Re: Please test a GENERIC.MP kernel for me

["Michael L. Hitch" <mhitch%lightning.msu.montana.edu@localhost>]

On Fri, 23 Oct 2009, Jason Thorpe wrote:

> While working on some changes to subr_pool.c recently, I discovered a bug in 
> the L1 page table allocation code in the Alpha pmap.  However, since I don't 
> have any Alpha hardware anymore, I can't test the fix myself.  I would greatly 
> appreciate it if someone could test out a GENERIC.MP kernel for me (built from 
> -current sources a couple of days old):
>
>    ftp://ftp.netbsd.org/pub/NetBSD/misc/thorpej/netbsd-alpha-mp.gz

  This kernel fails to boot on my ES40, presumably because since I've 
added drives to the iop(4) raid adapter, it seems the kernel needs to be 
loaded at a higher memory address.  I was also unable to boot a kernel 
from my -current tree - it kept crashing after calling the 
iop_config_interrupts() routine.  I then applied Jason's diffs to my 
netbsd-5 tree and that kernel booted and ran no worse that before.  There 
were some pool cache invalidation changes Jason made that weren't in the 
netbsd-5 branch, so I tried to add those diffs and got a kernel that 
didn't do so well.  I then added the other pool cache invalidation changes 
and got a kernel that seemed to work no worse than before.

  I was also able to finally get a -current kernel booting and running (no 
idea why I was having the problems in iop_config_interrupts() before). 
That also seems to run no worse than before.

  Both my netbsd-5 and -current kernels still run into problems once in a 
while.  I get segment violations in a couple of the build tools (install, 
and I think grotty), and have had some kernel memory trap panics.

--
Michael L. Hitch                        mhitch%montana.edu@localhost
Computer Consultant
Information Technology Center
Montana State University        Bozeman, MT     USA