From: Konstantin Khlebnikov Subject: Re: [PATCH 3/3] ipvlan: set dev_id for l2 ports to generate unique IPv6 addresses Date: Thu, 21 May 2015 14:38:20 +0300 Message-ID: <555DC3AC.5020604@yandex-team.ru> References: <20150514134657.14062.87579.stgit@buzz> <20150514135620.14062.1034.stgit@buzz> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Cc: linux-netdev , "David S. Miller" , Jiri Benc , Hannes Frederic Sowa To: Mahesh Bandewar Return-path: Received: from forward-corp1m.cmail.yandex.net ([5.255.216.100]:37260 "EHLO forward-corp1m.cmail.yandex.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751765AbbEULi0 (ORCPT ); Thu, 21 May 2015 07:38:26 -0400 In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On 20.05.2015 02:59, Mahesh Bandewar wrote: > On Thu, May 14, 2015 at 6:56 AM, Konstantin Khlebnikov > wrote: >> All ipvlan ports use one MAC address, this way ipv6 RA tries to assign >> one ipv6 address to all of them. This patch assigns unique dev_id to each >> ipvlan port. This field is used instead of common FF-FE in Modified EUI-64. >> >> Signed-off-by: Konstantin Khlebnikov >> --- >> Documentation/networking/ipvlan.txt | 12 +++++++++++- >> drivers/net/ipvlan/ipvlan.h | 1 + >> drivers/net/ipvlan/ipvlan_main.c | 20 ++++++++++++++++++++ >> 3 files changed, 32 insertions(+), 1 deletion(-) >> >> diff --git a/Documentation/networking/ipvlan.txt b/Documentation/networking/ipvlan.txt >> index cf996394e466..cb0b777bce58 100644 >> --- a/Documentation/networking/ipvlan.txt >> +++ b/Documentation/networking/ipvlan.txt >> @@ -24,7 +24,7 @@ using IProute2/ip utility. >> >> ip link add link type ipvlan mode { l2 | L3 } >> >> - e.g. ip link add link ipvl0 eth0 type ipvlan mode l2 >> + e.g. ip link add link eth0 ipvl0 type ipvlan mode l2 >> >> >> 4. Operating modes: >> @@ -41,6 +41,15 @@ slave device and packets are switched and queued to the master device to send >> out. In this mode the slaves will RX/TX multicast and broadcast (if applicable) >> as well. >> >> + In L2 mode slave devices receive Router Advertisements from the network >> +and perform autoconfiguration as well as master device. Each port has unique >> +16-bit device id which is used for filling octets 4-5 of Modified EUI-64. >> +That gives 65533 addresses (FF-FE used by master, FF-FF/00-00 reserved/not used). >> + > This is nice, thanks for fixing this! However how is "unique" id > guaranteed? Especially when multiple virtual drivers are stacked? Not > necessarily all of them may use the dev_id, but to avoid any possible > collision, shouldn't the device hierarchy (especially lower_dev) be > traversed before settling on the initial value? Well, uniqueness isn't guaranteed but that should work in most cases. ipv6 anyway checks for duplicate addresses after configuration. As I see creation of ipvlan on ipvlan just creates slave at original master device, so this will work as expected. And ipvlan cannot share physical device with bonding/bridge/macvlan, so I don't see how to stack more than one layer of ipvlans accidentally. > >> + Also lower half of IPv6 address could be set as interface token: >> + >> + ip token set ::aaaa:bbbb:cccc:dddd dev ipvl0 >> + >> 4.2 L3 mode: >> In this mode TX processing upto L3 happens on the stack instance attached >> to the slave device and packets are switched to the stack instance of the >> @@ -105,3 +114,4 @@ namespace where L2 on the slave could be changed / misused. >> (4) ip -4 addr add 127.0.0.1 dev lo >> (5) ip -4 addr add $IPADDR dev ipvl1 >> (6) ip -4 route add default via $ROUTER dev ipvl1 >> + >> diff --git a/drivers/net/ipvlan/ipvlan.h b/drivers/net/ipvlan/ipvlan.h >> index 54549a6223dd..1ebab84e7a0e 100644 >> --- a/drivers/net/ipvlan/ipvlan.h >> +++ b/drivers/net/ipvlan/ipvlan.h >> @@ -95,6 +95,7 @@ struct ipvl_port { >> struct rcu_head rcu; >> int count; >> u16 mode; >> + struct ida ida; >> }; >> >> static inline struct ipvl_port *ipvlan_port_get_rcu(const struct net_device *d) >> diff --git a/drivers/net/ipvlan/ipvlan_main.c b/drivers/net/ipvlan/ipvlan_main.c >> index 0cafd3e6f02d..dee0e8441150 100644 >> --- a/drivers/net/ipvlan/ipvlan_main.c >> +++ b/drivers/net/ipvlan/ipvlan_main.c >> @@ -53,6 +53,7 @@ static int ipvlan_port_create(struct net_device *dev) >> INIT_LIST_HEAD(&port->ipvlans); >> for (idx = 0; idx < IPVLAN_HASH_SIZE; idx++) >> INIT_HLIST_HEAD(&port->hlhead[idx]); >> + ida_init(&port->ida); >> >> err = netdev_rx_handler_register(dev, ipvlan_handle_frame, port); >> if (err) >> @@ -72,6 +73,7 @@ static void ipvlan_port_destroy(struct net_device *dev) >> >> dev->priv_flags &= ~IFF_IPVLAN_MASTER; >> netdev_rx_handler_unregister(dev); >> + ida_destroy(&port->ida); >> kfree_rcu(port, rcu); >> } >> >> @@ -484,6 +486,18 @@ static int ipvlan_link_new(struct net *src_net, struct net_device *dev, >> */ >> memcpy(dev->dev_addr, phy_dev->dev_addr, ETH_ALEN); >> >> + if (port->mode == IPVLAN_MODE_L2) { >> + /* >> + * IPv6 addrconf uses it to produce unique addresses, >> + * see function addrconf_ifid_eui48. >> + */ >> + err = ida_simple_get(&port->ida, 1, 0xFFFE, GFP_KERNEL); >> + if (err > 0) >> + dev->dev_id = err; >> + else if (err != -ENOSPC) >> + goto ipvlan_destroy_port; >> + } >> + >> dev->priv_flags |= IFF_IPVLAN_SLAVE; >> >> port->count += 1; >> @@ -518,6 +532,12 @@ static void ipvlan_link_delete(struct net_device *dev, struct list_head *head) >> list_del(&addr->anode); >> } >> } >> + >> + if (dev->dev_id) { >> + ida_simple_remove(&ipvlan->port->ida, dev->dev_id); >> + dev->dev_id = 0; >> + } >> + >> list_del_rcu(&ipvlan->pnode); >> unregister_netdevice_queue(dev, head); >> netdev_upper_dev_unlink(ipvlan->phy_dev, dev); >> -- Konstantin