From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753126AbaBJRpP (ORCPT ); Mon, 10 Feb 2014 12:45:15 -0500 Received: from cam-admin0.cambridge.arm.com ([217.140.96.50]:63596 "EHLO cam-admin0.cambridge.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753673AbaBJRpD (ORCPT ); Mon, 10 Feb 2014 12:45:03 -0500 From: Mark Rutland To: linux-kernel@vger.kernel.org Cc: will.deacon@arm.com, dave.martin@arm.com, Mark Rutland , Peter Zijlstra , Ingo Molnar Subject: [PATCH 5/7] perf: kill pmu::hrtimer_interval_ms Date: Mon, 10 Feb 2014 17:44:22 +0000 Message-Id: <1392054264-23570-6-git-send-email-mark.rutland@arm.com> X-Mailer: git-send-email 1.8.1.1 In-Reply-To: <1392054264-23570-1-git-send-email-mark.rutland@arm.com> References: <1392054264-23570-1-git-send-email-mark.rutland@arm.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org The hrtimers affecting a pmu live in perf_cpu_context, which may be shared by multiple pmus. As such it does not make sense for the pmu to track the interval, which is a property of the shared context. This patch removes pmu::hrtimer_interval_ms, and reworks all existing users to use perf_cpu_context::hrtimer_interval instead. No pmu implementations currently override the default hrtimer_interval_ms statically, so we don't lose any existing implementation-specific intervals (which could themselves be problematic if two pmus had different hrtimer_interval_ms values). While we're in the area we also replace uses of ns_to_ktime(NSEC_PER_MSEC * $x) with ms_to_ktime($x), which does the same thing. Signed-off-by: Mark Rutland Reviewed-by: Dave Martin Cc: Peter Zijlstra Cc: Ingo Molnar --- include/linux/perf_event.h | 1 - kernel/events/core.c | 24 +++++++++++++----------- 2 files changed, 13 insertions(+), 12 deletions(-) diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index df3d34a..7794a39 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -180,7 +180,6 @@ struct pmu { int * __percpu pmu_disable_count; struct perf_cpu_context * __percpu pmu_cpu_context; int task_ctx_nr; - int hrtimer_interval_ms; /* * Fully disable/enable this PMU, can be used to protect from the PMI diff --git a/kernel/events/core.c b/kernel/events/core.c index 13ede70..710c3fe 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -816,14 +816,11 @@ static void __perf_cpu_hrtimer_init(struct perf_cpu_context *cpuctx, int cpu) return; /* - * check default is sane, if not set then force to - * default interval (1/tick) + * Set sane default interval (1/tick) */ - timer = pmu->hrtimer_interval_ms; - if (timer < 1) - timer = pmu->hrtimer_interval_ms = PERF_CPU_HRTIMER; + timer = PERF_CPU_HRTIMER; - cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer); + cpuctx->hrtimer_interval = ms_to_ktime(timer); hrtimer_init(hr, CLOCK_MONOTONIC, HRTIMER_MODE_REL_PINNED); hr->function = perf_cpu_hrtimer_handler; @@ -6346,14 +6343,21 @@ type_show(struct device *dev, struct device_attribute *attr, char *page) } static DEVICE_ATTR_RO(type); +static int get_pmu_hrtimer_interval(struct pmu *pmu) +{ + struct perf_cpu_context *cpu_ctx = this_cpu_ptr(pmu->pmu_cpu_context); + return ktime_to_ms(cpu_ctx->hrtimer_interval); +} + static ssize_t perf_event_mux_interval_ms_show(struct device *dev, struct device_attribute *attr, char *page) { struct pmu *pmu = dev_get_drvdata(dev); + int interval_ms = get_pmu_hrtimer_interval(pmu); - return snprintf(page, PAGE_SIZE-1, "%d\n", pmu->hrtimer_interval_ms); + return snprintf(page, PAGE_SIZE-1, "%d\n", interval_ms); } static ssize_t @@ -6372,16 +6376,14 @@ perf_event_mux_interval_ms_store(struct device *dev, return -EINVAL; /* same value, noting to do */ - if (timer == pmu->hrtimer_interval_ms) + if (timer == get_pmu_hrtimer_interval(pmu)) return count; - pmu->hrtimer_interval_ms = timer; - /* update all cpuctx for this PMU */ for_each_possible_cpu(cpu) { struct perf_cpu_context *cpuctx; cpuctx = per_cpu_ptr(pmu->pmu_cpu_context, cpu); - cpuctx->hrtimer_interval = ns_to_ktime(NSEC_PER_MSEC * timer); + cpuctx->hrtimer_interval = ms_to_ktime(timer); if (hrtimer_active(&cpuctx->hrtimer)) hrtimer_forward_now(&cpuctx->hrtimer, cpuctx->hrtimer_interval); -- 1.8.1.1