Path: utzoo!mnetor!uunet!husc6!think!ames!ptsfa!ihnp4!occrsh!uokmax!rmtodd
From: rmtodd@uokmax.UUCP (Richard Michael Todd)
Newsgroups: comp.os.minix
Subject: Re: Hard Disk Problems
Message-ID: <1055@uokmax.UUCP>
Date: 20 Feb 88 06:36:52 GMT
References: <998@bcd-dyn.UUCP>
Reply-To: rmtodd@uokmax.UUCP ()
Organization: University of Oklahoma, Norman, OK
Lines: 50
Keywords: Minix WD Hard Disk Problems
Summary: There is a bug in the autoconfig code.

In article <998@bcd-dyn.UUCP> sas@bcd-dyn.UUCP (432 2370.00) writes:
>I'm still having problems with my western digital wd1002s-wx2 hard disk
>and tandon tm262 combination.  Is anyone running MINIX on hard disk #2?
Yes and no.  I've got the second hard disk partitioned into two partitions,
18.5M for DOS and 1.5M for MINIX.  Currently I'm not doing anything much
with the MINIX partition, but I did do a mkfs on /dev/hd7 and it worked
fine.  I should note that there is a bug in the wd_wini.c driver code, as
I mention below.  I'm using a WD controller with two Seagate ST225 drives.

>Using drivers /dev/hd5-9?  Using standard 1.2 Minix, I can only get the
>os to recognize the drive as /dev/hd5, it doesn't even access the disk 
>when I try to use partitions 6-9.  Using hd5, I can only mkfs about 1/2
>of the 20 meg disk.  I fixed the AUTO_CONFIG definition, but my controller
>isn't an autoconfig version, so that doesn't help, although that version

There is a bug in the code to handle AUTO_CONFIG in a two-drive system.
The code in wd_wini.c gets the info on the configuration of drive C from
the first sector on the drive (the info is in a 16 byte table).  One would
expect that when the BIOS variable for "# of hard drives" was two, it would
proceed to read the first sector of drive D to get its config information.
It doesn't. It instead copies the next 16 bytes of the drive C config sector
(the ones after the table for drive C) into the structure for drive D.  It
thus uses garbage values for the drive D parameters and the attempt to
read drive D's partition table using these bogus parameters causes a
kernel panic.  I suspect that the intent of the code's author was to
handle WD's "logical drive" scheme, where a single physical drive is
partitioned into two logical drives, and apparently the config. info on
both logical drives is stored in the first sector of the physical drive.
But when you have two physical hard disk drives, the code bombs.  I took the
easy way out and patched wd_wini.c so that the parameters used for drive
D were simply copied from those for C.  I can get away with that since
my two drives are both ST225s and thus have the same parameters.
  Obviously this isn't quite the problem you're experiencing, since you
don't have an autoconfig controller and your system will at least boot,
even though it doesn't correctly access drive D.  However it shouldn't be
two surprising that if they got the auto_config code wrong, the code
for the non-auto-config case might be broken as well.  
  One thing you might do is install debugging code in wd_wini.c to see
just exactly what the driver thinks the drive parameters and partition
tables are for both drives.  There's an array of structures indexed
by hard drive number for drive parameters and an array of structures indexed
by partition number for the partition info.  Sorry I don't remember what
the names are, but they're fairly easy to find.  
  As I said, I haven't used drive D extensively under MINIX, but I did 
check to see that "od -c /dev/hd5", "od -c /dev/hd6", etc., did read
the right portions of the hard disk, and I did successfully mkfs on /dev/hd7.
--------------------------------------------------------------------------
Richard Todd
USSnail:820 Annie Court,Norman OK 73069
UUCP: {allegra!cbosgd|ihnp4}!occrsh!uokmax!rmtodd