linux-kernel - Re: [PATCH] KEYS: prevent creating a different user's keyrings

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <26475.1506340460@warthog.procyon.org.uk>
Date:   Mon, 25 Sep 2017 12:54:20 +0100
From:   David Howells <dhowells@...hat.com>
To:     Eric Biggers <ebiggers3@...il.com>
Cc:     dhowells@...hat.com, keyrings@...r.kernel.org,
        Michael Halcrow <mhalcrow@...gle.com>,
        linux-security-module@...r.kernel.org,
        linux-kernel@...r.kernel.org, Eric Biggers <ebiggers@...gle.com>,
        stable@...r.kernel.org
Subject: Re: [PATCH] KEYS: prevent creating a different user's keyrings

Eric Biggers <ebiggers3@...il.com> wrote:

> Well, maybe.  Whitelists are hard to get right, and it would be a bit ugly
> having to check the name in both add_key() and join_session_keyring().  And
> hopefully that would be everything?

Actually, having thought about it some more, I think your way is better.

> I think there's also a more fundamental problem with how keyring names work.
> If you try to join a keyring with a certain name, how are you supposed to
> know which one you're joining?  There can be many keyrings that have the
> same name; and any unprivileged user can create a keyring with the name, and
> they can grant everyone SEARCH permission so that their keyring can be
> joined.  So it can be the case that a user is wanting to join a particular
> keyring, but they actually get a keyring that a malicious user has crafted
> for them...

Yeah.  With hindsight, I think that firstly, joinable keyrings really need
enablement and, secondly, thread, process, session, user and user-session need
to have to be non-manually-creatable.

However, I'm not sure they can be renamed, since they're searchable and
joinable by name and fixing this might break something in userspace (though I
should hope that this is unlikely).

> Also, if period ('.') is meant to be the reserved character in keyring names,
> why do most of the special names actually start with underscore ('_')?

'.' wasn't a reserved char originally.

David