Messages in this thread Patch in this message | | | Date | Fri, 7 Nov 2008 08:59:25 +0100 | From | Ingo Molnar <> | Subject | Re: [patch] add /proc/pid/stack to dump task's stack trace |
| |
* Ingo Molnar <mingo@elte.hu> wrote:
> > > + * buffer size used for proc read. See proc_info_read(). > > > + * 4K page size but our output routines use some slack for overruns > > > + */ > > > +#define PROC_BLOCK_SIZE (3*1024) > > That sounds like a proper limit - the hard limit for this particular > printout function is 4096-170, so we are well within the > PROC_BLOCK_SIZE range.
ok, i've added Ken's patch to tip/core/stacktrace and started testing it.
Alexey, i've added your Acked-by because you appeared to like the patch - let me know if i should remove it.
i've done a few finishing touches to the patch as well - see the end result below.
Andrew, i remember that you found some sort of problem (crashes?) with stack-dumping tasks that are running on another CPU (or something like that) - do you remember the specifics? Could we run into any of those problems with the patch below, on some rare architecture?
Ingo
-------------------> From d3d23f82adeb5ec8a9546747d1df058dcaad0589 Mon Sep 17 00:00:00 2001 From: Ken Chen <kenchen@google.com> Date: Thu, 6 Nov 2008 16:30:23 -0800 Subject: [PATCH] stacktrace: add /proc/<pid>/stack to dump task's stack trace
Impact: add the new (root-only) /proc/<pid>/stack debug facility
This patch adds the ability to query a task's current stack trace via /proc/<pid>/stack.
It is considered to be more useful than /proc/pid/wchan as it provides full stack trace instead of single depth.
It is also more useful than sysrq-t because it does not overflow the dmesg like sysrq-t often does, can be read in a finegrained per-task way and can be read programmatically as well.
It works on sleeping and running tasks as well.
Also, move up PROC_BLOCK_SIZE a bit so that proc_pid_stack() can use it.
[ mingo@elte.hu: small cleanups, comments ]
Signed-off-by: Ken Chen <kenchen@google.com> Acked-by: Alexey Dobriyan <adobriyan@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> --- Documentation/filesystems/proc.txt | 1 + fs/proc/base.c | 52 ++++++++++++++++++++++++++++++++++- 2 files changed, 51 insertions(+), 2 deletions(-)
diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt index bcceb99..11f5b75 100644 --- a/Documentation/filesystems/proc.txt +++ b/Documentation/filesystems/proc.txt @@ -139,6 +139,7 @@ Table 1-1: Process specific entries in /proc statm Process memory status information status Process status in human readable form wchan If CONFIG_KALLSYMS is set, a pre-decoded wchan + stack Report full stack trace, enable via CONFIG_STACKTRACE smaps Extension based on maps, the rss size for each mapped file .............................................................................. diff --git a/fs/proc/base.c b/fs/proc/base.c index 486cf3f..805e514 100644 --- a/fs/proc/base.c +++ b/fs/proc/base.c @@ -65,6 +65,7 @@ #include <linux/mm.h> #include <linux/rcupdate.h> #include <linux/kallsyms.h> +#include <linux/stacktrace.h> #include <linux/resource.h> #include <linux/module.h> #include <linux/mount.h> @@ -130,6 +131,12 @@ struct pid_entry { { .proc_show = &proc_##OTYPE } ) /* + * buffer size used for proc read. See proc_info_read(). + * 4K page size but our output routines use some slack for overruns + */ +#define PROC_BLOCK_SIZE (3*1024) + +/* * Count the number of hardlinks for the pid_entry table, excluding the . * and .. links. */ @@ -340,6 +347,46 @@ static int proc_pid_wchan(struct task_struct *task, char *buffer) } #endif /* CONFIG_KALLSYMS */ +#ifdef CONFIG_STACKTRACE + +#define MAX_STACK_TRACE_DEPTH 64 + +static int proc_pid_stack(struct task_struct *task, char *buffer) +{ + struct stack_trace trace; + unsigned long *entries; + int i, len = 0; + + entries = kmalloc(sizeof(*entries)*MAX_STACK_TRACE_DEPTH, GFP_KERNEL); + if (!entries) + return -ENOMEM; + + trace.nr_entries = 0; + trace.max_entries = MAX_STACK_TRACE_DEPTH; + trace.entries = entries; + trace.skip = 0; + + /* + * Protect against the task exiting (and deallocating its + * stack, etc.) while we save its backtrace: + */ + read_lock(&tasklist_lock); + save_stack_trace_tsk(task, &trace); + read_unlock(&tasklist_lock); + + for (i = 0; i < trace.nr_entries; i++) { + len += snprintf(buffer + len, PROC_BLOCK_SIZE - len, + "[<%p>] %pS\n", + (void *)entries[i], (void *)entries[i]); + if (!len) + break; + } + kfree(entries); + + return len; +} +#endif + #ifdef CONFIG_SCHEDSTATS /* * Provides /proc/PID/schedstat @@ -688,8 +735,6 @@ static const struct file_operations proc_mountstats_operations = { .release = mounts_release, }; -#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */ - static ssize_t proc_info_read(struct file * file, char __user * buf, size_t count, loff_t *ppos) { @@ -2491,6 +2536,9 @@ static const struct pid_entry tgid_base_stuff[] = { #ifdef CONFIG_KALLSYMS INF("wchan", S_IRUGO, pid_wchan), #endif +#ifdef CONFIG_STACKTRACE + INF("stack", S_IRUSR, pid_stack), +#endif #ifdef CONFIG_SCHEDSTATS INF("schedstat", S_IRUGO, pid_schedstat), #endif
| |