[Intel-gfx] [PATCH 2/2] drm/ttm: revert "stop allocating dummy resources during BO creation"

[Matthew Auld]

On Wed, 25 Jan 2023 at 15:50, Christian König
<ckoenig.leichtzumerken at gmail.com> wrote:
>
> This reverts commit 00984ad39599bb2a1e6ec5d4e9c75a749f7f45c9.
>
> It seems to still breka i915.

We also need to revert the third patch:

b49323aa35d5 drm/ttm: prevent moving of pinned BOs

It introduces the side effect of no longer calling tt_create(true) in
ttm_bo_validate(), and I'm 99% sure that will break object clearing.
We rely on having a ttm_tt for the initial dummy placement, with
FLAG_ZERO_ALLOC set if clear is needed. Also I'm not sure who even
creates the ttm_tt now, if ttm_bo_validate() doesn't, and we don't
have the dummy move, like with this patch.

>
> Signed-off-by: Christian König <christian.koenig at amd.com>
> ---
>  drivers/gpu/drm/ttm/ttm_bo.c | 7 +++++++
>  1 file changed, 7 insertions(+)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
> index 33471e363ff4..9baccb2f6e99 100644
> --- a/drivers/gpu/drm/ttm/ttm_bo.c
> +++ b/drivers/gpu/drm/ttm/ttm_bo.c
> @@ -957,6 +957,7 @@ int ttm_bo_init_reserved(struct ttm_device *bdev, struct ttm_buffer_object *bo,
>                          struct sg_table *sg, struct dma_resv *resv,
>                          void (*destroy) (struct ttm_buffer_object *))
>  {
> +       static const struct ttm_place sys_mem = { .mem_type = TTM_PL_SYSTEM };
>         int ret;
>
>         kref_init(&bo->kref);
> @@ -973,6 +974,12 @@ int ttm_bo_init_reserved(struct ttm_device *bdev, struct ttm_buffer_object *bo,
>                 bo->base.resv = &bo->base._resv;
>         atomic_inc(&ttm_glob.bo_count);
>
> +       ret = ttm_resource_alloc(bo, &sys_mem, &bo->resource);
> +       if (unlikely(ret)) {
> +               ttm_bo_put(bo);
> +               return ret;
> +       }
> +
>         /*
>          * For ttm_bo_type_device buffers, allocate
>          * address space from the device.
> --
> 2.34.1
>