On Thu, Aug 10, 2017 at 9:33 PM, Tyler Hicks <tyhicks(a)canonical.com> wrote:
This patch set aims to improve logging in seccomp by:
1) Empowering administrators to be able to permit or quiet logging of
specific seccomp actions
2) Allowing applications to request logging of all actions, except for
RET_ALLOW, in the filter being loaded (subject to the
administrator's wishes in #1)
3) By making it possible for application developers to request logging
of specific syscalls while developing filters for their application
(subject to the administrator's wishes in #1)
With this patch set applied, the logic for deciding if an action will be
logged is as described in the commit message of the final patch.
* Changes since v5:
- Rebase onto
https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/?h=seccomp...
(kees)
+ Redefine the value of SECCOMP_FILTER_FLAG_LOG to account for the
new SECCOMP_FILTER_FLAG_KILL_PROCESS
+ Add SECCOMP_FILTER_FLAG_KILL_PROCESS to the
detect_seccomp_filter_flags selftest
- Reorder patches to move SECCOMP_GET_ACTION_AVAIL patch behind
actions_avail sysctl patch (kees)
- Reorder patches to move the selftest to detect filter flag support
before the patch that adds SECCOMP_FILTER_FLAG_LOG (kees)
- Add psuedo code showing the high level logic of when and when not to
log to the commit message of each patch that changes the logging
behavior (inspired by kees)
- Add Suggested-by to the SECCOMP_GET_ACTION_AVAIL patch to credit
Andy for the idea (tyhicks)
- Use sizeof(seccomp_actions_avail), instead of strlen(), to avoid
variable length "names" array in seccomp_actions_logged_handler()
(smatch)
- Only check the actions_logged sysctl value for "kill" when first
introducing the actions_logged sysctl since filters cannot yet set
the FILTER_FLAG_LOG flag (kees)
- Mention how the actions_logged sysctl could quiet SECCOMP_RET_LOG
actions in seccomp_filter.rst documentation (kees)
Thanks! This looks good; I made some minor comment style adjustments,
and changed the rebase slightly (as I continue to agonize over how to
do KILL_PROCESS best...)
I'll get this sent to James for -next.
-Kees
--
Kees Cook
Pixel Security