]> git.baikalelectronics.ru Git - kernel.git/commitdiff
drm/i915/gem: Reduce context termination list iteration guard to RCU
authorChris Wilson <chris@chris-wilson.co.uk>
Thu, 6 Aug 2020 10:59:54 +0000 (11:59 +0100)
committerJoonas Lahtinen <joonas.lahtinen@linux.intel.com>
Mon, 7 Sep 2020 11:28:14 +0000 (14:28 +0300)
As we now protect the timeline list using RCU, we can drop the
timeline->mutex for guarding the list iteration during context close, as
we are searching for an inflight request. Any new request will see the
context is banned and not be submitted. In doing so, pull the checks for
a concurrent submission of the request (notably the
i915_request_completed()) under the engine spinlock, to fully serialise
with __i915_request_submit()). That is in the case of preempt-to-busy
where the request may be completed during the __i915_request_submit(),
we need to be careful that we sample the request status after
serialising so that we don't miss the request the engine is actually
submitting.

Fixes: 4a3174152147 ("drm/i915/gem: Refine occupancy test in kill_context()")
References: d22d2d073ef8 ("drm/i915: Protect i915_request_await_start from early waits") # rcu protection of timeline->requests
References: https://gitlab.freedesktop.org/drm/intel/-/issues/1622
References: https://gitlab.freedesktop.org/drm/intel/-/issues/2158
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200806105954.7766-1-chris@chris-wilson.co.uk
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
drivers/gpu/drm/i915/gem/i915_gem_context.c

index d8cccbab7a51d7aa9dd6316a88659f17a05313a1..db893f6c516b4c14dcfd2d85d4a5945ef692c948 100644 (file)
@@ -439,29 +439,36 @@ static bool __cancel_engine(struct intel_engine_cs *engine)
        return __reset_engine(engine);
 }
 
-static struct intel_engine_cs *__active_engine(struct i915_request *rq)
+static bool
+__active_engine(struct i915_request *rq, struct intel_engine_cs **active)
 {
        struct intel_engine_cs *engine, *locked;
+       bool ret = false;
 
        /*
         * Serialise with __i915_request_submit() so that it sees
         * is-banned?, or we know the request is already inflight.
+        *
+        * Note that rq->engine is unstable, and so we double
+        * check that we have acquired the lock on the final engine.
         */
        locked = READ_ONCE(rq->engine);
        spin_lock_irq(&locked->active.lock);
        while (unlikely(locked != (engine = READ_ONCE(rq->engine)))) {
                spin_unlock(&locked->active.lock);
-               spin_lock(&engine->active.lock);
                locked = engine;
+               spin_lock(&locked->active.lock);
        }
 
-       engine = NULL;
-       if (i915_request_is_active(rq) && rq->fence.error != -EIO)
-               engine = rq->engine;
+       if (!i915_request_completed(rq)) {
+               if (i915_request_is_active(rq) && rq->fence.error != -EIO)
+                       *active = locked;
+               ret = true;
+       }
 
        spin_unlock_irq(&locked->active.lock);
 
-       return engine;
+       return ret;
 }
 
 static struct intel_engine_cs *active_engine(struct intel_context *ce)
@@ -472,17 +479,16 @@ static struct intel_engine_cs *active_engine(struct intel_context *ce)
        if (!ce->timeline)
                return NULL;
 
-       mutex_lock(&ce->timeline->mutex);
-       list_for_each_entry_reverse(rq, &ce->timeline->requests, link) {
-               if (i915_request_completed(rq))
-                       break;
+       rcu_read_lock();
+       list_for_each_entry_rcu(rq, &ce->timeline->requests, link) {
+               if (i915_request_is_active(rq) && i915_request_completed(rq))
+                       continue;
 
                /* Check with the backend if the request is inflight */
-               engine = __active_engine(rq);
-               if (engine)
+               if (__active_engine(rq, &engine))
                        break;
        }
-       mutex_unlock(&ce->timeline->mutex);
+       rcu_read_unlock();
 
        return engine;
 }