1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34#include <linux/kernel.h>
35#include <linux/module.h>
36#include <linux/types.h>
37#include <linux/fcntl.h>
38#include <linux/interrupt.h>
39#include <linux/ptrace.h>
40#include <linux/ioport.h>
41#include <linux/in.h>
42#include <net/ip.h>
43#include <linux/ip.h>
44#include <linux/tcp.h>
45#include <linux/udp.h>
46#include <linux/slab.h>
47#include <linux/string.h>
48#include <linux/init.h>
49#include <linux/timer.h>
50#include <linux/socket.h>
51#include <linux/ctype.h>
52#include <linux/inet.h>
53#include <linux/bitops.h>
54#include <linux/io.h>
55#include <asm/dma.h>
56#include <linux/uaccess.h>
57#include <linux/errno.h>
58#include <linux/netdevice.h>
59#include <linux/inetdevice.h>
60#include <linux/igmp.h>
61#include <linux/etherdevice.h>
62#include <linux/skbuff.h>
63#include <net/sock.h>
64#include <linux/rtnetlink.h>
65#include <linux/smp.h>
66#include <linux/if_ether.h>
67#include <net/arp.h>
68#include <linux/mii.h>
69#include <linux/ethtool.h>
70#include <linux/if_vlan.h>
71#include <linux/if_bonding.h>
72#include <linux/jiffies.h>
73#include <linux/preempt.h>
74#include <net/route.h>
75#include <net/net_namespace.h>
76#include <net/netns/generic.h>
77#include <net/pkt_sched.h>
78#include <linux/rculist.h>
79#include <net/flow_dissector.h>
80#include <net/bonding.h>
81#include <net/bond_3ad.h>
82#include <net/bond_alb.h>
83
84#include "bonding_priv.h"
85
86
87
88
89
90static int max_bonds = BOND_DEFAULT_MAX_BONDS;
91static int tx_queues = BOND_DEFAULT_TX_QUEUES;
92static int num_peer_notif = 1;
93static int miimon;
94static int updelay;
95static int downdelay;
96static int use_carrier = 1;
97static char *mode;
98static char *primary;
99static char *primary_reselect;
100static char *lacp_rate;
101static int min_links;
102static char *ad_select;
103static char *xmit_hash_policy;
104static int arp_interval;
105static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
106static char *arp_validate;
107static char *arp_all_targets;
108static char *fail_over_mac;
109static int all_slaves_active;
110static struct bond_params bonding_defaults;
111static int resend_igmp = BOND_DEFAULT_RESEND_IGMP;
112static int packets_per_slave = 1;
113static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
114
115module_param(max_bonds, int, 0);
116MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
117module_param(tx_queues, int, 0);
118MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)");
119module_param_named(num_grat_arp, num_peer_notif, int, 0644);
120MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on "
121 "failover event (alias of num_unsol_na)");
122module_param_named(num_unsol_na, num_peer_notif, int, 0644);
123MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on "
124 "failover event (alias of num_grat_arp)");
125module_param(miimon, int, 0);
126MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
127module_param(updelay, int, 0);
128MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
129module_param(downdelay, int, 0);
130MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
131 "in milliseconds");
132module_param(use_carrier, int, 0);
133MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
134 "0 for off, 1 for on (default)");
135module_param(mode, charp, 0);
136MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, "
137 "1 for active-backup, 2 for balance-xor, "
138 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
139 "6 for balance-alb");
140module_param(primary, charp, 0);
141MODULE_PARM_DESC(primary, "Primary network device to use");
142module_param(primary_reselect, charp, 0);
143MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
144 "once it comes up; "
145 "0 for always (default), "
146 "1 for only if speed of primary is "
147 "better, "
148 "2 for only on active slave "
149 "failure");
150module_param(lacp_rate, charp, 0);
151MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; "
152 "0 for slow, 1 for fast");
153module_param(ad_select, charp, 0);
154MODULE_PARM_DESC(ad_select, "802.3ad aggregation selection logic; "
155 "0 for stable (default), 1 for bandwidth, "
156 "2 for count");
157module_param(min_links, int, 0);
158MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier");
159
160module_param(xmit_hash_policy, charp, 0);
161MODULE_PARM_DESC(xmit_hash_policy, "balance-alb, balance-tlb, balance-xor, 802.3ad hashing method; "
162 "0 for layer 2 (default), 1 for layer 3+4, "
163 "2 for layer 2+3, 3 for encap layer 2+3, "
164 "4 for encap layer 3+4");
165module_param(arp_interval, int, 0);
166MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
167module_param_array(arp_ip_target, charp, NULL, 0);
168MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
169module_param(arp_validate, charp, 0);
170MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; "
171 "0 for none (default), 1 for active, "
172 "2 for backup, 3 for all");
173module_param(arp_all_targets, charp, 0);
174MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all");
175module_param(fail_over_mac, charp, 0);
176MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to "
177 "the same MAC; 0 for none (default), "
178 "1 for active, 2 for follow");
179module_param(all_slaves_active, int, 0);
180MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface "
181 "by setting active flag for all slaves; "
182 "0 for never (default), 1 for always.");
183module_param(resend_igmp, int, 0);
184MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on "
185 "link failure");
186module_param(packets_per_slave, int, 0);
187MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr "
188 "mode; 0 for a random slave, 1 packet per "
189 "slave (default), >1 packets per slave.");
190module_param(lp_interval, uint, 0);
191MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where "
192 "the bonding driver sends learning packets to "
193 "each slaves peer switch. The default is 1.");
194
195
196
197#ifdef CONFIG_NET_POLL_CONTROLLER
198atomic_t netpoll_block_tx = ATOMIC_INIT(0);
199#endif
200
201unsigned int bond_net_id __read_mostly;
202
203
204
205static int bond_init(struct net_device *bond_dev);
206static void bond_uninit(struct net_device *bond_dev);
207static void bond_get_stats(struct net_device *bond_dev,
208 struct rtnl_link_stats64 *stats);
209static void bond_slave_arr_handler(struct work_struct *work);
210static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
211 int mod);
212static void bond_netdev_notify_work(struct work_struct *work);
213
214
215
216const char *bond_mode_name(int mode)
217{
218 static const char *names[] = {
219 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
220 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
221 [BOND_MODE_XOR] = "load balancing (xor)",
222 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
223 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
224 [BOND_MODE_TLB] = "transmit load balancing",
225 [BOND_MODE_ALB] = "adaptive load balancing",
226 };
227
228 if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB)
229 return "unknown";
230
231 return names[mode];
232}
233
234
235
236
237
238
239
240
241
242
243void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
244 struct net_device *slave_dev)
245{
246 skb->dev = slave_dev;
247
248 BUILD_BUG_ON(sizeof(skb->queue_mapping) !=
249 sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping));
250 skb_set_queue_mapping(skb, qdisc_skb_cb(skb)->slave_dev_queue_mapping);
251
252 if (unlikely(netpoll_tx_running(bond->dev)))
253 bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb);
254 else
255 dev_queue_xmit(skb);
256}
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279static int bond_vlan_rx_add_vid(struct net_device *bond_dev,
280 __be16 proto, u16 vid)
281{
282 struct bonding *bond = netdev_priv(bond_dev);
283 struct slave *slave, *rollback_slave;
284 struct list_head *iter;
285 int res;
286
287 bond_for_each_slave(bond, slave, iter) {
288 res = vlan_vid_add(slave->dev, proto, vid);
289 if (res)
290 goto unwind;
291 }
292
293 return 0;
294
295unwind:
296
297 bond_for_each_slave(bond, rollback_slave, iter) {
298 if (rollback_slave == slave)
299 break;
300
301 vlan_vid_del(rollback_slave->dev, proto, vid);
302 }
303
304 return res;
305}
306
307
308
309
310
311
312static int bond_vlan_rx_kill_vid(struct net_device *bond_dev,
313 __be16 proto, u16 vid)
314{
315 struct bonding *bond = netdev_priv(bond_dev);
316 struct list_head *iter;
317 struct slave *slave;
318
319 bond_for_each_slave(bond, slave, iter)
320 vlan_vid_del(slave->dev, proto, vid);
321
322 if (bond_is_lb(bond))
323 bond_alb_clear_vlan(bond, vid);
324
325 return 0;
326}
327
328
329
330
331
332
333
334
335
336int bond_set_carrier(struct bonding *bond)
337{
338 struct list_head *iter;
339 struct slave *slave;
340
341 if (!bond_has_slaves(bond))
342 goto down;
343
344 if (BOND_MODE(bond) == BOND_MODE_8023AD)
345 return bond_3ad_set_carrier(bond);
346
347 bond_for_each_slave(bond, slave, iter) {
348 if (slave->link == BOND_LINK_UP) {
349 if (!netif_carrier_ok(bond->dev)) {
350 netif_carrier_on(bond->dev);
351 return 1;
352 }
353 return 0;
354 }
355 }
356
357down:
358 if (netif_carrier_ok(bond->dev)) {
359 netif_carrier_off(bond->dev);
360 return 1;
361 }
362 return 0;
363}
364
365
366
367
368
369
370
371static int bond_update_speed_duplex(struct slave *slave)
372{
373 struct net_device *slave_dev = slave->dev;
374 struct ethtool_link_ksettings ecmd;
375 int res;
376
377 slave->speed = SPEED_UNKNOWN;
378 slave->duplex = DUPLEX_UNKNOWN;
379
380 res = __ethtool_get_link_ksettings(slave_dev, &ecmd);
381 if (res < 0)
382 return 1;
383 if (ecmd.base.speed == 0 || ecmd.base.speed == ((__u32)-1))
384 return 1;
385 switch (ecmd.base.duplex) {
386 case DUPLEX_FULL:
387 case DUPLEX_HALF:
388 break;
389 default:
390 return 1;
391 }
392
393 slave->speed = ecmd.base.speed;
394 slave->duplex = ecmd.base.duplex;
395
396 return 0;
397}
398
399const char *bond_slave_link_status(s8 link)
400{
401 switch (link) {
402 case BOND_LINK_UP:
403 return "up";
404 case BOND_LINK_FAIL:
405 return "going down";
406 case BOND_LINK_DOWN:
407 return "down";
408 case BOND_LINK_BACK:
409 return "going back";
410 default:
411 return "unknown";
412 }
413}
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430static int bond_check_dev_link(struct bonding *bond,
431 struct net_device *slave_dev, int reporting)
432{
433 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
434 int (*ioctl)(struct net_device *, struct ifreq *, int);
435 struct ifreq ifr;
436 struct mii_ioctl_data *mii;
437
438 if (!reporting && !netif_running(slave_dev))
439 return 0;
440
441 if (bond->params.use_carrier)
442 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
443
444
445 if (slave_dev->ethtool_ops->get_link)
446 return slave_dev->ethtool_ops->get_link(slave_dev) ?
447 BMSR_LSTATUS : 0;
448
449
450 ioctl = slave_ops->ndo_do_ioctl;
451 if (ioctl) {
452
453
454
455
456
457
458
459
460
461
462
463
464
465 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
466 mii = if_mii(&ifr);
467 if (ioctl(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
468 mii->reg_num = MII_BMSR;
469 if (ioctl(slave_dev, &ifr, SIOCGMIIREG) == 0)
470 return mii->val_out & BMSR_LSTATUS;
471 }
472 }
473
474
475
476
477
478
479 return reporting ? -1 : BMSR_LSTATUS;
480}
481
482
483
484
485static int bond_set_promiscuity(struct bonding *bond, int inc)
486{
487 struct list_head *iter;
488 int err = 0;
489
490 if (bond_uses_primary(bond)) {
491 struct slave *curr_active = rtnl_dereference(bond->curr_active_slave);
492
493 if (curr_active)
494 err = dev_set_promiscuity(curr_active->dev, inc);
495 } else {
496 struct slave *slave;
497
498 bond_for_each_slave(bond, slave, iter) {
499 err = dev_set_promiscuity(slave->dev, inc);
500 if (err)
501 return err;
502 }
503 }
504 return err;
505}
506
507
508static int bond_set_allmulti(struct bonding *bond, int inc)
509{
510 struct list_head *iter;
511 int err = 0;
512
513 if (bond_uses_primary(bond)) {
514 struct slave *curr_active = rtnl_dereference(bond->curr_active_slave);
515
516 if (curr_active)
517 err = dev_set_allmulti(curr_active->dev, inc);
518 } else {
519 struct slave *slave;
520
521 bond_for_each_slave(bond, slave, iter) {
522 err = dev_set_allmulti(slave->dev, inc);
523 if (err)
524 return err;
525 }
526 }
527 return err;
528}
529
530
531
532
533
534static void bond_resend_igmp_join_requests_delayed(struct work_struct *work)
535{
536 struct bonding *bond = container_of(work, struct bonding,
537 mcast_work.work);
538
539 if (!rtnl_trylock()) {
540 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
541 return;
542 }
543 call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev);
544
545 if (bond->igmp_retrans > 1) {
546 bond->igmp_retrans--;
547 queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5);
548 }
549 rtnl_unlock();
550}
551
552
553static void bond_hw_addr_flush(struct net_device *bond_dev,
554 struct net_device *slave_dev)
555{
556 struct bonding *bond = netdev_priv(bond_dev);
557
558 dev_uc_unsync(slave_dev, bond_dev);
559 dev_mc_unsync(slave_dev, bond_dev);
560
561 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
562
563 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
564
565 dev_mc_del(slave_dev, lacpdu_multicast);
566 }
567}
568
569
570
571
572
573
574
575
576static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active,
577 struct slave *old_active)
578{
579 if (old_active) {
580 if (bond->dev->flags & IFF_PROMISC)
581 dev_set_promiscuity(old_active->dev, -1);
582
583 if (bond->dev->flags & IFF_ALLMULTI)
584 dev_set_allmulti(old_active->dev, -1);
585
586 bond_hw_addr_flush(bond->dev, old_active->dev);
587 }
588
589 if (new_active) {
590
591 if (bond->dev->flags & IFF_PROMISC)
592 dev_set_promiscuity(new_active->dev, 1);
593
594 if (bond->dev->flags & IFF_ALLMULTI)
595 dev_set_allmulti(new_active->dev, 1);
596
597 netif_addr_lock_bh(bond->dev);
598 dev_uc_sync(new_active->dev, bond->dev);
599 dev_mc_sync(new_active->dev, bond->dev);
600 netif_addr_unlock_bh(bond->dev);
601 }
602}
603
604
605
606
607
608
609
610
611static int bond_set_dev_addr(struct net_device *bond_dev,
612 struct net_device *slave_dev)
613{
614 int err;
615
616 netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->name=%s slave_dev->addr_len=%d\n",
617 bond_dev, slave_dev, slave_dev->name, slave_dev->addr_len);
618 err = dev_pre_changeaddr_notify(bond_dev, slave_dev->dev_addr, NULL);
619 if (err)
620 return err;
621
622 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
623 bond_dev->addr_assign_type = NET_ADDR_STOLEN;
624 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev);
625 return 0;
626}
627
628static struct slave *bond_get_old_active(struct bonding *bond,
629 struct slave *new_active)
630{
631 struct slave *slave;
632 struct list_head *iter;
633
634 bond_for_each_slave(bond, slave, iter) {
635 if (slave == new_active)
636 continue;
637
638 if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr))
639 return slave;
640 }
641
642 return NULL;
643}
644
645
646
647
648
649
650
651static void bond_do_fail_over_mac(struct bonding *bond,
652 struct slave *new_active,
653 struct slave *old_active)
654{
655 u8 tmp_mac[MAX_ADDR_LEN];
656 struct sockaddr_storage ss;
657 int rv;
658
659 switch (bond->params.fail_over_mac) {
660 case BOND_FOM_ACTIVE:
661 if (new_active) {
662 rv = bond_set_dev_addr(bond->dev, new_active->dev);
663 if (rv)
664 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
665 -rv, bond->dev->name);
666 }
667 break;
668 case BOND_FOM_FOLLOW:
669
670
671
672
673 if (!new_active)
674 return;
675
676 if (!old_active)
677 old_active = bond_get_old_active(bond, new_active);
678
679 if (old_active) {
680 bond_hw_addr_copy(tmp_mac, new_active->dev->dev_addr,
681 new_active->dev->addr_len);
682 bond_hw_addr_copy(ss.__data,
683 old_active->dev->dev_addr,
684 old_active->dev->addr_len);
685 ss.ss_family = new_active->dev->type;
686 } else {
687 bond_hw_addr_copy(ss.__data, bond->dev->dev_addr,
688 bond->dev->addr_len);
689 ss.ss_family = bond->dev->type;
690 }
691
692 rv = dev_set_mac_address(new_active->dev,
693 (struct sockaddr *)&ss, NULL);
694 if (rv) {
695 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
696 -rv, new_active->dev->name);
697 goto out;
698 }
699
700 if (!old_active)
701 goto out;
702
703 bond_hw_addr_copy(ss.__data, tmp_mac,
704 new_active->dev->addr_len);
705 ss.ss_family = old_active->dev->type;
706
707 rv = dev_set_mac_address(old_active->dev,
708 (struct sockaddr *)&ss, NULL);
709 if (rv)
710 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
711 -rv, new_active->dev->name);
712out:
713 break;
714 default:
715 netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n",
716 bond->params.fail_over_mac);
717 break;
718 }
719
720}
721
722static struct slave *bond_choose_primary_or_current(struct bonding *bond)
723{
724 struct slave *prim = rtnl_dereference(bond->primary_slave);
725 struct slave *curr = rtnl_dereference(bond->curr_active_slave);
726
727 if (!prim || prim->link != BOND_LINK_UP) {
728 if (!curr || curr->link != BOND_LINK_UP)
729 return NULL;
730 return curr;
731 }
732
733 if (bond->force_primary) {
734 bond->force_primary = false;
735 return prim;
736 }
737
738 if (!curr || curr->link != BOND_LINK_UP)
739 return prim;
740
741
742 switch (bond->params.primary_reselect) {
743 case BOND_PRI_RESELECT_ALWAYS:
744 return prim;
745 case BOND_PRI_RESELECT_BETTER:
746 if (prim->speed < curr->speed)
747 return curr;
748 if (prim->speed == curr->speed && prim->duplex <= curr->duplex)
749 return curr;
750 return prim;
751 case BOND_PRI_RESELECT_FAILURE:
752 return curr;
753 default:
754 netdev_err(bond->dev, "impossible primary_reselect %d\n",
755 bond->params.primary_reselect);
756 return curr;
757 }
758}
759
760
761
762
763
764static struct slave *bond_find_best_slave(struct bonding *bond)
765{
766 struct slave *slave, *bestslave = NULL;
767 struct list_head *iter;
768 int mintime = bond->params.updelay;
769
770 slave = bond_choose_primary_or_current(bond);
771 if (slave)
772 return slave;
773
774 bond_for_each_slave(bond, slave, iter) {
775 if (slave->link == BOND_LINK_UP)
776 return slave;
777 if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) &&
778 slave->delay < mintime) {
779 mintime = slave->delay;
780 bestslave = slave;
781 }
782 }
783
784 return bestslave;
785}
786
787static bool bond_should_notify_peers(struct bonding *bond)
788{
789 struct slave *slave;
790
791 rcu_read_lock();
792 slave = rcu_dereference(bond->curr_active_slave);
793 rcu_read_unlock();
794
795 netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n",
796 slave ? slave->dev->name : "NULL");
797
798 if (!slave || !bond->send_peer_notif ||
799 !netif_carrier_ok(bond->dev) ||
800 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
801 return false;
802
803 return true;
804}
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
822{
823 struct slave *old_active;
824
825 ASSERT_RTNL();
826
827 old_active = rtnl_dereference(bond->curr_active_slave);
828
829 if (old_active == new_active)
830 return;
831
832 if (new_active) {
833 new_active->last_link_up = jiffies;
834
835 if (new_active->link == BOND_LINK_BACK) {
836 if (bond_uses_primary(bond)) {
837 netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n",
838 new_active->dev->name,
839 (bond->params.updelay - new_active->delay) * bond->params.miimon);
840 }
841
842 new_active->delay = 0;
843 bond_set_slave_link_state(new_active, BOND_LINK_UP,
844 BOND_SLAVE_NOTIFY_NOW);
845
846 if (BOND_MODE(bond) == BOND_MODE_8023AD)
847 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
848
849 if (bond_is_lb(bond))
850 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
851 } else {
852 if (bond_uses_primary(bond)) {
853 netdev_info(bond->dev, "making interface %s the new active one\n",
854 new_active->dev->name);
855 }
856 }
857 }
858
859 if (bond_uses_primary(bond))
860 bond_hw_addr_swap(bond, new_active, old_active);
861
862 if (bond_is_lb(bond)) {
863 bond_alb_handle_active_change(bond, new_active);
864 if (old_active)
865 bond_set_slave_inactive_flags(old_active,
866 BOND_SLAVE_NOTIFY_NOW);
867 if (new_active)
868 bond_set_slave_active_flags(new_active,
869 BOND_SLAVE_NOTIFY_NOW);
870 } else {
871 rcu_assign_pointer(bond->curr_active_slave, new_active);
872 }
873
874 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) {
875 if (old_active)
876 bond_set_slave_inactive_flags(old_active,
877 BOND_SLAVE_NOTIFY_NOW);
878
879 if (new_active) {
880 bool should_notify_peers = false;
881
882 bond_set_slave_active_flags(new_active,
883 BOND_SLAVE_NOTIFY_NOW);
884
885 if (bond->params.fail_over_mac)
886 bond_do_fail_over_mac(bond, new_active,
887 old_active);
888
889 if (netif_running(bond->dev)) {
890 bond->send_peer_notif =
891 bond->params.num_peer_notif;
892 should_notify_peers =
893 bond_should_notify_peers(bond);
894 }
895
896 call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
897 if (should_notify_peers)
898 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
899 bond->dev);
900 }
901 }
902
903
904
905
906
907
908 if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) &&
909 ((bond_uses_primary(bond) && new_active) ||
910 BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) {
911 bond->igmp_retrans = bond->params.resend_igmp;
912 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
913 }
914}
915
916
917
918
919
920
921
922
923
924
925
926
927void bond_select_active_slave(struct bonding *bond)
928{
929 struct slave *best_slave;
930 int rv;
931
932 ASSERT_RTNL();
933
934 best_slave = bond_find_best_slave(bond);
935 if (best_slave != rtnl_dereference(bond->curr_active_slave)) {
936 bond_change_active_slave(bond, best_slave);
937 rv = bond_set_carrier(bond);
938 if (!rv)
939 return;
940
941 if (netif_carrier_ok(bond->dev))
942 netdev_info(bond->dev, "first active interface up!\n");
943 else
944 netdev_info(bond->dev, "now running without any active interface!\n");
945 }
946}
947
948#ifdef CONFIG_NET_POLL_CONTROLLER
949static inline int slave_enable_netpoll(struct slave *slave)
950{
951 struct netpoll *np;
952 int err = 0;
953
954 np = kzalloc(sizeof(*np), GFP_KERNEL);
955 err = -ENOMEM;
956 if (!np)
957 goto out;
958
959 err = __netpoll_setup(np, slave->dev);
960 if (err) {
961 kfree(np);
962 goto out;
963 }
964 slave->np = np;
965out:
966 return err;
967}
968static inline void slave_disable_netpoll(struct slave *slave)
969{
970 struct netpoll *np = slave->np;
971
972 if (!np)
973 return;
974
975 slave->np = NULL;
976
977 __netpoll_free(np);
978}
979
980static void bond_poll_controller(struct net_device *bond_dev)
981{
982 struct bonding *bond = netdev_priv(bond_dev);
983 struct slave *slave = NULL;
984 struct list_head *iter;
985 struct ad_info ad_info;
986
987 if (BOND_MODE(bond) == BOND_MODE_8023AD)
988 if (bond_3ad_get_active_agg_info(bond, &ad_info))
989 return;
990
991 bond_for_each_slave_rcu(bond, slave, iter) {
992 if (!bond_slave_is_up(slave))
993 continue;
994
995 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
996 struct aggregator *agg =
997 SLAVE_AD_INFO(slave)->port.aggregator;
998
999 if (agg &&
1000 agg->aggregator_identifier != ad_info.aggregator_id)
1001 continue;
1002 }
1003
1004 netpoll_poll_dev(slave->dev);
1005 }
1006}
1007
1008static void bond_netpoll_cleanup(struct net_device *bond_dev)
1009{
1010 struct bonding *bond = netdev_priv(bond_dev);
1011 struct list_head *iter;
1012 struct slave *slave;
1013
1014 bond_for_each_slave(bond, slave, iter)
1015 if (bond_slave_is_up(slave))
1016 slave_disable_netpoll(slave);
1017}
1018
1019static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
1020{
1021 struct bonding *bond = netdev_priv(dev);
1022 struct list_head *iter;
1023 struct slave *slave;
1024 int err = 0;
1025
1026 bond_for_each_slave(bond, slave, iter) {
1027 err = slave_enable_netpoll(slave);
1028 if (err) {
1029 bond_netpoll_cleanup(dev);
1030 break;
1031 }
1032 }
1033 return err;
1034}
1035#else
1036static inline int slave_enable_netpoll(struct slave *slave)
1037{
1038 return 0;
1039}
1040static inline void slave_disable_netpoll(struct slave *slave)
1041{
1042}
1043static void bond_netpoll_cleanup(struct net_device *bond_dev)
1044{
1045}
1046#endif
1047
1048
1049
1050static netdev_features_t bond_fix_features(struct net_device *dev,
1051 netdev_features_t features)
1052{
1053 struct bonding *bond = netdev_priv(dev);
1054 struct list_head *iter;
1055 netdev_features_t mask;
1056 struct slave *slave;
1057
1058 mask = features;
1059
1060 features &= ~NETIF_F_ONE_FOR_ALL;
1061 features |= NETIF_F_ALL_FOR_ALL;
1062
1063 bond_for_each_slave(bond, slave, iter) {
1064 features = netdev_increment_features(features,
1065 slave->dev->features,
1066 mask);
1067 }
1068 features = netdev_add_tso_features(features, mask);
1069
1070 return features;
1071}
1072
1073#define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \
1074 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
1075 NETIF_F_HIGHDMA | NETIF_F_LRO)
1076
1077#define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \
1078 NETIF_F_RXCSUM | NETIF_F_ALL_TSO)
1079
1080static void bond_compute_features(struct bonding *bond)
1081{
1082 unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE |
1083 IFF_XMIT_DST_RELEASE_PERM;
1084 netdev_features_t vlan_features = BOND_VLAN_FEATURES;
1085 netdev_features_t enc_features = BOND_ENC_FEATURES;
1086 struct net_device *bond_dev = bond->dev;
1087 struct list_head *iter;
1088 struct slave *slave;
1089 unsigned short max_hard_header_len = ETH_HLEN;
1090 unsigned int gso_max_size = GSO_MAX_SIZE;
1091 u16 gso_max_segs = GSO_MAX_SEGS;
1092
1093 if (!bond_has_slaves(bond))
1094 goto done;
1095 vlan_features &= NETIF_F_ALL_FOR_ALL;
1096
1097 bond_for_each_slave(bond, slave, iter) {
1098 vlan_features = netdev_increment_features(vlan_features,
1099 slave->dev->vlan_features, BOND_VLAN_FEATURES);
1100
1101 enc_features = netdev_increment_features(enc_features,
1102 slave->dev->hw_enc_features,
1103 BOND_ENC_FEATURES);
1104 dst_release_flag &= slave->dev->priv_flags;
1105 if (slave->dev->hard_header_len > max_hard_header_len)
1106 max_hard_header_len = slave->dev->hard_header_len;
1107
1108 gso_max_size = min(gso_max_size, slave->dev->gso_max_size);
1109 gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs);
1110 }
1111 bond_dev->hard_header_len = max_hard_header_len;
1112
1113done:
1114 bond_dev->vlan_features = vlan_features;
1115 bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL |
1116 NETIF_F_GSO_UDP_L4;
1117 bond_dev->gso_max_segs = gso_max_segs;
1118 netif_set_gso_max_size(bond_dev, gso_max_size);
1119
1120 bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1121 if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) &&
1122 dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM))
1123 bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE;
1124
1125 netdev_change_features(bond_dev);
1126}
1127
1128static void bond_setup_by_slave(struct net_device *bond_dev,
1129 struct net_device *slave_dev)
1130{
1131 bond_dev->header_ops = slave_dev->header_ops;
1132
1133 bond_dev->type = slave_dev->type;
1134 bond_dev->hard_header_len = slave_dev->hard_header_len;
1135 bond_dev->addr_len = slave_dev->addr_len;
1136
1137 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1138 slave_dev->addr_len);
1139}
1140
1141
1142
1143
1144static bool bond_should_deliver_exact_match(struct sk_buff *skb,
1145 struct slave *slave,
1146 struct bonding *bond)
1147{
1148 if (bond_is_slave_inactive(slave)) {
1149 if (BOND_MODE(bond) == BOND_MODE_ALB &&
1150 skb->pkt_type != PACKET_BROADCAST &&
1151 skb->pkt_type != PACKET_MULTICAST)
1152 return false;
1153 return true;
1154 }
1155 return false;
1156}
1157
1158static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
1159{
1160 struct sk_buff *skb = *pskb;
1161 struct slave *slave;
1162 struct bonding *bond;
1163 int (*recv_probe)(const struct sk_buff *, struct bonding *,
1164 struct slave *);
1165 int ret = RX_HANDLER_ANOTHER;
1166
1167 skb = skb_share_check(skb, GFP_ATOMIC);
1168 if (unlikely(!skb))
1169 return RX_HANDLER_CONSUMED;
1170
1171 *pskb = skb;
1172
1173 slave = bond_slave_get_rcu(skb->dev);
1174 bond = slave->bond;
1175
1176 recv_probe = READ_ONCE(bond->recv_probe);
1177 if (recv_probe) {
1178 ret = recv_probe(skb, bond, slave);
1179 if (ret == RX_HANDLER_CONSUMED) {
1180 consume_skb(skb);
1181 return ret;
1182 }
1183 }
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196 if (bond_should_deliver_exact_match(skb, slave, bond)) {
1197 if (is_link_local_ether_addr(eth_hdr(skb)->h_dest))
1198 return RX_HANDLER_PASS;
1199 return RX_HANDLER_EXACT;
1200 }
1201
1202 skb->dev = bond->dev;
1203
1204 if (BOND_MODE(bond) == BOND_MODE_ALB &&
1205 bond->dev->priv_flags & IFF_BRIDGE_PORT &&
1206 skb->pkt_type == PACKET_HOST) {
1207
1208 if (unlikely(skb_cow_head(skb,
1209 skb->data - skb_mac_header(skb)))) {
1210 kfree_skb(skb);
1211 return RX_HANDLER_CONSUMED;
1212 }
1213 bond_hw_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr,
1214 bond->dev->addr_len);
1215 }
1216
1217 return ret;
1218}
1219
1220static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond)
1221{
1222 switch (BOND_MODE(bond)) {
1223 case BOND_MODE_ROUNDROBIN:
1224 return NETDEV_LAG_TX_TYPE_ROUNDROBIN;
1225 case BOND_MODE_ACTIVEBACKUP:
1226 return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP;
1227 case BOND_MODE_BROADCAST:
1228 return NETDEV_LAG_TX_TYPE_BROADCAST;
1229 case BOND_MODE_XOR:
1230 case BOND_MODE_8023AD:
1231 return NETDEV_LAG_TX_TYPE_HASH;
1232 default:
1233 return NETDEV_LAG_TX_TYPE_UNKNOWN;
1234 }
1235}
1236
1237static enum netdev_lag_hash bond_lag_hash_type(struct bonding *bond,
1238 enum netdev_lag_tx_type type)
1239{
1240 if (type != NETDEV_LAG_TX_TYPE_HASH)
1241 return NETDEV_LAG_HASH_NONE;
1242
1243 switch (bond->params.xmit_policy) {
1244 case BOND_XMIT_POLICY_LAYER2:
1245 return NETDEV_LAG_HASH_L2;
1246 case BOND_XMIT_POLICY_LAYER34:
1247 return NETDEV_LAG_HASH_L34;
1248 case BOND_XMIT_POLICY_LAYER23:
1249 return NETDEV_LAG_HASH_L23;
1250 case BOND_XMIT_POLICY_ENCAP23:
1251 return NETDEV_LAG_HASH_E23;
1252 case BOND_XMIT_POLICY_ENCAP34:
1253 return NETDEV_LAG_HASH_E34;
1254 default:
1255 return NETDEV_LAG_HASH_UNKNOWN;
1256 }
1257}
1258
1259static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave,
1260 struct netlink_ext_ack *extack)
1261{
1262 struct netdev_lag_upper_info lag_upper_info;
1263 enum netdev_lag_tx_type type;
1264
1265 type = bond_lag_tx_type(bond);
1266 lag_upper_info.tx_type = type;
1267 lag_upper_info.hash_type = bond_lag_hash_type(bond, type);
1268
1269 return netdev_master_upper_dev_link(slave->dev, bond->dev, slave,
1270 &lag_upper_info, extack);
1271}
1272
1273static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave)
1274{
1275 netdev_upper_dev_unlink(slave->dev, bond->dev);
1276 slave->dev->flags &= ~IFF_SLAVE;
1277}
1278
1279static struct slave *bond_alloc_slave(struct bonding *bond)
1280{
1281 struct slave *slave = NULL;
1282
1283 slave = kzalloc(sizeof(*slave), GFP_KERNEL);
1284 if (!slave)
1285 return NULL;
1286
1287 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1288 SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info),
1289 GFP_KERNEL);
1290 if (!SLAVE_AD_INFO(slave)) {
1291 kfree(slave);
1292 return NULL;
1293 }
1294 }
1295 INIT_DELAYED_WORK(&slave->notify_work, bond_netdev_notify_work);
1296
1297 return slave;
1298}
1299
1300static void bond_free_slave(struct slave *slave)
1301{
1302 struct bonding *bond = bond_get_bond_by_slave(slave);
1303
1304 cancel_delayed_work_sync(&slave->notify_work);
1305 if (BOND_MODE(bond) == BOND_MODE_8023AD)
1306 kfree(SLAVE_AD_INFO(slave));
1307
1308 kfree(slave);
1309}
1310
1311static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info)
1312{
1313 info->bond_mode = BOND_MODE(bond);
1314 info->miimon = bond->params.miimon;
1315 info->num_slaves = bond->slave_cnt;
1316}
1317
1318static void bond_fill_ifslave(struct slave *slave, struct ifslave *info)
1319{
1320 strcpy(info->slave_name, slave->dev->name);
1321 info->link = slave->link;
1322 info->state = bond_slave_state(slave);
1323 info->link_failure_count = slave->link_failure_count;
1324}
1325
1326static void bond_netdev_notify_work(struct work_struct *_work)
1327{
1328 struct slave *slave = container_of(_work, struct slave,
1329 notify_work.work);
1330
1331 if (rtnl_trylock()) {
1332 struct netdev_bonding_info binfo;
1333
1334 bond_fill_ifslave(slave, &binfo.slave);
1335 bond_fill_ifbond(slave->bond, &binfo.master);
1336 netdev_bonding_info_change(slave->dev, &binfo);
1337 rtnl_unlock();
1338 } else {
1339 queue_delayed_work(slave->bond->wq, &slave->notify_work, 1);
1340 }
1341}
1342
1343void bond_queue_slave_event(struct slave *slave)
1344{
1345 queue_delayed_work(slave->bond->wq, &slave->notify_work, 0);
1346}
1347
1348void bond_lower_state_changed(struct slave *slave)
1349{
1350 struct netdev_lag_lower_state_info info;
1351
1352 info.link_up = slave->link == BOND_LINK_UP ||
1353 slave->link == BOND_LINK_FAIL;
1354 info.tx_enabled = bond_is_active_slave(slave);
1355 netdev_lower_state_changed(slave->dev, &info);
1356}
1357
1358
1359int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev,
1360 struct netlink_ext_ack *extack)
1361{
1362 struct bonding *bond = netdev_priv(bond_dev);
1363 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1364 struct slave *new_slave = NULL, *prev_slave;
1365 struct sockaddr_storage ss;
1366 int link_reporting;
1367 int res = 0, i;
1368
1369 if (!bond->params.use_carrier &&
1370 slave_dev->ethtool_ops->get_link == NULL &&
1371 slave_ops->ndo_do_ioctl == NULL) {
1372 netdev_warn(bond_dev, "no link monitoring support for %s\n",
1373 slave_dev->name);
1374 }
1375
1376
1377 if (netdev_is_rx_handler_busy(slave_dev)) {
1378 NL_SET_ERR_MSG(extack, "Device is in use and cannot be enslaved");
1379 netdev_err(bond_dev,
1380 "Error: Device is in use and cannot be enslaved\n");
1381 return -EBUSY;
1382 }
1383
1384 if (bond_dev == slave_dev) {
1385 NL_SET_ERR_MSG(extack, "Cannot enslave bond to itself.");
1386 netdev_err(bond_dev, "cannot enslave bond to itself.\n");
1387 return -EPERM;
1388 }
1389
1390
1391
1392 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
1393 netdev_dbg(bond_dev, "%s is NETIF_F_VLAN_CHALLENGED\n",
1394 slave_dev->name);
1395 if (vlan_uses_dev(bond_dev)) {
1396 NL_SET_ERR_MSG(extack, "Can not enslave VLAN challenged device to VLAN enabled bond");
1397 netdev_err(bond_dev, "Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1398 slave_dev->name, bond_dev->name);
1399 return -EPERM;
1400 } else {
1401 netdev_warn(bond_dev, "enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1402 slave_dev->name, slave_dev->name,
1403 bond_dev->name);
1404 }
1405 } else {
1406 netdev_dbg(bond_dev, "%s is !NETIF_F_VLAN_CHALLENGED\n",
1407 slave_dev->name);
1408 }
1409
1410
1411
1412
1413
1414
1415 if (slave_dev->flags & IFF_UP) {
1416 NL_SET_ERR_MSG(extack, "Device can not be enslaved while up");
1417 netdev_err(bond_dev, "%s is up - this may be due to an out of date ifenslave\n",
1418 slave_dev->name);
1419 return -EPERM;
1420 }
1421
1422
1423
1424
1425
1426
1427
1428
1429 if (!bond_has_slaves(bond)) {
1430 if (bond_dev->type != slave_dev->type) {
1431 netdev_dbg(bond_dev, "change device type from %d to %d\n",
1432 bond_dev->type, slave_dev->type);
1433
1434 res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE,
1435 bond_dev);
1436 res = notifier_to_errno(res);
1437 if (res) {
1438 netdev_err(bond_dev, "refused to change device type\n");
1439 return -EBUSY;
1440 }
1441
1442
1443 dev_uc_flush(bond_dev);
1444 dev_mc_flush(bond_dev);
1445
1446 if (slave_dev->type != ARPHRD_ETHER)
1447 bond_setup_by_slave(bond_dev, slave_dev);
1448 else {
1449 ether_setup(bond_dev);
1450 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1451 }
1452
1453 call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE,
1454 bond_dev);
1455 }
1456 } else if (bond_dev->type != slave_dev->type) {
1457 NL_SET_ERR_MSG(extack, "Device type is different from other slaves");
1458 netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n",
1459 slave_dev->name, slave_dev->type, bond_dev->type);
1460 return -EINVAL;
1461 }
1462
1463 if (slave_dev->type == ARPHRD_INFINIBAND &&
1464 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
1465 NL_SET_ERR_MSG(extack, "Only active-backup mode is supported for infiniband slaves");
1466 netdev_warn(bond_dev, "Type (%d) supports only active-backup mode\n",
1467 slave_dev->type);
1468 res = -EOPNOTSUPP;
1469 goto err_undo_flags;
1470 }
1471
1472 if (!slave_ops->ndo_set_mac_address ||
1473 slave_dev->type == ARPHRD_INFINIBAND) {
1474 netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n");
1475 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP &&
1476 bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
1477 if (!bond_has_slaves(bond)) {
1478 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
1479 netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n");
1480 } else {
1481 NL_SET_ERR_MSG(extack, "Slave device does not support setting the MAC address, but fail_over_mac is not set to active");
1482 netdev_err(bond_dev, "The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n");
1483 res = -EOPNOTSUPP;
1484 goto err_undo_flags;
1485 }
1486 }
1487 }
1488
1489 call_netdevice_notifiers(NETDEV_JOIN, slave_dev);
1490
1491
1492
1493
1494 if (!bond_has_slaves(bond) &&
1495 bond->dev->addr_assign_type == NET_ADDR_RANDOM) {
1496 res = bond_set_dev_addr(bond->dev, slave_dev);
1497 if (res)
1498 goto err_undo_flags;
1499 }
1500
1501 new_slave = bond_alloc_slave(bond);
1502 if (!new_slave) {
1503 res = -ENOMEM;
1504 goto err_undo_flags;
1505 }
1506
1507 new_slave->bond = bond;
1508 new_slave->dev = slave_dev;
1509
1510
1511
1512 new_slave->queue_id = 0;
1513
1514
1515 new_slave->original_mtu = slave_dev->mtu;
1516 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1517 if (res) {
1518 netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res);
1519 goto err_free;
1520 }
1521
1522
1523
1524
1525
1526 bond_hw_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr,
1527 slave_dev->addr_len);
1528
1529 if (!bond->params.fail_over_mac ||
1530 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
1531
1532
1533
1534 memcpy(ss.__data, bond_dev->dev_addr, bond_dev->addr_len);
1535 ss.ss_family = slave_dev->type;
1536 res = dev_set_mac_address(slave_dev, (struct sockaddr *)&ss,
1537 extack);
1538 if (res) {
1539 netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res);
1540 goto err_restore_mtu;
1541 }
1542 }
1543
1544
1545 slave_dev->flags |= IFF_SLAVE;
1546
1547
1548 res = dev_open(slave_dev, extack);
1549 if (res) {
1550 netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name);
1551 goto err_restore_mac;
1552 }
1553
1554 slave_dev->priv_flags |= IFF_BONDING;
1555
1556 dev_get_stats(new_slave->dev, &new_slave->slave_stats);
1557
1558 if (bond_is_lb(bond)) {
1559
1560
1561
1562 res = bond_alb_init_slave(bond, new_slave);
1563 if (res)
1564 goto err_close;
1565 }
1566
1567 res = vlan_vids_add_by_dev(slave_dev, bond_dev);
1568 if (res) {
1569 netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n",
1570 slave_dev->name);
1571 goto err_close;
1572 }
1573
1574 prev_slave = bond_last_slave(bond);
1575
1576 new_slave->delay = 0;
1577 new_slave->link_failure_count = 0;
1578
1579 if (bond_update_speed_duplex(new_slave) &&
1580 bond_needs_speed_duplex(bond))
1581 new_slave->link = BOND_LINK_DOWN;
1582
1583 new_slave->last_rx = jiffies -
1584 (msecs_to_jiffies(bond->params.arp_interval) + 1);
1585 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
1586 new_slave->target_last_arp_rx[i] = new_slave->last_rx;
1587
1588 if (bond->params.miimon && !bond->params.use_carrier) {
1589 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1590
1591 if ((link_reporting == -1) && !bond->params.arp_interval) {
1592
1593
1594
1595
1596
1597
1598
1599
1600 netdev_warn(bond_dev, "MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n",
1601 slave_dev->name);
1602 } else if (link_reporting == -1) {
1603
1604 netdev_warn(bond_dev, "can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface\n",
1605 slave_dev->name);
1606 }
1607 }
1608
1609
1610 new_slave->link = BOND_LINK_NOCHANGE;
1611 if (bond->params.miimon) {
1612 if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) {
1613 if (bond->params.updelay) {
1614 bond_set_slave_link_state(new_slave,
1615 BOND_LINK_BACK,
1616 BOND_SLAVE_NOTIFY_NOW);
1617 new_slave->delay = bond->params.updelay;
1618 } else {
1619 bond_set_slave_link_state(new_slave,
1620 BOND_LINK_UP,
1621 BOND_SLAVE_NOTIFY_NOW);
1622 }
1623 } else {
1624 bond_set_slave_link_state(new_slave, BOND_LINK_DOWN,
1625 BOND_SLAVE_NOTIFY_NOW);
1626 }
1627 } else if (bond->params.arp_interval) {
1628 bond_set_slave_link_state(new_slave,
1629 (netif_carrier_ok(slave_dev) ?
1630 BOND_LINK_UP : BOND_LINK_DOWN),
1631 BOND_SLAVE_NOTIFY_NOW);
1632 } else {
1633 bond_set_slave_link_state(new_slave, BOND_LINK_UP,
1634 BOND_SLAVE_NOTIFY_NOW);
1635 }
1636
1637 if (new_slave->link != BOND_LINK_DOWN)
1638 new_slave->last_link_up = jiffies;
1639 netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n",
1640 new_slave->link == BOND_LINK_DOWN ? "DOWN" :
1641 (new_slave->link == BOND_LINK_UP ? "UP" : "BACK"));
1642
1643 if (bond_uses_primary(bond) && bond->params.primary[0]) {
1644
1645 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1646 rcu_assign_pointer(bond->primary_slave, new_slave);
1647 bond->force_primary = true;
1648 }
1649 }
1650
1651 switch (BOND_MODE(bond)) {
1652 case BOND_MODE_ACTIVEBACKUP:
1653 bond_set_slave_inactive_flags(new_slave,
1654 BOND_SLAVE_NOTIFY_NOW);
1655 break;
1656 case BOND_MODE_8023AD:
1657
1658
1659
1660
1661 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1662
1663 if (!prev_slave) {
1664 SLAVE_AD_INFO(new_slave)->id = 1;
1665
1666
1667
1668 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL);
1669 } else {
1670 SLAVE_AD_INFO(new_slave)->id =
1671 SLAVE_AD_INFO(prev_slave)->id + 1;
1672 }
1673
1674 bond_3ad_bind_slave(new_slave);
1675 break;
1676 case BOND_MODE_TLB:
1677 case BOND_MODE_ALB:
1678 bond_set_active_slave(new_slave);
1679 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1680 break;
1681 default:
1682 netdev_dbg(bond_dev, "This slave is always active in trunk mode\n");
1683
1684
1685 bond_set_active_slave(new_slave);
1686
1687
1688
1689
1690
1691 if (!rcu_access_pointer(bond->curr_active_slave) &&
1692 new_slave->link == BOND_LINK_UP)
1693 rcu_assign_pointer(bond->curr_active_slave, new_slave);
1694
1695 break;
1696 }
1697
1698#ifdef CONFIG_NET_POLL_CONTROLLER
1699 if (bond->dev->npinfo) {
1700 if (slave_enable_netpoll(new_slave)) {
1701 netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n");
1702 res = -EBUSY;
1703 goto err_detach;
1704 }
1705 }
1706#endif
1707
1708 if (!(bond_dev->features & NETIF_F_LRO))
1709 dev_disable_lro(slave_dev);
1710
1711 res = netdev_rx_handler_register(slave_dev, bond_handle_frame,
1712 new_slave);
1713 if (res) {
1714 netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res);
1715 goto err_detach;
1716 }
1717
1718 res = bond_master_upper_dev_link(bond, new_slave, extack);
1719 if (res) {
1720 netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res);
1721 goto err_unregister;
1722 }
1723
1724 res = bond_sysfs_slave_add(new_slave);
1725 if (res) {
1726 netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res);
1727 goto err_upper_unlink;
1728 }
1729
1730 bond->nest_level = dev_get_nest_level(bond_dev) + 1;
1731
1732
1733
1734
1735 if (!bond_uses_primary(bond)) {
1736
1737 if (bond_dev->flags & IFF_PROMISC) {
1738 res = dev_set_promiscuity(slave_dev, 1);
1739 if (res)
1740 goto err_sysfs_del;
1741 }
1742
1743
1744 if (bond_dev->flags & IFF_ALLMULTI) {
1745 res = dev_set_allmulti(slave_dev, 1);
1746 if (res) {
1747 if (bond_dev->flags & IFF_PROMISC)
1748 dev_set_promiscuity(slave_dev, -1);
1749 goto err_sysfs_del;
1750 }
1751 }
1752
1753 netif_addr_lock_bh(bond_dev);
1754 dev_mc_sync_multiple(slave_dev, bond_dev);
1755 dev_uc_sync_multiple(slave_dev, bond_dev);
1756 netif_addr_unlock_bh(bond_dev);
1757
1758 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1759
1760 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1761
1762 dev_mc_add(slave_dev, lacpdu_multicast);
1763 }
1764 }
1765
1766 bond->slave_cnt++;
1767 bond_compute_features(bond);
1768 bond_set_carrier(bond);
1769
1770 if (bond_uses_primary(bond)) {
1771 block_netpoll_tx();
1772 bond_select_active_slave(bond);
1773 unblock_netpoll_tx();
1774 }
1775
1776 if (bond_mode_can_use_xmit_hash(bond))
1777 bond_update_slave_arr(bond, NULL);
1778
1779
1780 netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n",
1781 slave_dev->name,
1782 bond_is_active_slave(new_slave) ? "an active" : "a backup",
1783 new_slave->link != BOND_LINK_DOWN ? "an up" : "a down");
1784
1785
1786 bond_queue_slave_event(new_slave);
1787 return 0;
1788
1789
1790err_sysfs_del:
1791 bond_sysfs_slave_del(new_slave);
1792
1793err_upper_unlink:
1794 bond_upper_dev_unlink(bond, new_slave);
1795
1796err_unregister:
1797 netdev_rx_handler_unregister(slave_dev);
1798
1799err_detach:
1800 vlan_vids_del_by_dev(slave_dev, bond_dev);
1801 if (rcu_access_pointer(bond->primary_slave) == new_slave)
1802 RCU_INIT_POINTER(bond->primary_slave, NULL);
1803 if (rcu_access_pointer(bond->curr_active_slave) == new_slave) {
1804 block_netpoll_tx();
1805 bond_change_active_slave(bond, NULL);
1806 bond_select_active_slave(bond);
1807 unblock_netpoll_tx();
1808 }
1809
1810 synchronize_rcu();
1811 slave_disable_netpoll(new_slave);
1812
1813err_close:
1814 slave_dev->priv_flags &= ~IFF_BONDING;
1815 dev_close(slave_dev);
1816
1817err_restore_mac:
1818 slave_dev->flags &= ~IFF_SLAVE;
1819 if (!bond->params.fail_over_mac ||
1820 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
1821
1822
1823
1824
1825 bond_hw_addr_copy(ss.__data, new_slave->perm_hwaddr,
1826 new_slave->dev->addr_len);
1827 ss.ss_family = slave_dev->type;
1828 dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, NULL);
1829 }
1830
1831err_restore_mtu:
1832 dev_set_mtu(slave_dev, new_slave->original_mtu);
1833
1834err_free:
1835 bond_free_slave(new_slave);
1836
1837err_undo_flags:
1838
1839 if (!bond_has_slaves(bond)) {
1840 if (ether_addr_equal_64bits(bond_dev->dev_addr,
1841 slave_dev->dev_addr))
1842 eth_hw_addr_random(bond_dev);
1843 if (bond_dev->type != ARPHRD_ETHER) {
1844 dev_close(bond_dev);
1845 ether_setup(bond_dev);
1846 bond_dev->flags |= IFF_MASTER;
1847 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1848 }
1849 }
1850
1851 return res;
1852}
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865static int __bond_release_one(struct net_device *bond_dev,
1866 struct net_device *slave_dev,
1867 bool all, bool unregister)
1868{
1869 struct bonding *bond = netdev_priv(bond_dev);
1870 struct slave *slave, *oldcurrent;
1871 struct sockaddr_storage ss;
1872 int old_flags = bond_dev->flags;
1873 netdev_features_t old_features = bond_dev->features;
1874
1875
1876 if (!(slave_dev->flags & IFF_SLAVE) ||
1877 !netdev_has_upper_dev(slave_dev, bond_dev)) {
1878 netdev_dbg(bond_dev, "cannot release %s\n",
1879 slave_dev->name);
1880 return -EINVAL;
1881 }
1882
1883 block_netpoll_tx();
1884
1885 slave = bond_get_slave_by_dev(bond, slave_dev);
1886 if (!slave) {
1887
1888 netdev_info(bond_dev, "%s not enslaved\n",
1889 slave_dev->name);
1890 unblock_netpoll_tx();
1891 return -EINVAL;
1892 }
1893
1894 bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW);
1895
1896 bond_sysfs_slave_del(slave);
1897
1898
1899 bond_get_stats(bond->dev, &bond->bond_stats);
1900
1901 bond_upper_dev_unlink(bond, slave);
1902
1903
1904
1905 netdev_rx_handler_unregister(slave_dev);
1906
1907 if (BOND_MODE(bond) == BOND_MODE_8023AD)
1908 bond_3ad_unbind_slave(slave);
1909
1910 if (bond_mode_can_use_xmit_hash(bond))
1911 bond_update_slave_arr(bond, slave);
1912
1913 netdev_info(bond_dev, "Releasing %s interface %s\n",
1914 bond_is_active_slave(slave) ? "active" : "backup",
1915 slave_dev->name);
1916
1917 oldcurrent = rcu_access_pointer(bond->curr_active_slave);
1918
1919 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
1920
1921 if (!all && (!bond->params.fail_over_mac ||
1922 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) {
1923 if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) &&
1924 bond_has_slaves(bond))
1925 netdev_warn(bond_dev, "the permanent HWaddr of %s - %pM - is still in use by %s - set the HWaddr of %s to a different address to avoid conflicts\n",
1926 slave_dev->name, slave->perm_hwaddr,
1927 bond_dev->name, slave_dev->name);
1928 }
1929
1930 if (rtnl_dereference(bond->primary_slave) == slave)
1931 RCU_INIT_POINTER(bond->primary_slave, NULL);
1932
1933 if (oldcurrent == slave)
1934 bond_change_active_slave(bond, NULL);
1935
1936 if (bond_is_lb(bond)) {
1937
1938
1939
1940
1941
1942 bond_alb_deinit_slave(bond, slave);
1943 }
1944
1945 if (all) {
1946 RCU_INIT_POINTER(bond->curr_active_slave, NULL);
1947 } else if (oldcurrent == slave) {
1948
1949
1950
1951
1952 bond_select_active_slave(bond);
1953 }
1954
1955 if (!bond_has_slaves(bond)) {
1956 bond_set_carrier(bond);
1957 eth_hw_addr_random(bond_dev);
1958 bond->nest_level = SINGLE_DEPTH_NESTING;
1959 } else {
1960 bond->nest_level = dev_get_nest_level(bond_dev) + 1;
1961 }
1962
1963 unblock_netpoll_tx();
1964 synchronize_rcu();
1965 bond->slave_cnt--;
1966
1967 if (!bond_has_slaves(bond)) {
1968 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev);
1969 call_netdevice_notifiers(NETDEV_RELEASE, bond->dev);
1970 }
1971
1972 bond_compute_features(bond);
1973 if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1974 (old_features & NETIF_F_VLAN_CHALLENGED))
1975 netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n",
1976 slave_dev->name, bond_dev->name);
1977
1978 vlan_vids_del_by_dev(slave_dev, bond_dev);
1979
1980
1981
1982
1983 if (!bond_uses_primary(bond)) {
1984
1985
1986
1987
1988
1989
1990
1991 if (old_flags & IFF_PROMISC)
1992 dev_set_promiscuity(slave_dev, -1);
1993
1994
1995 if (old_flags & IFF_ALLMULTI)
1996 dev_set_allmulti(slave_dev, -1);
1997
1998 bond_hw_addr_flush(bond_dev, slave_dev);
1999 }
2000
2001 slave_disable_netpoll(slave);
2002
2003
2004 dev_close(slave_dev);
2005
2006 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE ||
2007 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
2008
2009 bond_hw_addr_copy(ss.__data, slave->perm_hwaddr,
2010 slave->dev->addr_len);
2011 ss.ss_family = slave_dev->type;
2012 dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, NULL);
2013 }
2014
2015 if (unregister)
2016 __dev_set_mtu(slave_dev, slave->original_mtu);
2017 else
2018 dev_set_mtu(slave_dev, slave->original_mtu);
2019
2020 slave_dev->priv_flags &= ~IFF_BONDING;
2021
2022 bond_free_slave(slave);
2023
2024 return 0;
2025}
2026
2027
2028int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
2029{
2030 return __bond_release_one(bond_dev, slave_dev, false, false);
2031}
2032
2033
2034
2035
2036static int bond_release_and_destroy(struct net_device *bond_dev,
2037 struct net_device *slave_dev)
2038{
2039 struct bonding *bond = netdev_priv(bond_dev);
2040 int ret;
2041
2042 ret = __bond_release_one(bond_dev, slave_dev, false, true);
2043 if (ret == 0 && !bond_has_slaves(bond)) {
2044 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
2045 netdev_info(bond_dev, "Destroying bond %s\n",
2046 bond_dev->name);
2047 bond_remove_proc_entry(bond);
2048 unregister_netdevice(bond_dev);
2049 }
2050 return ret;
2051}
2052
2053static void bond_info_query(struct net_device *bond_dev, struct ifbond *info)
2054{
2055 struct bonding *bond = netdev_priv(bond_dev);
2056 bond_fill_ifbond(bond, info);
2057}
2058
2059static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
2060{
2061 struct bonding *bond = netdev_priv(bond_dev);
2062 struct list_head *iter;
2063 int i = 0, res = -ENODEV;
2064 struct slave *slave;
2065
2066 bond_for_each_slave(bond, slave, iter) {
2067 if (i++ == (int)info->slave_id) {
2068 res = 0;
2069 bond_fill_ifslave(slave, info);
2070 break;
2071 }
2072 }
2073
2074 return res;
2075}
2076
2077
2078
2079
2080static int bond_miimon_inspect(struct bonding *bond)
2081{
2082 int link_state, commit = 0;
2083 struct list_head *iter;
2084 struct slave *slave;
2085 bool ignore_updelay;
2086
2087 ignore_updelay = !rcu_dereference(bond->curr_active_slave);
2088
2089 bond_for_each_slave_rcu(bond, slave, iter) {
2090 slave->new_link = BOND_LINK_NOCHANGE;
2091 slave->link_new_state = slave->link;
2092
2093 link_state = bond_check_dev_link(bond, slave->dev, 0);
2094
2095 switch (slave->link) {
2096 case BOND_LINK_UP:
2097 if (link_state)
2098 continue;
2099
2100 bond_propose_link_state(slave, BOND_LINK_FAIL);
2101 commit++;
2102 slave->delay = bond->params.downdelay;
2103 if (slave->delay) {
2104 netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n",
2105 (BOND_MODE(bond) ==
2106 BOND_MODE_ACTIVEBACKUP) ?
2107 (bond_is_active_slave(slave) ?
2108 "active " : "backup ") : "",
2109 slave->dev->name,
2110 bond->params.downdelay * bond->params.miimon);
2111 }
2112
2113 case BOND_LINK_FAIL:
2114 if (link_state) {
2115
2116 bond_propose_link_state(slave, BOND_LINK_UP);
2117 slave->last_link_up = jiffies;
2118 netdev_info(bond->dev, "link status up again after %d ms for interface %s\n",
2119 (bond->params.downdelay - slave->delay) *
2120 bond->params.miimon,
2121 slave->dev->name);
2122 commit++;
2123 continue;
2124 }
2125
2126 if (slave->delay <= 0) {
2127 slave->new_link = BOND_LINK_DOWN;
2128 commit++;
2129 continue;
2130 }
2131
2132 slave->delay--;
2133 break;
2134
2135 case BOND_LINK_DOWN:
2136 if (!link_state)
2137 continue;
2138
2139 bond_propose_link_state(slave, BOND_LINK_BACK);
2140 commit++;
2141 slave->delay = bond->params.updelay;
2142
2143 if (slave->delay) {
2144 netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n",
2145 slave->dev->name,
2146 ignore_updelay ? 0 :
2147 bond->params.updelay *
2148 bond->params.miimon);
2149 }
2150
2151 case BOND_LINK_BACK:
2152 if (!link_state) {
2153 bond_propose_link_state(slave, BOND_LINK_DOWN);
2154 netdev_info(bond->dev, "link status down again after %d ms for interface %s\n",
2155 (bond->params.updelay - slave->delay) *
2156 bond->params.miimon,
2157 slave->dev->name);
2158 commit++;
2159 continue;
2160 }
2161
2162 if (ignore_updelay)
2163 slave->delay = 0;
2164
2165 if (slave->delay <= 0) {
2166 slave->new_link = BOND_LINK_UP;
2167 commit++;
2168 ignore_updelay = false;
2169 continue;
2170 }
2171
2172 slave->delay--;
2173 break;
2174 }
2175 }
2176
2177 return commit;
2178}
2179
2180static void bond_miimon_link_change(struct bonding *bond,
2181 struct slave *slave,
2182 char link)
2183{
2184 switch (BOND_MODE(bond)) {
2185 case BOND_MODE_8023AD:
2186 bond_3ad_handle_link_change(slave, link);
2187 break;
2188 case BOND_MODE_TLB:
2189 case BOND_MODE_ALB:
2190 bond_alb_handle_link_change(bond, slave, link);
2191 break;
2192 case BOND_MODE_XOR:
2193 bond_update_slave_arr(bond, NULL);
2194 break;
2195 }
2196}
2197
2198static void bond_miimon_commit(struct bonding *bond)
2199{
2200 struct list_head *iter;
2201 struct slave *slave, *primary;
2202
2203 bond_for_each_slave(bond, slave, iter) {
2204 switch (slave->new_link) {
2205 case BOND_LINK_NOCHANGE:
2206 continue;
2207
2208 case BOND_LINK_UP:
2209 if (bond_update_speed_duplex(slave) &&
2210 bond_needs_speed_duplex(bond)) {
2211 slave->link = BOND_LINK_DOWN;
2212 if (net_ratelimit())
2213 netdev_warn(bond->dev,
2214 "failed to get link speed/duplex for %s\n",
2215 slave->dev->name);
2216 continue;
2217 }
2218 bond_set_slave_link_state(slave, BOND_LINK_UP,
2219 BOND_SLAVE_NOTIFY_NOW);
2220 slave->last_link_up = jiffies;
2221
2222 primary = rtnl_dereference(bond->primary_slave);
2223 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
2224
2225 bond_set_backup_slave(slave);
2226 } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
2227
2228 bond_set_active_slave(slave);
2229 } else if (slave != primary) {
2230
2231 bond_set_backup_slave(slave);
2232 }
2233
2234 netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n",
2235 slave->dev->name,
2236 slave->speed == SPEED_UNKNOWN ? 0 : slave->speed,
2237 slave->duplex ? "full" : "half");
2238
2239 bond_miimon_link_change(bond, slave, BOND_LINK_UP);
2240
2241 if (!bond->curr_active_slave || slave == primary)
2242 goto do_failover;
2243
2244 continue;
2245
2246 case BOND_LINK_DOWN:
2247 if (slave->link_failure_count < UINT_MAX)
2248 slave->link_failure_count++;
2249
2250 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2251 BOND_SLAVE_NOTIFY_NOW);
2252
2253 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP ||
2254 BOND_MODE(bond) == BOND_MODE_8023AD)
2255 bond_set_slave_inactive_flags(slave,
2256 BOND_SLAVE_NOTIFY_NOW);
2257
2258 netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n",
2259 slave->dev->name);
2260
2261 bond_miimon_link_change(bond, slave, BOND_LINK_DOWN);
2262
2263 if (slave == rcu_access_pointer(bond->curr_active_slave))
2264 goto do_failover;
2265
2266 continue;
2267
2268 default:
2269 netdev_err(bond->dev, "invalid new link %d on slave %s\n",
2270 slave->new_link, slave->dev->name);
2271 slave->new_link = BOND_LINK_NOCHANGE;
2272
2273 continue;
2274 }
2275
2276do_failover:
2277 block_netpoll_tx();
2278 bond_select_active_slave(bond);
2279 unblock_netpoll_tx();
2280 }
2281
2282 bond_set_carrier(bond);
2283}
2284
2285
2286
2287
2288
2289
2290
2291
2292static void bond_mii_monitor(struct work_struct *work)
2293{
2294 struct bonding *bond = container_of(work, struct bonding,
2295 mii_work.work);
2296 bool should_notify_peers = false;
2297 unsigned long delay;
2298 struct slave *slave;
2299 struct list_head *iter;
2300
2301 delay = msecs_to_jiffies(bond->params.miimon);
2302
2303 if (!bond_has_slaves(bond))
2304 goto re_arm;
2305
2306 rcu_read_lock();
2307
2308 should_notify_peers = bond_should_notify_peers(bond);
2309
2310 if (bond_miimon_inspect(bond)) {
2311 rcu_read_unlock();
2312
2313
2314 if (!rtnl_trylock()) {
2315 delay = 1;
2316 should_notify_peers = false;
2317 goto re_arm;
2318 }
2319
2320 bond_for_each_slave(bond, slave, iter) {
2321 bond_commit_link_state(slave, BOND_SLAVE_NOTIFY_LATER);
2322 }
2323 bond_miimon_commit(bond);
2324
2325 rtnl_unlock();
2326 } else
2327 rcu_read_unlock();
2328
2329re_arm:
2330 if (bond->params.miimon)
2331 queue_delayed_work(bond->wq, &bond->mii_work, delay);
2332
2333 if (should_notify_peers) {
2334 if (!rtnl_trylock())
2335 return;
2336 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2337 rtnl_unlock();
2338 }
2339}
2340
2341static int bond_upper_dev_walk(struct net_device *upper, void *data)
2342{
2343 __be32 ip = *((__be32 *)data);
2344
2345 return ip == bond_confirm_addr(upper, 0, ip);
2346}
2347
2348static bool bond_has_this_ip(struct bonding *bond, __be32 ip)
2349{
2350 bool ret = false;
2351
2352 if (ip == bond_confirm_addr(bond->dev, 0, ip))
2353 return true;
2354
2355 rcu_read_lock();
2356 if (netdev_walk_all_upper_dev_rcu(bond->dev, bond_upper_dev_walk, &ip))
2357 ret = true;
2358 rcu_read_unlock();
2359
2360 return ret;
2361}
2362
2363
2364
2365
2366
2367static void bond_arp_send(struct net_device *slave_dev, int arp_op,
2368 __be32 dest_ip, __be32 src_ip,
2369 struct bond_vlan_tag *tags)
2370{
2371 struct sk_buff *skb;
2372 struct bond_vlan_tag *outer_tag = tags;
2373
2374 netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n",
2375 arp_op, slave_dev->name, &dest_ip, &src_ip);
2376
2377 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2378 NULL, slave_dev->dev_addr, NULL);
2379
2380 if (!skb) {
2381 net_err_ratelimited("ARP packet allocation failed\n");
2382 return;
2383 }
2384
2385 if (!tags || tags->vlan_proto == VLAN_N_VID)
2386 goto xmit;
2387
2388 tags++;
2389
2390
2391 while (tags->vlan_proto != VLAN_N_VID) {
2392 if (!tags->vlan_id) {
2393 tags++;
2394 continue;
2395 }
2396
2397 netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n",
2398 ntohs(outer_tag->vlan_proto), tags->vlan_id);
2399 skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto,
2400 tags->vlan_id);
2401 if (!skb) {
2402 net_err_ratelimited("failed to insert inner VLAN tag\n");
2403 return;
2404 }
2405
2406 tags++;
2407 }
2408
2409 if (outer_tag->vlan_id) {
2410 netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n",
2411 ntohs(outer_tag->vlan_proto), outer_tag->vlan_id);
2412 __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto,
2413 outer_tag->vlan_id);
2414 }
2415
2416xmit:
2417 arp_xmit(skb);
2418}
2419
2420
2421
2422
2423
2424
2425
2426struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev,
2427 struct net_device *end_dev,
2428 int level)
2429{
2430 struct bond_vlan_tag *tags;
2431 struct net_device *upper;
2432 struct list_head *iter;
2433
2434 if (start_dev == end_dev) {
2435 tags = kcalloc(level + 1, sizeof(*tags), GFP_ATOMIC);
2436 if (!tags)
2437 return ERR_PTR(-ENOMEM);
2438 tags[level].vlan_proto = VLAN_N_VID;
2439 return tags;
2440 }
2441
2442 netdev_for_each_upper_dev_rcu(start_dev, upper, iter) {
2443 tags = bond_verify_device_path(upper, end_dev, level + 1);
2444 if (IS_ERR_OR_NULL(tags)) {
2445 if (IS_ERR(tags))
2446 return tags;
2447 continue;
2448 }
2449 if (is_vlan_dev(upper)) {
2450 tags[level].vlan_proto = vlan_dev_vlan_proto(upper);
2451 tags[level].vlan_id = vlan_dev_vlan_id(upper);
2452 }
2453
2454 return tags;
2455 }
2456
2457 return NULL;
2458}
2459
2460static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2461{
2462 struct rtable *rt;
2463 struct bond_vlan_tag *tags;
2464 __be32 *targets = bond->params.arp_targets, addr;
2465 int i;
2466
2467 for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) {
2468 netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]);
2469 tags = NULL;
2470
2471
2472 rt = ip_route_output(dev_net(bond->dev), targets[i], 0,
2473 RTO_ONLINK, 0);
2474 if (IS_ERR(rt)) {
2475
2476
2477
2478 if (bond->params.arp_validate)
2479 net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n",
2480 bond->dev->name,
2481 &targets[i]);
2482 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2483 0, tags);
2484 continue;
2485 }
2486
2487
2488 if (rt->dst.dev == bond->dev)
2489 goto found;
2490
2491 rcu_read_lock();
2492 tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0);
2493 rcu_read_unlock();
2494
2495 if (!IS_ERR_OR_NULL(tags))
2496 goto found;
2497
2498
2499 netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n",
2500 &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL");
2501
2502 ip_rt_put(rt);
2503 continue;
2504
2505found:
2506 addr = bond_confirm_addr(rt->dst.dev, targets[i], 0);
2507 ip_rt_put(rt);
2508 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2509 addr, tags);
2510 kfree(tags);
2511 }
2512}
2513
2514static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
2515{
2516 int i;
2517
2518 if (!sip || !bond_has_this_ip(bond, tip)) {
2519 netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n",
2520 &sip, &tip);
2521 return;
2522 }
2523
2524 i = bond_get_targets_ip(bond->params.arp_targets, sip);
2525 if (i == -1) {
2526 netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n",
2527 &sip);
2528 return;
2529 }
2530 slave->last_rx = jiffies;
2531 slave->target_last_arp_rx[i] = jiffies;
2532}
2533
2534int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
2535 struct slave *slave)
2536{
2537 struct arphdr *arp = (struct arphdr *)skb->data;
2538 struct slave *curr_active_slave, *curr_arp_slave;
2539 unsigned char *arp_ptr;
2540 __be32 sip, tip;
2541 int is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
2542 unsigned int alen;
2543
2544 if (!slave_do_arp_validate(bond, slave)) {
2545 if ((slave_do_arp_validate_only(bond) && is_arp) ||
2546 !slave_do_arp_validate_only(bond))
2547 slave->last_rx = jiffies;
2548 return RX_HANDLER_ANOTHER;
2549 } else if (!is_arp) {
2550 return RX_HANDLER_ANOTHER;
2551 }
2552
2553 alen = arp_hdr_len(bond->dev);
2554
2555 netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n",
2556 skb->dev->name);
2557
2558 if (alen > skb_headlen(skb)) {
2559 arp = kmalloc(alen, GFP_ATOMIC);
2560 if (!arp)
2561 goto out_unlock;
2562 if (skb_copy_bits(skb, 0, arp, alen) < 0)
2563 goto out_unlock;
2564 }
2565
2566 if (arp->ar_hln != bond->dev->addr_len ||
2567 skb->pkt_type == PACKET_OTHERHOST ||
2568 skb->pkt_type == PACKET_LOOPBACK ||
2569 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2570 arp->ar_pro != htons(ETH_P_IP) ||
2571 arp->ar_pln != 4)
2572 goto out_unlock;
2573
2574 arp_ptr = (unsigned char *)(arp + 1);
2575 arp_ptr += bond->dev->addr_len;
2576 memcpy(&sip, arp_ptr, 4);
2577 arp_ptr += 4 + bond->dev->addr_len;
2578 memcpy(&tip, arp_ptr, 4);
2579
2580 netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n",
2581 slave->dev->name, bond_slave_state(slave),
2582 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2583 &sip, &tip);
2584
2585 curr_active_slave = rcu_dereference(bond->curr_active_slave);
2586 curr_arp_slave = rcu_dereference(bond->current_arp_slave);
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611 if (bond_is_active_slave(slave))
2612 bond_validate_arp(bond, slave, sip, tip);
2613 else if (curr_active_slave &&
2614 time_after(slave_last_rx(bond, curr_active_slave),
2615 curr_active_slave->last_link_up))
2616 bond_validate_arp(bond, slave, tip, sip);
2617 else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) &&
2618 bond_time_in_interval(bond,
2619 dev_trans_start(curr_arp_slave->dev), 1))
2620 bond_validate_arp(bond, slave, sip, tip);
2621
2622out_unlock:
2623 if (arp != (struct arphdr *)skb->data)
2624 kfree(arp);
2625 return RX_HANDLER_ANOTHER;
2626}
2627
2628
2629
2630
2631
2632static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
2633 int mod)
2634{
2635 int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2636
2637 return time_in_range(jiffies,
2638 last_act - delta_in_ticks,
2639 last_act + mod * delta_in_ticks + delta_in_ticks/2);
2640}
2641
2642
2643
2644
2645
2646
2647
2648static void bond_loadbalance_arp_mon(struct bonding *bond)
2649{
2650 struct slave *slave, *oldcurrent;
2651 struct list_head *iter;
2652 int do_failover = 0, slave_state_changed = 0;
2653
2654 if (!bond_has_slaves(bond))
2655 goto re_arm;
2656
2657 rcu_read_lock();
2658
2659 oldcurrent = rcu_dereference(bond->curr_active_slave);
2660
2661
2662
2663
2664
2665
2666
2667
2668 bond_for_each_slave_rcu(bond, slave, iter) {
2669 unsigned long trans_start = dev_trans_start(slave->dev);
2670
2671 slave->new_link = BOND_LINK_NOCHANGE;
2672
2673 if (slave->link != BOND_LINK_UP) {
2674 if (bond_time_in_interval(bond, trans_start, 1) &&
2675 bond_time_in_interval(bond, slave->last_rx, 1)) {
2676
2677 slave->new_link = BOND_LINK_UP;
2678 slave_state_changed = 1;
2679
2680
2681
2682
2683
2684
2685 if (!oldcurrent) {
2686 netdev_info(bond->dev, "link status definitely up for interface %s\n",
2687 slave->dev->name);
2688 do_failover = 1;
2689 } else {
2690 netdev_info(bond->dev, "interface %s is now up\n",
2691 slave->dev->name);
2692 }
2693 }
2694 } else {
2695
2696
2697
2698
2699
2700
2701 if (!bond_time_in_interval(bond, trans_start, 2) ||
2702 !bond_time_in_interval(bond, slave->last_rx, 2)) {
2703
2704 slave->new_link = BOND_LINK_DOWN;
2705 slave_state_changed = 1;
2706
2707 if (slave->link_failure_count < UINT_MAX)
2708 slave->link_failure_count++;
2709
2710 netdev_info(bond->dev, "interface %s is now down\n",
2711 slave->dev->name);
2712
2713 if (slave == oldcurrent)
2714 do_failover = 1;
2715 }
2716 }
2717
2718
2719
2720
2721
2722
2723
2724
2725 if (bond_slave_is_up(slave))
2726 bond_arp_send_all(bond, slave);
2727 }
2728
2729 rcu_read_unlock();
2730
2731 if (do_failover || slave_state_changed) {
2732 if (!rtnl_trylock())
2733 goto re_arm;
2734
2735 bond_for_each_slave(bond, slave, iter) {
2736 if (slave->new_link != BOND_LINK_NOCHANGE)
2737 slave->link = slave->new_link;
2738 }
2739
2740 if (slave_state_changed) {
2741 bond_slave_state_change(bond);
2742 if (BOND_MODE(bond) == BOND_MODE_XOR)
2743 bond_update_slave_arr(bond, NULL);
2744 }
2745 if (do_failover) {
2746 block_netpoll_tx();
2747 bond_select_active_slave(bond);
2748 unblock_netpoll_tx();
2749 }
2750 rtnl_unlock();
2751 }
2752
2753re_arm:
2754 if (bond->params.arp_interval)
2755 queue_delayed_work(bond->wq, &bond->arp_work,
2756 msecs_to_jiffies(bond->params.arp_interval));
2757}
2758
2759
2760
2761
2762
2763
2764
2765
2766static int bond_ab_arp_inspect(struct bonding *bond)
2767{
2768 unsigned long trans_start, last_rx;
2769 struct list_head *iter;
2770 struct slave *slave;
2771 int commit = 0;
2772
2773 bond_for_each_slave_rcu(bond, slave, iter) {
2774 slave->new_link = BOND_LINK_NOCHANGE;
2775 last_rx = slave_last_rx(bond, slave);
2776
2777 if (slave->link != BOND_LINK_UP) {
2778 if (bond_time_in_interval(bond, last_rx, 1)) {
2779 slave->new_link = BOND_LINK_UP;
2780 commit++;
2781 }
2782 continue;
2783 }
2784
2785
2786
2787
2788
2789 if (bond_time_in_interval(bond, slave->last_link_up, 2))
2790 continue;
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804 if (!bond_is_active_slave(slave) &&
2805 !rcu_access_pointer(bond->current_arp_slave) &&
2806 !bond_time_in_interval(bond, last_rx, 3)) {
2807 slave->new_link = BOND_LINK_DOWN;
2808 commit++;
2809 }
2810
2811
2812
2813
2814
2815
2816 trans_start = dev_trans_start(slave->dev);
2817 if (bond_is_active_slave(slave) &&
2818 (!bond_time_in_interval(bond, trans_start, 2) ||
2819 !bond_time_in_interval(bond, last_rx, 2))) {
2820 slave->new_link = BOND_LINK_DOWN;
2821 commit++;
2822 }
2823 }
2824
2825 return commit;
2826}
2827
2828
2829
2830
2831
2832
2833static void bond_ab_arp_commit(struct bonding *bond)
2834{
2835 unsigned long trans_start;
2836 struct list_head *iter;
2837 struct slave *slave;
2838
2839 bond_for_each_slave(bond, slave, iter) {
2840 switch (slave->new_link) {
2841 case BOND_LINK_NOCHANGE:
2842 continue;
2843
2844 case BOND_LINK_UP:
2845 trans_start = dev_trans_start(slave->dev);
2846 if (rtnl_dereference(bond->curr_active_slave) != slave ||
2847 (!rtnl_dereference(bond->curr_active_slave) &&
2848 bond_time_in_interval(bond, trans_start, 1))) {
2849 struct slave *current_arp_slave;
2850
2851 current_arp_slave = rtnl_dereference(bond->current_arp_slave);
2852 bond_set_slave_link_state(slave, BOND_LINK_UP,
2853 BOND_SLAVE_NOTIFY_NOW);
2854 if (current_arp_slave) {
2855 bond_set_slave_inactive_flags(
2856 current_arp_slave,
2857 BOND_SLAVE_NOTIFY_NOW);
2858 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
2859 }
2860
2861 netdev_info(bond->dev, "link status definitely up for interface %s\n",
2862 slave->dev->name);
2863
2864 if (!rtnl_dereference(bond->curr_active_slave) ||
2865 slave == rtnl_dereference(bond->primary_slave))
2866 goto do_failover;
2867
2868 }
2869
2870 continue;
2871
2872 case BOND_LINK_DOWN:
2873 if (slave->link_failure_count < UINT_MAX)
2874 slave->link_failure_count++;
2875
2876 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2877 BOND_SLAVE_NOTIFY_NOW);
2878 bond_set_slave_inactive_flags(slave,
2879 BOND_SLAVE_NOTIFY_NOW);
2880
2881 netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n",
2882 slave->dev->name);
2883
2884 if (slave == rtnl_dereference(bond->curr_active_slave)) {
2885 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
2886 goto do_failover;
2887 }
2888
2889 continue;
2890
2891 default:
2892 netdev_err(bond->dev, "impossible: new_link %d on slave %s\n",
2893 slave->new_link, slave->dev->name);
2894 continue;
2895 }
2896
2897do_failover:
2898 block_netpoll_tx();
2899 bond_select_active_slave(bond);
2900 unblock_netpoll_tx();
2901 }
2902
2903 bond_set_carrier(bond);
2904}
2905
2906
2907
2908
2909
2910static bool bond_ab_arp_probe(struct bonding *bond)
2911{
2912 struct slave *slave, *before = NULL, *new_slave = NULL,
2913 *curr_arp_slave = rcu_dereference(bond->current_arp_slave),
2914 *curr_active_slave = rcu_dereference(bond->curr_active_slave);
2915 struct list_head *iter;
2916 bool found = false;
2917 bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER;
2918
2919 if (curr_arp_slave && curr_active_slave)
2920 netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n",
2921 curr_arp_slave->dev->name,
2922 curr_active_slave->dev->name);
2923
2924 if (curr_active_slave) {
2925 bond_arp_send_all(bond, curr_active_slave);
2926 return should_notify_rtnl;
2927 }
2928
2929
2930
2931
2932
2933
2934 if (!curr_arp_slave) {
2935 curr_arp_slave = bond_first_slave_rcu(bond);
2936 if (!curr_arp_slave)
2937 return should_notify_rtnl;
2938 }
2939
2940 bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER);
2941
2942 bond_for_each_slave_rcu(bond, slave, iter) {
2943 if (!found && !before && bond_slave_is_up(slave))
2944 before = slave;
2945
2946 if (found && !new_slave && bond_slave_is_up(slave))
2947 new_slave = slave;
2948
2949
2950
2951
2952
2953
2954
2955 if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
2956 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2957 BOND_SLAVE_NOTIFY_LATER);
2958 if (slave->link_failure_count < UINT_MAX)
2959 slave->link_failure_count++;
2960
2961 bond_set_slave_inactive_flags(slave,
2962 BOND_SLAVE_NOTIFY_LATER);
2963
2964 netdev_info(bond->dev, "backup interface %s is now down\n",
2965 slave->dev->name);
2966 }
2967 if (slave == curr_arp_slave)
2968 found = true;
2969 }
2970
2971 if (!new_slave && before)
2972 new_slave = before;
2973
2974 if (!new_slave)
2975 goto check_state;
2976
2977 bond_set_slave_link_state(new_slave, BOND_LINK_BACK,
2978 BOND_SLAVE_NOTIFY_LATER);
2979 bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER);
2980 bond_arp_send_all(bond, new_slave);
2981 new_slave->last_link_up = jiffies;
2982 rcu_assign_pointer(bond->current_arp_slave, new_slave);
2983
2984check_state:
2985 bond_for_each_slave_rcu(bond, slave, iter) {
2986 if (slave->should_notify || slave->should_notify_link) {
2987 should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW;
2988 break;
2989 }
2990 }
2991 return should_notify_rtnl;
2992}
2993
2994static void bond_activebackup_arp_mon(struct bonding *bond)
2995{
2996 bool should_notify_peers = false;
2997 bool should_notify_rtnl = false;
2998 int delta_in_ticks;
2999
3000 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
3001
3002 if (!bond_has_slaves(bond))
3003 goto re_arm;
3004
3005 rcu_read_lock();
3006
3007 should_notify_peers = bond_should_notify_peers(bond);
3008
3009 if (bond_ab_arp_inspect(bond)) {
3010 rcu_read_unlock();
3011
3012
3013 if (!rtnl_trylock()) {
3014 delta_in_ticks = 1;
3015 should_notify_peers = false;
3016 goto re_arm;
3017 }
3018
3019 bond_ab_arp_commit(bond);
3020
3021 rtnl_unlock();
3022 rcu_read_lock();
3023 }
3024
3025 should_notify_rtnl = bond_ab_arp_probe(bond);
3026 rcu_read_unlock();
3027
3028re_arm:
3029 if (bond->params.arp_interval)
3030 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
3031
3032 if (should_notify_peers || should_notify_rtnl) {
3033 if (!rtnl_trylock())
3034 return;
3035
3036 if (should_notify_peers)
3037 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
3038 bond->dev);
3039 if (should_notify_rtnl) {
3040 bond_slave_state_notify(bond);
3041 bond_slave_link_notify(bond);
3042 }
3043
3044 rtnl_unlock();
3045 }
3046}
3047
3048static void bond_arp_monitor(struct work_struct *work)
3049{
3050 struct bonding *bond = container_of(work, struct bonding,
3051 arp_work.work);
3052
3053 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
3054 bond_activebackup_arp_mon(bond);
3055 else
3056 bond_loadbalance_arp_mon(bond);
3057}
3058
3059
3060
3061
3062static int bond_event_changename(struct bonding *bond)
3063{
3064 bond_remove_proc_entry(bond);
3065 bond_create_proc_entry(bond);
3066
3067 bond_debug_reregister(bond);
3068
3069 return NOTIFY_DONE;
3070}
3071
3072static int bond_master_netdev_event(unsigned long event,
3073 struct net_device *bond_dev)
3074{
3075 struct bonding *event_bond = netdev_priv(bond_dev);
3076
3077 switch (event) {
3078 case NETDEV_CHANGENAME:
3079 return bond_event_changename(event_bond);
3080 case NETDEV_UNREGISTER:
3081 bond_remove_proc_entry(event_bond);
3082 break;
3083 case NETDEV_REGISTER:
3084 bond_create_proc_entry(event_bond);
3085 break;
3086 case NETDEV_NOTIFY_PEERS:
3087 if (event_bond->send_peer_notif)
3088 event_bond->send_peer_notif--;
3089 break;
3090 default:
3091 break;
3092 }
3093
3094 return NOTIFY_DONE;
3095}
3096
3097static int bond_slave_netdev_event(unsigned long event,
3098 struct net_device *slave_dev)
3099{
3100 struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary;
3101 struct bonding *bond;
3102 struct net_device *bond_dev;
3103
3104
3105
3106
3107
3108 if (!slave)
3109 return NOTIFY_DONE;
3110 bond_dev = slave->bond->dev;
3111 bond = slave->bond;
3112 primary = rtnl_dereference(bond->primary_slave);
3113
3114 switch (event) {
3115 case NETDEV_UNREGISTER:
3116 if (bond_dev->type != ARPHRD_ETHER)
3117 bond_release_and_destroy(bond_dev, slave_dev);
3118 else
3119 __bond_release_one(bond_dev, slave_dev, false, true);
3120 break;
3121 case NETDEV_UP:
3122 case NETDEV_CHANGE:
3123
3124
3125
3126
3127
3128
3129
3130 if (bond_update_speed_duplex(slave) &&
3131 BOND_MODE(bond) == BOND_MODE_8023AD) {
3132 if (slave->last_link_up)
3133 slave->link = BOND_LINK_FAIL;
3134 else
3135 slave->link = BOND_LINK_DOWN;
3136 }
3137
3138 if (BOND_MODE(bond) == BOND_MODE_8023AD)
3139 bond_3ad_adapter_speed_duplex_changed(slave);
3140
3141 case NETDEV_DOWN:
3142
3143
3144
3145
3146
3147
3148
3149
3150 if (bond_mode_can_use_xmit_hash(bond))
3151 bond_update_slave_arr(bond, NULL);
3152 break;
3153 case NETDEV_CHANGEMTU:
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165 break;
3166 case NETDEV_CHANGENAME:
3167
3168 if (!bond_uses_primary(bond) ||
3169 !bond->params.primary[0])
3170 break;
3171
3172 if (slave == primary) {
3173
3174 RCU_INIT_POINTER(bond->primary_slave, NULL);
3175 } else if (!strcmp(slave_dev->name, bond->params.primary)) {
3176
3177 rcu_assign_pointer(bond->primary_slave, slave);
3178 } else {
3179 break;
3180 }
3181
3182 netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n",
3183 primary ? slave_dev->name : "none");
3184
3185 block_netpoll_tx();
3186 bond_select_active_slave(bond);
3187 unblock_netpoll_tx();
3188 break;
3189 case NETDEV_FEAT_CHANGE:
3190 bond_compute_features(bond);
3191 break;
3192 case NETDEV_RESEND_IGMP:
3193
3194 call_netdevice_notifiers(event, slave->bond->dev);
3195 break;
3196 default:
3197 break;
3198 }
3199
3200 return NOTIFY_DONE;
3201}
3202
3203
3204
3205
3206
3207
3208
3209
3210static int bond_netdev_event(struct notifier_block *this,
3211 unsigned long event, void *ptr)
3212{
3213 struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
3214
3215 netdev_dbg(event_dev, "event: %lx\n", event);
3216
3217 if (!(event_dev->priv_flags & IFF_BONDING))
3218 return NOTIFY_DONE;
3219
3220 if (event_dev->flags & IFF_MASTER) {
3221 int ret;
3222
3223 netdev_dbg(event_dev, "IFF_MASTER\n");
3224 ret = bond_master_netdev_event(event, event_dev);
3225 if (ret != NOTIFY_DONE)
3226 return ret;
3227 }
3228
3229 if (event_dev->flags & IFF_SLAVE) {
3230 netdev_dbg(event_dev, "IFF_SLAVE\n");
3231 return bond_slave_netdev_event(event, event_dev);
3232 }
3233
3234 return NOTIFY_DONE;
3235}
3236
3237static struct notifier_block bond_netdev_notifier = {
3238 .notifier_call = bond_netdev_event,
3239};
3240
3241
3242
3243
3244static inline u32 bond_eth_hash(struct sk_buff *skb)
3245{
3246 struct ethhdr *ep, hdr_tmp;
3247
3248 ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp);
3249 if (ep)
3250 return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto;
3251 return 0;
3252}
3253
3254
3255static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
3256 struct flow_keys *fk)
3257{
3258 const struct ipv6hdr *iph6;
3259 const struct iphdr *iph;
3260 int noff, proto = -1;
3261
3262 if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
3263 return skb_flow_dissect_flow_keys(skb, fk, 0);
3264
3265 fk->ports.ports = 0;
3266 noff = skb_network_offset(skb);
3267 if (skb->protocol == htons(ETH_P_IP)) {
3268 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
3269 return false;
3270 iph = ip_hdr(skb);
3271 iph_to_flow_copy_v4addrs(fk, iph);
3272 noff += iph->ihl << 2;
3273 if (!ip_is_fragment(iph))
3274 proto = iph->protocol;
3275 } else if (skb->protocol == htons(ETH_P_IPV6)) {
3276 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
3277 return false;
3278 iph6 = ipv6_hdr(skb);
3279 iph_to_flow_copy_v6addrs(fk, iph6);
3280 noff += sizeof(*iph6);
3281 proto = iph6->nexthdr;
3282 } else {
3283 return false;
3284 }
3285 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
3286 fk->ports.ports = skb_flow_get_ports(skb, noff, proto);
3287
3288 return true;
3289}
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb)
3300{
3301 struct flow_keys flow;
3302 u32 hash;
3303
3304 if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 &&
3305 skb->l4_hash)
3306 return skb->hash;
3307
3308 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 ||
3309 !bond_flow_dissect(bond, skb, &flow))
3310 return bond_eth_hash(skb);
3311
3312 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 ||
3313 bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
3314 hash = bond_eth_hash(skb);
3315 else
3316 hash = (__force u32)flow.ports.ports;
3317 hash ^= (__force u32)flow_get_u32_dst(&flow) ^
3318 (__force u32)flow_get_u32_src(&flow);
3319 hash ^= (hash >> 16);
3320 hash ^= (hash >> 8);
3321
3322 return hash >> 1;
3323}
3324
3325
3326
3327void bond_work_init_all(struct bonding *bond)
3328{
3329 INIT_DELAYED_WORK(&bond->mcast_work,
3330 bond_resend_igmp_join_requests_delayed);
3331 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3332 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
3333 INIT_DELAYED_WORK(&bond->arp_work, bond_arp_monitor);
3334 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
3335 INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler);
3336}
3337
3338static void bond_work_cancel_all(struct bonding *bond)
3339{
3340 cancel_delayed_work_sync(&bond->mii_work);
3341 cancel_delayed_work_sync(&bond->arp_work);
3342 cancel_delayed_work_sync(&bond->alb_work);
3343 cancel_delayed_work_sync(&bond->ad_work);
3344 cancel_delayed_work_sync(&bond->mcast_work);
3345 cancel_delayed_work_sync(&bond->slave_arr_work);
3346}
3347
3348static int bond_open(struct net_device *bond_dev)
3349{
3350 struct bonding *bond = netdev_priv(bond_dev);
3351 struct list_head *iter;
3352 struct slave *slave;
3353
3354
3355 if (bond_has_slaves(bond)) {
3356 bond_for_each_slave(bond, slave, iter) {
3357 if (bond_uses_primary(bond) &&
3358 slave != rcu_access_pointer(bond->curr_active_slave)) {
3359 bond_set_slave_inactive_flags(slave,
3360 BOND_SLAVE_NOTIFY_NOW);
3361 } else if (BOND_MODE(bond) != BOND_MODE_8023AD) {
3362 bond_set_slave_active_flags(slave,
3363 BOND_SLAVE_NOTIFY_NOW);
3364 }
3365 }
3366 }
3367
3368 if (bond_is_lb(bond)) {
3369
3370
3371
3372 if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB)))
3373 return -ENOMEM;
3374 if (bond->params.tlb_dynamic_lb || BOND_MODE(bond) == BOND_MODE_ALB)
3375 queue_delayed_work(bond->wq, &bond->alb_work, 0);
3376 }
3377
3378 if (bond->params.miimon)
3379 queue_delayed_work(bond->wq, &bond->mii_work, 0);
3380
3381 if (bond->params.arp_interval) {
3382 queue_delayed_work(bond->wq, &bond->arp_work, 0);
3383 bond->recv_probe = bond_arp_rcv;
3384 }
3385
3386 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3387 queue_delayed_work(bond->wq, &bond->ad_work, 0);
3388
3389 bond->recv_probe = bond_3ad_lacpdu_recv;
3390 bond_3ad_initiate_agg_selection(bond, 1);
3391 }
3392
3393 if (bond_mode_can_use_xmit_hash(bond))
3394 bond_update_slave_arr(bond, NULL);
3395
3396 return 0;
3397}
3398
3399static int bond_close(struct net_device *bond_dev)
3400{
3401 struct bonding *bond = netdev_priv(bond_dev);
3402
3403 bond_work_cancel_all(bond);
3404 bond->send_peer_notif = 0;
3405 if (bond_is_lb(bond))
3406 bond_alb_deinitialize(bond);
3407 bond->recv_probe = NULL;
3408
3409 return 0;
3410}
3411
3412
3413
3414
3415static void bond_fold_stats(struct rtnl_link_stats64 *_res,
3416 const struct rtnl_link_stats64 *_new,
3417 const struct rtnl_link_stats64 *_old)
3418{
3419 const u64 *new = (const u64 *)_new;
3420 const u64 *old = (const u64 *)_old;
3421 u64 *res = (u64 *)_res;
3422 int i;
3423
3424 for (i = 0; i < sizeof(*_res) / sizeof(u64); i++) {
3425 u64 nv = new[i];
3426 u64 ov = old[i];
3427 s64 delta = nv - ov;
3428
3429
3430 if (((nv | ov) >> 32) == 0)
3431 delta = (s64)(s32)((u32)nv - (u32)ov);
3432
3433
3434
3435
3436 if (delta > 0)
3437 res[i] += delta;
3438 }
3439}
3440
3441static int bond_get_nest_level(struct net_device *bond_dev)
3442{
3443 struct bonding *bond = netdev_priv(bond_dev);
3444
3445 return bond->nest_level;
3446}
3447
3448static void bond_get_stats(struct net_device *bond_dev,
3449 struct rtnl_link_stats64 *stats)
3450{
3451 struct bonding *bond = netdev_priv(bond_dev);
3452 struct rtnl_link_stats64 temp;
3453 struct list_head *iter;
3454 struct slave *slave;
3455
3456 spin_lock_nested(&bond->stats_lock, bond_get_nest_level(bond_dev));
3457 memcpy(stats, &bond->bond_stats, sizeof(*stats));
3458
3459 rcu_read_lock();
3460 bond_for_each_slave_rcu(bond, slave, iter) {
3461 const struct rtnl_link_stats64 *new =
3462 dev_get_stats(slave->dev, &temp);
3463
3464 bond_fold_stats(stats, new, &slave->slave_stats);
3465
3466
3467 memcpy(&slave->slave_stats, new, sizeof(*new));
3468 }
3469 rcu_read_unlock();
3470
3471 memcpy(&bond->bond_stats, stats, sizeof(*stats));
3472 spin_unlock(&bond->stats_lock);
3473}
3474
3475static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3476{
3477 struct bonding *bond = netdev_priv(bond_dev);
3478 struct net_device *slave_dev = NULL;
3479 struct ifbond k_binfo;
3480 struct ifbond __user *u_binfo = NULL;
3481 struct ifslave k_sinfo;
3482 struct ifslave __user *u_sinfo = NULL;
3483 struct mii_ioctl_data *mii = NULL;
3484 struct bond_opt_value newval;
3485 struct net *net;
3486 int res = 0;
3487
3488 netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd);
3489
3490 switch (cmd) {
3491 case SIOCGMIIPHY:
3492 mii = if_mii(ifr);
3493 if (!mii)
3494 return -EINVAL;
3495
3496 mii->phy_id = 0;
3497
3498 case SIOCGMIIREG:
3499
3500
3501
3502 mii = if_mii(ifr);
3503 if (!mii)
3504 return -EINVAL;
3505
3506 if (mii->reg_num == 1) {
3507 mii->val_out = 0;
3508 if (netif_carrier_ok(bond->dev))
3509 mii->val_out = BMSR_LSTATUS;
3510 }
3511
3512 return 0;
3513 case BOND_INFO_QUERY_OLD:
3514 case SIOCBONDINFOQUERY:
3515 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3516
3517 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
3518 return -EFAULT;
3519
3520 bond_info_query(bond_dev, &k_binfo);
3521 if (copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3522 return -EFAULT;
3523
3524 return 0;
3525 case BOND_SLAVE_INFO_QUERY_OLD:
3526 case SIOCBONDSLAVEINFOQUERY:
3527 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3528
3529 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
3530 return -EFAULT;
3531
3532 res = bond_slave_info_query(bond_dev, &k_sinfo);
3533 if (res == 0 &&
3534 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3535 return -EFAULT;
3536
3537 return res;
3538 default:
3539 break;
3540 }
3541
3542 net = dev_net(bond_dev);
3543
3544 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
3545 return -EPERM;
3546
3547 slave_dev = __dev_get_by_name(net, ifr->ifr_slave);
3548
3549 netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev);
3550
3551 if (!slave_dev)
3552 return -ENODEV;
3553
3554 netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name);
3555 switch (cmd) {
3556 case BOND_ENSLAVE_OLD:
3557 case SIOCBONDENSLAVE:
3558 res = bond_enslave(bond_dev, slave_dev, NULL);
3559 break;
3560 case BOND_RELEASE_OLD:
3561 case SIOCBONDRELEASE:
3562 res = bond_release(bond_dev, slave_dev);
3563 break;
3564 case BOND_SETHWADDR_OLD:
3565 case SIOCBONDSETHWADDR:
3566 res = bond_set_dev_addr(bond_dev, slave_dev);
3567 break;
3568 case BOND_CHANGE_ACTIVE_OLD:
3569 case SIOCBONDCHANGEACTIVE:
3570 bond_opt_initstr(&newval, slave_dev->name);
3571 res = __bond_opt_set_notify(bond, BOND_OPT_ACTIVE_SLAVE,
3572 &newval);
3573 break;
3574 default:
3575 res = -EOPNOTSUPP;
3576 }
3577
3578 return res;
3579}
3580
3581static void bond_change_rx_flags(struct net_device *bond_dev, int change)
3582{
3583 struct bonding *bond = netdev_priv(bond_dev);
3584
3585 if (change & IFF_PROMISC)
3586 bond_set_promiscuity(bond,
3587 bond_dev->flags & IFF_PROMISC ? 1 : -1);
3588
3589 if (change & IFF_ALLMULTI)
3590 bond_set_allmulti(bond,
3591 bond_dev->flags & IFF_ALLMULTI ? 1 : -1);
3592}
3593
3594static void bond_set_rx_mode(struct net_device *bond_dev)
3595{
3596 struct bonding *bond = netdev_priv(bond_dev);
3597 struct list_head *iter;
3598 struct slave *slave;
3599
3600 rcu_read_lock();
3601 if (bond_uses_primary(bond)) {
3602 slave = rcu_dereference(bond->curr_active_slave);
3603 if (slave) {
3604 dev_uc_sync(slave->dev, bond_dev);
3605 dev_mc_sync(slave->dev, bond_dev);
3606 }
3607 } else {
3608 bond_for_each_slave_rcu(bond, slave, iter) {
3609 dev_uc_sync_multiple(slave->dev, bond_dev);
3610 dev_mc_sync_multiple(slave->dev, bond_dev);
3611 }
3612 }
3613 rcu_read_unlock();
3614}
3615
3616static int bond_neigh_init(struct neighbour *n)
3617{
3618 struct bonding *bond = netdev_priv(n->dev);
3619 const struct net_device_ops *slave_ops;
3620 struct neigh_parms parms;
3621 struct slave *slave;
3622 int ret;
3623
3624 slave = bond_first_slave(bond);
3625 if (!slave)
3626 return 0;
3627 slave_ops = slave->dev->netdev_ops;
3628 if (!slave_ops->ndo_neigh_setup)
3629 return 0;
3630
3631 parms.neigh_setup = NULL;
3632 parms.neigh_cleanup = NULL;
3633 ret = slave_ops->ndo_neigh_setup(slave->dev, &parms);
3634 if (ret)
3635 return ret;
3636
3637
3638
3639
3640
3641
3642 n->parms->neigh_cleanup = parms.neigh_cleanup;
3643
3644 if (!parms.neigh_setup)
3645 return 0;
3646
3647 return parms.neigh_setup(n);
3648}
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658static int bond_neigh_setup(struct net_device *dev,
3659 struct neigh_parms *parms)
3660{
3661
3662 if (parms->dev == dev)
3663 parms->neigh_setup = bond_neigh_init;
3664
3665 return 0;
3666}
3667
3668
3669static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3670{
3671 struct bonding *bond = netdev_priv(bond_dev);
3672 struct slave *slave, *rollback_slave;
3673 struct list_head *iter;
3674 int res = 0;
3675
3676 netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu);
3677
3678 bond_for_each_slave(bond, slave, iter) {
3679 netdev_dbg(bond_dev, "s %p c_m %p\n",
3680 slave, slave->dev->netdev_ops->ndo_change_mtu);
3681
3682 res = dev_set_mtu(slave->dev, new_mtu);
3683
3684 if (res) {
3685
3686
3687
3688
3689
3690
3691
3692
3693 netdev_dbg(bond_dev, "err %d %s\n", res,
3694 slave->dev->name);
3695 goto unwind;
3696 }
3697 }
3698
3699 bond_dev->mtu = new_mtu;
3700
3701 return 0;
3702
3703unwind:
3704
3705 bond_for_each_slave(bond, rollback_slave, iter) {
3706 int tmp_res;
3707
3708 if (rollback_slave == slave)
3709 break;
3710
3711 tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu);
3712 if (tmp_res) {
3713 netdev_dbg(bond_dev, "unwind err %d dev %s\n",
3714 tmp_res, rollback_slave->dev->name);
3715 }
3716 }
3717
3718 return res;
3719}
3720
3721
3722
3723
3724
3725
3726
3727static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
3728{
3729 struct bonding *bond = netdev_priv(bond_dev);
3730 struct slave *slave, *rollback_slave;
3731 struct sockaddr_storage *ss = addr, tmp_ss;
3732 struct list_head *iter;
3733 int res = 0;
3734
3735 if (BOND_MODE(bond) == BOND_MODE_ALB)
3736 return bond_alb_set_mac_address(bond_dev, addr);
3737
3738
3739 netdev_dbg(bond_dev, "bond=%p\n", bond);
3740
3741
3742
3743
3744 if (bond->params.fail_over_mac &&
3745 BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
3746 return 0;
3747
3748 if (!is_valid_ether_addr(ss->__data))
3749 return -EADDRNOTAVAIL;
3750
3751 bond_for_each_slave(bond, slave, iter) {
3752 netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name);
3753 res = dev_set_mac_address(slave->dev, addr, NULL);
3754 if (res) {
3755
3756
3757
3758
3759
3760
3761 netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name);
3762 goto unwind;
3763 }
3764 }
3765
3766
3767 memcpy(bond_dev->dev_addr, ss->__data, bond_dev->addr_len);
3768 return 0;
3769
3770unwind:
3771 memcpy(tmp_ss.__data, bond_dev->dev_addr, bond_dev->addr_len);
3772 tmp_ss.ss_family = bond_dev->type;
3773
3774
3775 bond_for_each_slave(bond, rollback_slave, iter) {
3776 int tmp_res;
3777
3778 if (rollback_slave == slave)
3779 break;
3780
3781 tmp_res = dev_set_mac_address(rollback_slave->dev,
3782 (struct sockaddr *)&tmp_ss, NULL);
3783 if (tmp_res) {
3784 netdev_dbg(bond_dev, "unwind err %d dev %s\n",
3785 tmp_res, rollback_slave->dev->name);
3786 }
3787 }
3788
3789 return res;
3790}
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id)
3803{
3804 struct list_head *iter;
3805 struct slave *slave;
3806 int i = slave_id;
3807
3808
3809 bond_for_each_slave_rcu(bond, slave, iter) {
3810 if (--i < 0) {
3811 if (bond_slave_can_tx(slave)) {
3812 bond_dev_queue_xmit(bond, skb, slave->dev);
3813 return;
3814 }
3815 }
3816 }
3817
3818
3819 i = slave_id;
3820 bond_for_each_slave_rcu(bond, slave, iter) {
3821 if (--i < 0)
3822 break;
3823 if (bond_slave_can_tx(slave)) {
3824 bond_dev_queue_xmit(bond, skb, slave->dev);
3825 return;
3826 }
3827 }
3828
3829 bond_tx_drop(bond->dev, skb);
3830}
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840static u32 bond_rr_gen_slave_id(struct bonding *bond)
3841{
3842 u32 slave_id;
3843 struct reciprocal_value reciprocal_packets_per_slave;
3844 int packets_per_slave = bond->params.packets_per_slave;
3845
3846 switch (packets_per_slave) {
3847 case 0:
3848 slave_id = prandom_u32();
3849 break;
3850 case 1:
3851 slave_id = bond->rr_tx_counter;
3852 break;
3853 default:
3854 reciprocal_packets_per_slave =
3855 bond->params.reciprocal_packets_per_slave;
3856 slave_id = reciprocal_divide(bond->rr_tx_counter,
3857 reciprocal_packets_per_slave);
3858 break;
3859 }
3860 bond->rr_tx_counter++;
3861
3862 return slave_id;
3863}
3864
3865static netdev_tx_t bond_xmit_roundrobin(struct sk_buff *skb,
3866 struct net_device *bond_dev)
3867{
3868 struct bonding *bond = netdev_priv(bond_dev);
3869 struct iphdr *iph = ip_hdr(skb);
3870 struct slave *slave;
3871 u32 slave_id;
3872
3873
3874
3875
3876
3877
3878
3879 if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) {
3880 slave = rcu_dereference(bond->curr_active_slave);
3881 if (slave)
3882 bond_dev_queue_xmit(bond, skb, slave->dev);
3883 else
3884 bond_xmit_slave_id(bond, skb, 0);
3885 } else {
3886 int slave_cnt = READ_ONCE(bond->slave_cnt);
3887
3888 if (likely(slave_cnt)) {
3889 slave_id = bond_rr_gen_slave_id(bond);
3890 bond_xmit_slave_id(bond, skb, slave_id % slave_cnt);
3891 } else {
3892 bond_tx_drop(bond_dev, skb);
3893 }
3894 }
3895
3896 return NETDEV_TX_OK;
3897}
3898
3899
3900
3901
3902static netdev_tx_t bond_xmit_activebackup(struct sk_buff *skb,
3903 struct net_device *bond_dev)
3904{
3905 struct bonding *bond = netdev_priv(bond_dev);
3906 struct slave *slave;
3907
3908 slave = rcu_dereference(bond->curr_active_slave);
3909 if (slave)
3910 bond_dev_queue_xmit(bond, skb, slave->dev);
3911 else
3912 bond_tx_drop(bond_dev, skb);
3913
3914 return NETDEV_TX_OK;
3915}
3916
3917
3918
3919
3920
3921void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay)
3922{
3923 queue_delayed_work(bond->wq, &bond->slave_arr_work, delay);
3924}
3925
3926
3927static void bond_slave_arr_handler(struct work_struct *work)
3928{
3929 struct bonding *bond = container_of(work, struct bonding,
3930 slave_arr_work.work);
3931 int ret;
3932
3933 if (!rtnl_trylock())
3934 goto err;
3935
3936 ret = bond_update_slave_arr(bond, NULL);
3937 rtnl_unlock();
3938 if (ret) {
3939 pr_warn_ratelimited("Failed to update slave array from WT\n");
3940 goto err;
3941 }
3942 return;
3943
3944err:
3945 bond_slave_arr_work_rearm(bond, 1);
3946}
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave)
3957{
3958 struct slave *slave;
3959 struct list_head *iter;
3960 struct bond_up_slave *new_arr, *old_arr;
3961 int agg_id = 0;
3962 int ret = 0;
3963
3964#ifdef CONFIG_LOCKDEP
3965 WARN_ON(lockdep_is_held(&bond->mode_lock));
3966#endif
3967
3968 new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]),
3969 GFP_KERNEL);
3970 if (!new_arr) {
3971 ret = -ENOMEM;
3972 pr_err("Failed to build slave-array.\n");
3973 goto out;
3974 }
3975 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3976 struct ad_info ad_info;
3977
3978 if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
3979 pr_debug("bond_3ad_get_active_agg_info failed\n");
3980 kfree_rcu(new_arr, rcu);
3981
3982
3983
3984 old_arr = rtnl_dereference(bond->slave_arr);
3985 if (old_arr) {
3986 RCU_INIT_POINTER(bond->slave_arr, NULL);
3987 kfree_rcu(old_arr, rcu);
3988 }
3989 goto out;
3990 }
3991 agg_id = ad_info.aggregator_id;
3992 }
3993 bond_for_each_slave(bond, slave, iter) {
3994 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3995 struct aggregator *agg;
3996
3997 agg = SLAVE_AD_INFO(slave)->port.aggregator;
3998 if (!agg || agg->aggregator_identifier != agg_id)
3999 continue;
4000 }
4001 if (!bond_slave_can_tx(slave))
4002 continue;
4003 if (skipslave == slave)
4004 continue;
4005
4006 netdev_dbg(bond->dev,
4007 "Adding slave dev %s to tx hash array[%d]\n",
4008 slave->dev->name, new_arr->count);
4009
4010 new_arr->arr[new_arr->count++] = slave;
4011 }
4012
4013 old_arr = rtnl_dereference(bond->slave_arr);
4014 rcu_assign_pointer(bond->slave_arr, new_arr);
4015 if (old_arr)
4016 kfree_rcu(old_arr, rcu);
4017out:
4018 if (ret != 0 && skipslave) {
4019 int idx;
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029 old_arr = rtnl_dereference(bond->slave_arr);
4030 for (idx = 0; idx < old_arr->count; idx++) {
4031 if (skipslave == old_arr->arr[idx]) {
4032 old_arr->arr[idx] =
4033 old_arr->arr[old_arr->count-1];
4034 old_arr->count--;
4035 break;
4036 }
4037 }
4038 }
4039 return ret;
4040}
4041
4042
4043
4044
4045
4046static netdev_tx_t bond_3ad_xor_xmit(struct sk_buff *skb,
4047 struct net_device *dev)
4048{
4049 struct bonding *bond = netdev_priv(dev);
4050 struct slave *slave;
4051 struct bond_up_slave *slaves;
4052 unsigned int count;
4053
4054 slaves = rcu_dereference(bond->slave_arr);
4055 count = slaves ? READ_ONCE(slaves->count) : 0;
4056 if (likely(count)) {
4057 slave = slaves->arr[bond_xmit_hash(bond, skb) % count];
4058 bond_dev_queue_xmit(bond, skb, slave->dev);
4059 } else {
4060 bond_tx_drop(dev, skb);
4061 }
4062
4063 return NETDEV_TX_OK;
4064}
4065
4066
4067static netdev_tx_t bond_xmit_broadcast(struct sk_buff *skb,
4068 struct net_device *bond_dev)
4069{
4070 struct bonding *bond = netdev_priv(bond_dev);
4071 struct slave *slave = NULL;
4072 struct list_head *iter;
4073
4074 bond_for_each_slave_rcu(bond, slave, iter) {
4075 if (bond_is_last_slave(bond, slave))
4076 break;
4077 if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
4078 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
4079
4080 if (!skb2) {
4081 net_err_ratelimited("%s: Error: %s: skb_clone() failed\n",
4082 bond_dev->name, __func__);
4083 continue;
4084 }
4085 bond_dev_queue_xmit(bond, skb2, slave->dev);
4086 }
4087 }
4088 if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP)
4089 bond_dev_queue_xmit(bond, skb, slave->dev);
4090 else
4091 bond_tx_drop(bond_dev, skb);
4092
4093 return NETDEV_TX_OK;
4094}
4095
4096
4097
4098
4099static inline int bond_slave_override(struct bonding *bond,
4100 struct sk_buff *skb)
4101{
4102 struct slave *slave = NULL;
4103 struct list_head *iter;
4104
4105 if (!skb_rx_queue_recorded(skb))
4106 return 1;
4107
4108
4109 bond_for_each_slave_rcu(bond, slave, iter) {
4110 if (slave->queue_id == skb_get_queue_mapping(skb)) {
4111 if (bond_slave_is_up(slave) &&
4112 slave->link == BOND_LINK_UP) {
4113 bond_dev_queue_xmit(bond, skb, slave->dev);
4114 return 0;
4115 }
4116
4117 break;
4118 }
4119 }
4120
4121 return 1;
4122}
4123
4124
4125static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb,
4126 struct net_device *sb_dev)
4127{
4128
4129
4130
4131
4132
4133 u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0;
4134
4135
4136 qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb_get_queue_mapping(skb);
4137
4138 if (unlikely(txq >= dev->real_num_tx_queues)) {
4139 do {
4140 txq -= dev->real_num_tx_queues;
4141 } while (txq >= dev->real_num_tx_queues);
4142 }
4143 return txq;
4144}
4145
4146static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
4147{
4148 struct bonding *bond = netdev_priv(dev);
4149
4150 if (bond_should_override_tx_queue(bond) &&
4151 !bond_slave_override(bond, skb))
4152 return NETDEV_TX_OK;
4153
4154 switch (BOND_MODE(bond)) {
4155 case BOND_MODE_ROUNDROBIN:
4156 return bond_xmit_roundrobin(skb, dev);
4157 case BOND_MODE_ACTIVEBACKUP:
4158 return bond_xmit_activebackup(skb, dev);
4159 case BOND_MODE_8023AD:
4160 case BOND_MODE_XOR:
4161 return bond_3ad_xor_xmit(skb, dev);
4162 case BOND_MODE_BROADCAST:
4163 return bond_xmit_broadcast(skb, dev);
4164 case BOND_MODE_ALB:
4165 return bond_alb_xmit(skb, dev);
4166 case BOND_MODE_TLB:
4167 return bond_tlb_xmit(skb, dev);
4168 default:
4169
4170 netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond));
4171 WARN_ON_ONCE(1);
4172 bond_tx_drop(dev, skb);
4173 return NETDEV_TX_OK;
4174 }
4175}
4176
4177static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
4178{
4179 struct bonding *bond = netdev_priv(dev);
4180 netdev_tx_t ret = NETDEV_TX_OK;
4181
4182
4183
4184
4185 if (unlikely(is_netpoll_tx_blocked(dev)))
4186 return NETDEV_TX_BUSY;
4187
4188 rcu_read_lock();
4189 if (bond_has_slaves(bond))
4190 ret = __bond_start_xmit(skb, dev);
4191 else
4192 bond_tx_drop(dev, skb);
4193 rcu_read_unlock();
4194
4195 return ret;
4196}
4197
4198static int bond_ethtool_get_link_ksettings(struct net_device *bond_dev,
4199 struct ethtool_link_ksettings *cmd)
4200{
4201 struct bonding *bond = netdev_priv(bond_dev);
4202 unsigned long speed = 0;
4203 struct list_head *iter;
4204 struct slave *slave;
4205
4206 cmd->base.duplex = DUPLEX_UNKNOWN;
4207 cmd->base.port = PORT_OTHER;
4208
4209
4210
4211
4212
4213
4214 bond_for_each_slave(bond, slave, iter) {
4215 if (bond_slave_can_tx(slave)) {
4216 if (slave->speed != SPEED_UNKNOWN)
4217 speed += slave->speed;
4218 if (cmd->base.duplex == DUPLEX_UNKNOWN &&
4219 slave->duplex != DUPLEX_UNKNOWN)
4220 cmd->base.duplex = slave->duplex;
4221 }
4222 }
4223 cmd->base.speed = speed ? : SPEED_UNKNOWN;
4224
4225 return 0;
4226}
4227
4228static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
4229 struct ethtool_drvinfo *drvinfo)
4230{
4231 strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
4232 strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
4233 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d",
4234 BOND_ABI_VERSION);
4235}
4236
4237static const struct ethtool_ops bond_ethtool_ops = {
4238 .get_drvinfo = bond_ethtool_get_drvinfo,
4239 .get_link = ethtool_op_get_link,
4240 .get_link_ksettings = bond_ethtool_get_link_ksettings,
4241};
4242
4243static const struct net_device_ops bond_netdev_ops = {
4244 .ndo_init = bond_init,
4245 .ndo_uninit = bond_uninit,
4246 .ndo_open = bond_open,
4247 .ndo_stop = bond_close,
4248 .ndo_start_xmit = bond_start_xmit,
4249 .ndo_select_queue = bond_select_queue,
4250 .ndo_get_stats64 = bond_get_stats,
4251 .ndo_do_ioctl = bond_do_ioctl,
4252 .ndo_change_rx_flags = bond_change_rx_flags,
4253 .ndo_set_rx_mode = bond_set_rx_mode,
4254 .ndo_change_mtu = bond_change_mtu,
4255 .ndo_set_mac_address = bond_set_mac_address,
4256 .ndo_neigh_setup = bond_neigh_setup,
4257 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
4258 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
4259 .ndo_get_lock_subclass = bond_get_nest_level,
4260#ifdef CONFIG_NET_POLL_CONTROLLER
4261 .ndo_netpoll_setup = bond_netpoll_setup,
4262 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
4263 .ndo_poll_controller = bond_poll_controller,
4264#endif
4265 .ndo_add_slave = bond_enslave,
4266 .ndo_del_slave = bond_release,
4267 .ndo_fix_features = bond_fix_features,
4268 .ndo_features_check = passthru_features_check,
4269};
4270
4271static const struct device_type bond_type = {
4272 .name = "bond",
4273};
4274
4275static void bond_destructor(struct net_device *bond_dev)
4276{
4277 struct bonding *bond = netdev_priv(bond_dev);
4278 if (bond->wq)
4279 destroy_workqueue(bond->wq);
4280}
4281
4282void bond_setup(struct net_device *bond_dev)
4283{
4284 struct bonding *bond = netdev_priv(bond_dev);
4285
4286 spin_lock_init(&bond->mode_lock);
4287 spin_lock_init(&bond->stats_lock);
4288 bond->params = bonding_defaults;
4289
4290
4291 bond->dev = bond_dev;
4292
4293
4294 ether_setup(bond_dev);
4295 bond_dev->max_mtu = ETH_MAX_MTU;
4296 bond_dev->netdev_ops = &bond_netdev_ops;
4297 bond_dev->ethtool_ops = &bond_ethtool_ops;
4298
4299 bond_dev->needs_free_netdev = true;
4300 bond_dev->priv_destructor = bond_destructor;
4301
4302 SET_NETDEV_DEVTYPE(bond_dev, &bond_type);
4303
4304
4305 bond_dev->flags |= IFF_MASTER;
4306 bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE;
4307 bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
4308
4309
4310 bond_dev->features |= NETIF_F_LLTX;
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320 bond_dev->features |= NETIF_F_NETNS_LOCAL;
4321
4322 bond_dev->hw_features = BOND_VLAN_FEATURES |
4323 NETIF_F_HW_VLAN_CTAG_RX |
4324 NETIF_F_HW_VLAN_CTAG_FILTER;
4325
4326 bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL | NETIF_F_GSO_UDP_L4;
4327 bond_dev->features |= bond_dev->hw_features;
4328 bond_dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
4329}
4330
4331
4332
4333
4334static void bond_uninit(struct net_device *bond_dev)
4335{
4336 struct bonding *bond = netdev_priv(bond_dev);
4337 struct list_head *iter;
4338 struct slave *slave;
4339 struct bond_up_slave *arr;
4340
4341 bond_netpoll_cleanup(bond_dev);
4342
4343
4344 bond_for_each_slave(bond, slave, iter)
4345 __bond_release_one(bond_dev, slave->dev, true, true);
4346 netdev_info(bond_dev, "Released all slaves\n");
4347
4348 arr = rtnl_dereference(bond->slave_arr);
4349 if (arr) {
4350 RCU_INIT_POINTER(bond->slave_arr, NULL);
4351 kfree_rcu(arr, rcu);
4352 }
4353
4354 list_del(&bond->bond_list);
4355
4356 bond_debug_unregister(bond);
4357}
4358
4359
4360
4361static int bond_check_params(struct bond_params *params)
4362{
4363 int arp_validate_value, fail_over_mac_value, primary_reselect_value, i;
4364 struct bond_opt_value newval;
4365 const struct bond_opt_value *valptr;
4366 int arp_all_targets_value = 0;
4367 u16 ad_actor_sys_prio = 0;
4368 u16 ad_user_port_key = 0;
4369 __be32 arp_target[BOND_MAX_ARP_TARGETS] = { 0 };
4370 int arp_ip_count;
4371 int bond_mode = BOND_MODE_ROUNDROBIN;
4372 int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
4373 int lacp_fast = 0;
4374 int tlb_dynamic_lb;
4375
4376
4377 if (mode) {
4378 bond_opt_initstr(&newval, mode);
4379 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval);
4380 if (!valptr) {
4381 pr_err("Error: Invalid bonding mode \"%s\"\n", mode);
4382 return -EINVAL;
4383 }
4384 bond_mode = valptr->value;
4385 }
4386
4387 if (xmit_hash_policy) {
4388 if (bond_mode == BOND_MODE_ROUNDROBIN ||
4389 bond_mode == BOND_MODE_ACTIVEBACKUP ||
4390 bond_mode == BOND_MODE_BROADCAST) {
4391 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
4392 bond_mode_name(bond_mode));
4393 } else {
4394 bond_opt_initstr(&newval, xmit_hash_policy);
4395 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH),
4396 &newval);
4397 if (!valptr) {
4398 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
4399 xmit_hash_policy);
4400 return -EINVAL;
4401 }
4402 xmit_hashtype = valptr->value;
4403 }
4404 }
4405
4406 if (lacp_rate) {
4407 if (bond_mode != BOND_MODE_8023AD) {
4408 pr_info("lacp_rate param is irrelevant in mode %s\n",
4409 bond_mode_name(bond_mode));
4410 } else {
4411 bond_opt_initstr(&newval, lacp_rate);
4412 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE),
4413 &newval);
4414 if (!valptr) {
4415 pr_err("Error: Invalid lacp rate \"%s\"\n",
4416 lacp_rate);
4417 return -EINVAL;
4418 }
4419 lacp_fast = valptr->value;
4420 }
4421 }
4422
4423 if (ad_select) {
4424 bond_opt_initstr(&newval, ad_select);
4425 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT),
4426 &newval);
4427 if (!valptr) {
4428 pr_err("Error: Invalid ad_select \"%s\"\n", ad_select);
4429 return -EINVAL;
4430 }
4431 params->ad_select = valptr->value;
4432 if (bond_mode != BOND_MODE_8023AD)
4433 pr_warn("ad_select param only affects 802.3ad mode\n");
4434 } else {
4435 params->ad_select = BOND_AD_STABLE;
4436 }
4437
4438 if (max_bonds < 0) {
4439 pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4440 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
4441 max_bonds = BOND_DEFAULT_MAX_BONDS;
4442 }
4443
4444 if (miimon < 0) {
4445 pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4446 miimon, INT_MAX);
4447 miimon = 0;
4448 }
4449
4450 if (updelay < 0) {
4451 pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4452 updelay, INT_MAX);
4453 updelay = 0;
4454 }
4455
4456 if (downdelay < 0) {
4457 pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4458 downdelay, INT_MAX);
4459 downdelay = 0;
4460 }
4461
4462 if ((use_carrier != 0) && (use_carrier != 1)) {
4463 pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4464 use_carrier);
4465 use_carrier = 1;
4466 }
4467
4468 if (num_peer_notif < 0 || num_peer_notif > 255) {
4469 pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4470 num_peer_notif);
4471 num_peer_notif = 1;
4472 }
4473
4474
4475 if (!bond_mode_uses_arp(bond_mode)) {
4476 if (!miimon) {
4477 pr_warn("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n");
4478 pr_warn("Forcing miimon to 100msec\n");
4479 miimon = BOND_DEFAULT_MIIMON;
4480 }
4481 }
4482
4483 if (tx_queues < 1 || tx_queues > 255) {
4484 pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n",
4485 tx_queues, BOND_DEFAULT_TX_QUEUES);
4486 tx_queues = BOND_DEFAULT_TX_QUEUES;
4487 }
4488
4489 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
4490 pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n",
4491 all_slaves_active);
4492 all_slaves_active = 0;
4493 }
4494
4495 if (resend_igmp < 0 || resend_igmp > 255) {
4496 pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n",
4497 resend_igmp, BOND_DEFAULT_RESEND_IGMP);
4498 resend_igmp = BOND_DEFAULT_RESEND_IGMP;
4499 }
4500
4501 bond_opt_initval(&newval, packets_per_slave);
4502 if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) {
4503 pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n",
4504 packets_per_slave, USHRT_MAX);
4505 packets_per_slave = 1;
4506 }
4507
4508 if (bond_mode == BOND_MODE_ALB) {
4509 pr_notice("In ALB mode you might experience client disconnections upon reconnection of a link if the bonding module updelay parameter (%d msec) is incompatible with the forwarding delay time of the switch\n",
4510 updelay);
4511 }
4512
4513 if (!miimon) {
4514 if (updelay || downdelay) {
4515
4516
4517
4518 pr_warn("Warning: miimon module parameter not set and updelay (%d) or downdelay (%d) module parameter is set; updelay and downdelay have no effect unless miimon is set\n",
4519 updelay, downdelay);
4520 }
4521 } else {
4522
4523 if (arp_interval) {
4524 pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4525 miimon, arp_interval);
4526 arp_interval = 0;
4527 }
4528
4529 if ((updelay % miimon) != 0) {
4530 pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4531 updelay, miimon, (updelay / miimon) * miimon);
4532 }
4533
4534 updelay /= miimon;
4535
4536 if ((downdelay % miimon) != 0) {
4537 pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4538 downdelay, miimon,
4539 (downdelay / miimon) * miimon);
4540 }
4541
4542 downdelay /= miimon;
4543 }
4544
4545 if (arp_interval < 0) {
4546 pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4547 arp_interval, INT_MAX);
4548 arp_interval = 0;
4549 }
4550
4551 for (arp_ip_count = 0, i = 0;
4552 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) {
4553 __be32 ip;
4554
4555
4556 if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) ||
4557 !bond_is_ip_target_ok(ip)) {
4558 pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4559 arp_ip_target[i]);
4560 arp_interval = 0;
4561 } else {
4562 if (bond_get_targets_ip(arp_target, ip) == -1)
4563 arp_target[arp_ip_count++] = ip;
4564 else
4565 pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n",
4566 &ip);
4567 }
4568 }
4569
4570 if (arp_interval && !arp_ip_count) {
4571
4572 pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4573 arp_interval);
4574 arp_interval = 0;
4575 }
4576
4577 if (arp_validate) {
4578 if (!arp_interval) {
4579 pr_err("arp_validate requires arp_interval\n");
4580 return -EINVAL;
4581 }
4582
4583 bond_opt_initstr(&newval, arp_validate);
4584 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE),
4585 &newval);
4586 if (!valptr) {
4587 pr_err("Error: invalid arp_validate \"%s\"\n",
4588 arp_validate);
4589 return -EINVAL;
4590 }
4591 arp_validate_value = valptr->value;
4592 } else {
4593 arp_validate_value = 0;
4594 }
4595
4596 if (arp_all_targets) {
4597 bond_opt_initstr(&newval, arp_all_targets);
4598 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS),
4599 &newval);
4600 if (!valptr) {
4601 pr_err("Error: invalid arp_all_targets_value \"%s\"\n",
4602 arp_all_targets);
4603 arp_all_targets_value = 0;
4604 } else {
4605 arp_all_targets_value = valptr->value;
4606 }
4607 }
4608
4609 if (miimon) {
4610 pr_info("MII link monitoring set to %d ms\n", miimon);
4611 } else if (arp_interval) {
4612 valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE,
4613 arp_validate_value);
4614 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
4615 arp_interval, valptr->string, arp_ip_count);
4616
4617 for (i = 0; i < arp_ip_count; i++)
4618 pr_cont(" %s", arp_ip_target[i]);
4619
4620 pr_cont("\n");
4621
4622 } else if (max_bonds) {
4623
4624
4625
4626 pr_debug("Warning: either miimon or arp_interval and arp_ip_target module parameters must be specified, otherwise bonding will not detect link failures! see bonding.txt for details\n");
4627 }
4628
4629 if (primary && !bond_mode_uses_primary(bond_mode)) {
4630
4631
4632
4633 pr_warn("Warning: %s primary device specified but has no effect in %s mode\n",
4634 primary, bond_mode_name(bond_mode));
4635 primary = NULL;
4636 }
4637
4638 if (primary && primary_reselect) {
4639 bond_opt_initstr(&newval, primary_reselect);
4640 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT),
4641 &newval);
4642 if (!valptr) {
4643 pr_err("Error: Invalid primary_reselect \"%s\"\n",
4644 primary_reselect);
4645 return -EINVAL;
4646 }
4647 primary_reselect_value = valptr->value;
4648 } else {
4649 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4650 }
4651
4652 if (fail_over_mac) {
4653 bond_opt_initstr(&newval, fail_over_mac);
4654 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC),
4655 &newval);
4656 if (!valptr) {
4657 pr_err("Error: invalid fail_over_mac \"%s\"\n",
4658 fail_over_mac);
4659 return -EINVAL;
4660 }
4661 fail_over_mac_value = valptr->value;
4662 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
4663 pr_warn("Warning: fail_over_mac only affects active-backup mode\n");
4664 } else {
4665 fail_over_mac_value = BOND_FOM_NONE;
4666 }
4667
4668 bond_opt_initstr(&newval, "default");
4669 valptr = bond_opt_parse(
4670 bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO),
4671 &newval);
4672 if (!valptr) {
4673 pr_err("Error: No ad_actor_sys_prio default value");
4674 return -EINVAL;
4675 }
4676 ad_actor_sys_prio = valptr->value;
4677
4678 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY),
4679 &newval);
4680 if (!valptr) {
4681 pr_err("Error: No ad_user_port_key default value");
4682 return -EINVAL;
4683 }
4684 ad_user_port_key = valptr->value;
4685
4686 bond_opt_initstr(&newval, "default");
4687 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_TLB_DYNAMIC_LB), &newval);
4688 if (!valptr) {
4689 pr_err("Error: No tlb_dynamic_lb default value");
4690 return -EINVAL;
4691 }
4692 tlb_dynamic_lb = valptr->value;
4693
4694 if (lp_interval == 0) {
4695 pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n",
4696 INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL);
4697 lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
4698 }
4699
4700
4701 params->mode = bond_mode;
4702 params->xmit_policy = xmit_hashtype;
4703 params->miimon = miimon;
4704 params->num_peer_notif = num_peer_notif;
4705 params->arp_interval = arp_interval;
4706 params->arp_validate = arp_validate_value;
4707 params->arp_all_targets = arp_all_targets_value;
4708 params->updelay = updelay;
4709 params->downdelay = downdelay;
4710 params->use_carrier = use_carrier;
4711 params->lacp_fast = lacp_fast;
4712 params->primary[0] = 0;
4713 params->primary_reselect = primary_reselect_value;
4714 params->fail_over_mac = fail_over_mac_value;
4715 params->tx_queues = tx_queues;
4716 params->all_slaves_active = all_slaves_active;
4717 params->resend_igmp = resend_igmp;
4718 params->min_links = min_links;
4719 params->lp_interval = lp_interval;
4720 params->packets_per_slave = packets_per_slave;
4721 params->tlb_dynamic_lb = tlb_dynamic_lb;
4722 params->ad_actor_sys_prio = ad_actor_sys_prio;
4723 eth_zero_addr(params->ad_actor_system);
4724 params->ad_user_port_key = ad_user_port_key;
4725 if (packets_per_slave > 0) {
4726 params->reciprocal_packets_per_slave =
4727 reciprocal_value(packets_per_slave);
4728 } else {
4729
4730
4731
4732 params->reciprocal_packets_per_slave =
4733 (struct reciprocal_value) { 0 };
4734 }
4735
4736 if (primary) {
4737 strncpy(params->primary, primary, IFNAMSIZ);
4738 params->primary[IFNAMSIZ - 1] = 0;
4739 }
4740
4741 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4742
4743 return 0;
4744}
4745
4746
4747static int bond_init(struct net_device *bond_dev)
4748{
4749 struct bonding *bond = netdev_priv(bond_dev);
4750 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
4751
4752 netdev_dbg(bond_dev, "Begin bond_init\n");
4753
4754 bond->wq = alloc_ordered_workqueue(bond_dev->name, WQ_MEM_RECLAIM);
4755 if (!bond->wq)
4756 return -ENOMEM;
4757
4758 bond->nest_level = SINGLE_DEPTH_NESTING;
4759 netdev_lockdep_set_classes(bond_dev);
4760
4761 list_add_tail(&bond->bond_list, &bn->dev_list);
4762
4763 bond_prepare_sysfs_group(bond);
4764
4765 bond_debug_register(bond);
4766
4767
4768 if (is_zero_ether_addr(bond_dev->dev_addr) &&
4769 bond_dev->addr_assign_type == NET_ADDR_PERM)
4770 eth_hw_addr_random(bond_dev);
4771
4772 return 0;
4773}
4774
4775unsigned int bond_get_num_tx_queues(void)
4776{
4777 return tx_queues;
4778}
4779
4780
4781
4782
4783
4784
4785int bond_create(struct net *net, const char *name)
4786{
4787 struct net_device *bond_dev;
4788 struct bonding *bond;
4789 struct alb_bond_info *bond_info;
4790 int res;
4791
4792 rtnl_lock();
4793
4794 bond_dev = alloc_netdev_mq(sizeof(struct bonding),
4795 name ? name : "bond%d", NET_NAME_UNKNOWN,
4796 bond_setup, tx_queues);
4797 if (!bond_dev) {
4798 pr_err("%s: eek! can't alloc netdev!\n", name);
4799 rtnl_unlock();
4800 return -ENOMEM;
4801 }
4802
4803
4804
4805
4806
4807 bond = netdev_priv(bond_dev);
4808 bond_info = &(BOND_ALB_INFO(bond));
4809 bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX;
4810
4811 dev_net_set(bond_dev, net);
4812 bond_dev->rtnl_link_ops = &bond_link_ops;
4813
4814 res = register_netdevice(bond_dev);
4815
4816 netif_carrier_off(bond_dev);
4817
4818 bond_work_init_all(bond);
4819
4820 rtnl_unlock();
4821 if (res < 0)
4822 free_netdev(bond_dev);
4823 return res;
4824}
4825
4826static int __net_init bond_net_init(struct net *net)
4827{
4828 struct bond_net *bn = net_generic(net, bond_net_id);
4829
4830 bn->net = net;
4831 INIT_LIST_HEAD(&bn->dev_list);
4832
4833 bond_create_proc_dir(bn);
4834 bond_create_sysfs(bn);
4835
4836 return 0;
4837}
4838
4839static void __net_exit bond_net_exit(struct net *net)
4840{
4841 struct bond_net *bn = net_generic(net, bond_net_id);
4842 struct bonding *bond, *tmp_bond;
4843 LIST_HEAD(list);
4844
4845 bond_destroy_sysfs(bn);
4846
4847
4848 rtnl_lock();
4849 list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list)
4850 unregister_netdevice_queue(bond->dev, &list);
4851 unregister_netdevice_many(&list);
4852 rtnl_unlock();
4853
4854 bond_destroy_proc_dir(bn);
4855}
4856
4857static struct pernet_operations bond_net_ops = {
4858 .init = bond_net_init,
4859 .exit = bond_net_exit,
4860 .id = &bond_net_id,
4861 .size = sizeof(struct bond_net),
4862};
4863
4864static int __init bonding_init(void)
4865{
4866 int i;
4867 int res;
4868
4869 pr_info("%s", bond_version);
4870
4871 res = bond_check_params(&bonding_defaults);
4872 if (res)
4873 goto out;
4874
4875 res = register_pernet_subsys(&bond_net_ops);
4876 if (res)
4877 goto out;
4878
4879 res = bond_netlink_init();
4880 if (res)
4881 goto err_link;
4882
4883 bond_create_debugfs();
4884
4885 for (i = 0; i < max_bonds; i++) {
4886 res = bond_create(&init_net, NULL);
4887 if (res)
4888 goto err;
4889 }
4890
4891 register_netdevice_notifier(&bond_netdev_notifier);
4892out:
4893 return res;
4894err:
4895 bond_destroy_debugfs();
4896 bond_netlink_fini();
4897err_link:
4898 unregister_pernet_subsys(&bond_net_ops);
4899 goto out;
4900
4901}
4902
4903static void __exit bonding_exit(void)
4904{
4905 unregister_netdevice_notifier(&bond_netdev_notifier);
4906
4907 bond_destroy_debugfs();
4908
4909 bond_netlink_fini();
4910 unregister_pernet_subsys(&bond_net_ops);
4911
4912#ifdef CONFIG_NET_POLL_CONTROLLER
4913
4914 WARN_ON(atomic_read(&netpoll_block_tx));
4915#endif
4916}
4917
4918module_init(bonding_init);
4919module_exit(bonding_exit);
4920MODULE_LICENSE("GPL");
4921MODULE_VERSION(DRV_VERSION);
4922MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
4923MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
4924