Subject: install/28841: sysinst code dumps on invalid extended partition (mbr)
To: None <,,>
From: None <kre@munnari.OZ.AU>
List: netbsd-bugs
Date: 01/02/2005 18:57:00
>Number:         28841
>Category:       install
>Synopsis:       sysinst code dumps on invalid extended partition (mbr)
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    install-manager
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sun Jan 02 18:57:00 +0000 2005
>Originator:     Robert Elz
>Release:        NetBSD 1.6U (irrelevant send-pr system...)
	Prince of Songkla University
System: NetBSD 1.6U NetBSD 1.6U (FUCHSIA) #51: Wed Jul 23 16:06:08 ICT 2003 i386
Architecture: i386
	All totally irrelevant.   It is an i386 running 2.99.11 (Dec 22).
Machine: i386
	sysinst (i386) core dumps when it references what seems like it
	should be an extended partition, but isn't.
	Create an invalid "extended" partition.   The way I almost certainly
	managed that was using sysinst from a 1.5 or very early 1.6 vintage
	system (may have even been 1.4) - what I really did was reinstall
	an old drive I'd had sitting on a shelf for a while, and attempt to
	use that to do a 2.99.11 install.

	The drive had an MBR with 3 partitions.   The first was a typical
	wintrash (fat32) type.   The second was "type 5".  The third was
	an old NetBSD partition.    The "type 5" was once a swap space
	area (for NetBSD - the NetBSD 'b' label used much of the space
	inside it - and any other systems that wanted it).   It got to be
	type "5" because sysinst insisted I give it some value (non-zero if
	I recall correctly)and I have absolutely no idea what the numbers
	actually mean (I know what the number scheme means, just not which
	values mean what) - so I just picked something I'd seen before.
	It worked for years & years...  (in fact, I thnk I still have a
	few drives installed that have layouts just like this).

	Use a -current (as of a week or so ago anyway) sysinst, attempt to
	delete the old partitions that are no longer wanted - watch sysinst
	dump core when it cannot find the internal label in the extended
	partition that's supposed to say what is really there.
	No idea, haven't looked ... simple workaround is to use fdisk to
	delete the partition, aside from issuing a warning about the
	invalid extended partition, it handled it with no problems.

 	This was from a 2.99.11 (-current) of December 22 (that is,
 	just before anoncvs took its holiday...)   I haven't seen any
 	sysinst commits during the intervening period though.