linux-kernel - Re: [PATCH] tools/memory-model: Provide extra ordering for unlock+lock pair on the same CPU

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: <YVakskMQGrnOUHsa@boqun-archlinux>
Date:   Fri, 1 Oct 2021 14:03:30 +0800
From:   Boqun Feng <boqun.feng@...il.com>
To:     Alan Stern <stern@...land.harvard.edu>
Cc:     "Paul E. McKenney" <paulmck@...nel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Dan Lustig <dlustig@...dia.com>, Will Deacon <will@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Peter Anvin <hpa@...or.com>,
        Andrea Parri <parri.andrea@...il.com>,
        Ingo Molnar <mingo@...nel.org>,
        Vince Weaver <vincent.weaver@...ne.edu>,
        Thomas Gleixner <tglx@...utronix.de>,
        Jiri Olsa <jolsa@...hat.com>,
        Arnaldo Carvalho de Melo <acme@...hat.com>,
        Stephane Eranian <eranian@...gle.com>, palmer@...belt.com,
        paul.walmsley@...ive.com, mpe@...erman.id.au
Subject: Re: [PATCH] tools/memory-model: Provide extra ordering for
 unlock+lock pair on the same CPU

On Thu, Sep 30, 2021 at 09:30:55PM -0400, Alan Stern wrote:
> On Fri, Oct 01, 2021 at 08:12:56AM +0800, Boqun Feng wrote:
> > On Thu, Sep 30, 2021 at 04:46:34PM -0400, Alan Stern wrote:
> > > On Thu, Sep 30, 2021 at 11:17:53AM -0700, Paul E. McKenney wrote:
> > > > On Thu, Sep 30, 2021 at 11:20:33AM -0400, Alan Stern wrote:
> > > > > On Thu, Sep 30, 2021 at 09:08:23PM +0800, Boqun Feng wrote:
> > > > > > A recent discussion[1] shows that we are in favor of strengthening the
> > > > > > ordering of unlock + lock on the same CPU: a unlock and a po-after lock
> > > > > > should provide the so-called RCtso ordering, that is a memory access S
> > > > > > po-before the unlock should be ordered against a memory access R
> > > > > > po-after the lock, unless S is a store and R is a load.
> > > > > > 
> > > > > > The strengthening meets programmers' expection that "sequence of two
> > > > > > locked regions to be ordered wrt each other" (from Linus), and can
> > > > > > reduce the mental burden when using locks. Therefore add it in LKMM.
> > > > > > 
> > > > > > [1]: https://lore.kernel.org/lkml/20210909185937.GA12379@rowland.harvard.edu/
> > > > > > 
> > > > > > Co-developed-by: Alan Stern <stern@...land.harvard.edu>
> > > > > > Signed-off-by: Alan Stern <stern@...land.harvard.edu>
> > > > > > Signed-off-by: Boqun Feng <boqun.feng@...il.com>
> > > > > > ---
> > > > > > Alan,
> > > > > > 
> > > > > > I added the "Co-developed-by" and "Signed-off-by" tags since most of the
> > > > > > work is done by you. Feel free to let me know if you want to change
> > > > > > anything.
> > > > > 
> > > > > It looks good to me.  However, do we really want to add these litmus
> > > > > tests to the kernel source, or would it be better to keep them with
> > > > > the thousands of other tests in Paul's archives?
> > > > 
> > > > Either way works for me.  But if they are referred to from within the
> > > > kernel, it is best to have them in the kernel source.  Which might be seen
> > > > as a reason to minimize referring to litmus tests from the kernel.  ;-)
> > > 
> > > In this case the litmus tests are not referred to within the kernel 
> > > source.
> > > 
> > 
> > I'm OK to drop the litmus tests, but the reason I add the two litmus
> > tests is that they directly describe one of our memory model rules. The
> > two litmus tests tells developers "you can use unlock+lock on the same
> > CPU to order READ->WRITE or WRITE->WRITE", so they are kind of part of
> > the manual of our memory model. Thoughts?
> 
> The explanation.txt file already contains example litmus tests (not 
> in a form acceptable to herd7, though) for these things.
> 

Agreed. I just think that runnable litmus tests may be more accurate
than words. But again, I'm OK to drop them.

While we are at it, maybe it's worthwhile to discuss what kind of litmus
tests should be put in tools/memory-model/litmus-tests. Selftests for
.cat changes? Documented patterns for developers to follow? Or something
else?

Regards,
Boqun

> Alan