Re: [PATCH nf 1/2] netfilter: nft_set_pipapo: split gc in unlink and reclaim phase

[Florian Westphal <fw@strlen.de>]

Florian Westphal <fw@strlen.de> wrote:
> index eaab422aa56a..4a5baebabaa5 100644
> --- a/net/netfilter/nft_set_pipapo.h
> +++ b/net/netfilter/nft_set_pipapo.h
> @@ -156,12 +156,14 @@ struct nft_pipapo_match {
>   * @clone:	Copy where pending insertions and deletions are kept
>   * @width:	Total bytes to be matched for one packet, including padding
>   * @last_gc:	Timestamp of last garbage collection run, jiffies
> + * @to_free:	single-linked list of elements to queue up for memory reclaim
>   */
>  struct nft_pipapo {
>  	struct nft_pipapo_match __rcu *match;
>  	struct nft_pipapo_match *clone;
>  	int width;
>  	unsigned long last_gc;
> +	struct nft_pipapo_elem *to_free;
>  };
>  
>  struct nft_pipapo_elem;
> @@ -169,10 +171,12 @@ struct nft_pipapo_elem;
>  /**
>   * struct nft_pipapo_elem - API-facing representation of single set element
>   * @priv:	element placeholder
> + * @to_free:	list of elements waiting for mem reclaim
>   * @ext:	nftables API extensions
>   */
>  struct nft_pipapo_elem {
>  	struct nft_elem_priv	priv;
> +	struct nft_pipapo_elem  *to_free;
>  	struct nft_set_ext	ext;

Pablo points out that we don't need this extra member.

Instead we'll chain nft_trans_gc containers via their list_head
members and just defer posting the gc containers to the gc engine.

So overall idea remains the same.