From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Matt Carlson" Subject: Re: [PATCH net-next 2/2] tg3: Scale back code that modifies MRRS Date: Mon, 28 Nov 2011 14:45:58 -0800 Message-ID: <20111128224558.GA20447@mcarlson.broadcom.com> References: <1322509264-19636-3-git-send-email-mcarlson@broadcom.com> <1322510036.26733.12.camel@bwh-desktop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: "Matthew Carlson" , "davem@davemloft.net" , "netdev@vger.kernel.org" , "Michael Chan" To: "Ben Hutchings" Return-path: Received: from mms1.broadcom.com ([216.31.210.17]:2675 "EHLO mms1.broadcom.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754975Ab1K1WpU (ORCPT ); Mon, 28 Nov 2011 17:45:20 -0500 In-Reply-To: <1322510036.26733.12.camel@bwh-desktop> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: On Mon, Nov 28, 2011 at 11:53:56AM -0800, Ben Hutchings wrote: > On Mon, 2011-11-28 at 11:41 -0800, Matt Carlson wrote: > > Tg3 normally gets a performance boost by increasing the PCI Maximum Read > > Request Size (MRRS) to 4k. Unfortunately, this is causing some problems > > on particular hardware platforms. This patch removes all code that > > modifies the MRRS except for one case. > > > > As part of a solution to fix an internal FIFO problem on the 5719, the > > driver artificially capped the MRRS to 2k for the entire 5719, and later > > 5720, ASIC revs. This was overly aggressive and only really needed to > > be done for the 5719 A0. In the spirit of the rest of this patch, the > > driver will only reprogram the MRRS for this device if the value exceeds > > the 2k cap. > [...] > > It may be better to do this as a PCI quirk, so that the generic MPS/MRRS > configuration code has this information. If at all possible, I'd rather do that as a follow-on patch. This patch fixes a critical problem that is blocking a lot of testing.