drm/i915: correctly order the ring init sequence
authorDaniel Vetter <daniel.vetter@ffwll.ch>
Tue, 7 Aug 2012 07:54:14 +0000 (09:54 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 26 Aug 2012 22:00:40 +0000 (15:00 -0700)
commit 0d8957c8a90bbb5d34fab9a304459448a5131e06 upstream.

We may only start to set up the new register values after having
confirmed that the ring is truely off. Otherwise the hw might lose the
newly written register values. This is caught later on in the init
sequence, when we check whether the register writes have stuck.

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50522
Tested-by: Yang Guang <guang.a.yang@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/gpu/drm/i915/intel_ringbuffer.c

index 302d3d5..12a9e5f 100644 (file)
@@ -258,8 +258,6 @@ static int init_ring_common(struct intel_ring_buffer *ring)
        I915_WRITE_HEAD(ring, 0);
        ring->write_tail(ring, 0);
 
-       /* Initialize the ring. */
-       I915_WRITE_START(ring, obj->gtt_offset);
        head = I915_READ_HEAD(ring) & HEAD_ADDR;
 
        /* G45 ring initialization fails to reset head to zero */
@@ -285,6 +283,11 @@ static int init_ring_common(struct intel_ring_buffer *ring)
                }
        }
 
+       /* Initialize the ring. This must happen _after_ we've cleared the ring
+        * registers with the above sequence (the readback of the HEAD registers
+        * also enforces ordering), otherwise the hw might lose the new ring
+        * register values. */
+       I915_WRITE_START(ring, obj->gtt_offset);
        I915_WRITE_CTL(ring,
                        ((ring->size - PAGE_SIZE) & RING_NR_PAGES)
                        | RING_VALID);