[CRIU] [PATCH 1/8] signals: dump alternate stack as misc parameters

[Cyrill Gorcunov]

On Thu, Jul 11, 2013 at 05:28:36PM +0400, Andrey Wagin wrote:
> 2013/7/11 Andrey Wagin <avagin at gmail.com>:
> > 2013/7/11 Pavel Emelyanov <xemul at parallels.com>:
> >> On 07/11/2013 01:46 PM, Andrey Vagin wrote:
> >>> The initilization stage is not good for that, because it can fail.
> >>
> >> IMO it should be done as early as possible, as the sas info is to
> >> be used by sigreturn rescue from parasite.
> >
> > Yes, you are rigth. I forgot about this. Thanks.
> 
> I was thinking a bit more and changed my mind.
> 
> If sas is NULL in sigframe, it will not changed by sigreturn. parasite
> doesn't change sas, so this patch is correct.

I must admit I didn't read this patch carefully but indeed, since we
don't touch sas in criu at all, it's safe to not restore sas on rescue
exit path, it'll remain the same as it was before parasite injected.

> 
> Currently sigframe is generated before executing a parasite. We should
> to do minimum actions before entering in daemon mode, because only
> here parasite can rollback itself.