On Fri, 21 Jun 2013, Konstantin Filatov wrote:
> > --- a/drivers/usb/host/uhci-q.c
> > +++ b/drivers/usb/host/uhci-q.c
> > @@ -977,6 +977,9 @@ static int uhci_submit_common(struct uhci_hcd *uhci,
> > struct urb *urb,
> > for (;;) { /* Allow zero length packets */
> > int pktsze = maxsze;
> >
> > + if (this_sg_len < pktsze) /* The short packet */
> > + pktsze = this_sg_len;
> > This will build a short packet transfer descriptor which is not the last
> > one, and it isn't correct.
> Why do you think so? Could you please refer to the specification of UHCI.
It _isn't_ correct. But the fault lies in the driver that submitted
the SG transfer in the first place, not in uhci-hcd. This is the right
thing to do.
Also, its the same as what ehci-hcd does. Oddly enough, it looks like
nobody ever added SG support to ohci-hcd.
By the way, the patch could be improved if it removed the line
pktsze = len;
This won't be needed, because the code is careful to guarantee that
this_sg_len <= len always.
> Your patch skips a chunk of data to transfer. This is a corruption of
> data. It's unacceptable. We can return an error code if we treat the
> request as malformed, but we can't transfer data selectively.
I agree. Fix up the patch description as Sergei suggested, and you can
add
Acked-by: Alan Stern <[email protected]>
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
