Re: a joint letter on low latency and Linux

[Andrew Morton]

Paul Barton-Davis wrote:
> 
>  I think Ingo would
> know more about this than anyone else, though others have done some
> kernel profiling that speak to this as well. I do know, for example,
> that an IDE-based disk subsystem has much worse impact on latency,
> almost certainly because of the long periods of times for which those
> drivers tend(ed?) to disable interrupts.

It's time to remind the world of the interrupt-latency measurement patch
(freshly ported to 2.4.0-test3-pre2):
http://www.uow.edu.au/~andrewm/linux/#intlat

Sample output: http://www.uow.edu.au/~andrewm/linux/26-mar-00-ann3.txt

MontaVista have a similar patch for 2.2: http://www.mvista.com/realtime

> As a reference point, I've been told that BeOS never disables
> interrupts for more than 50usec, except in some problematic drivers
> for not-too-common hardware.

Linux can do that if you're careful and use 'hdparm -u' on IDE. 
console/video/vt/fb is the main culprit.  If you're getting latency
problems, `intlat' will find them.

[ Note to net driver maintainers: don't do a spinlock_irqsave around
your mdio_read() functions.  Use spin_lock_bh(). 330uSecs.  You _were_
protecting your media timer from do_ioctl(), weren't you? ]