Re: [PATCH 0/7] Add rcu_sync infrastructure to avoid _expedited() in percpu-rwsem

[Linus Torvalds]

On Sat, Jul 11, 2015 at 4:35 PM, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
>
> Linus, I am mostly trying to convince you. Nobody else objected so far.
> Could you please comment?

I don't mind this part of the series.

It's the whole "do we really want to put the effort into percpu-rwsem
I worry about, as there just aren't that many users.

The conversions made that "too damn special" thing go away, but the
conversions (particularly the big _real_ user, namely fs/locks.c) seem
to have serious performance problems that are quite possibly not
fixable.

So my objection isn't to your change, my objection is to the whole
"right now there are two users, and they both use a global lock, so
*of course* they scale like shit, and this is all just papering over
that much more fundamental problem".

I hate one-off locking. One-off locking with one global lock? Yeah,
that just smells.

                  Linus
--
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/