2 * originally based on the dummy device.
4 * Copyright 1999, Thomas Davis, tadavis@lbl.gov.
5 * Licensed under the GPL. Based on dummy.c, and eql.c devices.
7 * bonding.c: an Ethernet Bonding driver
9 * This is useful to talk to a Cisco EtherChannel compatible equipment:
11 * Sun Trunking (Solaris)
12 * Alteon AceDirector Trunks
14 * and probably many L2 switches ...
17 * ifconfig bond0 ipaddress netmask up
18 * will setup a network device, with an ip address. No mac address
19 * will be assigned at this time. The hw mac address will come from
20 * the first slave bonded to the channel. All slaves will then use
21 * this hw mac address.
24 * will release all slaves, marking them as down.
26 * ifenslave bond0 eth0
27 * will attach eth0 to bond0 as a slave. eth0 hw mac address will either
28 * a: be used as initial mac address
29 * b: if a hw mac address already is there, eth0's hw mac address
30 * will then be set from bond0.
34 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
36 #include <linux/kernel.h>
37 #include <linux/module.h>
38 #include <linux/types.h>
39 #include <linux/fcntl.h>
40 #include <linux/interrupt.h>
41 #include <linux/ptrace.h>
42 #include <linux/ioport.h>
46 #include <linux/tcp.h>
47 #include <linux/udp.h>
48 #include <linux/slab.h>
49 #include <linux/string.h>
50 #include <linux/init.h>
51 #include <linux/timer.h>
52 #include <linux/socket.h>
53 #include <linux/ctype.h>
54 #include <linux/inet.h>
55 #include <linux/bitops.h>
57 #include <asm/system.h>
59 #include <linux/uaccess.h>
60 #include <linux/errno.h>
61 #include <linux/netdevice.h>
62 #include <linux/inetdevice.h>
63 #include <linux/igmp.h>
64 #include <linux/etherdevice.h>
65 #include <linux/skbuff.h>
67 #include <linux/rtnetlink.h>
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70 #include <linux/smp.h>
71 #include <linux/if_ether.h>
73 #include <linux/mii.h>
74 #include <linux/ethtool.h>
75 #include <linux/if_vlan.h>
76 #include <linux/if_bonding.h>
77 #include <linux/jiffies.h>
78 #include <net/route.h>
79 #include <net/net_namespace.h>
80 #include <net/netns/generic.h>
85 /*---------------------------- Module parameters ----------------------------*/
87 /* monitor all links that often (in milliseconds). <=0 disables monitoring */
88 #define BOND_LINK_MON_INTERV 0
89 #define BOND_LINK_ARP_INTERV 0
91 static int max_bonds
= BOND_DEFAULT_MAX_BONDS
;
92 static int num_grat_arp
= 1;
93 static int num_unsol_na
= 1;
94 static int miimon
= BOND_LINK_MON_INTERV
;
97 static int use_carrier
= 1;
100 static char *primary_reselect
;
101 static char *lacp_rate
;
102 static char *ad_select
;
103 static char *xmit_hash_policy
;
104 static int arp_interval
= BOND_LINK_ARP_INTERV
;
105 static char *arp_ip_target
[BOND_MAX_ARP_TARGETS
];
106 static char *arp_validate
;
107 static char *fail_over_mac
;
108 static struct bond_params bonding_defaults
;
110 module_param(max_bonds
, int, 0);
111 MODULE_PARM_DESC(max_bonds
, "Max number of bonded devices");
112 module_param(num_grat_arp
, int, 0644);
113 MODULE_PARM_DESC(num_grat_arp
, "Number of gratuitous ARP packets to send on failover event");
114 module_param(num_unsol_na
, int, 0644);
115 MODULE_PARM_DESC(num_unsol_na
, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event");
116 module_param(miimon
, int, 0);
117 MODULE_PARM_DESC(miimon
, "Link check interval in milliseconds");
118 module_param(updelay
, int, 0);
119 MODULE_PARM_DESC(updelay
, "Delay before considering link up, in milliseconds");
120 module_param(downdelay
, int, 0);
121 MODULE_PARM_DESC(downdelay
, "Delay before considering link down, "
123 module_param(use_carrier
, int, 0);
124 MODULE_PARM_DESC(use_carrier
, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
125 "0 for off, 1 for on (default)");
126 module_param(mode
, charp
, 0);
127 MODULE_PARM_DESC(mode
, "Mode of operation : 0 for balance-rr, "
128 "1 for active-backup, 2 for balance-xor, "
129 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
130 "6 for balance-alb");
131 module_param(primary
, charp
, 0);
132 MODULE_PARM_DESC(primary
, "Primary network device to use");
133 module_param(primary_reselect
, charp
, 0);
134 MODULE_PARM_DESC(primary_reselect
, "Reselect primary slave "
136 "0 for always (default), "
137 "1 for only if speed of primary is "
139 "2 for only on active slave "
141 module_param(lacp_rate
, charp
, 0);
142 MODULE_PARM_DESC(lacp_rate
, "LACPDU tx rate to request from 802.3ad partner "
144 module_param(ad_select
, charp
, 0);
145 MODULE_PARM_DESC(ad_select
, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)");
146 module_param(xmit_hash_policy
, charp
, 0);
147 MODULE_PARM_DESC(xmit_hash_policy
, "XOR hashing method: 0 for layer 2 (default)"
148 ", 1 for layer 3+4");
149 module_param(arp_interval
, int, 0);
150 MODULE_PARM_DESC(arp_interval
, "arp interval in milliseconds");
151 module_param_array(arp_ip_target
, charp
, NULL
, 0);
152 MODULE_PARM_DESC(arp_ip_target
, "arp targets in n.n.n.n form");
153 module_param(arp_validate
, charp
, 0);
154 MODULE_PARM_DESC(arp_validate
, "validate src/dst of ARP probes: none (default), active, backup or all");
155 module_param(fail_over_mac
, charp
, 0);
156 MODULE_PARM_DESC(fail_over_mac
, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow");
158 /*----------------------------- Global variables ----------------------------*/
160 static const char * const version
=
161 DRV_DESCRIPTION
": v" DRV_VERSION
" (" DRV_RELDATE
")\n";
163 int bond_net_id __read_mostly
;
165 static __be32 arp_target
[BOND_MAX_ARP_TARGETS
];
166 static int arp_ip_count
;
167 static int bond_mode
= BOND_MODE_ROUNDROBIN
;
168 static int xmit_hashtype
= BOND_XMIT_POLICY_LAYER2
;
169 static int lacp_fast
;
172 const struct bond_parm_tbl bond_lacp_tbl
[] = {
173 { "slow", AD_LACP_SLOW
},
174 { "fast", AD_LACP_FAST
},
178 const struct bond_parm_tbl bond_mode_tbl
[] = {
179 { "balance-rr", BOND_MODE_ROUNDROBIN
},
180 { "active-backup", BOND_MODE_ACTIVEBACKUP
},
181 { "balance-xor", BOND_MODE_XOR
},
182 { "broadcast", BOND_MODE_BROADCAST
},
183 { "802.3ad", BOND_MODE_8023AD
},
184 { "balance-tlb", BOND_MODE_TLB
},
185 { "balance-alb", BOND_MODE_ALB
},
189 const struct bond_parm_tbl xmit_hashtype_tbl
[] = {
190 { "layer2", BOND_XMIT_POLICY_LAYER2
},
191 { "layer3+4", BOND_XMIT_POLICY_LAYER34
},
192 { "layer2+3", BOND_XMIT_POLICY_LAYER23
},
196 const struct bond_parm_tbl arp_validate_tbl
[] = {
197 { "none", BOND_ARP_VALIDATE_NONE
},
198 { "active", BOND_ARP_VALIDATE_ACTIVE
},
199 { "backup", BOND_ARP_VALIDATE_BACKUP
},
200 { "all", BOND_ARP_VALIDATE_ALL
},
204 const struct bond_parm_tbl fail_over_mac_tbl
[] = {
205 { "none", BOND_FOM_NONE
},
206 { "active", BOND_FOM_ACTIVE
},
207 { "follow", BOND_FOM_FOLLOW
},
211 const struct bond_parm_tbl pri_reselect_tbl
[] = {
212 { "always", BOND_PRI_RESELECT_ALWAYS
},
213 { "better", BOND_PRI_RESELECT_BETTER
},
214 { "failure", BOND_PRI_RESELECT_FAILURE
},
218 struct bond_parm_tbl ad_select_tbl
[] = {
219 { "stable", BOND_AD_STABLE
},
220 { "bandwidth", BOND_AD_BANDWIDTH
},
221 { "count", BOND_AD_COUNT
},
225 /*-------------------------- Forward declarations ---------------------------*/
227 static void bond_send_gratuitous_arp(struct bonding
*bond
);
228 static int bond_init(struct net_device
*bond_dev
);
229 static void bond_uninit(struct net_device
*bond_dev
);
231 /*---------------------------- General routines -----------------------------*/
233 static const char *bond_mode_name(int mode
)
235 static const char *names
[] = {
236 [BOND_MODE_ROUNDROBIN
] = "load balancing (round-robin)",
237 [BOND_MODE_ACTIVEBACKUP
] = "fault-tolerance (active-backup)",
238 [BOND_MODE_XOR
] = "load balancing (xor)",
239 [BOND_MODE_BROADCAST
] = "fault-tolerance (broadcast)",
240 [BOND_MODE_8023AD
] = "IEEE 802.3ad Dynamic link aggregation",
241 [BOND_MODE_TLB
] = "transmit load balancing",
242 [BOND_MODE_ALB
] = "adaptive load balancing",
245 if (mode
< 0 || mode
> BOND_MODE_ALB
)
251 /*---------------------------------- VLAN -----------------------------------*/
254 * bond_add_vlan - add a new vlan id on bond
255 * @bond: bond that got the notification
256 * @vlan_id: the vlan id to add
258 * Returns -ENOMEM if allocation failed.
260 static int bond_add_vlan(struct bonding
*bond
, unsigned short vlan_id
)
262 struct vlan_entry
*vlan
;
264 pr_debug("bond: %s, vlan id %d\n",
265 (bond
? bond
->dev
->name
: "None"), vlan_id
);
267 vlan
= kzalloc(sizeof(struct vlan_entry
), GFP_KERNEL
);
271 INIT_LIST_HEAD(&vlan
->vlan_list
);
272 vlan
->vlan_id
= vlan_id
;
274 write_lock_bh(&bond
->lock
);
276 list_add_tail(&vlan
->vlan_list
, &bond
->vlan_list
);
278 write_unlock_bh(&bond
->lock
);
280 pr_debug("added VLAN ID %d on bond %s\n", vlan_id
, bond
->dev
->name
);
286 * bond_del_vlan - delete a vlan id from bond
287 * @bond: bond that got the notification
288 * @vlan_id: the vlan id to delete
290 * returns -ENODEV if @vlan_id was not found in @bond.
292 static int bond_del_vlan(struct bonding
*bond
, unsigned short vlan_id
)
294 struct vlan_entry
*vlan
;
297 pr_debug("bond: %s, vlan id %d\n", bond
->dev
->name
, vlan_id
);
299 write_lock_bh(&bond
->lock
);
301 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
302 if (vlan
->vlan_id
== vlan_id
) {
303 list_del(&vlan
->vlan_list
);
305 if (bond_is_lb(bond
))
306 bond_alb_clear_vlan(bond
, vlan_id
);
308 pr_debug("removed VLAN ID %d from bond %s\n",
309 vlan_id
, bond
->dev
->name
);
313 if (list_empty(&bond
->vlan_list
) &&
314 (bond
->slave_cnt
== 0)) {
315 /* Last VLAN removed and no slaves, so
316 * restore block on adding VLANs. This will
317 * be removed once new slaves that are not
318 * VLAN challenged will be added.
320 bond
->dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
328 pr_debug("couldn't find VLAN ID %d in bond %s\n",
329 vlan_id
, bond
->dev
->name
);
332 write_unlock_bh(&bond
->lock
);
337 * bond_has_challenged_slaves
338 * @bond: the bond we're working on
340 * Searches the slave list. Returns 1 if a vlan challenged slave
341 * was found, 0 otherwise.
343 * Assumes bond->lock is held.
345 static int bond_has_challenged_slaves(struct bonding
*bond
)
350 bond_for_each_slave(bond
, slave
, i
) {
351 if (slave
->dev
->features
& NETIF_F_VLAN_CHALLENGED
) {
352 pr_debug("found VLAN challenged slave - %s\n",
358 pr_debug("no VLAN challenged slaves found\n");
363 * bond_next_vlan - safely skip to the next item in the vlans list.
364 * @bond: the bond we're working on
365 * @curr: item we're advancing from
367 * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL,
368 * or @curr->next otherwise (even if it is @curr itself again).
370 * Caller must hold bond->lock
372 struct vlan_entry
*bond_next_vlan(struct bonding
*bond
, struct vlan_entry
*curr
)
374 struct vlan_entry
*next
, *last
;
376 if (list_empty(&bond
->vlan_list
))
380 next
= list_entry(bond
->vlan_list
.next
,
381 struct vlan_entry
, vlan_list
);
383 last
= list_entry(bond
->vlan_list
.prev
,
384 struct vlan_entry
, vlan_list
);
386 next
= list_entry(bond
->vlan_list
.next
,
387 struct vlan_entry
, vlan_list
);
389 next
= list_entry(curr
->vlan_list
.next
,
390 struct vlan_entry
, vlan_list
);
398 * bond_dev_queue_xmit - Prepare skb for xmit.
400 * @bond: bond device that got this skb for tx.
401 * @skb: hw accel VLAN tagged skb to transmit
402 * @slave_dev: slave that is supposed to xmit this skbuff
404 * When the bond gets an skb to transmit that is
405 * already hardware accelerated VLAN tagged, and it
406 * needs to relay this skb to a slave that is not
407 * hw accel capable, the skb needs to be "unaccelerated",
408 * i.e. strip the hwaccel tag and re-insert it as part
411 int bond_dev_queue_xmit(struct bonding
*bond
, struct sk_buff
*skb
,
412 struct net_device
*slave_dev
)
414 unsigned short uninitialized_var(vlan_id
);
416 if (!list_empty(&bond
->vlan_list
) &&
417 !(slave_dev
->features
& NETIF_F_HW_VLAN_TX
) &&
418 vlan_get_tag(skb
, &vlan_id
) == 0) {
419 skb
->dev
= slave_dev
;
420 skb
= vlan_put_tag(skb
, vlan_id
);
422 /* vlan_put_tag() frees the skb in case of error,
423 * so return success here so the calling functions
424 * won't attempt to free is again.
429 skb
->dev
= slave_dev
;
439 * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid
440 * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a
442 * a. This operation is performed in IOCTL context,
443 * b. The operation is protected by the RTNL semaphore in the 8021q code,
444 * c. Holding a lock with BH disabled while directly calling a base driver
445 * entry point is generally a BAD idea.
447 * The design of synchronization/protection for this operation in the 8021q
448 * module is good for one or more VLAN devices over a single physical device
449 * and cannot be extended for a teaming solution like bonding, so there is a
450 * potential race condition here where a net device from the vlan group might
451 * be referenced (either by a base driver or the 8021q code) while it is being
452 * removed from the system. However, it turns out we're not making matters
453 * worse, and if it works for regular VLAN usage it will work here too.
457 * bond_vlan_rx_register - Propagates registration to slaves
458 * @bond_dev: bonding net device that got called
459 * @grp: vlan group being registered
461 static void bond_vlan_rx_register(struct net_device
*bond_dev
,
462 struct vlan_group
*grp
)
464 struct bonding
*bond
= netdev_priv(bond_dev
);
470 bond_for_each_slave(bond
, slave
, i
) {
471 struct net_device
*slave_dev
= slave
->dev
;
472 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
474 if ((slave_dev
->features
& NETIF_F_HW_VLAN_RX
) &&
475 slave_ops
->ndo_vlan_rx_register
) {
476 slave_ops
->ndo_vlan_rx_register(slave_dev
, grp
);
482 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
483 * @bond_dev: bonding net device that got called
484 * @vid: vlan id being added
486 static void bond_vlan_rx_add_vid(struct net_device
*bond_dev
, uint16_t vid
)
488 struct bonding
*bond
= netdev_priv(bond_dev
);
492 bond_for_each_slave(bond
, slave
, i
) {
493 struct net_device
*slave_dev
= slave
->dev
;
494 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
496 if ((slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) &&
497 slave_ops
->ndo_vlan_rx_add_vid
) {
498 slave_ops
->ndo_vlan_rx_add_vid(slave_dev
, vid
);
502 res
= bond_add_vlan(bond
, vid
);
504 pr_err("%s: Error: Failed to add vlan id %d\n",
505 bond_dev
->name
, vid
);
510 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
511 * @bond_dev: bonding net device that got called
512 * @vid: vlan id being removed
514 static void bond_vlan_rx_kill_vid(struct net_device
*bond_dev
, uint16_t vid
)
516 struct bonding
*bond
= netdev_priv(bond_dev
);
518 struct net_device
*vlan_dev
;
521 bond_for_each_slave(bond
, slave
, i
) {
522 struct net_device
*slave_dev
= slave
->dev
;
523 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
525 if ((slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) &&
526 slave_ops
->ndo_vlan_rx_kill_vid
) {
527 /* Save and then restore vlan_dev in the grp array,
528 * since the slave's driver might clear it.
530 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vid
);
531 slave_ops
->ndo_vlan_rx_kill_vid(slave_dev
, vid
);
532 vlan_group_set_device(bond
->vlgrp
, vid
, vlan_dev
);
536 res
= bond_del_vlan(bond
, vid
);
538 pr_err("%s: Error: Failed to remove vlan id %d\n",
539 bond_dev
->name
, vid
);
543 static void bond_add_vlans_on_slave(struct bonding
*bond
, struct net_device
*slave_dev
)
545 struct vlan_entry
*vlan
;
546 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
548 write_lock_bh(&bond
->lock
);
550 if (list_empty(&bond
->vlan_list
))
553 if ((slave_dev
->features
& NETIF_F_HW_VLAN_RX
) &&
554 slave_ops
->ndo_vlan_rx_register
)
555 slave_ops
->ndo_vlan_rx_register(slave_dev
, bond
->vlgrp
);
557 if (!(slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) ||
558 !(slave_ops
->ndo_vlan_rx_add_vid
))
561 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
)
562 slave_ops
->ndo_vlan_rx_add_vid(slave_dev
, vlan
->vlan_id
);
565 write_unlock_bh(&bond
->lock
);
568 static void bond_del_vlans_from_slave(struct bonding
*bond
,
569 struct net_device
*slave_dev
)
571 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
572 struct vlan_entry
*vlan
;
573 struct net_device
*vlan_dev
;
575 write_lock_bh(&bond
->lock
);
577 if (list_empty(&bond
->vlan_list
))
580 if (!(slave_dev
->features
& NETIF_F_HW_VLAN_FILTER
) ||
581 !(slave_ops
->ndo_vlan_rx_kill_vid
))
584 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
585 /* Save and then restore vlan_dev in the grp array,
586 * since the slave's driver might clear it.
588 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
589 slave_ops
->ndo_vlan_rx_kill_vid(slave_dev
, vlan
->vlan_id
);
590 vlan_group_set_device(bond
->vlgrp
, vlan
->vlan_id
, vlan_dev
);
594 if ((slave_dev
->features
& NETIF_F_HW_VLAN_RX
) &&
595 slave_ops
->ndo_vlan_rx_register
)
596 slave_ops
->ndo_vlan_rx_register(slave_dev
, NULL
);
599 write_unlock_bh(&bond
->lock
);
602 /*------------------------------- Link status -------------------------------*/
605 * Set the carrier state for the master according to the state of its
606 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
607 * do special 802.3ad magic.
609 * Returns zero if carrier state does not change, nonzero if it does.
611 static int bond_set_carrier(struct bonding
*bond
)
616 if (bond
->slave_cnt
== 0)
619 if (bond
->params
.mode
== BOND_MODE_8023AD
)
620 return bond_3ad_set_carrier(bond
);
622 bond_for_each_slave(bond
, slave
, i
) {
623 if (slave
->link
== BOND_LINK_UP
) {
624 if (!netif_carrier_ok(bond
->dev
)) {
625 netif_carrier_on(bond
->dev
);
633 if (netif_carrier_ok(bond
->dev
)) {
634 netif_carrier_off(bond
->dev
);
641 * Get link speed and duplex from the slave's base driver
642 * using ethtool. If for some reason the call fails or the
643 * values are invalid, fake speed and duplex to 100/Full
646 static int bond_update_speed_duplex(struct slave
*slave
)
648 struct net_device
*slave_dev
= slave
->dev
;
649 struct ethtool_cmd etool
;
652 /* Fake speed and duplex */
653 slave
->speed
= SPEED_100
;
654 slave
->duplex
= DUPLEX_FULL
;
656 if (!slave_dev
->ethtool_ops
|| !slave_dev
->ethtool_ops
->get_settings
)
659 res
= slave_dev
->ethtool_ops
->get_settings(slave_dev
, &etool
);
663 switch (etool
.speed
) {
673 switch (etool
.duplex
) {
681 slave
->speed
= etool
.speed
;
682 slave
->duplex
= etool
.duplex
;
688 * if <dev> supports MII link status reporting, check its link status.
690 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
691 * depending upon the setting of the use_carrier parameter.
693 * Return either BMSR_LSTATUS, meaning that the link is up (or we
694 * can't tell and just pretend it is), or 0, meaning that the link is
697 * If reporting is non-zero, instead of faking link up, return -1 if
698 * both ETHTOOL and MII ioctls fail (meaning the device does not
699 * support them). If use_carrier is set, return whatever it says.
700 * It'd be nice if there was a good way to tell if a driver supports
701 * netif_carrier, but there really isn't.
703 static int bond_check_dev_link(struct bonding
*bond
,
704 struct net_device
*slave_dev
, int reporting
)
706 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
707 int (*ioctl
)(struct net_device
*, struct ifreq
*, int);
709 struct mii_ioctl_data
*mii
;
711 if (!reporting
&& !netif_running(slave_dev
))
714 if (bond
->params
.use_carrier
)
715 return netif_carrier_ok(slave_dev
) ? BMSR_LSTATUS
: 0;
717 /* Try to get link status using Ethtool first. */
718 if (slave_dev
->ethtool_ops
) {
719 if (slave_dev
->ethtool_ops
->get_link
) {
722 link
= slave_dev
->ethtool_ops
->get_link(slave_dev
);
724 return link
? BMSR_LSTATUS
: 0;
728 /* Ethtool can't be used, fallback to MII ioctls. */
729 ioctl
= slave_ops
->ndo_do_ioctl
;
731 /* TODO: set pointer to correct ioctl on a per team member */
732 /* bases to make this more efficient. that is, once */
733 /* we determine the correct ioctl, we will always */
734 /* call it and not the others for that team */
738 * We cannot assume that SIOCGMIIPHY will also read a
739 * register; not all network drivers (e.g., e100)
743 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
744 strncpy(ifr
.ifr_name
, slave_dev
->name
, IFNAMSIZ
);
746 if (IOCTL(slave_dev
, &ifr
, SIOCGMIIPHY
) == 0) {
747 mii
->reg_num
= MII_BMSR
;
748 if (IOCTL(slave_dev
, &ifr
, SIOCGMIIREG
) == 0)
749 return mii
->val_out
& BMSR_LSTATUS
;
754 * If reporting, report that either there's no dev->do_ioctl,
755 * or both SIOCGMIIREG and get_link failed (meaning that we
756 * cannot report link status). If not reporting, pretend
759 return reporting
? -1 : BMSR_LSTATUS
;
762 /*----------------------------- Multicast list ------------------------------*/
765 * Returns 0 if dmi1 and dmi2 are the same, non-0 otherwise
767 static inline int bond_is_dmi_same(const struct dev_mc_list
*dmi1
,
768 const struct dev_mc_list
*dmi2
)
770 return memcmp(dmi1
->dmi_addr
, dmi2
->dmi_addr
, dmi1
->dmi_addrlen
) == 0 &&
771 dmi1
->dmi_addrlen
== dmi2
->dmi_addrlen
;
775 * returns dmi entry if found, NULL otherwise
777 static struct dev_mc_list
*bond_mc_list_find_dmi(struct dev_mc_list
*dmi
,
778 struct dev_mc_list
*mc_list
)
780 struct dev_mc_list
*idmi
;
782 for (idmi
= mc_list
; idmi
; idmi
= idmi
->next
) {
783 if (bond_is_dmi_same(dmi
, idmi
))
791 * Push the promiscuity flag down to appropriate slaves
793 static int bond_set_promiscuity(struct bonding
*bond
, int inc
)
796 if (USES_PRIMARY(bond
->params
.mode
)) {
797 /* write lock already acquired */
798 if (bond
->curr_active_slave
) {
799 err
= dev_set_promiscuity(bond
->curr_active_slave
->dev
,
805 bond_for_each_slave(bond
, slave
, i
) {
806 err
= dev_set_promiscuity(slave
->dev
, inc
);
815 * Push the allmulti flag down to all slaves
817 static int bond_set_allmulti(struct bonding
*bond
, int inc
)
820 if (USES_PRIMARY(bond
->params
.mode
)) {
821 /* write lock already acquired */
822 if (bond
->curr_active_slave
) {
823 err
= dev_set_allmulti(bond
->curr_active_slave
->dev
,
829 bond_for_each_slave(bond
, slave
, i
) {
830 err
= dev_set_allmulti(slave
->dev
, inc
);
839 * Add a Multicast address to slaves
842 static void bond_mc_add(struct bonding
*bond
, void *addr
, int alen
)
844 if (USES_PRIMARY(bond
->params
.mode
)) {
845 /* write lock already acquired */
846 if (bond
->curr_active_slave
)
847 dev_mc_add(bond
->curr_active_slave
->dev
, addr
, alen
, 0);
852 bond_for_each_slave(bond
, slave
, i
)
853 dev_mc_add(slave
->dev
, addr
, alen
, 0);
858 * Remove a multicast address from slave
861 static void bond_mc_delete(struct bonding
*bond
, void *addr
, int alen
)
863 if (USES_PRIMARY(bond
->params
.mode
)) {
864 /* write lock already acquired */
865 if (bond
->curr_active_slave
)
866 dev_mc_delete(bond
->curr_active_slave
->dev
, addr
,
871 bond_for_each_slave(bond
, slave
, i
) {
872 dev_mc_delete(slave
->dev
, addr
, alen
, 0);
879 * Retrieve the list of registered multicast addresses for the bonding
880 * device and retransmit an IGMP JOIN request to the current active
883 static void bond_resend_igmp_join_requests(struct bonding
*bond
)
885 struct in_device
*in_dev
;
886 struct ip_mc_list
*im
;
889 in_dev
= __in_dev_get_rcu(bond
->dev
);
891 for (im
= in_dev
->mc_list
; im
; im
= im
->next
)
892 ip_mc_rejoin_group(im
);
899 * Totally destroys the mc_list in bond
901 static void bond_mc_list_destroy(struct bonding
*bond
)
903 struct dev_mc_list
*dmi
;
907 bond
->mc_list
= dmi
->next
;
912 bond
->mc_list
= NULL
;
916 * Copy all the Multicast addresses from src to the bonding device dst
918 static int bond_mc_list_copy(struct dev_mc_list
*mc_list
, struct bonding
*bond
,
921 struct dev_mc_list
*dmi
, *new_dmi
;
923 for (dmi
= mc_list
; dmi
; dmi
= dmi
->next
) {
924 new_dmi
= kmalloc(sizeof(struct dev_mc_list
), gfp_flag
);
927 /* FIXME: Potential memory leak !!! */
931 new_dmi
->next
= bond
->mc_list
;
932 bond
->mc_list
= new_dmi
;
933 new_dmi
->dmi_addrlen
= dmi
->dmi_addrlen
;
934 memcpy(new_dmi
->dmi_addr
, dmi
->dmi_addr
, dmi
->dmi_addrlen
);
935 new_dmi
->dmi_users
= dmi
->dmi_users
;
936 new_dmi
->dmi_gusers
= dmi
->dmi_gusers
;
943 * flush all members of flush->mc_list from device dev->mc_list
945 static void bond_mc_list_flush(struct net_device
*bond_dev
,
946 struct net_device
*slave_dev
)
948 struct bonding
*bond
= netdev_priv(bond_dev
);
949 struct dev_mc_list
*dmi
;
951 for (dmi
= bond_dev
->mc_list
; dmi
; dmi
= dmi
->next
)
952 dev_mc_delete(slave_dev
, dmi
->dmi_addr
, dmi
->dmi_addrlen
, 0);
954 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
955 /* del lacpdu mc addr from mc list */
956 u8 lacpdu_multicast
[ETH_ALEN
] = MULTICAST_LACPDU_ADDR
;
958 dev_mc_delete(slave_dev
, lacpdu_multicast
, ETH_ALEN
, 0);
962 /*--------------------------- Active slave change ---------------------------*/
965 * Update the mc list and multicast-related flags for the new and
966 * old active slaves (if any) according to the multicast mode, and
967 * promiscuous flags unconditionally.
969 static void bond_mc_swap(struct bonding
*bond
, struct slave
*new_active
,
970 struct slave
*old_active
)
972 struct dev_mc_list
*dmi
;
974 if (!USES_PRIMARY(bond
->params
.mode
))
975 /* nothing to do - mc list is already up-to-date on
981 if (bond
->dev
->flags
& IFF_PROMISC
)
982 dev_set_promiscuity(old_active
->dev
, -1);
984 if (bond
->dev
->flags
& IFF_ALLMULTI
)
985 dev_set_allmulti(old_active
->dev
, -1);
987 for (dmi
= bond
->dev
->mc_list
; dmi
; dmi
= dmi
->next
)
988 dev_mc_delete(old_active
->dev
, dmi
->dmi_addr
,
989 dmi
->dmi_addrlen
, 0);
993 /* FIXME: Signal errors upstream. */
994 if (bond
->dev
->flags
& IFF_PROMISC
)
995 dev_set_promiscuity(new_active
->dev
, 1);
997 if (bond
->dev
->flags
& IFF_ALLMULTI
)
998 dev_set_allmulti(new_active
->dev
, 1);
1000 for (dmi
= bond
->dev
->mc_list
; dmi
; dmi
= dmi
->next
)
1001 dev_mc_add(new_active
->dev
, dmi
->dmi_addr
,
1002 dmi
->dmi_addrlen
, 0);
1003 bond_resend_igmp_join_requests(bond
);
1008 * bond_do_fail_over_mac
1010 * Perform special MAC address swapping for fail_over_mac settings
1012 * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh.
1014 static void bond_do_fail_over_mac(struct bonding
*bond
,
1015 struct slave
*new_active
,
1016 struct slave
*old_active
)
1017 __releases(&bond
->curr_slave_lock
)
1018 __releases(&bond
->lock
)
1019 __acquires(&bond
->lock
)
1020 __acquires(&bond
->curr_slave_lock
)
1022 u8 tmp_mac
[ETH_ALEN
];
1023 struct sockaddr saddr
;
1026 switch (bond
->params
.fail_over_mac
) {
1027 case BOND_FOM_ACTIVE
:
1029 memcpy(bond
->dev
->dev_addr
, new_active
->dev
->dev_addr
,
1030 new_active
->dev
->addr_len
);
1032 case BOND_FOM_FOLLOW
:
1034 * if new_active && old_active, swap them
1035 * if just old_active, do nothing (going to no active slave)
1036 * if just new_active, set new_active to bond's MAC
1041 write_unlock_bh(&bond
->curr_slave_lock
);
1042 read_unlock(&bond
->lock
);
1045 memcpy(tmp_mac
, new_active
->dev
->dev_addr
, ETH_ALEN
);
1046 memcpy(saddr
.sa_data
, old_active
->dev
->dev_addr
,
1048 saddr
.sa_family
= new_active
->dev
->type
;
1050 memcpy(saddr
.sa_data
, bond
->dev
->dev_addr
, ETH_ALEN
);
1051 saddr
.sa_family
= bond
->dev
->type
;
1054 rv
= dev_set_mac_address(new_active
->dev
, &saddr
);
1056 pr_err("%s: Error %d setting MAC of slave %s\n",
1057 bond
->dev
->name
, -rv
, new_active
->dev
->name
);
1064 memcpy(saddr
.sa_data
, tmp_mac
, ETH_ALEN
);
1065 saddr
.sa_family
= old_active
->dev
->type
;
1067 rv
= dev_set_mac_address(old_active
->dev
, &saddr
);
1069 pr_err("%s: Error %d setting MAC of slave %s\n",
1070 bond
->dev
->name
, -rv
, new_active
->dev
->name
);
1072 read_lock(&bond
->lock
);
1073 write_lock_bh(&bond
->curr_slave_lock
);
1076 pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n",
1077 bond
->dev
->name
, bond
->params
.fail_over_mac
);
1083 static bool bond_should_change_active(struct bonding
*bond
)
1085 struct slave
*prim
= bond
->primary_slave
;
1086 struct slave
*curr
= bond
->curr_active_slave
;
1088 if (!prim
|| !curr
|| curr
->link
!= BOND_LINK_UP
)
1090 if (bond
->force_primary
) {
1091 bond
->force_primary
= false;
1094 if (bond
->params
.primary_reselect
== BOND_PRI_RESELECT_BETTER
&&
1095 (prim
->speed
< curr
->speed
||
1096 (prim
->speed
== curr
->speed
&& prim
->duplex
<= curr
->duplex
)))
1098 if (bond
->params
.primary_reselect
== BOND_PRI_RESELECT_FAILURE
)
1104 * find_best_interface - select the best available slave to be the active one
1105 * @bond: our bonding struct
1107 * Warning: Caller must hold curr_slave_lock for writing.
1109 static struct slave
*bond_find_best_slave(struct bonding
*bond
)
1111 struct slave
*new_active
, *old_active
;
1112 struct slave
*bestslave
= NULL
;
1113 int mintime
= bond
->params
.updelay
;
1116 new_active
= bond
->curr_active_slave
;
1118 if (!new_active
) { /* there were no active slaves left */
1119 if (bond
->slave_cnt
> 0) /* found one slave */
1120 new_active
= bond
->first_slave
;
1122 return NULL
; /* still no slave, return NULL */
1125 if ((bond
->primary_slave
) &&
1126 bond
->primary_slave
->link
== BOND_LINK_UP
&&
1127 bond_should_change_active(bond
)) {
1128 new_active
= bond
->primary_slave
;
1131 /* remember where to stop iterating over the slaves */
1132 old_active
= new_active
;
1134 bond_for_each_slave_from(bond
, new_active
, i
, old_active
) {
1135 if (new_active
->link
== BOND_LINK_UP
) {
1137 } else if (new_active
->link
== BOND_LINK_BACK
&&
1138 IS_UP(new_active
->dev
)) {
1139 /* link up, but waiting for stabilization */
1140 if (new_active
->delay
< mintime
) {
1141 mintime
= new_active
->delay
;
1142 bestslave
= new_active
;
1151 * change_active_interface - change the active slave into the specified one
1152 * @bond: our bonding struct
1153 * @new: the new slave to make the active one
1155 * Set the new slave to the bond's settings and unset them on the old
1156 * curr_active_slave.
1157 * Setting include flags, mc-list, promiscuity, allmulti, etc.
1159 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
1160 * because it is apparently the best available slave we have, even though its
1161 * updelay hasn't timed out yet.
1163 * If new_active is not NULL, caller must hold bond->lock for read and
1164 * curr_slave_lock for write_bh.
1166 void bond_change_active_slave(struct bonding
*bond
, struct slave
*new_active
)
1168 struct slave
*old_active
= bond
->curr_active_slave
;
1170 if (old_active
== new_active
)
1174 new_active
->jiffies
= jiffies
;
1176 if (new_active
->link
== BOND_LINK_BACK
) {
1177 if (USES_PRIMARY(bond
->params
.mode
)) {
1178 pr_info("%s: making interface %s the new active one %d ms earlier.\n",
1179 bond
->dev
->name
, new_active
->dev
->name
,
1180 (bond
->params
.updelay
- new_active
->delay
) * bond
->params
.miimon
);
1183 new_active
->delay
= 0;
1184 new_active
->link
= BOND_LINK_UP
;
1186 if (bond
->params
.mode
== BOND_MODE_8023AD
)
1187 bond_3ad_handle_link_change(new_active
, BOND_LINK_UP
);
1189 if (bond_is_lb(bond
))
1190 bond_alb_handle_link_change(bond
, new_active
, BOND_LINK_UP
);
1192 if (USES_PRIMARY(bond
->params
.mode
)) {
1193 pr_info("%s: making interface %s the new active one.\n",
1194 bond
->dev
->name
, new_active
->dev
->name
);
1199 if (USES_PRIMARY(bond
->params
.mode
))
1200 bond_mc_swap(bond
, new_active
, old_active
);
1202 if (bond_is_lb(bond
)) {
1203 bond_alb_handle_active_change(bond
, new_active
);
1205 bond_set_slave_inactive_flags(old_active
);
1207 bond_set_slave_active_flags(new_active
);
1209 bond
->curr_active_slave
= new_active
;
1212 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
) {
1214 bond_set_slave_inactive_flags(old_active
);
1217 bond_set_slave_active_flags(new_active
);
1219 if (bond
->params
.fail_over_mac
)
1220 bond_do_fail_over_mac(bond
, new_active
,
1223 bond
->send_grat_arp
= bond
->params
.num_grat_arp
;
1224 bond_send_gratuitous_arp(bond
);
1226 bond
->send_unsol_na
= bond
->params
.num_unsol_na
;
1227 bond_send_unsolicited_na(bond
);
1229 write_unlock_bh(&bond
->curr_slave_lock
);
1230 read_unlock(&bond
->lock
);
1232 netdev_bonding_change(bond
->dev
, NETDEV_BONDING_FAILOVER
);
1234 read_lock(&bond
->lock
);
1235 write_lock_bh(&bond
->curr_slave_lock
);
1241 * bond_select_active_slave - select a new active slave, if needed
1242 * @bond: our bonding struct
1244 * This functions should be called when one of the following occurs:
1245 * - The old curr_active_slave has been released or lost its link.
1246 * - The primary_slave has got its link back.
1247 * - A slave has got its link back and there's no old curr_active_slave.
1249 * Caller must hold bond->lock for read and curr_slave_lock for write_bh.
1251 void bond_select_active_slave(struct bonding
*bond
)
1253 struct slave
*best_slave
;
1256 best_slave
= bond_find_best_slave(bond
);
1257 if (best_slave
!= bond
->curr_active_slave
) {
1258 bond_change_active_slave(bond
, best_slave
);
1259 rv
= bond_set_carrier(bond
);
1263 if (netif_carrier_ok(bond
->dev
)) {
1264 pr_info("%s: first active interface up!\n",
1267 pr_info("%s: now running without any active interface !\n",
1273 /*--------------------------- slave list handling ---------------------------*/
1276 * This function attaches the slave to the end of list.
1278 * bond->lock held for writing by caller.
1280 static void bond_attach_slave(struct bonding
*bond
, struct slave
*new_slave
)
1282 if (bond
->first_slave
== NULL
) { /* attaching the first slave */
1283 new_slave
->next
= new_slave
;
1284 new_slave
->prev
= new_slave
;
1285 bond
->first_slave
= new_slave
;
1287 new_slave
->next
= bond
->first_slave
;
1288 new_slave
->prev
= bond
->first_slave
->prev
;
1289 new_slave
->next
->prev
= new_slave
;
1290 new_slave
->prev
->next
= new_slave
;
1297 * This function detaches the slave from the list.
1298 * WARNING: no check is made to verify if the slave effectively
1299 * belongs to <bond>.
1300 * Nothing is freed on return, structures are just unchained.
1301 * If any slave pointer in bond was pointing to <slave>,
1302 * it should be changed by the calling function.
1304 * bond->lock held for writing by caller.
1306 static void bond_detach_slave(struct bonding
*bond
, struct slave
*slave
)
1309 slave
->next
->prev
= slave
->prev
;
1312 slave
->prev
->next
= slave
->next
;
1314 if (bond
->first_slave
== slave
) { /* slave is the first slave */
1315 if (bond
->slave_cnt
> 1) { /* there are more slave */
1316 bond
->first_slave
= slave
->next
;
1318 bond
->first_slave
= NULL
; /* slave was the last one */
1327 /*---------------------------------- IOCTL ----------------------------------*/
1329 static int bond_sethwaddr(struct net_device
*bond_dev
,
1330 struct net_device
*slave_dev
)
1332 pr_debug("bond_dev=%p\n", bond_dev
);
1333 pr_debug("slave_dev=%p\n", slave_dev
);
1334 pr_debug("slave_dev->addr_len=%d\n", slave_dev
->addr_len
);
1335 memcpy(bond_dev
->dev_addr
, slave_dev
->dev_addr
, slave_dev
->addr_len
);
1339 #define BOND_VLAN_FEATURES \
1340 (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \
1341 NETIF_F_HW_VLAN_FILTER)
1344 * Compute the common dev->feature set available to all slaves. Some
1345 * feature bits are managed elsewhere, so preserve those feature bits
1346 * on the master device.
1348 static int bond_compute_features(struct bonding
*bond
)
1350 struct slave
*slave
;
1351 struct net_device
*bond_dev
= bond
->dev
;
1352 unsigned long features
= bond_dev
->features
;
1353 unsigned long vlan_features
= 0;
1354 unsigned short max_hard_header_len
= max((u16
)ETH_HLEN
,
1355 bond_dev
->hard_header_len
);
1358 features
&= ~(NETIF_F_ALL_CSUM
| BOND_VLAN_FEATURES
);
1359 features
|= NETIF_F_GSO_MASK
| NETIF_F_NO_CSUM
;
1361 if (!bond
->first_slave
)
1364 features
&= ~NETIF_F_ONE_FOR_ALL
;
1366 vlan_features
= bond
->first_slave
->dev
->vlan_features
;
1367 bond_for_each_slave(bond
, slave
, i
) {
1368 features
= netdev_increment_features(features
,
1369 slave
->dev
->features
,
1370 NETIF_F_ONE_FOR_ALL
);
1371 vlan_features
= netdev_increment_features(vlan_features
,
1372 slave
->dev
->vlan_features
,
1373 NETIF_F_ONE_FOR_ALL
);
1374 if (slave
->dev
->hard_header_len
> max_hard_header_len
)
1375 max_hard_header_len
= slave
->dev
->hard_header_len
;
1379 features
|= (bond_dev
->features
& BOND_VLAN_FEATURES
);
1380 bond_dev
->features
= netdev_fix_features(features
, NULL
);
1381 bond_dev
->vlan_features
= netdev_fix_features(vlan_features
, NULL
);
1382 bond_dev
->hard_header_len
= max_hard_header_len
;
1387 static void bond_setup_by_slave(struct net_device
*bond_dev
,
1388 struct net_device
*slave_dev
)
1390 struct bonding
*bond
= netdev_priv(bond_dev
);
1392 bond_dev
->header_ops
= slave_dev
->header_ops
;
1394 bond_dev
->type
= slave_dev
->type
;
1395 bond_dev
->hard_header_len
= slave_dev
->hard_header_len
;
1396 bond_dev
->addr_len
= slave_dev
->addr_len
;
1398 memcpy(bond_dev
->broadcast
, slave_dev
->broadcast
,
1399 slave_dev
->addr_len
);
1400 bond
->setup_by_slave
= 1;
1403 /* enslave device <slave> to bond device <master> */
1404 int bond_enslave(struct net_device
*bond_dev
, struct net_device
*slave_dev
)
1406 struct bonding
*bond
= netdev_priv(bond_dev
);
1407 const struct net_device_ops
*slave_ops
= slave_dev
->netdev_ops
;
1408 struct slave
*new_slave
= NULL
;
1409 struct dev_mc_list
*dmi
;
1410 struct sockaddr addr
;
1412 int old_features
= bond_dev
->features
;
1415 if (!bond
->params
.use_carrier
&& slave_dev
->ethtool_ops
== NULL
&&
1416 slave_ops
->ndo_do_ioctl
== NULL
) {
1417 pr_warning("%s: Warning: no link monitoring support for %s\n",
1418 bond_dev
->name
, slave_dev
->name
);
1421 /* bond must be initialized by bond_open() before enslaving */
1422 if (!(bond_dev
->flags
& IFF_UP
)) {
1423 pr_warning("%s: master_dev is not up in bond_enslave\n",
1427 /* already enslaved */
1428 if (slave_dev
->flags
& IFF_SLAVE
) {
1429 pr_debug("Error, Device was already enslaved\n");
1433 /* vlan challenged mutual exclusion */
1434 /* no need to lock since we're protected by rtnl_lock */
1435 if (slave_dev
->features
& NETIF_F_VLAN_CHALLENGED
) {
1436 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev
->name
);
1437 if (!list_empty(&bond
->vlan_list
)) {
1438 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1439 bond_dev
->name
, slave_dev
->name
, bond_dev
->name
);
1442 pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1443 bond_dev
->name
, slave_dev
->name
,
1444 slave_dev
->name
, bond_dev
->name
);
1445 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
1448 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev
->name
);
1449 if (bond
->slave_cnt
== 0) {
1450 /* First slave, and it is not VLAN challenged,
1451 * so remove the block of adding VLANs over the bond.
1453 bond_dev
->features
&= ~NETIF_F_VLAN_CHALLENGED
;
1458 * Old ifenslave binaries are no longer supported. These can
1459 * be identified with moderate accuracy by the state of the slave:
1460 * the current ifenslave will set the interface down prior to
1461 * enslaving it; the old ifenslave will not.
1463 if ((slave_dev
->flags
& IFF_UP
)) {
1464 pr_err("%s is up. This may be due to an out of date ifenslave.\n",
1467 goto err_undo_flags
;
1470 /* set bonding device ether type by slave - bonding netdevices are
1471 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1472 * there is a need to override some of the type dependent attribs/funcs.
1474 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1475 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1477 if (bond
->slave_cnt
== 0) {
1478 if (bond_dev
->type
!= slave_dev
->type
) {
1479 pr_debug("%s: change device type from %d to %d\n",
1481 bond_dev
->type
, slave_dev
->type
);
1483 res
= netdev_bonding_change(bond_dev
,
1484 NETDEV_PRE_TYPE_CHANGE
);
1485 res
= notifier_to_errno(res
);
1487 pr_err("%s: refused to change device type\n",
1490 goto err_undo_flags
;
1493 if (slave_dev
->type
!= ARPHRD_ETHER
)
1494 bond_setup_by_slave(bond_dev
, slave_dev
);
1496 ether_setup(bond_dev
);
1498 netdev_bonding_change(bond_dev
,
1499 NETDEV_POST_TYPE_CHANGE
);
1501 } else if (bond_dev
->type
!= slave_dev
->type
) {
1502 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n",
1504 slave_dev
->type
, bond_dev
->type
);
1506 goto err_undo_flags
;
1509 if (slave_ops
->ndo_set_mac_address
== NULL
) {
1510 if (bond
->slave_cnt
== 0) {
1511 pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.",
1513 bond
->params
.fail_over_mac
= BOND_FOM_ACTIVE
;
1514 } else if (bond
->params
.fail_over_mac
!= BOND_FOM_ACTIVE
) {
1515 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",
1518 goto err_undo_flags
;
1522 new_slave
= kzalloc(sizeof(struct slave
), GFP_KERNEL
);
1525 goto err_undo_flags
;
1528 /* save slave's original flags before calling
1529 * netdev_set_master and dev_open
1531 new_slave
->original_flags
= slave_dev
->flags
;
1534 * Save slave's original ("permanent") mac address for modes
1535 * that need it, and for restoring it upon release, and then
1536 * set it to the master's address
1538 memcpy(new_slave
->perm_hwaddr
, slave_dev
->dev_addr
, ETH_ALEN
);
1540 if (!bond
->params
.fail_over_mac
) {
1542 * Set slave to master's mac address. The application already
1543 * set the master's mac address to that of the first slave
1545 memcpy(addr
.sa_data
, bond_dev
->dev_addr
, bond_dev
->addr_len
);
1546 addr
.sa_family
= slave_dev
->type
;
1547 res
= dev_set_mac_address(slave_dev
, &addr
);
1549 pr_debug("Error %d calling set_mac_address\n", res
);
1554 res
= netdev_set_master(slave_dev
, bond_dev
);
1556 pr_debug("Error %d calling netdev_set_master\n", res
);
1557 goto err_restore_mac
;
1559 /* open the slave since the application closed it */
1560 res
= dev_open(slave_dev
);
1562 pr_debug("Opening slave %s failed\n", slave_dev
->name
);
1563 goto err_unset_master
;
1566 new_slave
->dev
= slave_dev
;
1567 slave_dev
->priv_flags
|= IFF_BONDING
;
1569 if (bond_is_lb(bond
)) {
1570 /* bond_alb_init_slave() must be called before all other stages since
1571 * it might fail and we do not want to have to undo everything
1573 res
= bond_alb_init_slave(bond
, new_slave
);
1578 /* If the mode USES_PRIMARY, then the new slave gets the
1579 * master's promisc (and mc) settings only if it becomes the
1580 * curr_active_slave, and that is taken care of later when calling
1581 * bond_change_active()
1583 if (!USES_PRIMARY(bond
->params
.mode
)) {
1584 /* set promiscuity level to new slave */
1585 if (bond_dev
->flags
& IFF_PROMISC
) {
1586 res
= dev_set_promiscuity(slave_dev
, 1);
1591 /* set allmulti level to new slave */
1592 if (bond_dev
->flags
& IFF_ALLMULTI
) {
1593 res
= dev_set_allmulti(slave_dev
, 1);
1598 netif_addr_lock_bh(bond_dev
);
1599 /* upload master's mc_list to new slave */
1600 for (dmi
= bond_dev
->mc_list
; dmi
; dmi
= dmi
->next
)
1601 dev_mc_add(slave_dev
, dmi
->dmi_addr
,
1602 dmi
->dmi_addrlen
, 0);
1603 netif_addr_unlock_bh(bond_dev
);
1606 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
1607 /* add lacpdu mc addr to mc list */
1608 u8 lacpdu_multicast
[ETH_ALEN
] = MULTICAST_LACPDU_ADDR
;
1610 dev_mc_add(slave_dev
, lacpdu_multicast
, ETH_ALEN
, 0);
1613 bond_add_vlans_on_slave(bond
, slave_dev
);
1615 write_lock_bh(&bond
->lock
);
1617 bond_attach_slave(bond
, new_slave
);
1619 new_slave
->delay
= 0;
1620 new_slave
->link_failure_count
= 0;
1622 bond_compute_features(bond
);
1624 write_unlock_bh(&bond
->lock
);
1626 read_lock(&bond
->lock
);
1628 new_slave
->last_arp_rx
= jiffies
;
1630 if (bond
->params
.miimon
&& !bond
->params
.use_carrier
) {
1631 link_reporting
= bond_check_dev_link(bond
, slave_dev
, 1);
1633 if ((link_reporting
== -1) && !bond
->params
.arp_interval
) {
1635 * miimon is set but a bonded network driver
1636 * does not support ETHTOOL/MII and
1637 * arp_interval is not set. Note: if
1638 * use_carrier is enabled, we will never go
1639 * here (because netif_carrier is always
1640 * supported); thus, we don't need to change
1641 * the messages for netif_carrier.
1643 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",
1644 bond_dev
->name
, slave_dev
->name
);
1645 } else if (link_reporting
== -1) {
1646 /* unable get link status using mii/ethtool */
1647 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",
1648 bond_dev
->name
, slave_dev
->name
);
1652 /* check for initial state */
1653 if (!bond
->params
.miimon
||
1654 (bond_check_dev_link(bond
, slave_dev
, 0) == BMSR_LSTATUS
)) {
1655 if (bond
->params
.updelay
) {
1656 pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n");
1657 new_slave
->link
= BOND_LINK_BACK
;
1658 new_slave
->delay
= bond
->params
.updelay
;
1660 pr_debug("Initial state of slave_dev is BOND_LINK_UP\n");
1661 new_slave
->link
= BOND_LINK_UP
;
1663 new_slave
->jiffies
= jiffies
;
1665 pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n");
1666 new_slave
->link
= BOND_LINK_DOWN
;
1669 if (bond_update_speed_duplex(new_slave
) &&
1670 (new_slave
->link
!= BOND_LINK_DOWN
)) {
1671 pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n",
1672 bond_dev
->name
, new_slave
->dev
->name
);
1674 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
1675 pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n",
1680 if (USES_PRIMARY(bond
->params
.mode
) && bond
->params
.primary
[0]) {
1681 /* if there is a primary slave, remember it */
1682 if (strcmp(bond
->params
.primary
, new_slave
->dev
->name
) == 0) {
1683 bond
->primary_slave
= new_slave
;
1684 bond
->force_primary
= true;
1688 write_lock_bh(&bond
->curr_slave_lock
);
1690 switch (bond
->params
.mode
) {
1691 case BOND_MODE_ACTIVEBACKUP
:
1692 bond_set_slave_inactive_flags(new_slave
);
1693 bond_select_active_slave(bond
);
1695 case BOND_MODE_8023AD
:
1696 /* in 802.3ad mode, the internal mechanism
1697 * will activate the slaves in the selected
1700 bond_set_slave_inactive_flags(new_slave
);
1701 /* if this is the first slave */
1702 if (bond
->slave_cnt
== 1) {
1703 SLAVE_AD_INFO(new_slave
).id
= 1;
1704 /* Initialize AD with the number of times that the AD timer is called in 1 second
1705 * can be called only after the mac address of the bond is set
1707 bond_3ad_initialize(bond
, 1000/AD_TIMER_INTERVAL
,
1708 bond
->params
.lacp_fast
);
1710 SLAVE_AD_INFO(new_slave
).id
=
1711 SLAVE_AD_INFO(new_slave
->prev
).id
+ 1;
1714 bond_3ad_bind_slave(new_slave
);
1718 new_slave
->state
= BOND_STATE_ACTIVE
;
1719 bond_set_slave_inactive_flags(new_slave
);
1720 bond_select_active_slave(bond
);
1723 pr_debug("This slave is always active in trunk mode\n");
1725 /* always active in trunk mode */
1726 new_slave
->state
= BOND_STATE_ACTIVE
;
1728 /* In trunking mode there is little meaning to curr_active_slave
1729 * anyway (it holds no special properties of the bond device),
1730 * so we can change it without calling change_active_interface()
1732 if (!bond
->curr_active_slave
)
1733 bond
->curr_active_slave
= new_slave
;
1736 } /* switch(bond_mode) */
1738 write_unlock_bh(&bond
->curr_slave_lock
);
1740 bond_set_carrier(bond
);
1742 read_unlock(&bond
->lock
);
1744 res
= bond_create_slave_symlinks(bond_dev
, slave_dev
);
1748 pr_info("%s: enslaving %s as a%s interface with a%s link.\n",
1749 bond_dev
->name
, slave_dev
->name
,
1750 new_slave
->state
== BOND_STATE_ACTIVE
? "n active" : " backup",
1751 new_slave
->link
!= BOND_LINK_DOWN
? "n up" : " down");
1753 /* enslave is successful */
1756 /* Undo stages on error */
1758 dev_close(slave_dev
);
1761 netdev_set_master(slave_dev
, NULL
);
1764 if (!bond
->params
.fail_over_mac
) {
1765 /* XXX TODO - fom follow mode needs to change master's
1766 * MAC if this slave's MAC is in use by the bond, or at
1767 * least print a warning.
1769 memcpy(addr
.sa_data
, new_slave
->perm_hwaddr
, ETH_ALEN
);
1770 addr
.sa_family
= slave_dev
->type
;
1771 dev_set_mac_address(slave_dev
, &addr
);
1778 bond_dev
->features
= old_features
;
1784 * Try to release the slave device <slave> from the bond device <master>
1785 * It is legal to access curr_active_slave without a lock because all the function
1788 * The rules for slave state should be:
1789 * for Active/Backup:
1790 * Active stays on all backups go down
1791 * for Bonded connections:
1792 * The first up interface should be left on and all others downed.
1794 int bond_release(struct net_device
*bond_dev
, struct net_device
*slave_dev
)
1796 struct bonding
*bond
= netdev_priv(bond_dev
);
1797 struct slave
*slave
, *oldcurrent
;
1798 struct sockaddr addr
;
1800 /* slave is not a slave or master is not master of this slave */
1801 if (!(slave_dev
->flags
& IFF_SLAVE
) ||
1802 (slave_dev
->master
!= bond_dev
)) {
1803 pr_err("%s: Error: cannot release %s.\n",
1804 bond_dev
->name
, slave_dev
->name
);
1808 write_lock_bh(&bond
->lock
);
1810 slave
= bond_get_slave_by_dev(bond
, slave_dev
);
1812 /* not a slave of this bond */
1813 pr_info("%s: %s not enslaved\n",
1814 bond_dev
->name
, slave_dev
->name
);
1815 write_unlock_bh(&bond
->lock
);
1819 if (!bond
->params
.fail_over_mac
) {
1820 if (!compare_ether_addr(bond_dev
->dev_addr
, slave
->perm_hwaddr
) &&
1821 bond
->slave_cnt
> 1)
1822 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",
1823 bond_dev
->name
, slave_dev
->name
,
1825 bond_dev
->name
, slave_dev
->name
);
1828 /* Inform AD package of unbinding of slave. */
1829 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
1830 /* must be called before the slave is
1831 * detached from the list
1833 bond_3ad_unbind_slave(slave
);
1836 pr_info("%s: releasing %s interface %s\n",
1838 (slave
->state
== BOND_STATE_ACTIVE
) ? "active" : "backup",
1841 oldcurrent
= bond
->curr_active_slave
;
1843 bond
->current_arp_slave
= NULL
;
1845 /* release the slave from its bond */
1846 bond_detach_slave(bond
, slave
);
1848 bond_compute_features(bond
);
1850 if (bond
->primary_slave
== slave
)
1851 bond
->primary_slave
= NULL
;
1853 if (oldcurrent
== slave
)
1854 bond_change_active_slave(bond
, NULL
);
1856 if (bond_is_lb(bond
)) {
1857 /* Must be called only after the slave has been
1858 * detached from the list and the curr_active_slave
1859 * has been cleared (if our_slave == old_current),
1860 * but before a new active slave is selected.
1862 write_unlock_bh(&bond
->lock
);
1863 bond_alb_deinit_slave(bond
, slave
);
1864 write_lock_bh(&bond
->lock
);
1867 if (oldcurrent
== slave
) {
1869 * Note that we hold RTNL over this sequence, so there
1870 * is no concern that another slave add/remove event
1873 write_unlock_bh(&bond
->lock
);
1874 read_lock(&bond
->lock
);
1875 write_lock_bh(&bond
->curr_slave_lock
);
1877 bond_select_active_slave(bond
);
1879 write_unlock_bh(&bond
->curr_slave_lock
);
1880 read_unlock(&bond
->lock
);
1881 write_lock_bh(&bond
->lock
);
1884 if (bond
->slave_cnt
== 0) {
1885 bond_set_carrier(bond
);
1887 /* if the last slave was removed, zero the mac address
1888 * of the master so it will be set by the application
1889 * to the mac address of the first slave
1891 memset(bond_dev
->dev_addr
, 0, bond_dev
->addr_len
);
1893 if (list_empty(&bond
->vlan_list
)) {
1894 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
1896 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
1897 bond_dev
->name
, bond_dev
->name
);
1898 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
1901 } else if ((bond_dev
->features
& NETIF_F_VLAN_CHALLENGED
) &&
1902 !bond_has_challenged_slaves(bond
)) {
1903 pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n",
1904 bond_dev
->name
, slave_dev
->name
, bond_dev
->name
);
1905 bond_dev
->features
&= ~NETIF_F_VLAN_CHALLENGED
;
1908 write_unlock_bh(&bond
->lock
);
1910 /* must do this from outside any spinlocks */
1911 bond_destroy_slave_symlinks(bond_dev
, slave_dev
);
1913 bond_del_vlans_from_slave(bond
, slave_dev
);
1915 /* If the mode USES_PRIMARY, then we should only remove its
1916 * promisc and mc settings if it was the curr_active_slave, but that was
1917 * already taken care of above when we detached the slave
1919 if (!USES_PRIMARY(bond
->params
.mode
)) {
1920 /* unset promiscuity level from slave */
1921 if (bond_dev
->flags
& IFF_PROMISC
)
1922 dev_set_promiscuity(slave_dev
, -1);
1924 /* unset allmulti level from slave */
1925 if (bond_dev
->flags
& IFF_ALLMULTI
)
1926 dev_set_allmulti(slave_dev
, -1);
1928 /* flush master's mc_list from slave */
1929 netif_addr_lock_bh(bond_dev
);
1930 bond_mc_list_flush(bond_dev
, slave_dev
);
1931 netif_addr_unlock_bh(bond_dev
);
1934 netdev_set_master(slave_dev
, NULL
);
1936 /* close slave before restoring its mac address */
1937 dev_close(slave_dev
);
1939 if (bond
->params
.fail_over_mac
!= BOND_FOM_ACTIVE
) {
1940 /* restore original ("permanent") mac address */
1941 memcpy(addr
.sa_data
, slave
->perm_hwaddr
, ETH_ALEN
);
1942 addr
.sa_family
= slave_dev
->type
;
1943 dev_set_mac_address(slave_dev
, &addr
);
1946 slave_dev
->priv_flags
&= ~(IFF_MASTER_8023AD
| IFF_MASTER_ALB
|
1947 IFF_SLAVE_INACTIVE
| IFF_BONDING
|
1952 return 0; /* deletion OK */
1956 * First release a slave and than destroy the bond if no more slaves are left.
1957 * Must be under rtnl_lock when this function is called.
1959 int bond_release_and_destroy(struct net_device
*bond_dev
,
1960 struct net_device
*slave_dev
)
1962 struct bonding
*bond
= netdev_priv(bond_dev
);
1965 ret
= bond_release(bond_dev
, slave_dev
);
1966 if ((ret
== 0) && (bond
->slave_cnt
== 0)) {
1967 pr_info("%s: destroying bond %s.\n",
1968 bond_dev
->name
, bond_dev
->name
);
1969 unregister_netdevice(bond_dev
);
1975 * This function releases all slaves.
1977 static int bond_release_all(struct net_device
*bond_dev
)
1979 struct bonding
*bond
= netdev_priv(bond_dev
);
1980 struct slave
*slave
;
1981 struct net_device
*slave_dev
;
1982 struct sockaddr addr
;
1984 write_lock_bh(&bond
->lock
);
1986 netif_carrier_off(bond_dev
);
1988 if (bond
->slave_cnt
== 0)
1991 bond
->current_arp_slave
= NULL
;
1992 bond
->primary_slave
= NULL
;
1993 bond_change_active_slave(bond
, NULL
);
1995 while ((slave
= bond
->first_slave
) != NULL
) {
1996 /* Inform AD package of unbinding of slave
1997 * before slave is detached from the list.
1999 if (bond
->params
.mode
== BOND_MODE_8023AD
)
2000 bond_3ad_unbind_slave(slave
);
2002 slave_dev
= slave
->dev
;
2003 bond_detach_slave(bond
, slave
);
2005 /* now that the slave is detached, unlock and perform
2006 * all the undo steps that should not be called from
2009 write_unlock_bh(&bond
->lock
);
2011 if (bond_is_lb(bond
)) {
2012 /* must be called only after the slave
2013 * has been detached from the list
2015 bond_alb_deinit_slave(bond
, slave
);
2018 bond_compute_features(bond
);
2020 bond_destroy_slave_symlinks(bond_dev
, slave_dev
);
2021 bond_del_vlans_from_slave(bond
, slave_dev
);
2023 /* If the mode USES_PRIMARY, then we should only remove its
2024 * promisc and mc settings if it was the curr_active_slave, but that was
2025 * already taken care of above when we detached the slave
2027 if (!USES_PRIMARY(bond
->params
.mode
)) {
2028 /* unset promiscuity level from slave */
2029 if (bond_dev
->flags
& IFF_PROMISC
)
2030 dev_set_promiscuity(slave_dev
, -1);
2032 /* unset allmulti level from slave */
2033 if (bond_dev
->flags
& IFF_ALLMULTI
)
2034 dev_set_allmulti(slave_dev
, -1);
2036 /* flush master's mc_list from slave */
2037 netif_addr_lock_bh(bond_dev
);
2038 bond_mc_list_flush(bond_dev
, slave_dev
);
2039 netif_addr_unlock_bh(bond_dev
);
2042 netdev_set_master(slave_dev
, NULL
);
2044 /* close slave before restoring its mac address */
2045 dev_close(slave_dev
);
2047 if (!bond
->params
.fail_over_mac
) {
2048 /* restore original ("permanent") mac address*/
2049 memcpy(addr
.sa_data
, slave
->perm_hwaddr
, ETH_ALEN
);
2050 addr
.sa_family
= slave_dev
->type
;
2051 dev_set_mac_address(slave_dev
, &addr
);
2054 slave_dev
->priv_flags
&= ~(IFF_MASTER_8023AD
| IFF_MASTER_ALB
|
2055 IFF_SLAVE_INACTIVE
);
2059 /* re-acquire the lock before getting the next slave */
2060 write_lock_bh(&bond
->lock
);
2063 /* zero the mac address of the master so it will be
2064 * set by the application to the mac address of the
2067 memset(bond_dev
->dev_addr
, 0, bond_dev
->addr_len
);
2069 if (list_empty(&bond
->vlan_list
))
2070 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
2072 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
2073 bond_dev
->name
, bond_dev
->name
);
2074 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
2078 pr_info("%s: released all slaves\n", bond_dev
->name
);
2081 write_unlock_bh(&bond
->lock
);
2087 * This function changes the active slave to slave <slave_dev>.
2088 * It returns -EINVAL in the following cases.
2089 * - <slave_dev> is not found in the list.
2090 * - There is not active slave now.
2091 * - <slave_dev> is already active.
2092 * - The link state of <slave_dev> is not BOND_LINK_UP.
2093 * - <slave_dev> is not running.
2094 * In these cases, this function does nothing.
2095 * In the other cases, current_slave pointer is changed and 0 is returned.
2097 static int bond_ioctl_change_active(struct net_device
*bond_dev
, struct net_device
*slave_dev
)
2099 struct bonding
*bond
= netdev_priv(bond_dev
);
2100 struct slave
*old_active
= NULL
;
2101 struct slave
*new_active
= NULL
;
2104 if (!USES_PRIMARY(bond
->params
.mode
))
2107 /* Verify that master_dev is indeed the master of slave_dev */
2108 if (!(slave_dev
->flags
& IFF_SLAVE
) || (slave_dev
->master
!= bond_dev
))
2111 read_lock(&bond
->lock
);
2113 read_lock(&bond
->curr_slave_lock
);
2114 old_active
= bond
->curr_active_slave
;
2115 read_unlock(&bond
->curr_slave_lock
);
2117 new_active
= bond_get_slave_by_dev(bond
, slave_dev
);
2120 * Changing to the current active: do nothing; return success.
2122 if (new_active
&& (new_active
== old_active
)) {
2123 read_unlock(&bond
->lock
);
2129 (new_active
->link
== BOND_LINK_UP
) &&
2130 IS_UP(new_active
->dev
)) {
2131 write_lock_bh(&bond
->curr_slave_lock
);
2132 bond_change_active_slave(bond
, new_active
);
2133 write_unlock_bh(&bond
->curr_slave_lock
);
2137 read_unlock(&bond
->lock
);
2142 static int bond_info_query(struct net_device
*bond_dev
, struct ifbond
*info
)
2144 struct bonding
*bond
= netdev_priv(bond_dev
);
2146 info
->bond_mode
= bond
->params
.mode
;
2147 info
->miimon
= bond
->params
.miimon
;
2149 read_lock(&bond
->lock
);
2150 info
->num_slaves
= bond
->slave_cnt
;
2151 read_unlock(&bond
->lock
);
2156 static int bond_slave_info_query(struct net_device
*bond_dev
, struct ifslave
*info
)
2158 struct bonding
*bond
= netdev_priv(bond_dev
);
2159 struct slave
*slave
;
2160 int i
, res
= -ENODEV
;
2162 read_lock(&bond
->lock
);
2164 bond_for_each_slave(bond
, slave
, i
) {
2165 if (i
== (int)info
->slave_id
) {
2167 strcpy(info
->slave_name
, slave
->dev
->name
);
2168 info
->link
= slave
->link
;
2169 info
->state
= slave
->state
;
2170 info
->link_failure_count
= slave
->link_failure_count
;
2175 read_unlock(&bond
->lock
);
2180 /*-------------------------------- Monitoring -------------------------------*/
2183 static int bond_miimon_inspect(struct bonding
*bond
)
2185 struct slave
*slave
;
2186 int i
, link_state
, commit
= 0;
2187 bool ignore_updelay
;
2189 ignore_updelay
= !bond
->curr_active_slave
? true : false;
2191 bond_for_each_slave(bond
, slave
, i
) {
2192 slave
->new_link
= BOND_LINK_NOCHANGE
;
2194 link_state
= bond_check_dev_link(bond
, slave
->dev
, 0);
2196 switch (slave
->link
) {
2201 slave
->link
= BOND_LINK_FAIL
;
2202 slave
->delay
= bond
->params
.downdelay
;
2204 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n",
2206 (bond
->params
.mode
==
2207 BOND_MODE_ACTIVEBACKUP
) ?
2208 ((slave
->state
== BOND_STATE_ACTIVE
) ?
2209 "active " : "backup ") : "",
2211 bond
->params
.downdelay
* bond
->params
.miimon
);
2214 case BOND_LINK_FAIL
:
2217 * recovered before downdelay expired
2219 slave
->link
= BOND_LINK_UP
;
2220 slave
->jiffies
= jiffies
;
2221 pr_info("%s: link status up again after %d ms for interface %s.\n",
2223 (bond
->params
.downdelay
- slave
->delay
) *
2224 bond
->params
.miimon
,
2229 if (slave
->delay
<= 0) {
2230 slave
->new_link
= BOND_LINK_DOWN
;
2238 case BOND_LINK_DOWN
:
2242 slave
->link
= BOND_LINK_BACK
;
2243 slave
->delay
= bond
->params
.updelay
;
2246 pr_info("%s: link status up for interface %s, enabling it in %d ms.\n",
2247 bond
->dev
->name
, slave
->dev
->name
,
2248 ignore_updelay
? 0 :
2249 bond
->params
.updelay
*
2250 bond
->params
.miimon
);
2253 case BOND_LINK_BACK
:
2255 slave
->link
= BOND_LINK_DOWN
;
2256 pr_info("%s: link status down again after %d ms for interface %s.\n",
2258 (bond
->params
.updelay
- slave
->delay
) *
2259 bond
->params
.miimon
,
2268 if (slave
->delay
<= 0) {
2269 slave
->new_link
= BOND_LINK_UP
;
2271 ignore_updelay
= false;
2283 static void bond_miimon_commit(struct bonding
*bond
)
2285 struct slave
*slave
;
2288 bond_for_each_slave(bond
, slave
, i
) {
2289 switch (slave
->new_link
) {
2290 case BOND_LINK_NOCHANGE
:
2294 slave
->link
= BOND_LINK_UP
;
2295 slave
->jiffies
= jiffies
;
2297 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
2298 /* prevent it from being the active one */
2299 slave
->state
= BOND_STATE_BACKUP
;
2300 } else if (bond
->params
.mode
!= BOND_MODE_ACTIVEBACKUP
) {
2301 /* make it immediately active */
2302 slave
->state
= BOND_STATE_ACTIVE
;
2303 } else if (slave
!= bond
->primary_slave
) {
2304 /* prevent it from being the active one */
2305 slave
->state
= BOND_STATE_BACKUP
;
2308 pr_info("%s: link status definitely up for interface %s.\n",
2309 bond
->dev
->name
, slave
->dev
->name
);
2311 /* notify ad that the link status has changed */
2312 if (bond
->params
.mode
== BOND_MODE_8023AD
)
2313 bond_3ad_handle_link_change(slave
, BOND_LINK_UP
);
2315 if (bond_is_lb(bond
))
2316 bond_alb_handle_link_change(bond
, slave
,
2319 if (!bond
->curr_active_slave
||
2320 (slave
== bond
->primary_slave
))
2325 case BOND_LINK_DOWN
:
2326 if (slave
->link_failure_count
< UINT_MAX
)
2327 slave
->link_failure_count
++;
2329 slave
->link
= BOND_LINK_DOWN
;
2331 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
||
2332 bond
->params
.mode
== BOND_MODE_8023AD
)
2333 bond_set_slave_inactive_flags(slave
);
2335 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2336 bond
->dev
->name
, slave
->dev
->name
);
2338 if (bond
->params
.mode
== BOND_MODE_8023AD
)
2339 bond_3ad_handle_link_change(slave
,
2342 if (bond_is_lb(bond
))
2343 bond_alb_handle_link_change(bond
, slave
,
2346 if (slave
== bond
->curr_active_slave
)
2352 pr_err("%s: invalid new link %d on slave %s\n",
2353 bond
->dev
->name
, slave
->new_link
,
2355 slave
->new_link
= BOND_LINK_NOCHANGE
;
2362 write_lock_bh(&bond
->curr_slave_lock
);
2363 bond_select_active_slave(bond
);
2364 write_unlock_bh(&bond
->curr_slave_lock
);
2367 bond_set_carrier(bond
);
2373 * Really a wrapper that splits the mii monitor into two phases: an
2374 * inspection, then (if inspection indicates something needs to be done)
2375 * an acquisition of appropriate locks followed by a commit phase to
2376 * implement whatever link state changes are indicated.
2378 void bond_mii_monitor(struct work_struct
*work
)
2380 struct bonding
*bond
= container_of(work
, struct bonding
,
2383 read_lock(&bond
->lock
);
2384 if (bond
->kill_timers
)
2387 if (bond
->slave_cnt
== 0)
2390 if (bond
->send_grat_arp
) {
2391 read_lock(&bond
->curr_slave_lock
);
2392 bond_send_gratuitous_arp(bond
);
2393 read_unlock(&bond
->curr_slave_lock
);
2396 if (bond
->send_unsol_na
) {
2397 read_lock(&bond
->curr_slave_lock
);
2398 bond_send_unsolicited_na(bond
);
2399 read_unlock(&bond
->curr_slave_lock
);
2402 if (bond_miimon_inspect(bond
)) {
2403 read_unlock(&bond
->lock
);
2405 read_lock(&bond
->lock
);
2407 bond_miimon_commit(bond
);
2409 read_unlock(&bond
->lock
);
2410 rtnl_unlock(); /* might sleep, hold no other locks */
2411 read_lock(&bond
->lock
);
2415 if (bond
->params
.miimon
)
2416 queue_delayed_work(bond
->wq
, &bond
->mii_work
,
2417 msecs_to_jiffies(bond
->params
.miimon
));
2419 read_unlock(&bond
->lock
);
2422 static __be32
bond_glean_dev_ip(struct net_device
*dev
)
2424 struct in_device
*idev
;
2425 struct in_ifaddr
*ifa
;
2432 idev
= __in_dev_get_rcu(dev
);
2436 ifa
= idev
->ifa_list
;
2440 addr
= ifa
->ifa_local
;
2446 static int bond_has_this_ip(struct bonding
*bond
, __be32 ip
)
2448 struct vlan_entry
*vlan
;
2450 if (ip
== bond
->master_ip
)
2453 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
2454 if (ip
== vlan
->vlan_ip
)
2462 * We go to the (large) trouble of VLAN tagging ARP frames because
2463 * switches in VLAN mode (especially if ports are configured as
2464 * "native" to a VLAN) might not pass non-tagged frames.
2466 static void bond_arp_send(struct net_device
*slave_dev
, int arp_op
, __be32 dest_ip
, __be32 src_ip
, unsigned short vlan_id
)
2468 struct sk_buff
*skb
;
2470 pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op
,
2471 slave_dev
->name
, dest_ip
, src_ip
, vlan_id
);
2473 skb
= arp_create(arp_op
, ETH_P_ARP
, dest_ip
, slave_dev
, src_ip
,
2474 NULL
, slave_dev
->dev_addr
, NULL
);
2477 pr_err("ARP packet allocation failed\n");
2481 skb
= vlan_put_tag(skb
, vlan_id
);
2483 pr_err("failed to insert VLAN tag\n");
2491 static void bond_arp_send_all(struct bonding
*bond
, struct slave
*slave
)
2494 __be32
*targets
= bond
->params
.arp_targets
;
2495 struct vlan_entry
*vlan
;
2496 struct net_device
*vlan_dev
;
2500 for (i
= 0; (i
< BOND_MAX_ARP_TARGETS
); i
++) {
2503 pr_debug("basa: target %x\n", targets
[i
]);
2504 if (list_empty(&bond
->vlan_list
)) {
2505 pr_debug("basa: empty vlan: arp_send\n");
2506 bond_arp_send(slave
->dev
, ARPOP_REQUEST
, targets
[i
],
2507 bond
->master_ip
, 0);
2512 * If VLANs are configured, we do a route lookup to
2513 * determine which VLAN interface would be used, so we
2514 * can tag the ARP with the proper VLAN tag.
2516 memset(&fl
, 0, sizeof(fl
));
2517 fl
.fl4_dst
= targets
[i
];
2518 fl
.fl4_tos
= RTO_ONLINK
;
2520 rv
= ip_route_output_key(dev_net(bond
->dev
), &rt
, &fl
);
2522 if (net_ratelimit()) {
2523 pr_warning("%s: no route to arp_ip_target %pI4\n",
2524 bond
->dev
->name
, &fl
.fl4_dst
);
2530 * This target is not on a VLAN
2532 if (rt
->u
.dst
.dev
== bond
->dev
) {
2534 pr_debug("basa: rtdev == bond->dev: arp_send\n");
2535 bond_arp_send(slave
->dev
, ARPOP_REQUEST
, targets
[i
],
2536 bond
->master_ip
, 0);
2541 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
2542 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
2543 if (vlan_dev
== rt
->u
.dst
.dev
) {
2544 vlan_id
= vlan
->vlan_id
;
2545 pr_debug("basa: vlan match on %s %d\n",
2546 vlan_dev
->name
, vlan_id
);
2553 bond_arp_send(slave
->dev
, ARPOP_REQUEST
, targets
[i
],
2554 vlan
->vlan_ip
, vlan_id
);
2558 if (net_ratelimit()) {
2559 pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2560 bond
->dev
->name
, &fl
.fl4_dst
,
2561 rt
->u
.dst
.dev
? rt
->u
.dst
.dev
->name
: "NULL");
2568 * Kick out a gratuitous ARP for an IP on the bonding master plus one
2569 * for each VLAN above us.
2571 * Caller must hold curr_slave_lock for read or better
2573 static void bond_send_gratuitous_arp(struct bonding
*bond
)
2575 struct slave
*slave
= bond
->curr_active_slave
;
2576 struct vlan_entry
*vlan
;
2577 struct net_device
*vlan_dev
;
2579 pr_debug("bond_send_grat_arp: bond %s slave %s\n",
2580 bond
->dev
->name
, slave
? slave
->dev
->name
: "NULL");
2582 if (!slave
|| !bond
->send_grat_arp
||
2583 test_bit(__LINK_STATE_LINKWATCH_PENDING
, &slave
->dev
->state
))
2586 bond
->send_grat_arp
--;
2588 if (bond
->master_ip
) {
2589 bond_arp_send(slave
->dev
, ARPOP_REPLY
, bond
->master_ip
,
2590 bond
->master_ip
, 0);
2593 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
2594 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
2595 if (vlan
->vlan_ip
) {
2596 bond_arp_send(slave
->dev
, ARPOP_REPLY
, vlan
->vlan_ip
,
2597 vlan
->vlan_ip
, vlan
->vlan_id
);
2602 static void bond_validate_arp(struct bonding
*bond
, struct slave
*slave
, __be32 sip
, __be32 tip
)
2605 __be32
*targets
= bond
->params
.arp_targets
;
2607 for (i
= 0; (i
< BOND_MAX_ARP_TARGETS
) && targets
[i
]; i
++) {
2608 pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n",
2609 &sip
, &tip
, i
, &targets
[i
],
2610 bond_has_this_ip(bond
, tip
));
2611 if (sip
== targets
[i
]) {
2612 if (bond_has_this_ip(bond
, tip
))
2613 slave
->last_arp_rx
= jiffies
;
2619 static int bond_arp_rcv(struct sk_buff
*skb
, struct net_device
*dev
, struct packet_type
*pt
, struct net_device
*orig_dev
)
2622 struct slave
*slave
;
2623 struct bonding
*bond
;
2624 unsigned char *arp_ptr
;
2627 if (dev
->priv_flags
& IFF_802_1Q_VLAN
) {
2629 * When using VLANS and bonding, dev and oriv_dev may be
2630 * incorrect if the physical interface supports VLAN
2631 * acceleration. With this change ARP validation now
2632 * works for hosts only reachable on the VLAN interface.
2634 dev
= vlan_dev_real_dev(dev
);
2635 orig_dev
= dev_get_by_index_rcu(dev_net(skb
->dev
),skb
->skb_iif
);
2638 if (!(dev
->priv_flags
& IFF_BONDING
) || !(dev
->flags
& IFF_MASTER
))
2641 bond
= netdev_priv(dev
);
2642 read_lock(&bond
->lock
);
2644 pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n",
2645 bond
->dev
->name
, skb
->dev
? skb
->dev
->name
: "NULL",
2646 orig_dev
? orig_dev
->name
: "NULL");
2648 slave
= bond_get_slave_by_dev(bond
, orig_dev
);
2649 if (!slave
|| !slave_do_arp_validate(bond
, slave
))
2652 if (!pskb_may_pull(skb
, arp_hdr_len(dev
)))
2656 if (arp
->ar_hln
!= dev
->addr_len
||
2657 skb
->pkt_type
== PACKET_OTHERHOST
||
2658 skb
->pkt_type
== PACKET_LOOPBACK
||
2659 arp
->ar_hrd
!= htons(ARPHRD_ETHER
) ||
2660 arp
->ar_pro
!= htons(ETH_P_IP
) ||
2664 arp_ptr
= (unsigned char *)(arp
+ 1);
2665 arp_ptr
+= dev
->addr_len
;
2666 memcpy(&sip
, arp_ptr
, 4);
2667 arp_ptr
+= 4 + dev
->addr_len
;
2668 memcpy(&tip
, arp_ptr
, 4);
2670 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
2671 bond
->dev
->name
, slave
->dev
->name
, slave
->state
,
2672 bond
->params
.arp_validate
, slave_do_arp_validate(bond
, slave
),
2676 * Backup slaves won't see the ARP reply, but do come through
2677 * here for each ARP probe (so we swap the sip/tip to validate
2678 * the probe). In a "redundant switch, common router" type of
2679 * configuration, the ARP probe will (hopefully) travel from
2680 * the active, through one switch, the router, then the other
2681 * switch before reaching the backup.
2683 if (slave
->state
== BOND_STATE_ACTIVE
)
2684 bond_validate_arp(bond
, slave
, sip
, tip
);
2686 bond_validate_arp(bond
, slave
, tip
, sip
);
2689 read_unlock(&bond
->lock
);
2692 return NET_RX_SUCCESS
;
2696 * this function is called regularly to monitor each slave's link
2697 * ensuring that traffic is being sent and received when arp monitoring
2698 * is used in load-balancing mode. if the adapter has been dormant, then an
2699 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2700 * arp monitoring in active backup mode.
2702 void bond_loadbalance_arp_mon(struct work_struct
*work
)
2704 struct bonding
*bond
= container_of(work
, struct bonding
,
2706 struct slave
*slave
, *oldcurrent
;
2707 int do_failover
= 0;
2711 read_lock(&bond
->lock
);
2713 delta_in_ticks
= msecs_to_jiffies(bond
->params
.arp_interval
);
2715 if (bond
->kill_timers
)
2718 if (bond
->slave_cnt
== 0)
2721 read_lock(&bond
->curr_slave_lock
);
2722 oldcurrent
= bond
->curr_active_slave
;
2723 read_unlock(&bond
->curr_slave_lock
);
2725 /* see if any of the previous devices are up now (i.e. they have
2726 * xmt and rcv traffic). the curr_active_slave does not come into
2727 * the picture unless it is null. also, slave->jiffies is not needed
2728 * here because we send an arp on each slave and give a slave as
2729 * long as it needs to get the tx/rx within the delta.
2730 * TODO: what about up/down delay in arp mode? it wasn't here before
2733 bond_for_each_slave(bond
, slave
, i
) {
2734 if (slave
->link
!= BOND_LINK_UP
) {
2735 if (time_before_eq(jiffies
, dev_trans_start(slave
->dev
) + delta_in_ticks
) &&
2736 time_before_eq(jiffies
, slave
->dev
->last_rx
+ delta_in_ticks
)) {
2738 slave
->link
= BOND_LINK_UP
;
2739 slave
->state
= BOND_STATE_ACTIVE
;
2741 /* primary_slave has no meaning in round-robin
2742 * mode. the window of a slave being up and
2743 * curr_active_slave being null after enslaving
2747 pr_info("%s: link status definitely up for interface %s, ",
2752 pr_info("%s: interface %s is now up\n",
2758 /* slave->link == BOND_LINK_UP */
2760 /* not all switches will respond to an arp request
2761 * when the source ip is 0, so don't take the link down
2762 * if we don't know our ip yet
2764 if (time_after_eq(jiffies
, dev_trans_start(slave
->dev
) + 2*delta_in_ticks
) ||
2765 (time_after_eq(jiffies
, slave
->dev
->last_rx
+ 2*delta_in_ticks
))) {
2767 slave
->link
= BOND_LINK_DOWN
;
2768 slave
->state
= BOND_STATE_BACKUP
;
2770 if (slave
->link_failure_count
< UINT_MAX
)
2771 slave
->link_failure_count
++;
2773 pr_info("%s: interface %s is now down.\n",
2777 if (slave
== oldcurrent
)
2782 /* note: if switch is in round-robin mode, all links
2783 * must tx arp to ensure all links rx an arp - otherwise
2784 * links may oscillate or not come up at all; if switch is
2785 * in something like xor mode, there is nothing we can
2786 * do - all replies will be rx'ed on same link causing slaves
2787 * to be unstable during low/no traffic periods
2789 if (IS_UP(slave
->dev
))
2790 bond_arp_send_all(bond
, slave
);
2794 write_lock_bh(&bond
->curr_slave_lock
);
2796 bond_select_active_slave(bond
);
2798 write_unlock_bh(&bond
->curr_slave_lock
);
2802 if (bond
->params
.arp_interval
)
2803 queue_delayed_work(bond
->wq
, &bond
->arp_work
, delta_in_ticks
);
2805 read_unlock(&bond
->lock
);
2809 * Called to inspect slaves for active-backup mode ARP monitor link state
2810 * changes. Sets new_link in slaves to specify what action should take
2811 * place for the slave. Returns 0 if no changes are found, >0 if changes
2812 * to link states must be committed.
2814 * Called with bond->lock held for read.
2816 static int bond_ab_arp_inspect(struct bonding
*bond
, int delta_in_ticks
)
2818 struct slave
*slave
;
2821 bond_for_each_slave(bond
, slave
, i
) {
2822 slave
->new_link
= BOND_LINK_NOCHANGE
;
2824 if (slave
->link
!= BOND_LINK_UP
) {
2825 if (time_before_eq(jiffies
, slave_last_rx(bond
, slave
) +
2827 slave
->new_link
= BOND_LINK_UP
;
2835 * Give slaves 2*delta after being enslaved or made
2836 * active. This avoids bouncing, as the last receive
2837 * times need a full ARP monitor cycle to be updated.
2839 if (!time_after_eq(jiffies
, slave
->jiffies
+
2840 2 * delta_in_ticks
))
2844 * Backup slave is down if:
2845 * - No current_arp_slave AND
2846 * - more than 3*delta since last receive AND
2847 * - the bond has an IP address
2849 * Note: a non-null current_arp_slave indicates
2850 * the curr_active_slave went down and we are
2851 * searching for a new one; under this condition
2852 * we only take the curr_active_slave down - this
2853 * gives each slave a chance to tx/rx traffic
2854 * before being taken out
2856 if (slave
->state
== BOND_STATE_BACKUP
&&
2857 !bond
->current_arp_slave
&&
2858 time_after(jiffies
, slave_last_rx(bond
, slave
) +
2859 3 * delta_in_ticks
)) {
2860 slave
->new_link
= BOND_LINK_DOWN
;
2865 * Active slave is down if:
2866 * - more than 2*delta since transmitting OR
2867 * - (more than 2*delta since receive AND
2868 * the bond has an IP address)
2870 if ((slave
->state
== BOND_STATE_ACTIVE
) &&
2871 (time_after_eq(jiffies
, dev_trans_start(slave
->dev
) +
2872 2 * delta_in_ticks
) ||
2873 (time_after_eq(jiffies
, slave_last_rx(bond
, slave
)
2874 + 2 * delta_in_ticks
)))) {
2875 slave
->new_link
= BOND_LINK_DOWN
;
2884 * Called to commit link state changes noted by inspection step of
2885 * active-backup mode ARP monitor.
2887 * Called with RTNL and bond->lock for read.
2889 static void bond_ab_arp_commit(struct bonding
*bond
, int delta_in_ticks
)
2891 struct slave
*slave
;
2894 bond_for_each_slave(bond
, slave
, i
) {
2895 switch (slave
->new_link
) {
2896 case BOND_LINK_NOCHANGE
:
2900 if ((!bond
->curr_active_slave
&&
2901 time_before_eq(jiffies
,
2902 dev_trans_start(slave
->dev
) +
2904 bond
->curr_active_slave
!= slave
) {
2905 slave
->link
= BOND_LINK_UP
;
2906 bond
->current_arp_slave
= NULL
;
2908 pr_info("%s: link status definitely up for interface %s.\n",
2909 bond
->dev
->name
, slave
->dev
->name
);
2911 if (!bond
->curr_active_slave
||
2912 (slave
== bond
->primary_slave
))
2919 case BOND_LINK_DOWN
:
2920 if (slave
->link_failure_count
< UINT_MAX
)
2921 slave
->link_failure_count
++;
2923 slave
->link
= BOND_LINK_DOWN
;
2924 bond_set_slave_inactive_flags(slave
);
2926 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2927 bond
->dev
->name
, slave
->dev
->name
);
2929 if (slave
== bond
->curr_active_slave
) {
2930 bond
->current_arp_slave
= NULL
;
2937 pr_err("%s: impossible: new_link %d on slave %s\n",
2938 bond
->dev
->name
, slave
->new_link
,
2945 write_lock_bh(&bond
->curr_slave_lock
);
2946 bond_select_active_slave(bond
);
2947 write_unlock_bh(&bond
->curr_slave_lock
);
2950 bond_set_carrier(bond
);
2954 * Send ARP probes for active-backup mode ARP monitor.
2956 * Called with bond->lock held for read.
2958 static void bond_ab_arp_probe(struct bonding
*bond
)
2960 struct slave
*slave
;
2963 read_lock(&bond
->curr_slave_lock
);
2965 if (bond
->current_arp_slave
&& bond
->curr_active_slave
)
2966 pr_info("PROBE: c_arp %s && cas %s BAD\n",
2967 bond
->current_arp_slave
->dev
->name
,
2968 bond
->curr_active_slave
->dev
->name
);
2970 if (bond
->curr_active_slave
) {
2971 bond_arp_send_all(bond
, bond
->curr_active_slave
);
2972 read_unlock(&bond
->curr_slave_lock
);
2976 read_unlock(&bond
->curr_slave_lock
);
2978 /* if we don't have a curr_active_slave, search for the next available
2979 * backup slave from the current_arp_slave and make it the candidate
2980 * for becoming the curr_active_slave
2983 if (!bond
->current_arp_slave
) {
2984 bond
->current_arp_slave
= bond
->first_slave
;
2985 if (!bond
->current_arp_slave
)
2989 bond_set_slave_inactive_flags(bond
->current_arp_slave
);
2991 /* search for next candidate */
2992 bond_for_each_slave_from(bond
, slave
, i
, bond
->current_arp_slave
->next
) {
2993 if (IS_UP(slave
->dev
)) {
2994 slave
->link
= BOND_LINK_BACK
;
2995 bond_set_slave_active_flags(slave
);
2996 bond_arp_send_all(bond
, slave
);
2997 slave
->jiffies
= jiffies
;
2998 bond
->current_arp_slave
= slave
;
3002 /* if the link state is up at this point, we
3003 * mark it down - this can happen if we have
3004 * simultaneous link failures and
3005 * reselect_active_interface doesn't make this
3006 * one the current slave so it is still marked
3007 * up when it is actually down
3009 if (slave
->link
== BOND_LINK_UP
) {
3010 slave
->link
= BOND_LINK_DOWN
;
3011 if (slave
->link_failure_count
< UINT_MAX
)
3012 slave
->link_failure_count
++;
3014 bond_set_slave_inactive_flags(slave
);
3016 pr_info("%s: backup interface %s is now down.\n",
3017 bond
->dev
->name
, slave
->dev
->name
);
3022 void bond_activebackup_arp_mon(struct work_struct
*work
)
3024 struct bonding
*bond
= container_of(work
, struct bonding
,
3028 read_lock(&bond
->lock
);
3030 if (bond
->kill_timers
)
3033 delta_in_ticks
= msecs_to_jiffies(bond
->params
.arp_interval
);
3035 if (bond
->slave_cnt
== 0)
3038 if (bond
->send_grat_arp
) {
3039 read_lock(&bond
->curr_slave_lock
);
3040 bond_send_gratuitous_arp(bond
);
3041 read_unlock(&bond
->curr_slave_lock
);
3044 if (bond
->send_unsol_na
) {
3045 read_lock(&bond
->curr_slave_lock
);
3046 bond_send_unsolicited_na(bond
);
3047 read_unlock(&bond
->curr_slave_lock
);
3050 if (bond_ab_arp_inspect(bond
, delta_in_ticks
)) {
3051 read_unlock(&bond
->lock
);
3053 read_lock(&bond
->lock
);
3055 bond_ab_arp_commit(bond
, delta_in_ticks
);
3057 read_unlock(&bond
->lock
);
3059 read_lock(&bond
->lock
);
3062 bond_ab_arp_probe(bond
);
3065 if (bond
->params
.arp_interval
)
3066 queue_delayed_work(bond
->wq
, &bond
->arp_work
, delta_in_ticks
);
3068 read_unlock(&bond
->lock
);
3071 /*------------------------------ proc/seq_file-------------------------------*/
3073 #ifdef CONFIG_PROC_FS
3075 static void *bond_info_seq_start(struct seq_file
*seq
, loff_t
*pos
)
3076 __acquires(&dev_base_lock
)
3077 __acquires(&bond
->lock
)
3079 struct bonding
*bond
= seq
->private;
3081 struct slave
*slave
;
3084 /* make sure the bond won't be taken away */
3085 read_lock(&dev_base_lock
);
3086 read_lock(&bond
->lock
);
3089 return SEQ_START_TOKEN
;
3091 bond_for_each_slave(bond
, slave
, i
) {
3099 static void *bond_info_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
3101 struct bonding
*bond
= seq
->private;
3102 struct slave
*slave
= v
;
3105 if (v
== SEQ_START_TOKEN
)
3106 return bond
->first_slave
;
3108 slave
= slave
->next
;
3110 return (slave
== bond
->first_slave
) ? NULL
: slave
;
3113 static void bond_info_seq_stop(struct seq_file
*seq
, void *v
)
3114 __releases(&bond
->lock
)
3115 __releases(&dev_base_lock
)
3117 struct bonding
*bond
= seq
->private;
3119 read_unlock(&bond
->lock
);
3120 read_unlock(&dev_base_lock
);
3123 static void bond_info_show_master(struct seq_file
*seq
)
3125 struct bonding
*bond
= seq
->private;
3129 read_lock(&bond
->curr_slave_lock
);
3130 curr
= bond
->curr_active_slave
;
3131 read_unlock(&bond
->curr_slave_lock
);
3133 seq_printf(seq
, "Bonding Mode: %s",
3134 bond_mode_name(bond
->params
.mode
));
3136 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
&&
3137 bond
->params
.fail_over_mac
)
3138 seq_printf(seq
, " (fail_over_mac %s)",
3139 fail_over_mac_tbl
[bond
->params
.fail_over_mac
].modename
);
3141 seq_printf(seq
, "\n");
3143 if (bond
->params
.mode
== BOND_MODE_XOR
||
3144 bond
->params
.mode
== BOND_MODE_8023AD
) {
3145 seq_printf(seq
, "Transmit Hash Policy: %s (%d)\n",
3146 xmit_hashtype_tbl
[bond
->params
.xmit_policy
].modename
,
3147 bond
->params
.xmit_policy
);
3150 if (USES_PRIMARY(bond
->params
.mode
)) {
3151 seq_printf(seq
, "Primary Slave: %s",
3152 (bond
->primary_slave
) ?
3153 bond
->primary_slave
->dev
->name
: "None");
3154 if (bond
->primary_slave
)
3155 seq_printf(seq
, " (primary_reselect %s)",
3156 pri_reselect_tbl
[bond
->params
.primary_reselect
].modename
);
3158 seq_printf(seq
, "\nCurrently Active Slave: %s\n",
3159 (curr
) ? curr
->dev
->name
: "None");
3162 seq_printf(seq
, "MII Status: %s\n", netif_carrier_ok(bond
->dev
) ?
3164 seq_printf(seq
, "MII Polling Interval (ms): %d\n", bond
->params
.miimon
);
3165 seq_printf(seq
, "Up Delay (ms): %d\n",
3166 bond
->params
.updelay
* bond
->params
.miimon
);
3167 seq_printf(seq
, "Down Delay (ms): %d\n",
3168 bond
->params
.downdelay
* bond
->params
.miimon
);
3171 /* ARP information */
3172 if (bond
->params
.arp_interval
> 0) {
3174 seq_printf(seq
, "ARP Polling Interval (ms): %d\n",
3175 bond
->params
.arp_interval
);
3177 seq_printf(seq
, "ARP IP target/s (n.n.n.n form):");
3179 for (i
= 0; (i
< BOND_MAX_ARP_TARGETS
); i
++) {
3180 if (!bond
->params
.arp_targets
[i
])
3183 seq_printf(seq
, ",");
3184 seq_printf(seq
, " %pI4", &bond
->params
.arp_targets
[i
]);
3187 seq_printf(seq
, "\n");
3190 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3191 struct ad_info ad_info
;
3193 seq_puts(seq
, "\n802.3ad info\n");
3194 seq_printf(seq
, "LACP rate: %s\n",
3195 (bond
->params
.lacp_fast
) ? "fast" : "slow");
3196 seq_printf(seq
, "Aggregator selection policy (ad_select): %s\n",
3197 ad_select_tbl
[bond
->params
.ad_select
].modename
);
3199 if (bond_3ad_get_active_agg_info(bond
, &ad_info
)) {
3200 seq_printf(seq
, "bond %s has no active aggregator\n",
3203 seq_printf(seq
, "Active Aggregator Info:\n");
3205 seq_printf(seq
, "\tAggregator ID: %d\n",
3206 ad_info
.aggregator_id
);
3207 seq_printf(seq
, "\tNumber of ports: %d\n",
3209 seq_printf(seq
, "\tActor Key: %d\n",
3211 seq_printf(seq
, "\tPartner Key: %d\n",
3212 ad_info
.partner_key
);
3213 seq_printf(seq
, "\tPartner Mac Address: %pM\n",
3214 ad_info
.partner_system
);
3219 static void bond_info_show_slave(struct seq_file
*seq
,
3220 const struct slave
*slave
)
3222 struct bonding
*bond
= seq
->private;
3224 seq_printf(seq
, "\nSlave Interface: %s\n", slave
->dev
->name
);
3225 seq_printf(seq
, "MII Status: %s\n",
3226 (slave
->link
== BOND_LINK_UP
) ? "up" : "down");
3227 seq_printf(seq
, "Link Failure Count: %u\n",
3228 slave
->link_failure_count
);
3230 seq_printf(seq
, "Permanent HW addr: %pM\n", slave
->perm_hwaddr
);
3232 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3233 const struct aggregator
*agg
3234 = SLAVE_AD_INFO(slave
).port
.aggregator
;
3237 seq_printf(seq
, "Aggregator ID: %d\n",
3238 agg
->aggregator_identifier
);
3240 seq_puts(seq
, "Aggregator ID: N/A\n");
3244 static int bond_info_seq_show(struct seq_file
*seq
, void *v
)
3246 if (v
== SEQ_START_TOKEN
) {
3247 seq_printf(seq
, "%s\n", version
);
3248 bond_info_show_master(seq
);
3250 bond_info_show_slave(seq
, v
);
3255 static const struct seq_operations bond_info_seq_ops
= {
3256 .start
= bond_info_seq_start
,
3257 .next
= bond_info_seq_next
,
3258 .stop
= bond_info_seq_stop
,
3259 .show
= bond_info_seq_show
,
3262 static int bond_info_open(struct inode
*inode
, struct file
*file
)
3264 struct seq_file
*seq
;
3265 struct proc_dir_entry
*proc
;
3268 res
= seq_open(file
, &bond_info_seq_ops
);
3270 /* recover the pointer buried in proc_dir_entry data */
3271 seq
= file
->private_data
;
3273 seq
->private = proc
->data
;
3279 static const struct file_operations bond_info_fops
= {
3280 .owner
= THIS_MODULE
,
3281 .open
= bond_info_open
,
3283 .llseek
= seq_lseek
,
3284 .release
= seq_release
,
3287 static void bond_create_proc_entry(struct bonding
*bond
)
3289 struct net_device
*bond_dev
= bond
->dev
;
3290 struct bond_net
*bn
= net_generic(dev_net(bond_dev
), bond_net_id
);
3293 bond
->proc_entry
= proc_create_data(bond_dev
->name
,
3294 S_IRUGO
, bn
->proc_dir
,
3295 &bond_info_fops
, bond
);
3296 if (bond
->proc_entry
== NULL
)
3297 pr_warning("Warning: Cannot create /proc/net/%s/%s\n",
3298 DRV_NAME
, bond_dev
->name
);
3300 memcpy(bond
->proc_file_name
, bond_dev
->name
, IFNAMSIZ
);
3304 static void bond_remove_proc_entry(struct bonding
*bond
)
3306 struct net_device
*bond_dev
= bond
->dev
;
3307 struct bond_net
*bn
= net_generic(dev_net(bond_dev
), bond_net_id
);
3309 if (bn
->proc_dir
&& bond
->proc_entry
) {
3310 remove_proc_entry(bond
->proc_file_name
, bn
->proc_dir
);
3311 memset(bond
->proc_file_name
, 0, IFNAMSIZ
);
3312 bond
->proc_entry
= NULL
;
3316 /* Create the bonding directory under /proc/net, if doesn't exist yet.
3317 * Caller must hold rtnl_lock.
3319 static void __net_init
bond_create_proc_dir(struct bond_net
*bn
)
3321 if (!bn
->proc_dir
) {
3322 bn
->proc_dir
= proc_mkdir(DRV_NAME
, bn
->net
->proc_net
);
3324 pr_warning("Warning: cannot create /proc/net/%s\n",
3329 /* Destroy the bonding directory under /proc/net, if empty.
3330 * Caller must hold rtnl_lock.
3332 static void __net_exit
bond_destroy_proc_dir(struct bond_net
*bn
)
3335 remove_proc_entry(DRV_NAME
, bn
->net
->proc_net
);
3336 bn
->proc_dir
= NULL
;
3340 #else /* !CONFIG_PROC_FS */
3342 static void bond_create_proc_entry(struct bonding
*bond
)
3346 static void bond_remove_proc_entry(struct bonding
*bond
)
3350 static inline void bond_create_proc_dir(struct bond_net
*bn
)
3354 static inline void bond_destroy_proc_dir(struct bond_net
*bn
)
3358 #endif /* CONFIG_PROC_FS */
3361 /*-------------------------- netdev event handling --------------------------*/
3364 * Change device name
3366 static int bond_event_changename(struct bonding
*bond
)
3368 bond_remove_proc_entry(bond
);
3369 bond_create_proc_entry(bond
);
3374 static int bond_master_netdev_event(unsigned long event
,
3375 struct net_device
*bond_dev
)
3377 struct bonding
*event_bond
= netdev_priv(bond_dev
);
3380 case NETDEV_CHANGENAME
:
3381 return bond_event_changename(event_bond
);
3389 static int bond_slave_netdev_event(unsigned long event
,
3390 struct net_device
*slave_dev
)
3392 struct net_device
*bond_dev
= slave_dev
->master
;
3393 struct bonding
*bond
= netdev_priv(bond_dev
);
3396 case NETDEV_UNREGISTER
:
3398 if (bond
->setup_by_slave
)
3399 bond_release_and_destroy(bond_dev
, slave_dev
);
3401 bond_release(bond_dev
, slave_dev
);
3405 if (bond
->params
.mode
== BOND_MODE_8023AD
|| bond_is_lb(bond
)) {
3406 struct slave
*slave
;
3408 slave
= bond_get_slave_by_dev(bond
, slave_dev
);
3410 u16 old_speed
= slave
->speed
;
3411 u16 old_duplex
= slave
->duplex
;
3413 bond_update_speed_duplex(slave
);
3415 if (bond_is_lb(bond
))
3418 if (old_speed
!= slave
->speed
)
3419 bond_3ad_adapter_speed_changed(slave
);
3420 if (old_duplex
!= slave
->duplex
)
3421 bond_3ad_adapter_duplex_changed(slave
);
3428 * ... Or is it this?
3431 case NETDEV_CHANGEMTU
:
3433 * TODO: Should slaves be allowed to
3434 * independently alter their MTU? For
3435 * an active-backup bond, slaves need
3436 * not be the same type of device, so
3437 * MTUs may vary. For other modes,
3438 * slaves arguably should have the
3439 * same MTUs. To do this, we'd need to
3440 * take over the slave's change_mtu
3441 * function for the duration of their
3445 case NETDEV_CHANGENAME
:
3447 * TODO: handle changing the primary's name
3450 case NETDEV_FEAT_CHANGE
:
3451 bond_compute_features(bond
);
3461 * bond_netdev_event: handle netdev notifier chain events.
3463 * This function receives events for the netdev chain. The caller (an
3464 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
3465 * locks for us to safely manipulate the slave devices (RTNL lock,
3468 static int bond_netdev_event(struct notifier_block
*this,
3469 unsigned long event
, void *ptr
)
3471 struct net_device
*event_dev
= (struct net_device
*)ptr
;
3473 pr_debug("event_dev: %s, event: %lx\n",
3474 event_dev
? event_dev
->name
: "None",
3477 if (!(event_dev
->priv_flags
& IFF_BONDING
))
3480 if (event_dev
->flags
& IFF_MASTER
) {
3481 pr_debug("IFF_MASTER\n");
3482 return bond_master_netdev_event(event
, event_dev
);
3485 if (event_dev
->flags
& IFF_SLAVE
) {
3486 pr_debug("IFF_SLAVE\n");
3487 return bond_slave_netdev_event(event
, event_dev
);
3494 * bond_inetaddr_event: handle inetaddr notifier chain events.
3496 * We keep track of device IPs primarily to use as source addresses in
3497 * ARP monitor probes (rather than spewing out broadcasts all the time).
3499 * We track one IP for the main device (if it has one), plus one per VLAN.
3501 static int bond_inetaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
3503 struct in_ifaddr
*ifa
= ptr
;
3504 struct net_device
*vlan_dev
, *event_dev
= ifa
->ifa_dev
->dev
;
3505 struct bond_net
*bn
= net_generic(dev_net(event_dev
), bond_net_id
);
3506 struct bonding
*bond
;
3507 struct vlan_entry
*vlan
;
3509 list_for_each_entry(bond
, &bn
->dev_list
, bond_list
) {
3510 if (bond
->dev
== event_dev
) {
3513 bond
->master_ip
= ifa
->ifa_local
;
3516 bond
->master_ip
= bond_glean_dev_ip(bond
->dev
);
3523 list_for_each_entry(vlan
, &bond
->vlan_list
, vlan_list
) {
3524 vlan_dev
= vlan_group_get_device(bond
->vlgrp
, vlan
->vlan_id
);
3525 if (vlan_dev
== event_dev
) {
3528 vlan
->vlan_ip
= ifa
->ifa_local
;
3532 bond_glean_dev_ip(vlan_dev
);
3543 static struct notifier_block bond_netdev_notifier
= {
3544 .notifier_call
= bond_netdev_event
,
3547 static struct notifier_block bond_inetaddr_notifier
= {
3548 .notifier_call
= bond_inetaddr_event
,
3551 /*-------------------------- Packet type handling ---------------------------*/
3553 /* register to receive lacpdus on a bond */
3554 static void bond_register_lacpdu(struct bonding
*bond
)
3556 struct packet_type
*pk_type
= &(BOND_AD_INFO(bond
).ad_pkt_type
);
3558 /* initialize packet type */
3559 pk_type
->type
= PKT_TYPE_LACPDU
;
3560 pk_type
->dev
= bond
->dev
;
3561 pk_type
->func
= bond_3ad_lacpdu_recv
;
3563 dev_add_pack(pk_type
);
3566 /* unregister to receive lacpdus on a bond */
3567 static void bond_unregister_lacpdu(struct bonding
*bond
)
3569 dev_remove_pack(&(BOND_AD_INFO(bond
).ad_pkt_type
));
3572 void bond_register_arp(struct bonding
*bond
)
3574 struct packet_type
*pt
= &bond
->arp_mon_pt
;
3579 pt
->type
= htons(ETH_P_ARP
);
3580 pt
->dev
= bond
->dev
;
3581 pt
->func
= bond_arp_rcv
;
3585 void bond_unregister_arp(struct bonding
*bond
)
3587 struct packet_type
*pt
= &bond
->arp_mon_pt
;
3589 dev_remove_pack(pt
);
3593 /*---------------------------- Hashing Policies -----------------------------*/
3596 * Hash for the output device based upon layer 2 and layer 3 data. If
3597 * the packet is not IP mimic bond_xmit_hash_policy_l2()
3599 static int bond_xmit_hash_policy_l23(struct sk_buff
*skb
, int count
)
3601 struct ethhdr
*data
= (struct ethhdr
*)skb
->data
;
3602 struct iphdr
*iph
= ip_hdr(skb
);
3604 if (skb
->protocol
== htons(ETH_P_IP
)) {
3605 return ((ntohl(iph
->saddr
^ iph
->daddr
) & 0xffff) ^
3606 (data
->h_dest
[5] ^ data
->h_source
[5])) % count
;
3609 return (data
->h_dest
[5] ^ data
->h_source
[5]) % count
;
3613 * Hash for the output device based upon layer 3 and layer 4 data. If
3614 * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is
3615 * altogether not IP, mimic bond_xmit_hash_policy_l2()
3617 static int bond_xmit_hash_policy_l34(struct sk_buff
*skb
, int count
)
3619 struct ethhdr
*data
= (struct ethhdr
*)skb
->data
;
3620 struct iphdr
*iph
= ip_hdr(skb
);
3621 __be16
*layer4hdr
= (__be16
*)((u32
*)iph
+ iph
->ihl
);
3624 if (skb
->protocol
== htons(ETH_P_IP
)) {
3625 if (!(iph
->frag_off
& htons(IP_MF
|IP_OFFSET
)) &&
3626 (iph
->protocol
== IPPROTO_TCP
||
3627 iph
->protocol
== IPPROTO_UDP
)) {
3628 layer4_xor
= ntohs((*layer4hdr
^ *(layer4hdr
+ 1)));
3630 return (layer4_xor
^
3631 ((ntohl(iph
->saddr
^ iph
->daddr
)) & 0xffff)) % count
;
3635 return (data
->h_dest
[5] ^ data
->h_source
[5]) % count
;
3639 * Hash for the output device based upon layer 2 data
3641 static int bond_xmit_hash_policy_l2(struct sk_buff
*skb
, int count
)
3643 struct ethhdr
*data
= (struct ethhdr
*)skb
->data
;
3645 return (data
->h_dest
[5] ^ data
->h_source
[5]) % count
;
3648 /*-------------------------- Device entry points ----------------------------*/
3650 static int bond_open(struct net_device
*bond_dev
)
3652 struct bonding
*bond
= netdev_priv(bond_dev
);
3654 bond
->kill_timers
= 0;
3656 if (bond_is_lb(bond
)) {
3657 /* bond_alb_initialize must be called before the timer
3660 if (bond_alb_initialize(bond
, (bond
->params
.mode
== BOND_MODE_ALB
))) {
3661 /* something went wrong - fail the open operation */
3665 INIT_DELAYED_WORK(&bond
->alb_work
, bond_alb_monitor
);
3666 queue_delayed_work(bond
->wq
, &bond
->alb_work
, 0);
3669 if (bond
->params
.miimon
) { /* link check interval, in milliseconds. */
3670 INIT_DELAYED_WORK(&bond
->mii_work
, bond_mii_monitor
);
3671 queue_delayed_work(bond
->wq
, &bond
->mii_work
, 0);
3674 if (bond
->params
.arp_interval
) { /* arp interval, in milliseconds. */
3675 if (bond
->params
.mode
== BOND_MODE_ACTIVEBACKUP
)
3676 INIT_DELAYED_WORK(&bond
->arp_work
,
3677 bond_activebackup_arp_mon
);
3679 INIT_DELAYED_WORK(&bond
->arp_work
,
3680 bond_loadbalance_arp_mon
);
3682 queue_delayed_work(bond
->wq
, &bond
->arp_work
, 0);
3683 if (bond
->params
.arp_validate
)
3684 bond_register_arp(bond
);
3687 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3688 INIT_DELAYED_WORK(&bond
->ad_work
, bond_3ad_state_machine_handler
);
3689 queue_delayed_work(bond
->wq
, &bond
->ad_work
, 0);
3690 /* register to receive LACPDUs */
3691 bond_register_lacpdu(bond
);
3692 bond_3ad_initiate_agg_selection(bond
, 1);
3698 static int bond_close(struct net_device
*bond_dev
)
3700 struct bonding
*bond
= netdev_priv(bond_dev
);
3702 if (bond
->params
.mode
== BOND_MODE_8023AD
) {
3703 /* Unregister the receive of LACPDUs */
3704 bond_unregister_lacpdu(bond
);
3707 if (bond
->params
.arp_validate
)
3708 bond_unregister_arp(bond
);
3710 write_lock_bh(&bond
->lock
);
3712 bond
->send_grat_arp
= 0;
3713 bond
->send_unsol_na
= 0;
3715 /* signal timers not to re-arm */
3716 bond
->kill_timers
= 1;
3718 write_unlock_bh(&bond
->lock
);
3720 if (bond
->params
.miimon
) { /* link check interval, in milliseconds. */
3721 cancel_delayed_work(&bond
->mii_work
);
3724 if (bond
->params
.arp_interval
) { /* arp interval, in milliseconds. */
3725 cancel_delayed_work(&bond
->arp_work
);
3728 switch (bond
->params
.mode
) {
3729 case BOND_MODE_8023AD
:
3730 cancel_delayed_work(&bond
->ad_work
);
3734 cancel_delayed_work(&bond
->alb_work
);
3741 if (bond_is_lb(bond
)) {
3742 /* Must be called only after all
3743 * slaves have been released
3745 bond_alb_deinitialize(bond
);
3751 static struct net_device_stats
*bond_get_stats(struct net_device
*bond_dev
)
3753 struct bonding
*bond
= netdev_priv(bond_dev
);
3754 struct net_device_stats
*stats
= &bond_dev
->stats
;
3755 struct net_device_stats local_stats
;
3756 struct slave
*slave
;
3759 memset(&local_stats
, 0, sizeof(struct net_device_stats
));
3761 read_lock_bh(&bond
->lock
);
3763 bond_for_each_slave(bond
, slave
, i
) {
3764 const struct net_device_stats
*sstats
= dev_get_stats(slave
->dev
);
3766 local_stats
.rx_packets
+= sstats
->rx_packets
;
3767 local_stats
.rx_bytes
+= sstats
->rx_bytes
;
3768 local_stats
.rx_errors
+= sstats
->rx_errors
;
3769 local_stats
.rx_dropped
+= sstats
->rx_dropped
;
3771 local_stats
.tx_packets
+= sstats
->tx_packets
;
3772 local_stats
.tx_bytes
+= sstats
->tx_bytes
;
3773 local_stats
.tx_errors
+= sstats
->tx_errors
;
3774 local_stats
.tx_dropped
+= sstats
->tx_dropped
;
3776 local_stats
.multicast
+= sstats
->multicast
;
3777 local_stats
.collisions
+= sstats
->collisions
;
3779 local_stats
.rx_length_errors
+= sstats
->rx_length_errors
;
3780 local_stats
.rx_over_errors
+= sstats
->rx_over_errors
;
3781 local_stats
.rx_crc_errors
+= sstats
->rx_crc_errors
;
3782 local_stats
.rx_frame_errors
+= sstats
->rx_frame_errors
;
3783 local_stats
.rx_fifo_errors
+= sstats
->rx_fifo_errors
;
3784 local_stats
.rx_missed_errors
+= sstats
->rx_missed_errors
;
3786 local_stats
.tx_aborted_errors
+= sstats
->tx_aborted_errors
;
3787 local_stats
.tx_carrier_errors
+= sstats
->tx_carrier_errors
;
3788 local_stats
.tx_fifo_errors
+= sstats
->tx_fifo_errors
;
3789 local_stats
.tx_heartbeat_errors
+= sstats
->tx_heartbeat_errors
;
3790 local_stats
.tx_window_errors
+= sstats
->tx_window_errors
;
3793 memcpy(stats
, &local_stats
, sizeof(struct net_device_stats
));
3795 read_unlock_bh(&bond
->lock
);
3800 static int bond_do_ioctl(struct net_device
*bond_dev
, struct ifreq
*ifr
, int cmd
)
3802 struct net_device
*slave_dev
= NULL
;
3803 struct ifbond k_binfo
;
3804 struct ifbond __user
*u_binfo
= NULL
;
3805 struct ifslave k_sinfo
;
3806 struct ifslave __user
*u_sinfo
= NULL
;
3807 struct mii_ioctl_data
*mii
= NULL
;
3810 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev
->name
, cmd
);
3822 * We do this again just in case we were called by SIOCGMIIREG
3823 * instead of SIOCGMIIPHY.
3830 if (mii
->reg_num
== 1) {
3831 struct bonding
*bond
= netdev_priv(bond_dev
);
3833 read_lock(&bond
->lock
);
3834 read_lock(&bond
->curr_slave_lock
);
3835 if (netif_carrier_ok(bond
->dev
))
3836 mii
->val_out
= BMSR_LSTATUS
;
3838 read_unlock(&bond
->curr_slave_lock
);
3839 read_unlock(&bond
->lock
);
3843 case BOND_INFO_QUERY_OLD
:
3844 case SIOCBONDINFOQUERY
:
3845 u_binfo
= (struct ifbond __user
*)ifr
->ifr_data
;
3847 if (copy_from_user(&k_binfo
, u_binfo
, sizeof(ifbond
)))
3850 res
= bond_info_query(bond_dev
, &k_binfo
);
3852 copy_to_user(u_binfo
, &k_binfo
, sizeof(ifbond
)))
3856 case BOND_SLAVE_INFO_QUERY_OLD
:
3857 case SIOCBONDSLAVEINFOQUERY
:
3858 u_sinfo
= (struct ifslave __user
*)ifr
->ifr_data
;
3860 if (copy_from_user(&k_sinfo
, u_sinfo
, sizeof(ifslave
)))
3863 res
= bond_slave_info_query(bond_dev
, &k_sinfo
);
3865 copy_to_user(u_sinfo
, &k_sinfo
, sizeof(ifslave
)))
3874 if (!capable(CAP_NET_ADMIN
))
3877 slave_dev
= dev_get_by_name(dev_net(bond_dev
), ifr
->ifr_slave
);
3879 pr_debug("slave_dev=%p:\n", slave_dev
);
3884 pr_debug("slave_dev->name=%s:\n", slave_dev
->name
);
3886 case BOND_ENSLAVE_OLD
:
3887 case SIOCBONDENSLAVE
:
3888 res
= bond_enslave(bond_dev
, slave_dev
);
3890 case BOND_RELEASE_OLD
:
3891 case SIOCBONDRELEASE
:
3892 res
= bond_release(bond_dev
, slave_dev
);
3894 case BOND_SETHWADDR_OLD
:
3895 case SIOCBONDSETHWADDR
:
3896 res
= bond_sethwaddr(bond_dev
, slave_dev
);
3898 case BOND_CHANGE_ACTIVE_OLD
:
3899 case SIOCBONDCHANGEACTIVE
:
3900 res
= bond_ioctl_change_active(bond_dev
, slave_dev
);
3912 static void bond_set_multicast_list(struct net_device
*bond_dev
)
3914 struct bonding
*bond
= netdev_priv(bond_dev
);
3915 struct dev_mc_list
*dmi
;
3918 * Do promisc before checking multicast_mode
3920 if ((bond_dev
->flags
& IFF_PROMISC
) && !(bond
->flags
& IFF_PROMISC
))
3922 * FIXME: Need to handle the error when one of the multi-slaves
3925 bond_set_promiscuity(bond
, 1);
3928 if (!(bond_dev
->flags
& IFF_PROMISC
) && (bond
->flags
& IFF_PROMISC
))
3929 bond_set_promiscuity(bond
, -1);
3932 /* set allmulti flag to slaves */
3933 if ((bond_dev
->flags
& IFF_ALLMULTI
) && !(bond
->flags
& IFF_ALLMULTI
))
3935 * FIXME: Need to handle the error when one of the multi-slaves
3938 bond_set_allmulti(bond
, 1);
3941 if (!(bond_dev
->flags
& IFF_ALLMULTI
) && (bond
->flags
& IFF_ALLMULTI
))
3942 bond_set_allmulti(bond
, -1);
3945 read_lock(&bond
->lock
);
3947 bond
->flags
= bond_dev
->flags
;
3949 /* looking for addresses to add to slaves' mc list */
3950 for (dmi
= bond_dev
->mc_list
; dmi
; dmi
= dmi
->next
) {
3951 if (!bond_mc_list_find_dmi(dmi
, bond
->mc_list
))
3952 bond_mc_add(bond
, dmi
->dmi_addr
, dmi
->dmi_addrlen
);
3955 /* looking for addresses to delete from slaves' list */
3956 for (dmi
= bond
->mc_list
; dmi
; dmi
= dmi
->next
) {
3957 if (!bond_mc_list_find_dmi(dmi
, bond_dev
->mc_list
))
3958 bond_mc_delete(bond
, dmi
->dmi_addr
, dmi
->dmi_addrlen
);
3961 /* save master's multicast list */
3962 bond_mc_list_destroy(bond
);
3963 bond_mc_list_copy(bond_dev
->mc_list
, bond
, GFP_ATOMIC
);
3965 read_unlock(&bond
->lock
);
3968 static int bond_neigh_setup(struct net_device
*dev
, struct neigh_parms
*parms
)
3970 struct bonding
*bond
= netdev_priv(dev
);
3971 struct slave
*slave
= bond
->first_slave
;
3974 const struct net_device_ops
*slave_ops
3975 = slave
->dev
->netdev_ops
;
3976 if (slave_ops
->ndo_neigh_setup
)
3977 return slave_ops
->ndo_neigh_setup(slave
->dev
, parms
);
3983 * Change the MTU of all of a master's slaves to match the master
3985 static int bond_change_mtu(struct net_device
*bond_dev
, int new_mtu
)
3987 struct bonding
*bond
= netdev_priv(bond_dev
);
3988 struct slave
*slave
, *stop_at
;
3992 pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond
,
3993 (bond_dev
? bond_dev
->name
: "None"), new_mtu
);
3995 /* Can't hold bond->lock with bh disabled here since
3996 * some base drivers panic. On the other hand we can't
3997 * hold bond->lock without bh disabled because we'll
3998 * deadlock. The only solution is to rely on the fact
3999 * that we're under rtnl_lock here, and the slaves
4000 * list won't change. This doesn't solve the problem
4001 * of setting the slave's MTU while it is
4002 * transmitting, but the assumption is that the base
4003 * driver can handle that.
4005 * TODO: figure out a way to safely iterate the slaves
4006 * list, but without holding a lock around the actual
4007 * call to the base driver.
4010 bond_for_each_slave(bond
, slave
, i
) {
4011 pr_debug("s %p s->p %p c_m %p\n",
4014 slave
->dev
->netdev_ops
->ndo_change_mtu
);
4016 res
= dev_set_mtu(slave
->dev
, new_mtu
);
4019 /* If we failed to set the slave's mtu to the new value
4020 * we must abort the operation even in ACTIVE_BACKUP
4021 * mode, because if we allow the backup slaves to have
4022 * different mtu values than the active slave we'll
4023 * need to change their mtu when doing a failover. That
4024 * means changing their mtu from timer context, which
4025 * is probably not a good idea.
4027 pr_debug("err %d %s\n", res
, slave
->dev
->name
);
4032 bond_dev
->mtu
= new_mtu
;
4037 /* unwind from head to the slave that failed */
4039 bond_for_each_slave_from_to(bond
, slave
, i
, bond
->first_slave
, stop_at
) {
4042 tmp_res
= dev_set_mtu(slave
->dev
, bond_dev
->mtu
);
4044 pr_debug("unwind err %d dev %s\n",
4045 tmp_res
, slave
->dev
->name
);
4055 * Note that many devices must be down to change the HW address, and
4056 * downing the master releases all slaves. We can make bonds full of
4057 * bonding devices to test this, however.
4059 static int bond_set_mac_address(struct net_device
*bond_dev
, void *addr
)
4061 struct bonding
*bond
= netdev_priv(bond_dev
);
4062 struct sockaddr
*sa
= addr
, tmp_sa
;
4063 struct slave
*slave
, *stop_at
;
4067 if (bond
->params
.mode
== BOND_MODE_ALB
)
4068 return bond_alb_set_mac_address(bond_dev
, addr
);
4071 pr_debug("bond=%p, name=%s\n",
4072 bond
, bond_dev
? bond_dev
->name
: "None");
4075 * If fail_over_mac is set to active, do nothing and return
4076 * success. Returning an error causes ifenslave to fail.
4078 if (bond
->params
.fail_over_mac
== BOND_FOM_ACTIVE
)
4081 if (!is_valid_ether_addr(sa
->sa_data
))
4082 return -EADDRNOTAVAIL
;
4084 /* Can't hold bond->lock with bh disabled here since
4085 * some base drivers panic. On the other hand we can't
4086 * hold bond->lock without bh disabled because we'll
4087 * deadlock. The only solution is to rely on the fact
4088 * that we're under rtnl_lock here, and the slaves
4089 * list won't change. This doesn't solve the problem
4090 * of setting the slave's hw address while it is
4091 * transmitting, but the assumption is that the base
4092 * driver can handle that.
4094 * TODO: figure out a way to safely iterate the slaves
4095 * list, but without holding a lock around the actual
4096 * call to the base driver.
4099 bond_for_each_slave(bond
, slave
, i
) {
4100 const struct net_device_ops
*slave_ops
= slave
->dev
->netdev_ops
;
4101 pr_debug("slave %p %s\n", slave
, slave
->dev
->name
);
4103 if (slave_ops
->ndo_set_mac_address
== NULL
) {
4105 pr_debug("EOPNOTSUPP %s\n", slave
->dev
->name
);
4109 res
= dev_set_mac_address(slave
->dev
, addr
);
4111 /* TODO: consider downing the slave
4113 * User should expect communications
4114 * breakage anyway until ARP finish
4117 pr_debug("err %d %s\n", res
, slave
->dev
->name
);
4123 memcpy(bond_dev
->dev_addr
, sa
->sa_data
, bond_dev
->addr_len
);
4127 memcpy(tmp_sa
.sa_data
, bond_dev
->dev_addr
, bond_dev
->addr_len
);
4128 tmp_sa
.sa_family
= bond_dev
->type
;
4130 /* unwind from head to the slave that failed */
4132 bond_for_each_slave_from_to(bond
, slave
, i
, bond
->first_slave
, stop_at
) {
4135 tmp_res
= dev_set_mac_address(slave
->dev
, &tmp_sa
);
4137 pr_debug("unwind err %d dev %s\n",
4138 tmp_res
, slave
->dev
->name
);
4145 static int bond_xmit_roundrobin(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4147 struct bonding
*bond
= netdev_priv(bond_dev
);
4148 struct slave
*slave
, *start_at
;
4149 int i
, slave_no
, res
= 1;
4151 read_lock(&bond
->lock
);
4153 if (!BOND_IS_OK(bond
))
4157 * Concurrent TX may collide on rr_tx_counter; we accept that
4158 * as being rare enough not to justify using an atomic op here
4160 slave_no
= bond
->rr_tx_counter
++ % bond
->slave_cnt
;
4162 bond_for_each_slave(bond
, slave
, i
) {
4169 bond_for_each_slave_from(bond
, slave
, i
, start_at
) {
4170 if (IS_UP(slave
->dev
) &&
4171 (slave
->link
== BOND_LINK_UP
) &&
4172 (slave
->state
== BOND_STATE_ACTIVE
)) {
4173 res
= bond_dev_queue_xmit(bond
, skb
, slave
->dev
);
4180 /* no suitable interface, frame not sent */
4183 read_unlock(&bond
->lock
);
4184 return NETDEV_TX_OK
;
4189 * in active-backup mode, we know that bond->curr_active_slave is always valid if
4190 * the bond has a usable interface.
4192 static int bond_xmit_activebackup(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4194 struct bonding
*bond
= netdev_priv(bond_dev
);
4197 read_lock(&bond
->lock
);
4198 read_lock(&bond
->curr_slave_lock
);
4200 if (!BOND_IS_OK(bond
))
4203 if (!bond
->curr_active_slave
)
4206 res
= bond_dev_queue_xmit(bond
, skb
, bond
->curr_active_slave
->dev
);
4210 /* no suitable interface, frame not sent */
4213 read_unlock(&bond
->curr_slave_lock
);
4214 read_unlock(&bond
->lock
);
4215 return NETDEV_TX_OK
;
4219 * In bond_xmit_xor() , we determine the output device by using a pre-
4220 * determined xmit_hash_policy(), If the selected device is not enabled,
4221 * find the next active slave.
4223 static int bond_xmit_xor(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4225 struct bonding
*bond
= netdev_priv(bond_dev
);
4226 struct slave
*slave
, *start_at
;
4231 read_lock(&bond
->lock
);
4233 if (!BOND_IS_OK(bond
))
4236 slave_no
= bond
->xmit_hash_policy(skb
, bond
->slave_cnt
);
4238 bond_for_each_slave(bond
, slave
, i
) {
4246 bond_for_each_slave_from(bond
, slave
, i
, start_at
) {
4247 if (IS_UP(slave
->dev
) &&
4248 (slave
->link
== BOND_LINK_UP
) &&
4249 (slave
->state
== BOND_STATE_ACTIVE
)) {
4250 res
= bond_dev_queue_xmit(bond
, skb
, slave
->dev
);
4257 /* no suitable interface, frame not sent */
4260 read_unlock(&bond
->lock
);
4261 return NETDEV_TX_OK
;
4265 * in broadcast mode, we send everything to all usable interfaces.
4267 static int bond_xmit_broadcast(struct sk_buff
*skb
, struct net_device
*bond_dev
)
4269 struct bonding
*bond
= netdev_priv(bond_dev
);
4270 struct slave
*slave
, *start_at
;
4271 struct net_device
*tx_dev
= NULL
;
4275 read_lock(&bond
->lock
);
4277 if (!BOND_IS_OK(bond
))
4280 read_lock(&bond
->curr_slave_lock
);
4281 start_at
= bond
->curr_active_slave
;
4282 read_unlock(&bond
->curr_slave_lock
);
4287 bond_for_each_slave_from(bond
, slave
, i
, start_at
) {
4288 if (IS_UP(slave
->dev
) &&
4289 (slave
->link
== BOND_LINK_UP
) &&
4290 (slave
->state
== BOND_STATE_ACTIVE
)) {
4292 struct sk_buff
*skb2
= skb_clone(skb
, GFP_ATOMIC
);
4294 pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n",
4299 res
= bond_dev_queue_xmit(bond
, skb2
, tx_dev
);
4301 dev_kfree_skb(skb2
);
4305 tx_dev
= slave
->dev
;
4310 res
= bond_dev_queue_xmit(bond
, skb
, tx_dev
);
4314 /* no suitable interface, frame not sent */
4317 /* frame sent to all suitable interfaces */
4318 read_unlock(&bond
->lock
);
4319 return NETDEV_TX_OK
;
4322 /*------------------------- Device initialization ---------------------------*/
4324 static void bond_set_xmit_hash_policy(struct bonding
*bond
)
4326 switch (bond
->params
.xmit_policy
) {
4327 case BOND_XMIT_POLICY_LAYER23
:
4328 bond
->xmit_hash_policy
= bond_xmit_hash_policy_l23
;
4330 case BOND_XMIT_POLICY_LAYER34
:
4331 bond
->xmit_hash_policy
= bond_xmit_hash_policy_l34
;
4333 case BOND_XMIT_POLICY_LAYER2
:
4335 bond
->xmit_hash_policy
= bond_xmit_hash_policy_l2
;
4340 static netdev_tx_t
bond_start_xmit(struct sk_buff
*skb
, struct net_device
*dev
)
4342 const struct bonding
*bond
= netdev_priv(dev
);
4344 switch (bond
->params
.mode
) {
4345 case BOND_MODE_ROUNDROBIN
:
4346 return bond_xmit_roundrobin(skb
, dev
);
4347 case BOND_MODE_ACTIVEBACKUP
:
4348 return bond_xmit_activebackup(skb
, dev
);
4350 return bond_xmit_xor(skb
, dev
);
4351 case BOND_MODE_BROADCAST
:
4352 return bond_xmit_broadcast(skb
, dev
);
4353 case BOND_MODE_8023AD
:
4354 return bond_3ad_xmit_xor(skb
, dev
);
4357 return bond_alb_xmit(skb
, dev
);
4359 /* Should never happen, mode already checked */
4360 pr_err("%s: Error: Unknown bonding mode %d\n",
4361 dev
->name
, bond
->params
.mode
);
4364 return NETDEV_TX_OK
;
4370 * set bond mode specific net device operations
4372 void bond_set_mode_ops(struct bonding
*bond
, int mode
)
4374 struct net_device
*bond_dev
= bond
->dev
;
4377 case BOND_MODE_ROUNDROBIN
:
4379 case BOND_MODE_ACTIVEBACKUP
:
4382 bond_set_xmit_hash_policy(bond
);
4384 case BOND_MODE_BROADCAST
:
4386 case BOND_MODE_8023AD
:
4387 bond_set_master_3ad_flags(bond
);
4388 bond_set_xmit_hash_policy(bond
);
4391 bond_set_master_alb_flags(bond
);
4396 /* Should never happen, mode already checked */
4397 pr_err("%s: Error: Unknown bonding mode %d\n",
4398 bond_dev
->name
, mode
);
4403 static void bond_ethtool_get_drvinfo(struct net_device
*bond_dev
,
4404 struct ethtool_drvinfo
*drvinfo
)
4406 strncpy(drvinfo
->driver
, DRV_NAME
, 32);
4407 strncpy(drvinfo
->version
, DRV_VERSION
, 32);
4408 snprintf(drvinfo
->fw_version
, 32, "%d", BOND_ABI_VERSION
);
4411 static const struct ethtool_ops bond_ethtool_ops
= {
4412 .get_drvinfo
= bond_ethtool_get_drvinfo
,
4413 .get_link
= ethtool_op_get_link
,
4414 .get_tx_csum
= ethtool_op_get_tx_csum
,
4415 .get_sg
= ethtool_op_get_sg
,
4416 .get_tso
= ethtool_op_get_tso
,
4417 .get_ufo
= ethtool_op_get_ufo
,
4418 .get_flags
= ethtool_op_get_flags
,
4421 static const struct net_device_ops bond_netdev_ops
= {
4422 .ndo_init
= bond_init
,
4423 .ndo_uninit
= bond_uninit
,
4424 .ndo_open
= bond_open
,
4425 .ndo_stop
= bond_close
,
4426 .ndo_start_xmit
= bond_start_xmit
,
4427 .ndo_get_stats
= bond_get_stats
,
4428 .ndo_do_ioctl
= bond_do_ioctl
,
4429 .ndo_set_multicast_list
= bond_set_multicast_list
,
4430 .ndo_change_mtu
= bond_change_mtu
,
4431 .ndo_set_mac_address
= bond_set_mac_address
,
4432 .ndo_neigh_setup
= bond_neigh_setup
,
4433 .ndo_vlan_rx_register
= bond_vlan_rx_register
,
4434 .ndo_vlan_rx_add_vid
= bond_vlan_rx_add_vid
,
4435 .ndo_vlan_rx_kill_vid
= bond_vlan_rx_kill_vid
,
4438 static void bond_setup(struct net_device
*bond_dev
)
4440 struct bonding
*bond
= netdev_priv(bond_dev
);
4442 /* initialize rwlocks */
4443 rwlock_init(&bond
->lock
);
4444 rwlock_init(&bond
->curr_slave_lock
);
4446 bond
->params
= bonding_defaults
;
4448 /* Initialize pointers */
4449 bond
->dev
= bond_dev
;
4450 INIT_LIST_HEAD(&bond
->vlan_list
);
4452 /* Initialize the device entry points */
4453 ether_setup(bond_dev
);
4454 bond_dev
->netdev_ops
= &bond_netdev_ops
;
4455 bond_dev
->ethtool_ops
= &bond_ethtool_ops
;
4456 bond_set_mode_ops(bond
, bond
->params
.mode
);
4458 bond_dev
->destructor
= free_netdev
;
4460 /* Initialize the device options */
4461 bond_dev
->tx_queue_len
= 0;
4462 bond_dev
->flags
|= IFF_MASTER
|IFF_MULTICAST
;
4463 bond_dev
->priv_flags
|= IFF_BONDING
;
4464 bond_dev
->priv_flags
&= ~IFF_XMIT_DST_RELEASE
;
4466 if (bond
->params
.arp_interval
)
4467 bond_dev
->priv_flags
|= IFF_MASTER_ARPMON
;
4469 /* At first, we block adding VLANs. That's the only way to
4470 * prevent problems that occur when adding VLANs over an
4471 * empty bond. The block will be removed once non-challenged
4472 * slaves are enslaved.
4474 bond_dev
->features
|= NETIF_F_VLAN_CHALLENGED
;
4476 /* don't acquire bond device's netif_tx_lock when
4478 bond_dev
->features
|= NETIF_F_LLTX
;
4480 /* By default, we declare the bond to be fully
4481 * VLAN hardware accelerated capable. Special
4482 * care is taken in the various xmit functions
4483 * when there are slaves that are not hw accel
4486 bond_dev
->features
|= (NETIF_F_HW_VLAN_TX
|
4487 NETIF_F_HW_VLAN_RX
|
4488 NETIF_F_HW_VLAN_FILTER
);
4492 static void bond_work_cancel_all(struct bonding
*bond
)
4494 write_lock_bh(&bond
->lock
);
4495 bond
->kill_timers
= 1;
4496 write_unlock_bh(&bond
->lock
);
4498 if (bond
->params
.miimon
&& delayed_work_pending(&bond
->mii_work
))
4499 cancel_delayed_work(&bond
->mii_work
);
4501 if (bond
->params
.arp_interval
&& delayed_work_pending(&bond
->arp_work
))
4502 cancel_delayed_work(&bond
->arp_work
);
4504 if (bond
->params
.mode
== BOND_MODE_ALB
&&
4505 delayed_work_pending(&bond
->alb_work
))
4506 cancel_delayed_work(&bond
->alb_work
);
4508 if (bond
->params
.mode
== BOND_MODE_8023AD
&&
4509 delayed_work_pending(&bond
->ad_work
))
4510 cancel_delayed_work(&bond
->ad_work
);
4514 * Destroy a bonding device.
4515 * Must be under rtnl_lock when this function is called.
4517 static void bond_uninit(struct net_device
*bond_dev
)
4519 struct bonding
*bond
= netdev_priv(bond_dev
);
4521 /* Release the bonded slaves */
4522 bond_release_all(bond_dev
);
4524 list_del(&bond
->bond_list
);
4526 bond_work_cancel_all(bond
);
4528 bond_remove_proc_entry(bond
);
4531 destroy_workqueue(bond
->wq
);
4533 netif_addr_lock_bh(bond_dev
);
4534 bond_mc_list_destroy(bond
);
4535 netif_addr_unlock_bh(bond_dev
);
4538 /*------------------------- Module initialization ---------------------------*/
4541 * Convert string input module parms. Accept either the
4542 * number of the mode or its string name. A bit complicated because
4543 * some mode names are substrings of other names, and calls from sysfs
4544 * may have whitespace in the name (trailing newlines, for example).
4546 int bond_parse_parm(const char *buf
, const struct bond_parm_tbl
*tbl
)
4548 int modeint
= -1, i
, rv
;
4549 char *p
, modestr
[BOND_MAX_MODENAME_LEN
+ 1] = { 0, };
4551 for (p
= (char *)buf
; *p
; p
++)
4552 if (!(isdigit(*p
) || isspace(*p
)))
4556 rv
= sscanf(buf
, "%20s", modestr
);
4558 rv
= sscanf(buf
, "%d", &modeint
);
4563 for (i
= 0; tbl
[i
].modename
; i
++) {
4564 if (modeint
== tbl
[i
].mode
)
4566 if (strcmp(modestr
, tbl
[i
].modename
) == 0)
4573 static int bond_check_params(struct bond_params
*params
)
4575 int arp_validate_value
, fail_over_mac_value
, primary_reselect_value
;
4578 * Convert string parameters.
4581 bond_mode
= bond_parse_parm(mode
, bond_mode_tbl
);
4582 if (bond_mode
== -1) {
4583 pr_err("Error: Invalid bonding mode \"%s\"\n",
4584 mode
== NULL
? "NULL" : mode
);
4589 if (xmit_hash_policy
) {
4590 if ((bond_mode
!= BOND_MODE_XOR
) &&
4591 (bond_mode
!= BOND_MODE_8023AD
)) {
4592 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
4593 bond_mode_name(bond_mode
));
4595 xmit_hashtype
= bond_parse_parm(xmit_hash_policy
,
4597 if (xmit_hashtype
== -1) {
4598 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
4599 xmit_hash_policy
== NULL
? "NULL" :
4607 if (bond_mode
!= BOND_MODE_8023AD
) {
4608 pr_info("lacp_rate param is irrelevant in mode %s\n",
4609 bond_mode_name(bond_mode
));
4611 lacp_fast
= bond_parse_parm(lacp_rate
, bond_lacp_tbl
);
4612 if (lacp_fast
== -1) {
4613 pr_err("Error: Invalid lacp rate \"%s\"\n",
4614 lacp_rate
== NULL
? "NULL" : lacp_rate
);
4621 params
->ad_select
= bond_parse_parm(ad_select
, ad_select_tbl
);
4622 if (params
->ad_select
== -1) {
4623 pr_err("Error: Invalid ad_select \"%s\"\n",
4624 ad_select
== NULL
? "NULL" : ad_select
);
4628 if (bond_mode
!= BOND_MODE_8023AD
) {
4629 pr_warning("ad_select param only affects 802.3ad mode\n");
4632 params
->ad_select
= BOND_AD_STABLE
;
4635 if (max_bonds
< 0) {
4636 pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4637 max_bonds
, 0, INT_MAX
, BOND_DEFAULT_MAX_BONDS
);
4638 max_bonds
= BOND_DEFAULT_MAX_BONDS
;
4642 pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n",
4643 miimon
, INT_MAX
, BOND_LINK_MON_INTERV
);
4644 miimon
= BOND_LINK_MON_INTERV
;
4648 pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4653 if (downdelay
< 0) {
4654 pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4655 downdelay
, INT_MAX
);
4659 if ((use_carrier
!= 0) && (use_carrier
!= 1)) {
4660 pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4665 if (num_grat_arp
< 0 || num_grat_arp
> 255) {
4666 pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1 \n",
4671 if (num_unsol_na
< 0 || num_unsol_na
> 255) {
4672 pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1 \n",
4677 /* reset values for 802.3ad */
4678 if (bond_mode
== BOND_MODE_8023AD
) {
4680 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");
4681 pr_warning("Forcing miimon to 100msec\n");
4686 /* reset values for TLB/ALB */
4687 if ((bond_mode
== BOND_MODE_TLB
) ||
4688 (bond_mode
== BOND_MODE_ALB
)) {
4690 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");
4691 pr_warning("Forcing miimon to 100msec\n");
4696 if (bond_mode
== BOND_MODE_ALB
) {
4697 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",
4702 if (updelay
|| downdelay
) {
4703 /* just warn the user the up/down delay will have
4704 * no effect since miimon is zero...
4706 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",
4707 updelay
, downdelay
);
4710 /* don't allow arp monitoring */
4712 pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4713 miimon
, arp_interval
);
4717 if ((updelay
% miimon
) != 0) {
4718 pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4720 (updelay
/ miimon
) * miimon
);
4725 if ((downdelay
% miimon
) != 0) {
4726 pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4728 (downdelay
/ miimon
) * miimon
);
4731 downdelay
/= miimon
;
4734 if (arp_interval
< 0) {
4735 pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n",
4736 arp_interval
, INT_MAX
, BOND_LINK_ARP_INTERV
);
4737 arp_interval
= BOND_LINK_ARP_INTERV
;
4740 for (arp_ip_count
= 0;
4741 (arp_ip_count
< BOND_MAX_ARP_TARGETS
) && arp_ip_target
[arp_ip_count
];
4743 /* not complete check, but should be good enough to
4745 if (!isdigit(arp_ip_target
[arp_ip_count
][0])) {
4746 pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4747 arp_ip_target
[arp_ip_count
]);
4750 __be32 ip
= in_aton(arp_ip_target
[arp_ip_count
]);
4751 arp_target
[arp_ip_count
] = ip
;
4755 if (arp_interval
&& !arp_ip_count
) {
4756 /* don't allow arping if no arp_ip_target given... */
4757 pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4763 if (bond_mode
!= BOND_MODE_ACTIVEBACKUP
) {
4764 pr_err("arp_validate only supported in active-backup mode\n");
4767 if (!arp_interval
) {
4768 pr_err("arp_validate requires arp_interval\n");
4772 arp_validate_value
= bond_parse_parm(arp_validate
,
4774 if (arp_validate_value
== -1) {
4775 pr_err("Error: invalid arp_validate \"%s\"\n",
4776 arp_validate
== NULL
? "NULL" : arp_validate
);
4780 arp_validate_value
= 0;
4783 pr_info("MII link monitoring set to %d ms\n", miimon
);
4784 } else if (arp_interval
) {
4787 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
4789 arp_validate_tbl
[arp_validate_value
].modename
,
4792 for (i
= 0; i
< arp_ip_count
; i
++)
4793 pr_info(" %s", arp_ip_target
[i
]);
4797 } else if (max_bonds
) {
4798 /* miimon and arp_interval not set, we need one so things
4799 * work as expected, see bonding.txt for details
4801 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");
4804 if (primary
&& !USES_PRIMARY(bond_mode
)) {
4805 /* currently, using a primary only makes sense
4806 * in active backup, TLB or ALB modes
4808 pr_warning("Warning: %s primary device specified but has no effect in %s mode\n",
4809 primary
, bond_mode_name(bond_mode
));
4813 if (primary
&& primary_reselect
) {
4814 primary_reselect_value
= bond_parse_parm(primary_reselect
,
4816 if (primary_reselect_value
== -1) {
4817 pr_err("Error: Invalid primary_reselect \"%s\"\n",
4819 NULL
? "NULL" : primary_reselect
);
4823 primary_reselect_value
= BOND_PRI_RESELECT_ALWAYS
;
4826 if (fail_over_mac
) {
4827 fail_over_mac_value
= bond_parse_parm(fail_over_mac
,
4829 if (fail_over_mac_value
== -1) {
4830 pr_err("Error: invalid fail_over_mac \"%s\"\n",
4831 arp_validate
== NULL
? "NULL" : arp_validate
);
4835 if (bond_mode
!= BOND_MODE_ACTIVEBACKUP
)
4836 pr_warning("Warning: fail_over_mac only affects active-backup mode.\n");
4838 fail_over_mac_value
= BOND_FOM_NONE
;
4841 /* fill params struct with the proper values */
4842 params
->mode
= bond_mode
;
4843 params
->xmit_policy
= xmit_hashtype
;
4844 params
->miimon
= miimon
;
4845 params
->num_grat_arp
= num_grat_arp
;
4846 params
->num_unsol_na
= num_unsol_na
;
4847 params
->arp_interval
= arp_interval
;
4848 params
->arp_validate
= arp_validate_value
;
4849 params
->updelay
= updelay
;
4850 params
->downdelay
= downdelay
;
4851 params
->use_carrier
= use_carrier
;
4852 params
->lacp_fast
= lacp_fast
;
4853 params
->primary
[0] = 0;
4854 params
->primary_reselect
= primary_reselect_value
;
4855 params
->fail_over_mac
= fail_over_mac_value
;
4858 strncpy(params
->primary
, primary
, IFNAMSIZ
);
4859 params
->primary
[IFNAMSIZ
- 1] = 0;
4862 memcpy(params
->arp_targets
, arp_target
, sizeof(arp_target
));
4867 static struct lock_class_key bonding_netdev_xmit_lock_key
;
4868 static struct lock_class_key bonding_netdev_addr_lock_key
;
4870 static void bond_set_lockdep_class_one(struct net_device
*dev
,
4871 struct netdev_queue
*txq
,
4874 lockdep_set_class(&txq
->_xmit_lock
,
4875 &bonding_netdev_xmit_lock_key
);
4878 static void bond_set_lockdep_class(struct net_device
*dev
)
4880 lockdep_set_class(&dev
->addr_list_lock
,
4881 &bonding_netdev_addr_lock_key
);
4882 netdev_for_each_tx_queue(dev
, bond_set_lockdep_class_one
, NULL
);
4886 * Called from registration process
4888 static int bond_init(struct net_device
*bond_dev
)
4890 struct bonding
*bond
= netdev_priv(bond_dev
);
4891 struct bond_net
*bn
= net_generic(dev_net(bond_dev
), bond_net_id
);
4893 pr_debug("Begin bond_init for %s\n", bond_dev
->name
);
4895 bond
->wq
= create_singlethread_workqueue(bond_dev
->name
);
4899 bond_set_lockdep_class(bond_dev
);
4901 netif_carrier_off(bond_dev
);
4903 bond_create_proc_entry(bond
);
4904 list_add_tail(&bond
->bond_list
, &bn
->dev_list
);
4906 bond_prepare_sysfs_group(bond
);
4910 static int bond_validate(struct nlattr
*tb
[], struct nlattr
*data
[])
4912 if (tb
[IFLA_ADDRESS
]) {
4913 if (nla_len(tb
[IFLA_ADDRESS
]) != ETH_ALEN
)
4915 if (!is_valid_ether_addr(nla_data(tb
[IFLA_ADDRESS
])))
4916 return -EADDRNOTAVAIL
;
4921 static struct rtnl_link_ops bond_link_ops __read_mostly
= {
4923 .priv_size
= sizeof(struct bonding
),
4924 .setup
= bond_setup
,
4925 .validate
= bond_validate
,
4928 /* Create a new bond based on the specified name and bonding parameters.
4929 * If name is NULL, obtain a suitable "bond%d" name for us.
4930 * Caller must NOT hold rtnl_lock; we need to release it here before we
4931 * set up our sysfs entries.
4933 int bond_create(struct net
*net
, const char *name
)
4935 struct net_device
*bond_dev
;
4940 bond_dev
= alloc_netdev(sizeof(struct bonding
), name
? name
: "",
4943 pr_err("%s: eek! can't alloc netdev!\n", name
);
4948 dev_net_set(bond_dev
, net
);
4949 bond_dev
->rtnl_link_ops
= &bond_link_ops
;
4952 res
= dev_alloc_name(bond_dev
, "bond%d");
4957 res
= register_netdevice(bond_dev
);
4965 free_netdev(bond_dev
);
4969 static int __net_init
bond_net_init(struct net
*net
)
4971 struct bond_net
*bn
= net_generic(net
, bond_net_id
);
4974 INIT_LIST_HEAD(&bn
->dev_list
);
4976 bond_create_proc_dir(bn
);
4981 static void __net_exit
bond_net_exit(struct net
*net
)
4983 struct bond_net
*bn
= net_generic(net
, bond_net_id
);
4985 bond_destroy_proc_dir(bn
);
4988 static struct pernet_operations bond_net_ops
= {
4989 .init
= bond_net_init
,
4990 .exit
= bond_net_exit
,
4992 .size
= sizeof(struct bond_net
),
4995 static int __init
bonding_init(void)
5000 pr_info("%s", version
);
5002 res
= bond_check_params(&bonding_defaults
);
5006 res
= register_pernet_subsys(&bond_net_ops
);
5010 res
= rtnl_link_register(&bond_link_ops
);
5014 for (i
= 0; i
< max_bonds
; i
++) {
5015 res
= bond_create(&init_net
, NULL
);
5020 res
= bond_create_sysfs();
5024 register_netdevice_notifier(&bond_netdev_notifier
);
5025 register_inetaddr_notifier(&bond_inetaddr_notifier
);
5026 bond_register_ipv6_notifier();
5030 rtnl_link_unregister(&bond_link_ops
);
5032 unregister_pernet_subsys(&bond_net_ops
);
5037 static void __exit
bonding_exit(void)
5039 unregister_netdevice_notifier(&bond_netdev_notifier
);
5040 unregister_inetaddr_notifier(&bond_inetaddr_notifier
);
5041 bond_unregister_ipv6_notifier();
5043 bond_destroy_sysfs();
5045 rtnl_link_unregister(&bond_link_ops
);
5046 unregister_pernet_subsys(&bond_net_ops
);
5049 module_init(bonding_init
);
5050 module_exit(bonding_exit
);
5051 MODULE_LICENSE("GPL");
5052 MODULE_VERSION(DRV_VERSION
);
5053 MODULE_DESCRIPTION(DRV_DESCRIPTION
", v" DRV_VERSION
);
5054 MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
5055 MODULE_ALIAS_RTNL_LINK("bond");