[2267] in linux-scsi channel archive
Re: Detecting Buslogic Flashpoint Controller with RedHat 4.2.
daemon@ATHENA.MIT.EDU (Jim Dennis)
Fri Aug 8 04:20:13 1997
To: Fred Harris <fah@uhheph.phys.hawaii.edu>
Cc: "Leonard N. Zubkoff" <lnz@dandelion.com>, linux-scsi@vger.rutgers.edu,
jimd@starshine.org, drew@PoohSticks.org
In-Reply-To: <Pine.OSF.3.91.970807102456.2603B-100000@uhhepk.phys.hawaii.edu>
Message Apparently From Fred Harris <fah@uhheph.phys.hawaii.edu>
Dated Thu, 07 Aug 1997 10:27:41 -1000.
Date: Fri, 08 Aug 1997 00:33:31 -0700
From: Jim Dennis <jimd@starshine.org>
My suggestions and comments are inserted below:
(JimD)
> Hi Leonard,
> I think that I tried that among all the other things that I have
> tried, but I will try it again. Note that the message says that no SCSI
> controllers were found. Should the Redhat Kernal be able to detect the
> SCSI as is? Thanks.
> Regards,
> Fred
You can always build your own kernel (and generally should).
> On Thu, 7 Aug 1997, Leonard N. Zubkoff wrote:
>
>> Date: Thu, 7 Aug 1997 09:21:32 -1000 (HST)
>> From: Fred Harris <fah@uhheph.phys.hawaii.edu>
>>
>> The problem is that I am unable to boot because linux is
>> unable to detect my SCSI controller.
>> I have a Micron Pentium Pro with an IDE hard drive and CDROM.
>> I have a Buslogic Flashpoint 930 SCSI controller and a Nomai 540
>> removable hard drive. I bought RedHat 4.2 linux and completed the
>> installation successfully. RedHat 4.2 has support for the Buslogic
>> SCSI controller. I partitioned the Nomai and installed linux there.
>> At the end of the installation, I chose to put Lilo on a floppy. I
>> don't want to partition my EIDE drive or change the MBR on the
>> EIDE drive.
I'd use LOADLIN.EXE (a DOS program that loads a Linux kernel
which you store in a standard DOS file and supply to LOADLIN
via a command line argument).
I don't know that the Buslogic 930 is supported -- but I
know that the 958 is very popular (David Parsons -- active
on in the newsgroups and a former co-worker -- swears by them).
If you were able to boot from the Red Hat installation floppies
and perform the installation than the card is supported. So
the trick is to use that same boot image (kernel plus, possibly
initrd -- initial ramdisk) as your LOADLIN image (or as the
basis for your boot floppy.
Unfortunately I don't know exactly how to do that, yet.
However you can look at the following two HOWTO's for more
details on creating a boot disk and for details about how
to pass additional parameters to your kernel (and what sorts
of parms to pass to it:
BootPrompt HOWTO
http://sunsite.unc.edu/LDP/HOWTO/BootPrompt-HOWTO.html
Bootdisk HOWTO
http://sunsite.unc.edu/LDP/HOWTO/Bootdisk-HOWTO.html
>> When I tried to boot from the floppy, I got 1's and 0's. I
>> had my doubts about using Lilo anyway with a mixed EIDE and SCSI
>> system because of all the restrictions of Lilo. I sent mail to
>> RedHat support about my problem but realy didn't get much in the
>> way of useful help.
That's characteristic of Red Hat. The simply don't have the
capacity to support their products in any meaningful way.
I'd consider calling Yggdrasil's 900 support line if you're
willing to pay for voice support. Yggdrasil will support any
distribution -- although their expertise would naturally be
better in their own "Plug & Play" product.
Eventually Red Hat may offer support -- probably they'll have
to charge for it in order for it to be financially feasible.
(Don't ask how I manage it).
>> I did boot the installation floppy in "rescue" mode. I was
>> able to mount the Nomai drive and look at the file system, which
>> is fine. So the installation was able to see the Buslogic controller
>> and Nomai harddrive ok. When booting with the installation disk,
>> you are asked if you have SCSI and what type of SCSI, so detection
>> is not automatic.
>> I then tried using LOADLIN. I added a folder on my C:
>> (EIDE) drive
>> with vmlinuz and loadlin. This works. LOADLIN uncompresses my
>> vmlinuz and finds everything on my system but the SCSI. It says that
>> "0 SCSI controllers were found" and then goes into panic mode since
>> it can't access /boot on sda2. (I put root on the 2nd partition).
It should be giving an error like:
'VFS kernel panic, cannot mount root'
... unless the issue has to do with an module that's
inaccessible...
Come to think of it -- that's probably the whole problem.
You probably need to build a kernel with that SCSI driver
built in -- rather than relying on the kerneld (dynamic
module loading feature). Or you need set up an initrd
image like I described earlier.
>> So I have been struggling with this for over 2 weeks.
>> Please help.
>> Thanks in advance.
>> Have you tried setting the "Map Removable As Fixed" option in AutoSCSI?
>> Without that, ths BIOS will not make the removable drive accessible
>> as a BIOS device, which could certainly account for not being able
>> to boot from it.
>> Leonard
The BIOS isn't the problem for LOADLIN (which he's tried).
After LOADLIN reads the kernel image into memory and passes
control to it -- you're in 32-bit mode and relying on Linux
drivers to access all devices (the BIOS is no longer useful
since it's all 16-bit (real mode) code.
The fact that the BIOS can't access the device as a boot
drive does affect LILO -- but not LOADLIN. The fact that
the kernel can't access the module that implements the device's
driver -- BECAUSE THE MODULE IS ON THE DEVICE -- is very likely
the big problem. The easy solution is to build a kernel with
that particular SCSI driver support built in.
The easy way to do that is to boot off of the Red Hat installation
diskettes passing it a root=/dev/sda? parameter. It can find the
module because it can access the floppy and because it probably
uses an initrd.
--
Jim Dennis, The Linux Gazette "Answer Guy"
Linux Gazette is Published under the GPL: http://www.ssc.com/lg/
tag@starshine.org
