ipvs: Fix reuse connection if real server is dead

Expire cached connection for new TCP/SCTP connection if real
server is down. Otherwise, IPVS uses the dead server for the
reused connection, instead of a new working one.

Signed-off-by: Grzegorz Lyczba <grzegorz.lyczba@gmail.com>
Acked-by: Hans Schillstrom <hans@schillstrom.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

1 files changed, 35 insertions, 0 deletions

diff --git a/net/netfilter/ipvs/ip_vs_core.c b/net/netfilter/ipvs/ip_vs_core.c
index 085b5880ab0d..05565d2b3a61 100644
--- a/net/netfilter/ipvs/ip_vs_core.c
+++ b/net/netfilter/ipvs/ip_vs_core.c
@@ -1001,6 +1001,32 @@ static inline int is_tcp_reset(const struct sk_buff *skb, int nh_len)
         return th->rst;
 }
 
+static inline bool is_new_conn(const struct sk_buff *skb,
+                               struct ip_vs_iphdr *iph)
+{
+        switch (iph->protocol) {
+        case IPPROTO_TCP: {
+                struct tcphdr _tcph, *th;
+
+                th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph);
+                if (th == NULL)
+                        return false;
+                return th->syn;
+        }
+        case IPPROTO_SCTP: {
+                sctp_chunkhdr_t *sch, schunk;
+
+                sch = skb_header_pointer(skb, iph->len + sizeof(sctp_sctphdr_t),
+                                         sizeof(schunk), &schunk);
+                if (sch == NULL)
+                        return false;
+                return sch->type == SCTP_CID_INIT;
+        }
+        default:
+                return false;
+        }
+}
+
 /* Handle response packets: rewrite addresses and send away...
  */
 static unsigned int
@@ -1612,6 +1638,15 @@ ip_vs_in(unsigned int hooknum, struct sk_buff *skb, int af)
          * Check if the packet belongs to an existing connection entry
          */
         cp = pp->conn_in_get(af, skb, &iph, 0);
+
+        if (unlikely(sysctl_expire_nodest_conn(ipvs)) && cp && cp->dest &&
+            unlikely(!atomic_read(&cp->dest->weight)) && !iph.fragoffs &&
+            is_new_conn(skb, &iph)) {
+                ip_vs_conn_expire_now(cp);
+                __ip_vs_conn_put(cp);
+                cp = NULL;
+        }
+
         if (unlikely(!cp) && !iph.fragoffs) {
                 /* No (second) fragments need to enter here, as nf_defrag_ipv6
                  * replayed fragment zero will already have created the cp