2 * net/switchdev/switchdev.c - Switch device API
3 * Copyright (c) 2014 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/if_bridge.h>
19 #include <net/ip_fib.h>
20 #include <net/switchdev.h>
23 * switchdev_port_attr_get - Get port attribute
26 * @attr: attribute to get
28 int switchdev_port_attr_get(struct net_device
*dev
, struct switchdev_attr
*attr
)
30 const struct switchdev_ops
*ops
= dev
->switchdev_ops
;
31 struct net_device
*lower_dev
;
32 struct list_head
*iter
;
33 struct switchdev_attr first
= {
34 .id
= SWITCHDEV_ATTR_UNDEFINED
36 int err
= -EOPNOTSUPP
;
38 if (ops
&& ops
->switchdev_port_attr_get
)
39 return ops
->switchdev_port_attr_get(dev
, attr
);
41 if (attr
->flags
& SWITCHDEV_F_NO_RECURSE
)
44 /* Switch device port(s) may be stacked under
45 * bond/team/vlan dev, so recurse down to get attr on
46 * each port. Return -ENODATA if attr values don't
47 * compare across ports.
50 netdev_for_each_lower_dev(dev
, lower_dev
, iter
) {
51 err
= switchdev_port_attr_get(lower_dev
, attr
);
54 if (first
.id
== SWITCHDEV_ATTR_UNDEFINED
)
56 else if (memcmp(&first
, attr
, sizeof(*attr
)))
62 EXPORT_SYMBOL_GPL(switchdev_port_attr_get
);
64 static int __switchdev_port_attr_set(struct net_device
*dev
,
65 struct switchdev_attr
*attr
)
67 const struct switchdev_ops
*ops
= dev
->switchdev_ops
;
68 struct net_device
*lower_dev
;
69 struct list_head
*iter
;
70 int err
= -EOPNOTSUPP
;
72 if (ops
&& ops
->switchdev_port_attr_set
)
73 return ops
->switchdev_port_attr_set(dev
, attr
);
75 if (attr
->flags
& SWITCHDEV_F_NO_RECURSE
)
78 /* Switch device port(s) may be stacked under
79 * bond/team/vlan dev, so recurse down to set attr on
83 netdev_for_each_lower_dev(dev
, lower_dev
, iter
) {
84 err
= __switchdev_port_attr_set(lower_dev
, attr
);
92 struct switchdev_attr_set_work
{
93 struct work_struct work
;
94 struct net_device
*dev
;
95 struct switchdev_attr attr
;
98 static void switchdev_port_attr_set_work(struct work_struct
*work
)
100 struct switchdev_attr_set_work
*asw
=
101 container_of(work
, struct switchdev_attr_set_work
, work
);
105 err
= switchdev_port_attr_set(asw
->dev
, &asw
->attr
);
106 if (err
&& err
!= -EOPNOTSUPP
)
107 netdev_err(asw
->dev
, "failed (err=%d) to set attribute (id=%d)\n",
115 static int switchdev_port_attr_set_defer(struct net_device
*dev
,
116 struct switchdev_attr
*attr
)
118 struct switchdev_attr_set_work
*asw
;
120 asw
= kmalloc(sizeof(*asw
), GFP_ATOMIC
);
124 INIT_WORK(&asw
->work
, switchdev_port_attr_set_work
);
128 memcpy(&asw
->attr
, attr
, sizeof(asw
->attr
));
130 schedule_work(&asw
->work
);
136 * switchdev_port_attr_set - Set port attribute
139 * @attr: attribute to set
141 * Use a 2-phase prepare-commit transaction model to ensure
142 * system is not left in a partially updated state due to
143 * failure from driver/device.
145 int switchdev_port_attr_set(struct net_device
*dev
, struct switchdev_attr
*attr
)
149 if (!rtnl_is_locked()) {
150 /* Running prepare-commit transaction across stacked
151 * devices requires nothing moves, so if rtnl_lock is
152 * not held, schedule a worker thread to hold rtnl_lock
153 * while setting attr.
156 return switchdev_port_attr_set_defer(dev
, attr
);
159 /* Phase I: prepare for attr set. Driver/device should fail
160 * here if there are going to be issues in the commit phase,
161 * such as lack of resources or support. The driver/device
162 * should reserve resources needed for the commit phase here,
163 * but should not commit the attr.
166 attr
->trans
= SWITCHDEV_TRANS_PREPARE
;
167 err
= __switchdev_port_attr_set(dev
, attr
);
169 /* Prepare phase failed: abort the transaction. Any
170 * resources reserved in the prepare phase are
174 attr
->trans
= SWITCHDEV_TRANS_ABORT
;
175 __switchdev_port_attr_set(dev
, attr
);
180 /* Phase II: commit attr set. This cannot fail as a fault
181 * of driver/device. If it does, it's a bug in the driver/device
182 * because the driver said everythings was OK in phase I.
185 attr
->trans
= SWITCHDEV_TRANS_COMMIT
;
186 err
= __switchdev_port_attr_set(dev
, attr
);
187 WARN(err
, "%s: Commit of attribute (id=%d) failed.\n",
188 dev
->name
, attr
->id
);
192 EXPORT_SYMBOL_GPL(switchdev_port_attr_set
);
194 static int __switchdev_port_obj_add(struct net_device
*dev
,
195 struct switchdev_obj
*obj
)
197 const struct switchdev_ops
*ops
= dev
->switchdev_ops
;
198 struct net_device
*lower_dev
;
199 struct list_head
*iter
;
200 int err
= -EOPNOTSUPP
;
202 if (ops
&& ops
->switchdev_port_obj_add
)
203 return ops
->switchdev_port_obj_add(dev
, obj
);
205 /* Switch device port(s) may be stacked under
206 * bond/team/vlan dev, so recurse down to add object on
210 netdev_for_each_lower_dev(dev
, lower_dev
, iter
) {
211 err
= __switchdev_port_obj_add(lower_dev
, obj
);
220 * switchdev_port_obj_add - Add port object
223 * @obj: object to add
225 * Use a 2-phase prepare-commit transaction model to ensure
226 * system is not left in a partially updated state due to
227 * failure from driver/device.
229 * rtnl_lock must be held.
231 int switchdev_port_obj_add(struct net_device
*dev
, struct switchdev_obj
*obj
)
237 /* Phase I: prepare for obj add. Driver/device should fail
238 * here if there are going to be issues in the commit phase,
239 * such as lack of resources or support. The driver/device
240 * should reserve resources needed for the commit phase here,
241 * but should not commit the obj.
244 obj
->trans
= SWITCHDEV_TRANS_PREPARE
;
245 err
= __switchdev_port_obj_add(dev
, obj
);
247 /* Prepare phase failed: abort the transaction. Any
248 * resources reserved in the prepare phase are
252 obj
->trans
= SWITCHDEV_TRANS_ABORT
;
253 __switchdev_port_obj_add(dev
, obj
);
258 /* Phase II: commit obj add. This cannot fail as a fault
259 * of driver/device. If it does, it's a bug in the driver/device
260 * because the driver said everythings was OK in phase I.
263 obj
->trans
= SWITCHDEV_TRANS_COMMIT
;
264 err
= __switchdev_port_obj_add(dev
, obj
);
265 WARN(err
, "%s: Commit of object (id=%d) failed.\n", dev
->name
, obj
->id
);
269 EXPORT_SYMBOL_GPL(switchdev_port_obj_add
);
272 * switchdev_port_obj_del - Delete port object
275 * @obj: object to delete
277 int switchdev_port_obj_del(struct net_device
*dev
, struct switchdev_obj
*obj
)
279 const struct switchdev_ops
*ops
= dev
->switchdev_ops
;
280 struct net_device
*lower_dev
;
281 struct list_head
*iter
;
282 int err
= -EOPNOTSUPP
;
284 if (ops
&& ops
->switchdev_port_obj_del
)
285 return ops
->switchdev_port_obj_del(dev
, obj
);
287 /* Switch device port(s) may be stacked under
288 * bond/team/vlan dev, so recurse down to delete object on
292 netdev_for_each_lower_dev(dev
, lower_dev
, iter
) {
293 err
= switchdev_port_obj_del(lower_dev
, obj
);
300 EXPORT_SYMBOL_GPL(switchdev_port_obj_del
);
303 * switchdev_port_obj_dump - Dump port objects
306 * @obj: object to dump
308 int switchdev_port_obj_dump(struct net_device
*dev
, struct switchdev_obj
*obj
)
310 const struct switchdev_ops
*ops
= dev
->switchdev_ops
;
311 struct net_device
*lower_dev
;
312 struct list_head
*iter
;
313 int err
= -EOPNOTSUPP
;
315 if (ops
&& ops
->switchdev_port_obj_dump
)
316 return ops
->switchdev_port_obj_dump(dev
, obj
);
318 /* Switch device port(s) may be stacked under
319 * bond/team/vlan dev, so recurse down to dump objects on
320 * first port at bottom of stack.
323 netdev_for_each_lower_dev(dev
, lower_dev
, iter
) {
324 err
= switchdev_port_obj_dump(lower_dev
, obj
);
330 EXPORT_SYMBOL_GPL(switchdev_port_obj_dump
);
332 static DEFINE_MUTEX(switchdev_mutex
);
333 static RAW_NOTIFIER_HEAD(switchdev_notif_chain
);
336 * register_switchdev_notifier - Register notifier
337 * @nb: notifier_block
339 * Register switch device notifier. This should be used by code
340 * which needs to monitor events happening in particular device.
341 * Return values are same as for atomic_notifier_chain_register().
343 int register_switchdev_notifier(struct notifier_block
*nb
)
347 mutex_lock(&switchdev_mutex
);
348 err
= raw_notifier_chain_register(&switchdev_notif_chain
, nb
);
349 mutex_unlock(&switchdev_mutex
);
352 EXPORT_SYMBOL_GPL(register_switchdev_notifier
);
355 * unregister_switchdev_notifier - Unregister notifier
356 * @nb: notifier_block
358 * Unregister switch device notifier.
359 * Return values are same as for atomic_notifier_chain_unregister().
361 int unregister_switchdev_notifier(struct notifier_block
*nb
)
365 mutex_lock(&switchdev_mutex
);
366 err
= raw_notifier_chain_unregister(&switchdev_notif_chain
, nb
);
367 mutex_unlock(&switchdev_mutex
);
370 EXPORT_SYMBOL_GPL(unregister_switchdev_notifier
);
373 * call_switchdev_notifiers - Call notifiers
374 * @val: value passed unmodified to notifier function
376 * @info: notifier information data
378 * Call all network notifier blocks. This should be called by driver
379 * when it needs to propagate hardware event.
380 * Return values are same as for atomic_notifier_call_chain().
382 int call_switchdev_notifiers(unsigned long val
, struct net_device
*dev
,
383 struct switchdev_notifier_info
*info
)
388 mutex_lock(&switchdev_mutex
);
389 err
= raw_notifier_call_chain(&switchdev_notif_chain
, val
, info
);
390 mutex_unlock(&switchdev_mutex
);
393 EXPORT_SYMBOL_GPL(call_switchdev_notifiers
);
395 struct switchdev_vlan_dump
{
396 struct switchdev_obj obj
;
404 static int switchdev_port_vlan_dump_put(struct net_device
*dev
,
405 struct switchdev_vlan_dump
*dump
)
407 struct bridge_vlan_info vinfo
;
409 vinfo
.flags
= dump
->flags
;
411 if (dump
->begin
== 0 && dump
->end
== 0) {
413 } else if (dump
->begin
== dump
->end
) {
414 vinfo
.vid
= dump
->begin
;
415 if (nla_put(dump
->skb
, IFLA_BRIDGE_VLAN_INFO
,
416 sizeof(vinfo
), &vinfo
))
419 vinfo
.vid
= dump
->begin
;
420 vinfo
.flags
|= BRIDGE_VLAN_INFO_RANGE_BEGIN
;
421 if (nla_put(dump
->skb
, IFLA_BRIDGE_VLAN_INFO
,
422 sizeof(vinfo
), &vinfo
))
424 vinfo
.vid
= dump
->end
;
425 vinfo
.flags
&= ~BRIDGE_VLAN_INFO_RANGE_BEGIN
;
426 vinfo
.flags
|= BRIDGE_VLAN_INFO_RANGE_END
;
427 if (nla_put(dump
->skb
, IFLA_BRIDGE_VLAN_INFO
,
428 sizeof(vinfo
), &vinfo
))
435 static int switchdev_port_vlan_dump_cb(struct net_device
*dev
,
436 struct switchdev_obj
*obj
)
438 struct switchdev_vlan_dump
*dump
=
439 container_of(obj
, struct switchdev_vlan_dump
, obj
);
440 struct switchdev_obj_vlan
*vlan
= &dump
->obj
.u
.vlan
;
443 if (vlan
->vid_begin
> vlan
->vid_end
)
446 if (dump
->filter_mask
& RTEXT_FILTER_BRVLAN
) {
447 dump
->flags
= vlan
->flags
;
448 for (dump
->begin
= dump
->end
= vlan
->vid_begin
;
449 dump
->begin
<= vlan
->vid_end
;
450 dump
->begin
++, dump
->end
++) {
451 err
= switchdev_port_vlan_dump_put(dev
, dump
);
455 } else if (dump
->filter_mask
& RTEXT_FILTER_BRVLAN_COMPRESSED
) {
456 if (dump
->begin
> vlan
->vid_begin
&&
457 dump
->begin
>= vlan
->vid_end
) {
458 if ((dump
->begin
- 1) == vlan
->vid_end
&&
459 dump
->flags
== vlan
->flags
) {
461 dump
->begin
= vlan
->vid_begin
;
463 err
= switchdev_port_vlan_dump_put(dev
, dump
);
464 dump
->flags
= vlan
->flags
;
465 dump
->begin
= vlan
->vid_begin
;
466 dump
->end
= vlan
->vid_end
;
468 } else if (dump
->end
<= vlan
->vid_begin
&&
469 dump
->end
< vlan
->vid_end
) {
470 if ((dump
->end
+ 1) == vlan
->vid_begin
&&
471 dump
->flags
== vlan
->flags
) {
473 dump
->end
= vlan
->vid_end
;
475 err
= switchdev_port_vlan_dump_put(dev
, dump
);
476 dump
->flags
= vlan
->flags
;
477 dump
->begin
= vlan
->vid_begin
;
478 dump
->end
= vlan
->vid_end
;
488 static int switchdev_port_vlan_fill(struct sk_buff
*skb
, struct net_device
*dev
,
491 struct switchdev_vlan_dump dump
= {
493 .id
= SWITCHDEV_OBJ_PORT_VLAN
,
494 .cb
= switchdev_port_vlan_dump_cb
,
497 .filter_mask
= filter_mask
,
501 if ((filter_mask
& RTEXT_FILTER_BRVLAN
) ||
502 (filter_mask
& RTEXT_FILTER_BRVLAN_COMPRESSED
)) {
503 err
= switchdev_port_obj_dump(dev
, &dump
.obj
);
506 if (filter_mask
& RTEXT_FILTER_BRVLAN_COMPRESSED
)
508 err
= switchdev_port_vlan_dump_put(dev
, &dump
);
512 return err
== -EOPNOTSUPP
? 0 : err
;
516 * switchdev_port_bridge_getlink - Get bridge port attributes
520 * Called for SELF on rtnl_bridge_getlink to get bridge port
523 int switchdev_port_bridge_getlink(struct sk_buff
*skb
, u32 pid
, u32 seq
,
524 struct net_device
*dev
, u32 filter_mask
,
527 struct switchdev_attr attr
= {
528 .id
= SWITCHDEV_ATTR_PORT_BRIDGE_FLAGS
,
530 u16 mode
= BRIDGE_MODE_UNDEF
;
531 u32 mask
= BR_LEARNING
| BR_LEARNING_SYNC
;
534 err
= switchdev_port_attr_get(dev
, &attr
);
535 if (err
&& err
!= -EOPNOTSUPP
)
538 return ndo_dflt_bridge_getlink(skb
, pid
, seq
, dev
, mode
,
539 attr
.u
.brport_flags
, mask
, nlflags
,
540 filter_mask
, switchdev_port_vlan_fill
);
542 EXPORT_SYMBOL_GPL(switchdev_port_bridge_getlink
);
544 static int switchdev_port_br_setflag(struct net_device
*dev
,
545 struct nlattr
*nlattr
,
546 unsigned long brport_flag
)
548 struct switchdev_attr attr
= {
549 .id
= SWITCHDEV_ATTR_PORT_BRIDGE_FLAGS
,
551 u8 flag
= nla_get_u8(nlattr
);
554 err
= switchdev_port_attr_get(dev
, &attr
);
559 attr
.u
.brport_flags
|= brport_flag
;
561 attr
.u
.brport_flags
&= ~brport_flag
;
563 return switchdev_port_attr_set(dev
, &attr
);
566 static const struct nla_policy
567 switchdev_port_bridge_policy
[IFLA_BRPORT_MAX
+ 1] = {
568 [IFLA_BRPORT_STATE
] = { .type
= NLA_U8
},
569 [IFLA_BRPORT_COST
] = { .type
= NLA_U32
},
570 [IFLA_BRPORT_PRIORITY
] = { .type
= NLA_U16
},
571 [IFLA_BRPORT_MODE
] = { .type
= NLA_U8
},
572 [IFLA_BRPORT_GUARD
] = { .type
= NLA_U8
},
573 [IFLA_BRPORT_PROTECT
] = { .type
= NLA_U8
},
574 [IFLA_BRPORT_FAST_LEAVE
] = { .type
= NLA_U8
},
575 [IFLA_BRPORT_LEARNING
] = { .type
= NLA_U8
},
576 [IFLA_BRPORT_LEARNING_SYNC
] = { .type
= NLA_U8
},
577 [IFLA_BRPORT_UNICAST_FLOOD
] = { .type
= NLA_U8
},
580 static int switchdev_port_br_setlink_protinfo(struct net_device
*dev
,
581 struct nlattr
*protinfo
)
587 err
= nla_validate_nested(protinfo
, IFLA_BRPORT_MAX
,
588 switchdev_port_bridge_policy
);
592 nla_for_each_nested(attr
, protinfo
, rem
) {
593 switch (nla_type(attr
)) {
594 case IFLA_BRPORT_LEARNING
:
595 err
= switchdev_port_br_setflag(dev
, attr
,
598 case IFLA_BRPORT_LEARNING_SYNC
:
599 err
= switchdev_port_br_setflag(dev
, attr
,
613 static int switchdev_port_br_afspec(struct net_device
*dev
,
614 struct nlattr
*afspec
,
615 int (*f
)(struct net_device
*dev
,
616 struct switchdev_obj
*obj
))
619 struct bridge_vlan_info
*vinfo
;
620 struct switchdev_obj obj
= {
621 .id
= SWITCHDEV_OBJ_PORT_VLAN
,
623 struct switchdev_obj_vlan
*vlan
= &obj
.u
.vlan
;
627 nla_for_each_nested(attr
, afspec
, rem
) {
628 if (nla_type(attr
) != IFLA_BRIDGE_VLAN_INFO
)
630 if (nla_len(attr
) != sizeof(struct bridge_vlan_info
))
632 vinfo
= nla_data(attr
);
633 vlan
->flags
= vinfo
->flags
;
634 if (vinfo
->flags
& BRIDGE_VLAN_INFO_RANGE_BEGIN
) {
637 vlan
->vid_begin
= vinfo
->vid
;
638 } else if (vinfo
->flags
& BRIDGE_VLAN_INFO_RANGE_END
) {
639 if (!vlan
->vid_begin
)
641 vlan
->vid_end
= vinfo
->vid
;
642 if (vlan
->vid_end
<= vlan
->vid_begin
)
647 memset(vlan
, 0, sizeof(*vlan
));
651 vlan
->vid_begin
= vinfo
->vid
;
652 vlan
->vid_end
= vinfo
->vid
;
656 memset(vlan
, 0, sizeof(*vlan
));
664 * switchdev_port_bridge_setlink - Set bridge port attributes
667 * @nlh: netlink header
668 * @flags: netlink flags
670 * Called for SELF on rtnl_bridge_setlink to set bridge port
673 int switchdev_port_bridge_setlink(struct net_device
*dev
,
674 struct nlmsghdr
*nlh
, u16 flags
)
676 struct nlattr
*protinfo
;
677 struct nlattr
*afspec
;
680 protinfo
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
),
683 err
= switchdev_port_br_setlink_protinfo(dev
, protinfo
);
688 afspec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
),
691 err
= switchdev_port_br_afspec(dev
, afspec
,
692 switchdev_port_obj_add
);
696 EXPORT_SYMBOL_GPL(switchdev_port_bridge_setlink
);
699 * switchdev_port_bridge_dellink - Set bridge port attributes
702 * @nlh: netlink header
703 * @flags: netlink flags
705 * Called for SELF on rtnl_bridge_dellink to set bridge port
708 int switchdev_port_bridge_dellink(struct net_device
*dev
,
709 struct nlmsghdr
*nlh
, u16 flags
)
711 struct nlattr
*afspec
;
713 afspec
= nlmsg_find_attr(nlh
, sizeof(struct ifinfomsg
),
716 return switchdev_port_br_afspec(dev
, afspec
,
717 switchdev_port_obj_del
);
721 EXPORT_SYMBOL_GPL(switchdev_port_bridge_dellink
);
724 * switchdev_port_fdb_add - Add FDB (MAC/VLAN) entry to port
726 * @ndmsg: netlink hdr
727 * @nlattr: netlink attributes
729 * @addr: MAC address to add
732 * Add FDB entry to switch device.
734 int switchdev_port_fdb_add(struct ndmsg
*ndm
, struct nlattr
*tb
[],
735 struct net_device
*dev
, const unsigned char *addr
,
736 u16 vid
, u16 nlm_flags
)
738 struct switchdev_obj obj
= {
739 .id
= SWITCHDEV_OBJ_PORT_FDB
,
746 return switchdev_port_obj_add(dev
, &obj
);
748 EXPORT_SYMBOL_GPL(switchdev_port_fdb_add
);
751 * switchdev_port_fdb_del - Delete FDB (MAC/VLAN) entry from port
753 * @ndmsg: netlink hdr
754 * @nlattr: netlink attributes
756 * @addr: MAC address to delete
757 * @vid: VLAN to delete
759 * Delete FDB entry from switch device.
761 int switchdev_port_fdb_del(struct ndmsg
*ndm
, struct nlattr
*tb
[],
762 struct net_device
*dev
, const unsigned char *addr
,
765 struct switchdev_obj obj
= {
766 .id
= SWITCHDEV_OBJ_PORT_FDB
,
773 return switchdev_port_obj_del(dev
, &obj
);
775 EXPORT_SYMBOL_GPL(switchdev_port_fdb_del
);
777 struct switchdev_fdb_dump
{
778 struct switchdev_obj obj
;
780 struct netlink_callback
*cb
;
784 static int switchdev_port_fdb_dump_cb(struct net_device
*dev
,
785 struct switchdev_obj
*obj
)
787 struct switchdev_fdb_dump
*dump
=
788 container_of(obj
, struct switchdev_fdb_dump
, obj
);
789 u32 portid
= NETLINK_CB(dump
->cb
->skb
).portid
;
790 u32 seq
= dump
->cb
->nlh
->nlmsg_seq
;
791 struct nlmsghdr
*nlh
;
794 if (dump
->idx
< dump
->cb
->args
[0])
797 nlh
= nlmsg_put(dump
->skb
, portid
, seq
, RTM_NEWNEIGH
,
798 sizeof(*ndm
), NLM_F_MULTI
);
802 ndm
= nlmsg_data(nlh
);
803 ndm
->ndm_family
= AF_BRIDGE
;
806 ndm
->ndm_flags
= NTF_SELF
;
808 ndm
->ndm_ifindex
= dev
->ifindex
;
809 ndm
->ndm_state
= NUD_REACHABLE
;
811 if (nla_put(dump
->skb
, NDA_LLADDR
, ETH_ALEN
, obj
->u
.fdb
.addr
))
812 goto nla_put_failure
;
814 if (obj
->u
.fdb
.vid
&& nla_put_u16(dump
->skb
, NDA_VLAN
, obj
->u
.fdb
.vid
))
815 goto nla_put_failure
;
817 nlmsg_end(dump
->skb
, nlh
);
824 nlmsg_cancel(dump
->skb
, nlh
);
829 * switchdev_port_fdb_dump - Dump port FDB (MAC/VLAN) entries
832 * @cb: netlink callback
834 * @filter_dev: filter device
837 * Delete FDB entry from switch device.
839 int switchdev_port_fdb_dump(struct sk_buff
*skb
, struct netlink_callback
*cb
,
840 struct net_device
*dev
,
841 struct net_device
*filter_dev
, int idx
)
843 struct switchdev_fdb_dump dump
= {
845 .id
= SWITCHDEV_OBJ_PORT_FDB
,
846 .cb
= switchdev_port_fdb_dump_cb
,
854 err
= switchdev_port_obj_dump(dev
, &dump
.obj
);
860 EXPORT_SYMBOL_GPL(switchdev_port_fdb_dump
);
862 static struct net_device
*switchdev_get_lowest_dev(struct net_device
*dev
)
864 const struct switchdev_ops
*ops
= dev
->switchdev_ops
;
865 struct net_device
*lower_dev
;
866 struct net_device
*port_dev
;
867 struct list_head
*iter
;
869 /* Recusively search down until we find a sw port dev.
870 * (A sw port dev supports switchdev_port_attr_get).
873 if (ops
&& ops
->switchdev_port_attr_get
)
876 netdev_for_each_lower_dev(dev
, lower_dev
, iter
) {
877 port_dev
= switchdev_get_lowest_dev(lower_dev
);
885 static struct net_device
*switchdev_get_dev_by_nhs(struct fib_info
*fi
)
887 struct switchdev_attr attr
= {
888 .id
= SWITCHDEV_ATTR_PORT_PARENT_ID
,
890 struct switchdev_attr prev_attr
;
891 struct net_device
*dev
= NULL
;
894 /* For this route, all nexthop devs must be on the same switch. */
896 for (nhsel
= 0; nhsel
< fi
->fib_nhs
; nhsel
++) {
897 const struct fib_nh
*nh
= &fi
->fib_nh
[nhsel
];
902 dev
= switchdev_get_lowest_dev(nh
->nh_dev
);
906 if (switchdev_port_attr_get(dev
, &attr
))
910 if (prev_attr
.u
.ppid
.id_len
!= attr
.u
.ppid
.id_len
)
912 if (memcmp(prev_attr
.u
.ppid
.id
, attr
.u
.ppid
.id
,
924 * switchdev_fib_ipv4_add - Add/modify switch IPv4 route entry
926 * @dst: route's IPv4 destination address
927 * @dst_len: destination address length (prefix length)
928 * @fi: route FIB info structure
931 * @nlflags: netlink flags passed in (NLM_F_*)
932 * @tb_id: route table ID
934 * Add/modify switch IPv4 route entry.
936 int switchdev_fib_ipv4_add(u32 dst
, int dst_len
, struct fib_info
*fi
,
937 u8 tos
, u8 type
, u32 nlflags
, u32 tb_id
)
939 struct switchdev_obj fib_obj
= {
940 .id
= SWITCHDEV_OBJ_IPV4_FIB
,
951 struct net_device
*dev
;
954 /* Don't offload route if using custom ip rules or if
955 * IPv4 FIB offloading has been disabled completely.
958 #ifdef CONFIG_IP_MULTIPLE_TABLES
959 if (fi
->fib_net
->ipv4
.fib_has_custom_rules
)
963 if (fi
->fib_net
->ipv4
.fib_offload_disabled
)
966 dev
= switchdev_get_dev_by_nhs(fi
);
970 err
= switchdev_port_obj_add(dev
, &fib_obj
);
972 fi
->fib_flags
|= RTNH_F_OFFLOAD
;
974 return err
== -EOPNOTSUPP
? 0 : err
;
976 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_add
);
979 * switchdev_fib_ipv4_del - Delete IPv4 route entry from switch
981 * @dst: route's IPv4 destination address
982 * @dst_len: destination address length (prefix length)
983 * @fi: route FIB info structure
986 * @tb_id: route table ID
988 * Delete IPv4 route entry from switch device.
990 int switchdev_fib_ipv4_del(u32 dst
, int dst_len
, struct fib_info
*fi
,
991 u8 tos
, u8 type
, u32 tb_id
)
993 struct switchdev_obj fib_obj
= {
994 .id
= SWITCHDEV_OBJ_IPV4_FIB
,
1005 struct net_device
*dev
;
1008 if (!(fi
->fib_flags
& RTNH_F_OFFLOAD
))
1011 dev
= switchdev_get_dev_by_nhs(fi
);
1015 err
= switchdev_port_obj_del(dev
, &fib_obj
);
1017 fi
->fib_flags
&= ~RTNH_F_OFFLOAD
;
1019 return err
== -EOPNOTSUPP
? 0 : err
;
1021 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_del
);
1024 * switchdev_fib_ipv4_abort - Abort an IPv4 FIB operation
1026 * @fi: route FIB info structure
1028 void switchdev_fib_ipv4_abort(struct fib_info
*fi
)
1030 /* There was a problem installing this route to the offload
1031 * device. For now, until we come up with more refined
1032 * policy handling, abruptly end IPv4 fib offloading for
1033 * for entire net by flushing offload device(s) of all
1034 * IPv4 routes, and mark IPv4 fib offloading broken from
1035 * this point forward.
1038 fib_flush_external(fi
->fib_net
);
1039 fi
->fib_net
->ipv4
.fib_offload_disabled
= true;
1041 EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort
);