On Fri, Nov 09, 2012 at 10:28:38AM -0700, Stephen Warren wrote:
> On 11/09/2012 10:10 AM, Mark Brown wrote:
> > On Fri, Nov 09, 2012 at 10:04:56AM -0700, Stephen Warren wrote:
> > 
> >> However just FYI, it should not be necessary for correctness; The
> >> DT matching order is supposed to be driven purely by the order of
> >> the compatible values in the DT now, and not affected by the
> >> order of values in the table. (This wasn't always the case, but
> >> was a bug that was fixed IIRC by Thierry Reding).
> > 
> > I guess the driver is being used backported in older kernels which
> > don't have that fix?
> 
> That sounds likely. Laxman, it'd be a good idea to track down the fix
> to the DT matching code and backport it, so that hard-to debug issues
> aren't caused by the lack of that patch!

Unfortunately the patch that was supposed to fixed this caused a
regression and was therefore reverted. Rob (Cc'ed) said there was a
patch to fix it properly and was supposed to go into 3.6 but it seems
that never happened. Rob, what's the status on this?

The revert is here: bc51b0c22cebf5c311a6f1895fcca9f78efd0478

Thierry