lkml.org 
[lkml]   [2015]   [Nov]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 4/4] locking: Introduce smp_cond_acquire()
Hi Oleg,

On Tue, Nov 03, 2015 at 06:59:58PM +0100, Oleg Nesterov wrote:
[snip]
>
> Unfortunately this doesn't look exactly right...
>
> spin_unlock_wait() is not equal to "while (locked) relax", the latter
> is live-lockable or at least sub-optimal: we do not really need to spin

Just be curious, should spin_unlock_wait() semantically be an ACQUIRE?
Because spin_unlock_wait() is used for us to wait for a certain lock to
RELEASE so that we can do something *after* we observe the RELEASE.
Considering the follow example:

CPU 0 CPU 1
============================ ===========================
{ X = 0 }
WRITE_ONCE(X, 1);
spin_unlock(&lock);
spin_unlock_wait(&lock)
r1 = READ_ONCE(X);

If spin_unlock_wait() is not an ACQUIRE, r1 can be 0 in this case,
right? Am I missing something subtle here? Or spin_unlock_wait() itself
doesn't have the ACQUIRE semantics, but it should always come with a
smp_mb() *following* it to achieve the ACQUIRE semantics? However in
do_exit(), an smp_mb() is preceding raw_spin_unlock_wait() rather than
following, which makes me confused... could you explain that? Thank you
;-)

Regards,
Boqun

> until we observe !spin_is_locked(), we only need to synchronize with the
> current owner of this lock. Once it drops the lock we can proceed, we
> do not care if another thread takes the same lock right after that.
>
> Oleg.
>
[unhandled content-type:application/pgp-signature]
\
 
 \ /
  Last update: 2015-11-11 11:01    [W:0.845 / U:0.556 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site