Here's the "fsck_ffs" picture:

# find / -name fsck_ffs -ls
 204397      2 drwxr-xr-x    4 root     wheel          512 Dec 27 04:23 /usr/src/sbin/fsck_ffs
6173090   4768 -r-xr-xr-x  129 root     wheel      2429184 Dec 27 08:35 /usr/src/destdir.i386/rescue/fsck_ffs
6173044    164 -r-xr-xr-x    1 root     wheel        83144 Dec 27 08:20 /usr/src/destdir.i386/sbin/fsck_ffs
5741595      2 drwxr-xr-x    2 root     wheel         1024 Dec 27 07:14 /usr/obj/sbin/fsck_ffs
5741994    180 -rwxr-xr-x    1 root     wheel        92077 Dec 27 07:14 /usr/obj/sbin/fsck_ffs/fsck_ffs
5721434      2 drwxr-xr-x    2 root     wheel         1024 Dec 27 07:56 /usr/obj/rescue/fsck_ffs
  40494    656 -r-xr-xr-x    1 root     wheel       322988 Sep  8  2002 /sbin/fsck_ffs

I presume "/sbin/fsck_ffs" is the problem and this implies that:

   ./build.sh install=/

didn't do what I expected per:

   http://www.netbsd.org/Documentation/current/#updating

This is my first attempt to track -current and the above looks
fool proof but it seems I may still be foolish enough.

I'm now concerned the userland install failed even though I took
the ending message as a success.

May safely assume "/sbin/fsck_ffs" should be replaced with the
"/usr/src/destdir.i386/sbin/fsck_ffs" (83144) even though it's
smaller than "/sbin/fsck_ffs" (322988)?

I'm concerned that if "/sbin/fsck_ffs" wasn't updated then what
else could have been missed.  Any suggestions on how to assure
the entire userland was updated?

Conrad

-----Original Message-----
From: Martin Husemann [mailto:martin@duskware.de]
Sent: Saturday, December 27, 2003 15:06
To: Conrad T. Pino
Cc: NetBSD Current
Subject: Re: FFS File System Corruption - 1.6ZG GENERIC.MP - Intel
N440BX Dual CPU


On Sat, Dec 27, 2003 at 02:52:17PM -0800, Conrad T. Pino wrote:
> No problem with 1.6.1 GENERIC kernel.
> 
> Repeats every time with 1.6ZG GENERIC.MP kernel:
> 
> # shutdown -r now
> 
> On reboot:
> 
> UNEXPECTED INCONSISTENCY; RUN fsck_ffs MANUALLY.

What vintage is your fsk_ffs binary?

If it's 1.6 you should get the latest 1.6.2_RC3 one and replace it, see
the entry in src/UPDATING copied below.

Martin

20030402:
        The superblock layout for FFS was changed.  If you have 1.6
        fsck binaries, they will signal a fatal superblock mismatch
        with the first alternate, because they compare too many
        fields (even ones that aren't useful).  If possible, upgrade
        your fsck_ffs binary before using a new kernel.
        None of this signals actual filesystem damage.