1 files changed, 63 insertions, 39 deletions

diff --git a/net/ipv4/nexthop.c b/net/ipv4/nexthop.c
index ec1282858cb7..ebafa5ed91ac 100644
--- a/net/ipv4/nexthop.c
+++ b/net/ipv4/nexthop.c
@@ -75,9 +75,16 @@ static void nexthop_free_mpath(struct nexthop *nh)
 	int i;
 
 	nhg = rcu_dereference_raw(nh->nh_grp);
-	for (i = 0; i < nhg->num_nh; ++i)
-		WARN_ON(nhg->nh_entries[i].nh);
+	for (i = 0; i < nhg->num_nh; ++i) {
+		struct nh_grp_entry *nhge = &nhg->nh_entries[i];
+
+		WARN_ON(!list_empty(&nhge->nh_list));
+		nexthop_put(nhge->nh);
+	}
+
+	WARN_ON(nhg->spare == nhg);
 
+	kfree(nhg->spare);
 	kfree(nhg);
 }
 
@@ -758,41 +765,56 @@ static void nh_group_rebalance(struct nh_group *nhg)
 	}
 }
 
-static void remove_nh_grp_entry(struct nh_grp_entry *nhge,
-				struct nh_group *nhg,
+static void remove_nh_grp_entry(struct net *net, struct nh_grp_entry *nhge,
 				struct nl_info *nlinfo)
 {
+	struct nh_grp_entry *nhges, *new_nhges;
+	struct nexthop *nhp = nhge->nh_parent;
 	struct nexthop *nh = nhge->nh;
-	struct nh_grp_entry *nhges;
-	bool found = false;
-	int i;
+	struct nh_group *nhg, *newg;
+	int i, j;
 
 	WARN_ON(!nh);
 
-	nhges = nhg->nh_entries;
-	for (i = 0; i < nhg->num_nh; ++i) {
-		if (found) {
-			nhges[i-1].nh = nhges[i].nh;
-			nhges[i-1].weight = nhges[i].weight;
-			list_del(&nhges[i].nh_list);
-			list_add(&nhges[i-1].nh_list, &nhges[i-1].nh->grp_list);
-		} else if (nhg->nh_entries[i].nh == nh) {
-			found = true;
-		}
-	}
+	nhg = rtnl_dereference(nhp->nh_grp);
+	newg = nhg->spare;
 
-	if (WARN_ON(!found))
+	/* last entry, keep it visible and remove the parent */
+	if (nhg->num_nh == 1) {
+		remove_nexthop(net, nhp, nlinfo);
 		return;
+	}
+
+	newg->has_v4 = nhg->has_v4;
+	newg->mpath = nhg->mpath;
+	newg->num_nh = nhg->num_nh;
 
-	nhg->num_nh--;
-	nhg->nh_entries[nhg->num_nh].nh = NULL;
+	/* copy old entries to new except the one getting removed */
+	nhges = nhg->nh_entries;
+	new_nhges = newg->nh_entries;
+	for (i = 0, j = 0; i < nhg->num_nh; ++i) {
+		/* current nexthop getting removed */
+		if (nhg->nh_entries[i].nh == nh) {
+			newg->num_nh--;
+			continue;
+		}
 
-	nh_group_rebalance(nhg);
+		list_del(&nhges[i].nh_list);
+		new_nhges[j].nh_parent = nhges[i].nh_parent;
+		new_nhges[j].nh = nhges[i].nh;
+		new_nhges[j].weight = nhges[i].weight;
+		list_add(&new_nhges[j].nh_list, &new_nhges[j].nh->grp_list);
+		j++;
+	}
 
-	nexthop_put(nh);
+	nh_group_rebalance(newg);
+	rcu_assign_pointer(nhp->nh_grp, newg);
+
+	list_del(&nhge->nh_list);
+	nexthop_put(nhge->nh);
 
 	if (nlinfo)
-		nexthop_notify(RTM_NEWNEXTHOP, nhge->nh_parent, nlinfo);
+		nexthop_notify(RTM_NEWNEXTHOP, nhp, nlinfo);
 }
 
 static void remove_nexthop_from_groups(struct net *net, struct nexthop *nh,
@@ -800,17 +822,11 @@ static void remove_nexthop_from_groups(struct net *net, struct nexthop *nh,
 {
 	struct nh_grp_entry *nhge, *tmp;
 
-	list_for_each_entry_safe(nhge, tmp, &nh->grp_list, nh_list) {
-		struct nh_group *nhg;
-
-		list_del(&nhge->nh_list);
-		nhg = rtnl_dereference(nhge->nh_parent->nh_grp);
-		remove_nh_grp_entry(nhge, nhg, nlinfo);
+	list_for_each_entry_safe(nhge, tmp, &nh->grp_list, nh_list)
+		remove_nh_grp_entry(net, nhge, nlinfo);
 
-		/* if this group has no more entries then remove it */
-		if (!nhg->num_nh)
-			remove_nexthop(net, nhge->nh_parent, nlinfo);
-	}
+	/* make sure all see the newly published array before releasing rtnl */
+	synchronize_rcu();
 }
 
 static void remove_nexthop_group(struct nexthop *nh, struct nl_info *nlinfo)
@@ -824,10 +840,7 @@ static void remove_nexthop_group(struct nexthop *nh, struct nl_info *nlinfo)
 		if (WARN_ON(!nhge->nh))
 			continue;
 
-		list_del(&nhge->nh_list);
-		nexthop_put(nhge->nh);
-		nhge->nh = NULL;
-		nhg->num_nh--;
+		list_del_init(&nhge->nh_list);
 	}
 }
 
@@ -1153,6 +1166,7 @@ static struct nexthop *nexthop_create_group(struct net *net,
 {
 	struct nlattr *grps_attr = cfg->nh_grp;
 	struct nexthop_grp *entry = nla_data(grps_attr);
+	u16 num_nh = nla_len(grps_attr) / sizeof(*entry);
 	struct nh_group *nhg;
 	struct nexthop *nh;
 	int i;
@@ -1163,12 +1177,21 @@ static struct nexthop *nexthop_create_group(struct net *net,
 
 	nh->is_group = 1;
 
-	nhg = nexthop_grp_alloc(nla_len(grps_attr) / sizeof(*entry));
+	nhg = nexthop_grp_alloc(num_nh);
 	if (!nhg) {
 		kfree(nh);
 		return ERR_PTR(-ENOMEM);
 	}
 
+	/* spare group used for removals */
+	nhg->spare = nexthop_grp_alloc(num_nh);
+	if (!nhg) {
+		kfree(nhg);
+		kfree(nh);
+		return NULL;
+	}
+	nhg->spare->spare = nhg;
+
 	for (i = 0; i < nhg->num_nh; ++i) {
 		struct nexthop *nhe;
 		struct nh_info *nhi;
@@ -1203,6 +1226,7 @@ out_no_nh:
 	for (; i >= 0; --i)
 		nexthop_put(nhg->nh_entries[i].nh);
 
+	kfree(nhg->spare);
 	kfree(nhg);
 	kfree(nh);