On Mon, Jan 13, 2014 at 6:56 PM, Eric Paris <eparis(a)redhat.com> wrote:
We have a helper function which writes out all of the interesting
identity information about tasks, audit_log_task_info(). We then have a
second helper, audit_log_task(), which is only used by audit_core_dumps()
and __audit_seccomp(). It is a light weight and only outputs some of the
information about the task. There does not appear to be rational for
its existence except audit_core_dumps() originally did it this way. At
the time audit_log_task_info() did not exist. When __audit_seccomp came
along audit_core_dumps() was split into this helper and reused. But
there was a better helper in audit.c.
This does reorder the records for audit_core_dumps() and
__audit_seccomp(). The new record order is below. The number in () is
the order in the old record. Entries without a () do not exist in the
old record.
audit_log_task_info:
ppid pid (6) auid (1) uid (2) gid (3) euid
suid fsuid egid sgid fsgid tty
ses (4) comm (7) exe (8) subj (5)
audit_log_task:
auid uid gid ses subj pid comm exe
It seems that reusing the task info pattern throughout records should
allow for faster simpler more streamlined userspace records parsing, but
changing order like this might be a deal breaker.
Signed-off-by: Eric Paris <eparis(a)redhat.com>
Sounds fine to me. Thanks!
Acked-by: Kees Cook <keescook(a)chromium.org>
-Kees
--
Kees Cook
Chrome OS Security