Re: dual CPU system hangs

[christos%astron.com@localhost (Christos Zoulas)]

In article <5236D1A1.8080802%groessler.org@localhost>,
Christian Groessler  <chris%groessler.org@localhost> wrote:
>On 09/11/13 23:04, Christian Groessler wrote:
>> On 09/11/13 22:37, Christos Zoulas wrote:
>>>
>>> This is quite strange. Is this kernel DIAGNOSTIC/DEBUG/LOCKDEBUG?
>>> I am wondering if we are seeing some mutex issue. I seem to recall
>>> that defining FULL in kern_mutex.c made the problem go away. But
>>> this was a while ago and if the kernel has these options, FULL is
>>> already set.
>>
>>
>> Should I set FULL in kern_mutex.c and retry?
>
>
>I've enabled FULL in kern_mutex.c and retried.
>
>The first build went through, but the second build now hangs again, 
>somehow more
>seriously.
>
>Now the system won't respond to pings anymore, and processes accessing
>the disk are hanging (tried 'df -h' on the command line).
>
>It's still somehow active, I can switch between 'sceen' terminals.
>
>Output of ps/n:

[output deleted]

I wish there was a way to see what locks those collect2's in tstile are
holding.

>
>The machine seems to run reliably under Linux, but now I'm remembering
>that I had freezes under OS-X, esp. when it ran for a few days or weeks.
>
>Maybe I've got a hardware problem here?

Could be, but what kind? Bad memory? Does not seem like it...
IIRC when you define DEBUG/DIAGNOSTIC/LOCKDEBUG you cannot reproduce the
problem right?

christos