drm/i915: Work around gen7 BLT ring synchronization issues.
authorEric Anholt <eric@anholt.net>
Thu, 22 Dec 2011 22:55:00 +0000 (14:55 -0800)
committerKeith Packard <keithp@keithp.com>
Tue, 3 Jan 2012 17:31:15 +0000 (09:31 -0800)
Previous to this commit, testing easily reproduced a failure where the
seqno would apparently arrive after the IRQ associated with it, with test programs as simple as:

for (;;) {
    glCopyPixels(0, 0, 1, 1);
    glFinish();
}

Various workarounds we've seen for previous generations didn't work to
fix this issue, so until new information comes in, replace the IRQ
waits on the BLT ring with polling.

Signed-off-by: Eric Anholt <eric@anholt.net>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
drivers/gpu/drm/i915/intel_ringbuffer.c

index f5dae5d..d0eb228 100644 (file)
@@ -791,6 +791,17 @@ ring_add_request(struct intel_ring_buffer *ring,
        return 0;
 }
 
+static bool
+gen7_blt_ring_get_irq(struct intel_ring_buffer *ring)
+{
+       /* The BLT ring on IVB appears to have broken synchronization
+        * between the seqno write and the interrupt, so that the
+        * interrupt appears first.  Returning false here makes
+        * i915_wait_request() do a polling loop, instead.
+        */
+       return false;
+}
+
 static bool
 gen6_ring_get_irq(struct intel_ring_buffer *ring, u32 gflag, u32 rflag)
 {
@@ -1557,5 +1568,8 @@ int intel_init_blt_ring_buffer(struct drm_device *dev)
 
        *ring = gen6_blt_ring;
 
+       if (IS_GEN7(dev))
+               ring->irq_get = gen7_blt_ring_get_irq;
+
        return intel_init_ring_buffer(dev, ring);
 }