Re: [PATCH] fix race in mark_mounts_for_expiry()

[David Howells <dhowells@redhat.com>]

Miklos Szeredi <miklos@szeredi.hu> wrote:

> 
> There's still the problem of cmpxchg meddling in the internals of an
> atomic_t.  Is that OK?  Will that work on all archs?

Probably. It might be worth defining an atomic_cmpxchg() op to formalise it
though:

	int atomic_cmpxchg(atomic_t *p, int old, int new);

The main reason for this is that atomic_t is mostly a 32-bit value, I think,
and cmpxchg() may sometimes enforce a 64-bit value.

It can be made to work on PPC, PPC64, x86, x86_64, frv. I imagine it'll work
on MIPS, sparc and alpha too without too much trouble. Dunno about the rest.

David