Re: kern/37881: repeatable crash in filesystem when running MP

[Antti Kantee <pooka%netbsd.org@localhost>]

The following reply was made to PR kern/37881; it has been noted by GNATS.

From: Antti Kantee <pooka%netbsd.org@localhost>
To: Andrew Doran <ad%netbsd.org@localhost>
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: kern/37881: repeatable crash in filesystem when running MP
Date: Mon, 28 Jan 2008 00:00:40 +0200

 On Sun Jan 27 2008 at 21:17:06 +0000, Andrew Doran wrote:
 > >  >  Hmm.  Can you see what the other CPU was doing?  Maybe nautilus is evil
 > >  >  enough to close the fd in another thread while the write is in progress
 > >  >  or something like that.
 > >  
 > >  I have 2 traces:
 > >  http://bsdusr.net/~dieter/NetBSD/37881-1.png and
 > >  http://bsdusr.net/~dieter/NetBSD/37881-2.png .
 > 
 > It's possible this is simply missing locking around vnode::v_klist.
 
 Most likely.  But I'm trying to figure out how it's hitting this.  Seems
 like there's no blocking path and the problem is the same every time.
 Seems like suspiciously bad luck.  But guess I the fd can be dropped
 while notifying.
 
 btw, the traces for the same pid on -2.png are "interesting".
 
 -- 
 Antti Kantee <pooka%iki.fi@localhost>                     Of course he runs 
NetBSD
 http://www.iki.fi/pooka/                          http://www.NetBSD.org/
     "la qualité la plus indispensable du cuisinier est l'exactitude"