2 * Copyright (c) 2007-2011 Nicira Networks.
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/dcache.h>
20 #include <linux/etherdevice.h>
22 #include <linux/if_vlan.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 <linux/version.h>
33 #include "vport-internal_dev.h"
35 /* List of statically compiled vport implementations. Don't forget to also
36 * add yours to the list at the bottom of vport.h. */
37 static const struct vport_ops *vport_ops_list[] = {
38 &ovs_netdev_vport_ops,
39 &ovs_internal_vport_ops,
42 /* Protected by RCU read lock for reading, RTNL lock for writing. */
43 static struct hlist_head *dev_table;
44 #define VPORT_HASH_BUCKETS 1024
47 * ovs_vport_init - initialize vport subsystem
49 * Called at module load time to initialize the vport subsystem.
51 int ovs_vport_init(void)
53 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
62 * ovs_vport_exit - shutdown vport subsystem
64 * Called at module exit time to shutdown the vport subsystem.
66 void ovs_vport_exit(void)
71 static struct hlist_head *hash_bucket(const char *name)
73 unsigned int hash = full_name_hash(name, strlen(name));
74 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
78 * ovs_vport_locate - find a port that has already been created
80 * @name: name of port to find
82 * Must be called with RTNL or RCU read lock.
84 struct vport *ovs_vport_locate(const char *name)
86 struct hlist_head *bucket = hash_bucket(name);
88 struct hlist_node *node;
90 hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
91 if (!strcmp(name, vport->ops->get_name(vport)))
98 * ovs_vport_alloc - allocate and initialize new vport
100 * @priv_size: Size of private data area to allocate.
101 * @ops: vport device ops
103 * Allocate and initialize a new vport defined by @ops. The vport will contain
104 * a private data area of size @priv_size that can be accessed using
105 * vport_priv(). vports that are no longer needed should be released with
108 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
109 const struct vport_parms *parms)
114 alloc_size = sizeof(struct vport);
116 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
117 alloc_size += priv_size;
120 vport = kzalloc(alloc_size, GFP_KERNEL);
122 return ERR_PTR(-ENOMEM);
124 vport->dp = parms->dp;
125 vport->port_no = parms->port_no;
126 vport->upcall_pid = parms->upcall_pid;
129 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
130 if (!vport->percpu_stats) {
132 return ERR_PTR(-ENOMEM);
135 spin_lock_init(&vport->stats_lock);
141 * ovs_vport_free - uninitialize and free vport
143 * @vport: vport to free
145 * Frees a vport allocated with vport_alloc() when it is no longer needed.
147 * The caller must ensure that an RCU grace period has passed since the last
148 * time @vport was in a datapath.
150 void ovs_vport_free(struct vport *vport)
152 free_percpu(vport->percpu_stats);
157 * ovs_vport_add - add vport device (for kernel callers)
159 * @parms: Information about new vport.
161 * Creates a new vport with the specified configuration (which is dependent on
162 * device type). RTNL lock must be held.
164 struct vport *ovs_vport_add(const struct vport_parms *parms)
172 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
173 if (vport_ops_list[i]->type == parms->type) {
174 vport = vport_ops_list[i]->create(parms);
176 err = PTR_ERR(vport);
180 hlist_add_head_rcu(&vport->hash_node,
181 hash_bucket(vport->ops->get_name(vport)));
193 * ovs_vport_set_options - modify existing vport device (for kernel callers)
195 * @vport: vport to modify.
196 * @port: New configuration.
198 * Modifies an existing device with the specified configuration (which is
199 * dependent on device type). RTNL lock must be held.
201 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
205 if (!vport->ops->set_options)
207 return vport->ops->set_options(vport, options);
211 * ovs_vport_del - delete existing vport device
213 * @vport: vport to delete.
215 * Detaches @vport from its datapath and destroys it. It is possible to fail
216 * for reasons such as lack of memory. RTNL lock must be held.
218 void ovs_vport_del(struct vport *vport)
222 hlist_del_rcu(&vport->hash_node);
224 vport->ops->destroy(vport);
228 * ovs_vport_get_stats - retrieve device stats
230 * @vport: vport from which to retrieve the stats
231 * @stats: location to store stats
233 * Retrieves transmit, receive, and error stats for the given device.
235 * Must be called with RTNL lock or rcu_read_lock.
237 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
241 memset(stats, 0, sizeof(*stats));
243 /* We potentially have 2 sources of stats that need to be combined:
244 * those we have collected (split into err_stats and percpu_stats) from
245 * set_stats() and device error stats from netdev->get_stats() (for
246 * errors that happen downstream and therefore aren't reported through
247 * our vport_record_error() function).
248 * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
249 * netdev-stats can be directly read over netlink-ioctl.
252 spin_lock_bh(&vport->stats_lock);
254 stats->rx_errors = vport->err_stats.rx_errors;
255 stats->tx_errors = vport->err_stats.tx_errors;
256 stats->tx_dropped = vport->err_stats.tx_dropped;
257 stats->rx_dropped = vport->err_stats.rx_dropped;
259 spin_unlock_bh(&vport->stats_lock);
261 for_each_possible_cpu(i) {
262 const struct vport_percpu_stats *percpu_stats;
263 struct vport_percpu_stats local_stats;
266 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
269 start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
270 local_stats = *percpu_stats;
271 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
273 stats->rx_bytes += local_stats.rx_bytes;
274 stats->rx_packets += local_stats.rx_packets;
275 stats->tx_bytes += local_stats.tx_bytes;
276 stats->tx_packets += local_stats.tx_packets;
281 * ovs_vport_get_options - retrieve device options
283 * @vport: vport from which to retrieve the options.
284 * @skb: sk_buff where options should be appended.
286 * Retrieves the configuration of the given device, appending an
287 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
288 * vport-specific attributes to @skb.
290 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
291 * negative error code if a real error occurred. If an error occurs, @skb is
294 * Must be called with RTNL lock or rcu_read_lock.
296 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
300 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
304 if (vport->ops->get_options) {
305 int err = vport->ops->get_options(vport, skb);
307 nla_nest_cancel(skb, nla);
312 nla_nest_end(skb, nla);
317 * ovs_vport_receive - pass up received packet to the datapath for processing
319 * @vport: vport that received the packet
320 * @skb: skb that was received
322 * Must be called with rcu_read_lock. The packet cannot be shared and
323 * skb->data should point to the Ethernet header. The caller must have already
324 * called compute_ip_summed() to initialize the checksumming fields.
326 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
328 struct vport_percpu_stats *stats;
330 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
332 u64_stats_update_begin(&stats->sync);
334 stats->rx_bytes += skb->len;
335 u64_stats_update_end(&stats->sync);
337 ovs_dp_process_received_packet(vport, skb);
341 * ovs_vport_send - send a packet on a device
343 * @vport: vport on which to send the packet
346 * Sends the given packet and returns the length of data sent. Either RTNL
347 * lock or rcu_read_lock must be held.
349 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
351 int sent = vport->ops->send(vport, skb);
354 struct vport_percpu_stats *stats;
356 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
358 u64_stats_update_begin(&stats->sync);
360 stats->tx_bytes += sent;
361 u64_stats_update_end(&stats->sync);
367 * ovs_vport_record_error - indicate device error to generic stats layer
369 * @vport: vport that encountered the error
370 * @err_type: one of enum vport_err_type types to indicate the error type
372 * If using the vport generic stats layer indicate that an error of the given
375 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
377 spin_lock(&vport->stats_lock);
380 case VPORT_E_RX_DROPPED:
381 vport->err_stats.rx_dropped++;
384 case VPORT_E_RX_ERROR:
385 vport->err_stats.rx_errors++;
388 case VPORT_E_TX_DROPPED:
389 vport->err_stats.tx_dropped++;
392 case VPORT_E_TX_ERROR:
393 vport->err_stats.tx_errors++;
397 spin_unlock(&vport->stats_lock);