Re: run interrupt driven hooks: still waiting for xpt_config

From: Attilio Rao <attilio_at_freebsd.org>
Date: Sat, 8 Aug 2009 17:51:19 +0200
2009/8/8 Robert Watson <rwatson_at_freebsd.org>:
> On Sat, 8 Aug 2009, Marc UBM Bocklet wrote:
>
>>>>>> I've got it narrowed down between "2009.06.30.06.00.00" and today. A
>>>>>> kernel with the "old" date boots, a freshly csupped and compiled kernel
>>>>>> hangs with the usual symptoms (waiting for interrupt driven hooks).
>>>>>>
>>>>>> I'll try csupping to just before the big cam commit to see if there is
>>>>>> any connection. When I said earlier that I was not running with the ahci
>>>>>> patch, I was partly wrong. I did not have device ahci in my kernel config
>>>>>> file nor had it loaded as a module, but I had the patch applied.
>>>>>
>>>>> "2009.07.09.06.00.00" fixes the problem. Could it be that there are
>>>>> some subtle interactions in the cam subsystem that are stirred by the recent
>>>>> mega-commit?
>>>>
>>>> Is there any other info I can / should provide to help debugging this?
>>>
>>> Any news on this? This pretty much prevents me from running 8.0 :-/
>>
>> I've a friend with a similar problem (also HighPoint controller, also "run
>> interrupt driven hooks: still waiting for xpt_config"), who gets a panic.
>> I'll try to get the panic string and a backtrace, if I do, I'll post it
>> here.
>
> xpt_config basically means that you're waiting on a device driver attached
> to CAM to finish probing, which could point at a number of potential problem
> sources (including things like interrupt routing problems).  At least in the
> 7.x line, I've seen firewire and USB at various times cause this issue.  It
> might be interesting to compile down to the smallest set of cam-related
> drivers required to support necessary hardware (omit firewire, for example)
> and see if you see any improvement.  Pinning it down to a specific driver
> would have significant debugging value :-).

Beyon that, I think that debugging in CAM can be improved by a good
factor using alredy existing general infrastructure  (KTR logging
points, new ddb commands, etc). We need someone which is willing to
spend some time on that.

Attilio


-- 
Peace can only be achieved by understanding - A. Einstein
Received on Sat Aug 08 2009 - 13:51:21 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:53 UTC