From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tejun Heo Subject: Re: [PATCH 2.6.19 2/3] sata_promise: new EH conversion Date: Wed, 06 Dec 2006 18:19:18 +0900 Message-ID: <45768B16.5020603@gmail.com> References: <200612060853.kB68r0Gg024641@harpo.it.uu.se> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: Received: from nz-out-0506.google.com ([64.233.162.234]:31837 "EHLO nz-out-0102.google.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1760365AbWLFJTZ (ORCPT ); Wed, 6 Dec 2006 04:19:25 -0500 Received: by nz-out-0102.google.com with SMTP id s1so89063nze for ; Wed, 06 Dec 2006 01:19:24 -0800 (PST) In-Reply-To: <200612060853.kB68r0Gg024641@harpo.it.uu.se> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Mikael Pettersson Cc: jeff@garzik.org, linux-ide@vger.kernel.org, linux-kernel@vger.kernel.org Mikael Pettersson wrote: >> Don't freeze port unconditionally. You'll end up hardresetting on every >> error. Just make sure DMA engine is stopped and the controller is in a >> sane state. If that fails, then, the port should be frozen. Sorry, s/hardresetting/resetting/ > I'm looking into this now, but so far it seems only a reset > (what Promise calls software reset, I don't know if libata > considers it a soft or hard reset) of the ATA channel will do. Errors properly reported by the device shouldn't cause resets. Think about ATAPI check condition. >>> + hardreset = NULL; >>> + if (sata_scr_valid(ap)) { >>> + ehc->i.action |= ATA_EH_HARDRESET; >> Why always force HARDRESET? > > I based that on sata_sil24: > > if (sil24_init_port(ap)) { > ata_eh_freeze_port(ap); > ehc->i.action |= ATA_EH_HARDRESET; > } > > I interpreted the ATA_EH_HARDRESET as being required due to > the ata_eh_freeze_port(), but perhaps it's only there because > sil24_init_port() returned failure? Yeap, that's right. > A different issue, but of practical importance, is which > libata branch I should base the EH conversion on: #upstream > or #ALL? Andrew Morton's -mm kernels include the ALL patches, > but they in turn include the promise-sata-pata patches, and > there is a conflict between the PATA patch and the EH conversion. > Currently my EH conversion is based on #upstream, and I've ported > the PATA patch to apply on top of it. #upstream, It is. #ALL is merge of all libata-dev devel branches and no development work occurs there directly. Thanks. -- tejun