X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fvport-netdev.c;h=4e7342c10bb090901ed07588fd88360366392706;hb=add037c6b692db9c66dd0c16dff2d4b1b46c7499;hp=05d7696faa0ee59e72fa6389ffd4ae176b2bb47d;hpb=9b80f761bed9a32c1b0eb22ee3361966057ea973;p=sliver-openvswitch.git diff --git a/datapath/vport-netdev.c b/datapath/vport-netdev.c index 05d7696fa..4e7342c10 100644 --- a/datapath/vport-netdev.c +++ b/datapath/vport-netdev.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2007-2012 Nicira Networks. + * Copyright (c) 2007-2012 Nicira, Inc. * * This program is free software; you can redistribute it and/or * modify it under the terms of version 2 of the GNU General Public @@ -45,6 +45,12 @@ MODULE_PARM_DESC(vlan_tso, "Enable TSO for VLAN packets"); #define vlan_tso true #endif +#ifdef HAVE_RHEL_OVS_HOOK +static atomic_t nr_bridges = ATOMIC_INIT(0); + +extern struct sk_buff *(*openvswitch_handle_frame_hook)(struct sk_buff *skb); +#endif + static void netdev_port_receive(struct vport *vport, struct sk_buff *skb); #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39) @@ -63,7 +69,8 @@ static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb) return RX_HANDLER_CONSUMED; } -#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) +#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \ + defined HAVE_RHEL_OVS_HOOK /* Called with rcu_read_lock and bottom-halves disabled. */ static struct sk_buff *netdev_frame_hook(struct sk_buff *skb) { @@ -105,21 +112,32 @@ static int netdev_frame_hook(struct net_bridge_port *p, struct sk_buff **pskb) #error #endif -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \ + defined HAVE_RHEL_OVS_HOOK static int netdev_init(void) { return 0; } static void netdev_exit(void) { } #else -static int netdev_init(void) +static int port_count; + +static void netdev_init(void) { + port_count++; + if (port_count > 1) + return; + /* Hook into callback used by the bridge to intercept packets. * Parasites we are. */ br_handle_frame_hook = netdev_frame_hook; - return 0; + return; } static void netdev_exit(void) { + port_count--; + if (port_count > 0) + return; + br_handle_frame_hook = NULL; } #endif @@ -152,19 +170,32 @@ static struct vport *netdev_create(const struct vport_parms *parms) goto error_put; } + rtnl_lock(); +#ifdef HAVE_RHEL_OVS_HOOK + rcu_assign_pointer(netdev_vport->dev->ax25_ptr, vport); + atomic_inc(&nr_bridges); + rcu_assign_pointer(openvswitch_handle_frame_hook, netdev_frame_hook); +#else err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook, vport); if (err) - goto error_put; + goto error_unlock; +#endif dev_set_promiscuity(netdev_vport->dev, 1); #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24) dev_disable_lro(netdev_vport->dev); #endif netdev_vport->dev->priv_flags |= IFF_OVS_DATAPATH; + rtnl_unlock(); + netdev_init(); return vport; +#ifndef HAVE_RHEL_OVS_HOOK +error_unlock: +#endif + rtnl_unlock(); error_put: dev_put(netdev_vport->dev); error_free_vport: @@ -173,29 +204,33 @@ error: return ERR_PTR(err); } -static void netdev_destroy(struct vport *vport) +static void free_port_rcu(struct rcu_head *rcu) { - struct netdev_vport *netdev_vport = netdev_vport_priv(vport); + struct netdev_vport *netdev_vport = container_of(rcu, + struct netdev_vport, rcu); - netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH; - netdev_rx_handler_unregister(netdev_vport->dev); - dev_set_promiscuity(netdev_vport->dev, -1); - - synchronize_rcu(); +#ifdef HAVE_RHEL_OVS_HOOK + rcu_assign_pointer(netdev_vport->dev->ax25_ptr, NULL); + if (atomic_dec_and_test(&nr_bridges)) + rcu_assign_pointer(openvswitch_handle_frame_hook, NULL); +#endif dev_put(netdev_vport->dev); - ovs_vport_free(vport); + ovs_vport_free(vport_from_priv(netdev_vport)); } -int ovs_netdev_set_addr(struct vport *vport, const unsigned char *addr) +static void netdev_destroy(struct vport *vport) { struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - struct sockaddr sa; - sa.sa_family = ARPHRD_ETHER; - memcpy(sa.sa_data, addr, ETH_ALEN); + netdev_exit(); + rtnl_lock(); + netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH; + netdev_rx_handler_unregister(netdev_vport->dev); + dev_set_promiscuity(netdev_vport->dev, -1); + rtnl_unlock(); - return dev_set_mac_address(netdev_vport->dev, &sa); + call_rcu(&netdev_vport->rcu, free_port_rcu); } const char *ovs_netdev_get_name(const struct vport *vport) @@ -204,55 +239,14 @@ const char *ovs_netdev_get_name(const struct vport *vport) return netdev_vport->dev->name; } -const unsigned char *ovs_netdev_get_addr(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->dev_addr; -} - -struct kobject *ovs_netdev_get_kobj(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return &netdev_vport->dev->NETDEV_DEV_MEMBER.kobj; -} - -unsigned ovs_netdev_get_dev_flags(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return dev_get_flags(netdev_vport->dev); -} - -int ovs_netdev_is_running(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netif_running(netdev_vport->dev); -} - -unsigned char ovs_netdev_get_operstate(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->operstate; -} - -int ovs_netdev_get_ifindex(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->ifindex; -} - -int ovs_netdev_get_mtu(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->mtu; -} - /* Must be called with rcu_read_lock. */ static void netdev_port_receive(struct vport *vport, struct sk_buff *skb) { - if (unlikely(!vport)) { - kfree_skb(skb); - return; - } + if (unlikely(!vport)) + goto error; + + if (unlikely(skb_warn_if_lro(skb))) + goto error; /* Make our own copy of the packet. Otherwise we will mangle the * packet for anyone who came before us (e.g. tcpdump via AF_PACKET). @@ -264,18 +258,21 @@ static void netdev_port_receive(struct vport *vport, struct sk_buff *skb) skb_push(skb, ETH_HLEN); - if (unlikely(compute_ip_summed(skb, false))) { - kfree_skb(skb); - return; - } + if (unlikely(compute_ip_summed(skb, false))) + goto error; + vlan_copy_skb_tci(skb); - ovs_vport_receive(vport, skb); + ovs_vport_receive(vport, skb, NULL); + return; + +error: + kfree_skb(skb); } -static unsigned packet_length(const struct sk_buff *skb) +static unsigned int packet_length(const struct sk_buff *skb) { - unsigned length = skb->len - ETH_HLEN; + unsigned int length = skb->len - ETH_HLEN; if (skb->protocol == htons(ETH_P_8021Q)) length -= VLAN_HLEN; @@ -303,15 +300,12 @@ static int netdev_send(struct vport *vport, struct sk_buff *skb) int len; if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) { - if (net_ratelimit()) - pr_warn("%s: dropped over-mtu packet: %d > %d\n", - ovs_dp_name(vport->dp), packet_length(skb), mtu); - goto error; + net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n", + netdev_vport->dev->name, + packet_length(skb), mtu); + goto drop; } - if (unlikely(skb_warn_if_lro(skb))) - goto error; - skb->dev = netdev_vport->dev; forward_ip_summed(skb, true); @@ -330,19 +324,15 @@ static int netdev_send(struct vport *vport, struct sk_buff *skb) nskb = skb_gso_segment(skb, features); if (!nskb) { if (unlikely(skb_cloned(skb) && - pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) { - kfree_skb(skb); - return 0; - } + pskb_expand_head(skb, 0, 0, GFP_ATOMIC))) + goto drop; skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY; goto tag; } - if (IS_ERR(nskb)) { - kfree_skb(skb); - return 0; - } + if (IS_ERR(nskb)) + goto drop; consume_skb(skb); skb = nskb; @@ -376,22 +366,26 @@ tag: return len; -error: +drop: kfree_skb(skb); - ovs_vport_record_error(vport, VPORT_E_TX_DROPPED); return 0; } /* Returns null if this device is not attached to a datapath. */ struct vport *ovs_netdev_get_vport(struct net_device *dev) { -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) || \ + defined HAVE_RHEL_OVS_HOOK #if IFF_OVS_DATAPATH != 0 if (likely(dev->priv_flags & IFF_OVS_DATAPATH)) #else if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook)) #endif +#ifdef HAVE_RHEL_OVS_HOOK + return (struct vport *)rcu_dereference_rtnl(dev->ax25_ptr); +#else return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data); +#endif else return NULL; #else @@ -401,29 +395,28 @@ struct vport *ovs_netdev_get_vport(struct net_device *dev) const struct vport_ops ovs_netdev_vport_ops = { .type = OVS_VPORT_TYPE_NETDEV, - .flags = VPORT_F_REQUIRED, - .init = netdev_init, - .exit = netdev_exit, .create = netdev_create, .destroy = netdev_destroy, - .set_addr = ovs_netdev_set_addr, .get_name = ovs_netdev_get_name, - .get_addr = ovs_netdev_get_addr, - .get_kobj = ovs_netdev_get_kobj, - .get_dev_flags = ovs_netdev_get_dev_flags, - .is_running = ovs_netdev_is_running, - .get_operstate = ovs_netdev_get_operstate, - .get_ifindex = ovs_netdev_get_ifindex, - .get_mtu = ovs_netdev_get_mtu, .send = netdev_send, }; -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \ + !defined HAVE_RHEL_OVS_HOOK /* - * In kernels earlier than 2.6.36, Open vSwitch cannot safely coexist with the - * Linux bridge module, because there is only a single bridge hook function and - * only a single br_port member in struct net_device, so this prevents loading - * both bridge and openvswitch at the same time. + * Enforces, mutual exclusion with the Linux bridge module, by declaring and + * exporting br_should_route_hook. Because the bridge module also exports the + * same symbol, the module loader will refuse to load both modules at the same + * time (e.g. "bridge: exports duplicate symbol br_should_route_hook (owned by + * openvswitch)"). + * + * Before Linux 2.6.36, Open vSwitch cannot safely coexist with the Linux + * bridge module, so openvswitch uses this macro in those versions. In + * Linux 2.6.36 and later, Open vSwitch can coexist with the bridge module. + * + * The use of "typeof" here avoids the need to track changes in the type of + * br_should_route_hook over various kernel versions. */ -BRIDGE_MUTUAL_EXCLUSION; +typeof(br_should_route_hook) br_should_route_hook; +EXPORT_SYMBOL(br_should_route_hook); #endif