On Tue, 24 Apr 2007, Greg A. Woods wrote:
> At Tue, 24 Apr 2007 10:08:37 +0100 (BST), Stephen Borrill wrote:
> Subject: Re: 3.1_STABLE and SMP
>>
>> I've given them the instructions and installed a LOCKDEBUG kernel for
>> them. So we're just waiting for the next crash that will allow them to get
>> to ddb (not all do apparently).
>
> If the hangs (crashes without DDB) are related to kernel locking then
> the LOCKDEBUG kernel should in fact detect them and dump into DDB much
> more readily.

I didn't even have to wait for their server to crash. Our test machine 
here died overnight (when not doing anything in particular):

http://projects.precedence.co.uk/netbsd/ddb2.jpg

Sorry for the photo ddb trace, when I manage to get a dump to a RAIDframe 
device I'll be happier.

The simple_lock_dump() output was as follows:
all simple locks:
0xc06beacc CPU 0 ../../../../kern/kern_lock.c:1487
0x80000000

-- 
Stephen