From: Herbert Xu <herbert.xu@redhat.com>
Date: Thu, 14 Aug 2008 20:22:21 +1000
Subject: [xen] xennet: coordinate ARP with backend network status
Message-id: 20080814102221.GA1300@gondor.apana.org.au
O-Subject: [RHEL5.3 PATCH] xennet: Coordinate gratuitous ARP with backend network status
Bugzilla: 458934
RH-Acked-by: Don Dutile <ddutile@redhat.com>
RH-Acked-by: Rik van Riel <riel@redhat.com>
RH-Acked-by: David S. Miller <davem@redhat.com>

Hi:

RHEL5.3 BZ 458934

xennet: Coordinate gratuitous ARP with backend network status

This patch delays the sending of the gratuitous ARP packet in
the Xen frontend network driver until the backend signals that
its carrier status has been processed by the stack.

The frontend part of the patch is not applicable in upstream
Linux as there is no gratuitous ARP packet support there at all.
While the backend upstream is dead as far as I know.  So this
is RHEL5-specific.

Both patches are needed to ensure the successful transmission
of the gratuitous ARP packet which is necessary for Xen live
migrations.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

diff --git a/drivers/xen/netback/common.h b/drivers/xen/netback/common.h
index 54a4095..75d82b1 100644
--- a/drivers/xen/netback/common.h
+++ b/drivers/xen/netback/common.h
@@ -98,6 +98,7 @@ typedef struct netif_st {
 	/* Miscellaneous private stuff. */
 	struct list_head list;  /* scheduling list */
 	atomic_t         refcnt;
+	struct xenbus_device *xendev;
 	struct net_device *dev;
 	struct net_device_stats stats;
 
@@ -121,6 +122,7 @@ int netif_map(netif_t *netif, unsigned long tx_ring_ref,
 	} while (0)
 
 void netif_xenbus_init(void);
+void netif_interfaces_init(void);
 
 #define netif_schedulable(dev) (netif_running(dev) && netif_carrier_ok(dev))
 
diff --git a/drivers/xen/netback/interface.c b/drivers/xen/netback/interface.c
index 646c737..04f5796 100644
--- a/drivers/xen/netback/interface.c
+++ b/drivers/xen/netback/interface.c
@@ -33,6 +33,7 @@
 #include "common.h"
 #include <linux/ethtool.h>
 #include <linux/rtnetlink.h>
+#include <xen/xenbus.h>
 
 /*
  * Module parameter 'queue_length':
@@ -347,3 +348,31 @@ void netif_disconnect(netif_t *netif)
 
 	free_netdev(netif->dev);
 }
+
+
+static int
+netdev_notify(struct notifier_block *this, unsigned long event, void *ptr)
+{
+	struct net_device *dev = ptr;
+
+	/* Carrier up event and is it one of our devices? */
+	if (event == NETDEV_CHANGE && netif_carrier_ok(dev) &&
+	    dev->open == net_open) {
+		netif_t *netif = netdev_priv(dev);
+
+		xenbus_switch_state(netif->xendev, XenbusStateConnected);
+	}
+
+	return NOTIFY_DONE;
+}
+
+
+static struct notifier_block notifier_netdev = {
+	.notifier_call  = netdev_notify,
+};
+
+
+void netif_interfaces_init(void)
+{
+	(void)register_netdevice_notifier(&notifier_netdev);
+}
diff --git a/drivers/xen/netback/netback.c b/drivers/xen/netback/netback.c
index 67900c9..bf2c6e7 100644
--- a/drivers/xen/netback/netback.c
+++ b/drivers/xen/netback/netback.c
@@ -1504,6 +1504,7 @@ static int __init netback_init(void)
 	spin_lock_init(&net_schedule_list_lock);
 	INIT_LIST_HEAD(&net_schedule_list);
 
+	netif_interfaces_init();
 	netif_xenbus_init();
 
 #ifdef NETBE_DEBUG_INTERRUPT
diff --git a/drivers/xen/netback/xenbus.c b/drivers/xen/netback/xenbus.c
index b81042d..d15ce95 100644
--- a/drivers/xen/netback/xenbus.c
+++ b/drivers/xen/netback/xenbus.c
@@ -186,6 +186,7 @@ static void backend_create_netif(struct backend_info *be)
 		return;
 	}
 
+	be->netif->xendev = dev;
 	kobject_uevent(&dev->dev.kobj, KOBJ_ONLINE);
 }
 
@@ -310,10 +311,6 @@ static void connect(struct backend_info *be)
 	int err;
 	struct xenbus_device *dev = be->dev;
 
-	err = connect_rings(be);
-	if (err)
-		return;
-
 	err = xen_net_read_mac(dev, be->netif->fe_dev_addr);
 	if (err) {
 		xenbus_dev_fatal(dev, err, "parsing %s/mac", dev->nodename);
@@ -324,7 +321,9 @@ static void connect(struct backend_info *be)
 			  &be->netif->credit_usec);
 	be->netif->remaining_credit = be->netif->credit_bytes;
 
-	xenbus_switch_state(dev, XenbusStateConnected);
+	err = connect_rings(be);
+	if (err)
+		return;
 
 	/* May not get a kick from the frontend, so start the tx_queue now. */
 	if (!netbk_can_queue(be->netif->dev))
diff --git a/drivers/xen/netfront/netfront.c b/drivers/xen/netfront/netfront.c
index 6310448..41d0178 100644
--- a/drivers/xen/netfront/netfront.c
+++ b/drivers/xen/netfront/netfront.c
@@ -542,18 +542,20 @@ static void backend_changed(struct xenbus_device *dev,
 	switch (backend_state) {
 	case XenbusStateInitialising:
 	case XenbusStateInitialised:
-	case XenbusStateConnected:
 	case XenbusStateUnknown:
 	case XenbusStateClosed:
 		break;
 
+	case XenbusStateConnected:
+		(void)send_fake_arp(netdev);
+		break;
+
 	case XenbusStateInitWait:
 		if (network_connect(netdev) != 0) {
 			netif_free(np);
 			break;
 		}
 		xenbus_switch_state(dev, XenbusStateConnected);
-		(void)send_fake_arp(netdev);
 		break;
 
 	case XenbusStateClosing: