From owner-freebsd-hackers Sat Oct 28 22:47:38 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id WAA00282 for hackers-outgoing; Sat, 28 Oct 1995 22:47:38 -0700 Received: from io.org (io.org [142.77.70.2]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id WAA00272 for ; Sat, 28 Oct 1995 22:47:36 -0700 Received: from trepan.io.org (taob@trepan.io.org [198.133.36.8]) by io.org (8.6.12/8.6.12) with SMTP id BAA14926; Sun, 29 Oct 1995 01:47:03 -0400 Date: Sun, 29 Oct 1995 01:47:02 -0400 (EDT) From: Brian Tao To: David Greenman cc: FREEBSD-HACKERS-L Subject: Re: panic: free vnode isn't In-Reply-To: <199510250616.XAA28141@corbin.Root.COM> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-hackers@freebsd.org Precedence: bulk On Tue, 24 Oct 1995, David Greenman wrote: > > Okay...but for DDB this really isn't necessary and only results in you > losing about 4-6MB of memory for all the unused gdb debugging symbols. So I noticed... sysctl -a reported 20 megs of hw.usermem on a 32-meg machine. :( > I think your problem is caused by bug(s) in the NFS client code. While > we do test it relatively thoroughly, we don't pond on it in the way > that you are doing and that likely explains why the bug hasn't been > seen before and fixed. :-) At least it has been isolated to an NFS-related bug. :) The server died again today (2.0.5R was installed on it), but this time it managed to send a message to the console before keeling over completely: biodone: page busy < 0, off: 3162112, foff: 3162112, resid: 4096, index: 0 iozone: 8192, lblkno: 386 valid: 0xff, dirty: 0x0, mapped: 0 panic: biodone: page busy < 0 Then a bunch of attempts at syncing which eventually failed, then it rebooted. I'm in the process of installing 2.1.0-951026-SNAP right now. I'm also going to move our FTP server over to another machine, and mount only the staff home directories (a single 2-gig drive). Since I'm really the only person who uses this machine for interactive use, NFS traffic will be significantly reduced. If it is a problem with NFS load, the machine should be stable. My machines back in Taiwan had everything but the base FreeBSD distribution NFS-mounted from an SGI server (including user home directories and work directories) without encountering this problem. If this incarnation survives for a week, then I'll try moving the FTP service back on and take it from there. Oh, I'm getting "nfs send error 55" in syslog quite often... is this a symptom of an overrun buffer? I should add that the collision rate on our LAN averages 50% and can spike above 100%. It's absolutely horrid, but I hope it isn't the cause of the kernel panics. > Debugging this problem will almost certainly be quite difficult. ... > Unfortunately, without a lot more information, it's unlikely that I'll > be able solve this problem. *sigh* I'll try the latest snapshot and see what happens then. -- Brian Tao (BT300, taob@io.org) Systems Administrator, Internex Online Inc. "Though this be madness, yet there is method in't"