linux-kernel - Re: Litmus test for question from Al Viro

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: <20201005153519.GJ2628@hirez.programming.kicks-ass.net>
Date:   Mon, 5 Oct 2020 17:35:19 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Alan Stern <stern@...land.harvard.edu>
Cc:     Will Deacon <will@...nel.org>,
        "Paul E. McKenney" <paulmck@...nel.org>, parri.andrea@...il.com,
        boqun.feng@...il.com, npiggin@...il.com, dhowells@...hat.com,
        j.alglave@....ac.uk, luc.maranget@...ia.fr, akiyks@...il.com,
        dlustig@...dia.com, joel@...lfernandes.org,
        viro@...iv.linux.org.uk, linux-kernel@...r.kernel.org,
        linux-arch@...r.kernel.org
Subject: Re: Litmus test for question from Al Viro

On Mon, Oct 05, 2020 at 11:16:39AM -0400, Alan Stern wrote:
> On Mon, Oct 05, 2020 at 04:13:13PM +0100, Will Deacon wrote:
> > > The failure to recognize the dependency in P0 should be considered a 
> > > combined limitation of the memory model and herd7.  It's not a simple 
> > > mistake that can be fixed by a small rewrite of herd7; rather it's a 
> > > deliberate choice we made based on herd7's inherent design.  We 
> > > explicitly said that control dependencies extend only to the code in the 
> > > branches of an "if" statement; anything beyond the end of the statement 
> > > is not considered to be dependent.
> > 
> > Interesting. How does this interact with loops that are conditionally broken
> > out of, e.g.  a relaxed cmpxchg() loop or an smp_cond_load_relaxed() call
> > prior to a WRITE_ONCE()?
> 
> Heh --  We finesse this issue by not supporting loops at all!  :-)

Right, so something like:

	smp_cond_load_relaxed(x, !VAL);
	WRITE_ONCE(*y, 1);

Would be modeled like:

	r1 = READ_ONCE(*x);
	if (!r1)
		WRITE_ONCE(*y, 1);

with an r1==0 constraint in the condition I suppose ?