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",
310 static int switchdev_port_attr_set_defer(struct net_device *dev,
311 const struct switchdev_attr *attr)
313 return switchdev_deferred_enqueue(dev, attr, sizeof(*attr),
314 switchdev_port_attr_set_deferred);
318 * switchdev_port_attr_set - Set port attribute
321 * @attr: attribute to set
323 * Use a 2-phase prepare-commit transaction model to ensure
324 * system is not left in a partially updated state due to
325 * failure from driver/device.
327 * rtnl_lock must be held and must not be in atomic section,
328 * in case SWITCHDEV_F_DEFER flag is not set.
330 int switchdev_port_attr_set(struct net_device *dev,
331 const struct switchdev_attr *attr)
333 if (attr->flags & SWITCHDEV_F_DEFER)
334 return switchdev_port_attr_set_defer(dev, attr);
336 return switchdev_port_attr_set_now(dev, attr);
338 EXPORT_SYMBOL_GPL(switchdev_port_attr_set);
340 static size_t switchdev_obj_size(const struct switchdev_obj *obj)
343 case SWITCHDEV_OBJ_ID_PORT_VLAN:
344 return sizeof(struct switchdev_obj_port_vlan);
345 case SWITCHDEV_OBJ_ID_IPV4_FIB:
346 return sizeof(struct switchdev_obj_ipv4_fib);
347 case SWITCHDEV_OBJ_ID_PORT_FDB:
348 return sizeof(struct switchdev_obj_port_fdb);
349 case SWITCHDEV_OBJ_ID_PORT_MDB:
350 return sizeof(struct switchdev_obj_port_mdb);
357 static int __switchdev_port_obj_add(struct net_device *dev,
358 const struct switchdev_obj *obj,
359 struct switchdev_trans *trans)
361 const struct switchdev_ops *ops = dev->switchdev_ops;
362 struct net_device *lower_dev;
363 struct list_head *iter;
364 int err = -EOPNOTSUPP;
366 if (ops && ops->switchdev_port_obj_add)
367 return ops->switchdev_port_obj_add(dev, obj, trans);
369 /* Switch device port(s) may be stacked under
370 * bond/team/vlan dev, so recurse down to add object on
374 netdev_for_each_lower_dev(dev, lower_dev, iter) {
375 err = __switchdev_port_obj_add(lower_dev, obj, trans);
383 static int switchdev_port_obj_add_now(struct net_device *dev,
384 const struct switchdev_obj *obj)
386 struct switchdev_trans trans;
391 switchdev_trans_init(&trans);
393 /* Phase I: prepare for obj add. Driver/device should fail
394 * here if there are going to be issues in the commit phase,
395 * such as lack of resources or support. The driver/device
396 * should reserve resources needed for the commit phase here,
397 * but should not commit the obj.
400 trans.ph_prepare = true;
401 err = __switchdev_port_obj_add(dev, obj, &trans);
403 /* Prepare phase failed: abort the transaction. Any
404 * resources reserved in the prepare phase are
408 if (err != -EOPNOTSUPP)
409 switchdev_trans_items_destroy(&trans);
414 /* Phase II: commit obj add. This cannot fail as a fault
415 * of driver/device. If it does, it's a bug in the driver/device
416 * because the driver said everythings was OK in phase I.
419 trans.ph_prepare = false;
420 err = __switchdev_port_obj_add(dev, obj, &trans);
421 WARN(err, "%s: Commit of object (id=%d) failed.\n", dev->name, obj->id);
422 switchdev_trans_items_warn_destroy(dev, &trans);
427 static void switchdev_port_obj_add_deferred(struct net_device *dev,
430 const struct switchdev_obj *obj = data;
433 err = switchdev_port_obj_add_now(dev, obj);
434 if (err && err != -EOPNOTSUPP)
435 netdev_err(dev, "failed (err=%d) to add object (id=%d)\n",
439 static int switchdev_port_obj_add_defer(struct net_device *dev,
440 const struct switchdev_obj *obj)
442 return switchdev_deferred_enqueue(dev, obj, switchdev_obj_size(obj),
443 switchdev_port_obj_add_deferred);
447 * switchdev_port_obj_add - Add port object
451 * @obj: object to add
453 * Use a 2-phase prepare-commit transaction model to ensure
454 * system is not left in a partially updated state due to
455 * failure from driver/device.
457 * rtnl_lock must be held and must not be in atomic section,
458 * in case SWITCHDEV_F_DEFER flag is not set.
460 int switchdev_port_obj_add(struct net_device *dev,
461 const struct switchdev_obj *obj)
463 if (obj->flags & SWITCHDEV_F_DEFER)
464 return switchdev_port_obj_add_defer(dev, obj);
466 return switchdev_port_obj_add_now(dev, obj);
468 EXPORT_SYMBOL_GPL(switchdev_port_obj_add);
470 static int switchdev_port_obj_del_now(struct net_device *dev,
471 const struct switchdev_obj *obj)
473 const struct switchdev_ops *ops = dev->switchdev_ops;
474 struct net_device *lower_dev;
475 struct list_head *iter;
476 int err = -EOPNOTSUPP;
478 if (ops && ops->switchdev_port_obj_del)
479 return ops->switchdev_port_obj_del(dev, obj);
481 /* Switch device port(s) may be stacked under
482 * bond/team/vlan dev, so recurse down to delete object on
486 netdev_for_each_lower_dev(dev, lower_dev, iter) {
487 err = switchdev_port_obj_del_now(lower_dev, obj);
495 static void switchdev_port_obj_del_deferred(struct net_device *dev,
498 const struct switchdev_obj *obj = data;
501 err = switchdev_port_obj_del_now(dev, obj);
502 if (err && err != -EOPNOTSUPP)
503 netdev_err(dev, "failed (err=%d) to del object (id=%d)\n",
507 static int switchdev_port_obj_del_defer(struct net_device *dev,
508 const struct switchdev_obj *obj)
510 return switchdev_deferred_enqueue(dev, obj, switchdev_obj_size(obj),
511 switchdev_port_obj_del_deferred);
515 * switchdev_port_obj_del - Delete port object
519 * @obj: object to delete
521 * rtnl_lock must be held and must not be in atomic section,
522 * in case SWITCHDEV_F_DEFER flag is not set.
524 int switchdev_port_obj_del(struct net_device *dev,
525 const struct switchdev_obj *obj)
527 if (obj->flags & SWITCHDEV_F_DEFER)
528 return switchdev_port_obj_del_defer(dev, obj);
530 return switchdev_port_obj_del_now(dev, obj);
532 EXPORT_SYMBOL_GPL(switchdev_port_obj_del);
535 * switchdev_port_obj_dump - Dump port objects
539 * @obj: object to dump
540 * @cb: function to call with a filled object
542 * rtnl_lock must be held.
544 int switchdev_port_obj_dump(struct net_device *dev, struct switchdev_obj *obj,
545 switchdev_obj_dump_cb_t *cb)
547 const struct switchdev_ops *ops = dev->switchdev_ops;
548 struct net_device *lower_dev;
549 struct list_head *iter;
550 int err = -EOPNOTSUPP;
554 if (ops && ops->switchdev_port_obj_dump)
555 return ops->switchdev_port_obj_dump(dev, obj, cb);
557 /* Switch device port(s) may be stacked under
558 * bond/team/vlan dev, so recurse down to dump objects on
559 * first port at bottom of stack.
562 netdev_for_each_lower_dev(dev, lower_dev, iter) {
563 err = switchdev_port_obj_dump(lower_dev, obj, cb);
569 EXPORT_SYMBOL_GPL(switchdev_port_obj_dump);
571 static RAW_NOTIFIER_HEAD(switchdev_notif_chain);
574 * register_switchdev_notifier - Register notifier
575 * @nb: notifier_block
577 * Register switch device notifier. This should be used by code
578 * which needs to monitor events happening in particular device.
579 * Return values are same as for atomic_notifier_chain_register().
581 int register_switchdev_notifier(struct notifier_block *nb)
586 err = raw_notifier_chain_register(&switchdev_notif_chain, nb);
590 EXPORT_SYMBOL_GPL(register_switchdev_notifier);
593 * unregister_switchdev_notifier - Unregister notifier
594 * @nb: notifier_block
596 * Unregister switch device notifier.
597 * Return values are same as for atomic_notifier_chain_unregister().
599 int unregister_switchdev_notifier(struct notifier_block *nb)
604 err = raw_notifier_chain_unregister(&switchdev_notif_chain, nb);
608 EXPORT_SYMBOL_GPL(unregister_switchdev_notifier);
611 * call_switchdev_notifiers - Call notifiers
612 * @val: value passed unmodified to notifier function
614 * @info: notifier information data
616 * Call all network notifier blocks. This should be called by driver
617 * when it needs to propagate hardware event.
618 * Return values are same as for atomic_notifier_call_chain().
619 * rtnl_lock must be held.
621 int call_switchdev_notifiers(unsigned long val, struct net_device *dev,
622 struct switchdev_notifier_info *info)
629 err = raw_notifier_call_chain(&switchdev_notif_chain, val, info);
632 EXPORT_SYMBOL_GPL(call_switchdev_notifiers);
634 struct switchdev_vlan_dump {
635 struct switchdev_obj_port_vlan vlan;
643 static int switchdev_port_vlan_dump_put(struct switchdev_vlan_dump *dump)
645 struct bridge_vlan_info vinfo;
647 vinfo.flags = dump->flags;
649 if (dump->begin == 0 && dump->end == 0) {
651 } else if (dump->begin == dump->end) {
652 vinfo.vid = dump->begin;
653 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
654 sizeof(vinfo), &vinfo))
657 vinfo.vid = dump->begin;
658 vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_BEGIN;
659 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
660 sizeof(vinfo), &vinfo))
662 vinfo.vid = dump->end;
663 vinfo.flags &= ~BRIDGE_VLAN_INFO_RANGE_BEGIN;
664 vinfo.flags |= BRIDGE_VLAN_INFO_RANGE_END;
665 if (nla_put(dump->skb, IFLA_BRIDGE_VLAN_INFO,
666 sizeof(vinfo), &vinfo))
673 static int switchdev_port_vlan_dump_cb(struct switchdev_obj *obj)
675 struct switchdev_obj_port_vlan *vlan = SWITCHDEV_OBJ_PORT_VLAN(obj);
676 struct switchdev_vlan_dump *dump =
677 container_of(vlan, struct switchdev_vlan_dump, vlan);
680 if (vlan->vid_begin > vlan->vid_end)
683 if (dump->filter_mask & RTEXT_FILTER_BRVLAN) {
684 dump->flags = vlan->flags;
685 for (dump->begin = dump->end = vlan->vid_begin;
686 dump->begin <= vlan->vid_end;
687 dump->begin++, dump->end++) {
688 err = switchdev_port_vlan_dump_put(dump);
692 } else if (dump->filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED) {
693 if (dump->begin > vlan->vid_begin &&
694 dump->begin >= vlan->vid_end) {
695 if ((dump->begin - 1) == vlan->vid_end &&
696 dump->flags == vlan->flags) {
698 dump->begin = vlan->vid_begin;
700 err = switchdev_port_vlan_dump_put(dump);
701 dump->flags = vlan->flags;
702 dump->begin = vlan->vid_begin;
703 dump->end = vlan->vid_end;
705 } else if (dump->end <= vlan->vid_begin &&
706 dump->end < vlan->vid_end) {
707 if ((dump->end + 1) == vlan->vid_begin &&
708 dump->flags == vlan->flags) {
710 dump->end = vlan->vid_end;
712 err = switchdev_port_vlan_dump_put(dump);
713 dump->flags = vlan->flags;
714 dump->begin = vlan->vid_begin;
715 dump->end = vlan->vid_end;
725 static int switchdev_port_vlan_fill(struct sk_buff *skb, struct net_device *dev,
728 struct switchdev_vlan_dump dump = {
729 .vlan.obj.orig_dev = dev,
730 .vlan.obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN,
732 .filter_mask = filter_mask,
736 if ((filter_mask & RTEXT_FILTER_BRVLAN) ||
737 (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)) {
738 err = switchdev_port_obj_dump(dev, &dump.vlan.obj,
739 switchdev_port_vlan_dump_cb);
742 if (filter_mask & RTEXT_FILTER_BRVLAN_COMPRESSED)
744 err = switchdev_port_vlan_dump_put(&dump);
748 return err == -EOPNOTSUPP ? 0 : err;
752 * switchdev_port_bridge_getlink - Get bridge port attributes
756 * Called for SELF on rtnl_bridge_getlink to get bridge port
759 int switchdev_port_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
760 struct net_device *dev, u32 filter_mask,
763 struct switchdev_attr attr = {
765 .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS,
767 u16 mode = BRIDGE_MODE_UNDEF;
768 u32 mask = BR_LEARNING | BR_LEARNING_SYNC | BR_FLOOD;
771 err = switchdev_port_attr_get(dev, &attr);
772 if (err && err != -EOPNOTSUPP)
775 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, mode,
776 attr.u.brport_flags, mask, nlflags,
777 filter_mask, switchdev_port_vlan_fill);
779 EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink);
781 static int switchdev_port_br_setflag(struct net_device *dev,
782 struct nlattr *nlattr,
783 unsigned long brport_flag)
785 struct switchdev_attr attr = {
787 .id = SWITCHDEV_ATTR_ID_PORT_BRIDGE_FLAGS,
789 u8 flag = nla_get_u8(nlattr);
792 err = switchdev_port_attr_get(dev, &attr);
797 attr.u.brport_flags |= brport_flag;
799 attr.u.brport_flags &= ~brport_flag;
801 return switchdev_port_attr_set(dev, &attr);
804 static const struct nla_policy
805 switchdev_port_bridge_policy[IFLA_BRPORT_MAX + 1] = {
806 [IFLA_BRPORT_STATE] = { .type = NLA_U8 },
807 [IFLA_BRPORT_COST] = { .type = NLA_U32 },
808 [IFLA_BRPORT_PRIORITY] = { .type = NLA_U16 },
809 [IFLA_BRPORT_MODE] = { .type = NLA_U8 },
810 [IFLA_BRPORT_GUARD] = { .type = NLA_U8 },
811 [IFLA_BRPORT_PROTECT] = { .type = NLA_U8 },
812 [IFLA_BRPORT_FAST_LEAVE] = { .type = NLA_U8 },
813 [IFLA_BRPORT_LEARNING] = { .type = NLA_U8 },
814 [IFLA_BRPORT_LEARNING_SYNC] = { .type = NLA_U8 },
815 [IFLA_BRPORT_UNICAST_FLOOD] = { .type = NLA_U8 },
818 static int switchdev_port_br_setlink_protinfo(struct net_device *dev,
819 struct nlattr *protinfo)
825 err = nla_validate_nested(protinfo, IFLA_BRPORT_MAX,
826 switchdev_port_bridge_policy);
830 nla_for_each_nested(attr, protinfo, rem) {
831 switch (nla_type(attr)) {
832 case IFLA_BRPORT_LEARNING:
833 err = switchdev_port_br_setflag(dev, attr,
836 case IFLA_BRPORT_LEARNING_SYNC:
837 err = switchdev_port_br_setflag(dev, attr,
840 case IFLA_BRPORT_UNICAST_FLOOD:
841 err = switchdev_port_br_setflag(dev, attr, BR_FLOOD);
854 static int switchdev_port_br_afspec(struct net_device *dev,
855 struct nlattr *afspec,
856 int (*f)(struct net_device *dev,
857 const struct switchdev_obj *obj))
860 struct bridge_vlan_info *vinfo;
861 struct switchdev_obj_port_vlan vlan = {
863 .obj.id = SWITCHDEV_OBJ_ID_PORT_VLAN,
868 nla_for_each_nested(attr, afspec, rem) {
869 if (nla_type(attr) != IFLA_BRIDGE_VLAN_INFO)
871 if (nla_len(attr) != sizeof(struct bridge_vlan_info))
873 vinfo = nla_data(attr);
874 if (!vinfo->vid || vinfo->vid >= VLAN_VID_MASK)
876 vlan.flags = vinfo->flags;
877 if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_BEGIN) {
880 vlan.vid_begin = vinfo->vid;
881 /* don't allow range of pvids */
882 if (vlan.flags & BRIDGE_VLAN_INFO_PVID)
884 } else if (vinfo->flags & BRIDGE_VLAN_INFO_RANGE_END) {
887 vlan.vid_end = vinfo->vid;
888 if (vlan.vid_end <= vlan.vid_begin)
890 err = f(dev, &vlan.obj);
897 vlan.vid_begin = vinfo->vid;
898 vlan.vid_end = vinfo->vid;
899 err = f(dev, &vlan.obj);
910 * switchdev_port_bridge_setlink - Set bridge port attributes
913 * @nlh: netlink header
914 * @flags: netlink flags
916 * Called for SELF on rtnl_bridge_setlink to set bridge port
919 int switchdev_port_bridge_setlink(struct net_device *dev,
920 struct nlmsghdr *nlh, u16 flags)
922 struct nlattr *protinfo;
923 struct nlattr *afspec;
926 protinfo = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
929 err = switchdev_port_br_setlink_protinfo(dev, protinfo);
934 afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
937 err = switchdev_port_br_afspec(dev, afspec,
938 switchdev_port_obj_add);
942 EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink);
945 * switchdev_port_bridge_dellink - Set bridge port attributes
948 * @nlh: netlink header
949 * @flags: netlink flags
951 * Called for SELF on rtnl_bridge_dellink to set bridge port
954 int switchdev_port_bridge_dellink(struct net_device *dev,
955 struct nlmsghdr *nlh, u16 flags)
957 struct nlattr *afspec;
959 afspec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg),
962 return switchdev_port_br_afspec(dev, afspec,
963 switchdev_port_obj_del);
967 EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink);
970 * switchdev_port_fdb_add - Add FDB (MAC/VLAN) entry to port
972 * @ndmsg: netlink hdr
973 * @nlattr: netlink attributes
975 * @addr: MAC address to add
978 * Add FDB entry to switch device.
980 int switchdev_port_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
981 struct net_device *dev, const unsigned char *addr,
982 u16 vid, u16 nlm_flags)
984 struct switchdev_obj_port_fdb fdb = {
986 .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
990 ether_addr_copy(fdb.addr, addr);
991 return switchdev_port_obj_add(dev, &fdb.obj);
993 EXPORT_SYMBOL_GPL(switchdev_port_fdb_add);
996 * switchdev_port_fdb_del - Delete FDB (MAC/VLAN) entry from port
998 * @ndmsg: netlink hdr
999 * @nlattr: netlink attributes
1001 * @addr: MAC address to delete
1002 * @vid: VLAN to delete
1004 * Delete FDB entry from switch device.
1006 int switchdev_port_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1007 struct net_device *dev, const unsigned char *addr,
1010 struct switchdev_obj_port_fdb fdb = {
1011 .obj.orig_dev = dev,
1012 .obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
1016 ether_addr_copy(fdb.addr, addr);
1017 return switchdev_port_obj_del(dev, &fdb.obj);
1019 EXPORT_SYMBOL_GPL(switchdev_port_fdb_del);
1021 struct switchdev_fdb_dump {
1022 struct switchdev_obj_port_fdb fdb;
1023 struct net_device *dev;
1024 struct sk_buff *skb;
1025 struct netlink_callback *cb;
1029 static int switchdev_port_fdb_dump_cb(struct switchdev_obj *obj)
1031 struct switchdev_obj_port_fdb *fdb = SWITCHDEV_OBJ_PORT_FDB(obj);
1032 struct switchdev_fdb_dump *dump =
1033 container_of(fdb, struct switchdev_fdb_dump, fdb);
1034 u32 portid = NETLINK_CB(dump->cb->skb).portid;
1035 u32 seq = dump->cb->nlh->nlmsg_seq;
1036 struct nlmsghdr *nlh;
1039 if (dump->idx < dump->cb->args[0])
1042 nlh = nlmsg_put(dump->skb, portid, seq, RTM_NEWNEIGH,
1043 sizeof(*ndm), NLM_F_MULTI);
1047 ndm = nlmsg_data(nlh);
1048 ndm->ndm_family = AF_BRIDGE;
1051 ndm->ndm_flags = NTF_SELF;
1053 ndm->ndm_ifindex = dump->dev->ifindex;
1054 ndm->ndm_state = fdb->ndm_state;
1056 if (nla_put(dump->skb, NDA_LLADDR, ETH_ALEN, fdb->addr))
1057 goto nla_put_failure;
1059 if (fdb->vid && nla_put_u16(dump->skb, NDA_VLAN, fdb->vid))
1060 goto nla_put_failure;
1062 nlmsg_end(dump->skb, nlh);
1069 nlmsg_cancel(dump->skb, nlh);
1074 * switchdev_port_fdb_dump - Dump port FDB (MAC/VLAN) entries
1077 * @cb: netlink callback
1079 * @filter_dev: filter device
1082 * Dump FDB entries from switch device.
1084 int switchdev_port_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1085 struct net_device *dev,
1086 struct net_device *filter_dev, int idx)
1088 struct switchdev_fdb_dump dump = {
1089 .fdb.obj.orig_dev = dev,
1090 .fdb.obj.id = SWITCHDEV_OBJ_ID_PORT_FDB,
1098 err = switchdev_port_obj_dump(dev, &dump.fdb.obj,
1099 switchdev_port_fdb_dump_cb);
1103 EXPORT_SYMBOL_GPL(switchdev_port_fdb_dump);
1105 static struct net_device *switchdev_get_lowest_dev(struct net_device *dev)
1107 const struct switchdev_ops *ops = dev->switchdev_ops;
1108 struct net_device *lower_dev;
1109 struct net_device *port_dev;
1110 struct list_head *iter;
1112 /* Recusively search down until we find a sw port dev.
1113 * (A sw port dev supports switchdev_port_attr_get).
1116 if (ops && ops->switchdev_port_attr_get)
1119 netdev_for_each_lower_dev(dev, lower_dev, iter) {
1120 port_dev = switchdev_get_lowest_dev(lower_dev);
1128 static struct net_device *switchdev_get_dev_by_nhs(struct fib_info *fi)
1130 struct switchdev_attr attr = {
1131 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1133 struct switchdev_attr prev_attr;
1134 struct net_device *dev = NULL;
1139 /* For this route, all nexthop devs must be on the same switch. */
1141 for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
1142 const struct fib_nh *nh = &fi->fib_nh[nhsel];
1147 dev = switchdev_get_lowest_dev(nh->nh_dev);
1151 attr.orig_dev = dev;
1152 if (switchdev_port_attr_get(dev, &attr))
1156 !netdev_phys_item_id_same(&prev_attr.u.ppid, &attr.u.ppid))
1166 * switchdev_fib_ipv4_add - Add/modify switch IPv4 route entry
1168 * @dst: route's IPv4 destination address
1169 * @dst_len: destination address length (prefix length)
1170 * @fi: route FIB info structure
1173 * @nlflags: netlink flags passed in (NLM_F_*)
1174 * @tb_id: route table ID
1176 * Add/modify switch IPv4 route entry.
1178 int switchdev_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi,
1179 u8 tos, u8 type, u32 nlflags, u32 tb_id)
1181 struct switchdev_obj_ipv4_fib ipv4_fib = {
1182 .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB,
1190 struct net_device *dev;
1193 memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi));
1195 /* Don't offload route if using custom ip rules or if
1196 * IPv4 FIB offloading has been disabled completely.
1199 #ifdef CONFIG_IP_MULTIPLE_TABLES
1200 if (fi->fib_net->ipv4.fib_has_custom_rules)
1204 if (fi->fib_net->ipv4.fib_offload_disabled)
1207 dev = switchdev_get_dev_by_nhs(fi);
1211 ipv4_fib.obj.orig_dev = dev;
1212 err = switchdev_port_obj_add(dev, &ipv4_fib.obj);
1214 fi->fib_flags |= RTNH_F_OFFLOAD;
1216 return err == -EOPNOTSUPP ? 0 : err;
1218 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add);
1221 * switchdev_fib_ipv4_del - Delete IPv4 route entry from switch
1223 * @dst: route's IPv4 destination address
1224 * @dst_len: destination address length (prefix length)
1225 * @fi: route FIB info structure
1228 * @tb_id: route table ID
1230 * Delete IPv4 route entry from switch device.
1232 int switchdev_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi,
1233 u8 tos, u8 type, u32 tb_id)
1235 struct switchdev_obj_ipv4_fib ipv4_fib = {
1236 .obj.id = SWITCHDEV_OBJ_ID_IPV4_FIB,
1244 struct net_device *dev;
1247 memcpy(&ipv4_fib.fi, fi, sizeof(ipv4_fib.fi));
1249 if (!(fi->fib_flags & RTNH_F_OFFLOAD))
1252 dev = switchdev_get_dev_by_nhs(fi);
1256 ipv4_fib.obj.orig_dev = dev;
1257 err = switchdev_port_obj_del(dev, &ipv4_fib.obj);
1259 fi->fib_flags &= ~RTNH_F_OFFLOAD;
1261 return err == -EOPNOTSUPP ? 0 : err;
1263 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del);
1266 * switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation
1268 * @fi: route FIB info structure
1270 void switchdev_fib_ipv4_abort(struct fib_info *fi)
1272 /* There was a problem installing this route to the offload
1273 * device. For now, until we come up with more refined
1274 * policy handling, abruptly end IPv4 fib offloading for
1275 * for entire net by flushing offload device(s) of all
1276 * IPv4 routes, and mark IPv4 fib offloading broken from
1277 * this point forward.
1280 fib_flush_external(fi->fib_net);
1281 fi->fib_net->ipv4.fib_offload_disabled = true;
1283 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort);
1285 static bool switchdev_port_same_parent_id(struct net_device *a,
1286 struct net_device *b)
1288 struct switchdev_attr a_attr = {
1290 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1291 .flags = SWITCHDEV_F_NO_RECURSE,
1293 struct switchdev_attr b_attr = {
1295 .id = SWITCHDEV_ATTR_ID_PORT_PARENT_ID,
1296 .flags = SWITCHDEV_F_NO_RECURSE,
1299 if (switchdev_port_attr_get(a, &a_attr) ||
1300 switchdev_port_attr_get(b, &b_attr))
1303 return netdev_phys_item_id_same(&a_attr.u.ppid, &b_attr.u.ppid);
1306 static u32 switchdev_port_fwd_mark_get(struct net_device *dev,
1307 struct net_device *group_dev)
1309 struct net_device *lower_dev;
1310 struct list_head *iter;
1312 netdev_for_each_lower_dev(group_dev, lower_dev, iter) {
1313 if (lower_dev == dev)
1315 if (switchdev_port_same_parent_id(dev, lower_dev))
1316 return lower_dev->offload_fwd_mark;
1317 return switchdev_port_fwd_mark_get(dev, lower_dev);
1320 return dev->ifindex;
1323 static void switchdev_port_fwd_mark_reset(struct net_device *group_dev,
1324 u32 old_mark, u32 *reset_mark)
1326 struct net_device *lower_dev;
1327 struct list_head *iter;
1329 netdev_for_each_lower_dev(group_dev, lower_dev, iter) {
1330 if (lower_dev->offload_fwd_mark == old_mark) {
1332 *reset_mark = lower_dev->ifindex;
1333 lower_dev->offload_fwd_mark = *reset_mark;
1335 switchdev_port_fwd_mark_reset(lower_dev, old_mark, reset_mark);
1340 * switchdev_port_fwd_mark_set - Set port offload forwarding mark
1343 * @group_dev: containing device
1344 * @joining: true if dev is joining group; false if leaving group
1346 * An ungrouped port's offload mark is just its ifindex. A grouped
1347 * port's (member of a bridge, for example) offload mark is the ifindex
1348 * of one of the ports in the group with the same parent (switch) ID.
1349 * Ports on the same device in the same group will have the same mark.
1354 * sw1p1 ifindex=2 mark=2
1355 * sw1p2 ifindex=3 mark=2
1356 * sw2p1 ifindex=4 mark=5
1357 * sw2p2 ifindex=5 mark=5
1359 * If sw2p2 leaves the bridge, we'll have:
1362 * sw1p1 ifindex=2 mark=2
1363 * sw1p2 ifindex=3 mark=2
1364 * sw2p1 ifindex=4 mark=4
1365 * sw2p2 ifindex=5 mark=5
1367 void switchdev_port_fwd_mark_set(struct net_device *dev,
1368 struct net_device *group_dev,
1371 u32 mark = dev->ifindex;
1377 mark = switchdev_port_fwd_mark_get(dev, group_dev);
1378 else if (dev->offload_fwd_mark == mark)
1379 /* Ohoh, this port was the mark reference port,
1380 * but it's leaving the group, so reset the
1381 * mark for the remaining ports in the group.
1383 switchdev_port_fwd_mark_reset(group_dev, mark,
1387 dev->offload_fwd_mark = mark;
1389 EXPORT_SYMBOL_GPL(switchdev_port_fwd_mark_set);