Four Disk Geometries and a Funeral

Four Disk Geometries and a Funeral

Post by johan antwe » Mon, 23 Jun 2003 13:18:52



Ultimate question: geometry B works but, is it correct?

I'm trying to install FreeBSD 4.8 on a Seagate ST380021A 80GB drive on
a SY-TISU mobo with latest bios ( Soyo on their web site claims this
mobo supports FreeBSD ).

When starting the partition phase of the installation, the installer
immediately reports that the geometry ( geomerty A - see list below )
for ad0 is incorrect and is using a more likely geometry ( geometry B
) however, I noticed this was not the geometry printed on the
sticker-label on the outside of the drive ( geometry C ) so I chose to
set the geometry myself. This is when the installer warned me that I
should  the geometry that is reported by the bios and nothing else. So
I aborted the installation, rebooted and checked the bios which
reported yet another disk geometry ( geometry D ).

B works but is it correct? A is not permitted by the installer. I
tried C and it won't boot. D is not permitted.

Disk Geometry List
   Cyl    / Head/ Sectors
A) 155061 / 16  / 63
B) 9729   / 255 / 63
C) 16383  / 16  / 63
D) 38309  / 16  / 255

The product pdf manual is here:
http://www.seagate.com/support/disc/manuals/ata/100129212b.pdf
This document claims that geometry C is the "default logical
geometry". I am confused. Any suggestions?

There is no funeral, it's just a subject line.

 
 
 

1. SCSI disk drive geometry limits hard disk choices?

In my attempts to get a HP C2247-300 1 Gb hard disk formatted
by Linux, I encountered a significant Linux limitation which
purchasers of hard disks should be made aware of (e.g. in the
FAQ).

In my setup fdisk wouldn't read the SCSI drive's geometry
parameters for reasons I'm still trying to understand; it may
be that the drive's SCSI ID number (7) conflicts with the
Future Domain (1680) controller's assignment for that number
which is usually reserved for SCSI tapes.  (It reported 0
heads, 0 cylinders, etc.)   Anyway, when I entered the
correct parameters into fdisk (13 heads, 1984 cyl, 56-96
sectors), I encountered a significant limitation: about 80
sectors are needed to achieve the disk's 1.05 Gb capacity,
yet Linux only supports <64 sectors per track).  At 63
sectors the disk formatted, but it gave me 75% of the blocks
that I anticipated.  This means that my 1.0 Gb disk is a 750
Mb disk under Linux!

Is my experience just an artifact of fdisk's inability to
correctly read the drive geometry as returned by the
controller's firmware?  Or, should users be told (in the FAQ)
that they need to carefully examine a drive's geometry to
determine if the 63 sector limit will limit their drive's
capacity?

In my case, for 1Gb, I will exchange the disk for one whose
cylinders * heads > 31,002.

Thanks for listening.


--

   candidate^H^H^H^H^H^H^H^H^H, (got it!) Environmental
   Planning, UC Berkeley.  Phone: 510-848-1241

2. New Users and the Walnut Creek CD Set

3. Disk Geometry on Second IDE Disk

4. OMT/UML OO modeler for free - here's one.

5. Getting disk geometry from SCSI disk

6. 3dfx for linuxppc

7. how to determine disk geometry for SCSI disk?

8. 'core' files?

9. Disk Geometry on Second IDE Disk

10. Matrox Mystique ands X.

11. Fourth hard disk

12. Help. More than four IDE disks.

13. Four E-IDE hard disks under Linux