Very Computer

After trying a number of configurations, I decided to install the
drive (a Western Digital AC2540, 540 Mb) on a second IDE card, so that
I could rule out the possibility of wierd interactions with AT BIOS.  The
drive geometry is hard coded into the kernel (1.1.52 w/ atdisk2-1.1.47+).

After partitioning the drive, fdisk 1.5 reports-

Command (m for help): p
Disk /dev/hd1a: 16 heads, 63 sectors, 1048 cylinders
Units = cylinders of 1008 * 512 bytes

    Device Boot  Begin   Start     End  Blocks   Id  System
/dev/hd1a1           1       1     246  123983+  83  Linux native
Partition 1 does not start on cylinder boundary:
     phys=(0, 0, 2) should be (0, 1, 1)
/dev/hd1a2         247     247     285   19656   83  Linux native
/dev/hd1a3         286     286     895  307440   83  Linux native
/dev/hd1a4         896     896    1048   77112   83  Linux native
Partition 4 has different physical/logical endings:
     phys=(1023, 15, 63) logical=(1047, 15, 63)

Is the message about different physical/logical endings significant?  The
logical information is correct.  I don't know where the physical information
is coming from.

In the words of an infamous operating system- abort, retry, or ignore?

   ] After trying a number of configurations, I decided to install the
   ] drive (a Western Digital AC2540, 540 Mb) on a second IDE card, so
   ] that I could rule out the possibility of wierd interactions with
   ] AT BIOS.  The drive geometry is hard coded into the kernel
   ] (1.1.52 w/ atdisk2-1.1.47+).

Oh atdisk2 patch against 1.1.x kernels can be bound in
mipgsun.mipg.upenn.edu:pub/delman

   ] Disk /dev/hd1a: 16 heads, 63 sectors, 1048 cylinders Units

I have only recently got a drive with more than 1024 cylinders. I
think the temporary fix is to go into expert mode in fdisk and set the
heads to 32, and cylinders to 524.

Will try and fix it in the atdisk2 code....

Delman.

--
______________________________________________________________________

  Delman Lee                                 Tel.: +1-215-662-6780
  Medical Image Processing Group,            Fax.: +1-215-898-9145
  University of Pennsylvania,
  4/F Blockley Hall, 423 Guardian Drive,                        
  Philadelphia, PA 19104-6021,

______________________________________________________________________

   ] Will try and fix it in the atdisk2 code....

The 1.1.52-diffs in mipgsun.mipg.upenn.edu:pub/delman should fix the
problem. It fakes the bios geometry from the real physical geometry by
halving the cylinders (and doubling the heads) until cyl<=1024.

Please give it a try.

Delman.
--
______________________________________________________________________

  Delman Lee                                 Tel.: +1-215-662-6780
  Medical Image Processing Group,            Fax.: +1-215-898-9145
  University of Pennsylvania,
  4/F Blockley Hall, 423 Guardian Drive,                        
  Philadelphia, PA 19104-6021,

______________________________________________________________________

Ugh!  This will cause trouble.

The correct solution for linux 1.1.x is to simply define the drive
as having 32 (or 64) heads in the BIOS CMOS setup.  Then DOS, linux,
and most other stuff will handle it just fine.
--

   ] <The 1.1.52-diffs in mipgsun.mipg.upenn.edu:pub/delman should fix the
   ] <problem. It fakes the bios geometry from the real physical geometry by
   ] <halving the cylinders (and doubling the heads) until cyl<=1024.
   ]
   ] Ugh!  This will cause trouble.
   ]
   ] The correct solution for linux 1.1.x is to simply define the
   ] drive as having 32 (or 64) heads in the BIOS CMOS setup.  Then
   ] DOS, linux, and most other stuff will handle it just fine.

The problem is that (at least for people with old BIOSes) there is no
CMOS setup for 3rd and 4th drives on the 2nd controller. That's why I
fake the logical geometry (bios_info[]) from the real physical
geometry (hd_info[]) in hd.c.

For people with the newer BIOS (EIDE), where it's possible to enter
geometries for 3rd and 4th drives on the 2nd controller, the correct
solution would be to enter the fake/logical geometry, EXCEPT Linux
doesn't yet read the EIDE BIOS entries. I was looking into it, I think
it means adding some assembly in setup.S. And assembly is not my
strong suit, so that's where I stopped.... Anybody?

Thanks, Delman.
--
______________________________________________________________________

  Delman Lee                                 Tel.: +1-215-662-6780
  Medical Image Processing Group,            Fax.: +1-215-898-9145
  University of Pennsylvania,
  4/F Blockley Hall, 423 Guardian Drive,                        
  Philadelphia, PA 19104-6021,

______________________________________________________________________

Allan Wind            Phone: (+45) 40 27 14 94

8230 Aabyhoej         www: http://www.daimi.aau.dk/~u930528/index.html
Denmark               ------------------------------------------------
--

Allan Wind            Phone: (+45) 40 27 14 94

8230 Aabyhoej         www: http://www.daimi.aau.dk/~u930528/index.html
Denmark               ------------------------------------------------