From nobody Mon May 11 07:05:14 2026 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id A24AFC433EF for ; Tue, 12 Apr 2022 20:55:39 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230450AbiDLU51 (ORCPT ); Tue, 12 Apr 2022 16:57:27 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:41942 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229379AbiDLU5T (ORCPT ); Tue, 12 Apr 2022 16:57:19 -0400 Received: from galois.linutronix.de (Galois.linutronix.de [IPv6:2a0a:51c0:0:12e:550::1]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 582F61480F3 for ; Tue, 12 Apr 2022 13:52:24 -0700 (PDT) Message-ID: <20220412204353.655068649@linutronix.de> DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linutronix.de; s=2020; t=1649796707; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: references:references; bh=fMHxW+I+DB5whhm2iNZXozQYgZvT3uBymP8I1OaqnYo=; b=Pm6IDzVTfJMphP8dUyw7QekdepbBvV0tCoeXvtoWvD7MB4gXsW8LI7rUNc2EvRQKBRIE4U XVg7rZUewqymqVzX7uQTFKhysxz5AEYTvvlB/3P7ftM8oLziYB+v10dQlMDx0LQvqkGfKi TnHVaBdklr+JuzXrTs2SA88miQvc01c67+53qb5ADsm4slqUtOkXNa+KCj2GDQldm40V/I Wy+AYssgsnXky2NctYFfEkZzXOItVlKfrXA5tyG1ibH+Cnk9ZFsOAqbgq5UfOBpSm4cMVL cVLWaLuD/+f+DfwCFJahq9b1m/ebemQv8ztAUcSmQaa2trg4Lj9uNL5LG/Fg9A== DKIM-Signature: v=1; a=ed25519-sha256; c=relaxed/relaxed; d=linutronix.de; s=2020e; t=1649796707; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: references:references; bh=fMHxW+I+DB5whhm2iNZXozQYgZvT3uBymP8I1OaqnYo=; b=B8LvWhv7tWGrAsm5jFRabEbTRYAjPohDeRTte93V6vyXkvGMaQzU4sz9Fxdq+NZ76P2IEN Fjuryv73X5VPedBg== From: Thomas Gleixner To: LKML Cc: Christoph Hellwig , Peter Zijlstra , Sebastian Andrzej Siewior Subject: [patch V4 1/2] smp: Rename flush_smp_call_function_from_idle() References: <20220412203649.956258017@linutronix.de> MIME-Version: 1.0 Date: Tue, 12 Apr 2022 22:51:46 +0200 (CEST) Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" This is invoked from the stopper thread too, which is definitely not idle. Rename it to flush_smp_call_function_queue() and fixup the callers. Signed-off-by: Thomas Gleixner Reported-by: kernel test robot --- V4: New patch --- kernel/sched/core.c | 2 +- kernel/sched/idle.c | 2 +- kernel/sched/sched.h | 4 ++-- kernel/smp.c | 27 ++++++++++++++++++++------- 4 files changed, 24 insertions(+), 11 deletions(-) --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2408,7 +2408,7 @@ static int migration_cpu_stop(void *data * __migrate_task() such that we will not miss enforcing cpus_ptr * during wakeups, see set_cpus_allowed_ptr()'s TASK_WAKING test. */ - flush_smp_call_function_from_idle(); + flush_smp_call_function_queue(); =20 raw_spin_lock(&p->pi_lock); rq_lock(rq, &rf); --- a/kernel/sched/idle.c +++ b/kernel/sched/idle.c @@ -327,7 +327,7 @@ static void do_idle(void) * RCU relies on this call to be done outside of an RCU read-side * critical section. */ - flush_smp_call_function_from_idle(); + flush_smp_call_function_queue(); schedule_idle(); =20 if (unlikely(klp_patch_pending(current))) --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -1834,10 +1834,10 @@ static inline void dirty_sched_domain_sy =20 extern int sched_update_scaling(void); =20 -extern void flush_smp_call_function_from_idle(void); +extern void flush_smp_call_function_queue(void); =20 #else /* !CONFIG_SMP: */ -static inline void flush_smp_call_function_from_idle(void) { } +static inline void flush_smp_call_function_queue(void) { } #endif =20 #include "stats.h" --- a/kernel/smp.c +++ b/kernel/smp.c @@ -96,7 +96,7 @@ static DEFINE_PER_CPU_ALIGNED(struct cal =20 static DEFINE_PER_CPU_SHARED_ALIGNED(struct llist_head, call_single_queue); =20 -static void flush_smp_call_function_queue(bool warn_cpu_offline); +static void __flush_smp_call_function_queue(bool warn_cpu_offline); =20 int smpcfd_prepare_cpu(unsigned int cpu) { @@ -141,7 +141,7 @@ int smpcfd_dying_cpu(unsigned int cpu) * ensure that the outgoing CPU doesn't go offline with work * still pending. */ - flush_smp_call_function_queue(false); + __flush_smp_call_function_queue(false); irq_work_run(); return 0; } @@ -541,11 +541,11 @@ void generic_smp_call_function_single_in { cfd_seq_store(this_cpu_ptr(&cfd_seq_local)->gotipi, CFD_SEQ_NOCPU, smp_processor_id(), CFD_SEQ_GOTIPI); - flush_smp_call_function_queue(true); + __flush_smp_call_function_queue(true); } =20 /** - * flush_smp_call_function_queue - Flush pending smp-call-function callbac= ks + * __flush_smp_call_function_queue - Flush pending smp-call-function callb= acks * * @warn_cpu_offline: If set to 'true', warn if callbacks were queued on an * offline CPU. Skip this check if set to 'false'. @@ -558,7 +558,7 @@ void generic_smp_call_function_single_in * Loop through the call_single_queue and run all the queued callbacks. * Must be called with interrupts disabled. */ -static void flush_smp_call_function_queue(bool warn_cpu_offline) +static void __flush_smp_call_function_queue(bool warn_cpu_offline) { call_single_data_t *csd, *csd_next; struct llist_node *entry, *prev; @@ -681,7 +681,20 @@ static void flush_smp_call_function_queu smp_processor_id(), CFD_SEQ_HDLEND); } =20 -void flush_smp_call_function_from_idle(void) + +/** + * flush_smp_call_function_queue - Flush pending smp-call-function callbac= ks + * from task context (idle, migration thread) + * + * When TIF_POLLING_NRFLAG is supported and a CPU is in idle and has it + * set, then remote CPUs can avoid sending IPIs and wake the idle CPU by + * setting TIF_NEED_RESCHED. The idle task on the woken up CPU has to + * handle queued SMP function calls before scheduling. + * + * The migration thread has to ensure that an eventually pending wakeup has + * been handled before it migrates a task. + */ +void flush_smp_call_function_queue(void) { unsigned long flags; =20 @@ -691,7 +704,7 @@ void flush_smp_call_function_from_idle(v cfd_seq_store(this_cpu_ptr(&cfd_seq_local)->idle, CFD_SEQ_NOCPU, smp_processor_id(), CFD_SEQ_IDLE); local_irq_save(flags); - flush_smp_call_function_queue(true); + __flush_smp_call_function_queue(true); if (local_softirq_pending()) do_softirq(); From nobody Mon May 11 07:05:14 2026 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id E699CC433EF for ; Tue, 12 Apr 2022 20:55:07 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230098AbiDLU5W (ORCPT ); Tue, 12 Apr 2022 16:57:22 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:41966 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229682AbiDLU5U (ORCPT ); Tue, 12 Apr 2022 16:57:20 -0400 Received: from galois.linutronix.de (Galois.linutronix.de [IPv6:2a0a:51c0:0:12e:550::1]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 86DB0C9B41 for ; Tue, 12 Apr 2022 13:52:25 -0700 (PDT) Message-ID: <20220412204353.712177025@linutronix.de> DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linutronix.de; s=2020; t=1649796708; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: references:references; bh=hndER118dmw7kYMFXyYuvsxJ9ycP9/DqyDamURXJ4cA=; b=AmRwcY/LhGb9vNhP+WyUubPJbFc+2etjHtIp065eY+h3wqeHg6fkEeELkQDzX6b8Gwryoe Iqmxq2EcM6m4fWPmRGDklO0XngB85nyLrEQNvuVhR8IyuiKGroEaUO0oDaYSfUpIkQqoaD G3zdaN69kbndjLQgIT7cfKZUj3JOaw95laXDMqttXE5sRBoiSam80uq0KbSURbjqYbMZcq 9ZSvI1ufcF3vOSEhpXUpnUWCRdCSZZ7rSWshUSn8OTyupuY4gLz+xXjJt/SFV1OENIR2cT xdopFgHYmq3GpawN5imMJD1pbEtE/jyndrweg1cclKmRz8tmdcgt3wkRVP7mGw== DKIM-Signature: v=1; a=ed25519-sha256; c=relaxed/relaxed; d=linutronix.de; s=2020e; t=1649796708; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: references:references; bh=hndER118dmw7kYMFXyYuvsxJ9ycP9/DqyDamURXJ4cA=; b=RKZPDyIMtrgo2N2+DgcPZfB7tgeC4aJOd42vo0AKwK+Ehwfs2V4TgjDl9RWxJChp5NVyFN /nm4BKohR+tNKjBw== From: Thomas Gleixner To: LKML Cc: Christoph Hellwig , Peter Zijlstra , Sebastian Andrzej Siewior Subject: [patch V4 2/2] smp: Make softirq handling RT safe in flush_smp_call_function_queue() References: <20220412203649.956258017@linutronix.de> MIME-Version: 1.0 Date: Tue, 12 Apr 2022 22:51:48 +0200 (CEST) Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" From: Sebastian Andrzej Siewior flush_smp_call_function_queue() invokes do_softirq() which is not available on PREEMPT_RT. flush_smp_call_function_queue() is invoked from the idle task and the migration task with preemption or interrupts disabled. So RT kernels cannot process soft interrupts in that context as that has to acquire 'sleeping spinlocks' which is not possible with preemption or interrupts disabled and forbidden from the idle task anyway. The currently known SMP function call which raises a soft interrupt is in the block layer, but this functionality is not enabled on RT kernels due to latency and performance reasons. RT could wake up ksoftirqd unconditionally, but this wants to be avoided if there were soft interrupts pending already when this is invoked in the context of the migration task. The migration task might have preempted a threaded interrupt handler which raised a soft interrupt, but did not reach the local_bh_enable() to process it. The "running" ksoftirqd might prevent the handling in the interrupt thread context which is causing latency issues. Add a new function which handles this case explicitely for RT and falls back to do_softirq() on !RT kernels. In the RT case this warns when one of the flushed SMP function calls raised a soft interrupt so this can be investigated. [ tglx: Moved the RT part out of SMP code ] Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Thomas Gleixner Link: https://lore.kernel.org/r/YgKgL6aPj8aBES6G@linutronix.de --- v4: - Move the RT logic into softirq.c which also avoids the wakeup when softinterrupts are disabled. The enable will handle them anyway. v3: - Only wake ksoftirqd if the softirqs were raised wthin flush_smp_call_function_queue(). - Add a warning in the wake case. v2: Drop an empty line. include/linux/interrupt.h | 9 +++++++++ kernel/smp.c | 5 ++++- kernel/softirq.c | 13 +++++++++++++ 3 files changed, 26 insertions(+), 1 deletion(-) --- --- a/include/linux/interrupt.h +++ b/include/linux/interrupt.h @@ -607,6 +607,15 @@ struct softirq_action asmlinkage void do_softirq(void); asmlinkage void __do_softirq(void); =20 +#ifdef CONFIG_PREEMPT_RT +extern void do_softirq_post_smp_call_flush(unsigned int was_pending); +#else +static inline void do_softirq_post_smp_call_flush(unsigned int unused) +{ + do_softirq(); +} +#endif + extern void open_softirq(int nr, void (*action)(struct softirq_action *)); extern void softirq_init(void); extern void __raise_softirq_irqoff(unsigned int nr); --- a/kernel/smp.c +++ b/kernel/smp.c @@ -696,6 +696,7 @@ static void __flush_smp_call_function_qu */ void flush_smp_call_function_queue(void) { + unsigned int was_pending; unsigned long flags; =20 if (llist_empty(this_cpu_ptr(&call_single_queue))) @@ -704,9 +705,11 @@ void flush_smp_call_function_queue(void) cfd_seq_store(this_cpu_ptr(&cfd_seq_local)->idle, CFD_SEQ_NOCPU, smp_processor_id(), CFD_SEQ_IDLE); local_irq_save(flags); + /* Get the already pending soft interrupts for RT enabled kernels */ + was_pending =3D local_softirq_pending(); __flush_smp_call_function_queue(true); if (local_softirq_pending()) - do_softirq(); + do_softirq_post_smp_call_flush(was_pending); =20 local_irq_restore(flags); } --- a/kernel/softirq.c +++ b/kernel/softirq.c @@ -294,6 +294,19 @@ static inline void invoke_softirq(void) wakeup_softirqd(); } =20 +/* + * flush_smp_call_function_queue() can raise a soft interrupt in a function + * call. On RT kernels this is undesired and the only known functionality + * in the block layer which does this is disabled on RT. If soft interrupts + * get raised which haven't been raised before the flush, warn so it can be + * investigated. + */ +void softirq_post_smp_call_flush(unsigned int was_pending) +{ + if (WARN_ON_ONCE(was_pending !=3D local_softirq_pending())) + invoke_softirq(); +} + #else /* CONFIG_PREEMPT_RT */ =20 /*