Re: [PATCH RFC memory-model 0/6] LKMM updates

[Alan Stern <stern@rowland.harvard.edu>]

On Thu, 10 Jan 2019, Paul E. McKenney wrote:

> On Thu, Jan 10, 2019 at 09:40:24AM +0100, Andrea Parri wrote:
> > > > > > It seems that
> > > > > > 
> > > > > >   1b52d0186177 ("tools/memory-model: Model smp_mb__after_unlock_lock()")
> > > > > >   
> > > > > > from linux-rcu/dev got lost; this also needs an ack (probably yours! ;D,
> > > > > > considered that, IIRC, you introduced the primitive and RCU is currently
> > > > > > its only user.)
> > > > > 
> > > > > That commit is in -tip:
> > > > > 
> > > > > 4607abbcf464 ("tools/memory-model: Model smp_mb__after_unlock_lock()")
> > > > > 
> > > > > So it has already left my -rcu tree.  ;-)
> > > > 
> > > > Oh, you're right: now I see the commit (e.g., with "git show"), but I
> > > > don't see the corresponding changes applied to the tree.
> > > > 
> > > >   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=locking/core&id=4607abbcf464ea2be14da444215d05c73025cf6e
> > > >   https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/tree/tools/memory-model/linux-kernel.bell?h=locking/core
> > > > 
> > > > Is this expected?
> > > 
> > > Are you asking why it is in -tip but not in mainline?  I am not sure,
> > > but given that the merge window was over the holiday season and that
> > > the length of the merge window proved to be shorter than many people
> > > expected it to be, I am not too surprised.  ;-)
> > 
> > Mmh, let me try again:
> > 
> >   $ git clone git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git
> >   $ cd tip
> >   $ git checkout -b locking/core origin/locking/core
> > 
> >   $ git show 4607abbcf464
> >   commit 4607abbcf464ea2be14da444215d05c73025cf6e
> >   Author: Andrea Parri <andrea.parri@amarulasolutions.com>
> >   Date:   Mon Dec 3 15:04:49 2018 -0800
> > 
> >       tools/memory-model: Model smp_mb__after_unlock_lock()
> > 
> >   $ cd tools/memory-model
> >   $ herd7 -conf linux-kernel.cfg after-unlock-lock-same-cpu.litmus
> >   File "after-unlock-lock-same-cpu.litmus": Unknown macro smp_mb__after_unlock_lock (User error)
> > 
> >   [aka, linux-kernel.def in tip:locking/core does not have the
> >    smp_mb__after_unlock_lock() added by 4607abbcf464]
> 
> Color me confused:
> 
> ------------------------------------------------------------------------
> 
> $ git checkout 4607abbcf464Checking out files: 100% (18397/18397), done.
> Previous HEAD position was 4e284b1bf15a rcu: Remove wrapper definitions for obsolete RCU update functions
> HEAD is now at 4607abbcf464 tools/memory-model: Model smp_mb__after_unlock_lock()
> $ grep smp_mb__after_unlock_lock tools/memory-model/linux-kernel.def 
> smp_mb__after_unlock_lock() { __fence{after-unlock-lock}; }
> 
> ------------------------------------------------------------------------
> 
> In addition, it handles this litmus test just fine:
> 
> ------------------------------------------------------------------------
> 
> C MP+polocks
> 
> (*
>  * Result: Never
>  *
>  * This litmus test demonstrates how lock acquisitions and releases can
>  * stand in for smp_load_acquire() and smp_store_release(), respectively.
>  * In other words, when holding a given lock (or indeed after releasing a
>  * given lock), a CPU is not only guaranteed to see the accesses that other
>  * CPUs made while previously holding that lock, it is also guaranteed
>  * to see all prior accesses by those other CPUs.
>  *)
> 
> {}
> 
> P0(int *x, int *y, spinlock_t *mylock)
> {
> 	WRITE_ONCE(*x, 1);
> 	spin_lock(mylock);
> 	smp_mb__after_unlock_lock();
> 	WRITE_ONCE(*y, 1);
> 	spin_unlock(mylock);
> }
> 
> P1(int *x, int *y, spinlock_t *mylock)
> {
> 	int r0;
> 	int r1;
> 
> 	spin_lock(mylock);
> 	r0 = READ_ONCE(*y);
> 	spin_unlock(mylock);
> 	r1 = READ_ONCE(*x);
> }
> 
> exists (1:r0=1 /\ 1:r1=0)
> 
> ------------------------------------------------------------------------
> 
> Again, color me confused.

Andrea's point is that while the 1b52d0186177 commit is present in the
tip repository, it isn't in the locking/core branch.

Alan