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/kconfig.h>
24 #include <linux/kernel.h>
25 #include <linux/list.h>
26 #include <linux/mutex.h>
27 #include <linux/percpu.h>
28 #include <linux/rcupdate.h>
29 #include <linux/rtnetlink.h>
30 #include <linux/compat.h>
31 #include <linux/version.h>
32 #include <net/net_namespace.h>
36 #include "vport-internal_dev.h"
38 static void ovs_vport_record_error(struct vport *,
39 enum vport_err_type err_type);
41 /* List of statically compiled vport implementations. Don't forget to also
42 * add yours to the list at the bottom of vport.h. */
43 static const struct vport_ops *vport_ops_list[] = {
44 &ovs_netdev_vport_ops,
45 &ovs_internal_vport_ops,
46 &ovs_geneve_vport_ops,
47 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
55 /* Protected by RCU read lock for reading, ovs_mutex 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.
64 int ovs_vport_init(void)
66 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
75 * ovs_vport_exit - shutdown vport subsystem
77 * Called at module exit time to shutdown the vport subsystem.
79 void ovs_vport_exit(void)
84 static struct hlist_head *hash_bucket(struct net *net, const char *name)
86 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
87 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
91 * ovs_vport_locate - find a port that has already been created
93 * @name: name of port to find
95 * Must be called with ovs or RCU read lock.
97 struct vport *ovs_vport_locate(struct net *net, const char *name)
99 struct hlist_head *bucket = hash_bucket(net, name);
102 hlist_for_each_entry_rcu(vport, bucket, hash_node)
103 if (!strcmp(name, vport->ops->get_name(vport)) &&
104 net_eq(ovs_dp_get_net(vport->dp), net))
111 * ovs_vport_alloc - allocate and initialize new vport
113 * @priv_size: Size of private data area to allocate.
114 * @ops: vport device ops
116 * Allocate and initialize a new vport defined by @ops. The vport will contain
117 * a private data area of size @priv_size that can be accessed using
118 * vport_priv(). vports that are no longer needed should be released with
121 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
122 const struct vport_parms *parms)
128 alloc_size = sizeof(struct vport);
130 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
131 alloc_size += priv_size;
134 vport = kzalloc(alloc_size, GFP_KERNEL);
136 return ERR_PTR(-ENOMEM);
138 vport->dp = parms->dp;
139 vport->port_no = parms->port_no;
141 INIT_HLIST_NODE(&vport->dp_hash_node);
143 if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids))
144 return ERR_PTR(-EINVAL);
146 vport->percpu_stats = alloc_percpu(struct pcpu_sw_netstats);
147 if (!vport->percpu_stats) {
149 return ERR_PTR(-ENOMEM);
152 for_each_possible_cpu(i) {
153 struct pcpu_sw_netstats *vport_stats;
154 vport_stats = per_cpu_ptr(vport->percpu_stats, i);
155 u64_stats_init(&vport_stats->syncp);
158 spin_lock_init(&vport->stats_lock);
164 * ovs_vport_free - uninitialize and free vport
166 * @vport: vport to free
168 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
170 * The caller must ensure that an RCU grace period has passed since the last
171 * time @vport was in a datapath.
173 void ovs_vport_free(struct vport *vport)
175 kfree((struct vport_portids __force *)vport->upcall_portids);
176 free_percpu(vport->percpu_stats);
181 * ovs_vport_add - add vport device (for kernel callers)
183 * @parms: Information about new vport.
185 * Creates a new vport with the specified configuration (which is dependent on
186 * device type). ovs_mutex must be held.
188 struct vport *ovs_vport_add(const struct vport_parms *parms)
194 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
195 if (vport_ops_list[i]->type == parms->type) {
196 struct hlist_head *bucket;
198 vport = vport_ops_list[i]->create(parms);
200 err = PTR_ERR(vport);
204 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
205 vport->ops->get_name(vport));
206 hlist_add_head_rcu(&vport->hash_node, bucket);
218 * ovs_vport_set_options - modify existing vport device (for kernel callers)
220 * @vport: vport to modify.
221 * @options: New configuration.
223 * Modifies an existing device with the specified configuration (which is
224 * dependent on device type). ovs_mutex must be held.
226 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
228 if (!vport->ops->set_options)
230 return vport->ops->set_options(vport, options);
234 * ovs_vport_del - delete existing vport device
236 * @vport: vport to delete.
238 * Detaches @vport from its datapath and destroys it. It is possible to fail
239 * for reasons such as lack of memory. ovs_mutex must be held.
241 void ovs_vport_del(struct vport *vport)
245 hlist_del_rcu(&vport->hash_node);
246 vport->ops->destroy(vport);
250 * ovs_vport_set_stats - sets offset device stats
252 * @vport: vport on which to set stats
253 * @stats: stats to set
255 * Provides a set of transmit, receive, and error stats to be added as an
256 * offset to the collected data when stats are retrieved. Some devices may not
257 * support setting the stats, in which case the result will always be
260 * Must be called with ovs_mutex.
262 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
264 spin_lock_bh(&vport->stats_lock);
265 vport->offset_stats = *stats;
266 spin_unlock_bh(&vport->stats_lock);
270 * ovs_vport_get_stats - retrieve device stats
272 * @vport: vport from which to retrieve the stats
273 * @stats: location to store stats
275 * Retrieves transmit, receive, and error stats for the given device.
277 * Must be called with ovs_mutex or rcu_read_lock.
279 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
283 /* We potentially have 3 sources of stats that need to be
284 * combined: those we have collected (split into err_stats and
285 * percpu_stats), offset_stats from set_stats(), and device
286 * error stats from netdev->get_stats() (for errors that happen
287 * downstream and therefore aren't reported through our
288 * vport_record_error() function).
289 * Stats from first two sources are merged and reported by ovs over
290 * OVS_VPORT_ATTR_STATS.
291 * netdev-stats can be directly read over netlink-ioctl.
294 spin_lock_bh(&vport->stats_lock);
296 *stats = vport->offset_stats;
298 stats->rx_errors += vport->err_stats.rx_errors;
299 stats->tx_errors += vport->err_stats.tx_errors;
300 stats->tx_dropped += vport->err_stats.tx_dropped;
301 stats->rx_dropped += vport->err_stats.rx_dropped;
303 spin_unlock_bh(&vport->stats_lock);
305 for_each_possible_cpu(i) {
306 const struct pcpu_sw_netstats *percpu_stats;
307 struct pcpu_sw_netstats local_stats;
310 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
313 start = u64_stats_fetch_begin_bh(&percpu_stats->syncp);
314 local_stats = *percpu_stats;
315 } while (u64_stats_fetch_retry_bh(&percpu_stats->syncp, start));
317 stats->rx_bytes += local_stats.rx_bytes;
318 stats->rx_packets += local_stats.rx_packets;
319 stats->tx_bytes += local_stats.tx_bytes;
320 stats->tx_packets += local_stats.tx_packets;
325 * ovs_vport_get_options - retrieve device options
327 * @vport: vport from which to retrieve the options.
328 * @skb: sk_buff where options should be appended.
330 * Retrieves the configuration of the given device, appending an
331 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
332 * vport-specific attributes to @skb.
334 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
335 * negative error code if a real error occurred. If an error occurs, @skb is
338 * Must be called with ovs_mutex or rcu_read_lock.
340 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
345 if (!vport->ops->get_options)
348 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
352 err = vport->ops->get_options(vport, skb);
354 nla_nest_cancel(skb, nla);
358 nla_nest_end(skb, nla);
362 static void vport_portids_destroy_rcu_cb(struct rcu_head *rcu)
364 struct vport_portids *ids = container_of(rcu, struct vport_portids,
371 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
373 * @vport: vport to modify.
374 * @ids: new configuration, an array of port ids.
376 * Sets the vport's upcall_portids to @ids.
378 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
379 * as an array of U32.
381 * Must be called with ovs_mutex.
383 int ovs_vport_set_upcall_portids(struct vport *vport, struct nlattr *ids)
385 struct vport_portids *old, *vport_portids;
387 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
390 old = ovsl_dereference(vport->upcall_portids);
392 vport_portids = kmalloc(sizeof *vport_portids + nla_len(ids),
394 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
395 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
396 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
398 rcu_assign_pointer(vport->upcall_portids, vport_portids);
401 call_rcu(&old->rcu, vport_portids_destroy_rcu_cb);
407 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
409 * @vport: vport from which to retrieve the portids.
410 * @skb: sk_buff where portids should be appended.
412 * Retrieves the configuration of the given vport, appending the
413 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
416 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
417 * If an error occurs, @skb is left unmodified. Must be called with
418 * ovs_mutex or rcu_read_lock.
420 int ovs_vport_get_upcall_portids(const struct vport *vport,
423 struct vport_portids *ids;
425 ids = rcu_dereference_ovsl(vport->upcall_portids);
427 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
428 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
429 ids->n_ids * sizeof(u32), (void *) ids->ids);
431 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
435 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
437 * @vport: vport from which the missed packet is received.
438 * @skb: skb that the missed packet was received.
440 * Uses the skb_get_hash() to select the upcall portid to send the
443 * Returns the portid of the target socket. Must be called with rcu_read_lock.
445 u32 ovs_vport_find_upcall_portid(const struct vport *p, struct sk_buff *skb)
447 struct vport_portids *ids;
450 ids = rcu_dereference(p->upcall_portids);
452 if (ids->n_ids == 1 && ids->ids[0] == 0)
455 hash = skb_get_hash(skb);
456 return ids->ids[hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids)];
460 * ovs_vport_receive - pass up received packet to the datapath for processing
462 * @vport: vport that received the packet
463 * @skb: skb that was received
464 * @tun_info: tunnel (if any) that carried packet
466 * Must be called with rcu_read_lock. The packet cannot be shared and
467 * skb->data should point to the Ethernet header. The caller must have already
468 * called compute_ip_summed() to initialize the checksumming fields.
470 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
471 struct ovs_tunnel_info *tun_info)
473 struct pcpu_sw_netstats *stats;
475 stats = this_cpu_ptr(vport->percpu_stats);
476 u64_stats_update_begin(&stats->syncp);
478 stats->rx_bytes += skb->len;
479 u64_stats_update_end(&stats->syncp);
481 OVS_CB(skb)->tun_info = tun_info;
482 ovs_dp_process_received_packet(vport, skb);
486 * ovs_vport_send - send a packet on a device
488 * @vport: vport on which to send the packet
491 * Sends the given packet and returns the length of data sent. Either ovs
492 * lock or rcu_read_lock must be held.
494 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
496 int sent = vport->ops->send(vport, skb);
498 if (likely(sent > 0)) {
499 struct pcpu_sw_netstats *stats;
501 stats = this_cpu_ptr(vport->percpu_stats);
503 u64_stats_update_begin(&stats->syncp);
505 stats->tx_bytes += sent;
506 u64_stats_update_end(&stats->syncp);
507 } else if (sent < 0) {
508 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
511 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
517 * ovs_vport_record_error - indicate device error to generic stats layer
519 * @vport: vport that encountered the error
520 * @err_type: one of enum vport_err_type types to indicate the error type
522 * If using the vport generic stats layer indicate that an error of the given
525 static void ovs_vport_record_error(struct vport *vport,
526 enum vport_err_type err_type)
528 spin_lock(&vport->stats_lock);
531 case VPORT_E_RX_DROPPED:
532 vport->err_stats.rx_dropped++;
535 case VPORT_E_RX_ERROR:
536 vport->err_stats.rx_errors++;
539 case VPORT_E_TX_DROPPED:
540 vport->err_stats.tx_dropped++;
543 case VPORT_E_TX_ERROR:
544 vport->err_stats.tx_errors++;
548 spin_unlock(&vport->stats_lock);
551 static void free_vport_rcu(struct rcu_head *rcu)
553 struct vport *vport = container_of(rcu, struct vport, rcu);
555 ovs_vport_free(vport);
558 void ovs_vport_deferred_free(struct vport *vport)
563 call_rcu(&vport->rcu, free_vport_rcu);