One of the few guarantees we want the busy ioctl to provide is that the
reported busy writer is included in the set of busy read engines. This
should be provided by the ordering of setting and retiring the active
trackers, but we can do better by explicitly setting the busy read
engine flag for the last writer.
Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
---
drivers/gpu/drm/i915/i915_gem.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index d71fa9a93afa..0478e4610da4 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3748,7 +3748,7 @@ static __always_inline unsigned int __busy_read_flag(unsigned int id)
static __always_inline unsigned int __busy_write_id(unsigned int id)
{
- return id;
+ return id | __busy_read_flag(id);
}
static __always_inline unsigned int
@@ -3857,9 +3857,11 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
args->busy |= busy_check_reader(&obj->last_read[idx]);
/* For ABI sanity, we only care that the write engine is in
- * the set of read engines. This is ensured by the ordering
- * of setting last_read/last_write in i915_vma_move_to_active,
- * and then in reverse in retire.
+ * the set of read engines. This should be ensured by the
+ * ordering of setting last_read/last_write in
+ * i915_vma_move_to_active(), and then in reverse in retire.
+ * However, for good measure, we always report the last_write
+ * request as a busy read as well as being a busy write.
*
* We don't care that the set of active read/write engines
* may change during construction of the result, as it is
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
