From owner-freebsd-current Tue Dec 21 10:12:52 1999 Delivered-To: freebsd-current@freebsd.org Received: from gatesrv.rz.unibw-muenchen.de (gatesrv.RZ.UniBw-Muenchen.de [137.193.10.21]) by hub.freebsd.org (Postfix) with ESMTP id E7D2014A12 for ; Tue, 21 Dec 1999 10:12:47 -0800 (PST) (envelope-from didi@Xterminator.STUDFB.UniBw-Muenchen.de) Received: from XTerminator.StudFb.UniBw-Muenchen.De (Xterminator.STUDFB.UniBw-Muenchen.de [137.193.211.64]) by gatesrv.rz.unibw-muenchen.de (8.9.1b+Sun/8.9.3) with ESMTP id TAA16607 for ; Tue, 21 Dec 1999 19:12:38 +0100 (MET) Received: from didi.STUDFB.UniBw-Muenchen.de (didi.STUDFB.UniBw-Muenchen.de [137.193.211.220]) by XTerminator.StudFb.UniBw-Muenchen.De (8.8.7/8.8.6) with SMTP id TAA24329 for ; Tue, 21 Dec 1999 19:11:50 +0100 From: Dieter Rothacker To: freebsd-current@FreeBSD.ORG Subject: Is AUTO_EOI better? [was:Re: ATA errors and AUTO_EOI] Date: Tue, 21 Dec 1999 19:12:37 +0100 Organization: Organized ? No way... Message-ID: References: <199912211700.SAA13589@dorifer.heim3.tu-clausthal.de> <385FA720.FAC2CB0E@dead-end.net> In-Reply-To: <385FA720.FAC2CB0E@dead-end.net> X-Mailer: Forte Agent 1.6/32.661 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Tue, 21 Dec 1999 17:13:20 +0100, D. Rock wrote: >Oliver Fromme schrieb: >> Doug White wrote in list.freebsd-current: >> > On Tue, 21 Dec 1999, Soren Schmidt wrote: >> > > It seems Dieter Rothacker wrote: >> > > > The solution for me was to recompile the kernel without AUTO_EOI1 and >> > > > AUTO_EOI2. >> > > >> > > Those options newer worked (for me at least) reliably with anything, could >> > > those that are seeing the hangs please check this ?? >> > >> > Although this isn't immediately related to ATA, I've found that Intel >> > L440GX+ boards *hate* AUTO_EOI_2 when running SMP. They freeze going into >> > multiuser mode. Took me quite a while to figure that out. >> >> I have always been using AUTO_EOI_1, but _not_ AUTO_EOI_2, and >> it has always worked very well. >> >> The comment in LINT about AUTO_EOI_2 sounds pretty suspicous, >> so I never even tried it: "it works for some clones and some >> integrated versions." That sounds to me like "it works on a >> very limited set of hardware (and if you're lucky)." >> >> AUTO_EOI_1 seems to be fine, though. >Same for me. Yeah, you are right. My system is now running with a kernel with AUTO_EOI_1. Seems like AUTO_EOI_2 really was the only problem... Does somebody have any actual evidence that the AUTO_EOI really boosts performance on modern integrated chipsets like the 440BX? -- Dieter Rothacker To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message