Skip to contentSkip to conversation

net: Fix userspace RTM_NEWLINK notifications.

PATCH Applied Eric W. Biederman 3 participants 5 messages Dec 13, 2009 – Dec 14, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net: Fix userspace RTM_NEWLINK notifications.

I received some bug reports about userspace programs having problems because after RTM_NEWLINK was received they could not immediate access files under /proc/sys/net/ because they had not been registered yet.

The original problem was trivially fixed by moving the userspace notification from rtnetlink_event() to the end of register_netdevice().

When testing that change I discovered I was still getting RTM_NEWLINK events before I could access proc and I was also getting RTM_NEWLINK events after I was seeing RTM_DELLINK. Things practically guaranteed to confuse userspace.

After a little more investigation these extra notifications proved to be from the new notifiers NETDEV_POST_INIT and NETDEV_UNREGISTER_BATCH hitting the default case in rtnetlink_event, and triggering unnecessary RTM_NEWLINK messages.

rtnetlink_event now explicitly handles NETDEV_UNREGISTER_BATCH and NETDEV_POST_INIT to avoid sending the incorrect userspace notifications.

---
 net/core/dev.c       |   11 +++++++++++
 net/core/rtnetlink.c |    6 +++---
 2 files changed, 14 insertions(+), 3 deletions(-)
net/core/dev.c +11 -0
index c36a17a..369e83a 100644
@@ -5033,6 +5033,11 @@ int register_netdevice(struct net_device *dev)
 		rollback_registered(dev);
 		dev->reg_state = NETREG_UNREGISTERED;
 	}
+	/*
+	 *	Prevent userspace races by waiting until the network
+	 *	device is fully setup before sending notifications.
+	 */
+	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
 
 out:
 	return ret;
@@ -5595,6 +5600,12 @@ int dev_change_net_namespace(struct net_device *dev, struct net *net, const char
 	/* Notify protocols, that a new device appeared. */
 	call_netdevice_notifiers(NETDEV_REGISTER, dev);
 
+	/*
+	 *	Prevent userspace races by waiting until the network
+	 *	device is fully setup before sending notifications.
+	 */
+	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
+
 	synchronize_net();
 	err = 0;
 out:
net/core/rtnetlink.c +3 -3
index 33148a5..794bcb8 100644
@@ -1364,15 +1364,15 @@ static int rtnetlink_event(struct notifier_block *this, unsigned long event, voi
 	case NETDEV_UNREGISTER:
 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
 		break;
-	case NETDEV_REGISTER:
-		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
-		break;
 	case NETDEV_UP:
 	case NETDEV_DOWN:
 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
 		break;
+	case NETDEV_POST_INIT:
+	case NETDEV_REGISTER:
 	case NETDEV_CHANGE:
 	case NETDEV_GOING_DOWN:
+	case NETDEV_UNREGISTER_BATCH:
 		break;
 	default:
 		rtmsg_ifinfo(RTM_NEWLINK, dev, 0);
Signature · 1 line
1.6.5.2.143.g8cc62
#1 lore raw

Hi Eric,

Quoted 13 lines

I received some bug reports about userspace programs having problems because after RTM_NEWLINK was received they could not immediate access files under /proc/sys/net/ because they had not been registered yet.

The original problem was trivially fixed by moving the userspace notification from rtnetlink_event() to the end of register_netdevice().

When testing that change I discovered I was still getting RTM_NEWLINK events before I could access proc and I was also getting RTM_NEWLINK events after I was seeing RTM_DELLINK. Things practically guaranteed to confuse userspace.

Quoted 8 lines

After a little more investigation these extra notifications proved to be from the new notifiers NETDEV_POST_INIT and NETDEV_UNREGISTER_BATCH hitting the default case in rtnetlink_event, and triggering unnecessary RTM_NEWLINK messages.

rtnetlink_event now explicitly handles NETDEV_UNREGISTER_BATCH and NETDEV_POST_INIT to avoid sending the incorrect userspace notifications.

the POST_INIT and UNREGISTER_BATCH are not 2.6.32 features. They are 2.6.33 features and thus you need to create a separate patch for -stable if you wanna fix the /proc access issue.

Regards

Marcel

maintainer
#2 lore raw
In reply to Marcel Holtmann

From: Marcel Holtmann <[email protected]>
Date: Mon, 14 Dec 2009 03:05:36 +0100

Quoted 3 lines

the POST_INIT and UNREGISTER_BATCH are not 2.6.32 features. They are 2.6.33 features and thus you need to create a separate patch for -stable if you wanna fix the /proc access issue.

Eric, please give me a suitable backport for -stable.

Thanks.

#3 lore raw

PATCH 1/1 net: Fix userspace RTM_NEWLINK notifications.

I received some bug reports about userspace programs having problems because after RTM_NEWLINK was received they could not immeidate access files under /proc/sys/net/ because they had not been registered yet.

The problem was trivailly fixed by moving the userspace notification from rtnetlink_event to the end of register_netdevice.

---

This patch is against 2.6.32 and should work back to 2.6.29 and probably earlier.

 net/core/dev.c       |   11 +++++++++++
 net/core/rtnetlink.c |    4 +---
 2 files changed, 12 insertions(+), 3 deletions(-)
net/core/dev.c +11 -0
index fe10551..584046e 100644
@@ -4860,6 +4860,11 @@ int register_netdevice(struct net_device *dev)
 		rollback_registered(dev);
 		dev->reg_state = NETREG_UNREGISTERED;
 	}
+	/*
+	 *	Prevent userspace races by waiting until the network
+	 *	device is fully setup before sending notifications.
+	 */
+	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
 
 out:
 	return ret;
@@ -5398,6 +5403,12 @@ int dev_change_net_namespace(struct net_device *dev, struct net *net, const char
 	/* Notify protocols, that a new device appeared. */
 	call_netdevice_notifiers(NETDEV_REGISTER, dev);
 
+	/*
+	 *	Prevent userspace races by waiting until the network
+	 *	device is fully setup before sending notifications.
+	 */
+	rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
+
 	synchronize_net();
 	err = 0;
 out:
net/core/rtnetlink.c +1 -3
index eb42873..d4fd895 100644
@@ -1334,13 +1334,11 @@ static int rtnetlink_event(struct notifier_block *this, unsigned long event, voi
 	case NETDEV_UNREGISTER:
 		rtmsg_ifinfo(RTM_DELLINK, dev, ~0U);
 		break;
-	case NETDEV_REGISTER:
-		rtmsg_ifinfo(RTM_NEWLINK, dev, ~0U);
-		break;
 	case NETDEV_UP:
 	case NETDEV_DOWN:
 		rtmsg_ifinfo(RTM_NEWLINK, dev, IFF_UP|IFF_RUNNING);
 		break;
+	case NETDEV_REGISTER:
 	case NETDEV_CHANGE:
 	case NETDEV_GOING_DOWN:
 		break;
Signature · 1 line
1.6.5.2.143.g8cc62