Commit | Line | Data |
---|---|---|
ea2ba7dc | 1 | /* |
4c9e1385 | 2 | * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com) |
1da177e4 LT |
3 | * Licensed under the GPL |
4 | */ | |
5 | ||
4c9e1385 JD |
6 | #include <linux/mm.h> |
7 | #include <linux/sched.h> | |
8 | #include <linux/hardirq.h> | |
73395a00 | 9 | #include <linux/module.h> |
fbc9f16a | 10 | #include <linux/uaccess.h> |
4c9e1385 JD |
11 | #include <asm/current.h> |
12 | #include <asm/pgtable.h> | |
13 | #include <asm/tlbflush.h> | |
37185b33 AV |
14 | #include <arch.h> |
15 | #include <as-layout.h> | |
16 | #include <kern_util.h> | |
17 | #include <os.h> | |
18 | #include <skas.h> | |
1da177e4 | 19 | |
4c9e1385 JD |
20 | /* |
21 | * Note this is constrained to return 0, -EFAULT, -EACCESS, -ENOMEM by | |
22 | * segv(). | |
23 | */ | |
1d3468a6 | 24 | int handle_page_fault(unsigned long address, unsigned long ip, |
1da177e4 LT |
25 | int is_write, int is_user, int *code_out) |
26 | { | |
27 | struct mm_struct *mm = current->mm; | |
28 | struct vm_area_struct *vma; | |
29 | pgd_t *pgd; | |
30 | pud_t *pud; | |
31 | pmd_t *pmd; | |
32 | pte_t *pte; | |
1da177e4 | 33 | int err = -EFAULT; |
759496ba | 34 | unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; |
1da177e4 LT |
35 | |
36 | *code_out = SEGV_MAPERR; | |
fea03cb4 | 37 | |
4c9e1385 | 38 | /* |
70ffdb93 | 39 | * If the fault was with pagefaults disabled, don't take the fault, just |
4c9e1385 JD |
40 | * fail. |
41 | */ | |
70ffdb93 | 42 | if (faulthandler_disabled()) |
fea03cb4 PBG |
43 | goto out_nosemaphore; |
44 | ||
759496ba JW |
45 | if (is_user) |
46 | flags |= FAULT_FLAG_USER; | |
1cefe28f | 47 | retry: |
1da177e4 LT |
48 | down_read(&mm->mmap_sem); |
49 | vma = find_vma(mm, address); | |
4c9e1385 | 50 | if (!vma) |
1da177e4 | 51 | goto out; |
4c9e1385 | 52 | else if (vma->vm_start <= address) |
1da177e4 | 53 | goto good_area; |
4c9e1385 | 54 | else if (!(vma->vm_flags & VM_GROWSDOWN)) |
1da177e4 | 55 | goto out; |
4c9e1385 | 56 | else if (is_user && !ARCH_IS_STACKGROW(address)) |
1da177e4 | 57 | goto out; |
4c9e1385 | 58 | else if (expand_stack(vma, address)) |
1da177e4 LT |
59 | goto out; |
60 | ||
3b52166c | 61 | good_area: |
1da177e4 | 62 | *code_out = SEGV_ACCERR; |
759496ba JW |
63 | if (is_write) { |
64 | if (!(vma->vm_flags & VM_WRITE)) | |
65 | goto out; | |
66 | flags |= FAULT_FLAG_WRITE; | |
67 | } else { | |
68 | /* Don't require VM_READ|VM_EXEC for write faults! */ | |
69 | if (!(vma->vm_flags & (VM_READ | VM_EXEC))) | |
70 | goto out; | |
71 | } | |
13479d52 | 72 | |
1da177e4 | 73 | do { |
83c54070 | 74 | int fault; |
1c0fe6e3 | 75 | |
1cefe28f KC |
76 | fault = handle_mm_fault(mm, vma, address, flags); |
77 | ||
78 | if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current)) | |
79 | goto out_nosemaphore; | |
80 | ||
83c54070 NP |
81 | if (unlikely(fault & VM_FAULT_ERROR)) { |
82 | if (fault & VM_FAULT_OOM) { | |
83c54070 | 83 | goto out_of_memory; |
33692f27 LT |
84 | } else if (fault & VM_FAULT_SIGSEGV) { |
85 | goto out; | |
83c54070 NP |
86 | } else if (fault & VM_FAULT_SIGBUS) { |
87 | err = -EACCES; | |
88 | goto out; | |
89 | } | |
1da177e4 LT |
90 | BUG(); |
91 | } | |
1cefe28f KC |
92 | if (flags & FAULT_FLAG_ALLOW_RETRY) { |
93 | if (fault & VM_FAULT_MAJOR) | |
94 | current->maj_flt++; | |
95 | else | |
96 | current->min_flt++; | |
97 | if (fault & VM_FAULT_RETRY) { | |
98 | flags &= ~FAULT_FLAG_ALLOW_RETRY; | |
45cac65b | 99 | flags |= FAULT_FLAG_TRIED; |
1cefe28f KC |
100 | |
101 | goto retry; | |
102 | } | |
103 | } | |
83c54070 | 104 | |
3b52166c PBG |
105 | pgd = pgd_offset(mm, address); |
106 | pud = pud_offset(pgd, address); | |
107 | pmd = pmd_offset(pud, address); | |
108 | pte = pte_offset_kernel(pmd, address); | |
4c9e1385 | 109 | } while (!pte_present(*pte)); |
1da177e4 | 110 | err = 0; |
4c9e1385 JD |
111 | /* |
112 | * The below warning was added in place of | |
cbc24afa PBG |
113 | * pte_mkyoung(); if (is_write) pte_mkdirty(); |
114 | * If it's triggered, we'd see normally a hang here (a clean pte is | |
115 | * marked read-only to emulate the dirty bit). | |
116 | * However, the generic code can mark a PTE writable but clean on a | |
117 | * concurrent read fault, triggering this harmlessly. So comment it out. | |
118 | */ | |
119 | #if 0 | |
16b03678 | 120 | WARN_ON(!pte_young(*pte) || (is_write && !pte_dirty(*pte))); |
cbc24afa | 121 | #endif |
3b52166c PBG |
122 | flush_tlb_page(vma, address); |
123 | out: | |
1da177e4 | 124 | up_read(&mm->mmap_sem); |
fea03cb4 | 125 | out_nosemaphore: |
4c9e1385 | 126 | return err; |
1da177e4 | 127 | |
1da177e4 | 128 | out_of_memory: |
1c0fe6e3 NP |
129 | /* |
130 | * We ran out of memory, call the OOM killer, and return the userspace | |
131 | * (which will retry the fault, or kill us if we got oom-killed). | |
132 | */ | |
133 | up_read(&mm->mmap_sem); | |
87134102 JW |
134 | if (!is_user) |
135 | goto out_nosemaphore; | |
1c0fe6e3 NP |
136 | pagefault_out_of_memory(); |
137 | return 0; | |
1da177e4 | 138 | } |
73395a00 | 139 | EXPORT_SYMBOL(handle_page_fault); |
1da177e4 | 140 | |
3ef6130a RW |
141 | static void show_segv_info(struct uml_pt_regs *regs) |
142 | { | |
143 | struct task_struct *tsk = current; | |
144 | struct faultinfo *fi = UPT_FAULTINFO(regs); | |
145 | ||
146 | if (!unhandled_signal(tsk, SIGSEGV)) | |
147 | return; | |
148 | ||
149 | if (!printk_ratelimit()) | |
150 | return; | |
151 | ||
152 | printk("%s%s[%d]: segfault at %lx ip %p sp %p error %x", | |
153 | task_pid_nr(tsk) > 1 ? KERN_INFO : KERN_EMERG, | |
154 | tsk->comm, task_pid_nr(tsk), FAULT_ADDRESS(*fi), | |
155 | (void *)UPT_IP(regs), (void *)UPT_SP(regs), | |
156 | fi->error_code); | |
157 | ||
158 | print_vma_addr(KERN_CONT " in ", UPT_IP(regs)); | |
159 | printk(KERN_CONT "\n"); | |
160 | } | |
161 | ||
27aa6ef3 JD |
162 | static void bad_segv(struct faultinfo fi, unsigned long ip) |
163 | { | |
164 | struct siginfo si; | |
165 | ||
166 | si.si_signo = SIGSEGV; | |
167 | si.si_code = SEGV_ACCERR; | |
168 | si.si_addr = (void __user *) FAULT_ADDRESS(fi); | |
169 | current->thread.arch.faultinfo = fi; | |
170 | force_sig_info(SIGSEGV, &si, current); | |
171 | } | |
172 | ||
3e6f2ac4 JD |
173 | void fatal_sigsegv(void) |
174 | { | |
175 | force_sigsegv(SIGSEGV, current); | |
176 | do_signal(); | |
177 | /* | |
178 | * This is to tell gcc that we're not returning - do_signal | |
179 | * can, in general, return, but in this case, it's not, since | |
180 | * we just got a fatal SIGSEGV queued. | |
181 | */ | |
182 | os_dump_core(); | |
183 | } | |
184 | ||
d3c1cfcd | 185 | void segv_handler(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs) |
c66fdd5e GS |
186 | { |
187 | struct faultinfo * fi = UPT_FAULTINFO(regs); | |
188 | ||
4c9e1385 | 189 | if (UPT_IS_USER(regs) && !SEGV_IS_FIXABLE(fi)) { |
3ef6130a | 190 | show_segv_info(regs); |
c66fdd5e GS |
191 | bad_segv(*fi, UPT_IP(regs)); |
192 | return; | |
193 | } | |
194 | segv(*fi, UPT_IP(regs), UPT_IS_USER(regs), regs); | |
195 | } | |
196 | ||
c578455a BS |
197 | /* |
198 | * We give a *copy* of the faultinfo in the regs to segv. | |
199 | * This must be done, since nesting SEGVs could overwrite | |
200 | * the info in the regs. A pointer to the info then would | |
201 | * give us bad data! | |
202 | */ | |
5d86456d | 203 | unsigned long segv(struct faultinfo fi, unsigned long ip, int is_user, |
77bf4400 | 204 | struct uml_pt_regs *regs) |
1da177e4 LT |
205 | { |
206 | struct siginfo si; | |
fab95c55 | 207 | jmp_buf *catcher; |
1da177e4 | 208 | int err; |
5d86456d JD |
209 | int is_write = FAULT_WRITE(fi); |
210 | unsigned long address = FAULT_ADDRESS(fi); | |
1da177e4 | 211 | |
bb6a1b2e | 212 | if (!is_user && regs) |
f72c22e4 RW |
213 | current->thread.segv_regs = container_of(regs, struct pt_regs, regs); |
214 | ||
4c9e1385 | 215 | if (!is_user && (address >= start_vm) && (address < end_vm)) { |
5d86456d | 216 | flush_tlb_kernel_vm(); |
f72c22e4 | 217 | goto out; |
5d86456d | 218 | } |
4c9e1385 | 219 | else if (current->mm == NULL) { |
377fad3a | 220 | show_regs(container_of(regs, struct pt_regs, regs)); |
4c9e1385 | 221 | panic("Segfault with no mm"); |
377fad3a | 222 | } |
d2313084 RW |
223 | else if (!is_user && address < TASK_SIZE) { |
224 | show_regs(container_of(regs, struct pt_regs, regs)); | |
225 | panic("Kernel tried to access user memory at addr 0x%lx, ip 0x%lx", | |
226 | address, ip); | |
227 | } | |
546fe1cb | 228 | |
d0b5e15f | 229 | if (SEGV_IS_FIXABLE(&fi)) |
4c9e1385 JD |
230 | err = handle_page_fault(address, ip, is_write, is_user, |
231 | &si.si_code); | |
546fe1cb PBG |
232 | else { |
233 | err = -EFAULT; | |
4c9e1385 JD |
234 | /* |
235 | * A thread accessed NULL, we get a fault, but CR2 is invalid. | |
236 | * This code is used in __do_copy_from_user() of TT mode. | |
237 | * XXX tt mode is gone, so maybe this isn't needed any more | |
238 | */ | |
546fe1cb PBG |
239 | address = 0; |
240 | } | |
1da177e4 LT |
241 | |
242 | catcher = current->thread.fault_catcher; | |
4c9e1385 | 243 | if (!err) |
f72c22e4 | 244 | goto out; |
4c9e1385 | 245 | else if (catcher != NULL) { |
1da177e4 | 246 | current->thread.fault_addr = (void *) address; |
fab95c55 | 247 | UML_LONGJMP(catcher, 1); |
1d3468a6 | 248 | } |
4c9e1385 | 249 | else if (current->thread.fault_addr != NULL) |
1da177e4 | 250 | panic("fault_addr set but no fault catcher"); |
4c9e1385 | 251 | else if (!is_user && arch_fixup(ip, regs)) |
f72c22e4 | 252 | goto out; |
1da177e4 | 253 | |
4c9e1385 | 254 | if (!is_user) { |
377fad3a | 255 | show_regs(container_of(regs, struct pt_regs, regs)); |
1d3468a6 | 256 | panic("Kernel mode fault at addr 0x%lx, ip 0x%lx", |
1da177e4 | 257 | address, ip); |
377fad3a | 258 | } |
1da177e4 | 259 | |
3ef6130a RW |
260 | show_segv_info(regs); |
261 | ||
3b52166c | 262 | if (err == -EACCES) { |
1da177e4 LT |
263 | si.si_signo = SIGBUS; |
264 | si.si_errno = 0; | |
265 | si.si_code = BUS_ADRERR; | |
4d338e1a | 266 | si.si_addr = (void __user *)address; |
5d86456d | 267 | current->thread.arch.faultinfo = fi; |
1da177e4 | 268 | force_sig_info(SIGBUS, &si, current); |
3b52166c PBG |
269 | } else { |
270 | BUG_ON(err != -EFAULT); | |
1da177e4 | 271 | si.si_signo = SIGSEGV; |
4d338e1a | 272 | si.si_addr = (void __user *) address; |
5d86456d | 273 | current->thread.arch.faultinfo = fi; |
1da177e4 LT |
274 | force_sig_info(SIGSEGV, &si, current); |
275 | } | |
f72c22e4 RW |
276 | |
277 | out: | |
278 | if (regs) | |
279 | current->thread.segv_regs = NULL; | |
280 | ||
5d86456d | 281 | return 0; |
1da177e4 LT |
282 | } |
283 | ||
d3c1cfcd | 284 | void relay_signal(int sig, struct siginfo *si, struct uml_pt_regs *regs) |
1da177e4 | 285 | { |
d3c1cfcd MP |
286 | struct faultinfo *fi; |
287 | struct siginfo clean_si; | |
288 | ||
4c9e1385 JD |
289 | if (!UPT_IS_USER(regs)) { |
290 | if (sig == SIGBUS) | |
291 | printk(KERN_ERR "Bus error - the host /dev/shm or /tmp " | |
292 | "mount likely just ran out of space\n"); | |
1da177e4 | 293 | panic("Kernel mode signal %d", sig); |
6edf428e JD |
294 | } |
295 | ||
9226b838 JD |
296 | arch_examine_signal(sig, regs); |
297 | ||
d3c1cfcd MP |
298 | memset(&clean_si, 0, sizeof(clean_si)); |
299 | clean_si.si_signo = si->si_signo; | |
300 | clean_si.si_errno = si->si_errno; | |
301 | clean_si.si_code = si->si_code; | |
302 | switch (sig) { | |
303 | case SIGILL: | |
304 | case SIGFPE: | |
305 | case SIGSEGV: | |
306 | case SIGBUS: | |
307 | case SIGTRAP: | |
308 | fi = UPT_FAULTINFO(regs); | |
309 | clean_si.si_addr = (void __user *) FAULT_ADDRESS(*fi); | |
310 | current->thread.arch.faultinfo = *fi; | |
311 | #ifdef __ARCH_SI_TRAPNO | |
312 | clean_si.si_trapno = si->si_trapno; | |
313 | #endif | |
314 | break; | |
315 | default: | |
316 | printk(KERN_ERR "Attempted to relay unknown signal %d (si_code = %d)\n", | |
317 | sig, si->si_code); | |
318 | } | |
319 | ||
320 | force_sig_info(sig, &clean_si, current); | |
1da177e4 LT |
321 | } |
322 | ||
d3c1cfcd | 323 | void bus_handler(int sig, struct siginfo *si, struct uml_pt_regs *regs) |
1da177e4 | 324 | { |
4c9e1385 | 325 | if (current->thread.fault_catcher != NULL) |
fab95c55 | 326 | UML_LONGJMP(current->thread.fault_catcher, 1); |
d3c1cfcd MP |
327 | else |
328 | relay_signal(sig, si, regs); | |
1da177e4 LT |
329 | } |
330 | ||
d3c1cfcd | 331 | void winch(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs) |
1da177e4 LT |
332 | { |
333 | do_IRQ(WINCH_IRQ, regs); | |
334 | } | |
335 | ||
336 | void trap_init(void) | |
337 | { | |
338 | } |