how to configure 18GB drive?

how to configure 18GB drive?

Post by Ed Koshimot » Sat, 09 Jan 1999 04:00:00



I have an 18GB ST118273W drive I'd like to use on
Linux.  Right now fdisk sees 17366MB out of a possible
18.21GB.  I'd like to change the cylinder/head/sector
parameters if possible since right now fdisk defaults
to 32 sectors.

Any pointers on how to do this?  I know linux only
supports 63 sectors in a partition table.  I'd like
to get more info on tradeoffs on using various
cylinder/head/sector configurations.

In particular I'd like to know the formula that
specifies the relation between a hard drive's
capacity and its cylinders/heads/sectors.

Thanks,

ed

 
 
 

how to configure 18GB drive?

Post by Dan Nguye » Sat, 09 Jan 1999 04:00:00


: I have an 18GB ST118273W drive I'd like to use on
: Linux.  Right now fdisk sees 17366MB out of a possible
: 18.21GB.  I'd like to change the cylinder/head/sector
: parameters if possible since right now fdisk defaults
: to 32 sectors.

Remember that 18.21 is the commercial size of the harddrive not the
actual.  18210MB does not actually equal 18.21GB.  Because 1024MB
equalls one 1GB.  And 1000000Bytes do not equal 1GB, 1048576Bytes do.
Try doing this 18.21*10^9 (18.21G) divided by 1048576 and you get
17366.  The correct size of your harddrive.

--
           Dan Nguyen            | There is only one happiness in

http://www.cse.msu.edu/~nguyend7 |                   -George Sand

 
 
 

how to configure 18GB drive?

Post by Ben Slom » Sun, 31 Jan 1999 04:00:00


I've just added an 14.4GB IBM *IDE* drive to my Linux system (running kernel
2.0.32 on a new PC i.e. a PC with a new BIOS). I wanted to partition the drive
into two partitions: one a Windows FAT32 partition, I'm afraid, and one for Linux.
After many experiments, I hit upon the following method.

I'm no expert so use this advice at your own risk: caveat installor!

    1) I let the BIOS autodetect the drive
        In my case it reports
        28229040 sectors and 14453MB in LBA mode
        16383 cyl, 16 hd, 63 sec, 8455M in CHS mode

    2) Create the Windows partition using windows FDISK. Reboot into Linux.

    3) Edit /etc/lilo.conf to pass the `linear' option to the kernel

    4) Edit the Linux kernel file /usr/src/linux/drivers/block/ide.c  
       and rebuild the kernel as suggested in the LargeDisk mini HOWTO  
       http://www.veryComputer.com/
       Reboot.

    5) Use Linux fdisk to add a second partition

The documentation with the disk tells me the geometry is 16 heads, 63 sectors
per track, and 16383 cylinders, and *28229040* sectors in total.
BUT, this doesn't seem to add up:

    16 * 63 * 16383 = 16514064 != 28229040

i.e. if I use the documented numbers I get to see only ~8GB out of the ~14GB disk.

Furthermore, Linux reports the disk geometry CHS as
    1024 cylinders, 255 heads, and 63 sectors
where
    1024 * 255 * 63 = 16450560  

The SCSI section (!) of the LargeDisk HOWTO suggests that 1023/255/63 might be
used to signal a drive accessed with LBA with a capacity of at least 1023*255*63
sectors (the actual capacity being found by other means). It also mentions that
the translation from CHS coordinates to a LBA number is independent
of the number of cylinders on the disk. So, assuming I've got the heads and
sectors set correctly, I can simply set the number of cylinders to give the required
overall capacity.

All that remained was to find out whether to use the documented head and sector
geometry, 16/63 or the geometry reported by Linux 255/63 (or something else).

I fiddled around a bit, and ended up plumping for 255/63 because it seems to work.
Then I chose C = floor(28229040 / (255 * 63)) = 1757

So I ran fdisk and used expert mode to set the number of cylinders to 1757, and
created a nice large Linux partition. Here is the partition table I ended up with:

        Disk /dev/hdb: 255 heads, 63 sectors, 1757 cylinders
        Units = cylinders of 16065 * 512 bytes

           Device Boot    Start      End   Blocks   Id  System
        /dev/hdb1             1      383  3076416    b  Win95 FAT32
        /dev/hdb2           384     1757 11036655   83  Linux native

I should probably edit /etc/lilo.conf to pass hdb=1757,255,63 to the kernel
at boot time, but I haven't done this yet (i.e. I don't pass any geometry
information to the kernel) so when I run fdisk without setting the cylinder
count explicitly I get:

        Disk /dev/hdb: 255 heads, 63 sectors, 1024 cylinders
        Units = cylinders of 16065 * 512 bytes

           Device Boot    Start      End   Blocks   Id  System
        /dev/hdb1             1      383  3076416    b  Win95 FAT32
        /dev/hdb2           384     1757 11036655   83  Linux native
        Partition 2 has different physical/logical endings:
             phys=(1023, 254, 63) logical=(1756, 254, 63)  

Apart from this I don't get any error messages and I'm able to read and write
the Windows partition from Linux.

I've installed LILO in the MBR of my original disk (/dev/hda) and told it
to boot the kernel found in /boot on /dev/hdb2. This works today, though I'm
a little nervous that I may run foul of the 1024 cylinder issue in future.

So, fingers crossed, I'm hoping everything is OK for the moment.
I've been running for nearly a week without problems, but both of my partitions
are less than half full so I can't be sure I'm safe ...

If anyone has a more definitive understanding of how to install the new large IDE disks
I'd be glad to hear their suggestions -- I've looked in the LargeDisk mini HOWTO
and it helps somewhat but not enough to really inspire confidence. And I note that
other people are also wrestling with these newer large disks.

Hope this helps.

Good Luck!

Ben

PS. As part of this process I also installed
    - a patch to add FAT32 support (see http://www.veryComputer.com/)
    - and an upgraded version of hdparm from ftp://sunsite.unc.edu/pub/Linux/system/hardware/
    and used it to enabled UDMA.

: Hi!

: This is interesting!

: Currently I am struggling to set up a 10.1GB disk. FDISK correctly
: recognizes the size when the disk is completely empty and no prtitions
: are configured. As soon as I add a partition, fdisk sets CHS =
: 1229/255/63 to CHS = 1024/255/63. The kernel does the same.
: I tried to configure the BIOS of my Asus P5A by giving the number of
: cylindes as given on the disk label, but that resultet in the BIOS
: incorrectly setting the size to approx. 8GB.

: This problem does not appear with Win98-FDISK / Win98.

: Any ideas on how to correctly set up linux so that it does not*up
: my disk?

: Thanks for the help.


: > : I have an 18GB ST118273W drive I'd like to use on
: > : Linux.  Right now fdisk sees 17366MB out of a possible
: > : 18.21GB.  I'd like to change the cylinder/head/sector
: > : parameters if possible since right now fdisk defaults
: > : to 32 sectors.
: >
: > Remember that 18.21 is the commercial size of the harddrive not the
: > actual.  18210MB does not actually equal 18.21GB.  Because 1024MB
: > equalls one 1GB.  And 1000000Bytes do not equal 1GB, 1048576Bytes do.
: > Try doing this 18.21*10^9 (18.21G) divided by 1048576 and you get
: > 17366.  The correct size of your harddrive.

--

phone: +44(0)117 9228946                         Hewlett Packard Laboratories
fax: +44(0)117 9228925                     Filton Road, Bristol, BS12 6QZ, UK
                Not speaking for Hewlett Packard

 
 
 

how to configure 18GB drive?

Post by Bradley M Kerya » Sun, 31 Jan 1999 04:00:00


One way is to use the expert menu in fdisk to change the number of
cylinders. Then repartition and the kernel will use whatever the partition
table says is in use. You don't have to pass "linear" to lilo or modify
your kernel.

Example: I have a Seagate ST39140A 9GB drive, which is slightly above the
8.4GB barrier. The BIOS says 1027 cylinders, but I used the expert menu in
fdisk to change the cylinder count to 1108 (which, incidentally, is what
the 2.2.x kernels report, although I did this under 2.1.129). Then I
partitioned, and got my extra 500MB or so. Just make sure not to use the
wrong number of cylinders, and to do a "badblocks -w" before formatting
your partitions to make sure you've set them up correctly (which,
incidentally, will destroy any data on those partitions).


> I've just added an 14.4GB IBM *IDE* drive to my Linux system (running kernel
> 2.0.32 on a new PC i.e. a PC with a new BIOS). I wanted to partition the drive
> into two partitions: one a Windows FAT32 partition, I'm afraid, and one for Linux.
> After many experiments, I hit upon the following method.

> I'm no expert so use this advice at your own risk: caveat installor!

>     1) I let the BIOS autodetect the drive
>         In my case it reports
>         28229040 sectors and 14453MB in LBA mode
>         16383 cyl, 16 hd, 63 sec, 8455M in CHS mode

>     2) Create the Windows partition using windows FDISK. Reboot into Linux.

>     3) Edit /etc/lilo.conf to pass the `linear' option to the kernel

>     4) Edit the Linux kernel file /usr/src/linux/drivers/block/ide.c  
>        and rebuild the kernel as suggested in the LargeDisk mini HOWTO  
>        http://www.veryComputer.com/
>        Reboot.

>     5) Use Linux fdisk to add a second partition

> The documentation with the disk tells me the geometry is 16 heads, 63 sectors
> per track, and 16383 cylinders, and *28229040* sectors in total.
> BUT, this doesn't seem to add up:

>     16 * 63 * 16383 = 16514064 != 28229040

> i.e. if I use the documented numbers I get to see only ~8GB out of the ~14GB disk.

> Furthermore, Linux reports the disk geometry CHS as
>     1024 cylinders, 255 heads, and 63 sectors
> where
>     1024 * 255 * 63 = 16450560  

> The SCSI section (!) of the LargeDisk HOWTO suggests that 1023/255/63 might be
> used to signal a drive accessed with LBA with a capacity of at least 1023*255*63
> sectors (the actual capacity being found by other means). It also mentions that
> the translation from CHS coordinates to a LBA number is independent
> of the number of cylinders on the disk. So, assuming I've got the heads and
> sectors set correctly, I can simply set the number of cylinders to give the required
> overall capacity.

> All that remained was to find out whether to use the documented head and sector
> geometry, 16/63 or the geometry reported by Linux 255/63 (or something else).

> I fiddled around a bit, and ended up plumping for 255/63 because it seems to work.
> Then I chose C = floor(28229040 / (255 * 63)) = 1757

> So I ran fdisk and used expert mode to set the number of cylinders to 1757, and
> created a nice large Linux partition. Here is the partition table I ended up with:

>         Disk /dev/hdb: 255 heads, 63 sectors, 1757 cylinders
>         Units = cylinders of 16065 * 512 bytes

>            Device Boot    Start      End   Blocks   Id  System
>         /dev/hdb1             1      383  3076416    b  Win95 FAT32
>         /dev/hdb2           384     1757 11036655   83  Linux native

> I should probably edit /etc/lilo.conf to pass hdb=1757,255,63 to the kernel
> at boot time, but I haven't done this yet (i.e. I don't pass any geometry
> information to the kernel) so when I run fdisk without setting the cylinder
> count explicitly I get:

>         Disk /dev/hdb: 255 heads, 63 sectors, 1024 cylinders
>         Units = cylinders of 16065 * 512 bytes

>            Device Boot    Start      End   Blocks   Id  System
>         /dev/hdb1             1      383  3076416    b  Win95 FAT32
>         /dev/hdb2           384     1757 11036655   83  Linux native
>         Partition 2 has different physical/logical endings:
>              phys=(1023, 254, 63) logical=(1756, 254, 63)  

> Apart from this I don't get any error messages and I'm able to read and write
> the Windows partition from Linux.

> I've installed LILO in the MBR of my original disk (/dev/hda) and told it
> to boot the kernel found in /boot on /dev/hdb2. This works today, though I'm
> a little nervous that I may run foul of the 1024 cylinder issue in future.

> So, fingers crossed, I'm hoping everything is OK for the moment.
> I've been running for nearly a week without problems, but both of my partitions
> are less than half full so I can't be sure I'm safe ...

> If anyone has a more definitive understanding of how to install the new large IDE disks
> I'd be glad to hear their suggestions -- I've looked in the LargeDisk mini HOWTO
> and it helps somewhat but not enough to really inspire confidence. And I note that
> other people are also wrestling with these newer large disks.

> Hope this helps.

> Good Luck!

> Ben

> PS. As part of this process I also installed
>     - a patch to add FAT32 support (see http://www.veryComputer.com/)
>     - and an upgraded version of hdparm from ftp://sunsite.unc.edu/pub/Linux/system/hardware/
>     and used it to enabled UDMA.


> : Hi!

> : This is interesting!

> : Currently I am struggling to set up a 10.1GB disk. FDISK correctly
> : recognizes the size when the disk is completely empty and no prtitions
> : are configured. As soon as I add a partition, fdisk sets CHS =
> : 1229/255/63 to CHS = 1024/255/63. The kernel does the same.
> : I tried to configure the BIOS of my Asus P5A by giving the number of
> : cylindes as given on the disk label, but that resultet in the BIOS
> : incorrectly setting the size to approx. 8GB.

> : This problem does not appear with Win98-FDISK / Win98.

> : Any ideas on how to correctly set up linux so that it does not*up
> : my disk?

> : Thanks for the help.



> : > : I have an 18GB ST118273W drive I'd like to use on
> : > : Linux.  Right now fdisk sees 17366MB out of a possible
> : > : 18.21GB.  I'd like to change the cylinder/head/sector
> : > : parameters if possible since right now fdisk defaults
> : > : to 32 sectors.
> : >
> : > Remember that 18.21 is the commercial size of the harddrive not the
> : > actual.  18210MB does not actually equal 18.21GB.  Because 1024MB
> : > equalls one 1GB.  And 1000000Bytes do not equal 1GB, 1048576Bytes do.
> : > Try doing this 18.21*10^9 (18.21G) divided by 1048576 and you get
> : > 17366.  The correct size of your harddrive.

> --

> phone: +44(0)117 9228946                         Hewlett Packard Laboratories
> fax: +44(0)117 9228925                     Filton Road, Bristol, BS12 6QZ, UK
>                 Not speaking for Hewlett Packard

        Brad
 
 
 

1. Trade: 8218 18GB SSA drives for 8209 9GB SSA drives

Has your company considered moving from 8209 9GB SSA drives to 8218
18GB SSA drives.  I would like to speak with someone who is considering
this move and would like to save money in the process. We have 8218 SSA
drives that are used for 7133 D40 disk subsystems.  We would like to
offer you credit on your 8209's in exchange for 8218 drives.  All of
our drives are 100% IBM, guaranteed IBM maintenance eligible and
contain a 30-day warranty.  Please contact me directly at:


Thanks in advance,

REID

--
Reid SV

Sent via Deja.com http://www.deja.com/
Before you buy.

2. Apache 2.0 on Caldera Openlinux 3.1.1

3. Ultra 450 & 18gb drives

4. can't open /dev/nst0

5. 2x VA Linux 18GB 10k rpm drives

6. [RESENT] fix bugs in HID driver

7. Severe h/w reliability problem on pSeries 640, 18Gb disks etc. ?

8. Kernel panic problem on Ultra1

9. Looking for format.dat info for 18gb Fujitsu

10. 18GB too much for 43p-132?

11. Partitioning of 18gb hd for Debian Linux/FreeBSD dual boot

12. Partitioning strategies on a 18GB fileserver

13. AH2940, Seagate 18Gb SCSI, Mandrake 7.1