On Fri, 2012-12-14 at 15:11 +0100, Bjørn Mork wrote:
> Herton Ronaldo Krzesinski <[email protected]> writes:
> 
> > There is a problem in stable kernels that included the commit cafbe85
> > ("USB: cdc-wdm: better allocate a buffer that is at least as big as we
> > tell the USB core")
> >
> > In that commit, the usb_alloc_coherent call was changed without also
> > adjusting usb_free_coherent calls. I got a report about it on
> > http://bugs.launchpad.net/bugs/1074157
> 
> Ouch!  That's embarrassing...  Thanks for finding and fixing it.
> 
> > But right after commit cafbe85, marked for stable, usb_alloc_coherent
> > was removed from the driver, so upstream is not affected by the issue,
> > while some stables are, the ones which included cafbe85, and one
> > solution must be considered for them.
> >
> > My question is, should stables pick commit 8457d99, which removes the
> > usb_alloc_coherent usage and will also solve the bug for them, or may be
> > for example use the version below, which also makes clear why it isn't
> > taking an upstream commit:
> 
> I'll leave the for the stable maintainers to decide, but one of these
> should be done, yes.
> 
> > Please advise, ack and consider for inclusion one of the solutions in
> > all stable kernels which have commit cafbe85, but still uses
> > usb_alloc_coherent.
> 
> That would be 3.0.y and 3.2.y AFAICS.  3.4.0 and later has commit
> 8457d99 which removes the problem as you say, and 2.6.32.y does not have
> cafbe85.

I've queued up Herton's patch for 3.2, thanks.

Ben.

-- 
Ben Hutchings
The world is coming to an end.  Please log off.

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to