2 * originally based on the dummy device.
4 * Copyright 1999, Thomas Davis, tadavis@lbl.gov.
5 * Licensed under the GPL. Based on dummy.c, and eql.c devices.
7 * bonding.c: an Ethernet Bonding driver
9 * This is useful to talk to a Cisco EtherChannel compatible equipment:
11 * Sun Trunking (Solaris)
12 * Alteon AceDirector Trunks
14 * and probably many L2 switches ...
17 * ifconfig bond0 ipaddress netmask up
18 * will setup a network device, with an ip address. No mac address
19 * will be assigned at this time. The hw mac address will come from
20 * the first slave bonded to the channel. All slaves will then use
21 * this hw mac address.
24 * will release all slaves, marking them as down.
26 * ifenslave bond0 eth0
27 * will attach eth0 to bond0 as a slave. eth0 hw mac address will either
28 * a: be used as initial mac address
29 * b: if a hw mac address already is there, eth0's hw mac address
30 * will then be set from bond0.
34 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
36 #include <linux/kernel.h>
37 #include <linux/module.h>
38 #include <linux/types.h>
39 #include <linux/fcntl.h>
40 #include <linux/interrupt.h>
41 #include <linux/ptrace.h>
42 #include <linux/ioport.h>
46 #include <linux/tcp.h>
47 #include <linux/udp.h>
48 #include <linux/slab.h>
49 #include <linux/string.h>
50 #include <linux/init.h>
51 #include <linux/timer.h>
52 #include <linux/socket.h>
53 #include <linux/ctype.h>
54 #include <linux/inet.h>
55 #include <linux/bitops.h>
57 #include <asm/system.h>
59 #include <linux/uaccess.h>
60 #include <linux/errno.h>
61 #include <linux/netdevice.h>
62 #include <linux/inetdevice.h>
63 #include <linux/igmp.h>
64 #include <linux/etherdevice.h>
65 #include <linux/skbuff.h>
67 #include <linux/rtnetlink.h>
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70 #include <linux/smp.h>
71 #include <linux/if_ether.h>
73 #include <linux/mii.h>
74 #include <linux/ethtool.h>
75 #include <linux/if_vlan.h>
76 #include <linux/if_bonding.h>
77 #include <linux/jiffies.h>
78 #include <net/route.h>
79 #include <net/net_namespace.h>
80 #include <net/netns/generic.h>
85 /*---------------------------- Module parameters ----------------------------*/
87 /* monitor all links that often (in milliseconds). <=0 disables monitoring */
88 #define BOND_LINK_MON_INTERV 0
89 #define BOND_LINK_ARP_INTERV 0
91 static int max_bonds = BOND_DEFAULT_MAX_BONDS;
92 static int num_grat_arp = 1;
93 static int num_unsol_na = 1;
94 static int miimon = BOND_LINK_MON_INTERV;
97 static int use_carrier = 1;
100 static char *primary_reselect;
101 static char *lacp_rate;
102 static char *ad_select;
103 static char *xmit_hash_policy;
104 static int arp_interval = BOND_LINK_ARP_INTERV;
105 static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
106 static char *arp_validate;
107 static char *fail_over_mac;
108 static struct bond_params bonding_defaults;
110 module_param(max_bonds, int, 0);
111 MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
112 module_param(num_grat_arp, int, 0644);
113 MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packets to send on failover event");
114 module_param(num_unsol_na, int, 0644);
115 MODULE_PARM_DESC(num_unsol_na, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event");
116 module_param(miimon, int, 0);
117 MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
118 module_param(updelay, int, 0);
119 MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
120 module_param(downdelay, int, 0);
121 MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
123 module_param(use_carrier, int, 0);
124 MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
125 "0 for off, 1 for on (default)");
126 module_param(mode, charp, 0);
127 MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, "
128 "1 for active-backup, 2 for balance-xor, "
129 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
130 "6 for balance-alb");
131 module_param(primary, charp, 0);
132 MODULE_PARM_DESC(primary, "Primary network device to use");
133 module_param(primary_reselect, charp, 0);
134 MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
136 "0 for always (default), "
137 "1 for only if speed of primary is "
139 "2 for only on active slave "
141 module_param(lacp_rate, charp, 0);
142 MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner "
144 module_param(ad_select, charp, 0);
145 MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)");
146 module_param(xmit_hash_policy, charp, 0);
147 MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)"
148 ", 1 for layer 3+4");
149 module_param(arp_interval, int, 0);
150 MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
151 module_param_array(arp_ip_target, charp, NULL, 0);
152 MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
153 module_param(arp_validate, charp, 0);
154 MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all");
155 module_param(fail_over_mac, charp, 0);
156 MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow");
158 /*----------------------------- Global variables ----------------------------*/
160 static const char * const version =
161 DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n";
163 int bond_net_id __read_mostly;
165 static __be32 arp_target[BOND_MAX_ARP_TARGETS];
166 static int arp_ip_count;
167 static int bond_mode = BOND_MODE_ROUNDROBIN;
168 static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
169 static int lacp_fast;
172 const struct bond_parm_tbl bond_lacp_tbl[] = {
173 { "slow", AD_LACP_SLOW},
174 { "fast", AD_LACP_FAST},
178 const struct bond_parm_tbl bond_mode_tbl[] = {
179 { "balance-rr", BOND_MODE_ROUNDROBIN},
180 { "active-backup", BOND_MODE_ACTIVEBACKUP},
181 { "balance-xor", BOND_MODE_XOR},
182 { "broadcast", BOND_MODE_BROADCAST},
183 { "802.3ad", BOND_MODE_8023AD},
184 { "balance-tlb", BOND_MODE_TLB},
185 { "balance-alb", BOND_MODE_ALB},
189 const struct bond_parm_tbl xmit_hashtype_tbl[] = {
190 { "layer2", BOND_XMIT_POLICY_LAYER2},
191 { "layer3+4", BOND_XMIT_POLICY_LAYER34},
192 { "layer2+3", BOND_XMIT_POLICY_LAYER23},
196 const struct bond_parm_tbl arp_validate_tbl[] = {
197 { "none", BOND_ARP_VALIDATE_NONE},
198 { "active", BOND_ARP_VALIDATE_ACTIVE},
199 { "backup", BOND_ARP_VALIDATE_BACKUP},
200 { "all", BOND_ARP_VALIDATE_ALL},
204 const struct bond_parm_tbl fail_over_mac_tbl[] = {
205 { "none", BOND_FOM_NONE},
206 { "active", BOND_FOM_ACTIVE},
207 { "follow", BOND_FOM_FOLLOW},
211 const struct bond_parm_tbl pri_reselect_tbl[] = {
212 { "always", BOND_PRI_RESELECT_ALWAYS},
213 { "better", BOND_PRI_RESELECT_BETTER},
214 { "failure", BOND_PRI_RESELECT_FAILURE},
218 struct bond_parm_tbl ad_select_tbl[] = {
219 { "stable", BOND_AD_STABLE},
220 { "bandwidth", BOND_AD_BANDWIDTH},
221 { "count", BOND_AD_COUNT},
225 /*-------------------------- Forward declarations ---------------------------*/
227 static void bond_send_gratuitous_arp(struct bonding *bond);
228 static int bond_init(struct net_device *bond_dev);
229 static void bond_uninit(struct net_device *bond_dev);
231 /*---------------------------- General routines -----------------------------*/
233 static const char *bond_mode_name(int mode)
235 static const char *names[] = {
236 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
237 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
238 [BOND_MODE_XOR] = "load balancing (xor)",
239 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
240 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
241 [BOND_MODE_TLB] = "transmit load balancing",
242 [BOND_MODE_ALB] = "adaptive load balancing",
245 if (mode < 0 || mode > BOND_MODE_ALB)
251 /*---------------------------------- VLAN -----------------------------------*/
254 * bond_add_vlan - add a new vlan id on bond
255 * @bond: bond that got the notification
256 * @vlan_id: the vlan id to add
258 * Returns -ENOMEM if allocation failed.
260 static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id)
262 struct vlan_entry *vlan;
264 pr_debug("bond: %s, vlan id %d\n",
265 (bond ? bond->dev->name : "None"), vlan_id);
267 vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL);
271 INIT_LIST_HEAD(&vlan->vlan_list);
272 vlan->vlan_id = vlan_id;
274 write_lock_bh(&bond->lock);
276 list_add_tail(&vlan->vlan_list, &bond->vlan_list);
278 write_unlock_bh(&bond->lock);
280 pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name);
286 * bond_del_vlan - delete a vlan id from bond
287 * @bond: bond that got the notification
288 * @vlan_id: the vlan id to delete
290 * returns -ENODEV if @vlan_id was not found in @bond.
292 static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id)
294 struct vlan_entry *vlan;
297 pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id);
299 write_lock_bh(&bond->lock);
301 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
302 if (vlan->vlan_id == vlan_id) {
303 list_del(&vlan->vlan_list);
305 if (bond_is_lb(bond))
306 bond_alb_clear_vlan(bond, vlan_id);
308 pr_debug("removed VLAN ID %d from bond %s\n",
309 vlan_id, bond->dev->name);
313 if (list_empty(&bond->vlan_list) &&
314 (bond->slave_cnt == 0)) {
315 /* Last VLAN removed and no slaves, so
316 * restore block on adding VLANs. This will
317 * be removed once new slaves that are not
318 * VLAN challenged will be added.
320 bond->dev->features |= NETIF_F_VLAN_CHALLENGED;
328 pr_debug("couldn't find VLAN ID %d in bond %s\n",
329 vlan_id, bond->dev->name);
332 write_unlock_bh(&bond->lock);
337 * bond_has_challenged_slaves
338 * @bond: the bond we're working on
340 * Searches the slave list. Returns 1 if a vlan challenged slave
341 * was found, 0 otherwise.
343 * Assumes bond->lock is held.
345 static int bond_has_challenged_slaves(struct bonding *bond)
350 bond_for_each_slave(bond, slave, i) {
351 if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) {
352 pr_debug("found VLAN challenged slave - %s\n",
358 pr_debug("no VLAN challenged slaves found\n");
363 * bond_next_vlan - safely skip to the next item in the vlans list.
364 * @bond: the bond we're working on
365 * @curr: item we're advancing from
367 * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL,
368 * or @curr->next otherwise (even if it is @curr itself again).
370 * Caller must hold bond->lock
372 struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr)
374 struct vlan_entry *next, *last;
376 if (list_empty(&bond->vlan_list))
380 next = list_entry(bond->vlan_list.next,
381 struct vlan_entry, vlan_list);
383 last = list_entry(bond->vlan_list.prev,
384 struct vlan_entry, vlan_list);
386 next = list_entry(bond->vlan_list.next,
387 struct vlan_entry, vlan_list);
389 next = list_entry(curr->vlan_list.next,
390 struct vlan_entry, vlan_list);
398 * bond_dev_queue_xmit - Prepare skb for xmit.
400 * @bond: bond device that got this skb for tx.
401 * @skb: hw accel VLAN tagged skb to transmit
402 * @slave_dev: slave that is supposed to xmit this skbuff
404 * When the bond gets an skb to transmit that is
405 * already hardware accelerated VLAN tagged, and it
406 * needs to relay this skb to a slave that is not
407 * hw accel capable, the skb needs to be "unaccelerated",
408 * i.e. strip the hwaccel tag and re-insert it as part
411 int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
412 struct net_device *slave_dev)
414 unsigned short uninitialized_var(vlan_id);
416 if (!list_empty(&bond->vlan_list) &&
417 !(slave_dev->features & NETIF_F_HW_VLAN_TX) &&
418 vlan_get_tag(skb, &vlan_id) == 0) {
419 skb->dev = slave_dev;
420 skb = vlan_put_tag(skb, vlan_id);
422 /* vlan_put_tag() frees the skb in case of error,
423 * so return success here so the calling functions
424 * won't attempt to free is again.
429 skb->dev = slave_dev;
439 * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid
440 * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a
442 * a. This operation is performed in IOCTL context,
443 * b. The operation is protected by the RTNL semaphore in the 8021q code,
444 * c. Holding a lock with BH disabled while directly calling a base driver
445 * entry point is generally a BAD idea.
447 * The design of synchronization/protection for this operation in the 8021q
448 * module is good for one or more VLAN devices over a single physical device
449 * and cannot be extended for a teaming solution like bonding, so there is a
450 * potential race condition here where a net device from the vlan group might
451 * be referenced (either by a base driver or the 8021q code) while it is being
452 * removed from the system. However, it turns out we're not making matters
453 * worse, and if it works for regular VLAN usage it will work here too.
457 * bond_vlan_rx_register - Propagates registration to slaves
458 * @bond_dev: bonding net device that got called
459 * @grp: vlan group being registered
461 static void bond_vlan_rx_register(struct net_device *bond_dev,
462 struct vlan_group *grp)
464 struct bonding *bond = netdev_priv(bond_dev);
470 bond_for_each_slave(bond, slave, i) {
471 struct net_device *slave_dev = slave->dev;
472 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
474 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
475 slave_ops->ndo_vlan_rx_register) {
476 slave_ops->ndo_vlan_rx_register(slave_dev, grp);
482 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
483 * @bond_dev: bonding net device that got called
484 * @vid: vlan id being added
486 static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid)
488 struct bonding *bond = netdev_priv(bond_dev);
492 bond_for_each_slave(bond, slave, i) {
493 struct net_device *slave_dev = slave->dev;
494 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
496 if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
497 slave_ops->ndo_vlan_rx_add_vid) {
498 slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid);
502 res = bond_add_vlan(bond, vid);
504 pr_err("%s: Error: Failed to add vlan id %d\n",
505 bond_dev->name, vid);
510 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
511 * @bond_dev: bonding net device that got called
512 * @vid: vlan id being removed
514 static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid)
516 struct bonding *bond = netdev_priv(bond_dev);
518 struct net_device *vlan_dev;
521 bond_for_each_slave(bond, slave, i) {
522 struct net_device *slave_dev = slave->dev;
523 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
525 if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
526 slave_ops->ndo_vlan_rx_kill_vid) {
527 /* Save and then restore vlan_dev in the grp array,
528 * since the slave's driver might clear it.
530 vlan_dev = vlan_group_get_device(bond->vlgrp, vid);
531 slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid);
532 vlan_group_set_device(bond->vlgrp, vid, vlan_dev);
536 res = bond_del_vlan(bond, vid);
538 pr_err("%s: Error: Failed to remove vlan id %d\n",
539 bond_dev->name, vid);
543 static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev)
545 struct vlan_entry *vlan;
546 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
548 write_lock_bh(&bond->lock);
550 if (list_empty(&bond->vlan_list))
553 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
554 slave_ops->ndo_vlan_rx_register)
555 slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp);
557 if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
558 !(slave_ops->ndo_vlan_rx_add_vid))
561 list_for_each_entry(vlan, &bond->vlan_list, vlan_list)
562 slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id);
565 write_unlock_bh(&bond->lock);
568 static void bond_del_vlans_from_slave(struct bonding *bond,
569 struct net_device *slave_dev)
571 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
572 struct vlan_entry *vlan;
573 struct net_device *vlan_dev;
575 write_lock_bh(&bond->lock);
577 if (list_empty(&bond->vlan_list))
580 if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
581 !(slave_ops->ndo_vlan_rx_kill_vid))
584 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
585 /* Save and then restore vlan_dev in the grp array,
586 * since the slave's driver might clear it.
588 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
589 slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id);
590 vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev);
594 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
595 slave_ops->ndo_vlan_rx_register)
596 slave_ops->ndo_vlan_rx_register(slave_dev, NULL);
599 write_unlock_bh(&bond->lock);
602 /*------------------------------- Link status -------------------------------*/
605 * Set the carrier state for the master according to the state of its
606 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
607 * do special 802.3ad magic.
609 * Returns zero if carrier state does not change, nonzero if it does.
611 static int bond_set_carrier(struct bonding *bond)
616 if (bond->slave_cnt == 0)
619 if (bond->params.mode == BOND_MODE_8023AD)
620 return bond_3ad_set_carrier(bond);
622 bond_for_each_slave(bond, slave, i) {
623 if (slave->link == BOND_LINK_UP) {
624 if (!netif_carrier_ok(bond->dev)) {
625 netif_carrier_on(bond->dev);
633 if (netif_carrier_ok(bond->dev)) {
634 netif_carrier_off(bond->dev);
641 * Get link speed and duplex from the slave's base driver
642 * using ethtool. If for some reason the call fails or the
643 * values are invalid, fake speed and duplex to 100/Full
646 static int bond_update_speed_duplex(struct slave *slave)
648 struct net_device *slave_dev = slave->dev;
649 struct ethtool_cmd etool;
652 /* Fake speed and duplex */
653 slave->speed = SPEED_100;
654 slave->duplex = DUPLEX_FULL;
656 if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings)
659 res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool);
663 switch (etool.speed) {
673 switch (etool.duplex) {
681 slave->speed = etool.speed;
682 slave->duplex = etool.duplex;
688 * if <dev> supports MII link status reporting, check its link status.
690 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
691 * depending upon the setting of the use_carrier parameter.
693 * Return either BMSR_LSTATUS, meaning that the link is up (or we
694 * can't tell and just pretend it is), or 0, meaning that the link is
697 * If reporting is non-zero, instead of faking link up, return -1 if
698 * both ETHTOOL and MII ioctls fail (meaning the device does not
699 * support them). If use_carrier is set, return whatever it says.
700 * It'd be nice if there was a good way to tell if a driver supports
701 * netif_carrier, but there really isn't.
703 static int bond_check_dev_link(struct bonding *bond,
704 struct net_device *slave_dev, int reporting)
706 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
707 int (*ioctl)(struct net_device *, struct ifreq *, int);
709 struct mii_ioctl_data *mii;
711 if (!reporting && !netif_running(slave_dev))
714 if (bond->params.use_carrier)
715 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
717 /* Try to get link status using Ethtool first. */
718 if (slave_dev->ethtool_ops) {
719 if (slave_dev->ethtool_ops->get_link) {
722 link = slave_dev->ethtool_ops->get_link(slave_dev);
724 return link ? BMSR_LSTATUS : 0;
728 /* Ethtool can't be used, fallback to MII ioctls. */
729 ioctl = slave_ops->ndo_do_ioctl;
731 /* TODO: set pointer to correct ioctl on a per team member */
732 /* bases to make this more efficient. that is, once */
733 /* we determine the correct ioctl, we will always */
734 /* call it and not the others for that team */
738 * We cannot assume that SIOCGMIIPHY will also read a
739 * register; not all network drivers (e.g., e100)
743 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
744 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
746 if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
747 mii->reg_num = MII_BMSR;
748 if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0)
749 return mii->val_out & BMSR_LSTATUS;
754 * If reporting, report that either there's no dev->do_ioctl,
755 * or both SIOCGMIIREG and get_link failed (meaning that we
756 * cannot report link status). If not reporting, pretend
759 return reporting ? -1 : BMSR_LSTATUS;
762 /*----------------------------- Multicast list ------------------------------*/
765 * Push the promiscuity flag down to appropriate slaves
767 static int bond_set_promiscuity(struct bonding *bond, int inc)
770 if (USES_PRIMARY(bond->params.mode)) {
771 /* write lock already acquired */
772 if (bond->curr_active_slave) {
773 err = dev_set_promiscuity(bond->curr_active_slave->dev,
779 bond_for_each_slave(bond, slave, i) {
780 err = dev_set_promiscuity(slave->dev, inc);
789 * Push the allmulti flag down to all slaves
791 static int bond_set_allmulti(struct bonding *bond, int inc)
794 if (USES_PRIMARY(bond->params.mode)) {
795 /* write lock already acquired */
796 if (bond->curr_active_slave) {
797 err = dev_set_allmulti(bond->curr_active_slave->dev,
803 bond_for_each_slave(bond, slave, i) {
804 err = dev_set_allmulti(slave->dev, inc);
813 * Add a Multicast address to slaves
816 static void bond_mc_add(struct bonding *bond, void *addr)
818 if (USES_PRIMARY(bond->params.mode)) {
819 /* write lock already acquired */
820 if (bond->curr_active_slave)
821 dev_mc_add(bond->curr_active_slave->dev, addr);
826 bond_for_each_slave(bond, slave, i)
827 dev_mc_add(slave->dev, addr);
832 * Remove a multicast address from slave
835 static void bond_mc_del(struct bonding *bond, void *addr)
837 if (USES_PRIMARY(bond->params.mode)) {
838 /* write lock already acquired */
839 if (bond->curr_active_slave)
840 dev_mc_del(bond->curr_active_slave->dev, addr);
844 bond_for_each_slave(bond, slave, i) {
845 dev_mc_del(slave->dev, addr);
852 * Retrieve the list of registered multicast addresses for the bonding
853 * device and retransmit an IGMP JOIN request to the current active
856 static void bond_resend_igmp_join_requests(struct bonding *bond)
858 struct in_device *in_dev;
859 struct ip_mc_list *im;
862 in_dev = __in_dev_get_rcu(bond->dev);
864 for (im = in_dev->mc_list; im; im = im->next)
865 ip_mc_rejoin_group(im);
872 * flush all members of flush->mc_list from device dev->mc_list
874 static void bond_mc_list_flush(struct net_device *bond_dev,
875 struct net_device *slave_dev)
877 struct bonding *bond = netdev_priv(bond_dev);
878 struct netdev_hw_addr *ha;
880 netdev_for_each_mc_addr(ha, bond_dev)
881 dev_mc_del(slave_dev, ha->addr);
883 if (bond->params.mode == BOND_MODE_8023AD) {
884 /* del lacpdu mc addr from mc list */
885 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
887 dev_mc_del(slave_dev, lacpdu_multicast);
891 /*--------------------------- Active slave change ---------------------------*/
894 * Update the mc list and multicast-related flags for the new and
895 * old active slaves (if any) according to the multicast mode, and
896 * promiscuous flags unconditionally.
898 static void bond_mc_swap(struct bonding *bond, struct slave *new_active,
899 struct slave *old_active)
901 struct netdev_hw_addr *ha;
903 if (!USES_PRIMARY(bond->params.mode))
904 /* nothing to do - mc list is already up-to-date on
910 if (bond->dev->flags & IFF_PROMISC)
911 dev_set_promiscuity(old_active->dev, -1);
913 if (bond->dev->flags & IFF_ALLMULTI)
914 dev_set_allmulti(old_active->dev, -1);
916 netdev_for_each_mc_addr(ha, bond->dev)
917 dev_mc_del(old_active->dev, ha->addr);
921 /* FIXME: Signal errors upstream. */
922 if (bond->dev->flags & IFF_PROMISC)
923 dev_set_promiscuity(new_active->dev, 1);
925 if (bond->dev->flags & IFF_ALLMULTI)
926 dev_set_allmulti(new_active->dev, 1);
928 netdev_for_each_mc_addr(ha, bond->dev)
929 dev_mc_add(new_active->dev, ha->addr);
930 bond_resend_igmp_join_requests(bond);
935 * bond_do_fail_over_mac
937 * Perform special MAC address swapping for fail_over_mac settings
939 * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh.
941 static void bond_do_fail_over_mac(struct bonding *bond,
942 struct slave *new_active,
943 struct slave *old_active)
944 __releases(&bond->curr_slave_lock)
945 __releases(&bond->lock)
946 __acquires(&bond->lock)
947 __acquires(&bond->curr_slave_lock)
949 u8 tmp_mac[ETH_ALEN];
950 struct sockaddr saddr;
953 switch (bond->params.fail_over_mac) {
954 case BOND_FOM_ACTIVE:
956 memcpy(bond->dev->dev_addr, new_active->dev->dev_addr,
957 new_active->dev->addr_len);
959 case BOND_FOM_FOLLOW:
961 * if new_active && old_active, swap them
962 * if just old_active, do nothing (going to no active slave)
963 * if just new_active, set new_active to bond's MAC
968 write_unlock_bh(&bond->curr_slave_lock);
969 read_unlock(&bond->lock);
972 memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN);
973 memcpy(saddr.sa_data, old_active->dev->dev_addr,
975 saddr.sa_family = new_active->dev->type;
977 memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN);
978 saddr.sa_family = bond->dev->type;
981 rv = dev_set_mac_address(new_active->dev, &saddr);
983 pr_err("%s: Error %d setting MAC of slave %s\n",
984 bond->dev->name, -rv, new_active->dev->name);
991 memcpy(saddr.sa_data, tmp_mac, ETH_ALEN);
992 saddr.sa_family = old_active->dev->type;
994 rv = dev_set_mac_address(old_active->dev, &saddr);
996 pr_err("%s: Error %d setting MAC of slave %s\n",
997 bond->dev->name, -rv, new_active->dev->name);
999 read_lock(&bond->lock);
1000 write_lock_bh(&bond->curr_slave_lock);
1003 pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n",
1004 bond->dev->name, bond->params.fail_over_mac);
1010 static bool bond_should_change_active(struct bonding *bond)
1012 struct slave *prim = bond->primary_slave;
1013 struct slave *curr = bond->curr_active_slave;
1015 if (!prim || !curr || curr->link != BOND_LINK_UP)
1017 if (bond->force_primary) {
1018 bond->force_primary = false;
1021 if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER &&
1022 (prim->speed < curr->speed ||
1023 (prim->speed == curr->speed && prim->duplex <= curr->duplex)))
1025 if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE)
1031 * find_best_interface - select the best available slave to be the active one
1032 * @bond: our bonding struct
1034 * Warning: Caller must hold curr_slave_lock for writing.
1036 static struct slave *bond_find_best_slave(struct bonding *bond)
1038 struct slave *new_active, *old_active;
1039 struct slave *bestslave = NULL;
1040 int mintime = bond->params.updelay;
1043 new_active = bond->curr_active_slave;
1045 if (!new_active) { /* there were no active slaves left */
1046 if (bond->slave_cnt > 0) /* found one slave */
1047 new_active = bond->first_slave;
1049 return NULL; /* still no slave, return NULL */
1052 if ((bond->primary_slave) &&
1053 bond->primary_slave->link == BOND_LINK_UP &&
1054 bond_should_change_active(bond)) {
1055 new_active = bond->primary_slave;
1058 /* remember where to stop iterating over the slaves */
1059 old_active = new_active;
1061 bond_for_each_slave_from(bond, new_active, i, old_active) {
1062 if (new_active->link == BOND_LINK_UP) {
1064 } else if (new_active->link == BOND_LINK_BACK &&
1065 IS_UP(new_active->dev)) {
1066 /* link up, but waiting for stabilization */
1067 if (new_active->delay < mintime) {
1068 mintime = new_active->delay;
1069 bestslave = new_active;
1078 * change_active_interface - change the active slave into the specified one
1079 * @bond: our bonding struct
1080 * @new: the new slave to make the active one
1082 * Set the new slave to the bond's settings and unset them on the old
1083 * curr_active_slave.
1084 * Setting include flags, mc-list, promiscuity, allmulti, etc.
1086 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
1087 * because it is apparently the best available slave we have, even though its
1088 * updelay hasn't timed out yet.
1090 * If new_active is not NULL, caller must hold bond->lock for read and
1091 * curr_slave_lock for write_bh.
1093 void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
1095 struct slave *old_active = bond->curr_active_slave;
1097 if (old_active == new_active)
1101 new_active->jiffies = jiffies;
1103 if (new_active->link == BOND_LINK_BACK) {
1104 if (USES_PRIMARY(bond->params.mode)) {
1105 pr_info("%s: making interface %s the new active one %d ms earlier.\n",
1106 bond->dev->name, new_active->dev->name,
1107 (bond->params.updelay - new_active->delay) * bond->params.miimon);
1110 new_active->delay = 0;
1111 new_active->link = BOND_LINK_UP;
1113 if (bond->params.mode == BOND_MODE_8023AD)
1114 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
1116 if (bond_is_lb(bond))
1117 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
1119 if (USES_PRIMARY(bond->params.mode)) {
1120 pr_info("%s: making interface %s the new active one.\n",
1121 bond->dev->name, new_active->dev->name);
1126 if (USES_PRIMARY(bond->params.mode))
1127 bond_mc_swap(bond, new_active, old_active);
1129 if (bond_is_lb(bond)) {
1130 bond_alb_handle_active_change(bond, new_active);
1132 bond_set_slave_inactive_flags(old_active);
1134 bond_set_slave_active_flags(new_active);
1136 bond->curr_active_slave = new_active;
1139 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
1141 bond_set_slave_inactive_flags(old_active);
1144 bond_set_slave_active_flags(new_active);
1146 if (bond->params.fail_over_mac)
1147 bond_do_fail_over_mac(bond, new_active,
1150 bond->send_grat_arp = bond->params.num_grat_arp;
1151 bond_send_gratuitous_arp(bond);
1153 bond->send_unsol_na = bond->params.num_unsol_na;
1154 bond_send_unsolicited_na(bond);
1156 write_unlock_bh(&bond->curr_slave_lock);
1157 read_unlock(&bond->lock);
1159 netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER);
1161 read_lock(&bond->lock);
1162 write_lock_bh(&bond->curr_slave_lock);
1166 /* resend IGMP joins since all were sent on curr_active_slave */
1167 if (bond->params.mode == BOND_MODE_ROUNDROBIN) {
1168 bond_resend_igmp_join_requests(bond);
1173 * bond_select_active_slave - select a new active slave, if needed
1174 * @bond: our bonding struct
1176 * This functions should be called when one of the following occurs:
1177 * - The old curr_active_slave has been released or lost its link.
1178 * - The primary_slave has got its link back.
1179 * - A slave has got its link back and there's no old curr_active_slave.
1181 * Caller must hold bond->lock for read and curr_slave_lock for write_bh.
1183 void bond_select_active_slave(struct bonding *bond)
1185 struct slave *best_slave;
1188 best_slave = bond_find_best_slave(bond);
1189 if (best_slave != bond->curr_active_slave) {
1190 bond_change_active_slave(bond, best_slave);
1191 rv = bond_set_carrier(bond);
1195 if (netif_carrier_ok(bond->dev)) {
1196 pr_info("%s: first active interface up!\n",
1199 pr_info("%s: now running without any active interface !\n",
1205 /*--------------------------- slave list handling ---------------------------*/
1208 * This function attaches the slave to the end of list.
1210 * bond->lock held for writing by caller.
1212 static void bond_attach_slave(struct bonding *bond, struct slave *new_slave)
1214 if (bond->first_slave == NULL) { /* attaching the first slave */
1215 new_slave->next = new_slave;
1216 new_slave->prev = new_slave;
1217 bond->first_slave = new_slave;
1219 new_slave->next = bond->first_slave;
1220 new_slave->prev = bond->first_slave->prev;
1221 new_slave->next->prev = new_slave;
1222 new_slave->prev->next = new_slave;
1229 * This function detaches the slave from the list.
1230 * WARNING: no check is made to verify if the slave effectively
1231 * belongs to <bond>.
1232 * Nothing is freed on return, structures are just unchained.
1233 * If any slave pointer in bond was pointing to <slave>,
1234 * it should be changed by the calling function.
1236 * bond->lock held for writing by caller.
1238 static void bond_detach_slave(struct bonding *bond, struct slave *slave)
1241 slave->next->prev = slave->prev;
1244 slave->prev->next = slave->next;
1246 if (bond->first_slave == slave) { /* slave is the first slave */
1247 if (bond->slave_cnt > 1) { /* there are more slave */
1248 bond->first_slave = slave->next;
1250 bond->first_slave = NULL; /* slave was the last one */
1259 /*---------------------------------- IOCTL ----------------------------------*/
1261 static int bond_sethwaddr(struct net_device *bond_dev,
1262 struct net_device *slave_dev)
1264 pr_debug("bond_dev=%p\n", bond_dev);
1265 pr_debug("slave_dev=%p\n", slave_dev);
1266 pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len);
1267 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
1271 #define BOND_VLAN_FEATURES \
1272 (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \
1273 NETIF_F_HW_VLAN_FILTER)
1276 * Compute the common dev->feature set available to all slaves. Some
1277 * feature bits are managed elsewhere, so preserve those feature bits
1278 * on the master device.
1280 static int bond_compute_features(struct bonding *bond)
1282 struct slave *slave;
1283 struct net_device *bond_dev = bond->dev;
1284 unsigned long features = bond_dev->features;
1285 unsigned long vlan_features = 0;
1286 unsigned short max_hard_header_len = max((u16)ETH_HLEN,
1287 bond_dev->hard_header_len);
1290 features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES);
1291 features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM;
1293 if (!bond->first_slave)
1296 features &= ~NETIF_F_ONE_FOR_ALL;
1298 vlan_features = bond->first_slave->dev->vlan_features;
1299 bond_for_each_slave(bond, slave, i) {
1300 features = netdev_increment_features(features,
1301 slave->dev->features,
1302 NETIF_F_ONE_FOR_ALL);
1303 vlan_features = netdev_increment_features(vlan_features,
1304 slave->dev->vlan_features,
1305 NETIF_F_ONE_FOR_ALL);
1306 if (slave->dev->hard_header_len > max_hard_header_len)
1307 max_hard_header_len = slave->dev->hard_header_len;
1311 features |= (bond_dev->features & BOND_VLAN_FEATURES);
1312 bond_dev->features = netdev_fix_features(features, NULL);
1313 bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL);
1314 bond_dev->hard_header_len = max_hard_header_len;
1319 static void bond_setup_by_slave(struct net_device *bond_dev,
1320 struct net_device *slave_dev)
1322 struct bonding *bond = netdev_priv(bond_dev);
1324 bond_dev->header_ops = slave_dev->header_ops;
1326 bond_dev->type = slave_dev->type;
1327 bond_dev->hard_header_len = slave_dev->hard_header_len;
1328 bond_dev->addr_len = slave_dev->addr_len;
1330 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1331 slave_dev->addr_len);
1332 bond->setup_by_slave = 1;
1335 /* enslave device <slave> to bond device <master> */
1336 int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1338 struct bonding *bond = netdev_priv(bond_dev);
1339 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1340 struct slave *new_slave = NULL;
1341 struct netdev_hw_addr *ha;
1342 struct sockaddr addr;
1344 int old_features = bond_dev->features;
1347 if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL &&
1348 slave_ops->ndo_do_ioctl == NULL) {
1349 pr_warning("%s: Warning: no link monitoring support for %s\n",
1350 bond_dev->name, slave_dev->name);
1353 /* bond must be initialized by bond_open() before enslaving */
1354 if (!(bond_dev->flags & IFF_UP)) {
1355 pr_warning("%s: master_dev is not up in bond_enslave\n",
1359 /* already enslaved */
1360 if (slave_dev->flags & IFF_SLAVE) {
1361 pr_debug("Error, Device was already enslaved\n");
1365 /* vlan challenged mutual exclusion */
1366 /* no need to lock since we're protected by rtnl_lock */
1367 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
1368 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1369 if (!list_empty(&bond->vlan_list)) {
1370 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1371 bond_dev->name, slave_dev->name, bond_dev->name);
1374 pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1375 bond_dev->name, slave_dev->name,
1376 slave_dev->name, bond_dev->name);
1377 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
1380 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1381 if (bond->slave_cnt == 0) {
1382 /* First slave, and it is not VLAN challenged,
1383 * so remove the block of adding VLANs over the bond.
1385 bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
1390 * Old ifenslave binaries are no longer supported. These can
1391 * be identified with moderate accuracy by the state of the slave:
1392 * the current ifenslave will set the interface down prior to
1393 * enslaving it; the old ifenslave will not.
1395 if ((slave_dev->flags & IFF_UP)) {
1396 pr_err("%s is up. This may be due to an out of date ifenslave.\n",
1399 goto err_undo_flags;
1402 /* set bonding device ether type by slave - bonding netdevices are
1403 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1404 * there is a need to override some of the type dependent attribs/funcs.
1406 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1407 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1409 if (bond->slave_cnt == 0) {
1410 if (bond_dev->type != slave_dev->type) {
1411 pr_debug("%s: change device type from %d to %d\n",
1413 bond_dev->type, slave_dev->type);
1415 res = netdev_bonding_change(bond_dev,
1416 NETDEV_PRE_TYPE_CHANGE);
1417 res = notifier_to_errno(res);
1419 pr_err("%s: refused to change device type\n",
1422 goto err_undo_flags;
1425 /* Flush unicast and multicast addresses */
1426 dev_uc_flush(bond_dev);
1427 dev_mc_flush(bond_dev);
1429 if (slave_dev->type != ARPHRD_ETHER)
1430 bond_setup_by_slave(bond_dev, slave_dev);
1432 ether_setup(bond_dev);
1434 netdev_bonding_change(bond_dev,
1435 NETDEV_POST_TYPE_CHANGE);
1437 } else if (bond_dev->type != slave_dev->type) {
1438 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n",
1440 slave_dev->type, bond_dev->type);
1442 goto err_undo_flags;
1445 if (slave_ops->ndo_set_mac_address == NULL) {
1446 if (bond->slave_cnt == 0) {
1447 pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.",
1449 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
1450 } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
1451 pr_err("%s: Error: The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active.\n",
1454 goto err_undo_flags;
1458 new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL);
1461 goto err_undo_flags;
1464 /* save slave's original flags before calling
1465 * netdev_set_master and dev_open
1467 new_slave->original_flags = slave_dev->flags;
1470 * Save slave's original ("permanent") mac address for modes
1471 * that need it, and for restoring it upon release, and then
1472 * set it to the master's address
1474 memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN);
1476 if (!bond->params.fail_over_mac) {
1478 * Set slave to master's mac address. The application already
1479 * set the master's mac address to that of the first slave
1481 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1482 addr.sa_family = slave_dev->type;
1483 res = dev_set_mac_address(slave_dev, &addr);
1485 pr_debug("Error %d calling set_mac_address\n", res);
1490 res = netdev_set_master(slave_dev, bond_dev);
1492 pr_debug("Error %d calling netdev_set_master\n", res);
1493 goto err_restore_mac;
1495 /* open the slave since the application closed it */
1496 res = dev_open(slave_dev);
1498 pr_debug("Opening slave %s failed\n", slave_dev->name);
1499 goto err_unset_master;
1502 new_slave->dev = slave_dev;
1503 slave_dev->priv_flags |= IFF_BONDING;
1505 if (bond_is_lb(bond)) {
1506 /* bond_alb_init_slave() must be called before all other stages since
1507 * it might fail and we do not want to have to undo everything
1509 res = bond_alb_init_slave(bond, new_slave);
1514 /* If the mode USES_PRIMARY, then the new slave gets the
1515 * master's promisc (and mc) settings only if it becomes the
1516 * curr_active_slave, and that is taken care of later when calling
1517 * bond_change_active()
1519 if (!USES_PRIMARY(bond->params.mode)) {
1520 /* set promiscuity level to new slave */
1521 if (bond_dev->flags & IFF_PROMISC) {
1522 res = dev_set_promiscuity(slave_dev, 1);
1527 /* set allmulti level to new slave */
1528 if (bond_dev->flags & IFF_ALLMULTI) {
1529 res = dev_set_allmulti(slave_dev, 1);
1534 netif_addr_lock_bh(bond_dev);
1535 /* upload master's mc_list to new slave */
1536 netdev_for_each_mc_addr(ha, bond_dev)
1537 dev_mc_add(slave_dev, ha->addr);
1538 netif_addr_unlock_bh(bond_dev);
1541 if (bond->params.mode == BOND_MODE_8023AD) {
1542 /* add lacpdu mc addr to mc list */
1543 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1545 dev_mc_add(slave_dev, lacpdu_multicast);
1548 bond_add_vlans_on_slave(bond, slave_dev);
1550 write_lock_bh(&bond->lock);
1552 bond_attach_slave(bond, new_slave);
1554 new_slave->delay = 0;
1555 new_slave->link_failure_count = 0;
1557 bond_compute_features(bond);
1559 write_unlock_bh(&bond->lock);
1561 read_lock(&bond->lock);
1563 new_slave->last_arp_rx = jiffies;
1565 if (bond->params.miimon && !bond->params.use_carrier) {
1566 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1568 if ((link_reporting == -1) && !bond->params.arp_interval) {
1570 * miimon is set but a bonded network driver
1571 * does not support ETHTOOL/MII and
1572 * arp_interval is not set. Note: if
1573 * use_carrier is enabled, we will never go
1574 * here (because netif_carrier is always
1575 * supported); thus, we don't need to change
1576 * the messages for netif_carrier.
1578 pr_warning("%s: Warning: 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",
1579 bond_dev->name, slave_dev->name);
1580 } else if (link_reporting == -1) {
1581 /* unable get link status using mii/ethtool */
1582 pr_warning("%s: Warning: 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",
1583 bond_dev->name, slave_dev->name);
1587 /* check for initial state */
1588 if (!bond->params.miimon ||
1589 (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) {
1590 if (bond->params.updelay) {
1591 pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n");
1592 new_slave->link = BOND_LINK_BACK;
1593 new_slave->delay = bond->params.updelay;
1595 pr_debug("Initial state of slave_dev is BOND_LINK_UP\n");
1596 new_slave->link = BOND_LINK_UP;
1598 new_slave->jiffies = jiffies;
1600 pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n");
1601 new_slave->link = BOND_LINK_DOWN;
1604 if (bond_update_speed_duplex(new_slave) &&
1605 (new_slave->link != BOND_LINK_DOWN)) {
1606 pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n",
1607 bond_dev->name, new_slave->dev->name);
1609 if (bond->params.mode == BOND_MODE_8023AD) {
1610 pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n",
1615 if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) {
1616 /* if there is a primary slave, remember it */
1617 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1618 bond->primary_slave = new_slave;
1619 bond->force_primary = true;
1623 write_lock_bh(&bond->curr_slave_lock);
1625 switch (bond->params.mode) {
1626 case BOND_MODE_ACTIVEBACKUP:
1627 bond_set_slave_inactive_flags(new_slave);
1628 bond_select_active_slave(bond);
1630 case BOND_MODE_8023AD:
1631 /* in 802.3ad mode, the internal mechanism
1632 * will activate the slaves in the selected
1635 bond_set_slave_inactive_flags(new_slave);
1636 /* if this is the first slave */
1637 if (bond->slave_cnt == 1) {
1638 SLAVE_AD_INFO(new_slave).id = 1;
1639 /* Initialize AD with the number of times that the AD timer is called in 1 second
1640 * can be called only after the mac address of the bond is set
1642 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL,
1643 bond->params.lacp_fast);
1645 SLAVE_AD_INFO(new_slave).id =
1646 SLAVE_AD_INFO(new_slave->prev).id + 1;
1649 bond_3ad_bind_slave(new_slave);
1653 new_slave->state = BOND_STATE_ACTIVE;
1654 bond_set_slave_inactive_flags(new_slave);
1655 bond_select_active_slave(bond);
1658 pr_debug("This slave is always active in trunk mode\n");
1660 /* always active in trunk mode */
1661 new_slave->state = BOND_STATE_ACTIVE;
1663 /* In trunking mode there is little meaning to curr_active_slave
1664 * anyway (it holds no special properties of the bond device),
1665 * so we can change it without calling change_active_interface()
1667 if (!bond->curr_active_slave)
1668 bond->curr_active_slave = new_slave;
1671 } /* switch(bond_mode) */
1673 write_unlock_bh(&bond->curr_slave_lock);
1675 bond_set_carrier(bond);
1677 read_unlock(&bond->lock);
1679 res = bond_create_slave_symlinks(bond_dev, slave_dev);
1683 pr_info("%s: enslaving %s as a%s interface with a%s link.\n",
1684 bond_dev->name, slave_dev->name,
1685 new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup",
1686 new_slave->link != BOND_LINK_DOWN ? "n up" : " down");
1688 /* enslave is successful */
1691 /* Undo stages on error */
1693 dev_close(slave_dev);
1696 netdev_set_master(slave_dev, NULL);
1699 if (!bond->params.fail_over_mac) {
1700 /* XXX TODO - fom follow mode needs to change master's
1701 * MAC if this slave's MAC is in use by the bond, or at
1702 * least print a warning.
1704 memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN);
1705 addr.sa_family = slave_dev->type;
1706 dev_set_mac_address(slave_dev, &addr);
1713 bond_dev->features = old_features;
1719 * Try to release the slave device <slave> from the bond device <master>
1720 * It is legal to access curr_active_slave without a lock because all the function
1723 * The rules for slave state should be:
1724 * for Active/Backup:
1725 * Active stays on all backups go down
1726 * for Bonded connections:
1727 * The first up interface should be left on and all others downed.
1729 int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1731 struct bonding *bond = netdev_priv(bond_dev);
1732 struct slave *slave, *oldcurrent;
1733 struct sockaddr addr;
1735 /* slave is not a slave or master is not master of this slave */
1736 if (!(slave_dev->flags & IFF_SLAVE) ||
1737 (slave_dev->master != bond_dev)) {
1738 pr_err("%s: Error: cannot release %s.\n",
1739 bond_dev->name, slave_dev->name);
1743 write_lock_bh(&bond->lock);
1745 slave = bond_get_slave_by_dev(bond, slave_dev);
1747 /* not a slave of this bond */
1748 pr_info("%s: %s not enslaved\n",
1749 bond_dev->name, slave_dev->name);
1750 write_unlock_bh(&bond->lock);
1754 if (!bond->params.fail_over_mac) {
1755 if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) &&
1756 bond->slave_cnt > 1)
1757 pr_warning("%s: Warning: 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",
1758 bond_dev->name, slave_dev->name,
1760 bond_dev->name, slave_dev->name);
1763 /* Inform AD package of unbinding of slave. */
1764 if (bond->params.mode == BOND_MODE_8023AD) {
1765 /* must be called before the slave is
1766 * detached from the list
1768 bond_3ad_unbind_slave(slave);
1771 pr_info("%s: releasing %s interface %s\n",
1773 (slave->state == BOND_STATE_ACTIVE) ? "active" : "backup",
1776 oldcurrent = bond->curr_active_slave;
1778 bond->current_arp_slave = NULL;
1780 /* release the slave from its bond */
1781 bond_detach_slave(bond, slave);
1783 bond_compute_features(bond);
1785 if (bond->primary_slave == slave)
1786 bond->primary_slave = NULL;
1788 if (oldcurrent == slave)
1789 bond_change_active_slave(bond, NULL);
1791 if (bond_is_lb(bond)) {
1792 /* Must be called only after the slave has been
1793 * detached from the list and the curr_active_slave
1794 * has been cleared (if our_slave == old_current),
1795 * but before a new active slave is selected.
1797 write_unlock_bh(&bond->lock);
1798 bond_alb_deinit_slave(bond, slave);
1799 write_lock_bh(&bond->lock);
1802 if (oldcurrent == slave) {
1804 * Note that we hold RTNL over this sequence, so there
1805 * is no concern that another slave add/remove event
1808 write_unlock_bh(&bond->lock);
1809 read_lock(&bond->lock);
1810 write_lock_bh(&bond->curr_slave_lock);
1812 bond_select_active_slave(bond);
1814 write_unlock_bh(&bond->curr_slave_lock);
1815 read_unlock(&bond->lock);
1816 write_lock_bh(&bond->lock);
1819 if (bond->slave_cnt == 0) {
1820 bond_set_carrier(bond);
1822 /* if the last slave was removed, zero the mac address
1823 * of the master so it will be set by the application
1824 * to the mac address of the first slave
1826 memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
1828 if (list_empty(&bond->vlan_list)) {
1829 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
1831 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
1832 bond_dev->name, bond_dev->name);
1833 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
1836 } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1837 !bond_has_challenged_slaves(bond)) {
1838 pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n",
1839 bond_dev->name, slave_dev->name, bond_dev->name);
1840 bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
1843 write_unlock_bh(&bond->lock);
1845 /* must do this from outside any spinlocks */
1846 bond_destroy_slave_symlinks(bond_dev, slave_dev);
1848 bond_del_vlans_from_slave(bond, slave_dev);
1850 /* If the mode USES_PRIMARY, then we should only remove its
1851 * promisc and mc settings if it was the curr_active_slave, but that was
1852 * already taken care of above when we detached the slave
1854 if (!USES_PRIMARY(bond->params.mode)) {
1855 /* unset promiscuity level from slave */
1856 if (bond_dev->flags & IFF_PROMISC)
1857 dev_set_promiscuity(slave_dev, -1);
1859 /* unset allmulti level from slave */
1860 if (bond_dev->flags & IFF_ALLMULTI)
1861 dev_set_allmulti(slave_dev, -1);
1863 /* flush master's mc_list from slave */
1864 netif_addr_lock_bh(bond_dev);
1865 bond_mc_list_flush(bond_dev, slave_dev);
1866 netif_addr_unlock_bh(bond_dev);
1869 netdev_set_master(slave_dev, NULL);
1871 /* close slave before restoring its mac address */
1872 dev_close(slave_dev);
1874 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
1875 /* restore original ("permanent") mac address */
1876 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
1877 addr.sa_family = slave_dev->type;
1878 dev_set_mac_address(slave_dev, &addr);
1881 slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB |
1882 IFF_SLAVE_INACTIVE | IFF_BONDING |
1887 return 0; /* deletion OK */
1891 * First release a slave and than destroy the bond if no more slaves are left.
1892 * Must be under rtnl_lock when this function is called.
1894 int bond_release_and_destroy(struct net_device *bond_dev,
1895 struct net_device *slave_dev)
1897 struct bonding *bond = netdev_priv(bond_dev);
1900 ret = bond_release(bond_dev, slave_dev);
1901 if ((ret == 0) && (bond->slave_cnt == 0)) {
1902 pr_info("%s: destroying bond %s.\n",
1903 bond_dev->name, bond_dev->name);
1904 unregister_netdevice(bond_dev);
1910 * This function releases all slaves.
1912 static int bond_release_all(struct net_device *bond_dev)
1914 struct bonding *bond = netdev_priv(bond_dev);
1915 struct slave *slave;
1916 struct net_device *slave_dev;
1917 struct sockaddr addr;
1919 write_lock_bh(&bond->lock);
1921 netif_carrier_off(bond_dev);
1923 if (bond->slave_cnt == 0)
1926 bond->current_arp_slave = NULL;
1927 bond->primary_slave = NULL;
1928 bond_change_active_slave(bond, NULL);
1930 while ((slave = bond->first_slave) != NULL) {
1931 /* Inform AD package of unbinding of slave
1932 * before slave is detached from the list.
1934 if (bond->params.mode == BOND_MODE_8023AD)
1935 bond_3ad_unbind_slave(slave);
1937 slave_dev = slave->dev;
1938 bond_detach_slave(bond, slave);
1940 /* now that the slave is detached, unlock and perform
1941 * all the undo steps that should not be called from
1944 write_unlock_bh(&bond->lock);
1946 if (bond_is_lb(bond)) {
1947 /* must be called only after the slave
1948 * has been detached from the list
1950 bond_alb_deinit_slave(bond, slave);
1953 bond_compute_features(bond);
1955 bond_destroy_slave_symlinks(bond_dev, slave_dev);
1956 bond_del_vlans_from_slave(bond, slave_dev);
1958 /* If the mode USES_PRIMARY, then we should only remove its
1959 * promisc and mc settings if it was the curr_active_slave, but that was
1960 * already taken care of above when we detached the slave
1962 if (!USES_PRIMARY(bond->params.mode)) {
1963 /* unset promiscuity level from slave */
1964 if (bond_dev->flags & IFF_PROMISC)
1965 dev_set_promiscuity(slave_dev, -1);
1967 /* unset allmulti level from slave */
1968 if (bond_dev->flags & IFF_ALLMULTI)
1969 dev_set_allmulti(slave_dev, -1);
1971 /* flush master's mc_list from slave */
1972 netif_addr_lock_bh(bond_dev);
1973 bond_mc_list_flush(bond_dev, slave_dev);
1974 netif_addr_unlock_bh(bond_dev);
1977 netdev_set_master(slave_dev, NULL);
1979 /* close slave before restoring its mac address */
1980 dev_close(slave_dev);
1982 if (!bond->params.fail_over_mac) {
1983 /* restore original ("permanent") mac address*/
1984 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
1985 addr.sa_family = slave_dev->type;
1986 dev_set_mac_address(slave_dev, &addr);
1989 slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB |
1990 IFF_SLAVE_INACTIVE);
1994 /* re-acquire the lock before getting the next slave */
1995 write_lock_bh(&bond->lock);
1998 /* zero the mac address of the master so it will be
1999 * set by the application to the mac address of the
2002 memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
2004 if (list_empty(&bond->vlan_list))
2005 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
2007 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
2008 bond_dev->name, bond_dev->name);
2009 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
2013 pr_info("%s: released all slaves\n", bond_dev->name);
2016 write_unlock_bh(&bond->lock);
2022 * This function changes the active slave to slave <slave_dev>.
2023 * It returns -EINVAL in the following cases.
2024 * - <slave_dev> is not found in the list.
2025 * - There is not active slave now.
2026 * - <slave_dev> is already active.
2027 * - The link state of <slave_dev> is not BOND_LINK_UP.
2028 * - <slave_dev> is not running.
2029 * In these cases, this function does nothing.
2030 * In the other cases, current_slave pointer is changed and 0 is returned.
2032 static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev)
2034 struct bonding *bond = netdev_priv(bond_dev);
2035 struct slave *old_active = NULL;
2036 struct slave *new_active = NULL;
2039 if (!USES_PRIMARY(bond->params.mode))
2042 /* Verify that master_dev is indeed the master of slave_dev */
2043 if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev))
2046 read_lock(&bond->lock);
2048 read_lock(&bond->curr_slave_lock);
2049 old_active = bond->curr_active_slave;
2050 read_unlock(&bond->curr_slave_lock);
2052 new_active = bond_get_slave_by_dev(bond, slave_dev);
2055 * Changing to the current active: do nothing; return success.
2057 if (new_active && (new_active == old_active)) {
2058 read_unlock(&bond->lock);
2064 (new_active->link == BOND_LINK_UP) &&
2065 IS_UP(new_active->dev)) {
2066 write_lock_bh(&bond->curr_slave_lock);
2067 bond_change_active_slave(bond, new_active);
2068 write_unlock_bh(&bond->curr_slave_lock);
2072 read_unlock(&bond->lock);
2077 static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
2079 struct bonding *bond = netdev_priv(bond_dev);
2081 info->bond_mode = bond->params.mode;
2082 info->miimon = bond->params.miimon;
2084 read_lock(&bond->lock);
2085 info->num_slaves = bond->slave_cnt;
2086 read_unlock(&bond->lock);
2091 static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
2093 struct bonding *bond = netdev_priv(bond_dev);
2094 struct slave *slave;
2095 int i, res = -ENODEV;
2097 read_lock(&bond->lock);
2099 bond_for_each_slave(bond, slave, i) {
2100 if (i == (int)info->slave_id) {
2102 strcpy(info->slave_name, slave->dev->name);
2103 info->link = slave->link;
2104 info->state = slave->state;
2105 info->link_failure_count = slave->link_failure_count;
2110 read_unlock(&bond->lock);
2115 /*-------------------------------- Monitoring -------------------------------*/
2118 static int bond_miimon_inspect(struct bonding *bond)
2120 struct slave *slave;
2121 int i, link_state, commit = 0;
2122 bool ignore_updelay;
2124 ignore_updelay = !bond->curr_active_slave ? true : false;
2126 bond_for_each_slave(bond, slave, i) {
2127 slave->new_link = BOND_LINK_NOCHANGE;
2129 link_state = bond_check_dev_link(bond, slave->dev, 0);
2131 switch (slave->link) {
2136 slave->link = BOND_LINK_FAIL;
2137 slave->delay = bond->params.downdelay;
2139 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n",
2141 (bond->params.mode ==
2142 BOND_MODE_ACTIVEBACKUP) ?
2143 ((slave->state == BOND_STATE_ACTIVE) ?
2144 "active " : "backup ") : "",
2146 bond->params.downdelay * bond->params.miimon);
2149 case BOND_LINK_FAIL:
2152 * recovered before downdelay expired
2154 slave->link = BOND_LINK_UP;
2155 slave->jiffies = jiffies;
2156 pr_info("%s: link status up again after %d ms for interface %s.\n",
2158 (bond->params.downdelay - slave->delay) *
2159 bond->params.miimon,
2164 if (slave->delay <= 0) {
2165 slave->new_link = BOND_LINK_DOWN;
2173 case BOND_LINK_DOWN:
2177 slave->link = BOND_LINK_BACK;
2178 slave->delay = bond->params.updelay;
2181 pr_info("%s: link status up for interface %s, enabling it in %d ms.\n",
2182 bond->dev->name, slave->dev->name,
2183 ignore_updelay ? 0 :
2184 bond->params.updelay *
2185 bond->params.miimon);
2188 case BOND_LINK_BACK:
2190 slave->link = BOND_LINK_DOWN;
2191 pr_info("%s: link status down again after %d ms for interface %s.\n",
2193 (bond->params.updelay - slave->delay) *
2194 bond->params.miimon,
2203 if (slave->delay <= 0) {
2204 slave->new_link = BOND_LINK_UP;
2206 ignore_updelay = false;
2218 static void bond_miimon_commit(struct bonding *bond)
2220 struct slave *slave;
2223 bond_for_each_slave(bond, slave, i) {
2224 switch (slave->new_link) {
2225 case BOND_LINK_NOCHANGE:
2229 slave->link = BOND_LINK_UP;
2230 slave->jiffies = jiffies;
2232 if (bond->params.mode == BOND_MODE_8023AD) {
2233 /* prevent it from being the active one */
2234 slave->state = BOND_STATE_BACKUP;
2235 } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2236 /* make it immediately active */
2237 slave->state = BOND_STATE_ACTIVE;
2238 } else if (slave != bond->primary_slave) {
2239 /* prevent it from being the active one */
2240 slave->state = BOND_STATE_BACKUP;
2243 pr_info("%s: link status definitely up for interface %s.\n",
2244 bond->dev->name, slave->dev->name);
2246 /* notify ad that the link status has changed */
2247 if (bond->params.mode == BOND_MODE_8023AD)
2248 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
2250 if (bond_is_lb(bond))
2251 bond_alb_handle_link_change(bond, slave,
2254 if (!bond->curr_active_slave ||
2255 (slave == bond->primary_slave))
2260 case BOND_LINK_DOWN:
2261 if (slave->link_failure_count < UINT_MAX)
2262 slave->link_failure_count++;
2264 slave->link = BOND_LINK_DOWN;
2266 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP ||
2267 bond->params.mode == BOND_MODE_8023AD)
2268 bond_set_slave_inactive_flags(slave);
2270 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2271 bond->dev->name, slave->dev->name);
2273 if (bond->params.mode == BOND_MODE_8023AD)
2274 bond_3ad_handle_link_change(slave,
2277 if (bond_is_lb(bond))
2278 bond_alb_handle_link_change(bond, slave,
2281 if (slave == bond->curr_active_slave)
2287 pr_err("%s: invalid new link %d on slave %s\n",
2288 bond->dev->name, slave->new_link,
2290 slave->new_link = BOND_LINK_NOCHANGE;
2297 write_lock_bh(&bond->curr_slave_lock);
2298 bond_select_active_slave(bond);
2299 write_unlock_bh(&bond->curr_slave_lock);
2302 bond_set_carrier(bond);
2308 * Really a wrapper that splits the mii monitor into two phases: an
2309 * inspection, then (if inspection indicates something needs to be done)
2310 * an acquisition of appropriate locks followed by a commit phase to
2311 * implement whatever link state changes are indicated.
2313 void bond_mii_monitor(struct work_struct *work)
2315 struct bonding *bond = container_of(work, struct bonding,
2318 read_lock(&bond->lock);
2319 if (bond->kill_timers)
2322 if (bond->slave_cnt == 0)
2325 if (bond->send_grat_arp) {
2326 read_lock(&bond->curr_slave_lock);
2327 bond_send_gratuitous_arp(bond);
2328 read_unlock(&bond->curr_slave_lock);
2331 if (bond->send_unsol_na) {
2332 read_lock(&bond->curr_slave_lock);
2333 bond_send_unsolicited_na(bond);
2334 read_unlock(&bond->curr_slave_lock);
2337 if (bond_miimon_inspect(bond)) {
2338 read_unlock(&bond->lock);
2340 read_lock(&bond->lock);
2342 bond_miimon_commit(bond);
2344 read_unlock(&bond->lock);
2345 rtnl_unlock(); /* might sleep, hold no other locks */
2346 read_lock(&bond->lock);
2350 if (bond->params.miimon)
2351 queue_delayed_work(bond->wq, &bond->mii_work,
2352 msecs_to_jiffies(bond->params.miimon));
2354 read_unlock(&bond->lock);
2357 static __be32 bond_glean_dev_ip(struct net_device *dev)
2359 struct in_device *idev;
2360 struct in_ifaddr *ifa;
2367 idev = __in_dev_get_rcu(dev);
2371 ifa = idev->ifa_list;
2375 addr = ifa->ifa_local;
2381 static int bond_has_this_ip(struct bonding *bond, __be32 ip)
2383 struct vlan_entry *vlan;
2385 if (ip == bond->master_ip)
2388 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
2389 if (ip == vlan->vlan_ip)
2397 * We go to the (large) trouble of VLAN tagging ARP frames because
2398 * switches in VLAN mode (especially if ports are configured as
2399 * "native" to a VLAN) might not pass non-tagged frames.
2401 static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id)
2403 struct sk_buff *skb;
2405 pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op,
2406 slave_dev->name, dest_ip, src_ip, vlan_id);
2408 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2409 NULL, slave_dev->dev_addr, NULL);
2412 pr_err("ARP packet allocation failed\n");
2416 skb = vlan_put_tag(skb, vlan_id);
2418 pr_err("failed to insert VLAN tag\n");
2426 static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2429 __be32 *targets = bond->params.arp_targets;
2430 struct vlan_entry *vlan;
2431 struct net_device *vlan_dev;
2435 for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) {
2438 pr_debug("basa: target %x\n", targets[i]);
2439 if (list_empty(&bond->vlan_list)) {
2440 pr_debug("basa: empty vlan: arp_send\n");
2441 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2442 bond->master_ip, 0);
2447 * If VLANs are configured, we do a route lookup to
2448 * determine which VLAN interface would be used, so we
2449 * can tag the ARP with the proper VLAN tag.
2451 memset(&fl, 0, sizeof(fl));
2452 fl.fl4_dst = targets[i];
2453 fl.fl4_tos = RTO_ONLINK;
2455 rv = ip_route_output_key(dev_net(bond->dev), &rt, &fl);
2457 if (net_ratelimit()) {
2458 pr_warning("%s: no route to arp_ip_target %pI4\n",
2459 bond->dev->name, &fl.fl4_dst);
2465 * This target is not on a VLAN
2467 if (rt->u.dst.dev == bond->dev) {
2469 pr_debug("basa: rtdev == bond->dev: arp_send\n");
2470 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2471 bond->master_ip, 0);
2476 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
2477 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
2478 if (vlan_dev == rt->u.dst.dev) {
2479 vlan_id = vlan->vlan_id;
2480 pr_debug("basa: vlan match on %s %d\n",
2481 vlan_dev->name, vlan_id);
2488 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2489 vlan->vlan_ip, vlan_id);
2493 if (net_ratelimit()) {
2494 pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2495 bond->dev->name, &fl.fl4_dst,
2496 rt->u.dst.dev ? rt->u.dst.dev->name : "NULL");
2503 * Kick out a gratuitous ARP for an IP on the bonding master plus one
2504 * for each VLAN above us.
2506 * Caller must hold curr_slave_lock for read or better
2508 static void bond_send_gratuitous_arp(struct bonding *bond)
2510 struct slave *slave = bond->curr_active_slave;
2511 struct vlan_entry *vlan;
2512 struct net_device *vlan_dev;
2514 pr_debug("bond_send_grat_arp: bond %s slave %s\n",
2515 bond->dev->name, slave ? slave->dev->name : "NULL");
2517 if (!slave || !bond->send_grat_arp ||
2518 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
2521 bond->send_grat_arp--;
2523 if (bond->master_ip) {
2524 bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip,
2525 bond->master_ip, 0);
2528 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
2529 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
2530 if (vlan->vlan_ip) {
2531 bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip,
2532 vlan->vlan_ip, vlan->vlan_id);
2537 static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
2540 __be32 *targets = bond->params.arp_targets;
2542 for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) {
2543 pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n",
2544 &sip, &tip, i, &targets[i],
2545 bond_has_this_ip(bond, tip));
2546 if (sip == targets[i]) {
2547 if (bond_has_this_ip(bond, tip))
2548 slave->last_arp_rx = jiffies;
2554 static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
2557 struct slave *slave;
2558 struct bonding *bond;
2559 unsigned char *arp_ptr;
2562 if (dev->priv_flags & IFF_802_1Q_VLAN) {
2564 * When using VLANS and bonding, dev and oriv_dev may be
2565 * incorrect if the physical interface supports VLAN
2566 * acceleration. With this change ARP validation now
2567 * works for hosts only reachable on the VLAN interface.
2569 dev = vlan_dev_real_dev(dev);
2570 orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif);
2573 if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER))
2576 bond = netdev_priv(dev);
2577 read_lock(&bond->lock);
2579 pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n",
2580 bond->dev->name, skb->dev ? skb->dev->name : "NULL",
2581 orig_dev ? orig_dev->name : "NULL");
2583 slave = bond_get_slave_by_dev(bond, orig_dev);
2584 if (!slave || !slave_do_arp_validate(bond, slave))
2587 if (!pskb_may_pull(skb, arp_hdr_len(dev)))
2591 if (arp->ar_hln != dev->addr_len ||
2592 skb->pkt_type == PACKET_OTHERHOST ||
2593 skb->pkt_type == PACKET_LOOPBACK ||
2594 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2595 arp->ar_pro != htons(ETH_P_IP) ||
2599 arp_ptr = (unsigned char *)(arp + 1);
2600 arp_ptr += dev->addr_len;
2601 memcpy(&sip, arp_ptr, 4);
2602 arp_ptr += 4 + dev->addr_len;
2603 memcpy(&tip, arp_ptr, 4);
2605 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
2606 bond->dev->name, slave->dev->name, slave->state,
2607 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2611 * Backup slaves won't see the ARP reply, but do come through
2612 * here for each ARP probe (so we swap the sip/tip to validate
2613 * the probe). In a "redundant switch, common router" type of
2614 * configuration, the ARP probe will (hopefully) travel from
2615 * the active, through one switch, the router, then the other
2616 * switch before reaching the backup.
2618 if (slave->state == BOND_STATE_ACTIVE)
2619 bond_validate_arp(bond, slave, sip, tip);
2621 bond_validate_arp(bond, slave, tip, sip);
2624 read_unlock(&bond->lock);
2627 return NET_RX_SUCCESS;
2631 * this function is called regularly to monitor each slave's link
2632 * ensuring that traffic is being sent and received when arp monitoring
2633 * is used in load-balancing mode. if the adapter has been dormant, then an
2634 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2635 * arp monitoring in active backup mode.
2637 void bond_loadbalance_arp_mon(struct work_struct *work)
2639 struct bonding *bond = container_of(work, struct bonding,
2641 struct slave *slave, *oldcurrent;
2642 int do_failover = 0;
2646 read_lock(&bond->lock);
2648 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2650 if (bond->kill_timers)
2653 if (bond->slave_cnt == 0)
2656 read_lock(&bond->curr_slave_lock);
2657 oldcurrent = bond->curr_active_slave;
2658 read_unlock(&bond->curr_slave_lock);
2660 /* see if any of the previous devices are up now (i.e. they have
2661 * xmt and rcv traffic). the curr_active_slave does not come into
2662 * the picture unless it is null. also, slave->jiffies is not needed
2663 * here because we send an arp on each slave and give a slave as
2664 * long as it needs to get the tx/rx within the delta.
2665 * TODO: what about up/down delay in arp mode? it wasn't here before
2668 bond_for_each_slave(bond, slave, i) {
2669 if (slave->link != BOND_LINK_UP) {
2670 if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) &&
2671 time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) {
2673 slave->link = BOND_LINK_UP;
2674 slave->state = BOND_STATE_ACTIVE;
2676 /* primary_slave has no meaning in round-robin
2677 * mode. the window of a slave being up and
2678 * curr_active_slave being null after enslaving
2682 pr_info("%s: link status definitely up for interface %s, ",
2687 pr_info("%s: interface %s is now up\n",
2693 /* slave->link == BOND_LINK_UP */
2695 /* not all switches will respond to an arp request
2696 * when the source ip is 0, so don't take the link down
2697 * if we don't know our ip yet
2699 if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) ||
2700 (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) {
2702 slave->link = BOND_LINK_DOWN;
2703 slave->state = BOND_STATE_BACKUP;
2705 if (slave->link_failure_count < UINT_MAX)
2706 slave->link_failure_count++;
2708 pr_info("%s: interface %s is now down.\n",
2712 if (slave == oldcurrent)
2717 /* note: if switch is in round-robin mode, all links
2718 * must tx arp to ensure all links rx an arp - otherwise
2719 * links may oscillate or not come up at all; if switch is
2720 * in something like xor mode, there is nothing we can
2721 * do - all replies will be rx'ed on same link causing slaves
2722 * to be unstable during low/no traffic periods
2724 if (IS_UP(slave->dev))
2725 bond_arp_send_all(bond, slave);
2729 write_lock_bh(&bond->curr_slave_lock);
2731 bond_select_active_slave(bond);
2733 write_unlock_bh(&bond->curr_slave_lock);
2737 if (bond->params.arp_interval)
2738 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
2740 read_unlock(&bond->lock);
2744 * Called to inspect slaves for active-backup mode ARP monitor link state
2745 * changes. Sets new_link in slaves to specify what action should take
2746 * place for the slave. Returns 0 if no changes are found, >0 if changes
2747 * to link states must be committed.
2749 * Called with bond->lock held for read.
2751 static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks)
2753 struct slave *slave;
2756 bond_for_each_slave(bond, slave, i) {
2757 slave->new_link = BOND_LINK_NOCHANGE;
2759 if (slave->link != BOND_LINK_UP) {
2760 if (time_before_eq(jiffies, slave_last_rx(bond, slave) +
2762 slave->new_link = BOND_LINK_UP;
2770 * Give slaves 2*delta after being enslaved or made
2771 * active. This avoids bouncing, as the last receive
2772 * times need a full ARP monitor cycle to be updated.
2774 if (!time_after_eq(jiffies, slave->jiffies +
2775 2 * delta_in_ticks))
2779 * Backup slave is down if:
2780 * - No current_arp_slave AND
2781 * - more than 3*delta since last receive AND
2782 * - the bond has an IP address
2784 * Note: a non-null current_arp_slave indicates
2785 * the curr_active_slave went down and we are
2786 * searching for a new one; under this condition
2787 * we only take the curr_active_slave down - this
2788 * gives each slave a chance to tx/rx traffic
2789 * before being taken out
2791 if (slave->state == BOND_STATE_BACKUP &&
2792 !bond->current_arp_slave &&
2793 time_after(jiffies, slave_last_rx(bond, slave) +
2794 3 * delta_in_ticks)) {
2795 slave->new_link = BOND_LINK_DOWN;
2800 * Active slave is down if:
2801 * - more than 2*delta since transmitting OR
2802 * - (more than 2*delta since receive AND
2803 * the bond has an IP address)
2805 if ((slave->state == BOND_STATE_ACTIVE) &&
2806 (time_after_eq(jiffies, dev_trans_start(slave->dev) +
2807 2 * delta_in_ticks) ||
2808 (time_after_eq(jiffies, slave_last_rx(bond, slave)
2809 + 2 * delta_in_ticks)))) {
2810 slave->new_link = BOND_LINK_DOWN;
2819 * Called to commit link state changes noted by inspection step of
2820 * active-backup mode ARP monitor.
2822 * Called with RTNL and bond->lock for read.
2824 static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks)
2826 struct slave *slave;
2829 bond_for_each_slave(bond, slave, i) {
2830 switch (slave->new_link) {
2831 case BOND_LINK_NOCHANGE:
2835 if ((!bond->curr_active_slave &&
2836 time_before_eq(jiffies,
2837 dev_trans_start(slave->dev) +
2839 bond->curr_active_slave != slave) {
2840 slave->link = BOND_LINK_UP;
2841 bond->current_arp_slave = NULL;
2843 pr_info("%s: link status definitely up for interface %s.\n",
2844 bond->dev->name, slave->dev->name);
2846 if (!bond->curr_active_slave ||
2847 (slave == bond->primary_slave))
2854 case BOND_LINK_DOWN:
2855 if (slave->link_failure_count < UINT_MAX)
2856 slave->link_failure_count++;
2858 slave->link = BOND_LINK_DOWN;
2859 bond_set_slave_inactive_flags(slave);
2861 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2862 bond->dev->name, slave->dev->name);
2864 if (slave == bond->curr_active_slave) {
2865 bond->current_arp_slave = NULL;
2872 pr_err("%s: impossible: new_link %d on slave %s\n",
2873 bond->dev->name, slave->new_link,
2880 write_lock_bh(&bond->curr_slave_lock);
2881 bond_select_active_slave(bond);
2882 write_unlock_bh(&bond->curr_slave_lock);
2885 bond_set_carrier(bond);
2889 * Send ARP probes for active-backup mode ARP monitor.
2891 * Called with bond->lock held for read.
2893 static void bond_ab_arp_probe(struct bonding *bond)
2895 struct slave *slave;
2898 read_lock(&bond->curr_slave_lock);
2900 if (bond->current_arp_slave && bond->curr_active_slave)
2901 pr_info("PROBE: c_arp %s && cas %s BAD\n",
2902 bond->current_arp_slave->dev->name,
2903 bond->curr_active_slave->dev->name);
2905 if (bond->curr_active_slave) {
2906 bond_arp_send_all(bond, bond->curr_active_slave);
2907 read_unlock(&bond->curr_slave_lock);
2911 read_unlock(&bond->curr_slave_lock);
2913 /* if we don't have a curr_active_slave, search for the next available
2914 * backup slave from the current_arp_slave and make it the candidate
2915 * for becoming the curr_active_slave
2918 if (!bond->current_arp_slave) {
2919 bond->current_arp_slave = bond->first_slave;
2920 if (!bond->current_arp_slave)
2924 bond_set_slave_inactive_flags(bond->current_arp_slave);
2926 /* search for next candidate */
2927 bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) {
2928 if (IS_UP(slave->dev)) {
2929 slave->link = BOND_LINK_BACK;
2930 bond_set_slave_active_flags(slave);
2931 bond_arp_send_all(bond, slave);
2932 slave->jiffies = jiffies;
2933 bond->current_arp_slave = slave;
2937 /* if the link state is up at this point, we
2938 * mark it down - this can happen if we have
2939 * simultaneous link failures and
2940 * reselect_active_interface doesn't make this
2941 * one the current slave so it is still marked
2942 * up when it is actually down
2944 if (slave->link == BOND_LINK_UP) {
2945 slave->link = BOND_LINK_DOWN;
2946 if (slave->link_failure_count < UINT_MAX)
2947 slave->link_failure_count++;
2949 bond_set_slave_inactive_flags(slave);
2951 pr_info("%s: backup interface %s is now down.\n",
2952 bond->dev->name, slave->dev->name);
2957 void bond_activebackup_arp_mon(struct work_struct *work)
2959 struct bonding *bond = container_of(work, struct bonding,
2963 read_lock(&bond->lock);
2965 if (bond->kill_timers)
2968 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2970 if (bond->slave_cnt == 0)
2973 if (bond->send_grat_arp) {
2974 read_lock(&bond->curr_slave_lock);
2975 bond_send_gratuitous_arp(bond);
2976 read_unlock(&bond->curr_slave_lock);
2979 if (bond->send_unsol_na) {
2980 read_lock(&bond->curr_slave_lock);
2981 bond_send_unsolicited_na(bond);
2982 read_unlock(&bond->curr_slave_lock);
2985 if (bond_ab_arp_inspect(bond, delta_in_ticks)) {
2986 read_unlock(&bond->lock);
2988 read_lock(&bond->lock);
2990 bond_ab_arp_commit(bond, delta_in_ticks);
2992 read_unlock(&bond->lock);
2994 read_lock(&bond->lock);
2997 bond_ab_arp_probe(bond);
3000 if (bond->params.arp_interval)
3001 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
3003 read_unlock(&bond->lock);
3006 /*------------------------------ proc/seq_file-------------------------------*/
3008 #ifdef CONFIG_PROC_FS
3010 static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos)
3011 __acquires(&dev_base_lock)
3012 __acquires(&bond->lock)
3014 struct bonding *bond = seq->private;
3016 struct slave *slave;
3019 /* make sure the bond won't be taken away */
3020 read_lock(&dev_base_lock);
3021 read_lock(&bond->lock);
3024 return SEQ_START_TOKEN;
3026 bond_for_each_slave(bond, slave, i) {
3034 static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3036 struct bonding *bond = seq->private;
3037 struct slave *slave = v;
3040 if (v == SEQ_START_TOKEN)
3041 return bond->first_slave;
3043 slave = slave->next;
3045 return (slave == bond->first_slave) ? NULL : slave;
3048 static void bond_info_seq_stop(struct seq_file *seq, void *v)
3049 __releases(&bond->lock)
3050 __releases(&dev_base_lock)
3052 struct bonding *bond = seq->private;
3054 read_unlock(&bond->lock);
3055 read_unlock(&dev_base_lock);
3058 static void bond_info_show_master(struct seq_file *seq)
3060 struct bonding *bond = seq->private;
3064 read_lock(&bond->curr_slave_lock);
3065 curr = bond->curr_active_slave;
3066 read_unlock(&bond->curr_slave_lock);
3068 seq_printf(seq, "Bonding Mode: %s",
3069 bond_mode_name(bond->params.mode));
3071 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP &&
3072 bond->params.fail_over_mac)
3073 seq_printf(seq, " (fail_over_mac %s)",
3074 fail_over_mac_tbl[bond->params.fail_over_mac].modename);
3076 seq_printf(seq, "\n");
3078 if (bond->params.mode == BOND_MODE_XOR ||
3079 bond->params.mode == BOND_MODE_8023AD) {
3080 seq_printf(seq, "Transmit Hash Policy: %s (%d)\n",
3081 xmit_hashtype_tbl[bond->params.xmit_policy].modename,
3082 bond->params.xmit_policy);
3085 if (USES_PRIMARY(bond->params.mode)) {
3086 seq_printf(seq, "Primary Slave: %s",
3087 (bond->primary_slave) ?
3088 bond->primary_slave->dev->name : "None");
3089 if (bond->primary_slave)
3090 seq_printf(seq, " (primary_reselect %s)",
3091 pri_reselect_tbl[bond->params.primary_reselect].modename);
3093 seq_printf(seq, "\nCurrently Active Slave: %s\n",
3094 (curr) ? curr->dev->name : "None");
3097 seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ?
3099 seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon);
3100 seq_printf(seq, "Up Delay (ms): %d\n",
3101 bond->params.updelay * bond->params.miimon);
3102 seq_printf(seq, "Down Delay (ms): %d\n",
3103 bond->params.downdelay * bond->params.miimon);
3106 /* ARP information */
3107 if (bond->params.arp_interval > 0) {
3109 seq_printf(seq, "ARP Polling Interval (ms): %d\n",
3110 bond->params.arp_interval);
3112 seq_printf(seq, "ARP IP target/s (n.n.n.n form):");
3114 for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) {
3115 if (!bond->params.arp_targets[i])
3118 seq_printf(seq, ",");
3119 seq_printf(seq, " %pI4", &bond->params.arp_targets[i]);
3122 seq_printf(seq, "\n");
3125 if (bond->params.mode == BOND_MODE_8023AD) {
3126 struct ad_info ad_info;
3128 seq_puts(seq, "\n802.3ad info\n");
3129 seq_printf(seq, "LACP rate: %s\n",
3130 (bond->params.lacp_fast) ? "fast" : "slow");
3131 seq_printf(seq, "Aggregator selection policy (ad_select): %s\n",
3132 ad_select_tbl[bond->params.ad_select].modename);
3134 if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
3135 seq_printf(seq, "bond %s has no active aggregator\n",
3138 seq_printf(seq, "Active Aggregator Info:\n");
3140 seq_printf(seq, "\tAggregator ID: %d\n",
3141 ad_info.aggregator_id);
3142 seq_printf(seq, "\tNumber of ports: %d\n",
3144 seq_printf(seq, "\tActor Key: %d\n",
3146 seq_printf(seq, "\tPartner Key: %d\n",
3147 ad_info.partner_key);
3148 seq_printf(seq, "\tPartner Mac Address: %pM\n",
3149 ad_info.partner_system);
3154 static void bond_info_show_slave(struct seq_file *seq,
3155 const struct slave *slave)
3157 struct bonding *bond = seq->private;
3159 seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name);
3160 seq_printf(seq, "MII Status: %s\n",
3161 (slave->link == BOND_LINK_UP) ? "up" : "down");
3162 seq_printf(seq, "Link Failure Count: %u\n",
3163 slave->link_failure_count);
3165 seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr);
3167 if (bond->params.mode == BOND_MODE_8023AD) {
3168 const struct aggregator *agg
3169 = SLAVE_AD_INFO(slave).port.aggregator;
3172 seq_printf(seq, "Aggregator ID: %d\n",
3173 agg->aggregator_identifier);
3175 seq_puts(seq, "Aggregator ID: N/A\n");
3179 static int bond_info_seq_show(struct seq_file *seq, void *v)
3181 if (v == SEQ_START_TOKEN) {
3182 seq_printf(seq, "%s\n", version);
3183 bond_info_show_master(seq);
3185 bond_info_show_slave(seq, v);
3190 static const struct seq_operations bond_info_seq_ops = {
3191 .start = bond_info_seq_start,
3192 .next = bond_info_seq_next,
3193 .stop = bond_info_seq_stop,
3194 .show = bond_info_seq_show,
3197 static int bond_info_open(struct inode *inode, struct file *file)
3199 struct seq_file *seq;
3200 struct proc_dir_entry *proc;
3203 res = seq_open(file, &bond_info_seq_ops);
3205 /* recover the pointer buried in proc_dir_entry data */
3206 seq = file->private_data;
3208 seq->private = proc->data;
3214 static const struct file_operations bond_info_fops = {
3215 .owner = THIS_MODULE,
3216 .open = bond_info_open,
3218 .llseek = seq_lseek,
3219 .release = seq_release,
3222 static void bond_create_proc_entry(struct bonding *bond)
3224 struct net_device *bond_dev = bond->dev;
3225 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
3228 bond->proc_entry = proc_create_data(bond_dev->name,
3229 S_IRUGO, bn->proc_dir,
3230 &bond_info_fops, bond);
3231 if (bond->proc_entry == NULL)
3232 pr_warning("Warning: Cannot create /proc/net/%s/%s\n",
3233 DRV_NAME, bond_dev->name);
3235 memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ);
3239 static void bond_remove_proc_entry(struct bonding *bond)
3241 struct net_device *bond_dev = bond->dev;
3242 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
3244 if (bn->proc_dir && bond->proc_entry) {
3245 remove_proc_entry(bond->proc_file_name, bn->proc_dir);
3246 memset(bond->proc_file_name, 0, IFNAMSIZ);
3247 bond->proc_entry = NULL;
3251 /* Create the bonding directory under /proc/net, if doesn't exist yet.
3252 * Caller must hold rtnl_lock.
3254 static void __net_init bond_create_proc_dir(struct bond_net *bn)
3256 if (!bn->proc_dir) {
3257 bn->proc_dir = proc_mkdir(DRV_NAME, bn->net->proc_net);
3259 pr_warning("Warning: cannot create /proc/net/%s\n",
3264 /* Destroy the bonding directory under /proc/net, if empty.
3265 * Caller must hold rtnl_lock.
3267 static void __net_exit bond_destroy_proc_dir(struct bond_net *bn)
3270 remove_proc_entry(DRV_NAME, bn->net->proc_net);
3271 bn->proc_dir = NULL;
3275 #else /* !CONFIG_PROC_FS */
3277 static void bond_create_proc_entry(struct bonding *bond)
3281 static void bond_remove_proc_entry(struct bonding *bond)
3285 static inline void bond_create_proc_dir(struct bond_net *bn)
3289 static inline void bond_destroy_proc_dir(struct bond_net *bn)
3293 #endif /* CONFIG_PROC_FS */
3296 /*-------------------------- netdev event handling --------------------------*/
3299 * Change device name
3301 static int bond_event_changename(struct bonding *bond)
3303 bond_remove_proc_entry(bond);
3304 bond_create_proc_entry(bond);
3309 static int bond_master_netdev_event(unsigned long event,
3310 struct net_device *bond_dev)
3312 struct bonding *event_bond = netdev_priv(bond_dev);
3315 case NETDEV_CHANGENAME:
3316 return bond_event_changename(event_bond);
3324 static int bond_slave_netdev_event(unsigned long event,
3325 struct net_device *slave_dev)
3327 struct net_device *bond_dev = slave_dev->master;
3328 struct bonding *bond = netdev_priv(bond_dev);
3331 case NETDEV_UNREGISTER:
3333 if (bond->setup_by_slave)
3334 bond_release_and_destroy(bond_dev, slave_dev);
3336 bond_release(bond_dev, slave_dev);
3340 if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) {
3341 struct slave *slave;
3343 slave = bond_get_slave_by_dev(bond, slave_dev);
3345 u16 old_speed = slave->speed;
3346 u16 old_duplex = slave->duplex;
3348 bond_update_speed_duplex(slave);
3350 if (bond_is_lb(bond))
3353 if (old_speed != slave->speed)
3354 bond_3ad_adapter_speed_changed(slave);
3355 if (old_duplex != slave->duplex)
3356 bond_3ad_adapter_duplex_changed(slave);
3363 * ... Or is it this?
3366 case NETDEV_CHANGEMTU:
3368 * TODO: Should slaves be allowed to
3369 * independently alter their MTU? For
3370 * an active-backup bond, slaves need
3371 * not be the same type of device, so
3372 * MTUs may vary. For other modes,
3373 * slaves arguably should have the
3374 * same MTUs. To do this, we'd need to
3375 * take over the slave's change_mtu
3376 * function for the duration of their
3380 case NETDEV_CHANGENAME:
3382 * TODO: handle changing the primary's name
3385 case NETDEV_FEAT_CHANGE:
3386 bond_compute_features(bond);
3396 * bond_netdev_event: handle netdev notifier chain events.
3398 * This function receives events for the netdev chain. The caller (an
3399 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
3400 * locks for us to safely manipulate the slave devices (RTNL lock,
3403 static int bond_netdev_event(struct notifier_block *this,
3404 unsigned long event, void *ptr)
3406 struct net_device *event_dev = (struct net_device *)ptr;
3408 pr_debug("event_dev: %s, event: %lx\n",
3409 event_dev ? event_dev->name : "None",
3412 if (!(event_dev->priv_flags & IFF_BONDING))
3415 if (event_dev->flags & IFF_MASTER) {
3416 pr_debug("IFF_MASTER\n");
3417 return bond_master_netdev_event(event, event_dev);
3420 if (event_dev->flags & IFF_SLAVE) {
3421 pr_debug("IFF_SLAVE\n");
3422 return bond_slave_netdev_event(event, event_dev);
3429 * bond_inetaddr_event: handle inetaddr notifier chain events.
3431 * We keep track of device IPs primarily to use as source addresses in
3432 * ARP monitor probes (rather than spewing out broadcasts all the time).
3434 * We track one IP for the main device (if it has one), plus one per VLAN.
3436 static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
3438 struct in_ifaddr *ifa = ptr;
3439 struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev;
3440 struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id);
3441 struct bonding *bond;
3442 struct vlan_entry *vlan;
3444 list_for_each_entry(bond, &bn->dev_list, bond_list) {
3445 if (bond->dev == event_dev) {
3448 bond->master_ip = ifa->ifa_local;
3451 bond->master_ip = bond_glean_dev_ip(bond->dev);
3458 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
3459 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
3460 if (vlan_dev == event_dev) {
3463 vlan->vlan_ip = ifa->ifa_local;
3467 bond_glean_dev_ip(vlan_dev);
3478 static struct notifier_block bond_netdev_notifier = {
3479 .notifier_call = bond_netdev_event,
3482 static struct notifier_block bond_inetaddr_notifier = {
3483 .notifier_call = bond_inetaddr_event,
3486 /*-------------------------- Packet type handling ---------------------------*/
3488 /* register to receive lacpdus on a bond */
3489 static void bond_register_lacpdu(struct bonding *bond)
3491 struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type);
3493 /* initialize packet type */
3494 pk_type->type = PKT_TYPE_LACPDU;
3495 pk_type->dev = bond->dev;
3496 pk_type->func = bond_3ad_lacpdu_recv;
3498 dev_add_pack(pk_type);
3501 /* unregister to receive lacpdus on a bond */
3502 static void bond_unregister_lacpdu(struct bonding *bond)
3504 dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type));
3507 void bond_register_arp(struct bonding *bond)
3509 struct packet_type *pt = &bond->arp_mon_pt;
3514 pt->type = htons(ETH_P_ARP);
3515 pt->dev = bond->dev;
3516 pt->func = bond_arp_rcv;
3520 void bond_unregister_arp(struct bonding *bond)
3522 struct packet_type *pt = &bond->arp_mon_pt;
3524 dev_remove_pack(pt);
3528 /*---------------------------- Hashing Policies -----------------------------*/
3531 * Hash for the output device based upon layer 2 and layer 3 data. If
3532 * the packet is not IP mimic bond_xmit_hash_policy_l2()
3534 static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count)
3536 struct ethhdr *data = (struct ethhdr *)skb->data;
3537 struct iphdr *iph = ip_hdr(skb);
3539 if (skb->protocol == htons(ETH_P_IP)) {
3540 return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^
3541 (data->h_dest[5] ^ data->h_source[5])) % count;
3544 return (data->h_dest[5] ^ data->h_source[5]) % count;
3548 * Hash for the output device based upon layer 3 and layer 4 data. If
3549 * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is
3550 * altogether not IP, mimic bond_xmit_hash_policy_l2()
3552 static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count)
3554 struct ethhdr *data = (struct ethhdr *)skb->data;
3555 struct iphdr *iph = ip_hdr(skb);
3556 __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl);
3559 if (skb->protocol == htons(ETH_P_IP)) {
3560 if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) &&
3561 (iph->protocol == IPPROTO_TCP ||
3562 iph->protocol == IPPROTO_UDP)) {
3563 layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1)));
3565 return (layer4_xor ^
3566 ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count;
3570 return (data->h_dest[5] ^ data->h_source[5]) % count;
3574 * Hash for the output device based upon layer 2 data
3576 static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count)
3578 struct ethhdr *data = (struct ethhdr *)skb->data;
3580 return (data->h_dest[5] ^ data->h_source[5]) % count;
3583 /*-------------------------- Device entry points ----------------------------*/
3585 static int bond_open(struct net_device *bond_dev)
3587 struct bonding *bond = netdev_priv(bond_dev);
3589 bond->kill_timers = 0;
3591 if (bond_is_lb(bond)) {
3592 /* bond_alb_initialize must be called before the timer
3595 if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) {
3596 /* something went wrong - fail the open operation */
3600 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3601 queue_delayed_work(bond->wq, &bond->alb_work, 0);
3604 if (bond->params.miimon) { /* link check interval, in milliseconds. */
3605 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
3606 queue_delayed_work(bond->wq, &bond->mii_work, 0);
3609 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
3610 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3611 INIT_DELAYED_WORK(&bond->arp_work,
3612 bond_activebackup_arp_mon);
3614 INIT_DELAYED_WORK(&bond->arp_work,
3615 bond_loadbalance_arp_mon);
3617 queue_delayed_work(bond->wq, &bond->arp_work, 0);
3618 if (bond->params.arp_validate)
3619 bond_register_arp(bond);
3622 if (bond->params.mode == BOND_MODE_8023AD) {
3623 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
3624 queue_delayed_work(bond->wq, &bond->ad_work, 0);
3625 /* register to receive LACPDUs */
3626 bond_register_lacpdu(bond);
3627 bond_3ad_initiate_agg_selection(bond, 1);
3633 static int bond_close(struct net_device *bond_dev)
3635 struct bonding *bond = netdev_priv(bond_dev);
3637 if (bond->params.mode == BOND_MODE_8023AD) {
3638 /* Unregister the receive of LACPDUs */
3639 bond_unregister_lacpdu(bond);
3642 if (bond->params.arp_validate)
3643 bond_unregister_arp(bond);
3645 write_lock_bh(&bond->lock);
3647 bond->send_grat_arp = 0;
3648 bond->send_unsol_na = 0;
3650 /* signal timers not to re-arm */
3651 bond->kill_timers = 1;
3653 write_unlock_bh(&bond->lock);
3655 if (bond->params.miimon) { /* link check interval, in milliseconds. */
3656 cancel_delayed_work(&bond->mii_work);
3659 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
3660 cancel_delayed_work(&bond->arp_work);
3663 switch (bond->params.mode) {
3664 case BOND_MODE_8023AD:
3665 cancel_delayed_work(&bond->ad_work);
3669 cancel_delayed_work(&bond->alb_work);
3676 if (bond_is_lb(bond)) {
3677 /* Must be called only after all
3678 * slaves have been released
3680 bond_alb_deinitialize(bond);
3686 static struct net_device_stats *bond_get_stats(struct net_device *bond_dev)
3688 struct bonding *bond = netdev_priv(bond_dev);
3689 struct net_device_stats *stats = &bond_dev->stats;
3690 struct net_device_stats local_stats;
3691 struct slave *slave;
3694 memset(&local_stats, 0, sizeof(struct net_device_stats));
3696 read_lock_bh(&bond->lock);
3698 bond_for_each_slave(bond, slave, i) {
3699 const struct net_device_stats *sstats = dev_get_stats(slave->dev);
3701 local_stats.rx_packets += sstats->rx_packets;
3702 local_stats.rx_bytes += sstats->rx_bytes;
3703 local_stats.rx_errors += sstats->rx_errors;
3704 local_stats.rx_dropped += sstats->rx_dropped;
3706 local_stats.tx_packets += sstats->tx_packets;
3707 local_stats.tx_bytes += sstats->tx_bytes;
3708 local_stats.tx_errors += sstats->tx_errors;
3709 local_stats.tx_dropped += sstats->tx_dropped;
3711 local_stats.multicast += sstats->multicast;
3712 local_stats.collisions += sstats->collisions;
3714 local_stats.rx_length_errors += sstats->rx_length_errors;
3715 local_stats.rx_over_errors += sstats->rx_over_errors;
3716 local_stats.rx_crc_errors += sstats->rx_crc_errors;
3717 local_stats.rx_frame_errors += sstats->rx_frame_errors;
3718 local_stats.rx_fifo_errors += sstats->rx_fifo_errors;
3719 local_stats.rx_missed_errors += sstats->rx_missed_errors;
3721 local_stats.tx_aborted_errors += sstats->tx_aborted_errors;
3722 local_stats.tx_carrier_errors += sstats->tx_carrier_errors;
3723 local_stats.tx_fifo_errors += sstats->tx_fifo_errors;
3724 local_stats.tx_heartbeat_errors += sstats->tx_heartbeat_errors;
3725 local_stats.tx_window_errors += sstats->tx_window_errors;
3728 memcpy(stats, &local_stats, sizeof(struct net_device_stats));
3730 read_unlock_bh(&bond->lock);
3735 static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3737 struct net_device *slave_dev = NULL;
3738 struct ifbond k_binfo;
3739 struct ifbond __user *u_binfo = NULL;
3740 struct ifslave k_sinfo;
3741 struct ifslave __user *u_sinfo = NULL;
3742 struct mii_ioctl_data *mii = NULL;
3745 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd);
3757 * We do this again just in case we were called by SIOCGMIIREG
3758 * instead of SIOCGMIIPHY.
3765 if (mii->reg_num == 1) {
3766 struct bonding *bond = netdev_priv(bond_dev);
3768 read_lock(&bond->lock);
3769 read_lock(&bond->curr_slave_lock);
3770 if (netif_carrier_ok(bond->dev))
3771 mii->val_out = BMSR_LSTATUS;
3773 read_unlock(&bond->curr_slave_lock);
3774 read_unlock(&bond->lock);
3778 case BOND_INFO_QUERY_OLD:
3779 case SIOCBONDINFOQUERY:
3780 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3782 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
3785 res = bond_info_query(bond_dev, &k_binfo);
3787 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3791 case BOND_SLAVE_INFO_QUERY_OLD:
3792 case SIOCBONDSLAVEINFOQUERY:
3793 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3795 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
3798 res = bond_slave_info_query(bond_dev, &k_sinfo);
3800 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3809 if (!capable(CAP_NET_ADMIN))
3812 slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave);
3814 pr_debug("slave_dev=%p:\n", slave_dev);
3819 pr_debug("slave_dev->name=%s:\n", slave_dev->name);
3821 case BOND_ENSLAVE_OLD:
3822 case SIOCBONDENSLAVE:
3823 res = bond_enslave(bond_dev, slave_dev);
3825 case BOND_RELEASE_OLD:
3826 case SIOCBONDRELEASE:
3827 res = bond_release(bond_dev, slave_dev);
3829 case BOND_SETHWADDR_OLD:
3830 case SIOCBONDSETHWADDR:
3831 res = bond_sethwaddr(bond_dev, slave_dev);
3833 case BOND_CHANGE_ACTIVE_OLD:
3834 case SIOCBONDCHANGEACTIVE:
3835 res = bond_ioctl_change_active(bond_dev, slave_dev);
3847 static bool bond_addr_in_mc_list(unsigned char *addr,
3848 struct netdev_hw_addr_list *list,
3851 struct netdev_hw_addr *ha;
3853 netdev_hw_addr_list_for_each(ha, list)
3854 if (!memcmp(ha->addr, addr, addrlen))
3860 static void bond_set_multicast_list(struct net_device *bond_dev)
3862 struct bonding *bond = netdev_priv(bond_dev);
3863 struct netdev_hw_addr *ha;
3867 * Do promisc before checking multicast_mode
3869 if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC))
3871 * FIXME: Need to handle the error when one of the multi-slaves
3874 bond_set_promiscuity(bond, 1);
3877 if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC))
3878 bond_set_promiscuity(bond, -1);
3881 /* set allmulti flag to slaves */
3882 if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI))
3884 * FIXME: Need to handle the error when one of the multi-slaves
3887 bond_set_allmulti(bond, 1);
3890 if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI))
3891 bond_set_allmulti(bond, -1);
3894 read_lock(&bond->lock);
3896 bond->flags = bond_dev->flags;
3898 /* looking for addresses to add to slaves' mc list */
3899 netdev_for_each_mc_addr(ha, bond_dev) {
3900 found = bond_addr_in_mc_list(ha->addr, &bond->mc_list,
3901 bond_dev->addr_len);
3903 bond_mc_add(bond, ha->addr);
3906 /* looking for addresses to delete from slaves' list */
3907 netdev_hw_addr_list_for_each(ha, &bond->mc_list) {
3908 found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc,
3909 bond_dev->addr_len);
3911 bond_mc_del(bond, ha->addr);
3914 /* save master's multicast list */
3915 __hw_addr_flush(&bond->mc_list);
3916 __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc,
3917 bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST);
3919 read_unlock(&bond->lock);
3922 static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms)
3924 struct bonding *bond = netdev_priv(dev);
3925 struct slave *slave = bond->first_slave;
3928 const struct net_device_ops *slave_ops
3929 = slave->dev->netdev_ops;
3930 if (slave_ops->ndo_neigh_setup)
3931 return slave_ops->ndo_neigh_setup(slave->dev, parms);
3937 * Change the MTU of all of a master's slaves to match the master
3939 static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3941 struct bonding *bond = netdev_priv(bond_dev);
3942 struct slave *slave, *stop_at;
3946 pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond,
3947 (bond_dev ? bond_dev->name : "None"), new_mtu);
3949 /* Can't hold bond->lock with bh disabled here since
3950 * some base drivers panic. On the other hand we can't
3951 * hold bond->lock without bh disabled because we'll
3952 * deadlock. The only solution is to rely on the fact
3953 * that we're under rtnl_lock here, and the slaves
3954 * list won't change. This doesn't solve the problem
3955 * of setting the slave's MTU while it is
3956 * transmitting, but the assumption is that the base
3957 * driver can handle that.
3959 * TODO: figure out a way to safely iterate the slaves
3960 * list, but without holding a lock around the actual
3961 * call to the base driver.
3964 bond_for_each_slave(bond, slave, i) {
3965 pr_debug("s %p s->p %p c_m %p\n",
3968 slave->dev->netdev_ops->ndo_change_mtu);
3970 res = dev_set_mtu(slave->dev, new_mtu);
3973 /* If we failed to set the slave's mtu to the new value
3974 * we must abort the operation even in ACTIVE_BACKUP
3975 * mode, because if we allow the backup slaves to have
3976 * different mtu values than the active slave we'll
3977 * need to change their mtu when doing a failover. That
3978 * means changing their mtu from timer context, which
3979 * is probably not a good idea.
3981 pr_debug("err %d %s\n", res, slave->dev->name);
3986 bond_dev->mtu = new_mtu;
3991 /* unwind from head to the slave that failed */
3993 bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
3996 tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu);
3998 pr_debug("unwind err %d dev %s\n",
3999 tmp_res, slave->dev->name);
4009 * Note that many devices must be down to change the HW address, and
4010 * downing the master releases all slaves. We can make bonds full of
4011 * bonding devices to test this, however.
4013 static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
4015 struct bonding *bond = netdev_priv(bond_dev);
4016 struct sockaddr *sa = addr, tmp_sa;
4017 struct slave *slave, *stop_at;
4021 if (bond->params.mode == BOND_MODE_ALB)
4022 return bond_alb_set_mac_address(bond_dev, addr);
4025 pr_debug("bond=%p, name=%s\n",
4026 bond, bond_dev ? bond_dev->name : "None");
4029 * If fail_over_mac is set to active, do nothing and return
4030 * success. Returning an error causes ifenslave to fail.
4032 if (bond->params.fail_over_mac == BOND_FOM_ACTIVE)
4035 if (!is_valid_ether_addr(sa->sa_data))
4036 return -EADDRNOTAVAIL;
4038 /* Can't hold bond->lock with bh disabled here since
4039 * some base drivers panic. On the other hand we can't
4040 * hold bond->lock without bh disabled because we'll
4041 * deadlock. The only solution is to rely on the fact
4042 * that we're under rtnl_lock here, and the slaves
4043 * list won't change. This doesn't solve the problem
4044 * of setting the slave's hw address while it is
4045 * transmitting, but the assumption is that the base
4046 * driver can handle that.
4048 * TODO: figure out a way to safely iterate the slaves
4049 * list, but without holding a lock around the actual
4050 * call to the base driver.
4053 bond_for_each_slave(bond, slave, i) {
4054 const struct net_device_ops *slave_ops = slave->dev->netdev_ops;
4055 pr_debug("slave %p %s\n", slave, slave->dev->name);
4057 if (slave_ops->ndo_set_mac_address == NULL) {
4059 pr_debug("EOPNOTSUPP %s\n", slave->dev->name);
4063 res = dev_set_mac_address(slave->dev, addr);
4065 /* TODO: consider downing the slave
4067 * User should expect communications
4068 * breakage anyway until ARP finish
4071 pr_debug("err %d %s\n", res, slave->dev->name);
4077 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
4081 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
4082 tmp_sa.sa_family = bond_dev->type;
4084 /* unwind from head to the slave that failed */
4086 bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
4089 tmp_res = dev_set_mac_address(slave->dev, &tmp_sa);
4091 pr_debug("unwind err %d dev %s\n",
4092 tmp_res, slave->dev->name);
4099 static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
4101 struct bonding *bond = netdev_priv(bond_dev);
4102 struct slave *slave, *start_at;
4103 int i, slave_no, res = 1;
4104 struct iphdr *iph = ip_hdr(skb);
4106 read_lock(&bond->lock);
4108 if (!BOND_IS_OK(bond))
4111 * Start with the curr_active_slave that joined the bond as the
4112 * default for sending IGMP traffic. For failover purposes one
4113 * needs to maintain some consistency for the interface that will
4114 * send the join/membership reports. The curr_active_slave found
4115 * will send all of this type of traffic.
4117 if ((iph->protocol == IPPROTO_IGMP) &&
4118 (skb->protocol == htons(ETH_P_IP))) {
4120 read_lock(&bond->curr_slave_lock);
4121 slave = bond->curr_active_slave;
4122 read_unlock(&bond->curr_slave_lock);
4128 * Concurrent TX may collide on rr_tx_counter; we accept
4129 * that as being rare enough not to justify using an
4132 slave_no = bond->rr_tx_counter++ % bond->slave_cnt;
4134 bond_for_each_slave(bond, slave, i) {
4142 bond_for_each_slave_from(bond, slave, i, start_at) {
4143 if (IS_UP(slave->dev) &&
4144 (slave->link == BOND_LINK_UP) &&
4145 (slave->state == BOND_STATE_ACTIVE)) {
4146 res = bond_dev_queue_xmit(bond, skb, slave->dev);
4153 /* no suitable interface, frame not sent */
4156 read_unlock(&bond->lock);
4157 return NETDEV_TX_OK;
4162 * in active-backup mode, we know that bond->curr_active_slave is always valid if
4163 * the bond has a usable interface.
4165 static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
4167 struct bonding *bond = netdev_priv(bond_dev);
4170 read_lock(&bond->lock);
4171 read_lock(&bond->curr_slave_lock);
4173 if (!BOND_IS_OK(bond))
4176 if (!bond->curr_active_slave)
4179 res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev);
4183 /* no suitable interface, frame not sent */
4186 read_unlock(&bond->curr_slave_lock);
4187 read_unlock(&bond->lock);
4188 return NETDEV_TX_OK;
4192 * In bond_xmit_xor() , we determine the output device by using a pre-
4193 * determined xmit_hash_policy(), If the selected device is not enabled,
4194 * find the next active slave.
4196 static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
4198 struct bonding *bond = netdev_priv(bond_dev);
4199 struct slave *slave, *start_at;
4204 read_lock(&bond->lock);
4206 if (!BOND_IS_OK(bond))
4209 slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt);
4211 bond_for_each_slave(bond, slave, i) {
4219 bond_for_each_slave_from(bond, slave, i, start_at) {
4220 if (IS_UP(slave->dev) &&
4221 (slave->link == BOND_LINK_UP) &&
4222 (slave->state == BOND_STATE_ACTIVE)) {
4223 res = bond_dev_queue_xmit(bond, skb, slave->dev);
4230 /* no suitable interface, frame not sent */
4233 read_unlock(&bond->lock);
4234 return NETDEV_TX_OK;
4238 * in broadcast mode, we send everything to all usable interfaces.
4240 static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
4242 struct bonding *bond = netdev_priv(bond_dev);
4243 struct slave *slave, *start_at;
4244 struct net_device *tx_dev = NULL;
4248 read_lock(&bond->lock);
4250 if (!BOND_IS_OK(bond))
4253 read_lock(&bond->curr_slave_lock);
4254 start_at = bond->curr_active_slave;
4255 read_unlock(&bond->curr_slave_lock);
4260 bond_for_each_slave_from(bond, slave, i, start_at) {
4261 if (IS_UP(slave->dev) &&
4262 (slave->link == BOND_LINK_UP) &&
4263 (slave->state == BOND_STATE_ACTIVE)) {
4265 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
4267 pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n",
4272 res = bond_dev_queue_xmit(bond, skb2, tx_dev);
4274 dev_kfree_skb(skb2);
4278 tx_dev = slave->dev;
4283 res = bond_dev_queue_xmit(bond, skb, tx_dev);
4287 /* no suitable interface, frame not sent */
4290 /* frame sent to all suitable interfaces */
4291 read_unlock(&bond->lock);
4292 return NETDEV_TX_OK;
4295 /*------------------------- Device initialization ---------------------------*/
4297 static void bond_set_xmit_hash_policy(struct bonding *bond)
4299 switch (bond->params.xmit_policy) {
4300 case BOND_XMIT_POLICY_LAYER23:
4301 bond->xmit_hash_policy = bond_xmit_hash_policy_l23;
4303 case BOND_XMIT_POLICY_LAYER34:
4304 bond->xmit_hash_policy = bond_xmit_hash_policy_l34;
4306 case BOND_XMIT_POLICY_LAYER2:
4308 bond->xmit_hash_policy = bond_xmit_hash_policy_l2;
4313 static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
4315 const struct bonding *bond = netdev_priv(dev);
4317 switch (bond->params.mode) {
4318 case BOND_MODE_ROUNDROBIN:
4319 return bond_xmit_roundrobin(skb, dev);
4320 case BOND_MODE_ACTIVEBACKUP:
4321 return bond_xmit_activebackup(skb, dev);
4323 return bond_xmit_xor(skb, dev);
4324 case BOND_MODE_BROADCAST:
4325 return bond_xmit_broadcast(skb, dev);
4326 case BOND_MODE_8023AD:
4327 return bond_3ad_xmit_xor(skb, dev);
4330 return bond_alb_xmit(skb, dev);
4332 /* Should never happen, mode already checked */
4333 pr_err("%s: Error: Unknown bonding mode %d\n",
4334 dev->name, bond->params.mode);
4337 return NETDEV_TX_OK;
4343 * set bond mode specific net device operations
4345 void bond_set_mode_ops(struct bonding *bond, int mode)
4347 struct net_device *bond_dev = bond->dev;
4350 case BOND_MODE_ROUNDROBIN:
4352 case BOND_MODE_ACTIVEBACKUP:
4355 bond_set_xmit_hash_policy(bond);
4357 case BOND_MODE_BROADCAST:
4359 case BOND_MODE_8023AD:
4360 bond_set_master_3ad_flags(bond);
4361 bond_set_xmit_hash_policy(bond);
4364 bond_set_master_alb_flags(bond);
4369 /* Should never happen, mode already checked */
4370 pr_err("%s: Error: Unknown bonding mode %d\n",
4371 bond_dev->name, mode);
4376 static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
4377 struct ethtool_drvinfo *drvinfo)
4379 strncpy(drvinfo->driver, DRV_NAME, 32);
4380 strncpy(drvinfo->version, DRV_VERSION, 32);
4381 snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION);
4384 static const struct ethtool_ops bond_ethtool_ops = {
4385 .get_drvinfo = bond_ethtool_get_drvinfo,
4386 .get_link = ethtool_op_get_link,
4387 .get_tx_csum = ethtool_op_get_tx_csum,
4388 .get_sg = ethtool_op_get_sg,
4389 .get_tso = ethtool_op_get_tso,
4390 .get_ufo = ethtool_op_get_ufo,
4391 .get_flags = ethtool_op_get_flags,
4394 static const struct net_device_ops bond_netdev_ops = {
4395 .ndo_init = bond_init,
4396 .ndo_uninit = bond_uninit,
4397 .ndo_open = bond_open,
4398 .ndo_stop = bond_close,
4399 .ndo_start_xmit = bond_start_xmit,
4400 .ndo_get_stats = bond_get_stats,
4401 .ndo_do_ioctl = bond_do_ioctl,
4402 .ndo_set_multicast_list = bond_set_multicast_list,
4403 .ndo_change_mtu = bond_change_mtu,
4404 .ndo_set_mac_address = bond_set_mac_address,
4405 .ndo_neigh_setup = bond_neigh_setup,
4406 .ndo_vlan_rx_register = bond_vlan_rx_register,
4407 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
4408 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
4411 static void bond_destructor(struct net_device *bond_dev)
4413 struct bonding *bond = netdev_priv(bond_dev);
4415 destroy_workqueue(bond->wq);
4416 free_netdev(bond_dev);
4419 static void bond_setup(struct net_device *bond_dev)
4421 struct bonding *bond = netdev_priv(bond_dev);
4423 /* initialize rwlocks */
4424 rwlock_init(&bond->lock);
4425 rwlock_init(&bond->curr_slave_lock);
4427 bond->params = bonding_defaults;
4429 /* Initialize pointers */
4430 bond->dev = bond_dev;
4431 INIT_LIST_HEAD(&bond->vlan_list);
4433 /* Initialize the device entry points */
4434 ether_setup(bond_dev);
4435 bond_dev->netdev_ops = &bond_netdev_ops;
4436 bond_dev->ethtool_ops = &bond_ethtool_ops;
4437 bond_set_mode_ops(bond, bond->params.mode);
4439 bond_dev->destructor = bond_destructor;
4441 /* Initialize the device options */
4442 bond_dev->tx_queue_len = 0;
4443 bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
4444 bond_dev->priv_flags |= IFF_BONDING;
4445 bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
4447 if (bond->params.arp_interval)
4448 bond_dev->priv_flags |= IFF_MASTER_ARPMON;
4450 /* At first, we block adding VLANs. That's the only way to
4451 * prevent problems that occur when adding VLANs over an
4452 * empty bond. The block will be removed once non-challenged
4453 * slaves are enslaved.
4455 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
4457 /* don't acquire bond device's netif_tx_lock when
4459 bond_dev->features |= NETIF_F_LLTX;
4461 /* By default, we declare the bond to be fully
4462 * VLAN hardware accelerated capable. Special
4463 * care is taken in the various xmit functions
4464 * when there are slaves that are not hw accel
4467 bond_dev->features |= (NETIF_F_HW_VLAN_TX |
4468 NETIF_F_HW_VLAN_RX |
4469 NETIF_F_HW_VLAN_FILTER);
4473 static void bond_work_cancel_all(struct bonding *bond)
4475 write_lock_bh(&bond->lock);
4476 bond->kill_timers = 1;
4477 write_unlock_bh(&bond->lock);
4479 if (bond->params.miimon && delayed_work_pending(&bond->mii_work))
4480 cancel_delayed_work(&bond->mii_work);
4482 if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work))
4483 cancel_delayed_work(&bond->arp_work);
4485 if (bond->params.mode == BOND_MODE_ALB &&
4486 delayed_work_pending(&bond->alb_work))
4487 cancel_delayed_work(&bond->alb_work);
4489 if (bond->params.mode == BOND_MODE_8023AD &&
4490 delayed_work_pending(&bond->ad_work))
4491 cancel_delayed_work(&bond->ad_work);
4495 * Destroy a bonding device.
4496 * Must be under rtnl_lock when this function is called.
4498 static void bond_uninit(struct net_device *bond_dev)
4500 struct bonding *bond = netdev_priv(bond_dev);
4502 /* Release the bonded slaves */
4503 bond_release_all(bond_dev);
4505 list_del(&bond->bond_list);
4507 bond_work_cancel_all(bond);
4509 bond_remove_proc_entry(bond);
4511 __hw_addr_flush(&bond->mc_list);
4514 /*------------------------- Module initialization ---------------------------*/
4517 * Convert string input module parms. Accept either the
4518 * number of the mode or its string name. A bit complicated because
4519 * some mode names are substrings of other names, and calls from sysfs
4520 * may have whitespace in the name (trailing newlines, for example).
4522 int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl)
4524 int modeint = -1, i, rv;
4525 char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, };
4527 for (p = (char *)buf; *p; p++)
4528 if (!(isdigit(*p) || isspace(*p)))
4532 rv = sscanf(buf, "%20s", modestr);
4534 rv = sscanf(buf, "%d", &modeint);
4539 for (i = 0; tbl[i].modename; i++) {
4540 if (modeint == tbl[i].mode)
4542 if (strcmp(modestr, tbl[i].modename) == 0)
4549 static int bond_check_params(struct bond_params *params)
4551 int arp_validate_value, fail_over_mac_value, primary_reselect_value;
4554 * Convert string parameters.
4557 bond_mode = bond_parse_parm(mode, bond_mode_tbl);
4558 if (bond_mode == -1) {
4559 pr_err("Error: Invalid bonding mode \"%s\"\n",
4560 mode == NULL ? "NULL" : mode);
4565 if (xmit_hash_policy) {
4566 if ((bond_mode != BOND_MODE_XOR) &&
4567 (bond_mode != BOND_MODE_8023AD)) {
4568 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
4569 bond_mode_name(bond_mode));
4571 xmit_hashtype = bond_parse_parm(xmit_hash_policy,
4573 if (xmit_hashtype == -1) {
4574 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
4575 xmit_hash_policy == NULL ? "NULL" :
4583 if (bond_mode != BOND_MODE_8023AD) {
4584 pr_info("lacp_rate param is irrelevant in mode %s\n",
4585 bond_mode_name(bond_mode));
4587 lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl);
4588 if (lacp_fast == -1) {
4589 pr_err("Error: Invalid lacp rate \"%s\"\n",
4590 lacp_rate == NULL ? "NULL" : lacp_rate);
4597 params->ad_select = bond_parse_parm(ad_select, ad_select_tbl);
4598 if (params->ad_select == -1) {
4599 pr_err("Error: Invalid ad_select \"%s\"\n",
4600 ad_select == NULL ? "NULL" : ad_select);
4604 if (bond_mode != BOND_MODE_8023AD) {
4605 pr_warning("ad_select param only affects 802.3ad mode\n");
4608 params->ad_select = BOND_AD_STABLE;
4611 if (max_bonds < 0) {
4612 pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4613 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
4614 max_bonds = BOND_DEFAULT_MAX_BONDS;
4618 pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n",
4619 miimon, INT_MAX, BOND_LINK_MON_INTERV);
4620 miimon = BOND_LINK_MON_INTERV;
4624 pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4629 if (downdelay < 0) {
4630 pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4631 downdelay, INT_MAX);
4635 if ((use_carrier != 0) && (use_carrier != 1)) {
4636 pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4641 if (num_grat_arp < 0 || num_grat_arp > 255) {
4642 pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n",
4647 if (num_unsol_na < 0 || num_unsol_na > 255) {
4648 pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4653 /* reset values for 802.3ad */
4654 if (bond_mode == BOND_MODE_8023AD) {
4656 pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n");
4657 pr_warning("Forcing miimon to 100msec\n");
4662 /* reset values for TLB/ALB */
4663 if ((bond_mode == BOND_MODE_TLB) ||
4664 (bond_mode == BOND_MODE_ALB)) {
4666 pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure and link speed which are essential for TLB/ALB load balancing\n");
4667 pr_warning("Forcing miimon to 100msec\n");
4672 if (bond_mode == BOND_MODE_ALB) {
4673 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",
4678 if (updelay || downdelay) {
4679 /* just warn the user the up/down delay will have
4680 * no effect since miimon is zero...
4682 pr_warning("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",
4683 updelay, downdelay);
4686 /* don't allow arp monitoring */
4688 pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4689 miimon, arp_interval);
4693 if ((updelay % miimon) != 0) {
4694 pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4696 (updelay / miimon) * miimon);
4701 if ((downdelay % miimon) != 0) {
4702 pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4704 (downdelay / miimon) * miimon);
4707 downdelay /= miimon;
4710 if (arp_interval < 0) {
4711 pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n",
4712 arp_interval, INT_MAX, BOND_LINK_ARP_INTERV);
4713 arp_interval = BOND_LINK_ARP_INTERV;
4716 for (arp_ip_count = 0;
4717 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count];
4719 /* not complete check, but should be good enough to
4721 if (!isdigit(arp_ip_target[arp_ip_count][0])) {
4722 pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4723 arp_ip_target[arp_ip_count]);
4726 __be32 ip = in_aton(arp_ip_target[arp_ip_count]);
4727 arp_target[arp_ip_count] = ip;
4731 if (arp_interval && !arp_ip_count) {
4732 /* don't allow arping if no arp_ip_target given... */
4733 pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4739 if (bond_mode != BOND_MODE_ACTIVEBACKUP) {
4740 pr_err("arp_validate only supported in active-backup mode\n");
4743 if (!arp_interval) {
4744 pr_err("arp_validate requires arp_interval\n");
4748 arp_validate_value = bond_parse_parm(arp_validate,
4750 if (arp_validate_value == -1) {
4751 pr_err("Error: invalid arp_validate \"%s\"\n",
4752 arp_validate == NULL ? "NULL" : arp_validate);
4756 arp_validate_value = 0;
4759 pr_info("MII link monitoring set to %d ms\n", miimon);
4760 } else if (arp_interval) {
4763 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
4765 arp_validate_tbl[arp_validate_value].modename,
4768 for (i = 0; i < arp_ip_count; i++)
4769 pr_info(" %s", arp_ip_target[i]);
4773 } else if (max_bonds) {
4774 /* miimon and arp_interval not set, we need one so things
4775 * work as expected, see bonding.txt for details
4777 pr_warning("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");
4780 if (primary && !USES_PRIMARY(bond_mode)) {
4781 /* currently, using a primary only makes sense
4782 * in active backup, TLB or ALB modes
4784 pr_warning("Warning: %s primary device specified but has no effect in %s mode\n",
4785 primary, bond_mode_name(bond_mode));
4789 if (primary && primary_reselect) {
4790 primary_reselect_value = bond_parse_parm(primary_reselect,
4792 if (primary_reselect_value == -1) {
4793 pr_err("Error: Invalid primary_reselect \"%s\"\n",
4795 NULL ? "NULL" : primary_reselect);
4799 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4802 if (fail_over_mac) {
4803 fail_over_mac_value = bond_parse_parm(fail_over_mac,
4805 if (fail_over_mac_value == -1) {
4806 pr_err("Error: invalid fail_over_mac \"%s\"\n",
4807 arp_validate == NULL ? "NULL" : arp_validate);
4811 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
4812 pr_warning("Warning: fail_over_mac only affects active-backup mode.\n");
4814 fail_over_mac_value = BOND_FOM_NONE;
4817 /* fill params struct with the proper values */
4818 params->mode = bond_mode;
4819 params->xmit_policy = xmit_hashtype;
4820 params->miimon = miimon;
4821 params->num_grat_arp = num_grat_arp;
4822 params->num_unsol_na = num_unsol_na;
4823 params->arp_interval = arp_interval;
4824 params->arp_validate = arp_validate_value;
4825 params->updelay = updelay;
4826 params->downdelay = downdelay;
4827 params->use_carrier = use_carrier;
4828 params->lacp_fast = lacp_fast;
4829 params->primary[0] = 0;
4830 params->primary_reselect = primary_reselect_value;
4831 params->fail_over_mac = fail_over_mac_value;
4834 strncpy(params->primary, primary, IFNAMSIZ);
4835 params->primary[IFNAMSIZ - 1] = 0;
4838 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4843 static struct lock_class_key bonding_netdev_xmit_lock_key;
4844 static struct lock_class_key bonding_netdev_addr_lock_key;
4846 static void bond_set_lockdep_class_one(struct net_device *dev,
4847 struct netdev_queue *txq,
4850 lockdep_set_class(&txq->_xmit_lock,
4851 &bonding_netdev_xmit_lock_key);
4854 static void bond_set_lockdep_class(struct net_device *dev)
4856 lockdep_set_class(&dev->addr_list_lock,
4857 &bonding_netdev_addr_lock_key);
4858 netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL);
4862 * Called from registration process
4864 static int bond_init(struct net_device *bond_dev)
4866 struct bonding *bond = netdev_priv(bond_dev);
4867 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
4869 pr_debug("Begin bond_init for %s\n", bond_dev->name);
4871 bond->wq = create_singlethread_workqueue(bond_dev->name);
4875 bond_set_lockdep_class(bond_dev);
4877 netif_carrier_off(bond_dev);
4879 bond_create_proc_entry(bond);
4880 list_add_tail(&bond->bond_list, &bn->dev_list);
4882 bond_prepare_sysfs_group(bond);
4884 __hw_addr_init(&bond->mc_list);
4888 static int bond_validate(struct nlattr *tb[], struct nlattr *data[])
4890 if (tb[IFLA_ADDRESS]) {
4891 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
4893 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
4894 return -EADDRNOTAVAIL;
4899 static struct rtnl_link_ops bond_link_ops __read_mostly = {
4901 .priv_size = sizeof(struct bonding),
4902 .setup = bond_setup,
4903 .validate = bond_validate,
4906 /* Create a new bond based on the specified name and bonding parameters.
4907 * If name is NULL, obtain a suitable "bond%d" name for us.
4908 * Caller must NOT hold rtnl_lock; we need to release it here before we
4909 * set up our sysfs entries.
4911 int bond_create(struct net *net, const char *name)
4913 struct net_device *bond_dev;
4918 bond_dev = alloc_netdev(sizeof(struct bonding), name ? name : "",
4921 pr_err("%s: eek! can't alloc netdev!\n", name);
4926 dev_net_set(bond_dev, net);
4927 bond_dev->rtnl_link_ops = &bond_link_ops;
4930 res = dev_alloc_name(bond_dev, "bond%d");
4935 res = register_netdevice(bond_dev);
4940 bond_destructor(bond_dev);
4944 static int __net_init bond_net_init(struct net *net)
4946 struct bond_net *bn = net_generic(net, bond_net_id);
4949 INIT_LIST_HEAD(&bn->dev_list);
4951 bond_create_proc_dir(bn);
4956 static void __net_exit bond_net_exit(struct net *net)
4958 struct bond_net *bn = net_generic(net, bond_net_id);
4960 bond_destroy_proc_dir(bn);
4963 static struct pernet_operations bond_net_ops = {
4964 .init = bond_net_init,
4965 .exit = bond_net_exit,
4967 .size = sizeof(struct bond_net),
4970 static int __init bonding_init(void)
4975 pr_info("%s", version);
4977 res = bond_check_params(&bonding_defaults);
4981 res = register_pernet_subsys(&bond_net_ops);
4985 res = rtnl_link_register(&bond_link_ops);
4989 for (i = 0; i < max_bonds; i++) {
4990 res = bond_create(&init_net, NULL);
4995 res = bond_create_sysfs();
4999 register_netdevice_notifier(&bond_netdev_notifier);
5000 register_inetaddr_notifier(&bond_inetaddr_notifier);
5001 bond_register_ipv6_notifier();
5005 rtnl_link_unregister(&bond_link_ops);
5007 unregister_pernet_subsys(&bond_net_ops);
5012 static void __exit bonding_exit(void)
5014 unregister_netdevice_notifier(&bond_netdev_notifier);
5015 unregister_inetaddr_notifier(&bond_inetaddr_notifier);
5016 bond_unregister_ipv6_notifier();
5018 bond_destroy_sysfs();
5020 rtnl_link_unregister(&bond_link_ops);
5021 unregister_pernet_subsys(&bond_net_ops);
5024 module_init(bonding_init);
5025 module_exit(bonding_exit);
5026 MODULE_LICENSE("GPL");
5027 MODULE_VERSION(DRV_VERSION);
5028 MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
5029 MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
5030 MODULE_ALIAS_RTNL_LINK("bond");