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);
151 spin_lock_init(&vport->stats_lock);
157 * ovs_vport_free - uninitialize and free vport
159 * @vport: vport to free
161 * Frees a vport allocated with vport_alloc() when it is no longer needed.
163 * The caller must ensure that an RCU grace period has passed since the last
164 * time @vport was in a datapath.
166 void ovs_vport_free(struct vport *vport)
168 /* vport is freed from RCU callback or error path, Therefore
169 * it is safe to use raw dereference.
171 kfree(rcu_dereference_raw(vport->upcall_portids));
172 free_percpu(vport->percpu_stats);
177 * ovs_vport_add - add vport device (for kernel callers)
179 * @parms: Information about new vport.
181 * Creates a new vport with the specified configuration (which is dependent on
182 * device type). ovs_mutex must be held.
184 struct vport *ovs_vport_add(const struct vport_parms *parms)
190 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
191 if (vport_ops_list[i]->type == parms->type) {
192 struct hlist_head *bucket;
194 vport = vport_ops_list[i]->create(parms);
196 err = PTR_ERR(vport);
200 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
201 vport->ops->get_name(vport));
202 hlist_add_head_rcu(&vport->hash_node, bucket);
214 * ovs_vport_set_options - modify existing vport device (for kernel callers)
216 * @vport: vport to modify.
217 * @options: New configuration.
219 * Modifies an existing device with the specified configuration (which is
220 * dependent on device type). ovs_mutex must be held.
222 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
224 if (!vport->ops->set_options)
226 return vport->ops->set_options(vport, options);
230 * ovs_vport_del - delete existing vport device
232 * @vport: vport to delete.
234 * Detaches @vport from its datapath and destroys it. It is possible to fail
235 * for reasons such as lack of memory. ovs_mutex must be held.
237 void ovs_vport_del(struct vport *vport)
241 hlist_del_rcu(&vport->hash_node);
243 vport->ops->destroy(vport);
247 * ovs_vport_get_stats - retrieve device stats
249 * @vport: vport from which to retrieve the stats
250 * @stats: location to store stats
252 * Retrieves transmit, receive, and error stats for the given device.
254 * Must be called with ovs_mutex or rcu_read_lock.
256 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
260 memset(stats, 0, sizeof(*stats));
262 /* We potentially have 2 sources of stats that need to be combined:
263 * those we have collected (split into err_stats and percpu_stats) from
264 * set_stats() and device error stats from netdev->get_stats() (for
265 * errors that happen downstream and therefore aren't reported through
266 * our vport_record_error() function).
267 * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
268 * netdev-stats can be directly read over netlink-ioctl.
271 spin_lock_bh(&vport->stats_lock);
273 stats->rx_errors = vport->err_stats.rx_errors;
274 stats->tx_errors = vport->err_stats.tx_errors;
275 stats->tx_dropped = vport->err_stats.tx_dropped;
276 stats->rx_dropped = vport->err_stats.rx_dropped;
278 spin_unlock_bh(&vport->stats_lock);
280 for_each_possible_cpu(i) {
281 const struct pcpu_sw_netstats *percpu_stats;
282 struct pcpu_sw_netstats local_stats;
285 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
288 start = u64_stats_fetch_begin_irq(&percpu_stats->syncp);
289 local_stats = *percpu_stats;
290 } while (u64_stats_fetch_retry_irq(&percpu_stats->syncp, start));
292 stats->rx_bytes += local_stats.rx_bytes;
293 stats->rx_packets += local_stats.rx_packets;
294 stats->tx_bytes += local_stats.tx_bytes;
295 stats->tx_packets += local_stats.tx_packets;
300 * ovs_vport_get_options - retrieve device options
302 * @vport: vport from which to retrieve the options.
303 * @skb: sk_buff where options should be appended.
305 * Retrieves the configuration of the given device, appending an
306 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
307 * vport-specific attributes to @skb.
309 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
310 * negative error code if a real error occurred. If an error occurs, @skb is
313 * Must be called with ovs_mutex or rcu_read_lock.
315 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
320 if (!vport->ops->get_options)
323 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
327 err = vport->ops->get_options(vport, skb);
329 nla_nest_cancel(skb, nla);
333 nla_nest_end(skb, nla);
338 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
340 * @vport: vport to modify.
341 * @ids: new configuration, an array of port ids.
343 * Sets the vport's upcall_portids to @ids.
345 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
346 * as an array of U32.
348 * Must be called with ovs_mutex.
350 int ovs_vport_set_upcall_portids(struct vport *vport, struct nlattr *ids)
352 struct vport_portids *old, *vport_portids;
354 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
357 old = ovsl_dereference(vport->upcall_portids);
359 vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
364 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
365 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
366 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
368 rcu_assign_pointer(vport->upcall_portids, vport_portids);
376 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
378 * @vport: vport from which to retrieve the portids.
379 * @skb: sk_buff where portids should be appended.
381 * Retrieves the configuration of the given vport, appending the
382 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
385 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
386 * If an error occurs, @skb is left unmodified. Must be called with
387 * ovs_mutex or rcu_read_lock.
389 int ovs_vport_get_upcall_portids(const struct vport *vport,
392 struct vport_portids *ids;
394 ids = rcu_dereference_ovsl(vport->upcall_portids);
396 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
397 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
398 ids->n_ids * sizeof(u32), (void *)ids->ids);
400 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
404 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
406 * @vport: vport from which the missed packet is received.
407 * @skb: skb that the missed packet was received.
409 * Uses the skb_get_hash() to select the upcall portid to send the
412 * Returns the portid of the target socket. Must be called with rcu_read_lock.
414 u32 ovs_vport_find_upcall_portid(const struct vport *p, struct sk_buff *skb)
416 struct vport_portids *ids;
420 ids = rcu_dereference(p->upcall_portids);
422 if (ids->n_ids == 1 && ids->ids[0] == 0)
425 hash = skb_get_hash(skb);
426 ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
427 return ids->ids[ids_index];
431 * ovs_vport_receive - pass up received packet to the datapath for processing
433 * @vport: vport that received the packet
434 * @skb: skb that was received
435 * @tun_key: tunnel (if any) that carried packet
437 * Must be called with rcu_read_lock. The packet cannot be shared and
438 * skb->data should point to the Ethernet header.
440 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
441 struct ovs_key_ipv4_tunnel *tun_key)
443 struct pcpu_sw_netstats *stats;
445 stats = this_cpu_ptr(vport->percpu_stats);
446 u64_stats_update_begin(&stats->syncp);
448 stats->rx_bytes += skb->len;
449 u64_stats_update_end(&stats->syncp);
451 OVS_CB(skb)->tun_key = tun_key;
452 ovs_dp_process_received_packet(vport, skb);
456 * ovs_vport_send - send a packet on a device
458 * @vport: vport on which to send the packet
461 * Sends the given packet and returns the length of data sent. Either ovs
462 * lock or rcu_read_lock must be held.
464 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
466 int sent = vport->ops->send(vport, skb);
468 if (likely(sent > 0)) {
469 struct pcpu_sw_netstats *stats;
471 stats = this_cpu_ptr(vport->percpu_stats);
473 u64_stats_update_begin(&stats->syncp);
475 stats->tx_bytes += sent;
476 u64_stats_update_end(&stats->syncp);
477 } else if (sent < 0) {
478 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
481 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
487 * ovs_vport_record_error - indicate device error to generic stats layer
489 * @vport: vport that encountered the error
490 * @err_type: one of enum vport_err_type types to indicate the error type
492 * If using the vport generic stats layer indicate that an error of the given
495 static void ovs_vport_record_error(struct vport *vport,
496 enum vport_err_type err_type)
498 spin_lock(&vport->stats_lock);
501 case VPORT_E_RX_DROPPED:
502 vport->err_stats.rx_dropped++;
505 case VPORT_E_RX_ERROR:
506 vport->err_stats.rx_errors++;
509 case VPORT_E_TX_DROPPED:
510 vport->err_stats.tx_dropped++;
513 case VPORT_E_TX_ERROR:
514 vport->err_stats.tx_errors++;
518 spin_unlock(&vport->stats_lock);
521 static void free_vport_rcu(struct rcu_head *rcu)
523 struct vport *vport = container_of(rcu, struct vport, rcu);
525 ovs_vport_free(vport);
528 void ovs_vport_deferred_free(struct vport *vport)
533 call_rcu(&vport->rcu, free_vport_rcu);