Skip to contentSkip to conversation

Staging: batman-adv for 2.6.35

PATCH New Sven Eckelmann 1 participant 5 messages May 22, 2010 Staging

lore
Share & tools
lore

Sven Eckelmann

author
#0 lore raw

cover Staging: batman-adv for 2.6.35

Hi,

here are the patches which fixes different bugs backported for 2.6.35

Marek Lindner (1):
Staging: batman-adv: fix rogue packets on shutdown

Sven Eckelmann (3):
Staging: batman-adv: Call unregister_netdev on failures to get rtnl lock
Staging: batman-adv: Don't call free_netdev twice
Staging: batman-adv: Don't allocate icmp packet with GFP_KERNEL

thanks,
Sven

author
#1 lore raw

PATCH 1/4 Staging: batman-adv: fix rogue packets on shutdown

From: Marek Lindner <[email protected]>

On module shutdown batman-adv would purge the internal packet queue by sending all remaining packets which could confuse other nodes. Now, the packets are silently discarded.

Signed-off-by Marek Lindner <[email protected]>
Signed-off-by Sven Eckelmann <[email protected]>
---
 drivers/staging/batman-adv/send.c |   24 +++++++++++++++---------
 1 files changed, 15 insertions(+), 9 deletions(-)
drivers/staging/batman-adv/send.c +15 -9
index d8536e2..ac69ed8 100644
@@ -440,6 +440,9 @@ void send_outstanding_bcast_packet(struct work_struct *work)
 	hlist_del(&forw_packet->list);
 	spin_unlock_irqrestore(&forw_bcast_list_lock, flags);
 
+	if (atomic_read(&module_state) == MODULE_DEACTIVATING)
+		goto out;
+
 	/* rebroadcast packet */
 	rcu_read_lock();
 	list_for_each_entry_rcu(batman_if, &if_list, list) {

@@ -453,15 +456,15 @@ void send_outstanding_bcast_packet(struct work_struct *work)
 
 	forw_packet->num_packets++;
 
-	/* if we still have some more bcasts to send and we are not shutting
-	 * down */
-	if ((forw_packet->num_packets < 3) &&
-	    (atomic_read(&module_state) != MODULE_DEACTIVATING))
+	/* if we still have some more bcasts to send */
+	if (forw_packet->num_packets < 3) {
 		_add_bcast_packet_to_list(forw_packet, ((5 * HZ) / 1000));
-	else {
-		forw_packet_free(forw_packet);
-		atomic_inc(&bcast_queue_left);
+		return;
 	}
+
+out:
+	forw_packet_free(forw_packet);
+	atomic_inc(&bcast_queue_left);
 }
 
 void send_outstanding_bat_packet(struct work_struct *work)

@@ -476,6 +479,9 @@ void send_outstanding_bat_packet(struct work_struct *work)
 	hlist_del(&forw_packet->list);
 	spin_unlock_irqrestore(&forw_bat_list_lock, flags);
 
+	if (atomic_read(&module_state) == MODULE_DEACTIVATING)
+		goto out;
+
 	send_packet(forw_packet);
 
 	/**

@@ -483,10 +489,10 @@ void send_outstanding_bat_packet(struct work_struct *work)
 	 * to determine the queues wake up time unless we are
 	 * shutting down
 	 */
-	if ((forw_packet->own) &&
-	    (atomic_read(&module_state) != MODULE_DEACTIVATING))
+	if (forw_packet->own)
 		schedule_own_packet(forw_packet->if_incoming);
 
+out:
 	/* don't count own packet */
 	if (!forw_packet->own)
 		atomic_inc(&batman_queue_left);
Signature · 1 line
1.7.1
author
#2 lore raw

PATCH 2/4 Staging: batman-adv: Call unregister_netdev on failures to get rtnl lock

In reply to Sven Eckelmann

We must call unregister_netdev when we couldn't initialise the batman-adv module and the soft_device was registered. There are two version of the function which we can use:
* unregister_netdevice - removes device
* unregister_netdev - takes rtnl semaphore and remove device

We don't hold the semaphore in an error situation. So we must use unregister_netdev.

Signed-off-by Sven Eckelmann <[email protected]>
---
 drivers/staging/batman-adv/main.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/staging/batman-adv/main.c +1 -1
index 9d13979..50a66b4 100644
@@ -127,7 +127,7 @@ int init_module(void)
 	return 0;
 
 unreg_soft_device:
-	unregister_netdevice(soft_device);
+	unregister_netdev(soft_device);
 free_soft_device:
 	free_netdev(soft_device);
 	soft_device = NULL;
Signature · 1 line
1.7.1
author
#3 lore raw

PATCH 3/4 Staging: batman-adv: Don't call free_netdev twice

In reply to Sven Eckelmann

Free_netdev is registered as destructor in interface_setup for every soft_device. This destructor is automatically called from unregister_netdev and we must not call it again for the freed net_device.

Signed-off-by Sven Eckelmann <[email protected]>
---
 drivers/staging/batman-adv/main.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)
drivers/staging/batman-adv/main.c +3 -0
index 50a66b4..74c70d5 100644
@@ -128,6 +128,9 @@ int init_module(void)
 
 unreg_soft_device:
 	unregister_netdev(soft_device);
+	soft_device = NULL;
+	return -ENOMEM;
+
 free_soft_device:
 	free_netdev(soft_device);
 	soft_device = NULL;
Signature · 1 line
1.7.1
author
#4 lore raw

PATCH 4/4 Staging: batman-adv: Don't allocate icmp packet with GFP_KERNEL

In reply to Sven Eckelmann

A new buffer for a packet is created when a icmp packet is received. This happens in a context with disabled irq. Thus we are not allowed to sleep or call function which might sleep. kmalloc must be called with GFP_ATOMIC instead of GFP_KERNEL to ensure that it does not sleep.

Signed-off-by Sven Eckelmann <[email protected]>
---
 drivers/staging/batman-adv/device.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/staging/batman-adv/device.c +1 -1
index ad82ec4..7eb6559 100644
@@ -309,7 +309,7 @@ void bat_device_add_packet(struct device_client *device_client,
 	struct device_packet *device_packet;
 	unsigned long flags;
 
-	device_packet = kmalloc(sizeof(struct device_packet), GFP_KERNEL);
+	device_packet = kmalloc(sizeof(struct device_packet), GFP_ATOMIC);
 
 	if (!device_packet)
 		return;
Signature · 1 line
1.7.1