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 <net/net_namespace.h>
34 #include "vport-internal_dev.h"
36 static void ovs_vport_record_error(struct vport *,
37 enum vport_err_type err_type);
39 /* List of statically compiled vport implementations. Don't forget to also
40 * add yours to the list at the bottom of vport.h. */
41 static const struct vport_ops *vport_ops_list[] = {
42 &ovs_netdev_vport_ops,
43 &ovs_internal_vport_ops,
45 #ifdef CONFIG_OPENVSWITCH_GRE
48 #ifdef CONFIG_OPENVSWITCH_VXLAN
53 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
54 static struct hlist_head *dev_table;
55 #define VPORT_HASH_BUCKETS 1024
58 * ovs_vport_init - initialize vport subsystem
60 * Called at module load time to initialize the vport subsystem.
62 int ovs_vport_init(void)
64 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
73 * ovs_vport_exit - shutdown vport subsystem
75 * Called at module exit time to shutdown the vport subsystem.
77 void ovs_vport_exit(void)
82 static struct hlist_head *hash_bucket(struct net *net, const char *name)
84 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
85 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
89 * ovs_vport_locate - find a port that has already been created
91 * @name: name of port to find
93 * Must be called with ovs or RCU read lock.
95 struct vport *ovs_vport_locate(struct net *net, const char *name)
97 struct hlist_head *bucket = hash_bucket(net, name);
100 hlist_for_each_entry_rcu(vport, bucket, hash_node)
101 if (!strcmp(name, vport->ops->get_name(vport)) &&
102 net_eq(ovs_dp_get_net(vport->dp), net))
109 * ovs_vport_alloc - allocate and initialize new vport
111 * @priv_size: Size of private data area to allocate.
112 * @ops: vport device ops
114 * Allocate and initialize a new vport defined by @ops. The vport will contain
115 * a private data area of size @priv_size that can be accessed using
116 * vport_priv(). vports that are no longer needed should be released with
119 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
120 const struct vport_parms *parms)
125 alloc_size = sizeof(struct vport);
127 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
128 alloc_size += priv_size;
131 vport = kzalloc(alloc_size, GFP_KERNEL);
133 return ERR_PTR(-ENOMEM);
135 vport->dp = parms->dp;
136 vport->port_no = parms->port_no;
138 INIT_HLIST_NODE(&vport->dp_hash_node);
140 if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids)) {
142 return ERR_PTR(-EINVAL);
145 vport->percpu_stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
146 if (!vport->percpu_stats) {
148 return ERR_PTR(-ENOMEM);
155 * ovs_vport_free - uninitialize and free vport
157 * @vport: vport to free
159 * Frees a vport allocated with vport_alloc() when it is no longer needed.
161 * The caller must ensure that an RCU grace period has passed since the last
162 * time @vport was in a datapath.
164 void ovs_vport_free(struct vport *vport)
166 /* vport is freed from RCU callback or error path, Therefore
167 * it is safe to use raw dereference.
169 kfree(rcu_dereference_raw(vport->upcall_portids));
170 free_percpu(vport->percpu_stats);
175 * ovs_vport_add - add vport device (for kernel callers)
177 * @parms: Information about new vport.
179 * Creates a new vport with the specified configuration (which is dependent on
180 * device type). ovs_mutex must be held.
182 struct vport *ovs_vport_add(const struct vport_parms *parms)
188 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
189 if (vport_ops_list[i]->type == parms->type) {
190 struct hlist_head *bucket;
192 vport = vport_ops_list[i]->create(parms);
194 err = PTR_ERR(vport);
198 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
199 vport->ops->get_name(vport));
200 hlist_add_head_rcu(&vport->hash_node, bucket);
212 * ovs_vport_set_options - modify existing vport device (for kernel callers)
214 * @vport: vport to modify.
215 * @options: New configuration.
217 * Modifies an existing device with the specified configuration (which is
218 * dependent on device type). ovs_mutex must be held.
220 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
222 if (!vport->ops->set_options)
224 return vport->ops->set_options(vport, options);
228 * ovs_vport_del - delete existing vport device
230 * @vport: vport to delete.
232 * Detaches @vport from its datapath and destroys it. It is possible to fail
233 * for reasons such as lack of memory. ovs_mutex must be held.
235 void ovs_vport_del(struct vport *vport)
239 hlist_del_rcu(&vport->hash_node);
241 vport->ops->destroy(vport);
245 * ovs_vport_get_stats - retrieve device stats
247 * @vport: vport from which to retrieve the stats
248 * @stats: location to store stats
250 * Retrieves transmit, receive, and error stats for the given device.
252 * Must be called with ovs_mutex or rcu_read_lock.
254 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
258 memset(stats, 0, sizeof(*stats));
260 /* We potentially have 2 sources of stats that need to be combined:
261 * those we have collected (split into err_stats and percpu_stats) from
262 * set_stats() and device error stats from netdev->get_stats() (for
263 * errors that happen downstream and therefore aren't reported through
264 * our vport_record_error() function).
265 * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
266 * netdev-stats can be directly read over netlink-ioctl.
269 stats->rx_errors = atomic_long_read(&vport->err_stats.rx_errors);
270 stats->tx_errors = atomic_long_read(&vport->err_stats.tx_errors);
271 stats->tx_dropped = atomic_long_read(&vport->err_stats.tx_dropped);
272 stats->rx_dropped = atomic_long_read(&vport->err_stats.rx_dropped);
274 for_each_possible_cpu(i) {
275 const struct pcpu_sw_netstats *percpu_stats;
276 struct pcpu_sw_netstats local_stats;
279 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
282 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
283 local_stats = *percpu_stats;
284 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
286 stats->rx_bytes += local_stats.rx_bytes;
287 stats->rx_packets += local_stats.rx_packets;
288 stats->tx_bytes += local_stats.tx_bytes;
289 stats->tx_packets += local_stats.tx_packets;
294 * ovs_vport_get_options - retrieve device options
296 * @vport: vport from which to retrieve the options.
297 * @skb: sk_buff where options should be appended.
299 * Retrieves the configuration of the given device, appending an
300 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
301 * vport-specific attributes to @skb.
303 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
304 * negative error code if a real error occurred. If an error occurs, @skb is
307 * Must be called with ovs_mutex or rcu_read_lock.
309 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
314 if (!vport->ops->get_options)
317 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
321 err = vport->ops->get_options(vport, skb);
323 nla_nest_cancel(skb, nla);
327 nla_nest_end(skb, nla);
332 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
334 * @vport: vport to modify.
335 * @ids: new configuration, an array of port ids.
337 * Sets the vport's upcall_portids to @ids.
339 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
340 * as an array of U32.
342 * Must be called with ovs_mutex.
344 int ovs_vport_set_upcall_portids(struct vport *vport, struct nlattr *ids)
346 struct vport_portids *old, *vport_portids;
348 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
351 old = ovsl_dereference(vport->upcall_portids);
353 vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
358 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
359 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
360 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
362 rcu_assign_pointer(vport->upcall_portids, vport_portids);
370 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
372 * @vport: vport from which to retrieve the portids.
373 * @skb: sk_buff where portids should be appended.
375 * Retrieves the configuration of the given vport, appending the
376 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
379 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
380 * If an error occurs, @skb is left unmodified. Must be called with
381 * ovs_mutex or rcu_read_lock.
383 int ovs_vport_get_upcall_portids(const struct vport *vport,
386 struct vport_portids *ids;
388 ids = rcu_dereference_ovsl(vport->upcall_portids);
390 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
391 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
392 ids->n_ids * sizeof(u32), (void *)ids->ids);
394 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
398 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
400 * @vport: vport from which the missed packet is received.
401 * @skb: skb that the missed packet was received.
403 * Uses the skb_get_hash() to select the upcall portid to send the
406 * Returns the portid of the target socket. Must be called with rcu_read_lock.
408 u32 ovs_vport_find_upcall_portid(const struct vport *p, struct sk_buff *skb)
410 struct vport_portids *ids;
414 ids = rcu_dereference(p->upcall_portids);
416 if (ids->n_ids == 1 && ids->ids[0] == 0)
419 hash = skb_get_hash(skb);
420 ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
421 return ids->ids[ids_index];
425 * ovs_vport_receive - pass up received packet to the datapath for processing
427 * @vport: vport that received the packet
428 * @skb: skb that was received
429 * @tun_key: tunnel (if any) that carried packet
431 * Must be called with rcu_read_lock. The packet cannot be shared and
432 * skb->data should point to the Ethernet header.
434 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
435 struct ovs_key_ipv4_tunnel *tun_key)
437 struct pcpu_sw_netstats *stats;
439 stats = this_cpu_ptr(vport->percpu_stats);
440 u64_stats_update_begin(&stats->syncp);
442 stats->rx_bytes += skb->len;
443 u64_stats_update_end(&stats->syncp);
445 OVS_CB(skb)->tun_key = tun_key;
446 ovs_dp_process_received_packet(vport, skb);
450 * ovs_vport_send - send a packet on a device
452 * @vport: vport on which to send the packet
455 * Sends the given packet and returns the length of data sent. Either ovs
456 * lock or rcu_read_lock must be held.
458 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
460 int sent = vport->ops->send(vport, skb);
462 if (likely(sent > 0)) {
463 struct pcpu_sw_netstats *stats;
465 stats = this_cpu_ptr(vport->percpu_stats);
467 u64_stats_update_begin(&stats->syncp);
469 stats->tx_bytes += sent;
470 u64_stats_update_end(&stats->syncp);
471 } else if (sent < 0) {
472 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
475 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
481 * ovs_vport_record_error - indicate device error to generic stats layer
483 * @vport: vport that encountered the error
484 * @err_type: one of enum vport_err_type types to indicate the error type
486 * If using the vport generic stats layer indicate that an error of the given
489 static void ovs_vport_record_error(struct vport *vport,
490 enum vport_err_type err_type)
493 case VPORT_E_RX_DROPPED:
494 atomic_long_inc(&vport->err_stats.rx_dropped);
497 case VPORT_E_RX_ERROR:
498 atomic_long_inc(&vport->err_stats.rx_errors);
501 case VPORT_E_TX_DROPPED:
502 atomic_long_inc(&vport->err_stats.tx_dropped);
505 case VPORT_E_TX_ERROR:
506 atomic_long_inc(&vport->err_stats.tx_errors);
512 static void free_vport_rcu(struct rcu_head *rcu)
514 struct vport *vport = container_of(rcu, struct vport, rcu);
516 ovs_vport_free(vport);
519 void ovs_vport_deferred_free(struct vport *vport)
524 call_rcu(&vport->rcu, free_vport_rcu);