[<prev] [next>] [day] [month] [year] [list]
Message-Id: <20160902190243.8414-1-eric@anholt.net>
Date: Fri, 2 Sep 2016 12:02:43 -0700
From: Eric Anholt <eric@...olt.net>
To: Dave Airlie <airlied@...ux.ie>, dri-devel@...ts.freedesktop.org
Cc: linux-kernel@...r.kernel.org, Eric Anholt <eric@...olt.net>
Subject: [PATCH] drm/vc4: Allow some more signals to be packed with uniform resets.
The intent was to make sure people don't sneak in a small immediate or
something to change the interpretation of the uniform update args, but
these signals are just fine.
Fixes a validation failure in the current X server on some Render
operation.
Signed-off-by: Eric Anholt <eric@...olt.net>
---
drivers/gpu/drm/vc4/vc4_validate_shaders.c | 10 ++++++++--
1 file changed, 8 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/vc4/vc4_validate_shaders.c b/drivers/gpu/drm/vc4/vc4_validate_shaders.c
index 46527e989ce3..2543cf5b8b51 100644
--- a/drivers/gpu/drm/vc4/vc4_validate_shaders.c
+++ b/drivers/gpu/drm/vc4/vc4_validate_shaders.c
@@ -309,8 +309,14 @@ validate_uniform_address_write(struct vc4_validated_shader_info *validated_shade
* of uniforms on each side. However, this scheme is easy to
* validate so it's all we allow for now.
*/
-
- if (QPU_GET_FIELD(inst, QPU_SIG) != QPU_SIG_NONE) {
+ switch (QPU_GET_FIELD(inst, QPU_SIG)) {
+ case QPU_SIG_NONE:
+ case QPU_SIG_SCOREBOARD_UNLOCK:
+ case QPU_SIG_COLOR_LOAD:
+ case QPU_SIG_LOAD_TMU0:
+ case QPU_SIG_LOAD_TMU1:
+ break;
+ default:
DRM_ERROR("uniforms address change must be "
"normal math\n");
return false;
--
2.9.3
Powered by blists - more mailing lists