[PATCH] net/fec: fix pm to survive to suspend/resume

[PATCH] net/fec: fix pm to survive to suspend/resume

[Eric Bénard]

* with this patch, if the connection if active before suspend, it
will be active after resume.
* before this patch, it was necessary to close the interface and
reopen it to recover the connection.

Signed-off-by: Eric Bénard <eric@eukrea.com>
Cc: s.hauer@pengutronix.de
Cc: sshtylyov@mvista.com
Cc: linux-arm-kernel@lists.infradead.org
Cc: fabioestevam@yahoo.com
Cc: netdev@vger.kernel.org
---
 drivers/net/fec.c |   10 +++++++---
 1 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/drivers/net/fec.c b/drivers/net/fec.c
index 9b4e8f7..1521972 100644
--- a/drivers/net/fec.c
+++ b/drivers/net/fec.c
@@ -1974,10 +1974,11 @@ fec_suspend(struct platform_device *dev, pm_message_t state)
 	struct fec_enet_private *fep;
 
 	if (ndev) {
-		fep = netdev_priv(ndev);
 		if (netif_running(ndev)) {
+			fep = netdev_priv(ndev);
+			netif_stop_queue(ndev);
 			netif_device_detach(ndev);
-			fec_stop(ndev);
+			clk_disable(fep->clk);
 		}
 	}
 	return 0;
@@ -1987,11 +1988,14 @@ static int
 fec_resume(struct platform_device *dev)
 {
 	struct net_device *ndev = platform_get_drvdata(dev);
+	struct fec_enet_private *fep;
 
 	if (ndev) {
 		if (netif_running(ndev)) {
-			fec_enet_init(ndev, 0);
+			fep = netdev_priv(ndev);
+			clk_enable(fep->clk);
 			netif_device_attach(ndev);
+			netif_start_queue(ndev);
 		}
 	}
 	return 0;
-- 
1.6.3.3

Re: [PATCH] net/fec: fix pm to survive to suspend/resume

[David Miller]

From: Eric Bénard <eric@eukrea.com>
Date: Thu, 27 May 2010 19:53:48 +0200

> * with this patch, if the connection if active before suspend, it
> will be active after resume.
> * before this patch, it was necessary to close the interface and
> reopen it to recover the connection.
> 
> Signed-off-by: Eric Bénard <eric@eukrea.com>

On resume the PHY should be reset and the link renegotiated (or set to
a fixed speed/duplex if set by the user via ethtool) just like as
happens on ->open().

You seem to be avoiding that here.

If the PHY is not functioning properly after the ->resume() triggered
reset, fix that instead.

I'm not applying this patch, it doesn't look the right thing to do at
all.

[PATCH v2] net/fec: fix pm to survive to suspend/resume

[Eric Bénard]

* in the actual driver, calling fec_stop and fec_enet_init doesn't
allow to have a working network interface at resume (where a
ifconfig down and up is required to recover the interface)
* by using fec_enet_close and fec_enet_open, this patch solves this
problem and handle the case where the link changed between suspend
and resume
* this patch also disable clock at suspend and reenable it at resume

Signed-off-by: Eric Bénard <eric@eukrea.com>
Cc: s.hauer@pengutronix.de
Cc: sshtylyov@mvista.com
Cc: linux-arm-kernel@lists.infradead.org
Cc: fabioestevam@yahoo.com
Cc: davem@davemloft.net
---
 drivers/net/fec.c |   16 ++++++++--------
 1 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/drivers/net/fec.c b/drivers/net/fec.c
index 9b4e8f7..40ffdb8 100644
--- a/drivers/net/fec.c
+++ b/drivers/net/fec.c
@@ -1975,10 +1975,9 @@ fec_suspend(struct platform_device *dev, pm_message_t state)
 
 	if (ndev) {
 		fep = netdev_priv(ndev);
-		if (netif_running(ndev)) {
-			netif_device_detach(ndev);
-			fec_stop(ndev);
-		}
+		if (netif_running(ndev))
+			fec_enet_close(ndev);
+		clk_disable(fep->clk);
 	}
 	return 0;
 }
@@ -1987,12 +1986,13 @@ static int
 fec_resume(struct platform_device *dev)
 {
 	struct net_device *ndev = platform_get_drvdata(dev);
+	struct fec_enet_private *fep;
 
 	if (ndev) {
-		if (netif_running(ndev)) {
-			fec_enet_init(ndev, 0);
-			netif_device_attach(ndev);
-		}
+		fep = netdev_priv(ndev);
+		clk_enable(fep->clk);
+		if (netif_running(ndev))
+			fec_enet_open(ndev);
 	}
 	return 0;
 }
-- 
1.6.3.3

Re: [PATCH v2] net/fec: fix pm to survive to suspend/resume

[David Miller]

From: Eric Bénard <eric@eukrea.com>
Date: Sat, 29 May 2010 22:04:50 +0200

> * in the actual driver, calling fec_stop and fec_enet_init doesn't
> allow to have a working network interface at resume (where a
> ifconfig down and up is required to recover the interface)
> * by using fec_enet_close and fec_enet_open, this patch solves this
> problem and handle the case where the link changed between suspend
> and resume
> * this patch also disable clock at suspend and reenable it at resume
> 
> Signed-off-by: Eric Bénard <eric@eukrea.com>

Applied, thanks.