Xref: news-dnh.mv.net comp.os.msdos.djgpp:2084
Path: news-dnh.mv.net!mv!news.sprintlink.net!news.bluesky.net!solaris.cc.vt.edu!news.mathworks.com!newsxfer.itd.umich.edu!gumby!yale!oitnews.harvard.edu!newsfeed.rice.edu!rice!news!sandmann
From: Charles Sandmann <sandmann AT clio DOT rice DOT edu>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: locking for HW ints/RMCB
Date: Mon, 18 Sep 1995 21:05:11 CDT
Organization: Rice University, Houston, Texas
Lines: 14
References: <434fn5$bgm AT newsie DOT wis DOT com> <43bf3v$6rm AT mrnews DOT mro DOT dec DOT com> <43b559$e6 AT alijku06 DOT edvz DOT uni-linz DOT ac DOT at> <43i3j7$fnq AT st-james DOT comp DOT vuw DOT ac DOT nz>
Reply-To: sandmann AT clio DOT rice DOT edu
Nntp-Posting-Host: clio.rice.edu
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Dj-Gateway: from newsgroup comp.os.msdos.djgpp

> How can you make sure that the stack in an assembly HW int handler is 
> locked. You can keep the code and data within boundaries, but the stack 
> is accessed before you get to change the stack registers (ie to save 
> CS:EIP), or does this one use of the stack not matter (page almost always 
> will be in memory as the program itself is using it)?

All Hardware interrupts and RMCBs are handled on a locked 4Kb stack provided
by the DPMI provider (so SS:ESP won't be in the DJGPP address space).  The
_go32_dpmi ... wrappers swap onto a local 64K stack in the DJGPP address
space so SS=DS and GCC's code won't get confused (then jump back on exit).

So, if you write in assembler, you know that DS != SS and can handle this,
also knowing you must minimize stack use (with only 4K available for all
nested interrupts/execeptions).