i386: prepare shared mm/fault.c
authorThomas Gleixner <tglx@linutronix.de>
Thu, 11 Oct 2007 09:13:57 +0000 (11:13 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Thu, 11 Oct 2007 09:13:57 +0000 (11:13 +0200)
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/i386/mm/Makefile
arch/i386/mm/fault.c [deleted file]
arch/i386/mm/fault_32.c [new file with mode: 0644]

index 4d7c5af4765d1eef065e7be2cf1b3b123bb64f06..362b4ad082de26dc807e1c6078ac79a05c438882 100644 (file)
@@ -2,7 +2,7 @@
 # Makefile for the linux i386-specific parts of the memory manager.
 #
 
-obj-y  := init_32.o pgtable_32.o fault.o ioremap_32.o extable_32.o pageattr_32.o mmap_32.o
+obj-y  := init_32.o pgtable_32.o fault_32.o ioremap_32.o extable_32.o pageattr_32.o mmap_32.o
 
 obj-$(CONFIG_NUMA) += discontig_32.o
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
diff --git a/arch/i386/mm/fault.c b/arch/i386/mm/fault.c
deleted file mode 100644 (file)
index fcb38e7..0000000
+++ /dev/null
@@ -1,657 +0,0 @@
-/*
- *  linux/arch/i386/mm/fault.c
- *
- *  Copyright (C) 1995  Linus Torvalds
- */
-
-#include <linux/signal.h>
-#include <linux/sched.h>
-#include <linux/kernel.h>
-#include <linux/errno.h>
-#include <linux/string.h>
-#include <linux/types.h>
-#include <linux/ptrace.h>
-#include <linux/mman.h>
-#include <linux/mm.h>
-#include <linux/smp.h>
-#include <linux/interrupt.h>
-#include <linux/init.h>
-#include <linux/tty.h>
-#include <linux/vt_kern.h>             /* For unblank_screen() */
-#include <linux/highmem.h>
-#include <linux/bootmem.h>             /* for max_low_pfn */
-#include <linux/vmalloc.h>
-#include <linux/module.h>
-#include <linux/kprobes.h>
-#include <linux/uaccess.h>
-#include <linux/kdebug.h>
-
-#include <asm/system.h>
-#include <asm/desc.h>
-#include <asm/segment.h>
-
-extern void die(const char *,struct pt_regs *,long);
-
-static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
-
-int register_page_fault_notifier(struct notifier_block *nb)
-{
-       vmalloc_sync_all();
-       return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
-}
-EXPORT_SYMBOL_GPL(register_page_fault_notifier);
-
-int unregister_page_fault_notifier(struct notifier_block *nb)
-{
-       return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
-}
-EXPORT_SYMBOL_GPL(unregister_page_fault_notifier);
-
-static inline int notify_page_fault(struct pt_regs *regs, long err)
-{
-       struct die_args args = {
-               .regs = regs,
-               .str = "page fault",
-               .err = err,
-               .trapnr = 14,
-               .signr = SIGSEGV
-       };
-       return atomic_notifier_call_chain(&notify_page_fault_chain,
-                                         DIE_PAGE_FAULT, &args);
-}
-
-/*
- * Return EIP plus the CS segment base.  The segment limit is also
- * adjusted, clamped to the kernel/user address space (whichever is
- * appropriate), and returned in *eip_limit.
- *
- * The segment is checked, because it might have been changed by another
- * task between the original faulting instruction and here.
- *
- * If CS is no longer a valid code segment, or if EIP is beyond the
- * limit, or if it is a kernel address when CS is not a kernel segment,
- * then the returned value will be greater than *eip_limit.
- * 
- * This is slow, but is very rarely executed.
- */
-static inline unsigned long get_segment_eip(struct pt_regs *regs,
-                                           unsigned long *eip_limit)
-{
-       unsigned long eip = regs->eip;
-       unsigned seg = regs->xcs & 0xffff;
-       u32 seg_ar, seg_limit, base, *desc;
-
-       /* Unlikely, but must come before segment checks. */
-       if (unlikely(regs->eflags & VM_MASK)) {
-               base = seg << 4;
-               *eip_limit = base + 0xffff;
-               return base + (eip & 0xffff);
-       }
-
-       /* The standard kernel/user address space limit. */
-       *eip_limit = user_mode(regs) ? USER_DS.seg : KERNEL_DS.seg;
-       
-       /* By far the most common cases. */
-       if (likely(SEGMENT_IS_FLAT_CODE(seg)))
-               return eip;
-
-       /* Check the segment exists, is within the current LDT/GDT size,
-          that kernel/user (ring 0..3) has the appropriate privilege,
-          that it's a code segment, and get the limit. */
-       __asm__ ("larl %3,%0; lsll %3,%1"
-                : "=&r" (seg_ar), "=r" (seg_limit) : "0" (0), "rm" (seg));
-       if ((~seg_ar & 0x9800) || eip > seg_limit) {
-               *eip_limit = 0;
-               return 1;        /* So that returned eip > *eip_limit. */
-       }
-
-       /* Get the GDT/LDT descriptor base. 
-          When you look for races in this code remember that
-          LDT and other horrors are only used in user space. */
-       if (seg & (1<<2)) {
-               /* Must lock the LDT while reading it. */
-               down(&current->mm->context.sem);
-               desc = current->mm->context.ldt;
-               desc = (void *)desc + (seg & ~7);
-       } else {
-               /* Must disable preemption while reading the GDT. */
-               desc = (u32 *)get_cpu_gdt_table(get_cpu());
-               desc = (void *)desc + (seg & ~7);
-       }
-
-       /* Decode the code segment base from the descriptor */
-       base = get_desc_base((unsigned long *)desc);
-
-       if (seg & (1<<2)) { 
-               up(&current->mm->context.sem);
-       } else
-               put_cpu();
-
-       /* Adjust EIP and segment limit, and clamp at the kernel limit.
-          It's legitimate for segments to wrap at 0xffffffff. */
-       seg_limit += base;
-       if (seg_limit < *eip_limit && seg_limit >= base)
-               *eip_limit = seg_limit;
-       return eip + base;
-}
-
-/* 
- * Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
- * Check that here and ignore it.
- */
-static int __is_prefetch(struct pt_regs *regs, unsigned long addr)
-{ 
-       unsigned long limit;
-       unsigned char *instr = (unsigned char *)get_segment_eip (regs, &limit);
-       int scan_more = 1;
-       int prefetch = 0; 
-       int i;
-
-       for (i = 0; scan_more && i < 15; i++) { 
-               unsigned char opcode;
-               unsigned char instr_hi;
-               unsigned char instr_lo;
-
-               if (instr > (unsigned char *)limit)
-                       break;
-               if (probe_kernel_address(instr, opcode))
-                       break; 
-
-               instr_hi = opcode & 0xf0; 
-               instr_lo = opcode & 0x0f; 
-               instr++;
-
-               switch (instr_hi) { 
-               case 0x20:
-               case 0x30:
-                       /* Values 0x26,0x2E,0x36,0x3E are valid x86 prefixes. */
-                       scan_more = ((instr_lo & 7) == 0x6);
-                       break;
-                       
-               case 0x60:
-                       /* 0x64 thru 0x67 are valid prefixes in all modes. */
-                       scan_more = (instr_lo & 0xC) == 0x4;
-                       break;          
-               case 0xF0:
-                       /* 0xF0, 0xF2, and 0xF3 are valid prefixes */
-                       scan_more = !instr_lo || (instr_lo>>1) == 1;
-                       break;                  
-               case 0x00:
-                       /* Prefetch instruction is 0x0F0D or 0x0F18 */
-                       scan_more = 0;
-                       if (instr > (unsigned char *)limit)
-                               break;
-                       if (probe_kernel_address(instr, opcode))
-                               break;
-                       prefetch = (instr_lo == 0xF) &&
-                               (opcode == 0x0D || opcode == 0x18);
-                       break;                  
-               default:
-                       scan_more = 0;
-                       break;
-               } 
-       }
-       return prefetch;
-}
-
-static inline int is_prefetch(struct pt_regs *regs, unsigned long addr,
-                             unsigned long error_code)
-{
-       if (unlikely(boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
-                    boot_cpu_data.x86 >= 6)) {
-               /* Catch an obscure case of prefetch inside an NX page. */
-               if (nx_enabled && (error_code & 16))
-                       return 0;
-               return __is_prefetch(regs, addr);
-       }
-       return 0;
-} 
-
-static noinline void force_sig_info_fault(int si_signo, int si_code,
-       unsigned long address, struct task_struct *tsk)
-{
-       siginfo_t info;
-
-       info.si_signo = si_signo;
-       info.si_errno = 0;
-       info.si_code = si_code;
-       info.si_addr = (void __user *)address;
-       force_sig_info(si_signo, &info, tsk);
-}
-
-fastcall void do_invalid_op(struct pt_regs *, unsigned long);
-
-static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
-{
-       unsigned index = pgd_index(address);
-       pgd_t *pgd_k;
-       pud_t *pud, *pud_k;
-       pmd_t *pmd, *pmd_k;
-
-       pgd += index;
-       pgd_k = init_mm.pgd + index;
-
-       if (!pgd_present(*pgd_k))
-               return NULL;
-
-       /*
-        * set_pgd(pgd, *pgd_k); here would be useless on PAE
-        * and redundant with the set_pmd() on non-PAE. As would
-        * set_pud.
-        */
-
-       pud = pud_offset(pgd, address);
-       pud_k = pud_offset(pgd_k, address);
-       if (!pud_present(*pud_k))
-               return NULL;
-
-       pmd = pmd_offset(pud, address);
-       pmd_k = pmd_offset(pud_k, address);
-       if (!pmd_present(*pmd_k))
-               return NULL;
-       if (!pmd_present(*pmd)) {
-               set_pmd(pmd, *pmd_k);
-               arch_flush_lazy_mmu_mode();
-       } else
-               BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));
-       return pmd_k;
-}
-
-/*
- * Handle a fault on the vmalloc or module mapping area
- *
- * This assumes no large pages in there.
- */
-static inline int vmalloc_fault(unsigned long address)
-{
-       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;
-}
-
-int show_unhandled_signals = 1;
-
-/*
- * This routine handles page faults.  It determines the address,
- * and the problem, and then passes it off to one of the appropriate
- * routines.
- *
- * error_code:
- *     bit 0 == 0 means no page found, 1 means protection fault
- *     bit 1 == 0 means read, 1 means write
- *     bit 2 == 0 means kernel, 1 means user-mode
- *     bit 3 == 1 means use of reserved bit detected
- *     bit 4 == 1 means fault was an instruction fetch
- */
-fastcall void __kprobes do_page_fault(struct pt_regs *regs,
-                                     unsigned long error_code)
-{
-       struct task_struct *tsk;
-       struct mm_struct *mm;
-       struct vm_area_struct * vma;
-       unsigned long address;
-       int write, si_code;
-       int fault;
-
-       /* get the address */
-        address = read_cr2();
-
-       tsk = current;
-
-       si_code = SEGV_MAPERR;
-
-       /*
-        * We fault-in kernel-space virtual memory on-demand. The
-        * 'reference' page table is init_mm.pgd.
-        *
-        * NOTE! We MUST NOT take any locks for this case. We may
-        * be in an interrupt or a critical region, and should
-        * only copy the information from the master page table,
-        * nothing more.
-        *
-        * This verifies that the fault happens in kernel space
-        * (error_code & 4) == 0, and that the fault was not a
-        * protection error (error_code & 9) == 0.
-        */
-       if (unlikely(address >= TASK_SIZE)) {
-               if (!(error_code & 0x0000000d) && vmalloc_fault(address) >= 0)
-                       return;
-               if (notify_page_fault(regs, error_code) == NOTIFY_STOP)
-                       return;
-               /*
-                * Don't take the mm semaphore here. If we fixup a prefetch
-                * fault we could otherwise deadlock.
-                */
-               goto bad_area_nosemaphore;
-       }
-
-       if (notify_page_fault(regs, error_code) == NOTIFY_STOP)
-               return;
-
-       /* It's safe to allow irq's after cr2 has been saved and the vmalloc
-          fault has been handled. */
-       if (regs->eflags & (X86_EFLAGS_IF|VM_MASK))
-               local_irq_enable();
-
-       mm = tsk->mm;
-
-       /*
-        * If we're in an interrupt, have no user context or are running in an
-        * atomic region then we must not take the fault..
-        */
-       if (in_atomic() || !mm)
-               goto bad_area_nosemaphore;
-
-       /* When running in the kernel we expect faults to occur only to
-        * addresses in user space.  All other faults represent errors in the
-        * kernel and should generate an OOPS.  Unfortunatly, in the case of an
-        * erroneous fault occurring in a code path which already holds mmap_sem
-        * we will deadlock attempting to validate the fault against the
-        * address space.  Luckily the kernel only validly references user
-        * space from well defined areas of code, which are listed in the
-        * exceptions table.
-        *
-        * As the vast majority of faults will be valid we will only perform
-        * the source reference check when there is a possibilty of a deadlock.
-        * Attempt to lock the address space, if we cannot we then validate the
-        * source.  If this is invalid we can skip the address space check,
-        * thus avoiding the deadlock.
-        */
-       if (!down_read_trylock(&mm->mmap_sem)) {
-               if ((error_code & 4) == 0 &&
-                   !search_exception_tables(regs->eip))
-                       goto bad_area_nosemaphore;
-               down_read(&mm->mmap_sem);
-       }
-
-       vma = find_vma(mm, address);
-       if (!vma)
-               goto bad_area;
-       if (vma->vm_start <= address)
-               goto good_area;
-       if (!(vma->vm_flags & VM_GROWSDOWN))
-               goto bad_area;
-       if (error_code & 4) {
-               /*
-                * Accessing the stack below %esp is always a bug.
-                * The large cushion allows instructions like enter
-                * and pusha to work.  ("enter $65535,$31" pushes
-                * 32 pointers and then decrements %esp by 65535.)
-                */
-               if (address + 65536 + 32 * sizeof(unsigned long) < regs->esp)
-                       goto bad_area;
-       }
-       if (expand_stack(vma, address))
-               goto bad_area;
-/*
- * Ok, we have a good vm_area for this memory access, so
- * we can handle it..
- */
-good_area:
-       si_code = SEGV_ACCERR;
-       write = 0;
-       switch (error_code & 3) {
-               default:        /* 3: write, present */
-                               /* fall through */
-               case 2:         /* write, not present */
-                       if (!(vma->vm_flags & VM_WRITE))
-                               goto bad_area;
-                       write++;
-                       break;
-               case 1:         /* read, present */
-                       goto bad_area;
-               case 0:         /* read, not present */
-                       if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
-                               goto bad_area;
-       }
-
- survive:
-       /*
-        * If for any reason at all we couldn't handle the fault,
-        * make sure we exit gracefully rather than endlessly redo
-        * the fault.
-        */
-       fault = handle_mm_fault(mm, vma, address, write);
-       if (unlikely(fault & VM_FAULT_ERROR)) {
-               if (fault & VM_FAULT_OOM)
-                       goto out_of_memory;
-               else if (fault & VM_FAULT_SIGBUS)
-                       goto do_sigbus;
-               BUG();
-       }
-       if (fault & VM_FAULT_MAJOR)
-               tsk->maj_flt++;
-       else
-               tsk->min_flt++;
-
-       /*
-        * Did it hit the DOS screen memory VA from vm86 mode?
-        */
-       if (regs->eflags & VM_MASK) {
-               unsigned long bit = (address - 0xA0000) >> PAGE_SHIFT;
-               if (bit < 32)
-                       tsk->thread.screen_bitmap |= 1 << bit;
-       }
-       up_read(&mm->mmap_sem);
-       return;
-
-/*
- * Something tried to access memory that isn't in our memory map..
- * Fix it, but check if it's kernel or user first..
- */
-bad_area:
-       up_read(&mm->mmap_sem);
-
-bad_area_nosemaphore:
-       /* User mode accesses just cause a SIGSEGV */
-       if (error_code & 4) {
-               /*
-                * It's possible to have interrupts off here.
-                */
-               local_irq_enable();
-
-               /* 
-                * Valid to do another page fault here because this one came 
-                * from user space.
-                */
-               if (is_prefetch(regs, address, error_code))
-                       return;
-
-               if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
-                   printk_ratelimit()) {
-                       printk("%s%s[%d]: segfault at %08lx eip %08lx "
-                           "esp %08lx error %lx\n",
-                           tsk->pid > 1 ? KERN_INFO : KERN_EMERG,
-                           tsk->comm, tsk->pid, address, regs->eip,
-                           regs->esp, error_code);
-               }
-               tsk->thread.cr2 = address;
-               /* Kernel addresses are always protection faults */
-               tsk->thread.error_code = error_code | (address >= TASK_SIZE);
-               tsk->thread.trap_no = 14;
-               force_sig_info_fault(SIGSEGV, si_code, address, tsk);
-               return;
-       }
-
-#ifdef CONFIG_X86_F00F_BUG
-       /*
-        * Pentium F0 0F C7 C8 bug workaround.
-        */
-       if (boot_cpu_data.f00f_bug) {
-               unsigned long nr;
-               
-               nr = (address - idt_descr.address) >> 3;
-
-               if (nr == 6) {
-                       do_invalid_op(regs, 0);
-                       return;
-               }
-       }
-#endif
-
-no_context:
-       /* Are we prepared to handle this kernel fault?  */
-       if (fixup_exception(regs))
-               return;
-
-       /* 
-        * Valid to do another page fault here, because if this fault
-        * had been triggered by is_prefetch fixup_exception would have 
-        * handled it.
-        */
-       if (is_prefetch(regs, address, error_code))
-               return;
-
-/*
- * Oops. The kernel tried to access some bad page. We'll have to
- * terminate things with extreme prejudice.
- */
-
-       bust_spinlocks(1);
-
-       if (oops_may_print()) {
-               __typeof__(pte_val(__pte(0))) page;
-
-#ifdef CONFIG_X86_PAE
-               if (error_code & 16) {
-                       pte_t *pte = lookup_address(address);
-
-                       if (pte && pte_present(*pte) && !pte_exec_kernel(*pte))
-                               printk(KERN_CRIT "kernel tried to execute "
-                                       "NX-protected page - exploit attempt? "
-                                       "(uid: %d)\n", current->uid);
-               }
-#endif
-               if (address < PAGE_SIZE)
-                       printk(KERN_ALERT "BUG: unable to handle kernel NULL "
-                                       "pointer dereference");
-               else
-                       printk(KERN_ALERT "BUG: unable to handle kernel paging"
-                                       " request");
-               printk(" at virtual address %08lx\n",address);
-               printk(KERN_ALERT " printing eip:\n");
-               printk("%08lx\n", regs->eip);
-
-               page = read_cr3();
-               page = ((__typeof__(page) *) __va(page))[address >> PGDIR_SHIFT];
-#ifdef CONFIG_X86_PAE
-               printk(KERN_ALERT "*pdpt = %016Lx\n", page);
-               if ((page >> PAGE_SHIFT) < max_low_pfn
-                   && page & _PAGE_PRESENT) {
-                       page &= PAGE_MASK;
-                       page = ((__typeof__(page) *) __va(page))[(address >> PMD_SHIFT)
-                                                                & (PTRS_PER_PMD - 1)];
-                       printk(KERN_ALERT "*pde = %016Lx\n", page);
-                       page &= ~_PAGE_NX;
-               }
-#else
-               printk(KERN_ALERT "*pde = %08lx\n", page);
-#endif
-
-               /*
-                * We must not directly access the pte in the highpte
-                * case if the page table is located in highmem.
-                * And let's rather not kmap-atomic the pte, just in case
-                * it's allocated already.
-                */
-               if ((page >> PAGE_SHIFT) < max_low_pfn
-                   && (page & _PAGE_PRESENT)) {
-                       page &= PAGE_MASK;
-                       page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
-                                                                & (PTRS_PER_PTE - 1)];
-                       printk(KERN_ALERT "*pte = %0*Lx\n", sizeof(page)*2, (u64)page);
-               }
-       }
-
-       tsk->thread.cr2 = address;
-       tsk->thread.trap_no = 14;
-       tsk->thread.error_code = error_code;
-       die("Oops", regs, error_code);
-       bust_spinlocks(0);
-       do_exit(SIGKILL);
-
-/*
- * We ran out of memory, or some other thing happened to us that made
- * us unable to handle the page fault gracefully.
- */
-out_of_memory:
-       up_read(&mm->mmap_sem);
-       if (is_init(tsk)) {
-               yield();
-               down_read(&mm->mmap_sem);
-               goto survive;
-       }
-       printk("VM: killing process %s\n", tsk->comm);
-       if (error_code & 4)
-               do_exit(SIGKILL);
-       goto no_context;
-
-do_sigbus:
-       up_read(&mm->mmap_sem);
-
-       /* Kernel mode? Handle exceptions or die */
-       if (!(error_code & 4))
-               goto no_context;
-
-       /* User space => ok to do another page fault */
-       if (is_prefetch(regs, address, error_code))
-               return;
-
-       tsk->thread.cr2 = address;
-       tsk->thread.error_code = error_code;
-       tsk->thread.trap_no = 14;
-       force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
-}
-
-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).
-        */
-       static DECLARE_BITMAP(insync, PTRS_PER_PGD);
-       static unsigned long start = TASK_SIZE;
-       unsigned long address;
-
-       if (SHARED_KERNEL_PMD)
-               return;
-
-       BUILD_BUG_ON(TASK_SIZE & ~PGDIR_MASK);
-       for (address = start; address >= TASK_SIZE; address += PGDIR_SIZE) {
-               if (!test_bit(pgd_index(address), insync)) {
-                       unsigned long flags;
-                       struct page *page;
-
-                       spin_lock_irqsave(&pgd_lock, flags);
-                       for (page = pgd_list; page; page =
-                                       (struct page *)page->index)
-                               if (!vmalloc_sync_one(page_address(page),
-                                                               address)) {
-                                       BUG_ON(page != pgd_list);
-                                       break;
-                               }
-                       spin_unlock_irqrestore(&pgd_lock, flags);
-                       if (!page)
-                               set_bit(pgd_index(address), insync);
-               }
-               if (address == start && test_bit(pgd_index(address), insync))
-                       start = address + PGDIR_SIZE;
-       }
-}
diff --git a/arch/i386/mm/fault_32.c b/arch/i386/mm/fault_32.c
new file mode 100644 (file)
index 0000000..fcb38e7
--- /dev/null
@@ -0,0 +1,657 @@
+/*
+ *  linux/arch/i386/mm/fault.c
+ *
+ *  Copyright (C) 1995  Linus Torvalds
+ */
+
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/smp.h>
+#include <linux/interrupt.h>
+#include <linux/init.h>
+#include <linux/tty.h>
+#include <linux/vt_kern.h>             /* For unblank_screen() */
+#include <linux/highmem.h>
+#include <linux/bootmem.h>             /* for max_low_pfn */
+#include <linux/vmalloc.h>
+#include <linux/module.h>
+#include <linux/kprobes.h>
+#include <linux/uaccess.h>
+#include <linux/kdebug.h>
+
+#include <asm/system.h>
+#include <asm/desc.h>
+#include <asm/segment.h>
+
+extern void die(const char *,struct pt_regs *,long);
+
+static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
+
+int register_page_fault_notifier(struct notifier_block *nb)
+{
+       vmalloc_sync_all();
+       return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
+}
+EXPORT_SYMBOL_GPL(register_page_fault_notifier);
+
+int unregister_page_fault_notifier(struct notifier_block *nb)
+{
+       return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
+}
+EXPORT_SYMBOL_GPL(unregister_page_fault_notifier);
+
+static inline int notify_page_fault(struct pt_regs *regs, long err)
+{
+       struct die_args args = {
+               .regs = regs,
+               .str = "page fault",
+               .err = err,
+               .trapnr = 14,
+               .signr = SIGSEGV
+       };
+       return atomic_notifier_call_chain(&notify_page_fault_chain,
+                                         DIE_PAGE_FAULT, &args);
+}
+
+/*
+ * Return EIP plus the CS segment base.  The segment limit is also
+ * adjusted, clamped to the kernel/user address space (whichever is
+ * appropriate), and returned in *eip_limit.
+ *
+ * The segment is checked, because it might have been changed by another
+ * task between the original faulting instruction and here.
+ *
+ * If CS is no longer a valid code segment, or if EIP is beyond the
+ * limit, or if it is a kernel address when CS is not a kernel segment,
+ * then the returned value will be greater than *eip_limit.
+ * 
+ * This is slow, but is very rarely executed.
+ */
+static inline unsigned long get_segment_eip(struct pt_regs *regs,
+                                           unsigned long *eip_limit)
+{
+       unsigned long eip = regs->eip;
+       unsigned seg = regs->xcs & 0xffff;
+       u32 seg_ar, seg_limit, base, *desc;
+
+       /* Unlikely, but must come before segment checks. */
+       if (unlikely(regs->eflags & VM_MASK)) {
+               base = seg << 4;
+               *eip_limit = base + 0xffff;
+               return base + (eip & 0xffff);
+       }
+
+       /* The standard kernel/user address space limit. */
+       *eip_limit = user_mode(regs) ? USER_DS.seg : KERNEL_DS.seg;
+       
+       /* By far the most common cases. */
+       if (likely(SEGMENT_IS_FLAT_CODE(seg)))
+               return eip;
+
+       /* Check the segment exists, is within the current LDT/GDT size,
+          that kernel/user (ring 0..3) has the appropriate privilege,
+          that it's a code segment, and get the limit. */
+       __asm__ ("larl %3,%0; lsll %3,%1"
+                : "=&r" (seg_ar), "=r" (seg_limit) : "0" (0), "rm" (seg));
+       if ((~seg_ar & 0x9800) || eip > seg_limit) {
+               *eip_limit = 0;
+               return 1;        /* So that returned eip > *eip_limit. */
+       }
+
+       /* Get the GDT/LDT descriptor base. 
+          When you look for races in this code remember that
+          LDT and other horrors are only used in user space. */
+       if (seg & (1<<2)) {
+               /* Must lock the LDT while reading it. */
+               down(&current->mm->context.sem);
+               desc = current->mm->context.ldt;
+               desc = (void *)desc + (seg & ~7);
+       } else {
+               /* Must disable preemption while reading the GDT. */
+               desc = (u32 *)get_cpu_gdt_table(get_cpu());
+               desc = (void *)desc + (seg & ~7);
+       }
+
+       /* Decode the code segment base from the descriptor */
+       base = get_desc_base((unsigned long *)desc);
+
+       if (seg & (1<<2)) { 
+               up(&current->mm->context.sem);
+       } else
+               put_cpu();
+
+       /* Adjust EIP and segment limit, and clamp at the kernel limit.
+          It's legitimate for segments to wrap at 0xffffffff. */
+       seg_limit += base;
+       if (seg_limit < *eip_limit && seg_limit >= base)
+               *eip_limit = seg_limit;
+       return eip + base;
+}
+
+/* 
+ * Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
+ * Check that here and ignore it.
+ */
+static int __is_prefetch(struct pt_regs *regs, unsigned long addr)
+{ 
+       unsigned long limit;
+       unsigned char *instr = (unsigned char *)get_segment_eip (regs, &limit);
+       int scan_more = 1;
+       int prefetch = 0; 
+       int i;
+
+       for (i = 0; scan_more && i < 15; i++) { 
+               unsigned char opcode;
+               unsigned char instr_hi;
+               unsigned char instr_lo;
+
+               if (instr > (unsigned char *)limit)
+                       break;
+               if (probe_kernel_address(instr, opcode))
+                       break; 
+
+               instr_hi = opcode & 0xf0; 
+               instr_lo = opcode & 0x0f; 
+               instr++;
+
+               switch (instr_hi) { 
+               case 0x20:
+               case 0x30:
+                       /* Values 0x26,0x2E,0x36,0x3E are valid x86 prefixes. */
+                       scan_more = ((instr_lo & 7) == 0x6);
+                       break;
+                       
+               case 0x60:
+                       /* 0x64 thru 0x67 are valid prefixes in all modes. */
+                       scan_more = (instr_lo & 0xC) == 0x4;
+                       break;          
+               case 0xF0:
+                       /* 0xF0, 0xF2, and 0xF3 are valid prefixes */
+                       scan_more = !instr_lo || (instr_lo>>1) == 1;
+                       break;                  
+               case 0x00:
+                       /* Prefetch instruction is 0x0F0D or 0x0F18 */
+                       scan_more = 0;
+                       if (instr > (unsigned char *)limit)
+                               break;
+                       if (probe_kernel_address(instr, opcode))
+                               break;
+                       prefetch = (instr_lo == 0xF) &&
+                               (opcode == 0x0D || opcode == 0x18);
+                       break;                  
+               default:
+                       scan_more = 0;
+                       break;
+               } 
+       }
+       return prefetch;
+}
+
+static inline int is_prefetch(struct pt_regs *regs, unsigned long addr,
+                             unsigned long error_code)
+{
+       if (unlikely(boot_cpu_data.x86_vendor == X86_VENDOR_AMD &&
+                    boot_cpu_data.x86 >= 6)) {
+               /* Catch an obscure case of prefetch inside an NX page. */
+               if (nx_enabled && (error_code & 16))
+                       return 0;
+               return __is_prefetch(regs, addr);
+       }
+       return 0;
+} 
+
+static noinline void force_sig_info_fault(int si_signo, int si_code,
+       unsigned long address, struct task_struct *tsk)
+{
+       siginfo_t info;
+
+       info.si_signo = si_signo;
+       info.si_errno = 0;
+       info.si_code = si_code;
+       info.si_addr = (void __user *)address;
+       force_sig_info(si_signo, &info, tsk);
+}
+
+fastcall void do_invalid_op(struct pt_regs *, unsigned long);
+
+static inline pmd_t *vmalloc_sync_one(pgd_t *pgd, unsigned long address)
+{
+       unsigned index = pgd_index(address);
+       pgd_t *pgd_k;
+       pud_t *pud, *pud_k;
+       pmd_t *pmd, *pmd_k;
+
+       pgd += index;
+       pgd_k = init_mm.pgd + index;
+
+       if (!pgd_present(*pgd_k))
+               return NULL;
+
+       /*
+        * set_pgd(pgd, *pgd_k); here would be useless on PAE
+        * and redundant with the set_pmd() on non-PAE. As would
+        * set_pud.
+        */
+
+       pud = pud_offset(pgd, address);
+       pud_k = pud_offset(pgd_k, address);
+       if (!pud_present(*pud_k))
+               return NULL;
+
+       pmd = pmd_offset(pud, address);
+       pmd_k = pmd_offset(pud_k, address);
+       if (!pmd_present(*pmd_k))
+               return NULL;
+       if (!pmd_present(*pmd)) {
+               set_pmd(pmd, *pmd_k);
+               arch_flush_lazy_mmu_mode();
+       } else
+               BUG_ON(pmd_page(*pmd) != pmd_page(*pmd_k));
+       return pmd_k;
+}
+
+/*
+ * Handle a fault on the vmalloc or module mapping area
+ *
+ * This assumes no large pages in there.
+ */
+static inline int vmalloc_fault(unsigned long address)
+{
+       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;
+}
+
+int show_unhandled_signals = 1;
+
+/*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ *
+ * error_code:
+ *     bit 0 == 0 means no page found, 1 means protection fault
+ *     bit 1 == 0 means read, 1 means write
+ *     bit 2 == 0 means kernel, 1 means user-mode
+ *     bit 3 == 1 means use of reserved bit detected
+ *     bit 4 == 1 means fault was an instruction fetch
+ */
+fastcall void __kprobes do_page_fault(struct pt_regs *regs,
+                                     unsigned long error_code)
+{
+       struct task_struct *tsk;
+       struct mm_struct *mm;
+       struct vm_area_struct * vma;
+       unsigned long address;
+       int write, si_code;
+       int fault;
+
+       /* get the address */
+        address = read_cr2();
+
+       tsk = current;
+
+       si_code = SEGV_MAPERR;
+
+       /*
+        * We fault-in kernel-space virtual memory on-demand. The
+        * 'reference' page table is init_mm.pgd.
+        *
+        * NOTE! We MUST NOT take any locks for this case. We may
+        * be in an interrupt or a critical region, and should
+        * only copy the information from the master page table,
+        * nothing more.
+        *
+        * This verifies that the fault happens in kernel space
+        * (error_code & 4) == 0, and that the fault was not a
+        * protection error (error_code & 9) == 0.
+        */
+       if (unlikely(address >= TASK_SIZE)) {
+               if (!(error_code & 0x0000000d) && vmalloc_fault(address) >= 0)
+                       return;
+               if (notify_page_fault(regs, error_code) == NOTIFY_STOP)
+                       return;
+               /*
+                * Don't take the mm semaphore here. If we fixup a prefetch
+                * fault we could otherwise deadlock.
+                */
+               goto bad_area_nosemaphore;
+       }
+
+       if (notify_page_fault(regs, error_code) == NOTIFY_STOP)
+               return;
+
+       /* It's safe to allow irq's after cr2 has been saved and the vmalloc
+          fault has been handled. */
+       if (regs->eflags & (X86_EFLAGS_IF|VM_MASK))
+               local_irq_enable();
+
+       mm = tsk->mm;
+
+       /*
+        * If we're in an interrupt, have no user context or are running in an
+        * atomic region then we must not take the fault..
+        */
+       if (in_atomic() || !mm)
+               goto bad_area_nosemaphore;
+
+       /* When running in the kernel we expect faults to occur only to
+        * addresses in user space.  All other faults represent errors in the
+        * kernel and should generate an OOPS.  Unfortunatly, in the case of an
+        * erroneous fault occurring in a code path which already holds mmap_sem
+        * we will deadlock attempting to validate the fault against the
+        * address space.  Luckily the kernel only validly references user
+        * space from well defined areas of code, which are listed in the
+        * exceptions table.
+        *
+        * As the vast majority of faults will be valid we will only perform
+        * the source reference check when there is a possibilty of a deadlock.
+        * Attempt to lock the address space, if we cannot we then validate the
+        * source.  If this is invalid we can skip the address space check,
+        * thus avoiding the deadlock.
+        */
+       if (!down_read_trylock(&mm->mmap_sem)) {
+               if ((error_code & 4) == 0 &&
+                   !search_exception_tables(regs->eip))
+                       goto bad_area_nosemaphore;
+               down_read(&mm->mmap_sem);
+       }
+
+       vma = find_vma(mm, address);
+       if (!vma)
+               goto bad_area;
+       if (vma->vm_start <= address)
+               goto good_area;
+       if (!(vma->vm_flags & VM_GROWSDOWN))
+               goto bad_area;
+       if (error_code & 4) {
+               /*
+                * Accessing the stack below %esp is always a bug.
+                * The large cushion allows instructions like enter
+                * and pusha to work.  ("enter $65535,$31" pushes
+                * 32 pointers and then decrements %esp by 65535.)
+                */
+               if (address + 65536 + 32 * sizeof(unsigned long) < regs->esp)
+                       goto bad_area;
+       }
+       if (expand_stack(vma, address))
+               goto bad_area;
+/*
+ * Ok, we have a good vm_area for this memory access, so
+ * we can handle it..
+ */
+good_area:
+       si_code = SEGV_ACCERR;
+       write = 0;
+       switch (error_code & 3) {
+               default:        /* 3: write, present */
+                               /* fall through */
+               case 2:         /* write, not present */
+                       if (!(vma->vm_flags & VM_WRITE))
+                               goto bad_area;
+                       write++;
+                       break;
+               case 1:         /* read, present */
+                       goto bad_area;
+               case 0:         /* read, not present */
+                       if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
+                               goto bad_area;
+       }
+
+ survive:
+       /*
+        * If for any reason at all we couldn't handle the fault,
+        * make sure we exit gracefully rather than endlessly redo
+        * the fault.
+        */
+       fault = handle_mm_fault(mm, vma, address, write);
+       if (unlikely(fault & VM_FAULT_ERROR)) {
+               if (fault & VM_FAULT_OOM)
+                       goto out_of_memory;
+               else if (fault & VM_FAULT_SIGBUS)
+                       goto do_sigbus;
+               BUG();
+       }
+       if (fault & VM_FAULT_MAJOR)
+               tsk->maj_flt++;
+       else
+               tsk->min_flt++;
+
+       /*
+        * Did it hit the DOS screen memory VA from vm86 mode?
+        */
+       if (regs->eflags & VM_MASK) {
+               unsigned long bit = (address - 0xA0000) >> PAGE_SHIFT;
+               if (bit < 32)
+                       tsk->thread.screen_bitmap |= 1 << bit;
+       }
+       up_read(&mm->mmap_sem);
+       return;
+
+/*
+ * Something tried to access memory that isn't in our memory map..
+ * Fix it, but check if it's kernel or user first..
+ */
+bad_area:
+       up_read(&mm->mmap_sem);
+
+bad_area_nosemaphore:
+       /* User mode accesses just cause a SIGSEGV */
+       if (error_code & 4) {
+               /*
+                * It's possible to have interrupts off here.
+                */
+               local_irq_enable();
+
+               /* 
+                * Valid to do another page fault here because this one came 
+                * from user space.
+                */
+               if (is_prefetch(regs, address, error_code))
+                       return;
+
+               if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) &&
+                   printk_ratelimit()) {
+                       printk("%s%s[%d]: segfault at %08lx eip %08lx "
+                           "esp %08lx error %lx\n",
+                           tsk->pid > 1 ? KERN_INFO : KERN_EMERG,
+                           tsk->comm, tsk->pid, address, regs->eip,
+                           regs->esp, error_code);
+               }
+               tsk->thread.cr2 = address;
+               /* Kernel addresses are always protection faults */
+               tsk->thread.error_code = error_code | (address >= TASK_SIZE);
+               tsk->thread.trap_no = 14;
+               force_sig_info_fault(SIGSEGV, si_code, address, tsk);
+               return;
+       }
+
+#ifdef CONFIG_X86_F00F_BUG
+       /*
+        * Pentium F0 0F C7 C8 bug workaround.
+        */
+       if (boot_cpu_data.f00f_bug) {
+               unsigned long nr;
+               
+               nr = (address - idt_descr.address) >> 3;
+
+               if (nr == 6) {
+                       do_invalid_op(regs, 0);
+                       return;
+               }
+       }
+#endif
+
+no_context:
+       /* Are we prepared to handle this kernel fault?  */
+       if (fixup_exception(regs))
+               return;
+
+       /* 
+        * Valid to do another page fault here, because if this fault
+        * had been triggered by is_prefetch fixup_exception would have 
+        * handled it.
+        */
+       if (is_prefetch(regs, address, error_code))
+               return;
+
+/*
+ * Oops. The kernel tried to access some bad page. We'll have to
+ * terminate things with extreme prejudice.
+ */
+
+       bust_spinlocks(1);
+
+       if (oops_may_print()) {
+               __typeof__(pte_val(__pte(0))) page;
+
+#ifdef CONFIG_X86_PAE
+               if (error_code & 16) {
+                       pte_t *pte = lookup_address(address);
+
+                       if (pte && pte_present(*pte) && !pte_exec_kernel(*pte))
+                               printk(KERN_CRIT "kernel tried to execute "
+                                       "NX-protected page - exploit attempt? "
+                                       "(uid: %d)\n", current->uid);
+               }
+#endif
+               if (address < PAGE_SIZE)
+                       printk(KERN_ALERT "BUG: unable to handle kernel NULL "
+                                       "pointer dereference");
+               else
+                       printk(KERN_ALERT "BUG: unable to handle kernel paging"
+                                       " request");
+               printk(" at virtual address %08lx\n",address);
+               printk(KERN_ALERT " printing eip:\n");
+               printk("%08lx\n", regs->eip);
+
+               page = read_cr3();
+               page = ((__typeof__(page) *) __va(page))[address >> PGDIR_SHIFT];
+#ifdef CONFIG_X86_PAE
+               printk(KERN_ALERT "*pdpt = %016Lx\n", page);
+               if ((page >> PAGE_SHIFT) < max_low_pfn
+                   && page & _PAGE_PRESENT) {
+                       page &= PAGE_MASK;
+                       page = ((__typeof__(page) *) __va(page))[(address >> PMD_SHIFT)
+                                                                & (PTRS_PER_PMD - 1)];
+                       printk(KERN_ALERT "*pde = %016Lx\n", page);
+                       page &= ~_PAGE_NX;
+               }
+#else
+               printk(KERN_ALERT "*pde = %08lx\n", page);
+#endif
+
+               /*
+                * We must not directly access the pte in the highpte
+                * case if the page table is located in highmem.
+                * And let's rather not kmap-atomic the pte, just in case
+                * it's allocated already.
+                */
+               if ((page >> PAGE_SHIFT) < max_low_pfn
+                   && (page & _PAGE_PRESENT)) {
+                       page &= PAGE_MASK;
+                       page = ((__typeof__(page) *) __va(page))[(address >> PAGE_SHIFT)
+                                                                & (PTRS_PER_PTE - 1)];
+                       printk(KERN_ALERT "*pte = %0*Lx\n", sizeof(page)*2, (u64)page);
+               }
+       }
+
+       tsk->thread.cr2 = address;
+       tsk->thread.trap_no = 14;
+       tsk->thread.error_code = error_code;
+       die("Oops", regs, error_code);
+       bust_spinlocks(0);
+       do_exit(SIGKILL);
+
+/*
+ * We ran out of memory, or some other thing happened to us that made
+ * us unable to handle the page fault gracefully.
+ */
+out_of_memory:
+       up_read(&mm->mmap_sem);
+       if (is_init(tsk)) {
+               yield();
+               down_read(&mm->mmap_sem);
+               goto survive;
+       }
+       printk("VM: killing process %s\n", tsk->comm);
+       if (error_code & 4)
+               do_exit(SIGKILL);
+       goto no_context;
+
+do_sigbus:
+       up_read(&mm->mmap_sem);
+
+       /* Kernel mode? Handle exceptions or die */
+       if (!(error_code & 4))
+               goto no_context;
+
+       /* User space => ok to do another page fault */
+       if (is_prefetch(regs, address, error_code))
+               return;
+
+       tsk->thread.cr2 = address;
+       tsk->thread.error_code = error_code;
+       tsk->thread.trap_no = 14;
+       force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
+}
+
+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).
+        */
+       static DECLARE_BITMAP(insync, PTRS_PER_PGD);
+       static unsigned long start = TASK_SIZE;
+       unsigned long address;
+
+       if (SHARED_KERNEL_PMD)
+               return;
+
+       BUILD_BUG_ON(TASK_SIZE & ~PGDIR_MASK);
+       for (address = start; address >= TASK_SIZE; address += PGDIR_SIZE) {
+               if (!test_bit(pgd_index(address), insync)) {
+                       unsigned long flags;
+                       struct page *page;
+
+                       spin_lock_irqsave(&pgd_lock, flags);
+                       for (page = pgd_list; page; page =
+                                       (struct page *)page->index)
+                               if (!vmalloc_sync_one(page_address(page),
+                                                               address)) {
+                                       BUG_ON(page != pgd_list);
+                                       break;
+                               }
+                       spin_unlock_irqrestore(&pgd_lock, flags);
+                       if (!page)
+                               set_bit(pgd_index(address), insync);
+               }
+               if (address == start && test_bit(pgd_index(address), insync))
+                       start = address + PGDIR_SIZE;
+       }
+}