On Sat, Jun 18, 2016 at 04:20:47PM +0100, Chris Wilson wrote:
> Rendering operations to the dma-buf are tracked implicitly via the
> reservation_object (dmabuf->resv). The dmabuf sync ioctl allows
> userspace to wait upon outstanding rendering and prepare the object for
> CPU access (provided by the prime dma_buf_ops.begin_cpu_access). Fill
> this out for the generic drm_gem_prime by waiting on outstanding
> rendering via dmabuf->resv. (This offers an alternative to using poll
> that is consistent with other drivers that may need to more work to
> prepare the object for access by the CPU.)
>
> Signed-off-by: Chris Wilson
> ---
> drivers/gpu/drm/drm_prime.c | 18 ++
> 1 file changed, 18 insertions(+)
>
> diff --git a/drivers/gpu/drm/drm_prime.c b/drivers/gpu/drm/drm_prime.c
> index 780589b420a4..479ff7cc3634 100644
> --- a/drivers/gpu/drm/drm_prime.c
> +++ b/drivers/gpu/drm/drm_prime.c
> @@ -28,6 +28,7 @@
>
> #include
> #include
> +#include
> #include
> #include
>
> @@ -288,6 +289,22 @@ static int drm_gem_dmabuf_mmap(struct dma_buf *dma_buf,
> return dev->driver->gem_prime_mmap(obj, vma);
> }
>
> +static int drm_gem_begin_cpu_access(struct dma_buf *dma_buf,
> + enum dma_data_direction direction)
> +{
> + bool write = (direction == DMA_BIDIRECTIONAL ||
> + direction == DMA_TO_DEVICE);
> + struct reservation_object *resv = dma_buf->resv;
> + long ret;
> +
> + ret = reservation_object_wait_timeout_rcu(resv, write, true,
> + MAX_SCHEDULE_TIMEOUT);
> + if (ret < 0)
> + return ret;
> +
> + return 0;
> +}
Maybe we even want this in the dma-buf layer as default function if
nothing else is provided? After all this one here is entirely generic, and
uses neither gem nor even drm_prime knowledge. Either way:
Reviewed-by: Daniel Vetter
> +
> static const struct dma_buf_ops drm_gem_prime_dmabuf_ops = {
> .attach = drm_gem_map_attach,
> .detach = drm_gem_map_detach,
> @@ -301,6 +318,7 @@ static const struct dma_buf_ops drm_gem_prime_dmabuf_ops
> = {
> .mmap = drm_gem_dmabuf_mmap,
> .vmap = drm_gem_dmabuf_vmap,
> .vunmap = drm_gem_dmabuf_vunmap,
> + .begin_cpu_access = drm_gem_begin_cpu_access,
> };
>
> /**
> --
> 2.8.1
>
> ___
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
