]>
Commit | Line | Data |
---|---|---|
007f790c JP |
1 | /* |
2 | * net/switchdev/switchdev.c - Switch device API | |
3 | * Copyright (c) 2014 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> |
47f8328b | 18 | #include <linux/if_bridge.h> |
5e8d9049 | 19 | #include <net/ip_fib.h> |
007f790c JP |
20 | #include <net/switchdev.h> |
21 | ||
3094333d SF |
22 | /** |
23 | * switchdev_port_attr_get - Get port attribute | |
24 | * | |
25 | * @dev: port device | |
26 | * @attr: attribute to get | |
27 | */ | |
28 | int switchdev_port_attr_get(struct net_device *dev, struct switchdev_attr *attr) | |
29 | { | |
30 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
31 | struct net_device *lower_dev; | |
32 | struct list_head *iter; | |
33 | struct switchdev_attr first = { | |
34 | .id = SWITCHDEV_ATTR_UNDEFINED | |
35 | }; | |
36 | int err = -EOPNOTSUPP; | |
37 | ||
38 | if (ops && ops->switchdev_port_attr_get) | |
39 | return ops->switchdev_port_attr_get(dev, attr); | |
40 | ||
41 | if (attr->flags & SWITCHDEV_F_NO_RECURSE) | |
42 | return err; | |
43 | ||
44 | /* Switch device port(s) may be stacked under | |
45 | * bond/team/vlan dev, so recurse down to get attr on | |
46 | * each port. Return -ENODATA if attr values don't | |
47 | * compare across ports. | |
48 | */ | |
49 | ||
50 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
51 | err = switchdev_port_attr_get(lower_dev, attr); | |
52 | if (err) | |
53 | break; | |
54 | if (first.id == SWITCHDEV_ATTR_UNDEFINED) | |
55 | first = *attr; | |
56 | else if (memcmp(&first, attr, sizeof(*attr))) | |
57 | return -ENODATA; | |
58 | } | |
59 | ||
60 | return err; | |
61 | } | |
62 | EXPORT_SYMBOL_GPL(switchdev_port_attr_get); | |
63 | ||
64 | static int __switchdev_port_attr_set(struct net_device *dev, | |
65 | struct switchdev_attr *attr) | |
66 | { | |
67 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
68 | struct net_device *lower_dev; | |
69 | struct list_head *iter; | |
70 | int err = -EOPNOTSUPP; | |
71 | ||
72 | if (ops && ops->switchdev_port_attr_set) | |
73 | return ops->switchdev_port_attr_set(dev, attr); | |
74 | ||
75 | if (attr->flags & SWITCHDEV_F_NO_RECURSE) | |
76 | return err; | |
77 | ||
78 | /* Switch device port(s) may be stacked under | |
79 | * bond/team/vlan dev, so recurse down to set attr on | |
80 | * each port. | |
81 | */ | |
82 | ||
83 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
84 | err = __switchdev_port_attr_set(lower_dev, attr); | |
85 | if (err) | |
86 | break; | |
87 | } | |
88 | ||
89 | return err; | |
90 | } | |
91 | ||
92 | struct switchdev_attr_set_work { | |
93 | struct work_struct work; | |
94 | struct net_device *dev; | |
95 | struct switchdev_attr attr; | |
96 | }; | |
97 | ||
98 | static void switchdev_port_attr_set_work(struct work_struct *work) | |
99 | { | |
100 | struct switchdev_attr_set_work *asw = | |
101 | container_of(work, struct switchdev_attr_set_work, work); | |
102 | int err; | |
103 | ||
104 | rtnl_lock(); | |
105 | err = switchdev_port_attr_set(asw->dev, &asw->attr); | |
106 | BUG_ON(err); | |
107 | rtnl_unlock(); | |
108 | ||
109 | dev_put(asw->dev); | |
110 | kfree(work); | |
111 | } | |
112 | ||
113 | static int switchdev_port_attr_set_defer(struct net_device *dev, | |
114 | struct switchdev_attr *attr) | |
115 | { | |
116 | struct switchdev_attr_set_work *asw; | |
117 | ||
118 | asw = kmalloc(sizeof(*asw), GFP_ATOMIC); | |
119 | if (!asw) | |
120 | return -ENOMEM; | |
121 | ||
122 | INIT_WORK(&asw->work, switchdev_port_attr_set_work); | |
123 | ||
124 | dev_hold(dev); | |
125 | asw->dev = dev; | |
126 | memcpy(&asw->attr, attr, sizeof(asw->attr)); | |
127 | ||
128 | schedule_work(&asw->work); | |
129 | ||
130 | return 0; | |
131 | } | |
132 | ||
133 | /** | |
134 | * switchdev_port_attr_set - Set port attribute | |
135 | * | |
136 | * @dev: port device | |
137 | * @attr: attribute to set | |
138 | * | |
139 | * Use a 2-phase prepare-commit transaction model to ensure | |
140 | * system is not left in a partially updated state due to | |
141 | * failure from driver/device. | |
142 | */ | |
143 | int switchdev_port_attr_set(struct net_device *dev, struct switchdev_attr *attr) | |
144 | { | |
145 | int err; | |
146 | ||
147 | if (!rtnl_is_locked()) { | |
148 | /* Running prepare-commit transaction across stacked | |
149 | * devices requires nothing moves, so if rtnl_lock is | |
150 | * not held, schedule a worker thread to hold rtnl_lock | |
151 | * while setting attr. | |
152 | */ | |
153 | ||
154 | return switchdev_port_attr_set_defer(dev, attr); | |
155 | } | |
156 | ||
157 | /* Phase I: prepare for attr set. Driver/device should fail | |
158 | * here if there are going to be issues in the commit phase, | |
159 | * such as lack of resources or support. The driver/device | |
160 | * should reserve resources needed for the commit phase here, | |
161 | * but should not commit the attr. | |
162 | */ | |
163 | ||
164 | attr->trans = SWITCHDEV_TRANS_PREPARE; | |
165 | err = __switchdev_port_attr_set(dev, attr); | |
166 | if (err) { | |
167 | /* Prepare phase failed: abort the transaction. Any | |
168 | * resources reserved in the prepare phase are | |
169 | * released. | |
170 | */ | |
171 | ||
172 | attr->trans = SWITCHDEV_TRANS_ABORT; | |
173 | __switchdev_port_attr_set(dev, attr); | |
174 | ||
175 | return err; | |
176 | } | |
177 | ||
178 | /* Phase II: commit attr set. This cannot fail as a fault | |
179 | * of driver/device. If it does, it's a bug in the driver/device | |
180 | * because the driver said everythings was OK in phase I. | |
181 | */ | |
182 | ||
183 | attr->trans = SWITCHDEV_TRANS_COMMIT; | |
184 | err = __switchdev_port_attr_set(dev, attr); | |
185 | BUG_ON(err); | |
186 | ||
187 | return err; | |
188 | } | |
189 | EXPORT_SYMBOL_GPL(switchdev_port_attr_set); | |
190 | ||
22c1f67e SF |
191 | static int __switchdev_port_obj_add(struct net_device *dev, |
192 | struct switchdev_obj *obj) | |
491d0f15 SF |
193 | { |
194 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
195 | struct net_device *lower_dev; | |
196 | struct list_head *iter; | |
197 | int err = -EOPNOTSUPP; | |
198 | ||
199 | if (ops && ops->switchdev_port_obj_add) | |
200 | return ops->switchdev_port_obj_add(dev, obj); | |
201 | ||
202 | /* Switch device port(s) may be stacked under | |
203 | * bond/team/vlan dev, so recurse down to add object on | |
204 | * each port. | |
205 | */ | |
206 | ||
207 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
208 | err = __switchdev_port_obj_add(lower_dev, obj); | |
209 | if (err) | |
210 | break; | |
211 | } | |
212 | ||
213 | return err; | |
214 | } | |
215 | ||
216 | /** | |
217 | * switchdev_port_obj_add - Add port object | |
218 | * | |
219 | * @dev: port device | |
220 | * @obj: object to add | |
221 | * | |
222 | * Use a 2-phase prepare-commit transaction model to ensure | |
223 | * system is not left in a partially updated state due to | |
224 | * failure from driver/device. | |
225 | * | |
226 | * rtnl_lock must be held. | |
227 | */ | |
228 | int switchdev_port_obj_add(struct net_device *dev, struct switchdev_obj *obj) | |
229 | { | |
230 | int err; | |
231 | ||
232 | ASSERT_RTNL(); | |
233 | ||
234 | /* Phase I: prepare for obj add. Driver/device should fail | |
235 | * here if there are going to be issues in the commit phase, | |
236 | * such as lack of resources or support. The driver/device | |
237 | * should reserve resources needed for the commit phase here, | |
238 | * but should not commit the obj. | |
239 | */ | |
240 | ||
241 | obj->trans = SWITCHDEV_TRANS_PREPARE; | |
242 | err = __switchdev_port_obj_add(dev, obj); | |
243 | if (err) { | |
244 | /* Prepare phase failed: abort the transaction. Any | |
245 | * resources reserved in the prepare phase are | |
246 | * released. | |
247 | */ | |
248 | ||
249 | obj->trans = SWITCHDEV_TRANS_ABORT; | |
250 | __switchdev_port_obj_add(dev, obj); | |
251 | ||
252 | return err; | |
253 | } | |
254 | ||
255 | /* Phase II: commit obj add. This cannot fail as a fault | |
256 | * of driver/device. If it does, it's a bug in the driver/device | |
257 | * because the driver said everythings was OK in phase I. | |
258 | */ | |
259 | ||
260 | obj->trans = SWITCHDEV_TRANS_COMMIT; | |
261 | err = __switchdev_port_obj_add(dev, obj); | |
262 | WARN(err, "%s: Commit of object (id=%d) failed.\n", dev->name, obj->id); | |
263 | ||
264 | return err; | |
265 | } | |
266 | EXPORT_SYMBOL_GPL(switchdev_port_obj_add); | |
267 | ||
268 | /** | |
269 | * switchdev_port_obj_del - Delete port object | |
270 | * | |
271 | * @dev: port device | |
272 | * @obj: object to delete | |
273 | */ | |
274 | int switchdev_port_obj_del(struct net_device *dev, struct switchdev_obj *obj) | |
275 | { | |
276 | const struct switchdev_ops *ops = dev->switchdev_ops; | |
277 | struct net_device *lower_dev; | |
278 | struct list_head *iter; | |
279 | int err = -EOPNOTSUPP; | |
280 | ||
281 | if (ops && ops->switchdev_port_obj_del) | |
282 | return ops->switchdev_port_obj_del(dev, obj); | |
283 | ||
284 | /* Switch device port(s) may be stacked under | |
285 | * bond/team/vlan dev, so recurse down to delete object on | |
286 | * each port. | |
287 | */ | |
288 | ||
289 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
290 | err = switchdev_port_obj_del(lower_dev, obj); | |
291 | if (err) | |
292 | break; | |
293 | } | |
294 | ||
295 | return err; | |
296 | } | |
297 | EXPORT_SYMBOL_GPL(switchdev_port_obj_del); | |
298 | ||
ebb9a03a JP |
299 | static DEFINE_MUTEX(switchdev_mutex); |
300 | static RAW_NOTIFIER_HEAD(switchdev_notif_chain); | |
03bf0c28 JP |
301 | |
302 | /** | |
ebb9a03a | 303 | * register_switchdev_notifier - Register notifier |
03bf0c28 JP |
304 | * @nb: notifier_block |
305 | * | |
306 | * Register switch device notifier. This should be used by code | |
307 | * which needs to monitor events happening in particular device. | |
308 | * Return values are same as for atomic_notifier_chain_register(). | |
309 | */ | |
ebb9a03a | 310 | int register_switchdev_notifier(struct notifier_block *nb) |
03bf0c28 JP |
311 | { |
312 | int err; | |
313 | ||
ebb9a03a JP |
314 | mutex_lock(&switchdev_mutex); |
315 | err = raw_notifier_chain_register(&switchdev_notif_chain, nb); | |
316 | mutex_unlock(&switchdev_mutex); | |
03bf0c28 JP |
317 | return err; |
318 | } | |
ebb9a03a | 319 | EXPORT_SYMBOL_GPL(register_switchdev_notifier); |
03bf0c28 JP |
320 | |
321 | /** | |
ebb9a03a | 322 | * unregister_switchdev_notifier - Unregister notifier |
03bf0c28 JP |
323 | * @nb: notifier_block |
324 | * | |
325 | * Unregister switch device notifier. | |
326 | * Return values are same as for atomic_notifier_chain_unregister(). | |
327 | */ | |
ebb9a03a | 328 | int unregister_switchdev_notifier(struct notifier_block *nb) |
03bf0c28 JP |
329 | { |
330 | int err; | |
331 | ||
ebb9a03a JP |
332 | mutex_lock(&switchdev_mutex); |
333 | err = raw_notifier_chain_unregister(&switchdev_notif_chain, nb); | |
334 | mutex_unlock(&switchdev_mutex); | |
03bf0c28 JP |
335 | return err; |
336 | } | |
ebb9a03a | 337 | EXPORT_SYMBOL_GPL(unregister_switchdev_notifier); |
03bf0c28 JP |
338 | |
339 | /** | |
ebb9a03a | 340 | * call_switchdev_notifiers - Call notifiers |
03bf0c28 JP |
341 | * @val: value passed unmodified to notifier function |
342 | * @dev: port device | |
343 | * @info: notifier information data | |
344 | * | |
345 | * Call all network notifier blocks. This should be called by driver | |
346 | * when it needs to propagate hardware event. | |
347 | * Return values are same as for atomic_notifier_call_chain(). | |
348 | */ | |
ebb9a03a JP |
349 | int call_switchdev_notifiers(unsigned long val, struct net_device *dev, |
350 | struct switchdev_notifier_info *info) | |
03bf0c28 JP |
351 | { |
352 | int err; | |
353 | ||
354 | info->dev = dev; | |
ebb9a03a JP |
355 | mutex_lock(&switchdev_mutex); |
356 | err = raw_notifier_call_chain(&switchdev_notif_chain, val, info); | |
357 | mutex_unlock(&switchdev_mutex); | |
03bf0c28 JP |
358 | return err; |
359 | } | |
ebb9a03a | 360 | EXPORT_SYMBOL_GPL(call_switchdev_notifiers); |
8a44dbb2 | 361 | |
8793d0a6 SF |
362 | /** |
363 | * switchdev_port_bridge_getlink - Get bridge port attributes | |
364 | * | |
365 | * @dev: port device | |
366 | * | |
367 | * Called for SELF on rtnl_bridge_getlink to get bridge port | |
368 | * attributes. | |
369 | */ | |
370 | int switchdev_port_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq, | |
371 | struct net_device *dev, u32 filter_mask, | |
372 | int nlflags) | |
373 | { | |
374 | struct switchdev_attr attr = { | |
375 | .id = SWITCHDEV_ATTR_PORT_BRIDGE_FLAGS, | |
376 | }; | |
377 | u16 mode = BRIDGE_MODE_UNDEF; | |
378 | u32 mask = BR_LEARNING | BR_LEARNING_SYNC; | |
379 | int err; | |
380 | ||
381 | err = switchdev_port_attr_get(dev, &attr); | |
382 | if (err) | |
383 | return err; | |
384 | ||
385 | return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode, | |
42275bd8 | 386 | attr.u.brport_flags, mask, nlflags); |
8793d0a6 SF |
387 | } |
388 | EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink); | |
389 | ||
47f8328b SF |
390 | static int switchdev_port_br_setflag(struct net_device *dev, |
391 | struct nlattr *nlattr, | |
392 | unsigned long brport_flag) | |
393 | { | |
394 | struct switchdev_attr attr = { | |
395 | .id = SWITCHDEV_ATTR_PORT_BRIDGE_FLAGS, | |
396 | }; | |
397 | u8 flag = nla_get_u8(nlattr); | |
398 | int err; | |
399 | ||
400 | err = switchdev_port_attr_get(dev, &attr); | |
401 | if (err) | |
402 | return err; | |
403 | ||
404 | if (flag) | |
42275bd8 | 405 | attr.u.brport_flags |= brport_flag; |
47f8328b | 406 | else |
42275bd8 | 407 | attr.u.brport_flags &= ~brport_flag; |
47f8328b SF |
408 | |
409 | return switchdev_port_attr_set(dev, &attr); | |
410 | } | |
411 | ||
412 | static const struct nla_policy | |
413 | switchdev_port_bridge_policy[IFLA_BRPORT_MAX + 1] = { | |
414 | [IFLA_BRPORT_STATE] = { .type = NLA_U8 }, | |
415 | [IFLA_BRPORT_COST] = { .type = NLA_U32 }, | |
416 | [IFLA_BRPORT_PRIORITY] = { .type = NLA_U16 }, | |
417 | [IFLA_BRPORT_MODE] = { .type = NLA_U8 }, | |
418 | [IFLA_BRPORT_GUARD] = { .type = NLA_U8 }, | |
419 | [IFLA_BRPORT_PROTECT] = { .type = NLA_U8 }, | |
420 | [IFLA_BRPORT_FAST_LEAVE] = { .type = NLA_U8 }, | |
421 | [IFLA_BRPORT_LEARNING] = { .type = NLA_U8 }, | |
422 | [IFLA_BRPORT_LEARNING_SYNC] = { .type = NLA_U8 }, | |
423 | [IFLA_BRPORT_UNICAST_FLOOD] = { .type = NLA_U8 }, | |
424 | }; | |
425 | ||
426 | static int switchdev_port_br_setlink_protinfo(struct net_device *dev, | |
427 | struct nlattr *protinfo) | |
428 | { | |
429 | struct nlattr *attr; | |
430 | int rem; | |
431 | int err; | |
432 | ||
433 | err = nla_validate_nested(protinfo, IFLA_BRPORT_MAX, | |
434 | switchdev_port_bridge_policy); | |
435 | if (err) | |
436 | return err; | |
437 | ||
438 | nla_for_each_nested(attr, protinfo, rem) { | |
439 | switch (nla_type(attr)) { | |
440 | case IFLA_BRPORT_LEARNING: | |
441 | err = switchdev_port_br_setflag(dev, attr, | |
442 | BR_LEARNING); | |
443 | break; | |
444 | case IFLA_BRPORT_LEARNING_SYNC: | |
445 | err = switchdev_port_br_setflag(dev, attr, | |
446 | BR_LEARNING_SYNC); | |
447 | break; | |
448 | default: | |
449 | err = -EOPNOTSUPP; | |
450 | break; | |
451 | } | |
452 | if (err) | |
453 | return err; | |
454 | } | |
455 | ||
456 | return 0; | |
457 | } | |
458 | ||
459 | static int switchdev_port_br_afspec(struct net_device *dev, | |
460 | struct nlattr *afspec, | |
461 | int (*f)(struct net_device *dev, | |
462 | struct switchdev_obj *obj)) | |
463 | { | |
464 | struct nlattr *attr; | |
465 | struct bridge_vlan_info *vinfo; | |
466 | struct switchdev_obj obj = { | |
467 | .id = SWITCHDEV_OBJ_PORT_VLAN, | |
468 | }; | |
42275bd8 | 469 | struct switchdev_obj_vlan *vlan = &obj.u.vlan; |
47f8328b SF |
470 | int rem; |
471 | int err; | |
472 | ||
473 | nla_for_each_nested(attr, afspec, rem) { | |
474 | if (nla_type(attr) != IFLA_BRIDGE_VLAN_INFO) | |
475 | continue; | |
476 | if (nla_len(attr) != sizeof(struct bridge_vlan_info)) | |
477 | return -EINVAL; | |
478 | vinfo = nla_data(attr); | |
42275bd8 | 479 | vlan->flags = vinfo->flags; |
47f8328b | 480 | if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_BEGIN) { |
42275bd8 | 481 | if (vlan->vid_start) |
47f8328b | 482 | return -EINVAL; |
42275bd8 | 483 | vlan->vid_start = vinfo->vid; |
47f8328b | 484 | } else if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_END) { |
42275bd8 | 485 | if (!vlan->vid_start) |
47f8328b | 486 | return -EINVAL; |
42275bd8 SF |
487 | vlan->vid_end = vinfo->vid; |
488 | if (vlan->vid_end <= vlan->vid_start) | |
47f8328b SF |
489 | return -EINVAL; |
490 | err = f(dev, &obj); | |
491 | if (err) | |
492 | return err; | |
42275bd8 | 493 | memset(vlan, 0, sizeof(*vlan)); |
47f8328b | 494 | } else { |
42275bd8 | 495 | if (vlan->vid_start) |
47f8328b | 496 | return -EINVAL; |
42275bd8 SF |
497 | vlan->vid_start = vinfo->vid; |
498 | vlan->vid_end = vinfo->vid; | |
47f8328b SF |
499 | err = f(dev, &obj); |
500 | if (err) | |
501 | return err; | |
42275bd8 | 502 | memset(vlan, 0, sizeof(*vlan)); |
47f8328b SF |
503 | } |
504 | } | |
505 | ||
506 | return 0; | |
507 | } | |
508 | ||
8a44dbb2 | 509 | /** |
47f8328b | 510 | * switchdev_port_bridge_setlink - Set bridge port attributes |
8a44dbb2 RP |
511 | * |
512 | * @dev: port device | |
47f8328b SF |
513 | * @nlh: netlink header |
514 | * @flags: netlink flags | |
8a44dbb2 | 515 | * |
47f8328b SF |
516 | * Called for SELF on rtnl_bridge_setlink to set bridge port |
517 | * attributes. | |
8a44dbb2 | 518 | */ |
ebb9a03a JP |
519 | int switchdev_port_bridge_setlink(struct net_device *dev, |
520 | struct nlmsghdr *nlh, u16 flags) | |
8a44dbb2 | 521 | { |
47f8328b SF |
522 | struct nlattr *protinfo; |
523 | struct nlattr *afspec; | |
524 | int err = 0; | |
525 | ||
526 | protinfo = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), | |
527 | IFLA_PROTINFO); | |
528 | if (protinfo) { | |
529 | err = switchdev_port_br_setlink_protinfo(dev, protinfo); | |
530 | if (err) | |
531 | return err; | |
532 | } | |
8a44dbb2 | 533 | |
47f8328b SF |
534 | afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), |
535 | IFLA_AF_SPEC); | |
536 | if (afspec) | |
537 | err = switchdev_port_br_afspec(dev, afspec, | |
538 | switchdev_port_obj_add); | |
8a44dbb2 | 539 | |
47f8328b | 540 | return err; |
8a44dbb2 | 541 | } |
ebb9a03a | 542 | EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink); |
8a44dbb2 RP |
543 | |
544 | /** | |
5c34e022 | 545 | * switchdev_port_bridge_dellink - Set bridge port attributes |
8a44dbb2 RP |
546 | * |
547 | * @dev: port device | |
5c34e022 SF |
548 | * @nlh: netlink header |
549 | * @flags: netlink flags | |
8a44dbb2 | 550 | * |
5c34e022 SF |
551 | * Called for SELF on rtnl_bridge_dellink to set bridge port |
552 | * attributes. | |
8a44dbb2 | 553 | */ |
ebb9a03a JP |
554 | int switchdev_port_bridge_dellink(struct net_device *dev, |
555 | struct nlmsghdr *nlh, u16 flags) | |
8a44dbb2 | 556 | { |
5c34e022 | 557 | struct nlattr *afspec; |
8a44dbb2 | 558 | |
5c34e022 SF |
559 | afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), |
560 | IFLA_AF_SPEC); | |
561 | if (afspec) | |
562 | return switchdev_port_br_afspec(dev, afspec, | |
563 | switchdev_port_obj_del); | |
8a44dbb2 | 564 | |
5c34e022 | 565 | return 0; |
8a44dbb2 | 566 | } |
ebb9a03a | 567 | EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink); |
8a44dbb2 | 568 | |
ebb9a03a | 569 | static struct net_device *switchdev_get_lowest_dev(struct net_device *dev) |
b5d6fbde | 570 | { |
9d47c0a2 | 571 | const struct switchdev_ops *ops = dev->switchdev_ops; |
b5d6fbde SF |
572 | struct net_device *lower_dev; |
573 | struct net_device *port_dev; | |
574 | struct list_head *iter; | |
575 | ||
576 | /* Recusively search down until we find a sw port dev. | |
f8e20a9f | 577 | * (A sw port dev supports switchdev_port_attr_get). |
b5d6fbde SF |
578 | */ |
579 | ||
f8e20a9f | 580 | if (ops && ops->switchdev_port_attr_get) |
b5d6fbde SF |
581 | return dev; |
582 | ||
583 | netdev_for_each_lower_dev(dev, lower_dev, iter) { | |
ebb9a03a | 584 | port_dev = switchdev_get_lowest_dev(lower_dev); |
b5d6fbde SF |
585 | if (port_dev) |
586 | return port_dev; | |
587 | } | |
588 | ||
589 | return NULL; | |
590 | } | |
591 | ||
ebb9a03a | 592 | static struct net_device *switchdev_get_dev_by_nhs(struct fib_info *fi) |
b5d6fbde | 593 | { |
f8e20a9f SF |
594 | struct switchdev_attr attr = { |
595 | .id = SWITCHDEV_ATTR_PORT_PARENT_ID, | |
596 | }; | |
597 | struct switchdev_attr prev_attr; | |
b5d6fbde SF |
598 | struct net_device *dev = NULL; |
599 | int nhsel; | |
600 | ||
601 | /* For this route, all nexthop devs must be on the same switch. */ | |
602 | ||
603 | for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) { | |
604 | const struct fib_nh *nh = &fi->fib_nh[nhsel]; | |
605 | ||
606 | if (!nh->nh_dev) | |
607 | return NULL; | |
608 | ||
ebb9a03a | 609 | dev = switchdev_get_lowest_dev(nh->nh_dev); |
b5d6fbde SF |
610 | if (!dev) |
611 | return NULL; | |
612 | ||
f8e20a9f | 613 | if (switchdev_port_attr_get(dev, &attr)) |
b5d6fbde SF |
614 | return NULL; |
615 | ||
616 | if (nhsel > 0) { | |
42275bd8 | 617 | if (prev_attr.u.ppid.id_len != attr.u.ppid.id_len) |
b5d6fbde | 618 | return NULL; |
42275bd8 SF |
619 | if (memcmp(prev_attr.u.ppid.id, attr.u.ppid.id, |
620 | attr.u.ppid.id_len)) | |
b5d6fbde SF |
621 | return NULL; |
622 | } | |
623 | ||
f8e20a9f | 624 | prev_attr = attr; |
b5d6fbde SF |
625 | } |
626 | ||
627 | return dev; | |
628 | } | |
629 | ||
5e8d9049 | 630 | /** |
ebb9a03a | 631 | * switchdev_fib_ipv4_add - Add IPv4 route entry to switch |
5e8d9049 SF |
632 | * |
633 | * @dst: route's IPv4 destination address | |
634 | * @dst_len: destination address length (prefix length) | |
635 | * @fi: route FIB info structure | |
636 | * @tos: route TOS | |
637 | * @type: route type | |
f8f21471 | 638 | * @nlflags: netlink flags passed in (NLM_F_*) |
5e8d9049 SF |
639 | * @tb_id: route table ID |
640 | * | |
641 | * Add IPv4 route entry to switch device. | |
642 | */ | |
ebb9a03a JP |
643 | int switchdev_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi, |
644 | u8 tos, u8 type, u32 nlflags, u32 tb_id) | |
5e8d9049 | 645 | { |
58c2cb16 SF |
646 | struct switchdev_obj fib_obj = { |
647 | .id = SWITCHDEV_OBJ_IPV4_FIB, | |
42275bd8 | 648 | .u.ipv4_fib = { |
7a7ee531 | 649 | .dst = dst, |
58c2cb16 SF |
650 | .dst_len = dst_len, |
651 | .fi = fi, | |
652 | .tos = tos, | |
653 | .type = type, | |
654 | .nlflags = nlflags, | |
655 | .tb_id = tb_id, | |
656 | }, | |
657 | }; | |
b5d6fbde | 658 | struct net_device *dev; |
b5d6fbde SF |
659 | int err = 0; |
660 | ||
8e05fd71 SF |
661 | /* Don't offload route if using custom ip rules or if |
662 | * IPv4 FIB offloading has been disabled completely. | |
663 | */ | |
664 | ||
e1315db1 SF |
665 | #ifdef CONFIG_IP_MULTIPLE_TABLES |
666 | if (fi->fib_net->ipv4.fib_has_custom_rules) | |
667 | return 0; | |
668 | #endif | |
669 | ||
670 | if (fi->fib_net->ipv4.fib_offload_disabled) | |
104616e7 SF |
671 | return 0; |
672 | ||
ebb9a03a | 673 | dev = switchdev_get_dev_by_nhs(fi); |
b5d6fbde SF |
674 | if (!dev) |
675 | return 0; | |
58c2cb16 SF |
676 | |
677 | err = switchdev_port_obj_add(dev, &fib_obj); | |
678 | if (!err) | |
679 | fi->fib_flags |= RTNH_F_EXTERNAL; | |
b5d6fbde SF |
680 | |
681 | return err; | |
5e8d9049 | 682 | } |
ebb9a03a | 683 | EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add); |
5e8d9049 SF |
684 | |
685 | /** | |
ebb9a03a | 686 | * switchdev_fib_ipv4_del - Delete IPv4 route entry from switch |
5e8d9049 SF |
687 | * |
688 | * @dst: route's IPv4 destination address | |
689 | * @dst_len: destination address length (prefix length) | |
690 | * @fi: route FIB info structure | |
691 | * @tos: route TOS | |
692 | * @type: route type | |
693 | * @tb_id: route table ID | |
694 | * | |
695 | * Delete IPv4 route entry from switch device. | |
696 | */ | |
ebb9a03a JP |
697 | int switchdev_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi, |
698 | u8 tos, u8 type, u32 tb_id) | |
5e8d9049 | 699 | { |
58c2cb16 SF |
700 | struct switchdev_obj fib_obj = { |
701 | .id = SWITCHDEV_OBJ_IPV4_FIB, | |
42275bd8 | 702 | .u.ipv4_fib = { |
7a7ee531 | 703 | .dst = dst, |
58c2cb16 SF |
704 | .dst_len = dst_len, |
705 | .fi = fi, | |
706 | .tos = tos, | |
707 | .type = type, | |
708 | .nlflags = 0, | |
709 | .tb_id = tb_id, | |
710 | }, | |
711 | }; | |
b5d6fbde | 712 | struct net_device *dev; |
b5d6fbde SF |
713 | int err = 0; |
714 | ||
715 | if (!(fi->fib_flags & RTNH_F_EXTERNAL)) | |
716 | return 0; | |
717 | ||
ebb9a03a | 718 | dev = switchdev_get_dev_by_nhs(fi); |
b5d6fbde SF |
719 | if (!dev) |
720 | return 0; | |
b5d6fbde | 721 | |
58c2cb16 SF |
722 | err = switchdev_port_obj_del(dev, &fib_obj); |
723 | if (!err) | |
724 | fi->fib_flags &= ~RTNH_F_EXTERNAL; | |
b5d6fbde SF |
725 | |
726 | return err; | |
5e8d9049 | 727 | } |
ebb9a03a | 728 | EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del); |
8e05fd71 SF |
729 | |
730 | /** | |
ebb9a03a | 731 | * switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation |
8e05fd71 SF |
732 | * |
733 | * @fi: route FIB info structure | |
734 | */ | |
ebb9a03a | 735 | void switchdev_fib_ipv4_abort(struct fib_info *fi) |
8e05fd71 SF |
736 | { |
737 | /* There was a problem installing this route to the offload | |
738 | * device. For now, until we come up with more refined | |
739 | * policy handling, abruptly end IPv4 fib offloading for | |
740 | * for entire net by flushing offload device(s) of all | |
741 | * IPv4 routes, and mark IPv4 fib offloading broken from | |
742 | * this point forward. | |
743 | */ | |
744 | ||
745 | fib_flush_external(fi->fib_net); | |
746 | fi->fib_net->ipv4.fib_offload_disabled = true; | |
747 | } | |
ebb9a03a | 748 | EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort); |