netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH -net 0/2] bonding: fix arp_validate desync state & race
@ 2013-09-06 19:41 Nikolay Aleksandrov
  2013-09-06 19:41 ` [PATCH -net 1/2] bonding: fix store_arp_validate race with mode change Nikolay Aleksandrov
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Nikolay Aleksandrov @ 2013-09-06 19:41 UTC (permalink / raw)
  To: netdev; +Cc: fubar, davem, andy, mleitner

Hello all,
These two patches aim to fix the possible de-sync state which the bond
can enter if we have arp_validate without arp_interval or the other way
around. They also fix a race condition between arp_validate setting and
mode changing.

Patch 01 - fixes the race condition between store_arp_validate and bond
mode change by using rtnl for sync
Patch 02 - fixes the possible de-sync state by setting/unsetting recv_probe
if arp_interval is set/unset and also if arp_validate is set/unset

Best regards,
 Nikolay Aleksandrov


Nikolay Aleksandrov (2):
  bonding: fix store_arp_validate race with mode change
  bonding: fix bond_arp_rcv setting and arp validate desync state

 drivers/net/bonding/bond_main.c  |  4 ++--
 drivers/net/bonding/bond_sysfs.c | 31 +++++++++++++++++++++++++------
 drivers/net/bonding/bonding.h    |  1 +
 3 files changed, 28 insertions(+), 8 deletions(-)

-- 
1.8.1.4

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH -net 1/2] bonding: fix store_arp_validate race with mode change
  2013-09-06 19:41 [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov
@ 2013-09-06 19:41 ` Nikolay Aleksandrov
  2013-09-06 19:41 ` [PATCH -net 2/2] bonding: fix bond_arp_rcv setting and arp validate desync state Nikolay Aleksandrov
  2013-09-06 20:10 ` [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov
  2 siblings, 0 replies; 4+ messages in thread
From: Nikolay Aleksandrov @ 2013-09-06 19:41 UTC (permalink / raw)
  To: netdev; +Cc: fubar, davem, andy, mleitner

We need to protect store_arp_validate via rtnl because it can race with
mode changing and we can end up having arp_validate set in a mode
different from active-backup.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
---
 drivers/net/bonding/bond_sysfs.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index ce46776..4e38683 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -419,27 +419,33 @@ static ssize_t bonding_store_arp_validate(struct device *d,
 					  struct device_attribute *attr,
 					  const char *buf, size_t count)
 {
-	int new_value;
+	int new_value, ret = count;
 	struct bonding *bond = to_bond(d);
 
+	if (!rtnl_trylock())
+		return restart_syscall();
 	new_value = bond_parse_parm(buf, arp_validate_tbl);
 	if (new_value < 0) {
 		pr_err("%s: Ignoring invalid arp_validate value %s\n",
 		       bond->dev->name, buf);
-		return -EINVAL;
+		ret = -EINVAL;
+		goto out;
 	}
 	if (new_value && (bond->params.mode != BOND_MODE_ACTIVEBACKUP)) {
 		pr_err("%s: arp_validate only supported in active-backup mode.\n",
 		       bond->dev->name);
-		return -EINVAL;
+		ret = -EINVAL;
+		goto out;
 	}
 	pr_info("%s: setting arp_validate to %s (%d).\n",
 		bond->dev->name, arp_validate_tbl[new_value].modename,
 		new_value);
 
 	bond->params.arp_validate = new_value;
+out:
+	rtnl_unlock();
 
-	return count;
+	return ret;
 }
 
 static DEVICE_ATTR(arp_validate, S_IRUGO | S_IWUSR, bonding_show_arp_validate,
-- 
1.8.1.4

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [PATCH -net 2/2] bonding: fix bond_arp_rcv setting and arp validate desync state
  2013-09-06 19:41 [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov
  2013-09-06 19:41 ` [PATCH -net 1/2] bonding: fix store_arp_validate race with mode change Nikolay Aleksandrov
@ 2013-09-06 19:41 ` Nikolay Aleksandrov
  2013-09-06 20:10 ` [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov
  2 siblings, 0 replies; 4+ messages in thread
From: Nikolay Aleksandrov @ 2013-09-06 19:41 UTC (permalink / raw)
  To: netdev; +Cc: fubar, davem, andy, mleitner

We make bond_arp_rcv global so it can be used in bond_sysfs if the bond
interface is up and arp_interval is being changed to a positive value
and cleared otherwise as per Jay's suggestion.
This also fixes a problem where bond_arp_rcv was set even though
arp_validate was disabled while the bond was up by unsetting recv_probe
in bond_store_arp_validate and respectively setting it if enabled.

Signed-off-by: Nikolay Aleksandrov <nikolay@redhat.com>
---
I've intentionally left the prototype line >80 chars, let me know if I
should break it.

 drivers/net/bonding/bond_main.c  |  4 ++--
 drivers/net/bonding/bond_sysfs.c | 19 ++++++++++++++++---
 drivers/net/bonding/bonding.h    |  1 +
 3 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 39e5b1c..72df399 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -2404,8 +2404,8 @@ static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32
 	slave->target_last_arp_rx[i] = jiffies;
 }
 
-static int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
-			struct slave *slave)
+int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
+		 struct slave *slave)
 {
 	struct arphdr *arp = (struct arphdr *)skb->data;
 	unsigned char *arp_ptr;
diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index 4e38683..9afb1c5 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -349,6 +349,8 @@ static ssize_t bonding_store_mode(struct device *d,
 		goto out;
 	}
 
+	/* don't cache arp_validate between modes */
+	bond->params.arp_validate = BOND_ARP_VALIDATE_NONE;
 	bond->params.mode = new_value;
 	bond_set_mode_ops(bond, bond->params.mode);
 	pr_info("%s: setting mode to %s (%d).\n",
@@ -419,8 +421,8 @@ static ssize_t bonding_store_arp_validate(struct device *d,
 					  struct device_attribute *attr,
 					  const char *buf, size_t count)
 {
-	int new_value, ret = count;
 	struct bonding *bond = to_bond(d);
+	int new_value, ret = count;
 
 	if (!rtnl_trylock())
 		return restart_syscall();
@@ -431,7 +433,7 @@ static ssize_t bonding_store_arp_validate(struct device *d,
 		ret = -EINVAL;
 		goto out;
 	}
-	if (new_value && (bond->params.mode != BOND_MODE_ACTIVEBACKUP)) {
+	if (new_value && bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
 		pr_err("%s: arp_validate only supported in active-backup mode.\n",
 		       bond->dev->name);
 		ret = -EINVAL;
@@ -441,6 +443,12 @@ static ssize_t bonding_store_arp_validate(struct device *d,
 		bond->dev->name, arp_validate_tbl[new_value].modename,
 		new_value);
 
+	if (bond->dev->flags & IFF_UP) {
+		if (!new_value)
+			bond->recv_probe = NULL;
+		else if (bond->params.arp_interval)
+			bond->recv_probe = bond_arp_rcv;
+	}
 	bond->params.arp_validate = new_value;
 out:
 	rtnl_unlock();
@@ -561,8 +569,8 @@ static ssize_t bonding_store_arp_interval(struct device *d,
 					  struct device_attribute *attr,
 					  const char *buf, size_t count)
 {
-	int new_value, ret = count;
 	struct bonding *bond = to_bond(d);
+	int new_value, ret = count;
 
 	if (!rtnl_trylock())
 		return restart_syscall();
@@ -605,8 +613,13 @@ static ssize_t bonding_store_arp_interval(struct device *d,
 		 * is called.
 		 */
 		if (!new_value) {
+			if (bond->params.arp_validate)
+				bond->recv_probe = NULL;
 			cancel_delayed_work_sync(&bond->arp_work);
 		} else {
+			/* arp_validate can be set only in active-backup mode */
+			if (bond->params.arp_validate)
+				bond->recv_probe = bond_arp_rcv;
 			cancel_delayed_work_sync(&bond->mii_work);
 			queue_delayed_work(bond->wq, &bond->arp_work, 0);
 		}
diff --git a/drivers/net/bonding/bonding.h b/drivers/net/bonding/bonding.h
index f7ab161..7ad8bd5 100644
--- a/drivers/net/bonding/bonding.h
+++ b/drivers/net/bonding/bonding.h
@@ -430,6 +430,7 @@ static inline bool slave_can_tx(struct slave *slave)
 
 struct bond_net;
 
+int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave);
 struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr);
 int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev);
 void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id);
-- 
1.8.1.4

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH -net 0/2] bonding: fix arp_validate desync state & race
  2013-09-06 19:41 [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov
  2013-09-06 19:41 ` [PATCH -net 1/2] bonding: fix store_arp_validate race with mode change Nikolay Aleksandrov
  2013-09-06 19:41 ` [PATCH -net 2/2] bonding: fix bond_arp_rcv setting and arp validate desync state Nikolay Aleksandrov
@ 2013-09-06 20:10 ` Nikolay Aleksandrov
  2 siblings, 0 replies; 4+ messages in thread
From: Nikolay Aleksandrov @ 2013-09-06 20:10 UTC (permalink / raw)
  To: netdev; +Cc: fubar, davem, andy, mleitner

On 09/06/2013 09:41 PM, Nikolay Aleksandrov wrote:
> Hello all,
> These two patches aim to fix the possible de-sync state which the bond
> can enter if we have arp_validate without arp_interval or the other way
> around. They also fix a race condition between arp_validate setting and
> mode changing.
> 
> Patch 01 - fixes the race condition between store_arp_validate and bond
> mode change by using rtnl for sync
> Patch 02 - fixes the possible de-sync state by setting/unsetting recv_probe
> if arp_interval is set/unset and also if arp_validate is set/unset
> 
> Best regards,
>  Nikolay Aleksandrov
> 
> 
> Nikolay Aleksandrov (2):
>   bonding: fix store_arp_validate race with mode change
>   bonding: fix bond_arp_rcv setting and arp validate desync state
> 
>  drivers/net/bonding/bond_main.c  |  4 ++--
>  drivers/net/bonding/bond_sysfs.c | 31 +++++++++++++++++++++++++------
>  drivers/net/bonding/bonding.h    |  1 +
>  3 files changed, 28 insertions(+), 8 deletions(-)
> 
Ugh, self-nack I just noticed the mode check in store_arp_validate passes
if new_value is 0.

I'll wait some time to see if there isn't anything else and re-post.

Cheers,
 Nik

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2013-09-06 20:11 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-09-06 19:41 [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov
2013-09-06 19:41 ` [PATCH -net 1/2] bonding: fix store_arp_validate race with mode change Nikolay Aleksandrov
2013-09-06 19:41 ` [PATCH -net 2/2] bonding: fix bond_arp_rcv setting and arp validate desync state Nikolay Aleksandrov
2013-09-06 20:10 ` [PATCH -net 0/2] bonding: fix arp_validate desync state & race Nikolay Aleksandrov

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).