| From: |
| Dmitry Sepp <dmitry.sepp-AT-opensynergy.com> |
| To: |
| <linux-media-AT-vger.kernel.org> |
| Subject: |
| [RFC] virtio video driver |
| Date: |
| Thu, 5 Dec 2019 17:11:37 +0100 |
| Message-ID: |
| <4595464.3jghpSLKuc@os-lin-dmo> |
| Cc: |
| <virtio-dev-AT-lists.oasis-open.org>, <kraxel-AT-redhat.com>, <tfiga-AT-chromium.org>, <keiichiw-AT-chromium.org>, <acourbot-AT-chromium.org>, <hverkuil-AT-xs4all.nl>, <posciak-AT-chromium.org>, <marcheu-AT-chromium.org>, <stevensd-AT-chromium.org>, <dgreid-AT-chromium.org>, <daniel-AT-ffwll.ch>, <egranata-AT-google.com> |
| Archive-link: |
| Article |
Hello,
My apologies for the long delay. The driver code is now available and provided
as a follow-up to the discussion from this thread [1].
The reference Linux kernel 5.4 driver implementation is located here:
https://github.com/OpenSynergy/linux/tree/virtio-video-dr...
The driver is implemented using the V4L2 API. It allocates a v4l2 device for
each probed virtio device and then creates a video device for each function
within the respective virtio device. The driver implements the stateful
decoder interface [2] and the stateful encoder interface (WIP) [3].
The DMA SG memory allocator tries to map buffers right away. As it is not
always suitable, and some implementations might need just a physical address,
we had to introduce a set of simple dma ops directly in the driver.
The driver is in the RFC state and currently a bit ahead of the spec that was
proposed in the discussion mentioned above. On the other hand, the driver
unfortunately does not yet include changes proposed in the recent comments
[4]. The driver currently supports encoder and decoder functions. Also, it
does not fully pass the v4l2-compliance yet, it has been a bit out of the
focus so far.
Any feedback and contribution would be greatly appreciated.
[1] https://markmail.org/message/gc6h25acct22niut
[2] https://www.kernel.org/doc/html/v5.4/media/uapi/v4l/dev-d...
[3] https://hverkuil.home.xs4all.nl/codec-api/uapi/v4l/dev-en...
[4] https://markmail.org/message/yy67elx2adbivdsp
Best regards,
Dmitry.
