[PATCH 1/2] E1000: Fix ifdown hang in git-2.6.24

[PATCH 1/2] E1000: Fix ifdown hang in git-2.6.24

[Krishna Kumar]

Doing napi_disable twice hangs "ifdown" of the device. e1000_down is the
common place to call napi_disable.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
---
 e1000_main.c |    4 ----
 1 files changed, 4 deletions(-)

diff -ruNp org/drivers/net/e1000/e1000_main.c new/drivers/net/e1000/e1000_main.c
--- org/drivers/net/e1000/e1000_main.c	2007-08-23 13:32:16.000000000 +0530
+++ new/drivers/net/e1000/e1000_main.c	2007-08-23 13:32:34.000000000 +0530
@@ -1477,10 +1477,6 @@ e1000_close(struct net_device *netdev)
 {
 	struct e1000_adapter *adapter = netdev_priv(netdev);
 
-#ifdef CONFIG_E1000_NAPI
-	napi_disable(&adapter->napi);
-#endif
-
 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->flags));
 	e1000_down(adapter);
 	e1000_power_down_phy(adapter);

[PATCH 2/2] [RFC] E1000: Fix hang in netdev_wait_allrefs()

[Krishna Kumar]

After applying patch1, I started getting "waiting for count" messages when
doing ifdown. Not sure if this is the right fix since the count was already
showing as -1 in that message, but this patch fixes the problem.

Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
---
 e1000_main.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletion(-)

diff -ruNp new/drivers/net/e1000/e1000_main.c new2/drivers/net/e1000/e1000_main.c
--- new/drivers/net/e1000/e1000_main.c	2007-08-23 13:32:34.000000000 +0530
+++ new2/drivers/net/e1000/e1000_main.c	2007-08-23 14:28:12.000000000 +0530
@@ -1219,12 +1219,13 @@ e1000_remove(struct pci_dev *pdev)
 	 * would have already happened in close and is redundant. */
 	e1000_release_hw_control(adapter);
 
-	unregister_netdev(netdev);
 #ifdef CONFIG_E1000_NAPI
 	for (i = 0; i < adapter->num_rx_queues; i++)
 		dev_put(&adapter->polling_netdev[i]);
 #endif
 
+	unregister_netdev(netdev);
+
 	if (!e1000_check_phy_reset_block(&adapter->hw))
 		e1000_phy_hw_reset(&adapter->hw);
 

Re: [PATCH 1/2] E1000: Fix ifdown hang in git-2.6.24

[Kok, Auke]

Krishna Kumar wrote:
> Doing napi_disable twice hangs "ifdown" of the device. e1000_down is the
> common place to call napi_disable.
> 
> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
> ---
>  e1000_main.c |    4 ----
>  1 files changed, 4 deletions(-)
> 
> diff -ruNp org/drivers/net/e1000/e1000_main.c new/drivers/net/e1000/e1000_main.c
> --- org/drivers/net/e1000/e1000_main.c	2007-08-23 13:32:16.000000000 +0530
> +++ new/drivers/net/e1000/e1000_main.c	2007-08-23 13:32:34.000000000 +0530
> @@ -1477,10 +1477,6 @@ e1000_close(struct net_device *netdev)
>  {
>  	struct e1000_adapter *adapter = netdev_priv(netdev);
>  
> -#ifdef CONFIG_E1000_NAPI
> -	napi_disable(&adapter->napi);
> -#endif
> -
>  	WARN_ON(test_bit(__E1000_RESETTING, &adapter->flags));
>  	e1000_down(adapter);
>  	e1000_power_down_phy(adapter);

Acked-by: Auke Kok <auke-jan.h.kok@intel.com>

I pushed this change to akpm for -mm as well in e1000e...

Thanks Krishna,

Auke

Re: [PATCH 1/2] E1000: Fix ifdown hang in git-2.6.24

[David Miller]

From: Krishna Kumar <krkumar2@in.ibm.com>
Date: Thu, 23 Aug 2007 14:34:18 +0530

> Doing napi_disable twice hangs "ifdown" of the device. e1000_down is the
> common place to call napi_disable.
> 
> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>

Applied, thanks a lot.

Re: [PATCH 2/2] [RFC] E1000: Fix hang in netdev_wait_allrefs()

[David Miller]

From: Krishna Kumar <krkumar2@in.ibm.com>
Date: Thu, 23 Aug 2007 14:34:31 +0530

> After applying patch1, I started getting "waiting for count" messages when
> doing ifdown. Not sure if this is the right fix since the count was already
> showing as -1 in that message, but this patch fixes the problem.
> 
> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>

I've applied this because is fixes the problem and there have
be no objections coming along with better fixes :)

If this is bogus we can rever it and put in a more proper fix.