Date: Sun, 11 Apr 1999 12:00:04 -0400 From: "David E. Cross" <crossd@cs.rpi.edu> To: Matthew Dillon <dillon@apollo.backplane.com> Cc: "David E. Cross" <crossd@cs.rpi.edu>, freebsd-hackers@freebsd.org, freebsd-security@freebsd.org, schimken@cs.rpi.edu Subject: Re: increased crashing in NFS server Message-ID: <199904111600.MAA24421@cs.rpi.edu> In-Reply-To: Message from Matthew Dillon <dillon@apollo.backplane.com> of "Sun, 11 Apr 1999 00:03:49 PDT." <199904110703.AAA03493@apollo.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> :I recently updated all of our FreeBSD3 clients to use NFSv3/UDP when > :contacting our servers (FreeBSD3 of the same build tree). We have > :noticed an increase in crashing of our main home directory server > :(which is the only server really handling RW mounts, our other servers > :are mostly RO, with some minor RW activity.) The first crash was > :obviously NFS. I traced it to one of 2 possible crash points in the > :kernel (sorry, no stack trace, we don't [yet] have a crashlogs enabled > :for that machine.). > : > :The panic was: > :mbuf siz=33476 > :panic: Bad nfs svc reply > > You are using a 32K file block size? If so, reduce it to 8K. > > I think you've just shown us a security hole in the NFS system -- it > panics if it is given too large a response packet. Oops. It should > just print a message and drop the packet. We are using AMD, so I have no idea what size it is using. I will again say that it would be really nice to be able to query the full mount options. I have tried statfs(2), but it does not seem to have what I want. I will tell amd to use rsize/wsize of 8192 to force the issue. It is becoming slightly annoying that amd would cdefault to unstable options ;) > :The second panic just happened, it claims to be softupdate related. I > :think it may have something to do with NFSv3 however since this machine > :used to be very stable (ie, not 2 crashes in a week). > : > :This panic was: > :panic: softdep_write_inodeblock: indirect pointer #0 mismatch 0 != 102192 > :8 > :syncing disks... panic: softdep_lock: locking against myself > :... > : The Regents of the University of California. All rights reserved. > :FreeBSD 3.1-STABLE #0: Sun Mar 21 02:23:19 EST 1999 > : schimken@wobble.cs.rpi.edu:/usr/src/sys/compile/STAGGER > :... > > This is when you built it. When did you check the source out of the CVS > tree? > > There have been a few commits in this area since Mar 21st, you definitely > want to update your sources, but I don't think any of the commits address > the above softupdates panic. I've never seen that panic before. > > I recommend fsck'ing all your filesystems from single-user just in case > there's some garbage in there. Ok, we will do all of that when we can schedule a convienient downtime. The sources were checked out arround noon, Fri, Mar 26. Is it possibly a NFS3/Softupdates problem? All the exported filesystems on that drive are softupdate mounted, and the only access to that filesystem is via NFS. -- David Cross | WinNT:Linux::Linux:FreeBSD Systems Administrator/Research Programmer | Rensselaer Polytechnic Institute, | This space intentionally Department of Computer Science | left unblank I speak only for myself. | To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199904111600.MAA24421>