From: Robin Grindley (grindley@helios.physics.utoronto.ca)
Date: 01/25/93 

From: grindley@helios.physics.utoronto.ca (Robin Grindley)
Subject: Re: ?? Why no out/in in assembler ??
Date: 25 Jan 1993 23:24:03 GMT

In article <C1E98M.4JE@ichaos.nullnet.fi> jlaiho@ichaos.nullnet.fi (Juha Laiho) writes:

>Here we go again. Linux is designed to be multi-user multi-processing
>environment. In such environment there is no reason to let anyone do direct
>port-I/O. Just think. There's unixes with shell access, and a C Compiler

Sorry, i should have been a little clearer. I realize that user tasks will
run at different privileges and IOPL's. Any normal code should not be allowed
to do what i want. But if one wants to debug a system one has to have access
to tools that can bypass normal security. My question is, IF i do setup
a task to run as root and i make sure i have the correct IOPL, the assembler
does not even have the capability to generate these instructions (they are
not in its "dictionary").

 I cannot test my mouse using the kernel, because that forces me to use the
existing serial drivers, and i have already tried all the combinations using
standard ioctl, it does not work. I have to try and duplicate what my mouse
driver is doing in DOS and the only way i can guarantee this is by direct
access.

I am NOT saying that this should be easy to do, just that it should be
possible for those pathological enough to want to bother. (Let's face it,
Josephine user is not too likely to bother using inline assembly code!)

And what about this undocumented -O6 flag in gcc? 

-- 
robin

                iNSANITY - It's only a state of mind