Subject: Perplexing halt/reboot core dumps in Pre-1.5 VAX ports?
To: None <>
From: NetBSD Bob <>
List: port-vax
Date: 01/17/2001 13:23:48
Since I have been the monkey's uncle lately, (or dunce, fool, [choose one])
and running up every port of NetBSD and previous BSD's that I could lay
my hands on, on the old MVII critter, I have noticed a particular habit
in NetBSD Pre-1.5 shutdowns.  They never shutdown gracefully.  This
includes 1.4.3.  The only early version that does shutdown gracefully
(after the tight loop lockup method from 4.3BSD through NetBSD-1.2)
is 1.3.  1.3.2, 1.4, 1.4.1, 1.4.2, 1.4.3 all go through the motions of
shutdown, but just pass through the hooks, then go off to never-never-land.
They then require a physical CTRL-C abort and manual sync;sync;halt.
At that time a halt.core or reboot.core is left behind at the current
tree level.  Interestingly, 1.5 and 1.3 both shutdown gracefully.
Can someone explain to me what is causing this?  If it is common knowledge,
the dunce, here has missed it.  If it is not common knowlege, a fix is
needed for 1.4.3 and later 1.4.x bits, so that a proper shutdown is done.
Since the MVII probably won't be able to keep up with the herd on 1.5,
I am realistically stuck with 1.4.x as the end of the line for the old
crate to still be a usable machine.

One other small nit to pick at.... the 1.4.3 bootfs does not allow
installation to occur by the install script.  It runs out of fs space
in the miniroot, complaining with a fs full error.  My workaround is
to use the 1.4.1 boot/install since that works fine.  Is there a proper
fix for this?  If the 1.4.x is the end of the usable line for the MVII
critters, it would be nice to get a good working tape install system
finalized and fixed for the last 1.4.x whatever.  For that matter,
it needs to be done for 1.5.x, and current, too.

There I go thinking out loud again....(:+}}...



Interesting tidbits I find on stripped MVII kernels and FTP timings:

FEATURE                          OS           TIME    FOOTPRINT
MVII stripped kernel compile     4.3BSD       ~1h     ~270k
                                 NBSD-1.4.3   ~9h     ~470k
                                 NBSD-1.5     ~24h    ~570k
FTP timing on 16mb file          4.3BSD       125k/s
                                 ULTRIX-4.2   110k/s
                                 NBSD-1.0A    150k/s
                                 NBSD-1.1A    180k/s
                                 NBSD-1.2     210k/s
                                 NBSD-1.3     210k/s
                                 NBSD-1.3.2   215k/s
                                 NBSD-1.4.1   215k/s
                                 NBSD-1.4.3   225k/s
                                 NBSD-1.5     155k/s

1.4.3 is the best on system throughput on the MVII, although
anything from 1.2 through 1.4.3 is fine.

What has happened to kernel sizes and compile times for identical
stripped kernel configs?  YIKES!