[lxc-devel] [PATCH 1/1] seccomp: add 'lxc.seccomp.optional'

Serge Hallyn serge.hallyn at ubuntu.com
Mon Feb 17 22:58:36 UTC 2014


Quoting Stéphane Graber (stgraber at ubuntu.com):
> On Mon, Feb 17, 2014 at 05:16:51PM -0500, Stéphane Graber wrote:
> > On Mon, Feb 17, 2014 at 02:05:10PM -0600, Serge Hallyn wrote:
> > > If that is set, then if reading the policy failed, we continue
> > > without trying to load seccomp.  (If reading the policy
> > > succeeded, then we do not ignore failure to load the policy;
> > > we could consider doing that as well, however the goal here
> > > is to have a generic container configuration work whether
> > > the host has seccompv2 support or not)
> > > 
> > > Signed-off-by: Serge Hallyn <serge.hallyn at ubuntu.com>
> >
> > Acked-by: Stéphane Graber <stgraber at ubuntu.com>
> 
> Actually, I'll wait a bit before pushing this, the code isn't
> technically wrong (that I could see) but it's pretty confusing.
> 
> The option isn't documented in lxc.container.conf.sgml and the behavior
> of the option is a bit odd too.
> 
> Let's take a lxc-clone example:
>  If .optional == 1 and .seccomp is invalid => neither option appear in target container
>  If .optional == 1 and .seccomp is valid => only .seccomp appears in the target
> 
> It's also not possible to query its value or set it through the API.

I should have put RFC in the subject.  I'm still not sure it's the right
thing to do in any case.  On the one hand, it helps for having a single
config usable in many places.  On the other hand, if I'm running
something I distrust enough to confine it with seccomp, I'd prefer to
not have to worry about whether 'lxc.seccomp.optional = 1' was set
somewhere...

Let's shelve this for now.  Thanks for looking.


More information about the lxc-devel mailing list