TencentOS-kernel: ipvs: avoid drop first packet by reusing conntrack

fix #29256237

commit a01a9445c00eca3e37523eb6b0d87f494eceeb4b TencentOS-kernel

Since 'commit f719e375 ("ipvs: drop first packet to
redirect conntrack")', when a new TCP connection meet
the conditions that need reschedule, the first syn packet
is dropped, this cause one second latency for the new
connection, more discussion about this problem can easy
search from google, such as:

1)One second connection delay in masque
https://marc.info/?t=151683118100004&r=1&w=2

2)IPVS low throughput #70747
https://github.com/kubernetes/kubernetes/issues/70747

3)Apache Bench can fill up ipvs service proxy in seconds #544
https://github.com/cloudnativelabs/kube-router/issues/544

4)Additional 1s latency in `host -> service IP -> pod`
https://github.com/kubernetes/kubernetes/issues/90854

5)kube-proxy ipvs conn_reuse_mode setting causes errors
with high load from single client
https://github.com/kubernetes/kubernetes/issues/81775

The root cause is when the old session is expired, the
conntrack related to the session is dropped by
ip_vs_conn_drop_conntrack. The code is as follows:
```
static void ip_vs_conn_expire(struct timer_list *t)
{
...

     if ((cp->flags & IP_VS_CONN_F_NFCT) &&
         !(cp->flags & IP_VS_CONN_F_ONE_PACKET)) {
             /* Do not access conntracks during subsys cleanup
              * because nf_conntrack_find_get can not be used after
              * conntrack cleanup for the net.
              */
             smp_rmb();
             if (ipvs->enable)
                     ip_vs_conn_drop_conntrack(cp);
     }
...
}
```
As shown in the code, only when condition (cp->flags & IP_VS_CONN_F_NFCT)
is true, the function ip_vs_conn_drop_conntrack will be called.

So we optimize this by following steps (Administrators
can choose the following optimization by setting
net.ipv4.vs.conn_reuse_old_conntrack=1):
1) erase the IP_VS_CONN_F_NFCT flag (it is safely because
   no packets will use the old session)
2) call ip_vs_conn_expire_now to release the old session,
   then the related conntrack will not be dropped
3) then ipvs unnecessary to drop the first syn packet, it
   just continue to pass the syn packet to the next process,
   create a new ipvs session, and the new session will related
   to the old conntrack(which is reopened by conntrack as a new
   one), the next whole things is just as normal as that the old
   session isn't used to exist.

The above processing has no problems except for passive FTP,
for passive FTP situation, ipvs can judging from
condition (atomic_read(&cp->n_control)) and condition (cp->control).
So, for other conditions(means not FTP), ipvs should give users
the right to choose，they can choose a high performance one processing
logical by setting net.ipv4.vs.conn_reuse_old_conntrack=1. It is necessary
because most business scenarios (such as kubernetes) are very sensitive
to TCP short connection latency.

This patch has been verified on our thousands of kubernets
node servers on Tencent Inc.
Signed-off-by: NYangYuxi <yx.atom1@gmail.com>
[Tony: add the missing sysctl knob and disable it by default]
Signed-off-by: NTony Lu <tonylu@linux.alibaba.com>
Acked-by: NDust Li <dust.li@linux.alibaba.com>

Documentation/networking/ipvs-sysctl.txt

@@ -43,6 +43,29 @@ conn_reuse_mode - INTEGER
 	balancer in Direct Routing mode. This bit helps on adding new
 	real servers to a very busy cluster.
 
+conn_reuse_old_conntrack - BOOLEAN
+	- 0 - disabled (default)
+	- not 0 - enabled
+
+	If set, when a new TCP syn packet hits an old ipvs connection
+	table and need reschedule to a new dest: if
+	  1) the packet uses conntrack
+	  2) the old ipvs connection table is not a master control
+		 connection (E.g. the command connection of passive FTP)
+	  3) the old ipvs connection table hasn't been controlled by any
+		 connection (E.g. the data connection of passive FTP)
+	ipvs will not release the old conntrack, just let the conntrack
+	reopen the old session as it is a new one. This is an optimization
+	option selectable by the system administrator.
+
+	If not set, when a new TCP syn packet hits an old ipvs connection
+	table and need reschedule to a new dest: if
+	  1) the packet uses conntrack
+	ipvs just drops this syn packet, expires the old connection by timer.
+	This will cause the client to retransmit TCP SYN.
+
+	Only has effect when conn_reuse_mode not 0.
+
 conntrack - BOOLEAN
 	0 - disabled (default)
 	not 0 - enabled

include/net/ip_vs.h

@@ -924,6 +924,7 @@ struct netns_ipvs {
 	int			sysctl_pmtu_disc;
 	int			sysctl_backup_only;
 	int			sysctl_conn_reuse_mode;
+	int			sysctl_conn_reuse_old_conntrack;
 	int			sysctl_schedule_icmp;
 	int			sysctl_ignore_tunneled;
 
@@ -1045,6 +1046,11 @@ static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
 	return ipvs->sysctl_conn_reuse_mode;
 }
 
+static inline int sysctl_conn_reuse_old_conntrack(struct netns_ipvs *ipvs)
+{
+	return ipvs->sysctl_conn_reuse_old_conntrack;
+}
+
 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
 {
 	return ipvs->sysctl_schedule_icmp;
@@ -1132,6 +1138,11 @@ static inline int sysctl_conn_reuse_mode(struct netns_ipvs *ipvs)
 	return 1;
 }
 
+static inline int sysctl_conn_reuse_old_conntrack(struct netns_ipvs *ipvs)
+{
+	return 0;
+}
+
 static inline int sysctl_schedule_icmp(struct netns_ipvs *ipvs)
 {
 	return 0;

net/netfilter/ipvs/ip_vs_core.c

@@ -1928,7 +1928,7 @@ ip_vs_in(struct netns_ipvs *ipvs, unsigned int hooknum, struct sk_buff *skb, int
 
 	conn_reuse_mode = sysctl_conn_reuse_mode(ipvs);
 	if (conn_reuse_mode && !iph.fragoffs && is_new_conn(skb, &iph) && cp) {
-		bool uses_ct = false, resched = false;
+		bool uses_ct = false, resched = false, drop = false;
 
 		if (unlikely(sysctl_expire_nodest_conn(ipvs)) && cp->dest &&
 		    unlikely(!atomic_read(&cp->dest->weight))) {
@@ -1948,10 +1948,17 @@ ip_vs_in(struct netns_ipvs *ipvs, unsigned int hooknum, struct sk_buff *skb, int
 		}
 
 		if (resched) {
+			if (uses_ct) {
+				if (likely(!atomic_read(&cp->n_control) && !cp->control) &&
+				   likely(sysctl_conn_reuse_old_conntrack(ipvs)))
+					cp->flags &= ~IP_VS_CONN_F_NFCT;
+				else
+					drop = true;
+			}
 			if (!atomic_read(&cp->n_control))
 				ip_vs_conn_expire_now(cp);
 			__ip_vs_conn_put(cp);
-			if (uses_ct)
+			if (drop)
 				return NF_DROP;
 			cp = NULL;
 		}

net/netfilter/ipvs/ip_vs_ctl.c

@@ -1891,6 +1891,12 @@ static struct ctl_table vs_vars[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
+	{
+		.procname	= "conn_reuse_old_conntrack",
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec,
+	},
 	{
 		.procname	= "schedule_icmp",
 		.maxlen		= sizeof(int),
@@ -3951,7 +3957,9 @@ static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs)
 	tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
 	tbl[idx++].data = &ipvs->sysctl_backup_only;
 	ipvs->sysctl_conn_reuse_mode = 1;
+	ipvs->sysctl_conn_reuse_old_conntrack = 0;
 	tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode;
+	tbl[idx++].data = &ipvs->sysctl_conn_reuse_old_conntrack;
 	tbl[idx++].data = &ipvs->sysctl_schedule_icmp;
 	tbl[idx++].data = &ipvs->sysctl_ignore_tunneled;