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:
> > I don't understand this response.  This isn't about rhashtable vs.
> > rhltable.  This is about my incompetence to flush an rhashtable or
> > rhashtable.  Simple version:
> > 
> >         rhashtable_walk_enter();
> >         rhashtable_walk_start();
> 
> You should never use rhashtable_walk for real work.  It was only ever
> intended for the very limited case of netlink dumping where stability
> or accuracy was not a requirement.
> 
> >         while ((he = rhashtable_walk_next())) {
> >                 if (IS_ERR(he)) {
> >                         if (PTR_ERR(he) != -EAGAIN) { ..  break; }
> >                         continue;
> >                 }
> > 
> > 		rhashtable_remove_fast()
> 
> So you want to free the entire table, right?

No, remove all elements. Concurrent insertion is not disabled.

> We already have rhashtable_free_and_destroy, any reason why it
> doesn't work?

It requires userspace or kernel don't add new elements.

AFAICS I can't "destroy, then re-init" without some external
mutex.  What I could do is add another indirection, i.e.

 struct htype {
-       struct rhashtable ht;   /* the hash table */
+       struct rhashtable *ht;   /* the hash table */
        struct net_prefixes __rcu *rnets[IPSET_NET_COUNT]; /* cidr prefixes */
 };

... and then alloc+init an new ht + free old one.
But its not nice either.