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>
37 #include "vport-internal_dev.h"
39 static void ovs_vport_record_error(struct vport *,
40 enum vport_err_type err_type);
42 /* List of statically compiled vport implementations. Don't forget to also
43 * add yours to the list at the bottom of vport.h. */
44 static const struct vport_ops *vport_ops_list[] = {
45 &ovs_netdev_vport_ops,
46 &ovs_internal_vport_ops,
47 &ovs_geneve_vport_ops,
48 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
56 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
57 static struct hlist_head *dev_table;
58 #define VPORT_HASH_BUCKETS 1024
61 * ovs_vport_init - initialize vport subsystem
63 * Called at module load time to initialize the vport subsystem.
65 int ovs_vport_init(void)
67 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
76 * ovs_vport_exit - shutdown vport subsystem
78 * Called at module exit time to shutdown the vport subsystem.
80 void ovs_vport_exit(void)
85 static struct hlist_head *hash_bucket(struct net *net, const char *name)
87 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
88 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
92 * ovs_vport_locate - find a port that has already been created
94 * @name: name of port to find
96 * Must be called with ovs or RCU read lock.
98 struct vport *ovs_vport_locate(struct net *net, const char *name)
100 struct hlist_head *bucket = hash_bucket(net, name);
103 hlist_for_each_entry_rcu(vport, bucket, hash_node)
104 if (!strcmp(name, vport->ops->get_name(vport)) &&
105 net_eq(ovs_dp_get_net(vport->dp), net))
112 * ovs_vport_alloc - allocate and initialize new vport
114 * @priv_size: Size of private data area to allocate.
115 * @ops: vport device ops
117 * Allocate and initialize a new vport defined by @ops. The vport will contain
118 * a private data area of size @priv_size that can be accessed using
119 * vport_priv(). vports that are no longer needed should be released with
122 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
123 const struct vport_parms *parms)
129 alloc_size = sizeof(struct vport);
131 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
132 alloc_size += priv_size;
135 vport = kzalloc(alloc_size, GFP_KERNEL);
137 return ERR_PTR(-ENOMEM);
139 vport->dp = parms->dp;
140 vport->port_no = parms->port_no;
142 INIT_HLIST_NODE(&vport->dp_hash_node);
144 if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids))
145 return ERR_PTR(-EINVAL);
147 vport->percpu_stats = alloc_percpu(struct pcpu_sw_netstats);
148 if (!vport->percpu_stats) {
150 return ERR_PTR(-ENOMEM);
153 for_each_possible_cpu(i) {
154 struct pcpu_sw_netstats *vport_stats;
155 vport_stats = per_cpu_ptr(vport->percpu_stats, i);
156 u64_stats_init(&vport_stats->syncp);
159 spin_lock_init(&vport->stats_lock);
165 * ovs_vport_free - uninitialize and free vport
167 * @vport: vport to free
169 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
171 * The caller must ensure that an RCU grace period has passed since the last
172 * time @vport was in a datapath.
174 void ovs_vport_free(struct vport *vport)
176 kfree((struct vport_portids __force *)vport->upcall_portids);
177 free_percpu(vport->percpu_stats);
182 * ovs_vport_add - add vport device (for kernel callers)
184 * @parms: Information about new vport.
186 * Creates a new vport with the specified configuration (which is dependent on
187 * device type). ovs_mutex must be held.
189 struct vport *ovs_vport_add(const struct vport_parms *parms)
195 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
196 if (vport_ops_list[i]->type == parms->type) {
197 struct hlist_head *bucket;
199 vport = vport_ops_list[i]->create(parms);
201 err = PTR_ERR(vport);
205 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
206 vport->ops->get_name(vport));
207 hlist_add_head_rcu(&vport->hash_node, bucket);
219 * ovs_vport_set_options - modify existing vport device (for kernel callers)
221 * @vport: vport to modify.
222 * @options: New configuration.
224 * Modifies an existing device with the specified configuration (which is
225 * dependent on device type). ovs_mutex must be held.
227 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
229 if (!vport->ops->set_options)
231 return vport->ops->set_options(vport, options);
235 * ovs_vport_del - delete existing vport device
237 * @vport: vport to delete.
239 * Detaches @vport from its datapath and destroys it. It is possible to fail
240 * for reasons such as lack of memory. ovs_mutex must be held.
242 void ovs_vport_del(struct vport *vport)
246 hlist_del_rcu(&vport->hash_node);
247 vport->ops->destroy(vport);
251 * ovs_vport_set_stats - sets offset device stats
253 * @vport: vport on which to set stats
254 * @stats: stats to set
256 * Provides a set of transmit, receive, and error stats to be added as an
257 * offset to the collected data when stats are retrieved. Some devices may not
258 * support setting the stats, in which case the result will always be
261 * Must be called with ovs_mutex.
263 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
265 spin_lock_bh(&vport->stats_lock);
266 vport->offset_stats = *stats;
267 spin_unlock_bh(&vport->stats_lock);
271 * ovs_vport_get_stats - retrieve device stats
273 * @vport: vport from which to retrieve the stats
274 * @stats: location to store stats
276 * Retrieves transmit, receive, and error stats for the given device.
278 * Must be called with ovs_mutex or rcu_read_lock.
280 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
284 /* We potentially have 3 sources of stats that need to be
285 * combined: those we have collected (split into err_stats and
286 * percpu_stats), offset_stats from set_stats(), and device
287 * error stats from netdev->get_stats() (for errors that happen
288 * downstream and therefore aren't reported through our
289 * vport_record_error() function).
290 * Stats from first two sources are merged and reported by ovs over
291 * OVS_VPORT_ATTR_STATS.
292 * netdev-stats can be directly read over netlink-ioctl.
295 spin_lock_bh(&vport->stats_lock);
297 *stats = vport->offset_stats;
299 stats->rx_errors += vport->err_stats.rx_errors;
300 stats->tx_errors += vport->err_stats.tx_errors;
301 stats->tx_dropped += vport->err_stats.tx_dropped;
302 stats->rx_dropped += vport->err_stats.rx_dropped;
304 spin_unlock_bh(&vport->stats_lock);
306 for_each_possible_cpu(i) {
307 const struct pcpu_sw_netstats *percpu_stats;
308 struct pcpu_sw_netstats local_stats;
311 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
314 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
315 local_stats = *percpu_stats;
316 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
318 stats->rx_bytes += local_stats.rx_bytes;
319 stats->rx_packets += local_stats.rx_packets;
320 stats->tx_bytes += local_stats.tx_bytes;
321 stats->tx_packets += local_stats.tx_packets;
326 * ovs_vport_get_options - retrieve device options
328 * @vport: vport from which to retrieve the options.
329 * @skb: sk_buff where options should be appended.
331 * Retrieves the configuration of the given device, appending an
332 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
333 * vport-specific attributes to @skb.
335 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
336 * negative error code if a real error occurred. If an error occurs, @skb is
339 * Must be called with ovs_mutex or rcu_read_lock.
341 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
346 if (!vport->ops->get_options)
349 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
353 err = vport->ops->get_options(vport, skb);
355 nla_nest_cancel(skb, nla);
359 nla_nest_end(skb, nla);
363 static void vport_portids_destroy_rcu_cb(struct rcu_head *rcu)
365 struct vport_portids *ids = container_of(rcu, struct vport_portids,
372 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
374 * @vport: vport to modify.
375 * @ids: new configuration, an array of port ids.
377 * Sets the vport's upcall_portids to @ids.
379 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
380 * as an array of U32.
382 * Must be called with ovs_mutex.
384 int ovs_vport_set_upcall_portids(struct vport *vport, struct nlattr *ids)
386 struct vport_portids *old, *vport_portids;
388 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
391 old = ovsl_dereference(vport->upcall_portids);
393 vport_portids = kmalloc(sizeof *vport_portids + nla_len(ids),
398 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
399 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
400 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
402 rcu_assign_pointer(vport->upcall_portids, vport_portids);
405 call_rcu(&old->rcu, vport_portids_destroy_rcu_cb);
411 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
413 * @vport: vport from which to retrieve the portids.
414 * @skb: sk_buff where portids should be appended.
416 * Retrieves the configuration of the given vport, appending the
417 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
420 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
421 * If an error occurs, @skb is left unmodified. Must be called with
422 * ovs_mutex or rcu_read_lock.
424 int ovs_vport_get_upcall_portids(const struct vport *vport,
427 struct vport_portids *ids;
429 ids = rcu_dereference_ovsl(vport->upcall_portids);
431 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
432 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
433 ids->n_ids * sizeof(u32), (void *) ids->ids);
435 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
439 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
441 * @vport: vport from which the missed packet is received.
442 * @skb: skb that the missed packet was received.
444 * Uses the skb_get_hash() to select the upcall portid to send the
447 * Returns the portid of the target socket. Must be called with rcu_read_lock.
449 u32 ovs_vport_find_upcall_portid(const struct vport *p, struct sk_buff *skb)
451 struct vport_portids *ids;
454 ids = rcu_dereference(p->upcall_portids);
456 if (ids->n_ids == 1 && ids->ids[0] == 0)
459 hash = skb_get_hash(skb);
460 return ids->ids[hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids)];
464 * ovs_vport_receive - pass up received packet to the datapath for processing
466 * @vport: vport that received the packet
467 * @skb: skb that was received
468 * @tun_info: tunnel (if any) that carried packet
470 * Must be called with rcu_read_lock. The packet cannot be shared and
471 * skb->data should point to the Ethernet header. The caller must have already
472 * called compute_ip_summed() to initialize the checksumming fields.
474 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
475 struct ovs_tunnel_info *tun_info)
477 struct pcpu_sw_netstats *stats;
478 struct sw_flow_key key;
481 stats = this_cpu_ptr(vport->percpu_stats);
482 u64_stats_update_begin(&stats->syncp);
484 stats->rx_bytes += skb->len;
485 u64_stats_update_end(&stats->syncp);
487 ovs_skb_init_inner_protocol(skb);
488 OVS_CB(skb)->input_vport = vport;
489 OVS_CB(skb)->egress_tun_info = NULL;
490 error = ovs_flow_key_extract(tun_info, skb, &key);
491 if (unlikely(error)) {
496 ovs_dp_process_packet(skb, false);
500 * ovs_vport_send - send a packet on a device
502 * @vport: vport on which to send the packet
505 * Sends the given packet and returns the length of data sent. Either ovs
506 * lock or rcu_read_lock must be held.
508 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
510 int sent = vport->ops->send(vport, skb);
512 if (likely(sent > 0)) {
513 struct pcpu_sw_netstats *stats;
515 stats = this_cpu_ptr(vport->percpu_stats);
517 u64_stats_update_begin(&stats->syncp);
519 stats->tx_bytes += sent;
520 u64_stats_update_end(&stats->syncp);
521 } else if (sent < 0) {
522 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
525 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
531 * ovs_vport_record_error - indicate device error to generic stats layer
533 * @vport: vport that encountered the error
534 * @err_type: one of enum vport_err_type types to indicate the error type
536 * If using the vport generic stats layer indicate that an error of the given
539 static void ovs_vport_record_error(struct vport *vport,
540 enum vport_err_type err_type)
542 spin_lock(&vport->stats_lock);
545 case VPORT_E_RX_DROPPED:
546 vport->err_stats.rx_dropped++;
549 case VPORT_E_RX_ERROR:
550 vport->err_stats.rx_errors++;
553 case VPORT_E_TX_DROPPED:
554 vport->err_stats.tx_dropped++;
557 case VPORT_E_TX_ERROR:
558 vport->err_stats.tx_errors++;
562 spin_unlock(&vport->stats_lock);
565 static void free_vport_rcu(struct rcu_head *rcu)
567 struct vport *vport = container_of(rcu, struct vport, rcu);
569 ovs_vport_free(vport);
572 void ovs_vport_deferred_free(struct vport *vport)
577 call_rcu(&vport->rcu, free_vport_rcu);