e100_shutdown: netif_poll_disable hang
Subsystems (2)
PATCH 1/1 e100_shutdown: netif_poll_disable hang
My machine annoyingly hangs while rebooting. I tracked it down to e100-fix-reboot-f-with-netconsole-enabled.patch in 2.6.18-rc2-mm2
I review the changes and it seemed to be calling netif_poll_disable one too many time. Once in e100_down(), and again in e100_shutdown().
The second one in e100_shutdown() caused the hang. So this patch removes it.
--- drivers/net/e100.c | 1 - 1 files changed, 1 deletion(-)
drivers/net/e100.c +0 -1
@@ -2709,7 +2709,6 @@ static void e100_shutdown(struct pci_devstruct net_device *netdev =pci_get_drvdata (pdev);struct nic *nic =netdev_priv (netdev); -netif_poll_disable (nic->netdev);del_timer_sync (&nic->watchdog);netif_carrier_off (nic->netdev);
--
Auke Kok
NAKQuoted 20 lines · Daniel Walker
My machine annoyingly hangs while rebooting. I tracked it down to e100-fix-reboot-f-with-netconsole-enabled.patch in 2.6.18-rc2-mm2
I review the changes and it seemed to be calling
netif_poll_disableone too many time. Once ine100_down(), and again ine100_shutdown().The second one in
e100_shutdown()caused the hang. So this patch removes it.Signed-off-by Daniel Walker <[email protected]>--- drivers/net/e100.c | 1 - 1 files changed, 1 deletion(-)Index:
linux-2.6.18/drivers/net/e100.c
===================================================================
drivers/net/e100.c static void e100_shutdown(struct pci_dev
Quoted 8 lines
drivers/net/e100.c +0
-2@@ -2709,7 +2709,6 @@ static void e100_shutdown(struct pci_devstruct net_device *netdev =pci_get_drvdata (pdev);struct nic *nic =netdev_priv (netdev); -netif_poll_disable (nic->netdev);del_timer_sync (&nic->watchdog);netif_carrier_off (nic->netdev); --
this won't help netconsole shutdown/reboot -f, probably locking it up again!
NAK
Also missing is the NAPI conditional, which I left out. Also missing is the same code for suspend.
Here's a better approach, allowing both normal shutdown code path and reboot -f to disable polling.
Cheers,
Auke
---
e100: disable polling only when up during suspend and shutdown
--- drivers/net/e100.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)
---
drivers/net/e100.c +8 -2
@@ -2719,7 +2719,10 @@ static int e100_suspend(struct pci_dev *struct net_device *netdev =pci_get_drvdata (pdev);struct nic *nic =netdev_priv (netdev); -netif_poll_disable (nic->netdev); +#ifdef CONFIG_E100_NAPI +if (netif_running (netdev)) +netif_poll_disable (nic->netdev); +#endif del_timer_sync (&nic->watchdog);netif_carrier_off (nic->netdev);
@@ -2763,7 +2766,10 @@ static void e100_shutdown(struct pci_devstruct net_device *netdev =pci_get_drvdata (pdev);struct nic *nic =netdev_priv (netdev); -netif_poll_disable (nic->netdev); +#ifdef CONFIG_E100_NAPI +if (netif_running (netdev)) +netif_poll_disable (nic->netdev); +#endif del_timer_sync (&nic->watchdog);netif_carrier_off (nic->netdev);
Quoted 9 lines · Auke Kok
Quoted 8 lines · Daniel Walker
My machine annoyingly hangs while rebooting. I tracked it down to e100-fix-reboot-f-with-netconsole-enabled.patch in 2.6.18-rc2-mm2
I review the changes and it seemed to be calling
netif_poll_disableone too many time. Once ine100_down(), and again ine100_shutdown().The second one in
e100_shutdown()caused the hang. So this patch removes it.
it doesn't even do harm to netif_poll_disable() twice as far as I can see, as it merely calls test_and_set_bit(), which will instantly succeed on the first attempt if the bit was already set.
did this change actually fix it for you? I'm wondering if the netif_carrier_off might not be the culprit here...
Auke
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
My machine annoyingly hangs while rebooting. I tracked it down to e100-fix-reboot-f-with-netconsole-enabled.patch in 2.6.18-rc2-mm2 I review the changes and it seemed to be calling
netif_poll_disableone too many time. Once ine100_down(), and again ine100_shutdown().
The second one ine100_shutdown()caused the hang. So this patch removes it.
Quoted 6 lines · Auke Kok
it doesn't even do harm to
netif_poll_disable()twice as far as I can see, as it merely callstest_and_set_bit(), which will instantly succeed on the first attempt if the bit was already set.did this change actually fix it for you? I'm wondering if the
netif_carrier_offmight not be the culprit here...
I can confirm the proposed original change of D. Walker fixed the problem for me. I did not test the change you proposed as a followup.
Signature · 1 line
Damien Wyart
Quoted 13 lines · Damien Wyart
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
My machine annoyingly hangs while rebooting. I tracked it down to e100-fix-reboot-f-with-netconsole-enabled.patch in 2.6.18-rc2-mm2 I review the changes and it seemed to be calling
netif_poll_disableone too many time. Once ine100_down(), and again ine100_shutdown().
The second one ine100_shutdown()caused the hang. So this patch removes it.Quoted 3 lines · Auke Kok
it doesn't even do harm to
netif_poll_disable()twice as far as I can see, as it merely callstest_and_set_bit(), which will instantly succeed on the first attempt if the bit was already set.
Quoted 5 lines
Quoted 2 lines
did this change actually fix it for you? I'm wondering if the
netif_carrier_offmight not be the culprit here...I can confirm the proposed original change of D. Walker fixed the problem for me. I did not test the change you proposed as a followup.
his change breaks something else (a reboot with netconsole, possibly suspend). Please give the latest version I sent a try. Daniel confirmed me that it works, but it's always nice to hear it from more people.
Thanks
Auke
* Auke Kok <[email protected]> [2006-10-21 10:54]: his change
Quoted 3 lines
breaks something else (a reboot with netconsole, possibly suspend). Please give the latest version I sent a try. Daniel confirmed me that it works, but it's always nice to hear it from more people.
Yes, your version works here too.
Signature · 1 line
Damien Wyart