On Friday 20 June 2008 18:01:09 Michael Buesch wrote:
quoted text
> On Friday 20 June 2008 17:55:41 Ingo Molnar wrote:
> > 
> > * Michael Buesch <mb@bu3sch.de> wrote:
> > 
> > > On Friday 20 June 2008 17:27:48 Ingo Molnar wrote:
> > > >  [<c012b361>] local_bh_enable_ip+0xd1/0xe0
> > > >  [<c08d9f9f>] _spin_unlock_bh+0x2f/0x40
> > > >  [<c0471b92>] vortex_timer+0xe2/0x3e0
> > > 
> > > > real bug or false positive?
> > > 
> > > Well, a timer runs with IRQs disabled, no? So this would be a bug.
> > 
> > indeed - agreed :) [no time for me to fix it, but can test any rfc patch.]
> 
> A quick workaround always is to convert the lock into an _irqsafe lock.
> Although it introduces higher overhead (interrupt-wise), it prevents the bug.
> A real fix would require to understand the locking in the driver, which I
> don't, as I never looked at the driver. :)


However, looking at the driver I think the fix actually is trivial:

Index: wireless-testing/drivers/net/3c59x.c
===================================================================
--- wireless-testing.orig/drivers/net/3c59x.c	2008-05-16 00:26:29.000000000 +0200
+++ wireless-testing/drivers/net/3c59x.c	2008-06-20 18:16:55.000000000 +0200
@@ -1768,9 +1768,10 @@ vortex_timer(unsigned long data)
 	case XCVR_MII: case XCVR_NWAY:
 		{
 			ok = 1;
-			spin_lock_bh(&vp->lock);
+			/* Interrupts are already disabled */
+			spin_lock(&vp->lock);
 			vortex_check_media(dev, 0);
-			spin_unlock_bh(&vp->lock);
+			spin_unlock(&vp->lock);
 		}
 		break;
 	  default:					/* Other media types handled by Tx timeouts. */


vp->lock is also taken in hardware IRQ context, so we _have_ to always
use irqsafe locking. As we run in a timer with IRQs disabled,
we can simply use spin_lock.

-- 
Greetings Michael.
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/