fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-07 02:48:06 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

vc4: flush write jobs before BO replacement in DISCARD_WHOLE path

Browse source 
The DISCARD_WHOLE_RESOURCE path in vc4_map_usage_prep() replaces the
resource's BO with vc4_resource_bo_alloc(). As the RCL resolves
rsc->bo at job submit in vc4_submit_setup_rcl_surface(), any pending
write job would store to the new BO instead of the old one, corrupting
the new written data.

This is the same bug that was fixed in v3d in the previous commit.

Fixes: 18ccda7b86 ("vc4: When asked to discard-map a whole resource, discard it.")
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
(cherry picked from commit ecb6c5d555)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40359>
This commit is contained in:
Jose Maria Casanova Crespo 2026-03-02 16:45:50 +01:00 • committed by Eric Engestrom
parent fb8f81a1d8
commit 7d25d214f5
3 changed files with 8 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
.pick_status.json
View file

@ -3914,7 +3914,7 @@
"description": "vc4: flush write jobs before BO replacement in DISCARD_WHOLE path",
"nominated": true,
"nomination_type": 2,
"resolution": 0,
"resolution": 1,
"main_sha": null,
"because_sha": "18ccda7b86b8f7ab7466265aefb3f3e773f4a757",
"notes": null

2
src/broadcom/ci/broadcom-rpi3-fails.txt
View file

@ -853,7 +853,6 @@ spec@!opengl 1.1@polygon-mode-offset@config 6: Expected blue pixel in center,Fai
spec@!opengl 1.1@polygon-mode-offset@config 6: Expected white pixel on right edge,Fail
spec@!opengl 1.1@polygon-mode-offset@config 6: Expected white pixel on top edge,Fail
spec@!opengl 1.1@texsubimage-unpack,Fail
spec@!opengl 1.1@texwrap 2d proj,Fail
spec@!opengl 1.1@texwrap 2d proj@GL_RGBA8- NPOT- projected,Fail
spec@!opengl 1.1@texwrap 2d proj@GL_RGBA8- projected,Fail
@ -953,7 +952,6 @@ spec@arb_occlusion_query@occlusion_query_conform,Fail
spec@arb_occlusion_query@occlusion_query_conform@GetObjivAval_multi2,Fail
spec@arb_pixel_buffer_object@fbo-pbo-readpixels-small,Fail
spec@arb_pixel_buffer_object@pbo-getteximage,Fail
spec@arb_pixel_buffer_object@texsubimage-unpack pbo,Fail
spec@arb_point_sprite@arb_point_sprite-mipmap,Fail
spec@arb_provoking_vertex@arb-provoking-vertex-render,Fail
spec@arb_sampler_objects@sampler-objects,Fail

7
src/gallium/drivers/vc4/vc4_resource.c
View file

@ -108,6 +108,13 @@ vc4_map_usage_prep(struct pipe_context *pctx,
MESA_TRACE_FUNC();
if (usage & PIPE_MAP_DISCARD_WHOLE_RESOURCE) {
/* Flush any pending write jobs before replacing the BO.
* The RCL reads rsc->bo at job submit time, so if we
* replace the BO first, a later flush of the pending
* job would resolve the new BO instead of the old one,
* corrupting the new data with a stale store.
*/
vc4_flush_jobs_writing_resource(vc4, prsc);
if (vc4_resource_bo_alloc(rsc)) {
/* If it might be bound as one of our vertex buffers,
* make sure we re-emit vertex buffer state.