[PATCHv4 net-next 05/14] ipvs: do not keep dest_dst after dest is removed

Before now dest->dest_dst is not released when server is moved into
dest_trash list after removal. As result, we can keep dst/dev
references for long time without actively using them.

It is better to avoid walking the dest_trash list when
ip_vs_dst_event() receives dev events. So, make sure we do not
hold dev references in dest_trash list. As packets can be flying
while server is being removed, check the IP_VS_DEST_F_AVAILABLE
flag in slow path to ensure we do not save new dev references to
removed servers.

Signed-off-by: Julian Anastasov <[email protected]>
---
 net/netfilter/ipvs/ip_vs_ctl.c  | 20 +++++++----------
 net/netfilter/ipvs/ip_vs_xmit.c | 39 ++++++++++++++++++++++++---------
 2 files changed, 37 insertions(+), 22 deletions(-)

diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c
index ac87675dc779..d377df9c7a37 100644
--- a/net/netfilter/ipvs/ip_vs_ctl.c
+++ b/net/netfilter/ipvs/ip_vs_ctl.c
@@ -810,7 +810,6 @@ static void ip_vs_dest_free(struct ip_vs_dest *dest)
 {
 	struct ip_vs_service *svc = rcu_dereference_protected(dest->svc, 1);
 
-	__ip_vs_dst_cache_reset(dest);
 	__ip_vs_svc_put(svc);
 	call_rcu(&dest->rcu_head, ip_vs_dest_rcu_free);
 }
@@ -1013,10 +1012,6 @@ __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 
 	dest->af = udest->af;
 
-	spin_lock_bh(&dest->dst_lock);
-	__ip_vs_dst_cache_reset(dest);
-	spin_unlock_bh(&dest->dst_lock);
-
 	if (add) {
 		list_add_rcu(&dest->n_list, &svc->destinations);
 		svc->num_dests++;
@@ -1024,6 +1019,10 @@ __ip_vs_update_dest(struct ip_vs_service *svc, struct ip_vs_dest *dest,
 		if (sched && sched->add_dest)
 			sched->add_dest(svc, dest);
 	} else {
+		spin_lock_bh(&dest->dst_lock);
+		__ip_vs_dst_cache_reset(dest);
+		spin_unlock_bh(&dest->dst_lock);
+
 		sched = rcu_dereference_protected(svc->scheduler, 1);
 		if (sched && sched->upd_dest)
 			sched->upd_dest(svc, dest);
@@ -1258,6 +1257,10 @@ static void __ip_vs_unlink_dest(struct ip_vs_service *svc,
 {
 	dest->flags &= ~IP_VS_DEST_F_AVAILABLE;
 
+	spin_lock_bh(&dest->dst_lock);
+	__ip_vs_dst_cache_reset(dest);
+	spin_unlock_bh(&dest->dst_lock);
+
 	/*
 	 *  Remove it from the d-linked destination list.
 	 */
@@ -1747,13 +1750,6 @@ static int ip_vs_dst_event(struct notifier_block *this, unsigned long event,
 	}
 	rcu_read_unlock();
 
-	mutex_lock(&ipvs->service_mutex);
-	spin_lock_bh(&ipvs->dest_trash_lock);
-	list_for_each_entry(dest, &ipvs->dest_trash, t_list) {
-		ip_vs_forget_dev(dest, dev);
-	}
-	spin_unlock_bh(&ipvs->dest_trash_lock);
-	mutex_unlock(&ipvs->service_mutex);
 	return NOTIFY_DONE;
 }
 
diff --git a/net/netfilter/ipvs/ip_vs_xmit.c b/net/netfilter/ipvs/ip_vs_xmit.c
index 3313bceb6cc9..28cd6ba52411 100644
--- a/net/netfilter/ipvs/ip_vs_xmit.c
+++ b/net/netfilter/ipvs/ip_vs_xmit.c
@@ -317,9 +317,11 @@ __ip_vs_get_out_rt(struct netns_ipvs *ipvs, int skb_af, struct sk_buff *skb,
 
 	if (dest) {
 		dest_dst = __ip_vs_dst_check(dest);
-		if (likely(dest_dst))
+		if (likely(dest_dst)) {
 			rt = dst_rtable(dest_dst->dst_cache);
-		else {
+			if (ret_saddr)
+				*ret_saddr = dest_dst->dst_saddr.ip;
+		} else {
 			dest_dst = ip_vs_dest_dst_alloc();
 			spin_lock_bh(&dest->dst_lock);
 			if (!dest_dst) {
@@ -335,14 +337,24 @@ __ip_vs_get_out_rt(struct netns_ipvs *ipvs, int skb_af, struct sk_buff *skb,
 				ip_vs_dest_dst_free(dest_dst);
 				goto err_unreach;
 			}
-			__ip_vs_dst_set(dest, dest_dst, &rt->dst, 0);
+			/* It is forbidden to attach dest->dest_dst if
+			 * server is deleted. We can see the flag going down,
+			 * for very short period and it must be checked under
+			 * dst_lock.
+			 */
+			if (dest->flags & IP_VS_DEST_F_AVAILABLE)
+				__ip_vs_dst_set(dest, dest_dst, &rt->dst, 0);
+			else
+				noref = 0;
 			spin_unlock_bh(&dest->dst_lock);
 			IP_VS_DBG(10, "new dst %pI4, src %pI4, refcnt=%d\n",
 				  &dest->addr.ip, &dest_dst->dst_saddr.ip,
 				  rcuref_read(&rt->dst.__rcuref));
+			if (ret_saddr)
+				*ret_saddr = dest_dst->dst_saddr.ip;
+			if (!noref)
+				ip_vs_dest_dst_free(dest_dst);
 		}
-		if (ret_saddr)
-			*ret_saddr = dest_dst->dst_saddr.ip;
 	} else {
 		__be32 saddr = htonl(INADDR_ANY);
 
@@ -480,9 +492,11 @@ __ip_vs_get_out_rt_v6(struct netns_ipvs *ipvs, int skb_af, struct sk_buff *skb,
 
 	if (dest) {
 		dest_dst = __ip_vs_dst_check(dest);
-		if (likely(dest_dst))
+		if (likely(dest_dst)) {
 			rt = dst_rt6_info(dest_dst->dst_cache);
-		else {
+			if (ret_saddr)
+				*ret_saddr = dest_dst->dst_saddr.in6;
+		} else {
 			u32 cookie;
 
 			dest_dst = ip_vs_dest_dst_alloc();
@@ -503,14 +517,19 @@ __ip_vs_get_out_rt_v6(struct netns_ipvs *ipvs, int skb_af, struct sk_buff *skb,
 			}
 			rt = dst_rt6_info(dst);
 			cookie = rt6_get_cookie(rt);
-			__ip_vs_dst_set(dest, dest_dst, &rt->dst, cookie);
+			if (dest->flags & IP_VS_DEST_F_AVAILABLE)
+				__ip_vs_dst_set(dest, dest_dst, &rt->dst, cookie);
+			else
+				noref = 0;
 			spin_unlock_bh(&dest->dst_lock);
 			IP_VS_DBG(10, "new dst %pI6, src %pI6, refcnt=%d\n",
 				  &dest->addr.in6, &dest_dst->dst_saddr.in6,
 				  rcuref_read(&rt->dst.__rcuref));
+			if (ret_saddr)
+				*ret_saddr = dest_dst->dst_saddr.in6;
+			if (!noref)
+				ip_vs_dest_dst_free(dest_dst);
 		}
-		if (ret_saddr)
-			*ret_saddr = dest_dst->dst_saddr.in6;
 	} else {
 		noref = 0;
 		dst = __ip_vs_route_output_v6(net, daddr, ret_saddr, do_xfrm,
-- 
2.44.0