[GIT PULL nf-next] IPVS

[GIT PULL nf-next] IPVS

[Simon Horman]

Hi Pablo,

please consider the following IPVS enhancements from Julian for 3.10.

----------------------------------------------------------------
The following changes since commit 1cdb09056b27b2a06b06dc7187d2c33d57082d20:

  netfilter: nfnetlink_queue: use xor hash function to distribute instances (2013-03-15 12:38:40 +0100)

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/horms/ipvs-next.git tags/ipvs-for-v3.10

for you to fetch changes up to e095c65caffc923a0d1fb27763c3f9ec1dcb57fc:

  ipvs: fix some sparse warnings (2013-03-18 19:25:14 +0900)

----------------------------------------------------------------
IPVS enhancements for v3.10 from Julian Anastasov

----------------------------------------------------------------
Julian Anastasov (5):
      ipvs: add backup_only flag to avoid loops
      ipvs: remove extra rcu lock
      ipvs: fix sctp chunk length order
      ipvs: fix hashing in ip_vs_svc_hashkey
      ipvs: fix some sparse warnings

 Documentation/networking/ipvs-sysctl.txt |    7 +++++++
 include/net/ip_vs.h                      |   14 +++++++++++++-
 net/netfilter/ipvs/ip_vs_core.c          |   22 +++++++++-------------
 net/netfilter/ipvs/ip_vs_ctl.c           |   15 ++++++++++++---
 net/netfilter/ipvs/ip_vs_est.c           |    2 +-
 net/netfilter/ipvs/ip_vs_proto_sctp.c    |   16 +++++++++-------
 6 files changed, 51 insertions(+), 25 deletions(-)

[PATCH 1/5] ipvs: add backup_only flag to avoid loops

[Simon Horman]

From: Julian Anastasov <ja@ssi.bg>

	Dmitry Akindinov is reporting for a problem where
SYNs are looping between the master and backup server
when the backup server is used as real server in DR mode
and has IPVS rules to function as director.

	Even when the backup function is enabled we
continue to forward traffic and schedule new connections
when the current master is using the backup server as
real server. While this is not a problem for NAT, for
DR and TUN method the backup server can not determine
if a request comes from client or from director.

	To avoid such loops add new sysctl flag
backup_only. It can be needed for DR/TUN setups that
do not need backup and director function at the
same time. When the backup function is enabled we
stop any forwarding and pass the traffic to the local
stack (real server mode). The flag disables the
director function when the backup function is enabled.

	For setups that enable backup function for
some virtual services and director function for
other virtual services there should be another more
complex solution to support DR/TUN mode, may be to
assign per-virtual service syncid value, so that
we can differentiate the requests.

Reported-by: Dmitry Akindinov <dimak@stalker.com>
Tested-by: German Myzovsky <lawyer@sipnet.ru>
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
---
 Documentation/networking/ipvs-sysctl.txt |    7 +++++++
 include/net/ip_vs.h                      |   12 ++++++++++++
 net/netfilter/ipvs/ip_vs_core.c          |   12 ++++++++----
 net/netfilter/ipvs/ip_vs_ctl.c           |    7 +++++++
 4 files changed, 34 insertions(+), 4 deletions(-)

diff --git a/Documentation/networking/ipvs-sysctl.txt b/Documentation/networking/ipvs-sysctl.txt
index f2a2488..9573d0c 100644
--- a/Documentation/networking/ipvs-sysctl.txt
+++ b/Documentation/networking/ipvs-sysctl.txt
@@ -15,6 +15,13 @@ amemthresh - INTEGER
         enabled and the variable is automatically set to 2, otherwise
         the strategy is disabled and the variable is  set  to 1.
 
+backup_only - BOOLEAN
+	0 - disabled (default)
+	not 0 - enabled
+
+	If set, disable the director function while the server is
+	in backup mode to avoid packet loops for DR/TUN methods.
+
 conntrack - BOOLEAN
 	0 - disabled (default)
 	not 0 - enabled
diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h
index 68c69d5..fce8e6b 100644
--- a/include/net/ip_vs.h
+++ b/include/net/ip_vs.h
@@ -976,6 +976,7 @@ struct netns_ipvs {
 	int			sysctl_sync_retries;
 	int			sysctl_nat_icmp_send;
 	int			sysctl_pmtu_disc;
+	int			sysctl_backup_only;
 
 	/* ip_vs_lblc */
 	int			sysctl_lblc_expiration;
@@ -1067,6 +1068,12 @@ static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
 	return ipvs->sysctl_pmtu_disc;
 }
 
+static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
+{
+	return ipvs->sync_state & IP_VS_STATE_BACKUP &&
+	       ipvs->sysctl_backup_only;
+}
+
 #else
 
 static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs)
@@ -1114,6 +1121,11 @@ static inline int sysctl_pmtu_disc(struct netns_ipvs *ipvs)
 	return 1;
 }
 
+static inline int sysctl_backup_only(struct netns_ipvs *ipvs)
+{
+	return 0;
+}
+
 #endif
 
 /*
diff --git a/net/netfilter/ipvs/ip_vs_core.c b/net/netfilter/ipvs/ip_vs_core.c
index 47edf5a..18b4bc5 100644
--- a/net/netfilter/ipvs/ip_vs_core.c
+++ b/net/netfilter/ipvs/ip_vs_core.c
@@ -1577,7 +1577,8 @@ ip_vs_in(unsigned int hooknum, struct sk_buff *skb, int af)
 	}
 	/* ipvs enabled in this netns ? */
 	net = skb_net(skb);
-	if (!net_ipvs(net)->enable)
+	ipvs = net_ipvs(net);
+	if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable))
 		return NF_ACCEPT;
 
 	ip_vs_fill_iph_skb(af, skb, &iph);
@@ -1654,7 +1655,6 @@ ip_vs_in(unsigned int hooknum, struct sk_buff *skb, int af)
 	}
 
 	IP_VS_DBG_PKT(11, af, pp, skb, 0, "Incoming packet");
-	ipvs = net_ipvs(net);
 	/* Check the server status */
 	if (cp->dest && !(cp->dest->flags & IP_VS_DEST_F_AVAILABLE)) {
 		/* the destination server is not available */
@@ -1815,13 +1815,15 @@ ip_vs_forward_icmp(unsigned int hooknum, struct sk_buff *skb,
 {
 	int r;
 	struct net *net;
+	struct netns_ipvs *ipvs;
 
 	if (ip_hdr(skb)->protocol != IPPROTO_ICMP)
 		return NF_ACCEPT;
 
 	/* ipvs enabled in this netns ? */
 	net = skb_net(skb);
-	if (!net_ipvs(net)->enable)
+	ipvs = net_ipvs(net);
+	if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable))
 		return NF_ACCEPT;
 
 	return ip_vs_in_icmp(skb, &r, hooknum);
@@ -1835,6 +1837,7 @@ ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb,
 {
 	int r;
 	struct net *net;
+	struct netns_ipvs *ipvs;
 	struct ip_vs_iphdr iphdr;
 
 	ip_vs_fill_iph_skb(AF_INET6, skb, &iphdr);
@@ -1843,7 +1846,8 @@ ip_vs_forward_icmp_v6(unsigned int hooknum, struct sk_buff *skb,
 
 	/* ipvs enabled in this netns ? */
 	net = skb_net(skb);
-	if (!net_ipvs(net)->enable)
+	ipvs = net_ipvs(net);
+	if (unlikely(sysctl_backup_only(ipvs) || !ipvs->enable))
 		return NF_ACCEPT;
 
 	return ip_vs_in_icmp_v6(skb, &r, hooknum, &iphdr);
diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c
index c68198b..9e2d1cc 100644
--- a/net/netfilter/ipvs/ip_vs_ctl.c
+++ b/net/netfilter/ipvs/ip_vs_ctl.c
@@ -1808,6 +1808,12 @@ static struct ctl_table vs_vars[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec,
 	},
+	{
+		.procname	= "backup_only",
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec,
+	},
 #ifdef CONFIG_IP_VS_DEBUG
 	{
 		.procname	= "debug_level",
@@ -3741,6 +3747,7 @@ static int __net_init ip_vs_control_net_init_sysctl(struct net *net)
 	tbl[idx++].data = &ipvs->sysctl_nat_icmp_send;
 	ipvs->sysctl_pmtu_disc = 1;
 	tbl[idx++].data = &ipvs->sysctl_pmtu_disc;
+	tbl[idx++].data = &ipvs->sysctl_backup_only;
 
 
 	ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl);
-- 
1.7.10.4

[PATCH 2/5] ipvs: remove extra rcu lock

[Simon Horman]

From: Julian Anastasov <ja@ssi.bg>

	In 3.7 we added code that uses ipv4_update_pmtu
but after commit c5ae7d4192 (ipv4: must use rcu protection
while calling fib_lookup) the RCU lock is not needed.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
---
 net/netfilter/ipvs/ip_vs_core.c |    2 --
 1 file changed, 2 deletions(-)

diff --git a/net/netfilter/ipvs/ip_vs_core.c b/net/netfilter/ipvs/ip_vs_core.c
index 18b4bc5..61f49d2 100644
--- a/net/netfilter/ipvs/ip_vs_core.c
+++ b/net/netfilter/ipvs/ip_vs_core.c
@@ -1394,10 +1394,8 @@ ip_vs_in_icmp(struct sk_buff *skb, int *related, unsigned int hooknum)
 			skb_reset_network_header(skb);
 			IP_VS_DBG(12, "ICMP for IPIP %pI4->%pI4: mtu=%u\n",
 				&ip_hdr(skb)->saddr, &ip_hdr(skb)->daddr, mtu);
-			rcu_read_lock();
 			ipv4_update_pmtu(skb, dev_net(skb->dev),
 					 mtu, 0, 0, 0, 0);
-			rcu_read_unlock();
 			/* Client uses PMTUD? */
 			if (!(cih->frag_off & htons(IP_DF)))
 				goto ignore_ipip;
-- 
1.7.10.4

[PATCH 3/5] ipvs: fix sctp chunk length order

[Simon Horman]

From: Julian Anastasov <ja@ssi.bg>

	Fix wrong but non-fatal access to chunk length.
sch->length should be in network order, next chunk should
be aligned to 4 bytes. Problem noticed in sparse output.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
---
 net/netfilter/ipvs/ip_vs_proto_sctp.c |   16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/net/netfilter/ipvs/ip_vs_proto_sctp.c b/net/netfilter/ipvs/ip_vs_proto_sctp.c
index ae8ec6f..cd1d729 100644
--- a/net/netfilter/ipvs/ip_vs_proto_sctp.c
+++ b/net/netfilter/ipvs/ip_vs_proto_sctp.c
@@ -906,7 +906,7 @@ set_sctp_state(struct ip_vs_proto_data *pd, struct ip_vs_conn *cp,
 	sctp_chunkhdr_t _sctpch, *sch;
 	unsigned char chunk_type;
 	int event, next_state;
-	int ihl;
+	int ihl, cofs;
 
 #ifdef CONFIG_IP_VS_IPV6
 	ihl = cp->af == AF_INET ? ip_hdrlen(skb) : sizeof(struct ipv6hdr);
@@ -914,8 +914,8 @@ set_sctp_state(struct ip_vs_proto_data *pd, struct ip_vs_conn *cp,
 	ihl = ip_hdrlen(skb);
 #endif
 
-	sch = skb_header_pointer(skb, ihl + sizeof(sctp_sctphdr_t),
-				sizeof(_sctpch), &_sctpch);
+	cofs = ihl + sizeof(sctp_sctphdr_t);
+	sch = skb_header_pointer(skb, cofs, sizeof(_sctpch), &_sctpch);
 	if (sch == NULL)
 		return;
 
@@ -933,10 +933,12 @@ set_sctp_state(struct ip_vs_proto_data *pd, struct ip_vs_conn *cp,
 	 */
 	if ((sch->type == SCTP_CID_COOKIE_ECHO) ||
 	    (sch->type == SCTP_CID_COOKIE_ACK)) {
-		sch = skb_header_pointer(skb, (ihl + sizeof(sctp_sctphdr_t) +
-				sch->length), sizeof(_sctpch), &_sctpch);
-		if (sch) {
-			if (sch->type == SCTP_CID_ABORT)
+		int clen = ntohs(sch->length);
+
+		if (clen >= sizeof(sctp_chunkhdr_t)) {
+			sch = skb_header_pointer(skb, cofs + ALIGN(clen, 4),
+						 sizeof(_sctpch), &_sctpch);
+			if (sch && sch->type == SCTP_CID_ABORT)
 				chunk_type = sch->type;
 		}
 	}
-- 
1.7.10.4

[PATCH 4/5] ipvs: fix hashing in ip_vs_svc_hashkey

[Simon Horman]

From: Julian Anastasov <ja@ssi.bg>

	net is a pointer in host order, mix it properly
with other keys in network order. Fixes sparse warning.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
---
 net/netfilter/ipvs/ip_vs_ctl.c |    8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c
index 9e2d1cc..8104120 100644
--- a/net/netfilter/ipvs/ip_vs_ctl.c
+++ b/net/netfilter/ipvs/ip_vs_ctl.c
@@ -271,16 +271,18 @@ ip_vs_svc_hashkey(struct net *net, int af, unsigned int proto,
 {
 	register unsigned int porth = ntohs(port);
 	__be32 addr_fold = addr->ip;
+	__u32 ahash;
 
 #ifdef CONFIG_IP_VS_IPV6
 	if (af == AF_INET6)
 		addr_fold = addr->ip6[0]^addr->ip6[1]^
 			    addr->ip6[2]^addr->ip6[3];
 #endif
-	addr_fold ^= ((size_t)net>>8);
+	ahash = ntohl(addr_fold);
+	ahash ^= ((size_t) net >> 8);
 
-	return (proto^ntohl(addr_fold)^(porth>>IP_VS_SVC_TAB_BITS)^porth)
-		& IP_VS_SVC_TAB_MASK;
+	return (proto ^ ahash ^ (porth >> IP_VS_SVC_TAB_BITS) ^ porth) &
+	       IP_VS_SVC_TAB_MASK;
 }
 
 /*
-- 
1.7.10.4

[PATCH 5/5] ipvs: fix some sparse warnings

[Simon Horman]

From: Julian Anastasov <ja@ssi.bg>

	Add missing __percpu annotations and make
ip_vs_net_id static.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
---
 include/net/ip_vs.h             |    2 +-
 net/netfilter/ipvs/ip_vs_core.c |    8 +-------
 net/netfilter/ipvs/ip_vs_est.c  |    2 +-
 3 files changed, 3 insertions(+), 9 deletions(-)

diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h
index fce8e6b..bee87ba 100644
--- a/include/net/ip_vs.h
+++ b/include/net/ip_vs.h
@@ -459,7 +459,7 @@ struct ip_vs_estimator {
 struct ip_vs_stats {
 	struct ip_vs_stats_user	ustats;		/* statistics */
 	struct ip_vs_estimator	est;		/* estimator */
-	struct ip_vs_cpu_stats	*cpustats;	/* per cpu counters */
+	struct ip_vs_cpu_stats __percpu	*cpustats;	/* per cpu counters */
 	spinlock_t		lock;		/* spin lock */
 	struct ip_vs_stats_user	ustats0;	/* reset values */
 };
diff --git a/net/netfilter/ipvs/ip_vs_core.c b/net/netfilter/ipvs/ip_vs_core.c
index 61f49d2..2aef23e 100644
--- a/net/netfilter/ipvs/ip_vs_core.c
+++ b/net/netfilter/ipvs/ip_vs_core.c
@@ -69,10 +69,7 @@ EXPORT_SYMBOL(ip_vs_conn_put);
 EXPORT_SYMBOL(ip_vs_get_debug_level);
 #endif
 
-int ip_vs_net_id __read_mostly;
-#ifdef IP_VS_GENERIC_NETNS
-EXPORT_SYMBOL(ip_vs_net_id);
-#endif
+static int ip_vs_net_id __read_mostly;
 /* netns cnt used for uniqueness */
 static atomic_t ipvs_netns_cnt = ATOMIC_INIT(0);
 
@@ -1181,9 +1178,6 @@ ip_vs_out(unsigned int hooknum, struct sk_buff *skb, int af)
 						iph.len)))) {
 #ifdef CONFIG_IP_VS_IPV6
 				if (af == AF_INET6) {
-					struct net *net =
-						dev_net(skb_dst(skb)->dev);
-
 					if (!skb->dev)
 						skb->dev = net->loopback_dev;
 					icmpv6_send(skb,
diff --git a/net/netfilter/ipvs/ip_vs_est.c b/net/netfilter/ipvs/ip_vs_est.c
index 0fac601..6bee6d0 100644
--- a/net/netfilter/ipvs/ip_vs_est.c
+++ b/net/netfilter/ipvs/ip_vs_est.c
@@ -56,7 +56,7 @@
  * Make a summary from each cpu
  */
 static void ip_vs_read_cpu_stats(struct ip_vs_stats_user *sum,
-				 struct ip_vs_cpu_stats *stats)
+				 struct ip_vs_cpu_stats __percpu *stats)
 {
 	int i;
 
-- 
1.7.10.4

Re: [GIT PULL nf-next] IPVS

[Pablo Neira Ayuso]

Hi Simon,

On Mon, Mar 18, 2013 at 10:15:38PM +0900, Simon Horman wrote:
[...]
> ----------------------------------------------------------------
> IPVS enhancements for v3.10 from Julian Anastasov
> 
> ----------------------------------------------------------------
> Julian Anastasov (5):
>       ipvs: add backup_only flag to avoid loops
>       ipvs: remove extra rcu lock
>       ipvs: fix sctp chunk length order
>       ipvs: fix hashing in ip_vs_svc_hashkey
>       ipvs: fix some sparse warnings

I think that these three fixes:

ipvs: add backup_only flag to avoid loops
ipvs: remove extra rcu lock
ipvs: fix sctp chunk length order

should find their path to the net tree.

The remaining two sparse fixes should go to net-next.

I can manually apply these patch if you want to the corresponding
trees.

Re: [GIT PULL nf-next] IPVS

[Simon Horman]

On Tue, Mar 19, 2013 at 12:31:15AM +0100, Pablo Neira Ayuso wrote:
> Hi Simon,
> 
> On Mon, Mar 18, 2013 at 10:15:38PM +0900, Simon Horman wrote:
> [...]
> > ----------------------------------------------------------------
> > IPVS enhancements for v3.10 from Julian Anastasov
> > 
> > ----------------------------------------------------------------
> > Julian Anastasov (5):
> >       ipvs: add backup_only flag to avoid loops
> >       ipvs: remove extra rcu lock
> >       ipvs: fix sctp chunk length order
> >       ipvs: fix hashing in ip_vs_svc_hashkey
> >       ipvs: fix some sparse warnings
> 
> I think that these three fixes:
> 
> ipvs: add backup_only flag to avoid loops
> ipvs: remove extra rcu lock
> ipvs: fix sctp chunk length order
> 
> should find their path to the net tree.
> 
> The remaining two sparse fixes should go to net-next.
> 
> I can manually apply these patch if you want to the corresponding
> trees.

Thanks. I'll send two fresh pull requests shortly.