On Fri, Jan 17, 2020 at 05:08:14PM -0800, Andrei Vagin wrote:
> On Wed, Jan 15, 2020 at 9:18 AM Christian Brauner
> <christian.brauner@xxxxxxxxxx> wrote:
> >
> > Commit 69f594a38967 ("ptrace: do not audit capability check when outputing /proc/pid/stat")
> > introduced the ability to opt out of audit messages for accesses to
> > various proc files since they are not violations of policy.
> > While doing so it somehow switched the check from ns_capable() to
> > has_ns_capability{_noaudit}(). That means it switched from checking the
> > subjective credentials of the task to using the objective credentials. I
> > couldn't find the original lkml thread and so I don't know why this switch
> > was done. But it seems wrong since ptrace_has_cap() is currently only used
> > in ptrace_may_access(). And it's used to check whether the calling task
> > (subject) has the CAP_SYS_PTRACE capability in the provided user namespace
> > to operate on the target task (object). According to the cred.h comments
> > this would mean the subjective credentials of the calling task need to be
> > used.
> > This switches it to use security_capable() because we only call
> > ptrace_has_cap() in ptrace_may_access() and in there we already have a
> > stable reference to the calling tasks creds under cred_guard_mutex so
> > there's no need to go through another series of dereferences and rcu
> > locking done in ns_capable{_noaudit}().
>
>
> The criu process is started with all capabilities in the root user namespace.
>
> I don't have time to investigate this issue right now, will provide
> more details next Tuesday.
Yeah, we've detected the issue. security_capable() indicates success by
returning 0 for whatever reason whereas has_ns_capability() returns 1.
So the logic was inverted. This is fixed in the new version. Sorry for
the noise!
Thanks!
Christian
