Received: by 2002:ac0:a581:0:0:0:0:0 with SMTP id m1-v6csp7968759imm; Thu, 28 Jun 2018 12:14:58 -0700 (PDT) X-Google-Smtp-Source: ADUXVKIGoUUoq6YwTPBIB9+fJ15y23a9rzwtL5HaPaudG3FGvtE4kzd7RLGTCGMYoL9dwpzpZnWy X-Received: by 2002:a63:7454:: with SMTP id e20-v6mr10148573pgn.410.1530213298649; Thu, 28 Jun 2018 12:14:58 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1530213298; cv=none; d=google.com; s=arc-20160816; b=ijXNi/Th3fUWB1iUWiJQbovGQ9JHJxddiI5zwGmBLLlo4KYIZHC6Hy4PcKlNQRxvYx S8d1yicQFWrAVSSJRI8VSvTg9xIyNQdWpWhyLCIOqeGh7BN/1MMUncLN8bq/6fuSo/W8 fcFtTBF2qisvcnFLP9d+5j3Up5GZ6K+lO5iw6V6gLaPF01S3UXM9qCKtn8zGrCKczzWR uWq4yUJWSGF2c4yoqYaJh4hX+Az79L/QlTwuLiT7FFF4feLv3ZcdbeEpheLxYR/v3/Gb Pl/p/uY2voHBnYQ95WkKwoo2bnGH/0RxNRNBJu93nPJIH0i8OEoo6nTz4QRgLr2kON10 UXHw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:sender:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:arc-authentication-results; bh=PNMCXAGYTbYNG9OSboAc9ufaaa3X/61Oq4OdnYqzSOI=; b=hVJhgagX189pYmRQYO5WXhLODNv7nD96+Sw+fiTII34gjRbWu46ICLMzbZ1Q27VbBY IxOpu23gbRuD2zq2XisL8ssLiptp5rEqCjD0n6HFjLOemPapIsK2bzXtgy5OwQulVu1K xBy6wd5A2ExU3wUGzlOjNkssR5ZSrdPfjsPFi/UVKgR4oDNAX/4AltsEo3aI49ixKIxG W3kv5JRSKXLFOaJBmVdBXO1gEsAwG/b00DOa+BUw3B1W/3DjbN9NwGFHI2RNcqiRq/YY YjPIwl0e9BONmKkHg5rGDrL2l+aoshmRsTfBQS5D4T/EAFEWq+VGOoGpwFtrUurrmxm9 tOcA== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id 193-v6si6302317pgh.407.2018.06.28.12.14.43; Thu, 28 Jun 2018 12:14:58 -0700 (PDT) Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S935650AbeF1QlM (ORCPT + 99 others); Thu, 28 Jun 2018 12:41:12 -0400 Received: from mx0a-00010702.pphosted.com ([148.163.156.75]:44608 "EHLO mx0b-00010702.pphosted.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S935611AbeF1QlK (ORCPT ); Thu, 28 Jun 2018 12:41:10 -0400 Received: from pps.filterd (m0098781.ppops.net [127.0.0.1]) by mx0a-00010702.pphosted.com (8.16.0.22/8.16.0.22) with SMTP id w5SGZuhL029674; Thu, 28 Jun 2018 11:40:47 -0500 Received: from ni.com (skprod2.natinst.com [130.164.80.23]) by mx0a-00010702.pphosted.com with ESMTP id 2jw1pr8d6v-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Thu, 28 Jun 2018 11:40:47 -0500 Received: from us-aus-exch1.ni.corp.natinst.com (us-aus-exch1.ni.corp.natinst.com [130.164.68.11]) by us-aus-skprod2.natinst.com (8.16.0.22/8.16.0.22) with ESMTPS id w5SGekuP018127 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-SHA384 bits=256 verify=NOT); Thu, 28 Jun 2018 11:40:46 -0500 Received: from us-aus-exhub1.ni.corp.natinst.com (130.164.68.41) by us-aus-exch1.ni.corp.natinst.com (130.164.68.11) with Microsoft SMTP Server (TLS) id 15.0.1156.6; Thu, 28 Jun 2018 11:40:46 -0500 Received: from harisdt.amer.corp.natinst.com (130.164.49.7) by us-aus-exhub1.ni.corp.natinst.com (130.164.68.41) with Microsoft SMTP Server id 15.0.1156.6 via Frontend Transport; Thu, 28 Jun 2018 11:40:46 -0500 From: Haris Okanovic To: , CC: , , , , , , , , Subject: [PATCH v5 1/2] timers: Don't wake ktimersoftd on every tick Date: Thu, 28 Jun 2018 11:40:35 -0500 Message-ID: <20180628164036.27378-1-haris.okanovic@ni.com> X-Mailer: git-send-email 2.17.1 In-Reply-To: <7d21ff63-3dce-b8ea-a0f1-81d53617747d@ni.com> References: <7d21ff63-3dce-b8ea-a0f1-81d53617747d@ni.com> MIME-Version: 1.0 Content-Type: text/plain X-Proofpoint-Virus-Version: vendor=fsecure engine=2.50.10434:,, definitions=2018-06-28_08:,, signatures=0 X-Proofpoint-Spam-Reason: safe Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Collect expired timers in interrupt context to avoid overhead of waking ktimersoftd on every scheduler tick. This is implemented by storing lists of expired timers in the timer_base struct, which is updated by the interrupt routing on each tick in run_local_timers(). TIMER softirq (ktimersoftd) is then raised only when one or more expired timers are collected. Performance impact on a 2core Intel Atom E3825 system: * reduction in small latency spikes measured by cyclictest * ~30% fewer context-switches measured by perf * run_local_timers() execution time increases by 0.2 measured by TSC Signed-off-by: Haris Okanovic --- [PATCH v2] Applied Thomas Gleixner's suggestions: - Fix expired_count race - Remove unneeded base->clk lookahead - Return expired_count in collect_expired_timers() - Add block_softirq - Rebase to v4.11.8-rt5 [PATCH v3] - Fix cosmetic issues - Rename "count" to "levels" in timer_base and various functions - Move expired_levels and block_softirq to fill holes in timer_base - Remove READ_ONCE/WRITE_ONCE around block_softirq [PATCH v4] - Rebase onto v4.14.20-rt17 [PATCH v5] - Fix hang when timer is rearmed with 0 offset in it's callback - Rewrite description https://github.com/harisokanovic/linux/tree/dev/hokanovi/timer-peek-v7 --- kernel/time/timer.c | 111 ++++++++++++++++++++++++++++++++------------ 1 file changed, 82 insertions(+), 29 deletions(-) diff --git a/kernel/time/timer.c b/kernel/time/timer.c index 96fd01c9f6b1..dd67c18c16d0 100644 --- a/kernel/time/timer.c +++ b/kernel/time/timer.c @@ -197,6 +197,7 @@ EXPORT_SYMBOL(jiffies_64); struct timer_base { raw_spinlock_t lock; + int expired_levels; struct timer_list *running_timer; #ifdef CONFIG_PREEMPT_RT_FULL struct swait_queue_head wait_for_running_timer; @@ -208,6 +209,7 @@ struct timer_base { bool must_forward_clk; DECLARE_BITMAP(pending_map, WHEEL_SIZE); struct hlist_head vectors[WHEEL_SIZE]; + struct hlist_head expired_lists[LVL_DEPTH]; } ____cacheline_aligned; static DEFINE_PER_CPU(struct timer_base, timer_bases[NR_BASES]); @@ -1342,7 +1344,8 @@ static void call_timer_fn(struct timer_list *timer, void (*fn)(unsigned long), } } -static void expire_timers(struct timer_base *base, struct hlist_head *head) +static void __expire_timers(struct timer_base *base, + struct hlist_head *head) { while (!hlist_empty(head)) { struct timer_list *timer; @@ -1372,21 +1375,49 @@ static void expire_timers(struct timer_base *base, struct hlist_head *head) } } -static int __collect_expired_timers(struct timer_base *base, - struct hlist_head *heads) +static void expire_timers(struct timer_base *base) +{ + struct hlist_head *head; + int levels = base->expired_levels; + + while (levels--) { + head = base->expired_lists + levels; + __expire_timers(base, head); + } + + /* + * Zero base->expired_levels after processing all base->expired_lists + * to signal it's ready to get re-populated. Otherwise, we race with + * tick_find_expired() when base->lock is temporarily dropped in + * __expire_timers() + */ + base->expired_levels = 0; +} + +static int __collect_expired_timers(struct timer_base *base) { - unsigned long clk = base->clk; struct hlist_head *vec; - int i, levels = 0; + struct hlist_head *expired_list = base->expired_lists; + unsigned long clk; + int i; unsigned int idx; + /* + * expire_timers() must be called at least once before we can + * collect more timers. + */ + if (base->expired_levels) + return base->expired_levels; + + clk = base->clk; for (i = 0; i < LVL_DEPTH; i++) { idx = (clk & LVL_MASK) + i * LVL_SIZE; if (__test_and_clear_bit(idx, base->pending_map)) { vec = base->vectors + idx; - hlist_move_list(vec, heads++); - levels++; + hlist_move_list(vec, expired_list); + base->expired_levels++; + expired_list++; } /* Is it time to look at the next level? */ if (clk & LVL_CLK_MASK) @@ -1394,7 +1425,8 @@ static int __collect_expired_timers(struct timer_base *base, /* Shift clock for the next level granularity */ clk >>= LVL_CLK_SHIFT; } - return levels; + + return base->expired_levels; } #ifdef CONFIG_NO_HZ_COMMON @@ -1593,8 +1625,7 @@ void timer_clear_idle(void) base->is_idle = false; } -static int collect_expired_timers(struct timer_base *base, - struct hlist_head *heads) +static int collect_expired_timers(struct timer_base *base) { /* * NOHZ optimization. After a long idle sleep we need to forward the @@ -1615,16 +1646,47 @@ static int collect_expired_timers(struct timer_base *base, } base->clk = next; } - return __collect_expired_timers(base, heads); + return __collect_expired_timers(base); } #else -static inline int collect_expired_timers(struct timer_base *base, - struct hlist_head *heads) +static inline int collect_expired_timers(struct timer_base *base) { - return __collect_expired_timers(base, heads); + return __collect_expired_timers(base); } #endif +/* + * Increments timer_base to current jiffies or until first expired + * timer is found. Return number of expired levels. + */ +static int find_expired_timers(struct timer_base *base) +{ + unsigned long int end_clk = jiffies; + int expired_levels = 0; + + while (time_after_eq(end_clk, base->clk) && !expired_levels) { + expired_levels = collect_expired_timers(base); + base->clk++; + } + + return expired_levels; +} + +/* + * Called from CPU tick routine to collect expired timers up to current + * jiffies. Return number of expired levels. + */ +static int tick_find_expired(struct timer_base *base) +{ + int levels; + + raw_spin_lock(&base->lock); + levels = find_expired_timers(base); + raw_spin_unlock(&base->lock); + + return levels; +} + /* * Called from the timer interrupt handler to charge one tick to the current * process. user_tick is 1 if the tick is user time, 0 for system. @@ -1652,22 +1714,12 @@ void update_process_times(int user_tick) */ static inline void __run_timers(struct timer_base *base) { - struct hlist_head heads[LVL_DEPTH]; - int levels; - - if (!time_after_eq(jiffies, base->clk)) - return; - raw_spin_lock_irq(&base->lock); - while (time_after_eq(jiffies, base->clk)) { + do { + expire_timers(base); + } while (find_expired_timers(base)); - levels = collect_expired_timers(base, heads); - base->clk++; - - while (levels--) - expire_timers(base, heads + levels); - } raw_spin_unlock_irq(&base->lock); wakeup_timer_waiters(base); } @@ -1707,12 +1759,12 @@ void run_local_timers(void) hrtimer_run_queues(); /* Raise the softirq only if required. */ - if (time_before(jiffies, base->clk)) { + if (time_before(jiffies, base->clk) || !tick_find_expired(base)) { if (!IS_ENABLED(CONFIG_NO_HZ_COMMON)) return; /* CPU is awake, so check the deferrable base. */ base++; - if (time_before(jiffies, base->clk)) + if (time_before(jiffies, base->clk) || !tick_find_expired(base)) return; } raise_softirq(TIMER_SOFTIRQ); @@ -1887,6 +1939,7 @@ int timers_dead_cpu(unsigned int cpu) raw_spin_lock_nested(&old_base->lock, SINGLE_DEPTH_NESTING); BUG_ON(old_base->running_timer); + BUG_ON(old_base->expired_levels); for (i = 0; i < WHEEL_SIZE; i++) migrate_timer_list(new_base, old_base->vectors + i); -- 2.17.1