Linux Audio

Check our new training course

Loading...
v4.6
  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
 
 33#include <asm/cpufeature.h>
 34#include <asm/exception.h>
 35#include <asm/debug-monitors.h>
 36#include <asm/esr.h>
 37#include <asm/sysreg.h>
 38#include <asm/system_misc.h>
 39#include <asm/pgtable.h>
 40#include <asm/tlbflush.h>
 41
 42static const char *fault_name(unsigned int esr);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 43
 44/*
 45 * Dump out the page tables associated with 'addr' in mm 'mm'.
 46 */
 47void show_pte(struct mm_struct *mm, unsigned long addr)
 48{
 49	pgd_t *pgd;
 50
 51	if (!mm)
 52		mm = &init_mm;
 53
 54	pr_alert("pgd = %p\n", mm->pgd);
 55	pgd = pgd_offset(mm, addr);
 56	pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
 57
 58	do {
 59		pud_t *pud;
 60		pmd_t *pmd;
 61		pte_t *pte;
 62
 63		if (pgd_none(*pgd) || pgd_bad(*pgd))
 64			break;
 65
 66		pud = pud_offset(pgd, addr);
 67		printk(", *pud=%016llx", pud_val(*pud));
 68		if (pud_none(*pud) || pud_bad(*pud))
 69			break;
 70
 71		pmd = pmd_offset(pud, addr);
 72		printk(", *pmd=%016llx", pmd_val(*pmd));
 73		if (pmd_none(*pmd) || pmd_bad(*pmd))
 74			break;
 75
 76		pte = pte_offset_map(pmd, addr);
 77		printk(", *pte=%016llx", pte_val(*pte));
 78		pte_unmap(pte);
 79	} while(0);
 80
 81	printk("\n");
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 82}
 83
 84/*
 85 * The kernel tried to access some page that wasn't present.
 86 */
 87static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
 88			      unsigned int esr, struct pt_regs *regs)
 89{
 90	/*
 91	 * Are we prepared to handle this kernel fault?
 
 92	 */
 93	if (fixup_exception(regs))
 94		return;
 95
 96	/*
 97	 * No handler, we'll have to terminate things with extreme prejudice.
 98	 */
 99	bust_spinlocks(1);
100	pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
101		 (addr < PAGE_SIZE) ? "NULL pointer dereference" :
102		 "paging request", addr);
103
104	show_pte(mm, addr);
105	die("Oops", regs, esr);
106	bust_spinlocks(0);
107	do_exit(SIGKILL);
108}
109
110/*
111 * Something tried to access memory that isn't in our memory map. User mode
112 * accesses just cause a SIGSEGV
113 */
114static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
115			    unsigned int esr, unsigned int sig, int code,
116			    struct pt_regs *regs)
117{
118	struct siginfo si;
 
119
120	if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
 
121		pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
122			tsk->comm, task_pid_nr(tsk), fault_name(esr), sig,
123			addr, esr);
124		show_pte(tsk->mm, addr);
125		show_regs(regs);
126	}
127
128	tsk->thread.fault_address = addr;
129	tsk->thread.fault_code = esr;
130	si.si_signo = sig;
131	si.si_errno = 0;
132	si.si_code = code;
133	si.si_addr = (void __user *)addr;
134	force_sig_info(sig, &si, tsk);
135}
136
137static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
138{
139	struct task_struct *tsk = current;
140	struct mm_struct *mm = tsk->active_mm;
 
141
142	/*
143	 * If we are in kernel mode at this point, we have no context to
144	 * handle this fault with.
145	 */
146	if (user_mode(regs))
147		__do_user_fault(tsk, addr, esr, SIGSEGV, SEGV_MAPERR, regs);
148	else
 
149		__do_kernel_fault(mm, addr, esr, regs);
150}
151
152#define VM_FAULT_BADMAP		0x010000
153#define VM_FAULT_BADACCESS	0x020000
154
155#define ESR_LNX_EXEC		(1 << 24)
156
157static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
158			   unsigned int mm_flags, unsigned long vm_flags,
159			   struct task_struct *tsk)
160{
161	struct vm_area_struct *vma;
162	int fault;
163
164	vma = find_vma(mm, addr);
165	fault = VM_FAULT_BADMAP;
166	if (unlikely(!vma))
167		goto out;
168	if (unlikely(vma->vm_start > addr))
169		goto check_stack;
170
171	/*
172	 * Ok, we have a good vm_area for this memory access, so we can handle
173	 * it.
174	 */
175good_area:
176	/*
177	 * Check that the permissions on the VMA allow for the fault which
178	 * occurred. If we encountered a write or exec fault, we must have
179	 * appropriate permissions, otherwise we allow any permission.
180	 */
181	if (!(vma->vm_flags & vm_flags)) {
182		fault = VM_FAULT_BADACCESS;
183		goto out;
184	}
185
186	return handle_mm_fault(mm, vma, addr & PAGE_MASK, mm_flags);
187
188check_stack:
189	if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
190		goto good_area;
191out:
192	return fault;
193}
194
195static inline int permission_fault(unsigned int esr)
196{
197	unsigned int ec       = (esr & ESR_ELx_EC_MASK) >> ESR_ELx_EC_SHIFT;
198	unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
199
200	return (ec == ESR_ELx_EC_DABT_CUR && fsc_type == ESR_ELx_FSC_PERM);
 
 
 
 
 
 
 
 
 
 
 
 
201}
202
203static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
204				   struct pt_regs *regs)
205{
206	struct task_struct *tsk;
207	struct mm_struct *mm;
208	int fault, sig, code;
209	unsigned long vm_flags = VM_READ | VM_WRITE | VM_EXEC;
210	unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
211
 
 
 
212	tsk = current;
213	mm  = tsk->mm;
214
215	/* Enable interrupts if they were enabled in the parent context. */
216	if (interrupts_enabled(regs))
217		local_irq_enable();
218
219	/*
220	 * If we're in an interrupt or have no user context, we must not take
221	 * the fault.
222	 */
223	if (faulthandler_disabled() || !mm)
224		goto no_context;
225
226	if (user_mode(regs))
227		mm_flags |= FAULT_FLAG_USER;
228
229	if (esr & ESR_LNX_EXEC) {
230		vm_flags = VM_EXEC;
231	} else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
232		vm_flags = VM_WRITE;
233		mm_flags |= FAULT_FLAG_WRITE;
234	}
235
236	if (permission_fault(esr) && (addr < USER_DS)) {
237		if (get_fs() == KERNEL_DS)
 
238			die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
239
 
 
 
240		if (!search_exception_tables(regs->pc))
241			die("Accessing user space memory outside uaccess.h routines", regs, esr);
242	}
243
244	/*
245	 * As per x86, we may deadlock here. However, since the kernel only
246	 * validly references user space from well defined areas of the code,
247	 * we can bug out early if this is from code which shouldn't.
248	 */
249	if (!down_read_trylock(&mm->mmap_sem)) {
250		if (!user_mode(regs) && !search_exception_tables(regs->pc))
251			goto no_context;
252retry:
253		down_read(&mm->mmap_sem);
254	} else {
255		/*
256		 * The above down_read_trylock() might have succeeded in which
257		 * case, we'll have missed the might_sleep() from down_read().
258		 */
259		might_sleep();
260#ifdef CONFIG_DEBUG_VM
261		if (!user_mode(regs) && !search_exception_tables(regs->pc))
262			goto no_context;
263#endif
264	}
265
266	fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
267
268	/*
269	 * If we need to retry but a fatal signal is pending, handle the
270	 * signal first. We do not need to release the mmap_sem because it
271	 * would already be released in __lock_page_or_retry in mm/filemap.c.
272	 */
273	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
274		return 0;
275
276	/*
277	 * Major/minor page fault accounting is only done on the initial
278	 * attempt. If we go through a retry, it is extremely likely that the
279	 * page will be found in page cache at that point.
280	 */
281
282	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
283	if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
284		if (fault & VM_FAULT_MAJOR) {
285			tsk->maj_flt++;
286			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
287				      addr);
288		} else {
289			tsk->min_flt++;
290			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
291				      addr);
292		}
293		if (fault & VM_FAULT_RETRY) {
294			/*
295			 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
296			 * starvation.
297			 */
298			mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
299			mm_flags |= FAULT_FLAG_TRIED;
300			goto retry;
301		}
302	}
303
304	up_read(&mm->mmap_sem);
305
306	/*
307	 * Handle the "normal" case first - VM_FAULT_MAJOR
308	 */
309	if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
310			      VM_FAULT_BADACCESS))))
311		return 0;
312
313	/*
314	 * If we are in kernel mode at this point, we have no context to
315	 * handle this fault with.
316	 */
317	if (!user_mode(regs))
318		goto no_context;
319
320	if (fault & VM_FAULT_OOM) {
321		/*
322		 * We ran out of memory, call the OOM killer, and return to
323		 * userspace (which will retry the fault, or kill us if we got
324		 * oom-killed).
325		 */
326		pagefault_out_of_memory();
327		return 0;
328	}
329
330	if (fault & VM_FAULT_SIGBUS) {
331		/*
332		 * We had some memory, but were unable to successfully fix up
333		 * this page fault.
334		 */
335		sig = SIGBUS;
336		code = BUS_ADRERR;
337	} else {
338		/*
339		 * Something tried to access memory that isn't in our memory
340		 * map.
341		 */
342		sig = SIGSEGV;
343		code = fault == VM_FAULT_BADACCESS ?
344			SEGV_ACCERR : SEGV_MAPERR;
345	}
346
347	__do_user_fault(tsk, addr, esr, sig, code, regs);
348	return 0;
349
350no_context:
351	__do_kernel_fault(mm, addr, esr, regs);
352	return 0;
353}
354
355/*
356 * First Level Translation Fault Handler
357 *
358 * We enter here because the first level page table doesn't contain a valid
359 * entry for the address.
360 *
361 * If the address is in kernel space (>= TASK_SIZE), then we are probably
362 * faulting in the vmalloc() area.
363 *
364 * If the init_task's first level page tables contains the relevant entry, we
365 * copy the it to this task.  If not, we send the process a signal, fixup the
366 * exception, or oops the kernel.
367 *
368 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
369 * or a critical region, and should only copy the information from the master
370 * page table, nothing more.
371 */
372static int __kprobes do_translation_fault(unsigned long addr,
373					  unsigned int esr,
374					  struct pt_regs *regs)
375{
376	if (addr < TASK_SIZE)
377		return do_page_fault(addr, esr, regs);
378
379	do_bad_area(addr, esr, regs);
380	return 0;
381}
382
383static int do_alignment_fault(unsigned long addr, unsigned int esr,
384			      struct pt_regs *regs)
385{
386	do_bad_area(addr, esr, regs);
387	return 0;
388}
389
390/*
391 * This abort handler always returns "fault".
392 */
393static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
394{
395	return 1;
396}
397
398static struct fault_info {
399	int	(*fn)(unsigned long addr, unsigned int esr, struct pt_regs *regs);
400	int	sig;
401	int	code;
402	const char *name;
403} fault_info[] = {
404	{ do_bad,		SIGBUS,  0,		"ttbr address size fault"	},
405	{ do_bad,		SIGBUS,  0,		"level 1 address size fault"	},
406	{ do_bad,		SIGBUS,  0,		"level 2 address size fault"	},
407	{ do_bad,		SIGBUS,  0,		"level 3 address size fault"	},
408	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 0 translation fault"	},
409	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 1 translation fault"	},
410	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 2 translation fault"	},
411	{ do_page_fault,	SIGSEGV, SEGV_MAPERR,	"level 3 translation fault"	},
412	{ do_bad,		SIGBUS,  0,		"unknown 8"			},
413	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 1 access flag fault"	},
414	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 2 access flag fault"	},
415	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 3 access flag fault"	},
416	{ do_bad,		SIGBUS,  0,		"unknown 12"			},
417	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 1 permission fault"	},
418	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 2 permission fault"	},
419	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 3 permission fault"	},
420	{ do_bad,		SIGBUS,  0,		"synchronous external abort"	},
421	{ do_bad,		SIGBUS,  0,		"unknown 17"			},
422	{ do_bad,		SIGBUS,  0,		"unknown 18"			},
423	{ do_bad,		SIGBUS,  0,		"unknown 19"			},
424	{ do_bad,		SIGBUS,  0,		"synchronous abort (translation table walk)" },
425	{ do_bad,		SIGBUS,  0,		"synchronous abort (translation table walk)" },
426	{ do_bad,		SIGBUS,  0,		"synchronous abort (translation table walk)" },
427	{ do_bad,		SIGBUS,  0,		"synchronous abort (translation table walk)" },
428	{ do_bad,		SIGBUS,  0,		"synchronous parity error"	},
429	{ do_bad,		SIGBUS,  0,		"unknown 25"			},
430	{ do_bad,		SIGBUS,  0,		"unknown 26"			},
431	{ do_bad,		SIGBUS,  0,		"unknown 27"			},
432	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
433	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
434	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
435	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
436	{ do_bad,		SIGBUS,  0,		"unknown 32"			},
437	{ do_alignment_fault,	SIGBUS,  BUS_ADRALN,	"alignment fault"		},
438	{ do_bad,		SIGBUS,  0,		"unknown 34"			},
439	{ do_bad,		SIGBUS,  0,		"unknown 35"			},
440	{ do_bad,		SIGBUS,  0,		"unknown 36"			},
441	{ do_bad,		SIGBUS,  0,		"unknown 37"			},
442	{ do_bad,		SIGBUS,  0,		"unknown 38"			},
443	{ do_bad,		SIGBUS,  0,		"unknown 39"			},
444	{ do_bad,		SIGBUS,  0,		"unknown 40"			},
445	{ do_bad,		SIGBUS,  0,		"unknown 41"			},
446	{ do_bad,		SIGBUS,  0,		"unknown 42"			},
447	{ do_bad,		SIGBUS,  0,		"unknown 43"			},
448	{ do_bad,		SIGBUS,  0,		"unknown 44"			},
449	{ do_bad,		SIGBUS,  0,		"unknown 45"			},
450	{ do_bad,		SIGBUS,  0,		"unknown 46"			},
451	{ do_bad,		SIGBUS,  0,		"unknown 47"			},
452	{ do_bad,		SIGBUS,  0,		"TLB conflict abort"		},
453	{ do_bad,		SIGBUS,  0,		"unknown 49"			},
454	{ do_bad,		SIGBUS,  0,		"unknown 50"			},
455	{ do_bad,		SIGBUS,  0,		"unknown 51"			},
456	{ do_bad,		SIGBUS,  0,		"implementation fault (lockdown abort)" },
457	{ do_bad,		SIGBUS,  0,		"implementation fault (unsupported exclusive)" },
458	{ do_bad,		SIGBUS,  0,		"unknown 54"			},
459	{ do_bad,		SIGBUS,  0,		"unknown 55"			},
460	{ do_bad,		SIGBUS,  0,		"unknown 56"			},
461	{ do_bad,		SIGBUS,  0,		"unknown 57"			},
462	{ do_bad,		SIGBUS,  0,		"unknown 58" 			},
463	{ do_bad,		SIGBUS,  0,		"unknown 59"			},
464	{ do_bad,		SIGBUS,  0,		"unknown 60"			},
465	{ do_bad,		SIGBUS,  0,		"section domain fault"		},
466	{ do_bad,		SIGBUS,  0,		"page domain fault"		},
467	{ do_bad,		SIGBUS,  0,		"unknown 63"			},
468};
469
470static const char *fault_name(unsigned int esr)
471{
472	const struct fault_info *inf = fault_info + (esr & 63);
473	return inf->name;
474}
475
476/*
477 * Dispatch a data abort to the relevant handler.
478 */
479asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
480					 struct pt_regs *regs)
481{
482	const struct fault_info *inf = fault_info + (esr & 63);
483	struct siginfo info;
484
485	if (!inf->fn(addr, esr, regs))
486		return;
487
488	pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
489		 inf->name, esr, addr);
490
491	info.si_signo = inf->sig;
492	info.si_errno = 0;
493	info.si_code  = inf->code;
494	info.si_addr  = (void __user *)addr;
495	arm64_notify_die("", regs, &info, esr);
496}
497
498/*
499 * Handle stack alignment exceptions.
500 */
501asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
502					   unsigned int esr,
503					   struct pt_regs *regs)
504{
505	struct siginfo info;
506	struct task_struct *tsk = current;
507
508	if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
509		pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
510				    tsk->comm, task_pid_nr(tsk),
511				    esr_get_class_string(esr), (void *)regs->pc,
512				    (void *)regs->sp);
513
514	info.si_signo = SIGBUS;
515	info.si_errno = 0;
516	info.si_code  = BUS_ADRALN;
517	info.si_addr  = (void __user *)addr;
518	arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
519}
520
521int __init early_brk64(unsigned long addr, unsigned int esr,
522		       struct pt_regs *regs);
523
524/*
525 * __refdata because early_brk64 is __init, but the reference to it is
526 * clobbered at arch_initcall time.
527 * See traps.c and debug-monitors.c:debug_traps_init().
528 */
529static struct fault_info __refdata debug_fault_info[] = {
530	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware breakpoint"	},
531	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware single-step"	},
532	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware watchpoint"	},
533	{ do_bad,	SIGBUS,		0,		"unknown 3"		},
534	{ do_bad,	SIGTRAP,	TRAP_BRKPT,	"aarch32 BKPT"		},
535	{ do_bad,	SIGTRAP,	0,		"aarch32 vector catch"	},
536	{ early_brk64,	SIGTRAP,	TRAP_BRKPT,	"aarch64 BRK"		},
537	{ do_bad,	SIGBUS,		0,		"unknown 7"		},
538};
539
540void __init hook_debug_fault_code(int nr,
541				  int (*fn)(unsigned long, unsigned int, struct pt_regs *),
542				  int sig, int code, const char *name)
543{
544	BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
545
546	debug_fault_info[nr].fn		= fn;
547	debug_fault_info[nr].sig	= sig;
548	debug_fault_info[nr].code	= code;
549	debug_fault_info[nr].name	= name;
550}
551
552asmlinkage int __exception do_debug_exception(unsigned long addr,
553					      unsigned int esr,
554					      struct pt_regs *regs)
555{
556	const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
557	struct siginfo info;
 
558
559	if (!inf->fn(addr, esr, regs))
560		return 1;
 
 
 
 
561
562	pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
563		 inf->name, esr, addr);
 
 
 
564
565	info.si_signo = inf->sig;
566	info.si_errno = 0;
567	info.si_code  = inf->code;
568	info.si_addr  = (void __user *)addr;
569	arm64_notify_die("", regs, &info, 0);
 
 
570
571	return 0;
 
 
 
572}
 
573
574#ifdef CONFIG_ARM64_PAN
575void cpu_enable_pan(void *__unused)
576{
 
 
 
 
 
 
577	config_sctlr_el1(SCTLR_EL1_SPAN, 0);
 
 
578}
579#endif /* CONFIG_ARM64_PAN */
580
581#ifdef CONFIG_ARM64_UAO
582/*
583 * Kernel threads have fs=KERNEL_DS by default, and don't need to call
584 * set_fs(), devtmpfs in particular relies on this behaviour.
585 * We need to enable the feature at runtime (instead of adding it to
586 * PSR_MODE_EL1h) as the feature may not be implemented by the cpu.
587 */
588void cpu_enable_uao(void *__unused)
589{
590	asm(SET_PSTATE_UAO(1));
 
591}
592#endif /* CONFIG_ARM64_UAO */
v4.10.11
  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.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
 44struct fault_info {
 45	int	(*fn)(unsigned long addr, unsigned int esr,
 46		      struct pt_regs *regs);
 47	int	sig;
 48	int	code;
 49	const char *name;
 50};
 51
 52static const struct fault_info fault_info[];
 53
 54static inline const struct fault_info *esr_to_fault_info(unsigned int esr)
 55{
 56	return fault_info + (esr & 63);
 57}
 58
 59#ifdef CONFIG_KPROBES
 60static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
 61{
 62	int ret = 0;
 63
 64	/* kprobe_running() needs smp_processor_id() */
 65	if (!user_mode(regs)) {
 66		preempt_disable();
 67		if (kprobe_running() && kprobe_fault_handler(regs, esr))
 68			ret = 1;
 69		preempt_enable();
 70	}
 71
 72	return ret;
 73}
 74#else
 75static inline int notify_page_fault(struct pt_regs *regs, unsigned int esr)
 76{
 77	return 0;
 78}
 79#endif
 80
 81/*
 82 * Dump out the page tables associated with 'addr' in mm 'mm'.
 83 */
 84void show_pte(struct mm_struct *mm, unsigned long addr)
 85{
 86	pgd_t *pgd;
 87
 88	if (!mm)
 89		mm = &init_mm;
 90
 91	pr_alert("pgd = %p\n", mm->pgd);
 92	pgd = pgd_offset(mm, addr);
 93	pr_alert("[%08lx] *pgd=%016llx", addr, pgd_val(*pgd));
 94
 95	do {
 96		pud_t *pud;
 97		pmd_t *pmd;
 98		pte_t *pte;
 99
100		if (pgd_none(*pgd) || pgd_bad(*pgd))
101			break;
102
103		pud = pud_offset(pgd, addr);
104		pr_cont(", *pud=%016llx", pud_val(*pud));
105		if (pud_none(*pud) || pud_bad(*pud))
106			break;
107
108		pmd = pmd_offset(pud, addr);
109		pr_cont(", *pmd=%016llx", pmd_val(*pmd));
110		if (pmd_none(*pmd) || pmd_bad(*pmd))
111			break;
112
113		pte = pte_offset_map(pmd, addr);
114		pr_cont(", *pte=%016llx", pte_val(*pte));
115		pte_unmap(pte);
116	} while(0);
117
118	pr_cont("\n");
119}
120
121#ifdef CONFIG_ARM64_HW_AFDBM
122/*
123 * This function sets the access flags (dirty, accessed), as well as write
124 * permission, and only to a more permissive setting.
125 *
126 * It needs to cope with hardware update of the accessed/dirty state by other
127 * agents in the system and can safely skip the __sync_icache_dcache() call as,
128 * like set_pte_at(), the PTE is never changed from no-exec to exec here.
129 *
130 * Returns whether or not the PTE actually changed.
131 */
132int ptep_set_access_flags(struct vm_area_struct *vma,
133			  unsigned long address, pte_t *ptep,
134			  pte_t entry, int dirty)
135{
136	pteval_t old_pteval;
137	unsigned int tmp;
138
139	if (pte_same(*ptep, entry))
140		return 0;
141
142	/* only preserve the access flags and write permission */
143	pte_val(entry) &= PTE_AF | PTE_WRITE | PTE_DIRTY;
144
145	/*
146	 * PTE_RDONLY is cleared by default in the asm below, so set it in
147	 * back if necessary (read-only or clean PTE).
148	 */
149	if (!pte_write(entry) || !pte_sw_dirty(entry))
150		pte_val(entry) |= PTE_RDONLY;
151
152	/*
153	 * Setting the flags must be done atomically to avoid racing with the
154	 * hardware update of the access/dirty state.
155	 */
156	asm volatile("//	ptep_set_access_flags\n"
157	"	prfm	pstl1strm, %2\n"
158	"1:	ldxr	%0, %2\n"
159	"	and	%0, %0, %3		// clear PTE_RDONLY\n"
160	"	orr	%0, %0, %4		// set flags\n"
161	"	stxr	%w1, %0, %2\n"
162	"	cbnz	%w1, 1b\n"
163	: "=&r" (old_pteval), "=&r" (tmp), "+Q" (pte_val(*ptep))
164	: "L" (~PTE_RDONLY), "r" (pte_val(entry)));
165
166	flush_tlb_fix_spurious_fault(vma, address);
167	return 1;
168}
169#endif
170
171static bool is_el1_instruction_abort(unsigned int esr)
172{
173	return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_CUR;
174}
175
176/*
177 * The kernel tried to access some page that wasn't present.
178 */
179static void __do_kernel_fault(struct mm_struct *mm, unsigned long addr,
180			      unsigned int esr, struct pt_regs *regs)
181{
182	/*
183	 * Are we prepared to handle this kernel fault?
184	 * We are almost certainly not prepared to handle instruction faults.
185	 */
186	if (!is_el1_instruction_abort(esr) && fixup_exception(regs))
187		return;
188
189	/*
190	 * No handler, we'll have to terminate things with extreme prejudice.
191	 */
192	bust_spinlocks(1);
193	pr_alert("Unable to handle kernel %s at virtual address %08lx\n",
194		 (addr < PAGE_SIZE) ? "NULL pointer dereference" :
195		 "paging request", addr);
196
197	show_pte(mm, addr);
198	die("Oops", regs, esr);
199	bust_spinlocks(0);
200	do_exit(SIGKILL);
201}
202
203/*
204 * Something tried to access memory that isn't in our memory map. User mode
205 * accesses just cause a SIGSEGV
206 */
207static void __do_user_fault(struct task_struct *tsk, unsigned long addr,
208			    unsigned int esr, unsigned int sig, int code,
209			    struct pt_regs *regs)
210{
211	struct siginfo si;
212	const struct fault_info *inf;
213
214	if (unhandled_signal(tsk, sig) && show_unhandled_signals_ratelimited()) {
215		inf = esr_to_fault_info(esr);
216		pr_info("%s[%d]: unhandled %s (%d) at 0x%08lx, esr 0x%03x\n",
217			tsk->comm, task_pid_nr(tsk), inf->name, sig,
218			addr, esr);
219		show_pte(tsk->mm, addr);
220		show_regs(regs);
221	}
222
223	tsk->thread.fault_address = addr;
224	tsk->thread.fault_code = esr;
225	si.si_signo = sig;
226	si.si_errno = 0;
227	si.si_code = code;
228	si.si_addr = (void __user *)addr;
229	force_sig_info(sig, &si, tsk);
230}
231
232static void do_bad_area(unsigned long addr, unsigned int esr, struct pt_regs *regs)
233{
234	struct task_struct *tsk = current;
235	struct mm_struct *mm = tsk->active_mm;
236	const struct fault_info *inf;
237
238	/*
239	 * If we are in kernel mode at this point, we have no context to
240	 * handle this fault with.
241	 */
242	if (user_mode(regs)) {
243		inf = esr_to_fault_info(esr);
244		__do_user_fault(tsk, addr, esr, inf->sig, inf->code, regs);
245	} else
246		__do_kernel_fault(mm, addr, esr, regs);
247}
248
249#define VM_FAULT_BADMAP		0x010000
250#define VM_FAULT_BADACCESS	0x020000
251
 
 
252static int __do_page_fault(struct mm_struct *mm, unsigned long addr,
253			   unsigned int mm_flags, unsigned long vm_flags,
254			   struct task_struct *tsk)
255{
256	struct vm_area_struct *vma;
257	int fault;
258
259	vma = find_vma(mm, addr);
260	fault = VM_FAULT_BADMAP;
261	if (unlikely(!vma))
262		goto out;
263	if (unlikely(vma->vm_start > addr))
264		goto check_stack;
265
266	/*
267	 * Ok, we have a good vm_area for this memory access, so we can handle
268	 * it.
269	 */
270good_area:
271	/*
272	 * Check that the permissions on the VMA allow for the fault which
273	 * occurred.
 
274	 */
275	if (!(vma->vm_flags & vm_flags)) {
276		fault = VM_FAULT_BADACCESS;
277		goto out;
278	}
279
280	return handle_mm_fault(vma, addr & PAGE_MASK, mm_flags);
281
282check_stack:
283	if (vma->vm_flags & VM_GROWSDOWN && !expand_stack(vma, addr))
284		goto good_area;
285out:
286	return fault;
287}
288
289static inline bool is_permission_fault(unsigned int esr, struct pt_regs *regs)
290{
291	unsigned int ec       = ESR_ELx_EC(esr);
292	unsigned int fsc_type = esr & ESR_ELx_FSC_TYPE;
293
294	if (ec != ESR_ELx_EC_DABT_CUR && ec != ESR_ELx_EC_IABT_CUR)
295		return false;
296
297	if (system_uses_ttbr0_pan())
298		return fsc_type == ESR_ELx_FSC_FAULT &&
299			(regs->pstate & PSR_PAN_BIT);
300	else
301		return fsc_type == ESR_ELx_FSC_PERM;
302}
303
304static bool is_el0_instruction_abort(unsigned int esr)
305{
306	return ESR_ELx_EC(esr) == ESR_ELx_EC_IABT_LOW;
307}
308
309static int __kprobes do_page_fault(unsigned long addr, unsigned int esr,
310				   struct pt_regs *regs)
311{
312	struct task_struct *tsk;
313	struct mm_struct *mm;
314	int fault, sig, code;
315	unsigned long vm_flags = VM_READ | VM_WRITE;
316	unsigned int mm_flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
317
318	if (notify_page_fault(regs, esr))
319		return 0;
320
321	tsk = current;
322	mm  = tsk->mm;
323
 
 
 
 
324	/*
325	 * If we're in an interrupt or have no user context, we must not take
326	 * the fault.
327	 */
328	if (faulthandler_disabled() || !mm)
329		goto no_context;
330
331	if (user_mode(regs))
332		mm_flags |= FAULT_FLAG_USER;
333
334	if (is_el0_instruction_abort(esr)) {
335		vm_flags = VM_EXEC;
336	} else if ((esr & ESR_ELx_WNR) && !(esr & ESR_ELx_CM)) {
337		vm_flags = VM_WRITE;
338		mm_flags |= FAULT_FLAG_WRITE;
339	}
340
341	if (addr < USER_DS && is_permission_fault(esr, regs)) {
342		/* regs->orig_addr_limit may be 0 if we entered from EL0 */
343		if (regs->orig_addr_limit == KERNEL_DS)
344			die("Accessing user space memory with fs=KERNEL_DS", regs, esr);
345
346		if (is_el1_instruction_abort(esr))
347			die("Attempting to execute userspace memory", regs, esr);
348
349		if (!search_exception_tables(regs->pc))
350			die("Accessing user space memory outside uaccess.h routines", regs, esr);
351	}
352
353	/*
354	 * As per x86, we may deadlock here. However, since the kernel only
355	 * validly references user space from well defined areas of the code,
356	 * we can bug out early if this is from code which shouldn't.
357	 */
358	if (!down_read_trylock(&mm->mmap_sem)) {
359		if (!user_mode(regs) && !search_exception_tables(regs->pc))
360			goto no_context;
361retry:
362		down_read(&mm->mmap_sem);
363	} else {
364		/*
365		 * The above down_read_trylock() might have succeeded in which
366		 * case, we'll have missed the might_sleep() from down_read().
367		 */
368		might_sleep();
369#ifdef CONFIG_DEBUG_VM
370		if (!user_mode(regs) && !search_exception_tables(regs->pc))
371			goto no_context;
372#endif
373	}
374
375	fault = __do_page_fault(mm, addr, mm_flags, vm_flags, tsk);
376
377	/*
378	 * If we need to retry but a fatal signal is pending, handle the
379	 * signal first. We do not need to release the mmap_sem because it
380	 * would already be released in __lock_page_or_retry in mm/filemap.c.
381	 */
382	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
383		return 0;
384
385	/*
386	 * Major/minor page fault accounting is only done on the initial
387	 * attempt. If we go through a retry, it is extremely likely that the
388	 * page will be found in page cache at that point.
389	 */
390
391	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, addr);
392	if (mm_flags & FAULT_FLAG_ALLOW_RETRY) {
393		if (fault & VM_FAULT_MAJOR) {
394			tsk->maj_flt++;
395			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, regs,
396				      addr);
397		} else {
398			tsk->min_flt++;
399			perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, regs,
400				      addr);
401		}
402		if (fault & VM_FAULT_RETRY) {
403			/*
404			 * Clear FAULT_FLAG_ALLOW_RETRY to avoid any risk of
405			 * starvation.
406			 */
407			mm_flags &= ~FAULT_FLAG_ALLOW_RETRY;
408			mm_flags |= FAULT_FLAG_TRIED;
409			goto retry;
410		}
411	}
412
413	up_read(&mm->mmap_sem);
414
415	/*
416	 * Handle the "normal" case first - VM_FAULT_MAJOR
417	 */
418	if (likely(!(fault & (VM_FAULT_ERROR | VM_FAULT_BADMAP |
419			      VM_FAULT_BADACCESS))))
420		return 0;
421
422	/*
423	 * If we are in kernel mode at this point, we have no context to
424	 * handle this fault with.
425	 */
426	if (!user_mode(regs))
427		goto no_context;
428
429	if (fault & VM_FAULT_OOM) {
430		/*
431		 * We ran out of memory, call the OOM killer, and return to
432		 * userspace (which will retry the fault, or kill us if we got
433		 * oom-killed).
434		 */
435		pagefault_out_of_memory();
436		return 0;
437	}
438
439	if (fault & VM_FAULT_SIGBUS) {
440		/*
441		 * We had some memory, but were unable to successfully fix up
442		 * this page fault.
443		 */
444		sig = SIGBUS;
445		code = BUS_ADRERR;
446	} else {
447		/*
448		 * Something tried to access memory that isn't in our memory
449		 * map.
450		 */
451		sig = SIGSEGV;
452		code = fault == VM_FAULT_BADACCESS ?
453			SEGV_ACCERR : SEGV_MAPERR;
454	}
455
456	__do_user_fault(tsk, addr, esr, sig, code, regs);
457	return 0;
458
459no_context:
460	__do_kernel_fault(mm, addr, esr, regs);
461	return 0;
462}
463
464/*
465 * First Level Translation Fault Handler
466 *
467 * We enter here because the first level page table doesn't contain a valid
468 * entry for the address.
469 *
470 * If the address is in kernel space (>= TASK_SIZE), then we are probably
471 * faulting in the vmalloc() area.
472 *
473 * If the init_task's first level page tables contains the relevant entry, we
474 * copy the it to this task.  If not, we send the process a signal, fixup the
475 * exception, or oops the kernel.
476 *
477 * NOTE! We MUST NOT take any locks for this case. We may be in an interrupt
478 * or a critical region, and should only copy the information from the master
479 * page table, nothing more.
480 */
481static int __kprobes do_translation_fault(unsigned long addr,
482					  unsigned int esr,
483					  struct pt_regs *regs)
484{
485	if (addr < TASK_SIZE)
486		return do_page_fault(addr, esr, regs);
487
488	do_bad_area(addr, esr, regs);
489	return 0;
490}
491
492static int do_alignment_fault(unsigned long addr, unsigned int esr,
493			      struct pt_regs *regs)
494{
495	do_bad_area(addr, esr, regs);
496	return 0;
497}
498
499/*
500 * This abort handler always returns "fault".
501 */
502static int do_bad(unsigned long addr, unsigned int esr, struct pt_regs *regs)
503{
504	return 1;
505}
506
507static const struct fault_info fault_info[] = {
 
 
 
 
 
508	{ do_bad,		SIGBUS,  0,		"ttbr address size fault"	},
509	{ do_bad,		SIGBUS,  0,		"level 1 address size fault"	},
510	{ do_bad,		SIGBUS,  0,		"level 2 address size fault"	},
511	{ do_bad,		SIGBUS,  0,		"level 3 address size fault"	},
512	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 0 translation fault"	},
513	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 1 translation fault"	},
514	{ do_translation_fault,	SIGSEGV, SEGV_MAPERR,	"level 2 translation fault"	},
515	{ do_page_fault,	SIGSEGV, SEGV_MAPERR,	"level 3 translation fault"	},
516	{ do_bad,		SIGBUS,  0,		"unknown 8"			},
517	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 1 access flag fault"	},
518	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 2 access flag fault"	},
519	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 3 access flag fault"	},
520	{ do_bad,		SIGBUS,  0,		"unknown 12"			},
521	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 1 permission fault"	},
522	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 2 permission fault"	},
523	{ do_page_fault,	SIGSEGV, SEGV_ACCERR,	"level 3 permission fault"	},
524	{ do_bad,		SIGBUS,  0,		"synchronous external abort"	},
525	{ do_bad,		SIGBUS,  0,		"unknown 17"			},
526	{ do_bad,		SIGBUS,  0,		"unknown 18"			},
527	{ do_bad,		SIGBUS,  0,		"unknown 19"			},
528	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
529	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
530	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
531	{ do_bad,		SIGBUS,  0,		"synchronous external abort (translation table walk)" },
532	{ do_bad,		SIGBUS,  0,		"synchronous parity error"	},
533	{ do_bad,		SIGBUS,  0,		"unknown 25"			},
534	{ do_bad,		SIGBUS,  0,		"unknown 26"			},
535	{ do_bad,		SIGBUS,  0,		"unknown 27"			},
536	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
537	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
538	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
539	{ do_bad,		SIGBUS,  0,		"synchronous parity error (translation table walk)" },
540	{ do_bad,		SIGBUS,  0,		"unknown 32"			},
541	{ do_alignment_fault,	SIGBUS,  BUS_ADRALN,	"alignment fault"		},
542	{ do_bad,		SIGBUS,  0,		"unknown 34"			},
543	{ do_bad,		SIGBUS,  0,		"unknown 35"			},
544	{ do_bad,		SIGBUS,  0,		"unknown 36"			},
545	{ do_bad,		SIGBUS,  0,		"unknown 37"			},
546	{ do_bad,		SIGBUS,  0,		"unknown 38"			},
547	{ do_bad,		SIGBUS,  0,		"unknown 39"			},
548	{ do_bad,		SIGBUS,  0,		"unknown 40"			},
549	{ do_bad,		SIGBUS,  0,		"unknown 41"			},
550	{ do_bad,		SIGBUS,  0,		"unknown 42"			},
551	{ do_bad,		SIGBUS,  0,		"unknown 43"			},
552	{ do_bad,		SIGBUS,  0,		"unknown 44"			},
553	{ do_bad,		SIGBUS,  0,		"unknown 45"			},
554	{ do_bad,		SIGBUS,  0,		"unknown 46"			},
555	{ do_bad,		SIGBUS,  0,		"unknown 47"			},
556	{ do_bad,		SIGBUS,  0,		"TLB conflict abort"		},
557	{ do_bad,		SIGBUS,  0,		"unknown 49"			},
558	{ do_bad,		SIGBUS,  0,		"unknown 50"			},
559	{ do_bad,		SIGBUS,  0,		"unknown 51"			},
560	{ do_bad,		SIGBUS,  0,		"implementation fault (lockdown abort)" },
561	{ do_bad,		SIGBUS,  0,		"implementation fault (unsupported exclusive)" },
562	{ do_bad,		SIGBUS,  0,		"unknown 54"			},
563	{ do_bad,		SIGBUS,  0,		"unknown 55"			},
564	{ do_bad,		SIGBUS,  0,		"unknown 56"			},
565	{ do_bad,		SIGBUS,  0,		"unknown 57"			},
566	{ do_bad,		SIGBUS,  0,		"unknown 58" 			},
567	{ do_bad,		SIGBUS,  0,		"unknown 59"			},
568	{ do_bad,		SIGBUS,  0,		"unknown 60"			},
569	{ do_bad,		SIGBUS,  0,		"section domain fault"		},
570	{ do_bad,		SIGBUS,  0,		"page domain fault"		},
571	{ do_bad,		SIGBUS,  0,		"unknown 63"			},
572};
573
 
 
 
 
 
 
574/*
575 * Dispatch a data abort to the relevant handler.
576 */
577asmlinkage void __exception do_mem_abort(unsigned long addr, unsigned int esr,
578					 struct pt_regs *regs)
579{
580	const struct fault_info *inf = esr_to_fault_info(esr);
581	struct siginfo info;
582
583	if (!inf->fn(addr, esr, regs))
584		return;
585
586	pr_alert("Unhandled fault: %s (0x%08x) at 0x%016lx\n",
587		 inf->name, esr, addr);
588
589	info.si_signo = inf->sig;
590	info.si_errno = 0;
591	info.si_code  = inf->code;
592	info.si_addr  = (void __user *)addr;
593	arm64_notify_die("", regs, &info, esr);
594}
595
596/*
597 * Handle stack alignment exceptions.
598 */
599asmlinkage void __exception do_sp_pc_abort(unsigned long addr,
600					   unsigned int esr,
601					   struct pt_regs *regs)
602{
603	struct siginfo info;
604	struct task_struct *tsk = current;
605
606	if (show_unhandled_signals && unhandled_signal(tsk, SIGBUS))
607		pr_info_ratelimited("%s[%d]: %s exception: pc=%p sp=%p\n",
608				    tsk->comm, task_pid_nr(tsk),
609				    esr_get_class_string(esr), (void *)regs->pc,
610				    (void *)regs->sp);
611
612	info.si_signo = SIGBUS;
613	info.si_errno = 0;
614	info.si_code  = BUS_ADRALN;
615	info.si_addr  = (void __user *)addr;
616	arm64_notify_die("Oops - SP/PC alignment exception", regs, &info, esr);
617}
618
619int __init early_brk64(unsigned long addr, unsigned int esr,
620		       struct pt_regs *regs);
621
622/*
623 * __refdata because early_brk64 is __init, but the reference to it is
624 * clobbered at arch_initcall time.
625 * See traps.c and debug-monitors.c:debug_traps_init().
626 */
627static struct fault_info __refdata debug_fault_info[] = {
628	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware breakpoint"	},
629	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware single-step"	},
630	{ do_bad,	SIGTRAP,	TRAP_HWBKPT,	"hardware watchpoint"	},
631	{ do_bad,	SIGBUS,		0,		"unknown 3"		},
632	{ do_bad,	SIGTRAP,	TRAP_BRKPT,	"aarch32 BKPT"		},
633	{ do_bad,	SIGTRAP,	0,		"aarch32 vector catch"	},
634	{ early_brk64,	SIGTRAP,	TRAP_BRKPT,	"aarch64 BRK"		},
635	{ do_bad,	SIGBUS,		0,		"unknown 7"		},
636};
637
638void __init hook_debug_fault_code(int nr,
639				  int (*fn)(unsigned long, unsigned int, struct pt_regs *),
640				  int sig, int code, const char *name)
641{
642	BUG_ON(nr < 0 || nr >= ARRAY_SIZE(debug_fault_info));
643
644	debug_fault_info[nr].fn		= fn;
645	debug_fault_info[nr].sig	= sig;
646	debug_fault_info[nr].code	= code;
647	debug_fault_info[nr].name	= name;
648}
649
650asmlinkage int __exception do_debug_exception(unsigned long addr,
651					      unsigned int esr,
652					      struct pt_regs *regs)
653{
654	const struct fault_info *inf = debug_fault_info + DBG_ESR_EVT(esr);
655	struct siginfo info;
656	int rv;
657
658	/*
659	 * Tell lockdep we disabled irqs in entry.S. Do nothing if they were
660	 * already disabled to preserve the last enabled/disabled addresses.
661	 */
662	if (interrupts_enabled(regs))
663		trace_hardirqs_off();
664
665	if (!inf->fn(addr, esr, regs)) {
666		rv = 1;
667	} else {
668		pr_alert("Unhandled debug exception: %s (0x%08x) at 0x%016lx\n",
669			 inf->name, esr, addr);
670
671		info.si_signo = inf->sig;
672		info.si_errno = 0;
673		info.si_code  = inf->code;
674		info.si_addr  = (void __user *)addr;
675		arm64_notify_die("", regs, &info, 0);
676		rv = 0;
677	}
678
679	if (interrupts_enabled(regs))
680		trace_hardirqs_on();
681
682	return rv;
683}
684NOKPROBE_SYMBOL(do_debug_exception);
685
686#ifdef CONFIG_ARM64_PAN
687int cpu_enable_pan(void *__unused)
688{
689	/*
690	 * We modify PSTATE. This won't work from irq context as the PSTATE
691	 * is discarded once we return from the exception.
692	 */
693	WARN_ON_ONCE(in_interrupt());
694
695	config_sctlr_el1(SCTLR_EL1_SPAN, 0);
696	asm(SET_PSTATE_PAN(1));
697	return 0;
698}
699#endif /* CONFIG_ARM64_PAN */
700
701#ifdef CONFIG_ARM64_UAO
702/*
703 * Kernel threads have fs=KERNEL_DS by default, and don't need to call
704 * set_fs(), devtmpfs in particular relies on this behaviour.
705 * We need to enable the feature at runtime (instead of adding it to
706 * PSR_MODE_EL1h) as the feature may not be implemented by the cpu.
707 */
708int cpu_enable_uao(void *__unused)
709{
710	asm(SET_PSTATE_UAO(1));
711	return 0;
712}
713#endif /* CONFIG_ARM64_UAO */