I'm capturing events on my x64 Dell XPS 13 (i7-6560U) laptop w/ Debian
Test and kernel 4.13.4. For the capture, I'm doing:
trace-cmd start -b 2000 -D -i -p function -l "*vblank*" -e i915:i915_pipe_update_vblank_evaded
-e sched:sched_switch -e sched:sched_process_fork -e sched:sched_process_exec
-e sched:sched_process_exit -e drm:drm_vblank_event -e drm:drm_vblank_event_queued
-e drm:drm_vblank_event_delivered -e *fence:*fence_signaled -e i915:i915_flip_request
-e i915:i915_flip_complete -e i915:intel_gpu_freq_change -e i915:i915_gem_request_add
-e i915:i915_gem_request_submit -e i915:i915_gem_request_in -e i915:i915_gem_request_out
-e i915:intel_engine_notify -e i915:i915_gem_request_wait_begin -e i915:i915_gem_request_wait_end
<... run app ...>
trace-cmd snapshot -s
trace-cm extract -s -k -o trace.dat
When I do a capture running just "glxgears -fullscreen", every
drm_vblank_event exists where expected, and has the expected events
ahead of it. I can grep sequence numbers in the trace and get this every
time:
trace-cmd report unloaded_system.dat | grep "seq=50856"
Xorg-666 [001] 850.248191: drm_vblank_event_queued: file=0xffff916525fcec00, crtc=0, seq=50856
Xorg-666 [001] 850.263652: drm_vblank_event_queued: file=0xffff916525fcec00, crtc=0, seq=50856
<idle>-0 [001] 850.264745: drm_vblank_event_delivered: file=0xffff916525fcec00, crtc=0, seq=50856
<idle>-0 [001] 850.264753: drm_vblank_event_delivered: file=0xffff916525fcec00, crtc=0, seq=50856
<idle>-0 [001] 850.264756: drm_vblank_event_delivered: file=0xffff916525fcec00, crtc=0, seq=50856
<idle>-0 [001] 850.264758: drm_vblank_event: crtc=0, seq=50856
On a loaded system (running Unigine Heaven demo + glxgears -fullscreen),
the above happens most of the time, but there are several missing
drm_vblank_events and several drm_vblank_events that don't have any
matching drm_vblank_event_queued or delivered events. Example:
trace-cmd report loaded_system.dat | grep "seq=93977"
<idle>-0 [001] 5161.664754: drm_vblank_event: crtc=0, seq=93977
This looks to me like we are missing events in the capture. Is this
possible? Is there any way to mitigate if so?
Thanks much.
-Mike
--
To unsubscribe from this list: send the line "unsubscribe linux-trace-users" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
|