Merge branch 'bonding-ipsec'

Taehee Yoo says:

====================
net: fix bonding ipsec offload problems

This series fixes some problems related to bonding ipsec offload.

The 1, 5, and 8th patches are to add a missing rcu_read_lock().
The 2nd patch is to add null check code to bond_ipsec_add_sa.
When bonding interface doesn't have an active real interface, the
bond->curr_active_slave pointer is null.
But bond_ipsec_add_sa() uses that pointer without null check.
So that it results in null-ptr-deref.
The 3 and 4th patches are to replace xs->xso.dev with xs->xso.real_dev.
The 6th patch is to disallow to set ipsec offload if a real interface
type is bonding.
The 7th patch is to add struct bond_ipsec to manage SA.
If bond mode is changed, or active real interface is changed, SA should
be removed from old current active real interface then it should be added
to new active real interface.
But it can't, because it doesn't manage SA.
The 9th patch is to fix incorrect return value of bond_ipsec_offload_ok().

v1 -> v2:
 - Add 9th patch.
 - Do not print warning when there is no SA in bond_ipsec_add_sa_all().
 - Add comment for ipsec_lock.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

4 files changed, 178 insertions, 40 deletions

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 0ff7567bd04f..d22d78303311 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -401,24 +401,85 @@ static int bond_vlan_rx_kill_vid(struct net_device *bond_dev,
 static int bond_ipsec_add_sa(struct xfrm_state *xs)
 {
 	struct net_device *bond_dev = xs->xso.dev;
+	struct bond_ipsec *ipsec;
 	struct bonding *bond;
 	struct slave *slave;
+	int err;
 
 	if (!bond_dev)
 		return -EINVAL;
 
+	rcu_read_lock();
 	bond = netdev_priv(bond_dev);
 	slave = rcu_dereference(bond->curr_active_slave);
-	xs->xso.real_dev = slave->dev;
-	bond->xs = xs;
+	if (!slave) {
+		rcu_read_unlock();
+		return -ENODEV;
+	}
 
-	if (!(slave->dev->xfrmdev_ops
-	      && slave->dev->xfrmdev_ops->xdo_dev_state_add)) {
+	if (!slave->dev->xfrmdev_ops ||
+	    !slave->dev->xfrmdev_ops->xdo_dev_state_add ||
+	    netif_is_bond_master(slave->dev)) {
 		slave_warn(bond_dev, slave->dev, "Slave does not support ipsec offload\n");
+		rcu_read_unlock();
 		return -EINVAL;
 	}
 
-	return slave->dev->xfrmdev_ops->xdo_dev_state_add(xs);
+	ipsec = kmalloc(sizeof(*ipsec), GFP_ATOMIC);
+	if (!ipsec) {
+		rcu_read_unlock();
+		return -ENOMEM;
+	}
+	xs->xso.real_dev = slave->dev;
+
+	err = slave->dev->xfrmdev_ops->xdo_dev_state_add(xs);
+	if (!err) {
+		ipsec->xs = xs;
+		INIT_LIST_HEAD(&ipsec->list);
+		spin_lock_bh(&bond->ipsec_lock);
+		list_add(&ipsec->list, &bond->ipsec_list);
+		spin_unlock_bh(&bond->ipsec_lock);
+	} else {
+		kfree(ipsec);
+	}
+	rcu_read_unlock();
+	return err;
+}
+
+static void bond_ipsec_add_sa_all(struct bonding *bond)
+{
+	struct net_device *bond_dev = bond->dev;
+	struct bond_ipsec *ipsec;
+	struct slave *slave;
+
+	rcu_read_lock();
+	slave = rcu_dereference(bond->curr_active_slave);
+	if (!slave)
+		goto out;
+
+	if (!slave->dev->xfrmdev_ops ||
+	    !slave->dev->xfrmdev_ops->xdo_dev_state_add ||
+	    netif_is_bond_master(slave->dev)) {
+		spin_lock_bh(&bond->ipsec_lock);
+		if (!list_empty(&bond->ipsec_list))
+			slave_warn(bond_dev, slave->dev,
+				   "%s: no slave xdo_dev_state_add\n",
+				   __func__);
+		spin_unlock_bh(&bond->ipsec_lock);
+		goto out;
+	}
+
+	spin_lock_bh(&bond->ipsec_lock);
+	list_for_each_entry(ipsec, &bond->ipsec_list, list) {
+		ipsec->xs->xso.real_dev = slave->dev;
+		if (slave->dev->xfrmdev_ops->xdo_dev_state_add(ipsec->xs)) {
+			slave_warn(bond_dev, slave->dev, "%s: failed to add SA\n", __func__);
+			ipsec->xs->xso.real_dev = NULL;
+		}
+	}
+	spin_unlock_bh(&bond->ipsec_lock);
+out:
+	rcu_read_unlock();
 }
 
 /**
@@ -428,27 +489,77 @@ static int bond_ipsec_add_sa(struct xfrm_state *xs)
 static void bond_ipsec_del_sa(struct xfrm_state *xs)
 {
 	struct net_device *bond_dev = xs->xso.dev;
+	struct bond_ipsec *ipsec;
 	struct bonding *bond;
 	struct slave *slave;
 
 	if (!bond_dev)
 		return;
 
+	rcu_read_lock();
 	bond = netdev_priv(bond_dev);
 	slave = rcu_dereference(bond->curr_active_slave);
 
 	if (!slave)
-		return;
+		goto out;
 
-	xs->xso.real_dev = slave->dev;
+	if (!xs->xso.real_dev)
+		goto out;
+
+	WARN_ON(xs->xso.real_dev != slave->dev);
 
-	if (!(slave->dev->xfrmdev_ops
-	      && slave->dev->xfrmdev_ops->xdo_dev_state_delete)) {
+	if (!slave->dev->xfrmdev_ops ||
+	    !slave->dev->xfrmdev_ops->xdo_dev_state_delete ||
+	    netif_is_bond_master(slave->dev)) {
 		slave_warn(bond_dev, slave->dev, "%s: no slave xdo_dev_state_delete\n", __func__);
-		return;
+		goto out;
 	}
 
 	slave->dev->xfrmdev_ops->xdo_dev_state_delete(xs);
+out:
+	spin_lock_bh(&bond->ipsec_lock);
+	list_for_each_entry(ipsec, &bond->ipsec_list, list) {
+		if (ipsec->xs == xs) {
+			list_del(&ipsec->list);
+			kfree(ipsec);
+			break;
+		}
+	}
+	spin_unlock_bh(&bond->ipsec_lock);
+	rcu_read_unlock();
+}
+
+static void bond_ipsec_del_sa_all(struct bonding *bond)
+{
+	struct net_device *bond_dev = bond->dev;
+	struct bond_ipsec *ipsec;
+	struct slave *slave;
+
+	rcu_read_lock();
+	slave = rcu_dereference(bond->curr_active_slave);
+	if (!slave) {
+		rcu_read_unlock();
+		return;
+	}
+
+	spin_lock_bh(&bond->ipsec_lock);
+	list_for_each_entry(ipsec, &bond->ipsec_list, list) {
+		if (!ipsec->xs->xso.real_dev)
+			continue;
+
+		if (!slave->dev->xfrmdev_ops ||
+		    !slave->dev->xfrmdev_ops->xdo_dev_state_delete ||
+		    netif_is_bond_master(slave->dev)) {
+			slave_warn(bond_dev, slave->dev,
+				   "%s: no slave xdo_dev_state_delete\n",
+				   __func__);
+		} else {
+			slave->dev->xfrmdev_ops->xdo_dev_state_delete(ipsec->xs);
+		}
+		ipsec->xs->xso.real_dev = NULL;
+	}
+	spin_unlock_bh(&bond->ipsec_lock);
+	rcu_read_unlock();
 }
 
 /**
@@ -459,21 +570,37 @@ static void bond_ipsec_del_sa(struct xfrm_state *xs)
 static bool bond_ipsec_offload_ok(struct sk_buff *skb, struct xfrm_state *xs)
 {
 	struct net_device *bond_dev = xs->xso.dev;
-	struct bonding *bond = netdev_priv(bond_dev);
-	struct slave *curr_active = rcu_dereference(bond->curr_active_slave);
-	struct net_device *slave_dev = curr_active->dev;
+	struct net_device *real_dev;
+	struct slave *curr_active;
+	struct bonding *bond;
+	int err;
 
-	if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)
-		return true;
+	bond = netdev_priv(bond_dev);
+	rcu_read_lock();
+	curr_active = rcu_dereference(bond->curr_active_slave);
+	real_dev = curr_active->dev;
 
-	if (!(slave_dev->xfrmdev_ops
-	      && slave_dev->xfrmdev_ops->xdo_dev_offload_ok)) {
-		slave_warn(bond_dev, slave_dev, "%s: no slave xdo_dev_offload_ok\n", __func__);
-		return false;
+	if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
+		err = false;
+		goto out;
 	}
 
-	xs->xso.real_dev = slave_dev;
-	return slave_dev->xfrmdev_ops->xdo_dev_offload_ok(skb, xs);
+	if (!xs->xso.real_dev) {
+		err = false;
+		goto out;
+	}
+
+	if (!real_dev->xfrmdev_ops ||
+	    !real_dev->xfrmdev_ops->xdo_dev_offload_ok ||
+	    netif_is_bond_master(real_dev)) {
+		err = false;
+		goto out;
+	}
+
+	err = real_dev->xfrmdev_ops->xdo_dev_offload_ok(skb, xs);
+out:
+	rcu_read_unlock();
+	return err;
 }
 
 static const struct xfrmdev_ops bond_xfrmdev_ops = {
@@ -990,8 +1117,7 @@ void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
 		return;
 
 #ifdef CONFIG_XFRM_OFFLOAD
-	if (old_active && bond->xs)
-		bond_ipsec_del_sa(bond->xs);
+	bond_ipsec_del_sa_all(bond);
 #endif /* CONFIG_XFRM_OFFLOAD */
 
 	if (new_active) {
@@ -1066,10 +1192,7 @@ void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
 	}
 
 #ifdef CONFIG_XFRM_OFFLOAD
-	if (new_active && bond->xs) {
-		xfrm_dev_state_flush(dev_net(bond->dev), bond->dev, true);
-		bond_ipsec_add_sa(bond->xs);
-	}
+	bond_ipsec_add_sa_all(bond);
 #endif /* CONFIG_XFRM_OFFLOAD */
 
 	/* resend IGMP joins since active slave has changed or
@@ -3327,6 +3450,7 @@ static int bond_master_netdev_event(unsigned long event,
 		return bond_event_changename(event_bond);
 	case NETDEV_UNREGISTER:
 		bond_remove_proc_entry(event_bond);
+		xfrm_dev_state_flush(dev_net(bond_dev), bond_dev, true);
 		break;
 	case NETDEV_REGISTER:
 		bond_create_proc_entry(event_bond);
@@ -4894,7 +5018,8 @@ void bond_setup(struct net_device *bond_dev)
 #ifdef CONFIG_XFRM_OFFLOAD
 	/* set up xfrm device ops (only supported in active-backup right now) */
 	bond_dev->xfrmdev_ops = &bond_xfrmdev_ops;
-	bond->xs = NULL;
+	INIT_LIST_HEAD(&bond->ipsec_list);
+	spin_lock_init(&bond->ipsec_lock);
 #endif /* CONFIG_XFRM_OFFLOAD */
 
 	/* don't acquire bond device's netif_tx_lock when transmitting */
diff --git a/drivers/net/ethernet/intel/ixgbevf/ipsec.c b/drivers/net/ethernet/intel/ixgbevf/ipsec.c
index caaea2c920a6..e3e4676af9e4 100644
--- a/drivers/net/ethernet/intel/ixgbevf/ipsec.c
+++ b/drivers/net/ethernet/intel/ixgbevf/ipsec.c
@@ -211,7 +211,7 @@ struct xfrm_state *ixgbevf_ipsec_find_rx_state(struct ixgbevf_ipsec *ipsec,
 static int ixgbevf_ipsec_parse_proto_keys(struct xfrm_state *xs,
 					  u32 *mykey, u32 *mysalt)
 {
-	struct net_device *dev = xs->xso.dev;
+	struct net_device *dev = xs->xso.real_dev;
 	unsigned char *key_data;
 	char *alg_name = NULL;
 	int key_len;
@@ -260,12 +260,15 @@ static int ixgbevf_ipsec_parse_proto_keys(struct xfrm_state *xs,
  **/
 static int ixgbevf_ipsec_add_sa(struct xfrm_state *xs)
 {
-	struct net_device *dev = xs->xso.dev;
-	struct ixgbevf_adapter *adapter = netdev_priv(dev);
-	struct ixgbevf_ipsec *ipsec = adapter->ipsec;
+	struct net_device *dev = xs->xso.real_dev;
+	struct ixgbevf_adapter *adapter;
+	struct ixgbevf_ipsec *ipsec;
 	u16 sa_idx;
 	int ret;
 
+	adapter = netdev_priv(dev);
+	ipsec = adapter->ipsec;
+
 	if (xs->id.proto != IPPROTO_ESP && xs->id.proto != IPPROTO_AH) {
 		netdev_err(dev, "Unsupported protocol 0x%04x for IPsec offload\n",
 			   xs->id.proto);
@@ -383,11 +386,14 @@ static int ixgbevf_ipsec_add_sa(struct xfrm_state *xs)
  **/
 static void ixgbevf_ipsec_del_sa(struct xfrm_state *xs)
 {
-	struct net_device *dev = xs->xso.dev;
-	struct ixgbevf_adapter *adapter = netdev_priv(dev);
-	struct ixgbevf_ipsec *ipsec = adapter->ipsec;
+	struct net_device *dev = xs->xso.real_dev;
+	struct ixgbevf_adapter *adapter;
+	struct ixgbevf_ipsec *ipsec;
 	u16 sa_idx;
 
+	adapter = netdev_priv(dev);
+	ipsec = adapter->ipsec;
+
 	if (xs->xso.flags & XFRM_OFFLOAD_INBOUND) {
 		sa_idx = xs->xso.offload_handle - IXGBE_IPSEC_BASE_RX_INDEX;
 
diff --git a/drivers/net/netdevsim/ipsec.c b/drivers/net/netdevsim/ipsec.c
index 3811f1bde84e..b80ed2ffd45e 100644
--- a/drivers/net/netdevsim/ipsec.c
+++ b/drivers/net/netdevsim/ipsec.c
@@ -85,7 +85,7 @@ static int nsim_ipsec_parse_proto_keys(struct xfrm_state *xs,
 				       u32 *mykey, u32 *mysalt)
 {
 	const char aes_gcm_name[] = "rfc4106(gcm(aes))";
-	struct net_device *dev = xs->xso.dev;
+	struct net_device *dev = xs->xso.real_dev;
 	unsigned char *key_data;
 	char *alg_name = NULL;
 	int key_len;
@@ -134,7 +134,7 @@ static int nsim_ipsec_add_sa(struct xfrm_state *xs)
 	u16 sa_idx;
 	int ret;
 
-	dev = xs->xso.dev;
+	dev = xs->xso.real_dev;
 	ns = netdev_priv(dev);
 	ipsec = &ns->ipsec;
 
@@ -194,7 +194,7 @@ static int nsim_ipsec_add_sa(struct xfrm_state *xs)
 
 static void nsim_ipsec_del_sa(struct xfrm_state *xs)
 {
-	struct netdevsim *ns = netdev_priv(xs->xso.dev);
+	struct netdevsim *ns = netdev_priv(xs->xso.real_dev);
 	struct nsim_ipsec *ipsec = &ns->ipsec;
 	u16 sa_idx;
 
@@ -211,7 +211,7 @@ static void nsim_ipsec_del_sa(struct xfrm_state *xs)
 
 static bool nsim_ipsec_offload_ok(struct sk_buff *skb, struct xfrm_state *xs)
 {
-	struct netdevsim *ns = netdev_priv(xs->xso.dev);
+	struct netdevsim *ns = netdev_priv(xs->xso.real_dev);
 	struct nsim_ipsec *ipsec = &ns->ipsec;
 
 	ipsec->ok++;
diff --git a/include/net/bonding.h b/include/net/bonding.h
index 15335732e166..625d9c72dee3 100644
--- a/include/net/bonding.h
+++ b/include/net/bonding.h
@@ -201,6 +201,11 @@ struct bond_up_slave {
  */
 #define BOND_LINK_NOCHANGE -1
 
+struct bond_ipsec {
+	struct list_head list;
+	struct xfrm_state *xs;
+};
+
 /*
  * Here are the locking policies for the two bonding locks:
  * Get rcu_read_lock when reading or RTNL when writing slave list.
@@ -249,7 +254,9 @@ struct bonding {
 #endif /* CONFIG_DEBUG_FS */
 	struct rtnl_link_stats64 bond_stats;
 #ifdef CONFIG_XFRM_OFFLOAD
-	struct xfrm_state *xs;
+	struct list_head ipsec_list;
+	/* protecting ipsec_list */
+	spinlock_t ipsec_lock;
 #endif /* CONFIG_XFRM_OFFLOAD */
 };