From owner-freebsd-current  Wed Jan 20 12:24:07 1999
Return-Path: <owner-freebsd-current@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id MAA13428
          for freebsd-current-outgoing; Wed, 20 Jan 1999 12:24:07 -0800 (PST)
          (envelope-from owner-freebsd-current@FreeBSD.ORG)
Received: from apollo.backplane.com (apollo.backplane.com [209.157.86.2])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id MAA13422
          for <current@FreeBSD.ORG>; Wed, 20 Jan 1999 12:24:06 -0800 (PST)
          (envelope-from dillon@apollo.backplane.com)
Received: (from dillon@localhost)
	by apollo.backplane.com (8.9.2/8.9.1) id MAA06937;
	Wed, 20 Jan 1999 12:23:59 -0800 (PST)
	(envelope-from dillon)
Date: Wed, 20 Jan 1999 12:23:59 -0800 (PST)
From: Matthew Dillon <dillon@apollo.backplane.com>
Message-Id: <199901202023.MAA06937@apollo.backplane.com>
To: Bjoern Fischer <bfischer@TechFak.Uni-Bielefeld.DE>, current@FreeBSD.ORG
Subject: Re: panic: vinvalbuf: dirty bufs (during reboot, several times)
References: <19990120073445.A402@titan.klemm.gtn.com> <199901200700.XAA76802@apollo.backplane.com> <19990120132700.A15521@voliere.TechFak.Uni-Bielefeld.DE> <199901201923.LAA05838@apollo.backplane.com>
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

    Try this.  In nfs_bioread(), nfs/nfs_bio.c:

    Before, it was:

                if (getpages && !(bp->b_flags & B_VMIO)) {
#ifdef DIAGNOSTIC
                        printf("nfs_bioread: non vmio buf found, discarding\n");


    Try changin the if() to this:

                if (
                    (getpages && !(bp->b_flags & B_VMIO)) ||
                    (bp->b_flags & (B_CACHE|B_DELWRI)) == B_DELWRI
                ) {
#ifdef DIAGNOSTIC
                        printf("nfs_bioread: non vmio buf found, discarding\n");
#endif

    I believe what is going on is that bioread() is misinterpreting B_CACHE
    to mean that it can discard the entire buffer.  If B_DELWRI is set,
    however, it must sync the buffer first.

    What is occuring is that when a program write()'s non-contiguously and
    then lseek's back and read()'s again, B_CACHE is getting cleared and the
    buffer is being re-read from NFS without first being flushed to NFS,
    causing the written data to be overwritten by the read.

    I have NOT tested this well.  It seems to solve the vi SEG fault problem.

					-Matt

					Matthew Dillon 
					<dillon@backplane.com>


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message