From owner-freebsd-bugs Thu Mar 7 12:08:11 1996 Return-Path: owner-bugs Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id MAA04653 for bugs-outgoing; Thu, 7 Mar 1996 12:08:11 -0800 (PST) Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id MAA04611 Thu, 7 Mar 1996 12:08:03 -0800 (PST) Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id NAA14613; Thu, 7 Mar 1996 13:04:36 -0700 From: Terry Lambert Message-Id: <199603072004.NAA14613@phaeton.artisoft.com> Subject: Re: ufs_valloc: dup whatever To: sysseh@devetir.qld.gov.au (Stephen Hocking) Date: Thu, 7 Mar 1996 13:04:36 -0700 (MST) Cc: bugs@FreeBSD.org, current@FreeBSD.org In-Reply-To: <199603070758.HAA10277@netfl15a.devetir.qld.gov.au> from "Stephen Hocking" at Mar 7, 96 05:58:39 pm X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-bugs@FreeBSD.org Precedence: bulk > Just noticed something really curious - I was running two makes at once, and > somehow the output of one compilation (an assembler file of all things) > appeared in the middle of a .depend file of a make depend! Help! Terry, John & > David, what's going on? At a guess, I'd say Bruce was the best one to ask. It looks like the VREF/vrele code on the directory traversal has caused a vnode with a non-zero reference count to get onto the free list. I can't say I didn't expect this. It looks like the new pager code being more agressive in recovery is what revealed the traversal race condition. These are just guesses... unfortunately, I can't do anything about any of this, except to recommend going back to a checked out tree of the 1st or 2nd at the latest and waiting for it to be fixed. This is obviously related to your other posting in this thread. Rather than posting each additional problem as you see it, you should work with Bruce, David, and John on resolving the problem using your system as a test bed. I think that otherwise the cascade factor will cause you to post 10 or more problems all related to the same two codependent code changes; the symptoms you've reported so far should be sufficient to resolve the problem. PS: Is it possible that you are running with my "free vnode isn't" kludge fix? I believe this will interact badly with the VREF/vrele changes... if you are, you should remove my kludge, and get around the problem by cranking up numvnodes where the calculation takes place to set it up in the first place ("numvnodes *= 3;" or something similar). Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.