RE: Kernel compiling problems

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

John Klos <john%klos.com@localhost> writs: 
> To that end, the (significant) changes I made to the kernel config file are:
> maxusers 256 (from 8) 
> options NMBCLUSTERS=4096 
> options BUFCACHEPERCENT=30
> 
> With maxusers set to 256, I get the following error:
> pmap_init:  0: 082b6000 - 0fffe000 ( 131366912)
> Maxproc 4116, mem_size 128 MB: allocating 37 KPT pages
> NetBSD 1.4.1 (reva_jk) #1: Tue Nov 30 05:21:58 EST 1999
>     root@reva:/usr/src/sys/arch/amiga/compile/reva_jk
> Amiga 4000 (68060 rev.1 CPU/MMU/FPU)
> real  mem = 134217728 (16384 pages)
> unexpected trap format 4 (vector offset 8) from 126b36
> panic: straytrap
> Stopped in  at  0xf34ee:        unlk    a6
> 
> However, if I reduce the MAXUSERS to 64, it compiles and boots.
> 
> I am currently running 22 virtual domains on Apache on AmigaOS, but will be
> colocating at the beginning of next week and cannot change anything
> thereafter. So does anyone know what else needs to be changed to allow a
> kernel with 256 MAXUSERS? Do I need to increase NKMEMCLUSTERS (kernel
> malloc space)? What exactly are NKPTADD and NKPTADDSHIFT for? The docs on
> those are somewhat vague.

  I suspect that if you really want to have MAXUSERS set to 256, you may have
to completely rewrite sys/arch/amiga/amiga/pmap.c.

  I think the pmap has to allocate kernel memory for the page tables for each
process.  With 256 MAXUSERS, that works out to be something like 4116 processes.
I don't recall how much memory is allocated for each page table, but I suspect
with 4116 processes, that's quite a bit of memory.

  [The "straytrap" message probably due to some trap occurring before the
trap vectors get adjusted for the 68060 traps, so the trap is not being 
correctly
recognized.]

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