[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

RE: Filesystem type survives formatting in debian installer?

On Wed, February 28, 2007 23:01, peter green said:
>> >this sounds like a bug in mke2fs to me, it is not cleaning up
>> the sectors
>> >before its header and as a result the partition gets misrecognised.
>> I think it's there because some arches and some partitioning schemes
>> actually have real data in the first sector which can't be overwritten
>> or bad things happen.
>> sparc is one such example.
> right so why are lvm and this crypto thing storing stuff there then?

I don't know (and it doesn't really matter, we just have to accept that it
is way things are ATM). Perhaps the authors of lvm and "this crypto thing"
were not aware of the problems a first-block signature might cause. xfs,
cramfs and romfs also have the signature at offset 0.

> maybe the answer is to write some kind of partition cleaner that is run
> before formatting a partition that looks for the signatures of things that
> don't leave a gap (lvm and this crypto thing, possiblly others) and if it
> finds them blow them away before formatting.

Looking for the signatures seem superflous, the partition is about to be
formatted so we could just run a dd if=/dev/zero of=/dev/partition bs=1k
count=65 a the last step before it is formatted.

Feel free to file a wishlist bug report on partman and I'll take a look at
it post-Etch.

(65k above because the signature with the greatest offset that I'm aware
of is some version of reiserfs which places the sig at a 64k offset).

David Härdeman

Reply to: