Re: patent on O_ATOMICLOOKUP - Warning actual technical content.

On Sat, May 25, 2002 at 05:39:00PM -0700, Linus Torvalds wrote:
> Side note: we could, of course, mark some spinlocks (and thus some
> code-paths) as being RT-safe, and then make sure that those spinlocks -
> when they disable interrupts - actually disable the _hw_ interrupts even
> with the RT patches.
> 
> That would make those sequences usable even from within a RT subset, but
> would obviously mean that those spinlocks have to be checked for latency
> issues - because any user (also non-RT ones) would obviously be truly
> uninterruptible within these spinlocks.

How about something more useful: interval progress assurances? Such as
"during any 5  second  period this process will be able to read X meg of
data from a file and write Y meg"

So if I have an RT task that dumps data to a DVD at millisecond intervals,
I can be sure that the non-RT task that reads the FS and puts data 
into a buffer will never let me run out of frames on a given shared memory
size.

This is useful in itself for nonRT Linux too. It seems quite hard, but it
could be relatively robust, once it was in place - making a 1 millisecond
worst case turn into a 10 millisecond worst case would not break it.