Re: [PATCH v2] ipvs: Fix reuse connection if RS weight is 0

[Julian Anastasov]

	Hello,

On Fri, 29 Oct 2021, yangxingwu wrote:

> Since commit dc7b3eb900aa ("ipvs: Fix reuse connection if real server is
> dead"), new connections to dead servers are redistributed immediately to
> new servers.
> 
> Then commit d752c3645717 ("ipvs: allow rescheduling of new connections when
> port reuse is detected") disable expire_nodest_conn if conn_reuse_mode is
> 0. And new connection may be distributed to a real server with weight 0.

	Can you better explain in commit message that we are changing 
expire_nodest_conn to work even for reused connections when
conn_reuse_mode=0 but without affecting the controlled/persistent
connections during the grace period while server is with weight=0.

	Even if you target -next trees adding commit d752c3645717
as Fixes line would be a good idea. Make sure the tree is specified
after the v3 tag.

> Co-developed-by: Chuanqi Liu <legend050709@xxxxxx>
> Signed-off-by: Chuanqi Liu <legend050709@xxxxxx>
> Signed-off-by: yangxingwu <xingwu.yang@xxxxxxxxx>
> ---
>  Documentation/networking/ipvs-sysctl.rst | 3 +--
>  net/netfilter/ipvs/ip_vs_core.c          | 7 ++++---
>  2 files changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/Documentation/networking/ipvs-sysctl.rst b/Documentation/networking/ipvs-sysctl.rst
> index 2afccc63856e..1cfbf1add2fc 100644
> --- a/Documentation/networking/ipvs-sysctl.rst
> +++ b/Documentation/networking/ipvs-sysctl.rst
> @@ -37,8 +37,7 @@ conn_reuse_mode - INTEGER
>  
>  	0: disable any special handling on port reuse. The new
>  	connection will be delivered to the same real server that was
> -	servicing the previous connection. This will effectively
> -	disable expire_nodest_conn.
> +	servicing the previous connection.
>  
>  	bit 1: enable rescheduling of new connections when it is safe.
>  	That is, whenever expire_nodest_conn and for TCP sockets, when
> diff --git a/net/netfilter/ipvs/ip_vs_core.c b/net/netfilter/ipvs/ip_vs_core.c
> index 128690c512df..374f4b0b7080 100644
> --- a/net/netfilter/ipvs/ip_vs_core.c
> +++ b/net/netfilter/ipvs/ip_vs_core.c
> @@ -2042,14 +2042,15 @@ ip_vs_in(struct netns_ipvs *ipvs, unsigned int hooknum, struct sk_buff *skb, int
>  			     ipvs, af, skb, &iph);
>  
>  	conn_reuse_mode = sysctl_conn_reuse_mode(ipvs);
> -	if (conn_reuse_mode && !iph.fragoffs && is_new_conn(skb, &iph) && cp) {
> +	if (!iph.fragoffs && is_new_conn(skb, &iph) && cp) {

	It is even better to move the !cp->control check above:

	if (!iph.fragoffs && is_new_conn(skb, &iph) && cp && !cp->control) {

	Then is not needed in is_new_conn_expected() anymore.

>  		bool old_ct = false, resched = false;

	And now you can move conn_reuse_mode here:

		int conn_reuse_mode = sysctl_conn_reuse_mode(ipvs);

>  		if (unlikely(sysctl_expire_nodest_conn(ipvs)) && cp->dest &&
> -		    unlikely(!atomic_read(&cp->dest->weight))) {
> +		    unlikely(!atomic_read(&cp->dest->weight)) && !cp->control) {
>  			resched = true;
>  			old_ct = ip_vs_conn_uses_old_conntrack(cp, skb);
> -		} else if (is_new_conn_expected(cp, conn_reuse_mode)) {
> +		} else if (conn_reuse_mode &&
> +			   is_new_conn_expected(cp, conn_reuse_mode)) {
>  			old_ct = ip_vs_conn_uses_old_conntrack(cp, skb);
>  			if (!atomic_read(&cp->n_control)) {
>  				resched = true;
> -- 
> 2.30.2

Regards

--
Julian Anastasov <ja@xxxxxx>