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;
From:   "Zhang, Qiang1" <qiang1.zhang@intel.com>
To:     "paulmck@kernel.org" <paulmck@kernel.org>
CC:     "frederic@kernel.org" <frederic@kernel.org>,
        "quic_neeraju@quicinc.com" <quic_neeraju@quicinc.com>,
        "joel@joelfernandes.org" <joel@joelfernandes.org>,
        "rcu@vger.kernel.org" <rcu@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: RE: [PATCH] rcu-tasks: Make shrink down to a single callback queue
 safely
Thread-Topic: [PATCH] rcu-tasks: Make shrink down to a single callback queue
 safely
Thread-Index: AQHZBHmgbK+Gas4s2UyDwnafgcYkO65bB6OAgAA6H0A=
Date:   Fri, 2 Dec 2022 23:35:09 +0000
Message-ID: <PH0PR11MB5880B8D62B5EA5643593EC62DA179@PH0PR11MB5880.namprd11.prod.outlook.com>
References: <20221130051253.1884572-1-qiang1.zhang@intel.com>
 <20221202195723.GB4001@paulmck-ThinkPad-P17-Gen-1>
In-Reply-To: <20221202195723.GB4001@paulmck-ThinkPad-P17-Gen-1>
Accept-Language: en-US
Content-Language: en-US
x-ms-exchange-antispam-messagedata-chunkcount: 1
x-ms-exchange-antispam-messagedata-0: =?us-ascii?Q?L9OSEyifURvuCZeu7pao9wQzrymdG7yWD96k7EfeMw8cbdHTGfA8xosN+IuR?=
 =?us-ascii?Q?FqFzv43f85vmoE41qu3slyEcl9Nr5yoqMCvFn1YEsKeAGecOB/6sE01eLqUV?=
 =?us-ascii?Q?FANNhQe25DuSGALtkl/Pwm1WFSrLYPTdX6pgBx24XRfD/ah7ifNsV/VY1J5b?=
 =?us-ascii?Q?QIRdT3Pa+0VFTfgkrW3KljlKfyV1g2076OpWghH9Ipo//z6E7H/BPCO3Au8n?=
 =?us-ascii?Q?1w27Wf7o8yBcWTKlU7UKcUN2cPEyoFPds9J1LBsdPUzyPX056rF+RmULfDUG?=
 =?us-ascii?Q?FwS4P3HOMHEUz7CT/RAxfVlJbjxW5pfiSNOvwjlRAHlXiHmbEjXSFxeeHfWo?=
 =?us-ascii?Q?Eh7M7ykU5vCUf+t8NeQnRZYBrOIk5ga38HrzlU+33MECgtduUai7VPtgNTnJ?=
 =?us-ascii?Q?ngTAKKpQOeKS171sI8m2MQ+FTd/oLUWFbsngGGdV7I3heztaAF2L1BHls6K1?=
 =?us-ascii?Q?7g+fi6J0RNoM3XNNOMVs8AgJbApdqJHnjx1OycgfiqKsdidlpjWc5MiaSfbs?=
 =?us-ascii?Q?iOy52ZJufqCN6ZhJZn+kR729cd30yAnDttKKj75MFkrxuCjP8bbrMxXyIJ7b?=
 =?us-ascii?Q?HAy/jQyReS2+YL9ljlMGpaxP7uGmMFov9bYDWN4swcqVsmZmTVkbv2p7eEk9?=
 =?us-ascii?Q?+nX2VWBAN9sp6XpvYsBMkwUrxTiWskbpqSA84K/d/e4b2NYcQE+PcxrZuquG?=
 =?us-ascii?Q?b4Z+bGs1R2r438QMdsuLtu0VvbOGoNXwuyS5Z2DbNsGZXQsSg0ZEFfbY8D+N?=
 =?us-ascii?Q?7VQwWTDWPS5VdMRK62/y9w+W0HTSVp/mgCdGibhuRuHKLeat48oTcam6kN8C?=
 =?us-ascii?Q?1uSfQ1QSfAbl1AtxuK19CcABL20CbXv2cnNVud1zRKhRhyCckPtRzqE8ZdOO?=
 =?us-ascii?Q?DBO1cVDswbkxVOLi2QH4OKd2n990PTlJI0yhftt0Hsw4IHGk6/CqmUYyW6/t?=
 =?us-ascii?Q?V9t7BMXvHVwf3sttnyBuW/JRfuw1FFe1Os137lcZYiAP6n2eIG9HKF5+BUGq?=
 =?us-ascii?Q?OMUIXKRowcTjbw8oWwlTIRpdetRGOP/Tudkq9hm203QSRuXc+xIalbEcH/v6?=
 =?us-ascii?Q?z3rJWH6jHeruQj6oj46TxmcH99K9mqnWeB3t9izUDc5eEWmcgrr+TXG75YS5?=
 =?us-ascii?Q?I1gil+kCxQQr5j00x6sk/Znl3YAqbQaW2CuW6qMxT+leiMKHpAmEgULLhHrP?=
 =?us-ascii?Q?cs2c7JxCXdT9Od+5DUfbz2KIJEHx8ynE+IlQlO8izA2wOsAu2lW0IKC4wmAl?=
 =?us-ascii?Q?9CvqbvzfZAaJb8lToQZkTTeb+B0WPmCIgX9ljB5fgHYOdHd00vtZ0N7z+vh2?=
 =?us-ascii?Q?K4EJOCq7NUVeMxW7M88db0P36SJPzJy2cXlp0kwH3ZykdnQQoYFPyQAQ2G4k?=
 =?us-ascii?Q?Q/uI2hPiUkznXRLazl0pjLNCYuIP4TKa6sp9T8WLfDUwIVvmYkTPM7ccoCmX?=
 =?us-ascii?Q?EN6/71gN1s8DcQYFCM4SB4SqhG3RHtKgvcu05nq8ImMXmAuH9ywbPhHy6M/k?=
 =?us-ascii?Q?eFhiV+uUFHJSjRwm3EtK8KY1DEAFv2eUV94KHrPToX4LU2zuq1DVoxnhvxOY?=
 =?us-ascii?Q?Vrdnh+IMW70YAbnfTlUZWgU9gGhpqZ4wFzSIcWTW?=
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
X-MS-Exchange-CrossTenant-AuthAs: Internal
X-MS-Exchange-CrossTenant-AuthSource: PH0PR11MB5880.namprd11.prod.outlook.com
X-MS-Exchange-CrossTenant-Network-Message-Id: 663f4cb8-ffcb-4b44-88f9-08dad4bdd9ec
X-MS-Exchange-CrossTenant-originalarrivaltime: 02 Dec 2022 23:35:09.6098
 (UTC)
X-MS-Exchange-CrossTenant-fromentityheader: Hosted
X-MS-Exchange-CrossTenant-id: 46c98d88-e344-4ed4-8496-4ed7712e255d
X-MS-Exchange-CrossTenant-mailboxtype: HOSTED
X-MS-Exchange-CrossTenant-userprincipalname: 5ciAfBO4Yvd2DO8G0SXImSrcqePmo8LA++stS5/TeARQTIguoFrzTJVhxWnpeRWW69act/HDadjcZrutyDtDaw==
X-MS-Exchange-Transport-CrossTenantHeadersStamped: CH0PR11MB5561
X-OriginatorOrg: intel.com
X-Spam-Status: No, score=-4.4 required=5.0 tests=BAYES_00,DKIMWL_WL_HIGH,
        DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,RCVD_IN_DNSWL_MED,
        SPF_HELO_NONE,SPF_NONE autolearn=ham autolearn_force=no version=3.4.6
X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on
        lindbergh.monkeyblade.net
Precedence: bulk
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Wed, Nov 30, 2022 at 01:12:53PM +0800, Zqiang wrote:
> Assume that the current RCU-task belongs to per-CPU callback queuing
> mode and the rcu_task_cb_adjust is true.
>=20
>       CPU0					CPU1
>=20
> rcu_tasks_need_gpcb()
>   ncbsnz =3D=3D 0 and
>   ncbs < rcu_task_collapse_lim
>=20
> 					      invoke call_rcu_tasks_generic()
> 						 enqueue callback to CPU1
> 					        (CPU1 n_cbs not equal zero)
>=20
>   if (rcu_task_cb_adjust &&
>   ncbs <=3D rcu_task_collapse_lim)
>     if (rtp->percpu_enqueue_lim > 1)
>       rtp->percpu_enqueue_lim =3D 1;
>       rtp->percpu_dequeue_gpseq =3D
>       get_state_synchronize_rcu();
>=20
>=20
>   A full RCU grace period has passed
>
>
>I don't see how this grace period can elapse.  The rcu_tasks_need_gpcb()
>function is invoked only from rcu_tasks_one_gp(), and while holding
>->tasks_gp_mutex.


Hi Paul

I mean that It's the RCU grace period instead of the RCU task grace period.

rtp->percpu_dequeue_gpseq =3D get_state_synchronize_rcu();

get_state_synchronize_rcu(rtp->percpu_dequeue_gpseq);

There is a window period between these two calls,  preemption may occur.
A full RCU grace period may have passed.
when we run it again we find get_state_synchronize_rcu() =3D=3D true.

Thanks
Zqiang


>
>What am I missing here?
>
>							Thanx, Paul

>   if (rcu_task_cb_adjust && !ncbsnz &&
>   poll_state_synchronize_rcu(
>     rtp->percpu_dequeue_gpseq) =3D=3D true
>     if (rtp->percpu_enqueue_lim <
> 	rtp->percpu_dequeue_lim)
>         rtp->percpu_dequeue_lim =3D 1
>     for (cpu =3D rtp->percpu_dequeue_lim;
>         cpu < nr_cpu_ids; cpu++)
>         find CPU1 n_cbs is not zero
>         trigger warning
>=20
> The above scenario will not only trigger WARN_ONCE(), but also set the
> rcu_tasks structure's->percpu_dequeue_lim is one when CPU1 still have
> callbacks, which will cause the callback of CPU1 to have no chance to be
> called.
>=20
> This commit add per-cpu callback check(except CPU0) before set the rcu_ta=
sks
> structure's->percpu_dequeue_lim to one, if other CPUs(except CPU0) still =
have
> callback, not set the rcu_tasks structure's->percpu_dequeue_lim to one, s=
et it
> until the all CPUs(except CPU0) has no callback.
>=20
> Signed-off-by: Zqiang <qiang1.zhang@intel.com>
> ---
>  kernel/rcu/tasks.h | 13 ++++++++-----
>  1 file changed, 8 insertions(+), 5 deletions(-)
>=20
> diff --git a/kernel/rcu/tasks.h b/kernel/rcu/tasks.h
> index e4f7d08bde64..690af479074f 100644
> --- a/kernel/rcu/tasks.h
> +++ b/kernel/rcu/tasks.h
> @@ -433,14 +433,17 @@ static int rcu_tasks_need_gpcb(struct rcu_tasks *rt=
p)
>  	    poll_state_synchronize_rcu(rtp->percpu_dequeue_gpseq)) {
>  		raw_spin_lock_irqsave(&rtp->cbs_gbl_lock, flags);
>  		if (rtp->percpu_enqueue_lim < rtp->percpu_dequeue_lim) {
> +			for (cpu =3D rtp->percpu_enqueue_lim; cpu < nr_cpu_ids; cpu++) {
> +				struct rcu_tasks_percpu *rtpcp =3D per_cpu_ptr(rtp->rtpcpu, cpu);
> +
> +				if(rcu_segcblist_n_cbs(&rtpcp->cblist)) {
> +					raw_spin_unlock_irqrestore(&rtp->cbs_gbl_lock, flags);
> +					return needgpcb;
> +				}
> +			}
>  			WRITE_ONCE(rtp->percpu_dequeue_lim, 1);
>  			pr_info("Completing switch %s to CPU-0 callback queuing.\n", rtp->nam=
e);
>  		}
> -		for (cpu =3D rtp->percpu_dequeue_lim; cpu < nr_cpu_ids; cpu++) {
> -			struct rcu_tasks_percpu *rtpcp =3D per_cpu_ptr(rtp->rtpcpu, cpu);
> -
> -			WARN_ON_ONCE(rcu_segcblist_n_cbs(&rtpcp->cblist));
> -		}
>  		raw_spin_unlock_irqrestore(&rtp->cbs_gbl_lock, flags);
>  	}
> =20
> --=20
> 2.25.1
>=20