Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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 | */ | |
c3fc5cd5 | 8 | #include <linux/context_tracking.h> |
1da177e4 LT |
9 | #include <linux/signal.h> |
10 | #include <linux/sched.h> | |
11 | #include <linux/interrupt.h> | |
12 | #include <linux/kernel.h> | |
13 | #include <linux/errno.h> | |
14 | #include <linux/string.h> | |
15 | #include <linux/types.h> | |
16 | #include <linux/ptrace.h> | |
17 | #include <linux/mman.h> | |
18 | #include <linux/mm.h> | |
19 | #include <linux/smp.h> | |
1da177e4 | 20 | #include <linux/module.h> |
c1bf207d | 21 | #include <linux/kprobes.h> |
7f788d2d | 22 | #include <linux/perf_event.h> |
1da177e4 LT |
23 | |
24 | #include <asm/branch.h> | |
25 | #include <asm/mmu_context.h> | |
1da177e4 LT |
26 | #include <asm/uaccess.h> |
27 | #include <asm/ptrace.h> | |
16033d61 | 28 | #include <asm/highmem.h> /* For VMALLOC_END */ |
c1bf207d | 29 | #include <linux/kdebug.h> |
1da177e4 LT |
30 | |
31 | /* | |
32 | * This routine handles page faults. It determines the address, | |
33 | * and the problem, and then passes it off to one of the appropriate | |
34 | * routines. | |
35 | */ | |
c3fc5cd5 RB |
36 | static void __kprobes __do_page_fault(struct pt_regs *regs, unsigned long write, |
37 | unsigned long address) | |
1da177e4 LT |
38 | { |
39 | struct vm_area_struct * vma = NULL; | |
40 | struct task_struct *tsk = current; | |
41 | struct mm_struct *mm = tsk->mm; | |
42 | const int field = sizeof(unsigned long) * 2; | |
43 | siginfo_t info; | |
83c54070 | 44 | int fault; |
759496ba | 45 | unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; |
1da177e4 LT |
46 | |
47 | #if 0 | |
d6f70360 | 48 | printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", raw_smp_processor_id(), |
1da177e4 LT |
49 | current->comm, current->pid, field, address, write, |
50 | field, regs->cp0_epc); | |
51 | #endif | |
52 | ||
c1bf207d DD |
53 | #ifdef CONFIG_KPROBES |
54 | /* | |
70342287 | 55 | * This is to notify the fault handler of the kprobes. The |
c1bf207d DD |
56 | * exception code is redundant as it is also carried in REGS, |
57 | * but we pass it anyhow. | |
58 | */ | |
59 | if (notify_die(DIE_PAGE_FAULT, "page fault", regs, -1, | |
60 | (regs->cp0_cause >> 2) & 0x1f, SIGSEGV) == NOTIFY_STOP) | |
61 | return; | |
62 | #endif | |
63 | ||
1da177e4 LT |
64 | info.si_code = SEGV_MAPERR; |
65 | ||
66 | /* | |
67 | * We fault-in kernel-space virtual memory on-demand. The | |
68 | * 'reference' page table is init_mm.pgd. | |
69 | * | |
70 | * NOTE! We MUST NOT take any locks for this case. We may | |
71 | * be in an interrupt or a critical region, and should | |
72 | * only copy the information from the master page table, | |
73 | * nothing more. | |
74 | */ | |
2ca2ebfd DD |
75 | #ifdef CONFIG_64BIT |
76 | # define VMALLOC_FAULT_TARGET no_context | |
77 | #else | |
78 | # define VMALLOC_FAULT_TARGET vmalloc_fault | |
79 | #endif | |
80 | ||
16033d61 | 81 | if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END)) |
2ca2ebfd | 82 | goto VMALLOC_FAULT_TARGET; |
656be92f AN |
83 | #ifdef MODULE_START |
84 | if (unlikely(address >= MODULE_START && address < MODULE_END)) | |
2ca2ebfd | 85 | goto VMALLOC_FAULT_TARGET; |
656be92f | 86 | #endif |
1da177e4 LT |
87 | |
88 | /* | |
89 | * If we're in an interrupt or have no user | |
90 | * context, we must not take the fault.. | |
91 | */ | |
92 | if (in_atomic() || !mm) | |
93 | goto bad_area_nosemaphore; | |
94 | ||
759496ba JW |
95 | if (user_mode(regs)) |
96 | flags |= FAULT_FLAG_USER; | |
43ca4957 | 97 | retry: |
1da177e4 LT |
98 | down_read(&mm->mmap_sem); |
99 | vma = find_vma(mm, address); | |
100 | if (!vma) | |
101 | goto bad_area; | |
102 | if (vma->vm_start <= address) | |
103 | goto good_area; | |
104 | if (!(vma->vm_flags & VM_GROWSDOWN)) | |
105 | goto bad_area; | |
106 | if (expand_stack(vma, address)) | |
107 | goto bad_area; | |
108 | /* | |
109 | * Ok, we have a good vm_area for this memory access, so | |
110 | * we can handle it.. | |
111 | */ | |
112 | good_area: | |
113 | info.si_code = SEGV_ACCERR; | |
114 | ||
115 | if (write) { | |
116 | if (!(vma->vm_flags & VM_WRITE)) | |
117 | goto bad_area; | |
759496ba | 118 | flags |= FAULT_FLAG_WRITE; |
1da177e4 | 119 | } else { |
05857c64 | 120 | if (cpu_has_rixi) { |
6dd9344c DD |
121 | if (address == regs->cp0_epc && !(vma->vm_flags & VM_EXEC)) { |
122 | #if 0 | |
123 | pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] XI violation\n", | |
124 | raw_smp_processor_id(), | |
125 | current->comm, current->pid, | |
126 | field, address, write, | |
127 | field, regs->cp0_epc); | |
128 | #endif | |
129 | goto bad_area; | |
130 | } | |
131 | if (!(vma->vm_flags & VM_READ)) { | |
132 | #if 0 | |
133 | pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] RI violation\n", | |
134 | raw_smp_processor_id(), | |
135 | current->comm, current->pid, | |
136 | field, address, write, | |
137 | field, regs->cp0_epc); | |
138 | #endif | |
139 | goto bad_area; | |
140 | } | |
141 | } else { | |
142 | if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC))) | |
143 | goto bad_area; | |
144 | } | |
1da177e4 LT |
145 | } |
146 | ||
1da177e4 LT |
147 | /* |
148 | * If for any reason at all we couldn't handle the fault, | |
149 | * make sure we exit gracefully rather than endlessly redo | |
150 | * the fault. | |
151 | */ | |
43ca4957 KC |
152 | fault = handle_mm_fault(mm, vma, address, flags); |
153 | ||
154 | if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current)) | |
155 | return; | |
156 | ||
a8b0ca17 | 157 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address); |
83c54070 NP |
158 | if (unlikely(fault & VM_FAULT_ERROR)) { |
159 | if (fault & VM_FAULT_OOM) | |
160 | goto out_of_memory; | |
33692f27 LT |
161 | else if (fault & VM_FAULT_SIGSEGV) |
162 | goto bad_area; | |
83c54070 NP |
163 | else if (fault & VM_FAULT_SIGBUS) |
164 | goto do_sigbus; | |
1da177e4 LT |
165 | BUG(); |
166 | } | |
43ca4957 KC |
167 | if (flags & FAULT_FLAG_ALLOW_RETRY) { |
168 | if (fault & VM_FAULT_MAJOR) { | |
169 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, | |
170 | regs, address); | |
171 | tsk->maj_flt++; | |
172 | } else { | |
173 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, | |
174 | regs, address); | |
175 | tsk->min_flt++; | |
176 | } | |
177 | if (fault & VM_FAULT_RETRY) { | |
178 | flags &= ~FAULT_FLAG_ALLOW_RETRY; | |
45cac65b | 179 | flags |= FAULT_FLAG_TRIED; |
43ca4957 KC |
180 | |
181 | /* | |
182 | * No need to up_read(&mm->mmap_sem) as we would | |
183 | * have already released it in __lock_page_or_retry | |
184 | * in mm/filemap.c. | |
185 | */ | |
186 | ||
187 | goto retry; | |
188 | } | |
7f788d2d | 189 | } |
1da177e4 LT |
190 | |
191 | up_read(&mm->mmap_sem); | |
192 | return; | |
193 | ||
194 | /* | |
195 | * Something tried to access memory that isn't in our memory map.. | |
196 | * Fix it, but check if it's kernel or user first.. | |
197 | */ | |
198 | bad_area: | |
199 | up_read(&mm->mmap_sem); | |
200 | ||
201 | bad_area_nosemaphore: | |
202 | /* User mode accesses just cause a SIGSEGV */ | |
203 | if (user_mode(regs)) { | |
204 | tsk->thread.cp0_badvaddr = address; | |
205 | tsk->thread.error_code = write; | |
206 | #if 0 | |
207 | printk("do_page_fault() #2: sending SIGSEGV to %s for " | |
208 | "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n", | |
209 | tsk->comm, | |
210 | write ? "write access to" : "read access from", | |
211 | field, address, | |
212 | field, (unsigned long) regs->cp0_epc, | |
213 | field, (unsigned long) regs->regs[31]); | |
214 | #endif | |
215 | info.si_signo = SIGSEGV; | |
216 | info.si_errno = 0; | |
217 | /* info.si_code has been set above */ | |
fe00f943 | 218 | info.si_addr = (void __user *) address; |
1da177e4 LT |
219 | force_sig_info(SIGSEGV, &info, tsk); |
220 | return; | |
221 | } | |
222 | ||
223 | no_context: | |
70342287 | 224 | /* Are we prepared to handle this kernel fault? */ |
1da177e4 LT |
225 | if (fixup_exception(regs)) { |
226 | current->thread.cp0_baduaddr = address; | |
227 | return; | |
228 | } | |
229 | ||
230 | /* | |
231 | * Oops. The kernel tried to access some bad page. We'll have to | |
232 | * terminate things with extreme prejudice. | |
233 | */ | |
1da177e4 LT |
234 | bust_spinlocks(1); |
235 | ||
236 | printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at " | |
237 | "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n", | |
d6f70360 | 238 | raw_smp_processor_id(), field, address, field, regs->cp0_epc, |
1da177e4 LT |
239 | field, regs->regs[31]); |
240 | die("Oops", regs); | |
241 | ||
1da177e4 | 242 | out_of_memory: |
c7c1e384 RB |
243 | /* |
244 | * We ran out of memory, call the OOM killer, and return the userspace | |
245 | * (which will retry the fault, or kill us if we got oom-killed). | |
246 | */ | |
a887b4da | 247 | up_read(&mm->mmap_sem); |
87134102 JW |
248 | if (!user_mode(regs)) |
249 | goto no_context; | |
c7c1e384 RB |
250 | pagefault_out_of_memory(); |
251 | return; | |
1da177e4 LT |
252 | |
253 | do_sigbus: | |
254 | up_read(&mm->mmap_sem); | |
255 | ||
256 | /* Kernel mode? Handle exceptions or die */ | |
257 | if (!user_mode(regs)) | |
258 | goto no_context; | |
41c594ab | 259 | else |
1da177e4 LT |
260 | /* |
261 | * Send a sigbus, regardless of whether we were in kernel | |
262 | * or user mode. | |
263 | */ | |
41c594ab RB |
264 | #if 0 |
265 | printk("do_page_fault() #3: sending SIGBUS to %s for " | |
266 | "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n", | |
267 | tsk->comm, | |
268 | write ? "write access to" : "read access from", | |
269 | field, address, | |
270 | field, (unsigned long) regs->cp0_epc, | |
271 | field, (unsigned long) regs->regs[31]); | |
272 | #endif | |
1da177e4 LT |
273 | tsk->thread.cp0_badvaddr = address; |
274 | info.si_signo = SIGBUS; | |
275 | info.si_errno = 0; | |
276 | info.si_code = BUS_ADRERR; | |
fe00f943 | 277 | info.si_addr = (void __user *) address; |
1da177e4 LT |
278 | force_sig_info(SIGBUS, &info, tsk); |
279 | ||
280 | return; | |
2ca2ebfd | 281 | #ifndef CONFIG_64BIT |
1da177e4 LT |
282 | vmalloc_fault: |
283 | { | |
284 | /* | |
285 | * Synchronize this task's top level page-table | |
286 | * with the 'reference' page table. | |
287 | * | |
288 | * Do _not_ use "tsk" here. We might be inside | |
289 | * an interrupt in the middle of a task switch.. | |
290 | */ | |
291 | int offset = __pgd_offset(address); | |
292 | pgd_t *pgd, *pgd_k; | |
c6e8b587 | 293 | pud_t *pud, *pud_k; |
1da177e4 LT |
294 | pmd_t *pmd, *pmd_k; |
295 | pte_t *pte_k; | |
296 | ||
d6f70360 | 297 | pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset; |
1da177e4 LT |
298 | pgd_k = init_mm.pgd + offset; |
299 | ||
300 | if (!pgd_present(*pgd_k)) | |
301 | goto no_context; | |
302 | set_pgd(pgd, *pgd_k); | |
303 | ||
c6e8b587 RB |
304 | pud = pud_offset(pgd, address); |
305 | pud_k = pud_offset(pgd_k, address); | |
306 | if (!pud_present(*pud_k)) | |
307 | goto no_context; | |
308 | ||
309 | pmd = pmd_offset(pud, address); | |
310 | pmd_k = pmd_offset(pud_k, address); | |
1da177e4 LT |
311 | if (!pmd_present(*pmd_k)) |
312 | goto no_context; | |
313 | set_pmd(pmd, *pmd_k); | |
314 | ||
315 | pte_k = pte_offset_kernel(pmd_k, address); | |
316 | if (!pte_present(*pte_k)) | |
317 | goto no_context; | |
318 | return; | |
319 | } | |
2ca2ebfd | 320 | #endif |
1da177e4 | 321 | } |
c3fc5cd5 RB |
322 | |
323 | asmlinkage void __kprobes do_page_fault(struct pt_regs *regs, | |
324 | unsigned long write, unsigned long address) | |
325 | { | |
326 | enum ctx_state prev_state; | |
327 | ||
328 | prev_state = exception_enter(); | |
329 | __do_page_fault(regs, write, address); | |
330 | exception_exit(prev_state); | |
331 | } |