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 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
188 if (!vport->percpu_stats)
189 return ERR_PTR(-ENOMEM);
191 spin_lock_init(&vport->stats_lock);
197 * vport_free - uninitialize and free vport
199 * @vport: vport to free
201 * Frees a vport allocated with vport_alloc() when it is no longer needed.
203 * The caller must ensure that an RCU grace period has passed since the last
204 * time @vport was in a datapath.
206 void vport_free(struct vport *vport)
208 free_percpu(vport->percpu_stats);
210 kobject_put(&vport->kobj);
214 * vport_add - add vport device (for kernel callers)
216 * @parms: Information about new vport.
218 * Creates a new vport with the specified configuration (which is dependent on
219 * device type). RTNL lock must be held.
221 struct vport *vport_add(const struct vport_parms *parms)
229 for (i = 0; i < n_vport_types; i++) {
230 if (vport_ops_list[i]->type == parms->type) {
231 vport = vport_ops_list[i]->create(parms);
233 err = PTR_ERR(vport);
237 hlist_add_head_rcu(&vport->hash_node,
238 hash_bucket(vport_get_name(vport)));
250 * vport_set_options - modify existing vport device (for kernel callers)
252 * @vport: vport to modify.
253 * @port: New configuration.
255 * Modifies an existing device with the specified configuration (which is
256 * dependent on device type). RTNL lock must be held.
258 int vport_set_options(struct vport *vport, struct nlattr *options)
262 if (!vport->ops->set_options)
264 return vport->ops->set_options(vport, options);
268 * vport_del - delete existing vport device
270 * @vport: vport to delete.
272 * Detaches @vport from its datapath and destroys it. It is possible to fail
273 * for reasons such as lack of memory. RTNL lock must be held.
275 void vport_del(struct vport *vport)
279 hlist_del_rcu(&vport->hash_node);
281 vport->ops->destroy(vport);
285 * vport_set_addr - set device Ethernet address (for kernel callers)
287 * @vport: vport on which to set Ethernet address.
288 * @addr: New address.
290 * Sets the Ethernet address of the given device. Some devices may not support
291 * setting the Ethernet address, in which case the result will always be
292 * -EOPNOTSUPP. RTNL lock must be held.
294 int vport_set_addr(struct vport *vport, const unsigned char *addr)
298 if (!is_valid_ether_addr(addr))
299 return -EADDRNOTAVAIL;
301 if (vport->ops->set_addr)
302 return vport->ops->set_addr(vport, addr);
308 * vport_set_stats - sets offset device stats
310 * @vport: vport on which to set stats
311 * @stats: stats to set
313 * Provides a set of transmit, receive, and error stats to be added as an
314 * offset to the collect data when stats are retreived. Some devices may not
315 * support setting the stats, in which case the result will always be
318 * Must be called with RTNL lock.
320 void vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
324 spin_lock_bh(&vport->stats_lock);
325 vport->offset_stats = *stats;
326 spin_unlock_bh(&vport->stats_lock);
330 * vport_get_name - retrieve device name
332 * @vport: vport from which to retrieve the name.
334 * Retrieves the name of the given device. Either RTNL lock or rcu_read_lock
335 * must be held for the entire duration that the name is in use.
337 const char *vport_get_name(const struct vport *vport)
339 return vport->ops->get_name(vport);
343 * vport_get_type - retrieve device type
345 * @vport: vport from which to retrieve the type.
347 * Retrieves the type of the given device.
349 enum ovs_vport_type vport_get_type(const struct vport *vport)
351 return vport->ops->type;
355 * vport_get_addr - retrieve device Ethernet address (for kernel callers)
357 * @vport: vport from which to retrieve the Ethernet address.
359 * Retrieves the Ethernet address of the given device. Either RTNL lock or
360 * rcu_read_lock must be held for the entire duration that the Ethernet address
363 const unsigned char *vport_get_addr(const struct vport *vport)
365 return vport->ops->get_addr(vport);
369 * vport_get_kobj - retrieve associated kobj
371 * @vport: vport from which to retrieve the associated kobj
373 * Retrieves the associated kobj or null if no kobj. The returned kobj is
374 * valid for as long as the vport exists.
376 struct kobject *vport_get_kobj(const struct vport *vport)
378 if (vport->ops->get_kobj)
379 return vport->ops->get_kobj(vport);
385 * vport_get_stats - retrieve device stats
387 * @vport: vport from which to retrieve the stats
388 * @stats: location to store stats
390 * Retrieves transmit, receive, and error stats for the given device.
392 * Must be called with RTNL lock or rcu_read_lock.
394 void vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
398 /* We potentially have 3 sources of stats that need to be
399 * combined: those we have collected (split into err_stats and
400 * percpu_stats), offset_stats from set_stats(), and device
401 * error stats from netdev->get_stats() (for errors that happen
402 * downstream and therefore aren't reported through our
403 * vport_record_error() function).
404 * Stats from first two sources are merged and reported by ovs over
405 * OVS_VPORT_ATTR_STATS.
406 * netdev-stats can be directly read over netlink-ioctl.
409 spin_lock_bh(&vport->stats_lock);
411 *stats = vport->offset_stats;
413 stats->rx_errors += vport->err_stats.rx_errors;
414 stats->tx_errors += vport->err_stats.tx_errors;
415 stats->tx_dropped += vport->err_stats.tx_dropped;
416 stats->rx_dropped += vport->err_stats.rx_dropped;
418 spin_unlock_bh(&vport->stats_lock);
420 for_each_possible_cpu(i) {
421 const struct vport_percpu_stats *percpu_stats;
422 struct vport_percpu_stats local_stats;
425 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
428 seqcount = read_seqcount_begin(&percpu_stats->seqlock);
429 local_stats = *percpu_stats;
430 } while (read_seqcount_retry(&percpu_stats->seqlock, seqcount));
432 stats->rx_bytes += local_stats.rx_bytes;
433 stats->rx_packets += local_stats.rx_packets;
434 stats->tx_bytes += local_stats.tx_bytes;
435 stats->tx_packets += local_stats.tx_packets;
440 * vport_get_flags - retrieve device flags
442 * @vport: vport from which to retrieve the flags
444 * Retrieves the flags of the given device.
446 * Must be called with RTNL lock or rcu_read_lock.
448 unsigned vport_get_flags(const struct vport *vport)
450 return vport->ops->get_dev_flags(vport);
454 * vport_get_flags - check whether device is running
456 * @vport: vport on which to check status.
458 * Checks whether the given device is running.
460 * Must be called with RTNL lock or rcu_read_lock.
462 int vport_is_running(const struct vport *vport)
464 return vport->ops->is_running(vport);
468 * vport_get_flags - retrieve device operating state
470 * @vport: vport from which to check status
472 * Retrieves the RFC2863 operstate of the given device.
474 * Must be called with RTNL lock or rcu_read_lock.
476 unsigned char vport_get_operstate(const struct vport *vport)
478 return vport->ops->get_operstate(vport);
482 * vport_get_ifindex - retrieve device system interface index
484 * @vport: vport from which to retrieve index
486 * Retrieves the system interface index of the given device or 0 if
487 * the device does not have one (in the case of virtual ports).
488 * Returns a negative index on error.
490 * Must be called with RTNL lock or rcu_read_lock.
492 int vport_get_ifindex(const struct vport *vport)
494 if (vport->ops->get_ifindex)
495 return vport->ops->get_ifindex(vport);
501 * vport_get_mtu - retrieve device MTU
503 * @vport: vport from which to retrieve MTU
505 * Retrieves the MTU of the given device. Returns 0 if @vport does not have an
506 * MTU (as e.g. some tunnels do not). Either RTNL lock or rcu_read_lock must
509 int vport_get_mtu(const struct vport *vport)
511 if (!vport->ops->get_mtu)
513 return vport->ops->get_mtu(vport);
517 * vport_get_options - retrieve device options
519 * @vport: vport from which to retrieve the options.
520 * @skb: sk_buff where options should be appended.
522 * Retrieves the configuration of the given device, appending an
523 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
524 * vport-specific attributes to @skb.
526 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
527 * negative error code if a real error occurred. If an error occurs, @skb is
530 * Must be called with RTNL lock or rcu_read_lock.
532 int vport_get_options(const struct vport *vport, struct sk_buff *skb)
536 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
540 if (vport->ops->get_options) {
541 int err = vport->ops->get_options(vport, skb);
543 nla_nest_cancel(skb, nla);
548 nla_nest_end(skb, nla);
553 * vport_receive - pass up received packet to the datapath for processing
555 * @vport: vport that received the packet
556 * @skb: skb that was received
558 * Must be called with rcu_read_lock. The packet cannot be shared and
559 * skb->data should point to the Ethernet header. The caller must have already
560 * called compute_ip_summed() to initialize the checksumming fields.
562 void vport_receive(struct vport *vport, struct sk_buff *skb)
564 struct vport_percpu_stats *stats;
567 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
569 write_seqcount_begin(&stats->seqlock);
571 stats->rx_bytes += skb->len;
572 write_seqcount_end(&stats->seqlock);
576 if (!(vport->ops->flags & VPORT_F_FLOW))
577 OVS_CB(skb)->flow = NULL;
579 if (!(vport->ops->flags & VPORT_F_TUN_ID))
580 OVS_CB(skb)->tun_id = 0;
582 dp_process_received_packet(vport, skb);
586 * vport_send - send a packet on a device
588 * @vport: vport on which to send the packet
591 * Sends the given packet and returns the length of data sent. Either RTNL
592 * lock or rcu_read_lock must be held.
594 int vport_send(struct vport *vport, struct sk_buff *skb)
596 struct vport_percpu_stats *stats;
597 int sent = vport->ops->send(vport, skb);
600 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
602 write_seqcount_begin(&stats->seqlock);
604 stats->tx_bytes += sent;
605 write_seqcount_end(&stats->seqlock);
613 * vport_record_error - indicate device error to generic stats layer
615 * @vport: vport that encountered the error
616 * @err_type: one of enum vport_err_type types to indicate the error type
618 * If using the vport generic stats layer indicate that an error of the given
621 void vport_record_error(struct vport *vport, enum vport_err_type err_type)
623 spin_lock_bh(&vport->stats_lock);
626 case VPORT_E_RX_DROPPED:
627 vport->err_stats.rx_dropped++;
630 case VPORT_E_RX_ERROR:
631 vport->err_stats.rx_errors++;
634 case VPORT_E_TX_DROPPED:
635 vport->err_stats.tx_dropped++;
638 case VPORT_E_TX_ERROR:
639 vport->err_stats.tx_errors++;
643 spin_unlock_bh(&vport->stats_lock);