[Intel-gfx] [PATCH 3/3] drm/i915; Preallocate the lazy request

[Chris Wilson]

It is possible for us to be forced to perform an allocation for the lazy
request whilst running the shrinker. This allocation may fail, leaving
us unable to reclaim any memory leading to premature OOM. A neat
solution to the problem is to preallocate the request at the same time
as acquiring the seqno for the ring transaction. This means that we can
report ENOMEM prior to touching the rings.

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem.c         |  9 ++++-----
 drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++++++++
 drivers/gpu/drm/i915/intel_ringbuffer.h |  1 +
 3 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 588fae9..e586c78 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2060,8 +2060,8 @@ int __i915_add_request(struct intel_ring_buffer *ring,
 	if (ret)
 		return ret;
 
-	request = kmalloc(sizeof(*request), GFP_KERNEL);
-	if (request == NULL)
+	request = ring->preallocated_lazy_request;
+	if (WARN_ON(request == NULL))
 		return -ENOMEM;
 
 	/* Record the position of the start of the request so that
@@ -2072,10 +2072,8 @@ int __i915_add_request(struct intel_ring_buffer *ring,
 	request_ring_position = intel_ring_get_tail(ring);
 
 	ret = ring->add_request(ring);
-	if (ret) {
-		kfree(request);
+	if (ret)
 		return ret;
-	}
 
 	request->seqno = intel_ring_get_seqno(ring);
 	request->ring = ring;
@@ -2114,6 +2112,7 @@ int __i915_add_request(struct intel_ring_buffer *ring,
 
 	trace_i915_gem_request_add(ring, request->seqno);
 	ring->outstanding_lazy_seqno = 0;
+	ring->preallocated_lazy_request = NULL;
 
 	if (!dev_priv->ums.mm_suspended) {
 		i915_queue_hangcheck(ring->dev);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index a83ff18..284afaf 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -1498,6 +1498,16 @@ intel_ring_alloc_seqno(struct intel_ring_buffer *ring)
 	if (ring->outstanding_lazy_seqno)
 		return 0;
 
+	if (ring->preallocated_lazy_request == NULL) {
+		struct drm_i915_gem_request *request;
+
+		request = kmalloc(sizeof(*request), GFP_KERNEL);
+		if (request == NULL)
+			return -ENOMEM;
+
+		ring->preallocated_lazy_request = request;
+	}
+
 	return i915_gem_get_seqno(ring->dev, &ring->outstanding_lazy_seqno);
 }
 
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index c6aa2b3..ad2dd65 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -140,6 +140,7 @@ struct  intel_ring_buffer {
 	/**
 	 * Do we have some not yet emitted requests outstanding?
 	 */
+	struct drm_i915_gem_request *preallocated_lazy_request;
 	u32 outstanding_lazy_seqno;
 	bool gpu_caches_dirty;
 	bool fbc_dirty;
-- 
1.8.4.rc3