[Intel-gfx] [PATCH] drm/i915: Flush the existing fence before GGTT read/write

[Chris Wilson]

Our fence management is lazy, very lazy. If the user marks an object as
untiled, we do not immediately flush the fence but merely mark it as
dirty. On the use we have to remember to check and remove the fence, by
which time we hope it is idle and we do not have to wait.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111468
Fixes: 1f7fd484fff1 ("drm/i915: Replace i915_vma_put_fence()")
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld at intel.com>
---
 drivers/gpu/drm/i915/i915_gem.c | 30 ++++++++++++++++++++++++++++--
 1 file changed, 28 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index eb31b69a316a..41b28f6d8620 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -324,6 +324,26 @@ gtt_user_read(struct io_mapping *mapping,
 	return unwritten;
 }
 
+static int linear_ggtt_offset(struct i915_vma *vma, u64 *offset)
+{
+	int ret;
+
+	GEM_BUG_ON(!i915_vma_is_map_and_fenceable(vma));
+
+	/* Open-coded i915_vma_pin_iomap() */
+
+	if (vma->fence) {
+		mutex_lock(&vma->vm->mutex);
+		ret = i915_vma_revoke_fence(vma);
+		mutex_unlock(&vma->vm->mutex);
+		if (ret)
+			return ret;
+	}
+
+	*offset = i915_ggtt_offset(vma);
+	return 0;
+}
+
 static int
 i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
 		   const struct drm_i915_gem_pread *args)
@@ -350,7 +370,10 @@ i915_gem_gtt_pread(struct drm_i915_gem_object *obj,
 					       PIN_NONBLOCK /* NOWARN */ |
 					       PIN_NOEVICT);
 	if (!IS_ERR(vma)) {
-		node.start = i915_ggtt_offset(vma);
+		ret = linear_ggtt_offset(vma, &node.start);
+		if (ret)
+			goto out_unpin;
+
 		node.allocated = false;
 	} else {
 		ret = insert_mappable_node(ggtt, &node, PAGE_SIZE);
@@ -560,7 +583,10 @@ i915_gem_gtt_pwrite_fast(struct drm_i915_gem_object *obj,
 					       PIN_NONBLOCK /* NOWARN */ |
 					       PIN_NOEVICT);
 	if (!IS_ERR(vma)) {
-		node.start = i915_ggtt_offset(vma);
+		ret = linear_ggtt_offset(vma, &node.start);
+		if (ret)
+			goto out_unpin;
+
 		node.allocated = false;
 	} else {
 		ret = insert_mappable_node(ggtt, &node, PAGE_SIZE);
-- 
2.23.0