From patchwork Wed Jan 19 10:04:12 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Eric Dumazet X-Patchwork-Id: 12717357 X-Patchwork-Delegate: kuba@kernel.org 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 3940FC433EF for ; Wed, 19 Jan 2022 10:04:21 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1353433AbiASKEU (ORCPT ); Wed, 19 Jan 2022 05:04:20 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:40694 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1352594AbiASKEU (ORCPT ); Wed, 19 Jan 2022 05:04:20 -0500 Received: from mail-pf1-x435.google.com (mail-pf1-x435.google.com [IPv6:2607:f8b0:4864:20::435]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id C545EC061574 for ; Wed, 19 Jan 2022 02:04:19 -0800 (PST) Received: by mail-pf1-x435.google.com with SMTP id s15so2125156pfw.1 for ; Wed, 19 Jan 2022 02:04:19 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=b1bj2taFyxGUc5k665U6Q4VTbJD5tAHU3fwgwPzDJy4=; b=IVHtKWX1kgUL+9FzM12ftJgJVk0Fi9RX0X1z7BlIK8KrEbH+izsXgUObp3LRZVazpG HRbaQ8AtMEt6JshDxbSN+yGwMPZATuGXPtgdzkzInVJLjfNVu9xw3N13FONGkuO0YCBQ 60hx6Qt1PpNdZoc9K334o/Uz9XRujJEbRSARUv19PMqRI2RQd6x0vDClVjPZh/qFKqBn 7qFNLM6VoQUQ3dU+BgQ47Ls3JEo2eDweyk3yqBg6tF1Vy96oiWeNi+39BC0au3dt74SN QTpgwaNwxwyrvG3ncZbIP5rm1hiOgVdZUON83X4tK1Ka/lesM/T1kudS72STWegT94DV 0g7g== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=b1bj2taFyxGUc5k665U6Q4VTbJD5tAHU3fwgwPzDJy4=; b=qJOiCsYFoETdJdEu///cVR5rudnRwR/AXwC7ffVgHOqPse/Xl/P5jYJMPpgd1PqCfP BXIloTcQYjhN9G8UdAloJY9xwV915jsiqK0n7MVMNNn3KWteYvImBLetKCR4iyysC9Io lirkTlHhothm2hvBT1eeR9NmcN4mImWYYsU8H4gEhnJDq2CGoUAoC2vbAQd44DrGhX/W kvIUVRnnmt26XPhoz2IlMX1+RZYvSZ7QJXtnr+4xw+ZT+7pSuQ1uSj+89Iftjr4PfM3d Lyhma3ebU+GvdyIuthPB4v0B7YUKBSjcjWlDVmOymIW7CAmPwGcAetNv5Pl/52lv294F q9cw== X-Gm-Message-State: AOAM531T0OkMZkp55zUoed1E3GW10//TDLPFGhL8d49CUB8k/MhWmMr4 Jw4VipuHxAsVGj1ph50mknZ2Yqm3Z4KCag== X-Google-Smtp-Source: ABdhPJwQ8d3zEHfLp6Gxp6ZZrrqgiagZ4QzMaMI2D8nQy5+FPI1BVpwpH56Mb2lkK4SwbS/BFsMUMQ== X-Received: by 2002:a63:9f0a:: with SMTP id g10mr26211479pge.387.1642586659325; Wed, 19 Jan 2022 02:04:19 -0800 (PST) Received: from edumazet1.svl.corp.google.com ([2620:15c:2c4:201:a634:e286:f865:178a]) by smtp.gmail.com with ESMTPSA id a20sm11695323pfv.122.2022.01.19.02.04.18 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 19 Jan 2022 02:04:18 -0800 (PST) From: Eric Dumazet To: "David S . Miller" , Jakub Kicinski Cc: David Ahern , netdev , Eric Dumazet , Eric Dumazet Subject: [PATCH v2 net 1/2] ipv4: avoid quadratic behavior in netns dismantle Date: Wed, 19 Jan 2022 02:04:12 -0800 Message-Id: <20220119100413.4077866-2-eric.dumazet@gmail.com> X-Mailer: git-send-email 2.34.1.703.g22d0c6ccf7-goog In-Reply-To: <20220119100413.4077866-1-eric.dumazet@gmail.com> References: <20220119100413.4077866-1-eric.dumazet@gmail.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org From: Eric Dumazet net/ipv4/fib_semantics.c uses an hash table of 256 slots, keyed by device ifindexes: fib_info_devhash[DEVINDEX_HASHSIZE] Problem is that with network namespaces, devices tend to use the same ifindex. lo device for instance has a fixed ifindex of one, for all network namespaces. This means that hosts with thousands of netns spend a lot of time looking at some hash buckets with thousands of elements, notably at netns dismantle. Simply add a per netns perturbation (net_hash_mix()) to spread elements more uniformely. Also change fib_devindex_hashfn() to use more entropy. Fixes: aa79e66eee5d ("net: Make ifindex generation per-net namespace") Signed-off-by: Eric Dumazet Reviewed-by: David Ahern --- net/ipv4/fib_semantics.c | 36 +++++++++++++++++------------------- 1 file changed, 17 insertions(+), 19 deletions(-) diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c index 45619c005b8dddd7ccd5c7029efa4ed69b6ce1de..9813949da10493de36b9db797b6a5d94fd9bd3b1 100644 --- a/net/ipv4/fib_semantics.c +++ b/net/ipv4/fib_semantics.c @@ -29,6 +29,7 @@ #include #include #include +#include #include #include @@ -319,11 +320,15 @@ static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi) static inline unsigned int fib_devindex_hashfn(unsigned int val) { - unsigned int mask = DEVINDEX_HASHSIZE - 1; + return hash_32(val, DEVINDEX_HASHBITS); +} - return (val ^ - (val >> DEVINDEX_HASHBITS) ^ - (val >> (DEVINDEX_HASHBITS * 2))) & mask; +static struct hlist_head * +fib_info_devhash_bucket(const struct net_device *dev) +{ + u32 val = net_hash_mix(dev_net(dev)) ^ dev->ifindex; + + return &fib_info_devhash[fib_devindex_hashfn(val)]; } static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope, @@ -433,12 +438,11 @@ int ip_fib_check_default(__be32 gw, struct net_device *dev) { struct hlist_head *head; struct fib_nh *nh; - unsigned int hash; spin_lock(&fib_info_lock); - hash = fib_devindex_hashfn(dev->ifindex); - head = &fib_info_devhash[hash]; + head = fib_info_devhash_bucket(dev); + hlist_for_each_entry(nh, head, nh_hash) { if (nh->fib_nh_dev == dev && nh->fib_nh_gw4 == gw && @@ -1609,12 +1613,10 @@ struct fib_info *fib_create_info(struct fib_config *cfg, } else { change_nexthops(fi) { struct hlist_head *head; - unsigned int hash; if (!nexthop_nh->fib_nh_dev) continue; - hash = fib_devindex_hashfn(nexthop_nh->fib_nh_dev->ifindex); - head = &fib_info_devhash[hash]; + head = fib_info_devhash_bucket(nexthop_nh->fib_nh_dev); hlist_add_head(&nexthop_nh->nh_hash, head); } endfor_nexthops(fi) } @@ -1966,8 +1968,7 @@ void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig) void fib_sync_mtu(struct net_device *dev, u32 orig_mtu) { - unsigned int hash = fib_devindex_hashfn(dev->ifindex); - struct hlist_head *head = &fib_info_devhash[hash]; + struct hlist_head *head = fib_info_devhash_bucket(dev); struct fib_nh *nh; hlist_for_each_entry(nh, head, nh_hash) { @@ -1986,12 +1987,11 @@ void fib_sync_mtu(struct net_device *dev, u32 orig_mtu) */ int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force) { - int ret = 0; - int scope = RT_SCOPE_NOWHERE; + struct hlist_head *head = fib_info_devhash_bucket(dev); struct fib_info *prev_fi = NULL; - unsigned int hash = fib_devindex_hashfn(dev->ifindex); - struct hlist_head *head = &fib_info_devhash[hash]; + int scope = RT_SCOPE_NOWHERE; struct fib_nh *nh; + int ret = 0; if (force) scope = -1; @@ -2136,7 +2136,6 @@ static void fib_select_default(const struct flowi4 *flp, struct fib_result *res) int fib_sync_up(struct net_device *dev, unsigned char nh_flags) { struct fib_info *prev_fi; - unsigned int hash; struct hlist_head *head; struct fib_nh *nh; int ret; @@ -2152,8 +2151,7 @@ int fib_sync_up(struct net_device *dev, unsigned char nh_flags) } prev_fi = NULL; - hash = fib_devindex_hashfn(dev->ifindex); - head = &fib_info_devhash[hash]; + head = fib_info_devhash_bucket(dev); ret = 0; hlist_for_each_entry(nh, head, nh_hash) { From patchwork Wed Jan 19 10:04:13 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Eric Dumazet X-Patchwork-Id: 12717358 X-Patchwork-Delegate: kuba@kernel.org 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 1C4ADC433EF for ; Wed, 19 Jan 2022 10:04:25 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1353437AbiASKEY (ORCPT ); Wed, 19 Jan 2022 05:04:24 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:40708 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1352594AbiASKEW (ORCPT ); Wed, 19 Jan 2022 05:04:22 -0500 Received: from mail-pf1-x42a.google.com (mail-pf1-x42a.google.com [IPv6:2607:f8b0:4864:20::42a]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id AEC95C06161C for ; Wed, 19 Jan 2022 02:04:21 -0800 (PST) Received: by mail-pf1-x42a.google.com with SMTP id 128so2074189pfe.12 for ; Wed, 19 Jan 2022 02:04:21 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20210112; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=WnGH7bBzUjG7H82yVnKfdVHGADQN8CNYwcunX2fHdZM=; b=W3lthLt7KSjV6eGJoJZFraXD2mhnxHSUF5EpSGGP4ZvOwDZxTe7FPDF9lNwudvmfRE XgsaarnIDJuSJrex7VYazWGqu/6WMi6IKpCLprB6MoVf8R5ppUNSpo8eFpnPhai2ZrEa 9Pqp59CCZYX2yVraXYnJiSwpjvwW6+hP7rhpN1SN3mdqhRgD9vOsCfCmszHcO6okgQEp W6k35ZKE+Wu+377dqBjHqWWugK53txmCSUAXVpEMtjehDyji/lYCLpn7mEkNBKEL6FTk yzpYtpWL5gWUh+MpVV5R47Q5hhtC8G2Ez46IbzxA7epiK6Ei5mGTAjumN/lq9JwZWzRh lZuQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=WnGH7bBzUjG7H82yVnKfdVHGADQN8CNYwcunX2fHdZM=; b=TO4gvV2JZmO8Jf709m2d2R1Ejae4Wu8TwSFIHJNlsZ1wptEKILJT4wF6+dqULJry2A nqmS3o8zhHeVITUkijoWnxpNi4PwLWeK7ZoKoZlQNQnH0yQkvYE9OrdmJJsaVpBgD5vk rwGjJXvToDvFxErh6Le6o12w+b+eyqKgtMT9YO6G/eJ9gVDnTIu30SnRI9nv7knhkiab jAJ++5Goghp53a9w2QoI4roeVEWD7u3b2Wc5U1yWLSyXs+i8QDEyXC8JdLmzXZRcFLWQ bsX/OPAZVxAmFb0GubE8Yq+tNyqRY87HX6mrvtxwQqBaBKA1BQGAHgzWJCCbGn5vze4f jCXQ== X-Gm-Message-State: AOAM5309dabSqaycNliA3nlaiDGzo7DMKTzXH7AUFHxMc7pOIPO6adMd 8O5Ruu85yd3a0EnBRPrcYpg= X-Google-Smtp-Source: ABdhPJzDxZLuU7yfkb8PyTx0d7/wdOHGFDg0jkF5VsogjUt7fR3PrSW4Wu/cTazV8HkEsq88P71LmQ== X-Received: by 2002:a63:b144:: with SMTP id g4mr26854527pgp.571.1642586661309; Wed, 19 Jan 2022 02:04:21 -0800 (PST) Received: from edumazet1.svl.corp.google.com ([2620:15c:2c4:201:a634:e286:f865:178a]) by smtp.gmail.com with ESMTPSA id a20sm11695323pfv.122.2022.01.19.02.04.20 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 19 Jan 2022 02:04:21 -0800 (PST) From: Eric Dumazet To: "David S . Miller" , Jakub Kicinski Cc: David Ahern , netdev , Eric Dumazet , Eric Dumazet Subject: [PATCH v2 net 2/2] ipv4: add net_hash_mix() dispersion to fib_info_laddrhash keys Date: Wed, 19 Jan 2022 02:04:13 -0800 Message-Id: <20220119100413.4077866-3-eric.dumazet@gmail.com> X-Mailer: git-send-email 2.34.1.703.g22d0c6ccf7-goog In-Reply-To: <20220119100413.4077866-1-eric.dumazet@gmail.com> References: <20220119100413.4077866-1-eric.dumazet@gmail.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org From: Eric Dumazet net/ipv4/fib_semantics.c uses a hash table (fib_info_laddrhash) in which fib_sync_down_addr() can locate fib_info based on IPv4 local address. This hash table is resized based on total number of hashed fib_info, but the hash function is only using the local address. For hosts having many active network namespaces, all fib_info for loopback devices (IPv4 address 127.0.0.1) are hashed into a single bucket, making netns dismantles very slow. Signed-off-by: Eric Dumazet Reviewed-by: David Ahern --- net/ipv4/fib_semantics.c | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-) diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c index 9813949da10493de36b9db797b6a5d94fd9bd3b1..b4589861b84c6bc4daa7149e078ad63749c7622f 100644 --- a/net/ipv4/fib_semantics.c +++ b/net/ipv4/fib_semantics.c @@ -52,6 +52,7 @@ static DEFINE_SPINLOCK(fib_info_lock); static struct hlist_head *fib_info_hash; static struct hlist_head *fib_info_laddrhash; static unsigned int fib_info_hash_size; +static unsigned int fib_info_hash_bits; static unsigned int fib_info_cnt; #define DEVINDEX_HASHBITS 8 @@ -1247,13 +1248,13 @@ int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope, return err; } -static inline unsigned int fib_laddr_hashfn(__be32 val) +static struct hlist_head * +fib_info_laddrhash_bucket(const struct net *net, __be32 val) { - unsigned int mask = (fib_info_hash_size - 1); + u32 slot = hash_32(net_hash_mix(net) ^ (__force u32)val, + fib_info_hash_bits); - return ((__force u32)val ^ - ((__force u32)val >> 7) ^ - ((__force u32)val >> 14)) & mask; + return &fib_info_laddrhash[slot]; } static struct hlist_head *fib_info_hash_alloc(int bytes) @@ -1289,6 +1290,7 @@ static void fib_info_hash_move(struct hlist_head *new_info_hash, old_info_hash = fib_info_hash; old_laddrhash = fib_info_laddrhash; fib_info_hash_size = new_size; + fib_info_hash_bits = ilog2(new_size); for (i = 0; i < old_size; i++) { struct hlist_head *head = &fib_info_hash[i]; @@ -1306,21 +1308,20 @@ static void fib_info_hash_move(struct hlist_head *new_info_hash, } fib_info_hash = new_info_hash; + fib_info_laddrhash = new_laddrhash; for (i = 0; i < old_size; i++) { - struct hlist_head *lhead = &fib_info_laddrhash[i]; + struct hlist_head *lhead = &old_laddrhash[i]; struct hlist_node *n; struct fib_info *fi; hlist_for_each_entry_safe(fi, n, lhead, fib_lhash) { struct hlist_head *ldest; - unsigned int new_hash; - new_hash = fib_laddr_hashfn(fi->fib_prefsrc); - ldest = &new_laddrhash[new_hash]; + ldest = fib_info_laddrhash_bucket(fi->fib_net, + fi->fib_prefsrc); hlist_add_head(&fi->fib_lhash, ldest); } } - fib_info_laddrhash = new_laddrhash; spin_unlock_bh(&fib_info_lock); @@ -1605,7 +1606,7 @@ struct fib_info *fib_create_info(struct fib_config *cfg, if (fi->fib_prefsrc) { struct hlist_head *head; - head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)]; + head = fib_info_laddrhash_bucket(net, fi->fib_prefsrc); hlist_add_head(&fi->fib_lhash, head); } if (fi->nh) { @@ -1877,16 +1878,16 @@ int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event, */ int fib_sync_down_addr(struct net_device *dev, __be32 local) { - int ret = 0; - unsigned int hash = fib_laddr_hashfn(local); - struct hlist_head *head = &fib_info_laddrhash[hash]; int tb_id = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN; struct net *net = dev_net(dev); + struct hlist_head *head; struct fib_info *fi; + int ret = 0; if (!fib_info_laddrhash || local == 0) return 0; + head = fib_info_laddrhash_bucket(net, local); hlist_for_each_entry(fi, head, fib_lhash) { if (!net_eq(fi->fib_net, net) || fi->fib_tb_id != tb_id)