refactor: separate syscall interfaces from kernel space, into posix compliant structure.

[lunaix-os.git] / lunaix-os / kernel / asm / x86 / pfault.c

diff --git a/lunaix-os/kernel/asm/x86/pfault.c b/lunaix-os/kernel/asm/x86/pfault.c
index 30fd498605d721f9c3d782210a37b186b4fa7ce9..7b8a12bbeb096da41545e5cb8c00e46e7e798ef3 100644 (file)
--- a/lunaix-os/kernel/asm/x86/pfault.c
+++ b/lunaix-os/kernel/asm/x86/pfault.c
@@ -5,9 +5,12 @@
 #include <lunaix/mm/region.h>
 #include <lunaix/mm/vmm.h>
 #include <lunaix/sched.h>
+#include <lunaix/signal.h>
 #include <lunaix/status.h>
 #include <lunaix/syslog.h>
 
+#include <klibc/string.h>
+
 static void
 kprintf(const char* fmt, ...)
 {
@@ -40,18 +43,18 @@ intr_routine_page_fault(const isr_param* param)
         if (do_kernel(&mapping)) {
             return;
         }
-        goto segv_term;
+        // 如果不是，那么看看内核是不是需要用户页。
     }
 
     struct mm_region* hit_region = region_get(&__current->mm.regions, ptr);
 
     if (!hit_region) {
-        // Into the void...
+        // 当你凝视深渊时……
         goto segv_term;
     }
 
-    x86_pte_t* pte = &PTE_MOUNTED(PD_REFERENCED, ptr >> 12);
-    if ((*pte & PG_PRESENT)) {
+    volatile x86_pte_t* pte = &PTE_MOUNTED(PD_REFERENCED, ptr >> 12);
+    if (PG_PRESENTED(*pte)) {
         if ((hit_region->attr & COW_MASK) == COW_MASK) {
             // normal page fault, do COW
             cpu_invplg(pte);
@@ -65,34 +68,64 @@ intr_routine_page_fault(const isr_param* param)
         goto segv_term;
     }
 
-    if (!(*pte)) {
-        // Invalid location
+    // an anonymous page and not present
+    //   -> a new page need to be alloc
+    if ((hit_region->attr & REGION_ANON)) {
+        if (!PG_PRESENTED(*pte)) {
+            cpu_invplg(pte);
+            uintptr_t pa = pmm_alloc_page(__current->pid, 0);
+            if (!pa) {
+                goto oom;
+            }
+
+            *pte = *pte | pa | PG_PRESENT;
+            goto resolved;
+        }
+        // permission denied on anon page
         goto segv_term;
     }
 
-    uintptr_t loc = *pte & ~0xfff;
+    // if mfile is set (Non-anonymous), then it is a mem map
+    if (hit_region->mfile && !PG_PRESENTED(*pte)) {
+        struct v_file* file = hit_region->mfile;
+        u32_t offset =
+          (ptr - hit_region->start) & (PG_SIZE - 1) + hit_region->offset;
+        uintptr_t pa = pmm_alloc_page(__current->pid, 0);
+
+        if (!pa) {
+            goto oom;
+        }
 
-    // a writable page, not present, not cached, pte attr is not null
-    //   -> a new page need to be alloc
-    if ((hit_region->attr & REGION_WRITE) && (*pte & 0xfff) && !loc) {
         cpu_invplg(pte);
-        uintptr_t pa = pmm_alloc_page(__current->pid, 0);
-        *pte = *pte | pa | PG_PRESENT;
+        *pte = (*pte & 0xFFF) | pa | PG_PRESENT;
+
+        ptr = ptr & ~(PG_SIZE - 1);
+        memset(ptr, 0, PG_SIZE);
+
+        int errno = file->ops->read_page(file->inode, ptr, PG_SIZE, offset);
+        if (errno < 0) {
+            kprintf(KERROR "fail to read page (%d)\n", errno);
+            goto segv_term;
+        }
+
         goto resolved;
     }
 
-    // page not present, bring it from disk or somewhere else
+    // page not present, might be a chance to introduce swap file?
     __print_panic_msg("WIP page fault route", param);
     while (1)
         ;
 
+oom:
+    kprintf(KERROR "out of memory\n");
 segv_term:
     kprintf(KERROR "(pid: %d) Segmentation fault on %p (%p:%p)\n",
             __current->pid,
             ptr,
             param->cs,
             param->eip);
-    terminate_proc(LXSEGFAULT);
+    __SIGSET(__current->sig_pending, _SIGSEGV);
+    schedule();
     // should not reach
     while (1)
         ;
@@ -106,14 +139,8 @@ int
 do_kernel(v_mapping* mapping)
 {
     uintptr_t addr = mapping->va;
-    if (addr >= KHEAP_START && addr < PROC_START) {
-        // This is kernel heap page
-        uintptr_t pa = pmm_alloc_page(KERNEL_PID, 0);
-        *mapping->pte = (*mapping->pte & 0xfff) | pa | PG_PRESENT;
-        cpu_invplg(mapping->pte);
-        cpu_invplg(addr);
-        goto done;
-    }
+
+    // TODO
 
     return 0;
 done: