While cleaning the stacktrace code I unintentially changed the skip depth of
trace_buffer_unlock_commit_regs() from 0 to 6. kprobes uses this function,
and with skipping 6 call backs, it can easily produce no stack.
Here's how I tested it:
# echo 'p:ext4_sync_fs ext4_sync_fs ' > /sys/kernel/debug/tracing/kprobe_events
# echo 1 > /sys/kernel/debug/tracing/events/kprobes/enable
# cat /sys/kernel/debug/trace
sync-2394 [005] 502.457060: ext4_sync_fs: (
ffffffff81317650)
sync-2394 [005] 502.457063: kernel_stack: <stack trace>
sync-2394 [005] 502.457086: ext4_sync_fs: (
ffffffff81317650)
sync-2394 [005] 502.457087: kernel_stack: <stack trace>
sync-2394 [005] 502.457091: ext4_sync_fs: (
ffffffff81317650)
After putting back the skip stack to zero, we have:
sync-2270 [000] 748.052693: ext4_sync_fs: (
ffffffff81317650)
sync-2270 [000] 748.052695: kernel_stack: <stack trace>
=> iterate_supers (
ffffffff8126412e)
=> sys_sync (
ffffffff8129c4b6)
=> entry_SYSCALL_64_fastpath (
ffffffff8181f0b2)
sync-2270 [000] 748.053017: ext4_sync_fs: (
ffffffff81317650)
sync-2270 [000] 748.053019: kernel_stack: <stack trace>
=> iterate_supers (
ffffffff8126412e)
=> sys_sync (
ffffffff8129c4b6)
=> entry_SYSCALL_64_fastpath (
ffffffff8181f0b2)
sync-2270 [000] 748.053381: ext4_sync_fs: (
ffffffff81317650)
sync-2270 [000] 748.053383: kernel_stack: <stack trace>
=> iterate_supers (
ffffffff8126412e)
=> sys_sync (
ffffffff8129c4b6)
=> entry_SYSCALL_64_fastpath (
ffffffff8181f0b2)
Cc: stable@vger.kernel.org # v4.4+
Fixes: 73dddbb57bb0 "tracing: Only create stacktrace option when STACKTRACE is configured"
Reported-by: Brendan Gregg <brendan.d.gregg@gmail.com>
Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
{
__buffer_unlock_commit(buffer, event);
- ftrace_trace_stack(tr, buffer, flags, 6, pc, regs);
+ ftrace_trace_stack(tr, buffer, flags, 0, pc, regs);
ftrace_trace_userstack(buffer, flags, pc);
}
EXPORT_SYMBOL_GPL(trace_buffer_unlock_commit_regs);