Re: [RFC][PATCH] make global bitlock waitqueues per-node

[Nicholas Piggin]

On Wed, 21 Dec 2016 11:50:49 -0800
Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:

> On Wed, Dec 21, 2016 at 11:01 AM, Nicholas Piggin <npiggin@xxxxxxxxx> wrote:
> > Peter's patch is less code and in that regard a bit nicer. I tried
> > going that way once, but I just thought it was a bit too sloppy to
> > do nicely with wait bit APIs.  
> 
> So I have to admit that when I read through your and PeterZ's patches
> back-to-back, yours was easier to understand.
> 
> PeterZ's is smaller but kind of subtle. The whole "return zero from
> lock_page_wait() and go around again" and the locking around that
> isn't exactly clear. In contrast, yours has the obvious waitqueue
> spinlock.
> 
> I'll think about it.  And yes, it would be good to have more testing,
> but at the same time xmas is imminent, and waiting around too much
> isn't going to help either..

Sure. Let's see if Dave and Mel get a chance to do some testing.

It might be a squeeze before Christmas. I realize we're going to fix
it anyway so on one hand might as well get something in. On the other
I didn't want to add a subtle bug then have everyone go on vacation.

How about I send up the page flag patch by Friday and that can bake
while the main patch gets more testing / review?

Thanks,
Nick