linux-kernel - Re: [PATCH memory-model 5/8] tools/memory-model: Add a glossary of LKMM terms

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20201106210413.GB3249@paulmck-ThinkPad-P72>
Date:   Fri, 6 Nov 2020 13:04:13 -0800
From:   "Paul E. McKenney" <paulmck@...nel.org>
To:     Alan Stern <stern@...land.harvard.edu>
Cc:     linux-kernel@...r.kernel.org, linux-arch@...r.kernel.org,
        kernel-team@...com, mingo@...nel.org, parri.andrea@...il.com,
        will@...nel.org, peterz@...radead.org, boqun.feng@...il.com,
        npiggin@...il.com, dhowells@...hat.com, j.alglave@....ac.uk,
        luc.maranget@...ia.fr, akiyks@...il.com
Subject: Re: [PATCH memory-model 5/8] tools/memory-model: Add a glossary of
 LKMM terms

On Fri, Nov 06, 2020 at 03:40:08PM -0500, Alan Stern wrote:
> On Fri, Nov 06, 2020 at 11:59:12AM -0800, Paul E. McKenney wrote:
> > On Fri, Nov 06, 2020 at 02:23:51PM -0500, Alan Stern wrote:
> > > On Fri, Nov 06, 2020 at 10:04:46AM -0800, Paul E. McKenney wrote:
> > > > On Fri, Nov 06, 2020 at 11:59:30AM -0500, Alan Stern wrote:
> > > > > > +	 See also "Control Dependency".
> > > > > 
> > > > > There should also be an entry for "Data Dependency", linked from here
> > > > > and from Control Dependency.
> > > > > 
> > > > > > +Marked Access:  An access to a variable that uses an special function or
> > > > > > +	macro such as "r1 = READ_ONCE()" or "smp_store_release(&a, 1)".
> > > > > 
> > > > > How about "r1 = READ_ONCE(x)"?
> > > > 
> > > > Good catches!  I am planning to squash the commit below into the
> > > > original.  Does that cover it?
> > > 
> > > No, because you didn't add a glossary entry for "Data Dependency" and 
> > > there's no link from "Control Dependency" to "Data Dependency".
> > 
> > Sigh.  I was thinking "entry in the list", and didn't even thing to
> > check for an entry in the glossary as a whole.  With the patch below
> > (on top of the one sent earlier), are we good?
> > 
> > 							Thanx, Paul
> > 
> > ------------------------------------------------------------------------
> > 
> > commit 5a49c32551e83d30e304d6c3fbb660737ba2654e
> > Author: Paul E. McKenney <paulmck@...nel.org>
> > Date:   Fri Nov 6 11:57:25 2020 -0800
> > 
> >     fixup! tools/memory-model: Add a glossary of LKMM terms
> >     
> >     Signed-off-by: Paul E. McKenney <paulmck@...nel.org>
> > 
> > diff --git a/tools/memory-model/Documentation/glossary.txt b/tools/memory-model/Documentation/glossary.txt
> > index 471bf13..b2da636 100644
> > --- a/tools/memory-model/Documentation/glossary.txt
> > +++ b/tools/memory-model/Documentation/glossary.txt
> > @@ -64,7 +64,7 @@ Control Dependency:  When a later store's execution depends on a test
> >  	 fragile, and can be easily destroyed by optimizing compilers.
> >  	 Please see control-dependencies.txt for more information.
> >  
> > -	 See also "Address Dependency".
> > +	 See also "Address Dependency" and "Data Dependency".
> >  
> >  Cycle:	Memory-barrier pairing is restricted to a pair of CPUs, as the
> >  	name suggests.	And in a great many cases, a pair of CPUs is all
> > @@ -85,6 +85,23 @@ Cycle:	Memory-barrier pairing is restricted to a pair of CPUs, as the
> >  
> >  	See also "Pairing".
> >  
> > +Data Dependency:  When the data written by a later store is computed based
> > +	on the value returned by an earlier load, a "data dependency"
> > +	extends from that load to that later store.  For example:
> > +
> > +	 1 r1 = READ_ONCE(x);
> > +	 2 WRITE_ONCE(y, r1 + 1);
> > +
> > +	In this case, the data dependency extends from the READ_ONCE()
> > +	on line 1 to the WRITE_ONCE() on line 2.  Data dependencies are
> > +	fragile and can be easily destroyed by optimizing compilers.
> > +	Because optimizing compilers put a great deal of effort into
> > +	working out what values integer variables might have, this is
> > +	especially true in cases where the dependency is carried through
> > +	an integer.
> > +
> > +	See also "Address Dependency" and "Control Dependency".
> > +
> >  From-Reads (fr):  When one CPU's store to a given variable happened
> >  	too late to affect the value returned by another CPU's
> >  	load from that same variable, there is said to be a from-reads
> 
> Yes, this is better.

Thank you for bearing with me on this!

> Is it really true that data dependencies are so easily destroyed?  I 
> would expect that a true "semantic" dependency (i.e., one where the 
> value written really does vary according to the value read) would be 
> rather hard to second guess.

The usual optimizations apply, for but one example:

	r1 = READ_ONCE(x);
	WRITE_ONCE(y, (r1 + 1) % MAX_ELEMENTS);

If MAX_ELEMENTS is 1, so long, data dependency!

With pointers, the compiler has fewer optimization opportunities,
but there are still cases where it can break the dependency.
Or transform it to a control dependency.

							Thanx, Paul