[PATCH] drm/amdgpu: Attach exclusive fence to prime exported bo's. (v2)

[Christian König]

Am 05.11.2016 um 01:56 schrieb Mario Kleiner:
> External clients which import our bo's wait only
> for exclusive dmabuf-fences, not on shared ones,
> so attach fences on such exported buffers as
> exclusive ones.
>
> See discussion in thread:
> https://lists.freedesktop.org/archives/dri-devel/2016-October/122370.html
>
> Tested on Intel iGPU + AMD Tonga dGPU as DRI3/Present
> Prime render offload, and with the Tonga standalone as
> primary gpu.
>
> v2: Add a wait for all shared fences before prime export,
>      as suggested by Christian Koenig.
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95472
> (v1) Tested-by: Mike Lothian <mike at fireburn.co.uk>
> Signed-off-by: Mario Kleiner <mario.kleiner.de at gmail.com>
> Cc: Christian König <christian.koenig at amd.com>
> Cc: Michel Dänzer <michel.daenzer at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu.h         |  1 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c |  2 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c   | 11 +++++++++++
>   3 files changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> index 039b57e..a337d56 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
> @@ -459,6 +459,7 @@ struct amdgpu_bo {
>   	u64				metadata_flags;
>   	void				*metadata;
>   	u32				metadata_size;
> +	bool				prime_exported;
>   	/* list of all virtual address to which this bo
>   	 * is associated to
>   	 */
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
> index 651115d..51c6f60 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c
> @@ -132,7 +132,7 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev,
>   		entry->priority = min(info[i].bo_priority,
>   				      AMDGPU_BO_LIST_MAX_PRIORITY);
>   		entry->tv.bo = &entry->robj->tbo;
> -		entry->tv.shared = true;
> +		entry->tv.shared = !entry->robj->prime_exported;
>   
>   		if (entry->robj->prefered_domains == AMDGPU_GEM_DOMAIN_GDS)
>   			gds_obj = entry->robj;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
> index 7700dc2..0ed92a1 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
> @@ -117,9 +117,20 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev,
>   					int flags)
>   {
>   	struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj);
> +	long ret;
>   
>   	if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm))
>   		return ERR_PTR(-EPERM);
>   
> +	/*
> +	 * Wait for all shared fences to complete before we switch to future
> +	 * use of exclusive fence on this prime_exported bo.
> +	 */
> +	ret = reservation_object_wait_timeout_rcu(bo->tbo.resv, true, false,
> +						  MAX_SCHEDULE_TIMEOUT);
> +	WARN(ret <= 0, "Fence wait returns %li\n", ret);
> +
> +	bo->prime_exported = true;
> +

As Michel noted as well this is the wrong function for this, cause this 
is used for buffer sharing without DMA-buf as well. Try to put it into 
amdgpu_gem_prime_pin() instead.

I would also improve the error handling so that we return the error when 
the wait failed.

Apart from that this looks really good to me.

Regards,
Christian.

>   	return drm_gem_prime_export(dev, gobj, flags);
>   }