Re: 2.6.25 regression: vivi - scheduling while atomic

[Mauro Carvalho Chehab <mchehab@infradead.org>]

> > > > This happenes on a vanilla 2.6.25 with loaded nvidia graphics module.
> > > > System architecture is x86_64. If it matters I'll try to reproduce this
> > > > error on a non tainted kernel.
> > > 
> > > No need to reproduce on a non-tainted Kernel.  This is a known issue
> > > with patches merged into the v4l-dvb tree several weeks ago but it seems
> > > to not have made it into 2.6.25 :(
> > > 
> > >  http://linuxtv.org/hg/v4l-dvb/rev/06eb92ed0b18
> > >  http://linuxtv.org/hg/v4l-dvb/rev/c50180f4ddfc
> > > 
> > > I can rebase the patches for 2.6.25 but they are too big to go into the
> > > stable 2.6.25 tree...

Unfortunately, the tests for the patches fixing several videobuf issues
finished later on -rc9, when Linus were releasing 2.6.25. 

I should send the videobuf patches to Linus tree, together with other patches,
probably today. After that, I think we should backport the fixes for 2.6.25,
and test it again, with vanilla 2.6.25.

The issue here is that videobuf suffered several changes on this development
cycle. Probably, only a few of those patches are needed to fix the issue. So,
we need to handle this with care, to avoid the risk of damaging other drivers that
relies on videobuf.

Cheers,
Mauro