Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 18 Nov 2014 20:35:49 -0500
From:      "Ellis H. Wilson III" <ellisw@panasas.com>
To:        Benjamin Kaduk <kaduk@MIT.EDU>
Cc:        freebsd-current@freebsd.org
Subject:   Re: WITNESS observes 2 LORs on Boot of Release 10.1
Message-ID:  <546BF3F5.8030109@panasas.com>
In-Reply-To: <alpine.GSO.1.10.1411181734520.19231@multics.mit.edu>
References:  <546BA9D3.6070007@panasas.com> <alpine.GSO.1.10.1411181734520.19231@multics.mit.edu>

next in thread | previous in thread | raw e-mail | index | archive | help
On 11/18/2014 05:39 PM, Benjamin Kaduk wrote:
> On Tue, 18 Nov 2014, Ellis H. Wilson III wrote:
>> I'm observing the following two WITNESS LORs being thrown upon boot-up of 10.0
>> and I was tracking current, hoping they would go away by 10.1, but it seems
>> they persist as shown below.  I suspect this is because current is being built
>> with WITNESS on but also with SKIPSPIN on.  So these issues are unlikely to
>> show up for any devs but those who specifically enable WITNESS and disable
>> SKIPSPIN like myself.  At my work we would greatly like to do our debugging
>> with checking of spin-locking order included and panicing upon LOR detection.
>> That's not possible with these in existence.
>
> However, I was under the impression that a kernel built with WITNESS and
> without WITNESS_SKIPSPIN would panic on boot on the cnputs_mutx (see,
> e.g.,
> https://lists.freebsd.org/pipermail/freebsd-stable/2014-January/076864.html).
> So, (1) I'm surprised you can boot it, and (2) that would explain why no
> one else has been using it.

That's a very interesting thread.  I've seen another where a fellow 
developer suggested just throwing on the WITNESS_SKIPSPIN flag to 
"solve" the issue.  I can't say that I agree with the approach, but I 
understand.  I'd be willing to tackle a bit of WITNESS massaging to help 
it be instructed about known false positives better, if that's desirable.

Why I'm able to boot however, is simple: I haven't enabled a full suite 
of debugging flags, KDB/DDB being the key ones that cause a panic to 
occur on a failure like the one I've seen.  We originally were seeing 
the panic so WITNESS in its entirety was shut off.  I was asked to try 
and get that back on-track, so to start I at least wanted to see how 
many LORs we were dealing with on boot.  2 is apparently that magic 
number, and maybe 3 if the cputs issue you refer to is still possible to 
be hit in the 10.1 wild.

If nobody has seen these before, I'll try and put together fixes for 
them.  Please somebody speak up if you have seen them or have useful 
information for me to go on in my patches.

Thanks,

ellis



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?546BF3F5.8030109>