2 * Copyright (c) 2010 Nicira Networks.
3 * Distributed under the terms of the GNU GPL version 2.
5 * Significant portions of this file may be copied from parts of the Linux
6 * kernel, by Linus Torvalds and others.
9 #include <linux/dcache.h>
10 #include <linux/etherdevice.h>
12 #include <linux/kernel.h>
13 #include <linux/list.h>
14 #include <linux/mutex.h>
15 #include <linux/percpu.h>
16 #include <linux/rtnetlink.h>
20 extern struct vport_ops netdev_vport_ops;
21 extern struct vport_ops internal_vport_ops;
22 extern struct vport_ops gre_vport_ops;
24 static struct vport_ops *base_vport_ops_list[] = {
30 static const struct vport_ops **vport_ops_list;
31 static int n_vport_types;
33 static struct hlist_head *dev_table;
34 #define VPORT_HASH_BUCKETS 1024
36 /* Both RTNL lock and vport_mutex need to be held when updating dev_table.
38 * If you use vport_locate and then perform some operations, you need to hold
39 * one of these locks if you don't want the vport to be deleted out from under
42 * If you get a reference to a vport through a dp_port, it is protected
43 * by RCU and you need to hold rcu_read_lock instead when reading.
45 * If multiple locks are taken, the hierarchy is:
50 static DEFINE_MUTEX(vport_mutex);
53 * vport_lock - acquire vport lock
55 * Acquire global vport lock. See above comment about locking requirements
56 * and specific function definitions. May sleep.
61 mutex_lock(&vport_mutex);
65 * vport_unlock - release vport lock
67 * Release lock acquired with vport_lock.
72 mutex_unlock(&vport_mutex);
75 #define ASSERT_VPORT() do { \
76 if (unlikely(!mutex_is_locked(&vport_mutex))) { \
77 printk(KERN_ERR "openvswitch: vport lock not held at %s (%d)\n", \
78 __FILE__, __LINE__); \
84 * vport_init - initialize vport subsystem
86 * Called at module load time to initialize the vport subsystem and any
87 * compiled in vport types.
95 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
102 vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
103 sizeof(struct vport_ops *), GFP_KERNEL);
104 if (!vport_ops_list) {
106 goto error_dev_table;
109 for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
110 struct vport_ops *new_ops = base_vport_ops_list[i];
112 if (new_ops->get_stats && new_ops->flags & VPORT_F_GEN_STATS) {
113 printk(KERN_INFO "openvswitch: both get_stats() and VPORT_F_GEN_STATS defined on vport %s, dropping VPORT_F_GEN_STATS\n", new_ops->type);
114 new_ops->flags &= ~VPORT_F_GEN_STATS;
118 err = new_ops->init();
123 vport_ops_list[n_vport_types++] = new_ops;
124 else if (new_ops->flags & VPORT_F_REQUIRED) {
146 for (i = 0; i < VPORT_HASH_BUCKETS; i++) {
147 struct hlist_head *bucket = &dev_table[i];
149 struct hlist_node *node, *next;
151 hlist_for_each_entry_safe(vport, node, next, bucket, hash_node)
160 * vport_exit - shutdown vport subsystem
162 * Called at module exit time to shutdown the vport subsystem and any
163 * initialized vport types.
172 for (i = 0; i < n_vport_types; i++) {
173 if (vport_ops_list[i]->exit)
174 vport_ops_list[i]->exit();
177 kfree(vport_ops_list);
182 * vport_add - add vport device (for userspace callers)
184 * @uvport_config: New port configuration.
186 * Creates a new vport with the specified configuration (which is dependent
187 * on device type). This function is for userspace callers and assumes no
191 vport_add(const struct odp_vport_add __user *uvport_config)
193 struct odp_vport_add vport_config;
197 if (copy_from_user(&vport_config, uvport_config, sizeof(struct odp_vport_add)))
200 vport_config.port_type[VPORT_TYPE_SIZE - 1] = '\0';
201 vport_config.devname[IFNAMSIZ - 1] = '\0';
205 vport = vport_locate(vport_config.devname);
212 vport = __vport_add(vport_config.devname, vport_config.port_type,
213 vport_config.config);
217 err = PTR_ERR(vport);
225 * vport_mod - modify existing vport device (for userspace callers)
227 * @uvport_config: New configuration for vport
229 * Modifies an existing device with the specified configuration (which is
230 * dependent on device type). This function is for userspace callers and
231 * assumes no locks are held.
234 vport_mod(const struct odp_vport_mod __user *uvport_config)
236 struct odp_vport_mod vport_config;
240 if (copy_from_user(&vport_config, uvport_config, sizeof(struct odp_vport_mod)))
243 vport_config.devname[IFNAMSIZ - 1] = '\0';
247 vport = vport_locate(vport_config.devname);
254 err = __vport_mod(vport, vport_config.config);
263 * vport_del - delete existing vport device (for userspace callers)
265 * @udevname: Name of device to delete
267 * Deletes the specified device. Detaches the device from a datapath first
268 * if it is attached. Deleting the device will fail if it does not exist or it
269 * is the datapath local port. It is also possible to fail for less obvious
270 * reasons, such as lack of memory. This function is for userspace callers and
271 * assumes no locks are held.
274 vport_del(const char __user *udevname)
276 char devname[IFNAMSIZ];
278 struct dp_port *dp_port;
281 if (strncpy_from_user(devname, udevname, IFNAMSIZ - 1) < 0)
283 devname[IFNAMSIZ - 1] = '\0';
287 vport = vport_locate(devname);
293 dp_port = vport_get_dp_port(vport);
295 struct datapath *dp = dp_port->dp;
297 mutex_lock(&dp->mutex);
299 if (!strcmp(dp_name(dp), devname)) {
304 err = dp_detach_port(dp_port, 0);
307 mutex_unlock(&dp->mutex);
314 err = __vport_del(vport);
323 * vport_stats_get - retrieve device stats (for userspace callers)
325 * @ustats_req: Stats request parameters.
327 * Retrieves transmit, receive, and error stats for the given device. This
328 * function is for userspace callers and assumes no locks are held.
331 vport_stats_get(struct odp_vport_stats_req __user *ustats_req)
333 struct odp_vport_stats_req stats_req;
337 if (copy_from_user(&stats_req, ustats_req, sizeof(struct odp_vport_stats_req)))
340 stats_req.devname[IFNAMSIZ - 1] = '\0';
344 vport = vport_locate(stats_req.devname);
350 if (vport->ops->get_stats)
351 err = vport->ops->get_stats(vport, &stats_req.stats);
352 else if (vport->ops->flags & VPORT_F_GEN_STATS) {
355 memset(&stats_req.stats, 0, sizeof(struct odp_vport_stats));
357 for_each_possible_cpu(i) {
358 const struct vport_percpu_stats *percpu_stats;
360 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
361 stats_req.stats.rx_bytes += percpu_stats->rx_bytes;
362 stats_req.stats.rx_packets += percpu_stats->rx_packets;
363 stats_req.stats.tx_bytes += percpu_stats->tx_bytes;
364 stats_req.stats.tx_packets += percpu_stats->tx_packets;
367 spin_lock_bh(&vport->err_stats.lock);
369 stats_req.stats.rx_dropped = vport->err_stats.rx_dropped;
370 stats_req.stats.rx_errors = vport->err_stats.rx_errors
371 + vport->err_stats.rx_frame_err
372 + vport->err_stats.rx_over_err
373 + vport->err_stats.rx_crc_err;
374 stats_req.stats.rx_frame_err = vport->err_stats.rx_frame_err;
375 stats_req.stats.rx_over_err = vport->err_stats.rx_over_err;
376 stats_req.stats.rx_crc_err = vport->err_stats.rx_crc_err;
377 stats_req.stats.tx_dropped = vport->err_stats.tx_dropped;
378 stats_req.stats.tx_errors = vport->err_stats.tx_errors;
379 stats_req.stats.collisions = vport->err_stats.collisions;
381 spin_unlock_bh(&vport->err_stats.lock);
391 if (copy_to_user(ustats_req, &stats_req, sizeof(struct odp_vport_stats_req)))
398 * vport_ether_get - retrieve device Ethernet address (for userspace callers)
400 * @uvport_ether: Ethernet address request parameters.
402 * Retrieves the Ethernet address of the given device. This function is for
403 * userspace callers and assumes no locks are held.
406 vport_ether_get(struct odp_vport_ether __user *uvport_ether)
408 struct odp_vport_ether vport_ether;
412 if (copy_from_user(&vport_ether, uvport_ether, sizeof(struct odp_vport_ether)))
415 vport_ether.devname[IFNAMSIZ - 1] = '\0';
419 vport = vport_locate(vport_ether.devname);
425 memcpy(vport_ether.ether_addr, vport_get_addr(vport), ETH_ALEN);
431 if (copy_to_user(uvport_ether, &vport_ether, sizeof(struct odp_vport_ether)))
438 * vport_ether_set - set device Ethernet address (for userspace callers)
440 * @uvport_ether: Ethernet address request parameters.
442 * Sets the Ethernet address of the given device. Some devices may not support
443 * setting the Ethernet address, in which case the result will always be
444 * -EOPNOTSUPP. This function is for userspace callers and assumes no locks
448 vport_ether_set(struct odp_vport_ether __user *uvport_ether)
450 struct odp_vport_ether vport_ether;
454 if (copy_from_user(&vport_ether, uvport_ether, sizeof(struct odp_vport_ether)))
457 vport_ether.devname[IFNAMSIZ - 1] = '\0';
462 vport = vport_locate(vport_ether.devname);
468 err = vport_set_addr(vport, vport_ether.ether_addr);
477 * vport_mut_get - retrieve device MTU (for userspace callers)
479 * @uvport_mtu: MTU request parameters.
481 * Retrieves the MTU of the given device. This function is for userspace
482 * callers and assumes no locks are held.
485 vport_mtu_get(struct odp_vport_mtu __user *uvport_mtu)
487 struct odp_vport_mtu vport_mtu;
491 if (copy_from_user(&vport_mtu, uvport_mtu, sizeof(struct odp_vport_mtu)))
494 vport_mtu.devname[IFNAMSIZ - 1] = '\0';
498 vport = vport_locate(vport_mtu.devname);
504 vport_mtu.mtu = vport_get_mtu(vport);
510 if (copy_to_user(uvport_mtu, &vport_mtu, sizeof(struct odp_vport_mtu)))
517 * vport_mtu_set - set device MTU (for userspace callers)
519 * @uvport_mtu: MTU request parameters.
521 * Sets the MTU of the given device. Some devices may not support setting the
522 * MTU, in which case the result will always be -EOPNOTSUPP. This function is
523 * for userspace callers and assumes no locks are held.
526 vport_mtu_set(struct odp_vport_mtu __user *uvport_mtu)
528 struct odp_vport_mtu vport_mtu;
532 if (copy_from_user(&vport_mtu, uvport_mtu, sizeof(struct odp_vport_mtu)))
535 vport_mtu.devname[IFNAMSIZ - 1] = '\0';
540 vport = vport_locate(vport_mtu.devname);
546 err = vport_set_mtu(vport, vport_mtu.mtu);
554 static struct hlist_head *
555 hash_bucket(const char *name)
557 unsigned int hash = full_name_hash(name, strlen(name));
558 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
562 * vport_locate - find a port that has already been created
564 * @name: name of port to find
566 * Either RTNL or vport lock must be acquired before calling this function
567 * and held while using the found port. See the locking comments at the
571 vport_locate(const char *name)
573 struct hlist_head *bucket = hash_bucket(name);
575 struct hlist_node *node;
577 if (unlikely(!mutex_is_locked(&vport_mutex) && !rtnl_is_locked())) {
578 printk(KERN_ERR "openvswitch: neither RTNL nor vport lock held in vport_locate\n");
582 hlist_for_each_entry(vport, node, bucket, hash_node)
583 if (!strcmp(name, vport_get_name(vport)))
590 register_vport(struct vport *vport)
592 hlist_add_head(&vport->hash_node, hash_bucket(vport_get_name(vport)));
596 unregister_vport(struct vport *vport)
598 hlist_del(&vport->hash_node);
602 * vport_alloc - allocate and initialize new vport
604 * @priv_size: Size of private data area to allocate.
605 * @ops: vport device ops
607 * Allocate and initialize a new vport defined by @ops. The vport will contain
608 * a private data area of size @priv_size that can be accessed using
609 * vport_priv(). vports that are no longer needed should be released with
613 vport_alloc(int priv_size, const struct vport_ops *ops)
618 alloc_size = sizeof(struct vport);
620 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
621 alloc_size += priv_size;
624 vport = kzalloc(alloc_size, GFP_KERNEL);
626 return ERR_PTR(-ENOMEM);
630 if (vport->ops->flags & VPORT_F_GEN_STATS) {
631 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
632 if (!vport->percpu_stats)
633 return ERR_PTR(-ENOMEM);
635 spin_lock_init(&vport->err_stats.lock);
642 * vport_free - uninitialize and free vport
644 * @vport: vport to free
646 * Frees a vport allocated with vport_alloc() when it is no longer needed.
649 vport_free(struct vport *vport)
651 if (vport->ops->flags & VPORT_F_GEN_STATS)
652 free_percpu(vport->percpu_stats);
658 * __vport_add - add vport device (for kernel callers)
660 * @name: Name of new device.
661 * @type: Type of new device (to be matched against types in registered vport
663 * @config: Device type specific configuration. Userspace pointer.
665 * Creates a new vport with the specified configuration (which is dependent
666 * on device type). Both RTNL and vport locks must be held.
669 __vport_add(const char *name, const char *type, const void __user *config)
678 for (i = 0; i < n_vport_types; i++) {
679 if (!strcmp(vport_ops_list[i]->type, type)) {
680 vport = vport_ops_list[i]->create(name, config);
682 err = PTR_ERR(vport);
686 register_vport(vport);
698 * __vport_mod - modify existing vport device (for kernel callers)
700 * @vport: vport to modify.
701 * @config: Device type specific configuration. Userspace pointer.
703 * Modifies an existing device with the specified configuration (which is
704 * dependent on device type). Both RTNL and vport locks must be held.
707 __vport_mod(struct vport *vport, const void __user *config)
712 if (vport->ops->modify)
713 return vport->ops->modify(vport, config);
719 * __vport_del - delete existing vport device (for kernel callers)
721 * @vport: vport to delete.
723 * Deletes the specified device. The device must not be currently attached to
724 * a datapath. It is possible to fail for reasons such as lack of memory.
725 * Both RTNL and vport locks must be held.
728 __vport_del(struct vport *vport)
732 BUG_ON(vport_get_dp_port(vport));
734 unregister_vport(vport);
736 return vport->ops->destroy(vport);
740 * vport_attach - attach a vport to a datapath
742 * @vport: vport to attach.
743 * @dp_port: Datapath port to attach the vport to.
745 * Attaches a vport to a specific datapath so that packets may be exchanged.
746 * Both ports must be currently unattached. @dp_port must be successfully
747 * attached to a vport before it is connected to a datapath and must not be
748 * modified while connected. RTNL lock and the appropriate DP mutex must be held.
751 vport_attach(struct vport *vport, struct dp_port *dp_port)
758 if (vport_get_dp_port(vport))
761 if (vport->ops->attach) {
764 err = vport->ops->attach(vport);
769 dp_port->vport = vport;
770 rcu_assign_pointer(vport->dp_port, dp_port);
776 * vport_detach - detach a vport from a datapath
778 * @vport: vport to detach.
780 * Detaches a vport from a datapath. May fail for a variety of reasons,
781 * including lack of memory. RTNL lock and the appropriate DP mutex must be held.
784 vport_detach(struct vport *vport)
786 struct dp_port *dp_port;
790 dp_port = vport_get_dp_port(vport);
794 dp_port->vport = NULL;
795 rcu_assign_pointer(vport->dp_port, NULL);
797 if (vport->ops->detach)
798 return vport->ops->detach(vport);
804 * vport_set_mtu - set device MTU (for kernel callers)
806 * @vport: vport on which to set MTU.
809 * Sets the MTU of the given device. Some devices may not support setting the
810 * MTU, in which case the result will always be -EOPNOTSUPP. RTNL lock must
814 vport_set_mtu(struct vport *vport, int mtu)
821 if (vport->ops->set_mtu)
822 return vport->ops->set_mtu(vport, mtu);
828 * vport_set_addr - set device Ethernet address (for kernel callers)
830 * @vport: vport on which to set Ethernet address.
831 * @addr: New address.
833 * Sets the Ethernet address of the given device. Some devices may not support
834 * setting the Ethernet address, in which case the result will always be
835 * -EOPNOTSUPP. RTNL lock must be held.
838 vport_set_addr(struct vport *vport, const unsigned char *addr)
842 if (!is_valid_ether_addr(addr))
843 return -EADDRNOTAVAIL;
845 if (vport->ops->set_addr)
846 return vport->ops->set_addr(vport, addr);
852 * vport_get_name - retrieve device name
854 * @vport: vport from which to retrieve the name.
856 * Retrieves the name of the given device. Either RTNL lock or rcu_read_lock
857 * must be held for the entire duration that the name is in use.
860 vport_get_name(const struct vport *vport)
862 return vport->ops->get_name(vport);
866 * vport_get_type - retrieve device type
868 * @vport: vport from which to retrieve the type.
870 * Retrieves the type of the given device. Either RTNL lock or rcu_read_lock
871 * must be held for the entire duration that the type is in use.
874 vport_get_type(const struct vport *vport)
876 return vport->ops->type;
880 * vport_get_addr - retrieve device Ethernet address (for kernel callers)
882 * @vport: vport from which to retrieve the Ethernet address.
884 * Retrieves the Ethernet address of the given device. Either RTNL lock or
885 * rcu_read_lock must be held for the entire duration that the Ethernet address
888 const unsigned char *
889 vport_get_addr(const struct vport *vport)
891 return vport->ops->get_addr(vport);
895 * vport_get_dp_port - retrieve attached datapath port
897 * @vport: vport from which to retrieve the datapath port.
899 * Retrieves the attached datapath port or null if not attached. Either RTNL
900 * lock or rcu_read_lock must be held for the entire duration that the datapath
901 * port is being accessed.
904 vport_get_dp_port(const struct vport *vport)
906 return rcu_dereference(vport->dp_port);
910 * vport_get_kobj - retrieve associated kobj
912 * @vport: vport from which to retrieve the associated kobj
914 * Retrieves the associated kobj or null if no kobj. The returned kobj is
915 * valid for as long as the vport exists.
918 vport_get_kobj(const struct vport *vport)
920 if (vport->ops->get_kobj)
921 return vport->ops->get_kobj(vport);
927 * vport_get_flags - retrieve device flags
929 * @vport: vport from which to retrieve the flags
931 * Retrieves the flags of the given device. Either RTNL lock or rcu_read_lock
935 vport_get_flags(const struct vport *vport)
937 return vport->ops->get_dev_flags(vport);
941 * vport_get_flags - check whether device is running
943 * @vport: vport on which to check status.
945 * Checks whether the given device is running. Either RTNL lock or
946 * rcu_read_lock must be held.
949 vport_is_running(const struct vport *vport)
951 return vport->ops->is_running(vport);
955 * vport_get_flags - retrieve device operating state
957 * @vport: vport from which to check status
959 * Retrieves the RFC2863 operstate of the given device. Either RTNL lock or
960 * rcu_read_lock must be held.
963 vport_get_operstate(const struct vport *vport)
965 return vport->ops->get_operstate(vport);
969 * vport_get_ifindex - retrieve device system interface index
971 * @vport: vport from which to retrieve index
973 * Retrieves the system interface index of the given device. Not all devices
974 * will have system indexes, in which case the index of the datapath local
975 * port is returned. Returns a negative index on error. Either RTNL lock or
976 * rcu_read_lock must be held.
979 vport_get_ifindex(const struct vport *vport)
981 const struct dp_port *dp_port;
983 if (vport->ops->get_ifindex)
984 return vport->ops->get_ifindex(vport);
986 /* If we don't actually have an ifindex, use the local port's.
987 * Userspace doesn't check it anyways. */
988 dp_port = vport_get_dp_port(vport);
992 return vport_get_ifindex(dp_port->dp->ports[ODPP_LOCAL]->vport);
996 * vport_get_iflink - retrieve device system link index
998 * @vport: vport from which to retrieve index
1000 * Retrieves the system link index of the given device. The link is the index
1001 * of the interface on which the packet will actually be sent. In most cases
1002 * this is the same as the ifindex but may be different for tunnel devices.
1003 * Returns a negative index on error. Either RTNL lock or rcu_read_lock must
1007 vport_get_iflink(const struct vport *vport)
1009 if (vport->ops->get_iflink)
1010 return vport->ops->get_iflink(vport);
1012 /* If we don't have an iflink, use the ifindex. In most cases they
1014 return vport_get_ifindex(vport);
1018 * vport_get_mtu - retrieve device MTU (for kernel callers)
1020 * @vport: vport from which to retrieve MTU
1022 * Retrieves the MTU of the given device. Either RTNL lock or rcu_read_lock
1026 vport_get_mtu(const struct vport *vport)
1028 return vport->ops->get_mtu(vport);
1032 * vport_receive - pass up received packet to the datapath for processing
1034 * @vport: vport that received the packet
1035 * @skb: skb that was received
1037 * Must be called with rcu_read_lock and bottom halves disabled. The packet
1038 * cannot be shared and skb->data should point to the Ethernet header.
1041 vport_receive(struct vport *vport, struct sk_buff *skb)
1043 struct dp_port *dp_port = vport_get_dp_port(vport);
1048 if (vport->ops->flags & VPORT_F_GEN_STATS) {
1049 struct vport_percpu_stats *stats;
1053 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
1054 stats->rx_packets++;
1055 stats->rx_bytes += skb->len;
1060 if (!(vport->ops->flags & VPORT_F_TUN_ID))
1061 OVS_CB(skb)->tun_id = 0;
1063 dp_process_received_packet(dp_port, skb);
1067 * vport_send - send a packet on a device
1069 * @vport: vport on which to send the packet
1072 * Sends the given packet and returns the length of data sent. Either RTNL
1073 * lock or rcu_read_lock must be held.
1076 vport_send(struct vport *vport, struct sk_buff *skb)
1080 sent = vport->ops->send(vport, skb);
1082 if (vport->ops->flags & VPORT_F_GEN_STATS && sent > 0) {
1083 struct vport_percpu_stats *stats;
1087 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
1088 stats->tx_packets++;
1089 stats->tx_bytes += sent;
1098 * vport_record_error - indicate device error to generic stats layer
1100 * @vport: vport that encountered the error
1101 * @err_type: one of enum vport_err_type types to indicate the error type
1103 * If using the vport generic stats layer indicate that an error of the given
1107 vport_record_error(struct vport *vport, enum vport_err_type err_type)
1109 if (vport->ops->flags & VPORT_F_GEN_STATS) {
1111 spin_lock_bh(&vport->err_stats.lock);
1114 case VPORT_E_RX_DROPPED:
1115 vport->err_stats.rx_dropped++;
1118 case VPORT_E_RX_ERROR:
1119 vport->err_stats.rx_errors++;
1122 case VPORT_E_RX_FRAME:
1123 vport->err_stats.rx_frame_err++;
1126 case VPORT_E_RX_OVER:
1127 vport->err_stats.rx_over_err++;
1130 case VPORT_E_RX_CRC:
1131 vport->err_stats.rx_crc_err++;
1134 case VPORT_E_TX_DROPPED:
1135 vport->err_stats.tx_dropped++;
1138 case VPORT_E_TX_ERROR:
1139 vport->err_stats.tx_errors++;
1142 case VPORT_E_COLLISION:
1143 vport->err_stats.collisions++;
1147 spin_unlock_bh(&vport->err_stats.lock);
1152 * vport_gen_ether_addr - generate an Ethernet address
1154 * @addr: location to store generated address
1156 * Generates a random Ethernet address for use when creating a device that
1157 * has no natural address.
1160 vport_gen_ether_addr(u8 *addr)
1162 random_ether_addr(addr);
1164 /* Set the OUI to the Nicira one. */
1169 /* Set the top bit to indicate random address. */