Hi Gilbert,
> diff --git a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
> b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
> index e8f42ec6b1b4..cbc4ecf26d8b 100644
> --- a/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
> +++ b/drivers/gpu/drm/i915/gt/intel_execlists_submission.c
> @@ -1751,7 +1751,7 @@ static bool gen8_csb_parse(const u64 csb)
> static noinline u64
> wa_csb_read(const struct intel_engine_cs *engine, u64 * const csb)
> {
> - u64 entry;
> + u64 entry = READ_ONCE(*csb);
>
> /*
> * Reading from the HWSP has one particular advantage: we can detect
> @@ -1763,7 +1763,7 @@ wa_csb_read(const struct intel_engine_cs *engine, u64 *
> const csb)
> * tgl,dg1:HSDES#22011327657
> */
> preempt_disable();
> - if (wait_for_atomic_us((entry = READ_ONCE(*csb)) != -1, 10)) {
> + if (wait_for_atomic_us(entry != -1, 10)) {
this doesn't work. Hav you looked inside the wait_for_atomic_us?
It's busy looping for 10us waiting for *csb to be different from
-1. Because the entry value won't have a chance to change, then
we would be waiting 10us for nothing.
Not all the checkpatch errors need to be fixed and we can live
with this one.
Andi
> int idx = csb - engine->execlists.csb_status;
> int status;
>
> --
> 2.34.1
