From owner-freebsd-current Wed Dec 8 8:26:24 1999 Delivered-To: freebsd-current@freebsd.org Received: from pulsar.dead-end.net (pulsar.high-performance.com [216.15.153.82]) by hub.freebsd.org (Postfix) with ESMTP id 2699E14BF8 for ; Wed, 8 Dec 1999 08:26:18 -0800 (PST) (envelope-from rock@dead-end.net) Received: from mailto.dead-end.net (dead-end.net [216.15.153.82] (may be forged)) by pulsar.dead-end.net (8.9.3/8.9.3/DEAD-END/1999020900) with ESMTP id RAA49583; Wed, 8 Dec 1999 17:24:55 +0100 (CET) (envelope-from rock@dead-end.net) Received: from server.rock.net (p3E9E0E3C.dip.t-dialin.net [62.158.14.60]) by mailto.dead-end.net (8.9.3/DEAD-END/1999102400-Customer) with ESMTP id RAA49577; Wed, 8 Dec 1999 17:24:50 +0100 (CET) (envelope-from rock@dead-end.net) Received: from dead-end.net (doom.rock.net [172.23.7.254]) by server.rock.net (8.9.3/8.9.3/ROCK/1999053100) with ESMTP id RAA75135 for ; Wed, 8 Dec 1999 17:23:46 +0100 (CET) (envelope-from rock@dead-end.net) Message-ID: <384E8633.BD23E6CC@dead-end.net> Date: Wed, 08 Dec 1999 17:24:19 +0100 From: "D. Rock" X-Mailer: Mozilla 4.7 [de] (Win98; U) X-Accept-Language: de MIME-Version: 1.0 To: Doug Ambrisko Cc: Subject: Re: ATA driver as the default References: <199912071957.LAA31946@whistle.com> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Doug Ambrisko schrieb: > > D. Rock writes: > | I just re-enabled the ATA driver again after reading the change log > | of better error handling and automatic falldown DMA->PIO under specific > | circumstances. > | But a few days later, while making world (with the ata driver), the > | system > | crashed quite heavily. The file system was totally screwed up afterwards > | (I found my /usr/local after some heavy searching: It magically > | moved to /usr/obj/usr/src/tmp/usr/share/zoneinfo (!) and got tons of > | fsck > | messages). The file system had softupdates enabled. I don't know the > | last kernel messages before the crash (was running X at that time). > > You might want to look at ata-disk.c and the timeout value around line > 438: > > /* start timeout for this transfer */ > if (panicstr) > request->timeout_handle.callout = NULL; > else > request->timeout_handle = > timeout((timeout_t*)ad_timeout, request, 5*hz); > > Originally it was 3s and recently increased to 5s. Personally > I switched it to 30s after it trashed my filesystem when it was 3s. > The issue was that 3s, is that it is to short to wait for my laptop's > drive to spin back up. Sometimes I would get a corrupted read sometimes on > a write it would trash things. I noticed in the old wd driver that > it tried 10s first then a couple 3s timeouts. After making this > change my system has been rock solid when the drive spins down. > Note I haven't tried to tune this value since trashing a 14G filesystem > is pain full. I don't think I have the same problem. My drive definitely doesn't spin down. It sometimes occurs during heavy usage, so the drive should still be very alive. With PIO mode I also don't have any timeout problems. I also had the same DMA problems with the old wd driver and under Windows. The problem is, that the new driver doesn't allow to selectively turn off DMA for problematic devices. I now had commented out the DMA activation code in ata-disk.c (DMA is still activited on the CD-ROM drive though) and I will see how the system behaves. Daniel To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message