An interesting snippet from Sandybridge's prm:
"Although a Ring Buffer can be enabled in the non-empty state, it must
not be disabled unless it is empty. Attempting to disable a Ring Buffer
in the non-empty state is UNDEFINED."
Let's avoid the undefined behaviour as we disable the rings prior to
reset and resume.
v2: Tell HEAD to catch up to TAIL (empty ring) first, then reset both to
0 (supposedly while stopped).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20171027094311.30380-1-chris@chris-wilson.co.uk
 
                }
        }
 
-       I915_WRITE_CTL(engine, 0);
+       I915_WRITE_HEAD(engine, I915_READ_TAIL(engine));
+
        I915_WRITE_HEAD(engine, 0);
        I915_WRITE_TAIL(engine, 0);
 
+       /* The ring must be empty before it is disabled */
+       I915_WRITE_CTL(engine, 0);
+
        return (I915_READ_HEAD(engine) & HEAD_ADDR) == 0;
 }
 
 
                DRM_DEBUG_DRIVER("%s: timed out on STOP_RING\n",
                                 engine->name);
 
-       I915_WRITE_FW(RING_CTL(base), 0);
+       I915_WRITE_FW(RING_HEAD(base), I915_READ_FW(RING_TAIL(base)));
+
        I915_WRITE_FW(RING_HEAD(base), 0);
        I915_WRITE_FW(RING_TAIL(base), 0);
 
+       /* The ring must be empty before it is disabled */
+       I915_WRITE_FW(RING_CTL(base), 0);
+
        /* Check acts as a post */
        if (I915_READ_FW(RING_HEAD(base)) != 0)
                DRM_DEBUG_DRIVER("%s: ring head not parked\n",