blob: e97a7a2fb2c0980346c75791fce09b4026ee334b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * This file is subject to the terms and conditions of the GNU General Public
3 * License. See the file "COPYING" in the main directory of this archive
4 * for more details.
5 *
6 * Copyright (C) 1995 - 2000 by Ralf Baechle
7 */
8#include <linux/signal.h>
9#include <linux/sched.h>
10#include <linux/interrupt.h>
11#include <linux/kernel.h>
12#include <linux/errno.h>
13#include <linux/string.h>
14#include <linux/types.h>
15#include <linux/ptrace.h>
16#include <linux/mman.h>
17#include <linux/mm.h>
18#include <linux/smp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/vt_kern.h> /* For unblank_screen() */
20#include <linux/module.h>
21
22#include <asm/branch.h>
23#include <asm/mmu_context.h>
24#include <asm/system.h>
25#include <asm/uaccess.h>
26#include <asm/ptrace.h>
Thiemo Seufer16033d62005-02-19 13:56:04 +000027#include <asm/highmem.h> /* For VMALLOC_END */
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
29/*
30 * This routine handles page faults. It determines the address,
31 * and the problem, and then passes it off to one of the appropriate
32 * routines.
33 */
34asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long write,
35 unsigned long address)
36{
37 struct vm_area_struct * vma = NULL;
38 struct task_struct *tsk = current;
39 struct mm_struct *mm = tsk->mm;
40 const int field = sizeof(unsigned long) * 2;
41 siginfo_t info;
Nick Piggin83c54072007-07-19 01:47:05 -070042 int fault;
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44#if 0
Ralf Baechled6f70362007-03-29 22:30:01 +010045 printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", raw_smp_processor_id(),
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 current->comm, current->pid, field, address, write,
47 field, regs->cp0_epc);
48#endif
49
50 info.si_code = SEGV_MAPERR;
51
52 /*
53 * We fault-in kernel-space virtual memory on-demand. The
54 * 'reference' page table is init_mm.pgd.
55 *
56 * NOTE! We MUST NOT take any locks for this case. We may
57 * be in an interrupt or a critical region, and should
58 * only copy the information from the master page table,
59 * nothing more.
60 */
David Daney2ca2ebf2009-09-02 15:47:34 -070061#ifdef CONFIG_64BIT
62# define VMALLOC_FAULT_TARGET no_context
63#else
64# define VMALLOC_FAULT_TARGET vmalloc_fault
65#endif
66
Thiemo Seufer16033d62005-02-19 13:56:04 +000067 if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END))
David Daney2ca2ebf2009-09-02 15:47:34 -070068 goto VMALLOC_FAULT_TARGET;
Atsushi Nemoto656be922006-10-26 00:08:31 +090069#ifdef MODULE_START
70 if (unlikely(address >= MODULE_START && address < MODULE_END))
David Daney2ca2ebf2009-09-02 15:47:34 -070071 goto VMALLOC_FAULT_TARGET;
Atsushi Nemoto656be922006-10-26 00:08:31 +090072#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070073
74 /*
75 * If we're in an interrupt or have no user
76 * context, we must not take the fault..
77 */
78 if (in_atomic() || !mm)
79 goto bad_area_nosemaphore;
80
81 down_read(&mm->mmap_sem);
82 vma = find_vma(mm, address);
83 if (!vma)
84 goto bad_area;
85 if (vma->vm_start <= address)
86 goto good_area;
87 if (!(vma->vm_flags & VM_GROWSDOWN))
88 goto bad_area;
89 if (expand_stack(vma, address))
90 goto bad_area;
91/*
92 * Ok, we have a good vm_area for this memory access, so
93 * we can handle it..
94 */
95good_area:
96 info.si_code = SEGV_ACCERR;
97
98 if (write) {
99 if (!(vma->vm_flags & VM_WRITE))
100 goto bad_area;
101 } else {
Ralf Baechle00932ba2006-09-16 01:29:37 +0100102 if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 goto bad_area;
104 }
105
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 /*
107 * If for any reason at all we couldn't handle the fault,
108 * make sure we exit gracefully rather than endlessly redo
109 * the fault.
110 */
Linus Torvaldsd06063c2009-04-10 09:01:23 -0700111 fault = handle_mm_fault(mm, vma, address, write ? FAULT_FLAG_WRITE : 0);
Nick Piggin83c54072007-07-19 01:47:05 -0700112 if (unlikely(fault & VM_FAULT_ERROR)) {
113 if (fault & VM_FAULT_OOM)
114 goto out_of_memory;
115 else if (fault & VM_FAULT_SIGBUS)
116 goto do_sigbus;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 BUG();
118 }
Nick Piggin83c54072007-07-19 01:47:05 -0700119 if (fault & VM_FAULT_MAJOR)
120 tsk->maj_flt++;
121 else
122 tsk->min_flt++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
124 up_read(&mm->mmap_sem);
125 return;
126
127/*
128 * Something tried to access memory that isn't in our memory map..
129 * Fix it, but check if it's kernel or user first..
130 */
131bad_area:
132 up_read(&mm->mmap_sem);
133
134bad_area_nosemaphore:
135 /* User mode accesses just cause a SIGSEGV */
136 if (user_mode(regs)) {
137 tsk->thread.cp0_badvaddr = address;
138 tsk->thread.error_code = write;
139#if 0
140 printk("do_page_fault() #2: sending SIGSEGV to %s for "
141 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
142 tsk->comm,
143 write ? "write access to" : "read access from",
144 field, address,
145 field, (unsigned long) regs->cp0_epc,
146 field, (unsigned long) regs->regs[31]);
147#endif
148 info.si_signo = SIGSEGV;
149 info.si_errno = 0;
150 /* info.si_code has been set above */
Ralf Baechlefe00f942005-03-01 19:22:29 +0000151 info.si_addr = (void __user *) address;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 force_sig_info(SIGSEGV, &info, tsk);
153 return;
154 }
155
156no_context:
157 /* Are we prepared to handle this kernel fault? */
158 if (fixup_exception(regs)) {
159 current->thread.cp0_baduaddr = address;
160 return;
161 }
162
163 /*
164 * Oops. The kernel tried to access some bad page. We'll have to
165 * terminate things with extreme prejudice.
166 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167 bust_spinlocks(1);
168
169 printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
170 "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
Ralf Baechled6f70362007-03-29 22:30:01 +0100171 raw_smp_processor_id(), field, address, field, regs->cp0_epc,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 field, regs->regs[31]);
173 die("Oops", regs);
174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175out_of_memory:
Ralf Baechlec7c1e382009-01-12 00:09:13 +0000176 /*
177 * We ran out of memory, call the OOM killer, and return the userspace
178 * (which will retry the fault, or kill us if we got oom-killed).
179 */
Akinobu Mitaa887b4d2009-07-04 01:33:09 +0900180 up_read(&mm->mmap_sem);
Ralf Baechlec7c1e382009-01-12 00:09:13 +0000181 pagefault_out_of_memory();
182 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183
184do_sigbus:
185 up_read(&mm->mmap_sem);
186
187 /* Kernel mode? Handle exceptions or die */
188 if (!user_mode(regs))
189 goto no_context;
Ralf Baechle41c594a2006-04-05 09:45:45 +0100190 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 /*
192 * Send a sigbus, regardless of whether we were in kernel
193 * or user mode.
194 */
Ralf Baechle41c594a2006-04-05 09:45:45 +0100195#if 0
196 printk("do_page_fault() #3: sending SIGBUS to %s for "
197 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
198 tsk->comm,
199 write ? "write access to" : "read access from",
200 field, address,
201 field, (unsigned long) regs->cp0_epc,
202 field, (unsigned long) regs->regs[31]);
203#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 tsk->thread.cp0_badvaddr = address;
205 info.si_signo = SIGBUS;
206 info.si_errno = 0;
207 info.si_code = BUS_ADRERR;
Ralf Baechlefe00f942005-03-01 19:22:29 +0000208 info.si_addr = (void __user *) address;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 force_sig_info(SIGBUS, &info, tsk);
210
211 return;
David Daney2ca2ebf2009-09-02 15:47:34 -0700212#ifndef CONFIG_64BIT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213vmalloc_fault:
214 {
215 /*
216 * Synchronize this task's top level page-table
217 * with the 'reference' page table.
218 *
219 * Do _not_ use "tsk" here. We might be inside
220 * an interrupt in the middle of a task switch..
221 */
222 int offset = __pgd_offset(address);
223 pgd_t *pgd, *pgd_k;
Ralf Baechlec6e8b582005-02-10 12:19:59 +0000224 pud_t *pud, *pud_k;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 pmd_t *pmd, *pmd_k;
226 pte_t *pte_k;
227
Ralf Baechled6f70362007-03-29 22:30:01 +0100228 pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 pgd_k = init_mm.pgd + offset;
230
231 if (!pgd_present(*pgd_k))
232 goto no_context;
233 set_pgd(pgd, *pgd_k);
234
Ralf Baechlec6e8b582005-02-10 12:19:59 +0000235 pud = pud_offset(pgd, address);
236 pud_k = pud_offset(pgd_k, address);
237 if (!pud_present(*pud_k))
238 goto no_context;
239
240 pmd = pmd_offset(pud, address);
241 pmd_k = pmd_offset(pud_k, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242 if (!pmd_present(*pmd_k))
243 goto no_context;
244 set_pmd(pmd, *pmd_k);
245
246 pte_k = pte_offset_kernel(pmd_k, address);
247 if (!pte_present(*pte_k))
248 goto no_context;
249 return;
250 }
David Daney2ca2ebf2009-09-02 15:47:34 -0700251#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252}