Re: interrupt latency

Chris Wedgwood (chris@cybernet.co.nz)
Wed, 12 Aug 1998 19:04:17 +1200


On Wed, Aug 12, 1998 at 12:18:11AM +0000, Linus Torvalds wrote:

> Yes. Allowing user mode programs to simulate hardware interrupts
> is a major security misfeature (imagine somebody fooling with the
> system by simulating lots of timer interrupts), and as such Linux
> doesn't allow it on the x86 even though the CPU itself would be
> able to do it fairly transparently.

Fair 'nuf.

> (However, even if Linux _did_ allow it, you'd have to use
>
> asm("int $15+32");
>
> to get IRQ15 - the kernel offsets the interrupts from 8592 by 32 to
> make sure that they don't clash with the internally generated x86
> exceptions).

I tried it before in a module, with just plan int $15, and it didn't
work.

With the offset of 32, things work as expected and I am able to
`generate interrupts'. Cool bananas.

It means the guts of what I wanted to do will now have to be done in
the module which means the interrupt times are going to be ring 0 ->
ring 0 -> ring 0, instead of ring 3 -> ring 0 -> ring3, are the
timings likely to be vastly different?

-cw

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html