Re: [PATCH] userns: honour no_new_privs for cap_bset during user ns creation/switch

From: ebiederm@xmission.com (Eric W. Biederman)
To: Aleksa Sarai <asarai@suse.de>
Cc: "Maciej Żenczykowski" <zenczykowski@gmail.com>,
	"Linux Containers" <containers@lists.linux-foundation.org>,
	linux-security-module@vger.kernel.org,
	"Mahesh Bandewar" <maheshb@google.com>,
	"Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>,
	"Willem de Bruijn" <willemb@google.com>
Subject: Re: [PATCH] userns: honour no_new_privs for cap_bset during user ns creation/switch
Date: Fri, 22 Dec 2017 08:21:38 -0600	[thread overview]
Message-ID: <87r2rmg7wt.fsf@xmission.com> (raw)
In-Reply-To: <20171222021733.rerkt6mhpf3cb3oe@gordon> (Aleksa Sarai's message of "Fri, 22 Dec 2017 13:17:34 +1100")

Aleksa Sarai <asarai@suse.de> writes:

> On 2017-12-21, Eric W. Biederman <ebiederm@xmission.com> wrote:
>> Good point about CAP_DAC_OVERRIDE on files you own.
>> 
>> I think there is an argument that you are playing dangerous games with
>> the permission system there, as it isn't effectively a file you own if
>> you can't read it, and you can't change it's permissions.
>
> This problem reminds me of the whole "unmapped group" problem. If you
> have access to a file through an unmapped group you can still access a
> file -- which to me is wrong. I understand the need for checking
> unmapped groups in order to fix the "chmod 707" problem, but I think
> that unmapped groups should only *block* access and never *grant* it.
>
> I was working on a patch for that issue a while ago but it touched more
> VFS than I was comfortable with. Eric, is that a fix you would be
> interested in?

I am not certain.  I don't see how there is a problem with an unmapped
group granting permissions.  You are talking about a scenario where a
more privileged login program set your groups, and uid and gid.  The
process despite being a user namespace does not have permission to
transition them.  As such I don't see the harm.

But spell it out for me, and deal with ensuring we don't have user space
regressions and I will take a patch that improves the security of user
namespaces.

I think the issue that raised all of this is that dropping a group can
in rare instances increase permissions.  I have heard people grumble at
me that the way I handle it with /etc/subuid might break things on some
people's systems.  AKA don't allow it by default but allow root to
configure a way for people using user namespaces to do that.  I have yet
to see someone come forward and say that is a problem in the real world.
If it actually is a problem I want to hear about it.

Eric