;
+static spin_lock_t procs_lock = SPIN_LOCK_INIT;
+
+static bool proc_check_elf (uint8_t* elf) {
+  if (!((elf[0] == 0x7F) && (elf[1] == 'E') && (elf[2] == 'L') && (elf[3] == 'F')))
+    return false;
+  return true;
+}
+
+void proc_map (struct proc* proc, uintptr_t start_paddr, uintptr_t start_vaddr, size_t pages,
+               uint32_t flags) {
+  struct proc_mapping* mapping = malloc (sizeof (*mapping));
+  mapping->paddr = start_paddr;
+  mapping->vaddr = start_vaddr;
+  mapping->size = pages * PAGE_SIZE;
+
+  flags &= ~MM_PD_LOCK; /* clear LOCK flag if present, because we lock manualy */
+
+  spin_lock (&proc->pd.lock);
+
+  linklist_append (struct proc_mapping*, proc->mappings, mapping);
+
+  for (uintptr_t vpage = start_vaddr, ppage = start_paddr; vpage < start_vaddr + pages * PAGE_SIZE;
+       vpage += PAGE_SIZE, ppage += PAGE_SIZE) {
+    mm_map_page (&proc->pd, ppage, vpage, flags);
+  }
+
+  spin_unlock (&proc->pd.lock);
+}
+
+static struct elf_aux proc_load_segments (struct proc* proc, uint8_t* elf) {
+  struct elf_aux aux;
+
+  Elf64_Ehdr* ehdr = (Elf64_Ehdr*)elf;
+  aux.entry = ehdr->e_entry;
+  aux.phnum = ehdr->e_phnum;
+  aux.phent = ehdr->e_phentsize;
+
+  struct limine_hhdm_response* hhdm = limine_hhdm_request.response;
+
+  for (uint64_t segment = 0; segment < ehdr->e_phnum; segment++) {
+    Elf64_Phdr* phdr =
+        (Elf64_Phdr*)((uintptr_t)elf + ehdr->e_phoff + (ehdr->e_phentsize * segment));
+
+    switch (phdr->p_type) {
+    case PT_PHDR: {
+      aux.phdr = (uint64_t)phdr->p_vaddr;
+    } break;
+    case PT_LOAD: {
+      uintptr_t v_addr = align_down (phdr->p_vaddr, PAGE_SIZE);
+      uintptr_t off = phdr->p_vaddr - v_addr;
+
+      size_t blks = div_align_up (phdr->p_memsz + off, PAGE_SIZE);
+
+      uintptr_t p_addr = pmm_alloc (blks);
+      if (p_addr == PMM_ALLOC_ERR)
+        DEBUG ("pmm oom error while loading ELF segments! (tried to alloc %zu blks)\n", blks);
+
+      memset ((void*)((uintptr_t)hhdm->offset + p_addr), 0, blks * PAGE_SIZE);
+      memcpy ((void*)((uintptr_t)hhdm->offset + p_addr + off),
+              (void*)((uintptr_t)elf + phdr->p_offset), phdr->p_filesz);
+
+      uint32_t pg_flags = MM_PG_USER | MM_PG_PRESENT;
+      if (phdr->p_flags & PF_W)
+        pg_flags |= MM_PG_RW;
+
+      proc_map (proc, p_addr, v_addr, blks, pg_flags);
+    } break;
+    }
+  }
+
+  return aux;
+}
+
+static struct proc* proc_spawn_rd (char* name) {
+  struct rd_file* rd_file = rd_get_file (name);
+
+  bool ok = proc_check_elf (rd_file->content);
+  DEBUG ("ELF magic %s\n", (ok ? "OK" : "BAD"));
+
+  if (!ok)
+    return NULL;
+
+  struct proc* proc = malloc (sizeof (*proc));
+  if (proc == NULL)
+    return NULL;
+
+  memset (proc, 0, sizeof (*proc));
+
+#if defined(__x86_64__)
+  proc->pd.lock = SPIN_LOCK_INIT;
+  proc->pd.cr3_paddr = mm_alloc_user_pd_phys ();
+  if (proc->pd.cr3_paddr == 0) {
+    free (proc);
+    return NULL;
+  }
+
+  proc->pdata.syscall_stack = pmm_alloc (KSTACK_SIZE / PAGE_SIZE);
+  if (proc->pdata.syscall_stack == PMM_ALLOC_ERR) {
+    free (proc);
+    return NULL;
+  }
+
+  proc->pdata.user_stack = pmm_alloc (USTACK_SIZE / PAGE_SIZE);
+  if (proc->pdata.user_stack == PMM_ALLOC_ERR) {
+    free (proc);
+    pmm_free (proc->pdata.syscall_stack, USTACK_SIZE / PAGE_SIZE);
+    return NULL;
+  }
+
+  uintptr_t user_stack = proc->pdata.user_stack;
+
+  proc->pdata.syscall_stack += KSTACK_SIZE;
+  proc->pdata.user_stack += USTACK_SIZE;
+
+  proc_map (proc, user_stack, PROC_USTACK_TOP - USTACK_SIZE, USTACK_SIZE / PAGE_SIZE,
+            MM_PG_USER | MM_PG_PRESENT | MM_PG_RW);
+
+  struct elf_aux aux = proc_load_segments (proc, rd_file->content);
+
+  proc->pdata.regs.ss = 0x20 | 0x03;
+  proc->pdata.regs.rsp = (uint64_t)PROC_USTACK_TOP;
+  proc->pdata.regs.rflags = 0x202;
+  proc->pdata.regs.cs = 0x18 | 0x03;
+  proc->pdata.regs.rip = aux.entry;
+  proc->lock = SPIN_LOCK_INIT;
+#endif
+
+  return proc;
+}
+
+static void proc_register (struct proc* proc) {
+  /* make available globally. */
+  struct procw* procw = malloc (sizeof (*procw));
+  if (procw == NULL)
+    return;
+  procw->proc = proc;
+  proc->procw = procw;
+
+  spin_lock (&procs_lock);
+
+  spin_lock (&thiscpu->lock);
+
+  linklist_append (struct procw*, procs, procw);
+  linklist_append (struct proc*, thiscpu->proc_run_q, proc);
+
+  if (thiscpu->proc_current == NULL)
+    thiscpu->proc_current = proc;
+
+  spin_unlock (&thiscpu->lock);
+
+  spin_unlock (&procs_lock);
+}
+
+void proc_sched (void) {
+  spin_lock (&thiscpu->lock);
+
+  if (thiscpu->proc_run_q == NULL || thiscpu->proc_current == NULL) {
+    goto done;
+  }
+
+  thiscpu->proc_current = thiscpu->proc_current->next;
+
+  if (thiscpu->proc_current == NULL) {
+    thiscpu->proc_current = thiscpu->proc_run_q;
+  }
+
+done:
+  spin_unlock (&thiscpu->lock);
+
+  if (thiscpu->proc_current != NULL) {
+    do_sched (&thiscpu->proc_current->pdata.regs, &thiscpu->proc_current->pd);
+  }
+
+#if defined(__x86_64__)
+  extern void amd64_spin (void);
+
+  amd64_spin ();
+#endif
+}
+
+void proc_kill (struct proc* proc) {
+  spin_lock (&procs_lock);
+
+  spin_lock (&thiscpu->lock);
+
+  linklist_remove (struct procw*, procs, proc->procw);
+  linklist_remove (struct proc*, thiscpu->proc_run_q, proc);
+
+  if (thiscpu->proc_current == proc)
+    thiscpu->proc_current = NULL;
+
+  spin_unlock (&thiscpu->lock);
+
+  spin_unlock (&procs_lock);
+
+  /* clean up */
+  free (proc->procw);
+  free (proc);
+}
+
+static void proc_irq_sched (void* arg, void* regs) {
+  (void)arg, (void)regs;
+  proc_sched ();
+}
+
+void proc_init (void) {
+  struct proc* init = proc_spawn_rd ("init.exe");
+  proc_register (init);
+
+#if defined(__x86_64__)
+  irq_attach (&proc_irq_sched, NULL, SCHED_PREEMPT_TIMER, 0);
+#endif
+
+  do_sched (&init->pdata.regs, &init->pd);
+}

kernel/proc/proc.h

@@ -0,0 +1,49 @@
+#ifndef _KERNEL_PROC_PROC_H
+#define _KERNEL_PROC_PROC_H
+
+#include <aux/compiler.h>
+#include <libk/std.h>
+#include <sync/spin_lock.h>
+#include <sys/mm.h>
+
+#if defined(__x86_64__)
+#include <amd64/gdt.h>  /* KSTACK_SIZE */
+#include <amd64/proc.h> /* USTACK_SIZE */
+#endif
+
+struct cpu;
+
+struct proc_mapping {
+  struct proc_mapping* next;
+  uintptr_t paddr;
+  uintptr_t vaddr;
+  size_t size;
+} PACKED;
+
+struct procw;
+
+struct proc {
+  struct proc* next;
+  struct proc_mapping* mappings; /* pd.lock implicitly protects this field */
+  struct proc_platformdata pdata;
+  struct pd pd;
+  spin_lock_t lock;
+  struct cpu* cpu;
+  struct procw* procw; /* link to it's global struct */
+};
+
+/*
+ * struct proc is a member of a CPU's proc_run_q.
+ * struct procw is a process wrapper that is a member of
+ * a global process list.
+ */
+struct procw {
+  struct procw* next;
+  struct proc* proc;
+};
+
+void proc_sched (void);
+void proc_kill (struct proc* proc);
+void proc_init (void);
+
+#endif // _KERNEL_PROC_PROC_H

kernel/proc/src.mk

@@ -0,0 +1,3 @@
+c += proc/proc.c
+
+o += proc/proc.o