Re: [patch] "HT scheduler", sched-2.5.63-B3

[Linus Torvalds]

On Thu, 6 Mar 2003, Ingo Molnar wrote:
> > Oh, well. I didn't actually even verify that UNIX domain sockets will
> > cause synchronous wakeups, so the patch may literally be doing nothing
> > at all. You can try that theory out by just removing the test for
> > "in_interrupt()".
> 
> you are not referring to the 'synchronous wakeups' as used by fs/pipe.c,
> right?

No, sorry. Bad choice of words.

The traditional "synchronous wakeups" as used by fs/pipe.c is a hint to 
the scheduler that the waker will go to sleep.

And no, that's not the hint I'm using at all. I'm only interested in 
"process-synchronous", since if the wakeup isn't process-synchronous then 
"current" doesn't make much sense to me.

> so i think your current patch should cover unix domain sockets just as
> well, they certain dont use IRQ-context wakeups.

Note that "in_interrupt()" will also trigger for callers that call from
bh-atomic regions as well as actual BH handlers. Which is correct - they 
are both "interrupt contexts" as far as most users should be concerned.

The unix domain case may well be bh-atomic, I haven't looked at the code. 
I'm pretty much certain that the TCP case _will_ be BH-atomic, even for 
loopback.

David?

		Linus