[PATCH] sungem: fix suspend regression due to NAPI changes

[PATCH] sungem: fix suspend regression due to NAPI changes

[Johannes Berg]

Commit bea3348e (the NAPI changes) made sungem unconditionally enable
NAPI when resuming and unconditionally disable when suspending, this,
however, makes napi_disable() hang when suspending when the interface
was taken down before suspend because taking the interface down also
disables NAPI. This patch makes touching the napi struct in
suspend/resume code paths depend on having the interface up, thereby
fixing the hang on suspend.

The patch also moves the napi_disable() in gem_close() under the lock so
that the NAPI state is always modified atomically together with the
"opened" variable.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>

---
 drivers/net/sungem.c |   11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

--- everything.orig/drivers/net/sungem.c	2007-11-12 18:22:49.948748047 +0100
+++ everything/drivers/net/sungem.c	2007-11-12 18:24:30.708748481 +0100
@@ -2333,10 +2333,10 @@ static int gem_close(struct net_device *
 {
 	struct gem *gp = dev->priv;
 
-	napi_disable(&gp->napi);
-
 	mutex_lock(&gp->pm_mutex);
 
+	napi_disable(&gp->napi);
+
 	gp->opened = 0;
 	if (!gp->asleep)
 		gem_do_stop(dev, 0);
@@ -2355,8 +2355,6 @@ static int gem_suspend(struct pci_dev *p
 
 	mutex_lock(&gp->pm_mutex);
 
-	napi_disable(&gp->napi);
-
 	printk(KERN_INFO "%s: suspending, WakeOnLan %s\n",
 	       dev->name,
 	       (gp->wake_on_lan && gp->opened) ? "enabled" : "disabled");
@@ -2370,6 +2368,8 @@ static int gem_suspend(struct pci_dev *p
 
 	/* If the driver is opened, we stop the MAC */
 	if (gp->opened) {
+		napi_disable(&gp->napi);
+
 		/* Stop traffic, mark us closed */
 		netif_device_detach(dev);
 
@@ -2460,6 +2460,7 @@ static int gem_resume(struct pci_dev *pd
 		/* Re-attach net device */
 		netif_device_attach(dev);
 
+		napi_enable(&gp->napi);
 	}
 
 	spin_lock_irqsave(&gp->lock, flags);
@@ -2479,8 +2480,6 @@ static int gem_resume(struct pci_dev *pd
 	spin_unlock(&gp->tx_lock);
 	spin_unlock_irqrestore(&gp->lock, flags);
 
-	napi_enable(&gp->napi);
-
 	mutex_unlock(&gp->pm_mutex);
 
 	return 0;

Re: [PATCH] sungem: fix suspend regression due to NAPI changes

[Benjamin Herrenschmidt]

On Mon, 2007-11-12 at 18:45 +0100, Johannes Berg wrote:
> Commit bea3348e (the NAPI changes) made sungem unconditionally enable
> NAPI when resuming and unconditionally disable when suspending, this,
> however, makes napi_disable() hang when suspending when the interface
> was taken down before suspend because taking the interface down also
> disables NAPI. This patch makes touching the napi struct in
> suspend/resume code paths depend on having the interface up, thereby
> fixing the hang on suspend.
> 
> The patch also moves the napi_disable() in gem_close() under the lock so
> that the NAPI state is always modified atomically together with the
> "opened" variable.
> 
> Signed-off-by: Johannes Berg <johannes@sipsolutions.net>

Thanks for fixing that !

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---

> ---
>  drivers/net/sungem.c |   11 +++++------
>  1 file changed, 5 insertions(+), 6 deletions(-)
> 
> --- everything.orig/drivers/net/sungem.c	2007-11-12 18:22:49.948748047 +0100
> +++ everything/drivers/net/sungem.c	2007-11-12 18:24:30.708748481 +0100
> @@ -2333,10 +2333,10 @@ static int gem_close(struct net_device *
>  {
>  	struct gem *gp = dev->priv;
>  
> -	napi_disable(&gp->napi);
> -
>  	mutex_lock(&gp->pm_mutex);
>  
> +	napi_disable(&gp->napi);
> +
>  	gp->opened = 0;
>  	if (!gp->asleep)
>  		gem_do_stop(dev, 0);
> @@ -2355,8 +2355,6 @@ static int gem_suspend(struct pci_dev *p
>  
>  	mutex_lock(&gp->pm_mutex);
>  
> -	napi_disable(&gp->napi);
> -
>  	printk(KERN_INFO "%s: suspending, WakeOnLan %s\n",
>  	       dev->name,
>  	       (gp->wake_on_lan && gp->opened) ? "enabled" : "disabled");
> @@ -2370,6 +2368,8 @@ static int gem_suspend(struct pci_dev *p
>  
>  	/* If the driver is opened, we stop the MAC */
>  	if (gp->opened) {
> +		napi_disable(&gp->napi);
> +
>  		/* Stop traffic, mark us closed */
>  		netif_device_detach(dev);
>  
> @@ -2460,6 +2460,7 @@ static int gem_resume(struct pci_dev *pd
>  		/* Re-attach net device */
>  		netif_device_attach(dev);
>  
> +		napi_enable(&gp->napi);
>  	}
>  
>  	spin_lock_irqsave(&gp->lock, flags);
> @@ -2479,8 +2480,6 @@ static int gem_resume(struct pci_dev *pd
>  	spin_unlock(&gp->tx_lock);
>  	spin_unlock_irqrestore(&gp->lock, flags);
>  
> -	napi_enable(&gp->napi);
> -
>  	mutex_unlock(&gp->pm_mutex);
>  
>  	return 0;
> 

Re: [PATCH] sungem: fix suspend regression due to NAPI changes

[David Miller]

From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Date: Tue, 13 Nov 2007 09:32:46 +1100

> 
> On Mon, 2007-11-12 at 18:45 +0100, Johannes Berg wrote:
> > Commit bea3348e (the NAPI changes) made sungem unconditionally enable
> > NAPI when resuming and unconditionally disable when suspending, this,
> > however, makes napi_disable() hang when suspending when the interface
> > was taken down before suspend because taking the interface down also
> > disables NAPI. This patch makes touching the napi struct in
> > suspend/resume code paths depend on having the interface up, thereby
> > fixing the hang on suspend.
> > 
> > The patch also moves the napi_disable() in gem_close() under the lock so
> > that the NAPI state is always modified atomically together with the
> > "opened" variable.
> > 
> > Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
> 
> Thanks for fixing that !
> 
> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

Indeed, thanks a lot Johannes.

Patch applied.