Messages in this thread | | | Date | Sun, 1 Dec 2013 13:46:11 -0800 | Subject | Re: [RFC patch 0/5] futex: Allow lockless empty check of hashbucket plist in futex_wake() | From | Linus Torvalds <> |
| |
On Sun, Dec 1, 2013 at 12:39 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote: > > Last time I checked, private futexes were used when appropriate. > > "strace -e futex" mainly show _PRIVATE uses.
Yeah, pthread mutexes seem to do it. Sadly we don't do it for mm_release(), so the case of clear_child_tid doesn't trigger it, and that happened to be what I tested (I did "git diff" with the threaded pre-population of the stat data). And we can't change that because it's ABI. Sad.
So we do have a couple of corner cases where we *could* use the private ones but don't. But I guess that thread creation/exit is heavy-weight enough that that particular one doesn't really matter.
Linus
| |