[GIT PULL] Samsung devel-2 for v3.2

[Russell King - ARM Linux]

On Mon, Oct 03, 2011 at 11:36:26AM +0100, Mark Brown wrote:
> On Mon, Oct 03, 2011 at 12:27:22PM +0900, Kukjin Kim wrote:
> > Mark Brown wrote:
> 
> > > As I reported the other day this change breaks the boot early on on the
> > > Cragganmore S3C6410 based systems (I've not actually fired up my
> > > SMDK6410 but I suspect it may have issues too).  I did make some brief
> > > attempts to debug but as there's no console output at all and the diff
> > > is rather large it's not easy to see what's gone wrong.
> 
> > Hmm, I thought it has no problem when I saw Thomas Abraham's 'serial:
> > samsung' patches because he said it works fine with my for-next plus his
> > patches on SMDK6410...Anyway, I need to sort it out.
> 
> It's possible it's something that doesn't manifest on that board, I've
> no real idea what's exploding or why.  There's also a bunch of other
> early boot problems right now like try_to_freeze() which aren't helping.

I don't know whether I can solve that before the next merge window.
There's lots of corner cases, and I'm no longer convinced that my
previous patch doesn't introduce more problems than it solves (such
as completely breaking the -ERESTART_RESTARTBLOCK handing.)

While it's possible to shut up the warning by enabling IRQs, that makes
the race for restarting (or not restarting) syscalls *huge*, and that
could cause unexpected userspace behaviour depending on whether they
handle EINTR and _how_ they handle EINTR on system calls which they
expect would've been restarted by the kernel.