Date: Wed, 6 Aug 2014 22:33:27 +0100 From: "Steven Hartland" <killing@multiplay.co.uk> To: <wollman@bimajority.org> Cc: freebsd-scsi@freebsd.org, freebsd-stable@freebsd.org Subject: Re: 9.3-RELEASE still instapanics on multi-mps(4) servers Message-ID: <FE12D1B47D154D7FB165B1531CB9635A@multiplay.co.uk> References: <21474.34330.572142.206098@hergotha.csail.mit.edu> <4DE72E01E8A64E98A31920FCFCAA5D26@multiplay.co.uk> <201408062110.s76LAhhE079487@hergotha.csail.mit.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
TBH that sounds like dodgy hardware. We had a similar thing a few years
back with a machine which would panic mfi badly all the time where as
other machines where solid as a rock.
If its random then you could be facing the same thing.
I our case it turned out to be a faulty Intel CPU. There where on other
signs of issues just random panic in mfi.
So given the similarity and you said it only effects one out of two machines
have the HW replaced and see if the problem goes away.
Regards
Stev
----- Original Message -----
From: <wollman@bimajority.org>
To: <killing@multiplay.co.uk>
Cc: <freebsd-stable@freebsd.org>; <freebsd-scsi@freebsd.org>
Sent: Wednesday, August 06, 2014 10:10 PM
Subject: Re: 9.3-RELEASE still instapanics on multi-mps(4) servers
> In article <4DE72E01E8A64E98A31920FCFCAA5D26@multiplay.co.uk>,
> killing@multiplay.co.uk writes:
>>The stack from the panic would be a good start.
>
> As I said, it's in the middle of the USB code, which does not appear,
> from my previous bisection, to be connected with the bug at all. (The
> panic is the result of an unhandled trap that happens during
> interrupt-driven probing, and it's nearly always in the USB code. By
> loading different modules I can make it happen at slightly different
> times and places.) Six months ago, I found that enabling any form of
> memory debugging suppresses the symptoms, although it also kills
> performance, of course. I haven't tried that yet this time around.
>
> Once I get a serial console hooked up I'll be in a better position to
> capture the full data (although obviously not a core dump).
>
> -GAWollman
>
>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?FE12D1B47D154D7FB165B1531CB9635A>