Re: kern/38762: panic: vwakeup: neg numoutput

[Antti Kantee <pooka%cs.hut.fi@localhost>]

The following reply was made to PR kern/38762; it has been noted by GNATS.

From: Antti Kantee <pooka%cs.hut.fi@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: Andreas Gustafsson <gson%gson.org@localhost>
Subject: Re: kern/38762: panic: vwakeup: neg numoutput
Date: Fri, 22 Aug 2008 14:35:21 +0300

 On Fri Aug 22 2008 at 11:25:01 +0000, Andreas Gustafsson wrote:
 >  After running a number of additional tests, some successfully
 >  reproducing the panic and some not, I think I have figured out a
 >  pattern: the panic is easily reproduced when the system is booted from
 >  a live CD (built using mklivecd from pkgsrc), but I can't reproduce it
 >  when booting from the network.
 >  
 >  I haven't tried reproducing the bug using a system booted from a hard
 >  disk yet, but based on the fact that this panic has not been widely
 >  reported or reproduced, I assume the panic does not occur when booting
 >  from a hard disk but is somehow specific to the live CD case. 
 
 Hmm, sounds suspicious.  Are you *sure* you're trying the exact same
 kernel, machine, usb stick, etc?
 
 Seems like the data flush triggers the problem, as the echo command has
 returned already.  So I'm guessing "sync" is enough instead of unmount.
 
 Can you do "show vnode" for the problematic vnode?
 
 If you have puffs/rump available, can you see if it tickles the problem
 also?  rump_ffs /dev/rsd0a /mnt ; echo foo > /mnt/bar ; sync