Date: Sun, 24 Mar 2002 09:49:19 +0200 (SAST) From: Willie Viljoen <will@laserfence.net> To: Robert Watson <rwatson@freebsd.org> Cc: John Hanley <jh_@yahoo.com>, <freebsd-fs@freebsd.org>, <re@freebsd.org> Subject: Re: Soft update instability with heavy IO and offboard IDE controller Message-ID: <20020324094355.X212-100000@phoenix.vh.laserfence.net> In-Reply-To: <Pine.NEB.3.96L.1020323190730.47668K-100000@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Well, I've yet to experiment with this, but I'm going to try and disable write caching in other operating systems where this configuration also has data loss and such problems. Might it be the problem is with the hardware, and not necessary the kernel? If anybody out there has anything similar to try and recreate this, the controller is a CMD-649 PCI ATA-100 (I know I said that before, but repeat it just incase) and the drive is a Seagate ST340810A 40GB 5400RPM ATA100. The drive is connected with standard 80-conductor cable which was supplied with the controller and runs at full speed (UDMA mode 4) I remember in the past some CMD controllers have given semi-serious troubles, specifically the 64x range, for which Linux developers have had to code several performance destroying kernel level workarounds. As for performance, I didn't really notice any significant hit on the system, so it might be that on this configuration write caching works so badly it doesn't really speed things up. Will On Sat, 23 Mar 2002, Robert Watson wrote: > > On Sat, 23 Mar 2002, John Hanley wrote: > > > > turn off write caching at boot time in > > > /boot/loader.conf by adding the line hw.ata.wc=0 > > > > Surely a default install should result in write caching disabled, right? > > Since any new filesystems will default to using soft update. > > We turned it off by default in 4.4-RELEASE, I believe, and got utterly > pounded in the benchmarks, magazine reviews, etc, and turned it back on > for later RELENG_4 releases. On 5.0-CURRENT, it's off again by default > due to the assumptions concerning background fsck, which is on by default. > At some point, we need to have that whole discussion again. Basically, > vendors ship hardware that don't conform to standards specifications, and > consistency guarantees are thrown out the window if you use the default > settings. If you disable write caching, you will see a very high > performance hit on most modern disks. > > I suppose we need to think about this for DP1 also. > > Robert N M Watson FreeBSD Core Team, TrustedBSD Project > robert@fledge.watson.org NAI Labs, Safeport Network Services > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-fs" in the body of the message > > > -- Willie Viljoen Private IT Consultant 214 Paul Kruger Avenue Universitas Bloemfontein 9321 South Africa +27 51 522 15 60, a/h +27 51 522 44 36 +27 82 404 03 27 will@laserfence.net To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-fs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020324094355.X212-100000>