when formatting the MMC card, why use expert mode to change geometry?

perhaps a dumb question but, when following the instructions to
format a dual-partition card here:

  http://code.google.com/p/beagleboard/wiki/LinuxBootDiskFormat

is it strictly necessary to go into fdisk's "expert" mode to change
the number of apparent number of heads and cylinders of the card?

  i've formatted a number of cards for various reasons (with both FAT
and ext2 filesystems), and in all those other situations, i've never
needed to change the geometry that way. what exactly is that
accomplishing? thanks.

rday

never mind, i think i know the answer to my own question. changing
the geometry is necessary only if the card is going to be used as a
boot device, yes?

rday

The internal bootrom of the omap3 needs a specific C/H/S layout to load MLO from the fat partition. If you only want to use the MLO and u-boot that are flashed into NAND then C/H/S doesn't matter.

regards,

Koen

ok, that makes sense, but i just hit another snag. i followed the
instructions at the link above, did all the partitioning, then ended
with "w" to finally modify the partition table and got:

  Command (m for help): w
  The partition table has been altered!

  Calling ioctl() to re-read partition table.

and it's hung. this is on a fedora 9 system, and a "ps -ef" command
shows that the command "fdisk /dev/mmcblk1" is still running. how
curious. and no amount of ^C or ^\ can break me out of it.

  i'm guessing i can just yank the card out as the above suggests that
the partitioning worked, but i'm puzzled as to why the hanging. has
anyone else noticed this?

rday

p.s. if i open another xterm and "fdisk -l", the partitioning looks
fine.

  Calling ioctl() to re-read partition table.

and it's hung. this is on a fedora 9 system, and a "ps -ef" command
shows that the command "fdisk /dev/mmcblk1" is still running. how
curious. and no amount of ^C or ^\ can break me out of it.

  i'm guessing i can just yank the card out as the above suggests that
the partitioning worked, but i'm puzzled as to why the hanging. has
anyone else noticed this?

rday

I had a similar behavior some months ago - Using Ubuntu in an coLinux
environment - I however never managed to figure out why, but I can confirm
to have seen the problem - I used a 2GB Kingston Card in case this is
relevant in any way...

Best regards
  Søren

PS: AFAIR the ROM code shouldn't infer any requirements on C/H/S - Only
requirement is that the first partition needs to be FAT12/16/32-formatted
and that MLO needs to be the first entry in the FAT table... That being said
I haven't tested many different MMC cards, but I think one should give it a
try "not going into the advanced settings" to see if it in general works
without modifying these parameters - I suppose it will...