]>
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> | |
29 | #include <linux/sched.h> | |
30 | #include <linux/highmem.h> | |
31 | #include <linux/perf_event.h> | |
7209c868 | 32 | #include <linux/preempt.h> |
1d18c47c | 33 | |
7209c868 | 34 | #include <asm/bug.h> |
338d4f49 | 35 | #include <asm/cpufeature.h> |
1d18c47c CM |
36 | #include <asm/exception.h> |
37 | #include <asm/debug-monitors.h> | |
9141300a | 38 | #include <asm/esr.h> |
338d4f49 | 39 | #include <asm/sysreg.h> |
1d18c47c CM |
40 | #include <asm/system_misc.h> |
41 | #include <asm/pgtable.h> | |
42 | #include <asm/tlbflush.h> | |
43 | ||
c46c9a74 TB |
44 | #include <acpi/ghes.h> |
45 | ||
eec2e726 VK |
46 | struct fault_info { |
47 | int (*fn)(unsigned long addr, unsigned int esr, | |
48 | struct pt_regs *regs); | |
49 | int sig; | |
50 | int code; | |
51 | const char *name; | |
52 | }; | |
53 | ||
54 | static const struct fault_info fault_info[]; | |
55 | ||
56 | static inline const struct fault_info *esr_to_fault_info(unsigned int esr) | |
57 | { | |
58 | return fault_info + (esr & 63); | |
59 | } | |
3495386b | 60 | |
2dd0e8d2 SP |
61 | #ifdef CONFIG_KPROBES |
62 | static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr) | |
63 | { | |
64 | int ret = 0; | |
65 | ||
66 | /* kprobe_running() needs smp_processor_id() */ | |
67 | if (!user_mode(regs)) { | |
68 | preempt_disable(); | |
69 | if (kprobe_running() && kprobe_fault_handler(regs, esr)) | |
70 | ret = 1; | |
71 | preempt_enable(); | |
72 | } | |
73 | ||
74 | return ret; | |
75 | } | |
76 | #else | |
77 | static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr) | |
78 | { | |
79 | return 0; | |
80 | } | |
81 | #endif | |
82 | ||
1d18c47c CM |
83 | /* |
84 | * Dump out the page tables associated with 'addr' in mm 'mm'. | |
85 | */ | |
86 | void show_pte(struct mm_struct *mm, unsigned long addr) | |
87 | { | |
88 | pgd_t *pgd; | |
89 | ||
90 | if (!mm) | |
91 | mm = &init_mm; | |
92 | ||
93 | pr_alert("pgd = %p\n", mm->pgd); | |
94 | pgd = pgd_offset(mm, addr); | |
95 | pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd)); | |
96 | ||
97 | do { | |
98 | pud_t *pud; | |
99 | pmd_t *pmd; | |
100 | pte_t *pte; | |
101 | ||
4339e3f3 | 102 | if (pgd_none(*pgd) || pgd_bad(*pgd)) |
1d18c47c CM |
103 | break; |
104 | ||
105 | pud = pud_offset(pgd, addr); | |
6ef4fb38 | 106 | pr_cont(", *pud=%016llx", pud_val(*pud)); |
4339e3f3 | 107 | if (pud_none(*pud) || pud_bad(*pud)) |
1d18c47c CM |
108 | break; |
109 | ||
110 | pmd = pmd_offset(pud, addr); | |
6ef4fb38 | 111 | pr_cont(", *pmd=%016llx", pmd_val(*pmd)); |
4339e3f3 | 112 | if (pmd_none(*pmd) || pmd_bad(*pmd)) |
1d18c47c CM |
113 | break; |
114 | ||
115 | pte = pte_offset_map(pmd, addr); | |
6ef4fb38 | 116 | pr_cont(", *pte=%016llx", pte_val(*pte)); |
1d18c47c CM |
117 | pte_unmap(pte); |
118 | } while(0); | |
119 | ||
6ef4fb38 | 120 | pr_cont("\n"); |
1d18c47c CM |
121 | } |
122 | ||
66dbd6e6 CM |
123 | #ifdef CONFIG_ARM64_HW_AFDBM |
124 | /* | |
125 | * This function sets the access flags (dirty, accessed), as well as write | |
126 | * permission, and only to a more permissive setting. | |
127 | * | |
128 | * It needs to cope with hardware update of the accessed/dirty state by other | |
129 | * agents in the system and can safely skip the __sync_icache_dcache() call as, | |
130 | * like set_pte_at(), the PTE is never changed from no-exec to exec here. | |
131 | * | |
132 | * Returns whether or not the PTE actually changed. | |
133 | */ | |
134 | int ptep_set_access_flags(struct vm_area_struct *vma, | |
135 | unsigned long address, pte_t *ptep, | |
136 | pte_t entry, int dirty) | |
137 | { | |
138 | pteval_t old_pteval; | |
139 | unsigned int tmp; | |
140 | ||
141 | if (pte_same(*ptep, entry)) | |
142 | return 0; | |
143 | ||
144 | /* only preserve the access flags and write permission */ | |
145 | pte_val(entry) &= PTE_AF | PTE_WRITE | PTE_DIRTY; | |
146 | ||
147 | /* | |
148 | * PTE_RDONLY is cleared by default in the asm below, so set it in | |
149 | * back if necessary (read-only or clean PTE). | |
150 | */ | |
0106d456 | 151 | if (!pte_write(entry) || !pte_sw_dirty(entry)) |
66dbd6e6 CM |
152 | pte_val(entry) |= PTE_RDONLY; |
153 | ||
154 | /* | |
155 | * Setting the flags must be done atomically to avoid racing with the | |
156 | * hardware update of the access/dirty state. | |
157 | */ | |
158 | asm volatile("// ptep_set_access_flags\n" | |
159 | " prfm pstl1strm, %2\n" | |
160 | "1: ldxr %0, %2\n" | |
161 | " and %0, %0, %3 // clear PTE_RDONLY\n" | |
162 | " orr %0, %0, %4 // set flags\n" | |
163 | " stxr %w1, %0, %2\n" | |
164 | " cbnz %w1, 1b\n" | |
165 | : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep)) | |
166 | : "L" (~PTE_RDONLY), "r" (pte_val(entry))); | |
167 | ||
168 | flush_tlb_fix_spurious_fault(vma, address); | |
169 | return 1; | |
170 | } | |
171 | #endif | |
172 | ||
9adeb8e7 LA |
173 | static bool is_el1_instruction_abort(unsigned int esr) |
174 | { | |
175 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR; | |
176 | } | |
177 | ||
1d18c47c CM |
178 | /* |
179 | * The kernel tried to access some page that wasn't present. | |
180 | */ | |
181 | static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr, | |
182 | unsigned int esr, struct pt_regs *regs) | |
183 | { | |
184 | /* | |
185 | * Are we prepared to handle this kernel fault? | |
9adeb8e7 | 186 | * We are almost certainly not prepared to handle instruction faults. |
1d18c47c | 187 | */ |
9adeb8e7 | 188 | if (!is_el1_instruction_abort(esr) && fixup_exception(regs)) |
1d18c47c CM |
189 | return; |
190 | ||
191 | /* | |
192 | * No handler, we'll have to terminate things with extreme prejudice. | |
193 | */ | |
194 | bust_spinlocks(1); | |
195 | pr_alert("Unable to handle kernel %s at virtual address %08lx\n", | |
196 | (addr < PAGE_SIZE) ? "NULL pointer dereference" : | |
197 | "paging request", addr); | |
198 | ||
199 | show_pte(mm, addr); | |
200 | die("Oops", regs, esr); | |
201 | bust_spinlocks(0); | |
202 | do_exit(SIGKILL); | |
203 | } | |
204 | ||
205 | /* | |
206 | * Something tried to access memory that isn't in our memory map. User mode | |
207 | * accesses just cause a SIGSEGV | |
208 | */ | |
209 | static void __do_user_fault(struct task_struct *tsk, unsigned long addr, | |
210 | unsigned int esr, unsigned int sig, int code, | |
211 | struct pt_regs *regs) | |
212 | { | |
213 | struct siginfo si; | |
eec2e726 | 214 | const struct fault_info *inf; |
1d18c47c | 215 | |
f871d268 | 216 | if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) { |
eec2e726 | 217 | inf = esr_to_fault_info(esr); |
3495386b | 218 | pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n", |
eec2e726 | 219 | tsk->comm, task_pid_nr(tsk), inf->name, sig, |
3495386b | 220 | addr, esr); |
1d18c47c CM |
221 | show_pte(tsk->mm, addr); |
222 | show_regs(regs); | |
223 | } | |
224 | ||
225 | tsk->thread.fault_address = addr; | |
9141300a | 226 | tsk->thread.fault_code = esr; |
1d18c47c CM |
227 | si.si_signo = sig; |
228 | si.si_errno = 0; | |
229 | si.si_code = code; | |
230 | si.si_addr = (void __user *)addr; | |
231 | force_sig_info(sig, &si, tsk); | |
232 | } | |
233 | ||
59f67e16 | 234 | static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs) |
1d18c47c CM |
235 | { |
236 | struct task_struct *tsk = current; | |
237 | struct mm_struct *mm = tsk->active_mm; | |
eec2e726 | 238 | const struct fault_info *inf; |
1d18c47c CM |
239 | |
240 | /* | |
241 | * If we are in kernel mode at this point, we have no context to | |
242 | * handle this fault with. | |
243 | */ | |
eec2e726 VK |
244 | if (user_mode(regs)) { |
245 | inf = esr_to_fault_info(esr); | |
246 | __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs); | |
247 | } else | |
1d18c47c CM |
248 | __do_kernel_fault(mm, addr, esr, regs); |
249 | } | |
250 | ||
251 | #define VM_FAULT_BADMAP 0x010000 | |
252 | #define VM_FAULT_BADACCESS 0x020000 | |
253 | ||
1d18c47c | 254 | static int __do_page_fault(struct mm_struct *mm, unsigned long addr, |
db6f4106 | 255 | unsigned int mm_flags, unsigned long vm_flags, |
1d18c47c CM |
256 | struct task_struct *tsk) |
257 | { | |
258 | struct vm_area_struct *vma; | |
259 | int fault; | |
260 | ||
261 | vma = find_vma(mm, addr); | |
262 | fault = VM_FAULT_BADMAP; | |
263 | if (unlikely(!vma)) | |
264 | goto out; | |
265 | if (unlikely(vma->vm_start > addr)) | |
266 | goto check_stack; | |
267 | ||
268 | /* | |
269 | * Ok, we have a good vm_area for this memory access, so we can handle | |
270 | * it. | |
271 | */ | |
272 | good_area: | |
db6f4106 WD |
273 | /* |
274 | * Check that the permissions on the VMA allow for the fault which | |
cab15ce6 | 275 | * occurred. |
db6f4106 WD |
276 | */ |
277 | if (!(vma->vm_flags & vm_flags)) { | |
1d18c47c CM |
278 | fault = VM_FAULT_BADACCESS; |
279 | goto out; | |
280 | } | |
281 | ||
dcddffd4 | 282 | return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags); |
1d18c47c CM |
283 | |
284 | check_stack: | |
285 | if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr)) | |
286 | goto good_area; | |
287 | out: | |
288 | return fault; | |
289 | } | |
290 | ||
78688963 | 291 | static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs) |
57f4959b | 292 | { |
275f344b | 293 | unsigned int ec = ESR_ELx_EC(esr); |
57f4959b JM |
294 | unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE; |
295 | ||
78688963 CM |
296 | if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR) |
297 | return false; | |
298 | ||
299 | if (system_uses_ttbr0_pan()) | |
300 | return fsc_type == ESR_ELx_FSC_FAULT && | |
301 | (regs->pstate & PSR_PAN_BIT); | |
302 | else | |
303 | return fsc_type == ESR_ELx_FSC_PERM; | |
57f4959b JM |
304 | } |
305 | ||
541ec870 MR |
306 | static bool is_el0_instruction_abort(unsigned int esr) |
307 | { | |
308 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW; | |
309 | } | |
310 | ||
1d18c47c CM |
311 | static int __kprobes do_page_fault(unsigned long addr, unsigned int esr, |
312 | struct pt_regs *regs) | |
313 | { | |
314 | struct task_struct *tsk; | |
315 | struct mm_struct *mm; | |
316 | int fault, sig, code; | |
cab15ce6 | 317 | unsigned long vm_flags = VM_READ | VM_WRITE; |
db6f4106 WD |
318 | unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; |
319 | ||
2dd0e8d2 SP |
320 | if (notify_page_fault(regs, esr)) |
321 | return 0; | |
322 | ||
1d18c47c CM |
323 | tsk = current; |
324 | mm = tsk->mm; | |
325 | ||
1d18c47c CM |
326 | /* |
327 | * If we're in an interrupt or have no user context, we must not take | |
328 | * the fault. | |
329 | */ | |
70ffdb93 | 330 | if (faulthandler_disabled() || !mm) |
1d18c47c CM |
331 | goto no_context; |
332 | ||
759496ba JW |
333 | if (user_mode(regs)) |
334 | mm_flags |= FAULT_FLAG_USER; | |
335 | ||
541ec870 | 336 | if (is_el0_instruction_abort(esr)) { |
759496ba | 337 | vm_flags = VM_EXEC; |
aed40e01 | 338 | } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) { |
759496ba JW |
339 | vm_flags = VM_WRITE; |
340 | mm_flags |= FAULT_FLAG_WRITE; | |
341 | } | |
342 | ||
78688963 | 343 | if (addr < USER_DS && is_permission_fault(esr, regs)) { |
e19a6ee2 JM |
344 | /* regs->orig_addr_limit may be 0 if we entered from EL0 */ |
345 | if (regs->orig_addr_limit == KERNEL_DS) | |
70c8abc2 | 346 | die("Accessing user space memory with fs=KERNEL_DS", regs, esr); |
70544196 | 347 | |
9adeb8e7 LA |
348 | if (is_el1_instruction_abort(esr)) |
349 | die("Attempting to execute userspace memory", regs, esr); | |
350 | ||
57f4959b | 351 | if (!search_exception_tables(regs->pc)) |
70c8abc2 | 352 | die("Accessing user space memory outside uaccess.h routines", regs, esr); |
57f4959b | 353 | } |
338d4f49 | 354 | |
1d18c47c CM |
355 | /* |
356 | * As per x86, we may deadlock here. However, since the kernel only | |
357 | * validly references user space from well defined areas of the code, | |
358 | * we can bug out early if this is from code which shouldn't. | |
359 | */ | |
360 | if (!down_read_trylock(&mm->mmap_sem)) { | |
361 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
362 | goto no_context; | |
363 | retry: | |
364 | down_read(&mm->mmap_sem); | |
365 | } else { | |
366 | /* | |
367 | * The above down_read_trylock() might have succeeded in which | |
368 | * case, we'll have missed the might_sleep() from down_read(). | |
369 | */ | |
370 | might_sleep(); | |
371 | #ifdef CONFIG_DEBUG_VM | |
372 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
373 | goto no_context; | |
374 | #endif | |
375 | } | |
376 | ||
db6f4106 | 377 | fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk); |
1d18c47c CM |
378 | |
379 | /* | |
380 | * If we need to retry but a fatal signal is pending, handle the | |
381 | * signal first. We do not need to release the mmap_sem because it | |
382 | * would already be released in __lock_page_or_retry in mm/filemap.c. | |
383 | */ | |
384 | if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current)) | |
385 | return 0; | |
386 | ||
387 | /* | |
388 | * Major/minor page fault accounting is only done on the initial | |
389 | * attempt. If we go through a retry, it is extremely likely that the | |
390 | * page will be found in page cache at that point. | |
391 | */ | |
392 | ||
393 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); | |
db6f4106 | 394 | if (mm_flags & FAULT_FLAG_ALLOW_RETRY) { |
1d18c47c CM |
395 | if (fault & VM_FAULT_MAJOR) { |
396 | tsk->maj_flt++; | |
397 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, | |
398 | addr); | |
399 | } else { | |
400 | tsk->min_flt++; | |
401 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, | |
402 | addr); | |
403 | } | |
404 | if (fault & VM_FAULT_RETRY) { | |
405 | /* | |
406 | * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of | |
407 | * starvation. | |
408 | */ | |
db6f4106 | 409 | mm_flags &= ~FAULT_FLAG_ALLOW_RETRY; |
569ba74a | 410 | mm_flags |= FAULT_FLAG_TRIED; |
1d18c47c CM |
411 | goto retry; |
412 | } | |
413 | } | |
414 | ||
415 | up_read(&mm->mmap_sem); | |
416 | ||
417 | /* | |
0e8fb931 | 418 | * Handle the "normal" case first - VM_FAULT_MAJOR |
1d18c47c CM |
419 | */ |
420 | if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP | | |
421 | VM_FAULT_BADACCESS)))) | |
422 | return 0; | |
423 | ||
87134102 JW |
424 | /* |
425 | * If we are in kernel mode at this point, we have no context to | |
426 | * handle this fault with. | |
427 | */ | |
428 | if (!user_mode(regs)) | |
429 | goto no_context; | |
430 | ||
1d18c47c CM |
431 | if (fault & VM_FAULT_OOM) { |
432 | /* | |
433 | * We ran out of memory, call the OOM killer, and return to | |
434 | * userspace (which will retry the fault, or kill us if we got | |
435 | * oom-killed). | |
436 | */ | |
437 | pagefault_out_of_memory(); | |
438 | return 0; | |
439 | } | |
440 | ||
1d18c47c CM |
441 | if (fault & VM_FAULT_SIGBUS) { |
442 | /* | |
443 | * We had some memory, but were unable to successfully fix up | |
444 | * this page fault. | |
445 | */ | |
446 | sig = SIGBUS; | |
447 | code = BUS_ADRERR; | |
448 | } else { | |
449 | /* | |
450 | * Something tried to access memory that isn't in our memory | |
451 | * map. | |
452 | */ | |
453 | sig = SIGSEGV; | |
454 | code = fault == VM_FAULT_BADACCESS ? | |
455 | SEGV_ACCERR : SEGV_MAPERR; | |
456 | } | |
457 | ||
458 | __do_user_fault(tsk, addr, esr, sig, code, regs); | |
459 | return 0; | |
460 | ||
461 | no_context: | |
462 | __do_kernel_fault(mm, addr, esr, regs); | |
463 | return 0; | |
464 | } | |
465 | ||
466 | /* | |
467 | * First Level Translation Fault Handler | |
468 | * | |
469 | * We enter here because the first level page table doesn't contain a valid | |
470 | * entry for the address. | |
471 | * | |
472 | * If the address is in kernel space (>= TASK_SIZE), then we are probably | |
473 | * faulting in the vmalloc() area. | |
474 | * | |
475 | * If the init_task's first level page tables contains the relevant entry, we | |
476 | * copy the it to this task. If not, we send the process a signal, fixup the | |
477 | * exception, or oops the kernel. | |
478 | * | |
479 | * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt | |
480 | * or a critical region, and should only copy the information from the master | |
481 | * page table, nothing more. | |
482 | */ | |
483 | static int __kprobes do_translation_fault(unsigned long addr, | |
484 | unsigned int esr, | |
485 | struct pt_regs *regs) | |
486 | { | |
487 | if (addr < TASK_SIZE) | |
488 | return do_page_fault(addr, esr, regs); | |
489 | ||
490 | do_bad_area(addr, esr, regs); | |
491 | return 0; | |
492 | } | |
493 | ||
52d7523d EL |
494 | static int do_alignment_fault(unsigned long addr, unsigned int esr, |
495 | struct pt_regs *regs) | |
496 | { | |
497 | do_bad_area(addr, esr, regs); | |
498 | return 0; | |
499 | } | |
500 | ||
1d18c47c CM |
501 | /* |
502 | * This abort handler always returns "fault". | |
503 | */ | |
504 | static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
505 | { | |
506 | return 1; | |
507 | } | |
508 | ||
a9bfe61a TB |
509 | /* |
510 | * This abort handler deals with Synchronous External Abort. | |
511 | * It calls notifiers, and then returns "fault". | |
512 | */ | |
513 | static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
514 | { | |
515 | struct siginfo info; | |
516 | const struct fault_info *inf; | |
6633b457 | 517 | int ret = 0; |
a9bfe61a TB |
518 | |
519 | inf = esr_to_fault_info(esr); | |
520 | pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n", | |
521 | inf->name, esr, addr); | |
522 | ||
c46c9a74 TB |
523 | /* |
524 | * Synchronous aborts may interrupt code which had interrupts masked. | |
525 | * Before calling out into the wider kernel tell the interested | |
526 | * subsystems. | |
527 | */ | |
528 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) { | |
529 | if (interrupts_enabled(regs)) | |
530 | nmi_enter(); | |
531 | ||
6633b457 | 532 | ret = ghes_notify_sea(); |
c46c9a74 TB |
533 | |
534 | if (interrupts_enabled(regs)) | |
535 | nmi_exit(); | |
536 | } | |
537 | ||
a9bfe61a TB |
538 | info.si_signo = SIGBUS; |
539 | info.si_errno = 0; | |
540 | info.si_code = 0; | |
541 | if (esr & ESR_ELx_FnV) | |
542 | info.si_addr = NULL; | |
543 | else | |
544 | info.si_addr = (void __user *)addr; | |
545 | arm64_notify_die("", regs, &info, esr); | |
546 | ||
6633b457 | 547 | return ret; |
a9bfe61a TB |
548 | } |
549 | ||
eec2e726 | 550 | static const struct fault_info fault_info[] = { |
1d18c47c CM |
551 | { do_bad, SIGBUS, 0, "ttbr address size fault" }, |
552 | { do_bad, SIGBUS, 0, "level 1 address size fault" }, | |
553 | { do_bad, SIGBUS, 0, "level 2 address size fault" }, | |
554 | { do_bad, SIGBUS, 0, "level 3 address size fault" }, | |
7f73f7ae | 555 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" }, |
1d18c47c CM |
556 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" }, |
557 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" }, | |
558 | { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" }, | |
c03784ee | 559 | { do_bad, SIGBUS, 0, "unknown 8" }, |
084bd298 SC |
560 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" }, |
561 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" }, | |
1d18c47c | 562 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" }, |
c03784ee | 563 | { do_bad, SIGBUS, 0, "unknown 12" }, |
084bd298 SC |
564 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" }, |
565 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" }, | |
1d18c47c | 566 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" }, |
a9bfe61a | 567 | { do_sea, SIGBUS, 0, "synchronous external abort" }, |
c03784ee | 568 | { do_bad, SIGBUS, 0, "unknown 17" }, |
1d18c47c CM |
569 | { do_bad, SIGBUS, 0, "unknown 18" }, |
570 | { do_bad, SIGBUS, 0, "unknown 19" }, | |
a9bfe61a TB |
571 | { do_sea, SIGBUS, 0, "level 0 (translation table walk)" }, |
572 | { do_sea, SIGBUS, 0, "level 1 (translation table walk)" }, | |
573 | { do_sea, SIGBUS, 0, "level 2 (translation table walk)" }, | |
574 | { do_sea, SIGBUS, 0, "level 3 (translation table walk)" }, | |
575 | { do_sea, SIGBUS, 0, "synchronous parity or ECC error" }, | |
c03784ee | 576 | { do_bad, SIGBUS, 0, "unknown 25" }, |
1d18c47c CM |
577 | { do_bad, SIGBUS, 0, "unknown 26" }, |
578 | { do_bad, SIGBUS, 0, "unknown 27" }, | |
a9bfe61a TB |
579 | { do_sea, SIGBUS, 0, "level 0 synchronous parity error (translation table walk)" }, |
580 | { do_sea, SIGBUS, 0, "level 1 synchronous parity error (translation table walk)" }, | |
581 | { do_sea, SIGBUS, 0, "level 2 synchronous parity error (translation table walk)" }, | |
582 | { do_sea, SIGBUS, 0, "level 3 synchronous parity error (translation table walk)" }, | |
1d18c47c | 583 | { do_bad, SIGBUS, 0, "unknown 32" }, |
52d7523d | 584 | { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" }, |
c03784ee | 585 | { do_bad, SIGBUS, 0, "unknown 34" }, |
1d18c47c CM |
586 | { do_bad, SIGBUS, 0, "unknown 35" }, |
587 | { do_bad, SIGBUS, 0, "unknown 36" }, | |
588 | { do_bad, SIGBUS, 0, "unknown 37" }, | |
589 | { do_bad, SIGBUS, 0, "unknown 38" }, | |
590 | { do_bad, SIGBUS, 0, "unknown 39" }, | |
591 | { do_bad, SIGBUS, 0, "unknown 40" }, | |
592 | { do_bad, SIGBUS, 0, "unknown 41" }, | |
593 | { do_bad, SIGBUS, 0, "unknown 42" }, | |
594 | { do_bad, SIGBUS, 0, "unknown 43" }, | |
595 | { do_bad, SIGBUS, 0, "unknown 44" }, | |
596 | { do_bad, SIGBUS, 0, "unknown 45" }, | |
597 | { do_bad, SIGBUS, 0, "unknown 46" }, | |
598 | { do_bad, SIGBUS, 0, "unknown 47" }, | |
c03784ee | 599 | { do_bad, SIGBUS, 0, "TLB conflict abort" }, |
1d18c47c CM |
600 | { do_bad, SIGBUS, 0, "unknown 49" }, |
601 | { do_bad, SIGBUS, 0, "unknown 50" }, | |
602 | { do_bad, SIGBUS, 0, "unknown 51" }, | |
603 | { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" }, | |
c03784ee | 604 | { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" }, |
1d18c47c CM |
605 | { do_bad, SIGBUS, 0, "unknown 54" }, |
606 | { do_bad, SIGBUS, 0, "unknown 55" }, | |
607 | { do_bad, SIGBUS, 0, "unknown 56" }, | |
608 | { do_bad, SIGBUS, 0, "unknown 57" }, | |
c03784ee | 609 | { do_bad, SIGBUS, 0, "unknown 58" }, |
1d18c47c CM |
610 | { do_bad, SIGBUS, 0, "unknown 59" }, |
611 | { do_bad, SIGBUS, 0, "unknown 60" }, | |
c03784ee MR |
612 | { do_bad, SIGBUS, 0, "section domain fault" }, |
613 | { do_bad, SIGBUS, 0, "page domain fault" }, | |
1d18c47c CM |
614 | { do_bad, SIGBUS, 0, "unknown 63" }, |
615 | }; | |
616 | ||
6633b457 TB |
617 | /* |
618 | * Handle Synchronous External Aborts that occur in a guest kernel. | |
619 | * | |
620 | * The return value will be zero if the SEA was successfully handled | |
621 | * and non-zero if there was an error processing the error or there was | |
622 | * no error to process. | |
623 | */ | |
624 | int handle_guest_sea(phys_addr_t addr, unsigned int esr) | |
625 | { | |
626 | int ret = -ENOENT; | |
627 | ||
628 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) | |
629 | ret = ghes_notify_sea(); | |
630 | ||
631 | return ret; | |
632 | } | |
633 | ||
1d18c47c CM |
634 | /* |
635 | * Dispatch a data abort to the relevant handler. | |
636 | */ | |
637 | asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr, | |
638 | struct pt_regs *regs) | |
639 | { | |
eec2e726 | 640 | const struct fault_info *inf = esr_to_fault_info(esr); |
1d18c47c CM |
641 | struct siginfo info; |
642 | ||
643 | if (!inf->fn(addr, esr, regs)) | |
644 | return; | |
645 | ||
646 | pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n", | |
647 | inf->name, esr, addr); | |
648 | ||
649 | info.si_signo = inf->sig; | |
650 | info.si_errno = 0; | |
651 | info.si_code = inf->code; | |
652 | info.si_addr = (void __user *)addr; | |
653 | arm64_notify_die("", regs, &info, esr); | |
654 | } | |
655 | ||
656 | /* | |
657 | * Handle stack alignment exceptions. | |
658 | */ | |
659 | asmlinkage void __exception do_sp_pc_abort(unsigned long addr, | |
660 | unsigned int esr, | |
661 | struct pt_regs *regs) | |
662 | { | |
663 | struct siginfo info; | |
9e793ab8 VM |
664 | struct task_struct *tsk = current; |
665 | ||
666 | if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS)) | |
667 | pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n", | |
668 | tsk->comm, task_pid_nr(tsk), | |
669 | esr_get_class_string(esr), (void *)regs->pc, | |
670 | (void *)regs->sp); | |
1d18c47c CM |
671 | |
672 | info.si_signo = SIGBUS; | |
673 | info.si_errno = 0; | |
674 | info.si_code = BUS_ADRALN; | |
675 | info.si_addr = (void __user *)addr; | |
9e793ab8 | 676 | arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr); |
1d18c47c CM |
677 | } |
678 | ||
9fb7410f DM |
679 | int __init early_brk64(unsigned long addr, unsigned int esr, |
680 | struct pt_regs *regs); | |
681 | ||
682 | /* | |
683 | * __refdata because early_brk64 is __init, but the reference to it is | |
684 | * clobbered at arch_initcall time. | |
685 | * See traps.c and debug-monitors.c:debug_traps_init(). | |
686 | */ | |
687 | static struct fault_info __refdata debug_fault_info[] = { | |
1d18c47c CM |
688 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" }, |
689 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" }, | |
690 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" }, | |
691 | { do_bad, SIGBUS, 0, "unknown 3" }, | |
692 | { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" }, | |
693 | { do_bad, SIGTRAP, 0, "aarch32 vector catch" }, | |
9fb7410f | 694 | { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" }, |
1d18c47c CM |
695 | { do_bad, SIGBUS, 0, "unknown 7" }, |
696 | }; | |
697 | ||
698 | void __init hook_debug_fault_code(int nr, | |
699 | int (*fn)(unsigned long, unsigned int, struct pt_regs *), | |
700 | int sig, int code, const char *name) | |
701 | { | |
702 | BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info)); | |
703 | ||
704 | debug_fault_info[nr].fn = fn; | |
705 | debug_fault_info[nr].sig = sig; | |
706 | debug_fault_info[nr].code = code; | |
707 | debug_fault_info[nr].name = name; | |
708 | } | |
709 | ||
710 | asmlinkage int __exception do_debug_exception(unsigned long addr, | |
711 | unsigned int esr, | |
712 | struct pt_regs *regs) | |
713 | { | |
714 | const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr); | |
715 | struct siginfo info; | |
6afedcd2 | 716 | int rv; |
1d18c47c | 717 | |
6afedcd2 JM |
718 | /* |
719 | * Tell lockdep we disabled irqs in entry.S. Do nothing if they were | |
720 | * already disabled to preserve the last enabled/disabled addresses. | |
721 | */ | |
722 | if (interrupts_enabled(regs)) | |
723 | trace_hardirqs_off(); | |
1d18c47c | 724 | |
6afedcd2 JM |
725 | if (!inf->fn(addr, esr, regs)) { |
726 | rv = 1; | |
727 | } else { | |
728 | pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n", | |
729 | inf->name, esr, addr); | |
730 | ||
731 | info.si_signo = inf->sig; | |
732 | info.si_errno = 0; | |
733 | info.si_code = inf->code; | |
734 | info.si_addr = (void __user *)addr; | |
735 | arm64_notify_die("", regs, &info, 0); | |
736 | rv = 0; | |
737 | } | |
1d18c47c | 738 | |
6afedcd2 JM |
739 | if (interrupts_enabled(regs)) |
740 | trace_hardirqs_on(); | |
1d18c47c | 741 | |
6afedcd2 | 742 | return rv; |
1d18c47c | 743 | } |
2dd0e8d2 | 744 | NOKPROBE_SYMBOL(do_debug_exception); |
338d4f49 JM |
745 | |
746 | #ifdef CONFIG_ARM64_PAN | |
2a6dcb2b | 747 | int cpu_enable_pan(void *__unused) |
338d4f49 | 748 | { |
7209c868 JM |
749 | /* |
750 | * We modify PSTATE. This won't work from irq context as the PSTATE | |
751 | * is discarded once we return from the exception. | |
752 | */ | |
753 | WARN_ON_ONCE(in_interrupt()); | |
754 | ||
338d4f49 | 755 | config_sctlr_el1(SCTLR_EL1_SPAN, 0); |
7209c868 | 756 | asm(SET_PSTATE_PAN(1)); |
2a6dcb2b | 757 | return 0; |
338d4f49 JM |
758 | } |
759 | #endif /* CONFIG_ARM64_PAN */ | |
57f4959b JM |
760 | |
761 | #ifdef CONFIG_ARM64_UAO | |
762 | /* | |
763 | * Kernel threads have fs=KERNEL_DS by default, and don't need to call | |
764 | * set_fs(), devtmpfs in particular relies on this behaviour. | |
765 | * We need to enable the feature at runtime (instead of adding it to | |
766 | * PSR_MODE_EL1h) as the feature may not be implemented by the cpu. | |
767 | */ | |
2a6dcb2b | 768 | int cpu_enable_uao(void *__unused) |
57f4959b JM |
769 | { |
770 | asm(SET_PSTATE_UAO(1)); | |
2a6dcb2b | 771 | return 0; |
57f4959b JM |
772 | } |
773 | #endif /* CONFIG_ARM64_UAO */ |