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 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
45 &ovs_capwap_vport_ops,
49 static const struct vport_ops **vport_ops_list;
50 static int n_vport_types;
52 /* Protected by RCU read lock for reading, RTNL lock for writing. */
53 static struct hlist_head *dev_table;
54 #define VPORT_HASH_BUCKETS 1024
57 * ovs_vport_init - initialize vport subsystem
59 * Called at module load time to initialize the vport subsystem and any
60 * compiled in vport types.
62 int ovs_vport_init(void)
67 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
74 vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
75 sizeof(struct vport_ops *), GFP_KERNEL);
76 if (!vport_ops_list) {
81 for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
82 const struct vport_ops *new_ops = base_vport_ops_list[i];
85 err = new_ops->init();
90 vport_ops_list[n_vport_types++] = new_ops;
91 else if (new_ops->flags & VPORT_F_REQUIRED) {
106 * ovs_vport_exit - shutdown vport subsystem
108 * Called at module exit time to shutdown the vport subsystem and any
109 * initialized vport types.
111 void ovs_vport_exit(void)
115 for (i = 0; i < n_vport_types; i++) {
116 if (vport_ops_list[i]->exit)
117 vport_ops_list[i]->exit();
120 kfree(vport_ops_list);
124 static struct hlist_head *hash_bucket(struct net *net, const char *name)
126 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
127 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
131 * ovs_vport_locate - find a port that has already been created
133 * @name: name of port to find
135 * Must be called with RTNL or RCU read lock.
137 struct vport *ovs_vport_locate(struct net *net, const char *name)
139 struct hlist_head *bucket = hash_bucket(net, name);
141 struct hlist_node *node;
143 hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
144 if (!strcmp(name, vport->ops->get_name(vport)) &&
145 net_eq(ovs_dp_get_net(vport->dp), net))
151 static void release_vport(struct kobject *kobj)
153 struct vport *p = container_of(kobj, struct vport, kobj);
157 static struct kobj_type brport_ktype = {
159 .sysfs_ops = &ovs_brport_sysfs_ops,
161 .release = release_vport
165 * ovs_vport_alloc - allocate and initialize new vport
167 * @priv_size: Size of private data area to allocate.
168 * @ops: vport device ops
170 * Allocate and initialize a new vport defined by @ops. The vport will contain
171 * a private data area of size @priv_size that can be accessed using
172 * vport_priv(). vports that are no longer needed should be released with
175 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
176 const struct vport_parms *parms)
181 alloc_size = sizeof(struct vport);
183 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
184 alloc_size += priv_size;
187 vport = kzalloc(alloc_size, GFP_KERNEL);
189 return ERR_PTR(-ENOMEM);
191 vport->dp = parms->dp;
192 vport->port_no = parms->port_no;
193 vport->upcall_pid = parms->upcall_pid;
195 INIT_HLIST_NODE(&vport->dp_hash_node);
197 /* Initialize kobject for bridge. This will be added as
198 * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
199 vport->kobj.kset = NULL;
200 kobject_init(&vport->kobj, &brport_ktype);
202 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
203 if (!vport->percpu_stats) {
205 return ERR_PTR(-ENOMEM);
208 spin_lock_init(&vport->stats_lock);
214 * ovs_vport_free - uninitialize and free vport
216 * @vport: vport to free
218 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
220 * The caller must ensure that an RCU grace period has passed since the last
221 * time @vport was in a datapath.
223 void ovs_vport_free(struct vport *vport)
225 free_percpu(vport->percpu_stats);
227 kobject_put(&vport->kobj);
231 * ovs_vport_add - add vport device (for kernel callers)
233 * @parms: Information about new vport.
235 * Creates a new vport with the specified configuration (which is dependent on
236 * device type). RTNL lock must be held.
238 struct vport *ovs_vport_add(const struct vport_parms *parms)
246 for (i = 0; i < n_vport_types; i++) {
247 if (vport_ops_list[i]->type == parms->type) {
248 struct hlist_head *bucket;
250 vport = vport_ops_list[i]->create(parms);
252 err = PTR_ERR(vport);
256 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
257 vport->ops->get_name(vport));
258 hlist_add_head_rcu(&vport->hash_node, bucket);
270 * ovs_vport_set_options - modify existing vport device (for kernel callers)
272 * @vport: vport to modify.
273 * @port: New configuration.
275 * Modifies an existing device with the specified configuration (which is
276 * dependent on device type). RTNL lock must be held.
278 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
282 if (!vport->ops->set_options)
284 return vport->ops->set_options(vport, options);
288 * ovs_vport_del - delete existing vport device
290 * @vport: vport to delete.
292 * Detaches @vport from its datapath and destroys it. It is possible to fail
293 * for reasons such as lack of memory. RTNL lock must be held.
295 void ovs_vport_del(struct vport *vport)
299 hlist_del_rcu(&vport->hash_node);
301 vport->ops->destroy(vport);
305 * ovs_vport_set_addr - set device Ethernet address (for kernel callers)
307 * @vport: vport on which to set Ethernet address.
308 * @addr: New address.
310 * Sets the Ethernet address of the given device. Some devices may not support
311 * setting the Ethernet address, in which case the result will always be
312 * -EOPNOTSUPP. RTNL lock must be held.
314 int ovs_vport_set_addr(struct vport *vport, const unsigned char *addr)
318 if (!is_valid_ether_addr(addr))
319 return -EADDRNOTAVAIL;
321 if (vport->ops->set_addr)
322 return vport->ops->set_addr(vport, addr);
328 * ovs_vport_set_stats - sets offset device stats
330 * @vport: vport on which to set stats
331 * @stats: stats to set
333 * Provides a set of transmit, receive, and error stats to be added as an
334 * offset to the collect data when stats are retreived. Some devices may not
335 * support setting the stats, in which case the result will always be
338 * Must be called with RTNL lock.
340 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
344 spin_lock_bh(&vport->stats_lock);
345 vport->offset_stats = *stats;
346 spin_unlock_bh(&vport->stats_lock);
350 * ovs_vport_get_stats - retrieve device stats
352 * @vport: vport from which to retrieve the stats
353 * @stats: location to store stats
355 * Retrieves transmit, receive, and error stats for the given device.
357 * Must be called with RTNL lock or rcu_read_lock.
359 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
363 /* We potentially have 3 sources of stats that need to be
364 * combined: those we have collected (split into err_stats and
365 * percpu_stats), offset_stats from set_stats(), and device
366 * error stats from netdev->get_stats() (for errors that happen
367 * downstream and therefore aren't reported through our
368 * vport_record_error() function).
369 * Stats from first two sources are merged and reported by ovs over
370 * OVS_VPORT_ATTR_STATS.
371 * netdev-stats can be directly read over netlink-ioctl.
374 spin_lock_bh(&vport->stats_lock);
376 *stats = vport->offset_stats;
378 stats->rx_errors += vport->err_stats.rx_errors;
379 stats->tx_errors += vport->err_stats.tx_errors;
380 stats->tx_dropped += vport->err_stats.tx_dropped;
381 stats->rx_dropped += vport->err_stats.rx_dropped;
383 spin_unlock_bh(&vport->stats_lock);
385 for_each_possible_cpu(i) {
386 const struct vport_percpu_stats *percpu_stats;
387 struct vport_percpu_stats local_stats;
390 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
393 start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
394 local_stats = *percpu_stats;
395 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
397 stats->rx_bytes += local_stats.rx_bytes;
398 stats->rx_packets += local_stats.rx_packets;
399 stats->tx_bytes += local_stats.tx_bytes;
400 stats->tx_packets += local_stats.tx_packets;
405 * ovs_vport_get_options - retrieve device options
407 * @vport: vport from which to retrieve the options.
408 * @skb: sk_buff where options should be appended.
410 * Retrieves the configuration of the given device, appending an
411 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
412 * vport-specific attributes to @skb.
414 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
415 * negative error code if a real error occurred. If an error occurs, @skb is
418 * Must be called with RTNL lock or rcu_read_lock.
420 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
424 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
428 if (vport->ops->get_options) {
429 int err = vport->ops->get_options(vport, skb);
431 nla_nest_cancel(skb, nla);
436 nla_nest_end(skb, nla);
441 * ovs_vport_receive - pass up received packet to the datapath for processing
443 * @vport: vport that received the packet
444 * @skb: skb that was received
446 * Must be called with rcu_read_lock. The packet cannot be shared and
447 * skb->data should point to the Ethernet header. The caller must have already
448 * called compute_ip_summed() to initialize the checksumming fields.
450 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
452 struct vport_percpu_stats *stats;
454 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
456 u64_stats_update_begin(&stats->sync);
458 stats->rx_bytes += skb->len;
459 u64_stats_update_end(&stats->sync);
461 if (!(vport->ops->flags & VPORT_F_FLOW))
462 OVS_CB(skb)->flow = NULL;
464 if (!(vport->ops->flags & VPORT_F_TUN_ID))
465 OVS_CB(skb)->tun_id = 0;
467 ovs_dp_process_received_packet(vport, skb);
471 * ovs_vport_send - send a packet on a device
473 * @vport: vport on which to send the packet
476 * Sends the given packet and returns the length of data sent. Either RTNL
477 * lock or rcu_read_lock must be held.
479 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
481 int sent = vport->ops->send(vport, skb);
484 struct vport_percpu_stats *stats;
486 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
488 u64_stats_update_begin(&stats->sync);
490 stats->tx_bytes += sent;
491 u64_stats_update_end(&stats->sync);
497 * ovs_vport_record_error - indicate device error to generic stats layer
499 * @vport: vport that encountered the error
500 * @err_type: one of enum vport_err_type types to indicate the error type
502 * If using the vport generic stats layer indicate that an error of the given
505 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
507 spin_lock(&vport->stats_lock);
510 case VPORT_E_RX_DROPPED:
511 vport->err_stats.rx_dropped++;
514 case VPORT_E_RX_ERROR:
515 vport->err_stats.rx_errors++;
518 case VPORT_E_TX_DROPPED:
519 vport->err_stats.tx_dropped++;
522 case VPORT_E_TX_ERROR:
523 vport->err_stats.tx_errors++;
527 spin_unlock(&vport->stats_lock);