[bug #27069] Unknown filesystem error with non-first partition of disk

[Daniel Richard G.]

Follow-up Comment #1, bug #27069 (project grub):

I ran into a similar problem recently. I think I know what's going on.

I installed (Ubuntu) Linux onto the end of a 500GB USB external hard drive,
connected to a laptop. When I rebooted after the install, I got the "unknown
filesystem" error. But if I then connected the same hard drive to a desktop
machine I had, the Linux system booted correctly.

The problem, I believe, is just a plain old BIOS barrier limitation---it
can't boot off the drive if the necessary /boot files are farther than a
certain amount (137GB?) from the first sector. BIOS barriers may seem like a
historical concern (137GB was an issue way back in 2001), but this came up on
a fairly recent-model laptop, new enough to have a Core 2 Duo processor. It's
plausible that BIOS code to boot off USB drives sometimes doesn't receive the
same amount of attention as code to boot off standard IDE/SATA drives.

Of course, the reason why I wanted to install Linux at the tail end of the
drive in the first place was so that the first partition would be a ~440GB
FAT32 filesystem. And the reason for _that_ is because, when the USB drive is
plugged into a Windows system, Windows will only take notice of the first
partition---and I wanted the drive to be useful as a data vehicle on non-Linux
systems. (I'm not sure how MacOS X handles things, but I wouldn't be surprised
if it did the same thing.)

The solution I eventually found online pointed out that Windows only cares
about the first partition---i.e. partition number 1 in the drive's partition
table (*)---but that partition does not also have to be the first one in the
drive's actual physical layout. That is to say, I created a partition table
(using fdisk(8)) that has the partitions out of order. The big FAT32 partition
goes right up to the end of the disk, yet the system lists it as /dev/sdx1.
Windows picks it up exactly as desired, and Linux boots happily, BIOS barrier
or no.

(*) Alternately, this could be read as "the lowest-numbered partition in the
partition table," but that is not something I took the time to confirm.

With all that said, I think this can be categorized not as a bug in GRUB2,
but nonetheless as an operating hazard that the program could probably deal
with in a better way---even if it's just printing a warning at an appropriate
point in time.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?27069>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/