Subject: Re: hacks/optimizations for netbsd-sun3 and Cnews
To: None <jim@reptiles.org>
From: Gordon W. Ross <gwr@mc.com>
List: port-sun3
Date: 12/06/1995 11:36:31
> From: jim@reptiles.org (Jim Mercer)
> Date: Wed, 6 Dec 1995 11:14:32 -0500 (EST)

> > I'm not a kernel-tuning expert, [ now obvious, no? ]
> > 	gdb -k --write /netbsd
> > 	set bufpages = 2048
> 
> hmmm, not really successful.
> this causes the following fatal error (comes up just after "real mem = ")
>	kmem_suballoc: bad status return of 3

As you can see, I didn't actually try it.  Maybe a smaller value for
bufpages would work.  I just tried bufpages=1024 and it survived.
I booted with the "-d" flag and used ddb to set the variable:

b sd() -d
Boot: sd(0,0,0) -d
>> NetBSD ufsboot [$Revision: 1.1.1.1 $]
631924+29992+95784+[53832+60330]=0xd4dba
Debug mode - enter c to continue
Abort at 0x002404C6.
>c
Starting program at 0x4000
sun3_save_symtab:  found 53832 + 60330
console on zs0 (ttya)
[ preserving 114166 bytes of netbsd symbol table ]
Stopped at      _Debugger+0x6:  unlk    a6
db> x/x bufpages
_bufpages:                 0
db> w bufpages 0x400
_bufpages                      0        =          0x400
db> c
Copyright (c) 1982, 1986, 1989, 1991, 1993
        The Regents of the University of California.  All rights
reserved.

NetBSD 1.1 (GENERIC) #23: Tue Nov 28 16:18:05 EST 1995
    gwr@venus:/home/gwr/src/sys/arch/sun3/compile/GENERIC
Model: Sun 3/260
real mem = 16760832
avail mem = 6987776
using 1024 buffers containing 8388608 bytes of memory

(That's half the memory on this machine!)

> note that i am using a serial console, and when i get the debug prompt,
> the system is non-responsive.  i guess i should check my parity etc settings.
> what settings do you have on your serial consoles?

I have everything set for 9600-N-8-1.

> (also, i have several suns on a serial switch, and if i switch to a different
> machine, the netbsd machine hangs.  this is different from sunos, in that the
> sunos machines would drop to monitor mode when i switched _back_ to them).

Maybe the switch is generating a break when you disconnect?

> this happens when i both do the gdb thing, or when i build a kernel fresh
> with "option BUFPAGES=2048".
> 
> here is the config file i am using:

Looks OK.

Gordon