*BSD News Article 17894


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!spool.mu.edu!howland.reston.ans.net!xlink.net!math.fu-berlin.de!informatik.tu-muenchen.de!lrz-muenchen.de!regent!angle!rse
From: rse@angle.cyvaned.com (Ralf S. Engelschall)
Newsgroups: comp.os.386bsd.questions
Subject: Still some SCSI/kernel problems
Message-ID: <Bv9at*YI6@angle.cyvaned.com>
Date: Fri, 02 Jul 1993 23:46:30 GMT
Reply-To: rse@angle.guug.de
X-Newsreader: Arn V1.02
Lines: 96


Hello,

  a few days ago I posted a request for help to install 386BSD on a 1GB
SCSI-drive with an Adaptec AHA-1742 attached to a i486/50 EISA board
which 32MB RAM.  Thanks for the replies - I removed 16MB and now I was
able to use NetBSD 0.8 kc-aha.fs + install1 to solve the bootstrap
problem of installing 386BSD on this machine (this was the only way, all
other booddisks failed!).

But there are problems and questions left:

                                   1.

    I have setup the Adaptec to "Standard Mode" and read in Julians
    bootblock code, that the kernel will switch to "Extended Mode"
    automatically after startup. Now I am confused: As I understand
    the hardware (but I have actually not very much knowledge of it),
    I had to remove 16MB because in "Standard Mode" the host adapter
    can access 16MB only. When the kernel create its DMA buffers
    outside this range -> PANIC! But if the kernel _automatically_
    switches to "Extended Mode", this couldn't be the actual problem.
??  Or is this "automatic switch" only included in 386BSD + pk0.2.x
??  and _not_ in the NetBSD 0.8 kc-aha.fs I had problems with?

??  Please, can anyone report me the _complete_ setup used to install
??  under AHA-1742 plus 32MB RAM on a EISA box? Am interested in
??     - the AHA-1742 EISA config options (mode, irq, etc)
??     - the CMOS BIOS settings (cache, wait-states, etc.)
??     - the kernal config and system used to install
??       (NetBSD, 386BSD+pk, special, etc.)

                                   2.

   The only way I am successful in installing is:
    - Boot from NetBSD 0.8 kc-aha.fs and install1.fs
    - disklabel, newfs from these disks
    - Install NetBSD 0.8 startup files to harddisk
    - Boot from harddisk and remove all NetBSD stuff
    - Read 386BSD distribution files via ethernet to harddisk
    - Extract 386BSD

    Now I wanted to replace the NetBSD kernel with a kernel I compiled
    on another 386BSD0.1+pk0.2.4 machine. This worked fine. The harddisk
    bootes up and the system runs. Replacing the bootblocks with
    new ones (also from the mentioned 386BSD+pk0.2.4 machine) also worked
    fine.

    But then I observed, that running under the NetBSD kernel,
    alle diskdrive access works fine, but under the new 386BSD0.1.2.4
    kernel the diskdrive is dead. Mounting or any other access to
    the diskdrives results in a system-dead. I think that someone
    posted a patch for these "floppy-dead-sitiuation" some months ago,
    but I thought this bugfix will already be included in the new patchkit.
    But it is not!
??  What must I do to get the floppies work under the 386BSD0.1.2.4 kernel?

    Another problem is, that using 386BSD0.1.2.4 kernels to setup
    a bootdisk with

       disklabel -r -w /dev/rfd0a floppy /usr/mdec/fdboot /usr/mdec/bootfd
       newfs /dev/rfd0a floppy
       mount /dev/fd0a /mnt
       cp /386bsd /mnt
       ...

    or changing a existing bootdisk by replacing the kernel always
    results in a correct startup (bootblock code and kernel start fine).
    But after the root filesytem change the system hangs also!

    The pronlem has to be the SCSI drivers in pk0.2.4, because the
    BSD0.1.2.4 machine runs fine with its kernel compiled under pk0.2.4.
    The only differnce is, thsi machine has a IDE drive, not a SCSI,
    so the problem has to be in the SCSI floopy support, I think.

??  What is the correct way of getting a 386BSD0.1.2.4 kernel run
??  with a 386BSD fixit.fs disk or something by replacing the existing
??  kernel with the new one?

                                   3.

    disklabel-ing the harddisk with _translated_ partition parameters
    results in no difference as disklabeling the physical parameters.
    Because I use the entire harddisk, I think it is OK to use the
    physical parameters. Only when installing with other coexisting
??  systems (like DOS) the translated parameters are needed? Is this
??  correct? Or is there a problem with my setup (i used physical params)?


Yes, I know... a lot of question marks, but what should I do ;_)
Please give me some advice! Thanks!

                                       Ralf S. Engelschall
                                       rse@cyvaned.com
                                       rse@guug.de
                                       rse@sdm.de