From: John Harrison <john.c.harrison@intel.com>
To: Matthew Brost <matthew.brost@intel.com>,
<intel-gfx@lists.freedesktop.org>,
<dri-devel@lists.freedesktop.org>
Cc: <daniel.vetter@ffwll.ch>, <tony.ye@intel.com>, <zhengguo.xu@intel.com>
Subject: Re: [Intel-gfx] [PATCH 16/27] drm/i915/guc: Insert submit fences between requests in parent-child relationship
Date: Mon, 20 Sep 2021 14:57:22 -0700 [thread overview]
Message-ID: <09932f6a-17e6-f226-2cc1-fad60781ce43@intel.com> (raw)
In-Reply-To: <20210820224446.30620-17-matthew.brost@intel.com>
On 8/20/2021 15:44, Matthew Brost wrote:
> The GuC must receive requests in the order submitted for contexts in a
> parent-child relationship to function correctly. To ensure this, insert
> a submit fence between the current request and last request submitted
> for requests / contexts in a parent child relationship. This is
> conceptually similar to a single timeline.
>
> Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> Cc: John Harrison <John.C.Harrison@Intel.com>
> ---
> drivers/gpu/drm/i915/gt/intel_context.h | 5 +
> drivers/gpu/drm/i915/gt/intel_context_types.h | 7 +
> .../gpu/drm/i915/gt/uc/intel_guc_submission.c | 5 +-
> drivers/gpu/drm/i915/i915_request.c | 120 ++++++++++++++----
> 4 files changed, 109 insertions(+), 28 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_context.h b/drivers/gpu/drm/i915/gt/intel_context.h
> index c2985822ab74..9dcc1b14697b 100644
> --- a/drivers/gpu/drm/i915/gt/intel_context.h
> +++ b/drivers/gpu/drm/i915/gt/intel_context.h
> @@ -75,6 +75,11 @@ intel_context_to_parent(struct intel_context *ce)
> }
> }
>
> +static inline bool intel_context_is_parallel(struct intel_context *ce)
> +{
> + return intel_context_is_child(ce) || intel_context_is_parent(ce);
> +}
> +
> void intel_context_bind_parent_child(struct intel_context *parent,
> struct intel_context *child);
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_context_types.h b/drivers/gpu/drm/i915/gt/intel_context_types.h
> index 6f567ebeb039..a63329520c35 100644
> --- a/drivers/gpu/drm/i915/gt/intel_context_types.h
> +++ b/drivers/gpu/drm/i915/gt/intel_context_types.h
> @@ -246,6 +246,13 @@ struct intel_context {
> * work queue descriptor
> */
> u8 parent_page;
> +
> + /**
> + * @last_rq: last request submitted on a parallel context, used
> + * to insert submit fences between request in the parallel
request -> requests
With that fixed:
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
> + * context.
> + */
> + struct i915_request *last_rq;
> };
>
> #ifdef CONFIG_DRM_I915_SELFTEST
> diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
> index b107ad095248..f0b60fecf253 100644
> --- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
> +++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
> @@ -672,8 +672,7 @@ static int rq_prio(const struct i915_request *rq)
>
> static bool is_multi_lrc_rq(struct i915_request *rq)
> {
> - return intel_context_is_child(rq->context) ||
> - intel_context_is_parent(rq->context);
> + return intel_context_is_parallel(rq->context);
> }
>
> static bool can_merge_rq(struct i915_request *rq,
> @@ -2843,6 +2842,8 @@ static void guc_parent_context_unpin(struct intel_context *ce)
> GEM_BUG_ON(!intel_context_is_parent(ce));
> GEM_BUG_ON(!intel_engine_is_virtual(ce->engine));
>
> + if (ce->last_rq)
> + i915_request_put(ce->last_rq);
> unpin_guc_id(guc, ce);
> lrc_unpin(ce);
> }
> diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
> index ce446716d092..2e51c8999088 100644
> --- a/drivers/gpu/drm/i915/i915_request.c
> +++ b/drivers/gpu/drm/i915/i915_request.c
> @@ -1546,36 +1546,62 @@ i915_request_await_object(struct i915_request *to,
> return ret;
> }
>
> +static inline bool is_parallel_rq(struct i915_request *rq)
> +{
> + return intel_context_is_parallel(rq->context);
> +}
> +
> +static inline struct intel_context *request_to_parent(struct i915_request *rq)
> +{
> + return intel_context_to_parent(rq->context);
> +}
> +
> static struct i915_request *
> -__i915_request_add_to_timeline(struct i915_request *rq)
> +__i915_request_ensure_parallel_ordering(struct i915_request *rq,
> + struct intel_timeline *timeline)
> {
> - struct intel_timeline *timeline = i915_request_timeline(rq);
> struct i915_request *prev;
>
> - /*
> - * Dependency tracking and request ordering along the timeline
> - * is special cased so that we can eliminate redundant ordering
> - * operations while building the request (we know that the timeline
> - * itself is ordered, and here we guarantee it).
> - *
> - * As we know we will need to emit tracking along the timeline,
> - * we embed the hooks into our request struct -- at the cost of
> - * having to have specialised no-allocation interfaces (which will
> - * be beneficial elsewhere).
> - *
> - * A second benefit to open-coding i915_request_await_request is
> - * that we can apply a slight variant of the rules specialised
> - * for timelines that jump between engines (such as virtual engines).
> - * If we consider the case of virtual engine, we must emit a dma-fence
> - * to prevent scheduling of the second request until the first is
> - * complete (to maximise our greedy late load balancing) and this
> - * precludes optimising to use semaphores serialisation of a single
> - * timeline across engines.
> - */
> + GEM_BUG_ON(!is_parallel_rq(rq));
> +
> + prev = request_to_parent(rq)->last_rq;
> + if (prev) {
> + if (!__i915_request_is_complete(prev)) {
> + i915_sw_fence_await_sw_fence(&rq->submit,
> + &prev->submit,
> + &rq->submitq);
> +
> + if (rq->engine->sched_engine->schedule)
> + __i915_sched_node_add_dependency(&rq->sched,
> + &prev->sched,
> + &rq->dep,
> + 0);
> + }
> + i915_request_put(prev);
> + }
> +
> + request_to_parent(rq)->last_rq = i915_request_get(rq);
> +
> + return to_request(__i915_active_fence_set(&timeline->last_request,
> + &rq->fence));
> +}
> +
> +static struct i915_request *
> +__i915_request_ensure_ordering(struct i915_request *rq,
> + struct intel_timeline *timeline)
> +{
> + struct i915_request *prev;
> +
> + GEM_BUG_ON(is_parallel_rq(rq));
> +
> prev = to_request(__i915_active_fence_set(&timeline->last_request,
> &rq->fence));
> +
> if (prev && !__i915_request_is_complete(prev)) {
> bool uses_guc = intel_engine_uses_guc(rq->engine);
> + bool pow2 = is_power_of_2(READ_ONCE(prev->engine)->mask |
> + rq->engine->mask);
> + bool same_context = prev->context == rq->context;
>
> /*
> * The requests are supposed to be kept in order. However,
> @@ -1583,13 +1609,11 @@ __i915_request_add_to_timeline(struct i915_request *rq)
> * is used as a barrier for external modification to this
> * context.
> */
> - GEM_BUG_ON(prev->context == rq->context &&
> + GEM_BUG_ON(same_context &&
> i915_seqno_passed(prev->fence.seqno,
> rq->fence.seqno));
>
> - if ((!uses_guc &&
> - is_power_of_2(READ_ONCE(prev->engine)->mask | rq->engine->mask)) ||
> - (uses_guc && prev->context == rq->context))
> + if ((same_context && uses_guc) || (!uses_guc && pow2))
> i915_sw_fence_await_sw_fence(&rq->submit,
> &prev->submit,
> &rq->submitq);
> @@ -1604,6 +1628,50 @@ __i915_request_add_to_timeline(struct i915_request *rq)
> 0);
> }
>
> + return prev;
> +}
> +
> +static struct i915_request *
> +__i915_request_add_to_timeline(struct i915_request *rq)
> +{
> + struct intel_timeline *timeline = i915_request_timeline(rq);
> + struct i915_request *prev;
> +
> + /*
> + * Dependency tracking and request ordering along the timeline
> + * is special cased so that we can eliminate redundant ordering
> + * operations while building the request (we know that the timeline
> + * itself is ordered, and here we guarantee it).
> + *
> + * As we know we will need to emit tracking along the timeline,
> + * we embed the hooks into our request struct -- at the cost of
> + * having to have specialised no-allocation interfaces (which will
> + * be beneficial elsewhere).
> + *
> + * A second benefit to open-coding i915_request_await_request is
> + * that we can apply a slight variant of the rules specialised
> + * for timelines that jump between engines (such as virtual engines).
> + * If we consider the case of virtual engine, we must emit a dma-fence
> + * to prevent scheduling of the second request until the first is
> + * complete (to maximise our greedy late load balancing) and this
> + * precludes optimising to use semaphores serialisation of a single
> + * timeline across engines.
> + *
> + * We do not order parallel submission requests on the timeline as each
> + * parallel submission context has its own timeline and the ordering
> + * rules for parallel requests are that they must be submitted in the
> + * order received from the execbuf IOCTL. So rather than using the
> + * timeline we store a pointer to last request submitted in the
> + * relationship in the gem context and insert a submission fence
> + * between that request and request passed into this function or
> + * alternatively we use completion fence if gem context has a single
> + * timeline and this is the first submission of an execbuf IOCTL.
> + */
> + if (likely(!is_parallel_rq(rq)))
> + prev = __i915_request_ensure_ordering(rq, timeline);
> + else
> + prev = __i915_request_ensure_parallel_ordering(rq, timeline);
> +
> /*
> * Make sure that no request gazumped us - if it was allocated after
> * our i915_request_alloc() and called __i915_request_add() before
next prev parent reply other threads:[~2021-09-20 21:57 UTC|newest]
Thread overview: 109+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-20 22:44 [Intel-gfx] [PATCH 00/27] Parallel submission aka multi-bb execbuf Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 01/27] drm/i915/guc: Squash Clean up GuC CI failures, simplify locking, and kernel DOC Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 02/27] drm/i915/guc: Allow flexible number of context ids Matthew Brost
2021-09-09 22:13 ` John Harrison
2021-09-10 0:14 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 03/27] drm/i915/guc: Connect the number of guc_ids to debugfs Matthew Brost
2021-09-09 22:16 ` John Harrison
2021-09-10 0:16 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 04/27] drm/i915/guc: Take GT PM ref when deregistering context Matthew Brost
2021-09-09 22:28 ` John Harrison
2021-09-10 0:21 ` Matthew Brost
2021-09-13 9:55 ` Tvrtko Ursulin
2021-09-13 17:12 ` Matthew Brost
2021-09-14 8:41 ` Tvrtko Ursulin
2021-08-20 22:44 ` [Intel-gfx] [PATCH 05/27] drm/i915: Add GT PM unpark worker Matthew Brost
2021-09-09 22:36 ` John Harrison
2021-09-10 0:34 ` Matthew Brost
2021-09-10 8:36 ` Tvrtko Ursulin
2021-09-10 20:09 ` Matthew Brost
2021-09-13 10:33 ` Tvrtko Ursulin
2021-09-13 17:20 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 06/27] drm/i915/guc: Take engine PM when a context is pinned with GuC submission Matthew Brost
2021-09-09 22:46 ` John Harrison
2021-09-10 0:41 ` Matthew Brost
2021-09-13 22:26 ` John Harrison
2021-09-14 1:12 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 07/27] drm/i915/guc: Don't call switch_to_kernel_context " Matthew Brost
2021-09-09 22:51 ` John Harrison
2021-09-13 16:54 ` Matthew Brost
2021-09-13 22:38 ` John Harrison
2021-09-14 5:02 ` Matthew Brost
2021-09-13 16:55 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 08/27] drm/i915: Add logical engine mapping Matthew Brost
2021-09-10 11:12 ` Tvrtko Ursulin
2021-09-10 19:49 ` Matthew Brost
2021-09-13 9:24 ` Tvrtko Ursulin
2021-09-13 16:50 ` Matthew Brost
2021-09-14 8:34 ` Tvrtko Ursulin
2021-09-14 18:04 ` Matthew Brost
2021-09-15 8:24 ` Tvrtko Ursulin
2021-09-15 16:58 ` Matthew Brost
2021-09-16 8:31 ` Tvrtko Ursulin
2021-08-20 22:44 ` [Intel-gfx] [PATCH 09/27] drm/i915: Expose logical engine instance to user Matthew Brost
2021-09-13 23:06 ` John Harrison
2021-09-14 1:08 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 10/27] drm/i915/guc: Introduce context parent-child relationship Matthew Brost
2021-09-13 23:19 ` John Harrison
2021-09-14 1:18 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 11/27] drm/i915/guc: Implement parallel context pin / unpin functions Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 12/27] drm/i915/guc: Add multi-lrc context registration Matthew Brost
2021-09-15 19:21 ` John Harrison
2021-09-15 19:31 ` Matthew Brost
2021-09-15 20:23 ` John Harrison
2021-09-15 20:33 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 13/27] drm/i915/guc: Ensure GuC schedule operations do not operate on child contexts Matthew Brost
2021-09-15 19:24 ` John Harrison
2021-09-15 19:34 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 14/27] drm/i915/guc: Assign contexts in parent-child relationship consecutive guc_ids Matthew Brost
2021-09-15 20:04 ` John Harrison
2021-09-15 20:55 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 15/27] drm/i915/guc: Implement multi-lrc submission Matthew Brost
2021-08-21 14:04 ` kernel test robot
2021-08-22 2:18 ` kernel test robot
2021-09-20 21:48 ` John Harrison
2021-09-22 16:25 ` Matthew Brost
2021-09-22 20:15 ` John Harrison
2021-09-23 2:44 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 16/27] drm/i915/guc: Insert submit fences between requests in parent-child relationship Matthew Brost
2021-09-20 21:57 ` John Harrison [this message]
2021-08-20 22:44 ` [Intel-gfx] [PATCH 17/27] drm/i915/guc: Implement multi-lrc reset Matthew Brost
2021-09-20 22:44 ` John Harrison
2021-09-22 16:16 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 18/27] drm/i915/guc: Update debugfs for GuC multi-lrc Matthew Brost
2021-09-20 22:48 ` John Harrison
2021-09-21 19:13 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 19/27] drm/i915: Fix bug in user proto-context creation that leaked contexts Matthew Brost
2021-09-20 22:57 ` John Harrison
2021-09-21 14:49 ` Tvrtko Ursulin
2021-09-21 19:28 ` Matthew Brost
2021-09-21 19:28 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 20/27] drm/i915/guc: Connect UAPI to GuC multi-lrc interface Matthew Brost
2021-08-29 4:00 ` kernel test robot
2021-08-29 19:59 ` kernel test robot
2021-09-21 0:09 ` John Harrison
2021-09-22 16:38 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 21/27] drm/i915/doc: Update parallel submit doc to point to i915_drm.h Matthew Brost
2021-09-21 0:12 ` John Harrison
2021-08-20 22:44 ` [Intel-gfx] [PATCH 22/27] drm/i915/guc: Add basic GuC multi-lrc selftest Matthew Brost
2021-09-28 20:47 ` John Harrison
2021-08-20 22:44 ` [Intel-gfx] [PATCH 23/27] drm/i915/guc: Implement no mid batch preemption for multi-lrc Matthew Brost
2021-09-10 11:25 ` Tvrtko Ursulin
2021-09-10 20:49 ` Matthew Brost
2021-09-13 10:52 ` Tvrtko Ursulin
2021-09-28 22:20 ` John Harrison
2021-09-28 22:33 ` Matthew Brost
2021-09-28 23:33 ` John Harrison
2021-09-29 0:22 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 24/27] drm/i915: Multi-BB execbuf Matthew Brost
2021-08-21 19:01 ` kernel test robot
2021-08-30 3:46 ` kernel test robot
2021-09-30 22:16 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 25/27] drm/i915/guc: Handle errors in multi-lrc requests Matthew Brost
2021-09-29 20:44 ` John Harrison
2021-09-29 20:58 ` Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 26/27] drm/i915: Enable multi-bb execbuf Matthew Brost
2021-08-20 22:44 ` [Intel-gfx] [PATCH 27/27] drm/i915/execlists: Weak parallel submission support for execlists Matthew Brost
2021-08-20 23:13 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for Parallel submission aka multi-bb execbuf (rev3) Patchwork
2021-08-20 23:14 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-08-20 23:45 ` [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=09932f6a-17e6-f226-2cc1-fad60781ce43@intel.com \
--to=john.c.harrison@intel.com \
--cc=daniel.vetter@ffwll.ch \
--cc=dri-devel@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.org \
--cc=matthew.brost@intel.com \
--cc=tony.ye@intel.com \
--cc=zhengguo.xu@intel.com \
/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: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).