Prism54 USB Wireless Adapters

[irchat at free.fr]

> The only thing unusual I  could see about the Prism54 driver
> is that it does a lot of opening and closing of pipes

Yeah, Prism54 devices are actually a pain :(

> so maybe something like that is tickling EHCI bugs.

> >Testing my driver with or without USBD_NO_COPY and the EHCI makes the
> > interrupt callback never called. With UHCI and without USBD_NO_COPY, my
> > code fails when the xfers using the buffers that don't support DMA are
> > submitted.
>
> Are you saying that something doesn't work with UHCI too? If you
> avoid the USBD_NO_COPY flag you shouldn't have to worry about DMA
> buffers, since the USB code will handle them itself. Or do you mean
> something different?

No, I meant that UHCI worked fine as it reports the error if I try to use
non-DMA-capable buffers with DMA, unlike EHCI which doesn't report anything
wrong and behaves improperly then.

I have worked a lot on the driver lately, see
http://yoshiyo.ath.cx/seb/prism54_freebsd.html
The DMA problems are fixed, I now use usbd_alloc_buffer().
Sebastien