From mboxrd@z Thu Jan 1 00:00:00 1970 From: Aldo Gavioli Subject: Re: sd driver Date: Tue, 24 May 2005 13:58:13 +0200 Message-ID: <429316D5.6050000@netstorage.it> References: <42919ABC.9030508@netstorage.it> <4292C7C4.2030204@torque.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from smtpout08.attiva.biz ([85.37.16.6]:42003 "EHLO smtpout08.attiva.biz") by vger.kernel.org with ESMTP id S262003AbVEXL6R (ORCPT ); Tue, 24 May 2005 07:58:17 -0400 In-Reply-To: <4292C7C4.2030204@torque.net> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: dougg@torque.net Cc: linux-scsi@vger.kernel.org Gilbert, thanks for your answer. I'm thinking regarding your ideas which could be the best. In any case I think that this behaviour is a malfunction of kernel 2.6. We cannot denied to load different drivers than sd for type MOD. This behaviour will generate other troubles in developper community. Aldo > Aldo Gavioli wrote: > >> Hello All, >> >> Hello All, >> I'm trying to attach one proprietary SCSI driver for optical disk on >> kernel 2.6. >> I noted that if sd attach the same device before, my driver will >> never be called in the probe function. How can I by-pass this >> situation ? >> If anyone can help me I appreciate very much. > > > Aldo, > Upper Level Drivers (ULDs of which sd is one) discriminate > on the basis of the peripheral device type returned by an > INQUIRY. Currently sd_probe() takes control of TYPE_DISK > (pdt=0x0) and TYPE_MOD (pdt=0x7 "Optical memory device") > and will soon take control of TYPE_RBC. > > If the class of devices you are interested in is TYPE_MOD > then (at least for testing) just comment out that bit of > code in sd_probe(). > > The st and osst (tape ULDs) co-operate to do a finer > grain division of labour (see their respective *_probe() > functions). These ULDs show you one way of solving the > problem. Another way would be a kernel/module option > to sd (and possibly other ULDs) telling it to leave > device (or better some world wide unique > identifier) alone. > > > As for Arjan's comment about extending sd to cope with > your class of devices, recent experience with RBC devices > indicates that such a move may not add to the coherency > of the sd driver. > > Doug Gilbert > > > > > > -- Aldo Gavioli Product Manager NetStorage Software s.r.l Viale Europa,74 - 20090 Cusago (Milan) Italy - Web: www.netstorage.it