Re: [PATCH v3 4/4] futex: Avoid taking hb lock if nothing to wakeup

[Darren Hart]

On Thu, 2013-12-19 at 16:04 -0800, Linus Torvalds wrote:
> On Thu, Dec 19, 2013 at 3:53 PM, Linus Torvalds
> <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
> >
> >  - in queue_lock(), immediately before getting the spinlock (which
> > will do the SAME ATOMIC INCREMENT, except it's just doing it on a
> > different member of the structure, namely the spinlock head)
> 
> Ok, so there's the "q->lock_ptr = &hb->lock" assignment in between,
> but if the ordering of that is critical, it should be documented in
> the memory ordering rules, because it sure is subtle and not obviously
> visible anywhere..

I'm trying to look into this to confirm, but it will require some time
for me to page that all in. There are some issues with respect to the
lock_ptr that are documented in the other functions (like unqueue_me),
but I *think* they are mostly on the other end of the operation and may
not be an issue here... 

q->lock_ptr == 0 also indicates "woken" state.

I will spend some time on this to try and get a definitive answer.

-- 
Darren Hart
Intel Open Source Technology Center
Yocto Project - Linux Kernel


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/