[PATCH 05/91] drm/i915: Only skip requests once a context is banned
Chris Wilson
chris at chris-wilson.co.uk
Thu Jan 5 13:32:21 UTC 2017
On Thu, Jan 05, 2017 at 01:25:23PM +0000, Tvrtko Ursulin wrote:
>
> On 05/01/2017 10:34, Chris Wilson wrote:
> >If we skip before banning, we have an inconsistent interface between
> >execbuf still queueing valid request but those requests already queued
> >being cancelled. If we only cancel the pending requests once we stop
> >accepting new requests, the interface is more consistent.
> >
> >Reported-by: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
> >Fixes: 821ed7df6e2a ("drm/i915: Update reset path to fix incomplete requests")
> >Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> >Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> >Cc: Mika Kuoppala <mika.kuoppala at intel.com>
> >---
> > drivers/gpu/drm/i915/i915_gem.c | 19 ++++++++++++-------
> > 1 file changed, 12 insertions(+), 7 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> >index 1fcce2db86f2..e7a7f404fd85 100644
> >--- a/drivers/gpu/drm/i915/i915_gem.c
> >+++ b/drivers/gpu/drm/i915/i915_gem.c
> >@@ -2736,7 +2736,7 @@ void i915_gem_reset_prepare(struct drm_i915_private *dev_priv)
> > static void i915_gem_reset_engine(struct intel_engine_cs *engine)
> > {
> > struct drm_i915_gem_request *request;
> >- struct i915_gem_context *incomplete_ctx;
> >+ struct i915_gem_context *hung_ctx;
> > struct intel_timeline *timeline;
> > unsigned long flags;
> > bool ring_hung;
> >@@ -2748,6 +2748,8 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
> > if (!request)
> > return;
> >
> >+ hung_ctx = request->ctx;
> >+
> > ring_hung = engine->hangcheck.stalled;
> > if (engine->hangcheck.seqno != intel_engine_get_seqno(engine)) {
> > DRM_DEBUG_DRIVER("%s pardoned, was guilty? %s\n",
> >@@ -2757,10 +2759,10 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
> > }
> >
> > if (ring_hung) {
> >- i915_gem_context_mark_guilty(request->ctx);
> >+ i915_gem_context_mark_guilty(hung_ctx);
> > dma_fence_set_error(&request->fence, -EIO);
> > } else {
> >- i915_gem_context_mark_innocent(request->ctx);
> >+ i915_gem_context_mark_innocent(hung_ctx);
> > }
> >
> > if (!ring_hung)
> >@@ -2772,6 +2774,10 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
> > /* Setup the CS to resume from the breadcrumb of the hung request */
> > engine->reset_hw(engine, request);
> >
> >+ /* If this context is now banned, skip all of its pending requests. */
> >+ if (!i915_gem_context_is_banned(hung_ctx))
> >+ return;
> >+
> > /* Users of the default context do not rely on logical state
> > * preserved between batches. They have to emit full state on
> > * every batch and so it is safe to execute queued requests following
> >@@ -2780,17 +2786,16 @@ static void i915_gem_reset_engine(struct intel_engine_cs *engine)
> > * Other contexts preserve state, now corrupt. We want to skip all
> > * queued requests that reference the corrupt context.
> > */
> >- incomplete_ctx = request->ctx;
> >- if (i915_gem_context_is_default(incomplete_ctx))
> >+ if (i915_gem_context_is_default(hung_ctx))
> > return;
> >
> >- timeline = i915_gem_context_lookup_timeline(incomplete_ctx, engine);
> >+ timeline = i915_gem_context_lookup_timeline(hung_ctx, engine);
> >
> > spin_lock_irqsave(&engine->timeline->lock, flags);
> > spin_lock(&timeline->lock);
> >
> > list_for_each_entry_continue(request, &engine->timeline->requests, link)
> >- if (request->ctx == incomplete_ctx)
> >+ if (request->ctx == hung_ctx)
> > reset_request(request);
> >
> > list_for_each_entry(request, &timeline->requests, link)
> >
>
> Thought I r-b'ed it already, anyway:
You had. I applied the r-b this morning but not until after trying to
get this series to trybot.
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
More information about the Intel-gfx-trybot
mailing list