From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alan Cox Subject: Re: Fun with HPT302N (Rocket Raid 1520 PCI card) Date: Thu, 02 Nov 2006 16:24:15 +0000 Message-ID: <1162484655.11965.212.camel@localhost.localdomain> References: <23724.1161684269@redhat.com> <454869C8.9000103@pobox.com> <1162383443.11965.109.camel@localhost.localdomain> <4548927F.10804@wasp.net.au> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from outpipe-village-512-1.bc.nu ([81.2.110.250]:6623 "EHLO lxorguk.ukuu.org.uk") by vger.kernel.org with ESMTP id S1751963AbWKBQ1s (ORCPT ); Thu, 2 Nov 2006 11:27:48 -0500 In-Reply-To: <4548927F.10804@wasp.net.au> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Brad Campbell Cc: Jeff Garzik , David Howells , linux-ide@vger.kernel.org Ar Mer, 2006-11-01 am 16:26 +0400, ysgrifennodd Brad Campbell: > Here is a complete dmesg from a Rocketraid 1540 (htp37x with bridges) if it helps. > (4 ports, 3 have drives and the 4th exhibits the failure David is seeing) > > http://www.nabble.com/2.6.19-rc3-pata_hpt37x-dying-with-%22irq-16:-nobody-cared%22-t2504272.html > > I still have this in a machine dedicated for testing if you need more info. That one is really very useful. The error handling is leaving some kind of IRQ status no cleared which in turn brings the house down. I think there are two different things in the log here - one is the empty port, the other is some kind of other failure. With a current -mm kernel can you stick ATA_FLAG_DETECT_POLLING on the various .flags = ATA_FLAG_SLAVE_POSS|.... lines for each ata_port_info structure in hpt37x_init_one then see if the identify behaviour is happier. That recipe is probably what we need to do for all the PATA chips that don't have reliable device presence bits. Alan