drm/i915: Embrace the race in busy-ioctl

Daniel Vetter proposed a new challenge to the serialisation inside the
busy-ioctl that exposed a flaw that could result in us reporting the
wrong engine as being busy. If the request is reallocated as we test
its busyness and then reassigned to this object by another thread, we
would not notice that the test itself was incorrect.

We are faced with a choice of using __i915_gem_active_get_request_rcu()
to first acquire a reference to the request preventing the race, or to
acknowledge the race and accept the limitations upon the accuracy of the
busy flags. Note that we guarantee that we never falsely report the
object as idle (providing userspace itself doesn't race), and so the
most important use of the busy-ioctl and its guarantees are fulfilled.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1471337440-16777-1-git-send-email-chris@chris-wilson.co.uk

1 files changed, 15 insertions, 1 deletions

diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h
index 452629de7a57..5501fe83ed92 100644
--- a/include/uapi/drm/i915_drm.h
+++ b/include/uapi/drm/i915_drm.h
@@ -855,7 +855,16 @@ struct drm_i915_gem_busy {
 	 * having flushed any pending activity), and a non-zero return that
 	 * the object is still in-flight on the GPU. (The GPU has not yet
 	 * signaled completion for all pending requests that reference the
-	 * object.)
+	 * object.) An object is guaranteed to become idle eventually (so
+	 * long as no new GPU commands are executed upon it). Due to the
+	 * asynchronous nature of the hardware, an object reported
+	 * as busy may become idle before the ioctl is completed.
+	 *
+	 * Furthermore, if the object is busy, which engine is busy is only
+	 * provided as a guide. There are race conditions which prevent the
+	 * report of which engines are busy from being always accurate.
+	 * However, the converse is not true. If the object is idle, the
+	 * result of the ioctl, that all engines are idle, is accurate.
 	 *
 	 * The returned dword is split into two fields to indicate both
 	 * the engines on which the object is being read, and the
@@ -878,6 +887,11 @@ struct drm_i915_gem_busy {
 	 * execution engines, e.g. multiple media engines, which are
 	 * mapped to the same identifier in the EXECBUFFER2 ioctl and
 	 * so are not separately reported for busyness.
+	 *
+	 * Caveat emptor:
+	 * Only the boolean result of this query is reliable; that is whether
+	 * the object is idle or busy. The report of which engines are busy
+	 * should be only used as a heuristic.
 	 */
 	__u32 busy;
 };