I don’t see why switching from ring 3 to ring 1 would be any slower
than switching from ring 3 to ring 0. Besides, if you really need
speed, look at the Unix Direct Rendering movement, that’s the way
to do it. And it’s going to be the odd driver that is processor bound
before it is bus or i/o bound.
Also, I/O does not need to belong in the trusted part of an OS -
interrupts included. My all-time favorite machine, the Sperry DCP
40, didn’t even have interrupts in the central processor. So, I see no
real problem in moving the whole i/o handling, interrupts et al, to
ring 1.
Alberto.
==========================
On 29 Apr 2002, at 9:27, Gregory G. Dyess wrote:
I could not agree with you more, Art. Bravo!!!
Greg
-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com]On Behalf Of Art Baker
Sent: Monday, April 29, 2002 9:05 AM
To: NT Developers Interest List
Subject: [ntdev] Re: Writing Drivers in Java…
> Actually, I believe that a Windows object oriented driver model,
> possibly running inside rings 1 and 2 instead of ring 0, would be
> a great step forward.
>
> Alberto.
During the original design of NT, there was some thought given to putting
I/O drivers in Ring 1. The idea was abandoned because (a) it made porting
the O/S to RISC platforms impossible, and (b) it proved to be really,
really, really SLOOOOOW.
> A driver is a piece of the operating system – a privileged extension of the
> I/O Manager. The correct approach is to make it “perfect” during the design,
> coding, and testing phases of its life (i.e., before it goes out into the
> world). It’s NOT appropriate to be sending out buggy drivers with the hope
> that, somehow, the operating system’s protection scheme will keep your
> driver from doing too much harm.
>
-Art Baker
You are currently subscribed to ntdev as: xxxxx@pdq.net
To unsubscribe send a blank email to %%email.unsub%%
You are currently subscribed to ntdev as: xxxxx@ieee.org
To unsubscribe send a blank email to %%email.unsub%%