2008-05-12 15:20:47 -04:00
|
|
|
/*
|
|
|
|
* trace stack traces
|
|
|
|
*
|
2008-05-12 15:20:54 -04:00
|
|
|
* Copyright (C) 2004-2008, Soeren Sandmann
|
2008-05-12 15:20:47 -04:00
|
|
|
* Copyright (C) 2007 Steven Rostedt <srostedt@redhat.com>
|
|
|
|
* Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
|
|
|
|
*/
|
|
|
|
#include <linux/kallsyms.h>
|
2008-05-12 15:20:47 -04:00
|
|
|
#include <linux/debugfs.h>
|
|
|
|
#include <linux/hrtimer.h>
|
2008-05-12 15:20:47 -04:00
|
|
|
#include <linux/uaccess.h>
|
|
|
|
#include <linux/ftrace.h>
|
2008-05-12 15:20:47 -04:00
|
|
|
#include <linux/module.h>
|
2008-05-12 15:20:47 -04:00
|
|
|
#include <linux/irq.h>
|
2008-05-12 15:20:47 -04:00
|
|
|
#include <linux/fs.h>
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-12 15:20:54 -04:00
|
|
|
#include <asm/stacktrace.h>
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
#include "trace.h"
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
static struct trace_array *sysprof_trace;
|
2008-05-12 15:20:47 -04:00
|
|
|
static int __read_mostly tracer_enabled;
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
/*
|
2008-05-12 15:20:49 -04:00
|
|
|
* 1 msec sample interval by default:
|
2008-05-12 15:20:47 -04:00
|
|
|
*/
|
2008-05-12 15:20:49 -04:00
|
|
|
static unsigned long sample_period = 1000000;
|
2008-05-12 15:20:47 -04:00
|
|
|
static const unsigned int sample_max_depth = 512;
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-12 15:20:49 -04:00
|
|
|
static DEFINE_MUTEX(sample_timer_lock);
|
2008-05-12 15:20:47 -04:00
|
|
|
/*
|
|
|
|
* Per CPU hrtimers that do the profiling:
|
|
|
|
*/
|
|
|
|
static DEFINE_PER_CPU(struct hrtimer, stack_trace_hrtimer);
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
struct stack_frame {
|
|
|
|
const void __user *next_fp;
|
|
|
|
unsigned long return_address;
|
|
|
|
};
|
|
|
|
|
|
|
|
static int copy_stack_frame(const void __user *fp, struct stack_frame *frame)
|
|
|
|
{
|
2008-05-12 15:20:48 -04:00
|
|
|
int ret;
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
if (!access_ok(VERIFY_READ, fp, sizeof(*frame)))
|
|
|
|
return 0;
|
|
|
|
|
2008-05-12 15:20:48 -04:00
|
|
|
ret = 1;
|
|
|
|
pagefault_disable();
|
|
|
|
if (__copy_from_user_inatomic(frame, fp, sizeof(*frame)))
|
|
|
|
ret = 0;
|
|
|
|
pagefault_enable();
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-12 15:20:48 -04:00
|
|
|
return ret;
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
2008-05-12 15:20:54 -04:00
|
|
|
struct backtrace_info {
|
|
|
|
struct trace_array_cpu *data;
|
|
|
|
struct trace_array *tr;
|
|
|
|
int pos;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void
|
|
|
|
backtrace_warning_symbol(void *data, char *msg, unsigned long symbol)
|
|
|
|
{
|
|
|
|
/* Ignore warnings */
|
|
|
|
}
|
|
|
|
|
|
|
|
static void backtrace_warning(void *data, char *msg)
|
|
|
|
{
|
|
|
|
/* Ignore warnings */
|
|
|
|
}
|
|
|
|
|
|
|
|
static int backtrace_stack(void *data, char *name)
|
|
|
|
{
|
|
|
|
/* Don't bother with IRQ stacks for now */
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void backtrace_address(void *data, unsigned long addr, int reliable)
|
|
|
|
{
|
|
|
|
struct backtrace_info *info = data;
|
|
|
|
|
|
|
|
if (info->pos < sample_max_depth && reliable) {
|
|
|
|
__trace_special(info->tr, info->data, 1, addr, 0);
|
|
|
|
|
|
|
|
info->pos++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
const static struct stacktrace_ops backtrace_ops = {
|
|
|
|
.warning = backtrace_warning,
|
|
|
|
.warning_symbol = backtrace_warning_symbol,
|
|
|
|
.stack = backtrace_stack,
|
|
|
|
.address = backtrace_address,
|
|
|
|
};
|
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
static int
|
2008-05-12 15:20:54 -04:00
|
|
|
trace_kernel(struct pt_regs *regs, struct trace_array *tr,
|
|
|
|
struct trace_array_cpu *data)
|
|
|
|
{
|
|
|
|
struct backtrace_info info;
|
|
|
|
unsigned long bp;
|
|
|
|
char *stack;
|
|
|
|
|
|
|
|
info.tr = tr;
|
|
|
|
info.data = data;
|
|
|
|
info.pos = 1;
|
|
|
|
|
|
|
|
__trace_special(info.tr, info.data, 1, regs->ip, 0);
|
|
|
|
|
|
|
|
stack = ((char *)regs + sizeof(struct pt_regs));
|
|
|
|
#ifdef CONFIG_FRAME_POINTER
|
|
|
|
bp = regs->bp;
|
|
|
|
#else
|
|
|
|
bp = 0;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
dump_trace(NULL, regs, (void *)stack, bp, &backtrace_ops, &info);
|
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
return info.pos;
|
2008-05-12 15:20:54 -04:00
|
|
|
}
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
static void timer_notify(struct pt_regs *regs, int cpu)
|
|
|
|
{
|
|
|
|
struct trace_array_cpu *data;
|
|
|
|
struct stack_frame frame;
|
|
|
|
struct trace_array *tr;
|
2008-05-12 15:20:48 -04:00
|
|
|
const void __user *fp;
|
2008-05-12 15:20:47 -04:00
|
|
|
int is_user;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (!regs)
|
|
|
|
return;
|
|
|
|
|
|
|
|
tr = sysprof_trace;
|
|
|
|
data = tr->data[cpu];
|
|
|
|
is_user = user_mode(regs);
|
|
|
|
|
|
|
|
if (!current || current->pid == 0)
|
|
|
|
return;
|
|
|
|
|
|
|
|
if (is_user && current->state != TASK_RUNNING)
|
|
|
|
return;
|
|
|
|
|
2008-05-12 15:20:54 -04:00
|
|
|
__trace_special(tr, data, 0, 0, current->pid);
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-12 15:20:54 -04:00
|
|
|
if (!is_user)
|
2008-05-11 23:28:50 -04:00
|
|
|
i = trace_kernel(regs, tr, data);
|
|
|
|
else
|
|
|
|
i = 0;
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
/*
|
|
|
|
* Trace user stack if we are not a kernel thread
|
|
|
|
*/
|
|
|
|
if (current->mm && i < sample_max_depth) {
|
|
|
|
regs = (struct pt_regs *)current->thread.sp0 - 1;
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
fp = (void __user *)regs->bp;
|
2008-05-12 15:20:54 -04:00
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
__trace_special(tr, data, 2, regs->ip, 0);
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
while (i < sample_max_depth) {
|
2008-07-25 04:48:39 -04:00
|
|
|
frame.next_fp = NULL;
|
2008-05-11 23:28:50 -04:00
|
|
|
frame.return_address = 0;
|
|
|
|
if (!copy_stack_frame(fp, &frame))
|
|
|
|
break;
|
|
|
|
if ((unsigned long)fp < regs->sp)
|
|
|
|
break;
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-11 23:28:50 -04:00
|
|
|
__trace_special(tr, data, 2, frame.return_address,
|
|
|
|
(unsigned long)fp);
|
|
|
|
fp = frame.next_fp;
|
|
|
|
|
|
|
|
i++;
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-05-12 15:20:48 -04:00
|
|
|
/*
|
|
|
|
* Special trace entry if we overflow the max depth:
|
|
|
|
*/
|
2008-05-12 15:20:47 -04:00
|
|
|
if (i == sample_max_depth)
|
2008-05-23 17:55:54 -04:00
|
|
|
__trace_special(tr, data, -1, -1, -1);
|
2008-05-11 23:28:50 -04:00
|
|
|
|
|
|
|
__trace_special(tr, data, 3, current->pid, i);
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
2008-05-12 15:20:47 -04:00
|
|
|
static enum hrtimer_restart stack_trace_timer_fn(struct hrtimer *hrtimer)
|
|
|
|
{
|
|
|
|
/* trace here */
|
2008-05-12 15:20:47 -04:00
|
|
|
timer_notify(get_irq_regs(), smp_processor_id());
|
2008-05-12 15:20:47 -04:00
|
|
|
|
|
|
|
hrtimer_forward_now(hrtimer, ns_to_ktime(sample_period));
|
|
|
|
|
|
|
|
return HRTIMER_RESTART;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void start_stack_timer(int cpu)
|
|
|
|
{
|
|
|
|
struct hrtimer *hrtimer = &per_cpu(stack_trace_hrtimer, cpu);
|
|
|
|
|
|
|
|
hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
|
|
|
|
hrtimer->function = stack_trace_timer_fn;
|
2008-09-29 09:47:42 -04:00
|
|
|
hrtimer->cb_mode = HRTIMER_CB_IRQSAFE_PERCPU;
|
2008-05-12 15:20:47 -04:00
|
|
|
|
|
|
|
hrtimer_start(hrtimer, ns_to_ktime(sample_period), HRTIMER_MODE_REL);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void start_stack_timers(void)
|
|
|
|
{
|
|
|
|
cpumask_t saved_mask = current->cpus_allowed;
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for_each_online_cpu(cpu) {
|
2008-07-24 21:21:31 -04:00
|
|
|
set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
|
2008-05-12 15:20:47 -04:00
|
|
|
start_stack_timer(cpu);
|
|
|
|
}
|
|
|
|
set_cpus_allowed_ptr(current, &saved_mask);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void stop_stack_timer(int cpu)
|
|
|
|
{
|
|
|
|
struct hrtimer *hrtimer = &per_cpu(stack_trace_hrtimer, cpu);
|
|
|
|
|
|
|
|
hrtimer_cancel(hrtimer);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void stop_stack_timers(void)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
for_each_online_cpu(cpu)
|
|
|
|
stop_stack_timer(cpu);
|
|
|
|
}
|
|
|
|
|
2008-05-23 17:50:41 -04:00
|
|
|
static void stack_reset(struct trace_array *tr)
|
2008-05-12 15:20:47 -04:00
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
|
|
|
|
tr->time_start = ftrace_now(tr->cpu);
|
|
|
|
|
|
|
|
for_each_online_cpu(cpu)
|
2008-09-29 23:02:41 -04:00
|
|
|
tracing_reset(tr, cpu);
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
2008-05-23 17:50:41 -04:00
|
|
|
static void start_stack_trace(struct trace_array *tr)
|
2008-05-12 15:20:47 -04:00
|
|
|
{
|
2008-05-12 15:20:49 -04:00
|
|
|
mutex_lock(&sample_timer_lock);
|
2008-05-12 15:20:47 -04:00
|
|
|
stack_reset(tr);
|
2008-05-12 15:20:47 -04:00
|
|
|
start_stack_timers();
|
2008-05-12 15:20:47 -04:00
|
|
|
tracer_enabled = 1;
|
2008-05-12 15:20:49 -04:00
|
|
|
mutex_unlock(&sample_timer_lock);
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
2008-05-23 17:50:41 -04:00
|
|
|
static void stop_stack_trace(struct trace_array *tr)
|
2008-05-12 15:20:47 -04:00
|
|
|
{
|
2008-05-12 15:20:49 -04:00
|
|
|
mutex_lock(&sample_timer_lock);
|
2008-05-12 15:20:47 -04:00
|
|
|
stop_stack_timers();
|
2008-05-12 15:20:47 -04:00
|
|
|
tracer_enabled = 0;
|
2008-05-12 15:20:49 -04:00
|
|
|
mutex_unlock(&sample_timer_lock);
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
2008-11-15 23:57:26 -05:00
|
|
|
static int stack_trace_init(struct trace_array *tr)
|
2008-05-12 15:20:47 -04:00
|
|
|
{
|
2008-05-12 15:20:47 -04:00
|
|
|
sysprof_trace = tr;
|
2008-05-12 15:20:47 -04:00
|
|
|
|
2008-11-07 22:36:02 -05:00
|
|
|
start_stack_trace(tr);
|
2008-11-15 23:57:26 -05:00
|
|
|
return 0;
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
2008-05-23 17:50:41 -04:00
|
|
|
static void stack_trace_reset(struct trace_array *tr)
|
2008-05-12 15:20:47 -04:00
|
|
|
{
|
2008-11-07 22:36:02 -05:00
|
|
|
stop_stack_trace(tr);
|
2008-05-12 15:20:47 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct tracer stack_trace __read_mostly =
|
|
|
|
{
|
|
|
|
.name = "sysprof",
|
|
|
|
.init = stack_trace_init,
|
|
|
|
.reset = stack_trace_reset,
|
|
|
|
#ifdef CONFIG_FTRACE_SELFTEST
|
2008-05-12 15:20:47 -04:00
|
|
|
.selftest = trace_selftest_startup_sysprof,
|
2008-05-12 15:20:47 -04:00
|
|
|
#endif
|
|
|
|
};
|
|
|
|
|
|
|
|
__init static int init_stack_trace(void)
|
|
|
|
{
|
|
|
|
return register_tracer(&stack_trace);
|
|
|
|
}
|
|
|
|
device_initcall(init_stack_trace);
|
2008-05-12 15:20:49 -04:00
|
|
|
|
|
|
|
#define MAX_LONG_DIGITS 22
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
sysprof_sample_read(struct file *filp, char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
|
|
|
char buf[MAX_LONG_DIGITS];
|
|
|
|
int r;
|
|
|
|
|
|
|
|
r = sprintf(buf, "%ld\n", nsecs_to_usecs(sample_period));
|
|
|
|
|
|
|
|
return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
sysprof_sample_write(struct file *filp, const char __user *ubuf,
|
|
|
|
size_t cnt, loff_t *ppos)
|
|
|
|
{
|
|
|
|
char buf[MAX_LONG_DIGITS];
|
|
|
|
unsigned long val;
|
|
|
|
|
|
|
|
if (cnt > MAX_LONG_DIGITS-1)
|
|
|
|
cnt = MAX_LONG_DIGITS-1;
|
|
|
|
|
|
|
|
if (copy_from_user(&buf, ubuf, cnt))
|
|
|
|
return -EFAULT;
|
|
|
|
|
|
|
|
buf[cnt] = 0;
|
|
|
|
|
|
|
|
val = simple_strtoul(buf, NULL, 10);
|
|
|
|
/*
|
|
|
|
* Enforce a minimum sample period of 100 usecs:
|
|
|
|
*/
|
|
|
|
if (val < 100)
|
|
|
|
val = 100;
|
|
|
|
|
|
|
|
mutex_lock(&sample_timer_lock);
|
|
|
|
stop_stack_timers();
|
|
|
|
sample_period = val * 1000;
|
|
|
|
start_stack_timers();
|
|
|
|
mutex_unlock(&sample_timer_lock);
|
|
|
|
|
|
|
|
return cnt;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct file_operations sysprof_sample_fops = {
|
|
|
|
.read = sysprof_sample_read,
|
|
|
|
.write = sysprof_sample_write,
|
|
|
|
};
|
|
|
|
|
|
|
|
void init_tracer_sysprof_debugfs(struct dentry *d_tracer)
|
|
|
|
{
|
|
|
|
struct dentry *entry;
|
|
|
|
|
|
|
|
entry = debugfs_create_file("sysprof_sample_period", 0644,
|
|
|
|
d_tracer, NULL, &sysprof_sample_fops);
|
|
|
|
if (entry)
|
|
|
|
return;
|
|
|
|
pr_warning("Could not create debugfs 'dyn_ftrace_total_info' entry\n");
|
|
|
|
}
|