On 2024/12/20 20:04, Kirill A. Korinsky wrote:
> On Fri, 20 Dec 2024 19:40:02 +0100,
> Martin Pieuchot <mpi@grenadille.net> wrote:
> > 
> > On 19/12/24(Thu) 14:18, Kirill A. Korinsky wrote:
> > > [...] 
> > > However, the situation is different for IN transfers. The kernel reads
> > > packets from the device, but they may be shorter than the expected size. As
> > > a result, the kernel tracks the total length of packets read using
> > > xfer->actlen and updates xfer->frlengths[i] accordingly.
> > 
> > Understood.  I'd prefer if we could find another solution than adding an
> > array to every USB transfer descriptor.
> > It seems to me that all isochronous USB drivers use frames of the same size,
> > which mean we could calculate the offset with:
> > 
> > 	(xfer->length / xfer->nframes)
> > 
> > Could that work?
> > 
> > On a related note, is their any OS and/or driver that use isochronous
> > transfers with frames of different sizes?
> > 
> 
> I made a quick grep through the Linux sources and was able to locate an
> example with a different offset and length in a driver:
> https://github.com/torvalds/linux/blob/v6.12/drivers/isdn/hardware/mISDN/hfcsusb.c#L1245-L1321

I highly doubt anyone will be writing an OpenBSD driver for those devices