Loading...
1/*
2 * Copyright (c) 2007-2011 Nicira Networks.
3 *
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.
7 *
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.
12 *
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
16 * 02110-1301, USA
17 */
18
19#include <linux/dcache.h>
20#include <linux/etherdevice.h>
21#include <linux/if.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
31#include "vport.h"
32#include "vport-internal_dev.h"
33
34/* List of statically compiled vport implementations. Don't forget to also
35 * add yours to the list at the bottom of vport.h. */
36static const struct vport_ops *vport_ops_list[] = {
37 &ovs_netdev_vport_ops,
38 &ovs_internal_vport_ops,
39};
40
41/* Protected by RCU read lock for reading, RTNL lock for writing. */
42static struct hlist_head *dev_table;
43#define VPORT_HASH_BUCKETS 1024
44
45/**
46 * ovs_vport_init - initialize vport subsystem
47 *
48 * Called at module load time to initialize the vport subsystem.
49 */
50int ovs_vport_init(void)
51{
52 dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
53 GFP_KERNEL);
54 if (!dev_table)
55 return -ENOMEM;
56
57 return 0;
58}
59
60/**
61 * ovs_vport_exit - shutdown vport subsystem
62 *
63 * Called at module exit time to shutdown the vport subsystem.
64 */
65void ovs_vport_exit(void)
66{
67 kfree(dev_table);
68}
69
70static struct hlist_head *hash_bucket(const char *name)
71{
72 unsigned int hash = full_name_hash(name, strlen(name));
73 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
74}
75
76/**
77 * ovs_vport_locate - find a port that has already been created
78 *
79 * @name: name of port to find
80 *
81 * Must be called with RTNL or RCU read lock.
82 */
83struct vport *ovs_vport_locate(const char *name)
84{
85 struct hlist_head *bucket = hash_bucket(name);
86 struct vport *vport;
87 struct hlist_node *node;
88
89 hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
90 if (!strcmp(name, vport->ops->get_name(vport)))
91 return vport;
92
93 return NULL;
94}
95
96/**
97 * ovs_vport_alloc - allocate and initialize new vport
98 *
99 * @priv_size: Size of private data area to allocate.
100 * @ops: vport device ops
101 *
102 * Allocate and initialize a new vport defined by @ops. The vport will contain
103 * a private data area of size @priv_size that can be accessed using
104 * vport_priv(). vports that are no longer needed should be released with
105 * vport_free().
106 */
107struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
108 const struct vport_parms *parms)
109{
110 struct vport *vport;
111 size_t alloc_size;
112
113 alloc_size = sizeof(struct vport);
114 if (priv_size) {
115 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
116 alloc_size += priv_size;
117 }
118
119 vport = kzalloc(alloc_size, GFP_KERNEL);
120 if (!vport)
121 return ERR_PTR(-ENOMEM);
122
123 vport->dp = parms->dp;
124 vport->port_no = parms->port_no;
125 vport->upcall_pid = parms->upcall_pid;
126 vport->ops = ops;
127
128 vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
129 if (!vport->percpu_stats) {
130 kfree(vport);
131 return ERR_PTR(-ENOMEM);
132 }
133
134 spin_lock_init(&vport->stats_lock);
135
136 return vport;
137}
138
139/**
140 * ovs_vport_free - uninitialize and free vport
141 *
142 * @vport: vport to free
143 *
144 * Frees a vport allocated with vport_alloc() when it is no longer needed.
145 *
146 * The caller must ensure that an RCU grace period has passed since the last
147 * time @vport was in a datapath.
148 */
149void ovs_vport_free(struct vport *vport)
150{
151 free_percpu(vport->percpu_stats);
152 kfree(vport);
153}
154
155/**
156 * ovs_vport_add - add vport device (for kernel callers)
157 *
158 * @parms: Information about new vport.
159 *
160 * Creates a new vport with the specified configuration (which is dependent on
161 * device type). RTNL lock must be held.
162 */
163struct vport *ovs_vport_add(const struct vport_parms *parms)
164{
165 struct vport *vport;
166 int err = 0;
167 int i;
168
169 ASSERT_RTNL();
170
171 for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
172 if (vport_ops_list[i]->type == parms->type) {
173 vport = vport_ops_list[i]->create(parms);
174 if (IS_ERR(vport)) {
175 err = PTR_ERR(vport);
176 goto out;
177 }
178
179 hlist_add_head_rcu(&vport->hash_node,
180 hash_bucket(vport->ops->get_name(vport)));
181 return vport;
182 }
183 }
184
185 err = -EAFNOSUPPORT;
186
187out:
188 return ERR_PTR(err);
189}
190
191/**
192 * ovs_vport_set_options - modify existing vport device (for kernel callers)
193 *
194 * @vport: vport to modify.
195 * @port: New configuration.
196 *
197 * Modifies an existing device with the specified configuration (which is
198 * dependent on device type). RTNL lock must be held.
199 */
200int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
201{
202 ASSERT_RTNL();
203
204 if (!vport->ops->set_options)
205 return -EOPNOTSUPP;
206 return vport->ops->set_options(vport, options);
207}
208
209/**
210 * ovs_vport_del - delete existing vport device
211 *
212 * @vport: vport to delete.
213 *
214 * Detaches @vport from its datapath and destroys it. It is possible to fail
215 * for reasons such as lack of memory. RTNL lock must be held.
216 */
217void ovs_vport_del(struct vport *vport)
218{
219 ASSERT_RTNL();
220
221 hlist_del_rcu(&vport->hash_node);
222
223 vport->ops->destroy(vport);
224}
225
226/**
227 * ovs_vport_get_stats - retrieve device stats
228 *
229 * @vport: vport from which to retrieve the stats
230 * @stats: location to store stats
231 *
232 * Retrieves transmit, receive, and error stats for the given device.
233 *
234 * Must be called with RTNL lock or rcu_read_lock.
235 */
236void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
237{
238 int i;
239
240 memset(stats, 0, sizeof(*stats));
241
242 /* We potentially have 2 sources of stats that need to be combined:
243 * those we have collected (split into err_stats and percpu_stats) from
244 * set_stats() and device error stats from netdev->get_stats() (for
245 * errors that happen downstream and therefore aren't reported through
246 * our vport_record_error() function).
247 * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
248 * netdev-stats can be directly read over netlink-ioctl.
249 */
250
251 spin_lock_bh(&vport->stats_lock);
252
253 stats->rx_errors = vport->err_stats.rx_errors;
254 stats->tx_errors = vport->err_stats.tx_errors;
255 stats->tx_dropped = vport->err_stats.tx_dropped;
256 stats->rx_dropped = vport->err_stats.rx_dropped;
257
258 spin_unlock_bh(&vport->stats_lock);
259
260 for_each_possible_cpu(i) {
261 const struct vport_percpu_stats *percpu_stats;
262 struct vport_percpu_stats local_stats;
263 unsigned int start;
264
265 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
266
267 do {
268 start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
269 local_stats = *percpu_stats;
270 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
271
272 stats->rx_bytes += local_stats.rx_bytes;
273 stats->rx_packets += local_stats.rx_packets;
274 stats->tx_bytes += local_stats.tx_bytes;
275 stats->tx_packets += local_stats.tx_packets;
276 }
277}
278
279/**
280 * ovs_vport_get_options - retrieve device options
281 *
282 * @vport: vport from which to retrieve the options.
283 * @skb: sk_buff where options should be appended.
284 *
285 * Retrieves the configuration of the given device, appending an
286 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
287 * vport-specific attributes to @skb.
288 *
289 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
290 * negative error code if a real error occurred. If an error occurs, @skb is
291 * left unmodified.
292 *
293 * Must be called with RTNL lock or rcu_read_lock.
294 */
295int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
296{
297 struct nlattr *nla;
298
299 nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
300 if (!nla)
301 return -EMSGSIZE;
302
303 if (vport->ops->get_options) {
304 int err = vport->ops->get_options(vport, skb);
305 if (err) {
306 nla_nest_cancel(skb, nla);
307 return err;
308 }
309 }
310
311 nla_nest_end(skb, nla);
312 return 0;
313}
314
315/**
316 * ovs_vport_receive - pass up received packet to the datapath for processing
317 *
318 * @vport: vport that received the packet
319 * @skb: skb that was received
320 *
321 * Must be called with rcu_read_lock. The packet cannot be shared and
322 * skb->data should point to the Ethernet header. The caller must have already
323 * called compute_ip_summed() to initialize the checksumming fields.
324 */
325void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
326{
327 struct vport_percpu_stats *stats;
328
329 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
330
331 u64_stats_update_begin(&stats->sync);
332 stats->rx_packets++;
333 stats->rx_bytes += skb->len;
334 u64_stats_update_end(&stats->sync);
335
336 ovs_dp_process_received_packet(vport, skb);
337}
338
339/**
340 * ovs_vport_send - send a packet on a device
341 *
342 * @vport: vport on which to send the packet
343 * @skb: skb to send
344 *
345 * Sends the given packet and returns the length of data sent. Either RTNL
346 * lock or rcu_read_lock must be held.
347 */
348int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
349{
350 int sent = vport->ops->send(vport, skb);
351
352 if (likely(sent)) {
353 struct vport_percpu_stats *stats;
354
355 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
356
357 u64_stats_update_begin(&stats->sync);
358 stats->tx_packets++;
359 stats->tx_bytes += sent;
360 u64_stats_update_end(&stats->sync);
361 }
362 return sent;
363}
364
365/**
366 * ovs_vport_record_error - indicate device error to generic stats layer
367 *
368 * @vport: vport that encountered the error
369 * @err_type: one of enum vport_err_type types to indicate the error type
370 *
371 * If using the vport generic stats layer indicate that an error of the given
372 * type has occured.
373 */
374void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
375{
376 spin_lock(&vport->stats_lock);
377
378 switch (err_type) {
379 case VPORT_E_RX_DROPPED:
380 vport->err_stats.rx_dropped++;
381 break;
382
383 case VPORT_E_RX_ERROR:
384 vport->err_stats.rx_errors++;
385 break;
386
387 case VPORT_E_TX_DROPPED:
388 vport->err_stats.tx_dropped++;
389 break;
390
391 case VPORT_E_TX_ERROR:
392 vport->err_stats.tx_errors++;
393 break;
394 };
395
396 spin_unlock(&vport->stats_lock);
397}
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Copyright (c) 2007-2014 Nicira, Inc.
4 */
5
6#include <linux/etherdevice.h>
7#include <linux/if.h>
8#include <linux/if_vlan.h>
9#include <linux/jhash.h>
10#include <linux/kernel.h>
11#include <linux/list.h>
12#include <linux/mutex.h>
13#include <linux/percpu.h>
14#include <linux/rcupdate.h>
15#include <linux/rtnetlink.h>
16#include <linux/compat.h>
17#include <net/net_namespace.h>
18#include <linux/module.h>
19
20#include "datapath.h"
21#include "vport.h"
22#include "vport-internal_dev.h"
23
24static LIST_HEAD(vport_ops_list);
25
26/* Protected by RCU read lock for reading, ovs_mutex for writing. */
27static struct hlist_head *dev_table;
28#define VPORT_HASH_BUCKETS 1024
29
30/**
31 * ovs_vport_init - initialize vport subsystem
32 *
33 * Called at module load time to initialize the vport subsystem.
34 */
35int ovs_vport_init(void)
36{
37 dev_table = kcalloc(VPORT_HASH_BUCKETS, sizeof(struct hlist_head),
38 GFP_KERNEL);
39 if (!dev_table)
40 return -ENOMEM;
41
42 return 0;
43}
44
45/**
46 * ovs_vport_exit - shutdown vport subsystem
47 *
48 * Called at module exit time to shutdown the vport subsystem.
49 */
50void ovs_vport_exit(void)
51{
52 kfree(dev_table);
53}
54
55static struct hlist_head *hash_bucket(const struct net *net, const char *name)
56{
57 unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
58 return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
59}
60
61int __ovs_vport_ops_register(struct vport_ops *ops)
62{
63 int err = -EEXIST;
64 struct vport_ops *o;
65
66 ovs_lock();
67 list_for_each_entry(o, &vport_ops_list, list)
68 if (ops->type == o->type)
69 goto errout;
70
71 list_add_tail(&ops->list, &vport_ops_list);
72 err = 0;
73errout:
74 ovs_unlock();
75 return err;
76}
77EXPORT_SYMBOL_GPL(__ovs_vport_ops_register);
78
79void ovs_vport_ops_unregister(struct vport_ops *ops)
80{
81 ovs_lock();
82 list_del(&ops->list);
83 ovs_unlock();
84}
85EXPORT_SYMBOL_GPL(ovs_vport_ops_unregister);
86
87/**
88 * ovs_vport_locate - find a port that has already been created
89 *
90 * @net: network namespace
91 * @name: name of port to find
92 *
93 * Must be called with ovs or RCU read lock.
94 */
95struct vport *ovs_vport_locate(const struct net *net, const char *name)
96{
97 struct hlist_head *bucket = hash_bucket(net, name);
98 struct vport *vport;
99
100 hlist_for_each_entry_rcu(vport, bucket, hash_node,
101 lockdep_ovsl_is_held())
102 if (!strcmp(name, ovs_vport_name(vport)) &&
103 net_eq(ovs_dp_get_net(vport->dp), net))
104 return vport;
105
106 return NULL;
107}
108
109/**
110 * ovs_vport_alloc - allocate and initialize new vport
111 *
112 * @priv_size: Size of private data area to allocate.
113 * @ops: vport device ops
114 *
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
118 * vport_free().
119 */
120struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
121 const struct vport_parms *parms)
122{
123 struct vport *vport;
124 size_t alloc_size;
125
126 alloc_size = sizeof(struct vport);
127 if (priv_size) {
128 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
129 alloc_size += priv_size;
130 }
131
132 vport = kzalloc(alloc_size, GFP_KERNEL);
133 if (!vport)
134 return ERR_PTR(-ENOMEM);
135
136 vport->dp = parms->dp;
137 vport->port_no = parms->port_no;
138 vport->ops = ops;
139 INIT_HLIST_NODE(&vport->dp_hash_node);
140
141 if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids)) {
142 kfree(vport);
143 return ERR_PTR(-EINVAL);
144 }
145
146 return vport;
147}
148EXPORT_SYMBOL_GPL(ovs_vport_alloc);
149
150/**
151 * ovs_vport_free - uninitialize and free vport
152 *
153 * @vport: vport to free
154 *
155 * Frees a vport allocated with vport_alloc() when it is no longer needed.
156 *
157 * The caller must ensure that an RCU grace period has passed since the last
158 * time @vport was in a datapath.
159 */
160void ovs_vport_free(struct vport *vport)
161{
162 /* vport is freed from RCU callback or error path, Therefore
163 * it is safe to use raw dereference.
164 */
165 kfree(rcu_dereference_raw(vport->upcall_portids));
166 kfree(vport);
167}
168EXPORT_SYMBOL_GPL(ovs_vport_free);
169
170static struct vport_ops *ovs_vport_lookup(const struct vport_parms *parms)
171{
172 struct vport_ops *ops;
173
174 list_for_each_entry(ops, &vport_ops_list, list)
175 if (ops->type == parms->type)
176 return ops;
177
178 return NULL;
179}
180
181/**
182 * ovs_vport_add - add vport device (for kernel callers)
183 *
184 * @parms: Information about new vport.
185 *
186 * Creates a new vport with the specified configuration (which is dependent on
187 * device type). ovs_mutex must be held.
188 */
189struct vport *ovs_vport_add(const struct vport_parms *parms)
190{
191 struct vport_ops *ops;
192 struct vport *vport;
193
194 ops = ovs_vport_lookup(parms);
195 if (ops) {
196 struct hlist_head *bucket;
197
198 if (!try_module_get(ops->owner))
199 return ERR_PTR(-EAFNOSUPPORT);
200
201 vport = ops->create(parms);
202 if (IS_ERR(vport)) {
203 module_put(ops->owner);
204 return vport;
205 }
206
207 bucket = hash_bucket(ovs_dp_get_net(vport->dp),
208 ovs_vport_name(vport));
209 hlist_add_head_rcu(&vport->hash_node, bucket);
210 return vport;
211 }
212
213 /* Unlock to attempt module load and return -EAGAIN if load
214 * was successful as we need to restart the port addition
215 * workflow.
216 */
217 ovs_unlock();
218 request_module("vport-type-%d", parms->type);
219 ovs_lock();
220
221 if (!ovs_vport_lookup(parms))
222 return ERR_PTR(-EAFNOSUPPORT);
223 else
224 return ERR_PTR(-EAGAIN);
225}
226
227/**
228 * ovs_vport_set_options - modify existing vport device (for kernel callers)
229 *
230 * @vport: vport to modify.
231 * @options: New configuration.
232 *
233 * Modifies an existing device with the specified configuration (which is
234 * dependent on device type). ovs_mutex must be held.
235 */
236int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
237{
238 if (!vport->ops->set_options)
239 return -EOPNOTSUPP;
240 return vport->ops->set_options(vport, options);
241}
242
243/**
244 * ovs_vport_del - delete existing vport device
245 *
246 * @vport: vport to delete.
247 *
248 * Detaches @vport from its datapath and destroys it. ovs_mutex must
249 * be held.
250 */
251void ovs_vport_del(struct vport *vport)
252{
253 hlist_del_rcu(&vport->hash_node);
254 module_put(vport->ops->owner);
255 vport->ops->destroy(vport);
256}
257
258/**
259 * ovs_vport_get_stats - retrieve device stats
260 *
261 * @vport: vport from which to retrieve the stats
262 * @stats: location to store stats
263 *
264 * Retrieves transmit, receive, and error stats for the given device.
265 *
266 * Must be called with ovs_mutex or rcu_read_lock.
267 */
268void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
269{
270 const struct rtnl_link_stats64 *dev_stats;
271 struct rtnl_link_stats64 temp;
272
273 dev_stats = dev_get_stats(vport->dev, &temp);
274 stats->rx_errors = dev_stats->rx_errors;
275 stats->tx_errors = dev_stats->tx_errors;
276 stats->tx_dropped = dev_stats->tx_dropped;
277 stats->rx_dropped = dev_stats->rx_dropped;
278
279 stats->rx_bytes = dev_stats->rx_bytes;
280 stats->rx_packets = dev_stats->rx_packets;
281 stats->tx_bytes = dev_stats->tx_bytes;
282 stats->tx_packets = dev_stats->tx_packets;
283}
284
285/**
286 * ovs_vport_get_options - retrieve device options
287 *
288 * @vport: vport from which to retrieve the options.
289 * @skb: sk_buff where options should be appended.
290 *
291 * Retrieves the configuration of the given device, appending an
292 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
293 * vport-specific attributes to @skb.
294 *
295 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
296 * negative error code if a real error occurred. If an error occurs, @skb is
297 * left unmodified.
298 *
299 * Must be called with ovs_mutex or rcu_read_lock.
300 */
301int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
302{
303 struct nlattr *nla;
304 int err;
305
306 if (!vport->ops->get_options)
307 return 0;
308
309 nla = nla_nest_start_noflag(skb, OVS_VPORT_ATTR_OPTIONS);
310 if (!nla)
311 return -EMSGSIZE;
312
313 err = vport->ops->get_options(vport, skb);
314 if (err) {
315 nla_nest_cancel(skb, nla);
316 return err;
317 }
318
319 nla_nest_end(skb, nla);
320 return 0;
321}
322
323/**
324 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
325 *
326 * @vport: vport to modify.
327 * @ids: new configuration, an array of port ids.
328 *
329 * Sets the vport's upcall_portids to @ids.
330 *
331 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
332 * as an array of U32.
333 *
334 * Must be called with ovs_mutex.
335 */
336int ovs_vport_set_upcall_portids(struct vport *vport, const struct nlattr *ids)
337{
338 struct vport_portids *old, *vport_portids;
339
340 if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
341 return -EINVAL;
342
343 old = ovsl_dereference(vport->upcall_portids);
344
345 vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
346 GFP_KERNEL);
347 if (!vport_portids)
348 return -ENOMEM;
349
350 vport_portids->n_ids = nla_len(ids) / sizeof(u32);
351 vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
352 nla_memcpy(vport_portids->ids, ids, nla_len(ids));
353
354 rcu_assign_pointer(vport->upcall_portids, vport_portids);
355
356 if (old)
357 kfree_rcu(old, rcu);
358 return 0;
359}
360
361/**
362 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
363 *
364 * @vport: vport from which to retrieve the portids.
365 * @skb: sk_buff where portids should be appended.
366 *
367 * Retrieves the configuration of the given vport, appending the
368 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
369 * portids to @skb.
370 *
371 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
372 * If an error occurs, @skb is left unmodified. Must be called with
373 * ovs_mutex or rcu_read_lock.
374 */
375int ovs_vport_get_upcall_portids(const struct vport *vport,
376 struct sk_buff *skb)
377{
378 struct vport_portids *ids;
379
380 ids = rcu_dereference_ovsl(vport->upcall_portids);
381
382 if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
383 return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
384 ids->n_ids * sizeof(u32), (void *)ids->ids);
385 else
386 return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
387}
388
389/**
390 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
391 *
392 * @vport: vport from which the missed packet is received.
393 * @skb: skb that the missed packet was received.
394 *
395 * Uses the skb_get_hash() to select the upcall portid to send the
396 * upcall.
397 *
398 * Returns the portid of the target socket. Must be called with rcu_read_lock.
399 */
400u32 ovs_vport_find_upcall_portid(const struct vport *vport, struct sk_buff *skb)
401{
402 struct vport_portids *ids;
403 u32 ids_index;
404 u32 hash;
405
406 ids = rcu_dereference(vport->upcall_portids);
407
408 /* If there is only one portid, select it in the fast-path. */
409 if (ids->n_ids == 1)
410 return ids->ids[0];
411
412 hash = skb_get_hash(skb);
413 ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
414 return ids->ids[ids_index];
415}
416
417/**
418 * ovs_vport_receive - pass up received packet to the datapath for processing
419 *
420 * @vport: vport that received the packet
421 * @skb: skb that was received
422 * @tun_info: tunnel (if any) that carried packet
423 *
424 * Must be called with rcu_read_lock. The packet cannot be shared and
425 * skb->data should point to the Ethernet header.
426 */
427int ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
428 const struct ip_tunnel_info *tun_info)
429{
430 struct sw_flow_key key;
431 int error;
432
433 OVS_CB(skb)->input_vport = vport;
434 OVS_CB(skb)->mru = 0;
435 OVS_CB(skb)->cutlen = 0;
436 if (unlikely(dev_net(skb->dev) != ovs_dp_get_net(vport->dp))) {
437 u32 mark;
438
439 mark = skb->mark;
440 skb_scrub_packet(skb, true);
441 skb->mark = mark;
442 tun_info = NULL;
443 }
444
445 /* Extract flow from 'skb' into 'key'. */
446 error = ovs_flow_key_extract(tun_info, skb, &key);
447 if (unlikely(error)) {
448 kfree_skb(skb);
449 return error;
450 }
451 ovs_dp_process_packet(skb, &key);
452 return 0;
453}
454
455static int packet_length(const struct sk_buff *skb,
456 struct net_device *dev)
457{
458 int length = skb->len - dev->hard_header_len;
459
460 if (!skb_vlan_tag_present(skb) &&
461 eth_type_vlan(skb->protocol))
462 length -= VLAN_HLEN;
463
464 /* Don't subtract for multiple VLAN tags. Most (all?) drivers allow
465 * (ETH_LEN + VLAN_HLEN) in addition to the mtu value, but almost none
466 * account for 802.1ad. e.g. is_skb_forwardable().
467 */
468
469 return length > 0 ? length : 0;
470}
471
472void ovs_vport_send(struct vport *vport, struct sk_buff *skb, u8 mac_proto)
473{
474 int mtu = vport->dev->mtu;
475
476 switch (vport->dev->type) {
477 case ARPHRD_NONE:
478 if (mac_proto == MAC_PROTO_ETHERNET) {
479 skb_reset_network_header(skb);
480 skb_reset_mac_len(skb);
481 skb->protocol = htons(ETH_P_TEB);
482 } else if (mac_proto != MAC_PROTO_NONE) {
483 WARN_ON_ONCE(1);
484 goto drop;
485 }
486 break;
487 case ARPHRD_ETHER:
488 if (mac_proto != MAC_PROTO_ETHERNET)
489 goto drop;
490 break;
491 default:
492 goto drop;
493 }
494
495 if (unlikely(packet_length(skb, vport->dev) > mtu &&
496 !skb_is_gso(skb))) {
497 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
498 vport->dev->name,
499 packet_length(skb, vport->dev), mtu);
500 vport->dev->stats.tx_errors++;
501 goto drop;
502 }
503
504 skb->dev = vport->dev;
505 vport->ops->send(skb);
506 return;
507
508drop:
509 kfree_skb(skb);
510}