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 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
54 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
55 static struct hlist_head *dev_table;
56 #define VPORT_HASH_BUCKETS 1024
59 * ovs_vport_init - initialize vport subsystem
61 * Called at module load time to initialize the vport subsystem.
63 int ovs_vport_init(void)
65 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
74 * ovs_vport_exit - shutdown vport subsystem
76 * Called at module exit time to shutdown the vport subsystem.
78 void ovs_vport_exit(void)
83 static struct hlist_head *hash_bucket(struct net *net, const char *name)
85 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
86 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
90 * ovs_vport_locate - find a port that has already been created
92 * @name: name of port to find
94 * Must be called with ovs or RCU read lock.
96 struct vport *ovs_vport_locate(struct net *net, const char *name)
98 struct hlist_head *bucket = hash_bucket(net, name);
101 hlist_for_each_entry_rcu(vport, bucket, hash_node)
102 if (!strcmp(name, vport->ops->get_name(vport)) &&
103 net_eq(ovs_dp_get_net(vport->dp), net))
110 * ovs_vport_alloc - allocate and initialize new vport
112 * @priv_size: Size of private data area to allocate.
113 * @ops: vport device ops
115 * Allocate and initialize a new vport defined by @ops. The vport will contain
116 * a private data area of size @priv_size that can be accessed using
117 * vport_priv(). vports that are no longer needed should be released with
120 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
121 const struct vport_parms *parms)
126 alloc_size = sizeof(struct vport);
128 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
129 alloc_size += priv_size;
132 vport = kzalloc(alloc_size, GFP_KERNEL);
134 return ERR_PTR(-ENOMEM);
136 vport->dp = parms->dp;
137 vport->port_no = parms->port_no;
138 vport->upcall_portid = parms->upcall_portid;
140 INIT_HLIST_NODE(&vport->dp_hash_node);
142 vport->percpu_stats = alloc_percpu(struct pcpu_tstats);
143 if (!vport->percpu_stats) {
145 return ERR_PTR(-ENOMEM);
148 spin_lock_init(&vport->stats_lock);
154 * ovs_vport_free - uninitialize and free vport
156 * @vport: vport to free
158 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
160 * The caller must ensure that an RCU grace period has passed since the last
161 * time @vport was in a datapath.
163 void ovs_vport_free(struct vport *vport)
165 free_percpu(vport->percpu_stats);
170 * ovs_vport_add - add vport device (for kernel callers)
172 * @parms: Information about new vport.
174 * Creates a new vport with the specified configuration (which is dependent on
175 * device type). ovs_mutex must be held.
177 struct vport *ovs_vport_add(const struct vport_parms *parms)
183 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
184 if (vport_ops_list[i]->type == parms->type) {
185 struct hlist_head *bucket;
187 vport = vport_ops_list[i]->create(parms);
189 err = PTR_ERR(vport);
193 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
194 vport->ops->get_name(vport));
195 hlist_add_head_rcu(&vport->hash_node, bucket);
207 * ovs_vport_set_options - modify existing vport device (for kernel callers)
209 * @vport: vport to modify.
210 * @options: New configuration.
212 * Modifies an existing device with the specified configuration (which is
213 * dependent on device type). ovs_mutex must be held.
215 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
217 if (!vport->ops->set_options)
219 return vport->ops->set_options(vport, options);
223 * ovs_vport_del - delete existing vport device
225 * @vport: vport to delete.
227 * Detaches @vport from its datapath and destroys it. It is possible to fail
228 * for reasons such as lack of memory. ovs_mutex must be held.
230 void ovs_vport_del(struct vport *vport)
234 hlist_del_rcu(&vport->hash_node);
235 vport->ops->destroy(vport);
239 * ovs_vport_set_stats - sets offset device stats
241 * @vport: vport on which to set stats
242 * @stats: stats to set
244 * Provides a set of transmit, receive, and error stats to be added as an
245 * offset to the collected data when stats are retrieved. Some devices may not
246 * support setting the stats, in which case the result will always be
249 * Must be called with ovs_mutex.
251 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
253 spin_lock_bh(&vport->stats_lock);
254 vport->offset_stats = *stats;
255 spin_unlock_bh(&vport->stats_lock);
259 * ovs_vport_get_stats - retrieve device stats
261 * @vport: vport from which to retrieve the stats
262 * @stats: location to store stats
264 * Retrieves transmit, receive, and error stats for the given device.
266 * Must be called with ovs_mutex or rcu_read_lock.
268 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
272 /* We potentially have 3 sources of stats that need to be
273 * combined: those we have collected (split into err_stats and
274 * percpu_stats), offset_stats from set_stats(), and device
275 * error stats from netdev->get_stats() (for errors that happen
276 * downstream and therefore aren't reported through our
277 * vport_record_error() function).
278 * Stats from first two sources are merged and reported by ovs over
279 * OVS_VPORT_ATTR_STATS.
280 * netdev-stats can be directly read over netlink-ioctl.
283 spin_lock_bh(&vport->stats_lock);
285 *stats = vport->offset_stats;
287 stats->rx_errors += vport->err_stats.rx_errors;
288 stats->tx_errors += vport->err_stats.tx_errors;
289 stats->tx_dropped += vport->err_stats.tx_dropped;
290 stats->rx_dropped += vport->err_stats.rx_dropped;
292 spin_unlock_bh(&vport->stats_lock);
294 for_each_possible_cpu(i) {
295 const struct pcpu_tstats *percpu_stats;
296 struct pcpu_tstats local_stats;
299 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
302 start = u64_stats_fetch_begin_bh(&percpu_stats->syncp);
303 local_stats = *percpu_stats;
304 } while (u64_stats_fetch_retry_bh(&percpu_stats->syncp, start));
306 stats->rx_bytes += local_stats.rx_bytes;
307 stats->rx_packets += local_stats.rx_packets;
308 stats->tx_bytes += local_stats.tx_bytes;
309 stats->tx_packets += local_stats.tx_packets;
314 * ovs_vport_get_options - retrieve device options
316 * @vport: vport from which to retrieve the options.
317 * @skb: sk_buff where options should be appended.
319 * Retrieves the configuration of the given device, appending an
320 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
321 * vport-specific attributes to @skb.
323 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
324 * negative error code if a real error occurred. If an error occurs, @skb is
327 * Must be called with ovs_mutex or rcu_read_lock.
329 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
334 if (!vport->ops->get_options)
337 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
341 err = vport->ops->get_options(vport, skb);
343 nla_nest_cancel(skb, nla);
347 nla_nest_end(skb, nla);
352 * ovs_vport_receive - pass up received packet to the datapath for processing
354 * @vport: vport that received the packet
355 * @skb: skb that was received
356 * @tun_key: tunnel (if any) that carried packet
358 * Must be called with rcu_read_lock. The packet cannot be shared and
359 * skb->data should point to the Ethernet header. The caller must have already
360 * called compute_ip_summed() to initialize the checksumming fields.
362 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
363 struct ovs_key_ipv4_tunnel *tun_key)
365 struct pcpu_tstats *stats;
367 stats = this_cpu_ptr(vport->percpu_stats);
368 u64_stats_update_begin(&stats->syncp);
370 stats->rx_bytes += skb->len;
371 u64_stats_update_end(&stats->syncp);
373 OVS_CB(skb)->tun_key = tun_key;
374 ovs_dp_process_received_packet(vport, skb);
378 * ovs_vport_send - send a packet on a device
380 * @vport: vport on which to send the packet
383 * Sends the given packet and returns the length of data sent. Either ovs
384 * lock or rcu_read_lock must be held.
386 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
388 int sent = vport->ops->send(vport, skb);
390 if (likely(sent > 0)) {
391 struct pcpu_tstats *stats;
393 stats = this_cpu_ptr(vport->percpu_stats);
395 u64_stats_update_begin(&stats->syncp);
397 stats->tx_bytes += sent;
398 u64_stats_update_end(&stats->syncp);
399 } else if (sent < 0) {
400 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
403 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
409 * ovs_vport_record_error - indicate device error to generic stats layer
411 * @vport: vport that encountered the error
412 * @err_type: one of enum vport_err_type types to indicate the error type
414 * If using the vport generic stats layer indicate that an error of the given
417 static void ovs_vport_record_error(struct vport *vport,
418 enum vport_err_type err_type)
420 spin_lock(&vport->stats_lock);
423 case VPORT_E_RX_DROPPED:
424 vport->err_stats.rx_dropped++;
427 case VPORT_E_RX_ERROR:
428 vport->err_stats.rx_errors++;
431 case VPORT_E_TX_DROPPED:
432 vport->err_stats.tx_dropped++;
435 case VPORT_E_TX_ERROR:
436 vport->err_stats.tx_errors++;
440 spin_unlock(&vport->stats_lock);
443 static void free_vport_rcu(struct rcu_head *rcu)
445 struct vport *vport = container_of(rcu, struct vport, rcu);
447 ovs_vport_free(vport);
450 void ovs_vport_deferred_free(struct vport *vport)
455 call_rcu(&vport->rcu, free_vport_rcu);