*BSD News Article 22617


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!agate!poisson.EECS.Berkeley.EDU!binwu
From: binwu@poisson.EECS.Berkeley.EDU (Bin Wu)
Newsgroups: comp.os.386bsd.questions
Subject: Re: Disklabel vs. DOS partition table
Date: 19 Oct 93 18:10:17
Organization: University of California, Berkeley
Lines: 64
Message-ID: <BINWU.93Oct19181017@poisson.EECS.Berkeley.EDU>
References: <1993Oct19.145248.22328@alw.nih.gov>
NNTP-Posting-Host: poisson.eecs.berkeley.edu
In-reply-to: crtb@helix.nih.gov's message of Tue, 19 Oct 1993 14:52:48 GMT

In article <1993Oct19.145248.22328@alw.nih.gov> crtb@helix.nih.gov (Chuck Bacon) writes:

   I've seen numerous posts complaining that the DOS or the BSD
   filesystem got zapped, either by FDISK or by disklabel.
   Having just lost both a DOS and a BSD partition while trying
   to get pcfs to mount, I'd like to get straightened out!
Having hosed and *recovered* a dos disk myself, I can certainly relate
the frustration you run into. I had to buy a book on PC hard disks and
read several chapters before I could understand what I did and how to
recover. :)

   Question 1:
   What is the relation between the DOS partition table's location
   (the very first sector on the disk, I think) and that of the
   disklabel?  I.e., where does the disk label get written?

Dos partition table(or master boot record, mbr) is the first sector on
disk, you need it no matter what operation system(s) you run. The
disklabel on the other hand, appears on the 2nd sector of you 386BSD
partition (the 1st sector is /usr/mdec/wdboot, and the rest is bootwd,
all together totals 16 sectors).

   Question 2:
   What is the minimum coordination between the DOS partition table
   and the disk label which will allow pcfs to mount?
I can sort of guess the sequence of access while you ask the computer
to mount a pcfs is:
	mbr, to find 386BSD partition
	disklabel, to find the dos partition
	mount it, having known the tracks/sectors/etc.
You can't mount a pcfs which resides on a disk w/o a 386 partition.
(well, I guess you can if you do some tricks in-core)

   Question 3:
   Is the c partition determined by wired-in logic, or does it
   depend on pc# and oc# values in disktab?  This appears to be
   important, since all the examples in /etc/disktab have defined
   c partitions, with oc#0.
c is the 386BSD partition, and d is the whole disk.
   Question 4:
   Just exactly what does disklabel do?  With -r, it reads from the
   disk, and without, it uses an in-core copy.  OK.  But what's
   getting edited with -e?  Since -w requires a disk name, apparently
   to look up in /etc/disktab, what's the point of editing?  I've
   apparently edited successfully by editing /etc/disktab, but not
   by use of -e.
-e let you manually edit the disklabel, be it with -r or not.
-w write out disklabel according to disktab and bootfiles you provide
on the command line.
   Question 5:
   I have found that FDISK will place the first DOS partition of the
   first disk exactly one track in from the beginning of the disk,
   but it will locate the first DOS partition of the second disk one
   *cylinder* in.  If I use Norton's diskedit, say, to revise this to
   start the second disk's DOS partition in the second track, will
   DOS work?  And what's the impact on the disk label?
it's just a convention. I think you can start it anywhere.
no impact on disklabel. remember, disklabel is within 386BSD
partition.

Hope this helps.
   --
	   Chuck Bacon - crtb@helix.nih.gov ( alas, not my 3b1 )-:
		   ABHOR SECRECY	-   DEFEND PRIVACY