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.
45 static const struct vport_ops *vport_ops_list[] = {
46 &ovs_netdev_vport_ops,
47 &ovs_internal_vport_ops,
48 &ovs_geneve_vport_ops,
49 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
57 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
58 static struct hlist_head *dev_table;
59 #define VPORT_HASH_BUCKETS 1024
62 * ovs_vport_init - initialize vport subsystem
64 * Called at module load time to initialize the vport subsystem.
66 int ovs_vport_init(void)
68 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
77 * ovs_vport_exit - shutdown vport subsystem
79 * Called at module exit time to shutdown the vport subsystem.
81 void ovs_vport_exit(void)
86 static struct hlist_head *hash_bucket(const struct net *net, const char *name)
88 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
89 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
93 * ovs_vport_locate - find a port that has already been created
95 * @name: name of port to find
97 * Must be called with ovs or RCU read lock.
99 struct vport *ovs_vport_locate(const struct net *net, const char *name)
101 struct hlist_head *bucket = hash_bucket(net, name);
104 hlist_for_each_entry_rcu(vport, bucket, hash_node)
105 if (!strcmp(name, vport->ops->get_name(vport)) &&
106 net_eq(ovs_dp_get_net(vport->dp), net))
113 * ovs_vport_alloc - allocate and initialize new vport
115 * @priv_size: Size of private data area to allocate.
116 * @ops: vport device ops
118 * Allocate and initialize a new vport defined by @ops. The vport will contain
119 * a private data area of size @priv_size that can be accessed using
120 * vport_priv(). vports that are no longer needed should be released with
123 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
124 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)) {
146 return ERR_PTR(-EINVAL);
149 vport->percpu_stats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
150 if (!vport->percpu_stats) {
152 return ERR_PTR(-ENOMEM);
159 * ovs_vport_free - uninitialize and free vport
161 * @vport: vport to free
163 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
165 * The caller must ensure that an RCU grace period has passed since the last
166 * time @vport was in a datapath.
168 void ovs_vport_free(struct vport *vport)
170 kfree(rcu_dereference_raw(vport->upcall_portids));
171 free_percpu(vport->percpu_stats);
176 * ovs_vport_add - add vport device (for kernel callers)
178 * @parms: Information about new vport.
180 * Creates a new vport with the specified configuration (which is dependent on
181 * device type). ovs_mutex must be held.
183 struct vport *ovs_vport_add(const struct vport_parms *parms)
189 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
190 if (vport_ops_list[i]->type == parms->type) {
191 struct hlist_head *bucket;
193 vport = vport_ops_list[i]->create(parms);
195 err = PTR_ERR(vport);
199 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
200 vport->ops->get_name(vport));
201 hlist_add_head_rcu(&vport->hash_node, bucket);
213 * ovs_vport_set_options - modify existing vport device (for kernel callers)
215 * @vport: vport to modify.
216 * @options: New configuration.
218 * Modifies an existing device with the specified configuration (which is
219 * dependent on device type). ovs_mutex must be held.
221 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
223 if (!vport->ops->set_options)
225 return vport->ops->set_options(vport, options);
229 * ovs_vport_del - delete existing vport device
231 * @vport: vport to delete.
233 * Detaches @vport from its datapath and destroys it. It is possible to fail
234 * for reasons such as lack of memory. ovs_mutex must be held.
236 void ovs_vport_del(struct vport *vport)
240 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 /* We potentially have two surces of stats that need to be
259 * combined: those we have collected (split into err_stats and
260 * percpu_stats), and device error stats from netdev->get_stats()
261 * (for errors that happen downstream and therefore aren't
262 * reported through our vport_record_error() function).
263 * Stats from first source are reported by ovs over
264 * OVS_VPORT_ATTR_STATS.
265 * netdev-stats can be directly read over netlink-ioctl.
268 stats->rx_errors = atomic_long_read(&vport->err_stats.rx_errors);
269 stats->tx_errors = atomic_long_read(&vport->err_stats.tx_errors);
270 stats->tx_dropped = atomic_long_read(&vport->err_stats.tx_dropped);
271 stats->rx_dropped = atomic_long_read(&vport->err_stats.rx_dropped);
274 stats->rx_packets = 0;
276 stats->tx_packets = 0;
278 for_each_possible_cpu(i) {
279 const struct pcpu_sw_netstats *percpu_stats;
280 struct pcpu_sw_netstats local_stats;
283 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
286 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
287 local_stats = *percpu_stats;
288 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
290 stats->rx_bytes += local_stats.rx_bytes;
291 stats->rx_packets += local_stats.rx_packets;
292 stats->tx_bytes += local_stats.tx_bytes;
293 stats->tx_packets += local_stats.tx_packets;
298 * ovs_vport_get_options - retrieve device options
300 * @vport: vport from which to retrieve the options.
301 * @skb: sk_buff where options should be appended.
303 * Retrieves the configuration of the given device, appending an
304 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
305 * vport-specific attributes to @skb.
307 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
308 * negative error code if a real error occurred. If an error occurs, @skb is
311 * Must be called with ovs_mutex or rcu_read_lock.
313 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
318 if (!vport->ops->get_options)
321 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
325 err = vport->ops->get_options(vport, skb);
327 nla_nest_cancel(skb, nla);
331 nla_nest_end(skb, nla);
335 static void vport_portids_destroy_rcu_cb(struct rcu_head *rcu)
337 struct vport_portids *ids = container_of(rcu, struct vport_portids,
344 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
346 * @vport: vport to modify.
347 * @ids: new configuration, an array of port ids.
349 * Sets the vport's upcall_portids to @ids.
351 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
352 * as an array of U32.
354 * Must be called with ovs_mutex.
356 int ovs_vport_set_upcall_portids(struct vport *vport, const struct nlattr *ids)
358 struct vport_portids *old, *vport_portids;
360 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
363 old = ovsl_dereference(vport->upcall_portids);
365 vport_portids = kmalloc(sizeof *vport_portids + nla_len(ids),
370 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
371 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
372 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
374 rcu_assign_pointer(vport->upcall_portids, vport_portids);
377 call_rcu(&old->rcu, vport_portids_destroy_rcu_cb);
383 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
385 * @vport: vport from which to retrieve the portids.
386 * @skb: sk_buff where portids should be appended.
388 * Retrieves the configuration of the given vport, appending the
389 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
392 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
393 * If an error occurs, @skb is left unmodified. Must be called with
394 * ovs_mutex or rcu_read_lock.
396 int ovs_vport_get_upcall_portids(const struct vport *vport,
399 struct vport_portids *ids;
401 ids = rcu_dereference_ovsl(vport->upcall_portids);
403 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
404 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
405 ids->n_ids * sizeof(u32), (void *) ids->ids);
407 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
411 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
413 * @vport: vport from which the missed packet is received.
414 * @skb: skb that the missed packet was received.
416 * Uses the skb_get_hash() to select the upcall portid to send the
419 * Returns the portid of the target socket. Must be called with rcu_read_lock.
421 u32 ovs_vport_find_upcall_portid(const struct vport *vport, struct sk_buff *skb)
423 struct vport_portids *ids;
426 ids = rcu_dereference(vport->upcall_portids);
428 if (ids->n_ids == 1 && ids->ids[0] == 0)
431 hash = skb_get_hash(skb);
432 return ids->ids[hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids)];
436 * ovs_vport_receive - pass up received packet to the datapath for processing
438 * @vport: vport that received the packet
439 * @skb: skb that was received
440 * @tun_info: tunnel (if any) that carried packet
442 * Must be called with rcu_read_lock. The packet cannot be shared and
443 * skb->data should point to the Ethernet header. The caller must have already
444 * called compute_ip_summed() to initialize the checksumming fields.
446 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
447 const struct ovs_tunnel_info *tun_info)
449 struct pcpu_sw_netstats *stats;
450 struct sw_flow_key key;
453 stats = this_cpu_ptr(vport->percpu_stats);
454 u64_stats_update_begin(&stats->syncp);
456 stats->rx_bytes += skb->len;
457 u64_stats_update_end(&stats->syncp);
459 ovs_skb_init_inner_protocol(skb);
460 OVS_CB(skb)->input_vport = vport;
461 OVS_CB(skb)->egress_tun_info = NULL;
462 error = ovs_flow_key_extract(tun_info, skb, &key);
463 if (unlikely(error)) {
468 ovs_dp_process_packet(skb, &key);
472 * ovs_vport_send - send a packet on a device
474 * @vport: vport on which to send the packet
477 * Sends the given packet and returns the length of data sent. Either ovs
478 * lock or rcu_read_lock must be held.
480 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
482 int sent = vport->ops->send(vport, skb);
484 if (likely(sent > 0)) {
485 struct pcpu_sw_netstats *stats;
487 stats = this_cpu_ptr(vport->percpu_stats);
489 u64_stats_update_begin(&stats->syncp);
491 stats->tx_bytes += sent;
492 u64_stats_update_end(&stats->syncp);
493 } else if (sent < 0) {
494 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
497 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
503 * ovs_vport_record_error - indicate device error to generic stats layer
505 * @vport: vport that encountered the error
506 * @err_type: one of enum vport_err_type types to indicate the error type
508 * If using the vport generic stats layer indicate that an error of the given
511 static void ovs_vport_record_error(struct vport *vport,
512 enum vport_err_type err_type)
515 case VPORT_E_RX_DROPPED:
516 atomic_long_inc(&vport->err_stats.rx_dropped);
519 case VPORT_E_RX_ERROR:
520 atomic_long_inc(&vport->err_stats.rx_errors);
523 case VPORT_E_TX_DROPPED:
524 atomic_long_inc(&vport->err_stats.tx_dropped);
527 case VPORT_E_TX_ERROR:
528 atomic_long_inc(&vport->err_stats.tx_errors);
533 static void free_vport_rcu(struct rcu_head *rcu)
535 struct vport *vport = container_of(rcu, struct vport, rcu);
537 ovs_vport_free(vport);
540 void ovs_vport_deferred_free(struct vport *vport)
545 call_rcu(&vport->rcu, free_vport_rcu);
548 int ovs_tunnel_get_egress_info(struct ovs_tunnel_info *egress_tun_info,
550 const struct ovs_tunnel_info *tun_info,
556 const struct ovs_key_ipv4_tunnel *tun_key;
560 if (unlikely(!tun_info))
563 tun_key = &tun_info->tunnel;
564 saddr = tun_key->ipv4_src;
565 /* Route lookup to get srouce IP address: saddr.
566 * The process may need to be changed if the corresponding process
567 * in vports ops changed.
580 /* Generate egress_tun_info based on tun_info,
581 * saddr, tp_src and tp_dst
583 __ovs_flow_tun_info_init(egress_tun_info,
584 saddr, tun_key->ipv4_dst,
591 tun_info->options_len);
596 int ovs_vport_get_egress_tun_info(struct vport *vport, struct sk_buff *skb,
597 struct ovs_tunnel_info *info)
599 /* get_egress_tun_info() is only implemented on tunnel ports. */
600 if (unlikely(!vport->ops->get_egress_tun_info))
603 return vport->ops->get_egress_tun_info(vport, skb, info);