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 #include <linux/dcache.h>
10 #include <linux/etherdevice.h>
12 #include <linux/if_vlan.h>
13 #include <linux/kernel.h>
14 #include <linux/list.h>
15 #include <linux/mutex.h>
16 #include <linux/percpu.h>
17 #include <linux/rcupdate.h>
18 #include <linux/rtnetlink.h>
19 #include <linux/compat.h>
20 #include <linux/version.h>
23 #include "vport-internal_dev.h"
25 /* List of statically compiled vport implementations. Don't forget to also
26 * add yours to the list at the bottom of vport.h. */
27 static const struct vport_ops *base_vport_ops_list[] = {
32 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
37 static const struct vport_ops **vport_ops_list;
38 static int n_vport_types;
40 /* Protected by RCU read lock for reading, RTNL lock for writing. */
41 static struct hlist_head *dev_table;
42 #define VPORT_HASH_BUCKETS 1024
45 * vport_init - initialize vport subsystem
47 * Called at module load time to initialize the vport subsystem and any
48 * compiled in vport types.
55 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
62 vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
63 sizeof(struct vport_ops *), GFP_KERNEL);
64 if (!vport_ops_list) {
69 for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
70 const struct vport_ops *new_ops = base_vport_ops_list[i];
73 err = new_ops->init();
78 vport_ops_list[n_vport_types++] = new_ops;
79 else if (new_ops->flags & VPORT_F_REQUIRED) {
94 * vport_exit - shutdown vport subsystem
96 * Called at module exit time to shutdown the vport subsystem and any
97 * initialized vport types.
103 for (i = 0; i < n_vport_types; i++) {
104 if (vport_ops_list[i]->exit)
105 vport_ops_list[i]->exit();
108 kfree(vport_ops_list);
112 static struct hlist_head *hash_bucket(const char *name)
114 unsigned int hash = full_name_hash(name, strlen(name));
115 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
119 * vport_locate - find a port that has already been created
121 * @name: name of port to find
123 * Must be called with RTNL or RCU read lock.
125 struct vport *vport_locate(const char *name)
127 struct hlist_head *bucket = hash_bucket(name);
129 struct hlist_node *node;
131 hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
132 if (!strcmp(name, vport_get_name(vport)))
138 static void release_vport(struct kobject *kobj)
140 struct vport *p = container_of(kobj, struct vport, kobj);
144 static struct kobj_type brport_ktype = {
146 .sysfs_ops = &brport_sysfs_ops,
148 .release = release_vport
152 * vport_alloc - allocate and initialize new vport
154 * @priv_size: Size of private data area to allocate.
155 * @ops: vport device ops
157 * Allocate and initialize a new vport defined by @ops. The vport will contain
158 * a private data area of size @priv_size that can be accessed using
159 * vport_priv(). vports that are no longer needed should be released with
162 struct vport *vport_alloc(int priv_size, const struct vport_ops *ops, const struct vport_parms *parms)
167 alloc_size = sizeof(struct vport);
169 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
170 alloc_size += priv_size;
173 vport = kzalloc(alloc_size, GFP_KERNEL);
175 return ERR_PTR(-ENOMEM);
177 vport->dp = parms->dp;
178 vport->port_no = parms->port_no;
179 atomic_set(&vport->sflow_pool, 0);
182 /* Initialize kobject for bridge. This will be added as
183 * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
184 vport->kobj.kset = NULL;
185 kobject_init(&vport->kobj, &brport_ktype);
187 if (vport->ops->flags & VPORT_F_GEN_STATS) {
188 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
189 if (!vport->percpu_stats)
190 return ERR_PTR(-ENOMEM);
192 spin_lock_init(&vport->stats_lock);
199 * vport_free - uninitialize and free vport
201 * @vport: vport to free
203 * Frees a vport allocated with vport_alloc() when it is no longer needed.
205 * The caller must ensure that an RCU grace period has passed since the last
206 * time @vport was in a datapath.
208 void vport_free(struct vport *vport)
210 if (vport->ops->flags & VPORT_F_GEN_STATS)
211 free_percpu(vport->percpu_stats);
213 kobject_put(&vport->kobj);
217 * vport_add - add vport device (for kernel callers)
219 * @parms: Information about new vport.
221 * Creates a new vport with the specified configuration (which is dependent on
222 * device type). RTNL lock must be held.
224 struct vport *vport_add(const struct vport_parms *parms)
232 for (i = 0; i < n_vport_types; i++) {
233 if (vport_ops_list[i]->type == parms->type) {
234 vport = vport_ops_list[i]->create(parms);
236 err = PTR_ERR(vport);
240 hlist_add_head_rcu(&vport->hash_node,
241 hash_bucket(vport_get_name(vport)));
253 * vport_set_options - modify existing vport device (for kernel callers)
255 * @vport: vport to modify.
256 * @port: New configuration.
258 * Modifies an existing device with the specified configuration (which is
259 * dependent on device type). RTNL lock must be held.
261 int vport_set_options(struct vport *vport, struct nlattr *options)
265 if (!vport->ops->set_options)
267 return vport->ops->set_options(vport, options);
271 * vport_del - delete existing vport device
273 * @vport: vport to delete.
275 * Detaches @vport from its datapath and destroys it. It is possible to fail
276 * for reasons such as lack of memory. RTNL lock must be held.
278 int vport_del(struct vport *vport)
282 hlist_del_rcu(&vport->hash_node);
284 return vport->ops->destroy(vport);
288 * vport_set_mtu - set device MTU (for kernel callers)
290 * @vport: vport on which to set MTU.
293 * Sets the MTU of the given device. Some devices may not support setting the
294 * MTU, in which case the result will always be -EOPNOTSUPP. RTNL lock must
297 int vport_set_mtu(struct vport *vport, int mtu)
304 if (vport->ops->set_mtu) {
307 ret = vport->ops->set_mtu(vport, mtu);
309 if (!ret && !is_internal_vport(vport))
310 set_internal_devs_mtu(vport->dp);
318 * vport_set_addr - set device Ethernet address (for kernel callers)
320 * @vport: vport on which to set Ethernet address.
321 * @addr: New address.
323 * Sets the Ethernet address of the given device. Some devices may not support
324 * setting the Ethernet address, in which case the result will always be
325 * -EOPNOTSUPP. RTNL lock must be held.
327 int vport_set_addr(struct vport *vport, const unsigned char *addr)
331 if (!is_valid_ether_addr(addr))
332 return -EADDRNOTAVAIL;
334 if (vport->ops->set_addr)
335 return vport->ops->set_addr(vport, addr);
341 * vport_set_stats - sets offset device stats
343 * @vport: vport on which to set stats
344 * @stats: stats to set
346 * Provides a set of transmit, receive, and error stats to be added as an
347 * offset to the collect data when stats are retreived. Some devices may not
348 * support setting the stats, in which case the result will always be
351 * Must be called with RTNL lock.
353 int vport_set_stats(struct vport *vport, struct rtnl_link_stats64 *stats)
357 if (vport->ops->flags & VPORT_F_GEN_STATS) {
358 spin_lock_bh(&vport->stats_lock);
359 vport->offset_stats = *stats;
360 spin_unlock_bh(&vport->stats_lock);
368 * vport_get_name - retrieve device name
370 * @vport: vport from which to retrieve the name.
372 * Retrieves the name of the given device. Either RTNL lock or rcu_read_lock
373 * must be held for the entire duration that the name is in use.
375 const char *vport_get_name(const struct vport *vport)
377 return vport->ops->get_name(vport);
381 * vport_get_type - retrieve device type
383 * @vport: vport from which to retrieve the type.
385 * Retrieves the type of the given device.
387 enum ovs_vport_type vport_get_type(const struct vport *vport)
389 return vport->ops->type;
393 * vport_get_addr - retrieve device Ethernet address (for kernel callers)
395 * @vport: vport from which to retrieve the Ethernet address.
397 * Retrieves the Ethernet address of the given device. Either RTNL lock or
398 * rcu_read_lock must be held for the entire duration that the Ethernet address
401 const unsigned char *vport_get_addr(const struct vport *vport)
403 return vport->ops->get_addr(vport);
407 * vport_get_kobj - retrieve associated kobj
409 * @vport: vport from which to retrieve the associated kobj
411 * Retrieves the associated kobj or null if no kobj. The returned kobj is
412 * valid for as long as the vport exists.
414 struct kobject *vport_get_kobj(const struct vport *vport)
416 if (vport->ops->get_kobj)
417 return vport->ops->get_kobj(vport);
422 static int vport_call_get_stats(struct vport *vport, struct rtnl_link_stats64 *stats)
427 err = vport->ops->get_stats(vport, stats);
434 * vport_get_stats - retrieve device stats
436 * @vport: vport from which to retrieve the stats
437 * @stats: location to store stats
439 * Retrieves transmit, receive, and error stats for the given device.
441 * Must be called with RTNL lock or rcu_read_lock.
443 int vport_get_stats(struct vport *vport, struct rtnl_link_stats64 *stats)
447 if (!(vport->ops->flags & VPORT_F_GEN_STATS))
448 return vport_call_get_stats(vport, stats);
450 /* We potentially have 3 sources of stats that need to be
451 * combined: those we have collected (split into err_stats and
452 * percpu_stats), offset_stats from set_stats(), and device
453 * error stats from get_stats() (for errors that happen
454 * downstream and therefore aren't reported through our
455 * vport_record_error() function). */
457 spin_lock_bh(&vport->stats_lock);
459 *stats = vport->offset_stats;
461 stats->rx_errors += vport->err_stats.rx_errors;
462 stats->tx_errors += vport->err_stats.tx_errors;
463 stats->tx_dropped += vport->err_stats.tx_dropped;
464 stats->rx_dropped += vport->err_stats.rx_dropped;
466 spin_unlock_bh(&vport->stats_lock);
468 if (vport->ops->get_stats) {
469 struct rtnl_link_stats64 dev_stats;
472 err = vport_call_get_stats(vport, &dev_stats);
476 stats->rx_errors += dev_stats.rx_errors;
477 stats->tx_errors += dev_stats.tx_errors;
478 stats->rx_dropped += dev_stats.rx_dropped;
479 stats->tx_dropped += dev_stats.tx_dropped;
480 stats->multicast += dev_stats.multicast;
481 stats->collisions += dev_stats.collisions;
482 stats->rx_length_errors += dev_stats.rx_length_errors;
483 stats->rx_over_errors += dev_stats.rx_over_errors;
484 stats->rx_crc_errors += dev_stats.rx_crc_errors;
485 stats->rx_frame_errors += dev_stats.rx_frame_errors;
486 stats->rx_fifo_errors += dev_stats.rx_fifo_errors;
487 stats->rx_missed_errors += dev_stats.rx_missed_errors;
488 stats->tx_aborted_errors += dev_stats.tx_aborted_errors;
489 stats->tx_carrier_errors += dev_stats.tx_carrier_errors;
490 stats->tx_fifo_errors += dev_stats.tx_fifo_errors;
491 stats->tx_heartbeat_errors += dev_stats.tx_heartbeat_errors;
492 stats->tx_window_errors += dev_stats.tx_window_errors;
493 stats->rx_compressed += dev_stats.rx_compressed;
494 stats->tx_compressed += dev_stats.tx_compressed;
497 for_each_possible_cpu(i) {
498 const struct vport_percpu_stats *percpu_stats;
499 struct vport_percpu_stats local_stats;
502 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
505 seqcount = read_seqcount_begin(&percpu_stats->seqlock);
506 local_stats = *percpu_stats;
507 } while (read_seqcount_retry(&percpu_stats->seqlock, seqcount));
509 stats->rx_bytes += local_stats.rx_bytes;
510 stats->rx_packets += local_stats.rx_packets;
511 stats->tx_bytes += local_stats.tx_bytes;
512 stats->tx_packets += local_stats.tx_packets;
519 * vport_get_flags - retrieve device flags
521 * @vport: vport from which to retrieve the flags
523 * Retrieves the flags of the given device.
525 * Must be called with RTNL lock or rcu_read_lock.
527 unsigned vport_get_flags(const struct vport *vport)
529 return vport->ops->get_dev_flags(vport);
533 * vport_get_flags - check whether device is running
535 * @vport: vport on which to check status.
537 * Checks whether the given device is running.
539 * Must be called with RTNL lock or rcu_read_lock.
541 int vport_is_running(const struct vport *vport)
543 return vport->ops->is_running(vport);
547 * vport_get_flags - retrieve device operating state
549 * @vport: vport from which to check status
551 * Retrieves the RFC2863 operstate of the given device.
553 * Must be called with RTNL lock or rcu_read_lock.
555 unsigned char vport_get_operstate(const struct vport *vport)
557 return vport->ops->get_operstate(vport);
561 * vport_get_ifindex - retrieve device system interface index
563 * @vport: vport from which to retrieve index
565 * Retrieves the system interface index of the given device or 0 if
566 * the device does not have one (in the case of virtual ports).
567 * Returns a negative index on error.
569 * Must be called with RTNL lock or rcu_read_lock.
571 int vport_get_ifindex(const struct vport *vport)
573 if (vport->ops->get_ifindex)
574 return vport->ops->get_ifindex(vport);
580 * vport_get_iflink - retrieve device system link index
582 * @vport: vport from which to retrieve index
584 * Retrieves the system link index of the given device. The link is the index
585 * of the interface on which the packet will actually be sent. In most cases
586 * this is the same as the ifindex but may be different for tunnel devices.
587 * Returns a negative index on error.
589 * Must be called with RTNL lock or rcu_read_lock.
591 int vport_get_iflink(const struct vport *vport)
593 if (vport->ops->get_iflink)
594 return vport->ops->get_iflink(vport);
596 /* If we don't have an iflink, use the ifindex. In most cases they
598 return vport_get_ifindex(vport);
602 * vport_get_mtu - retrieve device MTU
604 * @vport: vport from which to retrieve MTU
606 * Retrieves the MTU of the given device. Returns 0 if @vport does not have an
607 * MTU (as e.g. some tunnels do not). Either RTNL lock or rcu_read_lock must
610 int vport_get_mtu(const struct vport *vport)
612 if (!vport->ops->get_mtu)
614 return vport->ops->get_mtu(vport);
618 * vport_get_options - retrieve device options
620 * @vport: vport from which to retrieve the options.
621 * @skb: sk_buff where options should be appended.
623 * Retrieves the configuration of the given device, appending an
624 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
625 * vport-specific attributes to @skb.
627 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
628 * negative error code if a real error occurred. If an error occurs, @skb is
631 * Must be called with RTNL lock or rcu_read_lock.
633 int vport_get_options(const struct vport *vport, struct sk_buff *skb)
637 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
641 if (vport->ops->get_options) {
642 int err = vport->ops->get_options(vport, skb);
644 nla_nest_cancel(skb, nla);
649 nla_nest_end(skb, nla);
654 * vport_receive - pass up received packet to the datapath for processing
656 * @vport: vport that received the packet
657 * @skb: skb that was received
659 * Must be called with rcu_read_lock. The packet cannot be shared and
660 * skb->data should point to the Ethernet header. The caller must have already
661 * called compute_ip_summed() to initialize the checksumming fields.
663 void vport_receive(struct vport *vport, struct sk_buff *skb)
665 if (vport->ops->flags & VPORT_F_GEN_STATS) {
666 struct vport_percpu_stats *stats;
669 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
671 write_seqcount_begin(&stats->seqlock);
673 stats->rx_bytes += skb->len;
674 write_seqcount_end(&stats->seqlock);
679 if (!(vport->ops->flags & VPORT_F_FLOW))
680 OVS_CB(skb)->flow = NULL;
682 if (!(vport->ops->flags & VPORT_F_TUN_ID))
683 OVS_CB(skb)->tun_id = 0;
685 dp_process_received_packet(vport, skb);
689 * vport_send - send a packet on a device
691 * @vport: vport on which to send the packet
694 * Sends the given packet and returns the length of data sent. Either RTNL
695 * lock or rcu_read_lock must be held.
697 int vport_send(struct vport *vport, struct sk_buff *skb)
699 int sent = vport->ops->send(vport, skb);
701 if (vport->ops->flags & VPORT_F_GEN_STATS && sent > 0) {
702 struct vport_percpu_stats *stats;
705 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
707 write_seqcount_begin(&stats->seqlock);
709 stats->tx_bytes += sent;
710 write_seqcount_end(&stats->seqlock);
719 * vport_record_error - indicate device error to generic stats layer
721 * @vport: vport that encountered the error
722 * @err_type: one of enum vport_err_type types to indicate the error type
724 * If using the vport generic stats layer indicate that an error of the given
727 void vport_record_error(struct vport *vport, enum vport_err_type err_type)
729 if (vport->ops->flags & VPORT_F_GEN_STATS) {
731 spin_lock_bh(&vport->stats_lock);
734 case VPORT_E_RX_DROPPED:
735 vport->err_stats.rx_dropped++;
738 case VPORT_E_RX_ERROR:
739 vport->err_stats.rx_errors++;
742 case VPORT_E_TX_DROPPED:
743 vport->err_stats.tx_dropped++;
746 case VPORT_E_TX_ERROR:
747 vport->err_stats.tx_errors++;
751 spin_unlock_bh(&vport->stats_lock);