Messages in this thread | | | Date | Thu, 23 Apr 2009 23:24:34 +0200 | From | Ingo Molnar <> | Subject | Re: [PATCH] It may not be assumed that wake_up(), finish_wait() and co. imply a memory barrier |
| |
* Linus Torvalds <torvalds@linux-foundation.org> wrote:
> On Thu, 23 Apr 2009, David Howells wrote: > > > > Well, Ingo's point is it could be left up to the caller of > > wake_up() to supply the barrier:
Well, i mainly reacted to your documentation patch which was incorrect as it said wake-up implies a _FULL_ memory barrier.
I also suggested that lockless code should have its barriers clearly documented and they should not really rely on kernel facilities acting as memory barriers.
Then i also suggested that maybe in the future we could remove the smp_wmb() from try_to_wake_up(). That was just an afterthought, and a rather stupid one at that, as Linus quickly noted :-)
Ingo
| |