Re: [PATCH] netfilter: nft_dynset: support for element deletion

From: Ander Juaristi <a@juaristi.eus>
To: Florian Westphal <fw@strlen.de>
Cc: netfilter-devel@vger.kernel.org
Subject: Re: [PATCH] netfilter: nft_dynset: support for element deletion
Date: Mon, 15 Jul 2019 18:22:14 +0200	[thread overview]
Message-ID: <3db3e09d-5e1e-5d12-0f96-f911eb40c769@juaristi.eus> (raw)
In-Reply-To: <20190713165940.gyqrhab4z3eookgl@breakpoint.cc>

[-- Attachment #1: Type: text/plain, Size: 874 bytes --]

On 13/7/19 18:59, Florian Westphal wrote:
> 
>> +	if (he == NULL)
>> +		return false;
>> +
>> +	rhashtable_remove_fast(&priv->ht, &he->node, nft_rhash_params);
>> +	return true;
> 
> Perhaps add a small comment here that rhashtable_remove_fast retval
> is ignored intentionally?
> 
> I.e., don't make this return false in case two cpus race to remove same
> entry.

Hmm, this made me think. I don't know if this was all too intentional
from me.

Maybe rather than ignoring it, it would be better to return true only if
rhashtable_remove_fast returned 0, which will only happen if the element
was actually deleted (locking is done internally so two cpus cannot race
in there). Else, if return value is -ENOENT, we should return false.

And taking this reasoning further, maybe the initial call to
rhashtable_lookup wouldn't be needed either?

WDYT?

[-- Attachment #2: pEpkey.asc --]
[-- Type: application/pgp-keys, Size: 6263 bytes --]