From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261249AbTIYOoD (ORCPT ); Thu, 25 Sep 2003 10:44:03 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261260AbTIYOoD (ORCPT ); Thu, 25 Sep 2003 10:44:03 -0400 Received: from eliquid.com ([66.33.0.12]:54281 "EHLO ns.eliquid.com") by vger.kernel.org with ESMTP id S261249AbTIYOn7 (ORCPT ); Thu, 25 Sep 2003 10:43:59 -0400 Subject: Sil SATA Controlers From: Xavier Spriet To: LKML Content-Type: text/plain Organization: eliquidMEDIA International Inc. Message-Id: <1064501009.14931.42.camel@shadow.eliquid.com> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.4 Date: Thu, 25 Sep 2003 10:43:30 -0400 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org -- Sincerely, Xavier Spriet IT Director eliquidMEDIA International Inc. Phone: (1-519)-973-1930 1-800 : (1-800)-561-7525 Fax : (1-519)-253-0337 >>From what I've gathered reading the archives, the Sillicon Image SATA controlers that appear to ship with many newer boards is a rather problematic controller and apparently, the latest drivers for these controllers do not seem to provide the performance and stability expected. >>From the information I was able to find on this list, the development of a proper solution is a difficult challenge because of the NDA implemented by Sil. As I have not been able to find out the terms of this NDA, I'd like to know if this NDA would actually prevent kernel developers to come up with a proper driver. If this is the case as it appears to be, would it be a good idea to both complain to Sil for the terms of this NDA and ask them to be a little more reasonable, and let my board's manufacturer (ASUS) know about the issue and hope they will put pressure on Sil to change the terms of this NDA or come up with an open-source driver or documentation, etc..? Currently, I've been able to restore the high performance with this driver in my scenario by forcing rqsize to remain at 128 instead of switching to 15, and setting non-used interfaces back to 15. Unfortunately, this seems to have brought back some occasional hangs of the system. Thanks, Xavier.