From patchwork Thu Jan 26 16:38:02 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Pablo Neira Ayuso X-Patchwork-Id: 720257 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 3v8SLp6L2rz9tB1 for ; Fri, 27 Jan 2017 03:39:46 +1100 (AEDT) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753436AbdAZQja (ORCPT ); Thu, 26 Jan 2017 11:39:30 -0500 Received: from mail.us.es ([193.147.175.20]:48618 "EHLO mail.us.es" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753629AbdAZQjY (ORCPT ); Thu, 26 Jan 2017 11:39:24 -0500 Received: from antivirus1-rhel7.int (unknown [192.168.2.11]) by mail.us.es (Postfix) with ESMTP id 6395480FB for ; Thu, 26 Jan 2017 17:38:34 +0100 (CET) Received: from antivirus1-rhel7.int (localhost [127.0.0.1]) by antivirus1-rhel7.int (Postfix) with ESMTP id 529D7FEFC6 for ; Thu, 26 Jan 2017 17:38:34 +0100 (CET) Received: by antivirus1-rhel7.int (Postfix, from userid 99) id 47922A7BEC; Thu, 26 Jan 2017 17:38:34 +0100 (CET) X-Spam-Checker-Version: SpamAssassin 3.4.1 (2015-04-28) on antivirus1-rhel7.int X-Spam-Level: X-Spam-Status: No, score=-107.2 required=7.5 tests=BAYES_50,SMTPAUTH_US, USER_IN_WHITELIST autolearn=disabled version=3.4.1 Received: from antivirus1-rhel7.int (localhost [127.0.0.1]) by antivirus1-rhel7.int (Postfix) with ESMTP id 0343A1B3025 for ; Thu, 26 Jan 2017 17:38:32 +0100 (CET) Received: from 192.168.1.13 (192.168.1.13) by antivirus1-rhel7.int (F-Secure/fsigk_smtp/540/antivirus1-rhel7.int); Thu, 26 Jan 2017 17:38:32 +0100 (CET) X-Virus-Status: clean(F-Secure/fsigk_smtp/540/antivirus1-rhel7.int) Received: (qmail 4246 invoked from network); 26 Jan 2017 17:38:31 +0100 Received: from 129.166.216.87.static.jazztel.es (HELO salvia.redhat.com) (pneira@us.es@87.216.166.129) by mail.us.es with SMTP; 26 Jan 2017 17:38:31 +0100 From: Pablo Neira Ayuso To: netfilter-devel@vger.kernel.org Cc: davem@davemloft.net, netdev@vger.kernel.org Subject: [PATCH 09/14] netfilter: conntrack: refine gc worker heuristics, redux Date: Thu, 26 Jan 2017 17:38:02 +0100 Message-Id: <1485448687-6072-10-git-send-email-pablo@netfilter.org> X-Mailer: git-send-email 2.1.4 In-Reply-To: <1485448687-6072-1-git-send-email-pablo@netfilter.org> References: <1485448687-6072-1-git-send-email-pablo@netfilter.org> X-Virus-Scanned: ClamAV using ClamSMTP Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org From: Florian Westphal This further refines the changes made to conntrack gc_worker in commit e0df8cae6c16 ("netfilter: conntrack: refine gc worker heuristics"). The main idea of that change was to reduce the scan interval when evictions take place. However, on the reporters' setup, there are 1-2 million conntrack entries in total and roughly 8k new (and closing) connections per second. In this case we'll always evict at least one entry per gc cycle and scan interval is always at 1 jiffy because of this test: } else if (expired_count) { gc_work->next_gc_run /= 2U; next_run = msecs_to_jiffies(1); being true almost all the time. Given we scan ~10k entries per run its clearly wrong to reduce interval based on nonzero eviction count, it will only waste cpu cycles since a vast majorities of conntracks are not timed out. Thus only look at the ratio (scanned entries vs. evicted entries) to make a decision on whether to reduce or not. Because evictor is supposed to only kick in when system turns idle after a busy period, pick a high ratio -- this makes it 50%. We thus keep the idea of increasing scan rate when its likely that table contains many expired entries. In order to not let timed-out entries hang around for too long (important when using event logging, in which case we want to timely destroy events), we now scan the full table within at most GC_MAX_SCAN_JIFFIES (16 seconds) even in worst-case scenario where all timed-out entries sit in same slot. I tested this with a vm under synflood (with sysctl net.netfilter.nf_conntrack_tcp_timeout_syn_recv=3). While flood is ongoing, interval now stays at its max rate (GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV -> 125ms). With feedback from Nicolas Dichtel. Reported-by: Denys Fedoryshchenko Cc: Nicolas Dichtel Fixes: b87a2f9199ea82eaadc ("netfilter: conntrack: add gc worker to remove timed-out entries") Signed-off-by: Florian Westphal Tested-by: Nicolas Dichtel Acked-by: Nicolas Dichtel Tested-by: Denys Fedoryshchenko Signed-off-by: Pablo Neira Ayuso --- net/netfilter/nf_conntrack_core.c | 39 ++++++++++++++++++++------------------- 1 file changed, 20 insertions(+), 19 deletions(-) diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c index 6feb5d370319..4e8083c5e01d 100644 --- a/net/netfilter/nf_conntrack_core.c +++ b/net/netfilter/nf_conntrack_core.c @@ -85,9 +85,11 @@ static __read_mostly DEFINE_SPINLOCK(nf_conntrack_locks_all_lock); static __read_mostly bool nf_conntrack_locks_all; /* every gc cycle scans at most 1/GC_MAX_BUCKETS_DIV part of table */ -#define GC_MAX_BUCKETS_DIV 64u -/* upper bound of scan intervals */ -#define GC_INTERVAL_MAX (2 * HZ) +#define GC_MAX_BUCKETS_DIV 128u +/* upper bound of full table scan */ +#define GC_MAX_SCAN_JIFFIES (16u * HZ) +/* desired ratio of entries found to be expired */ +#define GC_EVICT_RATIO 50u static struct conntrack_gc_work conntrack_gc_work; @@ -936,6 +938,7 @@ static noinline int early_drop(struct net *net, unsigned int _hash) static void gc_worker(struct work_struct *work) { + unsigned int min_interval = max(HZ / GC_MAX_BUCKETS_DIV, 1u); unsigned int i, goal, buckets = 0, expired_count = 0; struct conntrack_gc_work *gc_work; unsigned int ratio, scanned = 0; @@ -994,27 +997,25 @@ static void gc_worker(struct work_struct *work) * 1. Minimize time until we notice a stale entry * 2. Maximize scan intervals to not waste cycles * - * Normally, expired_count will be 0, this increases the next_run time - * to priorize 2) above. + * Normally, expire ratio will be close to 0. * - * As soon as a timed-out entry is found, move towards 1) and increase - * the scan frequency. - * In case we have lots of evictions next scan is done immediately. + * As soon as a sizeable fraction of the entries have expired + * increase scan frequency. */ ratio = scanned ? expired_count * 100 / scanned : 0; - if (ratio >= 90) { - gc_work->next_gc_run = 0; - next_run = 0; - } else if (expired_count) { - gc_work->next_gc_run /= 2U; - next_run = msecs_to_jiffies(1); + if (ratio > GC_EVICT_RATIO) { + gc_work->next_gc_run = min_interval; } else { - if (gc_work->next_gc_run < GC_INTERVAL_MAX) - gc_work->next_gc_run += msecs_to_jiffies(1); + unsigned int max = GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV; - next_run = gc_work->next_gc_run; + BUILD_BUG_ON((GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV) == 0); + + gc_work->next_gc_run += min_interval; + if (gc_work->next_gc_run > max) + gc_work->next_gc_run = max; } + next_run = gc_work->next_gc_run; gc_work->last_bucket = i; queue_delayed_work(system_long_wq, &gc_work->dwork, next_run); } @@ -1022,7 +1023,7 @@ static void gc_worker(struct work_struct *work) static void conntrack_gc_work_init(struct conntrack_gc_work *gc_work) { INIT_DELAYED_WORK(&gc_work->dwork, gc_worker); - gc_work->next_gc_run = GC_INTERVAL_MAX; + gc_work->next_gc_run = HZ; gc_work->exiting = false; } @@ -1914,7 +1915,7 @@ int nf_conntrack_init_start(void) nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); conntrack_gc_work_init(&conntrack_gc_work); - queue_delayed_work(system_long_wq, &conntrack_gc_work.dwork, GC_INTERVAL_MAX); + queue_delayed_work(system_long_wq, &conntrack_gc_work.dwork, HZ); return 0;