Re: [PATCH 1/1] exec: do not sleep in TASK_TRACED under ->cred_guard_mutex

[David Howells <dhowells@redhat.com>]

Roland McGrath <roland@redhat.com> wrote:

> I certainly think it's right to hold the mutex only as long as necessary.
> Clearly holding it when we stop is wrong.

As far as I can tell, we don't hold it when we stop for debugging, or stop on
signal.

> I'm a bit concerned about holding it for arbitrary periods while we block
> in the filesystem code.  e.g., consider the scenario with a hangs-forever
> NFS server or suchlike.  But I'm not sure there is a reasonable way around
> that one.

If you drop the sem before committing the creds, you have to recalculate the
new credentials.

> The paired calls that leave the mutex locked in between should have some
> clear comments calling attention to their pairing.  Aside from that making
> sure that subtlety is clear, I don't see any problems in the patch off hand.
> But I haven't scoured the code path lately to have full confidence.
> I'd like to hear David's reactions.

Looking at the patch description, I don't see how the patch it relevant to the
problem.  There must be something else, either a call that's now being
skipped, or it's a matter of timing.

David