Re: [kernel-hardening] [RFC 2/5 v4] procfs: add hidepid= and gid= mount options

[Vasiliy Kulikov <segoon@openwall.com>]

Solar,

On Mon, Jun 20, 2011 at 18:25 +0400, Solar Designer wrote:
> On Mon, Jun 20, 2011 at 06:19:51PM +0400, Vasiliy Kulikov wrote:
> > On Mon, Jun 20, 2011 at 18:11 +0400, Solar Designer wrote:
> > > On Mon, Jun 20, 2011 at 04:58:10PM +0300, Alexey Dobriyan wrote:
> > > > gid= is bad choice because
> > > > a) e. g. VFAT uses uid=/gid= mount options to make all inodes to have
> > > > certain uid/gid
> > > > b) uid=/gid=, IIRC, will be added as generic VFS mount options (like ro)
> > > >    with semantics described in a)
> > > > 
> > > > so having different semantics for /proc won't be good.
> > > 
> > > I lost track of your proposals/patches.  Aren't you currently proposing
> > > that gid= would make all inodes have the specified gid?  If not, why
> > > not?  Such semantics sound fine to me.  That's what gid= does on procfs
> > > on Linux 2.4.x-ow.
> > 
> > With taskstats and similar mechanisms IMO it's better to use sysctls
> > instead of procfs mount options as it would influence not only on procfs
> > files.
> 
> OK, but why did we receive a comment about the gid= mount option then
> (quote above)? Are you still proposing it or not anymore?

No, till v4 I assumed there is no way to gather processes' information
except procfs:

http://www.openwall.com/lists/kernel-hardening/2011/06/16/6

>  Is the
> question "outdated" (resulting from your earlier proposal, not the
> latest one)?

Yes.

I didn't post a patch with taskstats and sysctl variables to LKML yet
(only the changes in ptrace/capabilities code).


Thanks,

-- 
Vasiliy