On Mon, May 1, 2017 at 7:41 PM, Tyler Hicks <tyhicks(a)canonical.com> wrote:
On 04/27/2017 07:42 PM, Kees Cook wrote:
> On Thu, Apr 27, 2017 at 3:17 PM, Tyler Hicks <tyhicks(a)canonical.com> wrote:
>> Quick update... I finished the move from the high-water mark
>> log_max_action sysctl to the bitmask based actions_logged sysctl.
>
> Awesome!
>
>> Unfortunately, I've just realized that SECCOMP_SET_LOGGING, or any
>> process-wide logging configuration mechanism, will not work. It is fine
>> for the situation where two unrelated processes set up seccomp filters
>> that should be logged differently. However, it fails when two closely
>> related processes, such as parent and child, need to set up seccomp
>> filters that should be logged differently. Imagine a launcher that sets
>> up an application sandbox (including a seccomp filter) and then launches
>> an electron app which will have its own seccomp filter for sandboxing
>> untrusted code that it runs. Unless the launcher and app completely
>> agree on actions that should be logged, the logging won't work as
>> intended for both processes.
>
> Oh, you mean the forked process sets up the logging it wants for the
> filters it just installed, then after exec a process sets up new
> logging requirements?
Yes - see below.
>
>> I think this needs to be configured at the filter level.
>
> I'm not sure that's even the right way to compose the logging desires.
>
> So, my initial thought was "whatever ran SECCOMP_SET_LOGGING knows
> what it's doing" and it should be the actual value.
>
> If the launcher wants logs of everything the application does with its
> filters, then a purely-tied-to-filter approach won't work either.
>
> Perhaps log bits can only be enabled? I.e. SECCOMP_SET_LOGGING
> performs an OR instead of an assignment?
The problem that I'm envisioning with this design is this:
1. Launcher is told to launch Chrome and forks off a process.
2. Launcher sets up a filter using RET_ERRNO for all unacceptable
syscalls and enables auditing of RET_ERRNO.
3. Launcher execs Chrome.
4. Chrome then sets up its own, more restrictive filter that uses
RET_ERRNO, among other actions, but does not want auditing of RET_ERRNO.
If we use process-wide auditing controls, the logs will be filled with
RET_ERRNO messages that were unintended and unrelated to the RET_ERRNO
actions set up in the launcher's filter.
Unfortunately, the OR'ing idea doesn't solve the problem.
Things like my more complicated solution solve this completely, I
think. The launcher would, by whatever means, say "RET_ERRNO and log
this". The more restrictive sandbox would say "RET_ERROR and don't
log this" and we'd just make sure that the composition rules mean the
inner rule wins.
--Andy