Willow Schlanger wrote:
>
> http://developer.intel.com/design/intarch/techinfo/pentium/specupdt/doc_ch.h
> appears to be Not Found.
>
> Here's from 24547103.pdf:
>
> Type Name Valid?
> 0 Reserved No
> 1 Available 16-bit TSS Yes
> 2 LDTYes
> 3 Busy 16-bit TSS Yes
> 4 16-bit call gate Yes
> 5 16-bit/32-bit task gate Yes
> 6 16-bit interrupt gate No ---
> 7 16-bit trap gate No ---
> 8 Reserved No
> 9 Available 32-bit TSS Yes <----
> A Reserved No
> B Busy 32-bit TSS Yes
> C 32-bit call gate Yes <----
> D Reserved No ----
> E 32-bit interrupt gate No ----
> F 32-bit trap gate No ----
All of these which you have labeled 'Yes' are in the switch
statement which can result in ZF = 1 if the protection checks
pass.
I created a simple LAR test. Ring3, cycling through system
segment types of 0..15. Tried on 486/P/PII/PIII. All reported
exactly as in the manual, which is exactly as in the plex86
code.
So I still have no idea what you are talking about, when you
say you were 'partially' right. The behaviour is supposed
to be as it is in the stock CVS code.
Is there a flaw in the protection checks? What is it exactly
that you changed that makes Windows work better?
-Kevin
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Kevin Lawton [EMAIL PROTECTED]
MandrakeSoft, Inc. Plex86 developer
http://www.linux-mandrake.com/ http://www.plex86.org/
