X-Git-Url: http://git.cascardo.eti.br/?a=blobdiff_plain;f=datapath%2Fvport-netdev.c;h=a1df79b2b9dd770afca437134b8a5df525a377af;hb=8b8ef592521e32d0e32581bf39c5d2a5cd445977;hp=4c7f1cb43692354d66cc863f1491026b2d9fa167;hpb=f613a0d72c521ca3a4eeb2c29ac523f6fdf72667;p=cascardo%2Fovs.git diff --git a/datapath/vport-netdev.c b/datapath/vport-netdev.c index 4c7f1cb43..a1df79b2b 100644 --- a/datapath/vport-netdev.c +++ b/datapath/vport-netdev.c @@ -1,9 +1,19 @@ /* - * Copyright (c) 2010, 2011 Nicira Networks. - * Distributed under the terms of the GNU GPL version 2. + * Copyright (c) 2007-2015 Nicira, Inc. * - * Significant portions of this file may be copied from parts of the Linux - * kernel, by Linus Torvalds and others. + * This program is free software; you can redistribute it and/or + * modify it under the terms of version 2 of the GNU General Public + * License as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301, USA */ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt @@ -15,60 +25,76 @@ #include #include #include +#include -#include +#include +#include -#include "checksum.h" #include "datapath.h" -#include "vlan.h" +#include "gso.h" +#include "vport.h" #include "vport-internal_dev.h" #include "vport-netdev.h" -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \ - !defined(HAVE_VLAN_BUG_WORKAROUND) -#include +static struct vport_ops ovs_netdev_vport_ops; + +/* Must be called with rcu_read_lock. */ +void netdev_port_receive(struct sk_buff *skb, struct ip_tunnel_info *tun_info) +{ + struct vport *vport; + + vport = ovs_netdev_get_vport(skb->dev); + if (unlikely(!vport)) + goto error; + + if (unlikely(skb_warn_if_lro(skb))) + goto error; -static int vlan_tso __read_mostly = 0; -module_param(vlan_tso, int, 0644); -MODULE_PARM_DESC(vlan_tso, "Enable TSO for VLAN packets"); + /* 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). + */ + skb = skb_share_check(skb, GFP_ATOMIC); + if (unlikely(!skb)) + return; + + skb_push(skb, ETH_HLEN); + ovs_skb_postpush_rcsum(skb, skb->data, ETH_HLEN); + ovs_vport_receive(vport, skb, tun_info); + return; +error: + kfree_skb(skb); +} + +#ifndef HAVE_METADATA_DST +#define port_receive(skb) netdev_port_receive(skb, NULL) #else -#define vlan_tso true +#define port_receive(skb) netdev_port_receive(skb, skb_tunnel_info(skb)) #endif -static void netdev_port_receive(struct vport *vport, struct sk_buff *skb); - -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39) +#if defined HAVE_RX_HANDLER_PSKB /* 2.6.39 and above or backports */ /* Called with rcu_read_lock and bottom-halves disabled. */ static rx_handler_result_t netdev_frame_hook(struct sk_buff **pskb) { struct sk_buff *skb = *pskb; - struct vport *vport; if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) return RX_HANDLER_PASS; - vport = netdev_get_vport(skb->dev); - - netdev_port_receive(vport, skb); - + port_receive(skb); 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) { - struct vport *vport; - if (unlikely(skb->pkt_type == PACKET_LOOPBACK)) return skb; - vport = netdev_get_vport(skb->dev); - - netdev_port_receive(vport, skb); - + port_receive(skb); return NULL; } -#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) +#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,32) /* * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on * different set of devices!) @@ -77,310 +103,152 @@ static struct sk_buff *netdev_frame_hook(struct sk_buff *skb) static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p, struct sk_buff *skb) { - netdev_port_receive((struct vport *)p, skb); + port_receive(skb); return NULL; } -#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0) -/* - * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on - * different set of devices!) - */ -/* Called with rcu_read_lock and bottom-halves disabled. */ -static int netdev_frame_hook(struct net_bridge_port *p, struct sk_buff **pskb) -{ - netdev_port_receive((struct vport *)p, *pskb); - return 1; -} #else #error #endif -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) -static int netdev_init(void) { return 0; } -static void netdev_exit(void) { } -#else -static int netdev_init(void) +static struct net_device *get_dpdev(const struct datapath *dp) { - /* Hook into callback used by the bridge to intercept packets. - * Parasites we are. */ - br_handle_frame_hook = netdev_frame_hook; + struct vport *local; - return 0; + local = ovs_vport_ovsl(dp, OVSP_LOCAL); + BUG_ON(!local); + return local->dev; } -static void netdev_exit(void) +struct vport *ovs_netdev_link(struct vport *vport, const char *name) { - br_handle_frame_hook = NULL; -} -#endif - -static struct vport *netdev_create(const struct vport_parms *parms) -{ - struct vport *vport; - struct netdev_vport *netdev_vport; int err; - vport = vport_alloc(sizeof(struct netdev_vport), &netdev_vport_ops, parms); - if (IS_ERR(vport)) { - err = PTR_ERR(vport); - goto error; - } - - netdev_vport = netdev_vport_priv(vport); - - netdev_vport->dev = dev_get_by_name(&init_net, parms->name); - if (!netdev_vport->dev) { + vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), name); + if (!vport->dev) { err = -ENODEV; goto error_free_vport; } - if (netdev_vport->dev->flags & IFF_LOOPBACK || - netdev_vport->dev->type != ARPHRD_ETHER || - is_internal_dev(netdev_vport->dev)) { + if (vport->dev->flags & IFF_LOOPBACK || + vport->dev->type != ARPHRD_ETHER || + ovs_is_internal_dev(vport->dev)) { err = -EINVAL; goto error_put; } - err = netdev_rx_handler_register(netdev_vport->dev, netdev_frame_hook, + rtnl_lock(); + err = netdev_master_upper_dev_link(vport->dev, + get_dpdev(vport->dp)); + if (err) + goto error_unlock; + + err = netdev_rx_handler_register(vport->dev, netdev_frame_hook, vport); if (err) - goto error_put; + goto error_master_upper_dev_unlink; - 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; + dev_disable_lro(vport->dev); + dev_set_promiscuity(vport->dev, 1); + vport->dev->priv_flags |= IFF_OVS_DATAPATH; + rtnl_unlock(); return vport; +error_master_upper_dev_unlink: + netdev_upper_dev_unlink(vport->dev, get_dpdev(vport->dp)); +error_unlock: + rtnl_unlock(); error_put: - dev_put(netdev_vport->dev); + dev_put(vport->dev); error_free_vport: - vport_free(vport); -error: + ovs_vport_free(vport); return ERR_PTR(err); } +EXPORT_SYMBOL_GPL(ovs_netdev_link); -static void netdev_destroy(struct vport *vport) -{ - struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - - netdev_vport->dev->priv_flags &= ~IFF_OVS_DATAPATH; - netdev_rx_handler_unregister(netdev_vport->dev); - dev_set_promiscuity(netdev_vport->dev, -1); - - synchronize_rcu(); - - dev_put(netdev_vport->dev); - vport_free(vport); -} - -int netdev_set_addr(struct vport *vport, const unsigned char *addr) -{ - struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - struct sockaddr sa; - - sa.sa_family = ARPHRD_ETHER; - memcpy(sa.sa_data, addr, ETH_ALEN); - - return dev_set_mac_address(netdev_vport->dev, &sa); -} - -const char *netdev_get_name(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->name; -} - -const unsigned char *netdev_get_addr(const struct vport *vport) +static struct vport *netdev_create(const struct vport_parms *parms) { - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->dev_addr; -} + struct vport *vport; -struct kobject *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; -} + vport = ovs_vport_alloc(0, &ovs_netdev_vport_ops, parms); + if (IS_ERR(vport)) + return vport; -unsigned 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); + return ovs_netdev_link(vport, parms->name); } -int netdev_is_running(const struct vport *vport) +static void vport_netdev_free(struct rcu_head *rcu) { - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netif_running(netdev_vport->dev); -} + struct vport *vport = container_of(rcu, struct vport, rcu); -unsigned char netdev_get_operstate(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->operstate; + if (vport->dev) + dev_put(vport->dev); + ovs_vport_free(vport); } -int netdev_get_ifindex(const struct vport *vport) +void ovs_netdev_detach_dev(struct vport *vport) { - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->ifindex; + ASSERT_RTNL(); + vport->dev->priv_flags &= ~IFF_OVS_DATAPATH; + netdev_rx_handler_unregister(vport->dev); + netdev_upper_dev_unlink(vport->dev, + netdev_master_upper_dev_get(vport->dev)); + dev_set_promiscuity(vport->dev, -1); } +EXPORT_SYMBOL_GPL(ovs_netdev_detach_dev); -int 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; - } - - /* 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). - * (No one comes after us, since we tell handle_bridge() that we took - * the packet.) */ - skb = skb_share_check(skb, GFP_ATOMIC); - if (unlikely(!skb)) - return; - - skb_push(skb, ETH_HLEN); - - if (unlikely(compute_ip_summed(skb, false))) { - kfree_skb(skb); - return; - } - vlan_copy_skb_tci(skb); - - vport_receive(vport, skb); -} - -static inline unsigned packet_length(const struct sk_buff *skb) +static void netdev_destroy(struct vport *vport) { - unsigned length = skb->len - ETH_HLEN; + rtnl_lock(); + if (vport->dev->priv_flags & IFF_OVS_DATAPATH) + ovs_netdev_detach_dev(vport); + rtnl_unlock(); - if (skb->protocol == htons(ETH_P_8021Q)) - length -= VLAN_HLEN; - - return length; + call_rcu(&vport->rcu, vport_netdev_free); } -static bool dev_supports_vlan_tx(struct net_device *dev) +void ovs_netdev_tunnel_destroy(struct vport *vport) { -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,37) - /* Software fallback means every device supports vlan_tci on TX. */ - return true; -#elif defined(HAVE_VLAN_BUG_WORKAROUND) - return dev->features & NETIF_F_HW_VLAN_TX; -#else - /* Assume that the driver is buggy. */ - return false; -#endif -} - -static int netdev_send(struct vport *vport, struct sk_buff *skb) -{ - struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - int mtu = netdev_vport->dev->mtu; - 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", - dp_name(vport->dp), packet_length(skb), mtu); - goto error; - } - - if (unlikely(skb_warn_if_lro(skb))) - goto error; + rtnl_lock(); + if (vport->dev->priv_flags & IFF_OVS_DATAPATH) + ovs_netdev_detach_dev(vport); - skb->dev = netdev_vport->dev; - forward_ip_summed(skb, true); - - if (vlan_tx_tag_present(skb) && !dev_supports_vlan_tx(skb->dev)) { - int features; - - features = netif_skb_features(skb); - - if (!vlan_tso) - features &= ~(NETIF_F_TSO | NETIF_F_TSO6 | - NETIF_F_UFO | NETIF_F_FSO); - - if (netif_needs_gso(skb, features)) { - struct sk_buff *nskb; - - 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; - } - - skb_shinfo(skb)->gso_type &= ~SKB_GSO_DODGY; - goto tag; - } - - if (IS_ERR(nskb)) { - kfree_skb(skb); - return 0; - } - consume_skb(skb); - skb = nskb; - - len = 0; - do { - nskb = skb->next; - skb->next = NULL; - - skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb)); - if (likely(skb)) { - len += skb->len; - vlan_set_tci(skb, 0); - dev_queue_xmit(skb); - } - - skb = nskb; - } while (skb); - - return len; - } - -tag: - skb = __vlan_put_tag(skb, vlan_tx_tag_get(skb)); - if (unlikely(!skb)) - return 0; - vlan_set_tci(skb, 0); - } + /* We can be invoked by both explicit vport deletion and + * underlying netdev deregistration; delete the link only + * if it's not already shutting down. + */ - len = skb->len; - dev_queue_xmit(skb); + if (vport->dev->reg_state == NETREG_REGISTERED) + rtnl_delete_link(vport->dev); - return len; + dev_put(vport->dev); + vport->dev = NULL; + rtnl_unlock(); -error: - kfree_skb(skb); - vport_record_error(vport, VPORT_E_TX_DROPPED); - return 0; + call_rcu(&vport->rcu, vport_netdev_free); } +EXPORT_SYMBOL_GPL(ovs_netdev_tunnel_destroy); /* Returns null if this device is not attached to a datapath. */ -struct vport *netdev_get_vport(struct net_device *dev) +struct vport *ovs_netdev_get_vport(struct net_device *dev) { -#ifdef IFF_BRIDGE_PORT -#if IFF_BRIDGE_PORT != IFF_OVS_DATAPATH +#if defined HAVE_NETDEV_RX_HANDLER_REGISTER || \ + defined HAVE_RHEL_OVS_HOOK +#ifdef HAVE_OVS_DATAPATH if (likely(dev->priv_flags & IFF_OVS_DATAPATH)) #else - if (likely(rcu_access_pointer(dev->rx_handler) == netdev_frame_hook)) + 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 +#ifdef HAVE_NET_DEVICE_EXTENDED + return (struct vport *) + rcu_dereference_rtnl(netdev_extended(dev)->rx_handler_data); +#else return (struct vport *)rcu_dereference_rtnl(dev->rx_handler_data); +#endif +#endif else return NULL; #else @@ -388,36 +256,35 @@ struct vport *netdev_get_vport(struct net_device *dev) #endif } -const struct vport_ops netdev_vport_ops = { +static 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 = netdev_set_addr, - .get_name = netdev_get_name, - .get_addr = netdev_get_addr, - .get_kobj = netdev_get_kobj, - .get_dev_flags = netdev_get_dev_flags, - .is_running = netdev_is_running, - .get_operstate = netdev_get_operstate, - .get_ifindex = netdev_get_ifindex, - .get_mtu = netdev_get_mtu, - .send = netdev_send, + .send = dev_queue_xmit, }; -#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) +int __init ovs_netdev_init(void) +{ + return ovs_vport_ops_register(&ovs_netdev_vport_ops); +} + +void ovs_netdev_exit(void) +{ + ovs_vport_ops_unregister(&ovs_netdev_vport_ops); +} + +#if !defined HAVE_NETDEV_RX_HANDLER_REGISTER && \ + !defined HAVE_RHEL_OVS_HOOK /* - * In kernels earlier than 2.6.36, Open vSwitch cannot safely coexist with - * the Linux bridge module on any released version of Linux, because there - * is only a single bridge hook function and only a single br_port member - * in struct net_device. + * 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)"). * - * Declaring and exporting this symbol enforces mutual exclusion. The bridge - * module also exports the same symbol, so 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_mod)"). + * 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.