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 f07c74faa54cc4f5afdaa4acd44a88f97195b4de..7b8a12bbeb096da41545e5cb8c00e46e7e798ef3 100644 (file)
--- a/lunaix-os/kernel/asm/x86/pfault.c
+++ b/lunaix-os/kernel/asm/x86/pfault.c
@@ -1,62 +1,148 @@
 #include <arch/x86/interrupts.h>
 #include <arch/x86/interrupts.h>

-#include <lunaix/mm/pmm.h>
+#include <lunaix/common.h>

 #include <lunaix/mm/mm.h>
 #include <lunaix/mm/mm.h>

+#include <lunaix/mm/pmm.h>

 #include <lunaix/mm/region.h>
 #include <lunaix/mm/vmm.h>
 #include <lunaix/mm/region.h>
 #include <lunaix/mm/vmm.h>

-#include <lunaix/common.h>
-#include <lunaix/syslog.h>
-#include <lunaix/status.h>

 #include <lunaix/sched.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, ...) { va_list args; va_start(args, fmt); __kprintf("PFAULT", fmt, args); va_end(args); }
+static void
+kprintf(const char* fmt, ...)
+{
+    va_list args;
+    va_start(args, fmt);
+    __kprintf("PFAULT", fmt, args);
+    va_end(args);
+}
+
+#define COW_MASK (REGION_RSHARED | REGION_READ | REGION_WRITE)

 
 

-extern void __print_panic_msg(const char* msg, const isr_param* param);
+extern void
+__print_panic_msg(const char* msg, const isr_param* param);

 
 void
 
 void

-intr_routine_page_fault (const isr_param* param) 
+intr_routine_page_fault(const isr_param* param)

 {
     uintptr_t ptr = cpu_rcr2();
     if (!ptr) {
         goto segv_term;
     }
 
 {
     uintptr_t ptr = cpu_rcr2();
     if (!ptr) {
         goto segv_term;
     }
 

-    struct mm_region* hit_region = region_get(__current, ptr);
+    v_mapping mapping;
+    if (!vmm_lookup(ptr, &mapping)) {
+        goto segv_term;
+    }
+
+    if (!SEL_RPL(param->cs)) {
+        // 如果是内核页错误……
+        if (do_kernel(&mapping)) {
+            return;
+        }
+        // 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) {
 
     if (!hit_region) {

-        // Into the void...
+        // 当你凝视深渊时……

         goto segv_term;
     }
 
         goto segv_term;
     }
 

-    if (param->eip == ptr && !(hit_region->attr & REGION_EXEC)) {
-        // Attempt to execute non-executable page
+    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);
+            uintptr_t pa =
+              (uintptr_t)vmm_dup_page(__current->pid, PG_ENTRY_ADDR(*pte));
+            pmm_free_page(__current->pid, *pte & ~0xFFF);
+            *pte = (*pte & 0xFFF) | pa | PG_WRITE;
+            goto resolved;
+        }
+        // impossible cases or accessing privileged page

         goto segv_term;
     }
 
         goto segv_term;
     }
 

-    x86_pte_t* pte = CURPROC_PTE(ptr >> 12);
-    if (*pte & PG_PRESENT) {
-        if ((hit_region->attr & REGION_PERM_MASK) == (REGION_RSHARED | REGION_READ)) {
-            // normal page fault, do COW
-                uintptr_t pa = (uintptr_t)vmm_dup_page(__current->pid, PG_ENTRY_ADDR(*pte));
-                pmm_free_page(__current->pid, *pte & ~0xFFF);
-                *pte = (*pte & 0xFFF) | pa | PG_WRITE;
-                return;
+    // 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;

         }
         }

-        else {
-            // impossible cases or accessing privileged page
-            goto segv_term;
+        // permission denied on anon page
+        goto segv_term;
+    }
+
+    // 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;

         }
         }

-    } else {
-        if (!(*pte)) {
-            // Invalid location
+
+        cpu_invplg(pte);
+        *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 segv_term;
         }

-        // page not present, bring it from disk or somewhere else
-        __print_panic_msg("WIP page fault route", param);
-        while (1);
+
+        goto resolved;

     }
 
     }
 

+    // 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:
 segv_term:

-    kprintf(KERROR "(pid: %d) Segmentation fault on %p\n", __current->pid, ptr);
-    terminate_proc(LXSEGFAULT);
+    kprintf(KERROR "(pid: %d) Segmentation fault on %p (%p:%p)\n",
+            __current->pid,
+            ptr,
+            param->cs,
+            param->eip);
+    __SIGSET(__current->sig_pending, _SIGSEGV);
+    schedule();

     // should not reach
     // should not reach

+    while (1)
+        ;
+
+resolved:
+    cpu_invplg(ptr);
+    return;
+}
+
+int
+do_kernel(v_mapping* mapping)
+{
+    uintptr_t addr = mapping->va;
+
+    // TODO
+
+    return 0;
+done:
+    return 1;

 }
\ No newline at end of file
 }
\ No newline at end of file