blob: 2d9624fd10ecb4ad61cc33f55901f2109d6a4137 [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>
19#include <linux/smp_lock.h>
20#include <linux/vt_kern.h> /* For unblank_screen() */
21#include <linux/module.h>
22
23#include <asm/branch.h>
24#include <asm/mmu_context.h>
25#include <asm/system.h>
26#include <asm/uaccess.h>
27#include <asm/ptrace.h>
Thiemo Seufer16033d62005-02-19 13:56:04 +000028#include <asm/highmem.h> /* For VMALLOC_END */
Linus Torvalds1da177e2005-04-16 15:20:36 -070029
30/*
31 * This routine handles page faults. It determines the address,
32 * and the problem, and then passes it off to one of the appropriate
33 * routines.
34 */
35asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long write,
36 unsigned long address)
37{
38 struct vm_area_struct * vma = NULL;
39 struct task_struct *tsk = current;
40 struct mm_struct *mm = tsk->mm;
41 const int field = sizeof(unsigned long) * 2;
42 siginfo_t info;
43
44#if 0
45 printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", smp_processor_id(),
46 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 */
Thiemo Seufer16033d62005-02-19 13:56:04 +000061 if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END))
Linus Torvalds1da177e2005-04-16 15:20:36 -070062 goto vmalloc_fault;
63
64 /*
65 * If we're in an interrupt or have no user
66 * context, we must not take the fault..
67 */
68 if (in_atomic() || !mm)
69 goto bad_area_nosemaphore;
70
71 down_read(&mm->mmap_sem);
72 vma = find_vma(mm, address);
73 if (!vma)
74 goto bad_area;
75 if (vma->vm_start <= address)
76 goto good_area;
77 if (!(vma->vm_flags & VM_GROWSDOWN))
78 goto bad_area;
79 if (expand_stack(vma, address))
80 goto bad_area;
81/*
82 * Ok, we have a good vm_area for this memory access, so
83 * we can handle it..
84 */
85good_area:
86 info.si_code = SEGV_ACCERR;
87
88 if (write) {
89 if (!(vma->vm_flags & VM_WRITE))
90 goto bad_area;
91 } else {
92 if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
93 goto bad_area;
94 }
95
96survive:
97 /*
98 * If for any reason at all we couldn't handle the fault,
99 * make sure we exit gracefully rather than endlessly redo
100 * the fault.
101 */
102 switch (handle_mm_fault(mm, vma, address, write)) {
103 case VM_FAULT_MINOR:
104 tsk->min_flt++;
105 break;
106 case VM_FAULT_MAJOR:
107 tsk->maj_flt++;
108 break;
109 case VM_FAULT_SIGBUS:
110 goto do_sigbus;
111 case VM_FAULT_OOM:
112 goto out_of_memory;
113 default:
114 BUG();
115 }
116
117 up_read(&mm->mmap_sem);
118 return;
119
120/*
121 * Something tried to access memory that isn't in our memory map..
122 * Fix it, but check if it's kernel or user first..
123 */
124bad_area:
125 up_read(&mm->mmap_sem);
126
127bad_area_nosemaphore:
128 /* User mode accesses just cause a SIGSEGV */
129 if (user_mode(regs)) {
130 tsk->thread.cp0_badvaddr = address;
131 tsk->thread.error_code = write;
132#if 0
133 printk("do_page_fault() #2: sending SIGSEGV to %s for "
134 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
135 tsk->comm,
136 write ? "write access to" : "read access from",
137 field, address,
138 field, (unsigned long) regs->cp0_epc,
139 field, (unsigned long) regs->regs[31]);
140#endif
141 info.si_signo = SIGSEGV;
142 info.si_errno = 0;
143 /* info.si_code has been set above */
Ralf Baechlefe00f942005-03-01 19:22:29 +0000144 info.si_addr = (void __user *) address;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 force_sig_info(SIGSEGV, &info, tsk);
146 return;
147 }
148
149no_context:
150 /* Are we prepared to handle this kernel fault? */
151 if (fixup_exception(regs)) {
152 current->thread.cp0_baduaddr = address;
153 return;
154 }
155
156 /*
157 * Oops. The kernel tried to access some bad page. We'll have to
158 * terminate things with extreme prejudice.
159 */
160
161 bust_spinlocks(1);
162
163 printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
164 "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
165 smp_processor_id(), field, address, field, regs->cp0_epc,
166 field, regs->regs[31]);
167 die("Oops", regs);
168
169/*
170 * We ran out of memory, or some other thing happened to us that made
171 * us unable to handle the page fault gracefully.
172 */
173out_of_memory:
174 up_read(&mm->mmap_sem);
175 if (tsk->pid == 1) {
176 yield();
177 down_read(&mm->mmap_sem);
178 goto survive;
179 }
180 printk("VM: killing process %s\n", tsk->comm);
181 if (user_mode(regs))
182 do_exit(SIGKILL);
183 goto no_context;
184
185do_sigbus:
186 up_read(&mm->mmap_sem);
187
188 /* Kernel mode? Handle exceptions or die */
189 if (!user_mode(regs))
190 goto no_context;
191
192 /*
193 * Send a sigbus, regardless of whether we were in kernel
194 * or user mode.
195 */
196 tsk->thread.cp0_badvaddr = address;
197 info.si_signo = SIGBUS;
198 info.si_errno = 0;
199 info.si_code = BUS_ADRERR;
Ralf Baechlefe00f942005-03-01 19:22:29 +0000200 info.si_addr = (void __user *) address;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 force_sig_info(SIGBUS, &info, tsk);
202
203 return;
204
205vmalloc_fault:
206 {
207 /*
208 * Synchronize this task's top level page-table
209 * with the 'reference' page table.
210 *
211 * Do _not_ use "tsk" here. We might be inside
212 * an interrupt in the middle of a task switch..
213 */
214 int offset = __pgd_offset(address);
215 pgd_t *pgd, *pgd_k;
Ralf Baechlec6e8b582005-02-10 12:19:59 +0000216 pud_t *pud, *pud_k;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 pmd_t *pmd, *pmd_k;
218 pte_t *pte_k;
219
220 pgd = (pgd_t *) pgd_current[smp_processor_id()] + offset;
221 pgd_k = init_mm.pgd + offset;
222
223 if (!pgd_present(*pgd_k))
224 goto no_context;
225 set_pgd(pgd, *pgd_k);
226
Ralf Baechlec6e8b582005-02-10 12:19:59 +0000227 pud = pud_offset(pgd, address);
228 pud_k = pud_offset(pgd_k, address);
229 if (!pud_present(*pud_k))
230 goto no_context;
231
232 pmd = pmd_offset(pud, address);
233 pmd_k = pmd_offset(pud_k, address);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 if (!pmd_present(*pmd_k))
235 goto no_context;
236 set_pmd(pmd, *pmd_k);
237
238 pte_k = pte_offset_kernel(pmd_k, address);
239 if (!pte_present(*pte_k))
240 goto no_context;
241 return;
242 }
243}