[Pkg-shadow-devel] Bug#470745: Bug#470745: Bug#470745: passwd: usermod loops and mem leaks

[Justin Pryzby]

On Thu, Mar 13, 2008 at 04:06:55PM -0400, Justin Pryzby wrote:
> On Thu, Mar 13, 2008 at 08:32:39PM +0100, Thomas Rasmussen wrote:
> > On Thu, Mar 13, 2008 at 4:30 PM, Nicolas François <nicolas.francois at centraliens.net> wrote:
> 
> > > On Thu, Mar 13, 2008 at 01:30:45PM +0100, rasmussen.thomas at gmail.comwrote:
> > > > If /etc/gshadow file has been changed so two otherwise non-identical
> > > > groups apear with the same groupname, usermod will loop and use all memory
> > > > on system if called.
> > > >
> > > > Reproducable by performing this:
> > > > # groupadd tr
> > > > # groupadd rtr
> > > > # useradd -g tr tr
> > > > # perl -pi -e 's/rtr/tr/g' /etc/gshadow
> > > > # usermod -G tr tr
> > > > <observe usermod using memory and proc time>
> > > >
> > > > Tested and reproduced on latest (4.0r3) netinst iso image and updated
> > > > with all packages.
> 
> > > I don't think this bug is that critical. I doesn't cause any
> > > corruption, and it occurs only in case of configuration errors,
> > > which do not appear if the administrators use the recommended
> > > tools for user/group administration. Thus I will not issue a fix
> > > for this bug.
> 
> > I agree that this should not be seen as admins should use the dedicated
> > toos, but I still see this as a quite serious bug because it will crash
> > either the system or random services running on the system when it hits the
> > maximum amount of memory.
> Really?  If usermod spins and sucks up memory, it *should* just get
> killed for OOM, perhaps after causing some disk usage.  Other
> processes might be affected by IO time, CPU usage and such, but I
> think neither the whole system nor any services should become
> unavailable (just their response time might suffer).
Nevermind, after trying it, I take that back.  It has a large effect
on the system probably because it necessarily runs as root, so OOM
probably hesitates to kill it.

Justin