From mboxrd@z Thu Jan 1 00:00:00 1970 From: Robert Hancock Subject: Re: [PATCH] sata_nv: Don't attempt using ADMA for (READ|SET)_MAX commands Date: Fri, 13 Apr 2007 16:52:38 -0600 Message-ID: <462009B6.4030406@shaw.ca> References: <20070413170831.GC27674@athena.road.mcmartin.ca> <20070413181935.6d60f230@the-village.bc.nu> <461FC892.6020602@rtr.ca> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from shawidc-mo1.cg.shawcable.net ([24.71.223.10]:41041 "EHLO pd4mo1so.prod.shaw.ca" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S965657AbXDMWxl (ORCPT ); Fri, 13 Apr 2007 18:53:41 -0400 Received: from pd2mr6so.prod.shaw.ca (pd2mr6so-qfe3.prod.shaw.ca [10.0.141.9]) by l-daemon (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0JGG009DCKW45500@l-daemon> for linux-ide@vger.kernel.org; Fri, 13 Apr 2007 16:52:52 -0600 (MDT) Received: from pn2ml10so.prod.shaw.ca ([10.0.121.80]) by pd2mr6so.prod.shaw.ca (Sun Java System Messaging Server 6.2-7.05 (built Sep 5 2006)) with ESMTP id <0JGG00JVZKW3OIW0@pd2mr6so.prod.shaw.ca> for linux-ide@vger.kernel.org; Fri, 13 Apr 2007 16:52:52 -0600 (MDT) Received: from [192.168.1.113] ([70.64.1.86]) by l-daemon (Sun ONE Messaging Server 6.0 HotFix 1.01 (built Mar 15 2004)) with ESMTP id <0JGG00K5GKW2A3Y1@l-daemon> for linux-ide@vger.kernel.org; Fri, 13 Apr 2007 16:52:51 -0600 (MDT) In-reply-to: <461FC892.6020602@rtr.ca> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Mark Lord Cc: Alan Cox , Kyle McMartin , linux-ide@vger.kernel.org, jgarzik@pobox.com, bcollins@ubuntu.com Mark Lord wrote: > Alan Cox wrote: >> On Fri, 13 Apr 2007 13:08:31 -0400 >> Kyle McMartin wrote: >> >>> READ_NATIVE_MAX and SET_MAX were causing timeouts on sata_nv >>> controllers. >>> Disabling ADMA helped, but that is quite a large hammer to use. >>> Reverting >>> 382a6652e91b34d5480cfc0ed840c196650493d4 also helped, but we might as >>> well >>> fix it right, instead of disabling the performance gain on cache flushes >>> by using ADMA mode. >> >> Probably not going to make any performance difference to blacklist all >> non-data commands or all but a few like cache-flush. > > I agree. The Pacific Digital ADMA stuff had some quirks for various > non-data > commands as well, and the sensible thing was to just not use ADMA for > anything > other than READs/WRITEs and possible CACHE FLUSHes. I'm told by NVidia that non-data commands should be fine in ADMA mode, except for those that require reading back a result taskfile (which can't be done in ADMA mode). The patch in libata-dev which I mentioned in another email will ensure that we don't try to use ADMA for such commands, so this patch should not be needed. -- Robert Hancock Saskatoon, SK, Canada To email, remove "nospam" from hancockr@nospamshaw.ca Home Page: http://www.roberthancock.com/