ipvs: avoid routing by TOS for real server
authorJulian Anastasov <ja@ssi.bg>
Thu, 21 Mar 2013 09:57:59 +0000 (11:57 +0200)
committerPablo Neira Ayuso <pablo@netfilter.org>
Mon, 1 Apr 2013 22:23:37 +0000 (00:23 +0200)
Avoid replacing the cached route for real server
on every packet with different TOS. I doubt that routing
by TOS for real server is used at all, so we should be
better with such optimization.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
include/net/ip_vs.h
net/netfilter/ipvs/ip_vs_xmit.c

index bee87badabef4820031dd444ca99cc2417705501..64db11769cc29296334026d5eb828dc0e0f3403d 100644 (file)
@@ -753,7 +753,6 @@ struct ip_vs_dest {
        /* for destination cache */
        spinlock_t              dst_lock;       /* lock of dst_cache */
        struct dst_entry        *dst_cache;     /* destination cache entry */
-       u32                     dst_rtos;       /* RT_TOS(tos) for dst */
        u32                     dst_cookie;
        union nf_inet_addr      dst_saddr;
 
index ee6b7a9f1ec24e4f0a0049ff3af4c6fc73a29752..4b0bd15ad7c176339b424e25821fe05d3dae873f 100644 (file)
@@ -57,27 +57,24 @@ enum {
  *      Destination cache to speed up outgoing route lookup
  */
 static inline void
-__ip_vs_dst_set(struct ip_vs_dest *dest, u32 rtos, struct dst_entry *dst,
-               u32 dst_cookie)
+__ip_vs_dst_set(struct ip_vs_dest *dest, struct dst_entry *dst, u32 dst_cookie)
 {
        struct dst_entry *old_dst;
 
        old_dst = dest->dst_cache;
        dest->dst_cache = dst;
-       dest->dst_rtos = rtos;
        dest->dst_cookie = dst_cookie;
        dst_release(old_dst);
 }
 
 static inline struct dst_entry *
-__ip_vs_dst_check(struct ip_vs_dest *dest, u32 rtos)
+__ip_vs_dst_check(struct ip_vs_dest *dest)
 {
        struct dst_entry *dst = dest->dst_cache;
 
        if (!dst)
                return NULL;
-       if ((dst->obsolete || rtos != dest->dst_rtos) &&
-           dst->ops->check(dst, dest->dst_cookie) == NULL) {
+       if (dst->obsolete && dst->ops->check(dst, dest->dst_cookie) == NULL) {
                dest->dst_cache = NULL;
                dst_release(dst);
                return NULL;
@@ -104,7 +101,7 @@ __mtu_check_toobig_v6(const struct sk_buff *skb, u32 mtu)
 
 /* Get route to daddr, update *saddr, optionally bind route to saddr */
 static struct rtable *do_output_route4(struct net *net, __be32 daddr,
-                                      u32 rtos, int rt_mode, __be32 *saddr)
+                                      int rt_mode, __be32 *saddr)
 {
        struct flowi4 fl4;
        struct rtable *rt;
@@ -113,7 +110,6 @@ static struct rtable *do_output_route4(struct net *net, __be32 daddr,
        memset(&fl4, 0, sizeof(fl4));
        fl4.daddr = daddr;
        fl4.saddr = (rt_mode & IP_VS_RT_MODE_CONNECT) ? *saddr : 0;
-       fl4.flowi4_tos = rtos;
        fl4.flowi4_flags = (rt_mode & IP_VS_RT_MODE_KNOWN_NH) ?
                           FLOWI_FLAG_KNOWN_NH : 0;
 
@@ -124,7 +120,7 @@ retry:
                if (PTR_ERR(rt) == -EINVAL && *saddr &&
                    rt_mode & IP_VS_RT_MODE_CONNECT && !loop) {
                        *saddr = 0;
-                       flowi4_update_output(&fl4, 0, rtos, daddr, 0);
+                       flowi4_update_output(&fl4, 0, 0, daddr, 0);
                        goto retry;
                }
                IP_VS_DBG_RL("ip_route_output error, dest: %pI4\n", &daddr);
@@ -132,7 +128,7 @@ retry:
        } else if (!*saddr && rt_mode & IP_VS_RT_MODE_CONNECT && fl4.saddr) {
                ip_rt_put(rt);
                *saddr = fl4.saddr;
-               flowi4_update_output(&fl4, 0, rtos, daddr, fl4.saddr);
+               flowi4_update_output(&fl4, 0, 0, daddr, fl4.saddr);
                loop++;
                goto retry;
        }
@@ -143,7 +139,7 @@ retry:
 /* Get route to destination or remote server */
 static struct rtable *
 __ip_vs_get_out_rt(struct sk_buff *skb, struct ip_vs_dest *dest,
-                  __be32 daddr, u32 rtos, int rt_mode, __be32 *ret_saddr)
+                  __be32 daddr, int rt_mode, __be32 *ret_saddr)
 {
        struct net *net = dev_net(skb_dst(skb)->dev);
        struct rtable *rt;                      /* Route to the other host */
@@ -152,19 +148,18 @@ __ip_vs_get_out_rt(struct sk_buff *skb, struct ip_vs_dest *dest,
 
        if (dest) {
                spin_lock(&dest->dst_lock);
-               if (!(rt = (struct rtable *)
-                     __ip_vs_dst_check(dest, rtos))) {
-                       rt = do_output_route4(net, dest->addr.ip, rtos,
-                                             rt_mode, &dest->dst_saddr.ip);
+               rt = (struct rtable *) __ip_vs_dst_check(dest);
+               if (!rt) {
+                       rt = do_output_route4(net, dest->addr.ip, rt_mode,
+                                             &dest->dst_saddr.ip);
                        if (!rt) {
                                spin_unlock(&dest->dst_lock);
                                return NULL;
                        }
-                       __ip_vs_dst_set(dest, rtos, dst_clone(&rt->dst), 0);
-                       IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d, "
-                                 "rtos=%X\n",
+                       __ip_vs_dst_set(dest, dst_clone(&rt->dst), 0);
+                       IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n",
                                  &dest->addr.ip, &dest->dst_saddr.ip,
-                                 atomic_read(&rt->dst.__refcnt), rtos);
+                                 atomic_read(&rt->dst.__refcnt));
                }
                daddr = dest->addr.ip;
                if (ret_saddr)
@@ -177,7 +172,7 @@ __ip_vs_get_out_rt(struct sk_buff *skb, struct ip_vs_dest *dest,
                 * for performance reasons because we do not remember saddr
                 */
                rt_mode &= ~IP_VS_RT_MODE_CONNECT;
-               rt = do_output_route4(net, daddr, rtos, rt_mode, &saddr);
+               rt = do_output_route4(net, daddr, rt_mode, &saddr);
                if (!rt)
                        return NULL;
                if (ret_saddr)
@@ -307,7 +302,7 @@ __ip_vs_get_out_rt_v6(struct sk_buff *skb, struct ip_vs_dest *dest,
 
        if (dest) {
                spin_lock(&dest->dst_lock);
-               rt = (struct rt6_info *)__ip_vs_dst_check(dest, 0);
+               rt = (struct rt6_info *)__ip_vs_dst_check(dest);
                if (!rt) {
                        u32 cookie;
 
@@ -320,7 +315,7 @@ __ip_vs_get_out_rt_v6(struct sk_buff *skb, struct ip_vs_dest *dest,
                        }
                        rt = (struct rt6_info *) dst;
                        cookie = rt->rt6i_node ? rt->rt6i_node->fn_sernum : 0;
-                       __ip_vs_dst_set(dest, 0, dst_clone(&rt->dst), cookie);
+                       __ip_vs_dst_set(dest, dst_clone(&rt->dst), cookie);
                        IP_VS_DBG(10, "new dst %pI6, src %pI6, refcnt=%d\n",
                                  &dest->addr.in6, &dest->dst_saddr.in6,
                                  atomic_read(&rt->dst.__refcnt));
@@ -449,8 +444,9 @@ ip_vs_bypass_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
 
        EnterFunction(10);
 
-       if (!(rt = __ip_vs_get_out_rt(skb, NULL, iph->daddr, RT_TOS(iph->tos),
-                                     IP_VS_RT_MODE_NON_LOCAL, NULL)))
+       rt = __ip_vs_get_out_rt(skb, NULL, iph->daddr, IP_VS_RT_MODE_NON_LOCAL,
+                               NULL);
+       if (!rt)
                goto tx_error_icmp;
 
        /* MTU checking */
@@ -581,10 +577,9 @@ ip_vs_nat_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        }
 
        if (!(rt = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
-                                     RT_TOS(iph->tos),
                                      IP_VS_RT_MODE_LOCAL |
-                                       IP_VS_RT_MODE_NON_LOCAL |
-                                       IP_VS_RT_MODE_RDR, NULL)))
+                                     IP_VS_RT_MODE_NON_LOCAL |
+                                     IP_VS_RT_MODE_RDR, NULL)))
                goto tx_error_icmp;
        local = rt->rt_flags & RTCF_LOCAL;
        /*
@@ -832,10 +827,9 @@ ip_vs_tunnel_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        EnterFunction(10);
 
        if (!(rt = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
-                                     RT_TOS(tos), IP_VS_RT_MODE_LOCAL |
-                                                  IP_VS_RT_MODE_NON_LOCAL |
-                                                  IP_VS_RT_MODE_CONNECT,
-                                                  &saddr)))
+                                     IP_VS_RT_MODE_LOCAL |
+                                     IP_VS_RT_MODE_NON_LOCAL |
+                                     IP_VS_RT_MODE_CONNECT, &saddr)))
                goto tx_error_icmp;
        if (rt->rt_flags & RTCF_LOCAL) {
                ip_rt_put(rt);
@@ -1067,7 +1061,6 @@ ip_vs_dr_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
        EnterFunction(10);
 
        if (!(rt = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
-                                     RT_TOS(iph->tos),
                                      IP_VS_RT_MODE_LOCAL |
                                      IP_VS_RT_MODE_NON_LOCAL |
                                      IP_VS_RT_MODE_KNOWN_NH, NULL)))
@@ -1223,7 +1216,6 @@ ip_vs_icmp_xmit(struct sk_buff *skb, struct ip_vs_conn *cp,
                  IP_VS_RT_MODE_LOCAL | IP_VS_RT_MODE_NON_LOCAL |
                  IP_VS_RT_MODE_RDR : IP_VS_RT_MODE_NON_LOCAL;
        if (!(rt = __ip_vs_get_out_rt(skb, cp->dest, cp->daddr.ip,
-                                     RT_TOS(ip_hdr(skb)->tos),
                                      rt_mode, NULL)))
                goto tx_error_icmp;
        local = rt->rt_flags & RTCF_LOCAL;