Short description of the situation:
- embedded linux device running 2.4.19 montavista (Archos PMA430)
- redesigning the firmware to as open source as possible
- moved system partition from an loop mounted ext3 image on a fat32
partition to a real partition.
- device supports usb-mass-storage mode (complete hdd gets handed over
to a usb-bridge)
Now we're trying to avoid unmounting that ext3 partition because we use
it with unionfs to have a writeable root filesystem.
The problem is that when the hdd is accessed via usb the operating
system of that desktop PC might mount that ext3 partition and change it.
What would be a clean way to hide the partition from systems accessing
it? What would be a clean way of making it ReadOnly to those systems?
We had some ideas like using a different magic for the filesystem or a
high and unique revision number we could patch our kernel for, but I'm
not sure if there would be bad side effects to that. Marking the
partition as e.g. hidden FAT16 (0x16) won't keep a linux system from
recognizing and automounting it.
Any ideas or pointers?
TIA & Cheers
Thomas B. R?cker
http://www.openpma.org
On Fri, 2006-11-10 at 21:31 +0100, "Thomas B. R?cker" wrote:
> Short description of the situation:
> - embedded linux device running 2.4.19 montavista (Archos PMA430)
> - redesigning the firmware to as open source as possible
> - moved system partition from an loop mounted ext3 image on a fat32
> partition to a real partition.
> - device supports usb-mass-storage mode (complete hdd gets handed over
> to a usb-bridge)
>
> Now we're trying to avoid unmounting that ext3 partition because we use
> it with unionfs to have a writeable root filesystem.
> The problem is that when the hdd is accessed via usb the operating
> system of that desktop PC might mount that ext3 partition and change it.
> What would be a clean way to hide the partition from systems accessing
> it? What would be a clean way of making it ReadOnly to those systems?
>
> We had some ideas like using a different magic for the filesystem or a
> high and unique revision number we could patch our kernel for, but I'm
> not sure if there would be bad side effects to that. Marking the
> partition as e.g. hidden FAT16 (0x16) won't keep a linux system from
> recognizing and automounting it.
>
> Any ideas or pointers?
Using a unique superblock magic sounds simple and is probably as
foolproof as anything.
Shaggy
--
David Kleikamp
IBM Linux Technology Center