If a semaphore is waiting on another ring, which in turn happens to be waiting on the first ring, but that second semaphore has been signalled, we will be able to kick the second ring and so can treat the first ring as a valid WAIT and not as HUNG. v2: Be paranoid and cap the potential recursion depth whilst visiting the semaphore signallers. (Mika) References: https://bugs.freedesktop.org/show_bug.cgi?id=54226 References: https://bugs.freedesktop.org/show_bug.cgi?id=75502 Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> Cc: Mika Kuoppala <mika.kuoppala@xxxxxxxxxxxxxxx> Cc: stable@xxxxxxxxxxxxxxx --- drivers/gpu/drm/i915/i915_irq.c | 18 ++++++++++++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 2 +- 2 files changed, 15 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index a702303eab08..2974698f8f27 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -2831,10 +2831,14 @@ static int semaphore_passed(struct intel_engine_cs *ring) struct intel_engine_cs *signaller; u32 seqno, ctl; - ring->hangcheck.deadlock = true; + ring->hangcheck.deadlock++; signaller = semaphore_waits_for(ring, &seqno); - if (signaller == NULL || signaller->hangcheck.deadlock) + if (signaller == NULL) + return -1; + + /* Prevent pathological recursion due to driver bugs */ + if (signaller->hangcheck.deadlock >= I915_NUM_RINGS) return -1; /* cursory check for an unkickable deadlock */ @@ -2842,7 +2846,13 @@ static int semaphore_passed(struct intel_engine_cs *ring) if (ctl & RING_WAIT_SEMAPHORE && semaphore_passed(signaller) < 0) return -1; - return i915_seqno_passed(signaller->get_seqno(signaller, false), seqno); + if (i915_seqno_passed(signaller->get_seqno(signaller, false), seqno)) + return 1; + + if (signaller->hangcheck.deadlock) + return -1; + + return 0; } static void semaphore_clear_deadlocks(struct drm_i915_private *dev_priv) @@ -2851,7 +2861,7 @@ static void semaphore_clear_deadlocks(struct drm_i915_private *dev_priv) int i; for_each_ring(ring, dev_priv, i) - ring->hangcheck.deadlock = false; + ring->hangcheck.deadlock = 0; } static enum intel_ring_hangcheck_action diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 24357c597b54..31c2aab8ad2c 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -55,7 +55,7 @@ struct intel_ring_hangcheck { u32 seqno; int score; enum intel_ring_hangcheck_action action; - bool deadlock; + int deadlock; }; struct intel_ringbuffer { -- 2.0.0 -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html