X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fvport-netdev.c;h=808ec34e2554c4e9f8c70189cfe93ee0ad410a06;hb=f285d3e715512571c4b2f92a4d1c65022bbcc9d5;hp=8e7847efda923f26959238d9d44db20f9e72e67d;hpb=5953c70e61897996e8b05fadea988b3289e133de;p=sliver-openvswitch.git diff --git a/datapath/vport-netdev.c b/datapath/vport-netdev.c index 8e7847efd..808ec34e2 100644 --- a/datapath/vport-netdev.c +++ b/datapath/vport-netdev.c @@ -1,11 +1,23 @@ /* - * Copyright (c) 2010 Nicira Networks. - * Distributed under the terms of the GNU GPL version 2. + * Copyright (c) 2007-2012 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 + #include #include #include @@ -16,40 +28,96 @@ #include +#include "checksum.h" #include "datapath.h" +#include "vlan.h" #include "vport-internal_dev.h" #include "vport-netdev.h" -#include "compat.h" +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,37) && \ + !defined(HAVE_VLAN_BUG_WORKAROUND) +#include + +static int vlan_tso __read_mostly; +module_param(vlan_tso, int, 0644); +MODULE_PARM_DESC(vlan_tso, "Enable TSO for VLAN packets"); +#else +#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 net_bridge_port *, struct sk_buff *); +static void netdev_port_receive(struct vport *vport, struct sk_buff *skb); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,39) +/* 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 = ovs_netdev_get_vport(skb->dev); + + netdev_port_receive(vport, skb); + + return RX_HANDLER_CONSUMED; +} +#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 = ovs_netdev_get_vport(skb->dev); + + netdev_port_receive(vport, skb); + + return NULL; +} +#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) /* * Used as br_handle_frame_hook. (Cannot run bridge at the same time, even on * different set of devices!) */ -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22) /* Called with rcu_read_lock and bottom-halves disabled. */ -static struct sk_buff * -netdev_frame_hook(struct net_bridge_port *p, struct sk_buff *skb) +static struct sk_buff *netdev_frame_hook(struct net_bridge_port *p, + struct sk_buff *skb) { - netdev_port_receive(p, skb); + netdev_port_receive((struct vport *)p, 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) +static int netdev_frame_hook(struct net_bridge_port *p, struct sk_buff **pskb) { - netdev_port_receive(p, *pskb); + netdev_port_receive((struct vport *)p, *pskb); return 1; } #else #error #endif -static int -netdev_init(void) +#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) { /* Hook into callback used by the bridge to intercept packets. * Parasites we are. */ @@ -58,20 +126,20 @@ netdev_init(void) return 0; } -static void -netdev_exit(void) +static void netdev_exit(void) { br_handle_frame_hook = NULL; } +#endif -static struct vport * -netdev_create(const char *name, const void __user *config) +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); + vport = ovs_vport_alloc(sizeof(struct netdev_vport), + &ovs_netdev_vport_ops, parms); if (IS_ERR(vport)) { err = PTR_ERR(vport); goto error; @@ -79,7 +147,7 @@ netdev_create(const char *name, const void __user *config) netdev_vport = netdev_vport_priv(vport); - netdev_vport->dev = dev_get_by_name(&init_net, name); + netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name); if (!netdev_vport->dev) { err = -ENODEV; goto error_free_vport; @@ -87,244 +155,271 @@ netdev_create(const char *name, const void __user *config) if (netdev_vport->dev->flags & IFF_LOOPBACK || netdev_vport->dev->type != ARPHRD_ETHER || - is_internal_dev(netdev_vport->dev)) { + ovs_is_internal_dev(netdev_vport->dev)) { err = -EINVAL; goto error_put; } - if (netdev_vport->dev->br_port) { - err = -EBUSY; - 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_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(); return vport; +#ifndef HAVE_RHEL_OVS_HOOK +error_unlock: +#endif + rtnl_unlock(); error_put: dev_put(netdev_vport->dev); error_free_vport: - vport_free(vport); + ovs_vport_free(vport); error: return ERR_PTR(err); } -static int -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); - dev_put(netdev_vport->dev); - vport_free(vport); - - return 0; -} +#ifdef HAVE_RHEL_OVS_HOOK + rcu_assign_pointer(netdev_vport->dev->ax25_ptr, NULL); -static int -netdev_attach(struct vport *vport) -{ - struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - - dev_set_promiscuity(netdev_vport->dev, 1); - dev_disable_lro(netdev_vport->dev); - rcu_assign_pointer(netdev_vport->dev->br_port, (struct net_bridge_port *)vport); - - return 0; + 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_from_priv(netdev_vport)); } -static int -netdev_detach(struct vport *vport) +static void netdev_destroy(struct vport *vport) { struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - rcu_assign_pointer(netdev_vport->dev->br_port, NULL); + 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 0; -} - -int -netdev_set_mtu(struct vport *vport, int mtu) -{ - struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return dev_set_mtu(netdev_vport->dev, mtu); -} - -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); + call_rcu(&netdev_vport->rcu, free_port_rcu); } -const char * -netdev_get_name(const struct vport *vport) +const char *ovs_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) +int ovs_netdev_get_ifindex(const struct vport *vport) { const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->dev_addr; + return netdev_vport->dev->ifindex; } -struct kobject * -netdev_get_kobj(const struct vport *vport) +/* Must be called with rcu_read_lock. */ +static void netdev_port_receive(struct vport *vport, struct sk_buff *skb) { - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return &netdev_vport->dev->NETDEV_DEV_MEMBER.kobj; -} + if (unlikely(!vport)) + goto error; -int -netdev_get_stats(const struct vport *vport, struct odp_vport_stats *stats) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - const struct net_device_stats *netdev_stats; - - netdev_stats = dev_get_stats(netdev_vport->dev); - - stats->rx_bytes = netdev_stats->rx_bytes; - stats->rx_packets = netdev_stats->rx_packets; - stats->tx_bytes = netdev_stats->tx_bytes; - stats->tx_packets = netdev_stats->tx_packets; - stats->rx_dropped = netdev_stats->rx_dropped; - stats->rx_errors = netdev_stats->rx_errors; - stats->rx_frame_err = netdev_stats->rx_frame_errors; - stats->rx_over_err = netdev_stats->rx_over_errors; - stats->rx_crc_err = netdev_stats->rx_crc_errors; - stats->tx_dropped = netdev_stats->tx_dropped; - stats->tx_errors = netdev_stats->tx_errors; - stats->collisions = netdev_stats->collisions; + if (unlikely(skb_warn_if_lro(skb))) + goto error; - return 0; -} + /* 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; -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); -} + skb_push(skb, ETH_HLEN); -int -netdev_is_running(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netif_running(netdev_vport->dev); -} + if (unlikely(compute_ip_summed(skb, false))) + goto error; -unsigned char -netdev_get_operstate(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->operstate; -} + vlan_copy_skb_tci(skb); -int -netdev_get_ifindex(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->ifindex; -} + ovs_vport_receive(vport, skb); + return; -int -netdev_get_iflink(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->iflink; -} - -int -netdev_get_mtu(const struct vport *vport) -{ - const struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - return netdev_vport->dev->mtu; +error: + kfree_skb(skb); } -/* Must be called with rcu_read_lock. */ -static void -netdev_port_receive(struct net_bridge_port *p, struct sk_buff *skb) +static unsigned int packet_length(const struct sk_buff *skb) { - struct vport *vport = (struct vport *)p; + unsigned int length = skb->len - ETH_HLEN; - /* 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 (!skb) - return; + if (skb->protocol == htons(ETH_P_8021Q)) + length -= VLAN_HLEN; - /* Push the Ethernet header back on. */ - skb_push(skb, ETH_HLEN); - skb_reset_mac_header(skb); - compute_ip_summed(skb, false); + return length; +} - vport_receive(vport, skb); +static bool dev_supports_vlan_tx(struct net_device *dev) +{ +#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) +static int netdev_send(struct vport *vport, struct sk_buff *skb) { struct netdev_vport *netdev_vport = netdev_vport_priv(vport); - int len = skb->len; + int mtu = netdev_vport->dev->mtu; + int len; + + if (unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) { + net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n", + netdev_vport->dev->name, + packet_length(skb), mtu); + goto error; + } skb->dev = netdev_vport->dev; - forward_ip_summed(skb); + 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); + } + + len = skb->len; dev_queue_xmit(skb); return len; + +error: + 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 * -netdev_get_vport(struct net_device *dev) +struct vport *ovs_netdev_get_vport(struct net_device *dev) { - return (struct vport *)dev->br_port; +#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 + return (struct vport *)rcu_dereference_rtnl(dev->br_port); +#endif } -struct vport_ops netdev_vport_ops = { - .type = "netdev", - .flags = VPORT_F_REQUIRED, +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, - .attach = netdev_attach, - .detach = netdev_detach, - .set_mtu = netdev_set_mtu, - .set_addr = netdev_set_addr, - .get_name = netdev_get_name, - .get_addr = netdev_get_addr, - .get_kobj = netdev_get_kobj, - .get_stats = netdev_get_stats, - .get_dev_flags = netdev_get_dev_flags, - .is_running = netdev_is_running, - .get_operstate = netdev_get_operstate, - .get_ifindex = netdev_get_ifindex, - .get_iflink = netdev_get_iflink, - .get_mtu = netdev_get_mtu, + .get_name = ovs_netdev_get_name, + .get_ifindex = ovs_netdev_get_ifindex, .send = netdev_send, }; +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,36) && \ + !defined HAVE_RHEL_OVS_HOOK /* - * 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. */ typeof(br_should_route_hook) br_should_route_hook; EXPORT_SYMBOL(br_should_route_hook); +#endif