* Request change in behaviour of capability inheritance.
@ 2007-03-08 8:32 rgarcia
2007-03-08 16:24 ` Casey Schaufler
0 siblings, 1 reply; 3+ messages in thread
From: rgarcia @ 2007-03-08 8:32 UTC (permalink / raw)
To: linux-kernel
I think that the current behaviour of capability inheritance across exec()
is not optimal.
The current behaviour consists in all effective and permitted capabilities
are cleared across a exec(). This is because it seems to be intended that
in the future the executable files have a set of "allowed" and "forced"
capabilities. If a capability is not allowed, it is not permitted in an
inherited process. But at pressent, there is no attribute of "allowed"
capabilities in the filesystem, and the behaviour is that no capability is
allowed.
I request to change this behaviour. By default all capabilities should be
allowed, that is, inherited across exec().
This is neccessary to protect a daemon without changing its source. With
the requested behaviour, it would be posible to launch a daemon intented
to run as root, as an unpriviledged user but with the neccessary
capabilities, by using a script or program that changes the uid and the
capabilities and then launches the daemon.
Thus the present behaviour of no "allowed" capabilities pretends to be
more secure, but is actually less secure. By default, all parent
priviledges should be inherited, because this allows one to create
programs and scripts that execute daemons or other programs in a
restricted environment.
I am modifying kernel source for our own setup. I think others should
----------------------------------------------
Ramon Garcia Fernandez
Kotasoft, your experts in Linux consulting
----------------------------------------------
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Request change in behaviour of capability inheritance.
2007-03-08 8:32 Request change in behaviour of capability inheritance rgarcia
@ 2007-03-08 16:24 ` Casey Schaufler
2007-03-08 17:00 ` rgarcia
0 siblings, 1 reply; 3+ messages in thread
From: Casey Schaufler @ 2007-03-08 16:24 UTC (permalink / raw)
To: rgarcia, linux-kernel
--- rgarcia@kotasoft.com wrote:
> I think that the current behaviour of capability
> inheritance across exec()
> is not optimal.
>
> The current behaviour consists in all effective and
> permitted capabilities
> are cleared across a exec(). This is because it
> seems to be intended that
> in the future the executable files have a set of
> "allowed" and "forced"
> capabilities.
File based capabilities are the same
effective, permitted, and inheritable
sets that the process has. Unless the
thinking has shifted over the last couple
weeks (it happens from time to time)
the inheritable set is what you're
missing.
Casey Schaufler
casey@schaufler-ca.com
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2007-03-08 17:01 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2007-03-08 8:32 Request change in behaviour of capability inheritance rgarcia
2007-03-08 16:24 ` Casey Schaufler
2007-03-08 17:00 ` rgarcia
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®