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 <linux/if_vlan.h>
23 #include <linux/rtnetlink.h>
24 #include <net/ip_fib.h>
25 #include <net/switchdev.h>
28 * switchdev_trans_item_enqueue - Enqueue data item to transaction queue
31 * @data: pointer to data being queued
32 * @destructor: data destructor
33 * @tritem: transaction item being queued
35 * Enqeueue data item to transaction queue. tritem is typically placed in
36 * cointainter pointed at by data pointer. Destructor is called on
37 * transaction abort and after successful commit phase in case
38 * the caller did not dequeue the item before.
40 void switchdev_trans_item_enqueue(struct switchdev_trans *trans,
41 void *data, void (*destructor)(void const *),
42 struct switchdev_trans_item *tritem)
45 tritem->destructor = destructor;
46 list_add_tail(&tritem->list, &trans->item_list);
48 EXPORT_SYMBOL_GPL(switchdev_trans_item_enqueue);
50 static struct switchdev_trans_item *
51 __switchdev_trans_item_dequeue(struct switchdev_trans *trans)
53 struct switchdev_trans_item *tritem;
55 if (list_empty(&trans->item_list))
57 tritem = list_first_entry(&trans->item_list,
58 struct switchdev_trans_item, list);
59 list_del(&tritem->list);
64 * switchdev_trans_item_dequeue - Dequeue data item from transaction queue
68 void *switchdev_trans_item_dequeue(struct switchdev_trans *trans)
70 struct switchdev_trans_item *tritem;
72 tritem = __switchdev_trans_item_dequeue(trans);
76 EXPORT_SYMBOL_GPL(switchdev_trans_item_dequeue);
78 static void switchdev_trans_init(struct switchdev_trans *trans)
80 INIT_LIST_HEAD(&trans->item_list);
83 static void switchdev_trans_items_destroy(struct switchdev_trans *trans)
85 struct switchdev_trans_item *tritem;
87 while ((tritem = __switchdev_trans_item_dequeue(trans)))
88 tritem->destructor(tritem->data);
91 static void switchdev_trans_items_warn_destroy(struct net_device *dev,
92 struct switchdev_trans *trans)
94 WARN(!list_empty(&trans->item_list), "%s: transaction item queue is not empty.\n",
96 switchdev_trans_items_destroy(trans);
99 static LIST_HEAD(deferred);
100 static DEFINE_SPINLOCK(deferred_lock);
102 typedef void switchdev_deferred_func_t(struct net_device *dev,
105 struct switchdev_deferred_item {
106 struct list_head list;
107 struct net_device *dev;
108 switchdev_deferred_func_t *func;
109 unsigned long data[0];
112 static struct switchdev_deferred_item *switchdev_deferred_dequeue(void)
114 struct switchdev_deferred_item *dfitem;
116 spin_lock_bh(&deferred_lock);
117 if (list_empty(&deferred)) {
121 dfitem = list_first_entry(&deferred,
122 struct switchdev_deferred_item, list);
123 list_del(&dfitem->list);
125 spin_unlock_bh(&deferred_lock);
130 * switchdev_deferred_process - Process ops in deferred queue
132 * Called to flush the ops currently queued in deferred ops queue.
133 * rtnl_lock must be held.
135 void switchdev_deferred_process(void)
137 struct switchdev_deferred_item *dfitem;
141 while ((dfitem = switchdev_deferred_dequeue())) {
142 dfitem->func(dfitem->dev, dfitem->data);
143 dev_put(dfitem->dev);
147 EXPORT_SYMBOL_GPL(switchdev_deferred_process);
149 static void switchdev_deferred_process_work(struct work_struct *work)
152 switchdev_deferred_process();
156 static DECLARE_WORK(deferred_process_work, switchdev_deferred_process_work);
158 static int switchdev_deferred_enqueue(struct net_device *dev,
159 const void *data, size_t data_len,
160 switchdev_deferred_func_t *func)
162 struct switchdev_deferred_item *dfitem;
164 dfitem = kmalloc(sizeof(*dfitem) + data_len, GFP_ATOMIC);
169 memcpy(dfitem->data, data, data_len);
171 spin_lock_bh(&deferred_lock);
172 list_add_tail(&dfitem->list, &deferred);
173 spin_unlock_bh(&deferred_lock);
174 schedule_work(&deferred_process_work);
179 * switchdev_port_attr_get - Get port attribute
182 * @attr: attribute to get
184 int switchdev_port_attr_get(struct net_device *dev, struct switchdev_attr *attr)
186 const struct switchdev_ops *ops = dev->switchdev_ops;
187 struct net_device *lower_dev;
188 struct list_head *iter;
189 struct switchdev_attr first = {
190 .id = SWITCHDEV_ATTR_ID_UNDEFINED
192 int err = -EOPNOTSUPP;
194 if (ops && ops->switchdev_port_attr_get)
195 return ops->switchdev_port_attr_get(dev, attr);
197 if (attr->flags & SWITCHDEV_F_NO_RECURSE)
200 /* Switch device port(s) may be stacked under
201 * bond/team/vlan dev, so recurse down to get attr on
202 * each port. Return -ENODATA if attr values don't
203 * compare across ports.
206 netdev_for_each_lower_dev(dev, lower_dev, iter) {
207 err = switchdev_port_attr_get(lower_dev, attr);
210 if (first.id == SWITCHDEV_ATTR_ID_UNDEFINED)
212 else if (memcmp(&first, attr, sizeof(*attr)))
218 EXPORT_SYMBOL_GPL(switchdev_port_attr_get);
220 static int __switchdev_port_attr_set(struct net_device *dev,
221 const struct switchdev_attr *attr,
222 struct switchdev_trans *trans)
224 const struct switchdev_ops *ops = dev->switchdev_ops;
225 struct net_device *lower_dev;
226 struct list_head *iter;
227 int err = -EOPNOTSUPP;
229 if (ops && ops->switchdev_port_attr_set) {
230 err = ops->switchdev_port_attr_set(dev, attr, trans);
234 if (attr->flags & SWITCHDEV_F_NO_RECURSE)
237 /* Switch device port(s) may be stacked under
238 * bond/team/vlan dev, so recurse down to set attr on
242 netdev_for_each_lower_dev(dev, lower_dev, iter) {
243 err = __switchdev_port_attr_set(lower_dev, attr, trans);
249 if (err == -EOPNOTSUPP && attr->flags & SWITCHDEV_F_SKIP_EOPNOTSUPP)
255 static int switchdev_port_attr_set_now(struct net_device *dev,
256 const struct switchdev_attr *attr)
258 struct switchdev_trans trans;
261 switchdev_trans_init(&trans);
263 /* Phase I: prepare for attr set. Driver/device should fail
264 * here if there are going to be issues in the commit phase,
265 * such as lack of resources or support. The driver/device
266 * should reserve resources needed for the commit phase here,
267 * but should not commit the attr.
270 trans.ph_prepare = true;
271 err = __switchdev_port_attr_set(dev, attr, &trans);
273 /* Prepare phase failed: abort the transaction. Any
274 * resources reserved in the prepare phase are
278 if (err != -EOPNOTSUPP)
279 switchdev_trans_items_destroy(&trans);
284 /* Phase II: commit attr set. This cannot fail as a fault
285 * of driver/device. If it does, it's a bug in the driver/device
286 * because the driver said everythings was OK in phase I.
289 trans.ph_prepare = false;
290 err = __switchdev_port_attr_set(dev, attr, &trans);
291 WARN(err, "%s: Commit of attribute (id=%d) failed.\n",
292 dev->name, attr->id);
293 switchdev_trans_items_warn_destroy(dev, &trans);
298 static void switchdev_port_attr_set_deferred(struct net_device *dev,
301 const struct switchdev_attr *attr = data;
304 err = switchdev_port_attr_set_now(dev, attr);
305 if (err && err != -EOPNOTSUPP)
306 netdev_err(dev, "failed (err=%d) to set attribute (id=%d)\n",
309 attr->complete(dev, err, attr->complete_priv);
312 static int switchdev_port_attr_set_defer(struct net_device *dev,
313 const struct switchdev_attr *attr)
315 return switchdev_deferred_enqueue(dev, attr, sizeof(*attr),
316 switchdev_port_attr_set_deferred);
320 * switchdev_port_attr_set - Set port attribute
323 * @attr: attribute to set
325 * Use a 2-phase prepare-commit transaction model to ensure
326 * system is not left in a partially updated state due to
327 * failure from driver/device.
329 * rtnl_lock must be held and must not be in atomic section,
330 * in case SWITCHDEV_F_DEFER flag is not set.
332 int switchdev_port_attr_set(struct net_device *dev,
333 const struct switchdev_attr *attr)
335 if (attr->flags & SWITCHDEV_F_DEFER)
336 return switchdev_port_attr_set_defer(dev, attr);
338 return switchdev_port_attr_set_now(dev, attr);
340 EXPORT_SYMBOL_GPL(switchdev_port_attr_set);
342 static size_t switchdev_obj_size(const struct switchdev_obj *obj)
345 case SWITCHDEV_OBJ_ID_PORT_VLAN:
346 return sizeof(struct switchdev_obj_port_vlan);
347 case SWITCHDEV_OBJ_ID_IPV4_FIB:
348 return sizeof(struct switchdev_obj_ipv4_fib);
349 case SWITCHDEV_OBJ_ID_PORT_FDB:
350 return sizeof(struct switchdev_obj_port_fdb);
351 case SWITCHDEV_OBJ_ID_PORT_MDB:
352 return sizeof(struct switchdev_obj_port_mdb);
359 static int __switchdev_port_obj_add(struct net_device *dev,
360 const struct switchdev_obj *obj,
361 struct switchdev_trans *trans)
363 const struct switchdev_ops *ops = dev->switchdev_ops;
364 struct net_device *lower_dev;
365 struct list_head *iter;
366 int err = -EOPNOTSUPP;
368 if (ops && ops->switchdev_port_obj_add)
369 return ops->switchdev_port_obj_add(dev, obj, trans);
371 /* Switch device port(s) may be stacked under
372 * bond/team/vlan dev, so recurse down to add object on
376 netdev_for_each_lower_dev(dev, lower_dev, iter) {
377 err = __switchdev_port_obj_add(lower_dev, obj, trans);
385 static int switchdev_port_obj_add_now(struct net_device *dev,
386 const struct switchdev_obj *obj)
388 struct switchdev_trans trans;
393 switchdev_trans_init(&trans);
395 /* Phase I: prepare for obj add. Driver/device should fail
396 * here if there are going to be issues in the commit phase,
397 * such as lack of resources or support. The driver/device
398 * should reserve resources needed for the commit phase here,
399 * but should not commit the obj.
402 trans.ph_prepare = true;
403 err = __switchdev_port_obj_add(dev, obj, &trans);
405 /* Prepare phase failed: abort the transaction. Any
406 * resources reserved in the prepare phase are
410 if (err != -EOPNOTSUPP)
411 switchdev_trans_items_destroy(&trans);
416 /* Phase II: commit obj add. This cannot fail as a fault
417 * of driver/device. If it does, it's a bug in the driver/device
418 * because the driver said everythings was OK in phase I.
421 trans.ph_prepare = false;
422 err = __switchdev_port_obj_add(dev, obj, &trans);
423 WARN(err, "%s: Commit of object (id=%d) failed.\n", dev->name, obj->id);
424 switchdev_trans_items_warn_destroy(dev, &trans);
429 static void switchdev_port_obj_add_deferred(struct net_device *dev,
432 const struct switchdev_obj *obj = data;
435 err = switchdev_port_obj_add_now(dev, obj);
436 if (err && err != -EOPNOTSUPP)
437 netdev_err(dev, "failed (err=%d) to add object (id=%d)\n",
440 obj->complete(dev, err, obj->complete_priv);
443 static int switchdev_port_obj_add_defer(struct net_device *dev,
444 const struct switchdev_obj *obj)
446 return switchdev_deferred_enqueue(dev, obj, switchdev_obj_size(obj),
447 switchdev_port_obj_add_deferred);
451 * switchdev_port_obj_add - Add port object
455 * @obj: object to add
457 * Use a 2-phase prepare-commit transaction model to ensure
458 * system is not left in a partially updated state due to
459 * failure from driver/device.
461 * rtnl_lock must be held and must not be in atomic section,
462 * in case SWITCHDEV_F_DEFER flag is not set.
464 int switchdev_port_obj_add(struct net_device *dev,
465 const struct switchdev_obj *obj)
467 if (obj->flags & SWITCHDEV_F_DEFER)
468 return switchdev_port_obj_add_defer(dev, obj);
470 return switchdev_port_obj_add_now(dev, obj);
472 EXPORT_SYMBOL_GPL(switchdev_port_obj_add);
474 static int switchdev_port_obj_del_now(struct net_device *dev,
475 const struct switchdev_obj *obj)
477 const struct switchdev_ops *ops = dev->switchdev_ops;
478 struct net_device *lower_dev;
479 struct list_head *iter;
480 int err = -EOPNOTSUPP;
482 if (ops && ops->switchdev_port_obj_del)
483 return ops->switchdev_port_obj_del(dev, obj);
485 /* Switch device port(s) may be stacked under
486 * bond/team/vlan dev, so recurse down to delete object on
490 netdev_for_each_lower_dev(dev, lower_dev, iter) {
491 err = switchdev_port_obj_del_now(lower_dev, obj);
499 static void switchdev_port_obj_del_deferred(struct net_device *dev,
502 const struct switchdev_obj *obj = data;
505 err = switchdev_port_obj_del_now(dev, obj);
506 if (err && err != -EOPNOTSUPP)
507 netdev_err(dev, "failed (err=%d) to del object (id=%d)\n",
510 obj->complete(dev, err, obj->complete_priv);
513 static int switchdev_port_obj_del_defer(struct net_device *dev,
514 const struct switchdev_obj *obj)
516 return switchdev_deferred_enqueue(dev, obj, switchdev_obj_size(obj),
517 switchdev_port_obj_del_deferred);
521 * switchdev_port_obj_del - Delete port object
525 * @obj: object to delete
527 * rtnl_lock must be held and must not be in atomic section,
528 * in case SWITCHDEV_F_DEFER flag is not set.
530 int switchdev_port_obj_del(struct net_device *dev,
531 const struct switchdev_obj *obj)
533 if (obj->flags & SWITCHDEV_F_DEFER)
534 return switchdev_port_obj_del_defer(dev, obj);
536 return switchdev_port_obj_del_now(dev, obj);
538 EXPORT_SYMBOL_GPL(switchdev_port_obj_del);
541 * switchdev_port_obj_dump - Dump port objects
545 * @obj: object to dump
546 * @cb: function to call with a filled object
548 * rtnl_lock must be held.
550 int switchdev_port_obj_dump(struct net_device *dev, struct switchdev_obj *obj,
551 switchdev_obj_dump_cb_t *cb)
553 const struct switchdev_ops *ops = dev->switchdev_ops;
554 struct net_device *lower_dev;
555 struct list_head *iter;
556 int err = -EOPNOTSUPP;
560 if (ops && ops->switchdev_port_obj_dump)
561 return ops->switchdev_port_obj_dump(dev, obj, cb);
563 /* Switch device port(s) may be stacked under
564 * bond/team/vlan dev, so recurse down to dump objects on
565 * first port at bottom of stack.
568 netdev_for_each_lower_dev(dev, lower_dev, iter) {
569 err = switchdev_port_obj_dump(lower_dev, obj, cb);
575 EXPORT_SYMBOL_GPL(switchdev_port_obj_dump);
577 static RAW_NOTIFIER_HEAD(switchdev_notif_chain);
580 * register_switchdev_notifier - Register notifier
581 * @nb: notifier_block
583 * Register switch device notifier. This should be used by code
584 * which needs to monitor events happening in particular device.
585 * Return values are same as for atomic_notifier_chain_register().
587 int register_switchdev_notifier(struct notifier_block *nb)
592 err = raw_notifier_chain_register(&switchdev_notif_chain, nb);
596 EXPORT_SYMBOL_GPL(register_switchdev_notifier);
599 * unregister_switchdev_notifier - Unregister notifier
600 * @nb: notifier_block
602 * Unregister switch device notifier.
603 * Return values are same as for atomic_notifier_chain_unregister().
605 int unregister_switchdev_notifier(struct notifier_block *nb)
610 err = raw_notifier_chain_unregister(&switchdev_notif_chain, nb);
614 EXPORT_SYMBOL_GPL(unregister_switchdev_notifier);
617 * call_switchdev_notifiers - Call notifiers
618 * @val: value passed unmodified to notifier function
620 * @info: notifier information data
622 * Call all network notifier blocks. This should be called by driver
623 * when it needs to propagate hardware event.
624 * Return values are same as for atomic_notifier_call_chain().
625 * rtnl_lock must be held.
627 int call_switchdev_notifiers(unsigned long val, struct net_device *dev,
628 struct switchdev_notifier_info *info)
635 err = raw_notifier_call_chain(&switchdev_notif_chain, val, info);
638 EXPORT_SYMBOL_GPL(call_switchdev_notifiers);
640 struct switchdev_vlan_dump {
641 struct switchdev_obj_port_vlan vlan;
649 static int switchdev_port_vlan_dump_put(struct switchdev_vlan_dump *dump)
651 struct bridge_vlan_info vinfo;
653 vinfo.flags = dump->flags;
655 if (dump->begin == 0 && dump->end == 0) {
657 } else if (dump->begin == dump->end) {
658 vinfo.vid = dump->begin;
659 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
660 sizeof(vinfo), &vinfo))
663 vinfo.vid = dump->begin;
664 vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_BEGIN;
665 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
666 sizeof(vinfo), &vinfo))
668 vinfo.vid = dump->end;
669 vinfo.flags &= ~BRIDGE_VLAN_INFO_RANGE_BEGIN;
670 vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_END;
671 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
672 sizeof(vinfo), &vinfo))
679 static int switchdev_port_vlan_dump_cb(struct switchdev_obj *obj)
681 struct switchdev_obj_port_vlan *vlan = SWITCHDEV_OBJ_PORT_VLAN(obj);
682 struct switchdev_vlan_dump *dump =
683 container_of(vlan, struct switchdev_vlan_dump, vlan);
686 if (vlan->vid_begin > vlan->vid_end)
689 if (dump->filter_mask & RTEXT_FILTER_BRVLAN) {
690 dump->flags = vlan->flags;
691 for (dump->begin = dump->end = vlan->vid_begin;
692 dump->begin <= vlan->vid_end;
693 dump->begin++, dump->end++) {
694 err = switchdev_port_vlan_dump_put(dump);
698 } else if (dump->filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED) {
699 if (dump->begin > vlan->vid_begin &&
700 dump->begin >= vlan->vid_end) {
701 if ((dump->begin - 1) == vlan->vid_end &&
702 dump->flags == vlan->flags) {
704 dump->begin = vlan->vid_begin;
706 err = switchdev_port_vlan_dump_put(dump);
707 dump->flags = vlan->flags;
708 dump->begin = vlan->vid_begin;
709 dump->end = vlan->vid_end;
711 } else if (dump->end <= vlan->vid_begin &&
712 dump->end < vlan->vid_end) {
713 if ((dump->end + 1) == vlan->vid_begin &&
714 dump->flags == vlan->flags) {
716 dump->end = vlan->vid_end;
718 err = switchdev_port_vlan_dump_put(dump);
719 dump->flags = vlan->flags;
720 dump->begin = vlan->vid_begin;
721 dump->end = vlan->vid_end;
731 static int switchdev_port_vlan_fill(struct sk_buff *skb, struct net_device *dev,
734 struct switchdev_vlan_dump dump = {
735 .vlan.obj.orig_dev = dev,
736 .vlan.obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN,
738 .filter_mask = filter_mask,
742 if ((filter_mask & RTEXT_FILTER_BRVLAN) ||
743 (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)) {
744 err = switchdev_port_obj_dump(dev, &dump.vlan.obj,
745 switchdev_port_vlan_dump_cb);
748 if (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)
750 err = switchdev_port_vlan_dump_put(&dump);
754 return err == -EOPNOTSUPP ? 0 : err;
758 * switchdev_port_bridge_getlink - Get bridge port attributes
762 * Called for SELF on rtnl_bridge_getlink to get bridge port
765 int switchdev_port_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
766 struct net_device *dev, u32 filter_mask,
769 struct switchdev_attr attr = {
771 .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS,
773 u16 mode = BRIDGE_MODE_UNDEF;
774 u32 mask = BR_LEARNING | BR_LEARNING_SYNC | BR_FLOOD;
777 err = switchdev_port_attr_get(dev, &attr);
778 if (err && err != -EOPNOTSUPP)
781 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode,
782 attr.u.brport_flags, mask, nlflags,
783 filter_mask, switchdev_port_vlan_fill);
785 EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink);
787 static int switchdev_port_br_setflag(struct net_device *dev,
788 struct nlattr *nlattr,
789 unsigned long brport_flag)
791 struct switchdev_attr attr = {
793 .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS,
795 u8 flag = nla_get_u8(nlattr);
798 err = switchdev_port_attr_get(dev, &attr);
803 attr.u.brport_flags |= brport_flag;
805 attr.u.brport_flags &= ~brport_flag;
807 return switchdev_port_attr_set(dev, &attr);
810 static const struct nla_policy
811 switchdev_port_bridge_policy[IFLA_BRPORT_MAX + 1] = {
812 [IFLA_BRPORT_STATE] = { .type = NLA_U8 },
813 [IFLA_BRPORT_COST] = { .type = NLA_U32 },
814 [IFLA_BRPORT_PRIORITY] = { .type = NLA_U16 },
815 [IFLA_BRPORT_MODE] = { .type = NLA_U8 },
816 [IFLA_BRPORT_GUARD] = { .type = NLA_U8 },
817 [IFLA_BRPORT_PROTECT] = { .type = NLA_U8 },
818 [IFLA_BRPORT_FAST_LEAVE] = { .type = NLA_U8 },
819 [IFLA_BRPORT_LEARNING] = { .type = NLA_U8 },
820 [IFLA_BRPORT_LEARNING_SYNC] = { .type = NLA_U8 },
821 [IFLA_BRPORT_UNICAST_FLOOD] = { .type = NLA_U8 },
824 static int switchdev_port_br_setlink_protinfo(struct net_device *dev,
825 struct nlattr *protinfo)
831 err = nla_validate_nested(protinfo, IFLA_BRPORT_MAX,
832 switchdev_port_bridge_policy);
836 nla_for_each_nested(attr, protinfo, rem) {
837 switch (nla_type(attr)) {
838 case IFLA_BRPORT_LEARNING:
839 err = switchdev_port_br_setflag(dev, attr,
842 case IFLA_BRPORT_LEARNING_SYNC:
843 err = switchdev_port_br_setflag(dev, attr,
846 case IFLA_BRPORT_UNICAST_FLOOD:
847 err = switchdev_port_br_setflag(dev, attr, BR_FLOOD);
860 static int switchdev_port_br_afspec(struct net_device *dev,
861 struct nlattr *afspec,
862 int (*f)(struct net_device *dev,
863 const struct switchdev_obj *obj))
866 struct bridge_vlan_info *vinfo;
867 struct switchdev_obj_port_vlan vlan = {
869 .obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN,
874 nla_for_each_nested(attr, afspec, rem) {
875 if (nla_type(attr) != IFLA_BRIDGE_VLAN_INFO)
877 if (nla_len(attr) != sizeof(struct bridge_vlan_info))
879 vinfo = nla_data(attr);
880 if (!vinfo->vid || vinfo->vid >= VLAN_VID_MASK)
882 vlan.flags = vinfo->flags;
883 if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_BEGIN) {
886 vlan.vid_begin = vinfo->vid;
887 /* don't allow range of pvids */
888 if (vlan.flags & BRIDGE_VLAN_INFO_PVID)
890 } else if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_END) {
893 vlan.vid_end = vinfo->vid;
894 if (vlan.vid_end <= vlan.vid_begin)
896 err = f(dev, &vlan.obj);
903 vlan.vid_begin = vinfo->vid;
904 vlan.vid_end = vinfo->vid;
905 err = f(dev, &vlan.obj);
916 * switchdev_port_bridge_setlink - Set bridge port attributes
919 * @nlh: netlink header
920 * @flags: netlink flags
922 * Called for SELF on rtnl_bridge_setlink to set bridge port
925 int switchdev_port_bridge_setlink(struct net_device *dev,
926 struct nlmsghdr *nlh, u16 flags)
928 struct nlattr *protinfo;
929 struct nlattr *afspec;
932 protinfo = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
935 err = switchdev_port_br_setlink_protinfo(dev, protinfo);
940 afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
943 err = switchdev_port_br_afspec(dev, afspec,
944 switchdev_port_obj_add);
948 EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink);
951 * switchdev_port_bridge_dellink - Set bridge port attributes
954 * @nlh: netlink header
955 * @flags: netlink flags
957 * Called for SELF on rtnl_bridge_dellink to set bridge port
960 int switchdev_port_bridge_dellink(struct net_device *dev,
961 struct nlmsghdr *nlh, u16 flags)
963 struct nlattr *afspec;
965 afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
968 return switchdev_port_br_afspec(dev, afspec,
969 switchdev_port_obj_del);
973 EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink);
976 * switchdev_port_fdb_add - Add FDB (MAC/VLAN) entry to port
978 * @ndmsg: netlink hdr
979 * @nlattr: netlink attributes
981 * @addr: MAC address to add
984 * Add FDB entry to switch device.
986 int switchdev_port_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
987 struct net_device *dev, const unsigned char *addr,
988 u16 vid, u16 nlm_flags)
990 struct switchdev_obj_port_fdb fdb = {
992 .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
996 ether_addr_copy(fdb.addr, addr);
997 return switchdev_port_obj_add(dev, &fdb.obj);
999 EXPORT_SYMBOL_GPL(switchdev_port_fdb_add);
1002 * switchdev_port_fdb_del - Delete FDB (MAC/VLAN) entry from port
1004 * @ndmsg: netlink hdr
1005 * @nlattr: netlink attributes
1007 * @addr: MAC address to delete
1008 * @vid: VLAN to delete
1010 * Delete FDB entry from switch device.
1012 int switchdev_port_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1013 struct net_device *dev, const unsigned char *addr,
1016 struct switchdev_obj_port_fdb fdb = {
1017 .obj.orig_dev = dev,
1018 .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
1022 ether_addr_copy(fdb.addr, addr);
1023 return switchdev_port_obj_del(dev, &fdb.obj);
1025 EXPORT_SYMBOL_GPL(switchdev_port_fdb_del);
1027 struct switchdev_fdb_dump {
1028 struct switchdev_obj_port_fdb fdb;
1029 struct net_device *dev;
1030 struct sk_buff *skb;
1031 struct netlink_callback *cb;
1035 static int switchdev_port_fdb_dump_cb(struct switchdev_obj *obj)
1037 struct switchdev_obj_port_fdb *fdb = SWITCHDEV_OBJ_PORT_FDB(obj);
1038 struct switchdev_fdb_dump *dump =
1039 container_of(fdb, struct switchdev_fdb_dump, fdb);
1040 u32 portid = NETLINK_CB(dump->cb->skb).portid;
1041 u32 seq = dump->cb->nlh->nlmsg_seq;
1042 struct nlmsghdr *nlh;
1045 if (dump->idx < dump->cb->args[0])
1048 nlh = nlmsg_put(dump->skb, portid, seq, RTM_NEWNEIGH,
1049 sizeof(*ndm), NLM_F_MULTI);
1053 ndm = nlmsg_data(nlh);
1054 ndm->ndm_family = AF_BRIDGE;
1057 ndm->ndm_flags = NTF_SELF;
1059 ndm->ndm_ifindex = dump->dev->ifindex;
1060 ndm->ndm_state = fdb->ndm_state;
1062 if (nla_put(dump->skb, NDA_LLADDR, ETH_ALEN, fdb->addr))
1063 goto nla_put_failure;
1065 if (fdb->vid && nla_put_u16(dump->skb, NDA_VLAN, fdb->vid))
1066 goto nla_put_failure;
1068 nlmsg_end(dump->skb, nlh);
1075 nlmsg_cancel(dump->skb, nlh);
1080 * switchdev_port_fdb_dump - Dump port FDB (MAC/VLAN) entries
1083 * @cb: netlink callback
1085 * @filter_dev: filter device
1088 * Dump FDB entries from switch device.
1090 int switchdev_port_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1091 struct net_device *dev,
1092 struct net_device *filter_dev, int idx)
1094 struct switchdev_fdb_dump dump = {
1095 .fdb.obj.orig_dev = dev,
1096 .fdb.obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
1104 err = switchdev_port_obj_dump(dev, &dump.fdb.obj,
1105 switchdev_port_fdb_dump_cb);
1109 EXPORT_SYMBOL_GPL(switchdev_port_fdb_dump);
1111 static struct net_device *switchdev_get_lowest_dev(struct net_device *dev)
1113 const struct switchdev_ops *ops = dev->switchdev_ops;
1114 struct net_device *lower_dev;
1115 struct net_device *port_dev;
1116 struct list_head *iter;
1118 /* Recusively search down until we find a sw port dev.
1119 * (A sw port dev supports switchdev_port_attr_get).
1122 if (ops && ops->switchdev_port_attr_get)
1125 netdev_for_each_lower_dev(dev, lower_dev, iter) {
1126 port_dev = switchdev_get_lowest_dev(lower_dev);
1134 static struct net_device *switchdev_get_dev_by_nhs(struct fib_info *fi)
1136 struct switchdev_attr attr = {
1137 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1139 struct switchdev_attr prev_attr;
1140 struct net_device *dev = NULL;
1145 /* For this route, all nexthop devs must be on the same switch. */
1147 for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
1148 const struct fib_nh *nh = &fi->fib_nh[nhsel];
1153 dev = switchdev_get_lowest_dev(nh->nh_dev);
1157 attr.orig_dev = dev;
1158 if (switchdev_port_attr_get(dev, &attr))
1162 !netdev_phys_item_id_same(&prev_attr.u.ppid, &attr.u.ppid))
1172 * switchdev_fib_ipv4_add - Add/modify switch IPv4 route entry
1174 * @dst: route's IPv4 destination address
1175 * @dst_len: destination address length (prefix length)
1176 * @fi: route FIB info structure
1179 * @nlflags: netlink flags passed in (NLM_F_*)
1180 * @tb_id: route table ID
1182 * Add/modify switch IPv4 route entry.
1184 int switchdev_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi,
1185 u8 tos, u8 type, u32 nlflags, u32 tb_id)
1187 struct switchdev_obj_ipv4_fib ipv4_fib = {
1188 .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB,
1196 struct net_device *dev;
1199 memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi));
1201 /* Don't offload route if using custom ip rules or if
1202 * IPv4 FIB offloading has been disabled completely.
1205 #ifdef CONFIG_IP_MULTIPLE_TABLES
1206 if (fi->fib_net->ipv4.fib_has_custom_rules)
1210 if (fi->fib_net->ipv4.fib_offload_disabled)
1213 dev = switchdev_get_dev_by_nhs(fi);
1217 ipv4_fib.obj.orig_dev = dev;
1218 err = switchdev_port_obj_add(dev, &ipv4_fib.obj);
1220 fi->fib_flags |= RTNH_F_OFFLOAD;
1222 return err == -EOPNOTSUPP ? 0 : err;
1224 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add);
1227 * switchdev_fib_ipv4_del - Delete IPv4 route entry from switch
1229 * @dst: route's IPv4 destination address
1230 * @dst_len: destination address length (prefix length)
1231 * @fi: route FIB info structure
1234 * @tb_id: route table ID
1236 * Delete IPv4 route entry from switch device.
1238 int switchdev_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi,
1239 u8 tos, u8 type, u32 tb_id)
1241 struct switchdev_obj_ipv4_fib ipv4_fib = {
1242 .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB,
1250 struct net_device *dev;
1253 memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi));
1255 if (!(fi->fib_flags & RTNH_F_OFFLOAD))
1258 dev = switchdev_get_dev_by_nhs(fi);
1262 ipv4_fib.obj.orig_dev = dev;
1263 err = switchdev_port_obj_del(dev, &ipv4_fib.obj);
1265 fi->fib_flags &= ~RTNH_F_OFFLOAD;
1267 return err == -EOPNOTSUPP ? 0 : err;
1269 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del);
1272 * switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation
1274 * @fi: route FIB info structure
1276 void switchdev_fib_ipv4_abort(struct fib_info *fi)
1278 /* There was a problem installing this route to the offload
1279 * device. For now, until we come up with more refined
1280 * policy handling, abruptly end IPv4 fib offloading for
1281 * for entire net by flushing offload device(s) of all
1282 * IPv4 routes, and mark IPv4 fib offloading broken from
1283 * this point forward.
1286 fib_flush_external(fi->fib_net);
1287 fi->fib_net->ipv4.fib_offload_disabled = true;
1289 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort);
1291 static bool switchdev_port_same_parent_id(struct net_device *a,
1292 struct net_device *b)
1294 struct switchdev_attr a_attr = {
1296 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1297 .flags = SWITCHDEV_F_NO_RECURSE,
1299 struct switchdev_attr b_attr = {
1301 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1302 .flags = SWITCHDEV_F_NO_RECURSE,
1305 if (switchdev_port_attr_get(a, &a_attr) ||
1306 switchdev_port_attr_get(b, &b_attr))
1309 return netdev_phys_item_id_same(&a_attr.u.ppid, &b_attr.u.ppid);
1312 static u32 switchdev_port_fwd_mark_get(struct net_device *dev,
1313 struct net_device *group_dev)
1315 struct net_device *lower_dev;
1316 struct list_head *iter;
1318 netdev_for_each_lower_dev(group_dev, lower_dev, iter) {
1319 if (lower_dev == dev)
1321 if (switchdev_port_same_parent_id(dev, lower_dev))
1322 return lower_dev->offload_fwd_mark;
1323 return switchdev_port_fwd_mark_get(dev, lower_dev);
1326 return dev->ifindex;
1329 static void switchdev_port_fwd_mark_reset(struct net_device *group_dev,
1330 u32 old_mark, u32 *reset_mark)
1332 struct net_device *lower_dev;
1333 struct list_head *iter;
1335 netdev_for_each_lower_dev(group_dev, lower_dev, iter) {
1336 if (lower_dev->offload_fwd_mark == old_mark) {
1338 *reset_mark = lower_dev->ifindex;
1339 lower_dev->offload_fwd_mark = *reset_mark;
1341 switchdev_port_fwd_mark_reset(lower_dev, old_mark, reset_mark);
1346 * switchdev_port_fwd_mark_set - Set port offload forwarding mark
1349 * @group_dev: containing device
1350 * @joining: true if dev is joining group; false if leaving group
1352 * An ungrouped port's offload mark is just its ifindex. A grouped
1353 * port's (member of a bridge, for example) offload mark is the ifindex
1354 * of one of the ports in the group with the same parent (switch) ID.
1355 * Ports on the same device in the same group will have the same mark.
1360 * sw1p1 ifindex=2 mark=2
1361 * sw1p2 ifindex=3 mark=2
1362 * sw2p1 ifindex=4 mark=5
1363 * sw2p2 ifindex=5 mark=5
1365 * If sw2p2 leaves the bridge, we'll have:
1368 * sw1p1 ifindex=2 mark=2
1369 * sw1p2 ifindex=3 mark=2
1370 * sw2p1 ifindex=4 mark=4
1371 * sw2p2 ifindex=5 mark=5
1373 void switchdev_port_fwd_mark_set(struct net_device *dev,
1374 struct net_device *group_dev,
1377 u32 mark = dev->ifindex;
1383 mark = switchdev_port_fwd_mark_get(dev, group_dev);
1384 else if (dev->offload_fwd_mark == mark)
1385 /* Ohoh, this port was the mark reference port,
1386 * but it's leaving the group, so reset the
1387 * mark for the remaining ports in the group.
1389 switchdev_port_fwd_mark_reset(group_dev, mark,
1393 dev->offload_fwd_mark = mark;
1395 EXPORT_SYMBOL_GPL(switchdev_port_fwd_mark_set);