[Intel-gfx] [PATCH] drm/i915: individualize fences before adding

Hellstrom, Thomas thomas.hellstrom at intel.com
Mon May 16 09:38:13 UTC 2022


Hi Nirmoy.


On Fri, 2022-05-13 at 13:37 +0200, Nirmoy Das wrote:
> _i915_vma_move_to_active() can receive > 1 fence for
> multiple batch buffer submission so make sure to
> individualize fences before adding to dma_resv obj
> 
> Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/5614
> Signed-off-by: Nirmoy Das <nirmoy.das at intel.com>

LGTM. 
Reviewed-by: Thomas Hellström <thomas.hellstrom at linux.intel.com>

> ---
>  drivers/gpu/drm/i915/i915_vma.c | 17 +++++++++++++----
>  1 file changed, 13 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_vma.c
> b/drivers/gpu/drm/i915/i915_vma.c
> index 4f6db539571a..f987fc1264c0 100644
> --- a/drivers/gpu/drm/i915/i915_vma.c
> +++ b/drivers/gpu/drm/i915/i915_vma.c
> @@ -23,6 +23,7 @@
>   */
>  
>  #include <linux/sched/mm.h>
> +#include <linux/dma-fence-array.h>
>  #include <drm/drm_gem.h>
>  
>  #include "display/intel_frontbuffer.h"
> @@ -1840,8 +1841,12 @@ int _i915_vma_move_to_active(struct i915_vma
> *vma,
>                 }
>  
>                 if (fence) {
> -                       dma_resv_add_fence(vma->obj->base.resv,
> fence,
> -                                          DMA_RESV_USAGE_WRITE);
> +                       int idx;
> +                       struct dma_fence *curr;
> +
> +                       dma_fence_array_for_each(curr, idx, fence)
> +                               dma_resv_add_fence(vma->obj-
> >base.resv, curr,
> +                                                 
> DMA_RESV_USAGE_WRITE);
>                         obj->write_domain = I915_GEM_DOMAIN_RENDER;
>                         obj->read_domains = 0;
>                 }
> @@ -1853,8 +1858,12 @@ int _i915_vma_move_to_active(struct i915_vma
> *vma,
>                 }
>  
>                 if (fence) {
> -                       dma_resv_add_fence(vma->obj->base.resv,
> fence,
> -                                          DMA_RESV_USAGE_READ);
> +                       int idx;
> +                       struct dma_fence *curr;
> +
> +                       dma_fence_array_for_each(curr, idx, fence)
> +                               dma_resv_add_fence(vma->obj-
> >base.resv, curr,
> +                                                 
> DMA_RESV_USAGE_READ);
>                         obj->write_domain = 0;
>                 }
>         }



More information about the Intel-gfx mailing list