From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tejun Heo Subject: [PATCHSET] libata: update EH speed down logic Date: Fri, 2 Nov 2007 00:51:36 +0900 Message-ID: <11939323042554-git-send-email-htejun@gmail.com> Return-path: Received: from [222.235.223.38] ([222.235.223.38]:20822 "EHLO htj.dyndns.org" rhost-flags-FAIL-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1754092AbXKAPvr (ORCPT ); Thu, 1 Nov 2007 11:51:47 -0400 Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: jeff@garzik.org, linux-ide@vger.kernel.org Hello, This patchset updates EH speed down logic such that... * Speed down rules are less conservative. For example, falling back to PIO was triggered if there have been more than 10 ATA_BUS, TOUT_HSM or UNK_DEV in 5 minutes. However, 11 timeouts can't happen in five minutes as each timeout is 30 secs long. * Consecutive errors occurring right after data transfer mode (NCQ on/off and xfermode) has been changed trigger fast back-to-back speed downs. This helps libata EH find the correct transfer mode faster such that the user is bored less and higher level probing (partition scanning) doesn't fail. This patchset contains 8 patches. The first five prepare for speed down logic update. The 6th one relaxes existing speed down rules. The last two implement fast speed down after transfer mode change. This patchset is on top of update-EH-configurat-behavior patchset[1]. Thanks. -- tejun [1] http://thread.gmane.org/gmane.linux.ide/24281