From owner-freebsd-questions Sat Oct 3 12:03:01 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id MAA13437 for freebsd-questions-outgoing; Sat, 3 Oct 1998 12:03:01 -0700 (PDT) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from localhost.localdomain (ppp-112-61.villette.club-internet.fr [194.158.112.61]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id MAA13352; Sat, 3 Oct 1998 12:02:40 -0700 (PDT) (envelope-from groudier@club-internet.fr) Received: from localhost (groudier@localhost) by localhost.localdomain (8.8.4/8.8.4) with SMTP id UAA01820; Sat, 3 Oct 1998 20:09:31 +0200 X-Authentication-Warning: localhost.localdomain: groudier owned process doing -bs Date: Sat, 3 Oct 1998 20:09:26 +0200 (MET DST) From: Gerard Roudier X-Sender: groudier@localhost To: Howard Lew cc: Dan Busarow , freebsd-questions@FreeBSD.ORG, freebsd-scsi@FreeBSD.ORG Subject: Re: NCR 53c875 SCSI Problems In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Wed, 30 Sep 1998, Howard Lew wrote: > On Sun, 27 Sep 1998, Gerard Roudier wrote: > > > The FreeBSD ncr driver does not read the user-setup from NVRAM and it is > > not possible to tell it about the BUS width at boot time. So, the WIDE > > negotiation has every chance to occur with great success and, as a result, > > will break any further data transfer between the controller and the > > device. > > > > Yes. This appears to be the problem. > > Because the driver doesn't seem to be able to correctly identify the bus > width, I think it would be a better idea to read the NVRAM values if they Wide and Narrow devices can share the same SCSI bus. This has been the result of T10 standards being careful about compatibility issues. But from device (a controller is a device) point of view, a Wide device is always connected to a Wide bus and a Narrow device is always connected to a Narrow bus. One of the tricks is that a Wide device can be driven using only the Narrow subset of signals of a real (or imaginary) Wide bus. > are available. Otherwise, those of us (anyone really) using a SCSI hard > drive that supports a WIDE bus but is running it on a NARROW bus connector My opinion is that connecting a Wide controller to a Wide device using a Narrow SCSI connector is beyond SCSI specifications. The NVRAM when present and read by the driver may help accomodate this configuration, but with controllers that haven't NVRAM, we are stuck when it is not possible to tell the driver about. > on a NCR/Symbios 53c875 card may run into this same problem. > > Unfortunately, it is not a simple problem to fix unless you already have a > FreeBSD system that is bootable off of another drive. If this is the > case, the fix is trivial by modifying ncr.c Using a working system to apply a work-around in order to be able to boot a broken system is quite usual. ;-) Modifying ncr.c, as I suggested, is just a hack for personnal convenience that has the effect to definitely narrow all 53C8XX controllers at driver level. It is not an acceptable fix at all. > But imagine a user trying to install FreeBSD and boot from boot.flp and > failing each time when the controller isn't able to talk to the hard drive > after switching to the WIDE bus. I think this is a serious enough problem > to require a fix... If you want to be nice with such users, then you must use a default driver setup that would neither negotiate Wide, nor accept Wide negotiation until some user intervention. I am not sure such a proposal will be accepted. A boot command option that would tell the driver that it must only drive devices (or part of devices) in Narrow mode (until it is told different) is probably the only acceptable solution. I donnot know if this is possible with FreeBSD. Regards, Gerard. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message