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> |
338d4f49 | 37 | #include <asm/cpufeature.h> |
1d18c47c CM |
38 | #include <asm/exception.h> |
39 | #include <asm/debug-monitors.h> | |
9141300a | 40 | #include <asm/esr.h> |
338d4f49 | 41 | #include <asm/sysreg.h> |
1d18c47c CM |
42 | #include <asm/system_misc.h> |
43 | #include <asm/pgtable.h> | |
44 | #include <asm/tlbflush.h> | |
45 | ||
7edda088 TB |
46 | #include <acpi/ghes.h> |
47 | ||
09a6adf5 VK |
48 | struct fault_info { |
49 | int (*fn)(unsigned long addr, unsigned int esr, | |
50 | struct pt_regs *regs); | |
51 | int sig; | |
52 | int code; | |
53 | const char *name; | |
54 | }; | |
55 | ||
56 | static const struct fault_info fault_info[]; | |
57 | ||
58 | static inline const struct fault_info *esr_to_fault_info(unsigned int esr) | |
59 | { | |
60 | return fault_info + (esr & 63); | |
61 | } | |
3495386b | 62 | |
2dd0e8d2 SP |
63 | #ifdef CONFIG_KPROBES |
64 | static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr) | |
65 | { | |
66 | int ret = 0; | |
67 | ||
68 | /* kprobe_running() needs smp_processor_id() */ | |
69 | if (!user_mode(regs)) { | |
70 | preempt_disable(); | |
71 | if (kprobe_running() && kprobe_fault_handler(regs, esr)) | |
72 | ret = 1; | |
73 | preempt_enable(); | |
74 | } | |
75 | ||
76 | return ret; | |
77 | } | |
78 | #else | |
79 | static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr) | |
80 | { | |
81 | return 0; | |
82 | } | |
83 | #endif | |
84 | ||
1f9b8936 JT |
85 | static void data_abort_decode(unsigned int esr) |
86 | { | |
87 | pr_alert("Data abort info:\n"); | |
88 | ||
89 | if (esr & ESR_ELx_ISV) { | |
90 | pr_alert(" Access size = %u byte(s)\n", | |
91 | 1U << ((esr & ESR_ELx_SAS) >> ESR_ELx_SAS_SHIFT)); | |
92 | pr_alert(" SSE = %lu, SRT = %lu\n", | |
93 | (esr & ESR_ELx_SSE) >> ESR_ELx_SSE_SHIFT, | |
94 | (esr & ESR_ELx_SRT_MASK) >> ESR_ELx_SRT_SHIFT); | |
95 | pr_alert(" SF = %lu, AR = %lu\n", | |
96 | (esr & ESR_ELx_SF) >> ESR_ELx_SF_SHIFT, | |
97 | (esr & ESR_ELx_AR) >> ESR_ELx_AR_SHIFT); | |
98 | } else { | |
99 | pr_alert(" ISV = 0, ISS = 0x%08lu\n", esr & ESR_ELx_ISS_MASK); | |
100 | } | |
101 | ||
102 | pr_alert(" CM = %lu, WnR = %lu\n", | |
103 | (esr & ESR_ELx_CM) >> ESR_ELx_CM_SHIFT, | |
104 | (esr & ESR_ELx_WNR) >> ESR_ELx_WNR_SHIFT); | |
105 | } | |
106 | ||
107 | /* | |
108 | * Decode mem abort information | |
109 | */ | |
110 | static void mem_abort_decode(unsigned int esr) | |
111 | { | |
112 | pr_alert("Mem abort info:\n"); | |
113 | ||
114 | pr_alert(" Exception class = %s, IL = %u bits\n", | |
115 | esr_get_class_string(esr), | |
116 | (esr & ESR_ELx_IL) ? 32 : 16); | |
117 | pr_alert(" SET = %lu, FnV = %lu\n", | |
118 | (esr & ESR_ELx_SET_MASK) >> ESR_ELx_SET_SHIFT, | |
119 | (esr & ESR_ELx_FnV) >> ESR_ELx_FnV_SHIFT); | |
120 | pr_alert(" EA = %lu, S1PTW = %lu\n", | |
121 | (esr & ESR_ELx_EA) >> ESR_ELx_EA_SHIFT, | |
122 | (esr & ESR_ELx_S1PTW) >> ESR_ELx_S1PTW_SHIFT); | |
123 | ||
124 | if (esr_is_data_abort(esr)) | |
125 | data_abort_decode(esr); | |
126 | } | |
127 | ||
1d18c47c | 128 | /* |
67ce16ec | 129 | * Dump out the page tables associated with 'addr' in the currently active mm. |
1d18c47c | 130 | */ |
67ce16ec | 131 | void show_pte(unsigned long addr) |
1d18c47c | 132 | { |
67ce16ec | 133 | struct mm_struct *mm; |
1d18c47c CM |
134 | pgd_t *pgd; |
135 | ||
67ce16ec KM |
136 | if (addr < TASK_SIZE) { |
137 | /* TTBR0 */ | |
138 | mm = current->active_mm; | |
139 | if (mm == &init_mm) { | |
140 | pr_alert("[%016lx] user address but active_mm is swapper\n", | |
141 | addr); | |
142 | return; | |
143 | } | |
144 | } else if (addr >= VA_START) { | |
145 | /* TTBR1 */ | |
1d18c47c | 146 | mm = &init_mm; |
67ce16ec KM |
147 | } else { |
148 | pr_alert("[%016lx] address between user and kernel address ranges\n", | |
149 | addr); | |
150 | return; | |
151 | } | |
1d18c47c | 152 | |
1eb34b6e WD |
153 | pr_alert("%s pgtable: %luk pages, %u-bit VAs, pgd = %p\n", |
154 | mm == &init_mm ? "swapper" : "user", PAGE_SIZE / SZ_1K, | |
155 | VA_BITS, mm->pgd); | |
1d18c47c | 156 | pgd = pgd_offset(mm, addr); |
67ce16ec | 157 | pr_alert("[%016lx] *pgd=%016llx", addr, pgd_val(*pgd)); |
1d18c47c CM |
158 | |
159 | do { | |
160 | pud_t *pud; | |
161 | pmd_t *pmd; | |
162 | pte_t *pte; | |
163 | ||
4339e3f3 | 164 | if (pgd_none(*pgd) || pgd_bad(*pgd)) |
1d18c47c CM |
165 | break; |
166 | ||
167 | pud = pud_offset(pgd, addr); | |
6ef4fb38 | 168 | pr_cont(", *pud=%016llx", pud_val(*pud)); |
4339e3f3 | 169 | if (pud_none(*pud) || pud_bad(*pud)) |
1d18c47c CM |
170 | break; |
171 | ||
172 | pmd = pmd_offset(pud, addr); | |
6ef4fb38 | 173 | pr_cont(", *pmd=%016llx", pmd_val(*pmd)); |
4339e3f3 | 174 | if (pmd_none(*pmd) || pmd_bad(*pmd)) |
1d18c47c CM |
175 | break; |
176 | ||
177 | pte = pte_offset_map(pmd, addr); | |
6ef4fb38 | 178 | pr_cont(", *pte=%016llx", pte_val(*pte)); |
1d18c47c CM |
179 | pte_unmap(pte); |
180 | } while(0); | |
181 | ||
6ef4fb38 | 182 | pr_cont("\n"); |
1d18c47c CM |
183 | } |
184 | ||
66dbd6e6 CM |
185 | #ifdef CONFIG_ARM64_HW_AFDBM |
186 | /* | |
187 | * This function sets the access flags (dirty, accessed), as well as write | |
188 | * permission, and only to a more permissive setting. | |
189 | * | |
190 | * It needs to cope with hardware update of the accessed/dirty state by other | |
191 | * agents in the system and can safely skip the __sync_icache_dcache() call as, | |
192 | * like set_pte_at(), the PTE is never changed from no-exec to exec here. | |
193 | * | |
194 | * Returns whether or not the PTE actually changed. | |
195 | */ | |
196 | int ptep_set_access_flags(struct vm_area_struct *vma, | |
197 | unsigned long address, pte_t *ptep, | |
198 | pte_t entry, int dirty) | |
199 | { | |
200 | pteval_t old_pteval; | |
201 | unsigned int tmp; | |
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)) |
66dbd6e6 CM |
211 | pte_val(entry) |= PTE_RDONLY; |
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; |
66dbd6e6 CM |
220 | asm volatile("// ptep_set_access_flags\n" |
221 | " prfm pstl1strm, %2\n" | |
222 | "1: ldxr %0, %2\n" | |
6d332747 | 223 | " eor %0, %0, %3 // negate PTE_RDONLY in *ptep\n" |
66dbd6e6 | 224 | " orr %0, %0, %4 // set flags\n" |
6d332747 | 225 | " eor %0, %0, %3 // negate final PTE_RDONLY\n" |
66dbd6e6 CM |
226 | " stxr %w1, %0, %2\n" |
227 | " cbnz %w1, 1b\n" | |
228 | : "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep)) | |
6d332747 | 229 | : "L" (PTE_RDONLY), "r" (pte_val(entry))); |
66dbd6e6 CM |
230 | |
231 | flush_tlb_fix_spurious_fault(vma, address); | |
232 | return 1; | |
233 | } | |
234 | #endif | |
235 | ||
9adeb8e7 LA |
236 | static bool is_el1_instruction_abort(unsigned int esr) |
237 | { | |
238 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR; | |
239 | } | |
240 | ||
b824b930 SB |
241 | static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs, |
242 | unsigned long addr) | |
243 | { | |
244 | unsigned int ec = ESR_ELx_EC(esr); | |
245 | unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE; | |
246 | ||
247 | if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR) | |
248 | return false; | |
249 | ||
250 | if (fsc_type == ESR_ELx_FSC_PERM) | |
251 | return true; | |
252 | ||
253 | if (addr < USER_DS && system_uses_ttbr0_pan()) | |
254 | return fsc_type == ESR_ELx_FSC_FAULT && | |
255 | (regs->pstate & PSR_PAN_BIT); | |
256 | ||
257 | return false; | |
258 | } | |
259 | ||
1d18c47c CM |
260 | /* |
261 | * The kernel tried to access some page that wasn't present. | |
262 | */ | |
67ce16ec KM |
263 | static void __do_kernel_fault(unsigned long addr, unsigned int esr, |
264 | struct pt_regs *regs) | |
1d18c47c | 265 | { |
b824b930 SB |
266 | const char *msg; |
267 | ||
1d18c47c CM |
268 | /* |
269 | * Are we prepared to handle this kernel fault? | |
9adeb8e7 | 270 | * We are almost certainly not prepared to handle instruction faults. |
1d18c47c | 271 | */ |
9adeb8e7 | 272 | if (!is_el1_instruction_abort(esr) && fixup_exception(regs)) |
1d18c47c CM |
273 | return; |
274 | ||
275 | /* | |
276 | * No handler, we'll have to terminate things with extreme prejudice. | |
277 | */ | |
278 | bust_spinlocks(1); | |
b824b930 SB |
279 | |
280 | if (is_permission_fault(esr, regs, addr)) { | |
281 | if (esr & ESR_ELx_WNR) | |
282 | msg = "write to read-only memory"; | |
283 | else | |
284 | msg = "read from unreadable memory"; | |
285 | } else if (addr < PAGE_SIZE) { | |
286 | msg = "NULL pointer dereference"; | |
287 | } else { | |
288 | msg = "paging request"; | |
289 | } | |
290 | ||
291 | pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg, | |
292 | addr); | |
1d18c47c | 293 | |
1f9b8936 JT |
294 | mem_abort_decode(esr); |
295 | ||
67ce16ec | 296 | show_pte(addr); |
1d18c47c CM |
297 | die("Oops", regs, esr); |
298 | bust_spinlocks(0); | |
299 | do_exit(SIGKILL); | |
300 | } | |
301 | ||
302 | /* | |
303 | * Something tried to access memory that isn't in our memory map. User mode | |
304 | * accesses just cause a SIGSEGV | |
305 | */ | |
306 | static void __do_user_fault(struct task_struct *tsk, unsigned long addr, | |
307 | unsigned int esr, unsigned int sig, int code, | |
e7c600f1 | 308 | struct pt_regs *regs, int fault) |
1d18c47c CM |
309 | { |
310 | struct siginfo si; | |
09a6adf5 | 311 | const struct fault_info *inf; |
e7c600f1 | 312 | unsigned int lsb = 0; |
1d18c47c | 313 | |
f871d268 | 314 | if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) { |
09a6adf5 | 315 | inf = esr_to_fault_info(esr); |
83016b20 | 316 | pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x", |
09a6adf5 | 317 | tsk->comm, task_pid_nr(tsk), inf->name, sig, |
3495386b | 318 | addr, esr); |
83016b20 KM |
319 | print_vma_addr(KERN_CONT ", in ", regs->pc); |
320 | pr_cont("\n"); | |
c07ab957 | 321 | __show_regs(regs); |
1d18c47c CM |
322 | } |
323 | ||
324 | tsk->thread.fault_address = addr; | |
9141300a | 325 | tsk->thread.fault_code = esr; |
1d18c47c CM |
326 | si.si_signo = sig; |
327 | si.si_errno = 0; | |
328 | si.si_code = code; | |
329 | si.si_addr = (void __user *)addr; | |
e7c600f1 JZZ |
330 | /* |
331 | * Either small page or large page may be poisoned. | |
332 | * In other words, VM_FAULT_HWPOISON_LARGE and | |
333 | * VM_FAULT_HWPOISON are mutually exclusive. | |
334 | */ | |
335 | if (fault & VM_FAULT_HWPOISON_LARGE) | |
336 | lsb = hstate_index_to_shift(VM_FAULT_GET_HINDEX(fault)); | |
337 | else if (fault & VM_FAULT_HWPOISON) | |
338 | lsb = PAGE_SHIFT; | |
339 | si.si_addr_lsb = lsb; | |
340 | ||
1d18c47c CM |
341 | force_sig_info(sig, &si, tsk); |
342 | } | |
343 | ||
59f67e16 | 344 | static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs) |
1d18c47c CM |
345 | { |
346 | struct task_struct *tsk = current; | |
09a6adf5 | 347 | const struct fault_info *inf; |
1d18c47c CM |
348 | |
349 | /* | |
350 | * If we are in kernel mode at this point, we have no context to | |
351 | * handle this fault with. | |
352 | */ | |
09a6adf5 VK |
353 | if (user_mode(regs)) { |
354 | inf = esr_to_fault_info(esr); | |
e7c600f1 | 355 | __do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs, 0); |
09a6adf5 | 356 | } else |
67ce16ec | 357 | __do_kernel_fault(addr, esr, regs); |
1d18c47c CM |
358 | } |
359 | ||
360 | #define VM_FAULT_BADMAP 0x010000 | |
361 | #define VM_FAULT_BADACCESS 0x020000 | |
362 | ||
1d18c47c | 363 | static int __do_page_fault(struct mm_struct *mm, unsigned long addr, |
db6f4106 | 364 | unsigned int mm_flags, unsigned long vm_flags, |
1d18c47c CM |
365 | struct task_struct *tsk) |
366 | { | |
367 | struct vm_area_struct *vma; | |
368 | int fault; | |
369 | ||
370 | vma = find_vma(mm, addr); | |
371 | fault = VM_FAULT_BADMAP; | |
372 | if (unlikely(!vma)) | |
373 | goto out; | |
374 | if (unlikely(vma->vm_start > addr)) | |
375 | goto check_stack; | |
376 | ||
377 | /* | |
378 | * Ok, we have a good vm_area for this memory access, so we can handle | |
379 | * it. | |
380 | */ | |
381 | good_area: | |
db6f4106 WD |
382 | /* |
383 | * Check that the permissions on the VMA allow for the fault which | |
cab15ce6 | 384 | * occurred. |
db6f4106 WD |
385 | */ |
386 | if (!(vma->vm_flags & vm_flags)) { | |
1d18c47c CM |
387 | fault = VM_FAULT_BADACCESS; |
388 | goto out; | |
389 | } | |
390 | ||
dcddffd4 | 391 | return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags); |
1d18c47c CM |
392 | |
393 | check_stack: | |
394 | if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr)) | |
395 | goto good_area; | |
396 | out: | |
397 | return fault; | |
398 | } | |
399 | ||
541ec870 MR |
400 | static bool is_el0_instruction_abort(unsigned int esr) |
401 | { | |
402 | return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW; | |
403 | } | |
404 | ||
1d18c47c CM |
405 | static int __kprobes do_page_fault(unsigned long addr, unsigned int esr, |
406 | struct pt_regs *regs) | |
407 | { | |
408 | struct task_struct *tsk; | |
409 | struct mm_struct *mm; | |
0e3a9026 | 410 | int fault, sig, code, major = 0; |
cab15ce6 | 411 | unsigned long vm_flags = VM_READ | VM_WRITE; |
db6f4106 WD |
412 | unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE; |
413 | ||
2dd0e8d2 SP |
414 | if (notify_page_fault(regs, esr)) |
415 | return 0; | |
416 | ||
1d18c47c CM |
417 | tsk = current; |
418 | mm = tsk->mm; | |
419 | ||
1d18c47c CM |
420 | /* |
421 | * If we're in an interrupt or have no user context, we must not take | |
422 | * the fault. | |
423 | */ | |
70ffdb93 | 424 | if (faulthandler_disabled() || !mm) |
1d18c47c CM |
425 | goto no_context; |
426 | ||
759496ba JW |
427 | if (user_mode(regs)) |
428 | mm_flags |= FAULT_FLAG_USER; | |
429 | ||
541ec870 | 430 | if (is_el0_instruction_abort(esr)) { |
759496ba | 431 | vm_flags = VM_EXEC; |
aed40e01 | 432 | } else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) { |
759496ba JW |
433 | vm_flags = VM_WRITE; |
434 | mm_flags |= FAULT_FLAG_WRITE; | |
435 | } | |
436 | ||
b824b930 | 437 | if (addr < USER_DS && is_permission_fault(esr, regs, addr)) { |
e19a6ee2 JM |
438 | /* regs->orig_addr_limit may be 0 if we entered from EL0 */ |
439 | if (regs->orig_addr_limit == KERNEL_DS) | |
70c8abc2 | 440 | die("Accessing user space memory with fs=KERNEL_DS", regs, esr); |
70544196 | 441 | |
9adeb8e7 LA |
442 | if (is_el1_instruction_abort(esr)) |
443 | die("Attempting to execute userspace memory", regs, esr); | |
444 | ||
57f4959b | 445 | if (!search_exception_tables(regs->pc)) |
70c8abc2 | 446 | die("Accessing user space memory outside uaccess.h routines", regs, esr); |
57f4959b | 447 | } |
338d4f49 | 448 | |
0e3a9026 PA |
449 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr); |
450 | ||
1d18c47c CM |
451 | /* |
452 | * As per x86, we may deadlock here. However, since the kernel only | |
453 | * validly references user space from well defined areas of the code, | |
454 | * we can bug out early if this is from code which shouldn't. | |
455 | */ | |
456 | if (!down_read_trylock(&mm->mmap_sem)) { | |
457 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
458 | goto no_context; | |
459 | retry: | |
460 | down_read(&mm->mmap_sem); | |
461 | } else { | |
462 | /* | |
463 | * The above down_read_trylock() might have succeeded in which | |
464 | * case, we'll have missed the might_sleep() from down_read(). | |
465 | */ | |
466 | might_sleep(); | |
467 | #ifdef CONFIG_DEBUG_VM | |
468 | if (!user_mode(regs) && !search_exception_tables(regs->pc)) | |
469 | goto no_context; | |
470 | #endif | |
471 | } | |
472 | ||
db6f4106 | 473 | fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk); |
0e3a9026 | 474 | major |= fault & VM_FAULT_MAJOR; |
1d18c47c | 475 | |
0e3a9026 PA |
476 | if (fault & VM_FAULT_RETRY) { |
477 | /* | |
478 | * If we need to retry but a fatal signal is pending, | |
479 | * handle the signal first. We do not need to release | |
480 | * the mmap_sem because it would already be released | |
481 | * in __lock_page_or_retry in mm/filemap.c. | |
482 | */ | |
483 | if (fatal_signal_pending(current)) | |
484 | return 0; | |
485 | ||
486 | /* | |
487 | * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of | |
488 | * starvation. | |
489 | */ | |
490 | if (mm_flags & FAULT_FLAG_ALLOW_RETRY) { | |
491 | mm_flags &= ~FAULT_FLAG_ALLOW_RETRY; | |
492 | mm_flags |= FAULT_FLAG_TRIED; | |
493 | goto retry; | |
494 | } | |
495 | } | |
496 | up_read(&mm->mmap_sem); | |
1d18c47c CM |
497 | |
498 | /* | |
0e3a9026 | 499 | * Handle the "normal" (no error) case first. |
1d18c47c | 500 | */ |
0e3a9026 PA |
501 | if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP | |
502 | VM_FAULT_BADACCESS)))) { | |
503 | /* | |
504 | * Major/minor page fault accounting is only done | |
505 | * once. If we go through a retry, it is extremely | |
506 | * likely that the page will be found in page cache at | |
507 | * that point. | |
508 | */ | |
509 | if (major) { | |
1d18c47c CM |
510 | tsk->maj_flt++; |
511 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs, | |
512 | addr); | |
513 | } else { | |
514 | tsk->min_flt++; | |
515 | perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs, | |
516 | addr); | |
517 | } | |
1d18c47c | 518 | |
1d18c47c | 519 | return 0; |
0e3a9026 | 520 | } |
1d18c47c | 521 | |
87134102 JW |
522 | /* |
523 | * If we are in kernel mode at this point, we have no context to | |
524 | * handle this fault with. | |
525 | */ | |
526 | if (!user_mode(regs)) | |
527 | goto no_context; | |
528 | ||
1d18c47c CM |
529 | if (fault & VM_FAULT_OOM) { |
530 | /* | |
531 | * We ran out of memory, call the OOM killer, and return to | |
532 | * userspace (which will retry the fault, or kill us if we got | |
533 | * oom-killed). | |
534 | */ | |
535 | pagefault_out_of_memory(); | |
536 | return 0; | |
537 | } | |
538 | ||
1d18c47c CM |
539 | if (fault & VM_FAULT_SIGBUS) { |
540 | /* | |
541 | * We had some memory, but were unable to successfully fix up | |
542 | * this page fault. | |
543 | */ | |
544 | sig = SIGBUS; | |
545 | code = BUS_ADRERR; | |
e7c600f1 JZZ |
546 | } else if (fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE)) { |
547 | sig = SIGBUS; | |
548 | code = BUS_MCEERR_AR; | |
1d18c47c CM |
549 | } else { |
550 | /* | |
551 | * Something tried to access memory that isn't in our memory | |
552 | * map. | |
553 | */ | |
554 | sig = SIGSEGV; | |
555 | code = fault == VM_FAULT_BADACCESS ? | |
556 | SEGV_ACCERR : SEGV_MAPERR; | |
557 | } | |
558 | ||
e7c600f1 | 559 | __do_user_fault(tsk, addr, esr, sig, code, regs, fault); |
1d18c47c CM |
560 | return 0; |
561 | ||
562 | no_context: | |
67ce16ec | 563 | __do_kernel_fault(addr, esr, regs); |
1d18c47c CM |
564 | return 0; |
565 | } | |
566 | ||
567 | /* | |
568 | * First Level Translation Fault Handler | |
569 | * | |
570 | * We enter here because the first level page table doesn't contain a valid | |
571 | * entry for the address. | |
572 | * | |
573 | * If the address is in kernel space (>= TASK_SIZE), then we are probably | |
574 | * faulting in the vmalloc() area. | |
575 | * | |
576 | * If the init_task's first level page tables contains the relevant entry, we | |
577 | * copy the it to this task. If not, we send the process a signal, fixup the | |
578 | * exception, or oops the kernel. | |
579 | * | |
580 | * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt | |
581 | * or a critical region, and should only copy the information from the master | |
582 | * page table, nothing more. | |
583 | */ | |
584 | static int __kprobes do_translation_fault(unsigned long addr, | |
585 | unsigned int esr, | |
586 | struct pt_regs *regs) | |
587 | { | |
588 | if (addr < TASK_SIZE) | |
589 | return do_page_fault(addr, esr, regs); | |
590 | ||
591 | do_bad_area(addr, esr, regs); | |
592 | return 0; | |
593 | } | |
594 | ||
52d7523d EL |
595 | static int do_alignment_fault(unsigned long addr, unsigned int esr, |
596 | struct pt_regs *regs) | |
597 | { | |
598 | do_bad_area(addr, esr, regs); | |
599 | return 0; | |
600 | } | |
601 | ||
1d18c47c CM |
602 | /* |
603 | * This abort handler always returns "fault". | |
604 | */ | |
605 | static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
606 | { | |
607 | return 1; | |
608 | } | |
609 | ||
32015c23 TB |
610 | /* |
611 | * This abort handler deals with Synchronous External Abort. | |
612 | * It calls notifiers, and then returns "fault". | |
613 | */ | |
614 | static int do_sea(unsigned long addr, unsigned int esr, struct pt_regs *regs) | |
615 | { | |
616 | struct siginfo info; | |
617 | const struct fault_info *inf; | |
621f48e4 | 618 | int ret = 0; |
32015c23 TB |
619 | |
620 | inf = esr_to_fault_info(esr); | |
621 | pr_err("Synchronous External Abort: %s (0x%08x) at 0x%016lx\n", | |
622 | inf->name, esr, addr); | |
623 | ||
7edda088 TB |
624 | /* |
625 | * Synchronous aborts may interrupt code which had interrupts masked. | |
626 | * Before calling out into the wider kernel tell the interested | |
627 | * subsystems. | |
628 | */ | |
629 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) { | |
630 | if (interrupts_enabled(regs)) | |
631 | nmi_enter(); | |
632 | ||
621f48e4 | 633 | ret = ghes_notify_sea(); |
7edda088 TB |
634 | |
635 | if (interrupts_enabled(regs)) | |
636 | nmi_exit(); | |
637 | } | |
638 | ||
32015c23 TB |
639 | info.si_signo = SIGBUS; |
640 | info.si_errno = 0; | |
641 | info.si_code = 0; | |
642 | if (esr & ESR_ELx_FnV) | |
643 | info.si_addr = NULL; | |
644 | else | |
645 | info.si_addr = (void __user *)addr; | |
646 | arm64_notify_die("", regs, &info, esr); | |
647 | ||
621f48e4 | 648 | return ret; |
32015c23 TB |
649 | } |
650 | ||
09a6adf5 | 651 | static const struct fault_info fault_info[] = { |
1d18c47c CM |
652 | { do_bad, SIGBUS, 0, "ttbr address size fault" }, |
653 | { do_bad, SIGBUS, 0, "level 1 address size fault" }, | |
654 | { do_bad, SIGBUS, 0, "level 2 address size fault" }, | |
655 | { do_bad, SIGBUS, 0, "level 3 address size fault" }, | |
7f73f7ae | 656 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 0 translation fault" }, |
1d18c47c CM |
657 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 1 translation fault" }, |
658 | { do_translation_fault, SIGSEGV, SEGV_MAPERR, "level 2 translation fault" }, | |
659 | { do_page_fault, SIGSEGV, SEGV_MAPERR, "level 3 translation fault" }, | |
c03784ee | 660 | { do_bad, SIGBUS, 0, "unknown 8" }, |
084bd298 SC |
661 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 access flag fault" }, |
662 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 access flag fault" }, | |
1d18c47c | 663 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 access flag fault" }, |
c03784ee | 664 | { do_bad, SIGBUS, 0, "unknown 12" }, |
084bd298 SC |
665 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 1 permission fault" }, |
666 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 2 permission fault" }, | |
1d18c47c | 667 | { do_page_fault, SIGSEGV, SEGV_ACCERR, "level 3 permission fault" }, |
32015c23 | 668 | { do_sea, SIGBUS, 0, "synchronous external abort" }, |
c03784ee | 669 | { do_bad, SIGBUS, 0, "unknown 17" }, |
1d18c47c CM |
670 | { do_bad, SIGBUS, 0, "unknown 18" }, |
671 | { do_bad, SIGBUS, 0, "unknown 19" }, | |
32015c23 TB |
672 | { do_sea, SIGBUS, 0, "level 0 (translation table walk)" }, |
673 | { do_sea, SIGBUS, 0, "level 1 (translation table walk)" }, | |
674 | { do_sea, SIGBUS, 0, "level 2 (translation table walk)" }, | |
675 | { do_sea, SIGBUS, 0, "level 3 (translation table walk)" }, | |
676 | { do_sea, SIGBUS, 0, "synchronous parity or ECC error" }, | |
c03784ee | 677 | { do_bad, SIGBUS, 0, "unknown 25" }, |
1d18c47c CM |
678 | { do_bad, SIGBUS, 0, "unknown 26" }, |
679 | { do_bad, SIGBUS, 0, "unknown 27" }, | |
32015c23 TB |
680 | { do_sea, SIGBUS, 0, "level 0 synchronous parity error (translation table walk)" }, |
681 | { do_sea, SIGBUS, 0, "level 1 synchronous parity error (translation table walk)" }, | |
682 | { do_sea, SIGBUS, 0, "level 2 synchronous parity error (translation table walk)" }, | |
683 | { do_sea, SIGBUS, 0, "level 3 synchronous parity error (translation table walk)" }, | |
1d18c47c | 684 | { do_bad, SIGBUS, 0, "unknown 32" }, |
52d7523d | 685 | { do_alignment_fault, SIGBUS, BUS_ADRALN, "alignment fault" }, |
c03784ee | 686 | { do_bad, SIGBUS, 0, "unknown 34" }, |
1d18c47c CM |
687 | { do_bad, SIGBUS, 0, "unknown 35" }, |
688 | { do_bad, SIGBUS, 0, "unknown 36" }, | |
689 | { do_bad, SIGBUS, 0, "unknown 37" }, | |
690 | { do_bad, SIGBUS, 0, "unknown 38" }, | |
691 | { do_bad, SIGBUS, 0, "unknown 39" }, | |
692 | { do_bad, SIGBUS, 0, "unknown 40" }, | |
693 | { do_bad, SIGBUS, 0, "unknown 41" }, | |
694 | { do_bad, SIGBUS, 0, "unknown 42" }, | |
695 | { do_bad, SIGBUS, 0, "unknown 43" }, | |
696 | { do_bad, SIGBUS, 0, "unknown 44" }, | |
697 | { do_bad, SIGBUS, 0, "unknown 45" }, | |
698 | { do_bad, SIGBUS, 0, "unknown 46" }, | |
699 | { do_bad, SIGBUS, 0, "unknown 47" }, | |
c03784ee | 700 | { do_bad, SIGBUS, 0, "TLB conflict abort" }, |
1d18c47c CM |
701 | { do_bad, SIGBUS, 0, "unknown 49" }, |
702 | { do_bad, SIGBUS, 0, "unknown 50" }, | |
703 | { do_bad, SIGBUS, 0, "unknown 51" }, | |
704 | { do_bad, SIGBUS, 0, "implementation fault (lockdown abort)" }, | |
c03784ee | 705 | { do_bad, SIGBUS, 0, "implementation fault (unsupported exclusive)" }, |
1d18c47c CM |
706 | { do_bad, SIGBUS, 0, "unknown 54" }, |
707 | { do_bad, SIGBUS, 0, "unknown 55" }, | |
708 | { do_bad, SIGBUS, 0, "unknown 56" }, | |
709 | { do_bad, SIGBUS, 0, "unknown 57" }, | |
c03784ee | 710 | { do_bad, SIGBUS, 0, "unknown 58" }, |
1d18c47c CM |
711 | { do_bad, SIGBUS, 0, "unknown 59" }, |
712 | { do_bad, SIGBUS, 0, "unknown 60" }, | |
c03784ee MR |
713 | { do_bad, SIGBUS, 0, "section domain fault" }, |
714 | { do_bad, SIGBUS, 0, "page domain fault" }, | |
1d18c47c CM |
715 | { do_bad, SIGBUS, 0, "unknown 63" }, |
716 | }; | |
717 | ||
621f48e4 TB |
718 | /* |
719 | * Handle Synchronous External Aborts that occur in a guest kernel. | |
720 | * | |
721 | * The return value will be zero if the SEA was successfully handled | |
722 | * and non-zero if there was an error processing the error or there was | |
723 | * no error to process. | |
724 | */ | |
725 | int handle_guest_sea(phys_addr_t addr, unsigned int esr) | |
726 | { | |
727 | int ret = -ENOENT; | |
728 | ||
729 | if (IS_ENABLED(CONFIG_ACPI_APEI_SEA)) | |
730 | ret = ghes_notify_sea(); | |
731 | ||
732 | return ret; | |
733 | } | |
734 | ||
1d18c47c CM |
735 | /* |
736 | * Dispatch a data abort to the relevant handler. | |
737 | */ | |
738 | asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr, | |
739 | struct pt_regs *regs) | |
740 | { | |
09a6adf5 | 741 | const struct fault_info *inf = esr_to_fault_info(esr); |
1d18c47c CM |
742 | struct siginfo info; |
743 | ||
744 | if (!inf->fn(addr, esr, regs)) | |
745 | return; | |
746 | ||
747 | pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n", | |
748 | inf->name, esr, addr); | |
749 | ||
1f9b8936 JT |
750 | mem_abort_decode(esr); |
751 | ||
1d18c47c CM |
752 | info.si_signo = inf->sig; |
753 | info.si_errno = 0; | |
754 | info.si_code = inf->code; | |
755 | info.si_addr = (void __user *)addr; | |
756 | arm64_notify_die("", regs, &info, esr); | |
757 | } | |
758 | ||
759 | /* | |
760 | * Handle stack alignment exceptions. | |
761 | */ | |
762 | asmlinkage void __exception do_sp_pc_abort(unsigned long addr, | |
763 | unsigned int esr, | |
764 | struct pt_regs *regs) | |
765 | { | |
766 | struct siginfo info; | |
9e793ab8 VM |
767 | struct task_struct *tsk = current; |
768 | ||
769 | if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS)) | |
770 | pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n", | |
771 | tsk->comm, task_pid_nr(tsk), | |
772 | esr_get_class_string(esr), (void *)regs->pc, | |
773 | (void *)regs->sp); | |
1d18c47c CM |
774 | |
775 | info.si_signo = SIGBUS; | |
776 | info.si_errno = 0; | |
777 | info.si_code = BUS_ADRALN; | |
778 | info.si_addr = (void __user *)addr; | |
9e793ab8 | 779 | arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr); |
1d18c47c CM |
780 | } |
781 | ||
9fb7410f DM |
782 | int __init early_brk64(unsigned long addr, unsigned int esr, |
783 | struct pt_regs *regs); | |
784 | ||
785 | /* | |
786 | * __refdata because early_brk64 is __init, but the reference to it is | |
787 | * clobbered at arch_initcall time. | |
788 | * See traps.c and debug-monitors.c:debug_traps_init(). | |
789 | */ | |
790 | static struct fault_info __refdata debug_fault_info[] = { | |
1d18c47c CM |
791 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware breakpoint" }, |
792 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware single-step" }, | |
793 | { do_bad, SIGTRAP, TRAP_HWBKPT, "hardware watchpoint" }, | |
794 | { do_bad, SIGBUS, 0, "unknown 3" }, | |
795 | { do_bad, SIGTRAP, TRAP_BRKPT, "aarch32 BKPT" }, | |
796 | { do_bad, SIGTRAP, 0, "aarch32 vector catch" }, | |
9fb7410f | 797 | { early_brk64, SIGTRAP, TRAP_BRKPT, "aarch64 BRK" }, |
1d18c47c CM |
798 | { do_bad, SIGBUS, 0, "unknown 7" }, |
799 | }; | |
800 | ||
801 | void __init hook_debug_fault_code(int nr, | |
802 | int (*fn)(unsigned long, unsigned int, struct pt_regs *), | |
803 | int sig, int code, const char *name) | |
804 | { | |
805 | BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info)); | |
806 | ||
807 | debug_fault_info[nr].fn = fn; | |
808 | debug_fault_info[nr].sig = sig; | |
809 | debug_fault_info[nr].code = code; | |
810 | debug_fault_info[nr].name = name; | |
811 | } | |
812 | ||
813 | asmlinkage int __exception do_debug_exception(unsigned long addr, | |
814 | unsigned int esr, | |
815 | struct pt_regs *regs) | |
816 | { | |
817 | const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr); | |
818 | struct siginfo info; | |
6afedcd2 | 819 | int rv; |
1d18c47c | 820 | |
6afedcd2 JM |
821 | /* |
822 | * Tell lockdep we disabled irqs in entry.S. Do nothing if they were | |
823 | * already disabled to preserve the last enabled/disabled addresses. | |
824 | */ | |
825 | if (interrupts_enabled(regs)) | |
826 | trace_hardirqs_off(); | |
1d18c47c | 827 | |
6afedcd2 JM |
828 | if (!inf->fn(addr, esr, regs)) { |
829 | rv = 1; | |
830 | } else { | |
831 | pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n", | |
832 | inf->name, esr, addr); | |
833 | ||
834 | info.si_signo = inf->sig; | |
835 | info.si_errno = 0; | |
836 | info.si_code = inf->code; | |
837 | info.si_addr = (void __user *)addr; | |
838 | arm64_notify_die("", regs, &info, 0); | |
839 | rv = 0; | |
840 | } | |
1d18c47c | 841 | |
6afedcd2 JM |
842 | if (interrupts_enabled(regs)) |
843 | trace_hardirqs_on(); | |
1d18c47c | 844 | |
6afedcd2 | 845 | return rv; |
1d18c47c | 846 | } |
2dd0e8d2 | 847 | NOKPROBE_SYMBOL(do_debug_exception); |
338d4f49 JM |
848 | |
849 | #ifdef CONFIG_ARM64_PAN | |
2a6dcb2b | 850 | int cpu_enable_pan(void *__unused) |
338d4f49 | 851 | { |
7209c868 JM |
852 | /* |
853 | * We modify PSTATE. This won't work from irq context as the PSTATE | |
854 | * is discarded once we return from the exception. | |
855 | */ | |
856 | WARN_ON_ONCE(in_interrupt()); | |
857 | ||
338d4f49 | 858 | config_sctlr_el1(SCTLR_EL1_SPAN, 0); |
7209c868 | 859 | asm(SET_PSTATE_PAN(1)); |
2a6dcb2b | 860 | return 0; |
338d4f49 JM |
861 | } |
862 | #endif /* CONFIG_ARM64_PAN */ |