Subject: VM trace tool
To: None <tech-kern@NetBSD.ORG>
From: VaX#n8 <>
List: tech-kern
Date: 06/17/1995 02:40:37
I've decided (due mainly to lack of documentation) to not jump directly into
the VM code to get a working CVM, but rather to do some simulations first.
My prof. has a virtual memory tracing tool VMTrace which I'd like to port to
NetBSD.  The rationale behind this is that the SunOS port they have is broken
due to strange interactions with register windows.  My prof said he talked
to Chris Torek (sp?), who said that the VM code that he wrote was similar
to SunOS, but not broken.  How relevant to NetBSD was this comment?  Is
NetBSD's code the same/similar/based upon Torek's?

This tool effectively access-protects almost all the pages of an app's memory
area and logs which are touched.  This should gather pre-stripped, pre-blocked
reference traces for use in some simulation runs.  This is pertinent to a CVM
design, since we can get some numbers on it's effectiveness before full
implementation.  Where would I start to learn about user-land interaction with
VM access protection mechanisms?

Also, I've heard:
> It's mostly the Mach 2.5 VM system, with some changes from Mach 3.0

However, it appears to still use a (pagedaemon) process, which makes me
think it's a little older.  Does it support external pagers?  I was told
this would be a good way to support CVM, and would require less hacking
to the kernel.
VaX#n8 (vak-sa-nate) - n, CS senior++ and Unix junkie -
Deal with evil through strength, yet encourage good through trust.    - PGP me