]>
Commit | Line | Data |
---|---|---|
1d18c47c CM |
1 | /* |
2 | * Based on arch/arm/mm/fault.c | |
3 | * | |
4 | * Copyright (C) 1995 Linus Torvalds | |
5 | * Copyright (C) 1995-2004 Russell King | |
6 | * Copyright (C) 2012 ARM Ltd. | |
7 | * | |
8 | * This program is free software; you can redistribute it and/or modify | |
9 | * it under the terms of the GNU General Public License version 2 as | |
10 | * published by the Free Software Foundation. | |
11 | * | |
12 | * This program is distributed in the hope that it will be useful, | |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
15 | * GNU General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU General Public License | |
18 | * along with this program. If not, see <http://www.gnu.org/licenses/>. | |
19 | */ | |
20 | ||
0edfa839 | 21 | #include <linux/extable.h> |
1d18c47c CM |
22 | #include <linux/signal.h> |
23 | #include <linux/mm.h> | |
24 | #include <linux/hardirq.h> | |
25 | #include <linux/init.h> | |
26 | #include <linux/kprobes.h> | |
27 | #include <linux/uaccess.h> | |
28 | #include <linux/page-flags.h> | |
3f07c014 | 29 | #include <linux/sched/signal.h> |
b17b0153 | 30 | #include <linux/sched/debug.h> |
1d18c47c CM |
31 | #include <linux/highmem.h> |
32 | #include <linux/perf_event.h> | |
7209c868 | 33 | #include <linux/preempt.h> |
e7c600f1 | 34 | #include <linux/hugetlb.h> |
1d18c47c | 35 | |
7209c868 | 36 | #include <asm/bug.h> |
4aae801e | 37 | #include <asm/cmpxchg.h> |
338d4f49 | 38 | #include <asm/cpufeature.h> |
1d18c47c CM |
39 | #include <asm/exception.h> |
40 | #include <asm/debug-monitors.h> | |
9141300a | 41 | #include <asm/esr.h> |
338d4f49 | 42 | #include <asm/sysreg.h> |
1d18c47c CM |
43 | #include <asm/system_misc.h> |
44 | #include <asm/pgtable.h> | |
45 | #include <asm/tlbflush.h> | |
46 | ||
7edda088 TB |
47 | #include <acpi/ghes.h> |
48 | ||
09a6adf5 VK |
49 | struct fault_info { |
50 | int (*fn)(unsigned long addr, unsigned int esr, | |
51 | struct pt_regs *regs); | |
52 | int sig; | |
53 | int code; | |
54 | const char *name; | |
55 | }; | |
56 | ||
57 | static const struct fault_info fault_info[]; | |
58 | ||
59 | static inline const struct fault_info *esr_to_fault_info(unsigned int esr) | |
60 | { | |
61 | return fault_info + (esr & 63); | |
62 | } | |
3495386b | 63 | |
2dd0e8d2 SP |
64 | #ifdef CONFIG_KPROBES |
65 | static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr) | |
66 | { | |
67 | int ret = 0; | |
68 | ||
69 | /* kprobe_running() needs smp_processor_id() */ | |
70 | if (!user_mode(regs)) { | |
71 | preempt_disable(); | |
72 | if (kprobe_running() && kprobe_fault_handler(regs, esr)) | |
73 | ret = 1; | |
74 | preempt_enable(); | |
75 | } | |
76 | ||
77 | return ret; | |
78 | } | |
79 | #else | |
80 | static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr) | |
81 | { | |
82 | return 0; | |
83 | } | |
84 | #endif | |
85 | ||
1d18c47c | 86 | /* |
67ce16ec | 87 | * Dump out the page tables associated with 'addr' in the currently active mm. |
1d18c47c | 88 | */ |
67ce16ec | 89 | void show_pte(unsigned long addr) |
1d18c47c | 90 | { |
67ce16ec | 91 | struct mm_struct *mm; |
1d18c47c CM |
92 | pgd_t *pgd; |
93 | ||
67ce16ec KM |
94 | if (addr < TASK_SIZE) { |
95 | /* TTBR0 */ | |
96 | mm = current->active_mm; | |
97 | if (mm == &init_mm) { | |
98 | pr_alert("[%016lx] user address but active_mm is swapper\n", | |
99 | addr); | |
100 | return; | |
101 | } | |
102 | } else if (addr >= VA_START) { | |
103 | /* TTBR1 */ | |
1d18c47c | 104 | mm = &init_mm; |
67ce16ec KM |
105 | } else { |
106 | pr_alert("[%016lx] address between user and kernel address ranges\n", | |
107 | addr); | |
108 | return; | |
109 | } | |
1d18c47c | 110 | |
1eb34b6e WD |
111 | pr_alert("%s pgtable: %luk pages, %u-bit VAs, pgd = %p\n", |
112 | mm == &init_mm ? "swapper" : "user", PAGE_SIZE / SZ_1K, | |
113 | VA_BITS, mm->pgd); | |
1d18c47c | 114 | pgd = pgd_offset(mm, addr); |
67ce16ec | 115 | pr_alert("[%016lx] *pgd=%016llx", addr, pgd_val(*pgd)); |
1d18c47c CM |
116 | |
117 | do { | |
118 | pud_t *pud; | |
119 | pmd_t *pmd; | |
120 | pte_t *pte; | |
121 | ||
4339e3f3 | 122 | if (pgd_none(*pgd) || pgd_bad(*pgd)) |
1d18c47c CM |
123 | break; |
124 | ||
125 | pud = pud_offset(pgd, addr); | |
6ef4fb38 | 126 | pr_cont(", *pud=%016llx", pud_val(*pud)); |
4339e3f3 | 127 | if (pud_none(*pud) || pud_bad(*pud)) |
1d18c47c CM |
128 | break; |
129 | ||
130 | pmd = pmd_offset(pud, addr); | |
6ef4fb38 | 131 | pr_cont(", *pmd=%016llx", pmd_val(*pmd)); |
4339e3f3 | 132 | if (pmd_none(*pmd) || pmd_bad(*pmd)) |
1d18c47c CM |
133 | break; |
134 | ||
135 | pte = pte_offset_map(pmd, addr); | |
6ef4fb38 | 136 | pr_cont(", *pte=%016llx", pte_val(*pte)); |
1d18c47c CM |
137 | pte_unmap(pte); |
138 | } while(0); | |
139 | ||
6ef4fb38 | 140 | pr_cont("\n"); |
1d18c47c CM |
141 | } |
142 | ||
66dbd6e6 CM |
143 | /* |
144 | * This function sets the access flags (dirty, accessed), as well as write | |
145 | * permission, and only to a more permissive setting. | |
146 | * | |
147 | * It needs to cope with hardware update of the accessed/dirty state by other | |
148 | * agents in the system and can safely skip the __sync_icache_dcache() call as, | |
149 | * like set_pte_at(), the PTE is never changed from no-exec to exec here. | |
150 | * | |
151 | * Returns whether or not the PTE actually changed. | |
152 | */ | |
153 | int ptep_set_access_flags(struct vm_area_struct *vma, | |
154 | unsigned long address, pte_t *ptep, | |
155 | pte_t entry, int dirty) | |
156 | { | |
4aae801e | 157 | pteval_t old_pteval, pteval; |
66dbd6e6 CM |
158 | |
159 | if (pte_same(*ptep, entry)) | |
160 | return 0; | |
161 | ||
162 | /* only preserve the access flags and write permission */ | |
2bbc76c3 | 163 | pte_val(entry) &= PTE_RDONLY | PTE_AF | PTE_WRITE | PTE_DIRTY; |
66dbd6e6 CM |
164 | |
165 | /* | |
166 | * Setting the flags must be done atomically to avoid racing with the | |
6d332747 CM |
167 | * hardware update of the access/dirty state. The PTE_RDONLY bit must |
168 | * be set to the most permissive (lowest value) of *ptep and entry | |
169 | * (calculated as: a & b == ~(~a | ~b)). | |
66dbd6e6 | 170 | */ |
6d332747 | 171 | pte_val(entry) ^= PTE_RDONLY; |
4aae801e CM |
172 | pteval = READ_ONCE(pte_val(*ptep)); |
173 | do { | |
174 | old_pteval = pteval; | |
175 | pteval ^= PTE_RDONLY; | |
176 | pteval |= pte_val(entry); | |
177 | pteval ^= PTE_RDONLY; | |
178 | pteval = cmpxchg_relaxed(&pte_val(*ptep), old_pteval, pteval); | |
179 | } while (pteval != old_pteval); | |
66dbd6e6 CM |
180 | |
181 | flush_tlb_fix_spurious_fault(vma, address); | |
182 | return 1; | |
183 | } | |
66dbd6e6 | 184 | |
9adeb8e7 LA |
185 | static bool is_el1_instruction_abort(unsigned int esr) |
186 | { | |
187 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR; | |
188 | } | |
189 | ||
b824b930 SB |
190 | static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs, |
191 | unsigned long addr) | |
192 | { | |
193 | unsigned int ec = ESR_ELx_EC(esr); | |
194 | unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE; | |
195 | ||
196 | if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR) | |
197 | return false; | |
198 | ||
199 | if (fsc_type == ESR_ELx_FSC_PERM) | |
200 | return true; | |
201 | ||
202 | if (addr < USER_DS && system_uses_ttbr0_pan()) | |
203 | return fsc_type == ESR_ELx_FSC_FAULT && | |
204 | (regs->pstate & PSR_PAN_BIT); | |
205 | ||
206 | return false; | |
207 | } | |
208 | ||
1d18c47c CM |
209 | /* |
210 | * The kernel tried to access some page that wasn't present. | |
211 | */ | |
67ce16ec KM |
212 | static void __do_kernel_fault(unsigned long addr, unsigned int esr, |
213 | struct pt_regs *regs) | |
1d18c47c | 214 | { |
b824b930 SB |
215 | const char *msg; |
216 | ||
1d18c47c CM |
217 | /* |
218 | * Are we prepared to handle this kernel fault? | |
9adeb8e7 | 219 | * We are almost certainly not prepared to handle instruction faults. |
1d18c47c | 220 | */ |
9adeb8e7 | 221 | if (!is_el1_instruction_abort(esr) && fixup_exception(regs)) |
1d18c47c CM |
222 | return; |
223 | ||
224 | /* | |
225 | * No handler, we'll have to terminate things with extreme prejudice. | |
226 | */ | |
227 | bust_spinlocks(1); | |
b824b930 SB |
228 | |
229 | if (is_permission_fault(esr, regs, addr)) { | |
230 | if (esr & ESR_ELx_WNR) | |
231 | msg = "write to read-only memory"; | |
232 | else | |
233 | msg = "read from unreadable memory"; | |
234 | } else if (addr < PAGE_SIZE) { | |
235 | msg = "NULL pointer dereference"; | |
236 | } else { | |
237 | msg = "paging request"; | |
238 | } | |
239 | ||
240 | pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg, | |
241 | addr); | |
1d18c47c | 242 | |
67ce16ec | 243 | show_pte(addr); |
1d18c47c CM |
244 | die("Oops", regs, esr); |
245 | bust_spinlocks(0); | |
246 | do_exit(SIGKILL); | |
247 | } | |
248 | ||
249 | /* | |
250 | * Something tried to access memory that isn't in our memory map. User mode | |
251 | * accesses just cause a SIGSEGV | |
252 | */ | |
253 | static void __do_user_fault(struct task_struct *tsk, unsigned long addr, | |
254 | unsigned int esr, unsigned int sig, int code, | |
e7c600f1 | 255 | struct pt_regs *regs, int fault) |
1d18c47c CM |
256 | { |
257 | struct siginfo si; | |
09a6adf5 | 258 | const struct fault_info *inf; |
e7c600f1 | 259 | unsigned int lsb = 0; |
1d18c47c | 260 | |
f871d268 | 261 | if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) { |
09a6adf5 | 262 | inf = esr_to_fault_info(esr); |
83016b20 | 263 | pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x", |
09a6adf5 | 264 | tsk->comm, task_pid_nr(tsk), inf->name, sig, |
3495386b | 265 | addr, esr); |
83016b20 KM |
266 | print_vma_addr(KERN_CONT ", in ", regs->pc); |
267 | pr_cont("\n"); | |
c07ab957 | 268 | __show_regs(regs); |
1d18c47c CM |
269 | } |
270 | ||
271 | tsk->thread.fault_address = addr; | |
9141300a | 272 | tsk->thread.fault_code = esr; |
1d18c47c CM |
273 | si.si_signo = sig; |
274 | si.si_errno = 0; | |
275 | si.si_code = code; | |
276 | si.si_addr = (void __user *)addr; | |
e7c600f1 JZZ |
277 | /* |
278 | * Either small page or large page may be poisoned. | |
279 | * In other words, VM_FAULT_HWPOISON_LARGE and | |
280 | * VM_FAULT_HWPOISON are mutually exclusive. | |
281 | */ | |
282 | if (fault & VM_FAULT_HWPOISON_LARGE) | |
283 | lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); | |
284 | else if (fault & VM_FAULT_HWPOISON) | |
285 | lsb = PAGE_SHIFT; | |
286 | si.si_addr_lsb = lsb; | |
287 | ||
1d18c47c CM |
288 | force_sig_info(sig, &si, tsk); |
289 | } | |
290 | ||
59f67e16 | 291 | static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs) |
1d18c47c CM |
292 | { |
293 | struct task_struct *tsk = current; | |
09a6adf5 | 294 | const struct fault_info *inf; |
1d18c47c CM |
295 | |
296 | /* | |
297 | * If we are in kernel mode at this point, we have no context to | |
298 | * handle this fault with. | |
299 | */ | |
09a6adf5 VK |
300 | if (user_mode(regs)) { |
301 | inf = esr_to_fault_info(esr); | |
e7c600f1 | 302 | __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0); |
09a6adf5 | 303 | } else |
67ce16ec | 304 | __do_kernel_fault(addr, esr, regs); |
1d18c47c CM |
305 | } |
306 | ||
307 | #define VM_FAULT_BADMAP 0x010000 | |
308 | #define VM_FAULT_BADACCESS 0x020000 | |
309 | ||
1d18c47c | 310 | static int __do_page_fault(struct mm_struct *mm, unsigned long addr, |
db6f4106 | 311 | unsigned int mm_flags, unsigned long vm_flags, |
1d18c47c CM |
312 | struct task_struct *tsk) |
313 | { | |
314 | struct vm_area_struct *vma; | |
315 | int fault; | |
316 | ||
317 | vma = find_vma(mm, addr); | |
318 | fault = VM_FAULT_BADMAP; | |
319 | if (unlikely(!vma)) | |
320 | goto out; | |
321 | if (unlikely(vma->vm_start > addr)) | |
322 | goto check_stack; | |
323 | ||
324 | /* | |
325 | * Ok, we have a good vm_area for this memory access, so we can handle | |
326 | * it. | |
327 | */ | |
328 | good_area: | |
db6f4106 WD |
329 | /* |
330 | * Check that the permissions on the VMA allow for the fault which | |
cab15ce6 | 331 | * occurred. |
db6f4106 WD |
332 | */ |
333 | if (!(vma->vm_flags & vm_flags)) { | |
1d18c47c CM |
334 | fault = VM_FAULT_BADACCESS; |
335 | goto out; | |
336 | } | |
337 | ||
dcddffd4 | 338 | return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags); |
1d18c47c CM |
339 | |
340 | check_stack: | |
341 | if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr)) | |
342 | goto good_area; | |
343 | out: | |
344 | return fault; | |
345 | } | |
346 | ||
541ec870 MR |
347 | static bool is_el0_instruction_abort(unsigned int esr) |
348 | { | |
349 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW; | |
350 | } | |
351 | ||
1d18c47c CM |
352 | static int __kprobes do_page_fault(unsigned long addr, unsigned int esr, |
353 | struct pt_regs *regs) | |
354 | { | |
355 | struct task_struct *tsk; | |
356 | struct mm_struct *mm; | |
0e3a9026 | 357 | int fault, sig, code, major = 0; |
cab15ce6 | 358 | unsigned long vm_flags = VM_READ | VM_WRITE; |
db6f4106 WD |
359 | unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; |
360 | ||
2dd0e8d2 SP |
361 | if (notify_page_fault(regs, esr)) |
362 | return 0; | |
363 | ||
1d18c47c CM |
364 | tsk = current; |
365 | mm = tsk->mm; | |
366 | ||
1d18c47c CM |
367 | /* |
368 | * If we're in an interrupt or have no user context, we must not take | |
369 | * the fault. | |
370 | */ | |
70ffdb93 | 371 | if (faulthandler_disabled() || !mm) |
1d18c47c CM |
372 | goto no_context; |
373 | ||
759496ba JW |
374 | if (user_mode(regs)) |
375 | mm_flags |= FAULT_FLAG_USER; | |
376 | ||
541ec870 | 377 | if (is_el0_instruction_abort(esr)) { |
759496ba | 378 | vm_flags = VM_EXEC; |
aed40e01 | 379 | } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) { |
759496ba JW |
380 | vm_flags = VM_WRITE; |
381 | mm_flags |= FAULT_FLAG_WRITE; | |
382 | } | |
383 | ||
b824b930 | 384 | if (addr < USER_DS && is_permission_fault(esr, regs, addr)) { |
e19a6ee2 JM |
385 | /* regs->orig_addr_limit may be 0 if we entered from EL0 */ |
386 | if (regs->orig_addr_limit == KERNEL_DS) | |
70c8abc2 | 387 | die("Accessing user space memory with fs=KERNEL_DS", regs, esr); |
70544196 | 388 | |
9adeb8e7 LA |
389 | if (is_el1_instruction_abort(esr)) |
390 | die("Attempting to execute userspace memory", regs, esr); | |
391 | ||
57f4959b | 392 | if (!search_exception_tables(regs->pc)) |
70c8abc2 | 393 | die("Accessing user space memory outside uaccess.h routines", regs, esr); |
57f4959b | 394 | } |
338d4f49 | 395 | |
0e3a9026 PA |
396 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); |
397 | ||
1d18c47c CM |
398 | /* |
399 | * As per x86, we may deadlock here. However, since the kernel only | |
400 | * validly references user space from well defined areas of the code, | |
401 | * we can bug out early if this is from code which shouldn't. | |
402 | */ | |
403 | if (!down_read_trylock(&mm->mmap_sem)) { | |
404 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
405 | goto no_context; | |
406 | retry: | |
407 | down_read(&mm->mmap_sem); | |
408 | } else { | |
409 | /* | |
410 | * The above down_read_trylock() might have succeeded in which | |
411 | * case, we'll have missed the might_sleep() from down_read(). | |
412 | */ | |
413 | might_sleep(); | |
414 | #ifdef CONFIG_DEBUG_VM | |
415 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
416 | goto no_context; | |
417 | #endif | |
418 | } | |
419 | ||
db6f4106 | 420 | fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk); |
0e3a9026 | 421 | major |= fault & VM_FAULT_MAJOR; |
1d18c47c | 422 | |
0e3a9026 PA |
423 | if (fault & VM_FAULT_RETRY) { |
424 | /* | |
425 | * If we need to retry but a fatal signal is pending, | |
426 | * handle the signal first. We do not need to release | |
427 | * the mmap_sem because it would already be released | |
428 | * in __lock_page_or_retry in mm/filemap.c. | |
429 | */ | |
289d07a2 MR |
430 | if (fatal_signal_pending(current)) { |
431 | if (!user_mode(regs)) | |
432 | goto no_context; | |
0e3a9026 | 433 | return 0; |
289d07a2 | 434 | } |
0e3a9026 PA |
435 | |
436 | /* | |
437 | * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of | |
438 | * starvation. | |
439 | */ | |
440 | if (mm_flags & FAULT_FLAG_ALLOW_RETRY) { | |
441 | mm_flags &= ~FAULT_FLAG_ALLOW_RETRY; | |
442 | mm_flags |= FAULT_FLAG_TRIED; | |
443 | goto retry; | |
444 | } | |
445 | } | |
446 | up_read(&mm->mmap_sem); | |
1d18c47c CM |
447 | |
448 | /* | |
0e3a9026 | 449 | * Handle the "normal" (no error) case first. |
1d18c47c | 450 | */ |
0e3a9026 PA |
451 | if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP | |
452 | VM_FAULT_BADACCESS)))) { | |
453 | /* | |
454 | * Major/minor page fault accounting is only done | |
455 | * once. If we go through a retry, it is extremely | |
456 | * likely that the page will be found in page cache at | |
457 | * that point. | |
458 | */ | |
459 | if (major) { | |
1d18c47c CM |
460 | tsk->maj_flt++; |
461 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, | |
462 | addr); | |
463 | } else { | |
464 | tsk->min_flt++; | |
465 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, | |
466 | addr); | |
467 | } | |
1d18c47c | 468 | |
1d18c47c | 469 | return 0; |
0e3a9026 | 470 | } |
1d18c47c | 471 | |
87134102 JW |
472 | /* |
473 | * If we are in kernel mode at this point, we have no context to | |
474 | * handle this fault with. | |
475 | */ | |
476 | if (!user_mode(regs)) | |
477 | goto no_context; | |
478 | ||
1d18c47c CM |
479 | if (fault & VM_FAULT_OOM) { |
480 | /* | |
481 | * We ran out of memory, call the OOM killer, and return to | |
482 | * userspace (which will retry the fault, or kill us if we got | |
483 | * oom-killed). | |
484 | */ | |
485 | pagefault_out_of_memory(); | |
486 | return 0; | |
487 | } | |
488 | ||
1d18c47c CM |
489 | if (fault & VM_FAULT_SIGBUS) { |
490 | /* | |
491 | * We had some memory, but were unable to successfully fix up | |
492 | * this page fault. | |
493 | */ | |
494 | sig = SIGBUS; | |
495 | code = BUS_ADRERR; | |
e7c600f1 JZZ |
496 | } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) { |
497 | sig = SIGBUS; | |
498 | code = BUS_MCEERR_AR; | |
1d18c47c CM |
499 | } else { |
500 | /* | |
501 | * Something tried to access memory that isn't in our memory | |
502 | * map. | |
503 | */ | |
504 | sig = SIGSEGV; | |
505 | code = fault == VM_FAULT_BADACCESS ? | |
506 | SEGV_ACCERR : SEGV_MAPERR; | |
507 | } | |
508 | ||
e7c600f1 | 509 | __do_user_fault(tsk, addr, esr, sig, code, regs, fault); |
1d18c47c CM |
510 | return 0; |
511 | ||
512 | no_context: | |
67ce16ec | 513 | __do_kernel_fault(addr, esr, regs); |
1d18c47c CM |
514 | return 0; |
515 | } | |
516 | ||
517 | /* | |
518 | * First Level Translation Fault Handler | |
519 | * | |
520 | * We enter here because the first level page table doesn't contain a valid | |
521 | * entry for the address. | |
522 | * | |
523 | * If the address is in kernel space (>= TASK_SIZE), then we are probably | |
524 | * faulting in the vmalloc() area. | |
525 | * | |
526 | * If the init_task's first level page tables contains the relevant entry, we | |
527 | * copy the it to this task. If not, we send the process a signal, fixup the | |
528 | * exception, or oops the kernel. | |
529 | * | |
530 | * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt | |
531 | * or a critical region, and should only copy the information from the master | |
532 | * page table, nothing more. | |
533 | */ | |
534 | static int __kprobes do_translation_fault(unsigned long addr, | |
535 | unsigned int esr, | |
536 | struct pt_regs *regs) | |
537 | { | |
538 | if (addr < TASK_SIZE) | |
539 | return do_page_fault(addr, esr, regs); | |
540 | ||
541 | do_bad_area(addr, esr, regs); | |
542 | return 0; | |
543 | } | |
544 | ||
52d7523d EL |
545 | static int do_alignment_fault(unsigned long addr, unsigned int esr, |
546 | struct pt_regs *regs) | |
547 | { | |
548 | do_bad_area(addr, esr, regs); | |
549 | return 0; | |
550 | } | |
551 | ||
1d18c47c CM |
552 | /* |
553 | * This abort handler always returns "fault". | |
554 | */ | |
555 | static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
556 | { | |
557 | return 1; | |
558 | } | |
559 | ||
32015c23 TB |
560 | /* |
561 | * This abort handler deals with Synchronous External Abort. | |
562 | * It calls notifiers, and then returns "fault". | |
563 | */ | |
564 | static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
565 | { | |
566 | struct siginfo info; | |
567 | const struct fault_info *inf; | |
621f48e4 | 568 | int ret = 0; |
32015c23 TB |
569 | |
570 | inf = esr_to_fault_info(esr); | |
571 | pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n", | |
572 | inf->name, esr, addr); | |
573 | ||
7edda088 TB |
574 | /* |
575 | * Synchronous aborts may interrupt code which had interrupts masked. | |
576 | * Before calling out into the wider kernel tell the interested | |
577 | * subsystems. | |
578 | */ | |
579 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) { | |
580 | if (interrupts_enabled(regs)) | |
581 | nmi_enter(); | |
582 | ||
621f48e4 | 583 | ret = ghes_notify_sea(); |
7edda088 TB |
584 | |
585 | if (interrupts_enabled(regs)) | |
586 | nmi_exit(); | |
587 | } | |
588 | ||
32015c23 TB |
589 | info.si_signo = SIGBUS; |
590 | info.si_errno = 0; | |
591 | info.si_code = 0; | |
592 | if (esr & ESR_ELx_FnV) | |
593 | info.si_addr = NULL; | |
594 | else | |
595 | info.si_addr = (void __user *)addr; | |
596 | arm64_notify_die("", regs, &info, esr); | |
597 | ||
621f48e4 | 598 | return ret; |
32015c23 TB |
599 | } |
600 | ||
09a6adf5 | 601 | static const struct fault_info fault_info[] = { |
1d18c47c CM |
602 | { do_bad, SIGBUS, 0, "ttbr address size fault" }, |
603 | { do_bad, SIGBUS, 0, "level 1 address size fault" }, | |
604 | { do_bad, SIGBUS, 0, "level 2 address size fault" }, | |
605 | { do_bad, SIGBUS, 0, "level 3 address size fault" }, | |
7f73f7ae | 606 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" }, |
1d18c47c CM |
607 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" }, |
608 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" }, | |
b60a0bac | 609 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" }, |
c03784ee | 610 | { do_bad, SIGBUS, 0, "unknown 8" }, |
084bd298 SC |
611 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" }, |
612 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" }, | |
1d18c47c | 613 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" }, |
c03784ee | 614 | { do_bad, SIGBUS, 0, "unknown 12" }, |
084bd298 SC |
615 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" }, |
616 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" }, | |
1d18c47c | 617 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" }, |
32015c23 | 618 | { do_sea, SIGBUS, 0, "synchronous external abort" }, |
c03784ee | 619 | { do_bad, SIGBUS, 0, "unknown 17" }, |
1d18c47c CM |
620 | { do_bad, SIGBUS, 0, "unknown 18" }, |
621 | { do_bad, SIGBUS, 0, "unknown 19" }, | |
32015c23 TB |
622 | { do_sea, SIGBUS, 0, "level 0 (translation table walk)" }, |
623 | { do_sea, SIGBUS, 0, "level 1 (translation table walk)" }, | |
624 | { do_sea, SIGBUS, 0, "level 2 (translation table walk)" }, | |
625 | { do_sea, SIGBUS, 0, "level 3 (translation table walk)" }, | |
626 | { do_sea, SIGBUS, 0, "synchronous parity or ECC error" }, | |
c03784ee | 627 | { do_bad, SIGBUS, 0, "unknown 25" }, |
1d18c47c CM |
628 | { do_bad, SIGBUS, 0, "unknown 26" }, |
629 | { do_bad, SIGBUS, 0, "unknown 27" }, | |
32015c23 TB |
630 | { do_sea, SIGBUS, 0, "level 0 synchronous parity error (translation table walk)" }, |
631 | { do_sea, SIGBUS, 0, "level 1 synchronous parity error (translation table walk)" }, | |
632 | { do_sea, SIGBUS, 0, "level 2 synchronous parity error (translation table walk)" }, | |
633 | { do_sea, SIGBUS, 0, "level 3 synchronous parity error (translation table walk)" }, | |
1d18c47c | 634 | { do_bad, SIGBUS, 0, "unknown 32" }, |
52d7523d | 635 | { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" }, |
c03784ee | 636 | { do_bad, SIGBUS, 0, "unknown 34" }, |
1d18c47c CM |
637 | { do_bad, SIGBUS, 0, "unknown 35" }, |
638 | { do_bad, SIGBUS, 0, "unknown 36" }, | |
639 | { do_bad, SIGBUS, 0, "unknown 37" }, | |
640 | { do_bad, SIGBUS, 0, "unknown 38" }, | |
641 | { do_bad, SIGBUS, 0, "unknown 39" }, | |
642 | { do_bad, SIGBUS, 0, "unknown 40" }, | |
643 | { do_bad, SIGBUS, 0, "unknown 41" }, | |
644 | { do_bad, SIGBUS, 0, "unknown 42" }, | |
645 | { do_bad, SIGBUS, 0, "unknown 43" }, | |
646 | { do_bad, SIGBUS, 0, "unknown 44" }, | |
647 | { do_bad, SIGBUS, 0, "unknown 45" }, | |
648 | { do_bad, SIGBUS, 0, "unknown 46" }, | |
649 | { do_bad, SIGBUS, 0, "unknown 47" }, | |
c03784ee | 650 | { do_bad, SIGBUS, 0, "TLB conflict abort" }, |
1d18c47c CM |
651 | { do_bad, SIGBUS, 0, "unknown 49" }, |
652 | { do_bad, SIGBUS, 0, "unknown 50" }, | |
653 | { do_bad, SIGBUS, 0, "unknown 51" }, | |
654 | { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" }, | |
c03784ee | 655 | { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" }, |
1d18c47c CM |
656 | { do_bad, SIGBUS, 0, "unknown 54" }, |
657 | { do_bad, SIGBUS, 0, "unknown 55" }, | |
658 | { do_bad, SIGBUS, 0, "unknown 56" }, | |
659 | { do_bad, SIGBUS, 0, "unknown 57" }, | |
c03784ee | 660 | { do_bad, SIGBUS, 0, "unknown 58" }, |
1d18c47c CM |
661 | { do_bad, SIGBUS, 0, "unknown 59" }, |
662 | { do_bad, SIGBUS, 0, "unknown 60" }, | |
c03784ee MR |
663 | { do_bad, SIGBUS, 0, "section domain fault" }, |
664 | { do_bad, SIGBUS, 0, "page domain fault" }, | |
1d18c47c CM |
665 | { do_bad, SIGBUS, 0, "unknown 63" }, |
666 | }; | |
667 | ||
621f48e4 TB |
668 | /* |
669 | * Handle Synchronous External Aborts that occur in a guest kernel. | |
670 | * | |
671 | * The return value will be zero if the SEA was successfully handled | |
672 | * and non-zero if there was an error processing the error or there was | |
673 | * no error to process. | |
674 | */ | |
675 | int handle_guest_sea(phys_addr_t addr, unsigned int esr) | |
676 | { | |
677 | int ret = -ENOENT; | |
678 | ||
679 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) | |
680 | ret = ghes_notify_sea(); | |
681 | ||
682 | return ret; | |
683 | } | |
684 | ||
1d18c47c CM |
685 | /* |
686 | * Dispatch a data abort to the relevant handler. | |
687 | */ | |
688 | asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr, | |
689 | struct pt_regs *regs) | |
690 | { | |
09a6adf5 | 691 | const struct fault_info *inf = esr_to_fault_info(esr); |
1d18c47c CM |
692 | struct siginfo info; |
693 | ||
694 | if (!inf->fn(addr, esr, regs)) | |
695 | return; | |
696 | ||
697 | pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n", | |
698 | inf->name, esr, addr); | |
699 | ||
700 | info.si_signo = inf->sig; | |
701 | info.si_errno = 0; | |
702 | info.si_code = inf->code; | |
703 | info.si_addr = (void __user *)addr; | |
704 | arm64_notify_die("", regs, &info, esr); | |
705 | } | |
706 | ||
707 | /* | |
708 | * Handle stack alignment exceptions. | |
709 | */ | |
710 | asmlinkage void __exception do_sp_pc_abort(unsigned long addr, | |
711 | unsigned int esr, | |
712 | struct pt_regs *regs) | |
713 | { | |
714 | struct siginfo info; | |
9e793ab8 VM |
715 | struct task_struct *tsk = current; |
716 | ||
717 | if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS)) | |
718 | pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n", | |
719 | tsk->comm, task_pid_nr(tsk), | |
720 | esr_get_class_string(esr), (void *)regs->pc, | |
721 | (void *)regs->sp); | |
1d18c47c CM |
722 | |
723 | info.si_signo = SIGBUS; | |
724 | info.si_errno = 0; | |
725 | info.si_code = BUS_ADRALN; | |
726 | info.si_addr = (void __user *)addr; | |
9e793ab8 | 727 | arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr); |
1d18c47c CM |
728 | } |
729 | ||
9fb7410f DM |
730 | int __init early_brk64(unsigned long addr, unsigned int esr, |
731 | struct pt_regs *regs); | |
732 | ||
733 | /* | |
734 | * __refdata because early_brk64 is __init, but the reference to it is | |
735 | * clobbered at arch_initcall time. | |
736 | * See traps.c and debug-monitors.c:debug_traps_init(). | |
737 | */ | |
738 | static struct fault_info __refdata debug_fault_info[] = { | |
1d18c47c CM |
739 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" }, |
740 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" }, | |
741 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" }, | |
742 | { do_bad, SIGBUS, 0, "unknown 3" }, | |
743 | { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" }, | |
744 | { do_bad, SIGTRAP, 0, "aarch32 vector catch" }, | |
9fb7410f | 745 | { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" }, |
1d18c47c CM |
746 | { do_bad, SIGBUS, 0, "unknown 7" }, |
747 | }; | |
748 | ||
749 | void __init hook_debug_fault_code(int nr, | |
750 | int (*fn)(unsigned long, unsigned int, struct pt_regs *), | |
751 | int sig, int code, const char *name) | |
752 | { | |
753 | BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info)); | |
754 | ||
755 | debug_fault_info[nr].fn = fn; | |
756 | debug_fault_info[nr].sig = sig; | |
757 | debug_fault_info[nr].code = code; | |
758 | debug_fault_info[nr].name = name; | |
759 | } | |
760 | ||
761 | asmlinkage int __exception do_debug_exception(unsigned long addr, | |
762 | unsigned int esr, | |
763 | struct pt_regs *regs) | |
764 | { | |
765 | const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr); | |
766 | struct siginfo info; | |
6afedcd2 | 767 | int rv; |
1d18c47c | 768 | |
6afedcd2 JM |
769 | /* |
770 | * Tell lockdep we disabled irqs in entry.S. Do nothing if they were | |
771 | * already disabled to preserve the last enabled/disabled addresses. | |
772 | */ | |
773 | if (interrupts_enabled(regs)) | |
774 | trace_hardirqs_off(); | |
1d18c47c | 775 | |
6afedcd2 JM |
776 | if (!inf->fn(addr, esr, regs)) { |
777 | rv = 1; | |
778 | } else { | |
779 | pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n", | |
780 | inf->name, esr, addr); | |
781 | ||
782 | info.si_signo = inf->sig; | |
783 | info.si_errno = 0; | |
784 | info.si_code = inf->code; | |
785 | info.si_addr = (void __user *)addr; | |
786 | arm64_notify_die("", regs, &info, 0); | |
787 | rv = 0; | |
788 | } | |
1d18c47c | 789 | |
6afedcd2 JM |
790 | if (interrupts_enabled(regs)) |
791 | trace_hardirqs_on(); | |
1d18c47c | 792 | |
6afedcd2 | 793 | return rv; |
1d18c47c | 794 | } |
2dd0e8d2 | 795 | NOKPROBE_SYMBOL(do_debug_exception); |
338d4f49 JM |
796 | |
797 | #ifdef CONFIG_ARM64_PAN | |
2a6dcb2b | 798 | int cpu_enable_pan(void *__unused) |
338d4f49 | 799 | { |
7209c868 JM |
800 | /* |
801 | * We modify PSTATE. This won't work from irq context as the PSTATE | |
802 | * is discarded once we return from the exception. | |
803 | */ | |
804 | WARN_ON_ONCE(in_interrupt()); | |
805 | ||
338d4f49 | 806 | config_sctlr_el1(SCTLR_EL1_SPAN, 0); |
7209c868 | 807 | asm(SET_PSTATE_PAN(1)); |
2a6dcb2b | 808 | return 0; |
338d4f49 JM |
809 | } |
810 | #endif /* CONFIG_ARM64_PAN */ |