From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tejun Heo Subject: Re: libata/SATA noprobe Date: Thu, 19 Apr 2007 01:28:57 +0900 Message-ID: <46264749.7080206@gmail.com> References: <6.1.1.1.2.20070418164107.02c87a30@192.168.6.12> <462645F6.50702@rtr.ca> 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]:50503 "EHLO wr-out-0506.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754123AbXDRQ3H (ORCPT ); Wed, 18 Apr 2007 12:29:07 -0400 Received: by wr-out-0506.google.com with SMTP id 76so205727wra for ; Wed, 18 Apr 2007 09:29:06 -0700 (PDT) In-Reply-To: <462645F6.50702@rtr.ca> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Mark Lord Cc: Roger While , linux-ide@vger.kernel.org, Alan Cox Mark Lord wrote: > Roger While wrote: >> Is there any knob/option to prevent libata >> probing non-existent channels ? >> Specifically how can I stop the kernel probing >> the second SATA? - > .. >> <6>ata1.00: ATA-7, max UDMA/133, 490234752 sectors: LBA48 NCQ (depth 0/1) >> <6>ata1.00: ata1: dev 0 multi count 8 >> <6>ata1.00: configured for UDMA/133 >> <6>scsi1 : ata_piix >> <4>ata2: port is slow to respond, please be patient (Status 0xff) >> <3>ata2: port failed to respond (30 secs, Status 0xff) >> <3>ata2: SRST failed (status 0xFF) >> <3>ata2: SRST failed (err_mask=0x100) >> <4>ata2: softreset failed, retrying in 5 secs >> <3>ata2: SRST failed (status 0xFF) >> <3>ata2: SRST failed (err_mask=0x100) >> <4>ata2: softreset failed, retrying in 5 secs >> <3>ata2: SRST failed (status 0xFF) >> <3>ata2: SRST failed (err_mask=0x100) >> <3>ata2: reset failed, giving up > .. > > Ugh. That could really slow down system startup. > > There is no parameter to avoid it, > just one to reduce the delay while it probes. Not ideal. > > But it really could be more clever here, and notice the 0xff patterns, > and have an early exit if there's obviously nothing attached. > > Or perhaps there's some register it could read to see if the > port was disabled in the BIOS (I'm betting it is still "enabled", > but it could be good to check if we don't already). > > Maybe just do it in the ata_piix subdriver. Tejun, Alan? We do consider 0xff as port empty these days. Roger, which kernel version are you using? Can you give a shot at 2.6.20.7? -- tejun