Re: [PATCH] drm/xe: Use ttm_uncached for BO with NEEDS_UC flag

["Thomas Hellström" <thomas.hellstrom@linux.intel.com>]

Hi, Michal,

On Thu, 2024-06-06 at 21:56 +0200, Michal Wajdeczko wrote:
> We should honor requested uncached mode also at the TTM layer.
> Otherwise, we risk losing updates to the memory based interrupts
> source or status vectors, as those require uncached memory.
> 
> Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
> Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
> Cc: Matt Roper <matthew.d.roper@intel.com>
> ---
>  drivers/gpu/drm/xe/xe_bo.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
> index 2bae01ce4e5b..2573cc118f29 100644
> --- a/drivers/gpu/drm/xe/xe_bo.c
> +++ b/drivers/gpu/drm/xe/xe_bo.c
> @@ -378,6 +378,9 @@ static struct ttm_tt *xe_ttm_tt_create(struct
> ttm_buffer_object *ttm_bo,
>  	    (xe->info.graphics_verx100 >= 1270 && bo->flags &
> XE_BO_FLAG_PAGETABLE))
>  		caching = ttm_write_combined;
>  
> +	if (bo->flags & XE_BO_FLAG_NEEDS_UC)
> +		caching = ttm_uncached;
> +
>  	err = ttm_tt_init(&tt->ttm, &bo->ttm, page_flags, caching,
> extra_pages);
>  	if (err) {
>  		kfree(tt);

To me the preferred method is to teach bo->cpu_caching about the
uncached mode and then include it in the switch statement above.

/Thomas