From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH] Regression: e100_phy_init() isolates even selected PHY, causes 10 seconds boot delay Date: Thu, 29 Oct 2009 06:02:39 -0700 (PDT) Message-ID: <20091029.060239.202729106.davem@davemloft.net> References: <4AC3C09B.3040203@gmx.net> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: bruce.w.allan@intel.com, jeffrey.t.kirsher@intel.com, netdev@vger.kernel.org To: bernhard.kaindl@gmx.net Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:44026 "EHLO sunset.davemloft.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752370AbZJ2NCP (ORCPT ); Thu, 29 Oct 2009 09:02:15 -0400 In-Reply-To: <4AC3C09B.3040203@gmx.net> Sender: netdev-owner@vger.kernel.org List-ID: From: Bernhard Kaindl Date: Wed, 30 Sep 2009 22:33:31 +0200 > The current e100.c:e100_phy_init() electrically isolates all > the PHYs (even the selected PHY -- for a short time!) from the MII. > > This happens only for a short duration before the isolation > of the selected PHY is reverted, but it's enough to cause a > major disturbance in the startup of our e100-based cards: > > On a number of Embedded/Industry Pentium boards which are in use, > the result is that the initial DHCP negotiation takes more > than 10 seconds to complete with 2.6.30 and .31, while it's > done in a fraction of a second with 2.6.29 and earlier > (kernels tested with no delay range from 2.6.23 to 2.6.29) > > That regression was introduced on March 31 in the by a patch > from Bruce which first appeared in 2.6.30-rc3: Bruce, can you give some feedback on this? I'd like to see this issue move forward. The only reason I haven't applied Bernhard's patch is because I haven't seen any feedback from Intel. But I will apply it anyways if I don't see reasonable feedback soon. Thanks.