Skip to content

RFC perf: remove __weak function hw_perf_group_sched_in()

Lin Ming lkml 2 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v2 RFC Latest4 patches 8 repliesIn review
  2. v1 RFC4 patches 7 repliesSuperseded

Patches in v2 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/4perf: core, add group scheduling transactional APIsinclude/linux/perf_event.h kernel/perf_event.c+37 -166 replies
2/4perf: x86, implement group scheduling transactional APIsarch/x86/kernel/cpu/perf_event.c+73 -1191 reply
3/4perf: powerpc, implement group scheduling transactional APIsarch/powerpc/kernel/perf_event.c+66 -640 replies
4/4perf: sparc, implement group scheduling transactional APIsarch/sparc/kernel/perf_event.c+65 -511 reply

Reviews trailers per patch

3 of 4 patches have a Reviewed-by or Acked-by.

PatchStephane EranianPeter ZijlstraFrederic WeisbeckerDavid Miller
1/4R-b R-b R-b R-b R-b
2/4R-b R-b R-b
3/4
4/4A-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes. Scroll the table sideways for all 4 reviewers.

Read the v2 thread

Cover letter [RFC][PATCH v2 0/4] perf: remove __weak function hw_perf_group_sched_in()

Hi, all

Quoted 1 line

From the discussion of thread "[RFC] perf_events: support for uncore

a.k.a. nest units", in order to support multiple hw pmu, Peter suggested the first thing that needs to be done is get rid of all the __weak functions (with exception of perf_callchain*, since that really is arch specific), includes

hw_perf_group_sched_in
hw_perf_event_init
hw_perf_disable
hw_perf_enable

Quoted 12 lines · Peter Zijlstra

One thing not on that list, which should happen first I guess, is to remove hw_perf_group_sched_in(). The idea is to add some sort of transactional API to the struct pmu, so that we can delay the schedulability check until commit time (and roll back when it fails).

Something as simple as:

 struct pmu {
   void start_txn(struct pmu *);
   void commit_txn(struct pmu *);

   ,,,
Quoted 6 lines

};

and then change group_sched_in() to use this instead of hw_perf_group_sched_in(), whose implementations mostly replicate group_sched_in() in various buggy ways anyway.

This patch series is to remove hw_perf_group_sched_in() first.

patch 1: core, add group scheduling transactional APIs patch 2: x86, implement group scheduling transactional APIs patch 3: powerpc, implement group scheduling transactional APIs patch 4: sparc, implement group scheduling transactional APIs

I have tested these patches on x86, but for powerpc and sparc it's only cross build tested.
It's very appreciated if anybody has powerpc or sparc can help to test.

Thanks for any comment.

Changes log:

v2.
Change transaction flag(cpu_hw_events->group_flag) from u8 to unsigned int. RISC architecture deals better with word sized variable than u8. (Peter Zijlstra)

Rename callback names, *_txn -> *_group_trans for better readability. Add a few comments to describe the callbacks. (Frederic Weisbecker)

v1.
remove __weak function hw_perf_group_sched_in() based on Peter's idea.

Thanks,
Lin Ming

Read the cover letter in its thread