Skip to contentSkip to conversation

e1000: Fix regression: garbled stats and irq allocation during swsusp

PATCH In review Auke Kok 4 participants 5 messages Nov 3, 2006 – Nov 6, 2006 Network drivers

lore
Share & tools
lore

Auke Kok

author
#0 lore raw

PATCH 1/1 e1000: Fix regression: garbled stats and irq allocation during swsusp

e1000: Fix regression: garbled stats and irq allocation during swsusp

After 7.0.33/2.6.16, e1000 suspend/resume left the user with an enabled device showing garbled statistics and undetermined irq allocation state, where ifconfig eth0 down would display trying to free already freed irq.

Explicitly free and allocate irq as well as powerup the PHY during resume fixes.

drivers/net/e1000/e1000_main.c +6 -0
index c0269b5..32cdccb 100644
@@ -5081,6 +5081,8 @@ #endif
 	if (adapter->hw.phy_type == e1000_phy_igp_3)
 		e1000_phy_powerdown_workaround(&adapter->hw);
 
+	e1000_free_irq(adapter);
+
 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
 	 * would have already happened in close and is redundant. */
 	e1000_release_hw_control(adapter);
@@ -5111,6 +5113,10 @@ e1000_resume(struct pci_dev *pdev)
 	pci_enable_wake(pdev, PCI_D3hot, 0);
 	pci_enable_wake(pdev, PCI_D3cold, 0);
 
+	if ((err = e1000_request_irq(adapter)))
+		return err;
+
+	e1000_power_up_phy(adapter);
 	e1000_reset(adapter);
 	E1000_WRITE_REG(&adapter->hw, WUS, ~0);
#1 lore raw

Hi Auke,

good news. I incorporated the changes manually b/cI couldn't apply your patch against the latest git tree from Linux with the program patch. The patch would solve http://bugzilla.kernel.org/show_bug.cgi?id=7207 :-) Here's the diff after my edit of the file e1000_main.c:

drivers/net/e1000/e1000_main.c +6 -1
index 8d04752..400bdee 100644
@@ -4799,7 +4799,8 @@ e1000_suspend(struct pci_dev *pdev, pm_m
 
        if (adapter->hw.phy_type == e1000_phy_igp_3)
                e1000_phy_powerdown_workaround(&adapter->hw);
-
+       e1000_free_irq(adapter);
+
        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
         * would have already happened in close and is redundant. */
        e1000_release_hw_control(adapter);
@@ -4830,6 +4831,10 @@ e1000_resume(struct pci_dev *pdev)
        pci_enable_wake(pdev, PCI_D3hot, 0);
        pci_enable_wake(pdev, PCI_D3cold, 0);
 
+       if ((err = e1000_request_irq(adapter)))
+               return err;
+
+       e1000_power_up_phy(adapter);
        e1000_reset(adapter);
        E1000_WRITE_REG(&adapter->hw, WUS, ~0);
 

Am Friday 03 November 2006 20:43 schrieb Auke Kok:

Quoted 37 lines

e1000: Fix regression: garbled stats and irq allocation during swsusp

After 7.0.33/2.6.16, e1000 suspend/resume left the user with an enabled device showing garbled statistics and undetermined irq allocation state, where ifconfig eth0 down would display trying to free already freed irq.

Explicitly free and allocate irq as well as powerup the PHY during resume fixes.

drivers/net/e1000/e1000_main.c +6 -0
index c0269b5..32cdccb 100644
@@ -5081,6 +5081,8 @@ #endif
  	if (adapter->hw.phy_type == e1000_phy_igp_3)
  		e1000_phy_powerdown_workaround(&adapter->hw);
 
+	e1000_free_irq(adapter);
+
  	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
  	 * would have already happened in close and is redundant. */
  	e1000_release_hw_control(adapter);
@@ -5111,6 +5113,10 @@ e1000_resume(struct pci_dev *pdev)
  	pci_enable_wake(pdev, PCI_D3hot, 0);
  	pci_enable_wake(pdev, PCI_D3cold, 0);
 
+	if ((err = e1000_request_irq(adapter)))
+		return err;
+
+	e1000_power_up_phy(adapter);
  	e1000_reset(adapter);
  	E1000_WRITE_REG(&adapter->hw, WUS, ~0);
 
Signature · 4 lines
MfG/Sincerely

Toralf Förster
+++ I'm not subscribed to the email list, please Cc: me +++
#2 lore raw
In reply to Auke Kok
Quoted 11 lines · Auke Kok

e1000: Fix regression: garbled stats and irq allocation during swsusp

After 7.0.33/2.6.16, e1000 suspend/resume left the user with an enabled device showing garbled statistics and undetermined irq allocation state, where ifconfig eth0 down would display trying to free already freed irq.

Explicitly free and allocate irq as well as powerup the PHY during resume fixes.

ACK, but:

Applying 'e1000: Fix regression: garbled stats and irq allocation during swsusp'

fatal: corrupt patch at line 8

#3 lore raw

Hi!

Quoted 10 lines
Quoted 10 lines

e1000: Fix regression: garbled stats and irq allocation during swsusp

After 7.0.33/2.6.16, e1000 suspend/resume left the user with an enabled device showing garbled statistics and undetermined irq allocation state, where ifconfig eth0 down would display trying to free already freed irq.

Explicitly free and allocate irq as well as powerup the PHY during resume fixes.

Quoted 8 lines
Quoted 1 line

ACK, but:

Applying 'e1000: Fix regression: garbled stats and irq allocation during swsusp'

fatal: corrupt patch at line 8

Toralf posted rediffed (manually applied) version... should I forward it to you?
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
author
#4 lore raw
In reply to Jeff Garzik
Quoted 11 lines · Jeff Garzik
Quoted 10 lines · Auke Kok

e1000: Fix regression: garbled stats and irq allocation during swsusp

After 7.0.33/2.6.16, e1000 suspend/resume left the user with an enabled device showing garbled statistics and undetermined irq allocation state, where ifconfig eth0 down would display trying to free already freed irq.

Explicitly free and allocate irq as well as powerup the PHY during resume fixes.

Quoted 7 lines
Quoted 1 line

ACK, but:

Applying 'e1000: Fix regression: garbled stats and irq allocation during swsusp'

fatal: corrupt patch at line 8

There's another problem with it, that needs attention as well. I will reformat and re-post the patch to everyone.

Cheers,

Auke

From: <[email protected]>
Also, would this cause a problem if I ifdown the interface to call ->close then suspend? Looks like it'd free_irq twice, maybe need a netif_running() check?