Re: Q: check_unsafe_exec() races (Was: [PATCH 2/4] fix setuid sometimes doesn't)

[Hugh Dickins <hugh@veritas.com>]

On Wed, 1 Apr 2009, Al Viro wrote:
> On Wed, Apr 01, 2009 at 01:28:01AM +0100, Hugh Dickins wrote:
>  
> > Otherwise it looks good to me, except I keep worrying about those
> > EAGAINs.  The more so once I noticed current->cred_exec_mutex is
> > already being used to handle a similar issue with ptrace.  What
> > do you think of this rather smaller patch?  which I'd much rather
> > send after having slept on it, since it may be embarrassingly and
> > obviously wrong, but tomorrow may be too late ...
>  
> Eh...  I'm not particulary happy with fork() growing heavier and heavier.

I don't see it as making fork() any heavier, but never mind.
The important thing is to get a fix out.

> Besides, there's a subtle problem avoided by another variant - think what
> happens if past the point of no return execve() will unshare fs_struct
> (e.g. by explicit unshare() from dynamic linker).

You're too far ahead of me there.

> 
> Frankly, -EAGAIN in situation when we have userland race is fine.  And
> we *do* have a userland race here - execve() will kill -9 those threads
> in case of success, so if they'd been doing something useful, they are
> about to be suddenly screwed.

Good point.  I found it quite odd the way the awkward case (shared
beyond the threadgroup) is allowed to go forward (with possibility
that setuid will be undone), but the easy case is -EAGAINed.  (And
I gave up on trying to find a better name for your "in_exec" flag,
which is rather more subtle than just that!)  But odd as it is,
there's good reason for doing it that way.

> 
> So I stand by my variant.

Fair enough.

> Note that if we have *other* tasks sharing
> fs_struct, your variant will block their clone() for the duration of
> execve() while mine will simply leave them alone (and accept that we
> have unsafe sharing).

Yes, intentional, consistent with the existing cred_exec_mutex technique.

Hugh