2 * Copyright (c) 2007-2012 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 #include <linux/etherdevice.h>
21 #include <linux/if_vlan.h>
22 #include <linux/jhash.h>
23 #include <linux/kernel.h>
24 #include <linux/list.h>
25 #include <linux/mutex.h>
26 #include <linux/percpu.h>
27 #include <linux/rcupdate.h>
28 #include <linux/rtnetlink.h>
29 #include <linux/compat.h>
30 #include <linux/version.h>
31 #include <net/net_namespace.h>
35 #include "vport-internal_dev.h"
37 /* List of statically compiled vport implementations. Don't forget to also
38 * add yours to the list at the bottom of vport.h. */
39 static const struct vport_ops *base_vport_ops_list[] = {
40 &ovs_netdev_vport_ops,
41 &ovs_internal_vport_ops,
44 &ovs_gre_ft_vport_ops,
46 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
47 &ovs_capwap_vport_ops,
52 static const struct vport_ops **vport_ops_list;
53 static int n_vport_types;
55 /* Protected by RCU read lock for reading, RTNL lock for writing. */
56 static struct hlist_head *dev_table;
57 #define VPORT_HASH_BUCKETS 1024
60 * ovs_vport_init - initialize vport subsystem
62 * Called at module load time to initialize the vport subsystem and any
63 * compiled in vport types.
65 int ovs_vport_init(void)
70 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
77 vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
78 sizeof(struct vport_ops *), GFP_KERNEL);
79 if (!vport_ops_list) {
84 for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
85 const struct vport_ops *new_ops = base_vport_ops_list[i];
88 err = new_ops->init();
93 vport_ops_list[n_vport_types++] = new_ops;
94 else if (new_ops->flags & VPORT_F_REQUIRED) {
109 * ovs_vport_exit - shutdown vport subsystem
111 * Called at module exit time to shutdown the vport subsystem and any
112 * initialized vport types.
114 void ovs_vport_exit(void)
118 for (i = 0; i < n_vport_types; i++) {
119 if (vport_ops_list[i]->exit)
120 vport_ops_list[i]->exit();
123 kfree(vport_ops_list);
127 static struct hlist_head *hash_bucket(struct net *net, const char *name)
129 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
130 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
134 * ovs_vport_locate - find a port that has already been created
136 * @name: name of port to find
138 * Must be called with RTNL or RCU read lock.
140 struct vport *ovs_vport_locate(struct net *net, const char *name)
142 struct hlist_head *bucket = hash_bucket(net, name);
144 struct hlist_node *node;
146 hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
147 if (!strcmp(name, vport->ops->get_name(vport)) &&
148 net_eq(ovs_dp_get_net(vport->dp), net))
154 static void release_vport(struct kobject *kobj)
156 struct vport *p = container_of(kobj, struct vport, kobj);
160 static struct kobj_type brport_ktype = {
162 .sysfs_ops = &ovs_brport_sysfs_ops,
164 .release = release_vport
168 * ovs_vport_alloc - allocate and initialize new vport
170 * @priv_size: Size of private data area to allocate.
171 * @ops: vport device ops
173 * Allocate and initialize a new vport defined by @ops. The vport will contain
174 * a private data area of size @priv_size that can be accessed using
175 * vport_priv(). vports that are no longer needed should be released with
178 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
179 const struct vport_parms *parms)
184 alloc_size = sizeof(struct vport);
186 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
187 alloc_size += priv_size;
190 vport = kzalloc(alloc_size, GFP_KERNEL);
192 return ERR_PTR(-ENOMEM);
194 vport->dp = parms->dp;
195 vport->port_no = parms->port_no;
196 vport->upcall_pid = parms->upcall_pid;
198 INIT_HLIST_NODE(&vport->dp_hash_node);
200 /* Initialize kobject for bridge. This will be added as
201 * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
202 vport->kobj.kset = NULL;
203 kobject_init(&vport->kobj, &brport_ktype);
205 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
206 if (!vport->percpu_stats) {
208 return ERR_PTR(-ENOMEM);
211 spin_lock_init(&vport->stats_lock);
217 * ovs_vport_free - uninitialize and free vport
219 * @vport: vport to free
221 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
223 * The caller must ensure that an RCU grace period has passed since the last
224 * time @vport was in a datapath.
226 void ovs_vport_free(struct vport *vport)
228 free_percpu(vport->percpu_stats);
230 kobject_put(&vport->kobj);
234 * ovs_vport_add - add vport device (for kernel callers)
236 * @parms: Information about new vport.
238 * Creates a new vport with the specified configuration (which is dependent on
239 * device type). RTNL lock must be held.
241 struct vport *ovs_vport_add(const struct vport_parms *parms)
249 for (i = 0; i < n_vport_types; i++) {
250 if (vport_ops_list[i]->type == parms->type) {
251 struct hlist_head *bucket;
253 vport = vport_ops_list[i]->create(parms);
255 err = PTR_ERR(vport);
259 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
260 vport->ops->get_name(vport));
261 hlist_add_head_rcu(&vport->hash_node, bucket);
273 * ovs_vport_set_options - modify existing vport device (for kernel callers)
275 * @vport: vport to modify.
276 * @port: New configuration.
278 * Modifies an existing device with the specified configuration (which is
279 * dependent on device type). RTNL lock must be held.
281 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
285 if (!vport->ops->set_options)
287 return vport->ops->set_options(vport, options);
291 * ovs_vport_del - delete existing vport device
293 * @vport: vport to delete.
295 * Detaches @vport from its datapath and destroys it. It is possible to fail
296 * for reasons such as lack of memory. RTNL lock must be held.
298 void ovs_vport_del(struct vport *vport)
302 hlist_del_rcu(&vport->hash_node);
304 vport->ops->destroy(vport);
308 * ovs_vport_set_addr - set device Ethernet address (for kernel callers)
310 * @vport: vport on which to set Ethernet address.
311 * @addr: New address.
313 * Sets the Ethernet address of the given device. Some devices may not support
314 * setting the Ethernet address, in which case the result will always be
315 * -EOPNOTSUPP. RTNL lock must be held.
317 int ovs_vport_set_addr(struct vport *vport, const unsigned char *addr)
321 if (!is_valid_ether_addr(addr))
322 return -EADDRNOTAVAIL;
324 if (vport->ops->set_addr)
325 return vport->ops->set_addr(vport, addr);
331 * ovs_vport_set_stats - sets offset device stats
333 * @vport: vport on which to set stats
334 * @stats: stats to set
336 * Provides a set of transmit, receive, and error stats to be added as an
337 * offset to the collect data when stats are retreived. Some devices may not
338 * support setting the stats, in which case the result will always be
341 * Must be called with RTNL lock.
343 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
347 spin_lock_bh(&vport->stats_lock);
348 vport->offset_stats = *stats;
349 spin_unlock_bh(&vport->stats_lock);
353 * ovs_vport_get_stats - retrieve device stats
355 * @vport: vport from which to retrieve the stats
356 * @stats: location to store stats
358 * Retrieves transmit, receive, and error stats for the given device.
360 * Must be called with RTNL lock or rcu_read_lock.
362 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
366 /* We potentially have 3 sources of stats that need to be
367 * combined: those we have collected (split into err_stats and
368 * percpu_stats), offset_stats from set_stats(), and device
369 * error stats from netdev->get_stats() (for errors that happen
370 * downstream and therefore aren't reported through our
371 * vport_record_error() function).
372 * Stats from first two sources are merged and reported by ovs over
373 * OVS_VPORT_ATTR_STATS.
374 * netdev-stats can be directly read over netlink-ioctl.
377 spin_lock_bh(&vport->stats_lock);
379 *stats = vport->offset_stats;
381 stats->rx_errors += vport->err_stats.rx_errors;
382 stats->tx_errors += vport->err_stats.tx_errors;
383 stats->tx_dropped += vport->err_stats.tx_dropped;
384 stats->rx_dropped += vport->err_stats.rx_dropped;
386 spin_unlock_bh(&vport->stats_lock);
388 for_each_possible_cpu(i) {
389 const struct vport_percpu_stats *percpu_stats;
390 struct vport_percpu_stats local_stats;
393 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
396 start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
397 local_stats = *percpu_stats;
398 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
400 stats->rx_bytes += local_stats.rx_bytes;
401 stats->rx_packets += local_stats.rx_packets;
402 stats->tx_bytes += local_stats.tx_bytes;
403 stats->tx_packets += local_stats.tx_packets;
408 * ovs_vport_get_options - retrieve device options
410 * @vport: vport from which to retrieve the options.
411 * @skb: sk_buff where options should be appended.
413 * Retrieves the configuration of the given device, appending an
414 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
415 * vport-specific attributes to @skb.
417 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
418 * negative error code if a real error occurred. If an error occurs, @skb is
421 * Must be called with RTNL lock or rcu_read_lock.
423 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
427 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
431 if (vport->ops->get_options) {
432 int err = vport->ops->get_options(vport, skb);
434 nla_nest_cancel(skb, nla);
439 nla_nest_end(skb, nla);
444 * ovs_vport_receive - pass up received packet to the datapath for processing
446 * @vport: vport that received the packet
447 * @skb: skb that was received
449 * Must be called with rcu_read_lock. The packet cannot be shared and
450 * skb->data should point to the Ethernet header. The caller must have already
451 * called compute_ip_summed() to initialize the checksumming fields.
453 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
455 struct vport_percpu_stats *stats;
457 stats = this_cpu_ptr(vport->percpu_stats);
458 u64_stats_update_begin(&stats->sync);
460 stats->rx_bytes += skb->len;
461 u64_stats_update_end(&stats->sync);
463 if (!(vport->ops->flags & VPORT_F_FLOW))
464 OVS_CB(skb)->flow = NULL;
466 if (!(vport->ops->flags & VPORT_F_TUN_ID))
467 OVS_CB(skb)->tun_key = NULL;
469 ovs_dp_process_received_packet(vport, skb);
473 * ovs_vport_send - send a packet on a device
475 * @vport: vport on which to send the packet
478 * Sends the given packet and returns the length of data sent. Either RTNL
479 * lock or rcu_read_lock must be held.
481 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
483 int sent = vport->ops->send(vport, skb);
486 struct vport_percpu_stats *stats;
488 stats = this_cpu_ptr(vport->percpu_stats);
490 u64_stats_update_begin(&stats->sync);
492 stats->tx_bytes += sent;
493 u64_stats_update_end(&stats->sync);
499 * ovs_vport_record_error - indicate device error to generic stats layer
501 * @vport: vport that encountered the error
502 * @err_type: one of enum vport_err_type types to indicate the error type
504 * If using the vport generic stats layer indicate that an error of the given
507 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
509 spin_lock(&vport->stats_lock);
512 case VPORT_E_RX_DROPPED:
513 vport->err_stats.rx_dropped++;
516 case VPORT_E_RX_ERROR:
517 vport->err_stats.rx_errors++;
520 case VPORT_E_TX_DROPPED:
521 vport->err_stats.tx_dropped++;
524 case VPORT_E_TX_ERROR:
525 vport->err_stats.tx_errors++;
529 spin_unlock(&vport->stats_lock);