Re: Adding landlock support to Suricata

Mickaël Salaün <[email protected]> Thu, 4 Aug 2022 18:37:21 +0200
Newsgroups dev.linux.lists.landlock
Message-ID <[email protected]>
On 03/08/2022 15:29, Eric Leblond wrote:
> Hello,

Hi!

> 
> I worked on implementing support for landlock in Suricata. Suricata is
> an open source network threat
> detection engine that reads packets from the network or files and
> outputs analysis in other files.
> As such it is a good candidate for Landlock sandboxing.

Indeed!

> 
> The MR on Suricata is here and feedbacks and reviews are welcome:
> https://github.com/OISF/suricata/pull/7688
> 
> With regards to the usage of Landlock API, it was overall really fine
> and it did not took me long to obtain something working.
> I did although struggle with a few things (yes I'm a newbie here):
> - I looked at first if I had to add a library to the build system
> before realizing that up to date headers were enough

There is no C library. I don't plan to write one but if someone wants to 
get on that boat I'd be please to help. We're focusing on a Go and Rust 
library (WIP) for now.


> - The examples and documentation I found don't cover multiple calls to
> setup a policy and it got me confused a bit:
>    - I had to test to verify that the logic is create ruleset/add
> multiple rules/enforce ruleset

Hmm, the doc example [1] only explains how to add one rule indeed. The 
sandboxer example [2] should help though. I'll update the doc to deal 
with an array of paths.

[1] https://docs.kernel.org/userspace-api/landlock.html
[2] 
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/samples/landlock/sandboxer.c

> - I did struggle a bit on the ruleset scope flags (handled_access_fs)
> as it was not clear to me this was a choice of the sandbox software
> implementation

Was it about the fact that we need to populate handled_access_fs or that 
we need to mask some bits?

Which modification could have helped you here? 
https://docs.kernel.org/userspace-api/landlock.html#defining-and-enforcing-a-security-policy

> 
> It is just small things that are easy to fix in documentation, for the
> rest it was really easy to implement (if you know the software you are
> adding support to).
> 
> Thanks a lot for Landlock, this is a great feature that is easy to implement.

Thanks for the feedback, it's valuable to improve documentation and help 
others.

> 
> Best regards,