x86: make sure initmem is writable

[linux-2.6] / arch / x86 / mm / fault_64.c

diff --git a/arch/x86/mm/fault_64.c b/arch/x86/mm/fault_64.c
index c6b3ad515cf12c5e071c0e2761fd049db4be914f..dd26e680a43166ca9abe0ced6390989edddf4349 100644 (file)
--- a/arch/x86/mm/fault_64.c
+++ b/arch/x86/mm/fault_64.c
@@ -223,6 +223,7 @@ KERN_ERR "******* Your BIOS seems to not contain a fix for K8 errata #93\n"
 KERN_ERR "******* Working around it, but it may cause SEGVs or burn power.\n"
 KERN_ERR "******* Please consider a BIOS update.\n"
 KERN_ERR "******* Disabling USB legacy in the BIOS may also help.\n";
+#endif
 
 /* Workaround for K8 erratum #93 & buggy BIOS.
    BIOS SMM functions are required to use a specific workaround
@@ -230,10 +231,12 @@ KERN_ERR "******* Disabling USB legacy in the BIOS may also help.\n";
    A lot of BIOS that didn't get tested properly miss this.
    The OS sees this as a page fault with the upper 32bits of RIP cleared.
    Try to work around it here.
-   Note we only handle faults in kernel here. */
-
+   Note we only handle faults in kernel here.
+   Does nothing for X86_32
+ */
 static int is_errata93(struct pt_regs *regs, unsigned long address)
 {
+#ifdef CONFIG_X86_64
        static int warned;
        if (address != regs->ip)
                return 0;
@@ -249,9 +252,9 @@ static int is_errata93(struct pt_regs *regs, unsigned long address)
                regs->ip = address;
                return 1;
        }
+#endif
        return 0;
 }
-#endif
 
 static noinline void pgtable_bad(unsigned long address, struct pt_regs *regs,
                                 unsigned long error_code)
@@ -278,6 +281,26 @@ static noinline void pgtable_bad(unsigned long address, struct pt_regs *regs,
  */
 static int vmalloc_fault(unsigned long address)
 {
+#ifdef CONFIG_X86_32
+       unsigned long pgd_paddr;
+       pmd_t *pmd_k;
+       pte_t *pte_k;
+       /*
+        * Synchronize this task's top level page-table
+        * with the 'reference' page table.
+        *
+        * Do _not_ use "current" here. We might be inside
+        * an interrupt in the middle of a task switch..
+        */
+       pgd_paddr = read_cr3();
+       pmd_k = vmalloc_sync_one(__va(pgd_paddr), address);
+       if (!pmd_k)
+               return -1;
+       pte_k = pte_offset_kernel(pmd_k, address);
+       if (!pte_present(*pte_k))
+               return -1;
+       return 0;
+#else
        pgd_t *pgd, *pgd_ref;
        pud_t *pud, *pud_ref;
        pmd_t *pmd, *pmd_ref;
@@ -321,6 +344,7 @@ static int vmalloc_fault(unsigned long address)
        if (!pte_present(*pte) || pte_pfn(*pte) != pte_pfn(*pte_ref))
                BUG();
        return 0;
+#endif
 }
 
 int show_unhandled_signals = 1;
@@ -440,8 +464,11 @@ asmlinkage void __kprobes do_page_fault(struct pt_regs *regs,
        if (!(vma->vm_flags & VM_GROWSDOWN))
                goto bad_area;
        if (error_code & PF_USER) {
-               /* Allow userspace just enough access below the stack pointer
-                * to let the 'enter' instruction work.
+               /*
+                * Accessing the stack below %sp is always a bug.
+                * The large cushion allows instructions like enter
+                * and pusha to work.  ("enter $65535,$31" pushes
+                * 32 pointers and then decrements %sp by 65535.)
                 */
                if (address + 65536 + 32 * sizeof(unsigned long) < regs->sp)
                        goto bad_area;
@@ -487,6 +514,17 @@ good_area:
                tsk->maj_flt++;
        else
                tsk->min_flt++;
+
+#ifdef CONFIG_X86_32
+       /*
+        * Did it hit the DOS screen memory VA from vm86 mode?
+        */
+       if (v8086_mode(regs)) {
+               unsigned long bit = (address - 0xA0000) >> PAGE_SHIFT;
+               if (bit < 32)
+                       tsk->thread.screen_bitmap |= 1 << bit;
+       }
+#endif
        up_read(&mm->mmap_sem);
        return;
 
@@ -522,10 +560,16 @@ bad_area_nosemaphore:
                if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
                    printk_ratelimit()) {
                        printk(
-                      "%s%s[%d]: segfault at %lx ip %lx sp %lx error %lx\n",
-                                       tsk->pid > 1 ? KERN_INFO : KERN_EMERG,
-                                       tsk->comm, tsk->pid, address, regs->ip,
-                                       regs->sp, error_code);
+#ifdef CONFIG_X86_32
+                       "%s%s[%d]: segfault at %lx ip %08lx sp %08lx error %lx",
+#else
+                       "%s%s[%d]: segfault at %lx ip %lx sp %lx error %lx",
+#endif
+                       task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG,
+                       tsk->comm, task_pid_nr(tsk), address, regs->ip,
+                       regs->sp, error_code);
+                       print_vma_addr(" in ", regs->ip);
+                       printk("\n");
                }
 
                tsk->thread.cr2 = address;
@@ -564,7 +608,7 @@ no_context:
        else
                printk(KERN_ALERT "Unable to handle kernel paging request");
        printk(" at %016lx RIP: \n" KERN_ALERT, address);
-       printk_address(regs->ip, regs->bp);
+       printk_address(regs->ip, 1);
        dump_pagetable(address);
        tsk->thread.cr2 = address;
        tsk->thread.trap_no = 14;
@@ -609,10 +653,12 @@ LIST_HEAD(pgd_list);
 
 void vmalloc_sync_all(void)
 {
-       /* Note that races in the updates of insync and start aren't
-          problematic:
-          insync can only get set bits added, and updates to start are only
-          improving performance (without affecting correctness if undone). */
+       /*
+        * Note that races in the updates of insync and start aren't
+        * problematic: insync can only get set bits added, and updates to
+        * start are only improving performance (without affecting correctness
+        * if undone).
+        */
        static DECLARE_BITMAP(insync, PTRS_PER_PGD);
        static unsigned long start = VMALLOC_START & PGDIR_MASK;
        unsigned long address;