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(const 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(const 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)
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)) {
145 return ERR_PTR(-EINVAL);
148 vport->percpu_stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
149 if (!vport->percpu_stats) {
151 return ERR_PTR(-ENOMEM);
158 * ovs_vport_free - uninitialize and free vport
160 * @vport: vport to free
162 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
164 * The caller must ensure that an RCU grace period has passed since the last
165 * time @vport was in a datapath.
167 void ovs_vport_free(struct vport *vport)
169 kfree((struct vport_portids __force *)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);
240 vport->ops->destroy(vport);
244 * ovs_vport_get_stats - retrieve device stats
246 * @vport: vport from which to retrieve the stats
247 * @stats: location to store stats
249 * Retrieves transmit, receive, and error stats for the given device.
251 * Must be called with ovs_mutex or rcu_read_lock.
253 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
257 /* We potentially have two surces of stats that need to be
258 * combined: those we have collected (split into err_stats and
259 * percpu_stats), and device error stats from netdev->get_stats()
260 * (for errors that happen downstream and therefore aren't
261 * reported through our vport_record_error() function).
262 * Stats from first source are reported by ovs over
263 * OVS_VPORT_ATTR_STATS.
264 * netdev-stats can be directly read over netlink-ioctl.
267 stats->rx_errors = atomic_long_read(&vport->err_stats.rx_errors);
268 stats->tx_errors = atomic_long_read(&vport->err_stats.tx_errors);
269 stats->tx_dropped = atomic_long_read(&vport->err_stats.tx_dropped);
270 stats->rx_dropped = atomic_long_read(&vport->err_stats.rx_dropped);
273 stats->rx_packets = 0;
275 stats->tx_packets = 0;
277 for_each_possible_cpu(i) {
278 const struct pcpu_sw_netstats *percpu_stats;
279 struct pcpu_sw_netstats local_stats;
282 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
285 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
286 local_stats = *percpu_stats;
287 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
289 stats->rx_bytes += local_stats.rx_bytes;
290 stats->rx_packets += local_stats.rx_packets;
291 stats->tx_bytes += local_stats.tx_bytes;
292 stats->tx_packets += local_stats.tx_packets;
297 * ovs_vport_get_options - retrieve device options
299 * @vport: vport from which to retrieve the options.
300 * @skb: sk_buff where options should be appended.
302 * Retrieves the configuration of the given device, appending an
303 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
304 * vport-specific attributes to @skb.
306 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
307 * negative error code if a real error occurred. If an error occurs, @skb is
310 * Must be called with ovs_mutex or rcu_read_lock.
312 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
317 if (!vport->ops->get_options)
320 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
324 err = vport->ops->get_options(vport, skb);
326 nla_nest_cancel(skb, nla);
330 nla_nest_end(skb, nla);
334 static void vport_portids_destroy_rcu_cb(struct rcu_head *rcu)
336 struct vport_portids *ids = container_of(rcu, struct vport_portids,
343 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
345 * @vport: vport to modify.
346 * @ids: new configuration, an array of port ids.
348 * Sets the vport's upcall_portids to @ids.
350 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
351 * as an array of U32.
353 * Must be called with ovs_mutex.
355 int ovs_vport_set_upcall_portids(struct vport *vport, const struct nlattr *ids)
357 struct vport_portids *old, *vport_portids;
359 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
362 old = ovsl_dereference(vport->upcall_portids);
364 vport_portids = kmalloc(sizeof *vport_portids + nla_len(ids),
369 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
370 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
371 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
373 rcu_assign_pointer(vport->upcall_portids, vport_portids);
376 call_rcu(&old->rcu, vport_portids_destroy_rcu_cb);
382 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
384 * @vport: vport from which to retrieve the portids.
385 * @skb: sk_buff where portids should be appended.
387 * Retrieves the configuration of the given vport, appending the
388 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
391 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
392 * If an error occurs, @skb is left unmodified. Must be called with
393 * ovs_mutex or rcu_read_lock.
395 int ovs_vport_get_upcall_portids(const struct vport *vport,
398 struct vport_portids *ids;
400 ids = rcu_dereference_ovsl(vport->upcall_portids);
402 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
403 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
404 ids->n_ids * sizeof(u32), (void *) ids->ids);
406 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
410 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
412 * @vport: vport from which the missed packet is received.
413 * @skb: skb that the missed packet was received.
415 * Uses the skb_get_hash() to select the upcall portid to send the
418 * Returns the portid of the target socket. Must be called with rcu_read_lock.
420 u32 ovs_vport_find_upcall_portid(const struct vport *vport, struct sk_buff *skb)
422 struct vport_portids *ids;
425 ids = rcu_dereference(vport->upcall_portids);
427 if (ids->n_ids == 1 && ids->ids[0] == 0)
430 hash = skb_get_hash(skb);
431 return ids->ids[hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids)];
435 * ovs_vport_receive - pass up received packet to the datapath for processing
437 * @vport: vport that received the packet
438 * @skb: skb that was received
439 * @tun_info: tunnel (if any) that carried packet
441 * Must be called with rcu_read_lock. The packet cannot be shared and
442 * skb->data should point to the Ethernet header. The caller must have already
443 * called compute_ip_summed() to initialize the checksumming fields.
445 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
446 const struct ovs_tunnel_info *tun_info)
448 struct pcpu_sw_netstats *stats;
449 struct sw_flow_key key;
452 stats = this_cpu_ptr(vport->percpu_stats);
453 u64_stats_update_begin(&stats->syncp);
455 stats->rx_bytes += skb->len;
456 u64_stats_update_end(&stats->syncp);
458 ovs_skb_init_inner_protocol(skb);
459 OVS_CB(skb)->input_vport = vport;
460 OVS_CB(skb)->egress_tun_info = NULL;
461 error = ovs_flow_key_extract(tun_info, skb, &key);
462 if (unlikely(error)) {
467 ovs_dp_process_packet(skb, &key);
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 ovs
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);
483 if (likely(sent > 0)) {
484 struct pcpu_sw_netstats *stats;
486 stats = this_cpu_ptr(vport->percpu_stats);
488 u64_stats_update_begin(&stats->syncp);
490 stats->tx_bytes += sent;
491 u64_stats_update_end(&stats->syncp);
492 } else if (sent < 0) {
493 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
496 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
502 * ovs_vport_record_error - indicate device error to generic stats layer
504 * @vport: vport that encountered the error
505 * @err_type: one of enum vport_err_type types to indicate the error type
507 * If using the vport generic stats layer indicate that an error of the given
510 static void ovs_vport_record_error(struct vport *vport,
511 enum vport_err_type err_type)
514 case VPORT_E_RX_DROPPED:
515 atomic_long_inc(&vport->err_stats.rx_dropped);
518 case VPORT_E_RX_ERROR:
519 atomic_long_inc(&vport->err_stats.rx_errors);
522 case VPORT_E_TX_DROPPED:
523 atomic_long_inc(&vport->err_stats.tx_dropped);
526 case VPORT_E_TX_ERROR:
527 atomic_long_inc(&vport->err_stats.tx_errors);
532 static void free_vport_rcu(struct rcu_head *rcu)
534 struct vport *vport = container_of(rcu, struct vport, rcu);
536 ovs_vport_free(vport);
539 void ovs_vport_deferred_free(struct vport *vport)
544 call_rcu(&vport->rcu, free_vport_rcu);
547 int ovs_tunnel_get_egress_info(struct ovs_tunnel_info *egress_tun_info,
549 const struct ovs_tunnel_info *tun_info,
555 const struct ovs_key_ipv4_tunnel *tun_key;
559 if (unlikely(!tun_info))
562 tun_key = &tun_info->tunnel;
563 saddr = tun_key->ipv4_src;
564 /* Route lookup to get srouce IP address: saddr.
565 * The process may need to be changed if the corresponding process
566 * in vports ops changed.
579 /* Generate egress_tun_info based on tun_info,
580 * saddr, tp_src and tp_dst
582 __ovs_flow_tun_info_init(egress_tun_info,
583 saddr, tun_key->ipv4_dst,
590 tun_info->options_len);
595 int ovs_vport_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
596 struct ovs_tunnel_info *info)
598 /* get_egress_tun_info() is only implemented on tunnel ports. */
599 if (unlikely(!vport->ops->get_egress_tun_info))
602 return vport->ops->get_egress_tun_info(vport, skb, info);