From: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> To: "Christian König" <ckoenig.leichtzumerken@gmail.com>, linaro-mm-sig@lists.linaro.org, dri-devel@lists.freedesktop.org, linux-media@vger.kernel.org, intel-gfx@lists.freedesktop.org Cc: daniel@ffwll.ch Subject: Re: [PATCH 02/28] dma-buf: add dma_resv_for_each_fence v2 Date: Wed, 6 Oct 2021 09:40:44 +0100 [thread overview] Message-ID: <fc9ac0dd-a069-7f34-d94d-ae6cd70d307f@linux.intel.com> (raw) In-Reply-To: <20211005113742.1101-3-christian.koenig@amd.com> On 05/10/2021 12:37, Christian König wrote: > A simpler version of the iterator to be used when the dma_resv object is > locked. > > v2: fix index check here as well > > Signed-off-by: Christian König <christian.koenig@amd.com> > --- > drivers/dma-buf/dma-resv.c | 49 ++++++++++++++++++++++++++++++++++++++ > include/linux/dma-resv.h | 19 +++++++++++++++ > 2 files changed, 68 insertions(+) > > diff --git a/drivers/dma-buf/dma-resv.c b/drivers/dma-buf/dma-resv.c > index 3cbcf66a137e..231bae173ef1 100644 > --- a/drivers/dma-buf/dma-resv.c > +++ b/drivers/dma-buf/dma-resv.c > @@ -423,6 +423,55 @@ struct dma_fence *dma_resv_iter_next_unlocked(struct dma_resv_iter *cursor) > } > EXPORT_SYMBOL(dma_resv_iter_next_unlocked); > > +/** > + * dma_resv_iter_first - first fence from a locked dma_resv object > + * @cursor: cursor to record the current position > + * > + * Return all the fences in the dma_resv object while holding the > + * &dma_resv.lock. > + */ > +struct dma_fence *dma_resv_iter_first(struct dma_resv_iter *cursor) > +{ > + struct dma_fence *fence; > + > + dma_resv_assert_held(cursor->obj); > + > + cursor->index = 0; > + cursor->fences = dma_resv_shared_list(cursor->obj); > + > + fence = dma_resv_excl_fence(cursor->obj); > + if (!fence) > + fence = dma_resv_iter_next(cursor); "Is restarted" probably does not matter hugely for the locked iterator but I think if it hits this path (no exclusive fence, returns first shared) then it will show it as false. Which is not consistent with the unlocked iterator. Bonus points if you make a debug build assert that makes querying "is restarted" warn when used with the locked iterator. > + > + cursor->is_restarted = true; > + return fence; > +} > +EXPORT_SYMBOL_GPL(dma_resv_iter_first); > + > +/** > + * dma_resv_iter_next - next fence from a locked dma_resv object > + * @cursor: cursor to record the current position > + * > + * Return all the fences in the dma_resv object while holding the > + * &dma_resv.lock. You probably want to replace "all the fences" with first and next, respectively, in here and in dma_resv_iter_first kerneldoc. > + */ > +struct dma_fence *dma_resv_iter_next(struct dma_resv_iter *cursor) > +{ > + unsigned int idx; > + > + dma_resv_assert_held(cursor->obj); > + > + cursor->is_restarted = false; > + if (!cursor->all_fences || !cursor->fences || > + cursor->index >= cursor->fences->shared_count) > + return NULL; Theoretically you could store the shared count in the cursor and so could have a single condition here (assuming initialized to zero when !all_fences and !cursor->fences). For some value of optimisation. :) Probably not worth it. But you could only assign cursor->fences if all_fences, in dma_resv_iter_first, so wouldn't have to duplicate the all_fences check here. > + > + idx = cursor->index++; > + return rcu_dereference_protected(cursor->fences->shared[idx], > + dma_resv_held(cursor->obj)); > +} > +EXPORT_SYMBOL_GPL(dma_resv_iter_next); > + > /** > * dma_resv_copy_fences - Copy all fences from src to dst. > * @dst: the destination reservation object > diff --git a/include/linux/dma-resv.h b/include/linux/dma-resv.h > index 764138ad8583..3df7ef23712d 100644 > --- a/include/linux/dma-resv.h > +++ b/include/linux/dma-resv.h > @@ -179,6 +179,8 @@ struct dma_resv_iter { > > struct dma_fence *dma_resv_iter_first_unlocked(struct dma_resv_iter *cursor); > struct dma_fence *dma_resv_iter_next_unlocked(struct dma_resv_iter *cursor); > +struct dma_fence *dma_resv_iter_first(struct dma_resv_iter *cursor); > +struct dma_fence *dma_resv_iter_next(struct dma_resv_iter *cursor); > > /** > * dma_resv_iter_begin - initialize a dma_resv_iter object > @@ -244,6 +246,23 @@ static inline bool dma_resv_iter_is_restarted(struct dma_resv_iter *cursor) > for (fence = dma_resv_iter_first_unlocked(cursor); \ > fence; fence = dma_resv_iter_next_unlocked(cursor)) > > +/** > + * dma_resv_for_each_fence - fence iterator > + * @cursor: a struct dma_resv_iter pointer > + * @obj: a dma_resv object pointer > + * @all_fences: true if all fences should be returned > + * @fence: the current fence > + * > + * Iterate over the fences in a struct dma_resv object while holding the > + * &dma_resv.lock. @all_fences controls if the shared fences are returned as > + * well. The cursor initialisation is part of the iterator and the fence stays > + * valid as long as the lock is held. I'd be super cautious and explicitly spell out that reference is not held in contrast to the unlocked iterator. > + */ > +#define dma_resv_for_each_fence(cursor, obj, all_fences, fence) \ > + for (dma_resv_iter_begin(cursor, obj, all_fences), \ > + fence = dma_resv_iter_first(cursor); fence; \ > + fence = dma_resv_iter_next(cursor)) > + > #define dma_resv_held(obj) lockdep_is_held(&(obj)->lock.base) > #define dma_resv_assert_held(obj) lockdep_assert_held(&(obj)->lock.base) > > Regards, Tvrtko
WARNING: multiple messages have this Message-ID (diff)
From: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> To: "Christian König" <ckoenig.leichtzumerken@gmail.com>, linaro-mm-sig@lists.linaro.org, dri-devel@lists.freedesktop.org, linux-media@vger.kernel.org, intel-gfx@lists.freedesktop.org Cc: daniel@ffwll.ch Subject: Re: [Intel-gfx] [PATCH 02/28] dma-buf: add dma_resv_for_each_fence v2 Date: Wed, 6 Oct 2021 09:40:44 +0100 [thread overview] Message-ID: <fc9ac0dd-a069-7f34-d94d-ae6cd70d307f@linux.intel.com> (raw) In-Reply-To: <20211005113742.1101-3-christian.koenig@amd.com> On 05/10/2021 12:37, Christian König wrote: > A simpler version of the iterator to be used when the dma_resv object is > locked. > > v2: fix index check here as well > > Signed-off-by: Christian König <christian.koenig@amd.com> > --- > drivers/dma-buf/dma-resv.c | 49 ++++++++++++++++++++++++++++++++++++++ > include/linux/dma-resv.h | 19 +++++++++++++++ > 2 files changed, 68 insertions(+) > > diff --git a/drivers/dma-buf/dma-resv.c b/drivers/dma-buf/dma-resv.c > index 3cbcf66a137e..231bae173ef1 100644 > --- a/drivers/dma-buf/dma-resv.c > +++ b/drivers/dma-buf/dma-resv.c > @@ -423,6 +423,55 @@ struct dma_fence *dma_resv_iter_next_unlocked(struct dma_resv_iter *cursor) > } > EXPORT_SYMBOL(dma_resv_iter_next_unlocked); > > +/** > + * dma_resv_iter_first - first fence from a locked dma_resv object > + * @cursor: cursor to record the current position > + * > + * Return all the fences in the dma_resv object while holding the > + * &dma_resv.lock. > + */ > +struct dma_fence *dma_resv_iter_first(struct dma_resv_iter *cursor) > +{ > + struct dma_fence *fence; > + > + dma_resv_assert_held(cursor->obj); > + > + cursor->index = 0; > + cursor->fences = dma_resv_shared_list(cursor->obj); > + > + fence = dma_resv_excl_fence(cursor->obj); > + if (!fence) > + fence = dma_resv_iter_next(cursor); "Is restarted" probably does not matter hugely for the locked iterator but I think if it hits this path (no exclusive fence, returns first shared) then it will show it as false. Which is not consistent with the unlocked iterator. Bonus points if you make a debug build assert that makes querying "is restarted" warn when used with the locked iterator. > + > + cursor->is_restarted = true; > + return fence; > +} > +EXPORT_SYMBOL_GPL(dma_resv_iter_first); > + > +/** > + * dma_resv_iter_next - next fence from a locked dma_resv object > + * @cursor: cursor to record the current position > + * > + * Return all the fences in the dma_resv object while holding the > + * &dma_resv.lock. You probably want to replace "all the fences" with first and next, respectively, in here and in dma_resv_iter_first kerneldoc. > + */ > +struct dma_fence *dma_resv_iter_next(struct dma_resv_iter *cursor) > +{ > + unsigned int idx; > + > + dma_resv_assert_held(cursor->obj); > + > + cursor->is_restarted = false; > + if (!cursor->all_fences || !cursor->fences || > + cursor->index >= cursor->fences->shared_count) > + return NULL; Theoretically you could store the shared count in the cursor and so could have a single condition here (assuming initialized to zero when !all_fences and !cursor->fences). For some value of optimisation. :) Probably not worth it. But you could only assign cursor->fences if all_fences, in dma_resv_iter_first, so wouldn't have to duplicate the all_fences check here. > + > + idx = cursor->index++; > + return rcu_dereference_protected(cursor->fences->shared[idx], > + dma_resv_held(cursor->obj)); > +} > +EXPORT_SYMBOL_GPL(dma_resv_iter_next); > + > /** > * dma_resv_copy_fences - Copy all fences from src to dst. > * @dst: the destination reservation object > diff --git a/include/linux/dma-resv.h b/include/linux/dma-resv.h > index 764138ad8583..3df7ef23712d 100644 > --- a/include/linux/dma-resv.h > +++ b/include/linux/dma-resv.h > @@ -179,6 +179,8 @@ struct dma_resv_iter { > > struct dma_fence *dma_resv_iter_first_unlocked(struct dma_resv_iter *cursor); > struct dma_fence *dma_resv_iter_next_unlocked(struct dma_resv_iter *cursor); > +struct dma_fence *dma_resv_iter_first(struct dma_resv_iter *cursor); > +struct dma_fence *dma_resv_iter_next(struct dma_resv_iter *cursor); > > /** > * dma_resv_iter_begin - initialize a dma_resv_iter object > @@ -244,6 +246,23 @@ static inline bool dma_resv_iter_is_restarted(struct dma_resv_iter *cursor) > for (fence = dma_resv_iter_first_unlocked(cursor); \ > fence; fence = dma_resv_iter_next_unlocked(cursor)) > > +/** > + * dma_resv_for_each_fence - fence iterator > + * @cursor: a struct dma_resv_iter pointer > + * @obj: a dma_resv object pointer > + * @all_fences: true if all fences should be returned > + * @fence: the current fence > + * > + * Iterate over the fences in a struct dma_resv object while holding the > + * &dma_resv.lock. @all_fences controls if the shared fences are returned as > + * well. The cursor initialisation is part of the iterator and the fence stays > + * valid as long as the lock is held. I'd be super cautious and explicitly spell out that reference is not held in contrast to the unlocked iterator. > + */ > +#define dma_resv_for_each_fence(cursor, obj, all_fences, fence) \ > + for (dma_resv_iter_begin(cursor, obj, all_fences), \ > + fence = dma_resv_iter_first(cursor); fence; \ > + fence = dma_resv_iter_next(cursor)) > + > #define dma_resv_held(obj) lockdep_is_held(&(obj)->lock.base) > #define dma_resv_assert_held(obj) lockdep_assert_held(&(obj)->lock.base) > > Regards, Tvrtko
next prev parent reply other threads:[~2021-10-06 8:40 UTC|newest] Thread overview: 131+ messages / expand[flat|nested] mbox.gz Atom feed top 2021-10-05 11:37 Deploying new iterator interface for dma-buf Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 01/28] dma-buf: add dma_resv_for_each_fence_unlocked v8 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 02/28] dma-buf: add dma_resv_for_each_fence v2 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-06 8:24 ` Christian König 2021-10-06 8:24 ` [Intel-gfx] " Christian König 2021-10-06 8:40 ` Tvrtko Ursulin [this message] 2021-10-06 8:40 ` Tvrtko Ursulin 2021-10-06 8:52 ` Tvrtko Ursulin 2021-10-06 8:52 ` [Intel-gfx] " Tvrtko Ursulin 2021-10-05 11:37 ` [PATCH 03/28] dma-buf: add dma_resv selftest v3 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:04 ` Daniel Vetter 2021-10-13 14:04 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 04/28] dma-buf: use new iterator in dma_resv_copy_fences Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 05/28] dma-buf: use new iterator in dma_resv_get_fences v3 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 06/28] dma-buf: use new iterator in dma_resv_wait_timeout Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 07/28] dma-buf: use new iterator in dma_resv_test_signaled Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 08/28] dma-buf: use the new iterator in dma_buf_debug_show Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 09/28] dma-buf: use the new iterator in dma_resv_poll Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 10/28] drm/ttm: use the new iterator in ttm_bo_flush_all_fences Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 11/28] drm/amdgpu: use the new iterator in amdgpu_sync_resv Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:06 ` Daniel Vetter 2021-10-13 14:06 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 12/28] drm/amdgpu: use new iterator in amdgpu_ttm_bo_eviction_valuable Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:07 ` Daniel Vetter 2021-10-13 14:07 ` [Intel-gfx] " Daniel Vetter 2021-10-19 11:36 ` Christian König 2021-10-19 11:36 ` [Intel-gfx] " Christian König 2021-10-19 16:30 ` Felix Kuehling 2021-10-19 16:30 ` [Intel-gfx] " Felix Kuehling 2021-10-21 11:29 ` Daniel Vetter 2021-10-21 11:29 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 13/28] drm/amdgpu: use new iterator in amdgpu_vm_prt_fini Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:12 ` Daniel Vetter 2021-10-13 14:12 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 14/28] drm/msm: use new iterator in msm_gem_describe Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:14 ` Daniel Vetter 2021-10-13 14:14 ` [Intel-gfx] " Daniel Vetter 2021-10-19 11:49 ` Christian König 2021-10-19 11:49 ` [Intel-gfx] " Christian König 2021-10-21 11:30 ` Daniel Vetter 2021-10-21 11:30 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 15/28] drm/radeon: use new iterator in radeon_sync_resv Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:15 ` Daniel Vetter 2021-10-13 14:15 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 16/28] drm/scheduler: use new iterator in drm_sched_job_add_implicit_dependencies v2 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-17 14:40 ` Nicolas Frattaroli 2021-10-17 14:40 ` [Intel-gfx] " Nicolas Frattaroli 2021-10-17 15:26 ` Christian König 2021-10-17 15:26 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 17/28] drm/i915: use the new iterator in i915_gem_busy_ioctl v2 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 12:40 ` Tvrtko Ursulin 2021-10-05 12:40 ` [Intel-gfx] " Tvrtko Ursulin 2021-10-05 12:44 ` Christian König 2021-10-05 12:44 ` [Intel-gfx] " Christian König 2021-10-13 14:19 ` Daniel Vetter 2021-10-13 14:19 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 18/28] drm/i915: use the new iterator in i915_sw_fence_await_reservation v3 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 19/28] drm/i915: use the new iterator in i915_request_await_object v2 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 20/28] drm/i915: use new iterator in i915_gem_object_wait_reservation Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-14 12:04 ` Maarten Lankhorst 2021-10-14 12:04 ` [Intel-gfx] " Maarten Lankhorst 2021-10-05 11:37 ` [PATCH 21/28] drm/i915: use new iterator in i915_gem_object_wait_priority Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 22/28] drm/i915: use new cursor in intel_prepare_plane_fb Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-05 11:37 ` [PATCH 23/28] drm: use new iterator in drm_gem_fence_array_add_implicit v3 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:21 ` Daniel Vetter 2021-10-13 14:21 ` [Intel-gfx] " Daniel Vetter 2021-10-19 12:54 ` Christian König 2021-10-19 12:54 ` [Intel-gfx] " Christian König 2021-10-19 13:59 ` Daniel Vetter 2021-10-19 13:59 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 24/28] drm: use new iterator in drm_gem_plane_helper_prepare_fb v2 Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:23 ` Daniel Vetter 2021-10-13 14:23 ` [Intel-gfx] " Daniel Vetter 2021-10-19 13:02 ` Christian König 2021-10-19 13:02 ` [Intel-gfx] " Christian König 2021-10-19 14:30 ` Daniel Vetter 2021-10-19 14:30 ` Daniel Vetter 2021-10-19 15:51 ` Christian König 2021-10-19 15:51 ` [Intel-gfx] " Christian König 2021-10-21 11:31 ` Daniel Vetter 2021-10-21 11:31 ` [Intel-gfx] " Daniel Vetter 2021-10-21 11:33 ` Daniel Vetter 2021-10-21 11:33 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 25/28] drm/nouveau: use the new iterator in nouveau_fence_sync Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:27 ` Daniel Vetter 2021-10-13 14:27 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 26/28] drm/nouveau: use the new interator in nv50_wndw_prepare_fb Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:29 ` Daniel Vetter 2021-10-13 14:29 ` [Intel-gfx] " Daniel Vetter 2021-10-22 13:17 ` Christian König 2021-10-22 13:17 ` [Intel-gfx] " Christian König 2021-10-28 15:26 ` Daniel Vetter 2021-10-28 15:26 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 27/28] drm/etnaviv: use new iterator in etnaviv_gem_describe Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:31 ` Daniel Vetter 2021-10-13 14:31 ` [Intel-gfx] " Daniel Vetter 2021-10-05 11:37 ` [PATCH 28/28] drm/etnaviv: replace dma_resv_get_excl_unlocked Christian König 2021-10-05 11:37 ` [Intel-gfx] " Christian König 2021-10-13 14:32 ` Daniel Vetter 2021-10-13 14:32 ` [Intel-gfx] " Daniel Vetter 2021-10-05 13:27 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/28] dma-buf: add dma_resv_for_each_fence_unlocked v8 Patchwork 2021-10-05 13:30 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork 2021-10-05 14:01 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=fc9ac0dd-a069-7f34-d94d-ae6cd70d307f@linux.intel.com \ --to=tvrtko.ursulin@linux.intel.com \ --cc=ckoenig.leichtzumerken@gmail.com \ --cc=daniel@ffwll.ch \ --cc=dri-devel@lists.freedesktop.org \ --cc=intel-gfx@lists.freedesktop.org \ --cc=linaro-mm-sig@lists.linaro.org \ --cc=linux-media@vger.kernel.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.