Re: [PATCH 10/13] ipvs: Rename kfree_rcu() to kfree_rcu_mightsleep()

[Uladzislau Rezki <urezki@gmail.com>]

On Wed, Feb 01, 2023 at 06:12:04PM +0200, Julian Anastasov wrote:
> 
> 	Hello,
> 
> On Wed, 1 Feb 2023, Pablo Neira Ayuso wrote:
> 
> > Hi,
> > 
> > On Wed, Feb 01, 2023 at 04:09:51PM +0100, Uladzislau Rezki (Sony) wrote:
> > > The kfree_rcu()'s single argument name is deprecated therefore
> > > rename it to kfree_rcu_mightsleep() variant. The goal is explicitly
> > > underline that it is for sleepable contexts.
> > > 
> > > Cc: Julian Anastasov <ja@ssi.bg>
> > > Cc: Pablo Neira Ayuso <pablo@netfilter.org>
> > > Cc: Jiri Wiesner <jwiesner@suse.de>
> > > Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
> > > ---
> > >  net/netfilter/ipvs/ip_vs_est.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/net/netfilter/ipvs/ip_vs_est.c b/net/netfilter/ipvs/ip_vs_est.c
> > > index ce2a1549b304..a39baf6d1367 100644
> > > --- a/net/netfilter/ipvs/ip_vs_est.c
> > > +++ b/net/netfilter/ipvs/ip_vs_est.c
> > > @@ -549,7 +549,7 @@ void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats)
> > >  	__set_bit(row, kd->avail);
> > >  	if (!kd->tick_len[row]) {
> > >  		RCU_INIT_POINTER(kd->ticks[row], NULL);
> > > -		kfree_rcu(td);
> > 
> > I also found this kfree_rcu() without rcu_head call a few weeks ago.
> > 
> > @Wiesner, @Julian: Any chance this can be turned into kfree_rcu(td, rcu_head); ?
> 
> 	Yes, as simple as this:
> 
> diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h
> index c6c61100d244..6d71a5ff52df 100644
> --- a/include/net/ip_vs.h
> +++ b/include/net/ip_vs.h
> @@ -461,6 +461,7 @@ void ip_vs_stats_free(struct ip_vs_stats *stats);
>  
>  /* Multiple chains processed in same tick */
>  struct ip_vs_est_tick_data {
> +	struct rcu_head		rcu_head;
>  	struct hlist_head	chains[IPVS_EST_TICK_CHAINS];
>  	DECLARE_BITMAP(present, IPVS_EST_TICK_CHAINS);
>  	DECLARE_BITMAP(full, IPVS_EST_TICK_CHAINS);
> diff --git a/net/netfilter/ipvs/ip_vs_est.c b/net/netfilter/ipvs/ip_vs_est.c
> index df56073bb282..25c7118d9348 100644
> --- a/net/netfilter/ipvs/ip_vs_est.c
> +++ b/net/netfilter/ipvs/ip_vs_est.c
> @@ -549,7 +549,7 @@ void ip_vs_stop_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats)
>  	__set_bit(row, kd->avail);
>  	if (!kd->tick_len[row]) {
>  		RCU_INIT_POINTER(kd->ticks[row], NULL);
> -		kfree_rcu(td);
> +		kfree_rcu(td, rcu_head);
>  	}
>  	kd->est_count--;
>  	if (kd->est_count) {
> 
> 	I was about to reply to Uladzislau Rezki but his patchset
> looks more like a renaming, so I'm not sure how we are about
> to integrate this change, as separate patch or as part of his
> patchset. I don't have preference, just let me know how to
> handle it.
> 
If you give on this patch an Ack i will resend it once i collect
all other ACKs. So it will be integrated via RCU-tree.

So this is just renaming.

If the ip_vs_stop_estimator() can be invoked from an atomic context
then it is a bug and you need to integrate rcu_head in your structure.

--
Uladzislau Rezki