Message-ID: <20090220200929.GR6960@linux.vnet.ibm.com>
Date:	Fri, 20 Feb 2009 12:09:29 -0800
From:	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To:	Mathieu Desnoyers <compudj@...stal.dyndns.org>
Cc:	ltt-dev@...ts.casi.polymtl.ca, bert.wesarg@...glemail.com,
	bob@...son.ibm.com, linux-kernel@...r.kernel.org
Subject: Re: [PATCH URCU formal] Add liveness checks to user-level RCU
	model.

On Fri, Feb 20, 2009 at 02:46:21PM -0500, Mathieu Desnoyers wrote:
> * Paul E. McKenney (paulmck@...ux.vnet.ibm.com) wrote:
> > On Fri, Feb 20, 2009 at 01:18:35PM -0500, Mathieu Desnoyers wrote:
> > > * Paul E. McKenney (paulmck@...ux.vnet.ibm.com) wrote:
> > > > Break all potentially infinite loops in both urcu_reader() and
> > > > urcu_updater(), ensure that urcu_reader() will process any memory barriers
> > > > that urcu_updater() might issue, and formulate a "never" claim that checks
> > > > to make sure that if either urcu_reader() or urcu_updater() completes,
> > > > then the other will eventually also complete.  Since urcu_reader()
> > > > now has a finite number of steps, it must eventually complete.
> > > > 
> > > > Also replace the code at the end of urcu_reader() that previously absorbed
> > > > late memory-barrier requests from urcu_updater with code in urcu_writer()
> > > > that checks to see if urcu_reader() has completed.
> > > > 
> > > > Signed-off-by: Paul E. McKenney <paulmck@...ux.vnet.ibm.com>
> > > 
> > > Thanks Paul, I'll merge it. However, I am currently reworking our spin
> > > tree so we can execute the tests in batch (rather that all at once,
> > > which consumes more memory than necessary) and also I am doing a nice
> > > build script which lets us create our own LTL formulaes for
> > > verification. The never claims will be automatically generated and
> > > verified. I'll keep you posted.
> > 
> > Sounds interesting!  Not sure what you mean by "execute the tests
> > in batch", but look forward to seeing it.
> > 
> > On the LTL, the formula "<>[] (reader_done != 0 && updater_done != 0)"
> > didn't do what I want.  The model would kick out an error with the
> > reader sitting just before the "reader_done = 1" and the updater spinning
> > waiting for the reader to respond to its memory-barrier request.
> > 
> > So I fell back to the hand-coded formula in the never clause, which
> > translates to English as "if either the reader or the updater complete,
> > then both the reader and the updater eventually complete".  There might
> > be a way to tranlate that into LTL, but I didn't immediately see one.
> > 
> > This morning I tried the weak fairness constraints (the "-f" argument
> > to ./pan) and that did allow LTL to do what I want, as shown in the
> > following patch (applied on top of my earlier patch).
> > 
> > I must confess that LTL is at best an acquired taste for me.
> > "Let's see...  '<>[](!reader_done || !updater_done)'...
> > That means eventually we always must have neither the reader or the
> > updater being done.  Huh???  Oh, yeah, this is supposed to say what
> > -cannot- happen..."  At this point, I have an easier time with the
> > hand-coded "never" claims.  ;-)
> > 
> > But I am quite happy to leave further hacking on this model in
> > your capable hands.  The other item on my todo list was making the
> > urcu_mbmin.spin model accurately handle omission of additional memory
> > barriers.  Are you willing to take that on as well?
> 
> I'll first get the translation of asserts into LTL formulaes, and try to
> see what should be fixed in the model. I have noticed that we would need
> to do this :
> 
> urcu_gp_ctr = (urcu_gp_ctr + RCU_GP_CTR_BIT) % (RCU_GP_CTR_BIT + 1);
> 
> Otherwise the overflow does not do what we expect (spin -f on the trail
> told me that it was overflowing to 1, which is not exactly what we want
> I guess). More to come on that side. When this will be settled, I'll dig
> further.

Hmmm...  The two legal values for urcu_gp_ctr are 1 and 129, so isn't
this in fact the desired behavior?  This is with your optimization that
cuts a half-cycle from rcu_read_lock() by making the initial value of
urcu_gp_ctr be 1 rather than 0.

One way of getting rid of the warning would be something like the
following:

	atomic {
		if
		:: urcu_gp_ctr == 1 -> urcu_gp_ctr = 129;
		:: urcu_gp_ctr == 129 -> urcu_gp_ctr = 1;
		fi;
	}

On converting the assert() to LTL, better you than me!  :-)

							Thanx, Paul

> Mathieu
> 
> 
> > 							Thanx, Paul
> > 
> > Signed-off-by: Paul E. McKenney <paulmck@...ux.vnet.ibm.com>
> > ---
> > 
> >  urcu.sh   |    4 ++--
> >  urcu.spin |   12 ------------
> >  2 files changed, 2 insertions(+), 14 deletions(-)
> > 
> > diff --git a/formal-model/urcu.sh b/formal-model/urcu.sh
> > index 5e525ec..3a6850c 100644
> > --- a/formal-model/urcu.sh
> > +++ b/formal-model/urcu.sh
> > @@ -20,6 +20,6 @@
> >  #
> >  # Authors: Paul E. McKenney <paulmck@...ux.vnet.ibm.com>
> >  
> > -spin -a urcu.spin 
> > +spin -a -f '<>[](!reader_done || !updater_done)' urcu.spin
> >  cc -o pan pan.c
> > -./pan -a
> > +./pan -a -f
> > diff --git a/formal-model/urcu.spin b/formal-model/urcu.spin
> > index cf1f670..851eb50 100644
> > --- a/formal-model/urcu.spin
> > +++ b/formal-model/urcu.spin
> > @@ -280,15 +280,3 @@ init {
> >  		run urcu_updater();
> >  	}
> >  }
> > -
> > -/* Require that both reader and updater eventually get done. */
> > -
> > -never {
> > -	do
> > -	:: skip;
> > -	:: reader_done != 0 || updater_done != 0 -> break;
> > -	od;
> > -accept:	do
> > -	:: reader_done == 0 || updater_done == 0;
> > -	od;
> > -}
> > 
> 
> -- 
> Mathieu Desnoyers
> OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists