I finally got around to implementing what I've been wanting to do for a long
time! That is to pass stacktraces from the start to the end event of a
synthetic event. Specifically, I need to get the stacktrace of a task as it
schedules out, but I don't want to see it until it schedules back in and I
check the timings to show that it is longer than normal. That is now done
with:
# echo 's:block_lat pid_t pid; u64 delta; unsigned long[] stack;' > dynamic_events
# echo 'hist:keys=next_pid:ts=common_timestamp.usecs,st=stacktrace if prev_state == 2' >> events/sched/sched_switch/trigger
# echo 'hist:keys=prev_pid:delta=common_timestamp.usecs-$ts,s=$st:onmax($delta).trace(block_lat,prev_pid,$delta,$s)' >> events/sched/sched_switch/trigger
The above creates a synthetic event that will trigger on the max delta of a
task blocked in an uninterruptible state. It will show you the stack trace
of where that occurred!
# echo 1 > events/synthetic/block_lat/enable
# ls -lR > /dev/null
# cat trace
# tracer: nop
#
# entries-in-buffer/entries-written: 2/2 #P:8
#
# _-----=> irqs-off/BH-disabled
# / _----=> need-resched
# | / _---=> hardirq/softirq
# || / _--=> preempt-depth
# ||| / _-=> migrate-disable
# |||| / delay
# TASK-PID CPU# ||||| TIMESTAMP FUNCTION
# | | | ||||| | |
<idle>-0 [005] d..4. 521.164922: block_lat: pid=0 delta=8322 stack=STACK:
=> __schedule+0x448/0x7b0
=> schedule+0x5a/0xb0
=> io_schedule+0x42/0x70
=> bit_wait_io+0xd/0x60
=> __wait_on_bit+0x4b/0x140
=> out_of_line_wait_on_bit+0x91/0xb0
=> jbd2_journal_commit_transaction+0x1679/0x1a70
=> kjournald2+0xa9/0x280
=> kthread+0xe9/0x110
=> ret_from_fork+0x2c/0x50
<...>-2 [004] d..4. 525.184257: block_lat: pid=2 delta=76 stack=STACK:
=> __schedule+0x448/0x7b0
=> schedule+0x5a/0xb0
=> schedule_timeout+0x11a/0x150
=> wait_for_completion_killable+0x144/0x1f0
=> __kthread_create_on_node+0xe7/0x1e0
=> kthread_create_on_node+0x51/0x70
=> create_worker+0xcc/0x1a0
=> worker_thread+0x2ad/0x380
=> kthread+0xe9/0x110
=> ret_from_fork+0x2c/0x50
Changes since v1: https://lore.kernel.org/linux-trace-kernel/[email protected]/
- Fixed an issue with passing a stacktrace variable to another variable.
- Can now use the stacktrace in the histogram as well:
# echo 'hist:delta.buckets=100,stack.stacktrace:sort=delta' > events/synthetic/block_lat/trigger
# cat events/synthetic/block_lat/hist
# event histogram
#
# trigger info: hist:keys=delta.buckets=100,stacktrace:vals=hitcount:sort=delta.buckets=100:size=2048 [active]
#
{ delta: ~ 0-99, stacktrace:
event_hist_trigger+0x464/0x480
event_triggers_call+0x52/0xe0
trace_event_buffer_commit+0x193/0x250
trace_event_raw_event_sched_switch+0xfc/0x150
__traceiter_sched_switch+0x41/0x60
__schedule+0x448/0x7b0
schedule_idle+0x26/0x40
cpu_startup_entry+0x19/0x20
start_secondary+0xed/0xf0
secondary_startup_64_no_verify+0xe0/0xeb
} hitcount: 6
{ delta: ~ 0-99, stacktrace:
event_hist_trigger+0x464/0x480
event_triggers_call+0x52/0xe0
trace_event_buffer_commit+0x193/0x250
trace_event_raw_event_sched_switch+0xfc/0x150
__traceiter_sched_switch+0x41/0x60
__schedule+0x448/0x7b0
schedule_idle+0x26/0x40
cpu_startup_entry+0x19/0x20
__pfx_kernel_init+0x0/0x10
arch_call_rest_init+0xa/0x24
start_kernel+0x964/0x98d
secondary_startup_64_no_verify+0xe0/0xeb
} hitcount: 3
[..]
{ delta: ~ 8500-8599, stacktrace:
event_hist_trigger+0x464/0x480
event_triggers_call+0x52/0xe0
trace_event_buffer_commit+0x193/0x250
trace_event_raw_event_sched_switch+0xfc/0x150
__traceiter_sched_switch+0x41/0x60
__schedule+0x448/0x7b0
schedule_idle+0x26/0x40
cpu_startup_entry+0x19/0x20
start_secondary+0xed/0xf0
secondary_startup_64_no_verify+0xe0/0xeb
} hitcount: 1
Totals:
Hits: 89
Entries: 11
Dropped: 0
Steven Rostedt (Google) (6):
tracing: Simplify calculating entry size using struct_size()
tracing: Allow stacktraces to be saved as histogram variables
tracing: Allow synthetic events to pass around stacktraces
tracing/histogram: Add stacktrace type
tracing/histogram: Document variable stacktrace
tracing/histogram: Add simple tests for stacktrace usage of synthetic events
----
Documentation/trace/histogram.rst | 129 +++++++++++++++++++++
kernel/trace/trace.c | 2 +-
kernel/trace/trace.h | 4 +
kernel/trace/trace_events_hist.c | 63 +++++++---
kernel/trace/trace_events_synth.c | 83 ++++++++++++-
kernel/trace/trace_synth.h | 1 +
.../inter-event/trigger-synthetic-event-stack.tc | 24 ++++
.../inter-event/trigger-synthetic-event-syntax.tc | 6 +
8 files changed, 294 insertions(+), 18 deletions(-)
create mode 100644 tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-synthetic-event-stack.tc
Hi Steve,
On Tue, 2023-01-17 at 10:21 -0500, Steven Rostedt wrote:
> I finally got around to implementing what I've been wanting to do for
> a long
> time! That is to pass stacktraces from the start to the end event of
> a
> synthetic event. Specifically, I need to get the stacktrace of a task
Hey, this is really great! Thanks for doing this, and sorry I haven't
gotten around to looking at it until now.
Anyway, while playing with it I found a few things that didn't work as
expected, and will send a few patches with some fixes shortly.
Thanks again, this is really useful!
Tom
> as it
> schedules out, but I don't want to see it until it schedules back in
> and I
> check the timings to show that it is longer than normal. That is now
> done
> with:
>
> # echo 's:block_lat pid_t pid; u64 delta; unsigned long[] stack;' >
> dynamic_events
> # echo 'hist:keys=next_pid:ts=common_timestamp.usecs,st=stacktrace
> if prev_state == 2' >> events/sched/sched_switch/trigger
> # echo 'hist:keys=prev_pid:delta=common_timestamp.usecs-
> $ts,s=$st:onmax($delta).trace(block_lat,prev_pid,$delta,$s)' >>
> events/sched/sched_switch/trigger
>
> The above creates a synthetic event that will trigger on the max
> delta of a
> task blocked in an uninterruptible state. It will show you the stack
> trace
> of where that occurred!
>
> # echo 1 > events/synthetic/block_lat/enable
> # ls -lR > /dev/null
> # cat trace
>
> # tracer: nop
> #
> # entries-in-buffer/entries-written: 2/2 #P:8
> #
> # _-----=> irqs-off/BH-disabled
> # / _----=> need-resched
> # | / _---=> hardirq/softirq
> # || / _--=> preempt-depth
> # ||| / _-=> migrate-disable
> # |||| / delay
> # TASK-PID CPU# ||||| TIMESTAMP FUNCTION
> # | | | ||||| | |
> <idle>-0 [005] d..4. 521.164922: block_lat: pid=0
> delta=8322 stack=STACK:
> => __schedule+0x448/0x7b0
> => schedule+0x5a/0xb0
> => io_schedule+0x42/0x70
> => bit_wait_io+0xd/0x60
> => __wait_on_bit+0x4b/0x140
> => out_of_line_wait_on_bit+0x91/0xb0
> => jbd2_journal_commit_transaction+0x1679/0x1a70
> => kjournald2+0xa9/0x280
> => kthread+0xe9/0x110
> => ret_from_fork+0x2c/0x50
>
> <...>-2 [004] d..4. 525.184257: block_lat: pid=2
> delta=76 stack=STACK:
> => __schedule+0x448/0x7b0
> => schedule+0x5a/0xb0
> => schedule_timeout+0x11a/0x150
> => wait_for_completion_killable+0x144/0x1f0
> => __kthread_create_on_node+0xe7/0x1e0
> => kthread_create_on_node+0x51/0x70
> => create_worker+0xcc/0x1a0
> => worker_thread+0x2ad/0x380
> => kthread+0xe9/0x110
> => ret_from_fork+0x2c/0x50
>
> Changes since v1:
> https://lore.kernel.org/linux-trace-kernel/[email protected]/
>
> - Fixed an issue with passing a stacktrace variable to another
> variable.
>
> - Can now use the stacktrace in the histogram as well:
>
> # echo 'hist:delta.buckets=100,stack.stacktrace:sort=delta' >
> events/synthetic/block_lat/trigger
> # cat events/synthetic/block_lat/hist
>
> # event histogram
> #
> # trigger info:
> hist:keys=delta.buckets=100,stacktrace:vals=hitcount:sort=delta.bucke
> ts=100:size=2048 [active]
> #
>
> { delta: ~ 0-99, stacktrace:
> event_hist_trigger+0x464/0x480
> event_triggers_call+0x52/0xe0
> trace_event_buffer_commit+0x193/0x250
> trace_event_raw_event_sched_switch+0xfc/0x150
> __traceiter_sched_switch+0x41/0x60
> __schedule+0x448/0x7b0
> schedule_idle+0x26/0x40
> cpu_startup_entry+0x19/0x20
> start_secondary+0xed/0xf0
> secondary_startup_64_no_verify+0xe0/0xeb
> } hitcount: 6
> { delta: ~ 0-99, stacktrace:
> event_hist_trigger+0x464/0x480
> event_triggers_call+0x52/0xe0
> trace_event_buffer_commit+0x193/0x250
> trace_event_raw_event_sched_switch+0xfc/0x150
> __traceiter_sched_switch+0x41/0x60
> __schedule+0x448/0x7b0
> schedule_idle+0x26/0x40
> cpu_startup_entry+0x19/0x20
> __pfx_kernel_init+0x0/0x10
> arch_call_rest_init+0xa/0x24
> start_kernel+0x964/0x98d
> secondary_startup_64_no_verify+0xe0/0xeb
> } hitcount: 3
> [..]
> { delta: ~ 8500-8599, stacktrace:
> event_hist_trigger+0x464/0x480
> event_triggers_call+0x52/0xe0
> trace_event_buffer_commit+0x193/0x250
> trace_event_raw_event_sched_switch+0xfc/0x150
> __traceiter_sched_switch+0x41/0x60
> __schedule+0x448/0x7b0
> schedule_idle+0x26/0x40
> cpu_startup_entry+0x19/0x20
> start_secondary+0xed/0xf0
> secondary_startup_64_no_verify+0xe0/0xeb
> } hitcount: 1
>
> Totals:
> Hits: 89
> Entries: 11
> Dropped: 0
>
>
>
> Steven Rostedt (Google) (6):
> tracing: Simplify calculating entry size using struct_size()
> tracing: Allow stacktraces to be saved as histogram variables
> tracing: Allow synthetic events to pass around stacktraces
> tracing/histogram: Add stacktrace type
> tracing/histogram: Document variable stacktrace
> tracing/histogram: Add simple tests for stacktrace usage of
> synthetic events
>
> ----
> Documentation/trace/histogram.rst | 129
> +++++++++++++++++++++
> kernel/trace/trace.c | 2 +-
> kernel/trace/trace.h | 4 +
> kernel/trace/trace_events_hist.c | 63 +++++++---
> kernel/trace/trace_events_synth.c | 83
> ++++++++++++-
> kernel/trace/trace_synth.h | 1 +
> .../inter-event/trigger-synthetic-event-stack.tc | 24 ++++
> .../inter-event/trigger-synthetic-event-syntax.tc | 6 +
> 8 files changed, 294 insertions(+), 18 deletions(-)
> create mode 100644
> tools/testing/selftests/ftrace/test.d/trigger/inter-event/trigger-
> synthetic-event-stack.tc