mpt request timed out

Ståle Kristoffersen staalebk at ifi.uio.no
Mon Jun 7 01:19:44 UTC 2010


On 2010-06-06 at 08:04, Matthew Jacob wrote:
> >
> >> Is it necessary to track further if all is okay after the first reset?
> >>      
> > Oh, I might have misunderstood what you where asking about;
> > It works fine untill the next timeout occurrs. Could be a few hours, could
> > be 5 minutes.
> >
> >    
> Oh- okay.  Then I guess we'd better work on fixing it.
> 
> Can you describe the actually physical topology here? Is this an 
> internal backplane?

One port on the LSI-controller is connected to the front internal backplane
of a supermicro 847E1 chassis[1]. The other port on the controller is not
connected, and the rear internal backplane is not connected to anything.

I have tried to change the cable between the backplane and controller,
without any effect. I've also tried to backport the changes in r198262 [2]
since that looked like it could be related. That did not fix it either.

I also tried to limit the depth of queue for tagged openings with
camcontrol:
camcontrol tags da0 -N 4
(and repeated for all the other disks connected)
and that had some effect, when I got a stall after that change, I did not
allways get the timeout message. 

[1]: http://www.supermicro.com/products/chassis/4U/847/SC847E1-R1400U.cfm
[2]: http://svn.freebsd.org/viewvc/base?view=revision&revision=198262
-- 
Ståle Kristoffersen
staalebk at ifi.uio.no


More information about the freebsd-scsi mailing list