[Intel-gfx] [PATCH 2/3] drm/i915: recheck ringspace after wait+retire

[Dave Gordon]

In {intel,logical}_ring_wait_request(), we try to find a request
whose completion will release the amount of ring space required.
If we find such a request, we wait for it, and then retire it, in
the expectation that we will now have at least the required amount
of free space in the ring. But it's good to check that this has
actually happened, so we can back out with a meaningful error
code if something unexpected has happened, such as wait_request
returning early.

This code was already in the execlist version, so the change to
intel_lrc.c is just to add a comment; but we want the same check
in the legacy ringbuffer mode too.

Signed-off-by: Dave Gordon <david.s.gordon at intel.com>
---
 drivers/gpu/drm/i915/intel_lrc.c        |    9 +++++++++
 drivers/gpu/drm/i915/intel_ringbuffer.c |   12 +++++++++++-
 2 files changed, 20 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 8545dbd..69b042f 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -956,6 +956,15 @@ static int logical_ring_wait_request(struct intel_ringbuffer *ringbuf,
 
 	i915_gem_retire_requests_ring(ring);
 
+	/*
+	 * According to our calculation above, retiring the request we just
+	 * waited for should have resulted in there being enough space in
+	 * the ringbuffer; but let's check.
+	 *
+	 * If there is not now enough space, something has gone horribly worng
+	 * (such as wait_request returning early, but with no error, or
+	 * retire_requests failing to retire the request we expected it to).
+	 */
 	return intel_ring_space(ringbuf) >= bytes ? 0 : -ENOSPC;
 }
 
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 9def83c..660d10d 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1912,6 +1912,7 @@ static int intel_ring_wait_request(struct intel_engine_cs *ring, int n)
 		return 0;
 
 	list_for_each_entry(request, &ring->request_list, list) {
+		/* Would completion of this request free enough space? */
 		if (__intel_ring_space(request->tail, ringbuf->tail,
 				       ringbuf->effective_size) >= n) {
 			break;
@@ -1927,7 +1928,16 @@ static int intel_ring_wait_request(struct intel_engine_cs *ring, int n)
 
 	i915_gem_retire_requests_ring(ring);
 
-	return 0;
+	/*
+	 * According to our calculation above, retiring the request we just
+	 * waited for should have resulted in there being enough space in
+	 * the ringbuffer; but let's check.
+	 *
+	 * If there is not now enough space, something has gone horribly worng
+	 * (such as wait_request returning early, but with no error, or
+	 * retire_requests failing to retire the request we expected it to).
+	 */
+	return intel_ring_space(ringbuf) >= n ? 0 : -ENOSPC;
 }
 
 static int ring_wait_for_space(struct intel_engine_cs *ring, int n)
-- 
1.7.9.5