Re: Another kassert in xbdback_xenbus.c

["Michael L. Hitch" <mhitch%lightning.msu.montana.edu@localhost>]

On Thu, 11 Aug 2011, Jeff Rizzo wrote:

> I updated my dom0 last night, and shortly after the domUs started up, it 
> triggered another KASSERT.  It's been running OK since a reboot, though, so 
> it looks like it's not quite as easy to trigger as the earlier ones (and may 
> or may not be a new problem).
>
> panic: kernel diagnostic assertion "xbdi->xbdi_pendingreqs == 0" failed: file 
> "/Users/riz/Documents/code/netbsd/src/sys/arch/xen/xen/xbdback_xenbus.c", 
> line 1017

  I had tried running a -current dom0 kernel and hit this panic twice 
fairly quickly.  It appears related to disk I/O, and the 2 VMs I was 
running are fairly heavy on disk I/O.  I switched back to a 5.1_STABLE 
dom0 for a while, but I looked at that KASSERT a bit, but wasn't able to 
follow all the code all that well.  The KASSERT had been added somewhat 
recently.  I changed the KASSERT to an #ifdef DIAGNOSTIC check so I at 
least wouldn't panic, and while testing the MP domu kernel had a couple of 
hits with this check last night:

xbdback_co_cache_flush: xbdi->xbdi_pendingreqs 10
xbdback_co_cache_flush: xbdi->xbdi_pendingreqs 1

  These were the only 2 occurances during an uptime of about an hour 
before the domu VM hung.

--
Michael L. Hitch                        mhitch%montana.edu@localhost
Computer Consultant
Information Technology Center
Montana State University        Bozeman, MT     USA