From owner-freebsd-current Sun Jan 10 11:34:05 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id LAA03464 for freebsd-current-outgoing; Sun, 10 Jan 1999 11:34:05 -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 LAA03455; Sun, 10 Jan 1999 11:34:01 -0800 (PST) (envelope-from dillon@apollo.backplane.com) Received: (from dillon@localhost) by apollo.backplane.com (8.9.1/8.9.1) id LAA86598; Sun, 10 Jan 1999 11:33:28 -0800 (PST) (envelope-from dillon) Date: Sun, 10 Jan 1999 11:33:28 -0800 (PST) From: Matthew Dillon Message-Id: <199901101933.LAA86598@apollo.backplane.com> To: Luoqi Chen Cc: dg@FreeBSD.ORG, syssgm@dtir.qld.gov.au, freebsd-current@FreeBSD.ORG Subject: Re: Hangs on "inode" and "thrd_sleep" Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG :I ran into the same problem before. If you search the -current archive :for deadlock, I posted a patch to solve this problem (I should have filed :a PR for this). There are more serious problems with locking in vm_fault(), :which are more difficult to fix, see PR 8416 by Tor Egge. : :-lq : :> My test machine hung last night during 'make -j5 buildworld' with 7 processes :> in "thrd_sleep" and 2 in "inode". Thus began a marathon DDB session :... :> Your patch to vm_fault.c in PR 8416 looks reasonable. If you have an additional patch ( or is the one in PR8416 the one you 'posted' ??) I'd appreciate it if you could email it to me. In general, a lot of the confusion in the VM system stems from it having to effectively bypass vm_object's and special-case vnode related ops. What should really happen is that all VM ops should run through the associated vm_object's and the backing store underlying those objects should be hidden. Also, it should be possible to handle other VM related deadlocks by locking-through to get a vm_page, lock the page, and then hold the lock on the page while releasing all the other locks. The PG_BUSY/vm_page->busy hacks could be used to effect here. Fixing this would also solve the lock ordering problem. I'll put cleaning this up on my list for 4.0. -Matt Matthew Dillon Engineering, HiWay Technologies, Inc. & BEST Internet Communications & God knows what else. (Please include original email in any response) To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message