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> |
3bbf7157 | 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 | ||
1f9b8936 JT |
86 | static void data_abort_decode(unsigned int esr) |
87 | { | |
88 | pr_alert("Data abort info:\n"); | |
89 | ||
90 | if (esr & ESR_ELx_ISV) { | |
91 | pr_alert(" Access size = %u byte(s)\n", | |
92 | 1U << ((esr & ESR_ELx_SAS) >> ESR_ELx_SAS_SHIFT)); | |
93 | pr_alert(" SSE = %lu, SRT = %lu\n", | |
94 | (esr & ESR_ELx_SSE) >> ESR_ELx_SSE_SHIFT, | |
95 | (esr & ESR_ELx_SRT_MASK) >> ESR_ELx_SRT_SHIFT); | |
96 | pr_alert(" SF = %lu, AR = %lu\n", | |
97 | (esr & ESR_ELx_SF) >> ESR_ELx_SF_SHIFT, | |
98 | (esr & ESR_ELx_AR) >> ESR_ELx_AR_SHIFT); | |
99 | } else { | |
100 | pr_alert(" ISV = 0, ISS = 0x%08lu\n", esr & ESR_ELx_ISS_MASK); | |
101 | } | |
102 | ||
103 | pr_alert(" CM = %lu, WnR = %lu\n", | |
104 | (esr & ESR_ELx_CM) >> ESR_ELx_CM_SHIFT, | |
105 | (esr & ESR_ELx_WNR) >> ESR_ELx_WNR_SHIFT); | |
106 | } | |
107 | ||
108 | /* | |
109 | * Decode mem abort information | |
110 | */ | |
111 | static void mem_abort_decode(unsigned int esr) | |
112 | { | |
113 | pr_alert("Mem abort info:\n"); | |
114 | ||
115 | pr_alert(" Exception class = %s, IL = %u bits\n", | |
116 | esr_get_class_string(esr), | |
117 | (esr & ESR_ELx_IL) ? 32 : 16); | |
118 | pr_alert(" SET = %lu, FnV = %lu\n", | |
119 | (esr & ESR_ELx_SET_MASK) >> ESR_ELx_SET_SHIFT, | |
120 | (esr & ESR_ELx_FnV) >> ESR_ELx_FnV_SHIFT); | |
121 | pr_alert(" EA = %lu, S1PTW = %lu\n", | |
122 | (esr & ESR_ELx_EA) >> ESR_ELx_EA_SHIFT, | |
123 | (esr & ESR_ELx_S1PTW) >> ESR_ELx_S1PTW_SHIFT); | |
124 | ||
125 | if (esr_is_data_abort(esr)) | |
126 | data_abort_decode(esr); | |
127 | } | |
128 | ||
1d18c47c | 129 | /* |
67ce16ec | 130 | * Dump out the page tables associated with 'addr' in the currently active mm. |
1d18c47c | 131 | */ |
67ce16ec | 132 | void show_pte(unsigned long addr) |
1d18c47c | 133 | { |
67ce16ec | 134 | struct mm_struct *mm; |
1d18c47c CM |
135 | pgd_t *pgd; |
136 | ||
67ce16ec KM |
137 | if (addr < TASK_SIZE) { |
138 | /* TTBR0 */ | |
139 | mm = current->active_mm; | |
140 | if (mm == &init_mm) { | |
141 | pr_alert("[%016lx] user address but active_mm is swapper\n", | |
142 | addr); | |
143 | return; | |
144 | } | |
145 | } else if (addr >= VA_START) { | |
146 | /* TTBR1 */ | |
1d18c47c | 147 | mm = &init_mm; |
67ce16ec KM |
148 | } else { |
149 | pr_alert("[%016lx] address between user and kernel address ranges\n", | |
150 | addr); | |
151 | return; | |
152 | } | |
1d18c47c | 153 | |
1eb34b6e WD |
154 | pr_alert("%s pgtable: %luk pages, %u-bit VAs, pgd = %p\n", |
155 | mm == &init_mm ? "swapper" : "user", PAGE_SIZE / SZ_1K, | |
156 | VA_BITS, mm->pgd); | |
1d18c47c | 157 | pgd = pgd_offset(mm, addr); |
67ce16ec | 158 | pr_alert("[%016lx] *pgd=%016llx", addr, pgd_val(*pgd)); |
1d18c47c CM |
159 | |
160 | do { | |
161 | pud_t *pud; | |
162 | pmd_t *pmd; | |
163 | pte_t *pte; | |
164 | ||
4339e3f3 | 165 | if (pgd_none(*pgd) || pgd_bad(*pgd)) |
1d18c47c CM |
166 | break; |
167 | ||
168 | pud = pud_offset(pgd, addr); | |
6ef4fb38 | 169 | pr_cont(", *pud=%016llx", pud_val(*pud)); |
4339e3f3 | 170 | if (pud_none(*pud) || pud_bad(*pud)) |
1d18c47c CM |
171 | break; |
172 | ||
173 | pmd = pmd_offset(pud, addr); | |
6ef4fb38 | 174 | pr_cont(", *pmd=%016llx", pmd_val(*pmd)); |
4339e3f3 | 175 | if (pmd_none(*pmd) || pmd_bad(*pmd)) |
1d18c47c CM |
176 | break; |
177 | ||
178 | pte = pte_offset_map(pmd, addr); | |
6ef4fb38 | 179 | pr_cont(", *pte=%016llx", pte_val(*pte)); |
1d18c47c CM |
180 | pte_unmap(pte); |
181 | } while(0); | |
182 | ||
6ef4fb38 | 183 | pr_cont("\n"); |
1d18c47c CM |
184 | } |
185 | ||
66dbd6e6 CM |
186 | #ifdef CONFIG_ARM64_HW_AFDBM |
187 | /* | |
188 | * This function sets the access flags (dirty, accessed), as well as write | |
189 | * permission, and only to a more permissive setting. | |
190 | * | |
191 | * It needs to cope with hardware update of the accessed/dirty state by other | |
192 | * agents in the system and can safely skip the __sync_icache_dcache() call as, | |
193 | * like set_pte_at(), the PTE is never changed from no-exec to exec here. | |
194 | * | |
195 | * Returns whether or not the PTE actually changed. | |
196 | */ | |
197 | int ptep_set_access_flags(struct vm_area_struct *vma, | |
198 | unsigned long address, pte_t *ptep, | |
199 | pte_t entry, int dirty) | |
200 | { | |
3bbf7157 | 201 | pteval_t old_pteval, pteval; |
66dbd6e6 CM |
202 | |
203 | if (pte_same(*ptep, entry)) | |
204 | return 0; | |
205 | ||
206 | /* only preserve the access flags and write permission */ | |
207 | pte_val(entry) &= PTE_AF | PTE_WRITE | PTE_DIRTY; | |
208 | ||
6d332747 | 209 | /* set PTE_RDONLY if actual read-only or clean PTE */ |
0106d456 | 210 | if (!pte_write(entry) || !pte_sw_dirty(entry)) |
3bbf7157 | 211 | entry = pte_set_rdonly(entry); |
66dbd6e6 CM |
212 | |
213 | /* | |
214 | * Setting the flags must be done atomically to avoid racing with the | |
6d332747 CM |
215 | * hardware update of the access/dirty state. The PTE_RDONLY bit must |
216 | * be set to the most permissive (lowest value) of *ptep and entry | |
217 | * (calculated as: a & b == ~(~a | ~b)). | |
66dbd6e6 | 218 | */ |
6d332747 | 219 | pte_val(entry) ^= PTE_RDONLY; |
3bbf7157 CM |
220 | pteval = READ_ONCE(pte_val(*ptep)); |
221 | do { | |
222 | old_pteval = pteval; | |
223 | pteval ^= PTE_RDONLY; | |
224 | pteval |= pte_val(entry); | |
225 | pteval ^= PTE_RDONLY; | |
226 | pteval = cmpxchg_relaxed(&pte_val(*ptep), old_pteval, pteval); | |
227 | } while (pteval != old_pteval); | |
66dbd6e6 CM |
228 | |
229 | flush_tlb_fix_spurious_fault(vma, address); | |
230 | return 1; | |
231 | } | |
232 | #endif | |
233 | ||
9adeb8e7 LA |
234 | static bool is_el1_instruction_abort(unsigned int esr) |
235 | { | |
236 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR; | |
237 | } | |
238 | ||
b824b930 SB |
239 | static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs, |
240 | unsigned long addr) | |
241 | { | |
242 | unsigned int ec = ESR_ELx_EC(esr); | |
243 | unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE; | |
244 | ||
245 | if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR) | |
246 | return false; | |
247 | ||
248 | if (fsc_type == ESR_ELx_FSC_PERM) | |
249 | return true; | |
250 | ||
251 | if (addr < USER_DS && system_uses_ttbr0_pan()) | |
252 | return fsc_type == ESR_ELx_FSC_FAULT && | |
253 | (regs->pstate & PSR_PAN_BIT); | |
254 | ||
255 | return false; | |
256 | } | |
257 | ||
1d18c47c CM |
258 | /* |
259 | * The kernel tried to access some page that wasn't present. | |
260 | */ | |
67ce16ec KM |
261 | static void __do_kernel_fault(unsigned long addr, unsigned int esr, |
262 | struct pt_regs *regs) | |
1d18c47c | 263 | { |
b824b930 SB |
264 | const char *msg; |
265 | ||
1d18c47c CM |
266 | /* |
267 | * Are we prepared to handle this kernel fault? | |
9adeb8e7 | 268 | * We are almost certainly not prepared to handle instruction faults. |
1d18c47c | 269 | */ |
9adeb8e7 | 270 | if (!is_el1_instruction_abort(esr) && fixup_exception(regs)) |
1d18c47c CM |
271 | return; |
272 | ||
273 | /* | |
274 | * No handler, we'll have to terminate things with extreme prejudice. | |
275 | */ | |
276 | bust_spinlocks(1); | |
b824b930 SB |
277 | |
278 | if (is_permission_fault(esr, regs, addr)) { | |
279 | if (esr & ESR_ELx_WNR) | |
280 | msg = "write to read-only memory"; | |
281 | else | |
282 | msg = "read from unreadable memory"; | |
283 | } else if (addr < PAGE_SIZE) { | |
284 | msg = "NULL pointer dereference"; | |
285 | } else { | |
286 | msg = "paging request"; | |
287 | } | |
288 | ||
289 | pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg, | |
290 | addr); | |
1d18c47c | 291 | |
1f9b8936 JT |
292 | mem_abort_decode(esr); |
293 | ||
67ce16ec | 294 | show_pte(addr); |
1d18c47c CM |
295 | die("Oops", regs, esr); |
296 | bust_spinlocks(0); | |
297 | do_exit(SIGKILL); | |
298 | } | |
299 | ||
300 | /* | |
301 | * Something tried to access memory that isn't in our memory map. User mode | |
302 | * accesses just cause a SIGSEGV | |
303 | */ | |
304 | static void __do_user_fault(struct task_struct *tsk, unsigned long addr, | |
305 | unsigned int esr, unsigned int sig, int code, | |
e7c600f1 | 306 | struct pt_regs *regs, int fault) |
1d18c47c CM |
307 | { |
308 | struct siginfo si; | |
09a6adf5 | 309 | const struct fault_info *inf; |
e7c600f1 | 310 | unsigned int lsb = 0; |
1d18c47c | 311 | |
f871d268 | 312 | if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) { |
09a6adf5 | 313 | inf = esr_to_fault_info(esr); |
83016b20 | 314 | pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x", |
09a6adf5 | 315 | tsk->comm, task_pid_nr(tsk), inf->name, sig, |
3495386b | 316 | addr, esr); |
83016b20 KM |
317 | print_vma_addr(KERN_CONT ", in ", regs->pc); |
318 | pr_cont("\n"); | |
c07ab957 | 319 | __show_regs(regs); |
1d18c47c CM |
320 | } |
321 | ||
322 | tsk->thread.fault_address = addr; | |
9141300a | 323 | tsk->thread.fault_code = esr; |
1d18c47c CM |
324 | si.si_signo = sig; |
325 | si.si_errno = 0; | |
326 | si.si_code = code; | |
327 | si.si_addr = (void __user *)addr; | |
e7c600f1 JZZ |
328 | /* |
329 | * Either small page or large page may be poisoned. | |
330 | * In other words, VM_FAULT_HWPOISON_LARGE and | |
331 | * VM_FAULT_HWPOISON are mutually exclusive. | |
332 | */ | |
333 | if (fault & VM_FAULT_HWPOISON_LARGE) | |
334 | lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); | |
335 | else if (fault & VM_FAULT_HWPOISON) | |
336 | lsb = PAGE_SHIFT; | |
337 | si.si_addr_lsb = lsb; | |
338 | ||
1d18c47c CM |
339 | force_sig_info(sig, &si, tsk); |
340 | } | |
341 | ||
59f67e16 | 342 | static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs) |
1d18c47c CM |
343 | { |
344 | struct task_struct *tsk = current; | |
09a6adf5 | 345 | const struct fault_info *inf; |
1d18c47c CM |
346 | |
347 | /* | |
348 | * If we are in kernel mode at this point, we have no context to | |
349 | * handle this fault with. | |
350 | */ | |
09a6adf5 VK |
351 | if (user_mode(regs)) { |
352 | inf = esr_to_fault_info(esr); | |
e7c600f1 | 353 | __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0); |
09a6adf5 | 354 | } else |
67ce16ec | 355 | __do_kernel_fault(addr, esr, regs); |
1d18c47c CM |
356 | } |
357 | ||
358 | #define VM_FAULT_BADMAP 0x010000 | |
359 | #define VM_FAULT_BADACCESS 0x020000 | |
360 | ||
1d18c47c | 361 | static int __do_page_fault(struct mm_struct *mm, unsigned long addr, |
db6f4106 | 362 | unsigned int mm_flags, unsigned long vm_flags, |
1d18c47c CM |
363 | struct task_struct *tsk) |
364 | { | |
365 | struct vm_area_struct *vma; | |
366 | int fault; | |
367 | ||
368 | vma = find_vma(mm, addr); | |
369 | fault = VM_FAULT_BADMAP; | |
370 | if (unlikely(!vma)) | |
371 | goto out; | |
372 | if (unlikely(vma->vm_start > addr)) | |
373 | goto check_stack; | |
374 | ||
375 | /* | |
376 | * Ok, we have a good vm_area for this memory access, so we can handle | |
377 | * it. | |
378 | */ | |
379 | good_area: | |
db6f4106 WD |
380 | /* |
381 | * Check that the permissions on the VMA allow for the fault which | |
cab15ce6 | 382 | * occurred. |
db6f4106 WD |
383 | */ |
384 | if (!(vma->vm_flags & vm_flags)) { | |
1d18c47c CM |
385 | fault = VM_FAULT_BADACCESS; |
386 | goto out; | |
387 | } | |
388 | ||
dcddffd4 | 389 | return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags); |
1d18c47c CM |
390 | |
391 | check_stack: | |
392 | if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr)) | |
393 | goto good_area; | |
394 | out: | |
395 | return fault; | |
396 | } | |
397 | ||
541ec870 MR |
398 | static bool is_el0_instruction_abort(unsigned int esr) |
399 | { | |
400 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW; | |
401 | } | |
402 | ||
1d18c47c CM |
403 | static int __kprobes do_page_fault(unsigned long addr, unsigned int esr, |
404 | struct pt_regs *regs) | |
405 | { | |
406 | struct task_struct *tsk; | |
407 | struct mm_struct *mm; | |
0e3a9026 | 408 | int fault, sig, code, major = 0; |
cab15ce6 | 409 | unsigned long vm_flags = VM_READ | VM_WRITE; |
db6f4106 WD |
410 | unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; |
411 | ||
2dd0e8d2 SP |
412 | if (notify_page_fault(regs, esr)) |
413 | return 0; | |
414 | ||
1d18c47c CM |
415 | tsk = current; |
416 | mm = tsk->mm; | |
417 | ||
1d18c47c CM |
418 | /* |
419 | * If we're in an interrupt or have no user context, we must not take | |
420 | * the fault. | |
421 | */ | |
70ffdb93 | 422 | if (faulthandler_disabled() || !mm) |
1d18c47c CM |
423 | goto no_context; |
424 | ||
759496ba JW |
425 | if (user_mode(regs)) |
426 | mm_flags |= FAULT_FLAG_USER; | |
427 | ||
541ec870 | 428 | if (is_el0_instruction_abort(esr)) { |
759496ba | 429 | vm_flags = VM_EXEC; |
aed40e01 | 430 | } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) { |
759496ba JW |
431 | vm_flags = VM_WRITE; |
432 | mm_flags |= FAULT_FLAG_WRITE; | |
433 | } | |
434 | ||
b824b930 | 435 | if (addr < USER_DS && is_permission_fault(esr, regs, addr)) { |
e19a6ee2 JM |
436 | /* regs->orig_addr_limit may be 0 if we entered from EL0 */ |
437 | if (regs->orig_addr_limit == KERNEL_DS) | |
70c8abc2 | 438 | die("Accessing user space memory with fs=KERNEL_DS", regs, esr); |
70544196 | 439 | |
9adeb8e7 LA |
440 | if (is_el1_instruction_abort(esr)) |
441 | die("Attempting to execute userspace memory", regs, esr); | |
442 | ||
57f4959b | 443 | if (!search_exception_tables(regs->pc)) |
70c8abc2 | 444 | die("Accessing user space memory outside uaccess.h routines", regs, esr); |
57f4959b | 445 | } |
338d4f49 | 446 | |
0e3a9026 PA |
447 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); |
448 | ||
1d18c47c CM |
449 | /* |
450 | * As per x86, we may deadlock here. However, since the kernel only | |
451 | * validly references user space from well defined areas of the code, | |
452 | * we can bug out early if this is from code which shouldn't. | |
453 | */ | |
454 | if (!down_read_trylock(&mm->mmap_sem)) { | |
455 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
456 | goto no_context; | |
457 | retry: | |
458 | down_read(&mm->mmap_sem); | |
459 | } else { | |
460 | /* | |
461 | * The above down_read_trylock() might have succeeded in which | |
462 | * case, we'll have missed the might_sleep() from down_read(). | |
463 | */ | |
464 | might_sleep(); | |
465 | #ifdef CONFIG_DEBUG_VM | |
466 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
467 | goto no_context; | |
468 | #endif | |
469 | } | |
470 | ||
db6f4106 | 471 | fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk); |
0e3a9026 | 472 | major |= fault & VM_FAULT_MAJOR; |
1d18c47c | 473 | |
0e3a9026 PA |
474 | if (fault & VM_FAULT_RETRY) { |
475 | /* | |
476 | * If we need to retry but a fatal signal is pending, | |
477 | * handle the signal first. We do not need to release | |
478 | * the mmap_sem because it would already be released | |
479 | * in __lock_page_or_retry in mm/filemap.c. | |
480 | */ | |
481 | if (fatal_signal_pending(current)) | |
482 | return 0; | |
483 | ||
484 | /* | |
485 | * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of | |
486 | * starvation. | |
487 | */ | |
488 | if (mm_flags & FAULT_FLAG_ALLOW_RETRY) { | |
489 | mm_flags &= ~FAULT_FLAG_ALLOW_RETRY; | |
490 | mm_flags |= FAULT_FLAG_TRIED; | |
491 | goto retry; | |
492 | } | |
493 | } | |
494 | up_read(&mm->mmap_sem); | |
1d18c47c CM |
495 | |
496 | /* | |
0e3a9026 | 497 | * Handle the "normal" (no error) case first. |
1d18c47c | 498 | */ |
0e3a9026 PA |
499 | if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP | |
500 | VM_FAULT_BADACCESS)))) { | |
501 | /* | |
502 | * Major/minor page fault accounting is only done | |
503 | * once. If we go through a retry, it is extremely | |
504 | * likely that the page will be found in page cache at | |
505 | * that point. | |
506 | */ | |
507 | if (major) { | |
1d18c47c CM |
508 | tsk->maj_flt++; |
509 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, | |
510 | addr); | |
511 | } else { | |
512 | tsk->min_flt++; | |
513 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, | |
514 | addr); | |
515 | } | |
1d18c47c | 516 | |
1d18c47c | 517 | return 0; |
0e3a9026 | 518 | } |
1d18c47c | 519 | |
87134102 JW |
520 | /* |
521 | * If we are in kernel mode at this point, we have no context to | |
522 | * handle this fault with. | |
523 | */ | |
524 | if (!user_mode(regs)) | |
525 | goto no_context; | |
526 | ||
1d18c47c CM |
527 | if (fault & VM_FAULT_OOM) { |
528 | /* | |
529 | * We ran out of memory, call the OOM killer, and return to | |
530 | * userspace (which will retry the fault, or kill us if we got | |
531 | * oom-killed). | |
532 | */ | |
533 | pagefault_out_of_memory(); | |
534 | return 0; | |
535 | } | |
536 | ||
1d18c47c CM |
537 | if (fault & VM_FAULT_SIGBUS) { |
538 | /* | |
539 | * We had some memory, but were unable to successfully fix up | |
540 | * this page fault. | |
541 | */ | |
542 | sig = SIGBUS; | |
543 | code = BUS_ADRERR; | |
e7c600f1 JZZ |
544 | } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) { |
545 | sig = SIGBUS; | |
546 | code = BUS_MCEERR_AR; | |
1d18c47c CM |
547 | } else { |
548 | /* | |
549 | * Something tried to access memory that isn't in our memory | |
550 | * map. | |
551 | */ | |
552 | sig = SIGSEGV; | |
553 | code = fault == VM_FAULT_BADACCESS ? | |
554 | SEGV_ACCERR : SEGV_MAPERR; | |
555 | } | |
556 | ||
e7c600f1 | 557 | __do_user_fault(tsk, addr, esr, sig, code, regs, fault); |
1d18c47c CM |
558 | return 0; |
559 | ||
560 | no_context: | |
67ce16ec | 561 | __do_kernel_fault(addr, esr, regs); |
1d18c47c CM |
562 | return 0; |
563 | } | |
564 | ||
565 | /* | |
566 | * First Level Translation Fault Handler | |
567 | * | |
568 | * We enter here because the first level page table doesn't contain a valid | |
569 | * entry for the address. | |
570 | * | |
571 | * If the address is in kernel space (>= TASK_SIZE), then we are probably | |
572 | * faulting in the vmalloc() area. | |
573 | * | |
574 | * If the init_task's first level page tables contains the relevant entry, we | |
575 | * copy the it to this task. If not, we send the process a signal, fixup the | |
576 | * exception, or oops the kernel. | |
577 | * | |
578 | * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt | |
579 | * or a critical region, and should only copy the information from the master | |
580 | * page table, nothing more. | |
581 | */ | |
582 | static int __kprobes do_translation_fault(unsigned long addr, | |
583 | unsigned int esr, | |
584 | struct pt_regs *regs) | |
585 | { | |
586 | if (addr < TASK_SIZE) | |
587 | return do_page_fault(addr, esr, regs); | |
588 | ||
589 | do_bad_area(addr, esr, regs); | |
590 | return 0; | |
591 | } | |
592 | ||
52d7523d EL |
593 | static int do_alignment_fault(unsigned long addr, unsigned int esr, |
594 | struct pt_regs *regs) | |
595 | { | |
596 | do_bad_area(addr, esr, regs); | |
597 | return 0; | |
598 | } | |
599 | ||
1d18c47c CM |
600 | /* |
601 | * This abort handler always returns "fault". | |
602 | */ | |
603 | static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
604 | { | |
605 | return 1; | |
606 | } | |
607 | ||
32015c23 TB |
608 | /* |
609 | * This abort handler deals with Synchronous External Abort. | |
610 | * It calls notifiers, and then returns "fault". | |
611 | */ | |
612 | static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
613 | { | |
614 | struct siginfo info; | |
615 | const struct fault_info *inf; | |
621f48e4 | 616 | int ret = 0; |
32015c23 TB |
617 | |
618 | inf = esr_to_fault_info(esr); | |
619 | pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n", | |
620 | inf->name, esr, addr); | |
621 | ||
7edda088 TB |
622 | /* |
623 | * Synchronous aborts may interrupt code which had interrupts masked. | |
624 | * Before calling out into the wider kernel tell the interested | |
625 | * subsystems. | |
626 | */ | |
627 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) { | |
628 | if (interrupts_enabled(regs)) | |
629 | nmi_enter(); | |
630 | ||
621f48e4 | 631 | ret = ghes_notify_sea(); |
7edda088 TB |
632 | |
633 | if (interrupts_enabled(regs)) | |
634 | nmi_exit(); | |
635 | } | |
636 | ||
32015c23 TB |
637 | info.si_signo = SIGBUS; |
638 | info.si_errno = 0; | |
639 | info.si_code = 0; | |
640 | if (esr & ESR_ELx_FnV) | |
641 | info.si_addr = NULL; | |
642 | else | |
643 | info.si_addr = (void __user *)addr; | |
644 | arm64_notify_die("", regs, &info, esr); | |
645 | ||
621f48e4 | 646 | return ret; |
32015c23 TB |
647 | } |
648 | ||
09a6adf5 | 649 | static const struct fault_info fault_info[] = { |
1d18c47c CM |
650 | { do_bad, SIGBUS, 0, "ttbr address size fault" }, |
651 | { do_bad, SIGBUS, 0, "level 1 address size fault" }, | |
652 | { do_bad, SIGBUS, 0, "level 2 address size fault" }, | |
653 | { do_bad, SIGBUS, 0, "level 3 address size fault" }, | |
7f73f7ae | 654 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" }, |
1d18c47c CM |
655 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" }, |
656 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" }, | |
657 | { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" }, | |
c03784ee | 658 | { do_bad, SIGBUS, 0, "unknown 8" }, |
084bd298 SC |
659 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" }, |
660 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" }, | |
1d18c47c | 661 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" }, |
c03784ee | 662 | { do_bad, SIGBUS, 0, "unknown 12" }, |
084bd298 SC |
663 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" }, |
664 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" }, | |
1d18c47c | 665 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" }, |
32015c23 | 666 | { do_sea, SIGBUS, 0, "synchronous external abort" }, |
c03784ee | 667 | { do_bad, SIGBUS, 0, "unknown 17" }, |
1d18c47c CM |
668 | { do_bad, SIGBUS, 0, "unknown 18" }, |
669 | { do_bad, SIGBUS, 0, "unknown 19" }, | |
32015c23 TB |
670 | { do_sea, SIGBUS, 0, "level 0 (translation table walk)" }, |
671 | { do_sea, SIGBUS, 0, "level 1 (translation table walk)" }, | |
672 | { do_sea, SIGBUS, 0, "level 2 (translation table walk)" }, | |
673 | { do_sea, SIGBUS, 0, "level 3 (translation table walk)" }, | |
674 | { do_sea, SIGBUS, 0, "synchronous parity or ECC error" }, | |
c03784ee | 675 | { do_bad, SIGBUS, 0, "unknown 25" }, |
1d18c47c CM |
676 | { do_bad, SIGBUS, 0, "unknown 26" }, |
677 | { do_bad, SIGBUS, 0, "unknown 27" }, | |
32015c23 TB |
678 | { do_sea, SIGBUS, 0, "level 0 synchronous parity error (translation table walk)" }, |
679 | { do_sea, SIGBUS, 0, "level 1 synchronous parity error (translation table walk)" }, | |
680 | { do_sea, SIGBUS, 0, "level 2 synchronous parity error (translation table walk)" }, | |
681 | { do_sea, SIGBUS, 0, "level 3 synchronous parity error (translation table walk)" }, | |
1d18c47c | 682 | { do_bad, SIGBUS, 0, "unknown 32" }, |
52d7523d | 683 | { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" }, |
c03784ee | 684 | { do_bad, SIGBUS, 0, "unknown 34" }, |
1d18c47c CM |
685 | { do_bad, SIGBUS, 0, "unknown 35" }, |
686 | { do_bad, SIGBUS, 0, "unknown 36" }, | |
687 | { do_bad, SIGBUS, 0, "unknown 37" }, | |
688 | { do_bad, SIGBUS, 0, "unknown 38" }, | |
689 | { do_bad, SIGBUS, 0, "unknown 39" }, | |
690 | { do_bad, SIGBUS, 0, "unknown 40" }, | |
691 | { do_bad, SIGBUS, 0, "unknown 41" }, | |
692 | { do_bad, SIGBUS, 0, "unknown 42" }, | |
693 | { do_bad, SIGBUS, 0, "unknown 43" }, | |
694 | { do_bad, SIGBUS, 0, "unknown 44" }, | |
695 | { do_bad, SIGBUS, 0, "unknown 45" }, | |
696 | { do_bad, SIGBUS, 0, "unknown 46" }, | |
697 | { do_bad, SIGBUS, 0, "unknown 47" }, | |
c03784ee | 698 | { do_bad, SIGBUS, 0, "TLB conflict abort" }, |
1d18c47c CM |
699 | { do_bad, SIGBUS, 0, "unknown 49" }, |
700 | { do_bad, SIGBUS, 0, "unknown 50" }, | |
701 | { do_bad, SIGBUS, 0, "unknown 51" }, | |
702 | { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" }, | |
c03784ee | 703 | { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" }, |
1d18c47c CM |
704 | { do_bad, SIGBUS, 0, "unknown 54" }, |
705 | { do_bad, SIGBUS, 0, "unknown 55" }, | |
706 | { do_bad, SIGBUS, 0, "unknown 56" }, | |
707 | { do_bad, SIGBUS, 0, "unknown 57" }, | |
c03784ee | 708 | { do_bad, SIGBUS, 0, "unknown 58" }, |
1d18c47c CM |
709 | { do_bad, SIGBUS, 0, "unknown 59" }, |
710 | { do_bad, SIGBUS, 0, "unknown 60" }, | |
c03784ee MR |
711 | { do_bad, SIGBUS, 0, "section domain fault" }, |
712 | { do_bad, SIGBUS, 0, "page domain fault" }, | |
1d18c47c CM |
713 | { do_bad, SIGBUS, 0, "unknown 63" }, |
714 | }; | |
715 | ||
621f48e4 TB |
716 | /* |
717 | * Handle Synchronous External Aborts that occur in a guest kernel. | |
718 | * | |
719 | * The return value will be zero if the SEA was successfully handled | |
720 | * and non-zero if there was an error processing the error or there was | |
721 | * no error to process. | |
722 | */ | |
723 | int handle_guest_sea(phys_addr_t addr, unsigned int esr) | |
724 | { | |
725 | int ret = -ENOENT; | |
726 | ||
727 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) | |
728 | ret = ghes_notify_sea(); | |
729 | ||
730 | return ret; | |
731 | } | |
732 | ||
1d18c47c CM |
733 | /* |
734 | * Dispatch a data abort to the relevant handler. | |
735 | */ | |
736 | asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr, | |
737 | struct pt_regs *regs) | |
738 | { | |
09a6adf5 | 739 | const struct fault_info *inf = esr_to_fault_info(esr); |
1d18c47c CM |
740 | struct siginfo info; |
741 | ||
742 | if (!inf->fn(addr, esr, regs)) | |
743 | return; | |
744 | ||
745 | pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n", | |
746 | inf->name, esr, addr); | |
747 | ||
1f9b8936 JT |
748 | mem_abort_decode(esr); |
749 | ||
1d18c47c CM |
750 | info.si_signo = inf->sig; |
751 | info.si_errno = 0; | |
752 | info.si_code = inf->code; | |
753 | info.si_addr = (void __user *)addr; | |
754 | arm64_notify_die("", regs, &info, esr); | |
755 | } | |
756 | ||
757 | /* | |
758 | * Handle stack alignment exceptions. | |
759 | */ | |
760 | asmlinkage void __exception do_sp_pc_abort(unsigned long addr, | |
761 | unsigned int esr, | |
762 | struct pt_regs *regs) | |
763 | { | |
764 | struct siginfo info; | |
9e793ab8 VM |
765 | struct task_struct *tsk = current; |
766 | ||
767 | if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS)) | |
768 | pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n", | |
769 | tsk->comm, task_pid_nr(tsk), | |
770 | esr_get_class_string(esr), (void *)regs->pc, | |
771 | (void *)regs->sp); | |
1d18c47c CM |
772 | |
773 | info.si_signo = SIGBUS; | |
774 | info.si_errno = 0; | |
775 | info.si_code = BUS_ADRALN; | |
776 | info.si_addr = (void __user *)addr; | |
9e793ab8 | 777 | arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr); |
1d18c47c CM |
778 | } |
779 | ||
9fb7410f DM |
780 | int __init early_brk64(unsigned long addr, unsigned int esr, |
781 | struct pt_regs *regs); | |
782 | ||
783 | /* | |
784 | * __refdata because early_brk64 is __init, but the reference to it is | |
785 | * clobbered at arch_initcall time. | |
786 | * See traps.c and debug-monitors.c:debug_traps_init(). | |
787 | */ | |
788 | static struct fault_info __refdata debug_fault_info[] = { | |
1d18c47c CM |
789 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" }, |
790 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" }, | |
791 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" }, | |
792 | { do_bad, SIGBUS, 0, "unknown 3" }, | |
793 | { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" }, | |
794 | { do_bad, SIGTRAP, 0, "aarch32 vector catch" }, | |
9fb7410f | 795 | { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" }, |
1d18c47c CM |
796 | { do_bad, SIGBUS, 0, "unknown 7" }, |
797 | }; | |
798 | ||
799 | void __init hook_debug_fault_code(int nr, | |
800 | int (*fn)(unsigned long, unsigned int, struct pt_regs *), | |
801 | int sig, int code, const char *name) | |
802 | { | |
803 | BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info)); | |
804 | ||
805 | debug_fault_info[nr].fn = fn; | |
806 | debug_fault_info[nr].sig = sig; | |
807 | debug_fault_info[nr].code = code; | |
808 | debug_fault_info[nr].name = name; | |
809 | } | |
810 | ||
811 | asmlinkage int __exception do_debug_exception(unsigned long addr, | |
812 | unsigned int esr, | |
813 | struct pt_regs *regs) | |
814 | { | |
815 | const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr); | |
816 | struct siginfo info; | |
6afedcd2 | 817 | int rv; |
1d18c47c | 818 | |
6afedcd2 JM |
819 | /* |
820 | * Tell lockdep we disabled irqs in entry.S. Do nothing if they were | |
821 | * already disabled to preserve the last enabled/disabled addresses. | |
822 | */ | |
823 | if (interrupts_enabled(regs)) | |
824 | trace_hardirqs_off(); | |
1d18c47c | 825 | |
6afedcd2 JM |
826 | if (!inf->fn(addr, esr, regs)) { |
827 | rv = 1; | |
828 | } else { | |
829 | pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n", | |
830 | inf->name, esr, addr); | |
831 | ||
832 | info.si_signo = inf->sig; | |
833 | info.si_errno = 0; | |
834 | info.si_code = inf->code; | |
835 | info.si_addr = (void __user *)addr; | |
836 | arm64_notify_die("", regs, &info, 0); | |
837 | rv = 0; | |
838 | } | |
1d18c47c | 839 | |
6afedcd2 JM |
840 | if (interrupts_enabled(regs)) |
841 | trace_hardirqs_on(); | |
1d18c47c | 842 | |
6afedcd2 | 843 | return rv; |
1d18c47c | 844 | } |
2dd0e8d2 | 845 | NOKPROBE_SYMBOL(do_debug_exception); |
338d4f49 JM |
846 | |
847 | #ifdef CONFIG_ARM64_PAN | |
2a6dcb2b | 848 | int cpu_enable_pan(void *__unused) |
338d4f49 | 849 | { |
7209c868 JM |
850 | /* |
851 | * We modify PSTATE. This won't work from irq context as the PSTATE | |
852 | * is discarded once we return from the exception. | |
853 | */ | |
854 | WARN_ON_ONCE(in_interrupt()); | |
855 | ||
338d4f49 | 856 | config_sctlr_el1(SCTLR_EL1_SPAN, 0); |
7209c868 | 857 | asm(SET_PSTATE_PAN(1)); |
2a6dcb2b | 858 | return 0; |
338d4f49 JM |
859 | } |
860 | #endif /* CONFIG_ARM64_PAN */ |