Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 19 Jul 2022 09:25:56 +0100
Message-ID: <87o7xlzey3.wl-maz@kernel.org>
From:   Marc Zyngier <maz@kernel.org>
To:     Neeraj Upadhyay <quic_neeraju@quicinc.com>
Cc:     <paulmck@kernel.org>, <frederic@kernel.org>,
        <josh@joshtriplett.org>, <rostedt@goodmis.org>,
        <mathieu.desnoyers@efficios.com>, <jiangshanlai@gmail.com>,
        <joel@joelfernandes.org>, <linux-kernel@vger.kernel.org>,
        <zhangfei.gao@foxmail.com>, <boqun.feng@gmail.com>,
        <urezki@gmail.com>, <shameerali.kolothum.thodi@huawei.com>,
        <pbonzini@redhat.com>, <mtosatti@redhat.com>,
        <eric.auger@redhat.com>, <chenxiang66@hisilicon.com>,
        <zhangfei.gao@linaro.org>, <rcu@vger.kernel.org>
Subject: Re: [PATCH v3] srcu: Reduce blocking agressiveness of expedited grace periods further
In-Reply-To: <20220701031545.9868-1-quic_neeraju@quicinc.com>
References: <20220701031545.9868-1-quic_neeraju@quicinc.com>
User-Agent: Wanderlust/2.15.9 (Almost Unreal) SEMI-EPG/1.14.7 (Harue)
 FLIM-LB/1.14.9 (=?UTF-8?B?R29qxY0=?=) APEL-LB/10.8 EasyPG/1.0.0 Emacs/27.1
 (x86_64-pc-linux-gnu) MULE/6.0 (HANACHIRUSATO)
MIME-Version: 1.0 (generated by SEMI-EPG 1.14.7 - "Harue")
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi folks,

On Fri, 01 Jul 2022 04:15:45 +0100,
Neeraj Upadhyay <quic_neeraju@quicinc.com> wrote:
>=20
> Commit 640a7d37c3f4 ("srcu: Block less aggressively for expedited
> grace periods") highlights a problem where aggressively blocking
> SRCU expedited grace periods, as was introduced in commit
> 282d8998e997 ("srcu: Prevent expedited GPs and blocking readers
> from consuming CPU"), introduces ~2 minutes delay to the overall
> ~3.5 minutes boot time, when starting VMs with "-bios QEMU_EFI.fd"
> cmdline on qemu, which results in very high rate of memslots
> add/remove, which causes > ~6000 synchronize_srcu() calls for
> kvm->srcu SRCU instance.
>=20
> Below table captures the experiments done by Zhangfei Gao and Shameer
> to measure the boottime impact with various values of non-sleeping
> per phase counts, with HZ_250 and preemption enabled:
>=20
> +=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=
=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=
=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=
=80=E2=94=80+=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=
=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=
=E2=94=80+
> | SRCU_MAX_NODELAY_PHASE   | Boot time (s)  |
> +=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=
=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=
=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=
=80=E2=94=80+=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=
=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=
=E2=94=80+
> | 100                      | 30.053         |
> | 150                      | 25.151         |
> | 200                      | 20.704         |
> | 250                      | 15.748         |
> | 500                      | 11.401         |
> | 1000                     | 11.443         |
> | 10000                    | 11.258         |
> | 1000000                  | 11.154         |
> +=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=
=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=
=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=
=80=E2=94=80+=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=
=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=E2=94=80=
=E2=94=80+
>=20
> Analysis on the experiment results showed improved boot time
> with non blocking delays close to one jiffy duration. This
> was also seen when number of per-phase iterations were scaled
> to one jiffy.
>=20
> So, this change scales per-grace-period phase number of non-sleeping
> polls, such that, non-sleeping polls are done for one jiffy. In addition
> to this, srcu_get_delay() call in srcu_gp_end(), which is used to calcula=
te
> the delay used for scheduling callbacks, is replaced with the check for
> expedited grace period. This is done, to schedule cbs for completed exped=
ited
> grace periods immediately, which results in improved boot time seen in
> experiments. Testing done by Marc and Zhangfei confirms that this change =
recovers
> most of the performance degradation in boottime; for CONFIG_HZ_250 config=
uration,
> boottime improves from 3m50s to 41s on Marc's setup; and from 2m40s to ~9=
.7s
> on Zhangfei's setup.
>=20
> In addition to the changes to default per phase delays, this change
> adds 3 new kernel parameters - srcutree.srcu_max_nodelay,
> srcutree.srcu_max_nodelay_phase, srcutree.srcu_retry_check_delay.
> This allows users to configure the srcu grace period scanning delays,
> depending on their system configuration requirements.
>=20
> Signed-off-by: Neeraj Upadhyay <quic_neeraju@quicinc.com>
> Tested-by: Marc Zyngier <maz@kernel.org>
> Tested-by: Zhangfei Gao <zhangfei.gao@linaro.org>

Is there any chance for this fix to make it into 5.19? The regression
is significant enough on low-end systems, and I'd rather see it
addressed.

Thanks,

	M.

--=20
Without deviation from the norm, progress is not possible.