MIME-Version: 1.0
References: <20230302062741.483079-1-jstultz@google.com> <20230302082414.77613351@gandalf.local.home>
 <CANDhNCo4ruC4pP+iDe49b3e1nAcWtYQj4bx82+oZhyLFYkdFJQ@mail.gmail.com>
 <20230302152103.2618f1b7@gandalf.local.home> <20230302163253.541ac3a8@gandalf.local.home>
 <20230302163603.223313ba@gandalf.local.home> <20230302165613.2dcc18ca@gandalf.local.home>
 <20230302200136.381468f0@gandalf.local.home> <20230303181134.GA1837196@google.com>
 <20230307140858.GE2017917@hirez.programming.kicks-ass.net>
In-Reply-To: <20230307140858.GE2017917@hirez.programming.kicks-ass.net>
From:   Joel Fernandes <joel@joelfernandes.org>
Date:   Tue, 7 Mar 2023 15:19:23 -0500
Message-ID: <CAEXW_YQY64XdhuxYuwqBkeYQFKTt4yDyHwUfBip4hf+A83mL-g@mail.gmail.com>
Subject: Re: [PATCH] pstore: Revert pmsg_lock back to a normal mutex
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Steven Rostedt <rostedt@goodmis.org>,
        John Stultz <jstultz@google.com>,
        LKML <linux-kernel@vger.kernel.org>, Wei Wang <wvw@google.com>,
        Midas Chien <midaschieh@google.com>,
        Kees Cook <keescook@chromium.org>,
        Anton Vorontsov <anton@enomsg.org>,
        "Guilherme G. Piccoli" <gpiccoli@igalia.com>,
        Tony Luck <tony.luck@intel.com>, kernel-team@android.com,
        Thomas Gleixner <tglx@linutronix.de>,
        Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, Mar 7, 2023 at 9:09=E2=80=AFAM Peter Zijlstra <peterz@infradead.org=
> wrote:
>
> On Fri, Mar 03, 2023 at 06:11:34PM +0000, Joel Fernandes wrote:
> > What  makes the rtmutex spin logic different from normal mutex in this
> > scenario, so that rtmutex wants to do that but normal ones dont?
>
> Regular mutex uses osq 'lock' to serialize waiters and only the top
> spinner gets to spin on the mutex itself, this greatly reduces the
> contention on the mutex.
>
> OSQ is FIFO, which is not what RT-mutex needs.

Got it, so basically OSQ ensures fairness as its FIFO and also reduces
lock contention because I am guessing the OSQ-spinners are spinning on
a per-spinner MCS node (that per-CPU optimistic_spin_node it appears).

This makes perfect sense now, thank you Peter!!!


 - Joel