Quoting Miles Lane <[EMAIL PROTECTED]>:
> On Sun, 22 Apr 2001, Dmitri wrote:
>
> <snip>
>
> > Updates to ibmcam are in CVS at SF. New driver supports two more
> > cameras (thus covering all IBM/Xirlink products that I know of),
> > moves processing into user context and is overall cleaner.
> > It is a "minidriver" that plugs into usbvideo library (module),
> > this makes about 1,500 lines of code shared among all V4L drivers
> > who care to use usbvideo.
>
> Does this include the se401 driver?
All drivers are welcomed :-) I recall, we discussed it briefly,
but since I never had any of se* cameras I couldn't work on that.
From what I know, only ibmdrv and ultradrv use usbvideo, and
both were written by me.
> If not, should the se401
> driver be modified to take advantage of this shared minidriver?
I would guess so because shared code is good if done properly.
Existing USB V4L drivers actually reimplement the same problem
again and again - allocation of mapped memory, dealing with
obscure V4L ioctls, USB Isoc data pump... this should all be done
in one place.
> How hard would the modifications be, if needed?
The biggest question is what data pipe you use (Isoc vs. Bulk)
and whether you need "blocks" of Isoc data (matching transfers)
or you just see it as one continuous stream. If former you need
to handle queuing differently, and this probably should be
added to videodev as a message queue instead of just ring queue.
Thanks,
Dmitri
--
"Never make any mistaeks."
(Anonymous, in a mail discussion about to a kernel bug report.)
PGP signature
