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