FW: Commit 81a43adae3b9 (locking/mutex: Use acquire/release semantics) causing failures on arm64 (ThunderX)

[Peter Zijlstra]

On Fri, Dec 11, 2015 at 02:35:40PM -0800, Paul E. McKenney wrote:
> On Fri, Dec 11, 2015 at 02:48:03PM +0100, Peter Zijlstra wrote:
> > On Fri, Dec 11, 2015 at 01:33:14PM +0000, Will Deacon wrote:
> > > On Fri, Dec 11, 2015 at 01:26:47PM +0100, Peter Zijlstra wrote:
> > 
> > > > While we're there, the acquire in osq_wait_next() seems somewhat ill
> > > > documented too.
> > > > 
> > > > I _think_ we need ACQUIRE semantics there because we want to strictly
> > > > order the lock-unqueue A,B,C steps and we get that with:
> > > > 
> > > >  A: SC
> > > >  B: ACQ
> > > >  C: Relaxed
> > > > 
> > > > Similarly for unlock we want the WRITE_ONCE to happen after
> > > > osq_wait_next, but in that case we can even rely on the control
> > > > dependency there.
> > > 
> > > Even for the lock-unqueue case, isn't B->C ordered by a control dependency
> > > because C consists only of stores?
> > 
> > Hmm, indeed. So we could go fully relaxed on it I suppose, since the
> > same is true for the unlock site.
> 
> I am probably missing quite a bit on this thread, but don't x86 MMIO
> accesses to frame buffers need to interact with something more heavyweight
> than an x86 release store or acquire load in order to remain confined
> to the resulting critical section?

So on x86 there really isn't a problem because every atomic op (and
there's plenty here) will be a full barrier.

That is, even if you were to replace everything with _relaxed() ops, it
would still work as 'expected' on x86.

ppc/arm64 will crash and burn, but that's another story.

But the important point here was that osq_wait_next() is never relied
upon to provide either the ACQUIRE semantics for osq_lock() not the
RELEASE semantics for osq_unlock(). Those are provided by other ops.