From patchwork Sat Jan 9 17:26:11 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Vladimir Oltean X-Patchwork-Id: 12008557 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 X-Spam-Level: X-Spam-Status: No, score=-15.7 required=3.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,FREEMAIL_FORGED_FROMDOMAIN,FREEMAIL_FROM, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,URIBL_BLOCKED,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id C27A9C433E0 for ; Sat, 9 Jan 2021 17:28:19 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 960AB23A7C for ; Sat, 9 Jan 2021 17:28:19 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726260AbhAIR2G (ORCPT ); Sat, 9 Jan 2021 12:28:06 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:55132 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725926AbhAIR2F (ORCPT ); Sat, 9 Jan 2021 12:28:05 -0500 Received: from mail-ej1-x632.google.com (mail-ej1-x632.google.com [IPv6:2a00:1450:4864:20::632]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 34A68C0617A2 for ; Sat, 9 Jan 2021 09:27:25 -0800 (PST) Received: by mail-ej1-x632.google.com with SMTP id q22so18901045eja.2 for ; Sat, 09 Jan 2021 09:27:25 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=/lWP95iyU9Sl+LT7D1Nh8pxkQcK19ngJ2NDvEv/R65E=; b=bdNQhJszjNT9Jv1lbad6SROnAHd8dd8xA6X3PflVEgSBy4kJ7/1FRNnjjKUBA2K/TL ORDzkacHn98nyc7vFH7VS6iWwJLSFiGVoIrJgX/EC+97FBIceRGQm/vHQZ5exs6M5Q96 yoVKXR8LbjWjO6+PyZ/SMfj50Hb/MjYIELpMX54FQTLWTOhF4WtJ8pEAtXhTG8QEPpch Gn3wg/BL+fROEudUN6FIc8iUUvS50XvsXib5m2+Eyp8tvCjL6chi9k/Op9FprM7pEhtO Gn1Je8lsUxyApAGaWG7kDRHuVqLiZ+0D8z6fF1GwR7bw8MC9TL746wZ8SqFOf1YxmLoH CdSw== 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:in-reply-to :references:mime-version:content-transfer-encoding; bh=/lWP95iyU9Sl+LT7D1Nh8pxkQcK19ngJ2NDvEv/R65E=; b=TW8gQq+DjZHjXop3KsNOjAQayQDLzIzFsE6lWtUfc1rc1DoGIENaywj9boWR+MiFwR diee3/D8gBmYWYjae1qLtKoFbjT9y+my7UN8jgqob36Egsffs4uXhtyqAvpLnDYoUe5q k9BNk55YlfZNz2MvoTapON7QkkkBaVBlJAnjd0ei89Wn31VhguYQTruwHkU1B2TIhhmv sbKX2n9ln7QYqXlMlov1PD9/Hr8UriGCaIolUbMW10esXHOECRijCzFZjZ6D2TKY/8h8 Vfb5Am/r7vAbLySimgizhTrbnwcxjpg/yMJJo7PlbjM9f6uDnuwg5N4md8dta7MYT1RZ yxbw== X-Gm-Message-State: AOAM532jbuBuNWWRd9+y2+E9fcjc6/QSyhjGHNS3VYleYcOB7s4RifoB xnkwF4FiaYZ/mPY8oMi9Ukc= X-Google-Smtp-Source: ABdhPJypmDLzuxCOqlVi6pBqSjKyNovLFhE6UADQwSSJjfNQIHi+yzwdoIDXCVFUaHxjCQCQ3YJHtg== X-Received: by 2002:a17:906:168f:: with SMTP id s15mr6026963ejd.180.1610213243852; Sat, 09 Jan 2021 09:27:23 -0800 (PST) Received: from localhost.localdomain (5-12-227-87.residential.rdsnet.ro. [5.12.227.87]) by smtp.gmail.com with ESMTPSA id h16sm4776714eji.110.2021.01.09.09.27.22 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sat, 09 Jan 2021 09:27:23 -0800 (PST) From: Vladimir Oltean To: "David S . Miller" , Jakub Kicinski Cc: netdev@vger.kernel.org, Andrew Lunn , Florian Fainelli , Cong Wang , Stephen Hemminger , Eric Dumazet , George McCollister , Oleksij Rempel , Jay Vosburgh , Veaceslav Falico , Andy Gospodarek , Arnd Bergmann , Taehee Yoo , Jiri Pirko , Florian Westphal , Nikolay Aleksandrov , Pravin B Shelar , Sridhar Samudrala , Saeed Mahameed Subject: [PATCH v6 net-next 02/15] net: introduce a mutex for the netns interface lists Date: Sat, 9 Jan 2021 19:26:11 +0200 Message-Id: <20210109172624.2028156-3-olteanv@gmail.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20210109172624.2028156-1-olteanv@gmail.com> References: <20210109172624.2028156-1-olteanv@gmail.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org X-Patchwork-Delegate: kuba@kernel.org From: Vladimir Oltean Currently, any writer that wants to alter the lists of network interfaces (either the plain list net->dev_base_head, or the hash tables net->dev_index_head and net->dev_name_head) can keep other writers at bay using the RTNL mutex. However, the RTNL mutex has become a very contended resource over the years, so there is a movement to do finer grained locking. This patch adds one more way for writers to the network interface lists to serialize themselves. We assume that all writers to the network interface lists are easily identifiable because the write side of dev_base_lock also needs to be held (note that some instances of that were deliberately skipped, since they only dealt with protecting the operational state of the netdev). Holding the RTNL mutex is now optional for new code that alters the lists, since all relevant writers were made to also hold the new lock. Signed-off-by: Vladimir Oltean --- Changes in v6: None. Changes in v5: None. Changes in v4: None. Changes in v3: None. include/linux/netdevice.h | 10 +++++++++ include/net/net_namespace.h | 6 +++++ net/core/dev.c | 44 +++++++++++++++++++++++++------------ 3 files changed, 46 insertions(+), 14 deletions(-) diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index 1ec3ac5d5bbf..8aae2386bd37 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -4376,6 +4376,16 @@ static inline void netif_addr_unlock_bh(struct net_device *dev) spin_unlock_bh(&dev->addr_list_lock); } +static inline void netif_lists_lock(struct net *net) +{ + mutex_lock(&net->netif_lists_lock); +} + +static inline void netif_lists_unlock(struct net *net) +{ + mutex_unlock(&net->netif_lists_lock); +} + /* * dev_addrs walker. Should be used only for read access. Call with * rcu_read_lock held. diff --git a/include/net/net_namespace.h b/include/net/net_namespace.h index 29567875f428..cac64c3c7ce0 100644 --- a/include/net/net_namespace.h +++ b/include/net/net_namespace.h @@ -183,6 +183,12 @@ struct net { struct sock *crypto_nlsk; #endif struct sock *diag_nlsk; + + /* Serializes writers to @dev_base_head, @dev_name_head and + * @dev_index_head. It does _not_ protect the netif adjacency lists + * (struct net_device::adj_list). + */ + struct mutex netif_lists_lock; } __randomize_layout; #include diff --git a/net/core/dev.c b/net/core/dev.c index 8e02240bb11c..53c12f92025c 100644 --- a/net/core/dev.c +++ b/net/core/dev.c @@ -175,13 +175,16 @@ static struct napi_struct *napi_by_id(unsigned int napi_id); * * Pure readers should hold rcu_read_lock() which should protect them against * concurrent changes to the interface lists made by the writers. Pure writers - * must serialize by holding the RTNL mutex while they loop through the list - * and make changes to it. + * must serialize by holding the @net->netif_lists_lock mutex while they loop + * through the list and make changes to it. + * + * It is possible to hold the RTNL mutex for serializing the writers too, but + * this should be avoided in new code due to lock contention. * * It is also possible to hold the global rwlock_t @dev_base_lock for * protection (holding its read side as an alternative to rcu_read_lock, and - * its write side as an alternative to the RTNL mutex), however this should not - * be done in new code, since it is deprecated and pending removal. + * its write side as an alternative to @net->netif_lists_lock), however this + * should not be done in new code, since it is deprecated and pending removal. * * One other role of @dev_base_lock is to protect against changes in the * operational state of a network interface. @@ -360,12 +363,14 @@ static void list_netdevice(struct net_device *dev) ASSERT_RTNL(); + netif_lists_lock(net); write_lock_bh(&dev_base_lock); list_add_tail_rcu(&dev->dev_list, &net->dev_base_head); netdev_name_node_add(net, dev->name_node); hlist_add_head_rcu(&dev->index_hlist, dev_index_hash(net, dev->ifindex)); write_unlock_bh(&dev_base_lock); + netif_lists_unlock(net); dev_base_seq_inc(net); } @@ -375,16 +380,20 @@ static void list_netdevice(struct net_device *dev) */ static void unlist_netdevice(struct net_device *dev) { + struct net *net = dev_net(dev); + ASSERT_RTNL(); /* Unlink dev from the device chain */ + netif_lists_lock(net); write_lock_bh(&dev_base_lock); list_del_rcu(&dev->dev_list); netdev_name_node_del(dev->name_node); hlist_del_rcu(&dev->index_hlist); write_unlock_bh(&dev_base_lock); + netif_lists_unlock(net); - dev_base_seq_inc(dev_net(dev)); + dev_base_seq_inc(net); } /* @@ -850,11 +859,11 @@ EXPORT_SYMBOL_GPL(dev_fill_metadata_dst); * @net: the applicable net namespace * @name: name to find * - * Find an interface by name. Must be called under RTNL semaphore - * or @dev_base_lock. If the name is found a pointer to the device - * is returned. If the name is not found then %NULL is returned. The - * reference counters are not incremented so the caller must be - * careful with locks. + * Find an interface by name. Must be called under RTNL semaphore, + * @net->netif_lists_lock or @dev_base_lock. If the name is found, + * a pointer to the device is returned. If the name is not found then + * %NULL is returned. The reference counters are not incremented so the + * caller must be careful with locks. */ struct net_device *__dev_get_by_name(struct net *net, const char *name) @@ -920,8 +929,8 @@ EXPORT_SYMBOL(dev_get_by_name); * Search for an interface by index. Returns %NULL if the device * is not found or a pointer to the device. The device has not * had its reference counter increased so the caller must be careful - * about locking. The caller must hold either the RTNL semaphore - * or @dev_base_lock. + * about locking. The caller must hold either the RTNL semaphore, + * @net->netif_lists_lock or @dev_base_lock. */ struct net_device *__dev_get_by_index(struct net *net, int ifindex) @@ -1330,15 +1339,19 @@ int dev_change_name(struct net_device *dev, const char *newname) netdev_adjacent_rename_links(dev, oldname); + netif_lists_lock(net); write_lock_bh(&dev_base_lock); netdev_name_node_del(dev->name_node); write_unlock_bh(&dev_base_lock); + netif_lists_unlock(net); synchronize_rcu(); + netif_lists_lock(net); write_lock_bh(&dev_base_lock); netdev_name_node_add(net, dev->name_node); write_unlock_bh(&dev_base_lock); + netif_lists_unlock(net); ret = call_netdevice_notifiers(NETDEV_CHANGENAME, dev); ret = notifier_to_errno(ret); @@ -9415,8 +9428,9 @@ int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack, * @net: the applicable net namespace * * Returns a suitable unique value for a new device interface - * number. The caller must hold the rtnl semaphore or the - * dev_base_lock to be sure it remains unique. + * number. + * The caller must hold the rtnl semaphore, @net->netif_lists_lock or the + * @dev_base_lock to be sure it remains unique. */ static int dev_new_index(struct net *net) { @@ -10999,6 +11013,8 @@ static int __net_init netdev_init(struct net *net) if (net->dev_index_head == NULL) goto err_idx; + mutex_init(&net->netif_lists_lock); + RAW_INIT_NOTIFIER_HEAD(&net->netdev_chain); return 0;