[Devel] [PATCH rh7 v2] ve/devpts: Force devpts mounting to use @newinstance inside VE

[Cyrill Gorcunov]

On Tue, Jul 21, 2015 at 03:25:34PM +0300, Vladimir Davydov wrote:
> > +	if (opts.newinstance && get_exec_env()->devpts_sb) {
> >  		root = mount_nodev(fs_type, flags, data, devpts_fill_super);
> > -	else
> > +	} else {
> > +		opts.newinstance = 0;
> >  		root = mount_ns(fs_type, flags, data, get_exec_env(), devpts_fill_super);
> > +	}
> 
> This looks like an abuse of mount_ns to me...
> 
> May be, we'd better simply revert to what we have in PCS6? I mean
> reverting commits 2c27d20125f51 and c77f3df733bfa. Less intrusive and
> more understandable if you ask me.

As far as I understand if we revert these commits we would
allow to mount host's devpts into container. At least I don't
see how this would be prevented.

The vanilla kernel relies on fact (again if I understand conrrectly)
that lxc and such tools would mount own devpts (as it does for example
centos-7's systemd) but for old containers and in particulary
ubuntu-14 this is not the case, they do quite the reverse -- mount
devpts without newinstance.