From patchwork Fri Mar 15 09:13:07 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Viresh Kumar X-Patchwork-Id: 1056910 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Authentication-Results: ozlabs.org; spf=none (mailfrom) smtp.mailfrom=vger.kernel.org (client-ip=209.132.180.67; helo=vger.kernel.org; envelope-from=sparclinux-owner@vger.kernel.org; receiver=) Authentication-Results: ozlabs.org; dmarc=pass (p=none dis=none) header.from=linaro.org Authentication-Results: ozlabs.org; dkim=pass (2048-bit key; unprotected) header.d=linaro.org header.i=@linaro.org header.b="lHEMOiHb"; dkim-atps=neutral Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 44LKcn3t0xz9s4Y for ; Fri, 15 Mar 2019 20:13:29 +1100 (AEDT) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1728667AbfCOJNU (ORCPT ); Fri, 15 Mar 2019 05:13:20 -0400 Received: from mail-pg1-f194.google.com ([209.85.215.194]:38311 "EHLO mail-pg1-f194.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1728612AbfCOJNU (ORCPT ); Fri, 15 Mar 2019 05:13:20 -0400 Received: by mail-pg1-f194.google.com with SMTP id v1so1307458pgi.5 for ; Fri, 15 Mar 2019 02:13:19 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=from:to:cc:subject:date:message-id:mime-version :content-transfer-encoding; bh=UkeUa0P2JUNBNYzTRx3rdebmZF7ZweVBbm012prHiz8=; b=lHEMOiHbFe0BNTK2ItHwbExU/Tvu6EIKyx/jO19kFW7dSZxhWK9hApaI9weYVRAJoP n+det+TdAJ1b0GtDbFfCLYG+6xv/zf9q9ul4z0+nyvP9T49UzcekknAQYLdnfP6BHl+2 WVp90Mhvpen6RALvShkCkdvJESbXOkEeSLJRVE0OvqZFKCfnkvKTaDChvSCDnhe7M8H1 T38O+M85nYVHNSbn6H1Ei2i8rehgdMW+ZkacA2r7VzLkoxL4JtKTfjFZassCM2vD4f97 WF+wnJnU3ST0rC2nPISkGbPJ1J6MoekasHgs7trMHd4/3aW5FuppwbhxwPXYZFYOYs7h 3FFA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:mime-version :content-transfer-encoding; bh=UkeUa0P2JUNBNYzTRx3rdebmZF7ZweVBbm012prHiz8=; b=Zfi9caFQ5ZtOCKbLIT6vAyNrFyWXnStgUPnJpe9Q3Ieb2rTBpfAx9m96ahPt4gJt2p NFk7AGdjRv4CQS2aeu+Fg+ijK4h3nNWqLADvbvEQjpzu2/BipC/+AIxp2+hFhf1Dq0cz 6S4USDGSGyZvOsJOz82/UP/jCtIsygYNy97XNgJgnYdsTKPp24hbRG3Tgj0s5izK0Whe 6CfdTaqRIkbTwGBrJhx7WzhqJRII0jjfoCHy8MctOMyZprVgD63+dCCZZ5naB5qjiSec UiEkAjSw4q4097m/OOQis/KgvGiX9UcLP6i5h86akeKZKcCDF4iP3OPAqsxyU3oOXNDk t35A== X-Gm-Message-State: APjAAAXyOOsh44JJJVzPJkR8dWiQLuHQFfSbN+jQHXewY8NIVJvvR6lS vhgDDYwJL5qy+1Bw+xPHF5K2Wg== X-Google-Smtp-Source: APXvYqworbbQYzaIoykaoNDYgD1OhlMyiiRgENEBdcAfTDg+DlYW8Eo1vfvAsi3yNPkkfMWLLduRrA== X-Received: by 2002:a17:902:b402:: with SMTP id x2mr3066111plr.53.1552641198787; Fri, 15 Mar 2019 02:13:18 -0700 (PDT) Received: from localhost ([122.166.134.37]) by smtp.gmail.com with ESMTPSA id f125sm2640436pfc.91.2019.03.15.02.13.17 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Fri, 15 Mar 2019 02:13:17 -0700 (PDT) From: Viresh Kumar To: Rafael Wysocki , Russell King , "David S. Miller" , Thomas Gleixner , Ingo Molnar , Borislav Petkov , "H. Peter Anvin" , x86@kernel.org, Paolo Bonzini , =?utf-8?b?UmFkaW0gS3LEjW3DocWZ?= Cc: Viresh Kumar , linux-pm@vger.kernel.org, Vincent Guittot , linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, sparclinux@vger.kernel.org, kvm@vger.kernel.org Subject: [PATCH V2] cpufreq: Call transition notifier only once for each policy Date: Fri, 15 Mar 2019 14:43:07 +0530 Message-Id: X-Mailer: git-send-email 2.21.0.rc0.269.g1a574e7a288b MIME-Version: 1.0 Sender: sparclinux-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: sparclinux@vger.kernel.org Currently we call these notifiers once for each CPU of the policy->cpus cpumask. It would be more optimal if the notifier can be called only once and all the relevant information be provided to it. Out of the 24 drivers that register for the transition notifiers today, only 5 of them do per-cpu updates and the callback for the rest can be called only once for the policy without any impact. This would also avoid multiple function calls to the notifier callbacks and reduce multiple iterations of notifier core's code (which does locking as well). This patch adds pointer to the cpufreq policy to the struct cpufreq_freqs, so the notifier callback has all the information available to it with a single call. The five drivers which perform per-cpu updates are updated to use the cpufreq policy. The freqs->cpu field is redundant now and is removed. Acked-by: David S. Miller (sparc) Signed-off-by: Viresh Kumar --- V1->V2: - Add cpufreq policy instead of cpus in struct cpufreq_freqs. - Use policy->cpus instead of related_cpus everywhere in order not to change the existing behavior. - Merged all 7 patches into a single patch. - Updated changlog and included Ack from David. arch/arm/kernel/smp.c | 24 +++++++++++++++--------- arch/arm/kernel/smp_twd.c | 9 ++++++--- arch/sparc/kernel/time_64.c | 28 ++++++++++++++++------------ arch/x86/kernel/tsc.c | 32 +++++++++++++++++++++----------- arch/x86/kvm/x86.c | 31 ++++++++++++++++++++----------- drivers/cpufreq/cpufreq.c | 19 ++++++++++--------- include/linux/cpufreq.h | 14 +++++++------- 7 files changed, 95 insertions(+), 62 deletions(-) diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c index 1d6f5ea522f4..6f6b981fecda 100644 --- a/arch/arm/kernel/smp.c +++ b/arch/arm/kernel/smp.c @@ -758,15 +758,20 @@ static int cpufreq_callback(struct notifier_block *nb, unsigned long val, void *data) { struct cpufreq_freqs *freq = data; - int cpu = freq->cpu; + struct cpumask *cpus = freq->policy->cpus; + int cpu, first = cpumask_first(cpus); + unsigned int lpj; if (freq->flags & CPUFREQ_CONST_LOOPS) return NOTIFY_OK; - if (!per_cpu(l_p_j_ref, cpu)) { - per_cpu(l_p_j_ref, cpu) = - per_cpu(cpu_data, cpu).loops_per_jiffy; - per_cpu(l_p_j_ref_freq, cpu) = freq->old; + if (!per_cpu(l_p_j_ref, first)) { + for_each_cpu(cpu, cpus) { + per_cpu(l_p_j_ref, cpu) = + per_cpu(cpu_data, cpu).loops_per_jiffy; + per_cpu(l_p_j_ref_freq, cpu) = freq->old; + } + if (!global_l_p_j_ref) { global_l_p_j_ref = loops_per_jiffy; global_l_p_j_ref_freq = freq->old; @@ -778,10 +783,11 @@ static int cpufreq_callback(struct notifier_block *nb, loops_per_jiffy = cpufreq_scale(global_l_p_j_ref, global_l_p_j_ref_freq, freq->new); - per_cpu(cpu_data, cpu).loops_per_jiffy = - cpufreq_scale(per_cpu(l_p_j_ref, cpu), - per_cpu(l_p_j_ref_freq, cpu), - freq->new); + + lpj = cpufreq_scale(per_cpu(l_p_j_ref, first), + per_cpu(l_p_j_ref_freq, first), freq->new); + for_each_cpu(cpu, cpus) + per_cpu(cpu_data, cpu).loops_per_jiffy = lpj; } return NOTIFY_OK; } diff --git a/arch/arm/kernel/smp_twd.c b/arch/arm/kernel/smp_twd.c index b30eafeef096..495cc7282096 100644 --- a/arch/arm/kernel/smp_twd.c +++ b/arch/arm/kernel/smp_twd.c @@ -162,15 +162,18 @@ static int twd_cpufreq_transition(struct notifier_block *nb, unsigned long state, void *data) { struct cpufreq_freqs *freqs = data; + int cpu; /* * The twd clock events must be reprogrammed to account for the new * frequency. The timer is local to a cpu, so cross-call to the * changing cpu. */ - if (state == CPUFREQ_POSTCHANGE) - smp_call_function_single(freqs->cpu, twd_update_frequency, - NULL, 1); + if (state != CPUFREQ_POSTCHANGE) + return NOTIFY_OK; + + for_each_cpu(cpu, freqs->policy->cpus) + smp_call_function_single(cpu, twd_update_frequency, NULL, 1); return NOTIFY_OK; } diff --git a/arch/sparc/kernel/time_64.c b/arch/sparc/kernel/time_64.c index 3eb77943ce12..89fb05f90609 100644 --- a/arch/sparc/kernel/time_64.c +++ b/arch/sparc/kernel/time_64.c @@ -653,19 +653,23 @@ static int sparc64_cpufreq_notifier(struct notifier_block *nb, unsigned long val void *data) { struct cpufreq_freqs *freq = data; - unsigned int cpu = freq->cpu; - struct freq_table *ft = &per_cpu(sparc64_freq_table, cpu); + unsigned int cpu; + struct freq_table *ft; - if (!ft->ref_freq) { - ft->ref_freq = freq->old; - ft->clock_tick_ref = cpu_data(cpu).clock_tick; - } - if ((val == CPUFREQ_PRECHANGE && freq->old < freq->new) || - (val == CPUFREQ_POSTCHANGE && freq->old > freq->new)) { - cpu_data(cpu).clock_tick = - cpufreq_scale(ft->clock_tick_ref, - ft->ref_freq, - freq->new); + for_each_cpu(cpu, freq->policy->cpus) { + ft = &per_cpu(sparc64_freq_table, cpu); + + if (!ft->ref_freq) { + ft->ref_freq = freq->old; + ft->clock_tick_ref = cpu_data(cpu).clock_tick; + } + + if ((val == CPUFREQ_PRECHANGE && freq->old < freq->new) || + (val == CPUFREQ_POSTCHANGE && freq->old > freq->new)) { + cpu_data(cpu).clock_tick = + cpufreq_scale(ft->clock_tick_ref, ft->ref_freq, + freq->new); + } } return 0; diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c index 3fae23834069..cff8779fc0d2 100644 --- a/arch/x86/kernel/tsc.c +++ b/arch/x86/kernel/tsc.c @@ -956,28 +956,38 @@ static int time_cpufreq_notifier(struct notifier_block *nb, unsigned long val, void *data) { struct cpufreq_freqs *freq = data; - unsigned long *lpj; - - lpj = &boot_cpu_data.loops_per_jiffy; -#ifdef CONFIG_SMP - if (!(freq->flags & CPUFREQ_CONST_LOOPS)) - lpj = &cpu_data(freq->cpu).loops_per_jiffy; -#endif + struct cpumask *cpus = freq->policy->cpus; + bool boot_cpu = !IS_ENABLED(CONFIG_SMP) || freq->flags & CPUFREQ_CONST_LOOPS; + unsigned long lpj; + int cpu; if (!ref_freq) { ref_freq = freq->old; - loops_per_jiffy_ref = *lpj; tsc_khz_ref = tsc_khz; + + if (boot_cpu) + loops_per_jiffy_ref = boot_cpu_data.loops_per_jiffy; + else + loops_per_jiffy_ref = cpu_data(cpumask_first(cpus)).loops_per_jiffy; } + if ((val == CPUFREQ_PRECHANGE && freq->old < freq->new) || (val == CPUFREQ_POSTCHANGE && freq->old > freq->new)) { - *lpj = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new); - + lpj = cpufreq_scale(loops_per_jiffy_ref, ref_freq, freq->new); tsc_khz = cpufreq_scale(tsc_khz_ref, ref_freq, freq->new); + if (!(freq->flags & CPUFREQ_CONST_LOOPS)) mark_tsc_unstable("cpufreq changes"); - set_cyc2ns_scale(tsc_khz, freq->cpu, rdtsc()); + if (boot_cpu) { + boot_cpu_data.loops_per_jiffy = lpj; + } else { + for_each_cpu(cpu, cpus) + cpu_data(cpu).loops_per_jiffy = lpj; + } + + for_each_cpu(cpu, cpus) + set_cyc2ns_scale(tsc_khz, cpu, rdtsc()); } return 0; diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index 941f932373d0..653c7da11647 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -6648,10 +6648,8 @@ static void kvm_hyperv_tsc_notifier(void) } #endif -static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long val, - void *data) +static void __kvmclock_cpufreq_notifier(struct cpufreq_freqs *freq, int cpu) { - struct cpufreq_freqs *freq = data; struct kvm *kvm; struct kvm_vcpu *vcpu; int i, send_ipi = 0; @@ -6695,17 +6693,12 @@ static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long va * */ - if (val == CPUFREQ_PRECHANGE && freq->old > freq->new) - return 0; - if (val == CPUFREQ_POSTCHANGE && freq->old < freq->new) - return 0; - - smp_call_function_single(freq->cpu, tsc_khz_changed, freq, 1); + smp_call_function_single(cpu, tsc_khz_changed, freq, 1); spin_lock(&kvm_lock); list_for_each_entry(kvm, &vm_list, vm_list) { kvm_for_each_vcpu(i, vcpu, kvm) { - if (vcpu->cpu != freq->cpu) + if (vcpu->cpu != cpu) continue; kvm_make_request(KVM_REQ_CLOCK_UPDATE, vcpu); if (vcpu->cpu != smp_processor_id()) @@ -6727,8 +6720,24 @@ static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long va * guest context is entered kvmclock will be updated, * so the guest will not see stale values. */ - smp_call_function_single(freq->cpu, tsc_khz_changed, freq, 1); + smp_call_function_single(cpu, tsc_khz_changed, freq, 1); } +} + +static int kvmclock_cpufreq_notifier(struct notifier_block *nb, unsigned long val, + void *data) +{ + struct cpufreq_freqs *freq = data; + int cpu; + + if (val == CPUFREQ_PRECHANGE && freq->old > freq->new) + return 0; + if (val == CPUFREQ_POSTCHANGE && freq->old < freq->new) + return 0; + + for_each_cpu(cpu, freq->policy->cpus) + __kvmclock_cpufreq_notifier(freq, cpu); + return 0; } diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index e10922709d13..fba38bf27d26 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -300,11 +300,14 @@ static void cpufreq_notify_transition(struct cpufreq_policy *policy, struct cpufreq_freqs *freqs, unsigned int state) { + int cpu; + BUG_ON(irqs_disabled()); if (cpufreq_disabled()) return; + freqs->policy = policy; freqs->flags = cpufreq_driver->flags; pr_debug("notification %u of frequency transition to %u kHz\n", state, freqs->new); @@ -324,10 +327,8 @@ static void cpufreq_notify_transition(struct cpufreq_policy *policy, } } - for_each_cpu(freqs->cpu, policy->cpus) { - srcu_notifier_call_chain(&cpufreq_transition_notifier_list, - CPUFREQ_PRECHANGE, freqs); - } + srcu_notifier_call_chain(&cpufreq_transition_notifier_list, + CPUFREQ_PRECHANGE, freqs); adjust_jiffies(CPUFREQ_PRECHANGE, freqs); break; @@ -337,11 +338,11 @@ static void cpufreq_notify_transition(struct cpufreq_policy *policy, pr_debug("FREQ: %u - CPUs: %*pbl\n", freqs->new, cpumask_pr_args(policy->cpus)); - for_each_cpu(freqs->cpu, policy->cpus) { - trace_cpu_frequency(freqs->new, freqs->cpu); - srcu_notifier_call_chain(&cpufreq_transition_notifier_list, - CPUFREQ_POSTCHANGE, freqs); - } + for_each_cpu(cpu, policy->cpus) + trace_cpu_frequency(freqs->new, cpu); + + srcu_notifier_call_chain(&cpufreq_transition_notifier_list, + CPUFREQ_POSTCHANGE, freqs); cpufreq_stats_record_transition(policy, freqs->new); policy->cur = freqs->new; diff --git a/include/linux/cpufreq.h b/include/linux/cpufreq.h index b160e98076e3..e327523ddff2 100644 --- a/include/linux/cpufreq.h +++ b/include/linux/cpufreq.h @@ -42,13 +42,6 @@ enum cpufreq_table_sorting { CPUFREQ_TABLE_SORTED_DESCENDING }; -struct cpufreq_freqs { - unsigned int cpu; /* cpu nr */ - unsigned int old; - unsigned int new; - u8 flags; /* flags of cpufreq_driver, see below. */ -}; - struct cpufreq_cpuinfo { unsigned int max_freq; unsigned int min_freq; @@ -156,6 +149,13 @@ struct cpufreq_policy { struct thermal_cooling_device *cdev; }; +struct cpufreq_freqs { + struct cpufreq_policy *policy; + unsigned int old; + unsigned int new; + u8 flags; /* flags of cpufreq_driver, see below. */ +}; + /* Only for ACPI */ #define CPUFREQ_SHARED_TYPE_NONE (0) /* None */ #define CPUFREQ_SHARED_TYPE_HW (1) /* HW does needed coordination */