From: Matthew Brost <matthew.brost@intel.com>
To: <intel-gfx@lists.freedesktop.org>, <dri-devel@lists.freedesktop.org>
Cc: <daniele.ceraolospurio@intel.com>
Subject: [PATCH 18/27] drm/i915/guc: Release submit fence from an irq_work
Date: Wed, 25 Aug 2021 20:23:18 -0700 [thread overview]
Message-ID: <20210826032327.18078-19-matthew.brost@intel.com> (raw)
In-Reply-To: <20210826032327.18078-1-matthew.brost@intel.com>
A subsequent patch will flip the locking hierarchy from
ce->guc_state.lock -> sched_engine->lock to sched_engine->lock ->
ce->guc_state.lock. As such we need to release the submit fence for a
request from an IRQ to break a lock inversion - i.e. the fence must be
release went holding ce->guc_state.lock and the releasing of the can
acquire sched_engine->lock.
v2:
(Daniele)
- Delete request from list before calling irq_work_queue
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
.../gpu/drm/i915/gt/uc/intel_guc_submission.c | 22 ++++++++++++++++---
drivers/gpu/drm/i915/i915_request.h | 5 +++++
2 files changed, 24 insertions(+), 3 deletions(-)
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 baf789f37d42..c86aae0899e5 100644
--- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
@@ -2035,17 +2035,32 @@ static const struct intel_context_ops guc_context_ops = {
.create_virtual = guc_create_virtual,
};
+static void submit_work_cb(struct irq_work *wrk)
+{
+ struct i915_request *rq = container_of(wrk, typeof(*rq), submit_work);
+
+ might_lock(&rq->engine->sched_engine->lock);
+ i915_sw_fence_complete(&rq->submit);
+}
+
static void __guc_signal_context_fence(struct intel_context *ce)
{
- struct i915_request *rq;
+ struct i915_request *rq, *rn;
lockdep_assert_held(&ce->guc_state.lock);
if (!list_empty(&ce->guc_state.fences))
trace_intel_context_fence_release(ce);
- list_for_each_entry(rq, &ce->guc_state.fences, guc_fence_link)
- i915_sw_fence_complete(&rq->submit);
+ /*
+ * Use an IRQ to ensure locking order of sched_engine->lock ->
+ * ce->guc_state.lock is preserved.
+ */
+ list_for_each_entry_safe(rq, rn, &ce->guc_state.fences,
+ guc_fence_link) {
+ list_del(&rq->guc_fence_link);
+ irq_work_queue(&rq->submit_work);
+ }
INIT_LIST_HEAD(&ce->guc_state.fences);
}
@@ -2155,6 +2170,7 @@ static int guc_request_alloc(struct i915_request *rq)
spin_lock_irqsave(&ce->guc_state.lock, flags);
if (context_wait_for_deregister_to_register(ce) ||
context_pending_disable(ce)) {
+ init_irq_work(&rq->submit_work, submit_work_cb);
i915_sw_fence_await(&rq->submit);
list_add_tail(&rq->guc_fence_link, &ce->guc_state.fences);
diff --git a/drivers/gpu/drm/i915/i915_request.h b/drivers/gpu/drm/i915/i915_request.h
index 1bc1349ba3c2..d818cfbfc41d 100644
--- a/drivers/gpu/drm/i915/i915_request.h
+++ b/drivers/gpu/drm/i915/i915_request.h
@@ -218,6 +218,11 @@ struct i915_request {
};
struct llist_head execute_cb;
struct i915_sw_fence semaphore;
+ /**
+ * @submit_work: complete submit fence from an IRQ if needed for
+ * locking hierarchy reasons.
+ */
+ struct irq_work submit_work;
/*
* A list of everyone we wait upon, and everyone who waits upon us.
--
2.32.0
next prev parent reply other threads:[~2021-08-26 3:29 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-26 3:23 [PATCH 00/27] Clean up GuC CI failures, simplify locking, and kernel DOC Matthew Brost
2021-08-26 3:23 ` [PATCH 01/27] drm/i915/guc: Fix blocked context accounting Matthew Brost
2021-08-26 3:23 ` [PATCH 02/27] drm/i915/guc: Fix outstanding G2H accounting Matthew Brost
2021-08-26 23:09 ` Daniele Ceraolo Spurio
2021-08-27 1:36 ` Matthew Brost
2021-08-26 3:23 ` [PATCH 03/27] drm/i915/guc: Unwind context requests in reverse order Matthew Brost
2021-08-26 3:23 ` [PATCH 04/27] drm/i915/guc: Don't drop ce->guc_active.lock when unwinding context Matthew Brost
2021-08-26 3:23 ` [PATCH 05/27] drm/i915/guc: Process all G2H message at once in work queue Matthew Brost
2021-08-26 3:23 ` [PATCH 06/27] drm/i915/guc: Workaround reset G2H is received after schedule done G2H Matthew Brost
2021-08-26 23:11 ` Daniele Ceraolo Spurio
2021-08-26 3:23 ` [PATCH 07/27] Revert "drm/i915/gt: Propagate change in error status to children on unhold" Matthew Brost
2021-08-26 3:23 ` [PATCH 08/27] drm/i915/selftests: Add a cancel request selftest that triggers a reset Matthew Brost
2021-08-26 9:32 ` [Intel-gfx] " Tvrtko Ursulin
2021-08-26 14:00 ` Matthew Brost
2021-08-26 3:23 ` [PATCH 09/27] drm/i915/guc: Kick tasklet after queuing a request Matthew Brost
2021-08-26 3:23 ` [PATCH 10/27] drm/i915/guc: Don't enable scheduling on a banned context, guc_id invalid, not registered Matthew Brost
2021-08-26 3:23 ` [PATCH 11/27] drm/i915/guc: Copy whole golden context, set engine state size of subset Matthew Brost
2021-08-26 23:21 ` Daniele Ceraolo Spurio
2021-08-26 23:33 ` [Intel-gfx] " John Harrison
2021-08-26 3:23 ` [PATCH 12/27] drm/i915/selftests: Add initial GuC selftest for scrubbing lost G2H Matthew Brost
2021-08-26 3:23 ` [PATCH 13/27] drm/i915/guc: Take context ref when cancelling request Matthew Brost
2021-08-26 3:23 ` [PATCH 14/27] drm/i915/guc: Don't touch guc_state.sched_state without a lock Matthew Brost
2021-08-26 3:23 ` [PATCH 15/27] drm/i915/guc: Reset LRC descriptor if register returns -ENODEV Matthew Brost
2021-08-26 3:23 ` [PATCH 16/27] drm/i915: Allocate error capture in nowait context Matthew Brost
2021-08-26 16:18 ` [Intel-gfx] " Matthew Brost
2021-08-26 16:21 ` Daniel Vetter
2021-08-26 3:23 ` [PATCH 17/27] drm/i915/guc: Flush G2H work queue during reset Matthew Brost
2021-08-26 3:23 ` Matthew Brost [this message]
2021-08-26 3:23 ` [PATCH 19/27] drm/i915/guc: Move guc_blocked fence to struct guc_state Matthew Brost
2021-08-26 3:23 ` [PATCH 20/27] drm/i915/guc: Rework and simplify locking Matthew Brost
2021-08-26 3:23 ` [PATCH 21/27] drm/i915/guc: Proper xarray usage for contexts_lookup Matthew Brost
2021-08-26 3:23 ` [PATCH 22/27] drm/i915/guc: Drop pin count check trick between sched_disable and re-pin Matthew Brost
2021-08-26 3:23 ` [PATCH 23/27] drm/i915/guc: Move GuC priority fields in context under guc_active Matthew Brost
2021-08-26 23:26 ` Daniele Ceraolo Spurio
2021-08-26 3:23 ` [PATCH 24/27] drm/i915/guc: Move fields protected by guc->contexts_lock into sub structure Matthew Brost
2021-08-26 3:23 ` [PATCH 25/27] drm/i915/guc: Drop guc_active move everything into guc_state Matthew Brost
2021-08-26 3:23 ` [PATCH 26/27] drm/i915/guc: Add GuC kernel doc Matthew Brost
2021-08-31 19:04 ` [Intel-gfx] " John Harrison
2021-08-26 3:23 ` [PATCH 27/27] drm/i915/guc: Drop static inline functions intel_guc_submission.c Matthew Brost
2021-08-31 19:09 ` [Intel-gfx] " John Harrison
-- strict thread matches above, loose matches on Subject: below --
2021-08-19 6:16 [PATCH 00/27] Clean up GuC CI failures, simplify locking, and kernel DOC Matthew Brost
2021-08-19 6:16 ` [PATCH 18/27] drm/i915/guc: Release submit fence from an irq_work Matthew Brost
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=20210826032327.18078-19-matthew.brost@intel.com \
--to=matthew.brost@intel.com \
--cc=daniele.ceraolospurio@intel.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.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: 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).