[v2] ipvs: drop first packet to dead server

Hello,

On Fri, 25 Sep 2015, Jiri Bohac wrote:

>  		if (!atomic_read(&cp->n_control))
>  			ip_vs_conn_expire_now(cp);
>  		__ip_vs_conn_put(cp);
> -		cp = NULL;
> +		return NF_DROP;

	So, at this point we do not know whether we have
one or many real servers, with same or different forwarding
method. For example, if we know that old real server is DR
and the new real server is again DR we can reuse the conntrack.

	Without such info we have to drop the connection
_only_ when conntrack is used. But if the connection is referenced
by others, eg. FTP:21 referenced by data connection, it is not
going to expire soon. We have to reuse it, otherwise client
will retransmit for long time. See below...

	Also, according to RFC 5961, 4.2. Mitigation, client can
send RST during its SYN retransmissions. So, may be we have to
allow rescheduling by adding IP_VS_TCP_S_CLOSE check after the
IP_VS_TCP_S_TIME_WAIT check in is_new_conn_expected(). This should
allow restarted client to be rescheduled. But this can be in
separate patch.

	Here is a RFC patch (not for latest tree, not tested):


Regards

--
Julian Anastasov <ja@ssi.bg>
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hi,

On Sun, Sep 27, 2015 at 08:25:18PM +0300, Julian Anastasov wrote:
> On Fri, 25 Sep 2015, Jiri Bohac wrote:
> 
> >  		if (!atomic_read(&cp->n_control))
> >  			ip_vs_conn_expire_now(cp);
> >  		__ip_vs_conn_put(cp);
> > -		cp = NULL;
> > +		return NF_DROP;
> 
> 	So, at this point we do not know whether we have
> one or many real servers, with same or different forwarding
> method. For example, if we know that old real server is DR
> and the new real server is again DR we can reuse the conntrack.
> 
> 	Without such info we have to drop the connection
> _only_ when conntrack is used.

right, good point!

> +static inline bool ip_vs_conn_uses_conntrack(struct ip_vs_conn *cp,
> +					struct sk_buff *skb)
> +{
> +#ifdef CONFIG_IP_VS_NFCT
> +	enum ip_conntrack_info ctinfo;
> +	struct nf_conn *ct;
> +
> +	if (!(cp->flags & IP_VS_CONN_F_NFCT))
> +		return false;
> +	ct = nf_ct_get(skb, &ctinfo);
> +	if (ct && !nf_ct_is_untracked(ct))
> +		return true;
> +#endif
> +	return false;
> +}
> +

I tested this part; we found the problem on an old (3.12) kernel,
we're missing the parts dealing with rescheduling on port reuse -
only dealing with the "weight == 0" case.

> +	if (conn_reuse_mode && !iph.fragoffs && is_new_conn(skb, &iph) && cp) {
> +		bool uses_ct = false, resched = false;
> +
> +		if (unlikely(sysctl_expire_nodest_conn(ipvs)) && cp->dest &&
> +		    unlikely(!atomic_read(&cp->dest->weight))) {
> +			resched = true;
> +			uses_ct = ip_vs_conn_uses_conntrack(cp, skb);
> +		} else if (is_new_conn_expected(cp, conn_reuse_mode)) {
> +			uses_ct = ip_vs_conn_uses_conntrack(cp, skb);
> +			if (!atomic_read(&cp->n_control)) {
> +				resched = true;
> +			} else {
> +				/* Do not reschedule controlling connection
> +				 * that uses conntrack while it is still
> +				 * referenced by controlled connection(s).
> +				 */
> +				resched = !uses_ct;
> +			}
> +		}
> +
> +		if (resched) {
> +			if (!atomic_read(&cp->n_control))
> +				ip_vs_conn_expire_now(cp);
> +			__ip_vs_conn_put(cp);
> +			if (uses_ct)
> +				return NF_DROP;
> +			cp = NULL;
> +		}

Looks good, but I can't easily test this.

Thanks,