On Thu, 26 May 2022 at 18:30, Simon Ser <[email protected]> wrote:
>
> Document how to obtain the size of a DMA-BUF. This is what
> Wayland compositors are doing.
>
> Mention the kernel version numbers from which DMA-BUF features are
> available.
>
> Signed-off-by: Simon Ser <[email protected]>
> Cc: Daniel Vetter <[email protected]>
> Cc: Jason Ekstrand <[email protected]>
> ---
> include/uapi/linux/dma-buf.h | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/include/uapi/linux/dma-buf.h b/include/uapi/linux/dma-buf.h
> index 8e4a2ca0bcbf..c95f6d3457d2 100644
> --- a/include/uapi/linux/dma-buf.h
> +++ b/include/uapi/linux/dma-buf.h
> @@ -47,9 +47,13 @@
> *
> * If the driver or API with which the client is interacting uses implicit
> * synchronization, waiting for prior work to complete can be done via
> - * poll() on the DMA buffer file descriptor. If the driver or API requires
> - * explicit synchronization, the client may have to wait on a sync_file or
> - * other synchronization primitive outside the scope of the DMA buffer API.
> + * poll() on the DMA buffer file descriptor from kernel version 3.17. If the
> + * driver or API requires explicit synchronization, the client may have to
> wait
> + * on a sync_file or other synchronization primitive outside the scope of the
> + * DMA buffer API.
This looks good, but you missed the DOC: implicit fence polling in
dma-buf.c. Probably good to update that too.
> + *
> + * From kernel version 3.12, user-space can use llseek(2) with the
> ``SEEK_END``
> + * whence to obtain the size of a DMA-BUF.
This feels misplaced, especially since this ends up under the "DMA
Buffer ioctls" heading.
Maybe simplest to just put a "DMA Buffer misc uAPI" section directly
into dma-buf.rst and put that section there? Also probably good to add
that the size is invariant.
Or if you don't want a new heading, put it somewhere sensible in the
mmap section? Also just noticed that the DOC: cpu access section could
perhaps link to the relevant ioctl docs here to connect things better.
-Daniel
> */
> struct dma_buf_sync {
> /**
> --
> 2.36.1
>
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
