linux-kernel - Re: [PATCH tip/locking/core v4 1/6] powerpc: atomic: Make *xchg and *cmpxchg a full barrier

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: <20151015162942.GI3910@linux.vnet.ibm.com>
Date:	Thu, 15 Oct 2015 09:29:42 -0700
From:	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To:	Will Deacon <will.deacon@....com>
Cc:	Peter Zijlstra <peterz@...radead.org>,
	Boqun Feng <boqun.feng@...il.com>,
	linux-kernel@...r.kernel.org, linuxppc-dev@...ts.ozlabs.org,
	Ingo Molnar <mingo@...nel.org>,
	Benjamin Herrenschmidt <benh@...nel.crashing.org>,
	Paul Mackerras <paulus@...ba.org>,
	Michael Ellerman <mpe@...erman.id.au>,
	Thomas Gleixner <tglx@...utronix.de>,
	Waiman Long <waiman.long@...com>,
	Davidlohr Bueso <dave@...olabs.net>, stable@...r.kernel.org
Subject: Re: [PATCH tip/locking/core v4 1/6] powerpc: atomic: Make *xchg and
 *cmpxchg a full barrier

On Thu, Oct 15, 2015 at 03:50:44PM +0100, Will Deacon wrote:
> On Thu, Oct 15, 2015 at 11:35:10AM +0100, Will Deacon wrote:
> > Dammit guys, it's never simple is it?
> 
> I re-read this and it's even more confusing than I first thought.
> 
> > On Wed, Oct 14, 2015 at 02:44:53PM -0700, Paul E. McKenney wrote:
> > > To that end, the herd tool can make a diagram of what it thought
> > > happened, and I have attached it.  I used this diagram to try and force
> > > this scenario at https://www.cl.cam.ac.uk/~pes20/ppcmem/index.html#PPC,
> > > and succeeded.  Here is the sequence of events:
> > > 
> > > o	Commit P0's write.  The model offers to propagate this write
> > > 	to the coherence point and to P1, but don't do so yet.
> > > 
> > > o	Commit P1's write.  Similar offers, but don't take them up yet.
> > > 
> > > o	Commit P0's lwsync.
> > > 
> > > o	Execute P0's lwarx, which reads a=0.  Then commit it.
> > > 
> > > o	Commit P0's stwcx. as successful.  This stores a=1.
> > 
> > On arm64, this is a conditional-store-*release* and therefore cannot be
> > observed before the initial write to x...
> > 
> > > o	Commit P0's branch (not taken).
> > > 
> > > o	Commit P0's final register-to-register move.
> > > 
> > > o	Commit P1's sync instruction.
> > > 
> > > o	There is now nothing that can happen in either processor.
> > > 	P0 is done, and P1 is waiting for its sync.  Therefore,
> > > 	propagate P1's a=2 write to the coherence point and to
> > > 	the other thread.
> > 
> > ... therefore this is illegal, because you haven't yet propagated that
> > prior write...
> 
> I misread this as a propagation of PO's conditional store. What actually
> happens on arm64, is that the early conditional store can only succeed
> once it is placed into the coherence order of the location which it is
> updating (but note that this is subtly different from multi-copy
> atomicity!).
> 
> So, given that the previous conditional store succeeded, the coherence
> order on A must be either {0, 1, 2} or {0, 2, 1}.
> 
> If it's {0, 1, 2} (as required by your complete example), that means
> P1's a=2 write "observes" the conditional store by P0, and therefore
> (because the conditional store has release semantics), also observes
> P0's x=1 write.
> 
> On the other hand, if P1's a=2 write propagates first and we have a
> coherence order of {0, 2, 1}, then P0 must have r3=2, because an
> exclusive load returning zero would have led to a failed conditional
> store thanks to the intervening write by P1.
> 
> I find it pretty weird if PPC allows the conditional store to succeed in
> this way, as I think that would break simple cases like two threads
> incrementing a shared variable in parallel:
> 
> 
> ""
> {
> 0:r1=1; 0:r3=3; 0:r10=0 ; 0:r11=0; 0:r12=a;
> 1:r1=1; 1:r3=3; 1:r10=0 ; 1:r11=0; 1:r12=a;
> }
> P0                 | P1                 ;
> lwarx  r11,r10,r12 | lwarx  r11,r10,r12 ;
> add r11,r1,r11     | add r11,r1,r11     ;
> stwcx. r11,r10,r12 | stwcx. r11,r10,r12 ;
> bne Fail0          | bne Fail1          ;
> mr r3,r1           | mr r3,r1           ;
> Fail0:             | Fail1:             ;
> exists
> (0:r3=1 /\ a=1 /\ 1:r3=1)
> 
> 
> Is also allowed by herd and forbidden by ppcmem, for example.

I had no idea that either herd or ppcmem knew about the add instruction.
It looks like they at least try to understand.  Needless to say, in this
case I agree with ppcmem.

							Thanx, Paul

--
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/