On Tue, Dec 8, 2015 at 1:39 AM, Herbert Xu <herb...@gondor.apana.org.au> wrote:
> David Miller <da...@davemloft.net> wrote:
>> From: Tom Herbert <t...@herbertland.com>
>> Date: Tue, 1 Dec 2015 15:11:09 -0800
>>
>>> + lock = rht_bucket_lock(tbl, hash);
>>> +
>>> + spin_lock_bh(lock);
>>> +
>>> + pprev = &tbl->buckets[hash];
>>> + rht_for_each(he, tbl, hash) {
>>> + if (he != obj_old) {
>>> + pprev = &he->next;
>>> + continue;
>>> + }
>>> +
>>> + rcu_assign_pointer(obj_new->next, obj_old->next);
>>> + rcu_assign_pointer(*pprev, obj_new);
>>> + err = 0;
>>> + break;
>>
>> Are you sure this works fine in the presence of both parallel readers and
>> table expansion passes?
>
> Good question.
>
> What's more this is something that can be easily implemented
> outside of rhashtable, i.e., by hashing a pointer to the actual
> object rather than the object itself. So I'd like to see some
> pretty good reasons for penny-pinching on memory and adding more
> complexity to rhashtable.
That creates one more level of indirection. I don't see how add an
atomic replace operation adds any complexity to the rhashtable, none
of the semantics for rhashtable need to be changed.
>
> Cheers,
> --
> Email: Herbert Xu <herb...@gondor.apana.org.au>
> Home Page: http://gondor.apana.org.au/~herbert/
> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
