All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] net: Handle NETREG_UNINITIALIZED devices correctly
@ 2009-12-09  8:26 Krishna Kumar
  2009-12-09  9:57 ` Eric Dumazet
  0 siblings, 1 reply; 3+ messages in thread
From: Krishna Kumar @ 2009-12-09  8:26 UTC (permalink / raw)
  To: davem, eric.dumazet, herbert; +Cc: netdev, Krishna Kumar

From: Krishna Kumar <krkumar2@in.ibm.com>

Fix two problems:

1. If unregister_netdevice_many() is called with both registered
   and unregistered devices, rollback_registered_many() bails out
   when it reaches the first unregistered device. The processing
   of the prior registered devices is unfinished, and the
   remaining devices are skipped, and possible registered netdev's
   are leaked/unregistered.

2. System hangs or panics depending on how the devices are passed,
   since when netdev_run_todo() runs, some devices were not fully
   processed.

Tested by passing intermingled unregistered and registered vlan
devices to unregister_netdevice_many() as follows:
	1. dev, fake_dev1, fake_dev2: hangs in run_todo
	   ("unregister_netdevice: waiting for eth1.100 to become
	    free. Usage count = 1")
	2. fake_dev1, dev, fake_dev2: failure during de-registration
	   and next registration, followed by a vlan driver Oops
	   during subsequent registration.

Confirmed that the patch fixes both cases.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
---
 net/core/dev.c |   10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

diff -ruNp org/net/core/dev.c new/net/core/dev.c
--- org/net/core/dev.c	2009-12-04 18:46:13.000000000 +0530
+++ new/net/core/dev.c	2009-12-05 10:23:25.000000000 +0530
@@ -4771,21 +4771,23 @@ static void net_set_todo(struct net_devi
 
 static void rollback_registered_many(struct list_head *head)
 {
-	struct net_device *dev;
+	struct net_device *dev, *tmp;
 
 	BUG_ON(dev_boot_phase);
 	ASSERT_RTNL();
 
-	list_for_each_entry(dev, head, unreg_list) {
+	list_for_each_entry_safe(dev, tmp, head, unreg_list) {
 		/* Some devices call without registering
-		 * for initialization unwind.
+		 * for initialization unwind. Remove those
+		 * devices and proceed with the remaining.
 		 */
 		if (dev->reg_state == NETREG_UNINITIALIZED) {
 			pr_debug("unregister_netdevice: device %s/%p never "
 				 "was registered\n", dev->name, dev);
 
 			WARN_ON(1);
-			return;
+			list_del(&dev->unreg_list);
+			continue;
 		}
 
 		BUG_ON(dev->reg_state != NETREG_REGISTERED);

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] net: Handle NETREG_UNINITIALIZED devices correctly
  2009-12-09  8:26 [PATCH] net: Handle NETREG_UNINITIALIZED devices correctly Krishna Kumar
@ 2009-12-09  9:57 ` Eric Dumazet
  2009-12-11 23:12   ` David Miller
  0 siblings, 1 reply; 3+ messages in thread
From: Eric Dumazet @ 2009-12-09  9:57 UTC (permalink / raw)
  To: Krishna Kumar; +Cc: davem, herbert, netdev

Krishna Kumar a écrit :
> From: Krishna Kumar <krkumar2@in.ibm.com>
> 
> Fix two problems:
> 
> 1. If unregister_netdevice_many() is called with both registered
>    and unregistered devices, rollback_registered_many() bails out
>    when it reaches the first unregistered device. The processing
>    of the prior registered devices is unfinished, and the
>    remaining devices are skipped, and possible registered netdev's
>    are leaked/unregistered.
> 
> 2. System hangs or panics depending on how the devices are passed,
>    since when netdev_run_todo() runs, some devices were not fully
>    processed.
> 
> Tested by passing intermingled unregistered and registered vlan
> devices to unregister_netdevice_many() as follows:
> 	1. dev, fake_dev1, fake_dev2: hangs in run_todo
> 	   ("unregister_netdevice: waiting for eth1.100 to become
> 	    free. Usage count = 1")
> 	2. fake_dev1, dev, fake_dev2: failure during de-registration
> 	   and next registration, followed by a vlan driver Oops
> 	   during subsequent registration.
> 
> Confirmed that the patch fixes both cases.
> 
> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>

Acked-by: Eric Dumazet <eric.dumazet@gmail.com>

Thanks !

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] net: Handle NETREG_UNINITIALIZED devices correctly
  2009-12-09  9:57 ` Eric Dumazet
@ 2009-12-11 23:12   ` David Miller
  0 siblings, 0 replies; 3+ messages in thread
From: David Miller @ 2009-12-11 23:12 UTC (permalink / raw)
  To: eric.dumazet; +Cc: krkumar2, herbert, netdev

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Wed, 09 Dec 2009 10:57:44 +0100

> Krishna Kumar a ^[$(D+1^[(Bcrit :
>> From: Krishna Kumar <krkumar2@in.ibm.com>
>> 
>> Fix two problems:
>> 
>> 1. If unregister_netdevice_many() is called with both registered
>>    and unregistered devices, rollback_registered_many() bails out
>>    when it reaches the first unregistered device. The processing
>>    of the prior registered devices is unfinished, and the
>>    remaining devices are skipped, and possible registered netdev's
>>    are leaked/unregistered.
>> 
>> 2. System hangs or panics depending on how the devices are passed,
>>    since when netdev_run_todo() runs, some devices were not fully
>>    processed.
>> 
>> Tested by passing intermingled unregistered and registered vlan
>> devices to unregister_netdevice_many() as follows:
>> 	1. dev, fake_dev1, fake_dev2: hangs in run_todo
>> 	   ("unregister_netdevice: waiting for eth1.100 to become
>> 	    free. Usage count = 1")
>> 	2. fake_dev1, dev, fake_dev2: failure during de-registration
>> 	   and next registration, followed by a vlan driver Oops
>> 	   during subsequent registration.
>> 
>> Confirmed that the patch fixes both cases.
>> 
>> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
> 
> Acked-by: Eric Dumazet <eric.dumazet@gmail.com>

Nice work Krishna, applied, thanks!

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2009-12-11 23:12 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-12-09  8:26 [PATCH] net: Handle NETREG_UNINITIALIZED devices correctly Krishna Kumar
2009-12-09  9:57 ` Eric Dumazet
2009-12-11 23:12   ` David Miller

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.