Re: [patch] revert: [NET]: Fix races in net_rx_action vs netpoll

[Ingo Molnar <mingo@elte.hu>]

* Ingo Molnar <mingo@elte.hu> wrote:

> ah! Just found the reason: the bug apparently depends on the precise 
> kernel command-line contents. I accidentally dropped ignore_loglevel 
> (found this while comparing with the older logs i sent to you), adding 
> it back in produces hung networking too. So it appears that a 
> netconsole printout while e1000 is initializing (or while some other 
> networking component is initializing) might be the culprit?

and the WARN_ON() below does not seem to trigger.

i'll now check whether removing ignore_on_loglevel (no other changes) 
makes the hang go away. Maybe ignore_on_loglevel is buggy - or it 
produces an immediate printk (going out to the interface) during a 
particularly sensitive period of network initialization.

	Ingo

----------------->
Index: linux/drivers/net/e1000/e1000_main.c
===================================================================
--- linux.orig/drivers/net/e1000/e1000_main.c
+++ linux/drivers/net/e1000/e1000_main.c
@@ -358,6 +358,7 @@ e1000_irq_enable(struct e1000_adapter *a
 		E1000_WRITE_REG(&adapter->hw, IMS, IMS_ENABLE_MASK);
 		E1000_WRITE_FLUSH(&adapter->hw);
 	}
+	WARN_ON_ONCE(atomic_read(&adapter->irq_sem) < 0);
 }
 
 static void