drm/i915/gt: Stage the transfer of the virtual breadcrumb
authorChris Wilson <chris@chris-wilson.co.uk>
Wed, 25 Mar 2020 13:00:59 +0000 (13:00 +0000)
committerRodrigo Vivi <rodrigo.vivi@intel.com>
Thu, 26 Mar 2020 17:21:30 +0000 (10:21 -0700)
We move the virtual breadcrumb from one physical engine to the next, if
the next virtual request is scheduled on a new physical engine. Since
the virtual context can only be in one signal queue, we need it to track
the current physical engine for the new breadcrumbs. However, to move
the list we need both breadcrumb locks -- and since we cannot take both
at the same time (unless we are careful and always ensure consistent
ordering) stage the movement of the signaler via the current virtual
request.

Closes: https://gitlab.freedesktop.org/drm/intel/issues/1510
Fixes: 6d06779e8672 ("drm/i915: Load balancing across a virtual engine")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200325130059.30600-1-chris@chris-wilson.co.uk
(cherry picked from commit 6c81e21a4742385c00713137c6fdcade0412e93c)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
drivers/gpu/drm/i915/gt/intel_lrc.c

index 112531b29f596abf4fa7203efbf5e7881a07d257..683014e7bc51dfff4e68f807e2668e592059fc2d 100644 (file)
@@ -1663,7 +1663,7 @@ static bool virtual_matches(const struct virtual_engine *ve,
 }
 
 static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
-                                    struct intel_engine_cs *engine)
+                                    struct i915_request *rq)
 {
        struct intel_engine_cs *old = ve->siblings[0];
 
@@ -1671,9 +1671,19 @@ static void virtual_xfer_breadcrumbs(struct virtual_engine *ve,
 
        spin_lock(&old->breadcrumbs.irq_lock);
        if (!list_empty(&ve->context.signal_link)) {
-               list_move_tail(&ve->context.signal_link,
-                              &engine->breadcrumbs.signalers);
-               intel_engine_signal_breadcrumbs(engine);
+               list_del_init(&ve->context.signal_link);
+
+               /*
+                * We cannot acquire the new engine->breadcrumbs.irq_lock
+                * (as we are holding a breadcrumbs.irq_lock already),
+                * so attach this request to the signaler on submission.
+                * The queued irq_work will occur when we finally drop
+                * the engine->active.lock after dequeue.
+                */
+               set_bit(DMA_FENCE_FLAG_ENABLE_SIGNAL_BIT, &rq->fence.flags);
+
+               /* Also transfer the pending irq_work for the old breadcrumb. */
+               intel_engine_signal_breadcrumbs(rq->engine);
        }
        spin_unlock(&old->breadcrumbs.irq_lock);
 }
@@ -2045,7 +2055,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
                                                                        engine);
 
                                if (!list_empty(&ve->context.signals))
-                                       virtual_xfer_breadcrumbs(ve, engine);
+                                       virtual_xfer_breadcrumbs(ve, rq);
 
                                /*
                                 * Move the bound engine to the top of the list