2 * Copyright (C) 1991, 1992 Linus Torvalds
3 * Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs
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>
13 #include <linux/sched/debug.h>
14 #include <linux/sched/task_stack.h>
15 #include <linux/ftrace.h>
16 #include <linux/kexec.h>
17 #include <linux/bug.h>
18 #include <linux/nmi.h>
19 #include <linux/sysfs.h>
21 #include <asm/cpu_entry_area.h>
22 #include <asm/stacktrace.h>
23 #include <asm/unwind.h>
25 int panic_on_unrecovered_nmi
;
27 unsigned int code_bytes
= 64;
28 static int die_counter
;
30 bool in_task_stack(unsigned long *stack
, struct task_struct
*task
,
31 struct stack_info
*info
)
33 unsigned long *begin
= task_stack_page(task
);
34 unsigned long *end
= task_stack_page(task
) + THREAD_SIZE
;
36 if (stack
< begin
|| stack
>= end
)
39 info
->type
= STACK_TYPE_TASK
;
47 bool in_entry_stack(unsigned long *stack
, struct stack_info
*info
)
49 struct entry_stack
*ss
= cpu_entry_stack(smp_processor_id());
54 if ((void *)stack
< begin
|| (void *)stack
>= end
)
57 info
->type
= STACK_TYPE_ENTRY
;
65 static void printk_stack_address(unsigned long address
, int reliable
,
69 printk("%s %s%pB\n", log_lvl
, reliable
? "" : "? ", (void *)address
);
72 void show_iret_regs(struct pt_regs
*regs
)
74 printk(KERN_DEFAULT
"RIP: %04x:%pS\n", (int)regs
->cs
, (void *)regs
->ip
);
75 printk(KERN_DEFAULT
"RSP: %04x:%016lx EFLAGS: %08lx", (int)regs
->ss
,
76 regs
->sp
, regs
->flags
);
79 static void show_regs_safe(struct stack_info
*info
, struct pt_regs
*regs
)
81 if (on_stack(info
, regs
, sizeof(*regs
)))
83 else if (on_stack(info
, (void *)regs
+ IRET_FRAME_OFFSET
,
86 * When an interrupt or exception occurs in entry code, the
87 * full pt_regs might not have been saved yet. In that case
88 * just print the iret frame.
94 void show_trace_log_lvl(struct task_struct
*task
, struct pt_regs
*regs
,
95 unsigned long *stack
, char *log_lvl
)
97 struct unwind_state state
;
98 struct stack_info stack_info
= {0};
99 unsigned long visit_mask
= 0;
102 printk("%sCall Trace:\n", log_lvl
);
104 unwind_start(&state
, task
, regs
, stack
);
105 stack
= stack
? : get_stack_pointer(task
, regs
);
108 * Iterate through the stacks, starting with the current stack pointer.
109 * Each stack has a pointer to the next one.
111 * x86-64 can have several stacks:
114 * - HW exception stacks (double fault, nmi, debug, mce)
117 * x86-32 can have up to four stacks:
123 for (regs
= NULL
; stack
; stack
= PTR_ALIGN(stack_info
.next_sp
, sizeof(long))) {
124 const char *stack_name
;
126 if (get_stack_info(stack
, task
, &stack_info
, &visit_mask
)) {
128 * We weren't on a valid stack. It's possible that
129 * we overflowed a valid stack into a guard page.
130 * See if the next page up is valid so that we can
131 * generate some kind of backtrace if this happens.
133 stack
= (unsigned long *)PAGE_ALIGN((unsigned long)stack
);
134 if (get_stack_info(stack
, task
, &stack_info
, &visit_mask
))
138 stack_name
= stack_type_name(stack_info
.type
);
140 printk("%s <%s>\n", log_lvl
, stack_name
);
143 show_regs_safe(&stack_info
, regs
);
146 * Scan the stack, printing any text addresses we find. At the
147 * same time, follow proper stack frames with the unwinder.
149 * Addresses found during the scan which are not reported by
150 * the unwinder are considered to be additional clues which are
151 * sometimes useful for debugging and are prefixed with '?'.
152 * This also serves as a failsafe option in case the unwinder
153 * goes off in the weeds.
155 for (; stack
< stack_info
.end
; stack
++) {
156 unsigned long real_addr
;
158 unsigned long addr
= READ_ONCE_NOCHECK(*stack
);
159 unsigned long *ret_addr_p
=
160 unwind_get_return_address_ptr(&state
);
162 if (!__kernel_text_address(addr
))
166 * Don't print regs->ip again if it was already printed
167 * by show_regs_safe() below.
169 if (regs
&& stack
== ®s
->ip
) {
170 unwind_next_frame(&state
);
174 if (stack
== ret_addr_p
)
178 * When function graph tracing is enabled for a
179 * function, its return address on the stack is
180 * replaced with the address of an ftrace handler
181 * (return_to_handler). In that case, before printing
182 * the "real" address, we want to print the handler
183 * address as an "unreliable" hint that function graph
184 * tracing was involved.
186 real_addr
= ftrace_graph_ret_addr(task
, &graph_idx
,
188 if (real_addr
!= addr
)
189 printk_stack_address(addr
, 0, log_lvl
);
190 printk_stack_address(real_addr
, reliable
, log_lvl
);
196 * Get the next frame from the unwinder. No need to
197 * check for an error: if anything goes wrong, the rest
198 * of the addresses will just be printed as unreliable.
200 unwind_next_frame(&state
);
202 /* if the frame has entry regs, print them */
203 regs
= unwind_get_entry_regs(&state
);
205 show_regs_safe(&stack_info
, regs
);
209 printk("%s </%s>\n", log_lvl
, stack_name
);
213 void show_stack(struct task_struct
*task
, unsigned long *sp
)
215 task
= task
? : current
;
218 * Stack frames below this one aren't interesting. Don't show them
219 * if we're printing for %current.
221 if (!sp
&& task
== current
)
222 sp
= get_stack_pointer(current
, NULL
);
224 show_trace_log_lvl(task
, NULL
, sp
, KERN_DEFAULT
);
227 void show_stack_regs(struct pt_regs
*regs
)
229 show_trace_log_lvl(current
, regs
, NULL
, KERN_DEFAULT
);
232 static arch_spinlock_t die_lock
= __ARCH_SPIN_LOCK_UNLOCKED
;
233 static int die_owner
= -1;
234 static unsigned int die_nest_count
;
236 unsigned long oops_begin(void)
243 /* racy, but better than risking deadlock. */
244 raw_local_irq_save(flags
);
245 cpu
= smp_processor_id();
246 if (!arch_spin_trylock(&die_lock
)) {
247 if (cpu
== die_owner
)
248 /* nested oops. should stop eventually */;
250 arch_spin_lock(&die_lock
);
258 EXPORT_SYMBOL_GPL(oops_begin
);
259 NOKPROBE_SYMBOL(oops_begin
);
261 void __noreturn
rewind_stack_do_exit(int signr
);
263 void oops_end(unsigned long flags
, struct pt_regs
*regs
, int signr
)
265 if (regs
&& kexec_should_crash(current
))
270 add_taint(TAINT_DIE
, LOCKDEP_NOW_UNRELIABLE
);
273 /* Nest count reaches zero, release the lock. */
274 arch_spin_unlock(&die_lock
);
275 raw_local_irq_restore(flags
);
281 panic("Fatal exception in interrupt");
283 panic("Fatal exception");
286 * We're not going to return, but we might be on an IST stack or
287 * have very little stack space left. Rewind the stack and kill
290 rewind_stack_do_exit(signr
);
292 NOKPROBE_SYMBOL(oops_end
);
294 int __die(const char *str
, struct pt_regs
*regs
, long err
)
301 "%s: %04lx [#%d]%s%s%s%s\n", str
, err
& 0xffff, ++die_counter
,
302 IS_ENABLED(CONFIG_PREEMPT
) ? " PREEMPT" : "",
303 IS_ENABLED(CONFIG_SMP
) ? " SMP" : "",
304 debug_pagealloc_enabled() ? " DEBUG_PAGEALLOC" : "",
305 IS_ENABLED(CONFIG_KASAN
) ? " KASAN" : "");
307 if (notify_die(DIE_OOPS
, str
, regs
, err
,
308 current
->thread
.trap_nr
, SIGSEGV
) == NOTIFY_STOP
)
314 if (user_mode(regs
)) {
316 ss
= regs
->ss
& 0xffff;
318 sp
= kernel_stack_pointer(regs
);
321 printk(KERN_EMERG
"EIP: %pS SS:ESP: %04x:%08lx\n",
322 (void *)regs
->ip
, ss
, sp
);
324 /* Executive summary in case the oops scrolled away */
325 printk(KERN_ALERT
"RIP: %pS RSP: %016lx\n", (void *)regs
->ip
, regs
->sp
);
329 NOKPROBE_SYMBOL(__die
);
332 * This is gone through when something in the kernel has done something bad
333 * and is about to be terminated:
335 void die(const char *str
, struct pt_regs
*regs
, long err
)
337 unsigned long flags
= oops_begin();
340 if (__die(str
, regs
, err
))
342 oops_end(flags
, regs
, sig
);
345 static int __init
code_bytes_setup(char *s
)
353 ret
= kstrtoul(s
, 0, &val
);
358 if (code_bytes
> 8192)
363 __setup("code_bytes=", code_bytes_setup
);