[parted-devel] [PATCH] Always allow at least two sectors for extended boot record

Phillip Susi psusi at cfl.rr.com
Mon Mar 29 17:15:21 UTC 2010


On 3/29/2010 12:49 PM, Jim Meyering wrote:
> Thanks!
> This looks like a fine patch.
> However, I would like a minimal test that fails without it,
> yet that succeeds with the patch.
> 
> Can you provide one?

I also ran into this over the weekend.  IIRC you just have to run parted
with -a none and put it in sector unit mode, then when you create a
logical partition with a starting sector +1 from the starting sector of
the extended partition, it fails.

It seems at some point along the line that the kernel was patched to
create a device node for the extended partition ( I swear there was a
time when it didn't ) and always sets the size to 2 sectors so that you
can install LILO to the extended partition.  I've posted a message to
LKML today about this because I think this is an error; the size should
be as much or as little hidden sectors there are before the start of the
logical partition.  In this test case where the logical starts on the
very next sector, then the extended partition should only have a length
of 1.  Since it gets a length of 2 instead, parted fails to add the
logical partition since it overlaps with the extended.



More information about the parted-devel mailing list