2 * Copyright (c) 2007-2014 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 <net/net_namespace.h>
31 #include <linux/module.h>
35 #include "vport-internal_dev.h"
37 static LIST_HEAD(vport_ops_list);
39 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
40 static struct hlist_head *dev_table;
41 #define VPORT_HASH_BUCKETS 1024
44 * ovs_vport_init - initialize vport subsystem
46 * Called at module load time to initialize the vport subsystem.
48 int ovs_vport_init(void)
50 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
59 * ovs_vport_exit - shutdown vport subsystem
61 * Called at module exit time to shutdown the vport subsystem.
63 void ovs_vport_exit(void)
68 static struct hlist_head *hash_bucket(const struct net *net, const char *name)
70 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
71 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
74 int ovs_vport_ops_register(struct vport_ops *ops)
80 list_for_each_entry(o, &vport_ops_list, list)
81 if (ops->type == o->type)
84 list_add_tail(&ops->list, &vport_ops_list);
90 EXPORT_SYMBOL_GPL(ovs_vport_ops_register);
92 void ovs_vport_ops_unregister(struct vport_ops *ops)
98 EXPORT_SYMBOL_GPL(ovs_vport_ops_unregister);
101 * ovs_vport_locate - find a port that has already been created
103 * @name: name of port to find
105 * Must be called with ovs or RCU read lock.
107 struct vport *ovs_vport_locate(const struct net *net, const char *name)
109 struct hlist_head *bucket = hash_bucket(net, name);
112 hlist_for_each_entry_rcu(vport, bucket, hash_node)
113 if (!strcmp(name, ovs_vport_name(vport)) &&
114 net_eq(ovs_dp_get_net(vport->dp), net))
121 * ovs_vport_alloc - allocate and initialize new vport
123 * @priv_size: Size of private data area to allocate.
124 * @ops: vport device ops
126 * Allocate and initialize a new vport defined by @ops. The vport will contain
127 * a private data area of size @priv_size that can be accessed using
128 * vport_priv(). vports that are no longer needed should be released with
131 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
132 const struct vport_parms *parms)
137 alloc_size = sizeof(struct vport);
139 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
140 alloc_size += priv_size;
143 vport = kzalloc(alloc_size, GFP_KERNEL);
145 return ERR_PTR(-ENOMEM);
147 vport->dp = parms->dp;
148 vport->port_no = parms->port_no;
150 INIT_HLIST_NODE(&vport->dp_hash_node);
152 if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids)) {
154 return ERR_PTR(-EINVAL);
159 EXPORT_SYMBOL_GPL(ovs_vport_alloc);
162 * ovs_vport_free - uninitialize and free vport
164 * @vport: vport to free
166 * Frees a vport allocated with vport_alloc() when it is no longer needed.
168 * The caller must ensure that an RCU grace period has passed since the last
169 * time @vport was in a datapath.
171 void ovs_vport_free(struct vport *vport)
173 /* vport is freed from RCU callback or error path, Therefore
174 * it is safe to use raw dereference.
176 kfree(rcu_dereference_raw(vport->upcall_portids));
179 EXPORT_SYMBOL_GPL(ovs_vport_free);
181 static struct vport_ops *ovs_vport_lookup(const struct vport_parms *parms)
183 struct vport_ops *ops;
185 list_for_each_entry(ops, &vport_ops_list, list)
186 if (ops->type == parms->type)
193 * ovs_vport_add - add vport device (for kernel callers)
195 * @parms: Information about new vport.
197 * Creates a new vport with the specified configuration (which is dependent on
198 * device type). ovs_mutex must be held.
200 struct vport *ovs_vport_add(const struct vport_parms *parms)
202 struct vport_ops *ops;
205 ops = ovs_vport_lookup(parms);
207 struct hlist_head *bucket;
209 if (!try_module_get(ops->owner))
210 return ERR_PTR(-EAFNOSUPPORT);
212 vport = ops->create(parms);
214 module_put(ops->owner);
218 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
219 ovs_vport_name(vport));
220 hlist_add_head_rcu(&vport->hash_node, bucket);
224 /* Unlock to attempt module load and return -EAGAIN if load
225 * was successful as we need to restart the port addition
229 request_module("vport-type-%d", parms->type);
232 if (!ovs_vport_lookup(parms))
233 return ERR_PTR(-EAFNOSUPPORT);
235 return ERR_PTR(-EAGAIN);
239 * ovs_vport_set_options - modify existing vport device (for kernel callers)
241 * @vport: vport to modify.
242 * @options: New configuration.
244 * Modifies an existing device with the specified configuration (which is
245 * dependent on device type). ovs_mutex must be held.
247 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
249 if (!vport->ops->set_options)
251 return vport->ops->set_options(vport, options);
255 * ovs_vport_del - delete existing vport device
257 * @vport: vport to delete.
259 * Detaches @vport from its datapath and destroys it. It is possible to fail
260 * for reasons such as lack of memory. ovs_mutex must be held.
262 void ovs_vport_del(struct vport *vport)
266 hlist_del_rcu(&vport->hash_node);
267 module_put(vport->ops->owner);
268 vport->ops->destroy(vport);
272 * ovs_vport_get_stats - retrieve device stats
274 * @vport: vport from which to retrieve the stats
275 * @stats: location to store stats
277 * Retrieves transmit, receive, and error stats for the given device.
279 * Must be called with ovs_mutex or rcu_read_lock.
281 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
283 struct net_device *dev = vport->dev;
286 memset(stats, 0, sizeof(*stats));
287 stats->rx_errors = dev->stats.rx_errors;
288 stats->tx_errors = dev->stats.tx_errors;
289 stats->tx_dropped = dev->stats.tx_dropped;
290 stats->rx_dropped = dev->stats.rx_dropped;
292 stats->rx_dropped += atomic_long_read(&dev->rx_dropped);
293 stats->tx_dropped += atomic_long_read(&dev->tx_dropped);
295 for_each_possible_cpu(i) {
296 const struct pcpu_sw_netstats *percpu_stats;
297 struct pcpu_sw_netstats local_stats;
300 percpu_stats = per_cpu_ptr(dev->tstats, i);
303 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
304 local_stats = *percpu_stats;
305 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
307 stats->rx_bytes += local_stats.rx_bytes;
308 stats->rx_packets += local_stats.rx_packets;
309 stats->tx_bytes += local_stats.tx_bytes;
310 stats->tx_packets += local_stats.tx_packets;
315 * ovs_vport_get_options - retrieve device options
317 * @vport: vport from which to retrieve the options.
318 * @skb: sk_buff where options should be appended.
320 * Retrieves the configuration of the given device, appending an
321 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
322 * vport-specific attributes to @skb.
324 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
325 * negative error code if a real error occurred. If an error occurs, @skb is
328 * Must be called with ovs_mutex or rcu_read_lock.
330 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
335 if (!vport->ops->get_options)
338 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
342 err = vport->ops->get_options(vport, skb);
344 nla_nest_cancel(skb, nla);
348 nla_nest_end(skb, nla);
353 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
355 * @vport: vport to modify.
356 * @ids: new configuration, an array of port ids.
358 * Sets the vport's upcall_portids to @ids.
360 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
361 * as an array of U32.
363 * Must be called with ovs_mutex.
365 int ovs_vport_set_upcall_portids(struct vport *vport, const struct nlattr *ids)
367 struct vport_portids *old, *vport_portids;
369 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
372 old = ovsl_dereference(vport->upcall_portids);
374 vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
379 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
380 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
381 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
383 rcu_assign_pointer(vport->upcall_portids, vport_portids);
391 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
393 * @vport: vport from which to retrieve the portids.
394 * @skb: sk_buff where portids should be appended.
396 * Retrieves the configuration of the given vport, appending the
397 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
400 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
401 * If an error occurs, @skb is left unmodified. Must be called with
402 * ovs_mutex or rcu_read_lock.
404 int ovs_vport_get_upcall_portids(const struct vport *vport,
407 struct vport_portids *ids;
409 ids = rcu_dereference_ovsl(vport->upcall_portids);
411 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
412 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
413 ids->n_ids * sizeof(u32), (void *)ids->ids);
415 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
419 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
421 * @vport: vport from which the missed packet is received.
422 * @skb: skb that the missed packet was received.
424 * Uses the skb_get_hash() to select the upcall portid to send the
427 * Returns the portid of the target socket. Must be called with rcu_read_lock.
429 u32 ovs_vport_find_upcall_portid(const struct vport *vport, struct sk_buff *skb)
431 struct vport_portids *ids;
435 ids = rcu_dereference(vport->upcall_portids);
437 if (ids->n_ids == 1 && ids->ids[0] == 0)
440 hash = skb_get_hash(skb);
441 ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
442 return ids->ids[ids_index];
446 * ovs_vport_receive - pass up received packet to the datapath for processing
448 * @vport: vport that received the packet
449 * @skb: skb that was received
450 * @tun_key: tunnel (if any) that carried packet
452 * Must be called with rcu_read_lock. The packet cannot be shared and
453 * skb->data should point to the Ethernet header.
455 int ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
456 const struct ip_tunnel_info *tun_info)
458 struct sw_flow_key key;
461 OVS_CB(skb)->input_vport = vport;
462 OVS_CB(skb)->mru = 0;
463 /* Extract flow from 'skb' into 'key'. */
464 error = ovs_flow_key_extract(tun_info, skb, &key);
465 if (unlikely(error)) {
469 ovs_dp_process_packet(skb, &key);
472 EXPORT_SYMBOL_GPL(ovs_vport_receive);
474 static void free_vport_rcu(struct rcu_head *rcu)
476 struct vport *vport = container_of(rcu, struct vport, rcu);
478 ovs_vport_free(vport);
481 void ovs_vport_deferred_free(struct vport *vport)
486 call_rcu(&vport->rcu, free_vport_rcu);
488 EXPORT_SYMBOL_GPL(ovs_vport_deferred_free);
490 int ovs_tunnel_get_egress_info(struct dp_upcall_info *upcall,
497 struct ip_tunnel_info *egress_tun_info = upcall->egress_tun_info;
498 const struct ip_tunnel_info *tun_info = skb_tunnel_info(skb);
499 const struct ip_tunnel_key *tun_key;
500 u32 skb_mark = skb->mark;
504 if (unlikely(!tun_info))
506 if (ip_tunnel_info_af(tun_info) != AF_INET)
509 tun_key = &tun_info->key;
511 /* Route lookup to get srouce IP address.
512 * The process may need to be changed if the corresponding process
513 * in vports ops changed.
515 rt = ovs_tunnel_route_lookup(net, tun_key, skb_mark, &fl, ipproto);
521 /* Generate egress_tun_info based on tun_info,
522 * saddr, tp_src and tp_dst
524 ip_tunnel_key_init(&egress_tun_info->key,
525 fl.saddr, tun_key->u.ipv4.dst,
531 egress_tun_info->options_len = tun_info->options_len;
532 egress_tun_info->mode = tun_info->mode;
533 upcall->egress_tun_opts = ip_tunnel_info_opts(egress_tun_info);
536 EXPORT_SYMBOL_GPL(ovs_tunnel_get_egress_info);
538 int ovs_vport_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
539 struct dp_upcall_info *upcall)
541 /* get_egress_tun_info() is only implemented on tunnel ports. */
542 if (unlikely(!vport->ops->get_egress_tun_info))
545 return vport->ops->get_egress_tun_info(vport, skb, upcall);