Commit | Line | Data |
---|---|---|
007f790c JP |
1 | /* |
2 | * net/switchdev/switchdev.c - Switch device API | |
7ea6eb3f | 3 | * Copyright (c) 2014-2015 Jiri Pirko <jiri@resnulli.us> |
f8f21471 | 4 | * Copyright (c) 2014-2015 Scott Feldman <sfeldma@gmail.com> |
007f790c JP |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License as published by | |
8 | * the Free Software Foundation; either version 2 of the License, or | |
9 | * (at your option) any later version. | |
10 | */ | |
11 | ||
12 | #include <linux/kernel.h> | |
13 | #include <linux/types.h> | |
14 | #include <linux/init.h> | |
03bf0c28 JP |
15 | #include <linux/mutex.h> |
16 | #include <linux/notifier.h> | |
007f790c | 17 | #include <linux/netdevice.h> |
850d0cbc | 18 | #include <linux/etherdevice.h> |
47f8328b | 19 | #include <linux/if_bridge.h> |
7ea6eb3f | 20 | #include <linux/list.h> |
793f4014 | 21 | #include <linux/workqueue.h> |
5e8d9049 | 22 | #include <net/ip_fib.h> |
007f790c JP |
23 | #include <net/switchdev.h> |
24 | ||
7ea6eb3f JP |
25 | /** |
26 | * switchdev_trans_item_enqueue - Enqueue data item to transaction queue | |
27 | * | |
28 | * @trans: transaction | |
29 | * @data: pointer to data being queued | |
30 | * @destructor: data destructor | |
31 | * @tritem: transaction item being queued | |
32 | * | |
33 | * Enqeueue data item to transaction queue. tritem is typically placed in | |
34 | * cointainter pointed at by data pointer. Destructor is called on | |
35 | * transaction abort and after successful commit phase in case | |
36 | * the caller did not dequeue the item before. | |
37 | */ | |
38 | void switchdev_trans_item_enqueue(struct switchdev_trans *trans, | |
39 | void *data, void (*destructor)(void const *), | |
40 | struct switchdev_trans_item *tritem) | |
41 | { | |
42 | tritem->data = data; | |
43 | tritem->destructor = destructor; | |
44 | list_add_tail(&tritem->list, &trans->item_list); | |
45 | } | |
46 | EXPORT_SYMBOL_GPL(switchdev_trans_item_enqueue); | |
47 | ||
48 | static struct switchdev_trans_item * | |
49 | __switchdev_trans_item_dequeue(struct switchdev_trans *trans) | |
50 | { | |
51 | struct switchdev_trans_item *tritem; | |
52 | ||
53 | if (list_empty(&trans->item_list)) | |
54 | return NULL; | |
55 | tritem = list_first_entry(&trans->item_list, | |
56 | struct switchdev_trans_item, list); | |
57 | list_del(&tritem->list); | |
58 | return tritem; | |
59 | } | |
60 | ||
61 | /** | |
62 | * switchdev_trans_item_dequeue - Dequeue data item from transaction queue | |
63 | * | |
64 | * @trans: transaction | |
65 | */ | |
66 | void *switchdev_trans_item_dequeue(struct switchdev_trans *trans) | |
67 | { | |
68 | struct switchdev_trans_item *tritem; | |
69 | ||
70 | tritem = __switchdev_trans_item_dequeue(trans); | |
71 | BUG_ON(!tritem); | |
72 | return tritem->data; | |
73 | } | |
74 | EXPORT_SYMBOL_GPL(switchdev_trans_item_dequeue); | |
75 | ||
76 | static void switchdev_trans_init(struct switchdev_trans *trans) | |
77 | { | |
78 | INIT_LIST_HEAD(&trans->item_list); | |
79 | } | |
80 | ||
81 | static void switchdev_trans_items_destroy(struct switchdev_trans *trans) | |
82 | { | |
83 | struct switchdev_trans_item *tritem; | |
84 | ||
85 | while ((tritem = __switchdev_trans_item_dequeue(trans))) | |
86 | tritem->destructor(tritem->data); | |
87 | } | |
88 | ||
89 | static void switchdev_trans_items_warn_destroy(struct net_device *dev, | |
90 | struct switchdev_trans *trans) | |
91 | { | |
92 | WARN(!list_empty(&trans->item_list), "%s: transaction item queue is not empty.\n", | |
93 | dev->name); | |
94 | switchdev_trans_items_destroy(trans); | |
95 | } | |
96 | ||
793f4014 JP |
97 | static LIST_HEAD(deferred); |
98 | static DEFINE_SPINLOCK(deferred_lock); | |
99 | ||
100 | typedef void switchdev_deferred_func_t(struct net_device *dev, | |
101 | const void *data); | |
102 | ||
103 | struct switchdev_deferred_item { | |
104 | struct list_head list; | |
105 | struct net_device *dev; | |
106 | switchdev_deferred_func_t *func; | |
107 | unsigned long data[0]; | |
108 | }; | |
109 | ||
110 | static struct switchdev_deferred_item *switchdev_deferred_dequeue(void) | |
111 | { | |
112 | struct switchdev_deferred_item *dfitem; | |
113 | ||
114 | spin_lock_bh(&deferred_lock); | |
115 | if (list_empty(&deferred)) { | |
116 | dfitem = NULL; | |
117 | goto unlock; | |
118 | } | |
119 | dfitem = list_first_entry(&deferred, | |
120 | struct switchdev_deferred_item, list); | |
121 | list_del(&dfitem->list); | |
122 | unlock: | |
123 | spin_unlock_bh(&deferred_lock); | |
124 | return dfitem; | |
125 | } | |
126 | ||
127 | /** | |
128 | * switchdev_deferred_process - Process ops in deferred queue | |
129 | * | |
130 | * Called to flush the ops currently queued in deferred ops queue. | |
131 | * rtnl_lock must be held. | |
132 | */ | |
133 | void switchdev_deferred_process(void) | |
134 | { | |
135 | struct switchdev_deferred_item *dfitem; | |
136 | ||
137 | ASSERT_RTNL(); | |
138 | ||
139 | while ((dfitem = switchdev_deferred_dequeue())) { | |
140 | dfitem->func(dfitem->dev, dfitem->data); | |
141 | dev_put(dfitem->dev); | |
142 | kfree(dfitem); | |
143 | } | |
144 | } | |
145 | EXPORT_SYMBOL_GPL(switchdev_deferred_process); | |
146 | ||
147 | static void switchdev_deferred_process_work(struct work_struct *work) | |
148 | { | |
149 | rtnl_lock(); | |
150 | switchdev_deferred_process(); | |
151 | rtnl_unlock(); | |
152 | } | |
153 | ||
154 | static DECLARE_WORK(deferred_process_work, switchdev_deferred_process_work); | |
155 | ||
156 | static int switchdev_deferred_enqueue(struct net_device *dev, | |
157 | const void *data, size_t data_len, | |
158 | switchdev_deferred_func_t *func) | |
159 | { | |
160 | struct switchdev_deferred_item *dfitem; | |
161 | ||
162 | dfitem = kmalloc(sizeof(*dfitem) + data_len, GFP_ATOMIC); | |
163 | if (!dfitem) | |
164 | return -ENOMEM; | |
165 | dfitem->dev = dev; | |
166 | dfitem->func = func; | |
167 | memcpy(dfitem->data, data, data_len); | |
168 | dev_hold(dev); | |
169 | spin_lock_bh(&deferred_lock); | |
170 | list_add_tail(&dfitem->list, &deferred); | |
171 | spin_unlock_bh(&deferred_lock); | |
172 | schedule_work(&deferred_process_work); | |
173 | return 0; | |
174 | } | |
175 | ||
3094333d SF |
176 | /** |
177 | * switchdev_port_attr_get - Get port attribute | |
178 | * | |
179 | * @dev: port device | |
180 | * @attr: attribute to get | |
181 | */ | |
182 | int switchdev_port_attr_get(struct net_device *dev, struct switchdev_attr *attr) | |
183 | { | |
184 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
185 | struct net_device *lower_dev; | |
186 | struct list_head *iter; | |
187 | struct switchdev_attr first = { | |
1f868398 | 188 | .id = SWITCHDEV_ATTR_ID_UNDEFINED |
3094333d SF |
189 | }; |
190 | int err = -EOPNOTSUPP; | |
191 | ||
192 | if (ops && ops->switchdev_port_attr_get) | |
193 | return ops->switchdev_port_attr_get(dev, attr); | |
194 | ||
195 | if (attr->flags & SWITCHDEV_F_NO_RECURSE) | |
196 | return err; | |
197 | ||
198 | /* Switch device port(s) may be stacked under | |
199 | * bond/team/vlan dev, so recurse down to get attr on | |
200 | * each port. Return -ENODATA if attr values don't | |
201 | * compare across ports. | |
202 | */ | |
203 | ||
204 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
205 | err = switchdev_port_attr_get(lower_dev, attr); | |
206 | if (err) | |
207 | break; | |
1f868398 | 208 | if (first.id == SWITCHDEV_ATTR_ID_UNDEFINED) |
3094333d SF |
209 | first = *attr; |
210 | else if (memcmp(&first, attr, sizeof(*attr))) | |
211 | return -ENODATA; | |
212 | } | |
213 | ||
214 | return err; | |
215 | } | |
216 | EXPORT_SYMBOL_GPL(switchdev_port_attr_get); | |
217 | ||
218 | static int __switchdev_port_attr_set(struct net_device *dev, | |
f7fadf30 | 219 | const struct switchdev_attr *attr, |
7ea6eb3f | 220 | struct switchdev_trans *trans) |
3094333d SF |
221 | { |
222 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
223 | struct net_device *lower_dev; | |
224 | struct list_head *iter; | |
225 | int err = -EOPNOTSUPP; | |
226 | ||
227 | if (ops && ops->switchdev_port_attr_set) | |
7ea6eb3f | 228 | return ops->switchdev_port_attr_set(dev, attr, trans); |
3094333d SF |
229 | |
230 | if (attr->flags & SWITCHDEV_F_NO_RECURSE) | |
464314ea | 231 | goto done; |
3094333d SF |
232 | |
233 | /* Switch device port(s) may be stacked under | |
234 | * bond/team/vlan dev, so recurse down to set attr on | |
235 | * each port. | |
236 | */ | |
237 | ||
238 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
7ea6eb3f | 239 | err = __switchdev_port_attr_set(lower_dev, attr, trans); |
464314ea SF |
240 | if (err == -EOPNOTSUPP && |
241 | attr->flags & SWITCHDEV_F_SKIP_EOPNOTSUPP) | |
242 | continue; | |
3094333d SF |
243 | if (err) |
244 | break; | |
245 | } | |
246 | ||
464314ea SF |
247 | done: |
248 | if (err == -EOPNOTSUPP && attr->flags & SWITCHDEV_F_SKIP_EOPNOTSUPP) | |
249 | err = 0; | |
250 | ||
3094333d SF |
251 | return err; |
252 | } | |
253 | ||
0bc05d58 JP |
254 | static int switchdev_port_attr_set_now(struct net_device *dev, |
255 | const struct switchdev_attr *attr) | |
3094333d | 256 | { |
7ea6eb3f | 257 | struct switchdev_trans trans; |
3094333d SF |
258 | int err; |
259 | ||
7ea6eb3f JP |
260 | switchdev_trans_init(&trans); |
261 | ||
3094333d SF |
262 | /* Phase I: prepare for attr set. Driver/device should fail |
263 | * here if there are going to be issues in the commit phase, | |
264 | * such as lack of resources or support. The driver/device | |
265 | * should reserve resources needed for the commit phase here, | |
266 | * but should not commit the attr. | |
267 | */ | |
268 | ||
f623ab7f | 269 | trans.ph_prepare = true; |
7ea6eb3f | 270 | err = __switchdev_port_attr_set(dev, attr, &trans); |
3094333d SF |
271 | if (err) { |
272 | /* Prepare phase failed: abort the transaction. Any | |
273 | * resources reserved in the prepare phase are | |
274 | * released. | |
275 | */ | |
276 | ||
9f6467cf | 277 | if (err != -EOPNOTSUPP) |
7ea6eb3f | 278 | switchdev_trans_items_destroy(&trans); |
3094333d SF |
279 | |
280 | return err; | |
281 | } | |
282 | ||
283 | /* Phase II: commit attr set. This cannot fail as a fault | |
284 | * of driver/device. If it does, it's a bug in the driver/device | |
285 | * because the driver said everythings was OK in phase I. | |
286 | */ | |
287 | ||
f623ab7f | 288 | trans.ph_prepare = false; |
7ea6eb3f | 289 | err = __switchdev_port_attr_set(dev, attr, &trans); |
e9fdaec0 SF |
290 | WARN(err, "%s: Commit of attribute (id=%d) failed.\n", |
291 | dev->name, attr->id); | |
7ea6eb3f | 292 | switchdev_trans_items_warn_destroy(dev, &trans); |
3094333d SF |
293 | |
294 | return err; | |
295 | } | |
0bc05d58 JP |
296 | |
297 | static void switchdev_port_attr_set_deferred(struct net_device *dev, | |
298 | const void *data) | |
299 | { | |
300 | const struct switchdev_attr *attr = data; | |
301 | int err; | |
302 | ||
303 | err = switchdev_port_attr_set_now(dev, attr); | |
304 | if (err && err != -EOPNOTSUPP) | |
305 | netdev_err(dev, "failed (err=%d) to set attribute (id=%d)\n", | |
306 | err, attr->id); | |
307 | } | |
308 | ||
309 | static int switchdev_port_attr_set_defer(struct net_device *dev, | |
310 | const struct switchdev_attr *attr) | |
311 | { | |
312 | return switchdev_deferred_enqueue(dev, attr, sizeof(*attr), | |
313 | switchdev_port_attr_set_deferred); | |
314 | } | |
315 | ||
316 | /** | |
317 | * switchdev_port_attr_set - Set port attribute | |
318 | * | |
319 | * @dev: port device | |
320 | * @attr: attribute to set | |
321 | * | |
322 | * Use a 2-phase prepare-commit transaction model to ensure | |
323 | * system is not left in a partially updated state due to | |
324 | * failure from driver/device. | |
325 | * | |
326 | * rtnl_lock must be held and must not be in atomic section, | |
327 | * in case SWITCHDEV_F_DEFER flag is not set. | |
328 | */ | |
329 | int switchdev_port_attr_set(struct net_device *dev, | |
330 | const struct switchdev_attr *attr) | |
331 | { | |
332 | if (attr->flags & SWITCHDEV_F_DEFER) | |
333 | return switchdev_port_attr_set_defer(dev, attr); | |
334 | ASSERT_RTNL(); | |
335 | return switchdev_port_attr_set_now(dev, attr); | |
336 | } | |
3094333d SF |
337 | EXPORT_SYMBOL_GPL(switchdev_port_attr_set); |
338 | ||
22c1f67e | 339 | static int __switchdev_port_obj_add(struct net_device *dev, |
648b4a99 | 340 | const struct switchdev_obj *obj, |
7ea6eb3f | 341 | struct switchdev_trans *trans) |
491d0f15 SF |
342 | { |
343 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
344 | struct net_device *lower_dev; | |
345 | struct list_head *iter; | |
346 | int err = -EOPNOTSUPP; | |
347 | ||
348 | if (ops && ops->switchdev_port_obj_add) | |
9e8f4a54 | 349 | return ops->switchdev_port_obj_add(dev, obj, trans); |
491d0f15 SF |
350 | |
351 | /* Switch device port(s) may be stacked under | |
352 | * bond/team/vlan dev, so recurse down to add object on | |
353 | * each port. | |
354 | */ | |
355 | ||
356 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
9e8f4a54 | 357 | err = __switchdev_port_obj_add(lower_dev, obj, trans); |
491d0f15 SF |
358 | if (err) |
359 | break; | |
360 | } | |
361 | ||
362 | return err; | |
363 | } | |
364 | ||
365 | /** | |
366 | * switchdev_port_obj_add - Add port object | |
367 | * | |
368 | * @dev: port device | |
ab069002 | 369 | * @id: object ID |
491d0f15 SF |
370 | * @obj: object to add |
371 | * | |
372 | * Use a 2-phase prepare-commit transaction model to ensure | |
373 | * system is not left in a partially updated state due to | |
374 | * failure from driver/device. | |
375 | * | |
376 | * rtnl_lock must be held. | |
377 | */ | |
9e8f4a54 | 378 | int switchdev_port_obj_add(struct net_device *dev, |
648b4a99 | 379 | const struct switchdev_obj *obj) |
491d0f15 | 380 | { |
7ea6eb3f | 381 | struct switchdev_trans trans; |
491d0f15 SF |
382 | int err; |
383 | ||
384 | ASSERT_RTNL(); | |
385 | ||
7ea6eb3f JP |
386 | switchdev_trans_init(&trans); |
387 | ||
491d0f15 SF |
388 | /* Phase I: prepare for obj add. Driver/device should fail |
389 | * here if there are going to be issues in the commit phase, | |
390 | * such as lack of resources or support. The driver/device | |
391 | * should reserve resources needed for the commit phase here, | |
392 | * but should not commit the obj. | |
393 | */ | |
394 | ||
f623ab7f | 395 | trans.ph_prepare = true; |
9e8f4a54 | 396 | err = __switchdev_port_obj_add(dev, obj, &trans); |
491d0f15 SF |
397 | if (err) { |
398 | /* Prepare phase failed: abort the transaction. Any | |
399 | * resources reserved in the prepare phase are | |
400 | * released. | |
401 | */ | |
402 | ||
9f6467cf | 403 | if (err != -EOPNOTSUPP) |
7ea6eb3f | 404 | switchdev_trans_items_destroy(&trans); |
491d0f15 SF |
405 | |
406 | return err; | |
407 | } | |
408 | ||
409 | /* Phase II: commit obj add. This cannot fail as a fault | |
410 | * of driver/device. If it does, it's a bug in the driver/device | |
411 | * because the driver said everythings was OK in phase I. | |
412 | */ | |
413 | ||
f623ab7f | 414 | trans.ph_prepare = false; |
9e8f4a54 JP |
415 | err = __switchdev_port_obj_add(dev, obj, &trans); |
416 | WARN(err, "%s: Commit of object (id=%d) failed.\n", dev->name, obj->id); | |
7ea6eb3f | 417 | switchdev_trans_items_warn_destroy(dev, &trans); |
491d0f15 SF |
418 | |
419 | return err; | |
420 | } | |
421 | EXPORT_SYMBOL_GPL(switchdev_port_obj_add); | |
422 | ||
423 | /** | |
424 | * switchdev_port_obj_del - Delete port object | |
425 | * | |
426 | * @dev: port device | |
ab069002 | 427 | * @id: object ID |
491d0f15 SF |
428 | * @obj: object to delete |
429 | */ | |
9e8f4a54 | 430 | int switchdev_port_obj_del(struct net_device *dev, |
648b4a99 | 431 | const struct switchdev_obj *obj) |
491d0f15 SF |
432 | { |
433 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
434 | struct net_device *lower_dev; | |
435 | struct list_head *iter; | |
436 | int err = -EOPNOTSUPP; | |
437 | ||
438 | if (ops && ops->switchdev_port_obj_del) | |
9e8f4a54 | 439 | return ops->switchdev_port_obj_del(dev, obj); |
491d0f15 SF |
440 | |
441 | /* Switch device port(s) may be stacked under | |
442 | * bond/team/vlan dev, so recurse down to delete object on | |
443 | * each port. | |
444 | */ | |
445 | ||
446 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
9e8f4a54 | 447 | err = switchdev_port_obj_del(lower_dev, obj); |
491d0f15 SF |
448 | if (err) |
449 | break; | |
450 | } | |
451 | ||
452 | return err; | |
453 | } | |
454 | EXPORT_SYMBOL_GPL(switchdev_port_obj_del); | |
455 | ||
45d4122c SS |
456 | /** |
457 | * switchdev_port_obj_dump - Dump port objects | |
458 | * | |
459 | * @dev: port device | |
25f07adc | 460 | * @id: object ID |
45d4122c | 461 | * @obj: object to dump |
25f07adc | 462 | * @cb: function to call with a filled object |
45d4122c | 463 | */ |
9e8f4a54 | 464 | int switchdev_port_obj_dump(struct net_device *dev, struct switchdev_obj *obj, |
648b4a99 | 465 | switchdev_obj_dump_cb_t *cb) |
45d4122c SS |
466 | { |
467 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
468 | struct net_device *lower_dev; | |
469 | struct list_head *iter; | |
470 | int err = -EOPNOTSUPP; | |
471 | ||
472 | if (ops && ops->switchdev_port_obj_dump) | |
9e8f4a54 | 473 | return ops->switchdev_port_obj_dump(dev, obj, cb); |
45d4122c SS |
474 | |
475 | /* Switch device port(s) may be stacked under | |
476 | * bond/team/vlan dev, so recurse down to dump objects on | |
477 | * first port at bottom of stack. | |
478 | */ | |
479 | ||
480 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
9e8f4a54 | 481 | err = switchdev_port_obj_dump(lower_dev, obj, cb); |
45d4122c SS |
482 | break; |
483 | } | |
484 | ||
485 | return err; | |
486 | } | |
487 | EXPORT_SYMBOL_GPL(switchdev_port_obj_dump); | |
488 | ||
ebb9a03a JP |
489 | static DEFINE_MUTEX(switchdev_mutex); |
490 | static RAW_NOTIFIER_HEAD(switchdev_notif_chain); | |
03bf0c28 JP |
491 | |
492 | /** | |
ebb9a03a | 493 | * register_switchdev_notifier - Register notifier |
03bf0c28 JP |
494 | * @nb: notifier_block |
495 | * | |
496 | * Register switch device notifier. This should be used by code | |
497 | * which needs to monitor events happening in particular device. | |
498 | * Return values are same as for atomic_notifier_chain_register(). | |
499 | */ | |
ebb9a03a | 500 | int register_switchdev_notifier(struct notifier_block *nb) |
03bf0c28 JP |
501 | { |
502 | int err; | |
503 | ||
ebb9a03a JP |
504 | mutex_lock(&switchdev_mutex); |
505 | err = raw_notifier_chain_register(&switchdev_notif_chain, nb); | |
506 | mutex_unlock(&switchdev_mutex); | |
03bf0c28 JP |
507 | return err; |
508 | } | |
ebb9a03a | 509 | EXPORT_SYMBOL_GPL(register_switchdev_notifier); |
03bf0c28 JP |
510 | |
511 | /** | |
ebb9a03a | 512 | * unregister_switchdev_notifier - Unregister notifier |
03bf0c28 JP |
513 | * @nb: notifier_block |
514 | * | |
515 | * Unregister switch device notifier. | |
516 | * Return values are same as for atomic_notifier_chain_unregister(). | |
517 | */ | |
ebb9a03a | 518 | int unregister_switchdev_notifier(struct notifier_block *nb) |
03bf0c28 JP |
519 | { |
520 | int err; | |
521 | ||
ebb9a03a JP |
522 | mutex_lock(&switchdev_mutex); |
523 | err = raw_notifier_chain_unregister(&switchdev_notif_chain, nb); | |
524 | mutex_unlock(&switchdev_mutex); | |
03bf0c28 JP |
525 | return err; |
526 | } | |
ebb9a03a | 527 | EXPORT_SYMBOL_GPL(unregister_switchdev_notifier); |
03bf0c28 JP |
528 | |
529 | /** | |
ebb9a03a | 530 | * call_switchdev_notifiers - Call notifiers |
03bf0c28 JP |
531 | * @val: value passed unmodified to notifier function |
532 | * @dev: port device | |
533 | * @info: notifier information data | |
534 | * | |
535 | * Call all network notifier blocks. This should be called by driver | |
536 | * when it needs to propagate hardware event. | |
537 | * Return values are same as for atomic_notifier_call_chain(). | |
538 | */ | |
ebb9a03a JP |
539 | int call_switchdev_notifiers(unsigned long val, struct net_device *dev, |
540 | struct switchdev_notifier_info *info) | |
03bf0c28 JP |
541 | { |
542 | int err; | |
543 | ||
544 | info->dev = dev; | |
ebb9a03a JP |
545 | mutex_lock(&switchdev_mutex); |
546 | err = raw_notifier_call_chain(&switchdev_notif_chain, val, info); | |
547 | mutex_unlock(&switchdev_mutex); | |
03bf0c28 JP |
548 | return err; |
549 | } | |
ebb9a03a | 550 | EXPORT_SYMBOL_GPL(call_switchdev_notifiers); |
8a44dbb2 | 551 | |
7d4f8d87 | 552 | struct switchdev_vlan_dump { |
8f24f309 | 553 | struct switchdev_obj_port_vlan vlan; |
7d4f8d87 SF |
554 | struct sk_buff *skb; |
555 | u32 filter_mask; | |
556 | u16 flags; | |
557 | u16 begin; | |
558 | u16 end; | |
559 | }; | |
560 | ||
e23b002b | 561 | static int switchdev_port_vlan_dump_put(struct switchdev_vlan_dump *dump) |
7d4f8d87 SF |
562 | { |
563 | struct bridge_vlan_info vinfo; | |
564 | ||
565 | vinfo.flags = dump->flags; | |
566 | ||
567 | if (dump->begin == 0 && dump->end == 0) { | |
568 | return 0; | |
569 | } else if (dump->begin == dump->end) { | |
570 | vinfo.vid = dump->begin; | |
571 | if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO, | |
572 | sizeof(vinfo), &vinfo)) | |
573 | return -EMSGSIZE; | |
574 | } else { | |
575 | vinfo.vid = dump->begin; | |
576 | vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_BEGIN; | |
577 | if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO, | |
578 | sizeof(vinfo), &vinfo)) | |
579 | return -EMSGSIZE; | |
580 | vinfo.vid = dump->end; | |
581 | vinfo.flags &= ~BRIDGE_VLAN_INFO_RANGE_BEGIN; | |
582 | vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_END; | |
583 | if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO, | |
584 | sizeof(vinfo), &vinfo)) | |
585 | return -EMSGSIZE; | |
586 | } | |
587 | ||
588 | return 0; | |
589 | } | |
590 | ||
648b4a99 | 591 | static int switchdev_port_vlan_dump_cb(struct switchdev_obj *obj) |
7d4f8d87 | 592 | { |
648b4a99 | 593 | struct switchdev_obj_port_vlan *vlan = SWITCHDEV_OBJ_PORT_VLAN(obj); |
7d4f8d87 | 594 | struct switchdev_vlan_dump *dump = |
25f07adc | 595 | container_of(vlan, struct switchdev_vlan_dump, vlan); |
7d4f8d87 SF |
596 | int err = 0; |
597 | ||
598 | if (vlan->vid_begin > vlan->vid_end) | |
599 | return -EINVAL; | |
600 | ||
601 | if (dump->filter_mask & RTEXT_FILTER_BRVLAN) { | |
602 | dump->flags = vlan->flags; | |
603 | for (dump->begin = dump->end = vlan->vid_begin; | |
604 | dump->begin <= vlan->vid_end; | |
605 | dump->begin++, dump->end++) { | |
e23b002b | 606 | err = switchdev_port_vlan_dump_put(dump); |
7d4f8d87 SF |
607 | if (err) |
608 | return err; | |
609 | } | |
610 | } else if (dump->filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED) { | |
611 | if (dump->begin > vlan->vid_begin && | |
612 | dump->begin >= vlan->vid_end) { | |
613 | if ((dump->begin - 1) == vlan->vid_end && | |
614 | dump->flags == vlan->flags) { | |
615 | /* prepend */ | |
616 | dump->begin = vlan->vid_begin; | |
617 | } else { | |
e23b002b | 618 | err = switchdev_port_vlan_dump_put(dump); |
7d4f8d87 SF |
619 | dump->flags = vlan->flags; |
620 | dump->begin = vlan->vid_begin; | |
621 | dump->end = vlan->vid_end; | |
622 | } | |
623 | } else if (dump->end <= vlan->vid_begin && | |
624 | dump->end < vlan->vid_end) { | |
625 | if ((dump->end + 1) == vlan->vid_begin && | |
626 | dump->flags == vlan->flags) { | |
627 | /* append */ | |
628 | dump->end = vlan->vid_end; | |
629 | } else { | |
e23b002b | 630 | err = switchdev_port_vlan_dump_put(dump); |
7d4f8d87 SF |
631 | dump->flags = vlan->flags; |
632 | dump->begin = vlan->vid_begin; | |
633 | dump->end = vlan->vid_end; | |
634 | } | |
635 | } else { | |
636 | err = -EINVAL; | |
637 | } | |
638 | } | |
639 | ||
640 | return err; | |
641 | } | |
642 | ||
643 | static int switchdev_port_vlan_fill(struct sk_buff *skb, struct net_device *dev, | |
644 | u32 filter_mask) | |
645 | { | |
646 | struct switchdev_vlan_dump dump = { | |
9e8f4a54 | 647 | .vlan.obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN, |
7d4f8d87 SF |
648 | .skb = skb, |
649 | .filter_mask = filter_mask, | |
650 | }; | |
651 | int err = 0; | |
652 | ||
653 | if ((filter_mask & RTEXT_FILTER_BRVLAN) || | |
654 | (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)) { | |
9e8f4a54 | 655 | err = switchdev_port_obj_dump(dev, &dump.vlan.obj, |
25f07adc | 656 | switchdev_port_vlan_dump_cb); |
7d4f8d87 SF |
657 | if (err) |
658 | goto err_out; | |
659 | if (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED) | |
660 | /* last one */ | |
e23b002b | 661 | err = switchdev_port_vlan_dump_put(&dump); |
7d4f8d87 SF |
662 | } |
663 | ||
664 | err_out: | |
665 | return err == -EOPNOTSUPP ? 0 : err; | |
666 | } | |
667 | ||
8793d0a6 SF |
668 | /** |
669 | * switchdev_port_bridge_getlink - Get bridge port attributes | |
670 | * | |
671 | * @dev: port device | |
672 | * | |
673 | * Called for SELF on rtnl_bridge_getlink to get bridge port | |
674 | * attributes. | |
675 | */ | |
676 | int switchdev_port_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, | |
677 | struct net_device *dev, u32 filter_mask, | |
678 | int nlflags) | |
679 | { | |
680 | struct switchdev_attr attr = { | |
1f868398 | 681 | .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS, |
8793d0a6 SF |
682 | }; |
683 | u16 mode = BRIDGE_MODE_UNDEF; | |
684 | u32 mask = BR_LEARNING | BR_LEARNING_SYNC; | |
685 | int err; | |
686 | ||
687 | err = switchdev_port_attr_get(dev, &attr); | |
5c8079d0 | 688 | if (err && err != -EOPNOTSUPP) |
8793d0a6 SF |
689 | return err; |
690 | ||
691 | return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode, | |
7d4f8d87 SF |
692 | attr.u.brport_flags, mask, nlflags, |
693 | filter_mask, switchdev_port_vlan_fill); | |
8793d0a6 SF |
694 | } |
695 | EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink); | |
696 | ||
47f8328b SF |
697 | static int switchdev_port_br_setflag(struct net_device *dev, |
698 | struct nlattr *nlattr, | |
699 | unsigned long brport_flag) | |
700 | { | |
701 | struct switchdev_attr attr = { | |
1f868398 | 702 | .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS, |
47f8328b SF |
703 | }; |
704 | u8 flag = nla_get_u8(nlattr); | |
705 | int err; | |
706 | ||
707 | err = switchdev_port_attr_get(dev, &attr); | |
708 | if (err) | |
709 | return err; | |
710 | ||
711 | if (flag) | |
42275bd8 | 712 | attr.u.brport_flags |= brport_flag; |
47f8328b | 713 | else |
42275bd8 | 714 | attr.u.brport_flags &= ~brport_flag; |
47f8328b SF |
715 | |
716 | return switchdev_port_attr_set(dev, &attr); | |
717 | } | |
718 | ||
719 | static const struct nla_policy | |
720 | switchdev_port_bridge_policy[IFLA_BRPORT_MAX + 1] = { | |
721 | [IFLA_BRPORT_STATE] = { .type = NLA_U8 }, | |
722 | [IFLA_BRPORT_COST] = { .type = NLA_U32 }, | |
723 | [IFLA_BRPORT_PRIORITY] = { .type = NLA_U16 }, | |
724 | [IFLA_BRPORT_MODE] = { .type = NLA_U8 }, | |
725 | [IFLA_BRPORT_GUARD] = { .type = NLA_U8 }, | |
726 | [IFLA_BRPORT_PROTECT] = { .type = NLA_U8 }, | |
727 | [IFLA_BRPORT_FAST_LEAVE] = { .type = NLA_U8 }, | |
728 | [IFLA_BRPORT_LEARNING] = { .type = NLA_U8 }, | |
729 | [IFLA_BRPORT_LEARNING_SYNC] = { .type = NLA_U8 }, | |
730 | [IFLA_BRPORT_UNICAST_FLOOD] = { .type = NLA_U8 }, | |
731 | }; | |
732 | ||
733 | static int switchdev_port_br_setlink_protinfo(struct net_device *dev, | |
734 | struct nlattr *protinfo) | |
735 | { | |
736 | struct nlattr *attr; | |
737 | int rem; | |
738 | int err; | |
739 | ||
740 | err = nla_validate_nested(protinfo, IFLA_BRPORT_MAX, | |
741 | switchdev_port_bridge_policy); | |
742 | if (err) | |
743 | return err; | |
744 | ||
745 | nla_for_each_nested(attr, protinfo, rem) { | |
746 | switch (nla_type(attr)) { | |
747 | case IFLA_BRPORT_LEARNING: | |
748 | err = switchdev_port_br_setflag(dev, attr, | |
749 | BR_LEARNING); | |
750 | break; | |
751 | case IFLA_BRPORT_LEARNING_SYNC: | |
752 | err = switchdev_port_br_setflag(dev, attr, | |
753 | BR_LEARNING_SYNC); | |
754 | break; | |
755 | default: | |
756 | err = -EOPNOTSUPP; | |
757 | break; | |
758 | } | |
759 | if (err) | |
760 | return err; | |
761 | } | |
762 | ||
763 | return 0; | |
764 | } | |
765 | ||
766 | static int switchdev_port_br_afspec(struct net_device *dev, | |
767 | struct nlattr *afspec, | |
768 | int (*f)(struct net_device *dev, | |
648b4a99 | 769 | const struct switchdev_obj *obj)) |
47f8328b SF |
770 | { |
771 | struct nlattr *attr; | |
772 | struct bridge_vlan_info *vinfo; | |
9e8f4a54 JP |
773 | struct switchdev_obj_port_vlan vlan = { |
774 | .obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN, | |
775 | }; | |
47f8328b SF |
776 | int rem; |
777 | int err; | |
778 | ||
779 | nla_for_each_nested(attr, afspec, rem) { | |
780 | if (nla_type(attr) != IFLA_BRIDGE_VLAN_INFO) | |
781 | continue; | |
782 | if (nla_len(attr) != sizeof(struct bridge_vlan_info)) | |
783 | return -EINVAL; | |
784 | vinfo = nla_data(attr); | |
ab069002 | 785 | vlan.flags = vinfo->flags; |
47f8328b | 786 | if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_BEGIN) { |
ab069002 | 787 | if (vlan.vid_begin) |
47f8328b | 788 | return -EINVAL; |
ab069002 | 789 | vlan.vid_begin = vinfo->vid; |
cc02aa8e NA |
790 | /* don't allow range of pvids */ |
791 | if (vlan.flags & BRIDGE_VLAN_INFO_PVID) | |
792 | return -EINVAL; | |
47f8328b | 793 | } else if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_END) { |
ab069002 | 794 | if (!vlan.vid_begin) |
47f8328b | 795 | return -EINVAL; |
ab069002 VD |
796 | vlan.vid_end = vinfo->vid; |
797 | if (vlan.vid_end <= vlan.vid_begin) | |
47f8328b | 798 | return -EINVAL; |
9e8f4a54 | 799 | err = f(dev, &vlan.obj); |
47f8328b SF |
800 | if (err) |
801 | return err; | |
ab069002 | 802 | memset(&vlan, 0, sizeof(vlan)); |
47f8328b | 803 | } else { |
ab069002 | 804 | if (vlan.vid_begin) |
47f8328b | 805 | return -EINVAL; |
ab069002 VD |
806 | vlan.vid_begin = vinfo->vid; |
807 | vlan.vid_end = vinfo->vid; | |
9e8f4a54 | 808 | err = f(dev, &vlan.obj); |
47f8328b SF |
809 | if (err) |
810 | return err; | |
ab069002 | 811 | memset(&vlan, 0, sizeof(vlan)); |
47f8328b SF |
812 | } |
813 | } | |
814 | ||
815 | return 0; | |
816 | } | |
817 | ||
8a44dbb2 | 818 | /** |
47f8328b | 819 | * switchdev_port_bridge_setlink - Set bridge port attributes |
8a44dbb2 RP |
820 | * |
821 | * @dev: port device | |
47f8328b SF |
822 | * @nlh: netlink header |
823 | * @flags: netlink flags | |
8a44dbb2 | 824 | * |
47f8328b SF |
825 | * Called for SELF on rtnl_bridge_setlink to set bridge port |
826 | * attributes. | |
8a44dbb2 | 827 | */ |
ebb9a03a JP |
828 | int switchdev_port_bridge_setlink(struct net_device *dev, |
829 | struct nlmsghdr *nlh, u16 flags) | |
8a44dbb2 | 830 | { |
47f8328b SF |
831 | struct nlattr *protinfo; |
832 | struct nlattr *afspec; | |
833 | int err = 0; | |
834 | ||
835 | protinfo = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), | |
836 | IFLA_PROTINFO); | |
837 | if (protinfo) { | |
838 | err = switchdev_port_br_setlink_protinfo(dev, protinfo); | |
839 | if (err) | |
840 | return err; | |
841 | } | |
8a44dbb2 | 842 | |
47f8328b SF |
843 | afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), |
844 | IFLA_AF_SPEC); | |
845 | if (afspec) | |
846 | err = switchdev_port_br_afspec(dev, afspec, | |
847 | switchdev_port_obj_add); | |
8a44dbb2 | 848 | |
47f8328b | 849 | return err; |
8a44dbb2 | 850 | } |
ebb9a03a | 851 | EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink); |
8a44dbb2 RP |
852 | |
853 | /** | |
5c34e022 | 854 | * switchdev_port_bridge_dellink - Set bridge port attributes |
8a44dbb2 RP |
855 | * |
856 | * @dev: port device | |
5c34e022 SF |
857 | * @nlh: netlink header |
858 | * @flags: netlink flags | |
8a44dbb2 | 859 | * |
5c34e022 SF |
860 | * Called for SELF on rtnl_bridge_dellink to set bridge port |
861 | * attributes. | |
8a44dbb2 | 862 | */ |
ebb9a03a JP |
863 | int switchdev_port_bridge_dellink(struct net_device *dev, |
864 | struct nlmsghdr *nlh, u16 flags) | |
8a44dbb2 | 865 | { |
5c34e022 | 866 | struct nlattr *afspec; |
8a44dbb2 | 867 | |
5c34e022 SF |
868 | afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), |
869 | IFLA_AF_SPEC); | |
870 | if (afspec) | |
871 | return switchdev_port_br_afspec(dev, afspec, | |
872 | switchdev_port_obj_del); | |
8a44dbb2 | 873 | |
5c34e022 | 874 | return 0; |
8a44dbb2 | 875 | } |
ebb9a03a | 876 | EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink); |
8a44dbb2 | 877 | |
45d4122c SS |
878 | /** |
879 | * switchdev_port_fdb_add - Add FDB (MAC/VLAN) entry to port | |
880 | * | |
881 | * @ndmsg: netlink hdr | |
882 | * @nlattr: netlink attributes | |
883 | * @dev: port device | |
884 | * @addr: MAC address to add | |
885 | * @vid: VLAN to add | |
886 | * | |
887 | * Add FDB entry to switch device. | |
888 | */ | |
889 | int switchdev_port_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], | |
890 | struct net_device *dev, const unsigned char *addr, | |
891 | u16 vid, u16 nlm_flags) | |
892 | { | |
52ba57cf | 893 | struct switchdev_obj_port_fdb fdb = { |
9e8f4a54 | 894 | .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB, |
ab069002 | 895 | .vid = vid, |
45d4122c SS |
896 | }; |
897 | ||
850d0cbc | 898 | ether_addr_copy(fdb.addr, addr); |
9e8f4a54 | 899 | return switchdev_port_obj_add(dev, &fdb.obj); |
45d4122c SS |
900 | } |
901 | EXPORT_SYMBOL_GPL(switchdev_port_fdb_add); | |
902 | ||
903 | /** | |
904 | * switchdev_port_fdb_del - Delete FDB (MAC/VLAN) entry from port | |
905 | * | |
906 | * @ndmsg: netlink hdr | |
907 | * @nlattr: netlink attributes | |
908 | * @dev: port device | |
909 | * @addr: MAC address to delete | |
910 | * @vid: VLAN to delete | |
911 | * | |
912 | * Delete FDB entry from switch device. | |
913 | */ | |
914 | int switchdev_port_fdb_del(struct ndmsg *ndm, struct nlattr *tb[], | |
915 | struct net_device *dev, const unsigned char *addr, | |
916 | u16 vid) | |
917 | { | |
52ba57cf | 918 | struct switchdev_obj_port_fdb fdb = { |
9e8f4a54 | 919 | .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB, |
ab069002 | 920 | .vid = vid, |
45d4122c SS |
921 | }; |
922 | ||
850d0cbc | 923 | ether_addr_copy(fdb.addr, addr); |
9e8f4a54 | 924 | return switchdev_port_obj_del(dev, &fdb.obj); |
45d4122c SS |
925 | } |
926 | EXPORT_SYMBOL_GPL(switchdev_port_fdb_del); | |
927 | ||
928 | struct switchdev_fdb_dump { | |
52ba57cf | 929 | struct switchdev_obj_port_fdb fdb; |
e02a06b2 | 930 | struct net_device *dev; |
45d4122c SS |
931 | struct sk_buff *skb; |
932 | struct netlink_callback *cb; | |
45d4122c SS |
933 | int idx; |
934 | }; | |
935 | ||
648b4a99 | 936 | static int switchdev_port_fdb_dump_cb(struct switchdev_obj *obj) |
45d4122c | 937 | { |
648b4a99 | 938 | struct switchdev_obj_port_fdb *fdb = SWITCHDEV_OBJ_PORT_FDB(obj); |
45d4122c | 939 | struct switchdev_fdb_dump *dump = |
25f07adc | 940 | container_of(fdb, struct switchdev_fdb_dump, fdb); |
45d4122c SS |
941 | u32 portid = NETLINK_CB(dump->cb->skb).portid; |
942 | u32 seq = dump->cb->nlh->nlmsg_seq; | |
943 | struct nlmsghdr *nlh; | |
944 | struct ndmsg *ndm; | |
45d4122c SS |
945 | |
946 | if (dump->idx < dump->cb->args[0]) | |
947 | goto skip; | |
948 | ||
45d4122c SS |
949 | nlh = nlmsg_put(dump->skb, portid, seq, RTM_NEWNEIGH, |
950 | sizeof(*ndm), NLM_F_MULTI); | |
951 | if (!nlh) | |
952 | return -EMSGSIZE; | |
953 | ||
954 | ndm = nlmsg_data(nlh); | |
955 | ndm->ndm_family = AF_BRIDGE; | |
956 | ndm->ndm_pad1 = 0; | |
957 | ndm->ndm_pad2 = 0; | |
958 | ndm->ndm_flags = NTF_SELF; | |
959 | ndm->ndm_type = 0; | |
e02a06b2 | 960 | ndm->ndm_ifindex = dump->dev->ifindex; |
25f07adc | 961 | ndm->ndm_state = fdb->ndm_state; |
45d4122c | 962 | |
25f07adc | 963 | if (nla_put(dump->skb, NDA_LLADDR, ETH_ALEN, fdb->addr)) |
45d4122c SS |
964 | goto nla_put_failure; |
965 | ||
25f07adc | 966 | if (fdb->vid && nla_put_u16(dump->skb, NDA_VLAN, fdb->vid)) |
45d4122c SS |
967 | goto nla_put_failure; |
968 | ||
969 | nlmsg_end(dump->skb, nlh); | |
970 | ||
971 | skip: | |
972 | dump->idx++; | |
973 | return 0; | |
974 | ||
975 | nla_put_failure: | |
976 | nlmsg_cancel(dump->skb, nlh); | |
977 | return -EMSGSIZE; | |
978 | } | |
979 | ||
980 | /** | |
981 | * switchdev_port_fdb_dump - Dump port FDB (MAC/VLAN) entries | |
982 | * | |
983 | * @skb: netlink skb | |
984 | * @cb: netlink callback | |
985 | * @dev: port device | |
986 | * @filter_dev: filter device | |
987 | * @idx: | |
988 | * | |
989 | * Delete FDB entry from switch device. | |
990 | */ | |
991 | int switchdev_port_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, | |
992 | struct net_device *dev, | |
993 | struct net_device *filter_dev, int idx) | |
994 | { | |
995 | struct switchdev_fdb_dump dump = { | |
9e8f4a54 | 996 | .fdb.obj.id = SWITCHDEV_OBJ_ID_PORT_FDB, |
e02a06b2 | 997 | .dev = dev, |
45d4122c SS |
998 | .skb = skb, |
999 | .cb = cb, | |
45d4122c SS |
1000 | .idx = idx, |
1001 | }; | |
45d4122c | 1002 | |
9e8f4a54 | 1003 | switchdev_port_obj_dump(dev, &dump.fdb.obj, switchdev_port_fdb_dump_cb); |
45d4122c SS |
1004 | return dump.idx; |
1005 | } | |
1006 | EXPORT_SYMBOL_GPL(switchdev_port_fdb_dump); | |
1007 | ||
ebb9a03a | 1008 | static struct net_device *switchdev_get_lowest_dev(struct net_device *dev) |
b5d6fbde | 1009 | { |
9d47c0a2 | 1010 | const struct switchdev_ops *ops = dev->switchdev_ops; |
b5d6fbde SF |
1011 | struct net_device *lower_dev; |
1012 | struct net_device *port_dev; | |
1013 | struct list_head *iter; | |
1014 | ||
1015 | /* Recusively search down until we find a sw port dev. | |
f8e20a9f | 1016 | * (A sw port dev supports switchdev_port_attr_get). |
b5d6fbde SF |
1017 | */ |
1018 | ||
f8e20a9f | 1019 | if (ops && ops->switchdev_port_attr_get) |
b5d6fbde SF |
1020 | return dev; |
1021 | ||
1022 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
ebb9a03a | 1023 | port_dev = switchdev_get_lowest_dev(lower_dev); |
b5d6fbde SF |
1024 | if (port_dev) |
1025 | return port_dev; | |
1026 | } | |
1027 | ||
1028 | return NULL; | |
1029 | } | |
1030 | ||
ebb9a03a | 1031 | static struct net_device *switchdev_get_dev_by_nhs(struct fib_info *fi) |
b5d6fbde | 1032 | { |
f8e20a9f | 1033 | struct switchdev_attr attr = { |
1f868398 | 1034 | .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID, |
f8e20a9f SF |
1035 | }; |
1036 | struct switchdev_attr prev_attr; | |
b5d6fbde SF |
1037 | struct net_device *dev = NULL; |
1038 | int nhsel; | |
1039 | ||
1040 | /* For this route, all nexthop devs must be on the same switch. */ | |
1041 | ||
1042 | for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) { | |
1043 | const struct fib_nh *nh = &fi->fib_nh[nhsel]; | |
1044 | ||
1045 | if (!nh->nh_dev) | |
1046 | return NULL; | |
1047 | ||
ebb9a03a | 1048 | dev = switchdev_get_lowest_dev(nh->nh_dev); |
b5d6fbde SF |
1049 | if (!dev) |
1050 | return NULL; | |
1051 | ||
f8e20a9f | 1052 | if (switchdev_port_attr_get(dev, &attr)) |
b5d6fbde SF |
1053 | return NULL; |
1054 | ||
d754f98b SF |
1055 | if (nhsel > 0 && |
1056 | !netdev_phys_item_id_same(&prev_attr.u.ppid, &attr.u.ppid)) | |
b5d6fbde | 1057 | return NULL; |
b5d6fbde | 1058 | |
f8e20a9f | 1059 | prev_attr = attr; |
b5d6fbde SF |
1060 | } |
1061 | ||
1062 | return dev; | |
1063 | } | |
1064 | ||
5e8d9049 | 1065 | /** |
7616dcbb | 1066 | * switchdev_fib_ipv4_add - Add/modify switch IPv4 route entry |
5e8d9049 SF |
1067 | * |
1068 | * @dst: route's IPv4 destination address | |
1069 | * @dst_len: destination address length (prefix length) | |
1070 | * @fi: route FIB info structure | |
1071 | * @tos: route TOS | |
1072 | * @type: route type | |
f8f21471 | 1073 | * @nlflags: netlink flags passed in (NLM_F_*) |
5e8d9049 SF |
1074 | * @tb_id: route table ID |
1075 | * | |
7616dcbb | 1076 | * Add/modify switch IPv4 route entry. |
5e8d9049 | 1077 | */ |
ebb9a03a JP |
1078 | int switchdev_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi, |
1079 | u8 tos, u8 type, u32 nlflags, u32 tb_id) | |
5e8d9049 | 1080 | { |
ab069002 | 1081 | struct switchdev_obj_ipv4_fib ipv4_fib = { |
9e8f4a54 | 1082 | .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB, |
ab069002 VD |
1083 | .dst = dst, |
1084 | .dst_len = dst_len, | |
ab069002 VD |
1085 | .tos = tos, |
1086 | .type = type, | |
1087 | .nlflags = nlflags, | |
1088 | .tb_id = tb_id, | |
58c2cb16 | 1089 | }; |
b5d6fbde | 1090 | struct net_device *dev; |
b5d6fbde SF |
1091 | int err = 0; |
1092 | ||
850d0cbc JP |
1093 | memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi)); |
1094 | ||
8e05fd71 SF |
1095 | /* Don't offload route if using custom ip rules or if |
1096 | * IPv4 FIB offloading has been disabled completely. | |
1097 | */ | |
1098 | ||
e1315db1 SF |
1099 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
1100 | if (fi->fib_net->ipv4.fib_has_custom_rules) | |
1101 | return 0; | |
1102 | #endif | |
1103 | ||
1104 | if (fi->fib_net->ipv4.fib_offload_disabled) | |
104616e7 SF |
1105 | return 0; |
1106 | ||
ebb9a03a | 1107 | dev = switchdev_get_dev_by_nhs(fi); |
b5d6fbde SF |
1108 | if (!dev) |
1109 | return 0; | |
58c2cb16 | 1110 | |
9e8f4a54 | 1111 | err = switchdev_port_obj_add(dev, &ipv4_fib.obj); |
58c2cb16 | 1112 | if (!err) |
36583eb5 | 1113 | fi->fib_flags |= RTNH_F_OFFLOAD; |
b5d6fbde | 1114 | |
af201f72 | 1115 | return err == -EOPNOTSUPP ? 0 : err; |
5e8d9049 | 1116 | } |
ebb9a03a | 1117 | EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add); |
5e8d9049 SF |
1118 | |
1119 | /** | |
ebb9a03a | 1120 | * switchdev_fib_ipv4_del - Delete IPv4 route entry from switch |
5e8d9049 SF |
1121 | * |
1122 | * @dst: route's IPv4 destination address | |
1123 | * @dst_len: destination address length (prefix length) | |
1124 | * @fi: route FIB info structure | |
1125 | * @tos: route TOS | |
1126 | * @type: route type | |
1127 | * @tb_id: route table ID | |
1128 | * | |
1129 | * Delete IPv4 route entry from switch device. | |
1130 | */ | |
ebb9a03a JP |
1131 | int switchdev_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi, |
1132 | u8 tos, u8 type, u32 tb_id) | |
5e8d9049 | 1133 | { |
ab069002 | 1134 | struct switchdev_obj_ipv4_fib ipv4_fib = { |
9e8f4a54 | 1135 | .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB, |
ab069002 VD |
1136 | .dst = dst, |
1137 | .dst_len = dst_len, | |
ab069002 VD |
1138 | .tos = tos, |
1139 | .type = type, | |
1140 | .nlflags = 0, | |
1141 | .tb_id = tb_id, | |
58c2cb16 | 1142 | }; |
b5d6fbde | 1143 | struct net_device *dev; |
b5d6fbde SF |
1144 | int err = 0; |
1145 | ||
850d0cbc JP |
1146 | memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi)); |
1147 | ||
eea39946 | 1148 | if (!(fi->fib_flags & RTNH_F_OFFLOAD)) |
b5d6fbde SF |
1149 | return 0; |
1150 | ||
ebb9a03a | 1151 | dev = switchdev_get_dev_by_nhs(fi); |
b5d6fbde SF |
1152 | if (!dev) |
1153 | return 0; | |
b5d6fbde | 1154 | |
9e8f4a54 | 1155 | err = switchdev_port_obj_del(dev, &ipv4_fib.obj); |
58c2cb16 | 1156 | if (!err) |
36583eb5 | 1157 | fi->fib_flags &= ~RTNH_F_OFFLOAD; |
b5d6fbde | 1158 | |
af201f72 | 1159 | return err == -EOPNOTSUPP ? 0 : err; |
5e8d9049 | 1160 | } |
ebb9a03a | 1161 | EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del); |
8e05fd71 SF |
1162 | |
1163 | /** | |
ebb9a03a | 1164 | * switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation |
8e05fd71 SF |
1165 | * |
1166 | * @fi: route FIB info structure | |
1167 | */ | |
ebb9a03a | 1168 | void switchdev_fib_ipv4_abort(struct fib_info *fi) |
8e05fd71 SF |
1169 | { |
1170 | /* There was a problem installing this route to the offload | |
1171 | * device. For now, until we come up with more refined | |
1172 | * policy handling, abruptly end IPv4 fib offloading for | |
1173 | * for entire net by flushing offload device(s) of all | |
1174 | * IPv4 routes, and mark IPv4 fib offloading broken from | |
1175 | * this point forward. | |
1176 | */ | |
1177 | ||
1178 | fib_flush_external(fi->fib_net); | |
1179 | fi->fib_net->ipv4.fib_offload_disabled = true; | |
1180 | } | |
ebb9a03a | 1181 | EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort); |
1a3b2ec9 SF |
1182 | |
1183 | static bool switchdev_port_same_parent_id(struct net_device *a, | |
1184 | struct net_device *b) | |
1185 | { | |
1186 | struct switchdev_attr a_attr = { | |
1f868398 | 1187 | .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID, |
1a3b2ec9 SF |
1188 | .flags = SWITCHDEV_F_NO_RECURSE, |
1189 | }; | |
1190 | struct switchdev_attr b_attr = { | |
1f868398 | 1191 | .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID, |
1a3b2ec9 SF |
1192 | .flags = SWITCHDEV_F_NO_RECURSE, |
1193 | }; | |
1194 | ||
1195 | if (switchdev_port_attr_get(a, &a_attr) || | |
1196 | switchdev_port_attr_get(b, &b_attr)) | |
1197 | return false; | |
1198 | ||
1199 | return netdev_phys_item_id_same(&a_attr.u.ppid, &b_attr.u.ppid); | |
1200 | } | |
1201 | ||
1202 | static u32 switchdev_port_fwd_mark_get(struct net_device *dev, | |
1203 | struct net_device *group_dev) | |
1204 | { | |
1205 | struct net_device *lower_dev; | |
1206 | struct list_head *iter; | |
1207 | ||
1208 | netdev_for_each_lower_dev(group_dev, lower_dev, iter) { | |
1209 | if (lower_dev == dev) | |
1210 | continue; | |
1211 | if (switchdev_port_same_parent_id(dev, lower_dev)) | |
1212 | return lower_dev->offload_fwd_mark; | |
1213 | return switchdev_port_fwd_mark_get(dev, lower_dev); | |
1214 | } | |
1215 | ||
1216 | return dev->ifindex; | |
1217 | } | |
1218 | ||
1219 | static void switchdev_port_fwd_mark_reset(struct net_device *group_dev, | |
1220 | u32 old_mark, u32 *reset_mark) | |
1221 | { | |
1222 | struct net_device *lower_dev; | |
1223 | struct list_head *iter; | |
1224 | ||
1225 | netdev_for_each_lower_dev(group_dev, lower_dev, iter) { | |
1226 | if (lower_dev->offload_fwd_mark == old_mark) { | |
1227 | if (!*reset_mark) | |
1228 | *reset_mark = lower_dev->ifindex; | |
1229 | lower_dev->offload_fwd_mark = *reset_mark; | |
1230 | } | |
1231 | switchdev_port_fwd_mark_reset(lower_dev, old_mark, reset_mark); | |
1232 | } | |
1233 | } | |
1234 | ||
1235 | /** | |
1236 | * switchdev_port_fwd_mark_set - Set port offload forwarding mark | |
1237 | * | |
1238 | * @dev: port device | |
1239 | * @group_dev: containing device | |
1240 | * @joining: true if dev is joining group; false if leaving group | |
1241 | * | |
1242 | * An ungrouped port's offload mark is just its ifindex. A grouped | |
1243 | * port's (member of a bridge, for example) offload mark is the ifindex | |
1244 | * of one of the ports in the group with the same parent (switch) ID. | |
1245 | * Ports on the same device in the same group will have the same mark. | |
1246 | * | |
1247 | * Example: | |
1248 | * | |
1249 | * br0 ifindex=9 | |
1250 | * sw1p1 ifindex=2 mark=2 | |
1251 | * sw1p2 ifindex=3 mark=2 | |
1252 | * sw2p1 ifindex=4 mark=5 | |
1253 | * sw2p2 ifindex=5 mark=5 | |
1254 | * | |
1255 | * If sw2p2 leaves the bridge, we'll have: | |
1256 | * | |
1257 | * br0 ifindex=9 | |
1258 | * sw1p1 ifindex=2 mark=2 | |
1259 | * sw1p2 ifindex=3 mark=2 | |
1260 | * sw2p1 ifindex=4 mark=4 | |
1261 | * sw2p2 ifindex=5 mark=5 | |
1262 | */ | |
1263 | void switchdev_port_fwd_mark_set(struct net_device *dev, | |
1264 | struct net_device *group_dev, | |
1265 | bool joining) | |
1266 | { | |
1267 | u32 mark = dev->ifindex; | |
1268 | u32 reset_mark = 0; | |
1269 | ||
1270 | if (group_dev && joining) { | |
1271 | mark = switchdev_port_fwd_mark_get(dev, group_dev); | |
1272 | } else if (group_dev && !joining) { | |
1273 | if (dev->offload_fwd_mark == mark) | |
1274 | /* Ohoh, this port was the mark reference port, | |
1275 | * but it's leaving the group, so reset the | |
1276 | * mark for the remaining ports in the group. | |
1277 | */ | |
1278 | switchdev_port_fwd_mark_reset(group_dev, mark, | |
1279 | &reset_mark); | |
1280 | } | |
1281 | ||
1282 | dev->offload_fwd_mark = mark; | |
1283 | } | |
1284 | EXPORT_SYMBOL_GPL(switchdev_port_fwd_mark_set); |