Re: [PATCH] sunrpc/cache: drop reference when sunrpc_cache_pipe_upcall() detects a race

["J. Bruce Fields" <bfields@fieldses.org>]

On Wed, Mar 16, 2016 at 08:31:03AM +0100, William Dauchy wrote:
> Hi Bruce,
> 
> Thank you for your reply.
> 
> On Tue, Mar 15, 2016 at 9:37 PM, NeilBrown <neilb@suse.com> wrote:
> > I agree that it isn't needed for stable - the race is tiny and the
> > consequence of losing the race is that entries get stuck in the cache
> > and possible an exported filesystem cannot be unmounted.
> 
> I believe in an industrial usage, this race will become more common;

Why is that?

> moreover the consequence of an exported filesystem which can't be
> unmounted is not something I can accept in some environment.
> 
> I accept your judgment about including it or not in -stable but I am
> just complaining about the lack of further consideration for those
> type of fixes in nfs generally speaking. In an industrial usage, we
> are hitting races way more easily than in a normal usage; lost of them
> are already fixed and we do backports ourself.

Do you have a list?

> We are at a point where
> it becomes almost impossible to come with a proof and say "we are
> hitting this race, and this patch fixes the issue, please backport in
> -stable".

I don't expect *proof*, necessarily, but some sort of argument would be
helpful.

--b.