From owner-freebsd-hackers Mon Mar 13 02:30:33 1995 Return-Path: hackers-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id CAA00295 for hackers-outgoing; Mon, 13 Mar 1995 02:30:33 -0800 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id CAA00274 for ; Mon, 13 Mar 1995 02:30:21 -0800 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id OAA21901; Mon, 13 Mar 1995 14:51:17 +1000 Date: Mon, 13 Mar 1995 14:51:17 +1000 From: Bruce Evans Message-Id: <199503130451.OAA21901@godzilla.zeta.org.au> To: freebsd-hackers@FreeBSD.org, jbryant@server.iadfw.net Subject: Re: rename bug Sender: hackers-owner@FreeBSD.org Precedence: bulk >> > The bug is when a rename(2) is called, the rename happens as per spec, >> > except that the kernel isn't being updated as to the available space on >> > the filesystem. A reboot is needed to recognize the space is available. >> > >> > For instance every time we add a user, our root filesystem APPEARS to >> > have lost 1508k, this is caused when pwd_mkdb renames pwd.db.tmp to >> > pwd.db, and spwd.db.tmp to spwd.db. This is fixed in FreeBSD-current and wasn't in FreeBSD-2.0. >> > I am confident that it is in the rename(2) system call. Yes, it forgot to clear a reference held by the vm layer, so the old file didn't go away. >> can you show me a script(1) of the bug occuring? cd /tmp; >x; >y; mv x y; fsck -n ... This only consumes one inode because `y' is empty. At first I thought you were saying that there was a problem rebuilding the password database when the disk is full for other reasons. Bruce