Re: [PATCH] Fast Userspace Mutexes III.

[Rusty Russell <rusty@rustcorp.com.au>]

In message <1015271393.15277.112.camel@phantasy> you write:
> > +static spinlock_t futex_lock = SPIN_LOCK_UNLOCKED;
> 
> Could we make this per-waitqueue?

Yes, once someone gives benchmarks proving it's worth doing the whole
"multiple locks and cache aligned" thing.  Until then, it's premature
optimization.

> We should do:
> 
> 	#define FUTEX_UP	1
> 	#define FUTEX_DOWN	-1

Ack.  Definitely.

> here.  The preempt statements compile away if CONFIG_PREEMPT is not set,
> so you can just put them in, even on arches that don't do preemption
> yet.

Oops, that code shouldn't have been in patch, and the only reason that
preempt_disable() was commented out is that I tested the patch on 2.4.

> ... oh, and I would love an example of using it in userspace ;)

I'll throw it in for patch IV. 8)

> Nice work, Rusty.

I don't know if I can accept the kudos: it's now hovering at about 70%
my code, but only 20% my ideas.

Cheers,
Rusty.
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.