Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 15 Sep 2013 09:52:38 -0500
From:      Diane Bruce <db@db.net>
To:        Paul Mather <pmather@vt.edu>
Cc:        freebsd-arm <freebsd-arm@freebsd.org>
Subject:   Re: sshd isnt working?
Message-ID:  <20130915145238.GB18382@night.db.net>
In-Reply-To: <92A05C0F-B921-4998-8AF5-E0B4994E37D5@g.mail.vt.edu>
References:  <CAHjFwoCK2qdgeyyQLJ6FEZxJFRs8gELP37LxoDvjOUsfv6PuBQ@mail.gmail.com> <20130915120005.GA16617@night.db.net> <92A05C0F-B921-4998-8AF5-E0B4994E37D5@g.mail.vt.edu>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Sep 15, 2013 at 10:23:08AM -0400, Paul Mather wrote:
> On Sep 15, 2013, at 8:00 AM, Diane Bruce wrote:
> 
> > On Sun, Sep 15, 2013 at 01:39:42PM +0800, Su Kang Yin wrote:
> >> Hi,
> >> 
> >> sshd throwing signal 6 on server side
...
> 
> 
> For me, I've only encountered this problem when connecting from certain clients---Mac OS X 10.7 and 10.8 so far.  I can SSH to my Raspberry Pi and BeagleBone Black just fine from another RELENG_9 FreeBSD system, and from a Debian Wheezy system.  The failure mode for those that don't work is the same as reported above---a signal 6 on the server side.  Also, this hasn't always been the case, but I can't recall a revision number when the problem began.

Yes, I have been trying to keep track of these types of reports as well.
This strongly smells of a timing related problem. I have a ktrace of
the entire process of sshd doing its thing. It would depend on how
fast the connection was when it was hit.

I initially thought we had a possible use after free so sent the ktrace to
jasone@ . The problem goes away using ElectricFence from ports to replace
the malloc code in sshd. Using jemalloc's opt.quarantine feature,
the problem will also go away. I think this the commit that triggered this bug.

http://svnweb.freebsd.org/base?view=revision&revision=250991

I'd be curious if quarantine use results in a successful world build, but
it is quite memory intensive using quarantine. 

The asm code that does our atomic operations for arm6 must be reviewed.
As sshd uses pthreads, jemalloc utilises pthread_mutex calls. 

> 
> Cheers,
> 
> Paul.
> 
> 

Diane
-- 
- db@FreeBSD.org db@db.net http://www.db.net/~db



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130915145238.GB18382>