Re: [PATCH nf] netfilter: nf_tables: clone set on flush only

[Florian Westphal <fw@strlen.de>]

Pablo Neira Ayuso <pablo@netfilter.org> wrote:
> Restrict set clone to the flush set command in the preparation phase.
> Add NFT_ITER_UPDATE_CLONE and use it for this purpose, update the rbtree
> and pipapo backends to only clone the set when this iteration type is
> used.
> 
> As for the existing NFT_ITER_UPDATE type, update the pipapo backend to
> use the existing set clone if available, otherwise use the existing set
> representation. After this update, there is no need to clone a set that
> is being deleted, this includes bound anonymous set.

Thanks Pablo, this looks good to me. Two small nits/suggestions:

1. Add to commit message that this affects fault injection resp.
   that this requires failing GFP_KERNEL allocation to trigger
   the WARN splat.

>  struct nft_set;
> diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
> index 834736237b09..618f79700f75 100644
> --- a/net/netfilter/nf_tables_api.c
> +++ b/net/netfilter/nf_tables_api.c
> @@ -7906,7 +7906,7 @@ static int nft_set_flush(struct nft_ctx *ctx, struct nft_set *set, u8 genmask)
>  {
>  	struct nft_set_iter iter = {
>  		.genmask	= genmask,
> -		.type		= NFT_ITER_UPDATE,
> +		.type		= NFT_ITER_UPDATE_CLONE,
>  		.fn		= nft_setelem_flush,
>  	};

2. I think it would help to add a comment to the existing
NFT_ITER_UPDATE users as to why they use UPDATE and not CLONE.

Or, alternatively, add comment here why this one needs the _CLONE
variant.

But this looks correct to me, nft_set_flush() makes clone mandatory
while others either can re-use the live copy OR are guaranteed to have
a clone, e.g. because the function is always called after a
delete/insert operation that already did the clone.