Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2.fluke 9/24/84; site fluke.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!ihnp4!houxm!vax135!cornell!uw-beaver!fluke!joe
From: joe@fluke.UUCP (Joe Kelsey)
Newsgroups: net.bugs.4bsd
Subject: 4.2 hangs with locked root inode after full file system
Message-ID: <2160@vax4.fluke.UUCP>
Date: Tue, 19-Feb-85 12:33:10 EST
Article-I.D.: vax4.2160
Posted: Tue Feb 19 12:33:10 1985
Date-Received: Fri, 22-Feb-85 08:16:00 EST
Organization: John Fluke Mfg. Co., Inc., Everett, WA
Lines: 26

Index:	sys/ufs_alloc.c 4.2BSD +FIX

Description:
	In a similar posting, Jim McGinness <decvax!jmcg>, notes that
	if the file system becomes full or nearly full, a bug in the
	buffer locking protocol in alloccg and ialloccg will cause
	an eventual hung system due to a locked buffer eventually
	manifesting itself as a locked root inode.  Unfortunately, he
	neglected to point out one more place to fix.

Repeat-By:
	Fill up the file system, then wait a while.

Fix:
	In ufs_alloc.c, function alloccg(), add the lines:

	if (bno < 0)
+++	{
+++		brelse(bp);
		return (NULL);
+++	}

	Jim pointed two other places in alloccg that also need to be
	fixed, and one place in ialloccg.

/Joe