disappointing upgrade to NetBSD 5.0

[Dan Engholm <mustang%engholm.org@localhost>]

To the experts,

I have been running NetBSD in a Xen user domain on top of NetBSD for
several years.  I finally decided to upgrade my user domain virtual machine
from NetBSD 4.0.1 to 5.0.  The installation went without a hitch but
shortly after booting to the upgraded system, the user domain kernel
crashed.  The message in the Xen console said:

panic: Unable to reduce memory reservation

Begin traceback...
rx_mcl(5e3a003,1,0,0,0,0,1,c04a1460,1,0) at netbsd:__kernassert+0xb0a90
End traceback...

dumping to dev 142,16 offset 1582207
dump device bad


rebooting...

I can't be sure what I am doing to get the problem to occur, but it seems
to be network related.  Looking around Google, I found one thread in
port-xen with a similar description but it dead ended with no resolution. 
(see http://mail-index.netbsd.org/port-xen/2009/01/18/msg004686.html)  I
thought that since I was still using Xen 2, I should look at this as an
opportunity to upgrade to Xen 3.  It was a further learning experience but
I did manage to get Domain0 up and running and also got my user domain
created and booting.  But, before it finished its rc script, it died:

Setting sysctl variables:
vfs.generic.usermount: 0 -> 1
kern.no_sa_support: 0 -> 1
ddb.onpanic: 1 -> 0
Starting network.
Hostname: xxxxx.xxxxxxx.org
IPv6 mode: host
Configuring network interfaces: xennet0.
Adding interface aliases:.
add net default: gateway yyyyyy
Building databases:panic: buf mem pool index 7
Begin traceback...
uvm_fault(0xc9d58b98, 0, 1) -> 0xe
fatal page fault in supervisor mode
trap type 6 code 0 eip c0396531 cs 9 eflags 10246 cr2 0 ilevel 0
panic: trap
Faulted in mid-traceback; aborting...
dumping to dev 142,16 offset 1582223
dump device bad


rebooting...

What it should have said after "Building databases:" is:

Building databases: dev, utmp, utmpx done
Starting syslogd.
Setting date via ntp.
Checking for core dump...

I'm stuck now and don't know where to go from here.  I can't move forward
and I can't stay where I am.  It would be very painful, but I could
downgrade back to 4.0.1.  I'm not doing anything strange here.  All of the
NetBSD kernels I use are generic, as in not built by me.  I would prefer to
move ahead to Xen 3 so I will say that the Xen kernel is from pkgsrc-2009Q1
xenkernel3 as are xentools3.  The NetBSD kernels were copied from the
NetBSD 5.0 i386 CD ISO image.  Any hints for getting this to work at all
under Xen 3 or even working around the reliability of Xen 2 would be
greatly appreciated.

Since the upgrade, my user domain no longer boots without intervention.  It
does not figure out on its own where the root partition is so I have to
manually enter "xbd0a" from the Xen console.  The configuration file has
'root = "/dev/xbd0a"' in it as it has since I started using Xen.  This is
true regardless of which version of Xen I use.

Also, I got burned by Xen telling me that the "nics" configuration option
is obsolete.  The man page says that it's still valid and that its default
value is 1.  If I don't have a "nics = 1" line in my configuration file, no
NICs are created and the creation fails because it can't configure the
network.  What gives?

--Dan