Todd Vierling wrote:
> 
> On Mon, 14 Sep 1998, Guenther Grau wrote:
> 
> : > : If we could count on the kernel having symbols in-core at load time,
> : > : any of these paradigms (/kern/kernel, /kern/kernelsymbols, etc)
> : > : becomes a SMOP.
> 
> : > Some machines (particularly lower-end mips, m68k, vax, even a real I386...)
> : > need all the memory they can get, and symbols take up a lot.
> :
> : Is it not possible to page the symbols out to swap space, once the
> : kernel is loaded?
> 
> No.  Swap is no longer configured at kernel boot time; it is configured as
> part of the bootup /etc/rc.  And that initial boot Must work in order to get
> the swap configured.

How about a sysctl that can be issued to page the symbols to 
the swapspace, as soon as it's available? Well, this sounds 
ugly :-), but then, I have only thought about it for a few
of seconds :-) Maybe someone else got a better idea? 

> : There is one thing I'd like to throw in here:
> : I really appreciate that people always try to think of every
> : (corner) case, when talking about an implementation.
> 
> Well, this is more than a corner case.  I have three machines that fit the
> lower-end definition above.  (And a lot of the "corner cases" have
> real-life backing cases--we're a diverse bunch.  ;)

I know, but how many of these machines don't have /netbsd (or the kernel
they were booted from) not available on a file system that can
be accessed by the kernel?

  Guenther