... 13:21:23 -0700 (PDT) > > bugme-daemon@bugzilla.kernel.org wrote: ... /show_bug.cgi?id=10326 > >> > >> Summary: inconsistent lock state in net_rx_action > >> Product: ... that would be a lockdep bug. -- To unsubscribe from this list ...

linux-kernel - Andrew Morton - Mar 26 2008 - 20:14

... 13:21:23 -0700 (PDT) > bugme-daemon@bugzilla.kernel.org wrote: ... show_bug.cgi?id=10326 >> >> Summary: inconsistent lock state in net_rx_action >> Product ... tracking?) imbalance and not networking bug. Regards, Jarek P. PS: linux ...

linux-kernel - Jarek Poplawski - Mar 26 2008 - 19:13

... bugzilla.kernel.org/show_bug.cgi?id=10326 > > ... > > > No, it's not an ... , but net_rx_action() takes the same lock from > > > within softirq context. > > > > > > If ... a lockdep bug. > > > > sky2_poll() doesn't take napi->poll_lock; this lock is taken ...

linux-kernel - Peter Zijlstra - Mar 27 2008 - 06:56

... ://bugzilla.kernel.org/show_bug.cgi?id=10326 ... > No, it's not an irq_disable() ... , but net_rx_action() takes the same lock from > within softirq context. > > If sky2_poll() ... a lockdep bug. sky2_poll() doesn't take napi->poll_lock; this lock is taken ...

linux-kernel - Jarek Poplawski - Mar 27 2008 - 04:55

... bugzilla.kernel.org/show_bug.cgi?id=10326 > ... > > No, it's not an irq_disable ... , but net_rx_action() takes the same lock from > > within softirq context. > > > > If sky2_poll ... lockdep bug. > > sky2_poll() doesn't take napi->poll_lock; this lock is ...

linux-kernel - Andrew Morton - Mar 27 2008 - 05:18

... do_softirq() > net_rx_action() > spin_lock(&napi->poll_lock) <--- Deadlock! > > Because we still hold the lock.... Yep, that will deadlock. Can some test if using local_irq_{save,restore}() in napi_complete() cures this ...

linux-kernel - David Miller - Mar 27 2008 - 07:03

... napi->poll_lock) > poll_one_napi() > napi->poll() := sky2_poll() > napi_complete() > local_irq_disable() > local_irq_enable() <--- *BUG* Yes! I missed it's unconditional here... Great catch! On the ...

linux-kernel - Jarek Poplawski - Mar 27 2008 - 08:22

... ) > > poll_one_napi() > > napi->poll() := sky2_poll() > > napi_complete() > > local_irq_disable() > > local_irq_enable() <--- *BUG* > > Yes! I missed it's unconditional here... Great catch! > > On ...

linux-kernel - Peter Zijlstra - Mar 27 2008 - 08:30

... 27, 2008 at 09:55:42AM +0100, Jarek Poplawski wrote: ... > sky2_poll() doesn't take napi->poll_lock; this lock is taken by > netpoll_poll() before calling sky2_poll(). And before this hardirqs > are disabled in write_msg(). Actually, hardirqs are ...

linux-kernel - Jarek Poplawski - Mar 27 2008 - 05:08

... :46 better kernel: #1: (target_list_lock){--..}, at: [] means lockdep saw hardirqs enabled while acquiring/holding this lock. Jarek P. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message ...

linux-kernel - Jarek Poplawski - Mar 27 2008 - 06:07

... : > Can some test if using local_irq_{save,restore}() in > napi_complete() cures this lockdep warning? I cannot reproduce the bug, so cannot help much. Marcus -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the ...

linux-kernel - Marcus Better - Mar 27 2008 - 07:59

... >> napi_complete() cures this lockdep warning? Isn't it now justified enough without lockdep's ack? > I cannot reproduce the bug, so cannot help much. Marcus, you've helped here very much and enough! Thanks, Jarek P. -- To unsubscribe from ...

linux-kernel - Jarek Poplawski - Mar 27 2008 - 08:32

On Thu, Mar 27, 2008 at 10:56:49PM +0100, Rafael J. Wysocki wrote: ... > I assume there will be a patch posted for this issue. Correct? I guess Peter should know the answer! Jarek P. -- To unsubscribe from this list: send the line "unsubscribe

linux-kernel - Jarek Poplawski - Mar 27 2008 - 18:22

From: Jarek Poplawski Date: Thu, 27 Mar 2008 23:22:48 +0100 > On Thu, Mar 27, 2008 at 10:56:49PM +0100, Rafael J. Wysocki wrote: > ... > > I assume there will be a patch posted for this issue. Correct? > > I guess Peter

linux-kernel - David Miller - Mar 27 2008 - 18:22

On Thu, Mar 27, 2008 at 03:22:18PM -0700, David Miller wrote: ... > Instead of going back and forth like this can someone post > just take it upon themselves to post such an obvious patch > already :-) > > All of this "he'll do it, she'll do it"

linux-kernel - Jarek Poplawski - Mar 27 2008 - 18:39

... admit to being a little out of my depth here) Probably you're right but there were really a lot of testing and stressing netconsole for misterious lockups for quite a long time with this new napi, so it's hard to believe this was hidden so ...

linux-kernel - Jarek Poplawski - Mar 27 2008 - 08:49

From: Jarek Poplawski Date: Thu, 27 Mar 2008 13:32:48 +0100 > On Thu, Mar 27, 2008 at 12:59:02PM +0100, Marcus Better wrote: > > David Miller wrote: > >> Can some test if using local_irq_{save,restore}() in > >> napi_complete()

linux-kernel - David Miller - Mar 27 2008 - 17:44

... admit to being a little out of my depth here) > > Probably you're right but there were really a lot of testing and > stressing netconsole for misterious lockups for quite a long time > with this new napi, so it's hard to believe this was hidden so > ...

linux-kernel - Rafael J. Wysocki - Mar 27 2008 - 17:56

From: Jarek Poplawski Date: Thu, 27 Mar 2008 23:39:13 +0100 > On Thu, Mar 27, 2008 at 03:22:18PM -0700, David Miller wrote: > ... > > Instead of going back and forth like this can someone post > > just take it upon themselves to

linux-kernel - David Miller - Mar 27 2008 - 20:59

On Thu, 2008-03-27 at 17:59 -0700, David Miller wrote: > From: Jarek Poplawski > Date: Thu, 27 Mar 2008 23:39:13 +0100 > > > On Thu, Mar 27, 2008 at 03:22:18PM -0700, David Miller wrote: > > ... > > > Instead of going back and

linux-kernel - Peter Zijlstra - Mar 28 2008 - 06:12