Subject: Re: [PATCH -v3 1/3] Capabilities: move cap_file_mmap to commoncap.c
From: Stephen Smalley <sds@tycho.nsa.gov>
To: Eric Paris <eparis@redhat.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>, linux-kernel@vger.kernel.org,
       linux-security-module@vger.kernel.org, selinux@tycho.nsa.gov,
       jwcart2@tycho.nsa.gov, spender@grsecurity.net, dwalsh@redhat.com,
       cl@linux-foundation.org, arjan@infradead.org, alan@lxorguk.ukuu.org.uk,
       kees@outflux.net, csellers@tresys.com,
       penguin-kernel@i-love.sakura.ne.jp
In-Reply-To: <1248982958.11627.187.camel@moss-pluto.epoch.ncsc.mil>
References: <20090729185620.21757.44366.stgit@paris.rdu.redhat.com>
	 <20090730051426.GA6082@us.ibm.com> <1248968428.2597.79.camel@localhost>
	 <20090730155403.GA13939@us.ibm.com>
	 <1248969492.11627.172.camel@moss-pluto.epoch.ncsc.mil>
	 <1248976252.2771.1.camel@localhost>
	 <1248982958.11627.187.camel@moss-pluto.epoch.ncsc.mil>
Content-Type: text/plain
Organization: National Security Agency
Date: Thu, 30 Jul 2009 15:54:52 -0400
Message-Id: <1248983692.11627.194.camel@moss-pluto.epoch.ncsc.mil>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3598
Lines: 81

On Thu, 2009-07-30 at 15:42 -0400, Stephen Smalley wrote:
> On Thu, 2009-07-30 at 13:50 -0400, Eric Paris wrote:
> > On Thu, 2009-07-30 at 11:58 -0400, Stephen Smalley wrote:
> > > On Thu, 2009-07-30 at 10:54 -0500, Serge E. Hallyn wrote:
> > > > Quoting Eric Paris (eparis@redhat.com):
> > > > > On Thu, 2009-07-30 at 00:14 -0500, Serge E. Hallyn wrote:
> > > > > > Quoting Eric Paris (eparis@redhat.com):
> > > > > > > Currently we duplicate the mmap_min_addr test in cap_file_mmap and in
> > > > > > > security_file_mmap if !CONFIG_SECURITY.  This patch moves cap_file_mmap
> > > > > > > into commoncap.c and then calls that function directly from
> > > > > > > security_file_mmap ifndef CONFIG_SECURITY like all of the other capability
> > > > > > > checks are done.
> > > > > > 
> > > > > > It also
> > > > > > 
> > > > > > 	1. changes the return value in error case from -EACCES to
> > > > > > 	   -EPERM
> > > > > > 	2. no onger sets PF_SUPERPRIV in t->flags if the capability
> > > > > > 	   is used.
> > > > > > 
> > > > > > Do we care about these?
> > > > > 
> > > > > Personally, not really, but I'll gladly put them back if you care.   #2
> > > > > seems more interesting to me than number 1.   I actually kinda like
> > > > > getting EPERM from caps rather than EACCES since them I know if I was
> > > > > denied by selinux or by caps.....
> > > > > 
> > > > > -Eric
> > > > 
> > > > Yup, I asked bc I didn't particularly care myself.
> > > > 
> > > > I think I agree with you about -EPERM being better anyway.  However I
> > > > (now) think in this case PF_SUPERPRIV definately should be set, as this
> > > > is a clear use of a capability to do something that couldn't have been
> > > > done without it.
> > > 
> > > On a related but different note, we should consider all current uses of
> > > cap_capable(), as they represent capability checks that will not be
> > > subject to a further restrictive check by other security modules.  In
> > > this case and in the vm_enough_memory case, that is intentional, but not
> > > so clear for other uses in commoncap.c.
> > 
> > Most of commoncap.c is called either as a secondary hook from the active
> > lsm (aka selinux calls the commoncap.c functions) or in the !
> > CONFIG_SECURITY case.
> > 
> > I'll audit this afternoon to see which of them might not fit these
> > rules....
> 
> That isn't what I meant.  Most of the commoncap functions call capable()
> rather than directly calling cap_capable(), thereby causing:
> - PF_SUPERPRIV to be set, and
> - The primary security module (e.g. SELinux) to apply its own
> restrictive check.
> 
> That is useful as it allows SELinux or AppArmor or TOMOYO to veto e.g.
> CAP_SYS_PTRACE without replicating the same logic within its own hook.
> 
> The current exceptions are:
> cap_inh_is_capped() called from cap_capset(),
> cap_task_prctl() in the PR_SET_SECUREBITS case,
> cap_vm_enough_memory(),
> cap_file_mmap() after your patch.
> 
> The latter two are indeed cases where we made a conscious choice that
> SELinux would not apply its capability check against policy.  But the
> first two are unclear to me.

Sorry, the vm_enough_memory case was about auditing, not about applying
the check itself.  Which might have been rendered moot by your later
changes to introduce SECURITY_CAP_NOAUDIT.

-- 
Stephen Smalley
National Security Agency

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/