zink: fix longstanding issue with active batch state recycling

the previous code could recycle a currently-submitting state by hitting
a race condition where zink_screen_check_last_finished(batch_id) returned
true because batch_id was 0

this can no longer recycle the current batch, but the race should still be
eliminated for consistency: check 'submitted' since this guarantees batch_id
is valid

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27729>
This commit is contained in:
Mike Blumenkrantz
2024-02-21 11:04:51 -05:00
committed by Marge Bot
parent 16103b61e7
commit 3283415bbd
+5 -2
View File
@@ -460,8 +460,11 @@ get_batch_state(struct zink_context *ctx, struct zink_batch *batch)
}
/* states are stored sequentially, so if the first one doesn't work, none of them will */
if (!bs && ctx->batch_states && ctx->batch_states->next) {
if (zink_screen_check_last_finished(screen, ctx->batch_states->fence.batch_id) ||
find_unused_state(ctx->batch_states)) {
/* only a submitted state can be reused */
if (p_atomic_read(&ctx->batch_states->fence.submitted) &&
/* a submitted state must have completed before it can be reused */
(zink_screen_check_last_finished(screen, ctx->batch_states->fence.batch_id) ||
p_atomic_read(&ctx->batch_states->fence.completed))) {
bs = ctx->batch_states;
pop_batch_state(ctx);
}