Re: [RFCv2] Request API

[Sakari Ailus <sakari.ailus@iki.fi>]

Hi Hans,

Thanks for writing this. I generally agree with the RFC to the level of
detail available here; a few comments below.

On Fri, Mar 23, 2018 at 09:46:00AM +0100, Hans Verkuil wrote:
> RFC Request API, version 2
> --------------------------
> 
> This document proposes the public API for handling requests.
> 
> There has been some confusion about how to do this, so this summarizes the
> current approach based on conversations with the various stakeholders today
> (Sakari, Alexandre Courbot, Thomasz Figa and myself).
> 
> 1) Additions to the media API
> 
>    Allocate an empty request object:
> 
>    #define MEDIA_IOC_REQUEST_ALLOC _IOW('|', 0x05, __s32 *)
> 
>    This will return a file descriptor representing the request or an error
>    if it can't allocate the request.
> 
>    If the pointer argument is NULL, then this will just return 0 (if this ioctl
>    is implemented) or -ENOTTY otherwise. This can be used to test whether this
>    ioctl is supported or not without actually having to allocate a request.
> 
> 2) Operations on the request fd
> 
>    You can queue or reinit a request by calling these ioctls on the request fd:
> 
>    #define MEDIA_REQUEST_IOC_QUEUE   _IO('|',  128)
>    #define MEDIA_REQUEST_IOC_REINIT  _IO('|',  129)
> 
>    With REINIT you reset the state of the request as if you had just allocated
>    it.
> 
>    You can poll the request fd to wait for it to complete.

I'm not sure I like having to release a reference (fd) to a request to know
whether or not it succeeded. I put this as an open question on my RFC
patchset.

> 
>    To free a request you close the request fd. Note that it may still be in
>    use internally, so the internal datastructures have to be refcounted.
> 
>    For this initial implementation only buffers and controls are contained
>    in a request. This is needed to implement stateless codecs. Supporting
>    complex camera pipelines will require more work.
> 
>    Requests only contain the changes to the state at request queue time
>    relative to the previously queued request(s) or the current hardware state
>    if no other requests were queued.
> 
>    Once a request is completed it will retain the state at completion
>    time.
> 
> 3) To associate a v4l2 buffer with a request the 'reserved' field in struct
>    v4l2_buffer is used to store the request fd. Buffers won't be 'prepared'
>    until the request is queued since the request may contain information that
>    is needed to prepare the buffer.
> 
>    To indicate that request_fd should be used this flag should be set by
>    userspace at QBUF time:
> 
> #define V4L2_BUF_FLAG_REQUEST			0x00800000
> 
>    This flag will also be returned by the driver to indicate that the buffer
>    is associated with a request.
> 
>    TBD: what should vb2 return as request_fd value if this flag is set?
>    This should act the same as the fence patch series and this is still
>    being tweaked so let's wait for that to be merged first, then we can
>    finalize this.
> 
> 4) To associate v4l2 controls with a request we take the first of the
>    'reserved[2]' array elements in struct v4l2_ext_controls and use it to store
>    the request fd.
> 
>    We also add a new WHICH value:
> 
> #define V4L2_CTRL_WHICH_REQUEST   0x0f010000
> 
>    This tells the control framework to get/set controls from the given
>    request fd.
> 
>    When querying a control value from a request it will return the newest
>    value in the list of pending requests, or the current hardware value if
>    is not set in any of the pending requests.
> 
>    When a request is completed the controls will no longer change. A copy
>    will be made of volatile controls at the time of completion (actually
>    it will be up to the driver to decide when to do that).
> 
>    Volatile controls and requests:
> 
>    - If you set a volatile control in a request, then that will be ignored,
>      unless the V4L2_CTRL_FLAG_EXECUTE_ON_WRITE flag is set as well.
> 
>    - If you get a volatile control from a request then:
>      1) If the request is completed it will return the value of the volatile
>         control at completion time.
>      2) Otherwise: if the V4L2_CTRL_FLAG_EXECUTE_ON_WRITE is set and it was
>         set in a request, then that value is returned.
>      3) Otherwise: return the current value from the hardware (i.e. normal
>         behavior).
> 
>    Read-only controls and requests:
> 
>    - If you get a read-only control from a request then:
>      1) If the request is completed it will return the value of the read-only
>         control at completion time.
>      2) Otherwise it will get the current value from the driver (i.e. normal
>         behavior).
> 
>    Open issue: should we receive control events if a control in a request is
>    added/changed? Currently there are no plans to support control events for
>    requests. I don't see a clear use-case and neither do I see an easy way
>    of implementing this (which fd would receive those events?).
> 
> Notes:
> 
> - Earlier versions of this API had a TRY command as well to validate the
>   request. I'm not sure that is useful so I dropped it, but it can easily

I'd think this would be mostly useful on complex pipelines where you'll
have interdependencies in configuration across entities. One such case is
the available V4L2 formats given a choice of a media bus format towards the
DMA engine: you can't use the current APIs to try that (dirty one-off fix
could be to provide the mbus code on the same IOCTL) but there are more
complex cases, too.

Currently this is a niche use case, but moving all IOCTLs to the media
device, depending a bit on how that would be done, is what I'd expect to
change this completely.

Because this is not very relevant right now, I left it out.

>   be added if there is a good use-case for it. Traditionally within V4L the
>   TRY ioctl will also update wrong values to something that works, but that
>   is not the intention here as far as I understand it. So the validation
>   step can also be done when the request is queued and, if it fails, it will
>   just return an error.
> 
> - If due to performance reasons we will have to allocate/queue/reinit multiple
>   requests with a single ioctl, then we will have to add new ioctls to the
>   media device. At this moment in time it is not clear that this is really
>   needed and it certainly isn't needed for the stateless codec support that
>   we are looking at now.

I think the time for this is later indeed, likely much later.

> 
> - The behavior of VIDIOC_G_EXT_CTRLS with which == V4L2_CTRL_WHICH_CUR_VAL
>   and VIDIOC_G_CTRL remains the same (i.e. it returns the current driver/HW
>   values). However, when combined with requests the documentation should make
>   clear that this returns a snapshot only and is racy w.r.t. applying values
>   from a request.
> 
> - There is a discussion whether there should be a VIDIOC_REQUEST_ALLOC ioctl
>   for V4L2 in addition to the media ioctl. The reason is that stateless codecs
>   do not need the media controller except for allocating requests. So a V4L2
>   ioctl would avoid applications from having to deal with a media device.
>   This would also add additional hassle w.r.t. SELinux as I understand it.
> 
>   Support for this can be added for now as a final patch in the Request API
>   patch series and we'll postpone the decision on this.

Works for me.

-- 
Kind regards,

Sakari Ailus
e-mail: sakari.ailus@iki.fi