2 * Copyright (c) 2010, 2011 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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11 #include <linux/dcache.h>
12 #include <linux/etherdevice.h>
14 #include <linux/if_vlan.h>
15 #include <linux/kernel.h>
16 #include <linux/list.h>
17 #include <linux/mutex.h>
18 #include <linux/percpu.h>
19 #include <linux/rcupdate.h>
20 #include <linux/rtnetlink.h>
21 #include <linux/compat.h>
22 #include <linux/version.h>
25 #include "vport-internal_dev.h"
27 /* List of statically compiled vport implementations. Don't forget to also
28 * add yours to the list at the bottom of vport.h. */
29 static const struct vport_ops *base_vport_ops_list[] = {
34 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
39 static const struct vport_ops **vport_ops_list;
40 static int n_vport_types;
42 /* Protected by RCU read lock for reading, RTNL lock for writing. */
43 static struct hlist_head *dev_table;
44 #define VPORT_HASH_BUCKETS 1024
47 * vport_init - initialize vport subsystem
49 * Called at module load time to initialize the vport subsystem and any
50 * compiled in vport types.
57 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
64 vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
65 sizeof(struct vport_ops *), GFP_KERNEL);
66 if (!vport_ops_list) {
71 for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
72 const struct vport_ops *new_ops = base_vport_ops_list[i];
75 err = new_ops->init();
80 vport_ops_list[n_vport_types++] = new_ops;
81 else if (new_ops->flags & VPORT_F_REQUIRED) {
96 * vport_exit - shutdown vport subsystem
98 * Called at module exit time to shutdown the vport subsystem and any
99 * initialized vport types.
101 void vport_exit(void)
105 for (i = 0; i < n_vport_types; i++) {
106 if (vport_ops_list[i]->exit)
107 vport_ops_list[i]->exit();
110 kfree(vport_ops_list);
114 static struct hlist_head *hash_bucket(const char *name)
116 unsigned int hash = full_name_hash(name, strlen(name));
117 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
121 * vport_locate - find a port that has already been created
123 * @name: name of port to find
125 * Must be called with RTNL or RCU read lock.
127 struct vport *vport_locate(const char *name)
129 struct hlist_head *bucket = hash_bucket(name);
131 struct hlist_node *node;
133 hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
134 if (!strcmp(name, vport_get_name(vport)))
140 static void release_vport(struct kobject *kobj)
142 struct vport *p = container_of(kobj, struct vport, kobj);
146 static struct kobj_type brport_ktype = {
148 .sysfs_ops = &brport_sysfs_ops,
150 .release = release_vport
154 * vport_alloc - allocate and initialize new vport
156 * @priv_size: Size of private data area to allocate.
157 * @ops: vport device ops
159 * Allocate and initialize a new vport defined by @ops. The vport will contain
160 * a private data area of size @priv_size that can be accessed using
161 * vport_priv(). vports that are no longer needed should be released with
164 struct vport *vport_alloc(int priv_size, const struct vport_ops *ops, const struct vport_parms *parms)
169 alloc_size = sizeof(struct vport);
171 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
172 alloc_size += priv_size;
175 vport = kzalloc(alloc_size, GFP_KERNEL);
177 return ERR_PTR(-ENOMEM);
179 vport->dp = parms->dp;
180 vport->port_no = parms->port_no;
181 atomic_set(&vport->sflow_pool, 0);
184 /* Initialize kobject for bridge. This will be added as
185 * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
186 vport->kobj.kset = NULL;
187 kobject_init(&vport->kobj, &brport_ktype);
189 if (vport->ops->flags & VPORT_F_GEN_STATS) {
190 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
191 if (!vport->percpu_stats)
192 return ERR_PTR(-ENOMEM);
194 spin_lock_init(&vport->stats_lock);
201 * vport_free - uninitialize and free vport
203 * @vport: vport to free
205 * Frees a vport allocated with vport_alloc() when it is no longer needed.
207 * The caller must ensure that an RCU grace period has passed since the last
208 * time @vport was in a datapath.
210 void vport_free(struct vport *vport)
212 if (vport->ops->flags & VPORT_F_GEN_STATS)
213 free_percpu(vport->percpu_stats);
215 kobject_put(&vport->kobj);
219 * vport_add - add vport device (for kernel callers)
221 * @parms: Information about new vport.
223 * Creates a new vport with the specified configuration (which is dependent on
224 * device type). RTNL lock must be held.
226 struct vport *vport_add(const struct vport_parms *parms)
234 for (i = 0; i < n_vport_types; i++) {
235 if (vport_ops_list[i]->type == parms->type) {
236 vport = vport_ops_list[i]->create(parms);
238 err = PTR_ERR(vport);
242 hlist_add_head_rcu(&vport->hash_node,
243 hash_bucket(vport_get_name(vport)));
255 * vport_set_options - modify existing vport device (for kernel callers)
257 * @vport: vport to modify.
258 * @port: New configuration.
260 * Modifies an existing device with the specified configuration (which is
261 * dependent on device type). RTNL lock must be held.
263 int vport_set_options(struct vport *vport, struct nlattr *options)
267 if (!vport->ops->set_options)
269 return vport->ops->set_options(vport, options);
273 * vport_del - delete existing vport device
275 * @vport: vport to delete.
277 * Detaches @vport from its datapath and destroys it. It is possible to fail
278 * for reasons such as lack of memory. RTNL lock must be held.
280 int vport_del(struct vport *vport)
284 hlist_del_rcu(&vport->hash_node);
286 return vport->ops->destroy(vport);
290 * vport_set_mtu - set device MTU (for kernel callers)
292 * @vport: vport on which to set MTU.
295 * Sets the MTU of the given device. Some devices may not support setting the
296 * MTU, in which case the result will always be -EOPNOTSUPP. RTNL lock must
299 int vport_set_mtu(struct vport *vport, int mtu)
306 if (vport->ops->set_mtu) {
309 ret = vport->ops->set_mtu(vport, mtu);
311 if (!ret && !is_internal_vport(vport))
312 set_internal_devs_mtu(vport->dp);
320 * vport_set_addr - set device Ethernet address (for kernel callers)
322 * @vport: vport on which to set Ethernet address.
323 * @addr: New address.
325 * Sets the Ethernet address of the given device. Some devices may not support
326 * setting the Ethernet address, in which case the result will always be
327 * -EOPNOTSUPP. RTNL lock must be held.
329 int vport_set_addr(struct vport *vport, const unsigned char *addr)
333 if (!is_valid_ether_addr(addr))
334 return -EADDRNOTAVAIL;
336 if (vport->ops->set_addr)
337 return vport->ops->set_addr(vport, addr);
343 * vport_set_stats - sets offset device stats
345 * @vport: vport on which to set stats
346 * @stats: stats to set
348 * Provides a set of transmit, receive, and error stats to be added as an
349 * offset to the collect data when stats are retreived. Some devices may not
350 * support setting the stats, in which case the result will always be
353 * Must be called with RTNL lock.
355 int vport_set_stats(struct vport *vport, struct rtnl_link_stats64 *stats)
359 if (vport->ops->flags & VPORT_F_GEN_STATS) {
360 spin_lock_bh(&vport->stats_lock);
361 vport->offset_stats = *stats;
362 spin_unlock_bh(&vport->stats_lock);
370 * vport_get_name - retrieve device name
372 * @vport: vport from which to retrieve the name.
374 * Retrieves the name of the given device. Either RTNL lock or rcu_read_lock
375 * must be held for the entire duration that the name is in use.
377 const char *vport_get_name(const struct vport *vport)
379 return vport->ops->get_name(vport);
383 * vport_get_type - retrieve device type
385 * @vport: vport from which to retrieve the type.
387 * Retrieves the type of the given device.
389 enum ovs_vport_type vport_get_type(const struct vport *vport)
391 return vport->ops->type;
395 * vport_get_addr - retrieve device Ethernet address (for kernel callers)
397 * @vport: vport from which to retrieve the Ethernet address.
399 * Retrieves the Ethernet address of the given device. Either RTNL lock or
400 * rcu_read_lock must be held for the entire duration that the Ethernet address
403 const unsigned char *vport_get_addr(const struct vport *vport)
405 return vport->ops->get_addr(vport);
409 * vport_get_kobj - retrieve associated kobj
411 * @vport: vport from which to retrieve the associated kobj
413 * Retrieves the associated kobj or null if no kobj. The returned kobj is
414 * valid for as long as the vport exists.
416 struct kobject *vport_get_kobj(const struct vport *vport)
418 if (vport->ops->get_kobj)
419 return vport->ops->get_kobj(vport);
424 static int vport_call_get_stats(struct vport *vport, struct rtnl_link_stats64 *stats)
429 err = vport->ops->get_stats(vport, stats);
436 * vport_get_stats - retrieve device stats
438 * @vport: vport from which to retrieve the stats
439 * @stats: location to store stats
441 * Retrieves transmit, receive, and error stats for the given device.
443 * Must be called with RTNL lock or rcu_read_lock.
445 int vport_get_stats(struct vport *vport, struct rtnl_link_stats64 *stats)
449 if (!(vport->ops->flags & VPORT_F_GEN_STATS))
450 return vport_call_get_stats(vport, stats);
452 /* We potentially have 3 sources of stats that need to be
453 * combined: those we have collected (split into err_stats and
454 * percpu_stats), offset_stats from set_stats(), and device
455 * error stats from get_stats() (for errors that happen
456 * downstream and therefore aren't reported through our
457 * vport_record_error() function). */
459 spin_lock_bh(&vport->stats_lock);
461 *stats = vport->offset_stats;
463 stats->rx_errors += vport->err_stats.rx_errors;
464 stats->tx_errors += vport->err_stats.tx_errors;
465 stats->tx_dropped += vport->err_stats.tx_dropped;
466 stats->rx_dropped += vport->err_stats.rx_dropped;
468 spin_unlock_bh(&vport->stats_lock);
470 if (vport->ops->get_stats) {
471 struct rtnl_link_stats64 dev_stats;
474 err = vport_call_get_stats(vport, &dev_stats);
478 stats->rx_errors += dev_stats.rx_errors;
479 stats->tx_errors += dev_stats.tx_errors;
480 stats->rx_dropped += dev_stats.rx_dropped;
481 stats->tx_dropped += dev_stats.tx_dropped;
482 stats->multicast += dev_stats.multicast;
483 stats->collisions += dev_stats.collisions;
484 stats->rx_length_errors += dev_stats.rx_length_errors;
485 stats->rx_over_errors += dev_stats.rx_over_errors;
486 stats->rx_crc_errors += dev_stats.rx_crc_errors;
487 stats->rx_frame_errors += dev_stats.rx_frame_errors;
488 stats->rx_fifo_errors += dev_stats.rx_fifo_errors;
489 stats->rx_missed_errors += dev_stats.rx_missed_errors;
490 stats->tx_aborted_errors += dev_stats.tx_aborted_errors;
491 stats->tx_carrier_errors += dev_stats.tx_carrier_errors;
492 stats->tx_fifo_errors += dev_stats.tx_fifo_errors;
493 stats->tx_heartbeat_errors += dev_stats.tx_heartbeat_errors;
494 stats->tx_window_errors += dev_stats.tx_window_errors;
495 stats->rx_compressed += dev_stats.rx_compressed;
496 stats->tx_compressed += dev_stats.tx_compressed;
499 for_each_possible_cpu(i) {
500 const struct vport_percpu_stats *percpu_stats;
501 struct vport_percpu_stats local_stats;
504 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
507 seqcount = read_seqcount_begin(&percpu_stats->seqlock);
508 local_stats = *percpu_stats;
509 } while (read_seqcount_retry(&percpu_stats->seqlock, seqcount));
511 stats->rx_bytes += local_stats.rx_bytes;
512 stats->rx_packets += local_stats.rx_packets;
513 stats->tx_bytes += local_stats.tx_bytes;
514 stats->tx_packets += local_stats.tx_packets;
521 * vport_get_flags - retrieve device flags
523 * @vport: vport from which to retrieve the flags
525 * Retrieves the flags of the given device.
527 * Must be called with RTNL lock or rcu_read_lock.
529 unsigned vport_get_flags(const struct vport *vport)
531 return vport->ops->get_dev_flags(vport);
535 * vport_get_flags - check whether device is running
537 * @vport: vport on which to check status.
539 * Checks whether the given device is running.
541 * Must be called with RTNL lock or rcu_read_lock.
543 int vport_is_running(const struct vport *vport)
545 return vport->ops->is_running(vport);
549 * vport_get_flags - retrieve device operating state
551 * @vport: vport from which to check status
553 * Retrieves the RFC2863 operstate of the given device.
555 * Must be called with RTNL lock or rcu_read_lock.
557 unsigned char vport_get_operstate(const struct vport *vport)
559 return vport->ops->get_operstate(vport);
563 * vport_get_ifindex - retrieve device system interface index
565 * @vport: vport from which to retrieve index
567 * Retrieves the system interface index of the given device or 0 if
568 * the device does not have one (in the case of virtual ports).
569 * Returns a negative index on error.
571 * Must be called with RTNL lock or rcu_read_lock.
573 int vport_get_ifindex(const struct vport *vport)
575 if (vport->ops->get_ifindex)
576 return vport->ops->get_ifindex(vport);
582 * vport_get_iflink - retrieve device system link index
584 * @vport: vport from which to retrieve index
586 * Retrieves the system link index of the given device. The link is the index
587 * of the interface on which the packet will actually be sent. In most cases
588 * this is the same as the ifindex but may be different for tunnel devices.
589 * Returns a negative index on error.
591 * Must be called with RTNL lock or rcu_read_lock.
593 int vport_get_iflink(const struct vport *vport)
595 if (vport->ops->get_iflink)
596 return vport->ops->get_iflink(vport);
598 /* If we don't have an iflink, use the ifindex. In most cases they
600 return vport_get_ifindex(vport);
604 * vport_get_mtu - retrieve device MTU
606 * @vport: vport from which to retrieve MTU
608 * Retrieves the MTU of the given device. Returns 0 if @vport does not have an
609 * MTU (as e.g. some tunnels do not). Either RTNL lock or rcu_read_lock must
612 int vport_get_mtu(const struct vport *vport)
614 if (!vport->ops->get_mtu)
616 return vport->ops->get_mtu(vport);
620 * vport_get_options - retrieve device options
622 * @vport: vport from which to retrieve the options.
623 * @skb: sk_buff where options should be appended.
625 * Retrieves the configuration of the given device, appending an
626 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
627 * vport-specific attributes to @skb.
629 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
630 * negative error code if a real error occurred. If an error occurs, @skb is
633 * Must be called with RTNL lock or rcu_read_lock.
635 int vport_get_options(const struct vport *vport, struct sk_buff *skb)
639 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
643 if (vport->ops->get_options) {
644 int err = vport->ops->get_options(vport, skb);
646 nla_nest_cancel(skb, nla);
651 nla_nest_end(skb, nla);
656 * vport_receive - pass up received packet to the datapath for processing
658 * @vport: vport that received the packet
659 * @skb: skb that was received
661 * Must be called with rcu_read_lock. The packet cannot be shared and
662 * skb->data should point to the Ethernet header. The caller must have already
663 * called compute_ip_summed() to initialize the checksumming fields.
665 void vport_receive(struct vport *vport, struct sk_buff *skb)
667 if (vport->ops->flags & VPORT_F_GEN_STATS) {
668 struct vport_percpu_stats *stats;
671 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
673 write_seqcount_begin(&stats->seqlock);
675 stats->rx_bytes += skb->len;
676 write_seqcount_end(&stats->seqlock);
681 if (!(vport->ops->flags & VPORT_F_FLOW))
682 OVS_CB(skb)->flow = NULL;
684 if (!(vport->ops->flags & VPORT_F_TUN_ID))
685 OVS_CB(skb)->tun_id = 0;
687 dp_process_received_packet(vport, skb);
690 static inline unsigned packet_length(const struct sk_buff *skb)
692 unsigned length = skb->len - ETH_HLEN;
694 if (skb->protocol == htons(ETH_P_8021Q))
701 * vport_send - send a packet on a device
703 * @vport: vport on which to send the packet
706 * Sends the given packet and returns the length of data sent. Either RTNL
707 * lock or rcu_read_lock must be held.
709 int vport_send(struct vport *vport, struct sk_buff *skb)
714 mtu = vport_get_mtu(vport);
715 if (mtu && unlikely(packet_length(skb) > mtu && !skb_is_gso(skb))) {
717 pr_warn("%s: dropped over-mtu packet: %d > %d\n",
718 dp_name(vport->dp), packet_length(skb), mtu);
722 sent = vport->ops->send(vport, skb);
724 if (vport->ops->flags & VPORT_F_GEN_STATS && sent > 0) {
725 struct vport_percpu_stats *stats;
728 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
730 write_seqcount_begin(&stats->seqlock);
732 stats->tx_bytes += sent;
733 write_seqcount_end(&stats->seqlock);
742 vport_record_error(vport, VPORT_E_TX_DROPPED);
747 * vport_record_error - indicate device error to generic stats layer
749 * @vport: vport that encountered the error
750 * @err_type: one of enum vport_err_type types to indicate the error type
752 * If using the vport generic stats layer indicate that an error of the given
755 void vport_record_error(struct vport *vport, enum vport_err_type err_type)
757 if (vport->ops->flags & VPORT_F_GEN_STATS) {
759 spin_lock_bh(&vport->stats_lock);
762 case VPORT_E_RX_DROPPED:
763 vport->err_stats.rx_dropped++;
766 case VPORT_E_RX_ERROR:
767 vport->err_stats.rx_errors++;
770 case VPORT_E_TX_DROPPED:
771 vport->err_stats.tx_dropped++;
774 case VPORT_E_TX_ERROR:
775 vport->err_stats.tx_errors++;
779 spin_unlock_bh(&vport->stats_lock);