Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 25 Apr 1995 17:54:01 +1000
From:      Bruce Evans <bde@zeta.org.au>
To:        julian@ref.tfs.com, phk@ref.tfs.com
Cc:        dgaudet@cs.ubc.ca, hackers@FreeBSD.org
Subject:   Re: [to Install people] Re: comments on an attempted install
Message-ID:  <199504250754.RAA21639@godzilla.zeta.org.au>

next in thread | raw e-mail | index | archive | help
>> 3/	we need to be able to create other partition types.
>no.  It may be handy to be able to, but we don't >need< it.

We can easily create partitions and set the partition type entry for
them, but we can't handle internal details.

DOS apparently removes stale tables when partitions are deleted, if
at all.  You can get into trouble deleting DOS partitions using non-
DOS fdisks because of this.  E.g.: start with a DOS partition covering
the whole disk.  Delete it and create a FreeBSD partition at the end
of the disk and a smaller DOS partition at the start of the disk.  Put
good stuff on the FreeBSD partition.  Format the DOS partition.  In
some cases the FreeBSD stuff will be destroyed because DOS uses its
old BPB at the front of its partition.

>> 5/	disklabel had old info from before teh new fdisk
>> 	(I understand how this happens, but we should look at how to
>> 	fix that when we re-write the MBD...(invalidate the label?))
>hopefully fixed for 2.1.

FreeBSD doesn't create its own partitions properly :(.  Stale disk
labels that appear when a new FreeBSD partition is created should be
rejected, e.g., by zeroing block 2 of a newly created FreeBSD partition.
Of course, old labels should not be clobbered without asking - fdisk
should normally not write anywhere except the partition table.

>> 7/	In the initial install boot, the -c option should be COMPULSORY
>> 	or at least better publisised (It is, but not well)
>Hmmm maybe.

All devices except the console and floppy should be initially off :-).

>> 13/	Slice code is pooly documented .. docs still refer to disklabel for 
>> 	DOS partitions.
>In the works these days.

I thought the drivers in the SNAPs didn't use new slice code.

>> > in a disklabel.  This seems like a really arbitrary restriction to me.
>> > Isn't /dev/wd0s2 the second slice on disk 0? 
>> No, wd0s1 is..... 0,1,2,3 right?

>No wd0s[1-4], wd0s0 is "compatibility slice" ie first type 0xa5 found,
>slice 5... are dos extended slices.

There is actually no slice s0.  Drivers go to some trouble to not print
a slice number (sN) for the compatbility slice.  wd0s1 actually has slice
number 2.  Slice number 0 is the compatibility slice and slice number 1
is the whole disk.  To avoid confusion you should rarely think of slice
numbers.  The numbers are also confusing in ls output:

brw-r-----  1 root    operator     0, 131074  Nov 10 19:04 /dev/wd0s1

Here 1301074 is 0x200002 where the first `2' is the slice number and
the second 2 is the partition number.

Bruce



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504250754.RAA21639>