On Thu, Jun 24, 2004 at 10:26:06AM -0400, Alan Stern wrote:
> > > ===== drivers/usb/host/uhci-hcd.c 1.115 vs edited =====
> > > --- 1.115/drivers/usb/host/uhci-hcd.c   Wed Jun 23 12:10:07 2004
> > > +++ edited/drivers/usb/host/uhci-hcd.c  Wed Jun 23 12:30:05 2004
> > > @@ -2261,6 +2261,11 @@
> > >                uhci->fl->frame[i] = cpu_to_le32(uhci->skelqh[irq]->dma_handle);
> > >        }
> > > 
> > > +       /*
> > > +        * Some architectures require a full mb() to enforce completion of
> > > +        * the memory writes above before the I/O transfers in start_hc().
> > > +        */
> > > +       mb();
> > >        start_hc(uhci);
> > > 
> > >        init_stall_timer(hcd);
> > 
> > Any reason why a wmb() is not sufficient?
> 
> See http://marc.theaimsgroup.com/?l=linux-usb-devel&m=108759528909973&w=2

Well I think this is something that we should document.

The question is what is the appropriate mechanism for synchronising
a direct write to DMA coherent memory and a subsequent writel().

include/asm-ppc/system.h says that mb() should be used, while
include/asm-ppc64/system.h says that wmb() should be and is used by
many drivers.

So which is it? Or perhaps we need something else?

Cheers,
-- 
Visit Openswan at http://www.openswan.org/
Email:  Herbert Xu ~{PmV>HI~} <[EMAIL PROTECTED]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to