From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tejun Heo Subject: Re: Resets on sil3124 & sil3726 PMP Date: Sat, 25 Aug 2007 11:19:38 +0900 Message-ID: <46CF91BA.7030608@gmail.com> References: <46C5E3A6.9020907@gmail.com> <98147300-A282-4909-A393-079365833BD8@infogears.com> <46CBA78D.2070405@gmail.com> <46CBCD18.7070507@gmail.com> <46CBDA17.6010009@gmail.com> <46CBDF82.6020206@gmail.com> <7644C0A3-8AEF-4C00-837A-429F73DD616D@infogears.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: Received: from py-out-1112.google.com ([64.233.166.179]:29134 "EHLO py-out-1112.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752193AbXHZKJ4 (ORCPT ); Sun, 26 Aug 2007 06:09:56 -0400 Received: by py-out-1112.google.com with SMTP id u77so272734pyb for ; Sun, 26 Aug 2007 03:09:55 -0700 (PDT) In-Reply-To: <7644C0A3-8AEF-4C00-837A-429F73DD616D@infogears.com> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Rusty Conover Cc: linux-ide@vger.kernel.org Hello, Rusty. Rusty Conover wrote: > I have some interesting results. > > I had a pair of Seagate 250 GB SATA disks (models below) and tried those > out rather then the WD's. At the 1.5 gbps rather they appear to work > just fine both being on the same PMP, at 3.0 gbps they timeout just like > the other disks did. Possibly, the code isn't detecting the max rate of > the disks correctly since these drives only do 1.5 and it attempted to > do 3.0 gbps? Hardware PHY layer is fully responsible of 1.5/3.0 negotiation. The driver can limit the maximum it can go but doesn't have much say in the actual negotiation. Unfortunately, some SATA PHY combinations have problems at 3.0bps. I saw quite a few weird cases. e.g. marvell PMP can't detect older Seagate drives on 3Gbps and can't talk to a first gen Maxtor even at 1.5. I'm still not quite sure how to solve this. Once detected, libata EH will do the right thing and speed down to 1.5 after a few transmission errors but it's a bit trickier before detection. It's probably not worth driving fan-out ports at 3.0Ghz in the first place so maybe limiting PMP fan-out ports to 1.5 by default can alleviate most situations. Oh well, users won't be happy tho. > Have you found that drives can be picky if they will work or not with > PMP ports? As I said above, yeap. Not only PMPs some host PHYs do that too. > I'll be trying to get a bunch of different drives to replace the ones > that don't work with PMP. Yeah, that will be the easiest way out for the moment. Thanks. -- tejun