xref: /openbmc/linux/arch/arm64/mm/fault.c (revision bb0eb050)
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 
21 #include <linux/extable.h>
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/signal.h>
30 #include <linux/sched/debug.h>
31 #include <linux/highmem.h>
32 #include <linux/perf_event.h>
33 #include <linux/preempt.h>
34 
35 #include <asm/bug.h>
36 #include <asm/cpufeature.h>
37 #include <asm/exception.h>
38 #include <asm/debug-monitors.h>
39 #include <asm/esr.h>
40 #include <asm/sysreg.h>
41 #include <asm/system_misc.h>
42 #include <asm/pgtable.h>
43 #include <asm/tlbflush.h>
44 
45 struct fault_info {
46 	int	(*fn)(unsigned long addr, unsigned int esr,
47 		      struct pt_regs *regs);
48 	int	sig;
49 	int	code;
50 	const char *name;
51 };
52 
53 static const struct fault_info fault_info[];
54 
55 static inline const struct fault_info *esr_to_fault_info(unsigned int esr)
56 {
57 	return fault_info + (esr & 63);
58 }
59 
60 #ifdef CONFIG_KPROBES
61 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
62 {
63 	int ret = 0;
64 
65 	/* kprobe_running() needs smp_processor_id() */
66 	if (!user_mode(regs)) {
67 		preempt_disable();
68 		if (kprobe_running() && kprobe_fault_handler(regs, esr))
69 			ret = 1;
70 		preempt_enable();
71 	}
72 
73 	return ret;
74 }
75 #else
76 static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
77 {
78 	return 0;
79 }
80 #endif
81 
82 /*
83  * Dump out the page tables associated with 'addr' in mm 'mm'.
84  */
85 void show_pte(struct mm_struct *mm, unsigned long addr)
86 {
87 	pgd_t *pgd;
88 
89 	if (!mm)
90 		mm = &init_mm;
91 
92 	pr_alert("pgd = %p\n", mm->pgd);
93 	pgd = pgd_offset(mm, addr);
94 	pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
95 
96 	do {
97 		pud_t *pud;
98 		pmd_t *pmd;
99 		pte_t *pte;
100 
101 		if (pgd_none(*pgd) || pgd_bad(*pgd))
102 			break;
103 
104 		pud = pud_offset(pgd, addr);
105 		pr_cont(", *pud=%016llx", pud_val(*pud));
106 		if (pud_none(*pud) || pud_bad(*pud))
107 			break;
108 
109 		pmd = pmd_offset(pud, addr);
110 		pr_cont(", *pmd=%016llx", pmd_val(*pmd));
111 		if (pmd_none(*pmd) || pmd_bad(*pmd))
112 			break;
113 
114 		pte = pte_offset_map(pmd, addr);
115 		pr_cont(", *pte=%016llx", pte_val(*pte));
116 		pte_unmap(pte);
117 	} while(0);
118 
119 	pr_cont("\n");
120 }
121 
122 #ifdef CONFIG_ARM64_HW_AFDBM
123 /*
124  * This function sets the access flags (dirty, accessed), as well as write
125  * permission, and only to a more permissive setting.
126  *
127  * It needs to cope with hardware update of the accessed/dirty state by other
128  * agents in the system and can safely skip the __sync_icache_dcache() call as,
129  * like set_pte_at(), the PTE is never changed from no-exec to exec here.
130  *
131  * Returns whether or not the PTE actually changed.
132  */
133 int ptep_set_access_flags(struct vm_area_struct *vma,
134 			  unsigned long address, pte_t *ptep,
135 			  pte_t entry, int dirty)
136 {
137 	pteval_t old_pteval;
138 	unsigned int tmp;
139 
140 	if (pte_same(*ptep, entry))
141 		return 0;
142 
143 	/* only preserve the access flags and write permission */
144 	pte_val(entry) &= PTE_AF | PTE_WRITE | PTE_DIRTY;
145 
146 	/*
147 	 * PTE_RDONLY is cleared by default in the asm below, so set it in
148 	 * back if necessary (read-only or clean PTE).
149 	 */
150 	if (!pte_write(entry) || !pte_sw_dirty(entry))
151 		pte_val(entry) |= PTE_RDONLY;
152 
153 	/*
154 	 * Setting the flags must be done atomically to avoid racing with the
155 	 * hardware update of the access/dirty state.
156 	 */
157 	asm volatile("//	ptep_set_access_flags\n"
158 	"	prfm	pstl1strm, %2\n"
159 	"1:	ldxr	%0, %2\n"
160 	"	and	%0, %0, %3		// clear PTE_RDONLY\n"
161 	"	orr	%0, %0, %4		// set flags\n"
162 	"	stxr	%w1, %0, %2\n"
163 	"	cbnz	%w1, 1b\n"
164 	: "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
165 	: "L" (~PTE_RDONLY), "r" (pte_val(entry)));
166 
167 	flush_tlb_fix_spurious_fault(vma, address);
168 	return 1;
169 }
170 #endif
171 
172 static bool is_el1_instruction_abort(unsigned int esr)
173 {
174 	return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
175 }
176 
177 static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs,
178 				       unsigned long addr)
179 {
180 	unsigned int ec       = ESR_ELx_EC(esr);
181 	unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
182 
183 	if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
184 		return false;
185 
186 	if (fsc_type == ESR_ELx_FSC_PERM)
187 		return true;
188 
189 	if (addr < USER_DS && system_uses_ttbr0_pan())
190 		return fsc_type == ESR_ELx_FSC_FAULT &&
191 			(regs->pstate & PSR_PAN_BIT);
192 
193 	return false;
194 }
195 
196 /*
197  * The kernel tried to access some page that wasn't present.
198  */
199 static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
200 			      unsigned int esr, struct pt_regs *regs)
201 {
202 	const char *msg;
203 
204 	/*
205 	 * Are we prepared to handle this kernel fault?
206 	 * We are almost certainly not prepared to handle instruction faults.
207 	 */
208 	if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
209 		return;
210 
211 	/*
212 	 * No handler, we'll have to terminate things with extreme prejudice.
213 	 */
214 	bust_spinlocks(1);
215 
216 	if (is_permission_fault(esr, regs, addr)) {
217 		if (esr & ESR_ELx_WNR)
218 			msg = "write to read-only memory";
219 		else
220 			msg = "read from unreadable memory";
221 	} else if (addr < PAGE_SIZE) {
222 		msg = "NULL pointer dereference";
223 	} else {
224 		msg = "paging request";
225 	}
226 
227 	pr_alert("Unable to handle kernel %s at virtual address %08lx\n", msg,
228 		 addr);
229 
230 	show_pte(mm, addr);
231 	die("Oops", regs, esr);
232 	bust_spinlocks(0);
233 	do_exit(SIGKILL);
234 }
235 
236 /*
237  * Something tried to access memory that isn't in our memory map. User mode
238  * accesses just cause a SIGSEGV
239  */
240 static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
241 			    unsigned int esr, unsigned int sig, int code,
242 			    struct pt_regs *regs)
243 {
244 	struct siginfo si;
245 	const struct fault_info *inf;
246 
247 	if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
248 		inf = esr_to_fault_info(esr);
249 		pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
250 			tsk->comm, task_pid_nr(tsk), inf->name, sig,
251 			addr, esr);
252 		show_pte(tsk->mm, addr);
253 		show_regs(regs);
254 	}
255 
256 	tsk->thread.fault_address = addr;
257 	tsk->thread.fault_code = esr;
258 	si.si_signo = sig;
259 	si.si_errno = 0;
260 	si.si_code = code;
261 	si.si_addr = (void __user *)addr;
262 	force_sig_info(sig, &si, tsk);
263 }
264 
265 static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
266 {
267 	struct task_struct *tsk = current;
268 	struct mm_struct *mm = tsk->active_mm;
269 	const struct fault_info *inf;
270 
271 	/*
272 	 * If we are in kernel mode at this point, we have no context to
273 	 * handle this fault with.
274 	 */
275 	if (user_mode(regs)) {
276 		inf = esr_to_fault_info(esr);
277 		__do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs);
278 	} else
279 		__do_kernel_fault(mm, addr, esr, regs);
280 }
281 
282 #define VM_FAULT_BADMAP		0x010000
283 #define VM_FAULT_BADACCESS	0x020000
284 
285 static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
286 			   unsigned int mm_flags, unsigned long vm_flags,
287 			   struct task_struct *tsk)
288 {
289 	struct vm_area_struct *vma;
290 	int fault;
291 
292 	vma = find_vma(mm, addr);
293 	fault = VM_FAULT_BADMAP;
294 	if (unlikely(!vma))
295 		goto out;
296 	if (unlikely(vma->vm_start > addr))
297 		goto check_stack;
298 
299 	/*
300 	 * Ok, we have a good vm_area for this memory access, so we can handle
301 	 * it.
302 	 */
303 good_area:
304 	/*
305 	 * Check that the permissions on the VMA allow for the fault which
306 	 * occurred.
307 	 */
308 	if (!(vma->vm_flags & vm_flags)) {
309 		fault = VM_FAULT_BADACCESS;
310 		goto out;
311 	}
312 
313 	return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
314 
315 check_stack:
316 	if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
317 		goto good_area;
318 out:
319 	return fault;
320 }
321 
322 static bool is_el0_instruction_abort(unsigned int esr)
323 {
324 	return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
325 }
326 
327 static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
328 				   struct pt_regs *regs)
329 {
330 	struct task_struct *tsk;
331 	struct mm_struct *mm;
332 	int fault, sig, code;
333 	unsigned long vm_flags = VM_READ | VM_WRITE;
334 	unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
335 
336 	if (notify_page_fault(regs, esr))
337 		return 0;
338 
339 	tsk = current;
340 	mm  = tsk->mm;
341 
342 	/*
343 	 * If we're in an interrupt or have no user context, we must not take
344 	 * the fault.
345 	 */
346 	if (faulthandler_disabled() || !mm)
347 		goto no_context;
348 
349 	if (user_mode(regs))
350 		mm_flags |= FAULT_FLAG_USER;
351 
352 	if (is_el0_instruction_abort(esr)) {
353 		vm_flags = VM_EXEC;
354 	} else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
355 		vm_flags = VM_WRITE;
356 		mm_flags |= FAULT_FLAG_WRITE;
357 	}
358 
359 	if (addr < USER_DS && is_permission_fault(esr, regs, addr)) {
360 		/* regs->orig_addr_limit may be 0 if we entered from EL0 */
361 		if (regs->orig_addr_limit == KERNEL_DS)
362 			die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
363 
364 		if (is_el1_instruction_abort(esr))
365 			die("Attempting to execute userspace memory", regs, esr);
366 
367 		if (!search_exception_tables(regs->pc))
368 			die("Accessing user space memory outside uaccess.h routines", regs, esr);
369 	}
370 
371 	/*
372 	 * As per x86, we may deadlock here. However, since the kernel only
373 	 * validly references user space from well defined areas of the code,
374 	 * we can bug out early if this is from code which shouldn't.
375 	 */
376 	if (!down_read_trylock(&mm->mmap_sem)) {
377 		if (!user_mode(regs) && !search_exception_tables(regs->pc))
378 			goto no_context;
379 retry:
380 		down_read(&mm->mmap_sem);
381 	} else {
382 		/*
383 		 * The above down_read_trylock() might have succeeded in which
384 		 * case, we'll have missed the might_sleep() from down_read().
385 		 */
386 		might_sleep();
387 #ifdef CONFIG_DEBUG_VM
388 		if (!user_mode(regs) && !search_exception_tables(regs->pc))
389 			goto no_context;
390 #endif
391 	}
392 
393 	fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
394 
395 	/*
396 	 * If we need to retry but a fatal signal is pending, handle the
397 	 * signal first. We do not need to release the mmap_sem because it
398 	 * would already be released in __lock_page_or_retry in mm/filemap.c.
399 	 */
400 	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
401 		return 0;
402 
403 	/*
404 	 * Major/minor page fault accounting is only done on the initial
405 	 * attempt. If we go through a retry, it is extremely likely that the
406 	 * page will be found in page cache at that point.
407 	 */
408 
409 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
410 	if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
411 		if (fault & VM_FAULT_MAJOR) {
412 			tsk->maj_flt++;
413 			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
414 				      addr);
415 		} else {
416 			tsk->min_flt++;
417 			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
418 				      addr);
419 		}
420 		if (fault & VM_FAULT_RETRY) {
421 			/*
422 			 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
423 			 * starvation.
424 			 */
425 			mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
426 			mm_flags |= FAULT_FLAG_TRIED;
427 			goto retry;
428 		}
429 	}
430 
431 	up_read(&mm->mmap_sem);
432 
433 	/*
434 	 * Handle the "normal" case first - VM_FAULT_MAJOR
435 	 */
436 	if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
437 			      VM_FAULT_BADACCESS))))
438 		return 0;
439 
440 	/*
441 	 * If we are in kernel mode at this point, we have no context to
442 	 * handle this fault with.
443 	 */
444 	if (!user_mode(regs))
445 		goto no_context;
446 
447 	if (fault & VM_FAULT_OOM) {
448 		/*
449 		 * We ran out of memory, call the OOM killer, and return to
450 		 * userspace (which will retry the fault, or kill us if we got
451 		 * oom-killed).
452 		 */
453 		pagefault_out_of_memory();
454 		return 0;
455 	}
456 
457 	if (fault & VM_FAULT_SIGBUS) {
458 		/*
459 		 * We had some memory, but were unable to successfully fix up
460 		 * this page fault.
461 		 */
462 		sig = SIGBUS;
463 		code = BUS_ADRERR;
464 	} else {
465 		/*
466 		 * Something tried to access memory that isn't in our memory
467 		 * map.
468 		 */
469 		sig = SIGSEGV;
470 		code = fault == VM_FAULT_BADACCESS ?
471 			SEGV_ACCERR : SEGV_MAPERR;
472 	}
473 
474 	__do_user_fault(tsk, addr, esr, sig, code, regs);
475 	return 0;
476 
477 no_context:
478 	__do_kernel_fault(mm, addr, esr, regs);
479 	return 0;
480 }
481 
482 /*
483  * First Level Translation Fault Handler
484  *
485  * We enter here because the first level page table doesn't contain a valid
486  * entry for the address.
487  *
488  * If the address is in kernel space (>= TASK_SIZE), then we are probably
489  * faulting in the vmalloc() area.
490  *
491  * If the init_task's first level page tables contains the relevant entry, we
492  * copy the it to this task.  If not, we send the process a signal, fixup the
493  * exception, or oops the kernel.
494  *
495  * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
496  * or a critical region, and should only copy the information from the master
497  * page table, nothing more.
498  */
499 static int __kprobes do_translation_fault(unsigned long addr,
500 					  unsigned int esr,
501 					  struct pt_regs *regs)
502 {
503 	if (addr < TASK_SIZE)
504 		return do_page_fault(addr, esr, regs);
505 
506 	do_bad_area(addr, esr, regs);
507 	return 0;
508 }
509 
510 static int do_alignment_fault(unsigned long addr, unsigned int esr,
511 			      struct pt_regs *regs)
512 {
513 	do_bad_area(addr, esr, regs);
514 	return 0;
515 }
516 
517 /*
518  * This abort handler always returns "fault".
519  */
520 static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
521 {
522 	return 1;
523 }
524 
525 static const struct fault_info fault_info[] = {
526 	{ do_bad,		SIGBUS,  0,		"ttbr address size fault"	},
527 	{ do_bad,		SIGBUS,  0,		"level 1 address size fault"	},
528 	{ do_bad,		SIGBUS,  0,		"level 2 address size fault"	},
529 	{ do_bad,		SIGBUS,  0,		"level 3 address size fault"	},
530 	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 0 translation fault"	},
531 	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 1 translation fault"	},
532 	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 2 translation fault"	},
533 	{ do_page_fault,	SIGSEGV, SEGV_MAPERR,	"level 3 translation fault"	},
534 	{ do_bad,		SIGBUS,  0,		"unknown 8"			},
535 	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 1 access flag fault"	},
536 	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 2 access flag fault"	},
537 	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 3 access flag fault"	},
538 	{ do_bad,		SIGBUS,  0,		"unknown 12"			},
539 	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 1 permission fault"	},
540 	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 2 permission fault"	},
541 	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 3 permission fault"	},
542 	{ do_bad,		SIGBUS,  0,		"synchronous external abort"	},
543 	{ do_bad,		SIGBUS,  0,		"unknown 17"			},
544 	{ do_bad,		SIGBUS,  0,		"unknown 18"			},
545 	{ do_bad,		SIGBUS,  0,		"unknown 19"			},
546 	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
547 	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
548 	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
549 	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
550 	{ do_bad,		SIGBUS,  0,		"synchronous parity error"	},
551 	{ do_bad,		SIGBUS,  0,		"unknown 25"			},
552 	{ do_bad,		SIGBUS,  0,		"unknown 26"			},
553 	{ do_bad,		SIGBUS,  0,		"unknown 27"			},
554 	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
555 	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
556 	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
557 	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
558 	{ do_bad,		SIGBUS,  0,		"unknown 32"			},
559 	{ do_alignment_fault,	SIGBUS,  BUS_ADRALN,	"alignment fault"		},
560 	{ do_bad,		SIGBUS,  0,		"unknown 34"			},
561 	{ do_bad,		SIGBUS,  0,		"unknown 35"			},
562 	{ do_bad,		SIGBUS,  0,		"unknown 36"			},
563 	{ do_bad,		SIGBUS,  0,		"unknown 37"			},
564 	{ do_bad,		SIGBUS,  0,		"unknown 38"			},
565 	{ do_bad,		SIGBUS,  0,		"unknown 39"			},
566 	{ do_bad,		SIGBUS,  0,		"unknown 40"			},
567 	{ do_bad,		SIGBUS,  0,		"unknown 41"			},
568 	{ do_bad,		SIGBUS,  0,		"unknown 42"			},
569 	{ do_bad,		SIGBUS,  0,		"unknown 43"			},
570 	{ do_bad,		SIGBUS,  0,		"unknown 44"			},
571 	{ do_bad,		SIGBUS,  0,		"unknown 45"			},
572 	{ do_bad,		SIGBUS,  0,		"unknown 46"			},
573 	{ do_bad,		SIGBUS,  0,		"unknown 47"			},
574 	{ do_bad,		SIGBUS,  0,		"TLB conflict abort"		},
575 	{ do_bad,		SIGBUS,  0,		"unknown 49"			},
576 	{ do_bad,		SIGBUS,  0,		"unknown 50"			},
577 	{ do_bad,		SIGBUS,  0,		"unknown 51"			},
578 	{ do_bad,		SIGBUS,  0,		"implementation fault (lockdown abort)" },
579 	{ do_bad,		SIGBUS,  0,		"implementation fault (unsupported exclusive)" },
580 	{ do_bad,		SIGBUS,  0,		"unknown 54"			},
581 	{ do_bad,		SIGBUS,  0,		"unknown 55"			},
582 	{ do_bad,		SIGBUS,  0,		"unknown 56"			},
583 	{ do_bad,		SIGBUS,  0,		"unknown 57"			},
584 	{ do_bad,		SIGBUS,  0,		"unknown 58" 			},
585 	{ do_bad,		SIGBUS,  0,		"unknown 59"			},
586 	{ do_bad,		SIGBUS,  0,		"unknown 60"			},
587 	{ do_bad,		SIGBUS,  0,		"section domain fault"		},
588 	{ do_bad,		SIGBUS,  0,		"page domain fault"		},
589 	{ do_bad,		SIGBUS,  0,		"unknown 63"			},
590 };
591 
592 /*
593  * Dispatch a data abort to the relevant handler.
594  */
595 asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
596 					 struct pt_regs *regs)
597 {
598 	const struct fault_info *inf = esr_to_fault_info(esr);
599 	struct siginfo info;
600 
601 	if (!inf->fn(addr, esr, regs))
602 		return;
603 
604 	pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
605 		 inf->name, esr, addr);
606 
607 	info.si_signo = inf->sig;
608 	info.si_errno = 0;
609 	info.si_code  = inf->code;
610 	info.si_addr  = (void __user *)addr;
611 	arm64_notify_die("", regs, &info, esr);
612 }
613 
614 /*
615  * Handle stack alignment exceptions.
616  */
617 asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
618 					   unsigned int esr,
619 					   struct pt_regs *regs)
620 {
621 	struct siginfo info;
622 	struct task_struct *tsk = current;
623 
624 	if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
625 		pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
626 				    tsk->comm, task_pid_nr(tsk),
627 				    esr_get_class_string(esr), (void *)regs->pc,
628 				    (void *)regs->sp);
629 
630 	info.si_signo = SIGBUS;
631 	info.si_errno = 0;
632 	info.si_code  = BUS_ADRALN;
633 	info.si_addr  = (void __user *)addr;
634 	arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
635 }
636 
637 int __init early_brk64(unsigned long addr, unsigned int esr,
638 		       struct pt_regs *regs);
639 
640 /*
641  * __refdata because early_brk64 is __init, but the reference to it is
642  * clobbered at arch_initcall time.
643  * See traps.c and debug-monitors.c:debug_traps_init().
644  */
645 static struct fault_info __refdata debug_fault_info[] = {
646 	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware breakpoint"	},
647 	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware single-step"	},
648 	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware watchpoint"	},
649 	{ do_bad,	SIGBUS,		0,		"unknown 3"		},
650 	{ do_bad,	SIGTRAP,	TRAP_BRKPT,	"aarch32 BKPT"		},
651 	{ do_bad,	SIGTRAP,	0,		"aarch32 vector catch"	},
652 	{ early_brk64,	SIGTRAP,	TRAP_BRKPT,	"aarch64 BRK"		},
653 	{ do_bad,	SIGBUS,		0,		"unknown 7"		},
654 };
655 
656 void __init hook_debug_fault_code(int nr,
657 				  int (*fn)(unsigned long, unsigned int, struct pt_regs *),
658 				  int sig, int code, const char *name)
659 {
660 	BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
661 
662 	debug_fault_info[nr].fn		= fn;
663 	debug_fault_info[nr].sig	= sig;
664 	debug_fault_info[nr].code	= code;
665 	debug_fault_info[nr].name	= name;
666 }
667 
668 asmlinkage int __exception do_debug_exception(unsigned long addr,
669 					      unsigned int esr,
670 					      struct pt_regs *regs)
671 {
672 	const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
673 	struct siginfo info;
674 	int rv;
675 
676 	/*
677 	 * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
678 	 * already disabled to preserve the last enabled/disabled addresses.
679 	 */
680 	if (interrupts_enabled(regs))
681 		trace_hardirqs_off();
682 
683 	if (!inf->fn(addr, esr, regs)) {
684 		rv = 1;
685 	} else {
686 		pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
687 			 inf->name, esr, addr);
688 
689 		info.si_signo = inf->sig;
690 		info.si_errno = 0;
691 		info.si_code  = inf->code;
692 		info.si_addr  = (void __user *)addr;
693 		arm64_notify_die("", regs, &info, 0);
694 		rv = 0;
695 	}
696 
697 	if (interrupts_enabled(regs))
698 		trace_hardirqs_on();
699 
700 	return rv;
701 }
702 NOKPROBE_SYMBOL(do_debug_exception);
703 
704 #ifdef CONFIG_ARM64_PAN
705 int cpu_enable_pan(void *__unused)
706 {
707 	/*
708 	 * We modify PSTATE. This won't work from irq context as the PSTATE
709 	 * is discarded once we return from the exception.
710 	 */
711 	WARN_ON_ONCE(in_interrupt());
712 
713 	config_sctlr_el1(SCTLR_EL1_SPAN, 0);
714 	asm(SET_PSTATE_PAN(1));
715 	return 0;
716 }
717 #endif /* CONFIG_ARM64_PAN */
718