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/netpoll.h>
63 #include <linux/inetdevice.h>
64 #include <linux/igmp.h>
65 #include <linux/etherdevice.h>
66 #include <linux/skbuff.h>
68 #include <linux/rtnetlink.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/smp.h>
72 #include <linux/if_ether.h>
74 #include <linux/mii.h>
75 #include <linux/ethtool.h>
76 #include <linux/if_vlan.h>
77 #include <linux/if_bonding.h>
78 #include <linux/jiffies.h>
79 #include <linux/preempt.h>
80 #include <net/route.h>
81 #include <net/net_namespace.h>
82 #include <net/netns/generic.h>
87 /*---------------------------- Module parameters ----------------------------*/
89 /* monitor all links that often (in milliseconds). <=0 disables monitoring */
90 #define BOND_LINK_MON_INTERV 0
91 #define BOND_LINK_ARP_INTERV 0
93 static int max_bonds
= BOND_DEFAULT_MAX_BONDS
;
94 static int tx_queues
= BOND_DEFAULT_TX_QUEUES
;
95 static int num_grat_arp
= 1;
96 static int num_unsol_na
= 1;
97 static int miimon
= BOND_LINK_MON_INTERV
;
100 static int use_carrier
= 1;
102 static char *primary
;
103 static char *primary_reselect
;
104 static char *lacp_rate
;
105 static char *ad_select
;
106 static char *xmit_hash_policy
;
107 static int arp_interval
= BOND_LINK_ARP_INTERV
;
108 static char *arp_ip_target
[BOND_MAX_ARP_TARGETS
];
109 static char *arp_validate
;
110 static char *fail_over_mac
;
111 static int all_slaves_active
= 0;
112 static struct bond_params bonding_defaults
;
113 static int resend_igmp
= BOND_DEFAULT_RESEND_IGMP
;
115 module_param(max_bonds
, int, 0);
116 MODULE_PARM_DESC(max_bonds
, "Max number of bonded devices");
117 module_param(tx_queues
, int, 0);
118 MODULE_PARM_DESC(tx_queues
, "Max number of transmit queues (default = 16)");
119 module_param(num_grat_arp
, int, 0644);
120 MODULE_PARM_DESC(num_grat_arp
, "Number of gratuitous ARP packets to send on failover event");
121 module_param(num_unsol_na
, int, 0644);
122 MODULE_PARM_DESC(num_unsol_na
, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event");
123 module_param(miimon
, int, 0);
124 MODULE_PARM_DESC(miimon
, "Link check interval in milliseconds");
125 module_param(updelay
, int, 0);
126 MODULE_PARM_DESC(updelay
, "Delay before considering link up, in milliseconds");
127 module_param(downdelay
, int, 0);
128 MODULE_PARM_DESC(downdelay
, "Delay before considering link down, "
130 module_param(use_carrier
, int, 0);
131 MODULE_PARM_DESC(use_carrier
, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
132 "0 for off, 1 for on (default)");
133 module_param(mode
, charp
, 0);
134 MODULE_PARM_DESC(mode
, "Mode of operation : 0 for balance-rr, "
135 "1 for active-backup, 2 for balance-xor, "
136 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
137 "6 for balance-alb");
138 module_param(primary
, charp
, 0);
139 MODULE_PARM_DESC(primary
, "Primary network device to use");
140 module_param(primary_reselect
, charp
, 0);
141 MODULE_PARM_DESC(primary_reselect
, "Reselect primary slave "
143 "0 for always (default), "
144 "1 for only if speed of primary is "
146 "2 for only on active slave "
148 module_param(lacp_rate
, charp
, 0);
149 MODULE_PARM_DESC(lacp_rate
, "LACPDU tx rate to request from 802.3ad partner "
151 module_param(ad_select
, charp
, 0);
152 MODULE_PARM_DESC(ad_select
, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)");
153 module_param(xmit_hash_policy
, charp
, 0);
154 MODULE_PARM_DESC(xmit_hash_policy
, "XOR hashing method: 0 for layer 2 (default)"
155 ", 1 for layer 3+4");
156 module_param(arp_interval
, int, 0);
157 MODULE_PARM_DESC(arp_interval
, "arp interval in milliseconds");
158 module_param_array(arp_ip_target
, charp
, NULL
, 0);
159 MODULE_PARM_DESC(arp_ip_target
, "arp targets in n.n.n.n form");
160 module_param(arp_validate
, charp
, 0);
161 MODULE_PARM_DESC(arp_validate
, "validate src/dst of ARP probes: none (default), active, backup or all");
162 module_param(fail_over_mac
, charp
, 0);
163 MODULE_PARM_DESC(fail_over_mac
, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow");
164 module_param(all_slaves_active
, int, 0);
165 MODULE_PARM_DESC(all_slaves_active
, "Keep all frames received on an interface"
166 "by setting active flag for all slaves. "
167 "0 for never (default), 1 for always.");
168 module_param(resend_igmp
, int, 0);
169 MODULE_PARM_DESC(resend_igmp
, "Number of IGMP membership reports to send on link failure");
171 /*----------------------------- Global variables ----------------------------*/
173 #ifdef CONFIG_NET_POLL_CONTROLLER
174 atomic_t netpoll_block_tx
= ATOMIC_INIT(0);
177 static const char * const version
=
178 DRV_DESCRIPTION
": v" DRV_VERSION
" (" DRV_RELDATE
")\n";
180 int bond_net_id __read_mostly
;
182 static __be32 arp_target
[BOND_MAX_ARP_TARGETS
];
183 static int arp_ip_count
;
184 static int bond_mode
= BOND_MODE_ROUNDROBIN
;
185 static int xmit_hashtype
= BOND_XMIT_POLICY_LAYER2
;
186 static int lacp_fast
;
188 const struct bond_parm_tbl bond_lacp_tbl
[] = {
189 { "slow", AD_LACP_SLOW
},
190 { "fast", AD_LACP_FAST
},
194 const struct bond_parm_tbl bond_mode_tbl
[] = {
195 { "balance-rr", BOND_MODE_ROUNDROBIN
},
196 { "active-backup", BOND_MODE_ACTIVEBACKUP
},
197 { "balance-xor", BOND_MODE_XOR
},
198 { "broadcast", BOND_MODE_BROADCAST
},
199 { "802.3ad", BOND_MODE_8023AD
},
200 { "balance-tlb", BOND_MODE_TLB
},
201 { "balance-alb", BOND_MODE_ALB
},
205 const struct bond_parm_tbl xmit_hashtype_tbl
[] = {
206 { "layer2", BOND_XMIT_POLICY_LAYER2
},
207 { "layer3+4", BOND_XMIT_POLICY_LAYER34
},
208 { "layer2+3", BOND_XMIT_POLICY_LAYER23
},
212 const struct bond_parm_tbl arp_validate_tbl
[] = {
213 { "none", BOND_ARP_VALIDATE_NONE
},
214 { "active", BOND_ARP_VALIDATE_ACTIVE
},
215 { "backup", BOND_ARP_VALIDATE_BACKUP
},
216 { "all", BOND_ARP_VALIDATE_ALL
},
220 const struct bond_parm_tbl fail_over_mac_tbl
[] = {
221 { "none", BOND_FOM_NONE
},
222 { "active", BOND_FOM_ACTIVE
},
223 { "follow", BOND_FOM_FOLLOW
},
227 const struct bond_parm_tbl pri_reselect_tbl
[] = {
228 { "always", BOND_PRI_RESELECT_ALWAYS
},
229 { "better", BOND_PRI_RESELECT_BETTER
},
230 { "failure", BOND_PRI_RESELECT_FAILURE
},
234 struct bond_parm_tbl ad_select_tbl
[] = {
235 { "stable", BOND_AD_STABLE
},
236 { "bandwidth", BOND_AD_BANDWIDTH
},
237 { "count", BOND_AD_COUNT
},
241 /*-------------------------- Forward declarations ---------------------------*/
243 static void bond_send_gratuitous_arp(struct bonding
*bond
);
244 static int bond_init(struct net_device
*bond_dev
);
245 static void bond_uninit(struct net_device
*bond_dev
);
247 /*---------------------------- General routines -----------------------------*/
249 static const char *bond_mode_name(int mode
)
251 static const char *names
[] = {
252 [BOND_MODE_ROUNDROBIN
] = "load balancing (round-robin)",
253 [BOND_MODE_ACTIVEBACKUP
] = "fault-tolerance (active-backup)",
254 [BOND_MODE_XOR
] = "load balancing (xor)",
255 [BOND_MODE_BROADCAST
] = "fault-tolerance (broadcast)",
256 [BOND_MODE_8023AD
] = "IEEE 802.3ad Dynamic link aggregation",
257 [BOND_MODE_TLB
] = "transmit load balancing",
258 [BOND_MODE_ALB
] = "adaptive load balancing",
261 if (mode
< 0 || mode
> BOND_MODE_ALB
)
267 /*---------------------------------- VLAN -----------------------------------*/
270 * bond_add_vlan - add a new vlan id on bond
271 * @bond: bond that got the notification
272 * @vlan_id: the vlan id to add
274 * Returns -ENOMEM if allocation failed.
276 static int bond_add_vlan(struct bonding
*bond
, unsigned short vlan_id
)
278 struct vlan_entry
*vlan
;
280 pr_debug("bond: %s, vlan id %d\n",
281 (bond
? bond
->dev
->name
: "None"), vlan_id
);
283 vlan
= kzalloc(sizeof(struct vlan_entry
), GFP_KERNEL
);
287 INIT_LIST_HEAD(&vlan
->vlan_list
);
288 vlan
->vlan_id
= vlan_id
;
290 write_lock_bh(&bond
->lock
);
292 list_add_tail(&vlan
->vlan_list
, &bond
->vlan_list
);
294 write_unlock_bh(&bond
->lock
);
296 pr_debug("added VLAN ID %d on bond %s\n", vlan_id
, bond
->dev
->name
);
302 * bond_del_vlan - delete a vlan id from bond
303 * @bond: bond that got the notification
304 * @vlan_id: the vlan id to delete
306 * returns -ENODEV if @vlan_id was not found in @bond.
308 static int bond_del_vlan(struct bonding
*bond
, unsigned short vlan_id
)
310 struct vlan_entry
*vlan
;
313 pr_debug("bond: %s, vlan id %d\n", bond
->dev
->name
, vlan_id
);
316 write_lock_bh(&bond
->lock
);
318 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
319 if (vlan
->vlan_id
== vlan_id
) {
320 list_del(&vlan
->vlan_list
);
322 if (bond_is_lb(bond
))
323 bond_alb_clear_vlan(bond
, vlan_id
);
325 pr_debug("removed VLAN ID %d from bond %s\n",
326 vlan_id
, bond
->dev
->name
);
330 if (list_empty(&bond
->vlan_list
) &&
331 (bond
->slave_cnt
== 0)) {
332 /* Last VLAN removed and no slaves, so
333 * restore block on adding VLANs. This will
334 * be removed once new slaves that are not
335 * VLAN challenged will be added.
337 bond
->dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
345 pr_debug("couldn't find VLAN ID %d in bond %s\n",
346 vlan_id
, bond
->dev
->name
);
349 write_unlock_bh(&bond
->lock
);
350 unblock_netpoll_tx();
355 * bond_has_challenged_slaves
356 * @bond: the bond we're working on
358 * Searches the slave list. Returns 1 if a vlan challenged slave
359 * was found, 0 otherwise.
361 * Assumes bond->lock is held.
363 static int bond_has_challenged_slaves(struct bonding
*bond
)
368 bond_for_each_slave(bond
, slave
, i
) {
369 if (slave
->dev
->features
& NETIF_F_VLAN_CHALLENGED
) {
370 pr_debug("found VLAN challenged slave - %s\n",
376 pr_debug("no VLAN challenged slaves found\n");
381 * bond_next_vlan - safely skip to the next item in the vlans list.
382 * @bond: the bond we're working on
383 * @curr: item we're advancing from
385 * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL,
386 * or @curr->next otherwise (even if it is @curr itself again).
388 * Caller must hold bond->lock
390 struct vlan_entry
*bond_next_vlan(struct bonding
*bond
, struct vlan_entry
*curr
)
392 struct vlan_entry
*next
, *last
;
394 if (list_empty(&bond
->vlan_list
))
398 next
= list_entry(bond
->vlan_list
.next
,
399 struct vlan_entry
, vlan_list
);
401 last
= list_entry(bond
->vlan_list
.prev
,
402 struct vlan_entry
, vlan_list
);
404 next
= list_entry(bond
->vlan_list
.next
,
405 struct vlan_entry
, vlan_list
);
407 next
= list_entry(curr
->vlan_list
.next
,
408 struct vlan_entry
, vlan_list
);
416 * bond_dev_queue_xmit - Prepare skb for xmit.
418 * @bond: bond device that got this skb for tx.
419 * @skb: hw accel VLAN tagged skb to transmit
420 * @slave_dev: slave that is supposed to xmit this skbuff
422 int bond_dev_queue_xmit(struct bonding
*bond
, struct sk_buff
*skb
,
423 struct net_device
*slave_dev
)
425 skb
->dev
= slave_dev
;
427 #ifdef CONFIG_NET_POLL_CONTROLLER
428 if (unlikely(bond
->dev
->priv_flags
& IFF_IN_NETPOLL
)) {
429 struct netpoll
*np
= bond
->dev
->npinfo
->netpoll
;
430 slave_dev
->npinfo
= bond
->dev
->npinfo
;
431 slave_dev
->priv_flags
|= IFF_IN_NETPOLL
;
432 netpoll_send_skb_on_dev(np
, skb
, slave_dev
);
433 slave_dev
->priv_flags
&= ~IFF_IN_NETPOLL
;
442 * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid
443 * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a
445 * a. This operation is performed in IOCTL context,
446 * b. The operation is protected by the RTNL semaphore in the 8021q code,
447 * c. Holding a lock with BH disabled while directly calling a base driver
448 * entry point is generally a BAD idea.
450 * The design of synchronization/protection for this operation in the 8021q
451 * module is good for one or more VLAN devices over a single physical device
452 * and cannot be extended for a teaming solution like bonding, so there is a
453 * potential race condition here where a net device from the vlan group might
454 * be referenced (either by a base driver or the 8021q code) while it is being
455 * removed from the system. However, it turns out we're not making matters
456 * worse, and if it works for regular VLAN usage it will work here too.
460 * bond_vlan_rx_register - Propagates registration to slaves
461 * @bond_dev: bonding net device that got called
462 * @grp: vlan group being registered
464 static void bond_vlan_rx_register(struct net_device
*bond_dev
,
465 struct vlan_group
*grp
)
467 struct bonding
*bond
= netdev_priv(bond_dev
);
471 write_lock_bh(&bond
->lock
);
473 write_unlock_bh(&bond
->lock
);
475 bond_for_each_slave(bond
, slave
, i
) {
476 struct net_device
*slave_dev
= slave
->dev
;
477 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
479 if ((slave_dev
->features
& NETIF_F_HW_VLAN_RX
) &&
480 slave_ops
->ndo_vlan_rx_register
) {
481 slave_ops
->ndo_vlan_rx_register(slave_dev
, grp
);
487 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
488 * @bond_dev: bonding net device that got called
489 * @vid: vlan id being added
491 static void bond_vlan_rx_add_vid(struct net_device
*bond_dev
, uint16_t vid
)
493 struct bonding
*bond
= netdev_priv(bond_dev
);
497 bond_for_each_slave(bond
, slave
, i
) {
498 struct net_device
*slave_dev
= slave
->dev
;
499 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
501 if ((slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) &&
502 slave_ops
->ndo_vlan_rx_add_vid
) {
503 slave_ops
->ndo_vlan_rx_add_vid(slave_dev
, vid
);
507 res
= bond_add_vlan(bond
, vid
);
509 pr_err("%s: Error: Failed to add vlan id %d\n",
510 bond_dev
->name
, vid
);
515 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
516 * @bond_dev: bonding net device that got called
517 * @vid: vlan id being removed
519 static void bond_vlan_rx_kill_vid(struct net_device
*bond_dev
, uint16_t vid
)
521 struct bonding
*bond
= netdev_priv(bond_dev
);
523 struct net_device
*vlan_dev
;
526 bond_for_each_slave(bond
, slave
, i
) {
527 struct net_device
*slave_dev
= slave
->dev
;
528 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
530 if ((slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) &&
531 slave_ops
->ndo_vlan_rx_kill_vid
) {
532 /* Save and then restore vlan_dev in the grp array,
533 * since the slave's driver might clear it.
535 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vid
);
536 slave_ops
->ndo_vlan_rx_kill_vid(slave_dev
, vid
);
537 vlan_group_set_device(bond
->vlgrp
, vid
, vlan_dev
);
541 res
= bond_del_vlan(bond
, vid
);
543 pr_err("%s: Error: Failed to remove vlan id %d\n",
544 bond_dev
->name
, vid
);
548 static void bond_add_vlans_on_slave(struct bonding
*bond
, struct net_device
*slave_dev
)
550 struct vlan_entry
*vlan
;
551 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
556 if ((slave_dev
->features
& NETIF_F_HW_VLAN_RX
) &&
557 slave_ops
->ndo_vlan_rx_register
)
558 slave_ops
->ndo_vlan_rx_register(slave_dev
, bond
->vlgrp
);
560 if (!(slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) ||
561 !(slave_ops
->ndo_vlan_rx_add_vid
))
564 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
)
565 slave_ops
->ndo_vlan_rx_add_vid(slave_dev
, vlan
->vlan_id
);
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
;
578 if (!(slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) ||
579 !(slave_ops
->ndo_vlan_rx_kill_vid
))
582 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 /*------------------------------- Link status -------------------------------*/
602 * Set the carrier state for the master according to the state of its
603 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
604 * do special 802.3ad magic.
606 * Returns zero if carrier state does not change, nonzero if it does.
608 static int bond_set_carrier(struct bonding
*bond
)
613 if (bond
->slave_cnt
== 0)
616 if (bond
->params
.mode
== BOND_MODE_8023AD
)
617 return bond_3ad_set_carrier(bond
);
619 bond_for_each_slave(bond
, slave
, i
) {
620 if (slave
->link
== BOND_LINK_UP
) {
621 if (!netif_carrier_ok(bond
->dev
)) {
622 netif_carrier_on(bond
->dev
);
630 if (netif_carrier_ok(bond
->dev
)) {
631 netif_carrier_off(bond
->dev
);
638 * Get link speed and duplex from the slave's base driver
639 * using ethtool. If for some reason the call fails or the
640 * values are invalid, fake speed and duplex to 100/Full
643 static int bond_update_speed_duplex(struct slave
*slave
)
645 struct net_device
*slave_dev
= slave
->dev
;
646 struct ethtool_cmd etool
;
649 /* Fake speed and duplex */
650 slave
->speed
= SPEED_100
;
651 slave
->duplex
= DUPLEX_FULL
;
653 if (!slave_dev
->ethtool_ops
|| !slave_dev
->ethtool_ops
->get_settings
)
656 res
= slave_dev
->ethtool_ops
->get_settings(slave_dev
, &etool
);
660 switch (etool
.speed
) {
670 switch (etool
.duplex
) {
678 slave
->speed
= etool
.speed
;
679 slave
->duplex
= etool
.duplex
;
685 * if <dev> supports MII link status reporting, check its link status.
687 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
688 * depending upon the setting of the use_carrier parameter.
690 * Return either BMSR_LSTATUS, meaning that the link is up (or we
691 * can't tell and just pretend it is), or 0, meaning that the link is
694 * If reporting is non-zero, instead of faking link up, return -1 if
695 * both ETHTOOL and MII ioctls fail (meaning the device does not
696 * support them). If use_carrier is set, return whatever it says.
697 * It'd be nice if there was a good way to tell if a driver supports
698 * netif_carrier, but there really isn't.
700 static int bond_check_dev_link(struct bonding
*bond
,
701 struct net_device
*slave_dev
, int reporting
)
703 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
704 int (*ioctl
)(struct net_device
*, struct ifreq
*, int);
706 struct mii_ioctl_data
*mii
;
708 if (!reporting
&& !netif_running(slave_dev
))
711 if (bond
->params
.use_carrier
)
712 return netif_carrier_ok(slave_dev
) ? BMSR_LSTATUS
: 0;
714 /* Try to get link status using Ethtool first. */
715 if (slave_dev
->ethtool_ops
) {
716 if (slave_dev
->ethtool_ops
->get_link
) {
719 link
= slave_dev
->ethtool_ops
->get_link(slave_dev
);
721 return link
? BMSR_LSTATUS
: 0;
725 /* Ethtool can't be used, fallback to MII ioctls. */
726 ioctl
= slave_ops
->ndo_do_ioctl
;
728 /* TODO: set pointer to correct ioctl on a per team member */
729 /* bases to make this more efficient. that is, once */
730 /* we determine the correct ioctl, we will always */
731 /* call it and not the others for that team */
735 * We cannot assume that SIOCGMIIPHY will also read a
736 * register; not all network drivers (e.g., e100)
740 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
741 strncpy(ifr
.ifr_name
, slave_dev
->name
, IFNAMSIZ
);
743 if (IOCTL(slave_dev
, &ifr
, SIOCGMIIPHY
) == 0) {
744 mii
->reg_num
= MII_BMSR
;
745 if (IOCTL(slave_dev
, &ifr
, SIOCGMIIREG
) == 0)
746 return mii
->val_out
& BMSR_LSTATUS
;
751 * If reporting, report that either there's no dev->do_ioctl,
752 * or both SIOCGMIIREG and get_link failed (meaning that we
753 * cannot report link status). If not reporting, pretend
756 return reporting
? -1 : BMSR_LSTATUS
;
759 /*----------------------------- Multicast list ------------------------------*/
762 * Push the promiscuity flag down to appropriate slaves
764 static int bond_set_promiscuity(struct bonding
*bond
, int inc
)
767 if (USES_PRIMARY(bond
->params
.mode
)) {
768 /* write lock already acquired */
769 if (bond
->curr_active_slave
) {
770 err
= dev_set_promiscuity(bond
->curr_active_slave
->dev
,
776 bond_for_each_slave(bond
, slave
, i
) {
777 err
= dev_set_promiscuity(slave
->dev
, inc
);
786 * Push the allmulti flag down to all slaves
788 static int bond_set_allmulti(struct bonding
*bond
, int inc
)
791 if (USES_PRIMARY(bond
->params
.mode
)) {
792 /* write lock already acquired */
793 if (bond
->curr_active_slave
) {
794 err
= dev_set_allmulti(bond
->curr_active_slave
->dev
,
800 bond_for_each_slave(bond
, slave
, i
) {
801 err
= dev_set_allmulti(slave
->dev
, inc
);
810 * Add a Multicast address to slaves
813 static void bond_mc_add(struct bonding
*bond
, void *addr
)
815 if (USES_PRIMARY(bond
->params
.mode
)) {
816 /* write lock already acquired */
817 if (bond
->curr_active_slave
)
818 dev_mc_add(bond
->curr_active_slave
->dev
, addr
);
823 bond_for_each_slave(bond
, slave
, i
)
824 dev_mc_add(slave
->dev
, addr
);
829 * Remove a multicast address from slave
832 static void bond_mc_del(struct bonding
*bond
, void *addr
)
834 if (USES_PRIMARY(bond
->params
.mode
)) {
835 /* write lock already acquired */
836 if (bond
->curr_active_slave
)
837 dev_mc_del(bond
->curr_active_slave
->dev
, addr
);
841 bond_for_each_slave(bond
, slave
, i
) {
842 dev_mc_del(slave
->dev
, addr
);
848 static void __bond_resend_igmp_join_requests(struct net_device
*dev
)
850 struct in_device
*in_dev
;
853 in_dev
= __in_dev_get_rcu(dev
);
855 ip_mc_rejoin_groups(in_dev
);
860 * Retrieve the list of registered multicast addresses for the bonding
861 * device and retransmit an IGMP JOIN request to the current active
864 static void bond_resend_igmp_join_requests(struct bonding
*bond
)
866 struct net_device
*vlan_dev
;
867 struct vlan_entry
*vlan
;
869 read_lock(&bond
->lock
);
871 /* rejoin all groups on bond device */
872 __bond_resend_igmp_join_requests(bond
->dev
);
874 /* rejoin all groups on vlan devices */
876 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
877 vlan_dev
= vlan_group_get_device(bond
->vlgrp
,
880 __bond_resend_igmp_join_requests(vlan_dev
);
884 if (--bond
->igmp_retrans
> 0)
885 queue_delayed_work(bond
->wq
, &bond
->mcast_work
, HZ
/5);
887 read_unlock(&bond
->lock
);
890 static void bond_resend_igmp_join_requests_delayed(struct work_struct
*work
)
892 struct bonding
*bond
= container_of(work
, struct bonding
,
894 bond_resend_igmp_join_requests(bond
);
898 * flush all members of flush->mc_list from device dev->mc_list
900 static void bond_mc_list_flush(struct net_device
*bond_dev
,
901 struct net_device
*slave_dev
)
903 struct bonding
*bond
= netdev_priv(bond_dev
);
904 struct netdev_hw_addr
*ha
;
906 netdev_for_each_mc_addr(ha
, bond_dev
)
907 dev_mc_del(slave_dev
, ha
->addr
);
909 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
910 /* del lacpdu mc addr from mc list */
911 u8 lacpdu_multicast
[ETH_ALEN
] = MULTICAST_LACPDU_ADDR
;
913 dev_mc_del(slave_dev
, lacpdu_multicast
);
917 /*--------------------------- Active slave change ---------------------------*/
920 * Update the mc list and multicast-related flags for the new and
921 * old active slaves (if any) according to the multicast mode, and
922 * promiscuous flags unconditionally.
924 static void bond_mc_swap(struct bonding
*bond
, struct slave
*new_active
,
925 struct slave
*old_active
)
927 struct netdev_hw_addr
*ha
;
929 if (!USES_PRIMARY(bond
->params
.mode
))
930 /* nothing to do - mc list is already up-to-date on
936 if (bond
->dev
->flags
& IFF_PROMISC
)
937 dev_set_promiscuity(old_active
->dev
, -1);
939 if (bond
->dev
->flags
& IFF_ALLMULTI
)
940 dev_set_allmulti(old_active
->dev
, -1);
942 netdev_for_each_mc_addr(ha
, bond
->dev
)
943 dev_mc_del(old_active
->dev
, ha
->addr
);
947 /* FIXME: Signal errors upstream. */
948 if (bond
->dev
->flags
& IFF_PROMISC
)
949 dev_set_promiscuity(new_active
->dev
, 1);
951 if (bond
->dev
->flags
& IFF_ALLMULTI
)
952 dev_set_allmulti(new_active
->dev
, 1);
954 netdev_for_each_mc_addr(ha
, bond
->dev
)
955 dev_mc_add(new_active
->dev
, ha
->addr
);
960 * bond_do_fail_over_mac
962 * Perform special MAC address swapping for fail_over_mac settings
964 * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh.
966 static void bond_do_fail_over_mac(struct bonding
*bond
,
967 struct slave
*new_active
,
968 struct slave
*old_active
)
969 __releases(&bond
->curr_slave_lock
)
970 __releases(&bond
->lock
)
971 __acquires(&bond
->lock
)
972 __acquires(&bond
->curr_slave_lock
)
974 u8 tmp_mac
[ETH_ALEN
];
975 struct sockaddr saddr
;
978 switch (bond
->params
.fail_over_mac
) {
979 case BOND_FOM_ACTIVE
:
981 memcpy(bond
->dev
->dev_addr
, new_active
->dev
->dev_addr
,
982 new_active
->dev
->addr_len
);
984 case BOND_FOM_FOLLOW
:
986 * if new_active && old_active, swap them
987 * if just old_active, do nothing (going to no active slave)
988 * if just new_active, set new_active to bond's MAC
993 write_unlock_bh(&bond
->curr_slave_lock
);
994 read_unlock(&bond
->lock
);
997 memcpy(tmp_mac
, new_active
->dev
->dev_addr
, ETH_ALEN
);
998 memcpy(saddr
.sa_data
, old_active
->dev
->dev_addr
,
1000 saddr
.sa_family
= new_active
->dev
->type
;
1002 memcpy(saddr
.sa_data
, bond
->dev
->dev_addr
, ETH_ALEN
);
1003 saddr
.sa_family
= bond
->dev
->type
;
1006 rv
= dev_set_mac_address(new_active
->dev
, &saddr
);
1008 pr_err("%s: Error %d setting MAC of slave %s\n",
1009 bond
->dev
->name
, -rv
, new_active
->dev
->name
);
1016 memcpy(saddr
.sa_data
, tmp_mac
, ETH_ALEN
);
1017 saddr
.sa_family
= old_active
->dev
->type
;
1019 rv
= dev_set_mac_address(old_active
->dev
, &saddr
);
1021 pr_err("%s: Error %d setting MAC of slave %s\n",
1022 bond
->dev
->name
, -rv
, new_active
->dev
->name
);
1024 read_lock(&bond
->lock
);
1025 write_lock_bh(&bond
->curr_slave_lock
);
1028 pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n",
1029 bond
->dev
->name
, bond
->params
.fail_over_mac
);
1035 static bool bond_should_change_active(struct bonding
*bond
)
1037 struct slave
*prim
= bond
->primary_slave
;
1038 struct slave
*curr
= bond
->curr_active_slave
;
1040 if (!prim
|| !curr
|| curr
->link
!= BOND_LINK_UP
)
1042 if (bond
->force_primary
) {
1043 bond
->force_primary
= false;
1046 if (bond
->params
.primary_reselect
== BOND_PRI_RESELECT_BETTER
&&
1047 (prim
->speed
< curr
->speed
||
1048 (prim
->speed
== curr
->speed
&& prim
->duplex
<= curr
->duplex
)))
1050 if (bond
->params
.primary_reselect
== BOND_PRI_RESELECT_FAILURE
)
1056 * find_best_interface - select the best available slave to be the active one
1057 * @bond: our bonding struct
1059 * Warning: Caller must hold curr_slave_lock for writing.
1061 static struct slave
*bond_find_best_slave(struct bonding
*bond
)
1063 struct slave
*new_active
, *old_active
;
1064 struct slave
*bestslave
= NULL
;
1065 int mintime
= bond
->params
.updelay
;
1068 new_active
= bond
->curr_active_slave
;
1070 if (!new_active
) { /* there were no active slaves left */
1071 if (bond
->slave_cnt
> 0) /* found one slave */
1072 new_active
= bond
->first_slave
;
1074 return NULL
; /* still no slave, return NULL */
1077 if ((bond
->primary_slave
) &&
1078 bond
->primary_slave
->link
== BOND_LINK_UP
&&
1079 bond_should_change_active(bond
)) {
1080 new_active
= bond
->primary_slave
;
1083 /* remember where to stop iterating over the slaves */
1084 old_active
= new_active
;
1086 bond_for_each_slave_from(bond
, new_active
, i
, old_active
) {
1087 if (new_active
->link
== BOND_LINK_UP
) {
1089 } else if (new_active
->link
== BOND_LINK_BACK
&&
1090 IS_UP(new_active
->dev
)) {
1091 /* link up, but waiting for stabilization */
1092 if (new_active
->delay
< mintime
) {
1093 mintime
= new_active
->delay
;
1094 bestslave
= new_active
;
1103 * change_active_interface - change the active slave into the specified one
1104 * @bond: our bonding struct
1105 * @new: the new slave to make the active one
1107 * Set the new slave to the bond's settings and unset them on the old
1108 * curr_active_slave.
1109 * Setting include flags, mc-list, promiscuity, allmulti, etc.
1111 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
1112 * because it is apparently the best available slave we have, even though its
1113 * updelay hasn't timed out yet.
1115 * If new_active is not NULL, caller must hold bond->lock for read and
1116 * curr_slave_lock for write_bh.
1118 void bond_change_active_slave(struct bonding
*bond
, struct slave
*new_active
)
1120 struct slave
*old_active
= bond
->curr_active_slave
;
1122 if (old_active
== new_active
)
1126 new_active
->jiffies
= jiffies
;
1128 if (new_active
->link
== BOND_LINK_BACK
) {
1129 if (USES_PRIMARY(bond
->params
.mode
)) {
1130 pr_info("%s: making interface %s the new active one %d ms earlier.\n",
1131 bond
->dev
->name
, new_active
->dev
->name
,
1132 (bond
->params
.updelay
- new_active
->delay
) * bond
->params
.miimon
);
1135 new_active
->delay
= 0;
1136 new_active
->link
= BOND_LINK_UP
;
1138 if (bond
->params
.mode
== BOND_MODE_8023AD
)
1139 bond_3ad_handle_link_change(new_active
, BOND_LINK_UP
);
1141 if (bond_is_lb(bond
))
1142 bond_alb_handle_link_change(bond
, new_active
, BOND_LINK_UP
);
1144 if (USES_PRIMARY(bond
->params
.mode
)) {
1145 pr_info("%s: making interface %s the new active one.\n",
1146 bond
->dev
->name
, new_active
->dev
->name
);
1151 if (USES_PRIMARY(bond
->params
.mode
))
1152 bond_mc_swap(bond
, new_active
, old_active
);
1154 if (bond_is_lb(bond
)) {
1155 bond_alb_handle_active_change(bond
, new_active
);
1157 bond_set_slave_inactive_flags(old_active
);
1159 bond_set_slave_active_flags(new_active
);
1161 bond
->curr_active_slave
= new_active
;
1164 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
) {
1166 bond_set_slave_inactive_flags(old_active
);
1169 bond_set_slave_active_flags(new_active
);
1171 if (bond
->params
.fail_over_mac
)
1172 bond_do_fail_over_mac(bond
, new_active
,
1175 if (netif_running(bond
->dev
)) {
1176 bond
->send_grat_arp
= bond
->params
.num_grat_arp
;
1177 bond_send_gratuitous_arp(bond
);
1179 bond
->send_unsol_na
= bond
->params
.num_unsol_na
;
1180 bond_send_unsolicited_na(bond
);
1183 write_unlock_bh(&bond
->curr_slave_lock
);
1184 read_unlock(&bond
->lock
);
1186 netdev_bonding_change(bond
->dev
, NETDEV_BONDING_FAILOVER
);
1188 read_lock(&bond
->lock
);
1189 write_lock_bh(&bond
->curr_slave_lock
);
1193 /* resend IGMP joins since active slave has changed or
1194 * all were sent on curr_active_slave */
1195 if (((USES_PRIMARY(bond
->params
.mode
) && new_active
) ||
1196 bond
->params
.mode
== BOND_MODE_ROUNDROBIN
) &&
1197 netif_running(bond
->dev
)) {
1198 bond
->igmp_retrans
= bond
->params
.resend_igmp
;
1199 queue_delayed_work(bond
->wq
, &bond
->mcast_work
, 0);
1204 * bond_select_active_slave - select a new active slave, if needed
1205 * @bond: our bonding struct
1207 * This functions should be called when one of the following occurs:
1208 * - The old curr_active_slave has been released or lost its link.
1209 * - The primary_slave has got its link back.
1210 * - A slave has got its link back and there's no old curr_active_slave.
1212 * Caller must hold bond->lock for read and curr_slave_lock for write_bh.
1214 void bond_select_active_slave(struct bonding
*bond
)
1216 struct slave
*best_slave
;
1219 best_slave
= bond_find_best_slave(bond
);
1220 if (best_slave
!= bond
->curr_active_slave
) {
1221 bond_change_active_slave(bond
, best_slave
);
1222 rv
= bond_set_carrier(bond
);
1226 if (netif_carrier_ok(bond
->dev
)) {
1227 pr_info("%s: first active interface up!\n",
1230 pr_info("%s: now running without any active interface !\n",
1236 /*--------------------------- slave list handling ---------------------------*/
1239 * This function attaches the slave to the end of list.
1241 * bond->lock held for writing by caller.
1243 static void bond_attach_slave(struct bonding
*bond
, struct slave
*new_slave
)
1245 if (bond
->first_slave
== NULL
) { /* attaching the first slave */
1246 new_slave
->next
= new_slave
;
1247 new_slave
->prev
= new_slave
;
1248 bond
->first_slave
= new_slave
;
1250 new_slave
->next
= bond
->first_slave
;
1251 new_slave
->prev
= bond
->first_slave
->prev
;
1252 new_slave
->next
->prev
= new_slave
;
1253 new_slave
->prev
->next
= new_slave
;
1260 * This function detaches the slave from the list.
1261 * WARNING: no check is made to verify if the slave effectively
1262 * belongs to <bond>.
1263 * Nothing is freed on return, structures are just unchained.
1264 * If any slave pointer in bond was pointing to <slave>,
1265 * it should be changed by the calling function.
1267 * bond->lock held for writing by caller.
1269 static void bond_detach_slave(struct bonding
*bond
, struct slave
*slave
)
1272 slave
->next
->prev
= slave
->prev
;
1275 slave
->prev
->next
= slave
->next
;
1277 if (bond
->first_slave
== slave
) { /* slave is the first slave */
1278 if (bond
->slave_cnt
> 1) { /* there are more slave */
1279 bond
->first_slave
= slave
->next
;
1281 bond
->first_slave
= NULL
; /* slave was the last one */
1290 #ifdef CONFIG_NET_POLL_CONTROLLER
1292 * You must hold read lock on bond->lock before calling this.
1294 static bool slaves_support_netpoll(struct net_device
*bond_dev
)
1296 struct bonding
*bond
= netdev_priv(bond_dev
);
1297 struct slave
*slave
;
1301 bond_for_each_slave(bond
, slave
, i
) {
1302 if ((slave
->dev
->priv_flags
& IFF_DISABLE_NETPOLL
) ||
1303 !slave
->dev
->netdev_ops
->ndo_poll_controller
)
1306 return i
!= 0 && ret
;
1309 static void bond_poll_controller(struct net_device
*bond_dev
)
1311 struct bonding
*bond
= netdev_priv(bond_dev
);
1312 struct slave
*slave
;
1315 bond_for_each_slave(bond
, slave
, i
) {
1316 if (slave
->dev
&& IS_UP(slave
->dev
))
1317 netpoll_poll_dev(slave
->dev
);
1321 static void bond_netpoll_cleanup(struct net_device
*bond_dev
)
1323 struct bonding
*bond
= netdev_priv(bond_dev
);
1324 struct slave
*slave
;
1325 const struct net_device_ops
*ops
;
1328 read_lock(&bond
->lock
);
1329 bond_dev
->npinfo
= NULL
;
1330 bond_for_each_slave(bond
, slave
, i
) {
1332 ops
= slave
->dev
->netdev_ops
;
1333 if (ops
->ndo_netpoll_cleanup
)
1334 ops
->ndo_netpoll_cleanup(slave
->dev
);
1336 slave
->dev
->npinfo
= NULL
;
1339 read_unlock(&bond
->lock
);
1344 static void bond_netpoll_cleanup(struct net_device
*bond_dev
)
1350 /*---------------------------------- IOCTL ----------------------------------*/
1352 static int bond_sethwaddr(struct net_device
*bond_dev
,
1353 struct net_device
*slave_dev
)
1355 pr_debug("bond_dev=%p\n", bond_dev
);
1356 pr_debug("slave_dev=%p\n", slave_dev
);
1357 pr_debug("slave_dev->addr_len=%d\n", slave_dev
->addr_len
);
1358 memcpy(bond_dev
->dev_addr
, slave_dev
->dev_addr
, slave_dev
->addr_len
);
1362 #define BOND_VLAN_FEATURES \
1363 (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \
1364 NETIF_F_HW_VLAN_FILTER)
1367 * Compute the common dev->feature set available to all slaves. Some
1368 * feature bits are managed elsewhere, so preserve those feature bits
1369 * on the master device.
1371 static int bond_compute_features(struct bonding
*bond
)
1373 struct slave
*slave
;
1374 struct net_device
*bond_dev
= bond
->dev
;
1375 unsigned long features
= bond_dev
->features
;
1376 unsigned long vlan_features
= 0;
1377 unsigned short max_hard_header_len
= max((u16
)ETH_HLEN
,
1378 bond_dev
->hard_header_len
);
1381 features
&= ~(NETIF_F_ALL_CSUM
| BOND_VLAN_FEATURES
);
1382 features
|= NETIF_F_GSO_MASK
| NETIF_F_NO_CSUM
;
1384 if (!bond
->first_slave
)
1387 features
&= ~NETIF_F_ONE_FOR_ALL
;
1389 vlan_features
= bond
->first_slave
->dev
->vlan_features
;
1390 bond_for_each_slave(bond
, slave
, i
) {
1391 features
= netdev_increment_features(features
,
1392 slave
->dev
->features
,
1393 NETIF_F_ONE_FOR_ALL
);
1394 vlan_features
= netdev_increment_features(vlan_features
,
1395 slave
->dev
->vlan_features
,
1396 NETIF_F_ONE_FOR_ALL
);
1397 if (slave
->dev
->hard_header_len
> max_hard_header_len
)
1398 max_hard_header_len
= slave
->dev
->hard_header_len
;
1402 features
|= (bond_dev
->features
& BOND_VLAN_FEATURES
);
1403 bond_dev
->features
= netdev_fix_features(features
, NULL
);
1404 bond_dev
->vlan_features
= netdev_fix_features(vlan_features
, NULL
);
1405 bond_dev
->hard_header_len
= max_hard_header_len
;
1410 static void bond_setup_by_slave(struct net_device
*bond_dev
,
1411 struct net_device
*slave_dev
)
1413 struct bonding
*bond
= netdev_priv(bond_dev
);
1415 bond_dev
->header_ops
= slave_dev
->header_ops
;
1417 bond_dev
->type
= slave_dev
->type
;
1418 bond_dev
->hard_header_len
= slave_dev
->hard_header_len
;
1419 bond_dev
->addr_len
= slave_dev
->addr_len
;
1421 memcpy(bond_dev
->broadcast
, slave_dev
->broadcast
,
1422 slave_dev
->addr_len
);
1423 bond
->setup_by_slave
= 1;
1426 /* enslave device <slave> to bond device <master> */
1427 int bond_enslave(struct net_device
*bond_dev
, struct net_device
*slave_dev
)
1429 struct bonding
*bond
= netdev_priv(bond_dev
);
1430 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
1431 struct slave
*new_slave
= NULL
;
1432 struct netdev_hw_addr
*ha
;
1433 struct sockaddr addr
;
1435 int old_features
= bond_dev
->features
;
1438 if (!bond
->params
.use_carrier
&& slave_dev
->ethtool_ops
== NULL
&&
1439 slave_ops
->ndo_do_ioctl
== NULL
) {
1440 pr_warning("%s: Warning: no link monitoring support for %s\n",
1441 bond_dev
->name
, slave_dev
->name
);
1444 /* bond must be initialized by bond_open() before enslaving */
1445 if (!(bond_dev
->flags
& IFF_UP
)) {
1446 pr_warning("%s: master_dev is not up in bond_enslave\n",
1450 /* already enslaved */
1451 if (slave_dev
->flags
& IFF_SLAVE
) {
1452 pr_debug("Error, Device was already enslaved\n");
1456 /* vlan challenged mutual exclusion */
1457 /* no need to lock since we're protected by rtnl_lock */
1458 if (slave_dev
->features
& NETIF_F_VLAN_CHALLENGED
) {
1459 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev
->name
);
1461 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1462 bond_dev
->name
, slave_dev
->name
, bond_dev
->name
);
1465 pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1466 bond_dev
->name
, slave_dev
->name
,
1467 slave_dev
->name
, bond_dev
->name
);
1468 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
1471 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev
->name
);
1472 if (bond
->slave_cnt
== 0) {
1473 /* First slave, and it is not VLAN challenged,
1474 * so remove the block of adding VLANs over the bond.
1476 bond_dev
->features
&= ~NETIF_F_VLAN_CHALLENGED
;
1481 * Old ifenslave binaries are no longer supported. These can
1482 * be identified with moderate accuracy by the state of the slave:
1483 * the current ifenslave will set the interface down prior to
1484 * enslaving it; the old ifenslave will not.
1486 if ((slave_dev
->flags
& IFF_UP
)) {
1487 pr_err("%s is up. This may be due to an out of date ifenslave.\n",
1490 goto err_undo_flags
;
1493 /* set bonding device ether type by slave - bonding netdevices are
1494 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1495 * there is a need to override some of the type dependent attribs/funcs.
1497 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1498 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1500 if (bond
->slave_cnt
== 0) {
1501 if (bond_dev
->type
!= slave_dev
->type
) {
1502 pr_debug("%s: change device type from %d to %d\n",
1504 bond_dev
->type
, slave_dev
->type
);
1506 res
= netdev_bonding_change(bond_dev
,
1507 NETDEV_PRE_TYPE_CHANGE
);
1508 res
= notifier_to_errno(res
);
1510 pr_err("%s: refused to change device type\n",
1513 goto err_undo_flags
;
1516 /* Flush unicast and multicast addresses */
1517 dev_uc_flush(bond_dev
);
1518 dev_mc_flush(bond_dev
);
1520 if (slave_dev
->type
!= ARPHRD_ETHER
)
1521 bond_setup_by_slave(bond_dev
, slave_dev
);
1523 ether_setup(bond_dev
);
1525 netdev_bonding_change(bond_dev
,
1526 NETDEV_POST_TYPE_CHANGE
);
1528 } else if (bond_dev
->type
!= slave_dev
->type
) {
1529 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n",
1531 slave_dev
->type
, bond_dev
->type
);
1533 goto err_undo_flags
;
1536 if (slave_ops
->ndo_set_mac_address
== NULL
) {
1537 if (bond
->slave_cnt
== 0) {
1538 pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.",
1540 bond
->params
.fail_over_mac
= BOND_FOM_ACTIVE
;
1541 } else if (bond
->params
.fail_over_mac
!= BOND_FOM_ACTIVE
) {
1542 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",
1545 goto err_undo_flags
;
1549 /* If this is the first slave, then we need to set the master's hardware
1550 * address to be the same as the slave's. */
1551 if (is_zero_ether_addr(bond
->dev
->dev_addr
))
1552 memcpy(bond
->dev
->dev_addr
, slave_dev
->dev_addr
,
1553 slave_dev
->addr_len
);
1556 new_slave
= kzalloc(sizeof(struct slave
), GFP_KERNEL
);
1559 goto err_undo_flags
;
1563 * Set the new_slave's queue_id to be zero. Queue ID mapping
1564 * is set via sysfs or module option if desired.
1566 new_slave
->queue_id
= 0;
1568 /* Save slave's original mtu and then set it to match the bond */
1569 new_slave
->original_mtu
= slave_dev
->mtu
;
1570 res
= dev_set_mtu(slave_dev
, bond
->dev
->mtu
);
1572 pr_debug("Error %d calling dev_set_mtu\n", res
);
1577 * Save slave's original ("permanent") mac address for modes
1578 * that need it, and for restoring it upon release, and then
1579 * set it to the master's address
1581 memcpy(new_slave
->perm_hwaddr
, slave_dev
->dev_addr
, ETH_ALEN
);
1583 if (!bond
->params
.fail_over_mac
) {
1585 * Set slave to master's mac address. The application already
1586 * set the master's mac address to that of the first slave
1588 memcpy(addr
.sa_data
, bond_dev
->dev_addr
, bond_dev
->addr_len
);
1589 addr
.sa_family
= slave_dev
->type
;
1590 res
= dev_set_mac_address(slave_dev
, &addr
);
1592 pr_debug("Error %d calling set_mac_address\n", res
);
1593 goto err_restore_mtu
;
1597 res
= netdev_set_master(slave_dev
, bond_dev
);
1599 pr_debug("Error %d calling netdev_set_master\n", res
);
1600 goto err_restore_mac
;
1602 /* open the slave since the application closed it */
1603 res
= dev_open(slave_dev
);
1605 pr_debug("Opening slave %s failed\n", slave_dev
->name
);
1606 goto err_unset_master
;
1609 new_slave
->dev
= slave_dev
;
1610 slave_dev
->priv_flags
|= IFF_BONDING
;
1612 if (bond_is_lb(bond
)) {
1613 /* bond_alb_init_slave() must be called before all other stages since
1614 * it might fail and we do not want to have to undo everything
1616 res
= bond_alb_init_slave(bond
, new_slave
);
1621 /* If the mode USES_PRIMARY, then the new slave gets the
1622 * master's promisc (and mc) settings only if it becomes the
1623 * curr_active_slave, and that is taken care of later when calling
1624 * bond_change_active()
1626 if (!USES_PRIMARY(bond
->params
.mode
)) {
1627 /* set promiscuity level to new slave */
1628 if (bond_dev
->flags
& IFF_PROMISC
) {
1629 res
= dev_set_promiscuity(slave_dev
, 1);
1634 /* set allmulti level to new slave */
1635 if (bond_dev
->flags
& IFF_ALLMULTI
) {
1636 res
= dev_set_allmulti(slave_dev
, 1);
1641 netif_addr_lock_bh(bond_dev
);
1642 /* upload master's mc_list to new slave */
1643 netdev_for_each_mc_addr(ha
, bond_dev
)
1644 dev_mc_add(slave_dev
, ha
->addr
);
1645 netif_addr_unlock_bh(bond_dev
);
1648 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
1649 /* add lacpdu mc addr to mc list */
1650 u8 lacpdu_multicast
[ETH_ALEN
] = MULTICAST_LACPDU_ADDR
;
1652 dev_mc_add(slave_dev
, lacpdu_multicast
);
1655 bond_add_vlans_on_slave(bond
, slave_dev
);
1657 write_lock_bh(&bond
->lock
);
1659 bond_attach_slave(bond
, new_slave
);
1661 new_slave
->delay
= 0;
1662 new_slave
->link_failure_count
= 0;
1664 bond_compute_features(bond
);
1666 write_unlock_bh(&bond
->lock
);
1668 read_lock(&bond
->lock
);
1670 new_slave
->last_arp_rx
= jiffies
;
1672 if (bond
->params
.miimon
&& !bond
->params
.use_carrier
) {
1673 link_reporting
= bond_check_dev_link(bond
, slave_dev
, 1);
1675 if ((link_reporting
== -1) && !bond
->params
.arp_interval
) {
1677 * miimon is set but a bonded network driver
1678 * does not support ETHTOOL/MII and
1679 * arp_interval is not set. Note: if
1680 * use_carrier is enabled, we will never go
1681 * here (because netif_carrier is always
1682 * supported); thus, we don't need to change
1683 * the messages for netif_carrier.
1685 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",
1686 bond_dev
->name
, slave_dev
->name
);
1687 } else if (link_reporting
== -1) {
1688 /* unable get link status using mii/ethtool */
1689 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",
1690 bond_dev
->name
, slave_dev
->name
);
1694 /* check for initial state */
1695 if (!bond
->params
.miimon
||
1696 (bond_check_dev_link(bond
, slave_dev
, 0) == BMSR_LSTATUS
)) {
1697 if (bond
->params
.updelay
) {
1698 pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n");
1699 new_slave
->link
= BOND_LINK_BACK
;
1700 new_slave
->delay
= bond
->params
.updelay
;
1702 pr_debug("Initial state of slave_dev is BOND_LINK_UP\n");
1703 new_slave
->link
= BOND_LINK_UP
;
1705 new_slave
->jiffies
= jiffies
;
1707 pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n");
1708 new_slave
->link
= BOND_LINK_DOWN
;
1711 if (bond_update_speed_duplex(new_slave
) &&
1712 (new_slave
->link
!= BOND_LINK_DOWN
)) {
1713 pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n",
1714 bond_dev
->name
, new_slave
->dev
->name
);
1716 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
1717 pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n",
1722 if (USES_PRIMARY(bond
->params
.mode
) && bond
->params
.primary
[0]) {
1723 /* if there is a primary slave, remember it */
1724 if (strcmp(bond
->params
.primary
, new_slave
->dev
->name
) == 0) {
1725 bond
->primary_slave
= new_slave
;
1726 bond
->force_primary
= true;
1730 write_lock_bh(&bond
->curr_slave_lock
);
1732 switch (bond
->params
.mode
) {
1733 case BOND_MODE_ACTIVEBACKUP
:
1734 bond_set_slave_inactive_flags(new_slave
);
1735 bond_select_active_slave(bond
);
1737 case BOND_MODE_8023AD
:
1738 /* in 802.3ad mode, the internal mechanism
1739 * will activate the slaves in the selected
1742 bond_set_slave_inactive_flags(new_slave
);
1743 /* if this is the first slave */
1744 if (bond
->slave_cnt
== 1) {
1745 SLAVE_AD_INFO(new_slave
).id
= 1;
1746 /* Initialize AD with the number of times that the AD timer is called in 1 second
1747 * can be called only after the mac address of the bond is set
1749 bond_3ad_initialize(bond
, 1000/AD_TIMER_INTERVAL
,
1750 bond
->params
.lacp_fast
);
1752 SLAVE_AD_INFO(new_slave
).id
=
1753 SLAVE_AD_INFO(new_slave
->prev
).id
+ 1;
1756 bond_3ad_bind_slave(new_slave
);
1760 new_slave
->state
= BOND_STATE_ACTIVE
;
1761 bond_set_slave_inactive_flags(new_slave
);
1762 bond_select_active_slave(bond
);
1765 pr_debug("This slave is always active in trunk mode\n");
1767 /* always active in trunk mode */
1768 new_slave
->state
= BOND_STATE_ACTIVE
;
1770 /* In trunking mode there is little meaning to curr_active_slave
1771 * anyway (it holds no special properties of the bond device),
1772 * so we can change it without calling change_active_interface()
1774 if (!bond
->curr_active_slave
)
1775 bond
->curr_active_slave
= new_slave
;
1778 } /* switch(bond_mode) */
1780 write_unlock_bh(&bond
->curr_slave_lock
);
1782 bond_set_carrier(bond
);
1784 #ifdef CONFIG_NET_POLL_CONTROLLER
1785 if (slaves_support_netpoll(bond_dev
)) {
1786 bond_dev
->priv_flags
&= ~IFF_DISABLE_NETPOLL
;
1787 if (bond_dev
->npinfo
)
1788 slave_dev
->npinfo
= bond_dev
->npinfo
;
1789 } else if (!(bond_dev
->priv_flags
& IFF_DISABLE_NETPOLL
)) {
1790 bond_dev
->priv_flags
|= IFF_DISABLE_NETPOLL
;
1791 pr_info("New slave device %s does not support netpoll\n",
1793 pr_info("Disabling netpoll support for %s\n", bond_dev
->name
);
1796 read_unlock(&bond
->lock
);
1798 res
= bond_create_slave_symlinks(bond_dev
, slave_dev
);
1802 pr_info("%s: enslaving %s as a%s interface with a%s link.\n",
1803 bond_dev
->name
, slave_dev
->name
,
1804 new_slave
->state
== BOND_STATE_ACTIVE
? "n active" : " backup",
1805 new_slave
->link
!= BOND_LINK_DOWN
? "n up" : " down");
1807 /* enslave is successful */
1810 /* Undo stages on error */
1812 dev_close(slave_dev
);
1815 netdev_set_master(slave_dev
, NULL
);
1818 if (!bond
->params
.fail_over_mac
) {
1819 /* XXX TODO - fom follow mode needs to change master's
1820 * MAC if this slave's MAC is in use by the bond, or at
1821 * least print a warning.
1823 memcpy(addr
.sa_data
, new_slave
->perm_hwaddr
, ETH_ALEN
);
1824 addr
.sa_family
= slave_dev
->type
;
1825 dev_set_mac_address(slave_dev
, &addr
);
1829 dev_set_mtu(slave_dev
, new_slave
->original_mtu
);
1835 bond_dev
->features
= old_features
;
1841 * Try to release the slave device <slave> from the bond device <master>
1842 * It is legal to access curr_active_slave without a lock because all the function
1845 * The rules for slave state should be:
1846 * for Active/Backup:
1847 * Active stays on all backups go down
1848 * for Bonded connections:
1849 * The first up interface should be left on and all others downed.
1851 int bond_release(struct net_device
*bond_dev
, struct net_device
*slave_dev
)
1853 struct bonding
*bond
= netdev_priv(bond_dev
);
1854 struct slave
*slave
, *oldcurrent
;
1855 struct sockaddr addr
;
1857 /* slave is not a slave or master is not master of this slave */
1858 if (!(slave_dev
->flags
& IFF_SLAVE
) ||
1859 (slave_dev
->master
!= bond_dev
)) {
1860 pr_err("%s: Error: cannot release %s.\n",
1861 bond_dev
->name
, slave_dev
->name
);
1866 netdev_bonding_change(bond_dev
, NETDEV_BONDING_DESLAVE
);
1867 write_lock_bh(&bond
->lock
);
1869 slave
= bond_get_slave_by_dev(bond
, slave_dev
);
1871 /* not a slave of this bond */
1872 pr_info("%s: %s not enslaved\n",
1873 bond_dev
->name
, slave_dev
->name
);
1874 write_unlock_bh(&bond
->lock
);
1875 unblock_netpoll_tx();
1879 if (!bond
->params
.fail_over_mac
) {
1880 if (!compare_ether_addr(bond_dev
->dev_addr
, slave
->perm_hwaddr
) &&
1881 bond
->slave_cnt
> 1)
1882 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",
1883 bond_dev
->name
, slave_dev
->name
,
1885 bond_dev
->name
, slave_dev
->name
);
1888 /* Inform AD package of unbinding of slave. */
1889 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
1890 /* must be called before the slave is
1891 * detached from the list
1893 bond_3ad_unbind_slave(slave
);
1896 pr_info("%s: releasing %s interface %s\n",
1898 (slave
->state
== BOND_STATE_ACTIVE
) ? "active" : "backup",
1901 oldcurrent
= bond
->curr_active_slave
;
1903 bond
->current_arp_slave
= NULL
;
1905 /* release the slave from its bond */
1906 bond_detach_slave(bond
, slave
);
1908 bond_compute_features(bond
);
1910 if (bond
->primary_slave
== slave
)
1911 bond
->primary_slave
= NULL
;
1913 if (oldcurrent
== slave
)
1914 bond_change_active_slave(bond
, NULL
);
1916 if (bond_is_lb(bond
)) {
1917 /* Must be called only after the slave has been
1918 * detached from the list and the curr_active_slave
1919 * has been cleared (if our_slave == old_current),
1920 * but before a new active slave is selected.
1922 write_unlock_bh(&bond
->lock
);
1923 bond_alb_deinit_slave(bond
, slave
);
1924 write_lock_bh(&bond
->lock
);
1927 if (oldcurrent
== slave
) {
1929 * Note that we hold RTNL over this sequence, so there
1930 * is no concern that another slave add/remove event
1933 write_unlock_bh(&bond
->lock
);
1934 read_lock(&bond
->lock
);
1935 write_lock_bh(&bond
->curr_slave_lock
);
1937 bond_select_active_slave(bond
);
1939 write_unlock_bh(&bond
->curr_slave_lock
);
1940 read_unlock(&bond
->lock
);
1941 write_lock_bh(&bond
->lock
);
1944 if (bond
->slave_cnt
== 0) {
1945 bond_set_carrier(bond
);
1947 /* if the last slave was removed, zero the mac address
1948 * of the master so it will be set by the application
1949 * to the mac address of the first slave
1951 memset(bond_dev
->dev_addr
, 0, bond_dev
->addr_len
);
1954 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
1956 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
1957 bond_dev
->name
, bond_dev
->name
);
1958 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
1961 } else if ((bond_dev
->features
& NETIF_F_VLAN_CHALLENGED
) &&
1962 !bond_has_challenged_slaves(bond
)) {
1963 pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n",
1964 bond_dev
->name
, slave_dev
->name
, bond_dev
->name
);
1965 bond_dev
->features
&= ~NETIF_F_VLAN_CHALLENGED
;
1968 write_unlock_bh(&bond
->lock
);
1969 unblock_netpoll_tx();
1971 /* must do this from outside any spinlocks */
1972 bond_destroy_slave_symlinks(bond_dev
, slave_dev
);
1974 bond_del_vlans_from_slave(bond
, slave_dev
);
1976 /* If the mode USES_PRIMARY, then we should only remove its
1977 * promisc and mc settings if it was the curr_active_slave, but that was
1978 * already taken care of above when we detached the slave
1980 if (!USES_PRIMARY(bond
->params
.mode
)) {
1981 /* unset promiscuity level from slave */
1982 if (bond_dev
->flags
& IFF_PROMISC
)
1983 dev_set_promiscuity(slave_dev
, -1);
1985 /* unset allmulti level from slave */
1986 if (bond_dev
->flags
& IFF_ALLMULTI
)
1987 dev_set_allmulti(slave_dev
, -1);
1989 /* flush master's mc_list from slave */
1990 netif_addr_lock_bh(bond_dev
);
1991 bond_mc_list_flush(bond_dev
, slave_dev
);
1992 netif_addr_unlock_bh(bond_dev
);
1995 netdev_set_master(slave_dev
, NULL
);
1997 #ifdef CONFIG_NET_POLL_CONTROLLER
1998 read_lock_bh(&bond
->lock
);
2000 if (slaves_support_netpoll(bond_dev
))
2001 bond_dev
->priv_flags
&= ~IFF_DISABLE_NETPOLL
;
2002 read_unlock_bh(&bond
->lock
);
2003 if (slave_dev
->netdev_ops
->ndo_netpoll_cleanup
)
2004 slave_dev
->netdev_ops
->ndo_netpoll_cleanup(slave_dev
);
2006 slave_dev
->npinfo
= NULL
;
2009 /* close slave before restoring its mac address */
2010 dev_close(slave_dev
);
2012 if (bond
->params
.fail_over_mac
!= BOND_FOM_ACTIVE
) {
2013 /* restore original ("permanent") mac address */
2014 memcpy(addr
.sa_data
, slave
->perm_hwaddr
, ETH_ALEN
);
2015 addr
.sa_family
= slave_dev
->type
;
2016 dev_set_mac_address(slave_dev
, &addr
);
2019 dev_set_mtu(slave_dev
, slave
->original_mtu
);
2021 slave_dev
->priv_flags
&= ~(IFF_MASTER_8023AD
| IFF_MASTER_ALB
|
2022 IFF_SLAVE_INACTIVE
| IFF_BONDING
|
2027 return 0; /* deletion OK */
2031 * First release a slave and than destroy the bond if no more slaves are left.
2032 * Must be under rtnl_lock when this function is called.
2034 static int bond_release_and_destroy(struct net_device
*bond_dev
,
2035 struct net_device
*slave_dev
)
2037 struct bonding
*bond
= netdev_priv(bond_dev
);
2040 ret
= bond_release(bond_dev
, slave_dev
);
2041 if ((ret
== 0) && (bond
->slave_cnt
== 0)) {
2042 pr_info("%s: destroying bond %s.\n",
2043 bond_dev
->name
, bond_dev
->name
);
2044 unregister_netdevice(bond_dev
);
2050 * This function releases all slaves.
2052 static int bond_release_all(struct net_device
*bond_dev
)
2054 struct bonding
*bond
= netdev_priv(bond_dev
);
2055 struct slave
*slave
;
2056 struct net_device
*slave_dev
;
2057 struct sockaddr addr
;
2059 write_lock_bh(&bond
->lock
);
2061 netif_carrier_off(bond_dev
);
2063 if (bond
->slave_cnt
== 0)
2066 bond
->current_arp_slave
= NULL
;
2067 bond
->primary_slave
= NULL
;
2068 bond_change_active_slave(bond
, NULL
);
2070 while ((slave
= bond
->first_slave
) != NULL
) {
2071 /* Inform AD package of unbinding of slave
2072 * before slave is detached from the list.
2074 if (bond
->params
.mode
== BOND_MODE_8023AD
)
2075 bond_3ad_unbind_slave(slave
);
2077 slave_dev
= slave
->dev
;
2078 bond_detach_slave(bond
, slave
);
2080 /* now that the slave is detached, unlock and perform
2081 * all the undo steps that should not be called from
2084 write_unlock_bh(&bond
->lock
);
2086 if (bond_is_lb(bond
)) {
2087 /* must be called only after the slave
2088 * has been detached from the list
2090 bond_alb_deinit_slave(bond
, slave
);
2093 bond_compute_features(bond
);
2095 bond_destroy_slave_symlinks(bond_dev
, slave_dev
);
2096 bond_del_vlans_from_slave(bond
, slave_dev
);
2098 /* If the mode USES_PRIMARY, then we should only remove its
2099 * promisc and mc settings if it was the curr_active_slave, but that was
2100 * already taken care of above when we detached the slave
2102 if (!USES_PRIMARY(bond
->params
.mode
)) {
2103 /* unset promiscuity level from slave */
2104 if (bond_dev
->flags
& IFF_PROMISC
)
2105 dev_set_promiscuity(slave_dev
, -1);
2107 /* unset allmulti level from slave */
2108 if (bond_dev
->flags
& IFF_ALLMULTI
)
2109 dev_set_allmulti(slave_dev
, -1);
2111 /* flush master's mc_list from slave */
2112 netif_addr_lock_bh(bond_dev
);
2113 bond_mc_list_flush(bond_dev
, slave_dev
);
2114 netif_addr_unlock_bh(bond_dev
);
2117 netdev_set_master(slave_dev
, NULL
);
2119 /* close slave before restoring its mac address */
2120 dev_close(slave_dev
);
2122 if (!bond
->params
.fail_over_mac
) {
2123 /* restore original ("permanent") mac address*/
2124 memcpy(addr
.sa_data
, slave
->perm_hwaddr
, ETH_ALEN
);
2125 addr
.sa_family
= slave_dev
->type
;
2126 dev_set_mac_address(slave_dev
, &addr
);
2129 slave_dev
->priv_flags
&= ~(IFF_MASTER_8023AD
| IFF_MASTER_ALB
|
2130 IFF_SLAVE_INACTIVE
);
2134 /* re-acquire the lock before getting the next slave */
2135 write_lock_bh(&bond
->lock
);
2138 /* zero the mac address of the master so it will be
2139 * set by the application to the mac address of the
2142 memset(bond_dev
->dev_addr
, 0, bond_dev
->addr_len
);
2145 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
2147 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
2148 bond_dev
->name
, bond_dev
->name
);
2149 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
2153 pr_info("%s: released all slaves\n", bond_dev
->name
);
2156 write_unlock_bh(&bond
->lock
);
2161 * This function changes the active slave to slave <slave_dev>.
2162 * It returns -EINVAL in the following cases.
2163 * - <slave_dev> is not found in the list.
2164 * - There is not active slave now.
2165 * - <slave_dev> is already active.
2166 * - The link state of <slave_dev> is not BOND_LINK_UP.
2167 * - <slave_dev> is not running.
2168 * In these cases, this function does nothing.
2169 * In the other cases, current_slave pointer is changed and 0 is returned.
2171 static int bond_ioctl_change_active(struct net_device
*bond_dev
, struct net_device
*slave_dev
)
2173 struct bonding
*bond
= netdev_priv(bond_dev
);
2174 struct slave
*old_active
= NULL
;
2175 struct slave
*new_active
= NULL
;
2178 if (!USES_PRIMARY(bond
->params
.mode
))
2181 /* Verify that master_dev is indeed the master of slave_dev */
2182 if (!(slave_dev
->flags
& IFF_SLAVE
) || (slave_dev
->master
!= bond_dev
))
2185 read_lock(&bond
->lock
);
2187 read_lock(&bond
->curr_slave_lock
);
2188 old_active
= bond
->curr_active_slave
;
2189 read_unlock(&bond
->curr_slave_lock
);
2191 new_active
= bond_get_slave_by_dev(bond
, slave_dev
);
2194 * Changing to the current active: do nothing; return success.
2196 if (new_active
&& (new_active
== old_active
)) {
2197 read_unlock(&bond
->lock
);
2203 (new_active
->link
== BOND_LINK_UP
) &&
2204 IS_UP(new_active
->dev
)) {
2206 write_lock_bh(&bond
->curr_slave_lock
);
2207 bond_change_active_slave(bond
, new_active
);
2208 write_unlock_bh(&bond
->curr_slave_lock
);
2209 unblock_netpoll_tx();
2213 read_unlock(&bond
->lock
);
2218 static int bond_info_query(struct net_device
*bond_dev
, struct ifbond
*info
)
2220 struct bonding
*bond
= netdev_priv(bond_dev
);
2222 info
->bond_mode
= bond
->params
.mode
;
2223 info
->miimon
= bond
->params
.miimon
;
2225 read_lock(&bond
->lock
);
2226 info
->num_slaves
= bond
->slave_cnt
;
2227 read_unlock(&bond
->lock
);
2232 static int bond_slave_info_query(struct net_device
*bond_dev
, struct ifslave
*info
)
2234 struct bonding
*bond
= netdev_priv(bond_dev
);
2235 struct slave
*slave
;
2236 int i
, res
= -ENODEV
;
2238 read_lock(&bond
->lock
);
2240 bond_for_each_slave(bond
, slave
, i
) {
2241 if (i
== (int)info
->slave_id
) {
2243 strcpy(info
->slave_name
, slave
->dev
->name
);
2244 info
->link
= slave
->link
;
2245 info
->state
= slave
->state
;
2246 info
->link_failure_count
= slave
->link_failure_count
;
2251 read_unlock(&bond
->lock
);
2256 /*-------------------------------- Monitoring -------------------------------*/
2259 static int bond_miimon_inspect(struct bonding
*bond
)
2261 struct slave
*slave
;
2262 int i
, link_state
, commit
= 0;
2263 bool ignore_updelay
;
2265 ignore_updelay
= !bond
->curr_active_slave
? true : false;
2267 bond_for_each_slave(bond
, slave
, i
) {
2268 slave
->new_link
= BOND_LINK_NOCHANGE
;
2270 link_state
= bond_check_dev_link(bond
, slave
->dev
, 0);
2272 switch (slave
->link
) {
2277 slave
->link
= BOND_LINK_FAIL
;
2278 slave
->delay
= bond
->params
.downdelay
;
2280 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n",
2282 (bond
->params
.mode
==
2283 BOND_MODE_ACTIVEBACKUP
) ?
2284 ((slave
->state
== BOND_STATE_ACTIVE
) ?
2285 "active " : "backup ") : "",
2287 bond
->params
.downdelay
* bond
->params
.miimon
);
2290 case BOND_LINK_FAIL
:
2293 * recovered before downdelay expired
2295 slave
->link
= BOND_LINK_UP
;
2296 slave
->jiffies
= jiffies
;
2297 pr_info("%s: link status up again after %d ms for interface %s.\n",
2299 (bond
->params
.downdelay
- slave
->delay
) *
2300 bond
->params
.miimon
,
2305 if (slave
->delay
<= 0) {
2306 slave
->new_link
= BOND_LINK_DOWN
;
2314 case BOND_LINK_DOWN
:
2318 slave
->link
= BOND_LINK_BACK
;
2319 slave
->delay
= bond
->params
.updelay
;
2322 pr_info("%s: link status up for interface %s, enabling it in %d ms.\n",
2323 bond
->dev
->name
, slave
->dev
->name
,
2324 ignore_updelay
? 0 :
2325 bond
->params
.updelay
*
2326 bond
->params
.miimon
);
2329 case BOND_LINK_BACK
:
2331 slave
->link
= BOND_LINK_DOWN
;
2332 pr_info("%s: link status down again after %d ms for interface %s.\n",
2334 (bond
->params
.updelay
- slave
->delay
) *
2335 bond
->params
.miimon
,
2344 if (slave
->delay
<= 0) {
2345 slave
->new_link
= BOND_LINK_UP
;
2347 ignore_updelay
= false;
2359 static void bond_miimon_commit(struct bonding
*bond
)
2361 struct slave
*slave
;
2364 bond_for_each_slave(bond
, slave
, i
) {
2365 switch (slave
->new_link
) {
2366 case BOND_LINK_NOCHANGE
:
2370 slave
->link
= BOND_LINK_UP
;
2371 slave
->jiffies
= jiffies
;
2373 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
2374 /* prevent it from being the active one */
2375 slave
->state
= BOND_STATE_BACKUP
;
2376 } else if (bond
->params
.mode
!= BOND_MODE_ACTIVEBACKUP
) {
2377 /* make it immediately active */
2378 slave
->state
= BOND_STATE_ACTIVE
;
2379 } else if (slave
!= bond
->primary_slave
) {
2380 /* prevent it from being the active one */
2381 slave
->state
= BOND_STATE_BACKUP
;
2384 bond_update_speed_duplex(slave
);
2386 pr_info("%s: link status definitely up for interface %s, %d Mbps %s duplex.\n",
2387 bond
->dev
->name
, slave
->dev
->name
,
2388 slave
->speed
, slave
->duplex
? "full" : "half");
2390 /* notify ad that the link status has changed */
2391 if (bond
->params
.mode
== BOND_MODE_8023AD
)
2392 bond_3ad_handle_link_change(slave
, BOND_LINK_UP
);
2394 if (bond_is_lb(bond
))
2395 bond_alb_handle_link_change(bond
, slave
,
2398 if (!bond
->curr_active_slave
||
2399 (slave
== bond
->primary_slave
))
2404 case BOND_LINK_DOWN
:
2405 if (slave
->link_failure_count
< UINT_MAX
)
2406 slave
->link_failure_count
++;
2408 slave
->link
= BOND_LINK_DOWN
;
2410 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
||
2411 bond
->params
.mode
== BOND_MODE_8023AD
)
2412 bond_set_slave_inactive_flags(slave
);
2414 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2415 bond
->dev
->name
, slave
->dev
->name
);
2417 if (bond
->params
.mode
== BOND_MODE_8023AD
)
2418 bond_3ad_handle_link_change(slave
,
2421 if (bond_is_lb(bond
))
2422 bond_alb_handle_link_change(bond
, slave
,
2425 if (slave
== bond
->curr_active_slave
)
2431 pr_err("%s: invalid new link %d on slave %s\n",
2432 bond
->dev
->name
, slave
->new_link
,
2434 slave
->new_link
= BOND_LINK_NOCHANGE
;
2442 write_lock_bh(&bond
->curr_slave_lock
);
2443 bond_select_active_slave(bond
);
2444 write_unlock_bh(&bond
->curr_slave_lock
);
2445 unblock_netpoll_tx();
2448 bond_set_carrier(bond
);
2454 * Really a wrapper that splits the mii monitor into two phases: an
2455 * inspection, then (if inspection indicates something needs to be done)
2456 * an acquisition of appropriate locks followed by a commit phase to
2457 * implement whatever link state changes are indicated.
2459 void bond_mii_monitor(struct work_struct
*work
)
2461 struct bonding
*bond
= container_of(work
, struct bonding
,
2464 read_lock(&bond
->lock
);
2465 if (bond
->kill_timers
)
2468 if (bond
->slave_cnt
== 0)
2471 if (bond
->send_grat_arp
) {
2472 read_lock(&bond
->curr_slave_lock
);
2473 bond_send_gratuitous_arp(bond
);
2474 read_unlock(&bond
->curr_slave_lock
);
2477 if (bond
->send_unsol_na
) {
2478 read_lock(&bond
->curr_slave_lock
);
2479 bond_send_unsolicited_na(bond
);
2480 read_unlock(&bond
->curr_slave_lock
);
2483 if (bond_miimon_inspect(bond
)) {
2484 read_unlock(&bond
->lock
);
2486 read_lock(&bond
->lock
);
2488 bond_miimon_commit(bond
);
2490 read_unlock(&bond
->lock
);
2491 rtnl_unlock(); /* might sleep, hold no other locks */
2492 read_lock(&bond
->lock
);
2496 if (bond
->params
.miimon
)
2497 queue_delayed_work(bond
->wq
, &bond
->mii_work
,
2498 msecs_to_jiffies(bond
->params
.miimon
));
2500 read_unlock(&bond
->lock
);
2503 static __be32
bond_glean_dev_ip(struct net_device
*dev
)
2505 struct in_device
*idev
;
2506 struct in_ifaddr
*ifa
;
2513 idev
= __in_dev_get_rcu(dev
);
2517 ifa
= idev
->ifa_list
;
2521 addr
= ifa
->ifa_local
;
2527 static int bond_has_this_ip(struct bonding
*bond
, __be32 ip
)
2529 struct vlan_entry
*vlan
;
2531 if (ip
== bond
->master_ip
)
2534 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
2535 if (ip
== vlan
->vlan_ip
)
2543 * We go to the (large) trouble of VLAN tagging ARP frames because
2544 * switches in VLAN mode (especially if ports are configured as
2545 * "native" to a VLAN) might not pass non-tagged frames.
2547 static void bond_arp_send(struct net_device
*slave_dev
, int arp_op
, __be32 dest_ip
, __be32 src_ip
, unsigned short vlan_id
)
2549 struct sk_buff
*skb
;
2551 pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op
,
2552 slave_dev
->name
, dest_ip
, src_ip
, vlan_id
);
2554 skb
= arp_create(arp_op
, ETH_P_ARP
, dest_ip
, slave_dev
, src_ip
,
2555 NULL
, slave_dev
->dev_addr
, NULL
);
2558 pr_err("ARP packet allocation failed\n");
2562 skb
= vlan_put_tag(skb
, vlan_id
);
2564 pr_err("failed to insert VLAN tag\n");
2572 static void bond_arp_send_all(struct bonding
*bond
, struct slave
*slave
)
2575 __be32
*targets
= bond
->params
.arp_targets
;
2576 struct vlan_entry
*vlan
;
2577 struct net_device
*vlan_dev
;
2581 for (i
= 0; (i
< BOND_MAX_ARP_TARGETS
); i
++) {
2584 pr_debug("basa: target %x\n", targets
[i
]);
2586 pr_debug("basa: empty vlan: arp_send\n");
2587 bond_arp_send(slave
->dev
, ARPOP_REQUEST
, targets
[i
],
2588 bond
->master_ip
, 0);
2593 * If VLANs are configured, we do a route lookup to
2594 * determine which VLAN interface would be used, so we
2595 * can tag the ARP with the proper VLAN tag.
2597 memset(&fl
, 0, sizeof(fl
));
2598 fl
.fl4_dst
= targets
[i
];
2599 fl
.fl4_tos
= RTO_ONLINK
;
2601 rv
= ip_route_output_key(dev_net(bond
->dev
), &rt
, &fl
);
2603 if (net_ratelimit()) {
2604 pr_warning("%s: no route to arp_ip_target %pI4\n",
2605 bond
->dev
->name
, &fl
.fl4_dst
);
2611 * This target is not on a VLAN
2613 if (rt
->dst
.dev
== bond
->dev
) {
2615 pr_debug("basa: rtdev == bond->dev: arp_send\n");
2616 bond_arp_send(slave
->dev
, ARPOP_REQUEST
, targets
[i
],
2617 bond
->master_ip
, 0);
2622 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
2623 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
2624 if (vlan_dev
== rt
->dst
.dev
) {
2625 vlan_id
= vlan
->vlan_id
;
2626 pr_debug("basa: vlan match on %s %d\n",
2627 vlan_dev
->name
, vlan_id
);
2634 bond_arp_send(slave
->dev
, ARPOP_REQUEST
, targets
[i
],
2635 vlan
->vlan_ip
, vlan_id
);
2639 if (net_ratelimit()) {
2640 pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2641 bond
->dev
->name
, &fl
.fl4_dst
,
2642 rt
->dst
.dev
? rt
->dst
.dev
->name
: "NULL");
2649 * Kick out a gratuitous ARP for an IP on the bonding master plus one
2650 * for each VLAN above us.
2652 * Caller must hold curr_slave_lock for read or better
2654 static void bond_send_gratuitous_arp(struct bonding
*bond
)
2656 struct slave
*slave
= bond
->curr_active_slave
;
2657 struct vlan_entry
*vlan
;
2658 struct net_device
*vlan_dev
;
2660 pr_debug("bond_send_grat_arp: bond %s slave %s\n",
2661 bond
->dev
->name
, slave
? slave
->dev
->name
: "NULL");
2663 if (!slave
|| !bond
->send_grat_arp
||
2664 test_bit(__LINK_STATE_LINKWATCH_PENDING
, &slave
->dev
->state
))
2667 bond
->send_grat_arp
--;
2669 if (bond
->master_ip
) {
2670 bond_arp_send(slave
->dev
, ARPOP_REPLY
, bond
->master_ip
,
2671 bond
->master_ip
, 0);
2677 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
2678 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
2679 if (vlan
->vlan_ip
) {
2680 bond_arp_send(slave
->dev
, ARPOP_REPLY
, vlan
->vlan_ip
,
2681 vlan
->vlan_ip
, vlan
->vlan_id
);
2686 static void bond_validate_arp(struct bonding
*bond
, struct slave
*slave
, __be32 sip
, __be32 tip
)
2689 __be32
*targets
= bond
->params
.arp_targets
;
2691 for (i
= 0; (i
< BOND_MAX_ARP_TARGETS
) && targets
[i
]; i
++) {
2692 pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n",
2693 &sip
, &tip
, i
, &targets
[i
],
2694 bond_has_this_ip(bond
, tip
));
2695 if (sip
== targets
[i
]) {
2696 if (bond_has_this_ip(bond
, tip
))
2697 slave
->last_arp_rx
= jiffies
;
2703 static int bond_arp_rcv(struct sk_buff
*skb
, struct net_device
*dev
, struct packet_type
*pt
, struct net_device
*orig_dev
)
2706 struct slave
*slave
;
2707 struct bonding
*bond
;
2708 unsigned char *arp_ptr
;
2711 if (dev
->priv_flags
& IFF_802_1Q_VLAN
) {
2713 * When using VLANS and bonding, dev and oriv_dev may be
2714 * incorrect if the physical interface supports VLAN
2715 * acceleration. With this change ARP validation now
2716 * works for hosts only reachable on the VLAN interface.
2718 dev
= vlan_dev_real_dev(dev
);
2719 orig_dev
= dev_get_by_index_rcu(dev_net(skb
->dev
),skb
->skb_iif
);
2722 if (!(dev
->priv_flags
& IFF_BONDING
) || !(dev
->flags
& IFF_MASTER
))
2725 bond
= netdev_priv(dev
);
2726 read_lock(&bond
->lock
);
2728 pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n",
2729 bond
->dev
->name
, skb
->dev
? skb
->dev
->name
: "NULL",
2730 orig_dev
? orig_dev
->name
: "NULL");
2732 slave
= bond_get_slave_by_dev(bond
, orig_dev
);
2733 if (!slave
|| !slave_do_arp_validate(bond
, slave
))
2736 if (!pskb_may_pull(skb
, arp_hdr_len(dev
)))
2740 if (arp
->ar_hln
!= dev
->addr_len
||
2741 skb
->pkt_type
== PACKET_OTHERHOST
||
2742 skb
->pkt_type
== PACKET_LOOPBACK
||
2743 arp
->ar_hrd
!= htons(ARPHRD_ETHER
) ||
2744 arp
->ar_pro
!= htons(ETH_P_IP
) ||
2748 arp_ptr
= (unsigned char *)(arp
+ 1);
2749 arp_ptr
+= dev
->addr_len
;
2750 memcpy(&sip
, arp_ptr
, 4);
2751 arp_ptr
+= 4 + dev
->addr_len
;
2752 memcpy(&tip
, arp_ptr
, 4);
2754 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
2755 bond
->dev
->name
, slave
->dev
->name
, slave
->state
,
2756 bond
->params
.arp_validate
, slave_do_arp_validate(bond
, slave
),
2760 * Backup slaves won't see the ARP reply, but do come through
2761 * here for each ARP probe (so we swap the sip/tip to validate
2762 * the probe). In a "redundant switch, common router" type of
2763 * configuration, the ARP probe will (hopefully) travel from
2764 * the active, through one switch, the router, then the other
2765 * switch before reaching the backup.
2767 if (slave
->state
== BOND_STATE_ACTIVE
)
2768 bond_validate_arp(bond
, slave
, sip
, tip
);
2770 bond_validate_arp(bond
, slave
, tip
, sip
);
2773 read_unlock(&bond
->lock
);
2776 return NET_RX_SUCCESS
;
2780 * this function is called regularly to monitor each slave's link
2781 * ensuring that traffic is being sent and received when arp monitoring
2782 * is used in load-balancing mode. if the adapter has been dormant, then an
2783 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2784 * arp monitoring in active backup mode.
2786 void bond_loadbalance_arp_mon(struct work_struct
*work
)
2788 struct bonding
*bond
= container_of(work
, struct bonding
,
2790 struct slave
*slave
, *oldcurrent
;
2791 int do_failover
= 0;
2795 read_lock(&bond
->lock
);
2797 delta_in_ticks
= msecs_to_jiffies(bond
->params
.arp_interval
);
2799 if (bond
->kill_timers
)
2802 if (bond
->slave_cnt
== 0)
2805 read_lock(&bond
->curr_slave_lock
);
2806 oldcurrent
= bond
->curr_active_slave
;
2807 read_unlock(&bond
->curr_slave_lock
);
2809 /* see if any of the previous devices are up now (i.e. they have
2810 * xmt and rcv traffic). the curr_active_slave does not come into
2811 * the picture unless it is null. also, slave->jiffies is not needed
2812 * here because we send an arp on each slave and give a slave as
2813 * long as it needs to get the tx/rx within the delta.
2814 * TODO: what about up/down delay in arp mode? it wasn't here before
2817 bond_for_each_slave(bond
, slave
, i
) {
2818 unsigned long trans_start
= dev_trans_start(slave
->dev
);
2820 if (slave
->link
!= BOND_LINK_UP
) {
2821 if (time_in_range(jiffies
,
2822 trans_start
- delta_in_ticks
,
2823 trans_start
+ delta_in_ticks
) &&
2824 time_in_range(jiffies
,
2825 slave
->dev
->last_rx
- delta_in_ticks
,
2826 slave
->dev
->last_rx
+ delta_in_ticks
)) {
2828 slave
->link
= BOND_LINK_UP
;
2829 slave
->state
= BOND_STATE_ACTIVE
;
2831 /* primary_slave has no meaning in round-robin
2832 * mode. the window of a slave being up and
2833 * curr_active_slave being null after enslaving
2837 pr_info("%s: link status definitely up for interface %s, ",
2842 pr_info("%s: interface %s is now up\n",
2848 /* slave->link == BOND_LINK_UP */
2850 /* not all switches will respond to an arp request
2851 * when the source ip is 0, so don't take the link down
2852 * if we don't know our ip yet
2854 if (!time_in_range(jiffies
,
2855 trans_start
- delta_in_ticks
,
2856 trans_start
+ 2 * delta_in_ticks
) ||
2857 !time_in_range(jiffies
,
2858 slave
->dev
->last_rx
- delta_in_ticks
,
2859 slave
->dev
->last_rx
+ 2 * delta_in_ticks
)) {
2861 slave
->link
= BOND_LINK_DOWN
;
2862 slave
->state
= BOND_STATE_BACKUP
;
2864 if (slave
->link_failure_count
< UINT_MAX
)
2865 slave
->link_failure_count
++;
2867 pr_info("%s: interface %s is now down.\n",
2871 if (slave
== oldcurrent
)
2876 /* note: if switch is in round-robin mode, all links
2877 * must tx arp to ensure all links rx an arp - otherwise
2878 * links may oscillate or not come up at all; if switch is
2879 * in something like xor mode, there is nothing we can
2880 * do - all replies will be rx'ed on same link causing slaves
2881 * to be unstable during low/no traffic periods
2883 if (IS_UP(slave
->dev
))
2884 bond_arp_send_all(bond
, slave
);
2889 write_lock_bh(&bond
->curr_slave_lock
);
2891 bond_select_active_slave(bond
);
2893 write_unlock_bh(&bond
->curr_slave_lock
);
2894 unblock_netpoll_tx();
2898 if (bond
->params
.arp_interval
)
2899 queue_delayed_work(bond
->wq
, &bond
->arp_work
, delta_in_ticks
);
2901 read_unlock(&bond
->lock
);
2905 * Called to inspect slaves for active-backup mode ARP monitor link state
2906 * changes. Sets new_link in slaves to specify what action should take
2907 * place for the slave. Returns 0 if no changes are found, >0 if changes
2908 * to link states must be committed.
2910 * Called with bond->lock held for read.
2912 static int bond_ab_arp_inspect(struct bonding
*bond
, int delta_in_ticks
)
2914 struct slave
*slave
;
2916 unsigned long trans_start
;
2918 bond_for_each_slave(bond
, slave
, i
) {
2919 slave
->new_link
= BOND_LINK_NOCHANGE
;
2921 if (slave
->link
!= BOND_LINK_UP
) {
2922 if (time_in_range(jiffies
,
2923 slave_last_rx(bond
, slave
) - delta_in_ticks
,
2924 slave_last_rx(bond
, slave
) + delta_in_ticks
)) {
2926 slave
->new_link
= BOND_LINK_UP
;
2934 * Give slaves 2*delta after being enslaved or made
2935 * active. This avoids bouncing, as the last receive
2936 * times need a full ARP monitor cycle to be updated.
2938 if (time_in_range(jiffies
,
2939 slave
->jiffies
- delta_in_ticks
,
2940 slave
->jiffies
+ 2 * delta_in_ticks
))
2944 * Backup slave is down if:
2945 * - No current_arp_slave AND
2946 * - more than 3*delta since last receive AND
2947 * - the bond has an IP address
2949 * Note: a non-null current_arp_slave indicates
2950 * the curr_active_slave went down and we are
2951 * searching for a new one; under this condition
2952 * we only take the curr_active_slave down - this
2953 * gives each slave a chance to tx/rx traffic
2954 * before being taken out
2956 if (slave
->state
== BOND_STATE_BACKUP
&&
2957 !bond
->current_arp_slave
&&
2958 !time_in_range(jiffies
,
2959 slave_last_rx(bond
, slave
) - delta_in_ticks
,
2960 slave_last_rx(bond
, slave
) + 3 * delta_in_ticks
)) {
2962 slave
->new_link
= BOND_LINK_DOWN
;
2967 * Active slave is down if:
2968 * - more than 2*delta since transmitting OR
2969 * - (more than 2*delta since receive AND
2970 * the bond has an IP address)
2972 trans_start
= dev_trans_start(slave
->dev
);
2973 if ((slave
->state
== BOND_STATE_ACTIVE
) &&
2974 (!time_in_range(jiffies
,
2975 trans_start
- delta_in_ticks
,
2976 trans_start
+ 2 * delta_in_ticks
) ||
2977 !time_in_range(jiffies
,
2978 slave_last_rx(bond
, slave
) - delta_in_ticks
,
2979 slave_last_rx(bond
, slave
) + 2 * delta_in_ticks
))) {
2981 slave
->new_link
= BOND_LINK_DOWN
;
2990 * Called to commit link state changes noted by inspection step of
2991 * active-backup mode ARP monitor.
2993 * Called with RTNL and bond->lock for read.
2995 static void bond_ab_arp_commit(struct bonding
*bond
, int delta_in_ticks
)
2997 struct slave
*slave
;
2999 unsigned long trans_start
;
3001 bond_for_each_slave(bond
, slave
, i
) {
3002 switch (slave
->new_link
) {
3003 case BOND_LINK_NOCHANGE
:
3007 trans_start
= dev_trans_start(slave
->dev
);
3008 if ((!bond
->curr_active_slave
&&
3009 time_in_range(jiffies
,
3010 trans_start
- delta_in_ticks
,
3011 trans_start
+ delta_in_ticks
)) ||
3012 bond
->curr_active_slave
!= slave
) {
3013 slave
->link
= BOND_LINK_UP
;
3014 bond
->current_arp_slave
= NULL
;
3016 pr_info("%s: link status definitely up for interface %s.\n",
3017 bond
->dev
->name
, slave
->dev
->name
);
3019 if (!bond
->curr_active_slave
||
3020 (slave
== bond
->primary_slave
))
3027 case BOND_LINK_DOWN
:
3028 if (slave
->link_failure_count
< UINT_MAX
)
3029 slave
->link_failure_count
++;
3031 slave
->link
= BOND_LINK_DOWN
;
3032 bond_set_slave_inactive_flags(slave
);
3034 pr_info("%s: link status definitely down for interface %s, disabling it\n",
3035 bond
->dev
->name
, slave
->dev
->name
);
3037 if (slave
== bond
->curr_active_slave
) {
3038 bond
->current_arp_slave
= NULL
;
3045 pr_err("%s: impossible: new_link %d on slave %s\n",
3046 bond
->dev
->name
, slave
->new_link
,
3054 write_lock_bh(&bond
->curr_slave_lock
);
3055 bond_select_active_slave(bond
);
3056 write_unlock_bh(&bond
->curr_slave_lock
);
3057 unblock_netpoll_tx();
3060 bond_set_carrier(bond
);
3064 * Send ARP probes for active-backup mode ARP monitor.
3066 * Called with bond->lock held for read.
3068 static void bond_ab_arp_probe(struct bonding
*bond
)
3070 struct slave
*slave
;
3073 read_lock(&bond
->curr_slave_lock
);
3075 if (bond
->current_arp_slave
&& bond
->curr_active_slave
)
3076 pr_info("PROBE: c_arp %s && cas %s BAD\n",
3077 bond
->current_arp_slave
->dev
->name
,
3078 bond
->curr_active_slave
->dev
->name
);
3080 if (bond
->curr_active_slave
) {
3081 bond_arp_send_all(bond
, bond
->curr_active_slave
);
3082 read_unlock(&bond
->curr_slave_lock
);
3086 read_unlock(&bond
->curr_slave_lock
);
3088 /* if we don't have a curr_active_slave, search for the next available
3089 * backup slave from the current_arp_slave and make it the candidate
3090 * for becoming the curr_active_slave
3093 if (!bond
->current_arp_slave
) {
3094 bond
->current_arp_slave
= bond
->first_slave
;
3095 if (!bond
->current_arp_slave
)
3099 bond_set_slave_inactive_flags(bond
->current_arp_slave
);
3101 /* search for next candidate */
3102 bond_for_each_slave_from(bond
, slave
, i
, bond
->current_arp_slave
->next
) {
3103 if (IS_UP(slave
->dev
)) {
3104 slave
->link
= BOND_LINK_BACK
;
3105 bond_set_slave_active_flags(slave
);
3106 bond_arp_send_all(bond
, slave
);
3107 slave
->jiffies
= jiffies
;
3108 bond
->current_arp_slave
= slave
;
3112 /* if the link state is up at this point, we
3113 * mark it down - this can happen if we have
3114 * simultaneous link failures and
3115 * reselect_active_interface doesn't make this
3116 * one the current slave so it is still marked
3117 * up when it is actually down
3119 if (slave
->link
== BOND_LINK_UP
) {
3120 slave
->link
= BOND_LINK_DOWN
;
3121 if (slave
->link_failure_count
< UINT_MAX
)
3122 slave
->link_failure_count
++;
3124 bond_set_slave_inactive_flags(slave
);
3126 pr_info("%s: backup interface %s is now down.\n",
3127 bond
->dev
->name
, slave
->dev
->name
);
3132 void bond_activebackup_arp_mon(struct work_struct
*work
)
3134 struct bonding
*bond
= container_of(work
, struct bonding
,
3138 read_lock(&bond
->lock
);
3140 if (bond
->kill_timers
)
3143 delta_in_ticks
= msecs_to_jiffies(bond
->params
.arp_interval
);
3145 if (bond
->slave_cnt
== 0)
3148 if (bond
->send_grat_arp
) {
3149 read_lock(&bond
->curr_slave_lock
);
3150 bond_send_gratuitous_arp(bond
);
3151 read_unlock(&bond
->curr_slave_lock
);
3154 if (bond
->send_unsol_na
) {
3155 read_lock(&bond
->curr_slave_lock
);
3156 bond_send_unsolicited_na(bond
);
3157 read_unlock(&bond
->curr_slave_lock
);
3160 if (bond_ab_arp_inspect(bond
, delta_in_ticks
)) {
3161 read_unlock(&bond
->lock
);
3163 read_lock(&bond
->lock
);
3165 bond_ab_arp_commit(bond
, delta_in_ticks
);
3167 read_unlock(&bond
->lock
);
3169 read_lock(&bond
->lock
);
3172 bond_ab_arp_probe(bond
);
3175 if (bond
->params
.arp_interval
)
3176 queue_delayed_work(bond
->wq
, &bond
->arp_work
, delta_in_ticks
);
3178 read_unlock(&bond
->lock
);
3181 /*------------------------------ proc/seq_file-------------------------------*/
3183 #ifdef CONFIG_PROC_FS
3185 static void *bond_info_seq_start(struct seq_file
*seq
, loff_t
*pos
)
3187 __acquires(&bond
->lock
)
3189 struct bonding
*bond
= seq
->private;
3191 struct slave
*slave
;
3194 /* make sure the bond won't be taken away */
3196 read_lock(&bond
->lock
);
3199 return SEQ_START_TOKEN
;
3201 bond_for_each_slave(bond
, slave
, i
) {
3209 static void *bond_info_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
3211 struct bonding
*bond
= seq
->private;
3212 struct slave
*slave
= v
;
3215 if (v
== SEQ_START_TOKEN
)
3216 return bond
->first_slave
;
3218 slave
= slave
->next
;
3220 return (slave
== bond
->first_slave
) ? NULL
: slave
;
3223 static void bond_info_seq_stop(struct seq_file
*seq
, void *v
)
3224 __releases(&bond
->lock
)
3227 struct bonding
*bond
= seq
->private;
3229 read_unlock(&bond
->lock
);
3233 static void bond_info_show_master(struct seq_file
*seq
)
3235 struct bonding
*bond
= seq
->private;
3239 read_lock(&bond
->curr_slave_lock
);
3240 curr
= bond
->curr_active_slave
;
3241 read_unlock(&bond
->curr_slave_lock
);
3243 seq_printf(seq
, "Bonding Mode: %s",
3244 bond_mode_name(bond
->params
.mode
));
3246 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
&&
3247 bond
->params
.fail_over_mac
)
3248 seq_printf(seq
, " (fail_over_mac %s)",
3249 fail_over_mac_tbl
[bond
->params
.fail_over_mac
].modename
);
3251 seq_printf(seq
, "\n");
3253 if (bond
->params
.mode
== BOND_MODE_XOR
||
3254 bond
->params
.mode
== BOND_MODE_8023AD
) {
3255 seq_printf(seq
, "Transmit Hash Policy: %s (%d)\n",
3256 xmit_hashtype_tbl
[bond
->params
.xmit_policy
].modename
,
3257 bond
->params
.xmit_policy
);
3260 if (USES_PRIMARY(bond
->params
.mode
)) {
3261 seq_printf(seq
, "Primary Slave: %s",
3262 (bond
->primary_slave
) ?
3263 bond
->primary_slave
->dev
->name
: "None");
3264 if (bond
->primary_slave
)
3265 seq_printf(seq
, " (primary_reselect %s)",
3266 pri_reselect_tbl
[bond
->params
.primary_reselect
].modename
);
3268 seq_printf(seq
, "\nCurrently Active Slave: %s\n",
3269 (curr
) ? curr
->dev
->name
: "None");
3272 seq_printf(seq
, "MII Status: %s\n", netif_carrier_ok(bond
->dev
) ?
3274 seq_printf(seq
, "MII Polling Interval (ms): %d\n", bond
->params
.miimon
);
3275 seq_printf(seq
, "Up Delay (ms): %d\n",
3276 bond
->params
.updelay
* bond
->params
.miimon
);
3277 seq_printf(seq
, "Down Delay (ms): %d\n",
3278 bond
->params
.downdelay
* bond
->params
.miimon
);
3281 /* ARP information */
3282 if (bond
->params
.arp_interval
> 0) {
3284 seq_printf(seq
, "ARP Polling Interval (ms): %d\n",
3285 bond
->params
.arp_interval
);
3287 seq_printf(seq
, "ARP IP target/s (n.n.n.n form):");
3289 for (i
= 0; (i
< BOND_MAX_ARP_TARGETS
); i
++) {
3290 if (!bond
->params
.arp_targets
[i
])
3293 seq_printf(seq
, ",");
3294 seq_printf(seq
, " %pI4", &bond
->params
.arp_targets
[i
]);
3297 seq_printf(seq
, "\n");
3300 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3301 struct ad_info ad_info
;
3303 seq_puts(seq
, "\n802.3ad info\n");
3304 seq_printf(seq
, "LACP rate: %s\n",
3305 (bond
->params
.lacp_fast
) ? "fast" : "slow");
3306 seq_printf(seq
, "Aggregator selection policy (ad_select): %s\n",
3307 ad_select_tbl
[bond
->params
.ad_select
].modename
);
3309 if (bond_3ad_get_active_agg_info(bond
, &ad_info
)) {
3310 seq_printf(seq
, "bond %s has no active aggregator\n",
3313 seq_printf(seq
, "Active Aggregator Info:\n");
3315 seq_printf(seq
, "\tAggregator ID: %d\n",
3316 ad_info
.aggregator_id
);
3317 seq_printf(seq
, "\tNumber of ports: %d\n",
3319 seq_printf(seq
, "\tActor Key: %d\n",
3321 seq_printf(seq
, "\tPartner Key: %d\n",
3322 ad_info
.partner_key
);
3323 seq_printf(seq
, "\tPartner Mac Address: %pM\n",
3324 ad_info
.partner_system
);
3329 static void bond_info_show_slave(struct seq_file
*seq
,
3330 const struct slave
*slave
)
3332 struct bonding
*bond
= seq
->private;
3334 seq_printf(seq
, "\nSlave Interface: %s\n", slave
->dev
->name
);
3335 seq_printf(seq
, "MII Status: %s\n",
3336 (slave
->link
== BOND_LINK_UP
) ? "up" : "down");
3337 seq_printf(seq
, "Speed: %d Mbps\n", slave
->speed
);
3338 seq_printf(seq
, "Duplex: %s\n", slave
->duplex
? "full" : "half");
3339 seq_printf(seq
, "Link Failure Count: %u\n",
3340 slave
->link_failure_count
);
3342 seq_printf(seq
, "Permanent HW addr: %pM\n", slave
->perm_hwaddr
);
3344 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3345 const struct aggregator
*agg
3346 = SLAVE_AD_INFO(slave
).port
.aggregator
;
3349 seq_printf(seq
, "Aggregator ID: %d\n",
3350 agg
->aggregator_identifier
);
3352 seq_puts(seq
, "Aggregator ID: N/A\n");
3354 seq_printf(seq
, "Slave queue ID: %d\n", slave
->queue_id
);
3357 static int bond_info_seq_show(struct seq_file
*seq
, void *v
)
3359 if (v
== SEQ_START_TOKEN
) {
3360 seq_printf(seq
, "%s\n", version
);
3361 bond_info_show_master(seq
);
3363 bond_info_show_slave(seq
, v
);
3368 static const struct seq_operations bond_info_seq_ops
= {
3369 .start
= bond_info_seq_start
,
3370 .next
= bond_info_seq_next
,
3371 .stop
= bond_info_seq_stop
,
3372 .show
= bond_info_seq_show
,
3375 static int bond_info_open(struct inode
*inode
, struct file
*file
)
3377 struct seq_file
*seq
;
3378 struct proc_dir_entry
*proc
;
3381 res
= seq_open(file
, &bond_info_seq_ops
);
3383 /* recover the pointer buried in proc_dir_entry data */
3384 seq
= file
->private_data
;
3386 seq
->private = proc
->data
;
3392 static const struct file_operations bond_info_fops
= {
3393 .owner
= THIS_MODULE
,
3394 .open
= bond_info_open
,
3396 .llseek
= seq_lseek
,
3397 .release
= seq_release
,
3400 static void bond_create_proc_entry(struct bonding
*bond
)
3402 struct net_device
*bond_dev
= bond
->dev
;
3403 struct bond_net
*bn
= net_generic(dev_net(bond_dev
), bond_net_id
);
3406 bond
->proc_entry
= proc_create_data(bond_dev
->name
,
3407 S_IRUGO
, bn
->proc_dir
,
3408 &bond_info_fops
, bond
);
3409 if (bond
->proc_entry
== NULL
)
3410 pr_warning("Warning: Cannot create /proc/net/%s/%s\n",
3411 DRV_NAME
, bond_dev
->name
);
3413 memcpy(bond
->proc_file_name
, bond_dev
->name
, IFNAMSIZ
);
3417 static void bond_remove_proc_entry(struct bonding
*bond
)
3419 struct net_device
*bond_dev
= bond
->dev
;
3420 struct bond_net
*bn
= net_generic(dev_net(bond_dev
), bond_net_id
);
3422 if (bn
->proc_dir
&& bond
->proc_entry
) {
3423 remove_proc_entry(bond
->proc_file_name
, bn
->proc_dir
);
3424 memset(bond
->proc_file_name
, 0, IFNAMSIZ
);
3425 bond
->proc_entry
= NULL
;
3429 /* Create the bonding directory under /proc/net, if doesn't exist yet.
3430 * Caller must hold rtnl_lock.
3432 static void __net_init
bond_create_proc_dir(struct bond_net
*bn
)
3434 if (!bn
->proc_dir
) {
3435 bn
->proc_dir
= proc_mkdir(DRV_NAME
, bn
->net
->proc_net
);
3437 pr_warning("Warning: cannot create /proc/net/%s\n",
3442 /* Destroy the bonding directory under /proc/net, if empty.
3443 * Caller must hold rtnl_lock.
3445 static void __net_exit
bond_destroy_proc_dir(struct bond_net
*bn
)
3448 remove_proc_entry(DRV_NAME
, bn
->net
->proc_net
);
3449 bn
->proc_dir
= NULL
;
3453 #else /* !CONFIG_PROC_FS */
3455 static void bond_create_proc_entry(struct bonding
*bond
)
3459 static void bond_remove_proc_entry(struct bonding
*bond
)
3463 static inline void bond_create_proc_dir(struct bond_net
*bn
)
3467 static inline void bond_destroy_proc_dir(struct bond_net
*bn
)
3471 #endif /* CONFIG_PROC_FS */
3474 /*-------------------------- netdev event handling --------------------------*/
3477 * Change device name
3479 static int bond_event_changename(struct bonding
*bond
)
3481 bond_remove_proc_entry(bond
);
3482 bond_create_proc_entry(bond
);
3484 bond_debug_reregister(bond
);
3489 static int bond_master_netdev_event(unsigned long event
,
3490 struct net_device
*bond_dev
)
3492 struct bonding
*event_bond
= netdev_priv(bond_dev
);
3495 case NETDEV_CHANGENAME
:
3496 return bond_event_changename(event_bond
);
3504 static int bond_slave_netdev_event(unsigned long event
,
3505 struct net_device
*slave_dev
)
3507 struct net_device
*bond_dev
= slave_dev
->master
;
3508 struct bonding
*bond
= netdev_priv(bond_dev
);
3511 case NETDEV_UNREGISTER
:
3513 if (bond
->setup_by_slave
)
3514 bond_release_and_destroy(bond_dev
, slave_dev
);
3516 bond_release(bond_dev
, slave_dev
);
3520 if (bond
->params
.mode
== BOND_MODE_8023AD
|| bond_is_lb(bond
)) {
3521 struct slave
*slave
;
3523 slave
= bond_get_slave_by_dev(bond
, slave_dev
);
3525 u16 old_speed
= slave
->speed
;
3526 u16 old_duplex
= slave
->duplex
;
3528 bond_update_speed_duplex(slave
);
3530 if (bond_is_lb(bond
))
3533 if (old_speed
!= slave
->speed
)
3534 bond_3ad_adapter_speed_changed(slave
);
3535 if (old_duplex
!= slave
->duplex
)
3536 bond_3ad_adapter_duplex_changed(slave
);
3543 * ... Or is it this?
3546 case NETDEV_CHANGEMTU
:
3548 * TODO: Should slaves be allowed to
3549 * independently alter their MTU? For
3550 * an active-backup bond, slaves need
3551 * not be the same type of device, so
3552 * MTUs may vary. For other modes,
3553 * slaves arguably should have the
3554 * same MTUs. To do this, we'd need to
3555 * take over the slave's change_mtu
3556 * function for the duration of their
3560 case NETDEV_CHANGENAME
:
3562 * TODO: handle changing the primary's name
3565 case NETDEV_FEAT_CHANGE
:
3566 bond_compute_features(bond
);
3576 * bond_netdev_event: handle netdev notifier chain events.
3578 * This function receives events for the netdev chain. The caller (an
3579 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
3580 * locks for us to safely manipulate the slave devices (RTNL lock,
3583 static int bond_netdev_event(struct notifier_block
*this,
3584 unsigned long event
, void *ptr
)
3586 struct net_device
*event_dev
= (struct net_device
*)ptr
;
3588 pr_debug("event_dev: %s, event: %lx\n",
3589 event_dev
? event_dev
->name
: "None",
3592 if (!(event_dev
->priv_flags
& IFF_BONDING
))
3595 if (event_dev
->flags
& IFF_MASTER
) {
3596 pr_debug("IFF_MASTER\n");
3597 return bond_master_netdev_event(event
, event_dev
);
3600 if (event_dev
->flags
& IFF_SLAVE
) {
3601 pr_debug("IFF_SLAVE\n");
3602 return bond_slave_netdev_event(event
, event_dev
);
3609 * bond_inetaddr_event: handle inetaddr notifier chain events.
3611 * We keep track of device IPs primarily to use as source addresses in
3612 * ARP monitor probes (rather than spewing out broadcasts all the time).
3614 * We track one IP for the main device (if it has one), plus one per VLAN.
3616 static int bond_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
3618 struct in_ifaddr
*ifa
= ptr
;
3619 struct net_device
*vlan_dev
, *event_dev
= ifa
->ifa_dev
->dev
;
3620 struct bond_net
*bn
= net_generic(dev_net(event_dev
), bond_net_id
);
3621 struct bonding
*bond
;
3622 struct vlan_entry
*vlan
;
3624 list_for_each_entry(bond
, &bn
->dev_list
, bond_list
) {
3625 if (bond
->dev
== event_dev
) {
3628 bond
->master_ip
= ifa
->ifa_local
;
3631 bond
->master_ip
= bond_glean_dev_ip(bond
->dev
);
3638 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
3641 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
3642 if (vlan_dev
== event_dev
) {
3645 vlan
->vlan_ip
= ifa
->ifa_local
;
3649 bond_glean_dev_ip(vlan_dev
);
3660 static struct notifier_block bond_netdev_notifier
= {
3661 .notifier_call
= bond_netdev_event
,
3664 static struct notifier_block bond_inetaddr_notifier
= {
3665 .notifier_call
= bond_inetaddr_event
,
3668 /*-------------------------- Packet type handling ---------------------------*/
3670 /* register to receive lacpdus on a bond */
3671 static void bond_register_lacpdu(struct bonding
*bond
)
3673 struct packet_type
*pk_type
= &(BOND_AD_INFO(bond
).ad_pkt_type
);
3675 /* initialize packet type */
3676 pk_type
->type
= PKT_TYPE_LACPDU
;
3677 pk_type
->dev
= bond
->dev
;
3678 pk_type
->func
= bond_3ad_lacpdu_recv
;
3680 dev_add_pack(pk_type
);
3683 /* unregister to receive lacpdus on a bond */
3684 static void bond_unregister_lacpdu(struct bonding
*bond
)
3686 dev_remove_pack(&(BOND_AD_INFO(bond
).ad_pkt_type
));
3689 void bond_register_arp(struct bonding
*bond
)
3691 struct packet_type
*pt
= &bond
->arp_mon_pt
;
3696 pt
->type
= htons(ETH_P_ARP
);
3697 pt
->dev
= bond
->dev
;
3698 pt
->func
= bond_arp_rcv
;
3702 void bond_unregister_arp(struct bonding
*bond
)
3704 struct packet_type
*pt
= &bond
->arp_mon_pt
;
3706 dev_remove_pack(pt
);
3710 /*---------------------------- Hashing Policies -----------------------------*/
3713 * Hash for the output device based upon layer 2 and layer 3 data. If
3714 * the packet is not IP mimic bond_xmit_hash_policy_l2()
3716 static int bond_xmit_hash_policy_l23(struct sk_buff
*skb
, int count
)
3718 struct ethhdr
*data
= (struct ethhdr
*)skb
->data
;
3719 struct iphdr
*iph
= ip_hdr(skb
);
3721 if (skb
->protocol
== htons(ETH_P_IP
)) {
3722 return ((ntohl(iph
->saddr
^ iph
->daddr
) & 0xffff) ^
3723 (data
->h_dest
[5] ^ data
->h_source
[5])) % count
;
3726 return (data
->h_dest
[5] ^ data
->h_source
[5]) % count
;
3730 * Hash for the output device based upon layer 3 and layer 4 data. If
3731 * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is
3732 * altogether not IP, mimic bond_xmit_hash_policy_l2()
3734 static int bond_xmit_hash_policy_l34(struct sk_buff
*skb
, int count
)
3736 struct ethhdr
*data
= (struct ethhdr
*)skb
->data
;
3737 struct iphdr
*iph
= ip_hdr(skb
);
3738 __be16
*layer4hdr
= (__be16
*)((u32
*)iph
+ iph
->ihl
);
3741 if (skb
->protocol
== htons(ETH_P_IP
)) {
3742 if (!(iph
->frag_off
& htons(IP_MF
|IP_OFFSET
)) &&
3743 (iph
->protocol
== IPPROTO_TCP
||
3744 iph
->protocol
== IPPROTO_UDP
)) {
3745 layer4_xor
= ntohs((*layer4hdr
^ *(layer4hdr
+ 1)));
3747 return (layer4_xor
^
3748 ((ntohl(iph
->saddr
^ iph
->daddr
)) & 0xffff)) % count
;
3752 return (data
->h_dest
[5] ^ data
->h_source
[5]) % count
;
3756 * Hash for the output device based upon layer 2 data
3758 static int bond_xmit_hash_policy_l2(struct sk_buff
*skb
, int count
)
3760 struct ethhdr
*data
= (struct ethhdr
*)skb
->data
;
3762 return (data
->h_dest
[5] ^ data
->h_source
[5]) % count
;
3765 /*-------------------------- Device entry points ----------------------------*/
3767 static int bond_open(struct net_device
*bond_dev
)
3769 struct bonding
*bond
= netdev_priv(bond_dev
);
3771 bond
->kill_timers
= 0;
3773 INIT_DELAYED_WORK(&bond
->mcast_work
, bond_resend_igmp_join_requests_delayed
);
3775 if (bond_is_lb(bond
)) {
3776 /* bond_alb_initialize must be called before the timer
3779 if (bond_alb_initialize(bond
, (bond
->params
.mode
== BOND_MODE_ALB
))) {
3780 /* something went wrong - fail the open operation */
3784 INIT_DELAYED_WORK(&bond
->alb_work
, bond_alb_monitor
);
3785 queue_delayed_work(bond
->wq
, &bond
->alb_work
, 0);
3788 if (bond
->params
.miimon
) { /* link check interval, in milliseconds. */
3789 INIT_DELAYED_WORK(&bond
->mii_work
, bond_mii_monitor
);
3790 queue_delayed_work(bond
->wq
, &bond
->mii_work
, 0);
3793 if (bond
->params
.arp_interval
) { /* arp interval, in milliseconds. */
3794 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
)
3795 INIT_DELAYED_WORK(&bond
->arp_work
,
3796 bond_activebackup_arp_mon
);
3798 INIT_DELAYED_WORK(&bond
->arp_work
,
3799 bond_loadbalance_arp_mon
);
3801 queue_delayed_work(bond
->wq
, &bond
->arp_work
, 0);
3802 if (bond
->params
.arp_validate
)
3803 bond_register_arp(bond
);
3806 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3807 INIT_DELAYED_WORK(&bond
->ad_work
, bond_3ad_state_machine_handler
);
3808 queue_delayed_work(bond
->wq
, &bond
->ad_work
, 0);
3809 /* register to receive LACPDUs */
3810 bond_register_lacpdu(bond
);
3811 bond_3ad_initiate_agg_selection(bond
, 1);
3817 static int bond_close(struct net_device
*bond_dev
)
3819 struct bonding
*bond
= netdev_priv(bond_dev
);
3821 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3822 /* Unregister the receive of LACPDUs */
3823 bond_unregister_lacpdu(bond
);
3826 if (bond
->params
.arp_validate
)
3827 bond_unregister_arp(bond
);
3829 write_lock_bh(&bond
->lock
);
3831 bond
->send_grat_arp
= 0;
3832 bond
->send_unsol_na
= 0;
3834 /* signal timers not to re-arm */
3835 bond
->kill_timers
= 1;
3837 write_unlock_bh(&bond
->lock
);
3839 if (bond
->params
.miimon
) { /* link check interval, in milliseconds. */
3840 cancel_delayed_work(&bond
->mii_work
);
3843 if (bond
->params
.arp_interval
) { /* arp interval, in milliseconds. */
3844 cancel_delayed_work(&bond
->arp_work
);
3847 switch (bond
->params
.mode
) {
3848 case BOND_MODE_8023AD
:
3849 cancel_delayed_work(&bond
->ad_work
);
3853 cancel_delayed_work(&bond
->alb_work
);
3859 if (delayed_work_pending(&bond
->mcast_work
))
3860 cancel_delayed_work(&bond
->mcast_work
);
3862 if (bond_is_lb(bond
)) {
3863 /* Must be called only after all
3864 * slaves have been released
3866 bond_alb_deinitialize(bond
);
3872 static struct rtnl_link_stats64
*bond_get_stats(struct net_device
*bond_dev
,
3873 struct rtnl_link_stats64
*stats
)
3875 struct bonding
*bond
= netdev_priv(bond_dev
);
3876 struct rtnl_link_stats64 temp
;
3877 struct slave
*slave
;
3880 memset(stats
, 0, sizeof(*stats
));
3882 read_lock_bh(&bond
->lock
);
3884 bond_for_each_slave(bond
, slave
, i
) {
3885 const struct rtnl_link_stats64
*sstats
=
3886 dev_get_stats(slave
->dev
, &temp
);
3888 stats
->rx_packets
+= sstats
->rx_packets
;
3889 stats
->rx_bytes
+= sstats
->rx_bytes
;
3890 stats
->rx_errors
+= sstats
->rx_errors
;
3891 stats
->rx_dropped
+= sstats
->rx_dropped
;
3893 stats
->tx_packets
+= sstats
->tx_packets
;
3894 stats
->tx_bytes
+= sstats
->tx_bytes
;
3895 stats
->tx_errors
+= sstats
->tx_errors
;
3896 stats
->tx_dropped
+= sstats
->tx_dropped
;
3898 stats
->multicast
+= sstats
->multicast
;
3899 stats
->collisions
+= sstats
->collisions
;
3901 stats
->rx_length_errors
+= sstats
->rx_length_errors
;
3902 stats
->rx_over_errors
+= sstats
->rx_over_errors
;
3903 stats
->rx_crc_errors
+= sstats
->rx_crc_errors
;
3904 stats
->rx_frame_errors
+= sstats
->rx_frame_errors
;
3905 stats
->rx_fifo_errors
+= sstats
->rx_fifo_errors
;
3906 stats
->rx_missed_errors
+= sstats
->rx_missed_errors
;
3908 stats
->tx_aborted_errors
+= sstats
->tx_aborted_errors
;
3909 stats
->tx_carrier_errors
+= sstats
->tx_carrier_errors
;
3910 stats
->tx_fifo_errors
+= sstats
->tx_fifo_errors
;
3911 stats
->tx_heartbeat_errors
+= sstats
->tx_heartbeat_errors
;
3912 stats
->tx_window_errors
+= sstats
->tx_window_errors
;
3915 read_unlock_bh(&bond
->lock
);
3920 static int bond_do_ioctl(struct net_device
*bond_dev
, struct ifreq
*ifr
, int cmd
)
3922 struct net_device
*slave_dev
= NULL
;
3923 struct ifbond k_binfo
;
3924 struct ifbond __user
*u_binfo
= NULL
;
3925 struct ifslave k_sinfo
;
3926 struct ifslave __user
*u_sinfo
= NULL
;
3927 struct mii_ioctl_data
*mii
= NULL
;
3930 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev
->name
, cmd
);
3942 * We do this again just in case we were called by SIOCGMIIREG
3943 * instead of SIOCGMIIPHY.
3950 if (mii
->reg_num
== 1) {
3951 struct bonding
*bond
= netdev_priv(bond_dev
);
3953 read_lock(&bond
->lock
);
3954 read_lock(&bond
->curr_slave_lock
);
3955 if (netif_carrier_ok(bond
->dev
))
3956 mii
->val_out
= BMSR_LSTATUS
;
3958 read_unlock(&bond
->curr_slave_lock
);
3959 read_unlock(&bond
->lock
);
3963 case BOND_INFO_QUERY_OLD
:
3964 case SIOCBONDINFOQUERY
:
3965 u_binfo
= (struct ifbond __user
*)ifr
->ifr_data
;
3967 if (copy_from_user(&k_binfo
, u_binfo
, sizeof(ifbond
)))
3970 res
= bond_info_query(bond_dev
, &k_binfo
);
3972 copy_to_user(u_binfo
, &k_binfo
, sizeof(ifbond
)))
3976 case BOND_SLAVE_INFO_QUERY_OLD
:
3977 case SIOCBONDSLAVEINFOQUERY
:
3978 u_sinfo
= (struct ifslave __user
*)ifr
->ifr_data
;
3980 if (copy_from_user(&k_sinfo
, u_sinfo
, sizeof(ifslave
)))
3983 res
= bond_slave_info_query(bond_dev
, &k_sinfo
);
3985 copy_to_user(u_sinfo
, &k_sinfo
, sizeof(ifslave
)))
3994 if (!capable(CAP_NET_ADMIN
))
3997 slave_dev
= dev_get_by_name(dev_net(bond_dev
), ifr
->ifr_slave
);
3999 pr_debug("slave_dev=%p:\n", slave_dev
);
4004 pr_debug("slave_dev->name=%s:\n", slave_dev
->name
);
4006 case BOND_ENSLAVE_OLD
:
4007 case SIOCBONDENSLAVE
:
4008 res
= bond_enslave(bond_dev
, slave_dev
);
4010 case BOND_RELEASE_OLD
:
4011 case SIOCBONDRELEASE
:
4012 res
= bond_release(bond_dev
, slave_dev
);
4014 case BOND_SETHWADDR_OLD
:
4015 case SIOCBONDSETHWADDR
:
4016 res
= bond_sethwaddr(bond_dev
, slave_dev
);
4018 case BOND_CHANGE_ACTIVE_OLD
:
4019 case SIOCBONDCHANGEACTIVE
:
4020 res
= bond_ioctl_change_active(bond_dev
, slave_dev
);
4032 static bool bond_addr_in_mc_list(unsigned char *addr
,
4033 struct netdev_hw_addr_list
*list
,
4036 struct netdev_hw_addr
*ha
;
4038 netdev_hw_addr_list_for_each(ha
, list
)
4039 if (!memcmp(ha
->addr
, addr
, addrlen
))
4045 static void bond_set_multicast_list(struct net_device
*bond_dev
)
4047 struct bonding
*bond
= netdev_priv(bond_dev
);
4048 struct netdev_hw_addr
*ha
;
4052 * Do promisc before checking multicast_mode
4054 if ((bond_dev
->flags
& IFF_PROMISC
) && !(bond
->flags
& IFF_PROMISC
))
4056 * FIXME: Need to handle the error when one of the multi-slaves
4059 bond_set_promiscuity(bond
, 1);
4062 if (!(bond_dev
->flags
& IFF_PROMISC
) && (bond
->flags
& IFF_PROMISC
))
4063 bond_set_promiscuity(bond
, -1);
4066 /* set allmulti flag to slaves */
4067 if ((bond_dev
->flags
& IFF_ALLMULTI
) && !(bond
->flags
& IFF_ALLMULTI
))
4069 * FIXME: Need to handle the error when one of the multi-slaves
4072 bond_set_allmulti(bond
, 1);
4075 if (!(bond_dev
->flags
& IFF_ALLMULTI
) && (bond
->flags
& IFF_ALLMULTI
))
4076 bond_set_allmulti(bond
, -1);
4079 read_lock(&bond
->lock
);
4081 bond
->flags
= bond_dev
->flags
;
4083 /* looking for addresses to add to slaves' mc list */
4084 netdev_for_each_mc_addr(ha
, bond_dev
) {
4085 found
= bond_addr_in_mc_list(ha
->addr
, &bond
->mc_list
,
4086 bond_dev
->addr_len
);
4088 bond_mc_add(bond
, ha
->addr
);
4091 /* looking for addresses to delete from slaves' list */
4092 netdev_hw_addr_list_for_each(ha
, &bond
->mc_list
) {
4093 found
= bond_addr_in_mc_list(ha
->addr
, &bond_dev
->mc
,
4094 bond_dev
->addr_len
);
4096 bond_mc_del(bond
, ha
->addr
);
4099 /* save master's multicast list */
4100 __hw_addr_flush(&bond
->mc_list
);
4101 __hw_addr_add_multiple(&bond
->mc_list
, &bond_dev
->mc
,
4102 bond_dev
->addr_len
, NETDEV_HW_ADDR_T_MULTICAST
);
4104 read_unlock(&bond
->lock
);
4107 static int bond_neigh_setup(struct net_device
*dev
, struct neigh_parms
*parms
)
4109 struct bonding
*bond
= netdev_priv(dev
);
4110 struct slave
*slave
= bond
->first_slave
;
4113 const struct net_device_ops
*slave_ops
4114 = slave
->dev
->netdev_ops
;
4115 if (slave_ops
->ndo_neigh_setup
)
4116 return slave_ops
->ndo_neigh_setup(slave
->dev
, parms
);
4122 * Change the MTU of all of a master's slaves to match the master
4124 static int bond_change_mtu(struct net_device
*bond_dev
, int new_mtu
)
4126 struct bonding
*bond
= netdev_priv(bond_dev
);
4127 struct slave
*slave
, *stop_at
;
4131 pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond
,
4132 (bond_dev
? bond_dev
->name
: "None"), new_mtu
);
4134 /* Can't hold bond->lock with bh disabled here since
4135 * some base drivers panic. On the other hand we can't
4136 * hold bond->lock without bh disabled because we'll
4137 * deadlock. The only solution is to rely on the fact
4138 * that we're under rtnl_lock here, and the slaves
4139 * list won't change. This doesn't solve the problem
4140 * of setting the slave's MTU while it is
4141 * transmitting, but the assumption is that the base
4142 * driver can handle that.
4144 * TODO: figure out a way to safely iterate the slaves
4145 * list, but without holding a lock around the actual
4146 * call to the base driver.
4149 bond_for_each_slave(bond
, slave
, i
) {
4150 pr_debug("s %p s->p %p c_m %p\n",
4153 slave
->dev
->netdev_ops
->ndo_change_mtu
);
4155 res
= dev_set_mtu(slave
->dev
, new_mtu
);
4158 /* If we failed to set the slave's mtu to the new value
4159 * we must abort the operation even in ACTIVE_BACKUP
4160 * mode, because if we allow the backup slaves to have
4161 * different mtu values than the active slave we'll
4162 * need to change their mtu when doing a failover. That
4163 * means changing their mtu from timer context, which
4164 * is probably not a good idea.
4166 pr_debug("err %d %s\n", res
, slave
->dev
->name
);
4171 bond_dev
->mtu
= new_mtu
;
4176 /* unwind from head to the slave that failed */
4178 bond_for_each_slave_from_to(bond
, slave
, i
, bond
->first_slave
, stop_at
) {
4181 tmp_res
= dev_set_mtu(slave
->dev
, bond_dev
->mtu
);
4183 pr_debug("unwind err %d dev %s\n",
4184 tmp_res
, slave
->dev
->name
);
4194 * Note that many devices must be down to change the HW address, and
4195 * downing the master releases all slaves. We can make bonds full of
4196 * bonding devices to test this, however.
4198 static int bond_set_mac_address(struct net_device
*bond_dev
, void *addr
)
4200 struct bonding
*bond
= netdev_priv(bond_dev
);
4201 struct sockaddr
*sa
= addr
, tmp_sa
;
4202 struct slave
*slave
, *stop_at
;
4206 if (bond
->params
.mode
== BOND_MODE_ALB
)
4207 return bond_alb_set_mac_address(bond_dev
, addr
);
4210 pr_debug("bond=%p, name=%s\n",
4211 bond
, bond_dev
? bond_dev
->name
: "None");
4214 * If fail_over_mac is set to active, do nothing and return
4215 * success. Returning an error causes ifenslave to fail.
4217 if (bond
->params
.fail_over_mac
== BOND_FOM_ACTIVE
)
4220 if (!is_valid_ether_addr(sa
->sa_data
))
4221 return -EADDRNOTAVAIL
;
4223 /* Can't hold bond->lock with bh disabled here since
4224 * some base drivers panic. On the other hand we can't
4225 * hold bond->lock without bh disabled because we'll
4226 * deadlock. The only solution is to rely on the fact
4227 * that we're under rtnl_lock here, and the slaves
4228 * list won't change. This doesn't solve the problem
4229 * of setting the slave's hw address while it is
4230 * transmitting, but the assumption is that the base
4231 * driver can handle that.
4233 * TODO: figure out a way to safely iterate the slaves
4234 * list, but without holding a lock around the actual
4235 * call to the base driver.
4238 bond_for_each_slave(bond
, slave
, i
) {
4239 const struct net_device_ops
*slave_ops
= slave
->dev
->netdev_ops
;
4240 pr_debug("slave %p %s\n", slave
, slave
->dev
->name
);
4242 if (slave_ops
->ndo_set_mac_address
== NULL
) {
4244 pr_debug("EOPNOTSUPP %s\n", slave
->dev
->name
);
4248 res
= dev_set_mac_address(slave
->dev
, addr
);
4250 /* TODO: consider downing the slave
4252 * User should expect communications
4253 * breakage anyway until ARP finish
4256 pr_debug("err %d %s\n", res
, slave
->dev
->name
);
4262 memcpy(bond_dev
->dev_addr
, sa
->sa_data
, bond_dev
->addr_len
);
4266 memcpy(tmp_sa
.sa_data
, bond_dev
->dev_addr
, bond_dev
->addr_len
);
4267 tmp_sa
.sa_family
= bond_dev
->type
;
4269 /* unwind from head to the slave that failed */
4271 bond_for_each_slave_from_to(bond
, slave
, i
, bond
->first_slave
, stop_at
) {
4274 tmp_res
= dev_set_mac_address(slave
->dev
, &tmp_sa
);
4276 pr_debug("unwind err %d dev %s\n",
4277 tmp_res
, slave
->dev
->name
);
4284 static int bond_xmit_roundrobin(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4286 struct bonding
*bond
= netdev_priv(bond_dev
);
4287 struct slave
*slave
, *start_at
;
4288 int i
, slave_no
, res
= 1;
4289 struct iphdr
*iph
= ip_hdr(skb
);
4291 read_lock(&bond
->lock
);
4293 if (!BOND_IS_OK(bond
))
4296 * Start with the curr_active_slave that joined the bond as the
4297 * default for sending IGMP traffic. For failover purposes one
4298 * needs to maintain some consistency for the interface that will
4299 * send the join/membership reports. The curr_active_slave found
4300 * will send all of this type of traffic.
4302 if ((iph
->protocol
== IPPROTO_IGMP
) &&
4303 (skb
->protocol
== htons(ETH_P_IP
))) {
4305 read_lock(&bond
->curr_slave_lock
);
4306 slave
= bond
->curr_active_slave
;
4307 read_unlock(&bond
->curr_slave_lock
);
4313 * Concurrent TX may collide on rr_tx_counter; we accept
4314 * that as being rare enough not to justify using an
4317 slave_no
= bond
->rr_tx_counter
++ % bond
->slave_cnt
;
4319 bond_for_each_slave(bond
, slave
, i
) {
4327 bond_for_each_slave_from(bond
, slave
, i
, start_at
) {
4328 if (IS_UP(slave
->dev
) &&
4329 (slave
->link
== BOND_LINK_UP
) &&
4330 (slave
->state
== BOND_STATE_ACTIVE
)) {
4331 res
= bond_dev_queue_xmit(bond
, skb
, slave
->dev
);
4338 /* no suitable interface, frame not sent */
4341 read_unlock(&bond
->lock
);
4342 return NETDEV_TX_OK
;
4347 * in active-backup mode, we know that bond->curr_active_slave is always valid if
4348 * the bond has a usable interface.
4350 static int bond_xmit_activebackup(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4352 struct bonding
*bond
= netdev_priv(bond_dev
);
4355 read_lock(&bond
->lock
);
4356 read_lock(&bond
->curr_slave_lock
);
4358 if (!BOND_IS_OK(bond
))
4361 if (!bond
->curr_active_slave
)
4364 res
= bond_dev_queue_xmit(bond
, skb
, bond
->curr_active_slave
->dev
);
4368 /* no suitable interface, frame not sent */
4371 read_unlock(&bond
->curr_slave_lock
);
4372 read_unlock(&bond
->lock
);
4373 return NETDEV_TX_OK
;
4377 * In bond_xmit_xor() , we determine the output device by using a pre-
4378 * determined xmit_hash_policy(), If the selected device is not enabled,
4379 * find the next active slave.
4381 static int bond_xmit_xor(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4383 struct bonding
*bond
= netdev_priv(bond_dev
);
4384 struct slave
*slave
, *start_at
;
4389 read_lock(&bond
->lock
);
4391 if (!BOND_IS_OK(bond
))
4394 slave_no
= bond
->xmit_hash_policy(skb
, bond
->slave_cnt
);
4396 bond_for_each_slave(bond
, slave
, i
) {
4404 bond_for_each_slave_from(bond
, slave
, i
, start_at
) {
4405 if (IS_UP(slave
->dev
) &&
4406 (slave
->link
== BOND_LINK_UP
) &&
4407 (slave
->state
== BOND_STATE_ACTIVE
)) {
4408 res
= bond_dev_queue_xmit(bond
, skb
, slave
->dev
);
4415 /* no suitable interface, frame not sent */
4418 read_unlock(&bond
->lock
);
4419 return NETDEV_TX_OK
;
4423 * in broadcast mode, we send everything to all usable interfaces.
4425 static int bond_xmit_broadcast(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4427 struct bonding
*bond
= netdev_priv(bond_dev
);
4428 struct slave
*slave
, *start_at
;
4429 struct net_device
*tx_dev
= NULL
;
4433 read_lock(&bond
->lock
);
4435 if (!BOND_IS_OK(bond
))
4438 read_lock(&bond
->curr_slave_lock
);
4439 start_at
= bond
->curr_active_slave
;
4440 read_unlock(&bond
->curr_slave_lock
);
4445 bond_for_each_slave_from(bond
, slave
, i
, start_at
) {
4446 if (IS_UP(slave
->dev
) &&
4447 (slave
->link
== BOND_LINK_UP
) &&
4448 (slave
->state
== BOND_STATE_ACTIVE
)) {
4450 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
4452 pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n",
4457 res
= bond_dev_queue_xmit(bond
, skb2
, tx_dev
);
4459 dev_kfree_skb(skb2
);
4463 tx_dev
= slave
->dev
;
4468 res
= bond_dev_queue_xmit(bond
, skb
, tx_dev
);
4472 /* no suitable interface, frame not sent */
4475 /* frame sent to all suitable interfaces */
4476 read_unlock(&bond
->lock
);
4477 return NETDEV_TX_OK
;
4480 /*------------------------- Device initialization ---------------------------*/
4482 static void bond_set_xmit_hash_policy(struct bonding
*bond
)
4484 switch (bond
->params
.xmit_policy
) {
4485 case BOND_XMIT_POLICY_LAYER23
:
4486 bond
->xmit_hash_policy
= bond_xmit_hash_policy_l23
;
4488 case BOND_XMIT_POLICY_LAYER34
:
4489 bond
->xmit_hash_policy
= bond_xmit_hash_policy_l34
;
4491 case BOND_XMIT_POLICY_LAYER2
:
4493 bond
->xmit_hash_policy
= bond_xmit_hash_policy_l2
;
4499 * Lookup the slave that corresponds to a qid
4501 static inline int bond_slave_override(struct bonding
*bond
,
4502 struct sk_buff
*skb
)
4505 struct slave
*slave
= NULL
;
4506 struct slave
*check_slave
;
4508 read_lock(&bond
->lock
);
4510 if (!BOND_IS_OK(bond
) || !skb
->queue_mapping
)
4513 /* Find out if any slaves have the same mapping as this skb. */
4514 bond_for_each_slave(bond
, check_slave
, i
) {
4515 if (check_slave
->queue_id
== skb
->queue_mapping
) {
4516 slave
= check_slave
;
4521 /* If the slave isn't UP, use default transmit policy. */
4522 if (slave
&& slave
->queue_id
&& IS_UP(slave
->dev
) &&
4523 (slave
->link
== BOND_LINK_UP
)) {
4524 res
= bond_dev_queue_xmit(bond
, skb
, slave
->dev
);
4528 read_unlock(&bond
->lock
);
4532 static u16
bond_select_queue(struct net_device
*dev
, struct sk_buff
*skb
)
4535 * This helper function exists to help dev_pick_tx get the correct
4536 * destination queue. Using a helper function skips the a call to
4537 * skb_tx_hash and will put the skbs in the queue we expect on their
4538 * way down to the bonding driver.
4540 return skb
->queue_mapping
;
4543 static netdev_tx_t
bond_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
4545 struct bonding
*bond
= netdev_priv(dev
);
4548 * If we risk deadlock from transmitting this in the
4549 * netpoll path, tell netpoll to queue the frame for later tx
4551 if (is_netpoll_tx_blocked(dev
))
4552 return NETDEV_TX_BUSY
;
4554 if (TX_QUEUE_OVERRIDE(bond
->params
.mode
)) {
4555 if (!bond_slave_override(bond
, skb
))
4556 return NETDEV_TX_OK
;
4559 switch (bond
->params
.mode
) {
4560 case BOND_MODE_ROUNDROBIN
:
4561 return bond_xmit_roundrobin(skb
, dev
);
4562 case BOND_MODE_ACTIVEBACKUP
:
4563 return bond_xmit_activebackup(skb
, dev
);
4565 return bond_xmit_xor(skb
, dev
);
4566 case BOND_MODE_BROADCAST
:
4567 return bond_xmit_broadcast(skb
, dev
);
4568 case BOND_MODE_8023AD
:
4569 return bond_3ad_xmit_xor(skb
, dev
);
4572 return bond_alb_xmit(skb
, dev
);
4574 /* Should never happen, mode already checked */
4575 pr_err("%s: Error: Unknown bonding mode %d\n",
4576 dev
->name
, bond
->params
.mode
);
4579 return NETDEV_TX_OK
;
4585 * set bond mode specific net device operations
4587 void bond_set_mode_ops(struct bonding
*bond
, int mode
)
4589 struct net_device
*bond_dev
= bond
->dev
;
4592 case BOND_MODE_ROUNDROBIN
:
4594 case BOND_MODE_ACTIVEBACKUP
:
4597 bond_set_xmit_hash_policy(bond
);
4599 case BOND_MODE_BROADCAST
:
4601 case BOND_MODE_8023AD
:
4602 bond_set_master_3ad_flags(bond
);
4603 bond_set_xmit_hash_policy(bond
);
4606 bond_set_master_alb_flags(bond
);
4611 /* Should never happen, mode already checked */
4612 pr_err("%s: Error: Unknown bonding mode %d\n",
4613 bond_dev
->name
, mode
);
4618 static void bond_ethtool_get_drvinfo(struct net_device
*bond_dev
,
4619 struct ethtool_drvinfo
*drvinfo
)
4621 strncpy(drvinfo
->driver
, DRV_NAME
, 32);
4622 strncpy(drvinfo
->version
, DRV_VERSION
, 32);
4623 snprintf(drvinfo
->fw_version
, 32, "%d", BOND_ABI_VERSION
);
4626 static const struct ethtool_ops bond_ethtool_ops
= {
4627 .get_drvinfo
= bond_ethtool_get_drvinfo
,
4628 .get_link
= ethtool_op_get_link
,
4629 .get_tx_csum
= ethtool_op_get_tx_csum
,
4630 .get_sg
= ethtool_op_get_sg
,
4631 .get_tso
= ethtool_op_get_tso
,
4632 .get_ufo
= ethtool_op_get_ufo
,
4633 .get_flags
= ethtool_op_get_flags
,
4636 static const struct net_device_ops bond_netdev_ops
= {
4637 .ndo_init
= bond_init
,
4638 .ndo_uninit
= bond_uninit
,
4639 .ndo_open
= bond_open
,
4640 .ndo_stop
= bond_close
,
4641 .ndo_start_xmit
= bond_start_xmit
,
4642 .ndo_select_queue
= bond_select_queue
,
4643 .ndo_get_stats64
= bond_get_stats
,
4644 .ndo_do_ioctl
= bond_do_ioctl
,
4645 .ndo_set_multicast_list
= bond_set_multicast_list
,
4646 .ndo_change_mtu
= bond_change_mtu
,
4647 .ndo_set_mac_address
= bond_set_mac_address
,
4648 .ndo_neigh_setup
= bond_neigh_setup
,
4649 .ndo_vlan_rx_register
= bond_vlan_rx_register
,
4650 .ndo_vlan_rx_add_vid
= bond_vlan_rx_add_vid
,
4651 .ndo_vlan_rx_kill_vid
= bond_vlan_rx_kill_vid
,
4652 #ifdef CONFIG_NET_POLL_CONTROLLER
4653 .ndo_netpoll_cleanup
= bond_netpoll_cleanup
,
4654 .ndo_poll_controller
= bond_poll_controller
,
4658 static void bond_destructor(struct net_device
*bond_dev
)
4660 struct bonding
*bond
= netdev_priv(bond_dev
);
4662 destroy_workqueue(bond
->wq
);
4663 free_netdev(bond_dev
);
4666 static void bond_setup(struct net_device
*bond_dev
)
4668 struct bonding
*bond
= netdev_priv(bond_dev
);
4670 /* initialize rwlocks */
4671 rwlock_init(&bond
->lock
);
4672 rwlock_init(&bond
->curr_slave_lock
);
4674 bond
->params
= bonding_defaults
;
4676 /* Initialize pointers */
4677 bond
->dev
= bond_dev
;
4678 INIT_LIST_HEAD(&bond
->vlan_list
);
4680 /* Initialize the device entry points */
4681 ether_setup(bond_dev
);
4682 bond_dev
->netdev_ops
= &bond_netdev_ops
;
4683 bond_dev
->ethtool_ops
= &bond_ethtool_ops
;
4684 bond_set_mode_ops(bond
, bond
->params
.mode
);
4686 bond_dev
->destructor
= bond_destructor
;
4688 /* Initialize the device options */
4689 bond_dev
->tx_queue_len
= 0;
4690 bond_dev
->flags
|= IFF_MASTER
|IFF_MULTICAST
;
4691 bond_dev
->priv_flags
|= IFF_BONDING
;
4692 bond_dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
4694 if (bond
->params
.arp_interval
)
4695 bond_dev
->priv_flags
|= IFF_MASTER_ARPMON
;
4697 /* At first, we block adding VLANs. That's the only way to
4698 * prevent problems that occur when adding VLANs over an
4699 * empty bond. The block will be removed once non-challenged
4700 * slaves are enslaved.
4702 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
4704 /* don't acquire bond device's netif_tx_lock when
4706 bond_dev
->features
|= NETIF_F_LLTX
;
4708 /* By default, we declare the bond to be fully
4709 * VLAN hardware accelerated capable. Special
4710 * care is taken in the various xmit functions
4711 * when there are slaves that are not hw accel
4714 bond_dev
->features
|= (NETIF_F_HW_VLAN_TX
|
4715 NETIF_F_HW_VLAN_RX
|
4716 NETIF_F_HW_VLAN_FILTER
);
4718 /* By default, we enable GRO on bonding devices.
4719 * Actual support requires lowlevel drivers are GRO ready.
4721 bond_dev
->features
|= NETIF_F_GRO
;
4724 static void bond_work_cancel_all(struct bonding
*bond
)
4726 write_lock_bh(&bond
->lock
);
4727 bond
->kill_timers
= 1;
4728 write_unlock_bh(&bond
->lock
);
4730 if (bond
->params
.miimon
&& delayed_work_pending(&bond
->mii_work
))
4731 cancel_delayed_work(&bond
->mii_work
);
4733 if (bond
->params
.arp_interval
&& delayed_work_pending(&bond
->arp_work
))
4734 cancel_delayed_work(&bond
->arp_work
);
4736 if (bond
->params
.mode
== BOND_MODE_ALB
&&
4737 delayed_work_pending(&bond
->alb_work
))
4738 cancel_delayed_work(&bond
->alb_work
);
4740 if (bond
->params
.mode
== BOND_MODE_8023AD
&&
4741 delayed_work_pending(&bond
->ad_work
))
4742 cancel_delayed_work(&bond
->ad_work
);
4744 if (delayed_work_pending(&bond
->mcast_work
))
4745 cancel_delayed_work(&bond
->mcast_work
);
4749 * Destroy a bonding device.
4750 * Must be under rtnl_lock when this function is called.
4752 static void bond_uninit(struct net_device
*bond_dev
)
4754 struct bonding
*bond
= netdev_priv(bond_dev
);
4755 struct vlan_entry
*vlan
, *tmp
;
4757 bond_netpoll_cleanup(bond_dev
);
4759 /* Release the bonded slaves */
4760 bond_release_all(bond_dev
);
4762 list_del(&bond
->bond_list
);
4764 bond_work_cancel_all(bond
);
4766 bond_remove_proc_entry(bond
);
4768 bond_debug_unregister(bond
);
4770 __hw_addr_flush(&bond
->mc_list
);
4772 list_for_each_entry_safe(vlan
, tmp
, &bond
->vlan_list
, vlan_list
) {
4773 list_del(&vlan
->vlan_list
);
4778 /*------------------------- Module initialization ---------------------------*/
4781 * Convert string input module parms. Accept either the
4782 * number of the mode or its string name. A bit complicated because
4783 * some mode names are substrings of other names, and calls from sysfs
4784 * may have whitespace in the name (trailing newlines, for example).
4786 int bond_parse_parm(const char *buf
, const struct bond_parm_tbl
*tbl
)
4788 int modeint
= -1, i
, rv
;
4789 char *p
, modestr
[BOND_MAX_MODENAME_LEN
+ 1] = { 0, };
4791 for (p
= (char *)buf
; *p
; p
++)
4792 if (!(isdigit(*p
) || isspace(*p
)))
4796 rv
= sscanf(buf
, "%20s", modestr
);
4798 rv
= sscanf(buf
, "%d", &modeint
);
4803 for (i
= 0; tbl
[i
].modename
; i
++) {
4804 if (modeint
== tbl
[i
].mode
)
4806 if (strcmp(modestr
, tbl
[i
].modename
) == 0)
4813 static int bond_check_params(struct bond_params
*params
)
4815 int arp_validate_value
, fail_over_mac_value
, primary_reselect_value
;
4818 * Convert string parameters.
4821 bond_mode
= bond_parse_parm(mode
, bond_mode_tbl
);
4822 if (bond_mode
== -1) {
4823 pr_err("Error: Invalid bonding mode \"%s\"\n",
4824 mode
== NULL
? "NULL" : mode
);
4829 if (xmit_hash_policy
) {
4830 if ((bond_mode
!= BOND_MODE_XOR
) &&
4831 (bond_mode
!= BOND_MODE_8023AD
)) {
4832 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
4833 bond_mode_name(bond_mode
));
4835 xmit_hashtype
= bond_parse_parm(xmit_hash_policy
,
4837 if (xmit_hashtype
== -1) {
4838 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
4839 xmit_hash_policy
== NULL
? "NULL" :
4847 if (bond_mode
!= BOND_MODE_8023AD
) {
4848 pr_info("lacp_rate param is irrelevant in mode %s\n",
4849 bond_mode_name(bond_mode
));
4851 lacp_fast
= bond_parse_parm(lacp_rate
, bond_lacp_tbl
);
4852 if (lacp_fast
== -1) {
4853 pr_err("Error: Invalid lacp rate \"%s\"\n",
4854 lacp_rate
== NULL
? "NULL" : lacp_rate
);
4861 params
->ad_select
= bond_parse_parm(ad_select
, ad_select_tbl
);
4862 if (params
->ad_select
== -1) {
4863 pr_err("Error: Invalid ad_select \"%s\"\n",
4864 ad_select
== NULL
? "NULL" : ad_select
);
4868 if (bond_mode
!= BOND_MODE_8023AD
) {
4869 pr_warning("ad_select param only affects 802.3ad mode\n");
4872 params
->ad_select
= BOND_AD_STABLE
;
4875 if (max_bonds
< 0) {
4876 pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4877 max_bonds
, 0, INT_MAX
, BOND_DEFAULT_MAX_BONDS
);
4878 max_bonds
= BOND_DEFAULT_MAX_BONDS
;
4882 pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n",
4883 miimon
, INT_MAX
, BOND_LINK_MON_INTERV
);
4884 miimon
= BOND_LINK_MON_INTERV
;
4888 pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4893 if (downdelay
< 0) {
4894 pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4895 downdelay
, INT_MAX
);
4899 if ((use_carrier
!= 0) && (use_carrier
!= 1)) {
4900 pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4905 if (num_grat_arp
< 0 || num_grat_arp
> 255) {
4906 pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n",
4911 if (num_unsol_na
< 0 || num_unsol_na
> 255) {
4912 pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4917 /* reset values for 802.3ad */
4918 if (bond_mode
== BOND_MODE_8023AD
) {
4920 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");
4921 pr_warning("Forcing miimon to 100msec\n");
4926 if (tx_queues
< 1 || tx_queues
> 255) {
4927 pr_warning("Warning: tx_queues (%d) should be between "
4928 "1 and 255, resetting to %d\n",
4929 tx_queues
, BOND_DEFAULT_TX_QUEUES
);
4930 tx_queues
= BOND_DEFAULT_TX_QUEUES
;
4933 if ((all_slaves_active
!= 0) && (all_slaves_active
!= 1)) {
4934 pr_warning("Warning: all_slaves_active module parameter (%d), "
4935 "not of valid value (0/1), so it was set to "
4936 "0\n", all_slaves_active
);
4937 all_slaves_active
= 0;
4940 if (resend_igmp
< 0 || resend_igmp
> 255) {
4941 pr_warning("Warning: resend_igmp (%d) should be between "
4942 "0 and 255, resetting to %d\n",
4943 resend_igmp
, BOND_DEFAULT_RESEND_IGMP
);
4944 resend_igmp
= BOND_DEFAULT_RESEND_IGMP
;
4947 /* reset values for TLB/ALB */
4948 if ((bond_mode
== BOND_MODE_TLB
) ||
4949 (bond_mode
== BOND_MODE_ALB
)) {
4951 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");
4952 pr_warning("Forcing miimon to 100msec\n");
4957 if (bond_mode
== BOND_MODE_ALB
) {
4958 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",
4963 if (updelay
|| downdelay
) {
4964 /* just warn the user the up/down delay will have
4965 * no effect since miimon is zero...
4967 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",
4968 updelay
, downdelay
);
4971 /* don't allow arp monitoring */
4973 pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4974 miimon
, arp_interval
);
4978 if ((updelay
% miimon
) != 0) {
4979 pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4981 (updelay
/ miimon
) * miimon
);
4986 if ((downdelay
% miimon
) != 0) {
4987 pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4989 (downdelay
/ miimon
) * miimon
);
4992 downdelay
/= miimon
;
4995 if (arp_interval
< 0) {
4996 pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n",
4997 arp_interval
, INT_MAX
, BOND_LINK_ARP_INTERV
);
4998 arp_interval
= BOND_LINK_ARP_INTERV
;
5001 for (arp_ip_count
= 0;
5002 (arp_ip_count
< BOND_MAX_ARP_TARGETS
) && arp_ip_target
[arp_ip_count
];
5004 /* not complete check, but should be good enough to
5006 if (!isdigit(arp_ip_target
[arp_ip_count
][0])) {
5007 pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
5008 arp_ip_target
[arp_ip_count
]);
5011 __be32 ip
= in_aton(arp_ip_target
[arp_ip_count
]);
5012 arp_target
[arp_ip_count
] = ip
;
5016 if (arp_interval
&& !arp_ip_count
) {
5017 /* don't allow arping if no arp_ip_target given... */
5018 pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
5024 if (bond_mode
!= BOND_MODE_ACTIVEBACKUP
) {
5025 pr_err("arp_validate only supported in active-backup mode\n");
5028 if (!arp_interval
) {
5029 pr_err("arp_validate requires arp_interval\n");
5033 arp_validate_value
= bond_parse_parm(arp_validate
,
5035 if (arp_validate_value
== -1) {
5036 pr_err("Error: invalid arp_validate \"%s\"\n",
5037 arp_validate
== NULL
? "NULL" : arp_validate
);
5041 arp_validate_value
= 0;
5044 pr_info("MII link monitoring set to %d ms\n", miimon
);
5045 } else if (arp_interval
) {
5048 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
5050 arp_validate_tbl
[arp_validate_value
].modename
,
5053 for (i
= 0; i
< arp_ip_count
; i
++)
5054 pr_info(" %s", arp_ip_target
[i
]);
5058 } else if (max_bonds
) {
5059 /* miimon and arp_interval not set, we need one so things
5060 * work as expected, see bonding.txt for details
5062 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");
5065 if (primary
&& !USES_PRIMARY(bond_mode
)) {
5066 /* currently, using a primary only makes sense
5067 * in active backup, TLB or ALB modes
5069 pr_warning("Warning: %s primary device specified but has no effect in %s mode\n",
5070 primary
, bond_mode_name(bond_mode
));
5074 if (primary
&& primary_reselect
) {
5075 primary_reselect_value
= bond_parse_parm(primary_reselect
,
5077 if (primary_reselect_value
== -1) {
5078 pr_err("Error: Invalid primary_reselect \"%s\"\n",
5080 NULL
? "NULL" : primary_reselect
);
5084 primary_reselect_value
= BOND_PRI_RESELECT_ALWAYS
;
5087 if (fail_over_mac
) {
5088 fail_over_mac_value
= bond_parse_parm(fail_over_mac
,
5090 if (fail_over_mac_value
== -1) {
5091 pr_err("Error: invalid fail_over_mac \"%s\"\n",
5092 arp_validate
== NULL
? "NULL" : arp_validate
);
5096 if (bond_mode
!= BOND_MODE_ACTIVEBACKUP
)
5097 pr_warning("Warning: fail_over_mac only affects active-backup mode.\n");
5099 fail_over_mac_value
= BOND_FOM_NONE
;
5102 /* fill params struct with the proper values */
5103 params
->mode
= bond_mode
;
5104 params
->xmit_policy
= xmit_hashtype
;
5105 params
->miimon
= miimon
;
5106 params
->num_grat_arp
= num_grat_arp
;
5107 params
->num_unsol_na
= num_unsol_na
;
5108 params
->arp_interval
= arp_interval
;
5109 params
->arp_validate
= arp_validate_value
;
5110 params
->updelay
= updelay
;
5111 params
->downdelay
= downdelay
;
5112 params
->use_carrier
= use_carrier
;
5113 params
->lacp_fast
= lacp_fast
;
5114 params
->primary
[0] = 0;
5115 params
->primary_reselect
= primary_reselect_value
;
5116 params
->fail_over_mac
= fail_over_mac_value
;
5117 params
->tx_queues
= tx_queues
;
5118 params
->all_slaves_active
= all_slaves_active
;
5119 params
->resend_igmp
= resend_igmp
;
5122 strncpy(params
->primary
, primary
, IFNAMSIZ
);
5123 params
->primary
[IFNAMSIZ
- 1] = 0;
5126 memcpy(params
->arp_targets
, arp_target
, sizeof(arp_target
));
5131 static struct lock_class_key bonding_netdev_xmit_lock_key
;
5132 static struct lock_class_key bonding_netdev_addr_lock_key
;
5134 static void bond_set_lockdep_class_one(struct net_device
*dev
,
5135 struct netdev_queue
*txq
,
5138 lockdep_set_class(&txq
->_xmit_lock
,
5139 &bonding_netdev_xmit_lock_key
);
5142 static void bond_set_lockdep_class(struct net_device
*dev
)
5144 lockdep_set_class(&dev
->addr_list_lock
,
5145 &bonding_netdev_addr_lock_key
);
5146 netdev_for_each_tx_queue(dev
, bond_set_lockdep_class_one
, NULL
);
5150 * Called from registration process
5152 static int bond_init(struct net_device
*bond_dev
)
5154 struct bonding
*bond
= netdev_priv(bond_dev
);
5155 struct bond_net
*bn
= net_generic(dev_net(bond_dev
), bond_net_id
);
5157 pr_debug("Begin bond_init for %s\n", bond_dev
->name
);
5159 bond
->wq
= create_singlethread_workqueue(bond_dev
->name
);
5163 bond_set_lockdep_class(bond_dev
);
5165 netif_carrier_off(bond_dev
);
5167 bond_create_proc_entry(bond
);
5168 list_add_tail(&bond
->bond_list
, &bn
->dev_list
);
5170 bond_prepare_sysfs_group(bond
);
5172 bond_debug_register(bond
);
5174 __hw_addr_init(&bond
->mc_list
);
5178 static int bond_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
5180 if (tb
[IFLA_ADDRESS
]) {
5181 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
)
5183 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
])))
5184 return -EADDRNOTAVAIL
;
5189 static struct rtnl_link_ops bond_link_ops __read_mostly
= {
5191 .priv_size
= sizeof(struct bonding
),
5192 .setup
= bond_setup
,
5193 .validate
= bond_validate
,
5196 /* Create a new bond based on the specified name and bonding parameters.
5197 * If name is NULL, obtain a suitable "bond%d" name for us.
5198 * Caller must NOT hold rtnl_lock; we need to release it here before we
5199 * set up our sysfs entries.
5201 int bond_create(struct net
*net
, const char *name
)
5203 struct net_device
*bond_dev
;
5208 bond_dev
= alloc_netdev_mq(sizeof(struct bonding
), name
? name
: "",
5209 bond_setup
, tx_queues
);
5211 pr_err("%s: eek! can't alloc netdev!\n", name
);
5216 dev_net_set(bond_dev
, net
);
5217 bond_dev
->rtnl_link_ops
= &bond_link_ops
;
5220 res
= dev_alloc_name(bond_dev
, "bond%d");
5225 * If we're given a name to register
5226 * we need to ensure that its not already
5230 if (__dev_get_by_name(net
, name
) != NULL
)
5234 res
= register_netdevice(bond_dev
);
5239 bond_destructor(bond_dev
);
5243 static int __net_init
bond_net_init(struct net
*net
)
5245 struct bond_net
*bn
= net_generic(net
, bond_net_id
);
5248 INIT_LIST_HEAD(&bn
->dev_list
);
5250 bond_create_proc_dir(bn
);
5255 static void __net_exit
bond_net_exit(struct net
*net
)
5257 struct bond_net
*bn
= net_generic(net
, bond_net_id
);
5259 bond_destroy_proc_dir(bn
);
5262 static struct pernet_operations bond_net_ops
= {
5263 .init
= bond_net_init
,
5264 .exit
= bond_net_exit
,
5266 .size
= sizeof(struct bond_net
),
5269 static int __init
bonding_init(void)
5274 pr_info("%s", version
);
5276 res
= bond_check_params(&bonding_defaults
);
5280 res
= register_pernet_subsys(&bond_net_ops
);
5284 res
= rtnl_link_register(&bond_link_ops
);
5288 bond_create_debugfs();
5290 for (i
= 0; i
< max_bonds
; i
++) {
5291 res
= bond_create(&init_net
, NULL
);
5296 res
= bond_create_sysfs();
5300 register_netdevice_notifier(&bond_netdev_notifier
);
5301 register_inetaddr_notifier(&bond_inetaddr_notifier
);
5302 bond_register_ipv6_notifier();
5306 rtnl_link_unregister(&bond_link_ops
);
5308 unregister_pernet_subsys(&bond_net_ops
);
5313 static void __exit
bonding_exit(void)
5315 unregister_netdevice_notifier(&bond_netdev_notifier
);
5316 unregister_inetaddr_notifier(&bond_inetaddr_notifier
);
5317 bond_unregister_ipv6_notifier();
5319 bond_destroy_sysfs();
5320 bond_destroy_debugfs();
5322 rtnl_link_unregister(&bond_link_ops
);
5323 unregister_pernet_subsys(&bond_net_ops
);
5325 #ifdef CONFIG_NET_POLL_CONTROLLER
5327 * Make sure we don't have an imbalance on our netpoll blocking
5329 WARN_ON(atomic_read(&netpoll_block_tx
));
5333 module_init(bonding_init
);
5334 module_exit(bonding_exit
);
5335 MODULE_LICENSE("GPL");
5336 MODULE_VERSION(DRV_VERSION
);
5337 MODULE_DESCRIPTION(DRV_DESCRIPTION
", v" DRV_VERSION
);
5338 MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
5339 MODULE_ALIAS_RTNL_LINK("bond");