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 vport->upcall_pid = parms->upcall_pid;
180 atomic_set(&vport->sflow_pool, 0);
183 /* Initialize kobject for bridge. This will be added as
184 * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
185 vport->kobj.kset = NULL;
186 kobject_init(&vport->kobj, &brport_ktype);
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);
198 * vport_free - uninitialize and free vport
200 * @vport: vport to free
202 * Frees a vport allocated with vport_alloc() when it is no longer needed.
204 * The caller must ensure that an RCU grace period has passed since the last
205 * time @vport was in a datapath.
207 void vport_free(struct vport *vport)
209 free_percpu(vport->percpu_stats);
211 kobject_put(&vport->kobj);
215 * vport_add - add vport device (for kernel callers)
217 * @parms: Information about new vport.
219 * Creates a new vport with the specified configuration (which is dependent on
220 * device type). RTNL lock must be held.
222 struct vport *vport_add(const struct vport_parms *parms)
230 for (i = 0; i < n_vport_types; i++) {
231 if (vport_ops_list[i]->type == parms->type) {
232 vport = vport_ops_list[i]->create(parms);
234 err = PTR_ERR(vport);
238 hlist_add_head_rcu(&vport->hash_node,
239 hash_bucket(vport_get_name(vport)));
251 * vport_set_options - modify existing vport device (for kernel callers)
253 * @vport: vport to modify.
254 * @port: New configuration.
256 * Modifies an existing device with the specified configuration (which is
257 * dependent on device type). RTNL lock must be held.
259 int vport_set_options(struct vport *vport, struct nlattr *options)
263 if (!vport->ops->set_options)
265 return vport->ops->set_options(vport, options);
269 * vport_del - delete existing vport device
271 * @vport: vport to delete.
273 * Detaches @vport from its datapath and destroys it. It is possible to fail
274 * for reasons such as lack of memory. RTNL lock must be held.
276 void vport_del(struct vport *vport)
280 hlist_del_rcu(&vport->hash_node);
282 vport->ops->destroy(vport);
286 * vport_set_addr - set device Ethernet address (for kernel callers)
288 * @vport: vport on which to set Ethernet address.
289 * @addr: New address.
291 * Sets the Ethernet address of the given device. Some devices may not support
292 * setting the Ethernet address, in which case the result will always be
293 * -EOPNOTSUPP. RTNL lock must be held.
295 int vport_set_addr(struct vport *vport, const unsigned char *addr)
299 if (!is_valid_ether_addr(addr))
300 return -EADDRNOTAVAIL;
302 if (vport->ops->set_addr)
303 return vport->ops->set_addr(vport, addr);
309 * vport_set_stats - sets offset device stats
311 * @vport: vport on which to set stats
312 * @stats: stats to set
314 * Provides a set of transmit, receive, and error stats to be added as an
315 * offset to the collect data when stats are retreived. Some devices may not
316 * support setting the stats, in which case the result will always be
319 * Must be called with RTNL lock.
321 void vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
325 spin_lock_bh(&vport->stats_lock);
326 vport->offset_stats = *stats;
327 spin_unlock_bh(&vport->stats_lock);
331 * vport_get_name - retrieve device name
333 * @vport: vport from which to retrieve the name.
335 * Retrieves the name of the given device. Either RTNL lock or rcu_read_lock
336 * must be held for the entire duration that the name is in use.
338 const char *vport_get_name(const struct vport *vport)
340 return vport->ops->get_name(vport);
344 * vport_get_type - retrieve device type
346 * @vport: vport from which to retrieve the type.
348 * Retrieves the type of the given device.
350 enum ovs_vport_type vport_get_type(const struct vport *vport)
352 return vport->ops->type;
356 * vport_get_addr - retrieve device Ethernet address (for kernel callers)
358 * @vport: vport from which to retrieve the Ethernet address.
360 * Retrieves the Ethernet address of the given device. Either RTNL lock or
361 * rcu_read_lock must be held for the entire duration that the Ethernet address
364 const unsigned char *vport_get_addr(const struct vport *vport)
366 return vport->ops->get_addr(vport);
370 * vport_get_kobj - retrieve associated kobj
372 * @vport: vport from which to retrieve the associated kobj
374 * Retrieves the associated kobj or null if no kobj. The returned kobj is
375 * valid for as long as the vport exists.
377 struct kobject *vport_get_kobj(const struct vport *vport)
379 if (vport->ops->get_kobj)
380 return vport->ops->get_kobj(vport);
386 * vport_get_stats - retrieve device stats
388 * @vport: vport from which to retrieve the stats
389 * @stats: location to store stats
391 * Retrieves transmit, receive, and error stats for the given device.
393 * Must be called with RTNL lock or rcu_read_lock.
395 void vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
399 /* We potentially have 3 sources of stats that need to be
400 * combined: those we have collected (split into err_stats and
401 * percpu_stats), offset_stats from set_stats(), and device
402 * error stats from netdev->get_stats() (for errors that happen
403 * downstream and therefore aren't reported through our
404 * vport_record_error() function).
405 * Stats from first two sources are merged and reported by ovs over
406 * OVS_VPORT_ATTR_STATS.
407 * netdev-stats can be directly read over netlink-ioctl.
410 spin_lock_bh(&vport->stats_lock);
412 *stats = vport->offset_stats;
414 stats->rx_errors += vport->err_stats.rx_errors;
415 stats->tx_errors += vport->err_stats.tx_errors;
416 stats->tx_dropped += vport->err_stats.tx_dropped;
417 stats->rx_dropped += vport->err_stats.rx_dropped;
419 spin_unlock_bh(&vport->stats_lock);
421 for_each_possible_cpu(i) {
422 const struct vport_percpu_stats *percpu_stats;
423 struct vport_percpu_stats local_stats;
426 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
429 seqcount = read_seqcount_begin(&percpu_stats->seqlock);
430 local_stats = *percpu_stats;
431 } while (read_seqcount_retry(&percpu_stats->seqlock, seqcount));
433 stats->rx_bytes += local_stats.rx_bytes;
434 stats->rx_packets += local_stats.rx_packets;
435 stats->tx_bytes += local_stats.tx_bytes;
436 stats->tx_packets += local_stats.tx_packets;
441 * vport_get_flags - retrieve device flags
443 * @vport: vport from which to retrieve the flags
445 * Retrieves the flags of the given device.
447 * Must be called with RTNL lock or rcu_read_lock.
449 unsigned vport_get_flags(const struct vport *vport)
451 return vport->ops->get_dev_flags(vport);
455 * vport_get_flags - check whether device is running
457 * @vport: vport on which to check status.
459 * Checks whether the given device is running.
461 * Must be called with RTNL lock or rcu_read_lock.
463 int vport_is_running(const struct vport *vport)
465 return vport->ops->is_running(vport);
469 * vport_get_flags - retrieve device operating state
471 * @vport: vport from which to check status
473 * Retrieves the RFC2863 operstate of the given device.
475 * Must be called with RTNL lock or rcu_read_lock.
477 unsigned char vport_get_operstate(const struct vport *vport)
479 return vport->ops->get_operstate(vport);
483 * vport_get_ifindex - retrieve device system interface index
485 * @vport: vport from which to retrieve index
487 * Retrieves the system interface index of the given device or 0 if
488 * the device does not have one (in the case of virtual ports).
489 * Returns a negative index on error.
491 * Must be called with RTNL lock or rcu_read_lock.
493 int vport_get_ifindex(const struct vport *vport)
495 if (vport->ops->get_ifindex)
496 return vport->ops->get_ifindex(vport);
502 * vport_get_mtu - retrieve device MTU
504 * @vport: vport from which to retrieve MTU
506 * Retrieves the MTU of the given device. Returns 0 if @vport does not have an
507 * MTU (as e.g. some tunnels do not). Either RTNL lock or rcu_read_lock must
510 int vport_get_mtu(const struct vport *vport)
512 if (!vport->ops->get_mtu)
514 return vport->ops->get_mtu(vport);
518 * vport_get_options - retrieve device options
520 * @vport: vport from which to retrieve the options.
521 * @skb: sk_buff where options should be appended.
523 * Retrieves the configuration of the given device, appending an
524 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
525 * vport-specific attributes to @skb.
527 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
528 * negative error code if a real error occurred. If an error occurs, @skb is
531 * Must be called with RTNL lock or rcu_read_lock.
533 int vport_get_options(const struct vport *vport, struct sk_buff *skb)
537 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
541 if (vport->ops->get_options) {
542 int err = vport->ops->get_options(vport, skb);
544 nla_nest_cancel(skb, nla);
549 nla_nest_end(skb, nla);
554 * vport_receive - pass up received packet to the datapath for processing
556 * @vport: vport that received the packet
557 * @skb: skb that was received
559 * Must be called with rcu_read_lock. The packet cannot be shared and
560 * skb->data should point to the Ethernet header. The caller must have already
561 * called compute_ip_summed() to initialize the checksumming fields.
563 void vport_receive(struct vport *vport, struct sk_buff *skb)
565 struct vport_percpu_stats *stats;
568 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
570 write_seqcount_begin(&stats->seqlock);
572 stats->rx_bytes += skb->len;
573 write_seqcount_end(&stats->seqlock);
577 if (!(vport->ops->flags & VPORT_F_FLOW))
578 OVS_CB(skb)->flow = NULL;
580 if (!(vport->ops->flags & VPORT_F_TUN_ID))
581 OVS_CB(skb)->tun_id = 0;
583 dp_process_received_packet(vport, skb);
587 * vport_send - send a packet on a device
589 * @vport: vport on which to send the packet
592 * Sends the given packet and returns the length of data sent. Either RTNL
593 * lock or rcu_read_lock must be held.
595 int vport_send(struct vport *vport, struct sk_buff *skb)
597 struct vport_percpu_stats *stats;
598 int sent = vport->ops->send(vport, skb);
601 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
603 write_seqcount_begin(&stats->seqlock);
605 stats->tx_bytes += sent;
606 write_seqcount_end(&stats->seqlock);
614 * vport_record_error - indicate device error to generic stats layer
616 * @vport: vport that encountered the error
617 * @err_type: one of enum vport_err_type types to indicate the error type
619 * If using the vport generic stats layer indicate that an error of the given
622 void vport_record_error(struct vport *vport, enum vport_err_type err_type)
624 spin_lock_bh(&vport->stats_lock);
627 case VPORT_E_RX_DROPPED:
628 vport->err_stats.rx_dropped++;
631 case VPORT_E_RX_ERROR:
632 vport->err_stats.rx_errors++;
635 case VPORT_E_TX_DROPPED:
636 vport->err_stats.tx_dropped++;
639 case VPORT_E_TX_ERROR:
640 vport->err_stats.tx_errors++;
644 spin_unlock_bh(&vport->stats_lock);