1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2008 Steven Rostedt <srostedt@redhat.com>
6 #include <linux/sched/task_stack.h>
7 #include <linux/stacktrace.h>
8 #include <linux/kallsyms.h>
9 #include <linux/seq_file.h>
10 #include <linux/spinlock.h>
11 #include <linux/uaccess.h>
12 #include <linux/ftrace.h>
13 #include <linux/module.h>
14 #include <linux/sysctl.h>
15 #include <linux/init.h>
17 #include <asm/setup.h>
21 #define STACK_TRACE_ENTRIES 500
23 static unsigned long stack_dump_trace[STACK_TRACE_ENTRIES];
24 static unsigned stack_trace_index[STACK_TRACE_ENTRIES];
26 static unsigned int stack_trace_nr_entries;
27 static unsigned long stack_trace_max_size;
28 static arch_spinlock_t stack_trace_max_lock =
29 (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
31 DEFINE_PER_CPU(int, disable_stack_tracer);
32 static DEFINE_MUTEX(stack_sysctl_mutex);
34 int stack_tracer_enabled;
36 static void print_max_stack(void)
41 pr_emerg(" Depth Size Location (%d entries)\n"
42 " ----- ---- --------\n",
43 stack_trace_nr_entries);
45 for (i = 0; i < stack_trace_nr_entries; i++) {
46 if (i + 1 == stack_trace_nr_entries)
47 size = stack_trace_index[i];
49 size = stack_trace_index[i] - stack_trace_index[i+1];
51 pr_emerg("%3ld) %8d %5d %pS\n", i, stack_trace_index[i],
52 size, (void *)stack_dump_trace[i]);
57 * The stack tracer looks for a maximum stack at each call from a function. It
58 * registers a callback from ftrace, and in that callback it examines the stack
59 * size. It determines the stack size from the variable passed in, which is the
60 * address of a local variable in the stack_trace_call() callback function.
61 * The stack size is calculated by the address of the local variable to the top
62 * of the current stack. If that size is smaller than the currently saved max
63 * stack size, nothing more is done.
65 * If the size of the stack is greater than the maximum recorded size, then the
66 * following algorithm takes place.
68 * For architectures (like x86) that store the function's return address before
69 * saving the function's local variables, the stack will look something like
73 * 0: sys call entry frame
74 * 10: return addr to entry code
75 * 11: start of sys_foo frame
76 * 20: return addr to sys_foo
77 * 21: start of kernel_func_bar frame
78 * 30: return addr to kernel_func_bar
79 * 31: [ do trace stack here ]
81 * The save_stack_trace() is called returning all the functions it finds in the
82 * current stack. Which would be (from the bottom of the stack to the top):
84 * return addr to kernel_func_bar
85 * return addr to sys_foo
86 * return addr to entry code
88 * Now to figure out how much each of these functions' local variable size is,
89 * a search of the stack is made to find these values. When a match is made, it
90 * is added to the stack_dump_trace[] array. The offset into the stack is saved
91 * in the stack_trace_index[] array. The above example would show:
93 * stack_dump_trace[] | stack_trace_index[]
94 * ------------------ + -------------------
95 * return addr to kernel_func_bar | 30
96 * return addr to sys_foo | 20
97 * return addr to entry | 10
99 * The print_max_stack() function above, uses these values to print the size of
100 * each function's portion of the stack.
102 * for (i = 0; i < nr_entries; i++) {
103 * size = i == nr_entries - 1 ? stack_trace_index[i] :
104 * stack_trace_index[i] - stack_trace_index[i+1]
105 * print "%d %d %d %s\n", i, stack_trace_index[i], size, stack_dump_trace[i]);
110 * depth size location
111 * ----- ---- --------
112 * 0 30 10 kernel_func_bar
116 * Now for architectures that might save the return address after the functions
117 * local variables (saving the link register before calling nested functions),
118 * this will cause the stack to look a little different:
121 * 0: sys call entry frame
122 * 10: start of sys_foo_frame
123 * 19: return addr to entry code << lr saved before calling kernel_func_bar
124 * 20: start of kernel_func_bar frame
125 * 29: return addr to sys_foo_frame << lr saved before calling next function
126 * 30: [ do trace stack here ]
128 * Although the functions returned by save_stack_trace() may be the same, the
129 * placement in the stack will be different. Using the same algorithm as above
132 * stack_dump_trace[] | stack_trace_index[]
133 * ------------------ + -------------------
134 * return addr to kernel_func_bar | 30
135 * return addr to sys_foo | 29
136 * return addr to entry | 19
138 * Where the mapping is off by one:
140 * kernel_func_bar stack frame size is 29 - 19 not 30 - 29!
142 * To fix this, if the architecture sets ARCH_RET_ADDR_AFTER_LOCAL_VARS the
143 * values in stack_trace_index[] are shifted by one to and the number of
144 * stack trace entries is decremented by one.
146 * stack_dump_trace[] | stack_trace_index[]
147 * ------------------ + -------------------
148 * return addr to kernel_func_bar | 29
149 * return addr to sys_foo | 19
151 * Although the entry function is not displayed, the first function (sys_foo)
152 * will still include the stack size of it.
154 static void check_stack(unsigned long ip, unsigned long *stack)
156 unsigned long this_size, flags; unsigned long *p, *top, *start;
157 static int tracer_frame;
158 int frame_size = READ_ONCE(tracer_frame);
161 this_size = ((unsigned long)stack) & (THREAD_SIZE-1);
162 this_size = THREAD_SIZE - this_size;
163 /* Remove the frame of the tracer */
164 this_size -= frame_size;
166 if (this_size <= stack_trace_max_size)
169 /* we do not handle interrupt stacks yet */
170 if (!object_is_on_stack(stack))
173 /* Can't do this from NMI context (can cause deadlocks) */
177 local_irq_save(flags);
178 arch_spin_lock(&stack_trace_max_lock);
180 /* In case another CPU set the tracer_frame on us */
181 if (unlikely(!frame_size))
182 this_size -= tracer_frame;
184 /* a race could have already updated it */
185 if (this_size <= stack_trace_max_size)
188 stack_trace_max_size = this_size;
190 stack_trace_nr_entries = stack_trace_save(stack_dump_trace,
191 ARRAY_SIZE(stack_dump_trace) - 1,
194 /* Skip over the overhead of the stack tracer itself */
195 for (i = 0; i < stack_trace_nr_entries; i++) {
196 if (stack_dump_trace[i] == ip)
201 * Some archs may not have the passed in ip in the dump.
202 * If that happens, we need to show everything.
204 if (i == stack_trace_nr_entries)
208 * Now find where in the stack these are.
212 top = (unsigned long *)
213 (((unsigned long)start & ~(THREAD_SIZE-1)) + THREAD_SIZE);
216 * Loop through all the entries. One of the entries may
217 * for some reason be missed on the stack, so we may
218 * have to account for them. If they are all there, this
219 * loop will only happen once. This code only takes place
220 * on a new max, so it is far from a fast path.
222 while (i < stack_trace_nr_entries) {
225 stack_trace_index[x] = this_size;
228 for (; p < top && i < stack_trace_nr_entries; p++) {
230 * The READ_ONCE_NOCHECK is used to let KASAN know that
231 * this is not a stack-out-of-bounds error.
233 if ((READ_ONCE_NOCHECK(*p)) == stack_dump_trace[i]) {
234 stack_dump_trace[x] = stack_dump_trace[i++];
235 this_size = stack_trace_index[x++] =
236 (top - p) * sizeof(unsigned long);
238 /* Start the search from here */
241 * We do not want to show the overhead
242 * of the stack tracer stack in the
243 * max stack. If we haven't figured
244 * out what that is, then figure it out
247 if (unlikely(!tracer_frame)) {
248 tracer_frame = (p - stack) *
249 sizeof(unsigned long);
250 stack_trace_max_size -= tracer_frame;
259 #ifdef ARCH_FTRACE_SHIFT_STACK_TRACER
261 * Some archs will store the link register before calling
262 * nested functions. This means the saved return address
263 * comes after the local storage, and we need to shift
267 memmove(&stack_trace_index[0], &stack_trace_index[1],
268 sizeof(stack_trace_index[0]) * (x - 1));
273 stack_trace_nr_entries = x;
275 if (task_stack_end_corrupted(current)) {
281 arch_spin_unlock(&stack_trace_max_lock);
282 local_irq_restore(flags);
286 stack_trace_call(unsigned long ip, unsigned long parent_ip,
287 struct ftrace_ops *op, struct pt_regs *pt_regs)
291 preempt_disable_notrace();
293 /* no atomic needed, we only modify this variable by this cpu */
294 __this_cpu_inc(disable_stack_tracer);
295 if (__this_cpu_read(disable_stack_tracer) != 1)
298 /* If rcu is not watching, then save stack trace can fail */
299 if (!rcu_is_watching())
302 ip += MCOUNT_INSN_SIZE;
304 check_stack(ip, &stack);
307 __this_cpu_dec(disable_stack_tracer);
308 /* prevent recursion in schedule */
309 preempt_enable_notrace();
312 static struct ftrace_ops trace_ops __read_mostly =
314 .func = stack_trace_call,
315 .flags = FTRACE_OPS_FL_RECURSION_SAFE,
319 stack_max_size_read(struct file *filp, char __user *ubuf,
320 size_t count, loff_t *ppos)
322 unsigned long *ptr = filp->private_data;
326 r = snprintf(buf, sizeof(buf), "%ld\n", *ptr);
329 return simple_read_from_buffer(ubuf, count, ppos, buf, r);
333 stack_max_size_write(struct file *filp, const char __user *ubuf,
334 size_t count, loff_t *ppos)
336 long *ptr = filp->private_data;
337 unsigned long val, flags;
340 ret = kstrtoul_from_user(ubuf, count, 10, &val);
344 local_irq_save(flags);
347 * In case we trace inside arch_spin_lock() or after (NMI),
348 * we will cause circular lock, so we also need to increase
349 * the percpu disable_stack_tracer here.
351 __this_cpu_inc(disable_stack_tracer);
353 arch_spin_lock(&stack_trace_max_lock);
355 arch_spin_unlock(&stack_trace_max_lock);
357 __this_cpu_dec(disable_stack_tracer);
358 local_irq_restore(flags);
363 static const struct file_operations stack_max_size_fops = {
364 .open = tracing_open_generic,
365 .read = stack_max_size_read,
366 .write = stack_max_size_write,
367 .llseek = default_llseek,
371 __next(struct seq_file *m, loff_t *pos)
375 if (n >= stack_trace_nr_entries)
378 m->private = (void *)n;
383 t_next(struct seq_file *m, void *v, loff_t *pos)
386 return __next(m, pos);
389 static void *t_start(struct seq_file *m, loff_t *pos)
393 __this_cpu_inc(disable_stack_tracer);
395 arch_spin_lock(&stack_trace_max_lock);
398 return SEQ_START_TOKEN;
400 return __next(m, pos);
403 static void t_stop(struct seq_file *m, void *p)
405 arch_spin_unlock(&stack_trace_max_lock);
407 __this_cpu_dec(disable_stack_tracer);
412 static void trace_lookup_stack(struct seq_file *m, long i)
414 unsigned long addr = stack_dump_trace[i];
416 seq_printf(m, "%pS\n", (void *)addr);
419 static void print_disabled(struct seq_file *m)
422 "# Stack tracer disabled\n"
424 "# To enable the stack tracer, either add 'stacktrace' to the\n"
425 "# kernel command line\n"
426 "# or 'echo 1 > /proc/sys/kernel/stack_tracer_enabled'\n"
430 static int t_show(struct seq_file *m, void *v)
435 if (v == SEQ_START_TOKEN) {
436 seq_printf(m, " Depth Size Location"
438 " ----- ---- --------\n",
439 stack_trace_nr_entries);
441 if (!stack_tracer_enabled && !stack_trace_max_size)
449 if (i >= stack_trace_nr_entries)
452 if (i + 1 == stack_trace_nr_entries)
453 size = stack_trace_index[i];
455 size = stack_trace_index[i] - stack_trace_index[i+1];
457 seq_printf(m, "%3ld) %8d %5d ", i, stack_trace_index[i], size);
459 trace_lookup_stack(m, i);
464 static const struct seq_operations stack_trace_seq_ops = {
471 static int stack_trace_open(struct inode *inode, struct file *file)
473 return seq_open(file, &stack_trace_seq_ops);
476 static const struct file_operations stack_trace_fops = {
477 .open = stack_trace_open,
480 .release = seq_release,
483 #ifdef CONFIG_DYNAMIC_FTRACE
486 stack_trace_filter_open(struct inode *inode, struct file *file)
488 struct ftrace_ops *ops = inode->i_private;
490 return ftrace_regex_open(ops, FTRACE_ITER_FILTER,
494 static const struct file_operations stack_trace_filter_fops = {
495 .open = stack_trace_filter_open,
497 .write = ftrace_filter_write,
498 .llseek = tracing_lseek,
499 .release = ftrace_regex_release,
502 #endif /* CONFIG_DYNAMIC_FTRACE */
505 stack_trace_sysctl(struct ctl_table *table, int write,
506 void __user *buffer, size_t *lenp,
512 mutex_lock(&stack_sysctl_mutex);
513 was_enabled = !!stack_tracer_enabled;
515 ret = proc_dointvec(table, write, buffer, lenp, ppos);
517 if (ret || !write || (was_enabled == !!stack_tracer_enabled))
520 if (stack_tracer_enabled)
521 register_ftrace_function(&trace_ops);
523 unregister_ftrace_function(&trace_ops);
525 mutex_unlock(&stack_sysctl_mutex);
529 static char stack_trace_filter_buf[COMMAND_LINE_SIZE+1] __initdata;
531 static __init int enable_stacktrace(char *str)
535 if ((len = str_has_prefix(str, "_filter=")))
536 strncpy(stack_trace_filter_buf, str + len, COMMAND_LINE_SIZE);
538 stack_tracer_enabled = 1;
541 __setup("stacktrace", enable_stacktrace);
543 static __init int stack_trace_init(void)
545 struct dentry *d_tracer;
547 d_tracer = tracing_init_dentry();
548 if (IS_ERR(d_tracer))
551 trace_create_file("stack_max_size", 0644, d_tracer,
552 &stack_trace_max_size, &stack_max_size_fops);
554 trace_create_file("stack_trace", 0444, d_tracer,
555 NULL, &stack_trace_fops);
557 #ifdef CONFIG_DYNAMIC_FTRACE
558 trace_create_file("stack_trace_filter", 0644, d_tracer,
559 &trace_ops, &stack_trace_filter_fops);
562 if (stack_trace_filter_buf[0])
563 ftrace_set_early_filter(&trace_ops, stack_trace_filter_buf, 1);
565 if (stack_tracer_enabled)
566 register_ftrace_function(&trace_ops);
571 device_initcall(stack_trace_init);