Commit | Line | Data |
---|---|---|
ccb1352e | 1 | /* |
caf2ee14 | 2 | * Copyright (c) 2007-2012 Nicira, Inc. |
ccb1352e JG |
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> | |
ccb1352e JG |
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. */ | |
36 | static 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. */ | |
42 | static 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 | */ | |
50 | int 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 | */ | |
65 | void ovs_vport_exit(void) | |
66 | { | |
67 | kfree(dev_table); | |
68 | } | |
69 | ||
70 | static 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 | */ | |
83 | struct 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 | */ | |
107 | struct 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); | |
f0a98ae8 DC |
129 | if (!vport->percpu_stats) { |
130 | kfree(vport); | |
ccb1352e | 131 | return ERR_PTR(-ENOMEM); |
f0a98ae8 | 132 | } |
ccb1352e JG |
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 | */ | |
149 | void 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 | */ | |
163 | struct 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 | ||
187 | out: | |
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 | */ | |
200 | int 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 | */ | |
217 | void 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 | */ | |
236 | void 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 | */ | |
295 | int 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 | */ | |
325 | void 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 | */ | |
348 | int 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 | */ | |
374 | void 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 | } |