On Tue, Dec 18, 2001 at 12:40:41PM +0200, Zwane Mwaikambo wrote:
> On Mon, 17 Dec 2001, Seth Arnold wrote:
>
> > Is this a feature of x86 not supporting interrupt levels natively, or is
> > there something deeper to understand? :)
>
> I think Christopher meant that interrupts can either be enabled (sti) or
> disabled (cli). Perhaps you need to elaborate on what you mean by
> interrupt priority. x86 interrupts are handled on the basis that the lower
> the interrupt number the higher the priority, but this is probably not
> what you're thinking about.
Zwane, thank you for the follow-up. I guess I do need to better specify
what I was after..
4BSD, and probably other unix-ish systems, support interrupt priority
levels to allow selectively blocking off specific interrupts when in
critical code sections. As an example, they could turn off block device
interrupts when dealing with network cards, and they could turn off
network cards and block devices when dealing with the clock..
Of course, this mostly requires support from the hardware to allow
blocking interrupts based on priority. It is my understanding that x86
doesn't have priorities on interrupts in this fashion, so I was curious
if the sti/cli functions were designed with x86 in mind, or if someone
along the way found that interrupt priorities were more work than they
were worth, and decided that linux would support 'on' and 'off' to keep
things simple...
I hope this makes sense. :) (If it doesn't, it probably isn't worth
trying to rephrase again... :)
--
"Soldiers quartered in a populous town will always occasion two mobs
where they prevent one. They are wretched conservators of the peace."
-- John Adams
PGP signature
