drm/i915: Reduce nested prepare_remote_context() to a trylock

On context retiring, we may invoke the kernel_context to unpin this
context. Elsewhere, we may use the kernel_context to modify this
context. This currently leads to an AB-BA lock inversion, so we need to
back-off from the contended lock, and repeat.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111732
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: a9877da2d6 ("drm/i915/oa: Reconfigure contexts on the fly")
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20191126065521.2331017-1-chris@chris-wilson.co.uk

drivers/gpu/drm/i915/gt/intel_context.c

@@ -310,10 +310,23 @@ int intel_context_prepare_remote_request(struct intel_context *ce,
 	GEM_BUG_ON(rq->hw_context == ce);
 
 	if (rcu_access_pointer(rq->timeline) != tl) { /* timeline sharing! */
-		err = mutex_lock_interruptible_nested(&tl->mutex,
+		/*
-						      SINGLE_DEPTH_NESTING);
+		 * Ideally, we just want to insert our foreign fence as
-		if (err)
+		 * a barrier into the remove context, such that this operation
-			return err;
+		 * occurs after all current operations in that context, and
+		 * all future operations must occur after this.
+		 *
+		 * Currently, the timeline->last_request tracking is guarded
+		 * by its mutex and so we must obtain that to atomically
+		 * insert our barrier. However, since we already hold our
+		 * timeline->mutex, we must be careful against potential
+		 * inversion if we are the kernel_context as the remote context
+		 * will itself poke at the kernel_context when it needs to
+		 * unpin. Ergo, if already locked, we drop both locks and
+		 * try again (through the magic of userspace repeating EAGAIN).
+		 */
+		if (!mutex_trylock(&tl->mutex))
+			return -EAGAIN;
 
 		/* Queue this switch after current activity by this context. */
 		err = i915_active_fence_set(&tl->last_request, rq);