Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190516080015.16033-1-duyuyang@gmail.com> <20190516080015.16033-12-duyuyang@gmail.com>
 <20190529114451.GA12812@tardis>
In-Reply-To: <20190529114451.GA12812@tardis>
From:   Yuyang Du <duyuyang@gmail.com>
Date:   Thu, 30 May 2019 15:37:50 +0800
Message-ID: <CAHttsrZ962Gw_6OA6J6GEhAx06yV70B5PEzqGEaYGDNSy57-3A@mail.gmail.com>
Subject: Re: [PATCH v2 11/17] locking/lockdep: Adjust lockdep selftest cases
To:     Boqun Feng <boqun.feng@gmail.com>
Cc:     Peter Zijlstra <peterz@infradead.org>, will.deacon@arm.com,
        Ingo Molnar <mingo@kernel.org>,
        Bart Van Assche <bvanassche@acm.org>, ming.lei@redhat.com,
        Frederic Weisbecker <frederic@kernel.org>, tglx@linutronix.de,
        paulmck@linux.ibm.com, LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Thanks for review.

On Wed, 29 May 2019 at 19:44, Boqun Feng <boqun.feng@gmail.com> wrote:
>
> > @@ -424,7 +424,7 @@ static void rwsem_ABBA2(void)
> >       ML(Y1);
> >       RSL(X1);
> >       RSU(X1);
> > -     MU(Y1); // should fail
> > +     MU(Y1); // should NOT fail
>
> I'm afraid you get this wrong ;-) reader of rwsem is non-recursive if I
> understand correctly, so case like:
>
>         Task 0                  Task 1
>
>         down_read(A);
>                                 mutex_lock(B);
>
>                                 down_read(A);
>         mutex_lock(B);
>
> can be a deadlock, if we consider a third independent task:
>
>         Task 0                  Task 1                  Task 2
>
>         down_read(A);
>                                 mutex_lock(B);
>                                                         down_write(A);
>                                 down_read(A);
>         mutex_lock(B);
>
> in this case, Task 1 can not get it's lock for A, therefore, deadlock.

Well, yes. This situation is damn counterintuitive and looks
suboptimal, but I guess I can understand why this is done so. It is a
shame read locks are not 100% concurrent. I wish I were bright enough
to have figured this out on my own.

Ok, now this perhaps can be easily remedied. it is merely a matter
that finally I can set straight the lock exclusiveness table, and then
from there the only change seems to be now only recursive-read locks
are no deadlock.

Thanks,
Yuyang