[Intel-gfx] [PATCH 11/15] drm/i915: First try the previous execbuffer location

[Chris Wilson]

When choosing a slot for an execbuffer, we ideally want to use the same
address as last time (so that we don't have to rebind it) and the same
address as expected by the user (so that we don't have to fixup any
relocations pointing to it). If we first try to bind the incoming
execbuffer->offset from the user, or the currently bound offset that
should hopefully achieve the goal of avoiding the rebind cost and the
relocation penalty. However, if the object is not currently bound there
we don't want to arbitrarily unbind an object in our chosen position and
so choose to rebind/relocate the incoming object instead. After we
report the new position back to the user, on the next pass the
relocations should have settled down.

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem_execbuffer.c | 12 ++++++++----
 drivers/gpu/drm/i915/i915_gem_gtt.c        |  6 ++++++
 drivers/gpu/drm/i915/i915_gem_gtt.h        |  1 +
 3 files changed, 15 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 6ff282c225d0..cdc57c1bd1db 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -289,10 +289,15 @@ eb_pin_vma(struct i915_execbuffer *eb,
 {
 	u64 flags;
 
-	flags = vma->node.start;
-	flags |= PIN_USER | PIN_NONBLOCK | PIN_OFFSET_FIXED;
+	if (vma->node.size)
+		flags = vma->node.start;
+	else
+		flags = entry->offset & PIN_OFFSET_MASK;
+
+	flags |= PIN_USER | PIN_NOEVICT | PIN_OFFSET_FIXED;
 	if (unlikely(entry->flags & EXEC_OBJECT_NEEDS_GTT))
 		flags |= PIN_GLOBAL;
+
 	if (unlikely(i915_vma_pin(vma, 0, 0, flags)))
 		return;
 
@@ -403,8 +408,7 @@ eb_add_vma(struct i915_execbuffer *eb,
 		entry->flags |= eb->context_flags;
 
 	ret = 0;
-	if (vma->node.size)
-		eb_pin_vma(eb, entry, vma);
+	eb_pin_vma(eb, entry, vma);
 	if (eb_vma_misplaced(entry, vma)) {
 		eb_unreserve_vma(vma, entry);
 
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 999f15455f48..d5cb1299f4f3 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3203,6 +3203,9 @@ int i915_gem_gtt_reserve(struct i915_address_space *vm,
 	if (err != -ENOSPC)
 		return err;
 
+	if (flags & PIN_NOEVICT)
+		return -ENOSPC;
+
 	err = i915_gem_evict_for_node(vm, node, flags);
 	if (err == 0)
 		err = drm_mm_reserve_node(&vm->mm, node);
@@ -3317,6 +3320,9 @@ int i915_gem_gtt_insert(struct i915_address_space *vm,
 	if (err != -ENOSPC)
 		return err;
 
+	if (flags & PIN_NOEVICT)
+		return -ENOSPC;
+
 	/* No free space, pick a slot at random.
 	 *
 	 * There is a pathological case here using a GTT shared between
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 069fc4e1be2a..6e68e7719ba4 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -582,6 +582,7 @@ int i915_gem_gtt_insert(struct i915_address_space *vm,
 #define PIN_MAPPABLE		BIT(1)
 #define PIN_ZONE_4G		BIT(2)
 #define PIN_NONFAULT		BIT(3)
+#define PIN_NOEVICT		BIT(4)
 
 #define PIN_MBZ			BIT(5) /* I915_VMA_PIN_OVERFLOW */
 #define PIN_GLOBAL		BIT(6) /* I915_VMA_GLOBAL_BIND */
-- 
2.11.0