| Message ID | 20150514135620.14062.1034.stgit@buzz |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | David Miller |
| Headers | show |
On Thu, May 14, 2015 at 6:56 AM, Konstantin Khlebnikov <khlebnikov@yandex-team.ru> 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 <khlebnikov@yandex-team.ru> > --- > 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 <master-dev> <slave-dev> 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? > + 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); > -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 20.05.2015 02:59, Mahesh Bandewar wrote: > On Thu, May 14, 2015 at 6:56 AM, Konstantin Khlebnikov > <khlebnikov@yandex-team.ru> 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 <khlebnikov@yandex-team.ru> >> --- >> 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 <master-dev> <slave-dev> 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); >>
On Thu, May 21, 2015, at 13:38, Konstantin Khlebnikov wrote: > On 20.05.2015 02:59, Mahesh Bandewar wrote: > > On Thu, May 14, 2015 at 6:56 AM, Konstantin Khlebnikov > > <khlebnikov@yandex-team.ru> 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 <khlebnikov@yandex-team.ru> > >> --- > >> 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 <master-dev> <slave-dev> 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. I hope that stable ipv6 privacy addresses will be used in future and old eui-48 based LL addresses will just disappear. That said, I would be even fine with a RNG generated dev_id, because we cannot really ensure uniqueness. Stable privacy addresses even use DAD retry counter to regenerate a new LL address in case the address becomes IFA_F_DADFAILED. Bye, Hannes -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
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 <master-dev> <slave-dev> 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). + + 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);
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 <khlebnikov@yandex-team.ru> --- 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(-) -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html