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/ftrace.h>
14 #include <linux/kexec.h>
15 #include <linux/bug.h>
16 #include <linux/nmi.h>
17 #include <linux/sysfs.h>
19 #include <asm/stacktrace.h>
20 #include <asm/unwind.h>
22 int panic_on_unrecovered_nmi;
24 unsigned int code_bytes = 64;
25 static int die_counter;
27 bool in_task_stack(unsigned long *stack, struct task_struct *task,
28 struct stack_info *info)
30 unsigned long *begin = task_stack_page(task);
31 unsigned long *end = task_stack_page(task) + THREAD_SIZE;
33 if (stack < begin || stack >= end)
36 info->type = STACK_TYPE_TASK;
44 static void printk_stack_address(unsigned long address, int reliable,
48 printk("%s %s%pB\n", log_lvl, reliable ? "" : "? ", (void *)address);
51 void show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs,
52 unsigned long *stack, char *log_lvl)
54 struct unwind_state state;
55 struct stack_info stack_info = {0};
56 unsigned long visit_mask = 0;
59 printk("%sCall Trace:\n", log_lvl);
61 unwind_start(&state, task, regs, stack);
62 stack = stack ? : get_stack_pointer(task, regs);
65 * Iterate through the stacks, starting with the current stack pointer.
66 * Each stack has a pointer to the next one.
68 * x86-64 can have several stacks:
71 * - HW exception stacks (double fault, nmi, debug, mce)
73 * x86-32 can have up to three stacks:
78 for (regs = NULL; stack; stack = stack_info.next_sp) {
79 const char *stack_name;
82 * If we overflowed the task stack into a guard page, jump back
83 * to the bottom of the usable stack.
85 if (task_stack_page(task) - (void *)stack < PAGE_SIZE)
86 stack = task_stack_page(task);
88 if (get_stack_info(stack, task, &stack_info, &visit_mask))
91 stack_name = stack_type_name(stack_info.type);
93 printk("%s <%s>\n", log_lvl, stack_name);
96 * Scan the stack, printing any text addresses we find. At the
97 * same time, follow proper stack frames with the unwinder.
99 * Addresses found during the scan which are not reported by
100 * the unwinder are considered to be additional clues which are
101 * sometimes useful for debugging and are prefixed with '?'.
102 * This also serves as a failsafe option in case the unwinder
103 * goes off in the weeds.
105 for (; stack < stack_info.end; stack++) {
106 unsigned long real_addr;
108 unsigned long addr = READ_ONCE_NOCHECK(*stack);
109 unsigned long *ret_addr_p =
110 unwind_get_return_address_ptr(&state);
112 if (!__kernel_text_address(addr))
116 * Don't print regs->ip again if it was already printed
117 * by __show_regs() below.
119 if (regs && stack == ®s->ip) {
120 unwind_next_frame(&state);
124 if (stack == ret_addr_p)
128 * When function graph tracing is enabled for a
129 * function, its return address on the stack is
130 * replaced with the address of an ftrace handler
131 * (return_to_handler). In that case, before printing
132 * the "real" address, we want to print the handler
133 * address as an "unreliable" hint that function graph
134 * tracing was involved.
136 real_addr = ftrace_graph_ret_addr(task, &graph_idx,
138 if (real_addr != addr)
139 printk_stack_address(addr, 0, log_lvl);
140 printk_stack_address(real_addr, reliable, log_lvl);
146 * Get the next frame from the unwinder. No need to
147 * check for an error: if anything goes wrong, the rest
148 * of the addresses will just be printed as unreliable.
150 unwind_next_frame(&state);
152 /* if the frame has entry regs, print them */
153 regs = unwind_get_entry_regs(&state);
155 __show_regs(regs, 0);
159 printk("%s </%s>\n", log_lvl, stack_name);
163 void show_stack(struct task_struct *task, unsigned long *sp)
165 task = task ? : current;
168 * Stack frames below this one aren't interesting. Don't show them
169 * if we're printing for %current.
171 if (!sp && task == current)
172 sp = get_stack_pointer(current, NULL);
174 show_trace_log_lvl(task, NULL, sp, KERN_DEFAULT);
177 void show_stack_regs(struct pt_regs *regs)
179 show_trace_log_lvl(current, regs, NULL, KERN_DEFAULT);
182 static arch_spinlock_t die_lock = __ARCH_SPIN_LOCK_UNLOCKED;
183 static int die_owner = -1;
184 static unsigned int die_nest_count;
186 unsigned long oops_begin(void)
193 /* racy, but better than risking deadlock. */
194 raw_local_irq_save(flags);
195 cpu = smp_processor_id();
196 if (!arch_spin_trylock(&die_lock)) {
197 if (cpu == die_owner)
198 /* nested oops. should stop eventually */;
200 arch_spin_lock(&die_lock);
208 EXPORT_SYMBOL_GPL(oops_begin);
209 NOKPROBE_SYMBOL(oops_begin);
211 void __noreturn rewind_stack_do_exit(int signr);
213 void oops_end(unsigned long flags, struct pt_regs *regs, int signr)
215 if (regs && kexec_should_crash(current))
220 add_taint(TAINT_DIE, LOCKDEP_NOW_UNRELIABLE);
223 /* Nest count reaches zero, release the lock. */
224 arch_spin_unlock(&die_lock);
225 raw_local_irq_restore(flags);
231 panic("Fatal exception in interrupt");
233 panic("Fatal exception");
236 * We're not going to return, but we might be on an IST stack or
237 * have very little stack space left. Rewind the stack and kill
240 rewind_stack_do_exit(signr);
242 NOKPROBE_SYMBOL(oops_end);
244 int __die(const char *str, struct pt_regs *regs, long err)
251 "%s: %04lx [#%d]%s%s%s%s\n", str, err & 0xffff, ++die_counter,
252 IS_ENABLED(CONFIG_PREEMPT) ? " PREEMPT" : "",
253 IS_ENABLED(CONFIG_SMP) ? " SMP" : "",
254 debug_pagealloc_enabled() ? " DEBUG_PAGEALLOC" : "",
255 IS_ENABLED(CONFIG_KASAN) ? " KASAN" : "");
257 if (notify_die(DIE_OOPS, str, regs, err,
258 current->thread.trap_nr, SIGSEGV) == NOTIFY_STOP)
264 if (user_mode(regs)) {
266 ss = regs->ss & 0xffff;
268 sp = kernel_stack_pointer(regs);
271 printk(KERN_EMERG "EIP: %pS SS:ESP: %04x:%08lx\n",
272 (void *)regs->ip, ss, sp);
274 /* Executive summary in case the oops scrolled away */
275 printk(KERN_ALERT "RIP: %pS RSP: %016lx\n", (void *)regs->ip, regs->sp);
279 NOKPROBE_SYMBOL(__die);
282 * This is gone through when something in the kernel has done something bad
283 * and is about to be terminated:
285 void die(const char *str, struct pt_regs *regs, long err)
287 unsigned long flags = oops_begin();
290 if (!user_mode(regs))
291 report_bug(regs->ip, regs);
293 if (__die(str, regs, err))
295 oops_end(flags, regs, sig);
298 static int __init code_bytes_setup(char *s)
306 ret = kstrtoul(s, 0, &val);
311 if (code_bytes > 8192)
316 __setup("code_bytes=", code_bytes_setup);