Skip to contentSkip to conversation

bonding: fix bond_arp_rcv() race of curr_active_slave

PATCH In review Veaceslav Falico 3 participants 4 messages Feb 20, 2014

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bonding: fix bond_arp_rcv() race of curr_active_slave

bond->curr_active_slave can be changed between its deferences, even to NULL, and thus we might panic.

We're always holding the rcu (rx_handler->bond_handle_frame()->bond_arp_rcv()) so fix this by rcu_dereferencing() it and using the saved.

Fixes aeea64a ("bonding: don't trust arp requests unless active slave really works")
CC Jay Vosburgh <[email protected]>
CC Andy Gospodarek <[email protected]>
---
 drivers/net/bonding/bond_main.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/net/bonding/bond_main.c +7 -3
index 71edf03..bd70bbc 100644
@@ -2254,6 +2254,7 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 		 struct slave *slave)
 {
 	struct arphdr *arp = (struct arphdr *)skb->data;
+	struct slave *curr_active_slave;
 	unsigned char *arp_ptr;
 	__be32 sip, tip;
 	int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
@@ -2299,6 +2300,8 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 		 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
 		 &sip, &tip);
 
+	curr_active_slave = rcu_dereference(bond->curr_active_slave);
+
 	/*
 	 * Backup slaves won't see the ARP reply, but do come through
 	 * here for each ARP probe (so we swap the sip/tip to validate
@@ -2312,11 +2315,12 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 	 * is done to avoid endless looping when we can't reach the
 	 * arp_ip_target and fool ourselves with our own arp requests.
 	 */
+
 	if (bond_is_active_slave(slave))
 		bond_validate_arp(bond, slave, sip, tip);
-	else if (bond->curr_active_slave &&
-		 time_after(slave_last_rx(bond, bond->curr_active_slave),
-			    bond->curr_active_slave->last_link_up))
+	else if (curr_active_slave &&
+		 time_after(slave_last_rx(bond, curr_active_slave),
+			    curr_active_slave->last_link_up))
 		bond_validate_arp(bond, slave, tip, sip);
 
 out_unlock:
Signature · 1 line
1.8.4
#1 lore raw
Quoted 11 lines · Veaceslav Falico

bond->curr_active_slave can be changed between its deferences, even to NULL, and thus we might panic.

We're always holding the rcu (rx_handler->bond_handle_frame()->bond_arp_rcv()) so fix this by rcu_dereferencing() it and using the saved.

Fixes aeea64a ("bonding: don't trust arp requests unless active slave really works")
CC Jay Vosburgh <[email protected]>
CC Andy Gospodarek <[email protected]>

Sorry David, it should have been net-next (I've based it on the net-next tree).

Can you apply it to net-next or should I resend it?

Sorry for the noise.

Quoted 44 lines

---
drivers/net/bonding/bond_main.c | 10 +++++++---

1 file changed, 7 insertions(+), 3 deletions(-)
drivers/net/bonding/bond_main.c +7 -3
index 71edf03..bd70bbc 100644
@@ -2254,6 +2254,7 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 		 struct slave *slave)
 {
 	struct arphdr *arp = (struct arphdr *)skb->data;
+	struct slave *curr_active_slave;
 	unsigned char *arp_ptr;
 	__be32 sip, tip;
 	int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
@@ -2299,6 +2300,8 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 		 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
 		 &sip, &tip);
 
+	curr_active_slave = rcu_dereference(bond->curr_active_slave);
+
 	/*
 	 * Backup slaves won't see the ARP reply, but do come through
 	 * here for each ARP probe (so we swap the sip/tip to validate
@@ -2312,11 +2315,12 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 	 * is done to avoid endless looping when we can't reach the
 	 * arp_ip_target and fool ourselves with our own arp requests.
 	 */
+
 	if (bond_is_active_slave(slave))
 		bond_validate_arp(bond, slave, sip, tip);
-	else if (bond->curr_active_slave &&
-		 time_after(slave_last_rx(bond, bond->curr_active_slave),
-			    bond->curr_active_slave->last_link_up))
+	else if (curr_active_slave &&
+		 time_after(slave_last_rx(bond, curr_active_slave),
+			    curr_active_slave->last_link_up))
 		bond_validate_arp(bond, slave, tip, sip);
 

out_unlock:

Signature · 1 line
1.8.4
maintainer
#2 lore raw

From: Veaceslav Falico <[email protected]>
Date: Thu, 20 Feb 2014 12:15:19 +0100

Quoted 15 lines · Veaceslav Falico
Quoted 7 lines

bond->curr_active_slave can be changed between its deferences, even to NULL, and thus we might panic.

We're always holding the rcu (rx_handler->bond_handle_frame()->bond_arp_rcv()) so fix this by rcu_dereferencing() it and using the saved.

Quoted 6 lines
Fixes aeea64a ("bonding: don't trust arp requests unless active slave

really works")

CC Jay Vosburgh <[email protected]>
CC Andy Gospodarek <[email protected]>
Quoted 7 lines

Sorry David, it should have been net-next (I've based it on the net-next
tree).

Can you apply it to net-next or should I resend it?

Sorry for the noise.

It's fine, applied to net-next, thanks.

Acked-by on v1 1/1
#3 lore raw
In reply to Veaceslav Falico
Quoted 43 lines · Veaceslav Falico

bond->curr_active_slave can be changed between its deferences, even to NULL, and thus we might panic.

We're always holding the rcu (rx_handler->bond_handle_frame()->bond_arp_rcv()) so fix this by rcu_dereferencing() it and using the saved.

Fixes aeea64a ("bonding: don't trust arp requests unless active slave really works")
CC Jay Vosburgh <[email protected]>
CC Andy Gospodarek <[email protected]>
---
 drivers/net/bonding/bond_main.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
drivers/net/bonding/bond_main.c +4 -0
index 71edf03..bd70bbc 100644
@@ -2254,6 +2254,7 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 		 struct slave *slave)
 {
 	struct arphdr *arp = (struct arphdr *)skb->data;
+	struct slave *curr_active_slave;
 	unsigned char *arp_ptr;
 	__be32 sip, tip;
 	int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
@@ -2299,6 +2300,8 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 		 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
 		 &sip, &tip);
 
+	curr_active_slave = rcu_dereference(bond->curr_active_slave);
+
 	/*
 	 * Backup slaves won't see the ARP reply, but do come through
 	 * here for each ARP probe (so we swap the sip/tip to validate
@@ -2312,11 +2315,12 @@ int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
 	 * is done to avoid endless looping when we can't reach the
 	 * arp_ip_target and fool ourselves with our own arp requests.
 	 */
+
 	if (bond_is_active_slave(slave))

drivers/net/bonding/bond_main.c int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, · line 2326

Quoted 10 lines
drivers/net/bonding/bond_main.c +3 -3
 		bond_validate_arp(bond, slave, sip, tip);
-	else if (bond->curr_active_slave &&
-		 time_after(slave_last_rx(bond, bond->curr_active_slave),
-			    bond->curr_active_slave->last_link_up))
+	else if (curr_active_slave &&
+		 time_after(slave_last_rx(bond, curr_active_slave),
+			    curr_active_slave->last_link_up))
 		bond_validate_arp(bond, slave, tip, sip);
 
 out_unlock: