From mboxrd@z Thu Jan 1 00:00:00 1970 Subject: Re: ide-scsi oops was: 2.6.0-test4-mm3 From: Alan Cox In-Reply-To: <20030911082057.GP1396@suse.de> References: <20030910114346.025fdb59.akpm@osdl.org> <10720000.1063224243@flay> <20030911082057.GP1396@suse.de> Content-Type: text/plain Content-Transfer-Encoding: 7bit Message-Id: <1063294049.2967.30.camel@dhcp23.swansea.linux.org.uk> Mime-Version: 1.0 Date: Thu, 11 Sep 2003 16:27:29 +0100 Sender: owner-linux-mm@kvack.org Return-Path: To: Jens Axboe Cc: "Martin J. Bligh" , Andrew Morton , Mike Fedyk , Linux Kernel Mailing List , linux-mm@kvack.org List-ID: On Iau, 2003-09-11 at 09:20, Jens Axboe wrote: > > need it. Is it unfixable? or just nobody's done it? > > It's not unfixable, there's just not a lot of motivation to fix it since > it's basically dead. Almost all IDE tape drives require ide-scsi/st modules for one. I'm not sure of the problems in the 2.5 case, in the 2.4 case the big one was that both IDE and SCSI want to control reset/recovery and reissue of commands. That turns into a nasty mess and 2.4 now lets the IDE layer do it, with SCSI just backing off. That may well be the right model for 2.5.x - ie the reset eh handler just waits for the IDE layer to kill the command. The other one was races in the reset code which 2.4 I think now has fixed, which will bite non scsi users but less often -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: aart@kvack.org