drm/i915/gt: Remove virtual breadcrumb before transfer

The issue with stale virtual breadcrumbs remain. Now we have the problem
that if the irq-signaler is still referencing the stale breadcrumb as we
transfer it to a new sibling, the list becomes spaghetti. This is a very
small window, but that doesn't stop it being hit infrequently. To
prevent the lists being tangled (the iterator starting on one engine's
b->signalers but walking onto another list), always decouple the virtual
breadcrumb on schedule-out and make sure that the walker has stepped out
of the lists.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201224135544.1713-5-chris@chris-wilson.co.uk

diff --git a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
index f96cd7d9b41929e87ad0d7c94f2d6bfbcdd12b02..8f7c8595ba08cbbf64293fb5bece2eab13b26ee7 100644 (file)
--- a/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
+++ b/drivers/gpu/drm/i915/gt/intel_breadcrumbs.c
@@ -451,15 +451,16 @@ void i915_request_cancel_breadcrumb(struct i915_request *rq)
 {
        struct intel_breadcrumbs *b = READ_ONCE(rq->engine)->breadcrumbs;
        struct intel_context *ce = rq->context;
+       unsigned long flags;
        bool release;
 
        if (!test_and_clear_bit(I915_FENCE_FLAG_SIGNAL, &rq->fence.flags))
                return;
 
-       spin_lock(&ce->signal_lock);
+       spin_lock_irqsave(&ce->signal_lock, flags);
        list_del_rcu(&rq->signal_link);
        release = remove_signaling_context(b, ce);
-       spin_unlock(&ce->signal_lock);
+       spin_unlock_irqrestore(&ce->signal_lock, flags);
        if (release)
                intel_context_put(ce);
 

diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
index b5f256be73dd1f02595e7fbd207d7806bda85204..1f6e9fe5bcc0fc777833b31ba05b9115021b04ce 100644 (file)
--- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
+++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
@@ -581,6 +581,21 @@ static inline void execlists_schedule_in(struct i915_request *rq, int idx)
 static void kick_siblings(struct i915_request *rq, struct intel_context *ce)
 {
        struct virtual_engine *ve = container_of(ce, typeof(*ve), context);
+       struct intel_engine_cs *engine = rq->engine;
+
+       /* Flush concurrent rcu iterators in signal_irq_work */
+       if (test_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &rq->fence.flags)) {
+               /*
+                * After this point, the rq may be transferred to a new
+                * sibling, so before we clear ce->inflight make sure that
+                * the context has been removed from the b->signalers and
+                * furthermore we need to make sure that the concurrent
+                * iterator in signal_irq_work is no longer following
+                * ce->signal_link.
+                */
+               i915_request_cancel_breadcrumb(rq);
+               irq_work_sync(&engine->breadcrumbs->irq_work);
+       }
 
        if (READ_ONCE(ve->request))
                tasklet_hi_schedule(&ve->base.execlists.tasklet);