Re: Thoughts on credential switching

[Andy Lutomirski <luto@amacapital.net>]

On Thu, Mar 27, 2014 at 11:56 AM, Jeremy Allison <jra@samba.org> wrote:
> On Thu, Mar 27, 2014 at 11:46:39AM -0700, Andy Lutomirski wrote:
>> On Thu, Mar 27, 2014 at 11:26 AM, Jeremy Allison <jra@samba.org> wrote:
>> >
>> > Amen to that :-).
>> >
>> > However, after talking with Jeff and Jim at CollabSummit,
>> > I was 'encouraged' to make my opinions known on the list.
>> >
>> > To me, calling the creds handle a file descriptor just
>> > feels wrong. IT *isn't* an fd, you can't read/write/poll
>> > on it, and it's only done as a convenience to get the
>> > close-on-exec semantics and the fact that the creds are
>> > already hung off the fd's in kernel space.
>>
>> Windows calls these things "handles."  Linux has "file descriptors,"
>> and there's plenty of precedent for things that aren't files.
>
> Sure, but there's a set of expectations around
> fd's that these things don't satisfy - IO-ops.

eventfd, timerfd, and signalfd barely satisfy those.  namespace fds
don't satisfy those expectations at all.  And /proc/pid/fd is really
quite useful for debugging.

>
>> > That way we can also make it clear this thing only has
>> > meaning to a thread group, and SHOULD NOT (and indeed
>> > preferably CAN NOT) be passed between processes.
>> >
>>
>> If you want those semantics, then stick a struct pid * in there for
>> the tgid of the cretor and make sure that current's tgid matches when
>> you try to use it.
>>
>> I think they'd be more useful without that check, though.
>
> I'm more worried about leakage and unintended consequences
> here.
>
>> BTW, what do you want to have happen on fork?  I think they should keep working.
>
> Yeah, that's true. I want them to keep
> working across fork, but not across exec
> or any other method of fd-passing.
>

This seems like an unfortunate restriction to put in the kernel to
prevent userspace from shooting itself in the foot.

--Andy