From owner-freebsd-current Mon Sep 16 20:17:36 2002 Delivered-To: freebsd-current@freebsd.org Received: from mx1.FreeBSD.org (mx1.FreeBSD.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 7225C37B400; Mon, 16 Sep 2002 20:17:34 -0700 (PDT) Received: from gull.mail.pas.earthlink.net (gull.mail.pas.earthlink.net [207.217.120.84]) by mx1.FreeBSD.org (Postfix) with ESMTP id 13A1643E72; Mon, 16 Sep 2002 20:17:34 -0700 (PDT) (envelope-from tlambert2@mindspring.com) Received: from pool0049.cvx22-bradley.dialup.earthlink.net ([209.179.198.49] helo=mindspring.com) by gull.mail.pas.earthlink.net with esmtp (Exim 3.33 #1) id 17r8re-0004dr-00; Mon, 16 Sep 2002 20:17:26 -0700 Message-ID: <3D869B5F.C48D7F10@mindspring.com> Date: Mon, 16 Sep 2002 20:02:55 -0700 From: Terry Lambert X-Mailer: Mozilla 4.79 [en] (Win98; U) X-Accept-Language: en MIME-Version: 1.0 To: Martin Blapp Cc: current@freebsd.org, Michael Reifenberger , Peter Wemm , julian@freebsd.org Subject: Re: filesystem corruption ? References: <20020917021615.D3162-100000@levais.imp.ch> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG Martin Blapp wrote: > It looks more and more to me that pmap does something wrong. > I get pmap related vm crashes or corruption, relating in > filesystem corruption. > > I had about 3-4 different panics. Mozilla build tends to prefer > "panic: bad link count", openoffice prefers page faults in ffs code ;) > > But these options here are enabled: > > options DISABLE_PSE > options DISABLE_PG_G > > Are there any options I could also add and try ? Page coloring etc ? > > You remember, I had SIG4 and SIG11 over and over until > I used these options. The builds run fine by then, > I had no panics at all anymore. I was happy. [ ... ] There have been a number of panics reported recently on -current, which seem to be repeatable on ATA drives, but not on SCSI drives. A casual perusal of the code seems to indicate that a soft failure in the paging path will be treated as a hard failure in the ATA case, instead of retried, but I haven't confirmed this by writing the code to tape and running the tape between my fillings (i.e. I have not intensly scrutinized the code, it just looks like it is that way). UnixWare had a similar bug in th ATA retry code, only it's FS, when a soft error was treated as hard, would start marking the sectors gone, so maybe you should consider yourself lucky... ;^). Last I heard,the ATA maintainers were looking into reproducing the problem, with no luck yet. Check the -current archives, and volunteer to be a guinea pig (that's my best suggestion at present, unless you can retry with a SCSI drive instead, and see if the problems disappear). You might also try playing with the ATA DMA and tag options (see NOTES/LINT for what you can turn off that way), and the ATA sysctl's (most are required to be done at boot time, but a "sysctl -a | grep ata" will list them out for you to try manually in the boot loader). -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message