netdev - Re: [PATCH net] bpf: expose netns inode to bpf programs

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 for Android: free password hash cracker in your pocket

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

Message-ID: <CALCETrUhE_Ga4PeobPFvhtLPVGKskTUaXe9H0v=kBXe4ii7dLw@mail.gmail.com>
Date:   Sat, 4 Feb 2017 19:27:01 -0800
From:   Andy Lutomirski <luto@...capital.net>
To:     Alexei Starovoitov <alexei.starovoitov@...il.com>
Cc:     "Eric W. Biederman" <ebiederm@...ssion.com>,
        Alexei Starovoitov <ast@...com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        "David S . Miller" <davem@...emloft.net>,
        Daniel Borkmann <daniel@...earbox.net>,
        David Ahern <dsa@...ulusnetworks.com>,
        Tejun Heo <tj@...nel.org>, Thomas Graf <tgraf@...g.ch>,
        Network Development <netdev@...r.kernel.org>
Subject: Re: [PATCH net] bpf: expose netns inode to bpf programs

On Sat, Feb 4, 2017 at 7:10 PM, Alexei Starovoitov
<alexei.starovoitov@...il.com> wrote:
> On Sat, Feb 04, 2017 at 09:07:19AM -0800, Andy Lutomirski wrote:
>> >> can see a namespaced view of the world.  For this to work, presumably
>> >> we need to make sure that eBPF programs that are installed by programs
>> >> that are in a container don't see traffic that isn't in that
>> >> container.
>> >
>> > such approach will break existing users.
>>
>> What existing users?  The whole feature has never been in a released kernel.
>
> the v3 patch commit log explains the use cases that work across netns.
> And they would break if netns boundary is suddenly started to be
> enforced for networking program types.

This is why these issues should be addressed carefully before this
feature is stabilized in 4.10, and I'm increasingly unconvinced that
this will happen.  There may be a single week left, and the cgroup
override issue is still entirely unaddressed, for example.

Also, the v3 commit log's example seems a bit weak, since seccomp can
handle that case (block non-init-netns raw sockets) and it's not clear
to me that a filter like that serves a purpose regardless besides
kernel attack surface reduction.

> If you're suggesting to limit root netns for cgroup_sock program type only
> then David explained weeks ago why it's not acceptable for vrf use
> case which was the main reason to add that program type in the first place.
> Also it would make one specific bpf program type to be different
> from all others which will be confusing.
>

I'm suggesting limiting it to the init netns for *all* cases.
(Specifically, limiting installing programs to the init netns.  But
perhaps limiting running of programs to the netns in which they're
installed would be the best solution after all.)  If there isn't a
well-understood solution that plays well with netns and that satisfies
ip vrf, the it's worth seriously thinking about whether ip vrf is
really ready for 4.10.

--Andy