2 * net/switchdev/switchdev.c - Switch device API
3 * Copyright (c) 2014-2015 Jiri Pirko <jiri@resnulli.us>
4 * Copyright (c) 2014-2015 Scott Feldman <sfeldma@gmail.com>
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.
12 #include <linux/kernel.h>
13 #include <linux/types.h>
14 #include <linux/init.h>
15 #include <linux/mutex.h>
16 #include <linux/notifier.h>
17 #include <linux/netdevice.h>
18 #include <linux/etherdevice.h>
19 #include <linux/if_bridge.h>
20 #include <linux/list.h>
21 #include <linux/workqueue.h>
22 #include <net/ip_fib.h>
23 #include <net/switchdev.h>
26 * switchdev_trans_item_enqueue - Enqueue data item to transaction queue
29 * @data: pointer to data being queued
30 * @destructor: data destructor
31 * @tritem: transaction item being queued
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.
38 void switchdev_trans_item_enqueue(struct switchdev_trans *trans,
39 void *data, void (*destructor)(void const *),
40 struct switchdev_trans_item *tritem)
43 tritem->destructor = destructor;
44 list_add_tail(&tritem->list, &trans->item_list);
46 EXPORT_SYMBOL_GPL(switchdev_trans_item_enqueue);
48 static struct switchdev_trans_item *
49 __switchdev_trans_item_dequeue(struct switchdev_trans *trans)
51 struct switchdev_trans_item *tritem;
53 if (list_empty(&trans->item_list))
55 tritem = list_first_entry(&trans->item_list,
56 struct switchdev_trans_item, list);
57 list_del(&tritem->list);
62 * switchdev_trans_item_dequeue - Dequeue data item from transaction queue
66 void *switchdev_trans_item_dequeue(struct switchdev_trans *trans)
68 struct switchdev_trans_item *tritem;
70 tritem = __switchdev_trans_item_dequeue(trans);
74 EXPORT_SYMBOL_GPL(switchdev_trans_item_dequeue);
76 static void switchdev_trans_init(struct switchdev_trans *trans)
78 INIT_LIST_HEAD(&trans->item_list);
81 static void switchdev_trans_items_destroy(struct switchdev_trans *trans)
83 struct switchdev_trans_item *tritem;
85 while ((tritem = __switchdev_trans_item_dequeue(trans)))
86 tritem->destructor(tritem->data);
89 static void switchdev_trans_items_warn_destroy(struct net_device *dev,
90 struct switchdev_trans *trans)
92 WARN(!list_empty(&trans->item_list), "%s: transaction item queue is not empty.\n",
94 switchdev_trans_items_destroy(trans);
97 static LIST_HEAD(deferred);
98 static DEFINE_SPINLOCK(deferred_lock);
100 typedef void switchdev_deferred_func_t(struct net_device *dev,
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];
110 static struct switchdev_deferred_item *switchdev_deferred_dequeue(void)
112 struct switchdev_deferred_item *dfitem;
114 spin_lock_bh(&deferred_lock);
115 if (list_empty(&deferred)) {
119 dfitem = list_first_entry(&deferred,
120 struct switchdev_deferred_item, list);
121 list_del(&dfitem->list);
123 spin_unlock_bh(&deferred_lock);
128 * switchdev_deferred_process - Process ops in deferred queue
130 * Called to flush the ops currently queued in deferred ops queue.
131 * rtnl_lock must be held.
133 void switchdev_deferred_process(void)
135 struct switchdev_deferred_item *dfitem;
139 while ((dfitem = switchdev_deferred_dequeue())) {
140 dfitem->func(dfitem->dev, dfitem->data);
141 dev_put(dfitem->dev);
145 EXPORT_SYMBOL_GPL(switchdev_deferred_process);
147 static void switchdev_deferred_process_work(struct work_struct *work)
150 switchdev_deferred_process();
154 static DECLARE_WORK(deferred_process_work, switchdev_deferred_process_work);
156 static int switchdev_deferred_enqueue(struct net_device *dev,
157 const void *data, size_t data_len,
158 switchdev_deferred_func_t *func)
160 struct switchdev_deferred_item *dfitem;
162 dfitem = kmalloc(sizeof(*dfitem) + data_len, GFP_ATOMIC);
167 memcpy(dfitem->data, data, data_len);
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);
177 * switchdev_port_attr_get - Get port attribute
180 * @attr: attribute to get
182 int switchdev_port_attr_get(struct net_device *dev, struct switchdev_attr *attr)
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 = {
188 .id = SWITCHDEV_ATTR_ID_UNDEFINED
190 int err = -EOPNOTSUPP;
192 if (ops && ops->switchdev_port_attr_get)
193 return ops->switchdev_port_attr_get(dev, attr);
195 if (attr->flags & SWITCHDEV_F_NO_RECURSE)
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.
204 netdev_for_each_lower_dev(dev, lower_dev, iter) {
205 err = switchdev_port_attr_get(lower_dev, attr);
208 if (first.id == SWITCHDEV_ATTR_ID_UNDEFINED)
210 else if (memcmp(&first, attr, sizeof(*attr)))
216 EXPORT_SYMBOL_GPL(switchdev_port_attr_get);
218 static int __switchdev_port_attr_set(struct net_device *dev,
219 const struct switchdev_attr *attr,
220 struct switchdev_trans *trans)
222 const struct switchdev_ops *ops = dev->switchdev_ops;
223 struct net_device *lower_dev;
224 struct list_head *iter;
225 int err = -EOPNOTSUPP;
227 if (ops && ops->switchdev_port_attr_set)
228 return ops->switchdev_port_attr_set(dev, attr, trans);
230 if (attr->flags & SWITCHDEV_F_NO_RECURSE)
233 /* Switch device port(s) may be stacked under
234 * bond/team/vlan dev, so recurse down to set attr on
238 netdev_for_each_lower_dev(dev, lower_dev, iter) {
239 err = __switchdev_port_attr_set(lower_dev, attr, trans);
240 if (err == -EOPNOTSUPP &&
241 attr->flags & SWITCHDEV_F_SKIP_EOPNOTSUPP)
248 if (err == -EOPNOTSUPP && attr->flags & SWITCHDEV_F_SKIP_EOPNOTSUPP)
254 static int switchdev_port_attr_set_now(struct net_device *dev,
255 const struct switchdev_attr *attr)
257 struct switchdev_trans trans;
260 switchdev_trans_init(&trans);
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.
269 trans.ph_prepare = true;
270 err = __switchdev_port_attr_set(dev, attr, &trans);
272 /* Prepare phase failed: abort the transaction. Any
273 * resources reserved in the prepare phase are
277 if (err != -EOPNOTSUPP)
278 switchdev_trans_items_destroy(&trans);
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.
288 trans.ph_prepare = false;
289 err = __switchdev_port_attr_set(dev, attr, &trans);
290 WARN(err, "%s: Commit of attribute (id=%d) failed.\n",
291 dev->name, attr->id);
292 switchdev_trans_items_warn_destroy(dev, &trans);
297 static void switchdev_port_attr_set_deferred(struct net_device *dev,
300 const struct switchdev_attr *attr = data;
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",
309 static int switchdev_port_attr_set_defer(struct net_device *dev,
310 const struct switchdev_attr *attr)
312 return switchdev_deferred_enqueue(dev, attr, sizeof(*attr),
313 switchdev_port_attr_set_deferred);
317 * switchdev_port_attr_set - Set port attribute
320 * @attr: attribute to set
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.
326 * rtnl_lock must be held and must not be in atomic section,
327 * in case SWITCHDEV_F_DEFER flag is not set.
329 int switchdev_port_attr_set(struct net_device *dev,
330 const struct switchdev_attr *attr)
332 if (attr->flags & SWITCHDEV_F_DEFER)
333 return switchdev_port_attr_set_defer(dev, attr);
335 return switchdev_port_attr_set_now(dev, attr);
337 EXPORT_SYMBOL_GPL(switchdev_port_attr_set);
339 static int __switchdev_port_obj_add(struct net_device *dev,
340 const struct switchdev_obj *obj,
341 struct switchdev_trans *trans)
343 const struct switchdev_ops *ops = dev->switchdev_ops;
344 struct net_device *lower_dev;
345 struct list_head *iter;
346 int err = -EOPNOTSUPP;
348 if (ops && ops->switchdev_port_obj_add)
349 return ops->switchdev_port_obj_add(dev, obj, trans);
351 /* Switch device port(s) may be stacked under
352 * bond/team/vlan dev, so recurse down to add object on
356 netdev_for_each_lower_dev(dev, lower_dev, iter) {
357 err = __switchdev_port_obj_add(lower_dev, obj, trans);
365 static int switchdev_port_obj_add_now(struct net_device *dev,
366 const struct switchdev_obj *obj)
368 struct switchdev_trans trans;
373 switchdev_trans_init(&trans);
375 /* Phase I: prepare for obj add. Driver/device should fail
376 * here if there are going to be issues in the commit phase,
377 * such as lack of resources or support. The driver/device
378 * should reserve resources needed for the commit phase here,
379 * but should not commit the obj.
382 trans.ph_prepare = true;
383 err = __switchdev_port_obj_add(dev, obj, &trans);
385 /* Prepare phase failed: abort the transaction. Any
386 * resources reserved in the prepare phase are
390 if (err != -EOPNOTSUPP)
391 switchdev_trans_items_destroy(&trans);
396 /* Phase II: commit obj add. This cannot fail as a fault
397 * of driver/device. If it does, it's a bug in the driver/device
398 * because the driver said everythings was OK in phase I.
401 trans.ph_prepare = false;
402 err = __switchdev_port_obj_add(dev, obj, &trans);
403 WARN(err, "%s: Commit of object (id=%d) failed.\n", dev->name, obj->id);
404 switchdev_trans_items_warn_destroy(dev, &trans);
409 static void switchdev_port_obj_add_deferred(struct net_device *dev,
412 const struct switchdev_obj *obj = data;
415 err = switchdev_port_obj_add_now(dev, obj);
416 if (err && err != -EOPNOTSUPP)
417 netdev_err(dev, "failed (err=%d) to add object (id=%d)\n",
421 static int switchdev_port_obj_add_defer(struct net_device *dev,
422 const struct switchdev_obj *obj)
424 return switchdev_deferred_enqueue(dev, obj, sizeof(*obj),
425 switchdev_port_obj_add_deferred);
429 * switchdev_port_obj_add - Add port object
433 * @obj: object to add
435 * Use a 2-phase prepare-commit transaction model to ensure
436 * system is not left in a partially updated state due to
437 * failure from driver/device.
439 * rtnl_lock must be held and must not be in atomic section,
440 * in case SWITCHDEV_F_DEFER flag is not set.
442 int switchdev_port_obj_add(struct net_device *dev,
443 const struct switchdev_obj *obj)
445 if (obj->flags & SWITCHDEV_F_DEFER)
446 return switchdev_port_obj_add_defer(dev, obj);
448 return switchdev_port_obj_add_now(dev, obj);
450 EXPORT_SYMBOL_GPL(switchdev_port_obj_add);
452 static int switchdev_port_obj_del_now(struct net_device *dev,
453 const struct switchdev_obj *obj)
455 const struct switchdev_ops *ops = dev->switchdev_ops;
456 struct net_device *lower_dev;
457 struct list_head *iter;
458 int err = -EOPNOTSUPP;
460 if (ops && ops->switchdev_port_obj_del)
461 return ops->switchdev_port_obj_del(dev, obj);
463 /* Switch device port(s) may be stacked under
464 * bond/team/vlan dev, so recurse down to delete object on
468 netdev_for_each_lower_dev(dev, lower_dev, iter) {
469 err = switchdev_port_obj_del_now(lower_dev, obj);
477 static void switchdev_port_obj_del_deferred(struct net_device *dev,
480 const struct switchdev_obj *obj = data;
483 err = switchdev_port_obj_del_now(dev, obj);
484 if (err && err != -EOPNOTSUPP)
485 netdev_err(dev, "failed (err=%d) to del object (id=%d)\n",
489 static int switchdev_port_obj_del_defer(struct net_device *dev,
490 const struct switchdev_obj *obj)
492 return switchdev_deferred_enqueue(dev, obj, sizeof(*obj),
493 switchdev_port_obj_del_deferred);
497 * switchdev_port_obj_del - Delete port object
501 * @obj: object to delete
503 * rtnl_lock must be held and must not be in atomic section,
504 * in case SWITCHDEV_F_DEFER flag is not set.
506 int switchdev_port_obj_del(struct net_device *dev,
507 const struct switchdev_obj *obj)
509 if (obj->flags & SWITCHDEV_F_DEFER)
510 return switchdev_port_obj_del_defer(dev, obj);
512 return switchdev_port_obj_del_now(dev, obj);
514 EXPORT_SYMBOL_GPL(switchdev_port_obj_del);
517 * switchdev_port_obj_dump - Dump port objects
521 * @obj: object to dump
522 * @cb: function to call with a filled object
524 * rtnl_lock must be held.
526 int switchdev_port_obj_dump(struct net_device *dev, struct switchdev_obj *obj,
527 switchdev_obj_dump_cb_t *cb)
529 const struct switchdev_ops *ops = dev->switchdev_ops;
530 struct net_device *lower_dev;
531 struct list_head *iter;
532 int err = -EOPNOTSUPP;
536 if (ops && ops->switchdev_port_obj_dump)
537 return ops->switchdev_port_obj_dump(dev, obj, cb);
539 /* Switch device port(s) may be stacked under
540 * bond/team/vlan dev, so recurse down to dump objects on
541 * first port at bottom of stack.
544 netdev_for_each_lower_dev(dev, lower_dev, iter) {
545 err = switchdev_port_obj_dump(lower_dev, obj, cb);
551 EXPORT_SYMBOL_GPL(switchdev_port_obj_dump);
553 static DEFINE_MUTEX(switchdev_mutex);
554 static RAW_NOTIFIER_HEAD(switchdev_notif_chain);
557 * register_switchdev_notifier - Register notifier
558 * @nb: notifier_block
560 * Register switch device notifier. This should be used by code
561 * which needs to monitor events happening in particular device.
562 * Return values are same as for atomic_notifier_chain_register().
564 int register_switchdev_notifier(struct notifier_block *nb)
568 mutex_lock(&switchdev_mutex);
569 err = raw_notifier_chain_register(&switchdev_notif_chain, nb);
570 mutex_unlock(&switchdev_mutex);
573 EXPORT_SYMBOL_GPL(register_switchdev_notifier);
576 * unregister_switchdev_notifier - Unregister notifier
577 * @nb: notifier_block
579 * Unregister switch device notifier.
580 * Return values are same as for atomic_notifier_chain_unregister().
582 int unregister_switchdev_notifier(struct notifier_block *nb)
586 mutex_lock(&switchdev_mutex);
587 err = raw_notifier_chain_unregister(&switchdev_notif_chain, nb);
588 mutex_unlock(&switchdev_mutex);
591 EXPORT_SYMBOL_GPL(unregister_switchdev_notifier);
594 * call_switchdev_notifiers - Call notifiers
595 * @val: value passed unmodified to notifier function
597 * @info: notifier information data
599 * Call all network notifier blocks. This should be called by driver
600 * when it needs to propagate hardware event.
601 * Return values are same as for atomic_notifier_call_chain().
603 int call_switchdev_notifiers(unsigned long val, struct net_device *dev,
604 struct switchdev_notifier_info *info)
609 mutex_lock(&switchdev_mutex);
610 err = raw_notifier_call_chain(&switchdev_notif_chain, val, info);
611 mutex_unlock(&switchdev_mutex);
614 EXPORT_SYMBOL_GPL(call_switchdev_notifiers);
616 struct switchdev_vlan_dump {
617 struct switchdev_obj_port_vlan vlan;
625 static int switchdev_port_vlan_dump_put(struct switchdev_vlan_dump *dump)
627 struct bridge_vlan_info vinfo;
629 vinfo.flags = dump->flags;
631 if (dump->begin == 0 && dump->end == 0) {
633 } else if (dump->begin == dump->end) {
634 vinfo.vid = dump->begin;
635 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
636 sizeof(vinfo), &vinfo))
639 vinfo.vid = dump->begin;
640 vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_BEGIN;
641 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
642 sizeof(vinfo), &vinfo))
644 vinfo.vid = dump->end;
645 vinfo.flags &= ~BRIDGE_VLAN_INFO_RANGE_BEGIN;
646 vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_END;
647 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
648 sizeof(vinfo), &vinfo))
655 static int switchdev_port_vlan_dump_cb(struct switchdev_obj *obj)
657 struct switchdev_obj_port_vlan *vlan = SWITCHDEV_OBJ_PORT_VLAN(obj);
658 struct switchdev_vlan_dump *dump =
659 container_of(vlan, struct switchdev_vlan_dump, vlan);
662 if (vlan->vid_begin > vlan->vid_end)
665 if (dump->filter_mask & RTEXT_FILTER_BRVLAN) {
666 dump->flags = vlan->flags;
667 for (dump->begin = dump->end = vlan->vid_begin;
668 dump->begin <= vlan->vid_end;
669 dump->begin++, dump->end++) {
670 err = switchdev_port_vlan_dump_put(dump);
674 } else if (dump->filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED) {
675 if (dump->begin > vlan->vid_begin &&
676 dump->begin >= vlan->vid_end) {
677 if ((dump->begin - 1) == vlan->vid_end &&
678 dump->flags == vlan->flags) {
680 dump->begin = vlan->vid_begin;
682 err = switchdev_port_vlan_dump_put(dump);
683 dump->flags = vlan->flags;
684 dump->begin = vlan->vid_begin;
685 dump->end = vlan->vid_end;
687 } else if (dump->end <= vlan->vid_begin &&
688 dump->end < vlan->vid_end) {
689 if ((dump->end + 1) == vlan->vid_begin &&
690 dump->flags == vlan->flags) {
692 dump->end = vlan->vid_end;
694 err = switchdev_port_vlan_dump_put(dump);
695 dump->flags = vlan->flags;
696 dump->begin = vlan->vid_begin;
697 dump->end = vlan->vid_end;
707 static int switchdev_port_vlan_fill(struct sk_buff *skb, struct net_device *dev,
710 struct switchdev_vlan_dump dump = {
711 .vlan.obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN,
713 .filter_mask = filter_mask,
717 if ((filter_mask & RTEXT_FILTER_BRVLAN) ||
718 (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)) {
719 err = switchdev_port_obj_dump(dev, &dump.vlan.obj,
720 switchdev_port_vlan_dump_cb);
723 if (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)
725 err = switchdev_port_vlan_dump_put(&dump);
729 return err == -EOPNOTSUPP ? 0 : err;
733 * switchdev_port_bridge_getlink - Get bridge port attributes
737 * Called for SELF on rtnl_bridge_getlink to get bridge port
740 int switchdev_port_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
741 struct net_device *dev, u32 filter_mask,
744 struct switchdev_attr attr = {
745 .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS,
747 u16 mode = BRIDGE_MODE_UNDEF;
748 u32 mask = BR_LEARNING | BR_LEARNING_SYNC;
751 err = switchdev_port_attr_get(dev, &attr);
752 if (err && err != -EOPNOTSUPP)
755 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode,
756 attr.u.brport_flags, mask, nlflags,
757 filter_mask, switchdev_port_vlan_fill);
759 EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink);
761 static int switchdev_port_br_setflag(struct net_device *dev,
762 struct nlattr *nlattr,
763 unsigned long brport_flag)
765 struct switchdev_attr attr = {
766 .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS,
768 u8 flag = nla_get_u8(nlattr);
771 err = switchdev_port_attr_get(dev, &attr);
776 attr.u.brport_flags |= brport_flag;
778 attr.u.brport_flags &= ~brport_flag;
780 return switchdev_port_attr_set(dev, &attr);
783 static const struct nla_policy
784 switchdev_port_bridge_policy[IFLA_BRPORT_MAX + 1] = {
785 [IFLA_BRPORT_STATE] = { .type = NLA_U8 },
786 [IFLA_BRPORT_COST] = { .type = NLA_U32 },
787 [IFLA_BRPORT_PRIORITY] = { .type = NLA_U16 },
788 [IFLA_BRPORT_MODE] = { .type = NLA_U8 },
789 [IFLA_BRPORT_GUARD] = { .type = NLA_U8 },
790 [IFLA_BRPORT_PROTECT] = { .type = NLA_U8 },
791 [IFLA_BRPORT_FAST_LEAVE] = { .type = NLA_U8 },
792 [IFLA_BRPORT_LEARNING] = { .type = NLA_U8 },
793 [IFLA_BRPORT_LEARNING_SYNC] = { .type = NLA_U8 },
794 [IFLA_BRPORT_UNICAST_FLOOD] = { .type = NLA_U8 },
797 static int switchdev_port_br_setlink_protinfo(struct net_device *dev,
798 struct nlattr *protinfo)
804 err = nla_validate_nested(protinfo, IFLA_BRPORT_MAX,
805 switchdev_port_bridge_policy);
809 nla_for_each_nested(attr, protinfo, rem) {
810 switch (nla_type(attr)) {
811 case IFLA_BRPORT_LEARNING:
812 err = switchdev_port_br_setflag(dev, attr,
815 case IFLA_BRPORT_LEARNING_SYNC:
816 err = switchdev_port_br_setflag(dev, attr,
830 static int switchdev_port_br_afspec(struct net_device *dev,
831 struct nlattr *afspec,
832 int (*f)(struct net_device *dev,
833 const struct switchdev_obj *obj))
836 struct bridge_vlan_info *vinfo;
837 struct switchdev_obj_port_vlan vlan = {
838 .obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN,
843 nla_for_each_nested(attr, afspec, rem) {
844 if (nla_type(attr) != IFLA_BRIDGE_VLAN_INFO)
846 if (nla_len(attr) != sizeof(struct bridge_vlan_info))
848 vinfo = nla_data(attr);
849 vlan.flags = vinfo->flags;
850 if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_BEGIN) {
853 vlan.vid_begin = vinfo->vid;
854 /* don't allow range of pvids */
855 if (vlan.flags & BRIDGE_VLAN_INFO_PVID)
857 } else if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_END) {
860 vlan.vid_end = vinfo->vid;
861 if (vlan.vid_end <= vlan.vid_begin)
863 err = f(dev, &vlan.obj);
866 memset(&vlan, 0, sizeof(vlan));
870 vlan.vid_begin = vinfo->vid;
871 vlan.vid_end = vinfo->vid;
872 err = f(dev, &vlan.obj);
875 memset(&vlan, 0, sizeof(vlan));
883 * switchdev_port_bridge_setlink - Set bridge port attributes
886 * @nlh: netlink header
887 * @flags: netlink flags
889 * Called for SELF on rtnl_bridge_setlink to set bridge port
892 int switchdev_port_bridge_setlink(struct net_device *dev,
893 struct nlmsghdr *nlh, u16 flags)
895 struct nlattr *protinfo;
896 struct nlattr *afspec;
899 protinfo = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
902 err = switchdev_port_br_setlink_protinfo(dev, protinfo);
907 afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
910 err = switchdev_port_br_afspec(dev, afspec,
911 switchdev_port_obj_add);
915 EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink);
918 * switchdev_port_bridge_dellink - Set bridge port attributes
921 * @nlh: netlink header
922 * @flags: netlink flags
924 * Called for SELF on rtnl_bridge_dellink to set bridge port
927 int switchdev_port_bridge_dellink(struct net_device *dev,
928 struct nlmsghdr *nlh, u16 flags)
930 struct nlattr *afspec;
932 afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
935 return switchdev_port_br_afspec(dev, afspec,
936 switchdev_port_obj_del);
940 EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink);
943 * switchdev_port_fdb_add - Add FDB (MAC/VLAN) entry to port
945 * @ndmsg: netlink hdr
946 * @nlattr: netlink attributes
948 * @addr: MAC address to add
951 * Add FDB entry to switch device.
953 int switchdev_port_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
954 struct net_device *dev, const unsigned char *addr,
955 u16 vid, u16 nlm_flags)
957 struct switchdev_obj_port_fdb fdb = {
958 .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
962 ether_addr_copy(fdb.addr, addr);
963 return switchdev_port_obj_add(dev, &fdb.obj);
965 EXPORT_SYMBOL_GPL(switchdev_port_fdb_add);
968 * switchdev_port_fdb_del - Delete FDB (MAC/VLAN) entry from port
970 * @ndmsg: netlink hdr
971 * @nlattr: netlink attributes
973 * @addr: MAC address to delete
974 * @vid: VLAN to delete
976 * Delete FDB entry from switch device.
978 int switchdev_port_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
979 struct net_device *dev, const unsigned char *addr,
982 struct switchdev_obj_port_fdb fdb = {
983 .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
987 ether_addr_copy(fdb.addr, addr);
988 return switchdev_port_obj_del(dev, &fdb.obj);
990 EXPORT_SYMBOL_GPL(switchdev_port_fdb_del);
992 struct switchdev_fdb_dump {
993 struct switchdev_obj_port_fdb fdb;
994 struct net_device *dev;
996 struct netlink_callback *cb;
1000 static int switchdev_port_fdb_dump_cb(struct switchdev_obj *obj)
1002 struct switchdev_obj_port_fdb *fdb = SWITCHDEV_OBJ_PORT_FDB(obj);
1003 struct switchdev_fdb_dump *dump =
1004 container_of(fdb, struct switchdev_fdb_dump, fdb);
1005 u32 portid = NETLINK_CB(dump->cb->skb).portid;
1006 u32 seq = dump->cb->nlh->nlmsg_seq;
1007 struct nlmsghdr *nlh;
1010 if (dump->idx < dump->cb->args[0])
1013 nlh = nlmsg_put(dump->skb, portid, seq, RTM_NEWNEIGH,
1014 sizeof(*ndm), NLM_F_MULTI);
1018 ndm = nlmsg_data(nlh);
1019 ndm->ndm_family = AF_BRIDGE;
1022 ndm->ndm_flags = NTF_SELF;
1024 ndm->ndm_ifindex = dump->dev->ifindex;
1025 ndm->ndm_state = fdb->ndm_state;
1027 if (nla_put(dump->skb, NDA_LLADDR, ETH_ALEN, fdb->addr))
1028 goto nla_put_failure;
1030 if (fdb->vid && nla_put_u16(dump->skb, NDA_VLAN, fdb->vid))
1031 goto nla_put_failure;
1033 nlmsg_end(dump->skb, nlh);
1040 nlmsg_cancel(dump->skb, nlh);
1045 * switchdev_port_fdb_dump - Dump port FDB (MAC/VLAN) entries
1048 * @cb: netlink callback
1050 * @filter_dev: filter device
1053 * Delete FDB entry from switch device.
1055 int switchdev_port_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1056 struct net_device *dev,
1057 struct net_device *filter_dev, int idx)
1059 struct switchdev_fdb_dump dump = {
1060 .fdb.obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
1067 switchdev_port_obj_dump(dev, &dump.fdb.obj, switchdev_port_fdb_dump_cb);
1070 EXPORT_SYMBOL_GPL(switchdev_port_fdb_dump);
1072 static struct net_device *switchdev_get_lowest_dev(struct net_device *dev)
1074 const struct switchdev_ops *ops = dev->switchdev_ops;
1075 struct net_device *lower_dev;
1076 struct net_device *port_dev;
1077 struct list_head *iter;
1079 /* Recusively search down until we find a sw port dev.
1080 * (A sw port dev supports switchdev_port_attr_get).
1083 if (ops && ops->switchdev_port_attr_get)
1086 netdev_for_each_lower_dev(dev, lower_dev, iter) {
1087 port_dev = switchdev_get_lowest_dev(lower_dev);
1095 static struct net_device *switchdev_get_dev_by_nhs(struct fib_info *fi)
1097 struct switchdev_attr attr = {
1098 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1100 struct switchdev_attr prev_attr;
1101 struct net_device *dev = NULL;
1106 /* For this route, all nexthop devs must be on the same switch. */
1108 for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
1109 const struct fib_nh *nh = &fi->fib_nh[nhsel];
1114 dev = switchdev_get_lowest_dev(nh->nh_dev);
1118 if (switchdev_port_attr_get(dev, &attr))
1122 !netdev_phys_item_id_same(&prev_attr.u.ppid, &attr.u.ppid))
1132 * switchdev_fib_ipv4_add - Add/modify switch IPv4 route entry
1134 * @dst: route's IPv4 destination address
1135 * @dst_len: destination address length (prefix length)
1136 * @fi: route FIB info structure
1139 * @nlflags: netlink flags passed in (NLM_F_*)
1140 * @tb_id: route table ID
1142 * Add/modify switch IPv4 route entry.
1144 int switchdev_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi,
1145 u8 tos, u8 type, u32 nlflags, u32 tb_id)
1147 struct switchdev_obj_ipv4_fib ipv4_fib = {
1148 .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB,
1156 struct net_device *dev;
1159 memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi));
1161 /* Don't offload route if using custom ip rules or if
1162 * IPv4 FIB offloading has been disabled completely.
1165 #ifdef CONFIG_IP_MULTIPLE_TABLES
1166 if (fi->fib_net->ipv4.fib_has_custom_rules)
1170 if (fi->fib_net->ipv4.fib_offload_disabled)
1173 dev = switchdev_get_dev_by_nhs(fi);
1177 err = switchdev_port_obj_add(dev, &ipv4_fib.obj);
1179 fi->fib_flags |= RTNH_F_OFFLOAD;
1181 return err == -EOPNOTSUPP ? 0 : err;
1183 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add);
1186 * switchdev_fib_ipv4_del - Delete IPv4 route entry from switch
1188 * @dst: route's IPv4 destination address
1189 * @dst_len: destination address length (prefix length)
1190 * @fi: route FIB info structure
1193 * @tb_id: route table ID
1195 * Delete IPv4 route entry from switch device.
1197 int switchdev_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi,
1198 u8 tos, u8 type, u32 tb_id)
1200 struct switchdev_obj_ipv4_fib ipv4_fib = {
1201 .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB,
1209 struct net_device *dev;
1212 memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi));
1214 if (!(fi->fib_flags & RTNH_F_OFFLOAD))
1217 dev = switchdev_get_dev_by_nhs(fi);
1221 err = switchdev_port_obj_del(dev, &ipv4_fib.obj);
1223 fi->fib_flags &= ~RTNH_F_OFFLOAD;
1225 return err == -EOPNOTSUPP ? 0 : err;
1227 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del);
1230 * switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation
1232 * @fi: route FIB info structure
1234 void switchdev_fib_ipv4_abort(struct fib_info *fi)
1236 /* There was a problem installing this route to the offload
1237 * device. For now, until we come up with more refined
1238 * policy handling, abruptly end IPv4 fib offloading for
1239 * for entire net by flushing offload device(s) of all
1240 * IPv4 routes, and mark IPv4 fib offloading broken from
1241 * this point forward.
1244 fib_flush_external(fi->fib_net);
1245 fi->fib_net->ipv4.fib_offload_disabled = true;
1247 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort);
1249 static bool switchdev_port_same_parent_id(struct net_device *a,
1250 struct net_device *b)
1252 struct switchdev_attr a_attr = {
1253 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1254 .flags = SWITCHDEV_F_NO_RECURSE,
1256 struct switchdev_attr b_attr = {
1257 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1258 .flags = SWITCHDEV_F_NO_RECURSE,
1261 if (switchdev_port_attr_get(a, &a_attr) ||
1262 switchdev_port_attr_get(b, &b_attr))
1265 return netdev_phys_item_id_same(&a_attr.u.ppid, &b_attr.u.ppid);
1268 static u32 switchdev_port_fwd_mark_get(struct net_device *dev,
1269 struct net_device *group_dev)
1271 struct net_device *lower_dev;
1272 struct list_head *iter;
1274 netdev_for_each_lower_dev(group_dev, lower_dev, iter) {
1275 if (lower_dev == dev)
1277 if (switchdev_port_same_parent_id(dev, lower_dev))
1278 return lower_dev->offload_fwd_mark;
1279 return switchdev_port_fwd_mark_get(dev, lower_dev);
1282 return dev->ifindex;
1285 static void switchdev_port_fwd_mark_reset(struct net_device *group_dev,
1286 u32 old_mark, u32 *reset_mark)
1288 struct net_device *lower_dev;
1289 struct list_head *iter;
1291 netdev_for_each_lower_dev(group_dev, lower_dev, iter) {
1292 if (lower_dev->offload_fwd_mark == old_mark) {
1294 *reset_mark = lower_dev->ifindex;
1295 lower_dev->offload_fwd_mark = *reset_mark;
1297 switchdev_port_fwd_mark_reset(lower_dev, old_mark, reset_mark);
1302 * switchdev_port_fwd_mark_set - Set port offload forwarding mark
1305 * @group_dev: containing device
1306 * @joining: true if dev is joining group; false if leaving group
1308 * An ungrouped port's offload mark is just its ifindex. A grouped
1309 * port's (member of a bridge, for example) offload mark is the ifindex
1310 * of one of the ports in the group with the same parent (switch) ID.
1311 * Ports on the same device in the same group will have the same mark.
1316 * sw1p1 ifindex=2 mark=2
1317 * sw1p2 ifindex=3 mark=2
1318 * sw2p1 ifindex=4 mark=5
1319 * sw2p2 ifindex=5 mark=5
1321 * If sw2p2 leaves the bridge, we'll have:
1324 * sw1p1 ifindex=2 mark=2
1325 * sw1p2 ifindex=3 mark=2
1326 * sw2p1 ifindex=4 mark=4
1327 * sw2p2 ifindex=5 mark=5
1329 void switchdev_port_fwd_mark_set(struct net_device *dev,
1330 struct net_device *group_dev,
1333 u32 mark = dev->ifindex;
1339 mark = switchdev_port_fwd_mark_get(dev, group_dev);
1340 else if (dev->offload_fwd_mark == mark)
1341 /* Ohoh, this port was the mark reference port,
1342 * but it's leaving the group, so reset the
1343 * mark for the remaining ports in the group.
1345 switchdev_port_fwd_mark_reset(group_dev, mark,
1349 dev->offload_fwd_mark = mark;
1351 EXPORT_SYMBOL_GPL(switchdev_port_fwd_mark_set);