Re: [Intel-gfx] [PATCH v8 0/7] drm/i915: use ref_tracker library for tracking wakerefs

From: Rodrigo Vivi
Date: Fri May 05 2023 - 16:06:48 EST

On Thu, May 04, 2023 at 06:27:53PM +0200, Andrzej Hajda wrote:
> Hi maintainers of net and i915,
>
> On 25.04.2023 00:05, Andrzej Hajda wrote:
> > This is revived patchset improving ref_tracker library and converting
> > i915 internal tracker to ref_tracker.
> > The old thread ended without consensus about small kernel allocations,
> > which are performed under spinlock.
> > I have tried to solve the problem by splitting the calls, but it results
> > in complicated API, so I went back to original solution.
> > If there are better solutions I am glad to discuss them.
> > Meanwhile I send original patchset with addressed remaining comments.
> >
> > To: Jani Nikula <jani.nikula@xxxxxxxxxxxxxxx>
> > To: Joonas Lahtinen <joonas.lahtinen@xxxxxxxxxxxxxxx>
> > To: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
> > To: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxxxxxxxx>
> > To: David Airlie <airlied@xxxxxxxxx>
> > To: Daniel Vetter <daniel@xxxxxxxx>
> > To: Eric Dumazet <edumazet@xxxxxxxxxx>
> > Cc: linux-kernel@xxxxxxxxxxxxxxx
> > Cc: intel-gfx@xxxxxxxxxxxxxxxxxxxxx
> > Cc: dri-devel@xxxxxxxxxxxxxxxxxxxxx
> > Cc: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> > Cc: netdev@xxxxxxxxxxxxxxx
> > Cc: Jakub Kicinski <kuba@xxxxxxxxxx>
> > Cc: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
> > Cc: "David S. Miller" <davem@xxxxxxxxxxxxx>
> > Cc: Andi Shyti <andi.shyti@xxxxxxxxxxxxxxx>
> > Cc: Das, Nirmoy <nirmoy.das@xxxxxxxxxxxxxxx>
> > Signed-off-by: Andrzej Hajda <andrzej.hajda@xxxxxxxxx>
> >
> > ---
> > Changes in v8:
> > - addressed comments from Eric, Zhou and CI, thanks,
> > - added ref_tracker_dir_init name argument to all callers in one patch
> > - moved intel_wakeref_tracker_show to *.c
> > - s/intel_wakeref_tracker_show/intel_ref_tracker_show/
> > - removed 'default n' from Kconfig
> > - changed strlcpy to strscpy,
> > - removed assignement from if condition,
> > - removed long lines from patch description
> > - added tags
> > - Link to v7: https://lore.kernel.org/r/20230224-track_gt-v7-0-11f08358c1ec@xxxxxxxxx
> >
> > Changes in v7:
> > - removed 8th patch (hold wakeref), as it was already merged
> > - added tags (thx Andi)
> > - Link to v6: https://lore.kernel.org/r/20230224-track_gt-v6-0-0dc8601fd02f@xxxxxxxxx
> >
> > Changes in v6:
> > - rebased to solve minor conflict and allow CI testing
> > - Link to v5: https://lore.kernel.org/r/20230224-track_gt-v5-0-77be86f2c872@xxxxxxxxx
> >
> > Changes in v5 (thx Andi for review):
> > - use *_locked convention instead of __*,
> > - improved commit messages,
> > - re-worked i915 patches, squashed separation and conversion patches,
> > - added tags,
> > - Link to v4: https://lore.kernel.org/r/20230224-track_gt-v4-0-464e8ab4c9ab@xxxxxxxxx
> >
> > Changes in v4:
> > - split "Separate wakeref tracking" to smaller parts
> > - fixed typos,
> > - Link to v1-v3: https://patchwork.freedesktop.org/series/100327/
> >
> > ---
> > Andrzej Hajda (7):
> > lib/ref_tracker: add unlocked leak print helper
> > lib/ref_tracker: improve printing stats
> > lib/ref_tracker: add printing to memory buffer
> > lib/ref_tracker: remove warnings in case of allocation failure
> > drm/i915: Correct type of wakeref variable
> > drm/i915: Replace custom intel runtime_pm tracker with ref_tracker library
> > drm/i915: Track gt pm wakerefs
>
> Finally all patches are reviewed.
> Question to network and i915 maintainers, how to merge this patchset:
> 1. Patches 1-4 belongs rather to network domain (especially patch 2).
> 2. Patches 5-7 are for i915.

Well, probably the easiest way to avoid conflicts would be to send
this right now through the net repo.

And hold patches 5-7 after drm-intel-next can backmerge them.

At this point I believe we would be looking at 6.5-rc2
backmerge to drm-intel-next in likely 11 weeks from now.

Do we have any urgency on them? Looking to all the changes in
i915 I believe we will get many conflicts if we let all these
i915 patches go through net tree as well.

>
> What would be the best way to do it?
>
> Regards
> Andrzej
>
>
>
> >
> > drivers/gpu/drm/i915/Kconfig.debug | 18 ++
> > drivers/gpu/drm/i915/display/intel_display_power.c | 2 +-
> > drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c | 7 +-
> > .../drm/i915/gem/selftests/i915_gem_coherency.c | 10 +-
> > drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c | 14 +-
> > drivers/gpu/drm/i915/gt/intel_breadcrumbs.c | 13 +-
> > drivers/gpu/drm/i915/gt/intel_breadcrumbs_types.h | 3 +-
> > drivers/gpu/drm/i915/gt/intel_context.h | 4 +-
> > drivers/gpu/drm/i915/gt/intel_context_types.h | 2 +
> > drivers/gpu/drm/i915/gt/intel_engine_pm.c | 7 +-
> > drivers/gpu/drm/i915/gt/intel_engine_types.h | 2 +
> > .../gpu/drm/i915/gt/intel_execlists_submission.c | 2 +-
> > drivers/gpu/drm/i915/gt/intel_gt_pm.c | 12 +-
> > drivers/gpu/drm/i915/gt/intel_gt_pm.h | 38 +++-
> > drivers/gpu/drm/i915/gt/intel_gt_pm_debugfs.c | 4 +-
> > drivers/gpu/drm/i915/gt/selftest_engine_cs.c | 20 +-
> > drivers/gpu/drm/i915/gt/selftest_gt_pm.c | 5 +-
> > drivers/gpu/drm/i915/gt/selftest_reset.c | 10 +-
> > drivers/gpu/drm/i915/gt/selftest_rps.c | 17 +-
> > drivers/gpu/drm/i915/gt/selftest_slpc.c | 5 +-
> > drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 12 +-
> > drivers/gpu/drm/i915/i915_driver.c | 2 +-
> > drivers/gpu/drm/i915/i915_pmu.c | 16 +-
> > drivers/gpu/drm/i915/intel_runtime_pm.c | 221 ++-------------------
> > drivers/gpu/drm/i915/intel_runtime_pm.h | 11 +-
> > drivers/gpu/drm/i915/intel_wakeref.c | 35 +++-
> > drivers/gpu/drm/i915/intel_wakeref.h | 73 ++++++-
> > include/linux/ref_tracker.h | 25 ++-
> > lib/ref_tracker.c | 179 ++++++++++++++---
> > lib/test_ref_tracker.c | 2 +-
> > net/core/dev.c | 2 +-
> > net/core/net_namespace.c | 4 +-
> > 32 files changed, 445 insertions(+), 332 deletions(-)
> > ---
> > base-commit: 4d0066a1c0763d50b6fb017e27d12b081ce21b57
> > change-id: 20230224-track_gt-1b3da8bdacd7
> >
> > Best regards,
>