Re: [PATCH] fix NULL pointer with MUSB and DMA on ISO ep

[Tony Lindgren <tony@atomide.com>]

* David Brownell <david-b@pacbell.net> [071102 09:37]:
> On Thursday 01 November 2007, Hunyue Yau wrote:
> > Hi,
> > 
> > The MUSB driver fails to pass a proper buffer pointer to
> > musb_write_fifo if DMA is enabled and it is an ISO endpoint. 
> > This removes the check that was preventing a proper buffer
> > pointer from being passed. 
> > 
> > Tested with a bluetooth headset and USB bluetooth dongle
> > using SCO audio on the 2430SDP.
> > 
> > -- Hunyue 
> > ---------------------------------------------------------------------------------
> > Description:
> >     For the case of an isochonous endpoint, musb_write_fifo may be 
> >     called with a NULL buffer from musb_host_tx if DMA is enabled.
> >     Remove DMA check and always supply a valid buffer as long as 
> >     the URB contains a valid buffer.
> > 
> > Signed-off-by: Hunyue Yau <hyau@mvista.com>
> > 
> > diff --git a/drivers/usb/musb/musb_host.c b/drivers/usb/musb/musb_host.c
> > index e488ac7..a7fe072 100644
> > --- a/drivers/usb/musb/musb_host.c
> > +++ b/drivers/usb/musb/musb_host.c
> > @@ -1258,7 +1258,7 @@ void musb_host_tx(struct musb *musb, u8 epnum)
> >  			d->actual_length = qh->segsize;
> >  			if (++qh->iso_idx >= urb->number_of_packets) {
> >  				done = true;
> > -			} else if (!dma) {
> > +			} else {
> 
> ... looks right.  ACK.
> 
> >  				d++;
> >  				buf = urb->transfer_buffer + d->offset;
> >  				wLength = d->length;
> > 

Pushing today.

Tony