blob: 1dd3f533d78c869a6c5bbb2ef38d7bb9a2370915 [file] [log] [blame]
Neil Horman878719e2008-10-23 10:40:06 -04001/*
2 * Copyright (C) 1991, 1992 Linus Torvalds
3 * Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
4 */
5#include <linux/kallsyms.h>
6#include <linux/kprobes.h>
7#include <linux/uaccess.h>
8#include <linux/utsname.h>
9#include <linux/hardirq.h>
10#include <linux/kdebug.h>
11#include <linux/module.h>
12#include <linux/ptrace.h>
Ingo Molnarb17b0152017-02-08 18:51:35 +010013#include <linux/sched/debug.h>
Ingo Molnar68db0cf2017-02-08 18:51:37 +010014#include <linux/sched/task_stack.h>
Steven Rostedt712406a2009-02-09 10:54:03 -080015#include <linux/ftrace.h>
Neil Horman878719e2008-10-23 10:40:06 -040016#include <linux/kexec.h>
17#include <linux/bug.h>
18#include <linux/nmi.h>
19#include <linux/sysfs.h>
20
21#include <asm/stacktrace.h>
Josh Poimboeufe18bccc2016-09-16 14:18:16 -050022#include <asm/unwind.h>
Neil Horman878719e2008-10-23 10:40:06 -040023
24int panic_on_unrecovered_nmi;
Kurt Garloff5211a242009-06-24 14:32:11 -070025int panic_on_io_nmi;
Neil Horman878719e2008-10-23 10:40:06 -040026unsigned int code_bytes = 64;
Neil Horman878719e2008-10-23 10:40:06 -040027static int die_counter;
28
Josh Poimboeufcb76c932016-09-14 21:07:42 -050029bool in_task_stack(unsigned long *stack, struct task_struct *task,
30 struct stack_info *info)
31{
32 unsigned long *begin = task_stack_page(task);
33 unsigned long *end = task_stack_page(task) + THREAD_SIZE;
34
35 if (stack < begin || stack >= end)
36 return false;
37
38 info->type = STACK_TYPE_TASK;
39 info->begin = begin;
40 info->end = end;
41 info->next_sp = NULL;
42
43 return true;
44}
45
Dave Hansen4fe2d8b2017-12-04 17:25:07 -080046bool in_entry_stack(unsigned long *stack, struct stack_info *info)
Andy Lutomirski33a2f1a2017-12-04 15:07:13 +010047{
Dave Hansen4fe2d8b2017-12-04 17:25:07 -080048 struct entry_stack *ss = cpu_entry_stack(smp_processor_id());
Andy Lutomirski33a2f1a2017-12-04 15:07:13 +010049
Andy Lutomirski0f9a4812017-12-04 15:07:28 +010050 void *begin = ss;
51 void *end = ss + 1;
Andy Lutomirski33a2f1a2017-12-04 15:07:13 +010052
53 if ((void *)stack < begin || (void *)stack >= end)
54 return false;
55
Dave Hansen4fe2d8b2017-12-04 17:25:07 -080056 info->type = STACK_TYPE_ENTRY;
Andy Lutomirski33a2f1a2017-12-04 15:07:13 +010057 info->begin = begin;
58 info->end = end;
59 info->next_sp = NULL;
60
61 return true;
62}
63
Adrien Schildknecht1fc7f612015-02-20 03:34:21 +010064static void printk_stack_address(unsigned long address, int reliable,
Josh Poimboeufd438f5f2016-08-24 11:50:16 -050065 char *log_lvl)
Neil Horman878719e2008-10-23 10:40:06 -040066{
Josh Poimboeufd438f5f2016-08-24 11:50:16 -050067 touch_nmi_watchdog();
Josh Poimboeufbb5e5ce2016-10-25 09:51:12 -050068 printk("%s %s%pB\n", log_lvl, reliable ? "" : "? ", (void *)address);
Jiri Slaby5f01c982013-10-25 15:06:58 +020069}
70
Josh Poimboeufb02fcf92017-12-04 15:07:09 +010071void show_iret_regs(struct pt_regs *regs)
72{
73 printk(KERN_DEFAULT "RIP: %04x:%pS\n", (int)regs->cs, (void *)regs->ip);
74 printk(KERN_DEFAULT "RSP: %04x:%016lx EFLAGS: %08lx", (int)regs->ss,
75 regs->sp, regs->flags);
76}
77
78static void show_regs_safe(struct stack_info *info, struct pt_regs *regs)
79{
80 if (on_stack(info, regs, sizeof(*regs)))
81 __show_regs(regs, 0);
82 else if (on_stack(info, (void *)regs + IRET_FRAME_OFFSET,
83 IRET_FRAME_SIZE)) {
84 /*
85 * When an interrupt or exception occurs in entry code, the
86 * full pt_regs might not have been saved yet. In that case
87 * just print the iret frame.
88 */
89 show_iret_regs(regs);
90 }
91}
92
Josh Poimboeufe18bccc2016-09-16 14:18:16 -050093void show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
94 unsigned long *stack, char *log_lvl)
Neil Horman878719e2008-10-23 10:40:06 -040095{
Josh Poimboeufe18bccc2016-09-16 14:18:16 -050096 struct unwind_state state;
97 struct stack_info stack_info = {0};
98 unsigned long visit_mask = 0;
99 int graph_idx = 0;
Neil Horman878719e2008-10-23 10:40:06 -0400100
Neil Horman878719e2008-10-23 10:40:06 -0400101 printk("%sCall Trace:\n", log_lvl);
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500102
103 unwind_start(&state, task, regs, stack);
Josh Poimboeuff4474c92016-11-17 00:04:58 -0600104 stack = stack ? : get_stack_pointer(task, regs);
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500105
106 /*
107 * Iterate through the stacks, starting with the current stack pointer.
108 * Each stack has a pointer to the next one.
109 *
110 * x86-64 can have several stacks:
111 * - task stack
112 * - interrupt stack
113 * - HW exception stacks (double fault, nmi, debug, mce)
Dave Hansen4fe2d8b2017-12-04 17:25:07 -0800114 * - entry stack
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500115 *
Andy Lutomirski6e60e582017-12-04 15:07:18 +0100116 * x86-32 can have up to four stacks:
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500117 * - task stack
118 * - softirq stack
119 * - hardirq stack
Dave Hansen4fe2d8b2017-12-04 17:25:07 -0800120 * - entry stack
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500121 */
Josh Poimboeufe335bb52017-04-17 08:44:00 -0500122 for (regs = NULL; stack; stack = PTR_ALIGN(stack_info.next_sp, sizeof(long))) {
Josh Poimboeuf3d02a9c2016-11-18 11:46:23 -0600123 const char *stack_name;
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500124
Andy Lutomirski6e60e582017-12-04 15:07:18 +0100125 if (get_stack_info(stack, task, &stack_info, &visit_mask)) {
126 /*
127 * We weren't on a valid stack. It's possible that
128 * we overflowed a valid stack into a guard page.
129 * See if the next page up is valid so that we can
130 * generate some kind of backtrace if this happens.
131 */
132 stack = (unsigned long *)PAGE_ALIGN((unsigned long)stack);
133 if (get_stack_info(stack, task, &stack_info, &visit_mask))
134 break;
135 }
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500136
Josh Poimboeuf3d02a9c2016-11-18 11:46:23 -0600137 stack_name = stack_type_name(stack_info.type);
138 if (stack_name)
139 printk("%s <%s>\n", log_lvl, stack_name);
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500140
Josh Poimboeufb02fcf92017-12-04 15:07:09 +0100141 if (regs)
142 show_regs_safe(&stack_info, regs);
Josh Poimboeufb0529be2017-07-11 10:33:40 -0500143
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500144 /*
145 * Scan the stack, printing any text addresses we find. At the
146 * same time, follow proper stack frames with the unwinder.
147 *
148 * Addresses found during the scan which are not reported by
149 * the unwinder are considered to be additional clues which are
150 * sometimes useful for debugging and are prefixed with '?'.
151 * This also serves as a failsafe option in case the unwinder
152 * goes off in the weeds.
153 */
154 for (; stack < stack_info.end; stack++) {
155 unsigned long real_addr;
156 int reliable = 0;
Josh Poimboeuf91e08ab2016-11-17 09:57:24 -0600157 unsigned long addr = READ_ONCE_NOCHECK(*stack);
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500158 unsigned long *ret_addr_p =
159 unwind_get_return_address_ptr(&state);
160
161 if (!__kernel_text_address(addr))
162 continue;
163
Josh Poimboeuf3b3fa112016-10-20 11:34:43 -0500164 /*
165 * Don't print regs->ip again if it was already printed
Josh Poimboeufb02fcf92017-12-04 15:07:09 +0100166 * by show_regs_safe() below.
Josh Poimboeuf3b3fa112016-10-20 11:34:43 -0500167 */
Josh Poimboeufb0529be2017-07-11 10:33:40 -0500168 if (regs && stack == &regs->ip)
169 goto next;
Josh Poimboeuf3b3fa112016-10-20 11:34:43 -0500170
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500171 if (stack == ret_addr_p)
172 reliable = 1;
173
174 /*
175 * When function graph tracing is enabled for a
176 * function, its return address on the stack is
177 * replaced with the address of an ftrace handler
178 * (return_to_handler). In that case, before printing
179 * the "real" address, we want to print the handler
180 * address as an "unreliable" hint that function graph
181 * tracing was involved.
182 */
183 real_addr = ftrace_graph_ret_addr(task, &graph_idx,
184 addr, stack);
185 if (real_addr != addr)
186 printk_stack_address(addr, 0, log_lvl);
187 printk_stack_address(real_addr, reliable, log_lvl);
188
189 if (!reliable)
190 continue;
191
Josh Poimboeufb0529be2017-07-11 10:33:40 -0500192next:
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500193 /*
194 * Get the next frame from the unwinder. No need to
195 * check for an error: if anything goes wrong, the rest
196 * of the addresses will just be printed as unreliable.
197 */
198 unwind_next_frame(&state);
Josh Poimboeuf3b3fa112016-10-20 11:34:43 -0500199
200 /* if the frame has entry regs, print them */
201 regs = unwind_get_entry_regs(&state);
Josh Poimboeufb02fcf92017-12-04 15:07:09 +0100202 if (regs)
203 show_regs_safe(&stack_info, regs);
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500204 }
205
Josh Poimboeuf3d02a9c2016-11-18 11:46:23 -0600206 if (stack_name)
207 printk("%s </%s>\n", log_lvl, stack_name);
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500208 }
Neil Horman878719e2008-10-23 10:40:06 -0400209}
210
Neil Horman878719e2008-10-23 10:40:06 -0400211void show_stack(struct task_struct *task, unsigned long *sp)
212{
Josh Poimboeuf81539162016-09-16 08:05:20 -0500213 task = task ? : current;
214
Tejun Heoa77f2a42013-04-30 15:27:09 -0700215 /*
216 * Stack frames below this one aren't interesting. Don't show them
217 * if we're printing for %current.
218 */
Josh Poimboeufe18bccc2016-09-16 14:18:16 -0500219 if (!sp && task == current)
Josh Poimboeuf4b8afaf2016-08-24 11:50:17 -0500220 sp = get_stack_pointer(current, NULL);
Tejun Heoa77f2a42013-04-30 15:27:09 -0700221
Josh Poimboeuf0ee1dd92016-10-25 09:51:13 -0500222 show_trace_log_lvl(task, NULL, sp, KERN_DEFAULT);
Neil Horman878719e2008-10-23 10:40:06 -0400223}
224
Borislav Petkov81c29492016-07-05 00:31:27 +0200225void show_stack_regs(struct pt_regs *regs)
226{
Josh Poimboeuf0ee1dd92016-10-25 09:51:13 -0500227 show_trace_log_lvl(current, regs, NULL, KERN_DEFAULT);
Borislav Petkov81c29492016-07-05 00:31:27 +0200228}
229
Thomas Gleixneredc35bd2009-12-03 12:38:57 +0100230static arch_spinlock_t die_lock = __ARCH_SPIN_LOCK_UNLOCKED;
Neil Horman878719e2008-10-23 10:40:06 -0400231static int die_owner = -1;
232static unsigned int die_nest_count;
233
Masami Hiramatsu93266382014-04-17 17:18:14 +0900234unsigned long oops_begin(void)
Neil Horman878719e2008-10-23 10:40:06 -0400235{
236 int cpu;
237 unsigned long flags;
238
Neil Horman878719e2008-10-23 10:40:06 -0400239 oops_enter();
240
241 /* racy, but better than risking deadlock. */
242 raw_local_irq_save(flags);
243 cpu = smp_processor_id();
Thomas Gleixner0199c4e2009-12-02 20:01:25 +0100244 if (!arch_spin_trylock(&die_lock)) {
Neil Horman878719e2008-10-23 10:40:06 -0400245 if (cpu == die_owner)
246 /* nested oops. should stop eventually */;
247 else
Thomas Gleixner0199c4e2009-12-02 20:01:25 +0100248 arch_spin_lock(&die_lock);
Neil Horman878719e2008-10-23 10:40:06 -0400249 }
250 die_nest_count++;
251 die_owner = cpu;
252 console_verbose();
253 bust_spinlocks(1);
254 return flags;
255}
Huang Ying81e88fd2011-01-12 14:44:55 +0800256EXPORT_SYMBOL_GPL(oops_begin);
Masami Hiramatsu93266382014-04-17 17:18:14 +0900257NOKPROBE_SYMBOL(oops_begin);
Neil Horman878719e2008-10-23 10:40:06 -0400258
Andy Lutomirski2deb4be2016-07-14 13:22:55 -0700259void __noreturn rewind_stack_do_exit(int signr);
260
Masami Hiramatsu93266382014-04-17 17:18:14 +0900261void oops_end(unsigned long flags, struct pt_regs *regs, int signr)
Neil Horman878719e2008-10-23 10:40:06 -0400262{
263 if (regs && kexec_should_crash(current))
264 crash_kexec(regs);
265
266 bust_spinlocks(0);
267 die_owner = -1;
Rusty Russell373d4d02013-01-21 17:17:39 +1030268 add_taint(TAINT_DIE, LOCKDEP_NOW_UNRELIABLE);
Neil Horman878719e2008-10-23 10:40:06 -0400269 die_nest_count--;
270 if (!die_nest_count)
271 /* Nest count reaches zero, release the lock. */
Thomas Gleixner0199c4e2009-12-02 20:01:25 +0100272 arch_spin_unlock(&die_lock);
Neil Horman878719e2008-10-23 10:40:06 -0400273 raw_local_irq_restore(flags);
274 oops_exit();
275
276 if (!signr)
277 return;
278 if (in_interrupt())
279 panic("Fatal exception in interrupt");
280 if (panic_on_oops)
281 panic("Fatal exception");
Andy Lutomirski2deb4be2016-07-14 13:22:55 -0700282
283 /*
284 * We're not going to return, but we might be on an IST stack or
285 * have very little stack space left. Rewind the stack and kill
286 * the task.
287 */
288 rewind_stack_do_exit(signr);
Neil Horman878719e2008-10-23 10:40:06 -0400289}
Masami Hiramatsu93266382014-04-17 17:18:14 +0900290NOKPROBE_SYMBOL(oops_end);
Neil Horman878719e2008-10-23 10:40:06 -0400291
Masami Hiramatsu93266382014-04-17 17:18:14 +0900292int __die(const char *str, struct pt_regs *regs, long err)
Neil Horman878719e2008-10-23 10:40:06 -0400293{
294#ifdef CONFIG_X86_32
295 unsigned short ss;
296 unsigned long sp;
297#endif
Prarit Bhargavab0f4c4b2012-01-26 08:55:34 -0500298 printk(KERN_DEFAULT
Rasmus Villemoes8fad7ec2016-03-26 21:40:16 +0100299 "%s: %04lx [#%d]%s%s%s%s\n", str, err & 0xffff, ++die_counter,
300 IS_ENABLED(CONFIG_PREEMPT) ? " PREEMPT" : "",
301 IS_ENABLED(CONFIG_SMP) ? " SMP" : "",
302 debug_pagealloc_enabled() ? " DEBUG_PAGEALLOC" : "",
303 IS_ENABLED(CONFIG_KASAN) ? " KASAN" : "");
304
Neil Horman878719e2008-10-23 10:40:06 -0400305 if (notify_die(DIE_OOPS, str, regs, err,
Srikar Dronamraju51e7dc72012-03-12 14:55:55 +0530306 current->thread.trap_nr, SIGSEGV) == NOTIFY_STOP)
Neil Horman878719e2008-10-23 10:40:06 -0400307 return 1;
308
Jan Beulich0fa0e2f2012-06-18 11:40:04 +0100309 print_modules();
Jan Beulich57da8b92012-05-09 08:47:37 +0100310 show_regs(regs);
Neil Horman878719e2008-10-23 10:40:06 -0400311#ifdef CONFIG_X86_32
Andy Lutomirskif39b6f02015-03-18 18:33:33 -0700312 if (user_mode(regs)) {
Neil Horman878719e2008-10-23 10:40:06 -0400313 sp = regs->sp;
Andy Lutomirski99504812017-07-28 06:00:32 -0700314 ss = regs->ss;
H. Peter Anvina343c752009-10-12 14:11:09 -0700315 } else {
316 sp = kernel_stack_pointer(regs);
317 savesegment(ss, ss);
Neil Horman878719e2008-10-23 10:40:06 -0400318 }
Josh Poimboeufbb5e5ce2016-10-25 09:51:12 -0500319 printk(KERN_EMERG "EIP: %pS SS:ESP: %04x:%08lx\n",
320 (void *)regs->ip, ss, sp);
Neil Horman878719e2008-10-23 10:40:06 -0400321#else
322 /* Executive summary in case the oops scrolled away */
Josh Poimboeufbb5e5ce2016-10-25 09:51:12 -0500323 printk(KERN_ALERT "RIP: %pS RSP: %016lx\n", (void *)regs->ip, regs->sp);
Neil Horman878719e2008-10-23 10:40:06 -0400324#endif
325 return 0;
326}
Masami Hiramatsu93266382014-04-17 17:18:14 +0900327NOKPROBE_SYMBOL(__die);
Neil Horman878719e2008-10-23 10:40:06 -0400328
329/*
330 * This is gone through when something in the kernel has done something bad
331 * and is about to be terminated:
332 */
333void die(const char *str, struct pt_regs *regs, long err)
334{
335 unsigned long flags = oops_begin();
336 int sig = SIGSEGV;
337
Neil Horman878719e2008-10-23 10:40:06 -0400338 if (__die(str, regs, err))
339 sig = 0;
340 oops_end(flags, regs, sig);
341}
342
Neil Horman878719e2008-10-23 10:40:06 -0400343static int __init code_bytes_setup(char *s)
344{
Shuah Khan363f7ce2012-05-06 11:58:04 -0600345 ssize_t ret;
346 unsigned long val;
347
348 if (!s)
349 return -EINVAL;
350
351 ret = kstrtoul(s, 0, &val);
352 if (ret)
353 return ret;
354
355 code_bytes = val;
Neil Horman878719e2008-10-23 10:40:06 -0400356 if (code_bytes > 8192)
357 code_bytes = 8192;
358
359 return 1;
360}
361__setup("code_bytes=", code_bytes_setup);