Skip to content

perf/x86/uncore: Overflow handling enhancements

Sandipan Das linux-perf-users lkml 2 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v2 Latest5 patches 4 repliesIn review
  2. v14 patches 3 repliesSuperseded

Changes in v2 from the cover letter

- Add a Fixes tag for the first patch.
- Change the hrtimer mode for both the Intel and the AMD uncore drivers based on Peter's suggestion.
- Fix an issue in the UMC workaround where prev_count is not zeroed out after a counter is reset.

Patches in v2 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/5perf/x86/amd/uncore: Remove unused member from amd_uncore_ctxarch/x86/events/amd/uncore.c+0 -10 replies
2/5perf/x86/intel/uncore: Use HRTIMER_MODE_HARD for detecting overflowsarch/x86/events/intel/uncore.c+2 -100 replies
3/5perf/x86/amd/uncore: Use hrtimer for handling overflowsarch/x86/events/amd/uncore.c+63 -02 replies
4/5perf/x86/amd/uncore: Add parameter to configure hrtimerarch/x86/events/amd/uncore.c+5 -10 replies
5/5perf/x86/amd/uncore: Prevent UMC counters from saturatingarch/x86/events/amd/uncore.c+34 -10 replies

Reviews trailers per patch

1 of 5 patches has a Reviewed-by or Acked-by.

PatchSong Liu
1/5
2/5
3/5
4/5
5/5R-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v2 thread

Cover letter [PATCH v2 0/5] perf/x86/uncore: Overflow handling enhancements

Uncore counters on AMD processors either roll over or saturate on overflow and the amd-uncore driver has no way of knowing if multiple overflows have occurred between two successive pmu->read() requests for an event. This makes the user-visible counter values inaccurate. Solve this by periodically initiating pmu->read() in order to keep prev_count up-to-date. The approach follows Intel's precedent in handling uncore counters.

Previous versions can be found at:
v1: https://lore.kernel.org/all/[email protected]/

Changes in v2:
- Add a Fixes tag for the first patch.
- Change the hrtimer mode for both the Intel and the AMD uncore drivers based on Peter's suggestion.
- Fix an issue in the UMC workaround where prev_count is not zeroed out after a counter is reset.

Sandipan Das (5):
perf/x86/amd/uncore: Remove unused member from amd_uncore_ctx perf/x86/intel/uncore: Use HRTIMER_MODE_HARD for detecting overflows perf/x86/amd/uncore: Use hrtimer for handling overflows perf/x86/amd/uncore: Add parameter to configure hrtimer perf/x86/amd/uncore: Prevent UMC counters from saturating

 arch/x86/events/amd/uncore.c   | 103 ++++++++++++++++++++++++++++++++-
 arch/x86/events/intel/uncore.c |  12 +---
 2 files changed, 103 insertions(+), 12 deletions(-)
Signature · 1 line
2.43.0