Re: [PATCH 1/6] drm/xe/xe_guc_ct: WRITE_ONCE g2h_fence done in g2h_fence_cancel

["Dixit, Ashutosh" <ashutosh.dixit@intel.com>]

On Thu, 18 Dec 2025 07:35:29 -0800, Jonathan Cavitt wrote:
>
> Use WRITE_ONCE when operating on g2h_fence->done in g2h_fence_cancel to
> prevent the compiler from ignoring this operation.
>
> Fixes: 94de94d24ea8 ("drm/xe/guc: Cancel ongoing H2G requests when stopping CT")
> Suggested-by: Matthew Brost <matthew.brost@intel.com>
> Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_guc_ct.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_guc_ct.c b/drivers/gpu/drm/xe/xe_guc_ct.c
> index c3df9b3f1b4d..d177df166d2a 100644
> --- a/drivers/gpu/drm/xe/xe_guc_ct.c
> +++ b/drivers/gpu/drm/xe/xe_guc_ct.c
> @@ -206,7 +206,9 @@ static void g2h_fence_cancel(struct g2h_fence *g2h_fence)
>  {
>	g2h_fence->cancel = true;
>	g2h_fence->fail = true;
> -	g2h_fence->done = true;
> +
> +	/* WRITE_ONCE pairs with wait_event_timeout in guc_ct_send_recv */
> +	WRITE_ONCE(g2h_fence->done, true);
>  }

So the compiler just optimizes ->done? It doesn't optimize ->cancel and
->fail, e.g.?

Generally ok to do this, but sometimes they are also unnecessary. Have we
seen evidence of the compiler optimizing this out? Also in other places
where we are doing this in other patches?


>
>  static bool g2h_fence_needs_alloc(struct g2h_fence *g2h_fence)
> --
> 2.43.0
>