From owner-freebsd-current Mon Jan 12 13:07:21 1998 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id NAA00491 for current-outgoing; Mon, 12 Jan 1998 13:07:21 -0800 (PST) (envelope-from owner-freebsd-current) Received: from dyson.iquest.net (dyson.iquest.net [198.70.144.127]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id NAA00396; Mon, 12 Jan 1998 13:06:32 -0800 (PST) (envelope-from toor@dyson.iquest.net) Received: (from root@localhost) by dyson.iquest.net (8.8.8/8.8.8) id QAA01060; Mon, 12 Jan 1998 16:05:11 -0500 (EST) (envelope-from toor) From: "John S. Dyson" Message-Id: <199801122105.QAA01060@dyson.iquest.net> Subject: Re: random (?) SIGBUS in -current In-Reply-To: <1160.884638463@critter.freebsd.dk> from Poul-Henning Kamp at "Jan 12, 98 09:54:23 pm" To: phk@critter.freebsd.dk (Poul-Henning Kamp) Date: Mon, 12 Jan 1998 16:05:11 -0500 (EST) Cc: toor@dyson.iquest.net, julian@whistle.com, dyson@freebsd.org, current@freebsd.org X-Mailer: ELM [version 2.4ME+ PL31 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Poul-Henning Kamp said: > In message <199801122047.PAA00534@dyson.iquest.net>, "John S. Dyson" writes: > >Poul-Henning Kamp said: > >> > >> It looks like some page gets stolen, and the object isn't updated > >> accordingly... > >> > >> >Note that the new code follows the "rules" fairly precisely regarding the > >> >vnode free list. Some parts of the system might not be able to deal with > >> >it yet (maybe.) > >> > >> But this would be related to pages, they die with a pagefault... > >> > >Vnodes, objects and pages are all intimately related to each other. > > Yes, but the program runs part of the way before it dies, so we're > not talking an entire vnode, object, but only a page or two... > When the vnodes are on the wrong list, exactly the same kind of problem can occur. Pages are and always have been protected off before freeing. If that is working correctly (which it should be), a new fault should bring a missing page back. If the pointers to the object or vnode are messed up, then a page will likely be demand zeroed, the wrong file will be referenced, or an error will be detected. It most likely isn't a problem with pages, but a problem with dangling references due to the new mechanism for vnode mangement. I already have seen a potential problem with the vnode management, and am trying to get it working before my flight to the Bay area. -- John | Never try to teach a pig to sing, dyson@freebsd.org | it just makes you look stupid, jdyson@nc.com | and it irritates the pig.