Subject: Re: NKPDE changes
To: Frank van der Linden <>
From: Jonathan Stone <jonathan@DSG.Stanford.EDU>
List: port-i386
Date: 11/19/1996 18:47:18
[[What is the  calculatation of  nkpde ignoring that it should
  take into account?]
Quoting frank van der Linden:

>..or if you define a large value for NBUFPAGES, or..

Or if you use lots of cluster mbufs, or 4k clusters.
Or if you  turn on kernel profiling.

For now I'e halved the resolution of the pc-samples; this
needs a change to MI code to allow overriding the relevant

>As Chris said, the main win is that you don't get punished with
>a panic just because you happen to have a lot of memory in your machine..
>And it's fixed in a reasonable way. It's not the Ultimate Solution,
>but it's most certainly better than the previous situation, which
>had been present for too long.

Doesn't FreeBSD dynamically add kpde's  as necessary -- going back
and adding the relevant entries to existing processes, if I recall
John Dyson's message correctly?  Could/should  we  do that, or size
nkpde at boot time?

> If it's any consolation, if all of my changes were applied:

>They were.

How are users supposed to find out what nkpde actually gets set to
(e.g., for users who alter parameters that need more or fewer

I can't find anything (e.g., in /usr/sbin/sysctl) that shows what the
calculated value of nkpde is.  Can we get add it to /usr/sbin/sysctl,
or  to kernfs, or something?

Also, is there a simple way to compute if nkpde and kernel_map are