Message-ID: <20190728152806.GB26905@tardis>
Date:   Sun, 28 Jul 2019 23:28:06 +0800
From:   Boqun Feng <boqun.feng@...il.com>
To:     Joel Fernandes <joel@...lfernandes.org>
Cc:     Alan Stern <stern@...land.harvard.edu>,
        linux-kernel@...r.kernel.org, kernel-team@...roid.com,
        Akira Yokosawa <akiyks@...il.com>,
        Andrea Parri <andrea.parri@...rulasolutions.com>,
        Daniel Lustig <dlustig@...dia.com>,
        David Howells <dhowells@...hat.com>,
        Ingo Molnar <mingo@...nel.org>,
        Jade Alglave <j.alglave@....ac.uk>, linux-arch@...r.kernel.org,
        Luc Maranget <luc.maranget@...ia.fr>,
        Nicholas Piggin <npiggin@...il.com>,
        "Paul E. McKenney" <paulmck@...ux.ibm.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Will Deacon <will@...nel.org>
Subject: Re: [PATCH v2] lkmm/docs: Correct ->prop example with additional rfe
 link

On Sun, Jul 28, 2019 at 11:19:59AM -0400, Joel Fernandes wrote:
> On Sun, Jul 28, 2019 at 10:48:51AM -0400, Alan Stern wrote:
> > On Sat, 27 Jul 2019, Joel Fernandes (Google) wrote:
> > 
> > > The lkmm example about ->prop relation should describe an additional rfe
> > > link between P1's store to y and P2's load of y, which should be
> > > critical to establishing the ordering resulting in the ->prop ordering
> > > on P0. IOW, there are 2 rfe links, not one.
> > > 
> > > Correct these in the docs to make the ->prop ordering on P0 more clear.
> > > 
> > > Cc: kernel-team@...roid.com
> > > Reviewed-by: Boqun Feng <boqun.feng@...il.com>
> > > Signed-off-by: Joel Fernandes (Google) <joel@...lfernandes.org>
> > > ---
> > 
> > This is not a good update.  See below...
> 
> No problem, thanks for the feedback. I am new to the LKMM so please bear
> with me.. I should have tagged this RFC.
> 
> > >  .../memory-model/Documentation/explanation.txt  | 17 ++++++++++-------
> > >  1 file changed, 10 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/tools/memory-model/Documentation/explanation.txt b/tools/memory-model/Documentation/explanation.txt
> > > index 68caa9a976d0..aa84fce854cc 100644
> > > --- a/tools/memory-model/Documentation/explanation.txt
> > > +++ b/tools/memory-model/Documentation/explanation.txt
> > > @@ -1302,8 +1302,8 @@ followed by an arbitrary number of cumul-fence links, ending with an
> > >  rfe link.  You can concoct more exotic examples, containing more than
> > >  one fence, although this quickly leads to diminishing returns in terms
> > >  of complexity.  For instance, here's an example containing a coe link
> > > -followed by two fences and an rfe link, utilizing the fact that
> > > -release fences are A-cumulative:
> > > +followed by a fence, an rfe link, another fence and and a final rfe link,
> >                                                    ^---^
> > > +utilizing the fact that release fences are A-cumulative:
> > 
> > I don't like this, for two reasons.  First is the repeated "and" typo.
> 
> Will fix the trivial typo, sorry about that.
> 
> > More importantly, it's not necessary to go into this level of detail; a
> > better revision would be:
> > 
> > +followed by two cumul-fences and an rfe link, utilizing the fact that
> > 
> > This is appropriate because the cumul-fence relation is defined to 
> > contain the rfe link which you noticed wasn't mentioned explicitly.
> 
> No, I am talking about the P1's store to Y and P2's load of Y. That is not
> through a cumul-fence so I don't understand what you meant? That _is_ missing
> the rfe link I am referring to, that is left out.
> 
> The example says r2 = 1 and then we work backwards from that. r2 could very
> well have been 0, there's no fence or anything involved, it just happens to
> be the executation pattern causing r2 = 1 and hence the rfe link. Right?
> 
> > >  	int x, y, z;
> > >  
> > > @@ -1334,11 +1334,14 @@ If x = 2, r0 = 1, and r2 = 1 after this code runs then there is a prop
> > >  link from P0's store to its load.  This is because P0's store gets
> > >  overwritten by P1's store since x = 2 at the end (a coe link), the
> > >  smp_wmb() ensures that P1's store to x propagates to P2 before the
> > > -store to y does (the first fence), the store to y propagates to P2
> > > -before P2's load and store execute, P2's smp_store_release()
> > > -guarantees that the stores to x and y both propagate to P0 before the
> > > -store to z does (the second fence), and P0's load executes after the
> > > -store to z has propagated to P0 (an rfe link).
> > > +store to y does (the first fence), P2's store to y happens before P2's
> > ---------------------------------------^
> > 
> > This makes no sense, since P2 doesn't store to y.  You meant P1's store
> > to y.  Also, the use of "happens before" is here unnecessarily
> > ambiguous (is it an informal usage or does it refer to the formal
> > happens-before relation?).  The original "propagates to" is better.
> 
> Will reword this.
> 
> > > +load of y (rfe link), P2's smp_store_release() ensures that P2's load
> > > +of y executes before P2's store to z (second fence), which implies that
> > > +that stores to x and y propagate to P2 before the smp_store_release(), which
> > > +means that P2's smp_store_release() will propagate stores to x and y to all
> > > +CPUs before the store to z propagates (A-cumulative property of this fence).
> > > +Finally P0's load of z executes after P2's store to z has propagated to
> > > +P0 (rfe link).
> > 
> > Again, a better change would be simply to replace the two instances of
> > "fence" in the original text with "cumul-fence".
> 
> Ok that's fine. But I still feel the rfe is not a part of the cumul-fence.
> The fences have nothing to do with the rfe. Or, I am missing something quite
> badly.
> 
> Boqun, did you understand what Alan is saying?
> 

I think 'cumul-fence' that Alan mentioned is not a fence, but a
relation, which could be the result of combining a rfe relation and a
A-cumulative fence relation. Please see the section "PROPAGATION ORDER
RELATION: cumul-fence" or the definition of cumul-fence in
linux-kernel.cat.

Did I get you right, Alan? If so, your suggestion is indeed a better
change.

Regards,
Boqun

> thanks,
> 
>  - Joel
> 

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)

Powered by blists - more mailing lists