Re: extended partitions
Wed, 4 Oct 1995 10:49:06 +0100

: >It is this line in genhd.c:
: >
: > /* prevent someone doing mkfs or mkswap on
: > an extended partition */
: > hd->part[minor].nr_sects = 0;
: >
: >Since nothing else depends on it, you can take it out.
: >But I think the advantages outweigh the disadvantages.

: Bad idea, IMO. On systems with DOS occupying the entire first drive and
: Linux installed on the 2nd drive, installing Lilo on the extended
: partition is the only alternative to installing it in the MBR, which is
: also a frequent cause of problems for newbies.


: I also dislike the idea of making the Lilo installer access /dev/hda and
: interpret the partition tables in order to find the extended partition.
: Interpreting the contents of a partition table is foreign to its
: function.

: Wouldn't it be more reasonable to make the partition type available to
: 'mkfs', 'mkswap', and the like, ...

I think I disagree on both counts. I had some correspondence with Ted
about e2fsck and he didnt like the idea that e2fsck should go outside
its partition and study partition tables, and I agree with him.

But LILO on the other hand already does study the partition tables -
LILO is very much aware of these details - it has to convert
"block 0 of /dev/hda4" into LBA and C/H/S addresses anyway.
It would be a rather minor change.

Anyway - there are many ways to solve this "problem" - but perhaps
there is no problem.
Reiterating: in my view it is very bad that writing to /dev/hda4
can destroy /dev/hda5, and it is only slightly inconvenient that
presently the number of ways to boot Linux has diminished a little.
(There is the MBR, there is LOADLIN, there is fips to make a small
Linux partition, etc. etc. And a small change in LILO will restore
the original freedom.)