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

/*
 *  Copyright (C) 1995  Linus Torvalds
 *  Copyright (C) 2001,2002 Andi Kleen, SuSE Labs.
 */

#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/compiler.h>
#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/pgalloc.h>
#include <asm/smp.h>
#include <asm/tlbflush.h>
#include <asm/proto.h>
#include <asm-generic/sections.h>

/*
 * Page fault error code bits
 *      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
 */
#define PF_PROT (1<<0)
#define PF_WRITE        (1<<1)
#define PF_USER (1<<2)
#define PF_RSVD (1<<3)
#define PF_INSTR        (1<<4)

static inline int notify_page_fault(struct pt_regs *regs)
{
#ifdef CONFIG_KPROBES
        int ret = 0;

        /* kprobe_running() needs smp_processor_id() */
        if (!user_mode(regs)) {
                preempt_disable();
                if (kprobe_running() && kprobe_fault_handler(regs, 14))
                        ret = 1;
                preempt_enable();
        }

        return ret;
#else
        return 0;
#endif
}

#ifdef CONFIG_X86_32
/*
 * 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 ip = regs->ip;
        unsigned seg = regs->cs & 0xffff;
        u32 seg_ar, seg_limit, base, *desc;

        /* Unlikely, but must come before segment checks. */
        if (unlikely(regs->flags & VM_MASK)) {
                base = seg << 4;
                *eip_limit = base + 0xffff;
                return base + (ip & 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 ip;

        /* 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) || ip > seg_limit) {
                *eip_limit = 0;
                return 1;        /* So that returned ip > *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. */
                mutex_lock(&current->mm->context.lock);
                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((struct desc_struct *)desc);

        if (seg & (1<<2))
                mutex_unlock(&current->mm->context.lock);
        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 ip + base;
}
#endif

/*
 * X86_32
 * Sometimes AMD Athlon/Opteron CPUs report invalid exceptions on prefetch.
 * Check that here and ignore it.
 *
 * X86_64
 * Sometimes the CPU reports invalid exceptions on prefetch.
 * Check that here and ignore it.
 *
 * Opcode checker based on code by Richard Brunner
 */
static int is_prefetch(struct pt_regs *regs, unsigned long addr,
                       unsigned long error_code)
{
        unsigned char *instr;
        int scan_more = 1;
        int prefetch = 0;
        unsigned char *max_instr;

#ifdef CONFIG_X86_32
        unsigned long limit;
        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 & PF_INSTR))
                        return 0;
        } else {
                return 0;
        }
        instr = (unsigned char *)get_segment_eip(regs, &limit);
#else
        /* If it was a exec fault ignore */
        if (error_code & PF_INSTR)
                return 0;
        instr = (unsigned char __user *)convert_rip_to_linear(current, regs);
#endif

        max_instr = instr + 15;

#ifdef CONFIG_X86_64
        if (user_mode(regs) && instr >= (unsigned char *)TASK_SIZE)
                return 0;
#endif

        while (scan_more && instr < max_instr) {
                unsigned char opcode;
                unsigned char instr_hi;
                unsigned char instr_lo;

#ifdef CONFIG_X86_32
                if (instr > (unsigned char *)limit)
                        break;
#endif
                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.
                         * In X86_64 long mode, the CPU will signal invalid
                         * opcode if some of these prefixes are present so
                         * X86_64 will never get here anyway
                         */
                        scan_more = ((instr_lo & 7) == 0x6);
                        break;
#ifdef CONFIG_X86_64
                case 0x40:
                        /*
                         * In AMD64 long mode 0x40..0x4F are valid REX prefixes
                         * Need to figure out under what instruction mode the
                         * instruction was issued. Could check the LDT for lm,
                         * but for now it's good enough to assume that long
                         * mode only uses well known segments or kernel.
                         */
                        scan_more = (!user_mode(regs)) || (regs->cs == __USER_CS);
                        break;
#endif
                case 0x60:
                        /* 0x64 thru 0x67 are valid prefixes in all modes. */
                        scan_more = (instr_lo & 0xC) == 0x4;
                        break;
                case 0xF0:
                        /* 0xF0, 0xF2, 0xF3 are valid prefixes in all modes. */
                        scan_more = !instr_lo || (instr_lo>>1) == 1;
                        break;
                case 0x00:
                        /* Prefetch instruction is 0x0F0D or 0x0F18 */
                        scan_more = 0;
#ifdef CONFIG_X86_32
                        if (instr > (unsigned char *)limit)
                                break;
#endif
                        if (probe_kernel_address(instr, opcode))
                                break;
                        prefetch = (instr_lo == 0xF) &&
                                (opcode == 0x0D || opcode == 0x18);
                        break;
                default:
                        scan_more = 0;
                        break;
                }
        }
        return prefetch;
}

static 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);
}

static int bad_address(void *p)
{
        unsigned long dummy;
        return probe_kernel_address((unsigned long *)p, dummy);
}

void dump_pagetable(unsigned long address)
{
        pgd_t *pgd;
        pud_t *pud;
        pmd_t *pmd;
        pte_t *pte;

        pgd = (pgd_t *)read_cr3();

        pgd = __va((unsigned long)pgd & PHYSICAL_PAGE_MASK);
        pgd += pgd_index(address);
        if (bad_address(pgd)) goto bad;
        printk("PGD %lx ", pgd_val(*pgd));
        if (!pgd_present(*pgd)) goto ret;

        pud = pud_offset(pgd, address);
        if (bad_address(pud)) goto bad;
        printk("PUD %lx ", pud_val(*pud));
        if (!pud_present(*pud)) goto ret;

        pmd = pmd_offset(pud, address);
        if (bad_address(pmd)) goto bad;
        printk("PMD %lx ", pmd_val(*pmd));
        if (!pmd_present(*pmd) || pmd_large(*pmd)) goto ret;

        pte = pte_offset_kernel(pmd, address);
        if (bad_address(pte)) goto bad;
        printk("PTE %lx", pte_val(*pte));
ret:
        printk("\n");
        return;
bad:
        printk("BAD\n");
}

#ifdef CONFIG_X86_64
static const char errata93_warning[] =
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";

/* Workaround for K8 erratum #93 & buggy BIOS.
   BIOS SMM functions are required to use a specific workaround
   to avoid corruption of the 64bit RIP register on C stepping K8.
   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. */

static int is_errata93(struct pt_regs *regs, unsigned long address)
{
        static int warned;
        if (address != regs->ip)
                return 0;
        if ((address >> 32) != 0)
                return 0;
        address |= 0xffffffffUL << 32;
        if ((address >= (u64)_stext && address <= (u64)_etext) ||
            (address >= MODULES_VADDR && address <= MODULES_END)) {
                if (!warned) {
                        printk(errata93_warning);
                        warned = 1;
                }
                regs->ip = address;
                return 1;
        }
        return 0;
}
#endif

static noinline void pgtable_bad(unsigned long address, struct pt_regs *regs,
                                 unsigned long error_code)
{
        unsigned long flags = oops_begin();
        struct task_struct *tsk;

        printk(KERN_ALERT "%s: Corrupted page table at address %lx\n",
               current->comm, address);
        dump_pagetable(address);
        tsk = current;
        tsk->thread.cr2 = address;
        tsk->thread.trap_no = 14;
        tsk->thread.error_code = error_code;
        if (__die("Bad pagetable", regs, error_code))
                regs = NULL;
        oops_end(flags, regs, SIGKILL);
}

/*
 * Handle a fault on the vmalloc area
 *
 * This assumes no large pages in there.
 */
static int vmalloc_fault(unsigned long address)
{
        pgd_t *pgd, *pgd_ref;
        pud_t *pud, *pud_ref;
        pmd_t *pmd, *pmd_ref;
        pte_t *pte, *pte_ref;

        /* Copy kernel mappings over when needed. This can also
           happen within a race in page table update. In the later
           case just flush. */

        pgd = pgd_offset(current->mm ?: &init_mm, address);
        pgd_ref = pgd_offset_k(address);
        if (pgd_none(*pgd_ref))
                return -1;
        if (pgd_none(*pgd))
                set_pgd(pgd, *pgd_ref);
        else
                BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));

        /* Below here mismatches are bugs because these lower tables
           are shared */

        pud = pud_offset(pgd, address);
        pud_ref = pud_offset(pgd_ref, address);
        if (pud_none(*pud_ref))
                return -1;
        if (pud_none(*pud) || pud_page_vaddr(*pud) != pud_page_vaddr(*pud_ref))
                BUG();
        pmd = pmd_offset(pud, address);
        pmd_ref = pmd_offset(pud_ref, address);
        if (pmd_none(*pmd_ref))
                return -1;
        if (pmd_none(*pmd) || pmd_page(*pmd) != pmd_page(*pmd_ref))
                BUG();
        pte_ref = pte_offset_kernel(pmd_ref, address);
        if (!pte_present(*pte_ref))
                return -1;
        pte = pte_offset_kernel(pmd, address);
        /* Don't use pte_page here, because the mappings can point
           outside mem_map, and the NUMA hash lookup cannot handle
           that. */
        if (!pte_present(*pte) || pte_pfn(*pte) != pte_pfn(*pte_ref))
                BUG();
        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.
 */
asmlinkage 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, fault;
        unsigned long flags;
        int si_code;

        /*
         * We can fault from pretty much anywhere, with unknown IRQ state.
         */
        trace_hardirqs_fixup();

        tsk = current;
        mm = tsk->mm;
        prefetchw(&mm->mmap_sem);

        /* get the address */
        address = read_cr2();

        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_SIZE64)) {
                /*
                 * Don't check for the module range here: its PML4
                 * is always initialized because it's shared with the main
                 * kernel text. Only vmalloc may need PML4 syncups.
                 */
                if (!(error_code & (PF_RSVD|PF_USER|PF_PROT)) &&
                      ((address >= VMALLOC_START && address < VMALLOC_END))) {
                        if (vmalloc_fault(address) >= 0)
                                return;
                }
                if (notify_page_fault(regs))
                        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))
                return;

        if (likely(regs->flags & X86_EFLAGS_IF))
                local_irq_enable();

        if (unlikely(error_code & PF_RSVD))
                pgtable_bad(address, regs, error_code);

        /*
         * 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 (unlikely(in_atomic() || !mm))
                goto bad_area_nosemaphore;

        /*
         * User-mode registers count as a user access even for any
         * potential system fault or CPU buglet.
         */
        if (user_mode_vm(regs))
                error_code |= PF_USER;

 again:
        /* 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.  Unfortunately, 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 possibility 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 & PF_USER) == 0 &&
                    !search_exception_tables(regs->ip))
                        goto bad_area_nosemaphore;
                down_read(&mm->mmap_sem);
        }

        vma = find_vma(mm, address);
        if (!vma)
                goto bad_area;
        if (likely(vma->vm_start <= address))
                goto good_area;
        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.
                 */
                if (address + 65536 + 32 * sizeof(unsigned long) < regs->sp)
                        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 & (PF_PROT|PF_WRITE)) {
        default:        /* 3: write, present */
                /* fall through */
        case PF_WRITE:          /* write, not present */
                if (!(vma->vm_flags & VM_WRITE))
                        goto bad_area;
                write++;
                break;
        case PF_PROT:           /* read, present */
                goto bad_area;
        case 0:                 /* read, not present */
                if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
                        goto bad_area;
        }

        /*
         * 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++;
        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 & PF_USER) {

                /*
                 * It's possible to have interrupts off here.
                 */
                local_irq_enable();

                if (is_prefetch(regs, address, error_code))
                        return;

                /* Work around K8 erratum #100 K8 in compat mode
                   occasionally jumps to illegal addresses >4GB.  We
                   catch this here in the page fault handler because
                   these addresses are not reachable. Just detect this
                   case and return.  Any code segment in LDT is
                   compatibility mode. */
                if ((regs->cs == __USER32_CS || (regs->cs & (1<<2))) &&
                    (address >> 32))
                        return;

                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);
                }

                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;
        }

no_context:
        /* Are we prepared to handle this kernel fault?  */
        if (fixup_exception(regs))
                return;

        /*
         * Hall of shame of CPU/BIOS bugs.
         */

        if (is_prefetch(regs, address, error_code))
                return;

        if (is_errata93(regs, address))
                return;

/*
 * Oops. The kernel tried to access some bad page. We'll have to
 * terminate things with extreme prejudice.
 */

        flags = oops_begin();

        if (address < PAGE_SIZE)
                printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference");
        else
                printk(KERN_ALERT "Unable to handle kernel paging request");
        printk(" at %016lx RIP: \n" KERN_ALERT, address);
        printk_address(regs->ip);
        dump_pagetable(address);
        tsk->thread.cr2 = address;
        tsk->thread.trap_no = 14;
        tsk->thread.error_code = error_code;
        if (__die("Oops", regs, error_code))
                regs = NULL;
        /* Executive summary in case the body of the oops scrolled away */
        printk(KERN_EMERG "CR2: %016lx\n", address);
        oops_end(flags, regs, 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_global_init(current)) {
                yield();
                goto again;
        }
        printk("VM: killing process %s\n", tsk->comm);
        if (error_code & 4)
                do_group_exit(SIGKILL);
        goto no_context;

do_sigbus:
        up_read(&mm->mmap_sem);

        /* Kernel mode? Handle exceptions or die */
        if (!(error_code & PF_USER))
                goto no_context;

        tsk->thread.cr2 = address;
        tsk->thread.error_code = error_code;
        tsk->thread.trap_no = 14;
        force_sig_info_fault(SIGBUS, BUS_ADRERR, address, tsk);
        return;
}

DEFINE_SPINLOCK(pgd_lock);
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). */
        static DECLARE_BITMAP(insync, PTRS_PER_PGD);
        static unsigned long start = VMALLOC_START & PGDIR_MASK;
        unsigned long address;

        for (address = start; address <= VMALLOC_END; address += PGDIR_SIZE) {
                if (!test_bit(pgd_index(address), insync)) {
                        const pgd_t *pgd_ref = pgd_offset_k(address);
                        struct page *page;

                        if (pgd_none(*pgd_ref))
                                continue;
                        spin_lock(&pgd_lock);
                        list_for_each_entry(page, &pgd_list, lru) {
                                pgd_t *pgd;
                                pgd = (pgd_t *)page_address(page) + pgd_index(address);
                                if (pgd_none(*pgd))
                                        set_pgd(pgd, *pgd_ref);
                                else
                                        BUG_ON(pgd_page_vaddr(*pgd) != pgd_page_vaddr(*pgd_ref));
                        }
                        spin_unlock(&pgd_lock);
                        set_bit(pgd_index(address), insync);
                }
                if (address == start)
                        start = address + PGDIR_SIZE;
        }
        /* Check that there is no need to do the same for the modules area. */
        BUILD_BUG_ON(!(MODULES_VADDR > __START_KERNEL));
        BUILD_BUG_ON(!(((MODULES_END - 1) & PGDIR_MASK) ==
                                (__START_KERNEL & PGDIR_MASK)));
}