netdev - Re: Possible mistake in commit 3ca459eaba1b ("tun: fix group permission check")

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

Message-ID: <e8b6c6f9-9647-4ab6-8bbb-ccc94b04ade4@yandex.ru>
Date: Mon, 27 Jan 2025 13:00:03 +0300
From: stsp <stsp2@...dex.ru>
To: Ondrej Mosnacek <omosnace@...hat.com>,
 Willem de Bruijn <willemb@...gle.com>, Jason Wang <jasowang@...hat.com>
Cc: Jakub Kicinski <kuba@...nel.org>, network dev <netdev@...r.kernel.org>,
 Linux Security Module list <linux-security-module@...r.kernel.org>,
 SElinux list <selinux@...r.kernel.org>
Subject: Re: Possible mistake in commit 3ca459eaba1b ("tun: fix group
 permission check")

27.01.2025 12:10, Ondrej Mosnacek пишет:
> Hello,
>
> It looks like the commit in $SUBJ may have introduced an unintended
> change in behavior. According to the commit message, the intent was to
> require just one of {user, group} to match instead of both, which
> sounds reasonable, but the commit also changes the behavior for when
> neither of tun->owner and tun->group is set. Before the commit the
> access was always allowed, while after the commit CAP_NET_ADMIN is
> required in this case.
>
> I'm asking because the tun_tap subtest of selinux-testuite [1] started
> to fail after this commit (it assumed CAP_NET_ADMIN was not needed),
> so I'm trying to figure out if we need to change the test or if it
> needs to be fixed in the kernel.
>
> Thanks,
>
> [1] https://github.com/SELinuxProject/selinux-testsuite/
>
Hi, IMHO having the persistent
TAP device inaccessible by anyone
but the CAP_NET_ADMIN is rather
useless, so the compatibility should
be restored on the kernel side.
I'd raise the questions about adding
the CAP_NET_ADMIN checks into
TUNSETOWNER and/or TUNSETPERSIST,
but this particular change to TUNSETIFF,
at least on my side, was unintentional.

Sorry about that. :(