Re: [PATCH] futex: Ensure get_futex_key_refs() always implies a barrier

[Catalin Marinas <catalin.marinas@arm.com>]

On Sat, Oct 18, 2014 at 09:19:50PM +0100, Davidlohr Bueso wrote:
> On Sat, 2014-10-18 at 14:32 -0500, Darren Hart wrote:
> > Which is not incomplete (lacking the explicit smp_mb()) added by this
> > patch. Perhaps the MB implementation of get_futex_key_refs() need not be
> > explicitly enumerated here?
> 
> Agreed, how about this:
> 
> diff --git a/kernel/futex.c b/kernel/futex.c
> index 21f7e41..7a0805a 100644
> --- a/kernel/futex.c
> +++ b/kernel/futex.c
> @@ -143,9 +143,8 @@
>   *
>   * Where (A) orders the waiters increment and the futex value read through
>   * atomic operations (see hb_waiters_inc) and where (B) orders the write
> - * to futex and the waiters read -- this is done by the barriers in
> - * get_futex_key_refs(), through either ihold or atomic_inc, depending on the
> - * futex type.
> + * to futex and the waiters read -- this is done by the barriers for both
> + * shared and private futexes in get_futex_key_refs().
>   *
>   * This yields the following case (where X:=waiters, Y:=futex):

Looks fine to me. Since Linus already picked the original patch, if you
plan to send an update for the comments please also mention the
"explicit MB (B) for private futexes" in get_futex_key_refs().

Thanks.

-- 
Catalin