From mboxrd@z Thu Jan  1 00:00:00 1970
From: swise@opengridcomputing.com (Steve Wise)
Date: Tue, 21 Jun 2016 09:18:43 -0500
Subject: target crash / host hang with nvme-all.3 branch of nvme-fabrics
In-Reply-To: <57668896.7090308@grimberg.me>
References: <00d801d1c7de$e17fc7d0$a47f5770$@opengridcomputing.com>
 <20160616145724.GA32635@infradead.org>
 <017001d1c7e7$95057270$bf105750$@opengridcomputing.com>
 <5763044A.9090206@grimberg.me>
 <005f01d1c8a1$5a229240$0e67b6c0$@opengridcomputing.com>
 <007801d1c8a2$d6056530$82102f90$@opengridcomputing.com>
 <CAF1ivSZNJhOtORceAfk9Ym9SPpYOGQ4q0KOD8eEuy0ovU2Eefw@mail.gmail.com>
 <57668896.7090308@grimberg.me>
Message-ID: <006d01d1cbc7$d0d3b1c0$727b1540$@opengridcomputing.com>

> On 17/06/16 20:20, Ming Lin wrote:
> > On Fri, Jun 17, 2016 at 7:16 AM, Steve Wise <swise at opengridcomputing.com>
> wrote:
> >>>
> >>>>
> >>>> Steve, is this something that started happening recently? does the
> >>>> 4.6-rc3 tag suffer from the same phenomenon?
> >>>
> >>> Where is this tag?
> >>
> >> Never mind.  I found it (needed 'git pull -t' for pull the tags from the
gitlab
> >> nvmef repo).
> >
> > I run this overnight,
> >
> > #!/bin/bash
> >
> > while [ 1 ] ; do
> >      ifconfig eth5 down ; sleep 15; ifconfig eth5 up; sleep 15
> > done
> >
> > Although the crash is not reproduced, but it triggers another OOM bug.
> 
> Which code-base is this? it looks like this code is just leaking queues.
> Obviously something changed...
> 

Yoichi has hit what is apparently the same OOM bug:

Kernel panic - not syncing: Out of memory and no killable processes...

He hit it with NVMF on the host, _and_ with NVME/PCI with a local SSD.  This is
all with using nvmf-all.3 + Christoph's fix for the target queue deletion crash
(http://lists.infradead.org/pipermail/linux-nvme/2016-June/005075.html).

Ming, any luck on isolating this?  I'm going to enable kernel memory leak
detection and see if I can figure this out.  

Steve.