Re: [PATCH nf-next v4 01/13] rhashtable: add rhashtable_flush_and_free helper

[Florian Westphal <fw@strlen.de>]

Herbert Xu <herbert@gondor.apana.org.au> wrote:
> > No, remove all elements. Concurrent insertion is not disabled.
> 
> So what's the desired semantics for insertions that occur while
> you're flushing the table?

In current ipset home-grown hashes:

If the insertion is into a region not yet scanned -> will be dropped
If the insertion is into a region already scanned -> will exist after flush.

i.e. no guarantee that table is empty after flush, but all
elements that existed before the flush are gone.

For the rhashtable variant I had proposed:

If the insertion is before the moment future_tbl is updated: will be dropped
If the insertion is after future_tbl update: will exist after flush

which I think is matches existing semantics of ipsets homegrown hash
tables (that I'd like to replace with rhashtable/rhltable).

Thanks,
Florian