🚀 go-pugleaf

"Stephen Schwetz" <darteng@aljan.com.au> wrote:

>Low level formats are safe I have done them recently on about 7 drives with
>no problems

Low Level formats on modern drives is a misnomer. Instead, the
*manufacturer* *specific* routines, some confusingly labelled as Low Level
Formatters, actually use a combination of zeroing and/or sector
reallocation to clear/find/hide media defects. A low-level format should
actually re-write all of a track, including the headers, id's, sync fields,
data areas, etc.

Quote from the Fujitsu Low-Level Formatter (argh!) manual:

  HOW THE PROGRAM WORKS

  When the program file is executed, it first tries to identify all the IDE
  drives connected in the system. All the necessary parameters and drive 
  information are then automatically obtained by the program for its own
  use. There are two major functions which this program implements:

    a) Defect Scan; and
    b) Low Level Format

  The Defect Scan only verifies the drive for errors using the READ VERIFY
  command. This command is issued starting from the minimum CHS/LBA ,
  depending on the current addressing mode (LBA being the default), to the
  maximum CHS/LBA address. When an uncorrectable (UNC) or Address Mark Not
  Found (AMNF) error occurs ( this error also means a Sync Byte Not Found
  error), the program treats the sector where the error occurred as a bad
  sector. After the verification process is completed, the number of bad
  sectors found is then displayed.

  In the Low Level Format function, the program first performs the same
  function with the Defect Scan. However, all the bad sectors found during
  the verification process are then converted to their equivalent Physical
  Address using the ADDRESS TRANSLATION factory command, and then added to
  the Primary Defect List Table in the SA Area. This new PLIST information
  will then be used during the Low Level Format routine, which executes
  another factory command, the FORMAT UNIT command. The FORMAT UNIT command
  will effectively remap all these bad sectors and also reinitialize the
  drive to its factory format with all the user data being 0.


I can't remember exactly where I got I, but this bookmark seems promising:
http://www.fujitsu-europe.com/file_lib.htm

If not, try www.fujitsu.co.jp.

Anyway, back to the problem (which was pruned off, earlier).

>Drive is a Fujitsu 3.2 Gb Ultra DMA type and runs in mode 4 on
>the BIOS which is configured as LBA. (I use a disk cage and swap
>drives as to which O/S I want to use).
>
>Drive is configured as 1200Mb DOS partition drive C: and the D:
>as 2000Mb.
>C: is primary and D: is extended DOS and logical drive.
>
>I used disk druid from the CD to delete the D: and reconfigure as
>Linux /root of 1800Mb and swap of 200 Mb (approx). The disk druid
>then asks to format and block check drives as I would expect, and
>then goes on as per the handbook to load software with the
>options etc.
>
>Normally this works OK and all software loads and system goes to
>boot but comes up with file errors during the fsck.ext2 checks,
>and tels me I have illegal blocks within the INODE (whatever that
>is) with many instances of this, and then drops to the bash.
>
>OR
>    Software will not load correctly and gets about a third of
>the way through and 'runs out of space' and wont decompress
>files. (loads to /temp/???).
>

First try reducing the speed of the drive (eg, mode 3, mode 2, etc), and/or
(preferably) separate your hdd and cdrom so they are on separate IDE
channels.

If that still doesn't work, then it might be that the end of the root
(boot) partition is beyond the 1023 cylinder BIOS startup limitation. In
other words, If the start of the Linux root (boot) partition is below 1023
cylinders, but the end is beyond 1023 cylinders, then you have a
possibility of parts of the kernel/modules will be unreachable during the
boot phase.

Solution ?

Reduce the size of the root (boot) partition so that it is wholly within
the 1023 cylinder boundary. ie make separate partitions for /usr, etc,
allowing root fs to be smaller. Once Linux (or rather the disk drivers) are
running, the 1023 cylinder boundary no longer applies.

You could also try putting the kernel (boot files) on a floppy (if they
fit) and boot off that.
  __   __   _______________________________
 //)) //)) | Richard RUDEK. MicroDek.      | Hey, it's Friday night...
//\\ //\\  | Chatswood, Sydney. Australia. | C:\WORK> CD \  
           `-------------------------------'