From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tejun Heo Subject: Re: ICH7m problem using libata Date: Tue, 16 Jan 2007 20:42:23 +0900 Message-ID: <45ACBA1F.6030007@gmail.com> References: <1165253799.16504.37.camel@laundromat.jangutter.com> <200701100817.28738.matthew4196@gmail.com> <45AB0F07.5020705@gmail.com> <200701160958.33561.matthew4196@gmail.com> <45AC933B.1000201@gmail.com> <1168947105.14636.11.camel@laundromat.jangutter.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: Received: from wr-out-0506.google.com ([64.233.184.224]:14379 "EHLO wr-out-0506.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750962AbXAPLmd (ORCPT ); Tue, 16 Jan 2007 06:42:33 -0500 Received: by wr-out-0506.google.com with SMTP id i20so150011wra for ; Tue, 16 Jan 2007 03:42:33 -0800 (PST) In-Reply-To: <1168947105.14636.11.camel@laundromat.jangutter.com> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Jan Gutter Cc: Matthew Stapleton , linux-ide@vger.kernel.org Jan Gutter wrote: > On Tue, 2007-01-16 at 17:56 +0900, Tejun Heo wrote: >> Matthew Stapleton wrote: >>> Tejun Heo wrote: >>>> Does the problem still persist? > > Sorry for the delayed reply: Holidays attacked before I could apply the > patch. > >> I got confused between your report and Jan's. Yours (not so sure about >> Jan's) seems to be drive firmware bug which hald was successful to >> undiscover. Does hald clear CDO_USE_FFLAGS using ioctl >> CDROM_CLEAR_OPTIONS? The failed commands seem to be from cdrom >> open_for_data(). Considering general quality of ATAPI devices, I >> wouldn't be too surprised if some device fails after hours of repeated >> poll sequence containing READ_TOC and ALLOW_MEDIUM_REMOVAL. > > Hmmm. I think I might have slightly different symptoms: Heavy disk use > definitely causes the bug to appear more frequently. Compiling often > causes it to kick in after about 15 minutes. I also get slightly > different error messages. > > With the latest patch set (2.6.19-gentoo-r4 + cocktail) the resets still > occur, but at least the drive doesn't go down to PIO so soon anymore. > I'll keep watching it for the rest of the day. Attached dmesg of the > first hour or so. (If Evolution borks this mail, I'll resort to telnet and > SMTP ;-) > > Thanks for the feedback, and hope the sucker gets nailed. Can you try 2.6.20-rc5? It has better error reporting and will tell us which SCSI command is timing out. -- tejun