On Tue, 31 Mar 2009, Jean-Michel Barbet wrote:

> Michael Mansour wrote:
>
>> For this (on a 5.2 i386 box), I have:
>> options bond0 mode=1 miimon=100 use_carrier=1
>> and it's never failed.
>
> Thanks Michael, it could be x86_64 specific maybe...

Or it could be pure statistics. It also depends on usage. We're seeing 
this problem frequently lately (x86_64, kernels 91.1.6 and 92.1.22 at 
least). Out of a few dozen nodes with two bonded GbE links, one or two 
per week develop such a soft lockup. It seems to be much more likely to 
happen when the links are very busy.

The latest kernel (128.1.1) has a fix in the locking code in the bonding 
module, so this one may fix it.

Any insights very welcome, since this problem is quite serious for 
us as well.

Regards,
 	Stephan

-- 
Stephan Wiesand
   DESY - DV -
   Platanenallee 6
   15738 Zeuthen, Germany