]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/net/bonding/bond_main.c
bonding: make speed, duplex setting consistent with link state
[mirror_ubuntu-bionic-kernel.git] / drivers / net / bonding / bond_main.c
CommitLineData
1da177e4
LT
1/*
2 * originally based on the dummy device.
3 *
4 * Copyright 1999, Thomas Davis, tadavis@lbl.gov.
5 * Licensed under the GPL. Based on dummy.c, and eql.c devices.
6 *
7 * bonding.c: an Ethernet Bonding driver
8 *
9 * This is useful to talk to a Cisco EtherChannel compatible equipment:
10 * Cisco 5500
11 * Sun Trunking (Solaris)
12 * Alteon AceDirector Trunks
13 * Linux Bonding
14 * and probably many L2 switches ...
15 *
16 * How it works:
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.
22 *
23 * ifconfig bond0 down
24 * will release all slaves, marking them as down.
25 *
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.
31 *
1da177e4
LT
32 */
33
1da177e4
LT
34#include <linux/kernel.h>
35#include <linux/module.h>
1da177e4
LT
36#include <linux/types.h>
37#include <linux/fcntl.h>
38#include <linux/interrupt.h>
39#include <linux/ptrace.h>
40#include <linux/ioport.h>
41#include <linux/in.h>
169a3e66 42#include <net/ip.h>
1da177e4 43#include <linux/ip.h>
169a3e66
JV
44#include <linux/tcp.h>
45#include <linux/udp.h>
1da177e4
LT
46#include <linux/slab.h>
47#include <linux/string.h>
48#include <linux/init.h>
49#include <linux/timer.h>
50#include <linux/socket.h>
51#include <linux/ctype.h>
52#include <linux/inet.h>
53#include <linux/bitops.h>
3d632c3f 54#include <linux/io.h>
1da177e4 55#include <asm/dma.h>
3d632c3f 56#include <linux/uaccess.h>
1da177e4
LT
57#include <linux/errno.h>
58#include <linux/netdevice.h>
59#include <linux/inetdevice.h>
a816c7c7 60#include <linux/igmp.h>
1da177e4
LT
61#include <linux/etherdevice.h>
62#include <linux/skbuff.h>
63#include <net/sock.h>
64#include <linux/rtnetlink.h>
1da177e4
LT
65#include <linux/smp.h>
66#include <linux/if_ether.h>
67#include <net/arp.h>
68#include <linux/mii.h>
69#include <linux/ethtool.h>
70#include <linux/if_vlan.h>
71#include <linux/if_bonding.h>
b63bb739 72#include <linux/jiffies.h>
e843fa50 73#include <linux/preempt.h>
c3ade5ca 74#include <net/route.h>
457c4cbc 75#include <net/net_namespace.h>
ec87fd3b 76#include <net/netns/generic.h>
5ee31c68 77#include <net/pkt_sched.h>
278b2083 78#include <linux/rculist.h>
1bd758eb 79#include <net/flow_dissector.h>
c158cba3 80#include <net/switchdev.h>
1ef8019b
DM
81#include <net/bonding.h>
82#include <net/bond_3ad.h>
83#include <net/bond_alb.h>
1da177e4 84
73b5a6f2
MB
85#include "bonding_priv.h"
86
1da177e4
LT
87/*---------------------------- Module parameters ----------------------------*/
88
89/* monitor all links that often (in milliseconds). <=0 disables monitoring */
1da177e4
LT
90
91static int max_bonds = BOND_DEFAULT_MAX_BONDS;
bb1d9123 92static int tx_queues = BOND_DEFAULT_TX_QUEUES;
ad246c99 93static int num_peer_notif = 1;
b98d9c66 94static int miimon;
3d632c3f
SH
95static int updelay;
96static int downdelay;
1da177e4 97static int use_carrier = 1;
3d632c3f
SH
98static char *mode;
99static char *primary;
a549952a 100static char *primary_reselect;
3d632c3f 101static char *lacp_rate;
655f8919 102static int min_links;
3d632c3f
SH
103static char *ad_select;
104static char *xmit_hash_policy;
7bdb04ed 105static int arp_interval;
3d632c3f
SH
106static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
107static char *arp_validate;
8599b52e 108static char *arp_all_targets;
3d632c3f 109static char *fail_over_mac;
b07ea07b 110static int all_slaves_active;
d2991f75 111static struct bond_params bonding_defaults;
c2952c31 112static int resend_igmp = BOND_DEFAULT_RESEND_IGMP;
73958329 113static int packets_per_slave = 1;
3a7129e5 114static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
1da177e4
LT
115
116module_param(max_bonds, int, 0);
117MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
bb1d9123
AG
118module_param(tx_queues, int, 0);
119MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)");
ad246c99 120module_param_named(num_grat_arp, num_peer_notif, int, 0644);
90e62474
AG
121MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on "
122 "failover event (alias of num_unsol_na)");
ad246c99 123module_param_named(num_unsol_na, num_peer_notif, int, 0644);
90e62474
AG
124MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on "
125 "failover event (alias of num_grat_arp)");
1da177e4
LT
126module_param(miimon, int, 0);
127MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
128module_param(updelay, int, 0);
129MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
130module_param(downdelay, int, 0);
2ac47660
MW
131MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
132 "in milliseconds");
1da177e4 133module_param(use_carrier, int, 0);
2ac47660
MW
134MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
135 "0 for off, 1 for on (default)");
1da177e4 136module_param(mode, charp, 0);
90e62474 137MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, "
2ac47660
MW
138 "1 for active-backup, 2 for balance-xor, "
139 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
140 "6 for balance-alb");
1da177e4
LT
141module_param(primary, charp, 0);
142MODULE_PARM_DESC(primary, "Primary network device to use");
a549952a
JP
143module_param(primary_reselect, charp, 0);
144MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
145 "once it comes up; "
146 "0 for always (default), "
147 "1 for only if speed of primary is "
148 "better, "
149 "2 for only on active slave "
150 "failure");
1da177e4 151module_param(lacp_rate, charp, 0);
90e62474
AG
152MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; "
153 "0 for slow, 1 for fast");
fd989c83 154module_param(ad_select, charp, 0);
0d039f33 155MODULE_PARM_DESC(ad_select, "802.3ad aggregation selection logic; "
90e62474
AG
156 "0 for stable (default), 1 for bandwidth, "
157 "2 for count");
655f8919 158module_param(min_links, int, 0);
159MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier");
160
169a3e66 161module_param(xmit_hash_policy, charp, 0);
90e62474
AG
162MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; "
163 "0 for layer 2 (default), 1 for layer 3+4, "
32819dc1
NA
164 "2 for layer 2+3, 3 for encap layer 2+3, "
165 "4 for encap layer 3+4");
1da177e4
LT
166module_param(arp_interval, int, 0);
167MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
168module_param_array(arp_ip_target, charp, NULL, 0);
169MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
f5b2b966 170module_param(arp_validate, charp, 0);
90e62474
AG
171MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; "
172 "0 for none (default), 1 for active, "
173 "2 for backup, 3 for all");
8599b52e
VF
174module_param(arp_all_targets, charp, 0);
175MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all");
3915c1e8 176module_param(fail_over_mac, charp, 0);
90e62474
AG
177MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to "
178 "the same MAC; 0 for none (default), "
179 "1 for active, 2 for follow");
ebd8e497 180module_param(all_slaves_active, int, 0);
37b7021d 181MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface "
90e62474 182 "by setting active flag for all slaves; "
ebd8e497 183 "0 for never (default), 1 for always.");
c2952c31 184module_param(resend_igmp, int, 0);
90e62474
AG
185MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on "
186 "link failure");
73958329
NA
187module_param(packets_per_slave, int, 0);
188MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr "
189 "mode; 0 for a random slave, 1 packet per "
190 "slave (default), >1 packets per slave.");
3a7129e5 191module_param(lp_interval, uint, 0);
192MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where "
193 "the bonding driver sends learning packets to "
194 "each slaves peer switch. The default is 1.");
1da177e4
LT
195
196/*----------------------------- Global variables ----------------------------*/
197
e843fa50 198#ifdef CONFIG_NET_POLL_CONTROLLER
fb4fa76a 199atomic_t netpoll_block_tx = ATOMIC_INIT(0);
e843fa50
NH
200#endif
201
c7d03a00 202unsigned int bond_net_id __read_mostly;
1da177e4 203
1da177e4
LT
204/*-------------------------- Forward declarations ---------------------------*/
205
181470fc 206static int bond_init(struct net_device *bond_dev);
c67dfb29 207static void bond_uninit(struct net_device *bond_dev);
bc1f4470 208static void bond_get_stats(struct net_device *bond_dev,
209 struct rtnl_link_stats64 *stats);
ee637714 210static void bond_slave_arr_handler(struct work_struct *work);
21a75f09
JV
211static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
212 int mod);
1da177e4
LT
213
214/*---------------------------- General routines -----------------------------*/
215
bd33acc3 216const char *bond_mode_name(int mode)
1da177e4 217{
77afc92b
HE
218 static const char *names[] = {
219 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
220 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
221 [BOND_MODE_XOR] = "load balancing (xor)",
222 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
3d632c3f 223 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
77afc92b
HE
224 [BOND_MODE_TLB] = "transmit load balancing",
225 [BOND_MODE_ALB] = "adaptive load balancing",
226 };
227
f5280948 228 if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB)
1da177e4 229 return "unknown";
77afc92b
HE
230
231 return names[mode];
1da177e4
LT
232}
233
234/*---------------------------------- VLAN -----------------------------------*/
235
1da177e4
LT
236/**
237 * bond_dev_queue_xmit - Prepare skb for xmit.
3d632c3f 238 *
1da177e4
LT
239 * @bond: bond device that got this skb for tx.
240 * @skb: hw accel VLAN tagged skb to transmit
241 * @slave_dev: slave that is supposed to xmit this skbuff
1da177e4 242 */
d316dedd 243void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
3d632c3f 244 struct net_device *slave_dev)
1da177e4 245{
8387451e 246 skb->dev = slave_dev;
374eeb5a 247
5ee31c68 248 BUILD_BUG_ON(sizeof(skb->queue_mapping) !=
df4ab5b3
JP
249 sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping));
250 skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping;
374eeb5a 251
e15c3c22 252 if (unlikely(netpoll_tx_running(bond->dev)))
8a8efa22 253 bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb);
080e4130 254 else
f6dc31a8 255 dev_queue_xmit(skb);
1da177e4
LT
256}
257
547942ca 258/* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid,
cc0e4070 259 * We don't protect the slave list iteration with a lock because:
1da177e4
LT
260 * a. This operation is performed in IOCTL context,
261 * b. The operation is protected by the RTNL semaphore in the 8021q code,
262 * c. Holding a lock with BH disabled while directly calling a base driver
263 * entry point is generally a BAD idea.
3d632c3f 264 *
1da177e4
LT
265 * The design of synchronization/protection for this operation in the 8021q
266 * module is good for one or more VLAN devices over a single physical device
267 * and cannot be extended for a teaming solution like bonding, so there is a
268 * potential race condition here where a net device from the vlan group might
269 * be referenced (either by a base driver or the 8021q code) while it is being
270 * removed from the system. However, it turns out we're not making matters
271 * worse, and if it works for regular VLAN usage it will work here too.
272*/
273
1da177e4
LT
274/**
275 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
276 * @bond_dev: bonding net device that got called
277 * @vid: vlan id being added
278 */
80d5c368
PM
279static int bond_vlan_rx_add_vid(struct net_device *bond_dev,
280 __be16 proto, u16 vid)
1da177e4 281{
454d7c9b 282 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 283 struct slave *slave, *rollback_slave;
9caff1e7 284 struct list_head *iter;
dec1e90e 285 int res;
1da177e4 286
9caff1e7 287 bond_for_each_slave(bond, slave, iter) {
80d5c368 288 res = vlan_vid_add(slave->dev, proto, vid);
87002b03
JP
289 if (res)
290 goto unwind;
1da177e4
LT
291 }
292
8e586137 293 return 0;
87002b03
JP
294
295unwind:
81f23b13 296 /* unwind to the slave that failed */
9caff1e7 297 bond_for_each_slave(bond, rollback_slave, iter) {
81f23b13
VF
298 if (rollback_slave == slave)
299 break;
300
301 vlan_vid_del(rollback_slave->dev, proto, vid);
302 }
87002b03
JP
303
304 return res;
1da177e4
LT
305}
306
307/**
308 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
309 * @bond_dev: bonding net device that got called
310 * @vid: vlan id being removed
311 */
80d5c368
PM
312static int bond_vlan_rx_kill_vid(struct net_device *bond_dev,
313 __be16 proto, u16 vid)
1da177e4 314{
454d7c9b 315 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 316 struct list_head *iter;
1da177e4 317 struct slave *slave;
1da177e4 318
9caff1e7 319 bond_for_each_slave(bond, slave, iter)
80d5c368 320 vlan_vid_del(slave->dev, proto, vid);
1da177e4 321
e868b0c9
VF
322 if (bond_is_lb(bond))
323 bond_alb_clear_vlan(bond, vid);
8e586137
JP
324
325 return 0;
1da177e4
LT
326}
327
1da177e4
LT
328/*------------------------------- Link status -------------------------------*/
329
547942ca 330/* Set the carrier state for the master according to the state of its
ff59c456
JV
331 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
332 * do special 802.3ad magic.
333 *
334 * Returns zero if carrier state does not change, nonzero if it does.
335 */
2477bc9a 336int bond_set_carrier(struct bonding *bond)
ff59c456 337{
9caff1e7 338 struct list_head *iter;
ff59c456 339 struct slave *slave;
ff59c456 340
0965a1f3 341 if (!bond_has_slaves(bond))
ff59c456
JV
342 goto down;
343
01844098 344 if (BOND_MODE(bond) == BOND_MODE_8023AD)
ff59c456
JV
345 return bond_3ad_set_carrier(bond);
346
9caff1e7 347 bond_for_each_slave(bond, slave, iter) {
ff59c456
JV
348 if (slave->link == BOND_LINK_UP) {
349 if (!netif_carrier_ok(bond->dev)) {
350 netif_carrier_on(bond->dev);
351 return 1;
352 }
353 return 0;
354 }
355 }
356
357down:
358 if (netif_carrier_ok(bond->dev)) {
359 netif_carrier_off(bond->dev);
360 return 1;
361 }
362 return 0;
363}
364
547942ca 365/* Get link speed and duplex from the slave's base driver
1da177e4 366 * using ethtool. If for some reason the call fails or the
98f41f69 367 * values are invalid, set speed and duplex to -1,
c4adfc82
MB
368 * and return. Return 1 if speed or duplex settings are
369 * UNKNOWN; 0 otherwise.
1da177e4 370 */
c4adfc82 371static int bond_update_speed_duplex(struct slave *slave)
1da177e4
LT
372{
373 struct net_device *slave_dev = slave->dev;
9856909c 374 struct ethtool_link_ksettings ecmd;
61a44b9c 375 int res;
1da177e4 376
589665f5
DC
377 slave->speed = SPEED_UNKNOWN;
378 slave->duplex = DUPLEX_UNKNOWN;
1da177e4 379
9856909c 380 res = __ethtool_get_link_ksettings(slave_dev, &ecmd);
c4adfc82
MB
381 if (res < 0) {
382 slave->link = BOND_LINK_DOWN;
383 return 1;
384 }
385 if (ecmd.base.speed == 0 || ecmd.base.speed == ((__u32)-1)) {
386 slave->link = BOND_LINK_DOWN;
387 return 1;
388 }
9856909c 389 switch (ecmd.base.duplex) {
1da177e4
LT
390 case DUPLEX_FULL:
391 case DUPLEX_HALF:
392 break;
393 default:
c4adfc82
MB
394 slave->link = BOND_LINK_DOWN;
395 return 1;
1da177e4
LT
396 }
397
9856909c
DD
398 slave->speed = ecmd.base.speed;
399 slave->duplex = ecmd.base.duplex;
1da177e4 400
c4adfc82 401 return 0;
1da177e4
LT
402}
403
07699f9a 404const char *bond_slave_link_status(s8 link)
405{
406 switch (link) {
407 case BOND_LINK_UP:
408 return "up";
409 case BOND_LINK_FAIL:
410 return "going down";
411 case BOND_LINK_DOWN:
412 return "down";
413 case BOND_LINK_BACK:
414 return "going back";
415 default:
416 return "unknown";
417 }
418}
419
547942ca 420/* if <dev> supports MII link status reporting, check its link status.
1da177e4
LT
421 *
422 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
3d632c3f 423 * depending upon the setting of the use_carrier parameter.
1da177e4
LT
424 *
425 * Return either BMSR_LSTATUS, meaning that the link is up (or we
426 * can't tell and just pretend it is), or 0, meaning that the link is
427 * down.
428 *
429 * If reporting is non-zero, instead of faking link up, return -1 if
430 * both ETHTOOL and MII ioctls fail (meaning the device does not
431 * support them). If use_carrier is set, return whatever it says.
432 * It'd be nice if there was a good way to tell if a driver supports
433 * netif_carrier, but there really isn't.
434 */
3d632c3f
SH
435static int bond_check_dev_link(struct bonding *bond,
436 struct net_device *slave_dev, int reporting)
1da177e4 437{
eb7cc59a 438 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
d9d52832 439 int (*ioctl)(struct net_device *, struct ifreq *, int);
1da177e4
LT
440 struct ifreq ifr;
441 struct mii_ioctl_data *mii;
1da177e4 442
6c988853
PG
443 if (!reporting && !netif_running(slave_dev))
444 return 0;
445
eb7cc59a 446 if (bond->params.use_carrier)
1da177e4 447 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
1da177e4 448
29112f4e 449 /* Try to get link status using Ethtool first. */
c772dde3
BH
450 if (slave_dev->ethtool_ops->get_link)
451 return slave_dev->ethtool_ops->get_link(slave_dev) ?
452 BMSR_LSTATUS : 0;
29112f4e 453
3d632c3f 454 /* Ethtool can't be used, fallback to MII ioctls. */
eb7cc59a 455 ioctl = slave_ops->ndo_do_ioctl;
1da177e4 456 if (ioctl) {
547942ca
NA
457 /* TODO: set pointer to correct ioctl on a per team member
458 * bases to make this more efficient. that is, once
459 * we determine the correct ioctl, we will always
460 * call it and not the others for that team
461 * member.
462 */
463
464 /* We cannot assume that SIOCGMIIPHY will also read a
1da177e4
LT
465 * register; not all network drivers (e.g., e100)
466 * support that.
467 */
468
469 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
470 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
471 mii = if_mii(&ifr);
4ad41c1e 472 if (ioctl(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
1da177e4 473 mii->reg_num = MII_BMSR;
4ad41c1e 474 if (ioctl(slave_dev, &ifr, SIOCGMIIREG) == 0)
3d632c3f 475 return mii->val_out & BMSR_LSTATUS;
1da177e4
LT
476 }
477 }
478
547942ca 479 /* If reporting, report that either there's no dev->do_ioctl,
61a44b9c 480 * or both SIOCGMIIREG and get_link failed (meaning that we
1da177e4
LT
481 * cannot report link status). If not reporting, pretend
482 * we're ok.
483 */
3d632c3f 484 return reporting ? -1 : BMSR_LSTATUS;
1da177e4
LT
485}
486
487/*----------------------------- Multicast list ------------------------------*/
488
547942ca 489/* Push the promiscuity flag down to appropriate slaves */
7e1a1ac1 490static int bond_set_promiscuity(struct bonding *bond, int inc)
1da177e4 491{
9caff1e7 492 struct list_head *iter;
7e1a1ac1 493 int err = 0;
9caff1e7 494
ec0865a9 495 if (bond_uses_primary(bond)) {
14056e79 496 struct slave *curr_active = rtnl_dereference(bond->curr_active_slave);
4740d638 497
4740d638
ED
498 if (curr_active)
499 err = dev_set_promiscuity(curr_active->dev, inc);
1da177e4
LT
500 } else {
501 struct slave *slave;
dec1e90e 502
9caff1e7 503 bond_for_each_slave(bond, slave, iter) {
7e1a1ac1
WC
504 err = dev_set_promiscuity(slave->dev, inc);
505 if (err)
506 return err;
1da177e4
LT
507 }
508 }
7e1a1ac1 509 return err;
1da177e4
LT
510}
511
547942ca 512/* Push the allmulti flag down to all slaves */
7e1a1ac1 513static int bond_set_allmulti(struct bonding *bond, int inc)
1da177e4 514{
9caff1e7 515 struct list_head *iter;
7e1a1ac1 516 int err = 0;
9caff1e7 517
ec0865a9 518 if (bond_uses_primary(bond)) {
14056e79 519 struct slave *curr_active = rtnl_dereference(bond->curr_active_slave);
4740d638 520
4740d638
ED
521 if (curr_active)
522 err = dev_set_allmulti(curr_active->dev, inc);
1da177e4
LT
523 } else {
524 struct slave *slave;
dec1e90e 525
9caff1e7 526 bond_for_each_slave(bond, slave, iter) {
7e1a1ac1
WC
527 err = dev_set_allmulti(slave->dev, inc);
528 if (err)
529 return err;
1da177e4
LT
530 }
531 }
7e1a1ac1 532 return err;
1da177e4
LT
533}
534
547942ca 535/* Retrieve the list of registered multicast addresses for the bonding
5a37e8ca
FL
536 * device and retransmit an IGMP JOIN request to the current active
537 * slave.
538 */
f2369109 539static void bond_resend_igmp_join_requests_delayed(struct work_struct *work)
5a37e8ca 540{
f2369109 541 struct bonding *bond = container_of(work, struct bonding,
542 mcast_work.work);
543
4aa5dee4 544 if (!rtnl_trylock()) {
4beac029 545 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
4aa5dee4 546 return;
5a37e8ca 547 }
4aa5dee4 548 call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev);
5a37e8ca 549
4f5474e7
NA
550 if (bond->igmp_retrans > 1) {
551 bond->igmp_retrans--;
c2952c31 552 queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5);
4f5474e7 553 }
f2369109 554 rtnl_unlock();
5a37e8ca
FL
555}
556
547942ca 557/* Flush bond's hardware addresses from slave */
303d1cbf 558static void bond_hw_addr_flush(struct net_device *bond_dev,
3d632c3f 559 struct net_device *slave_dev)
1da177e4 560{
454d7c9b 561 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 562
303d1cbf
JV
563 dev_uc_unsync(slave_dev, bond_dev);
564 dev_mc_unsync(slave_dev, bond_dev);
1da177e4 565
01844098 566 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1da177e4
LT
567 /* del lacpdu mc addr from mc list */
568 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
569
22bedad3 570 dev_mc_del(slave_dev, lacpdu_multicast);
1da177e4
LT
571 }
572}
573
574/*--------------------------- Active slave change ---------------------------*/
575
303d1cbf 576/* Update the hardware address list and promisc/allmulti for the new and
ec0865a9
VF
577 * old active slaves (if any). Modes that are not using primary keep all
578 * slaves up date at all times; only the modes that use primary need to call
303d1cbf 579 * this function to swap these settings during a failover.
1da177e4 580 */
303d1cbf
JV
581static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active,
582 struct slave *old_active)
1da177e4 583{
1da177e4 584 if (old_active) {
3d632c3f 585 if (bond->dev->flags & IFF_PROMISC)
1da177e4 586 dev_set_promiscuity(old_active->dev, -1);
1da177e4 587
3d632c3f 588 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 589 dev_set_allmulti(old_active->dev, -1);
1da177e4 590
303d1cbf 591 bond_hw_addr_flush(bond->dev, old_active->dev);
1da177e4
LT
592 }
593
594 if (new_active) {
7e1a1ac1 595 /* FIXME: Signal errors upstream. */
3d632c3f 596 if (bond->dev->flags & IFF_PROMISC)
1da177e4 597 dev_set_promiscuity(new_active->dev, 1);
1da177e4 598
3d632c3f 599 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 600 dev_set_allmulti(new_active->dev, 1);
1da177e4 601
d632ce98 602 netif_addr_lock_bh(bond->dev);
303d1cbf
JV
603 dev_uc_sync(new_active->dev, bond->dev);
604 dev_mc_sync(new_active->dev, bond->dev);
d632ce98 605 netif_addr_unlock_bh(bond->dev);
1da177e4
LT
606 }
607}
608
ae0d6750 609/**
610 * bond_set_dev_addr - clone slave's address to bond
611 * @bond_dev: bond net device
612 * @slave_dev: slave net device
613 *
614 * Should be called with RTNL held.
615 */
616static void bond_set_dev_addr(struct net_device *bond_dev,
617 struct net_device *slave_dev)
618{
c6140a29
ZS
619 netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->name=%s slave_dev->addr_len=%d\n",
620 bond_dev, slave_dev, slave_dev->name, slave_dev->addr_len);
ae0d6750 621 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
622 bond_dev->addr_assign_type = NET_ADDR_STOLEN;
623 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev);
624}
625
a951bc1e 626static struct slave *bond_get_old_active(struct bonding *bond,
627 struct slave *new_active)
628{
629 struct slave *slave;
630 struct list_head *iter;
631
632 bond_for_each_slave(bond, slave, iter) {
633 if (slave == new_active)
634 continue;
635
636 if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr))
637 return slave;
638 }
639
640 return NULL;
641}
642
547942ca 643/* bond_do_fail_over_mac
3915c1e8
JV
644 *
645 * Perform special MAC address swapping for fail_over_mac settings
646 *
1c72cfdc 647 * Called with RTNL
3915c1e8
JV
648 */
649static void bond_do_fail_over_mac(struct bonding *bond,
650 struct slave *new_active,
651 struct slave *old_active)
652{
653 u8 tmp_mac[ETH_ALEN];
654 struct sockaddr saddr;
655 int rv;
656
657 switch (bond->params.fail_over_mac) {
658 case BOND_FOM_ACTIVE:
1c72cfdc 659 if (new_active)
ae0d6750 660 bond_set_dev_addr(bond->dev, new_active->dev);
3915c1e8
JV
661 break;
662 case BOND_FOM_FOLLOW:
547942ca 663 /* if new_active && old_active, swap them
3915c1e8
JV
664 * if just old_active, do nothing (going to no active slave)
665 * if just new_active, set new_active to bond's MAC
666 */
667 if (!new_active)
668 return;
669
a951bc1e 670 if (!old_active)
671 old_active = bond_get_old_active(bond, new_active);
672
3915c1e8 673 if (old_active) {
ada0f863 674 ether_addr_copy(tmp_mac, new_active->dev->dev_addr);
2a7c183b
JP
675 ether_addr_copy(saddr.sa_data,
676 old_active->dev->dev_addr);
3915c1e8
JV
677 saddr.sa_family = new_active->dev->type;
678 } else {
ada0f863 679 ether_addr_copy(saddr.sa_data, bond->dev->dev_addr);
3915c1e8
JV
680 saddr.sa_family = bond->dev->type;
681 }
682
683 rv = dev_set_mac_address(new_active->dev, &saddr);
684 if (rv) {
76444f50
VF
685 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
686 -rv, new_active->dev->name);
3915c1e8
JV
687 goto out;
688 }
689
690 if (!old_active)
691 goto out;
692
ada0f863 693 ether_addr_copy(saddr.sa_data, tmp_mac);
3915c1e8
JV
694 saddr.sa_family = old_active->dev->type;
695
696 rv = dev_set_mac_address(old_active->dev, &saddr);
697 if (rv)
76444f50
VF
698 netdev_err(bond->dev, "Error %d setting MAC of slave %s\n",
699 -rv, new_active->dev->name);
3915c1e8 700out:
3915c1e8
JV
701 break;
702 default:
76444f50
VF
703 netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n",
704 bond->params.fail_over_mac);
3915c1e8
JV
705 break;
706 }
707
708}
709
b5a983f3 710static struct slave *bond_choose_primary_or_current(struct bonding *bond)
a549952a 711{
059b47e8 712 struct slave *prim = rtnl_dereference(bond->primary_slave);
1c72cfdc 713 struct slave *curr = rtnl_dereference(bond->curr_active_slave);
a549952a 714
b5a983f3
MR
715 if (!prim || prim->link != BOND_LINK_UP) {
716 if (!curr || curr->link != BOND_LINK_UP)
717 return NULL;
718 return curr;
719 }
720
a549952a
JP
721 if (bond->force_primary) {
722 bond->force_primary = false;
b5a983f3
MR
723 return prim;
724 }
725
726 if (!curr || curr->link != BOND_LINK_UP)
727 return prim;
728
729 /* At this point, prim and curr are both up */
730 switch (bond->params.primary_reselect) {
731 case BOND_PRI_RESELECT_ALWAYS:
732 return prim;
733 case BOND_PRI_RESELECT_BETTER:
734 if (prim->speed < curr->speed)
735 return curr;
736 if (prim->speed == curr->speed && prim->duplex <= curr->duplex)
737 return curr;
738 return prim;
739 case BOND_PRI_RESELECT_FAILURE:
740 return curr;
741 default:
742 netdev_err(bond->dev, "impossible primary_reselect %d\n",
743 bond->params.primary_reselect);
744 return curr;
a549952a 745 }
a549952a 746}
3915c1e8 747
1da177e4 748/**
b5a983f3 749 * bond_find_best_slave - select the best available slave to be the active one
1da177e4 750 * @bond: our bonding struct
1da177e4
LT
751 */
752static struct slave *bond_find_best_slave(struct bonding *bond)
753{
b5a983f3 754 struct slave *slave, *bestslave = NULL;
77140d29 755 struct list_head *iter;
1da177e4 756 int mintime = bond->params.updelay;
1da177e4 757
b5a983f3
MR
758 slave = bond_choose_primary_or_current(bond);
759 if (slave)
760 return slave;
1da177e4 761
77140d29
VF
762 bond_for_each_slave(bond, slave, iter) {
763 if (slave->link == BOND_LINK_UP)
764 return slave;
b6adc610 765 if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) &&
77140d29
VF
766 slave->delay < mintime) {
767 mintime = slave->delay;
768 bestslave = slave;
1da177e4
LT
769 }
770 }
771
772 return bestslave;
773}
774
ad246c99
BH
775static bool bond_should_notify_peers(struct bonding *bond)
776{
4cb4f97b 777 struct slave *slave;
778
779 rcu_read_lock();
780 slave = rcu_dereference(bond->curr_active_slave);
781 rcu_read_unlock();
ad246c99 782
76444f50
VF
783 netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n",
784 slave ? slave->dev->name : "NULL");
ad246c99
BH
785
786 if (!slave || !bond->send_peer_notif ||
b02e3e94 787 !netif_carrier_ok(bond->dev) ||
ad246c99
BH
788 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
789 return false;
790
ad246c99
BH
791 return true;
792}
793
1da177e4
LT
794/**
795 * change_active_interface - change the active slave into the specified one
796 * @bond: our bonding struct
797 * @new: the new slave to make the active one
798 *
799 * Set the new slave to the bond's settings and unset them on the old
800 * curr_active_slave.
801 * Setting include flags, mc-list, promiscuity, allmulti, etc.
802 *
803 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
804 * because it is apparently the best available slave we have, even though its
805 * updelay hasn't timed out yet.
806 *
1c72cfdc 807 * Caller must hold RTNL.
1da177e4 808 */
a77b5325 809void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
1da177e4 810{
4740d638
ED
811 struct slave *old_active;
812
1c72cfdc
NA
813 ASSERT_RTNL();
814
815 old_active = rtnl_dereference(bond->curr_active_slave);
1da177e4 816
3d632c3f 817 if (old_active == new_active)
1da177e4 818 return;
1da177e4
LT
819
820 if (new_active) {
8e603460 821 new_active->last_link_up = jiffies;
b2220cad 822
1da177e4 823 if (new_active->link == BOND_LINK_BACK) {
ec0865a9 824 if (bond_uses_primary(bond)) {
76444f50
VF
825 netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n",
826 new_active->dev->name,
827 (bond->params.updelay - new_active->delay) * bond->params.miimon);
1da177e4
LT
828 }
829
830 new_active->delay = 0;
5d397061
JP
831 bond_set_slave_link_state(new_active, BOND_LINK_UP,
832 BOND_SLAVE_NOTIFY_NOW);
1da177e4 833
01844098 834 if (BOND_MODE(bond) == BOND_MODE_8023AD)
1da177e4 835 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
1da177e4 836
58402054 837 if (bond_is_lb(bond))
1da177e4 838 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
1da177e4 839 } else {
ec0865a9 840 if (bond_uses_primary(bond)) {
76444f50
VF
841 netdev_info(bond->dev, "making interface %s the new active one\n",
842 new_active->dev->name);
1da177e4
LT
843 }
844 }
845 }
846
ec0865a9 847 if (bond_uses_primary(bond))
303d1cbf 848 bond_hw_addr_swap(bond, new_active, old_active);
1da177e4 849
58402054 850 if (bond_is_lb(bond)) {
1da177e4 851 bond_alb_handle_active_change(bond, new_active);
8f903c70 852 if (old_active)
5e5b0665 853 bond_set_slave_inactive_flags(old_active,
854 BOND_SLAVE_NOTIFY_NOW);
8f903c70 855 if (new_active)
5e5b0665 856 bond_set_slave_active_flags(new_active,
857 BOND_SLAVE_NOTIFY_NOW);
1da177e4 858 } else {
278b2083 859 rcu_assign_pointer(bond->curr_active_slave, new_active);
1da177e4 860 }
c3ade5ca 861
01844098 862 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) {
3d632c3f 863 if (old_active)
5e5b0665 864 bond_set_slave_inactive_flags(old_active,
865 BOND_SLAVE_NOTIFY_NOW);
c3ade5ca
JV
866
867 if (new_active) {
ad246c99
BH
868 bool should_notify_peers = false;
869
5e5b0665 870 bond_set_slave_active_flags(new_active,
871 BOND_SLAVE_NOTIFY_NOW);
2ab82852 872
709f8a45
OG
873 if (bond->params.fail_over_mac)
874 bond_do_fail_over_mac(bond, new_active,
875 old_active);
3915c1e8 876
ad246c99
BH
877 if (netif_running(bond->dev)) {
878 bond->send_peer_notif =
879 bond->params.num_peer_notif;
880 should_notify_peers =
881 bond_should_notify_peers(bond);
882 }
883
b7bc2a5b 884 call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
ad246c99 885 if (should_notify_peers)
b7bc2a5b
AW
886 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
887 bond->dev);
7893b249 888 }
c3ade5ca 889 }
a2fd940f 890
5a37e8ca 891 /* resend IGMP joins since active slave has changed or
94265cf5
FL
892 * all were sent on curr_active_slave.
893 * resend only if bond is brought up with the affected
547942ca
NA
894 * bonding modes and the retransmission is enabled
895 */
94265cf5 896 if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) &&
ec0865a9 897 ((bond_uses_primary(bond) && new_active) ||
01844098 898 BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) {
c2952c31 899 bond->igmp_retrans = bond->params.resend_igmp;
4beac029 900 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
a2fd940f 901 }
1da177e4
LT
902}
903
904/**
905 * bond_select_active_slave - select a new active slave, if needed
906 * @bond: our bonding struct
907 *
3d632c3f 908 * This functions should be called when one of the following occurs:
1da177e4
LT
909 * - The old curr_active_slave has been released or lost its link.
910 * - The primary_slave has got its link back.
911 * - A slave has got its link back and there's no old curr_active_slave.
912 *
1c72cfdc 913 * Caller must hold RTNL.
1da177e4 914 */
a77b5325 915void bond_select_active_slave(struct bonding *bond)
1da177e4
LT
916{
917 struct slave *best_slave;
ff59c456 918 int rv;
1da177e4 919
e0974585
NA
920 ASSERT_RTNL();
921
1da177e4 922 best_slave = bond_find_best_slave(bond);
1c72cfdc 923 if (best_slave != rtnl_dereference(bond->curr_active_slave)) {
1da177e4 924 bond_change_active_slave(bond, best_slave);
ff59c456
JV
925 rv = bond_set_carrier(bond);
926 if (!rv)
927 return;
928
d66bd905 929 if (netif_carrier_ok(bond->dev))
76444f50 930 netdev_info(bond->dev, "first active interface up!\n");
d66bd905 931 else
76444f50 932 netdev_info(bond->dev, "now running without any active interface!\n");
1da177e4
LT
933 }
934}
935
f6dc31a8 936#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 937static inline int slave_enable_netpoll(struct slave *slave)
f6dc31a8 938{
8a8efa22
AW
939 struct netpoll *np;
940 int err = 0;
f6dc31a8 941
a8779ec1 942 np = kzalloc(sizeof(*np), GFP_KERNEL);
8a8efa22
AW
943 err = -ENOMEM;
944 if (!np)
945 goto out;
946
a8779ec1 947 err = __netpoll_setup(np, slave->dev);
8a8efa22
AW
948 if (err) {
949 kfree(np);
950 goto out;
f6dc31a8 951 }
8a8efa22
AW
952 slave->np = np;
953out:
954 return err;
955}
956static inline void slave_disable_netpoll(struct slave *slave)
957{
958 struct netpoll *np = slave->np;
959
960 if (!np)
961 return;
962
963 slave->np = NULL;
2cde6acd 964 __netpoll_free_async(np);
8a8efa22 965}
f6dc31a8
WC
966
967static void bond_poll_controller(struct net_device *bond_dev)
968{
616f4541
MB
969 struct bonding *bond = netdev_priv(bond_dev);
970 struct slave *slave = NULL;
971 struct list_head *iter;
972 struct ad_info ad_info;
973 struct netpoll_info *ni;
974 const struct net_device_ops *ops;
975
976 if (BOND_MODE(bond) == BOND_MODE_8023AD)
977 if (bond_3ad_get_active_agg_info(bond, &ad_info))
978 return;
979
616f4541
MB
980 bond_for_each_slave_rcu(bond, slave, iter) {
981 ops = slave->dev->netdev_ops;
982 if (!bond_slave_is_up(slave) || !ops->ndo_poll_controller)
983 continue;
984
985 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
986 struct aggregator *agg =
987 SLAVE_AD_INFO(slave)->port.aggregator;
988
989 if (agg &&
990 agg->aggregator_identifier != ad_info.aggregator_id)
991 continue;
992 }
993
994 ni = rcu_dereference_bh(slave->dev->npinfo);
995 if (down_trylock(&ni->dev_lock))
996 continue;
997 ops->ndo_poll_controller(slave->dev);
998 up(&ni->dev_lock);
999 }
8a8efa22
AW
1000}
1001
c4cdef9b 1002static void bond_netpoll_cleanup(struct net_device *bond_dev)
8a8efa22 1003{
c4cdef9b 1004 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 1005 struct list_head *iter;
c2355e1a 1006 struct slave *slave;
c2355e1a 1007
9caff1e7 1008 bond_for_each_slave(bond, slave, iter)
b6adc610 1009 if (bond_slave_is_up(slave))
8a8efa22 1010 slave_disable_netpoll(slave);
f6dc31a8 1011}
8a8efa22 1012
a8779ec1 1013static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
8a8efa22
AW
1014{
1015 struct bonding *bond = netdev_priv(dev);
9caff1e7 1016 struct list_head *iter;
f6dc31a8 1017 struct slave *slave;
dec1e90e 1018 int err = 0;
f6dc31a8 1019
9caff1e7 1020 bond_for_each_slave(bond, slave, iter) {
8a8efa22
AW
1021 err = slave_enable_netpoll(slave);
1022 if (err) {
c4cdef9b 1023 bond_netpoll_cleanup(dev);
8a8efa22 1024 break;
f6dc31a8
WC
1025 }
1026 }
8a8efa22 1027 return err;
f6dc31a8 1028}
8a8efa22
AW
1029#else
1030static inline int slave_enable_netpoll(struct slave *slave)
1031{
1032 return 0;
1033}
1034static inline void slave_disable_netpoll(struct slave *slave)
1035{
1036}
f6dc31a8
WC
1037static void bond_netpoll_cleanup(struct net_device *bond_dev)
1038{
1039}
f6dc31a8
WC
1040#endif
1041
1da177e4
LT
1042/*---------------------------------- IOCTL ----------------------------------*/
1043
c8f44aff 1044static netdev_features_t bond_fix_features(struct net_device *dev,
9b7b165a 1045 netdev_features_t features)
8531c5ff 1046{
b2a103e6 1047 struct bonding *bond = netdev_priv(dev);
9caff1e7 1048 struct list_head *iter;
c8f44aff 1049 netdev_features_t mask;
9b7b165a 1050 struct slave *slave;
b63365a2 1051
7889cbee 1052 mask = features;
c158cba3 1053
b63365a2 1054 features &= ~NETIF_F_ONE_FOR_ALL;
b2a103e6 1055 features |= NETIF_F_ALL_FOR_ALL;
7f353bf2 1056
9caff1e7 1057 bond_for_each_slave(bond, slave, iter) {
b63365a2
HX
1058 features = netdev_increment_features(features,
1059 slave->dev->features,
b2a103e6
MM
1060 mask);
1061 }
b0ce3508 1062 features = netdev_add_tso_features(features, mask);
b2a103e6 1063
b2a103e6
MM
1064 return features;
1065}
1066
a188222b 1067#define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \
62f2a3a4
MM
1068 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
1069 NETIF_F_HIGHDMA | NETIF_F_LRO)
b2a103e6 1070
a188222b
TH
1071#define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \
1072 NETIF_F_RXCSUM | NETIF_F_ALL_TSO)
5a7baa78 1073
b2a103e6
MM
1074static void bond_compute_features(struct bonding *bond)
1075{
02875878
ED
1076 unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE |
1077 IFF_XMIT_DST_RELEASE_PERM;
c8f44aff 1078 netdev_features_t vlan_features = BOND_VLAN_FEATURES;
5a7baa78 1079 netdev_features_t enc_features = BOND_ENC_FEATURES;
9caff1e7
VF
1080 struct net_device *bond_dev = bond->dev;
1081 struct list_head *iter;
1082 struct slave *slave;
b2a103e6 1083 unsigned short max_hard_header_len = ETH_HLEN;
0e376bd0
SB
1084 unsigned int gso_max_size = GSO_MAX_SIZE;
1085 u16 gso_max_segs = GSO_MAX_SEGS;
b2a103e6 1086
0965a1f3 1087 if (!bond_has_slaves(bond))
b2a103e6 1088 goto done;
a9b3ace4 1089 vlan_features &= NETIF_F_ALL_FOR_ALL;
b2a103e6 1090
9caff1e7 1091 bond_for_each_slave(bond, slave, iter) {
278339a4 1092 vlan_features = netdev_increment_features(vlan_features,
b2a103e6
MM
1093 slave->dev->vlan_features, BOND_VLAN_FEATURES);
1094
5a7baa78
OG
1095 enc_features = netdev_increment_features(enc_features,
1096 slave->dev->hw_enc_features,
1097 BOND_ENC_FEATURES);
b6fe83e9 1098 dst_release_flag &= slave->dev->priv_flags;
54ef3137
JV
1099 if (slave->dev->hard_header_len > max_hard_header_len)
1100 max_hard_header_len = slave->dev->hard_header_len;
0e376bd0
SB
1101
1102 gso_max_size = min(gso_max_size, slave->dev->gso_max_size);
1103 gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs);
54ef3137 1104 }
8531c5ff 1105
b63365a2 1106done:
b2a103e6 1107 bond_dev->vlan_features = vlan_features;
24f87d4c 1108 bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL;
54ef3137 1109 bond_dev->hard_header_len = max_hard_header_len;
0e376bd0
SB
1110 bond_dev->gso_max_segs = gso_max_segs;
1111 netif_set_gso_max_size(bond_dev, gso_max_size);
8531c5ff 1112
02875878
ED
1113 bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1114 if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) &&
1115 dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM))
1116 bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE;
b6fe83e9 1117
b2a103e6 1118 netdev_change_features(bond_dev);
8531c5ff
AK
1119}
1120
872254dd
MS
1121static void bond_setup_by_slave(struct net_device *bond_dev,
1122 struct net_device *slave_dev)
1123{
00829823 1124 bond_dev->header_ops = slave_dev->header_ops;
872254dd
MS
1125
1126 bond_dev->type = slave_dev->type;
1127 bond_dev->hard_header_len = slave_dev->hard_header_len;
1128 bond_dev->addr_len = slave_dev->addr_len;
1129
1130 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1131 slave_dev->addr_len);
1132}
1133
5b2c4dd2 1134/* On bonding slaves other than the currently active slave, suppress
3aba891d 1135 * duplicates except for alb non-mcast/bcast.
5b2c4dd2
JP
1136 */
1137static bool bond_should_deliver_exact_match(struct sk_buff *skb,
0bd80dad
JP
1138 struct slave *slave,
1139 struct bonding *bond)
5b2c4dd2 1140{
2d7011ca 1141 if (bond_is_slave_inactive(slave)) {
01844098 1142 if (BOND_MODE(bond) == BOND_MODE_ALB &&
5b2c4dd2
JP
1143 skb->pkt_type != PACKET_BROADCAST &&
1144 skb->pkt_type != PACKET_MULTICAST)
5b2c4dd2 1145 return false;
5b2c4dd2
JP
1146 return true;
1147 }
1148 return false;
1149}
1150
8a4eb573 1151static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
5b2c4dd2 1152{
8a4eb573 1153 struct sk_buff *skb = *pskb;
f1c1775a 1154 struct slave *slave;
0bd80dad 1155 struct bonding *bond;
de063b70
ED
1156 int (*recv_probe)(const struct sk_buff *, struct bonding *,
1157 struct slave *);
13a8e0c8 1158 int ret = RX_HANDLER_ANOTHER;
5b2c4dd2 1159
8a4eb573
JP
1160 skb = skb_share_check(skb, GFP_ATOMIC);
1161 if (unlikely(!skb))
1162 return RX_HANDLER_CONSUMED;
1163
1164 *pskb = skb;
5b2c4dd2 1165
35d48903
JP
1166 slave = bond_slave_get_rcu(skb->dev);
1167 bond = slave->bond;
0bd80dad 1168
4d97480b
MH
1169 recv_probe = ACCESS_ONCE(bond->recv_probe);
1170 if (recv_probe) {
de063b70
ED
1171 ret = recv_probe(skb, bond, slave);
1172 if (ret == RX_HANDLER_CONSUMED) {
1173 consume_skb(skb);
1174 return ret;
3aba891d
JP
1175 }
1176 }
1177
d66bd905 1178 if (bond_should_deliver_exact_match(skb, slave, bond))
8a4eb573 1179 return RX_HANDLER_EXACT;
5b2c4dd2 1180
35d48903 1181 skb->dev = bond->dev;
5b2c4dd2 1182
01844098 1183 if (BOND_MODE(bond) == BOND_MODE_ALB &&
35d48903 1184 bond->dev->priv_flags & IFF_BRIDGE_PORT &&
5b2c4dd2 1185 skb->pkt_type == PACKET_HOST) {
5b2c4dd2 1186
541ac7c9
CG
1187 if (unlikely(skb_cow_head(skb,
1188 skb->data - skb_mac_header(skb)))) {
1189 kfree_skb(skb);
8a4eb573 1190 return RX_HANDLER_CONSUMED;
541ac7c9 1191 }
2a7c183b 1192 ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr);
5b2c4dd2
JP
1193 }
1194
13a8e0c8 1195 return ret;
5b2c4dd2
JP
1196}
1197
41f0b049 1198static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond)
471cb5a3 1199{
41f0b049
JP
1200 switch (BOND_MODE(bond)) {
1201 case BOND_MODE_ROUNDROBIN:
1202 return NETDEV_LAG_TX_TYPE_ROUNDROBIN;
1203 case BOND_MODE_ACTIVEBACKUP:
1204 return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP;
1205 case BOND_MODE_BROADCAST:
1206 return NETDEV_LAG_TX_TYPE_BROADCAST;
1207 case BOND_MODE_XOR:
1208 case BOND_MODE_8023AD:
1209 return NETDEV_LAG_TX_TYPE_HASH;
1210 default:
1211 return NETDEV_LAG_TX_TYPE_UNKNOWN;
1212 }
1213}
1214
1215static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave)
1216{
1217 struct netdev_lag_upper_info lag_upper_info;
471cb5a3
JP
1218 int err;
1219
41f0b049
JP
1220 lag_upper_info.tx_type = bond_lag_tx_type(bond);
1221 err = netdev_master_upper_dev_link(slave->dev, bond->dev, slave,
1222 &lag_upper_info);
471cb5a3
JP
1223 if (err)
1224 return err;
41f0b049 1225 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP
1226 return 0;
1227}
1228
41f0b049 1229static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave)
471cb5a3 1230{
41f0b049
JP
1231 netdev_upper_dev_unlink(slave->dev, bond->dev);
1232 slave->dev->flags &= ~IFF_SLAVE;
1233 rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP
1234}
1235
3fdddd85 1236static struct slave *bond_alloc_slave(struct bonding *bond)
1237{
1238 struct slave *slave = NULL;
1239
d66bd905 1240 slave = kzalloc(sizeof(*slave), GFP_KERNEL);
3fdddd85 1241 if (!slave)
1242 return NULL;
1243
01844098 1244 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3fdddd85 1245 SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info),
1246 GFP_KERNEL);
1247 if (!SLAVE_AD_INFO(slave)) {
1248 kfree(slave);
1249 return NULL;
1250 }
1251 }
1252 return slave;
1253}
1254
1255static void bond_free_slave(struct slave *slave)
1256{
1257 struct bonding *bond = bond_get_bond_by_slave(slave);
1258
01844098 1259 if (BOND_MODE(bond) == BOND_MODE_8023AD)
3fdddd85 1260 kfree(SLAVE_AD_INFO(slave));
1261
1262 kfree(slave);
1263}
1264
69a2338e
MS
1265static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info)
1266{
1267 info->bond_mode = BOND_MODE(bond);
1268 info->miimon = bond->params.miimon;
1269 info->num_slaves = bond->slave_cnt;
1270}
1271
1272static void bond_fill_ifslave(struct slave *slave, struct ifslave *info)
1273{
1274 strcpy(info->slave_name, slave->dev->name);
1275 info->link = slave->link;
1276 info->state = bond_slave_state(slave);
1277 info->link_failure_count = slave->link_failure_count;
1278}
1279
92e584fe
MS
1280static void bond_netdev_notify(struct net_device *dev,
1281 struct netdev_bonding_info *info)
69e61133 1282{
69e61133 1283 rtnl_lock();
92e584fe 1284 netdev_bonding_info_change(dev, info);
69e61133
MS
1285 rtnl_unlock();
1286}
1287
1288static void bond_netdev_notify_work(struct work_struct *_work)
1289{
1290 struct netdev_notify_work *w =
1291 container_of(_work, struct netdev_notify_work, work.work);
1292
92e584fe 1293 bond_netdev_notify(w->dev, &w->bonding_info);
69e61133 1294 dev_put(w->dev);
92e584fe 1295 kfree(w);
69e61133
MS
1296}
1297
1298void bond_queue_slave_event(struct slave *slave)
1299{
92e584fe 1300 struct bonding *bond = slave->bond;
69e61133
MS
1301 struct netdev_notify_work *nnw = kzalloc(sizeof(*nnw), GFP_ATOMIC);
1302
1303 if (!nnw)
1304 return;
1305
92e584fe 1306 dev_hold(slave->dev);
69e61133 1307 nnw->dev = slave->dev;
92e584fe
MS
1308 bond_fill_ifslave(slave, &nnw->bonding_info.slave);
1309 bond_fill_ifbond(bond, &nnw->bonding_info.master);
1310 INIT_DELAYED_WORK(&nnw->work, bond_netdev_notify_work);
69e61133 1311
92e584fe 1312 queue_delayed_work(slave->bond->wq, &nnw->work, 0);
69e61133
MS
1313}
1314
f7c7eb7f
JP
1315void bond_lower_state_changed(struct slave *slave)
1316{
1317 struct netdev_lag_lower_state_info info;
1318
1319 info.link_up = slave->link == BOND_LINK_UP ||
1320 slave->link == BOND_LINK_FAIL;
1321 info.tx_enabled = bond_is_active_slave(slave);
1322 netdev_lower_state_changed(slave->dev, &info);
1323}
1324
1da177e4 1325/* enslave device <slave> to bond device <master> */
a77b5325 1326int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1327{
454d7c9b 1328 struct bonding *bond = netdev_priv(bond_dev);
eb7cc59a 1329 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
c8c23903 1330 struct slave *new_slave = NULL, *prev_slave;
1da177e4
LT
1331 struct sockaddr addr;
1332 int link_reporting;
8599b52e 1333 int res = 0, i;
1da177e4 1334
c772dde3
BH
1335 if (!bond->params.use_carrier &&
1336 slave_dev->ethtool_ops->get_link == NULL &&
1337 slave_ops->ndo_do_ioctl == NULL) {
76444f50
VF
1338 netdev_warn(bond_dev, "no link monitoring support for %s\n",
1339 slave_dev->name);
1da177e4
LT
1340 }
1341
24b27fc4
MB
1342 /* already in-use? */
1343 if (netdev_is_rx_handler_busy(slave_dev)) {
1344 netdev_err(bond_dev,
1345 "Error: Device is in use and cannot be enslaved\n");
1da177e4
LT
1346 return -EBUSY;
1347 }
1348
09a89c21 1349 if (bond_dev == slave_dev) {
76444f50 1350 netdev_err(bond_dev, "cannot enslave bond to itself.\n");
09a89c21
JB
1351 return -EPERM;
1352 }
1353
1da177e4
LT
1354 /* vlan challenged mutual exclusion */
1355 /* no need to lock since we're protected by rtnl_lock */
1356 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
76444f50
VF
1357 netdev_dbg(bond_dev, "%s is NETIF_F_VLAN_CHALLENGED\n",
1358 slave_dev->name);
55462cf3 1359 if (vlan_uses_dev(bond_dev)) {
76444f50
VF
1360 netdev_err(bond_dev, "Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1361 slave_dev->name, bond_dev->name);
1da177e4
LT
1362 return -EPERM;
1363 } else {
76444f50
VF
1364 netdev_warn(bond_dev, "enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1365 slave_dev->name, slave_dev->name,
1366 bond_dev->name);
1da177e4
LT
1367 }
1368 } else {
76444f50
VF
1369 netdev_dbg(bond_dev, "%s is !NETIF_F_VLAN_CHALLENGED\n",
1370 slave_dev->name);
1da177e4
LT
1371 }
1372
547942ca 1373 /* Old ifenslave binaries are no longer supported. These can
3d632c3f 1374 * be identified with moderate accuracy by the state of the slave:
217df670
JV
1375 * the current ifenslave will set the interface down prior to
1376 * enslaving it; the old ifenslave will not.
1377 */
ce3ea1c7 1378 if (slave_dev->flags & IFF_UP) {
76444f50
VF
1379 netdev_err(bond_dev, "%s is up - this may be due to an out of date ifenslave\n",
1380 slave_dev->name);
1e2a8868 1381 return -EPERM;
217df670 1382 }
1da177e4 1383
872254dd
MS
1384 /* set bonding device ether type by slave - bonding netdevices are
1385 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1386 * there is a need to override some of the type dependent attribs/funcs.
1387 *
1388 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1389 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1390 */
0965a1f3 1391 if (!bond_has_slaves(bond)) {
e36b9d16 1392 if (bond_dev->type != slave_dev->type) {
76444f50
VF
1393 netdev_dbg(bond_dev, "change device type from %d to %d\n",
1394 bond_dev->type, slave_dev->type);
75c78500 1395
b7bc2a5b
AW
1396 res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE,
1397 bond_dev);
3ca5b404
JP
1398 res = notifier_to_errno(res);
1399 if (res) {
76444f50 1400 netdev_err(bond_dev, "refused to change device type\n");
1e2a8868 1401 return -EBUSY;
3ca5b404 1402 }
75c78500 1403
32a806c1 1404 /* Flush unicast and multicast addresses */
a748ee24 1405 dev_uc_flush(bond_dev);
22bedad3 1406 dev_mc_flush(bond_dev);
32a806c1 1407
e36b9d16
MS
1408 if (slave_dev->type != ARPHRD_ETHER)
1409 bond_setup_by_slave(bond_dev, slave_dev);
550fd08c 1410 else {
e36b9d16 1411 ether_setup(bond_dev);
550fd08c
NH
1412 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1413 }
75c78500 1414
b7bc2a5b
AW
1415 call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE,
1416 bond_dev);
e36b9d16 1417 }
872254dd 1418 } else if (bond_dev->type != slave_dev->type) {
76444f50
VF
1419 netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n",
1420 slave_dev->name, slave_dev->type, bond_dev->type);
1e2a8868 1421 return -EINVAL;
872254dd
MS
1422 }
1423
1533e773
MB
1424 if (slave_dev->type == ARPHRD_INFINIBAND &&
1425 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
1426 netdev_warn(bond_dev, "Type (%d) supports only active-backup mode\n",
1427 slave_dev->type);
1428 res = -EOPNOTSUPP;
1429 goto err_undo_flags;
1430 }
1431
1432 if (!slave_ops->ndo_set_mac_address ||
1433 slave_dev->type == ARPHRD_INFINIBAND) {
76444f50 1434 netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n");
f5442441
VF
1435 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP &&
1436 bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
1437 if (!bond_has_slaves(bond)) {
00503b6f 1438 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
76444f50 1439 netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n");
f5442441 1440 } else {
76444f50 1441 netdev_err(bond_dev, "The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n");
f5442441
VF
1442 res = -EOPNOTSUPP;
1443 goto err_undo_flags;
00503b6f 1444 }
2ab82852 1445 }
1da177e4
LT
1446 }
1447
8d8fc29d
AW
1448 call_netdevice_notifiers(NETDEV_JOIN, slave_dev);
1449
c20811a7 1450 /* If this is the first slave, then we need to set the master's hardware
547942ca
NA
1451 * address to be the same as the slave's.
1452 */
0965a1f3 1453 if (!bond_has_slaves(bond) &&
dec1e90e 1454 bond->dev->addr_assign_type == NET_ADDR_RANDOM)
409cc1f8 1455 bond_set_dev_addr(bond->dev, slave_dev);
c20811a7 1456
3fdddd85 1457 new_slave = bond_alloc_slave(bond);
1da177e4
LT
1458 if (!new_slave) {
1459 res = -ENOMEM;
1460 goto err_undo_flags;
1461 }
3fdddd85 1462
dc73c41f
VF
1463 new_slave->bond = bond;
1464 new_slave->dev = slave_dev;
547942ca 1465 /* Set the new_slave's queue_id to be zero. Queue ID mapping
bb1d9123
AG
1466 * is set via sysfs or module option if desired.
1467 */
1468 new_slave->queue_id = 0;
1469
b15ba0fb
JP
1470 /* Save slave's original mtu and then set it to match the bond */
1471 new_slave->original_mtu = slave_dev->mtu;
1472 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1473 if (res) {
76444f50 1474 netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res);
b15ba0fb
JP
1475 goto err_free;
1476 }
1477
547942ca 1478 /* Save slave's original ("permanent") mac address for modes
217df670
JV
1479 * that need it, and for restoring it upon release, and then
1480 * set it to the master's address
1481 */
ada0f863 1482 ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr);
1da177e4 1483
00503b6f 1484 if (!bond->params.fail_over_mac ||
01844098 1485 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
547942ca 1486 /* Set slave to master's mac address. The application already
2ab82852
MS
1487 * set the master's mac address to that of the first slave
1488 */
1489 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1490 addr.sa_family = slave_dev->type;
1491 res = dev_set_mac_address(slave_dev, &addr);
1492 if (res) {
76444f50 1493 netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res);
b15ba0fb 1494 goto err_restore_mtu;
2ab82852 1495 }
217df670 1496 }
1da177e4 1497
03d84a5f
KH
1498 /* set slave flag before open to prevent IPv6 addrconf */
1499 slave_dev->flags |= IFF_SLAVE;
1500
217df670
JV
1501 /* open the slave since the application closed it */
1502 res = dev_open(slave_dev);
1503 if (res) {
76444f50 1504 netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name);
1f718f0f 1505 goto err_restore_mac;
1da177e4
LT
1506 }
1507
0b680e75 1508 slave_dev->priv_flags |= IFF_BONDING;
5f0c5f73
AG
1509 /* initialize slave stats */
1510 dev_get_stats(new_slave->dev, &new_slave->slave_stats);
1da177e4 1511
58402054 1512 if (bond_is_lb(bond)) {
1da177e4
LT
1513 /* bond_alb_init_slave() must be called before all other stages since
1514 * it might fail and we do not want to have to undo everything
1515 */
1516 res = bond_alb_init_slave(bond, new_slave);
3d632c3f 1517 if (res)
569f0c4d 1518 goto err_close;
1da177e4
LT
1519 }
1520
ec0865a9 1521 /* If the mode uses primary, then the following is handled by
303d1cbf 1522 * bond_change_active_slave().
1da177e4 1523 */
ec0865a9 1524 if (!bond_uses_primary(bond)) {
1da177e4
LT
1525 /* set promiscuity level to new slave */
1526 if (bond_dev->flags & IFF_PROMISC) {
7e1a1ac1
WC
1527 res = dev_set_promiscuity(slave_dev, 1);
1528 if (res)
1529 goto err_close;
1da177e4
LT
1530 }
1531
1532 /* set allmulti level to new slave */
1533 if (bond_dev->flags & IFF_ALLMULTI) {
7e1a1ac1
WC
1534 res = dev_set_allmulti(slave_dev, 1);
1535 if (res)
1536 goto err_close;
1da177e4
LT
1537 }
1538
b9e40857 1539 netif_addr_lock_bh(bond_dev);
303d1cbf
JV
1540
1541 dev_mc_sync_multiple(slave_dev, bond_dev);
1542 dev_uc_sync_multiple(slave_dev, bond_dev);
1543
b9e40857 1544 netif_addr_unlock_bh(bond_dev);
1da177e4
LT
1545 }
1546
01844098 1547 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1da177e4
LT
1548 /* add lacpdu mc addr to mc list */
1549 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1550
22bedad3 1551 dev_mc_add(slave_dev, lacpdu_multicast);
1da177e4
LT
1552 }
1553
b8e2fde4
WY
1554 res = vlan_vids_add_by_dev(slave_dev, bond_dev);
1555 if (res) {
76444f50
VF
1556 netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n",
1557 slave_dev->name);
1ff412ad 1558 goto err_close;
1559 }
1da177e4 1560
c8c23903 1561 prev_slave = bond_last_slave(bond);
1da177e4
LT
1562
1563 new_slave->delay = 0;
1564 new_slave->link_failure_count = 0;
1565
876254ae
VF
1566 bond_update_speed_duplex(new_slave);
1567
49f17de7 1568 new_slave->last_rx = jiffies -
f31c7937 1569 (msecs_to_jiffies(bond->params.arp_interval) + 1);
8599b52e 1570 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
49f17de7 1571 new_slave->target_last_arp_rx[i] = new_slave->last_rx;
f5b2b966 1572
1da177e4
LT
1573 if (bond->params.miimon && !bond->params.use_carrier) {
1574 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1575
1576 if ((link_reporting == -1) && !bond->params.arp_interval) {
547942ca 1577 /* miimon is set but a bonded network driver
1da177e4
LT
1578 * does not support ETHTOOL/MII and
1579 * arp_interval is not set. Note: if
1580 * use_carrier is enabled, we will never go
1581 * here (because netif_carrier is always
1582 * supported); thus, we don't need to change
1583 * the messages for netif_carrier.
1584 */
76444f50
VF
1585 netdev_warn(bond_dev, "MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n",
1586 slave_dev->name);
1da177e4
LT
1587 } else if (link_reporting == -1) {
1588 /* unable get link status using mii/ethtool */
76444f50
VF
1589 netdev_warn(bond_dev, "can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface\n",
1590 slave_dev->name);
1da177e4
LT
1591 }
1592 }
1593
1594 /* check for initial state */
a30b0168 1595 new_slave->link = BOND_LINK_NOCHANGE;
f31c7937
MK
1596 if (bond->params.miimon) {
1597 if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) {
1598 if (bond->params.updelay) {
69a2338e 1599 bond_set_slave_link_state(new_slave,
5d397061
JP
1600 BOND_LINK_BACK,
1601 BOND_SLAVE_NOTIFY_NOW);
f31c7937
MK
1602 new_slave->delay = bond->params.updelay;
1603 } else {
69a2338e 1604 bond_set_slave_link_state(new_slave,
5d397061
JP
1605 BOND_LINK_UP,
1606 BOND_SLAVE_NOTIFY_NOW);
f31c7937 1607 }
1da177e4 1608 } else {
5d397061
JP
1609 bond_set_slave_link_state(new_slave, BOND_LINK_DOWN,
1610 BOND_SLAVE_NOTIFY_NOW);
1da177e4 1611 }
f31c7937 1612 } else if (bond->params.arp_interval) {
69a2338e
MS
1613 bond_set_slave_link_state(new_slave,
1614 (netif_carrier_ok(slave_dev) ?
5d397061
JP
1615 BOND_LINK_UP : BOND_LINK_DOWN),
1616 BOND_SLAVE_NOTIFY_NOW);
1da177e4 1617 } else {
5d397061
JP
1618 bond_set_slave_link_state(new_slave, BOND_LINK_UP,
1619 BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT
1620 }
1621
f31c7937 1622 if (new_slave->link != BOND_LINK_DOWN)
8e603460 1623 new_slave->last_link_up = jiffies;
76444f50
VF
1624 netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n",
1625 new_slave->link == BOND_LINK_DOWN ? "DOWN" :
1626 (new_slave->link == BOND_LINK_UP ? "UP" : "BACK"));
f31c7937 1627
ec0865a9 1628 if (bond_uses_primary(bond) && bond->params.primary[0]) {
1da177e4 1629 /* if there is a primary slave, remember it */
a549952a 1630 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
059b47e8 1631 rcu_assign_pointer(bond->primary_slave, new_slave);
a549952a
JP
1632 bond->force_primary = true;
1633 }
1da177e4
LT
1634 }
1635
01844098 1636 switch (BOND_MODE(bond)) {
1da177e4 1637 case BOND_MODE_ACTIVEBACKUP:
5e5b0665 1638 bond_set_slave_inactive_flags(new_slave,
1639 BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT
1640 break;
1641 case BOND_MODE_8023AD:
1642 /* in 802.3ad mode, the internal mechanism
1643 * will activate the slaves in the selected
1644 * aggregator
1645 */
5e5b0665 1646 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1da177e4 1647 /* if this is the first slave */
23c147e0 1648 if (!prev_slave) {
3fdddd85 1649 SLAVE_AD_INFO(new_slave)->id = 1;
1da177e4
LT
1650 /* Initialize AD with the number of times that the AD timer is called in 1 second
1651 * can be called only after the mac address of the bond is set
1652 */
56d00c67 1653 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL);
1da177e4 1654 } else {
3fdddd85 1655 SLAVE_AD_INFO(new_slave)->id =
1656 SLAVE_AD_INFO(prev_slave)->id + 1;
1da177e4
LT
1657 }
1658
1659 bond_3ad_bind_slave(new_slave);
1660 break;
1661 case BOND_MODE_TLB:
1662 case BOND_MODE_ALB:
e30bc066 1663 bond_set_active_slave(new_slave);
5e5b0665 1664 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT
1665 break;
1666 default:
76444f50 1667 netdev_dbg(bond_dev, "This slave is always active in trunk mode\n");
1da177e4
LT
1668
1669 /* always active in trunk mode */
e30bc066 1670 bond_set_active_slave(new_slave);
1da177e4
LT
1671
1672 /* In trunking mode there is little meaning to curr_active_slave
1673 * anyway (it holds no special properties of the bond device),
1674 * so we can change it without calling change_active_interface()
1675 */
4740d638
ED
1676 if (!rcu_access_pointer(bond->curr_active_slave) &&
1677 new_slave->link == BOND_LINK_UP)
278b2083 1678 rcu_assign_pointer(bond->curr_active_slave, new_slave);
3d632c3f 1679
1da177e4
LT
1680 break;
1681 } /* switch(bond_mode) */
1682
f6dc31a8 1683#ifdef CONFIG_NET_POLL_CONTROLLER
10eccb46 1684 slave_dev->npinfo = bond->dev->npinfo;
8a8efa22
AW
1685 if (slave_dev->npinfo) {
1686 if (slave_enable_netpoll(new_slave)) {
76444f50 1687 netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n");
8a8efa22 1688 res = -EBUSY;
f7d9821a 1689 goto err_detach;
8a8efa22 1690 }
f6dc31a8
WC
1691 }
1692#endif
8a8efa22 1693
fbe168ba
MK
1694 if (!(bond_dev->features & NETIF_F_LRO))
1695 dev_disable_lro(slave_dev);
1696
35d48903
JP
1697 res = netdev_rx_handler_register(slave_dev, bond_handle_frame,
1698 new_slave);
1699 if (res) {
76444f50 1700 netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res);
5831d66e 1701 goto err_detach;
35d48903
JP
1702 }
1703
41f0b049 1704 res = bond_master_upper_dev_link(bond, new_slave);
1f718f0f 1705 if (res) {
76444f50 1706 netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res);
1f718f0f
VF
1707 goto err_unregister;
1708 }
1709
07699f9a 1710 res = bond_sysfs_slave_add(new_slave);
1711 if (res) {
76444f50 1712 netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res);
07699f9a 1713 goto err_upper_unlink;
1714 }
1715
5378c2e6
VF
1716 bond->slave_cnt++;
1717 bond_compute_features(bond);
1718 bond_set_carrier(bond);
1719
ec0865a9 1720 if (bond_uses_primary(bond)) {
f80889a5 1721 block_netpoll_tx();
5378c2e6 1722 bond_select_active_slave(bond);
f80889a5 1723 unblock_netpoll_tx();
5378c2e6 1724 }
1f718f0f 1725
ee637714
MB
1726 if (bond_mode_uses_xmit_hash(bond))
1727 bond_update_slave_arr(bond, NULL);
1728
76444f50
VF
1729 netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n",
1730 slave_dev->name,
1731 bond_is_active_slave(new_slave) ? "an active" : "a backup",
1732 new_slave->link != BOND_LINK_DOWN ? "an up" : "a down");
1da177e4
LT
1733
1734 /* enslave is successful */
69e61133 1735 bond_queue_slave_event(new_slave);
1da177e4
LT
1736 return 0;
1737
1738/* Undo stages on error */
07699f9a 1739err_upper_unlink:
41f0b049 1740 bond_upper_dev_unlink(bond, new_slave);
07699f9a 1741
1f718f0f
VF
1742err_unregister:
1743 netdev_rx_handler_unregister(slave_dev);
1744
f7d9821a 1745err_detach:
ec0865a9 1746 if (!bond_uses_primary(bond))
303d1cbf
JV
1747 bond_hw_addr_flush(bond_dev, slave_dev);
1748
1ff412ad 1749 vlan_vids_del_by_dev(slave_dev, bond_dev);
059b47e8
NA
1750 if (rcu_access_pointer(bond->primary_slave) == new_slave)
1751 RCU_INIT_POINTER(bond->primary_slave, NULL);
4740d638 1752 if (rcu_access_pointer(bond->curr_active_slave) == new_slave) {
f80889a5 1753 block_netpoll_tx();
c8517035 1754 bond_change_active_slave(bond, NULL);
3c5913b5 1755 bond_select_active_slave(bond);
f80889a5 1756 unblock_netpoll_tx();
3c5913b5 1757 }
059b47e8
NA
1758 /* either primary_slave or curr_active_slave might've changed */
1759 synchronize_rcu();
fc7a72ac 1760 slave_disable_netpoll(new_slave);
f7d9821a 1761
1da177e4 1762err_close:
b6a5a7b9 1763 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4
LT
1764 dev_close(slave_dev);
1765
1766err_restore_mac:
03d84a5f 1767 slave_dev->flags &= ~IFF_SLAVE;
00503b6f 1768 if (!bond->params.fail_over_mac ||
01844098 1769 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
3915c1e8
JV
1770 /* XXX TODO - fom follow mode needs to change master's
1771 * MAC if this slave's MAC is in use by the bond, or at
1772 * least print a warning.
1773 */
ada0f863 1774 ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr);
2ab82852
MS
1775 addr.sa_family = slave_dev->type;
1776 dev_set_mac_address(slave_dev, &addr);
1777 }
1da177e4 1778
b15ba0fb
JP
1779err_restore_mtu:
1780 dev_set_mtu(slave_dev, new_slave->original_mtu);
1781
1da177e4 1782err_free:
3fdddd85 1783 bond_free_slave(new_slave);
1da177e4
LT
1784
1785err_undo_flags:
b8fad459 1786 /* Enslave of first slave has failed and we need to fix master's mac */
7d5cd2ce
NA
1787 if (!bond_has_slaves(bond)) {
1788 if (ether_addr_equal_64bits(bond_dev->dev_addr,
1789 slave_dev->dev_addr))
1790 eth_hw_addr_random(bond_dev);
1791 if (bond_dev->type != ARPHRD_ETHER) {
40baec22 1792 dev_close(bond_dev);
7d5cd2ce
NA
1793 ether_setup(bond_dev);
1794 bond_dev->flags |= IFF_MASTER;
1795 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1796 }
1797 }
3d632c3f 1798
1da177e4
LT
1799 return res;
1800}
1801
547942ca 1802/* Try to release the slave device <slave> from the bond device <master>
1da177e4 1803 * It is legal to access curr_active_slave without a lock because all the function
8c0bc550 1804 * is RTNL-locked. If "all" is true it means that the function is being called
0896341a 1805 * while destroying a bond interface and all slaves are being released.
1da177e4
LT
1806 *
1807 * The rules for slave state should be:
1808 * for Active/Backup:
1809 * Active stays on all backups go down
1810 * for Bonded connections:
1811 * The first up interface should be left on and all others downed.
1812 */
0896341a 1813static int __bond_release_one(struct net_device *bond_dev,
1814 struct net_device *slave_dev,
1815 bool all)
1da177e4 1816{
454d7c9b 1817 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT
1818 struct slave *slave, *oldcurrent;
1819 struct sockaddr addr;
5a0068de 1820 int old_flags = bond_dev->flags;
c8f44aff 1821 netdev_features_t old_features = bond_dev->features;
1da177e4
LT
1822
1823 /* slave is not a slave or master is not master of this slave */
1824 if (!(slave_dev->flags & IFF_SLAVE) ||
471cb5a3 1825 !netdev_has_upper_dev(slave_dev, bond_dev)) {
a22a9e41 1826 netdev_dbg(bond_dev, "cannot release %s\n",
76444f50 1827 slave_dev->name);
1da177e4
LT
1828 return -EINVAL;
1829 }
1830
e843fa50 1831 block_netpoll_tx();
1da177e4
LT
1832
1833 slave = bond_get_slave_by_dev(bond, slave_dev);
1834 if (!slave) {
1835 /* not a slave of this bond */
76444f50
VF
1836 netdev_info(bond_dev, "%s not enslaved\n",
1837 slave_dev->name);
e843fa50 1838 unblock_netpoll_tx();
1da177e4
LT
1839 return -EINVAL;
1840 }
1841
57beaca8
JP
1842 bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW);
1843
07699f9a 1844 bond_sysfs_slave_del(slave);
1845
5f0c5f73
AG
1846 /* recompute stats just before removing the slave */
1847 bond_get_stats(bond->dev, &bond->bond_stats);
1848
41f0b049 1849 bond_upper_dev_unlink(bond, slave);
35d48903
JP
1850 /* unregister rx_handler early so bond_handle_frame wouldn't be called
1851 * for this slave anymore.
1852 */
1853 netdev_rx_handler_unregister(slave_dev);
35d48903 1854
e470259f 1855 if (BOND_MODE(bond) == BOND_MODE_8023AD)
1da177e4 1856 bond_3ad_unbind_slave(slave);
1da177e4 1857
ee637714
MB
1858 if (bond_mode_uses_xmit_hash(bond))
1859 bond_update_slave_arr(bond, slave);
1860
76444f50
VF
1861 netdev_info(bond_dev, "Releasing %s interface %s\n",
1862 bond_is_active_slave(slave) ? "active" : "backup",
1863 slave_dev->name);
1da177e4 1864
4740d638 1865 oldcurrent = rcu_access_pointer(bond->curr_active_slave);
1da177e4 1866
85741718 1867 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
1da177e4 1868
00503b6f 1869 if (!all && (!bond->params.fail_over_mac ||
01844098 1870 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) {
844223ab 1871 if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) &&
0965a1f3 1872 bond_has_slaves(bond))
76444f50
VF
1873 netdev_warn(bond_dev, "the permanent HWaddr of %s - %pM - is still in use by %s - set the HWaddr of %s to a different address to avoid conflicts\n",
1874 slave_dev->name, slave->perm_hwaddr,
1875 bond_dev->name, slave_dev->name);
dec1e90e 1876 }
1877
059b47e8
NA
1878 if (rtnl_dereference(bond->primary_slave) == slave)
1879 RCU_INIT_POINTER(bond->primary_slave, NULL);
1da177e4 1880
1c72cfdc 1881 if (oldcurrent == slave)
1da177e4 1882 bond_change_active_slave(bond, NULL);
1da177e4 1883
58402054 1884 if (bond_is_lb(bond)) {
1da177e4
LT
1885 /* Must be called only after the slave has been
1886 * detached from the list and the curr_active_slave
1887 * has been cleared (if our_slave == old_current),
1888 * but before a new active slave is selected.
1889 */
1890 bond_alb_deinit_slave(bond, slave);
1891 }
1892
0896341a 1893 if (all) {
36708b89 1894 RCU_INIT_POINTER(bond->curr_active_slave, NULL);
0896341a 1895 } else if (oldcurrent == slave) {
547942ca 1896 /* Note that we hold RTNL over this sequence, so there
059fe7a5
JV
1897 * is no concern that another slave add/remove event
1898 * will interfere.
1899 */
1da177e4 1900 bond_select_active_slave(bond);
059fe7a5
JV
1901 }
1902
0965a1f3 1903 if (!bond_has_slaves(bond)) {
ff59c456 1904 bond_set_carrier(bond);
409cc1f8 1905 eth_hw_addr_random(bond_dev);
1da177e4
LT
1906 }
1907
e843fa50 1908 unblock_netpoll_tx();
278b2083 1909 synchronize_rcu();
ee6154e1 1910 bond->slave_cnt--;
1da177e4 1911
0965a1f3 1912 if (!bond_has_slaves(bond)) {
2af73d4b 1913 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev);
80028ea1
VF
1914 call_netdevice_notifiers(NETDEV_RELEASE, bond->dev);
1915 }
2af73d4b 1916
b2a103e6
MM
1917 bond_compute_features(bond);
1918 if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1919 (old_features & NETIF_F_VLAN_CHALLENGED))
76444f50
VF
1920 netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n",
1921 slave_dev->name, bond_dev->name);
b2a103e6 1922
1ff412ad 1923 vlan_vids_del_by_dev(slave_dev, bond_dev);
1da177e4 1924
547942ca 1925 /* If the mode uses primary, then this case was handled above by
303d1cbf 1926 * bond_change_active_slave(..., NULL)
1da177e4 1927 */
ec0865a9 1928 if (!bond_uses_primary(bond)) {
5a0068de
NH
1929 /* unset promiscuity level from slave
1930 * NOTE: The NETDEV_CHANGEADDR call above may change the value
1931 * of the IFF_PROMISC flag in the bond_dev, but we need the
1932 * value of that flag before that change, as that was the value
1933 * when this slave was attached, so we cache at the start of the
1934 * function and use it here. Same goes for ALLMULTI below
1935 */
1936 if (old_flags & IFF_PROMISC)
1da177e4 1937 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT
1938
1939 /* unset allmulti level from slave */
5a0068de 1940 if (old_flags & IFF_ALLMULTI)
1da177e4 1941 dev_set_allmulti(slave_dev, -1);
1da177e4 1942
303d1cbf 1943 bond_hw_addr_flush(bond_dev, slave_dev);
1da177e4
LT
1944 }
1945
8a8efa22 1946 slave_disable_netpoll(slave);
f6dc31a8 1947
1da177e4
LT
1948 /* close slave before restoring its mac address */
1949 dev_close(slave_dev);
1950
00503b6f 1951 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE ||
01844098 1952 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
2ab82852 1953 /* restore original ("permanent") mac address */
ada0f863 1954 ether_addr_copy(addr.sa_data, slave->perm_hwaddr);
2ab82852
MS
1955 addr.sa_family = slave_dev->type;
1956 dev_set_mac_address(slave_dev, &addr);
1957 }
1da177e4 1958
b15ba0fb
JP
1959 dev_set_mtu(slave_dev, slave->original_mtu);
1960
2d7011ca 1961 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4 1962
3fdddd85 1963 bond_free_slave(slave);
1da177e4 1964
547942ca 1965 return 0;
1da177e4
LT
1966}
1967
0896341a 1968/* A wrapper used because of ndo_del_link */
1969int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1970{
1971 return __bond_release_one(bond_dev, slave_dev, false);
1972}
1973
547942ca
NA
1974/* First release a slave and then destroy the bond if no more slaves are left.
1975 * Must be under rtnl_lock when this function is called.
1976 */
26d8ee75 1977static int bond_release_and_destroy(struct net_device *bond_dev,
1978 struct net_device *slave_dev)
d90a162a 1979{
454d7c9b 1980 struct bonding *bond = netdev_priv(bond_dev);
d90a162a
MS
1981 int ret;
1982
1983 ret = bond_release(bond_dev, slave_dev);
0965a1f3 1984 if (ret == 0 && !bond_has_slaves(bond)) {
8a8efa22 1985 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
76444f50
VF
1986 netdev_info(bond_dev, "Destroying bond %s\n",
1987 bond_dev->name);
06f6d109 1988 bond_remove_proc_entry(bond);
9e71626c 1989 unregister_netdevice(bond_dev);
d90a162a
MS
1990 }
1991 return ret;
1992}
1993
3d67576d 1994static void bond_info_query(struct net_device *bond_dev, struct ifbond *info)
1da177e4 1995{
454d7c9b 1996 struct bonding *bond = netdev_priv(bond_dev);
69a2338e 1997 bond_fill_ifbond(bond, info);
1da177e4
LT
1998}
1999
2000static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
2001{
454d7c9b 2002 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 2003 struct list_head *iter;
dec1e90e 2004 int i = 0, res = -ENODEV;
1da177e4 2005 struct slave *slave;
1da177e4 2006
9caff1e7 2007 bond_for_each_slave(bond, slave, iter) {
dec1e90e 2008 if (i++ == (int)info->slave_id) {
689c96cc 2009 res = 0;
69a2338e 2010 bond_fill_ifslave(slave, info);
1da177e4
LT
2011 break;
2012 }
2013 }
1da177e4 2014
689c96cc 2015 return res;
1da177e4
LT
2016}
2017
2018/*-------------------------------- Monitoring -------------------------------*/
2019
4740d638 2020/* called with rcu_read_lock() */
f0c76d61
JV
2021static int bond_miimon_inspect(struct bonding *bond)
2022{
dec1e90e 2023 int link_state, commit = 0;
9caff1e7 2024 struct list_head *iter;
f0c76d61 2025 struct slave *slave;
41f89100
JP
2026 bool ignore_updelay;
2027
4740d638 2028 ignore_updelay = !rcu_dereference(bond->curr_active_slave);
1da177e4 2029
4cb4f97b 2030 bond_for_each_slave_rcu(bond, slave, iter) {
f0c76d61 2031 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 2032
f0c76d61 2033 link_state = bond_check_dev_link(bond, slave->dev, 0);
1da177e4
LT
2034
2035 switch (slave->link) {
f0c76d61
JV
2036 case BOND_LINK_UP:
2037 if (link_state)
2038 continue;
1da177e4 2039
de77ecd4 2040 bond_propose_link_state(slave, BOND_LINK_FAIL);
f0c76d61
JV
2041 slave->delay = bond->params.downdelay;
2042 if (slave->delay) {
76444f50
VF
2043 netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n",
2044 (BOND_MODE(bond) ==
2045 BOND_MODE_ACTIVEBACKUP) ?
2046 (bond_is_active_slave(slave) ?
2047 "active " : "backup ") : "",
2048 slave->dev->name,
2049 bond->params.downdelay * bond->params.miimon);
1da177e4 2050 }
f0c76d61
JV
2051 /*FALLTHRU*/
2052 case BOND_LINK_FAIL:
2053 if (link_state) {
547942ca 2054 /* recovered before downdelay expired */
de77ecd4 2055 bond_propose_link_state(slave, BOND_LINK_UP);
8e603460 2056 slave->last_link_up = jiffies;
76444f50
VF
2057 netdev_info(bond->dev, "link status up again after %d ms for interface %s\n",
2058 (bond->params.downdelay - slave->delay) *
2059 bond->params.miimon,
2060 slave->dev->name);
f0c76d61 2061 continue;
1da177e4 2062 }
f0c76d61
JV
2063
2064 if (slave->delay <= 0) {
2065 slave->new_link = BOND_LINK_DOWN;
2066 commit++;
2067 continue;
1da177e4 2068 }
1da177e4 2069
f0c76d61
JV
2070 slave->delay--;
2071 break;
2072
2073 case BOND_LINK_DOWN:
2074 if (!link_state)
2075 continue;
2076
de77ecd4 2077 bond_propose_link_state(slave, BOND_LINK_BACK);
f0c76d61
JV
2078 slave->delay = bond->params.updelay;
2079
2080 if (slave->delay) {
76444f50
VF
2081 netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n",
2082 slave->dev->name,
2083 ignore_updelay ? 0 :
2084 bond->params.updelay *
2085 bond->params.miimon);
f0c76d61
JV
2086 }
2087 /*FALLTHRU*/
2088 case BOND_LINK_BACK:
2089 if (!link_state) {
de77ecd4 2090 bond_propose_link_state(slave, BOND_LINK_DOWN);
76444f50
VF
2091 netdev_info(bond->dev, "link status down again after %d ms for interface %s\n",
2092 (bond->params.updelay - slave->delay) *
2093 bond->params.miimon,
2094 slave->dev->name);
f0c76d61
JV
2095
2096 continue;
2097 }
2098
41f89100
JP
2099 if (ignore_updelay)
2100 slave->delay = 0;
2101
f0c76d61
JV
2102 if (slave->delay <= 0) {
2103 slave->new_link = BOND_LINK_UP;
2104 commit++;
41f89100 2105 ignore_updelay = false;
f0c76d61 2106 continue;
1da177e4 2107 }
f0c76d61
JV
2108
2109 slave->delay--;
1da177e4 2110 break;
f0c76d61
JV
2111 }
2112 }
1da177e4 2113
f0c76d61
JV
2114 return commit;
2115}
1da177e4 2116
f0c76d61
JV
2117static void bond_miimon_commit(struct bonding *bond)
2118{
9caff1e7 2119 struct list_head *iter;
059b47e8 2120 struct slave *slave, *primary;
f0c76d61 2121
9caff1e7 2122 bond_for_each_slave(bond, slave, iter) {
f0c76d61
JV
2123 switch (slave->new_link) {
2124 case BOND_LINK_NOCHANGE:
2125 continue;
1da177e4 2126
f0c76d61 2127 case BOND_LINK_UP:
266b495f 2128 bond_update_speed_duplex(slave);
5d397061
JP
2129 bond_set_slave_link_state(slave, BOND_LINK_UP,
2130 BOND_SLAVE_NOTIFY_NOW);
8e603460 2131 slave->last_link_up = jiffies;
f0c76d61 2132
059b47e8 2133 primary = rtnl_dereference(bond->primary_slave);
01844098 2134 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
f0c76d61 2135 /* prevent it from being the active one */
e30bc066 2136 bond_set_backup_slave(slave);
01844098 2137 } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
f0c76d61 2138 /* make it immediately active */
e30bc066 2139 bond_set_active_slave(slave);
059b47e8 2140 } else if (slave != primary) {
f0c76d61 2141 /* prevent it from being the active one */
e30bc066 2142 bond_set_backup_slave(slave);
1da177e4 2143 }
1da177e4 2144
76444f50
VF
2145 netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n",
2146 slave->dev->name,
2147 slave->speed == SPEED_UNKNOWN ? 0 : slave->speed,
2148 slave->duplex ? "full" : "half");
1da177e4 2149
f0c76d61 2150 /* notify ad that the link status has changed */
01844098 2151 if (BOND_MODE(bond) == BOND_MODE_8023AD)
f0c76d61 2152 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
059fe7a5 2153
58402054 2154 if (bond_is_lb(bond))
f0c76d61
JV
2155 bond_alb_handle_link_change(bond, slave,
2156 BOND_LINK_UP);
1da177e4 2157
ee637714
MB
2158 if (BOND_MODE(bond) == BOND_MODE_XOR)
2159 bond_update_slave_arr(bond, NULL);
2160
059b47e8 2161 if (!bond->curr_active_slave || slave == primary)
f0c76d61 2162 goto do_failover;
1da177e4 2163
f0c76d61 2164 continue;
059fe7a5 2165
f0c76d61 2166 case BOND_LINK_DOWN:
fba4acda
JV
2167 if (slave->link_failure_count < UINT_MAX)
2168 slave->link_failure_count++;
2169
5d397061
JP
2170 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2171 BOND_SLAVE_NOTIFY_NOW);
1da177e4 2172
01844098
VF
2173 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP ||
2174 BOND_MODE(bond) == BOND_MODE_8023AD)
5e5b0665 2175 bond_set_slave_inactive_flags(slave,
2176 BOND_SLAVE_NOTIFY_NOW);
f0c76d61 2177
76444f50
VF
2178 netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n",
2179 slave->dev->name);
f0c76d61 2180
01844098 2181 if (BOND_MODE(bond) == BOND_MODE_8023AD)
f0c76d61
JV
2182 bond_3ad_handle_link_change(slave,
2183 BOND_LINK_DOWN);
2184
ae63e808 2185 if (bond_is_lb(bond))
f0c76d61
JV
2186 bond_alb_handle_link_change(bond, slave,
2187 BOND_LINK_DOWN);
2188
ee637714
MB
2189 if (BOND_MODE(bond) == BOND_MODE_XOR)
2190 bond_update_slave_arr(bond, NULL);
2191
4740d638 2192 if (slave == rcu_access_pointer(bond->curr_active_slave))
f0c76d61
JV
2193 goto do_failover;
2194
2195 continue;
2196
2197 default:
76444f50
VF
2198 netdev_err(bond->dev, "invalid new link %d on slave %s\n",
2199 slave->new_link, slave->dev->name);
f0c76d61
JV
2200 slave->new_link = BOND_LINK_NOCHANGE;
2201
2202 continue;
2203 }
2204
2205do_failover:
e843fa50 2206 block_netpoll_tx();
f0c76d61 2207 bond_select_active_slave(bond);
e843fa50 2208 unblock_netpoll_tx();
f0c76d61
JV
2209 }
2210
2211 bond_set_carrier(bond);
1da177e4
LT
2212}
2213
547942ca 2214/* bond_mii_monitor
0b0eef66
JV
2215 *
2216 * Really a wrapper that splits the mii monitor into two phases: an
f0c76d61
JV
2217 * inspection, then (if inspection indicates something needs to be done)
2218 * an acquisition of appropriate locks followed by a commit phase to
2219 * implement whatever link state changes are indicated.
0b0eef66 2220 */
6da67d26 2221static void bond_mii_monitor(struct work_struct *work)
0b0eef66
JV
2222{
2223 struct bonding *bond = container_of(work, struct bonding,
2224 mii_work.work);
ad246c99 2225 bool should_notify_peers = false;
1f2cd845 2226 unsigned long delay;
de77ecd4
MB
2227 struct slave *slave;
2228 struct list_head *iter;
0b0eef66 2229
1f2cd845
DM
2230 delay = msecs_to_jiffies(bond->params.miimon);
2231
2232 if (!bond_has_slaves(bond))
f0c76d61 2233 goto re_arm;
b59f9f74 2234
4cb4f97b 2235 rcu_read_lock();
2236
ad246c99
BH
2237 should_notify_peers = bond_should_notify_peers(bond);
2238
1f2cd845 2239 if (bond_miimon_inspect(bond)) {
4cb4f97b 2240 rcu_read_unlock();
f0c76d61 2241
1f2cd845
DM
2242 /* Race avoidance with bond_close cancel of workqueue */
2243 if (!rtnl_trylock()) {
1f2cd845
DM
2244 delay = 1;
2245 should_notify_peers = false;
2246 goto re_arm;
2247 }
6b6c5261 2248
de77ecd4
MB
2249 bond_for_each_slave(bond, slave, iter) {
2250 bond_commit_link_state(slave, BOND_SLAVE_NOTIFY_LATER);
2251 }
1f2cd845
DM
2252 bond_miimon_commit(bond);
2253
1f2cd845 2254 rtnl_unlock(); /* might sleep, hold no other locks */
4cb4f97b 2255 } else
2256 rcu_read_unlock();
0b0eef66 2257
f0c76d61 2258re_arm:
e6d265e8 2259 if (bond->params.miimon)
1f2cd845
DM
2260 queue_delayed_work(bond->wq, &bond->mii_work, delay);
2261
1f2cd845
DM
2262 if (should_notify_peers) {
2263 if (!rtnl_trylock())
2264 return;
2265 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2266 rtnl_unlock();
2267 }
0b0eef66 2268}
c3ade5ca 2269
b3208b20
DA
2270static int bond_upper_dev_walk(struct net_device *upper, void *data)
2271{
2272 __be32 ip = *((__be32 *)data);
2273
2274 return ip == bond_confirm_addr(upper, 0, ip);
2275}
2276
50223ce4 2277static bool bond_has_this_ip(struct bonding *bond, __be32 ip)
f5b2b966 2278{
50223ce4 2279 bool ret = false;
f5b2b966 2280
eaddcd76 2281 if (ip == bond_confirm_addr(bond->dev, 0, ip))
50223ce4 2282 return true;
f5b2b966 2283
50223ce4 2284 rcu_read_lock();
b3208b20
DA
2285 if (netdev_walk_all_upper_dev_rcu(bond->dev, bond_upper_dev_walk, &ip))
2286 ret = true;
50223ce4 2287 rcu_read_unlock();
f5b2b966 2288
50223ce4 2289 return ret;
f5b2b966
JV
2290}
2291
547942ca 2292/* We go to the (large) trouble of VLAN tagging ARP frames because
c3ade5ca
JV
2293 * switches in VLAN mode (especially if ports are configured as
2294 * "native" to a VLAN) might not pass non-tagged frames.
2295 */
fbd929f2 2296static void bond_arp_send(struct net_device *slave_dev, int arp_op,
2297 __be32 dest_ip, __be32 src_ip,
44a40855 2298 struct bond_vlan_tag *tags)
c3ade5ca
JV
2299{
2300 struct sk_buff *skb;
3e403a77 2301 struct bond_vlan_tag *outer_tag = tags;
c3ade5ca 2302
76444f50
VF
2303 netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n",
2304 arp_op, slave_dev->name, &dest_ip, &src_ip);
3d632c3f 2305
c3ade5ca
JV
2306 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2307 NULL, slave_dev->dev_addr, NULL);
2308
2309 if (!skb) {
4873ac3c 2310 net_err_ratelimited("ARP packet allocation failed\n");
c3ade5ca
JV
2311 return;
2312 }
fbd929f2 2313
3e403a77
VF
2314 if (!tags || tags->vlan_proto == VLAN_N_VID)
2315 goto xmit;
2316
2317 tags++;
2318
44a40855 2319 /* Go through all the tags backwards and add them to the packet */
3e403a77
VF
2320 while (tags->vlan_proto != VLAN_N_VID) {
2321 if (!tags->vlan_id) {
2322 tags++;
44a40855 2323 continue;
3e403a77 2324 }
44a40855 2325
76444f50 2326 netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n",
3e403a77 2327 ntohs(outer_tag->vlan_proto), tags->vlan_id);
62749e2c
JP
2328 skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto,
2329 tags->vlan_id);
44a40855
VY
2330 if (!skb) {
2331 net_err_ratelimited("failed to insert inner VLAN tag\n");
2332 return;
2333 }
3e403a77
VF
2334
2335 tags++;
44a40855
VY
2336 }
2337 /* Set the outer tag */
3e403a77 2338 if (outer_tag->vlan_id) {
76444f50 2339 netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n",
3e403a77 2340 ntohs(outer_tag->vlan_proto), outer_tag->vlan_id);
b4bef1b5
JP
2341 __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto,
2342 outer_tag->vlan_id);
c3ade5ca 2343 }
3e403a77
VF
2344
2345xmit:
c3ade5ca
JV
2346 arp_xmit(skb);
2347}
2348
44a40855
VY
2349/* Validate the device path between the @start_dev and the @end_dev.
2350 * The path is valid if the @end_dev is reachable through device
2351 * stacking.
2352 * When the path is validated, collect any vlan information in the
2353 * path.
2354 */
3e403a77
VF
2355struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev,
2356 struct net_device *end_dev,
2357 int level)
44a40855 2358{
3e403a77 2359 struct bond_vlan_tag *tags;
44a40855
VY
2360 struct net_device *upper;
2361 struct list_head *iter;
44a40855 2362
3e403a77
VF
2363 if (start_dev == end_dev) {
2364 tags = kzalloc(sizeof(*tags) * (level + 1), GFP_ATOMIC);
2365 if (!tags)
2366 return ERR_PTR(-ENOMEM);
2367 tags[level].vlan_proto = VLAN_N_VID;
2368 return tags;
2369 }
44a40855
VY
2370
2371 netdev_for_each_upper_dev_rcu(start_dev, upper, iter) {
3e403a77
VF
2372 tags = bond_verify_device_path(upper, end_dev, level + 1);
2373 if (IS_ERR_OR_NULL(tags)) {
2374 if (IS_ERR(tags))
2375 return tags;
2376 continue;
44a40855 2377 }
3e403a77
VF
2378 if (is_vlan_dev(upper)) {
2379 tags[level].vlan_proto = vlan_dev_vlan_proto(upper);
2380 tags[level].vlan_id = vlan_dev_vlan_id(upper);
2381 }
2382
2383 return tags;
44a40855
VY
2384 }
2385
3e403a77 2386 return NULL;
44a40855 2387}
c3ade5ca 2388
1da177e4
LT
2389static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2390{
c3ade5ca 2391 struct rtable *rt;
3e403a77 2392 struct bond_vlan_tag *tags;
27bc11e6 2393 __be32 *targets = bond->params.arp_targets, addr;
fbd929f2 2394 int i;
1da177e4 2395
27bc11e6 2396 for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) {
76444f50 2397 netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]);
3e403a77 2398 tags = NULL;
c3ade5ca 2399
27bc11e6 2400 /* Find out through which dev should the packet go */
78fbfd8a
DM
2401 rt = ip_route_output(dev_net(bond->dev), targets[i], 0,
2402 RTO_ONLINK, 0);
b23dd4fe 2403 if (IS_ERR(rt)) {
28572760
VF
2404 /* there's no route to target - try to send arp
2405 * probe to generate any traffic (arp_validate=0)
2406 */
4873ac3c 2407 if (bond->params.arp_validate)
2408 net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n",
2409 bond->dev->name,
2410 &targets[i]);
44a40855
VY
2411 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2412 0, tags);
c3ade5ca
JV
2413 continue;
2414 }
2415
27bc11e6
VF
2416 /* bond device itself */
2417 if (rt->dst.dev == bond->dev)
2418 goto found;
2419
2420 rcu_read_lock();
3e403a77 2421 tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0);
27bc11e6 2422 rcu_read_unlock();
c3ade5ca 2423
3e403a77 2424 if (!IS_ERR_OR_NULL(tags))
44a40855
VY
2425 goto found;
2426
27bc11e6 2427 /* Not our device - skip */
76444f50
VF
2428 netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n",
2429 &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL");
3e32582f 2430
ed4b9f80 2431 ip_rt_put(rt);
27bc11e6
VF
2432 continue;
2433
2434found:
2435 addr = bond_confirm_addr(rt->dst.dev, targets[i], 0);
2436 ip_rt_put(rt);
2437 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
44a40855 2438 addr, tags);
a67eed57 2439 kfree(tags);
c3ade5ca
JV
2440 }
2441}
2442
d3bb52b0 2443static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
f5b2b966 2444{
8599b52e
VF
2445 int i;
2446
87a7b84b 2447 if (!sip || !bond_has_this_ip(bond, tip)) {
76444f50
VF
2448 netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n",
2449 &sip, &tip);
87a7b84b
VF
2450 return;
2451 }
f5b2b966 2452
8599b52e
VF
2453 i = bond_get_targets_ip(bond->params.arp_targets, sip);
2454 if (i == -1) {
76444f50
VF
2455 netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n",
2456 &sip);
87a7b84b 2457 return;
f5b2b966 2458 }
49f17de7 2459 slave->last_rx = jiffies;
8599b52e 2460 slave->target_last_arp_rx[i] = jiffies;
f5b2b966
JV
2461}
2462
5bb9e0b5 2463int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
2464 struct slave *slave)
f5b2b966 2465{
de063b70 2466 struct arphdr *arp = (struct arphdr *)skb->data;
21a75f09 2467 struct slave *curr_active_slave, *curr_arp_slave;
f5b2b966 2468 unsigned char *arp_ptr;
d3bb52b0 2469 __be32 sip, tip;
f2cb691a 2470 int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
f5b2b966 2471
f2cb691a 2472 if (!slave_do_arp_validate(bond, slave)) {
bedabf90 2473 if ((slave_do_arp_validate_only(bond) && is_arp) ||
2474 !slave_do_arp_validate_only(bond))
49f17de7 2475 slave->last_rx = jiffies;
13a8e0c8 2476 return RX_HANDLER_ANOTHER;
f2cb691a
VF
2477 } else if (!is_arp) {
2478 return RX_HANDLER_ANOTHER;
2479 }
2c146102 2480
de063b70 2481 alen = arp_hdr_len(bond->dev);
f5b2b966 2482
76444f50
VF
2483 netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n",
2484 skb->dev->name);
f5b2b966 2485
de063b70
ED
2486 if (alen > skb_headlen(skb)) {
2487 arp = kmalloc(alen, GFP_ATOMIC);
2488 if (!arp)
2489 goto out_unlock;
2490 if (skb_copy_bits(skb, 0, arp, alen) < 0)
2491 goto out_unlock;
2492 }
f5b2b966 2493
3aba891d 2494 if (arp->ar_hln != bond->dev->addr_len ||
f5b2b966
JV
2495 skb->pkt_type == PACKET_OTHERHOST ||
2496 skb->pkt_type == PACKET_LOOPBACK ||
2497 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2498 arp->ar_pro != htons(ETH_P_IP) ||
2499 arp->ar_pln != 4)
2500 goto out_unlock;
2501
2502 arp_ptr = (unsigned char *)(arp + 1);
3aba891d 2503 arp_ptr += bond->dev->addr_len;
f5b2b966 2504 memcpy(&sip, arp_ptr, 4);
3aba891d 2505 arp_ptr += 4 + bond->dev->addr_len;
f5b2b966
JV
2506 memcpy(&tip, arp_ptr, 4);
2507
76444f50
VF
2508 netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n",
2509 slave->dev->name, bond_slave_state(slave),
2510 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2511 &sip, &tip);
f5b2b966 2512
010d3c39 2513 curr_active_slave = rcu_dereference(bond->curr_active_slave);
21a75f09 2514 curr_arp_slave = rcu_dereference(bond->current_arp_slave);
010d3c39 2515
21a75f09 2516 /* We 'trust' the received ARP enough to validate it if:
aeea64ac 2517 *
21a75f09
JV
2518 * (a) the slave receiving the ARP is active (which includes the
2519 * current ARP slave, if any), or
2520 *
2521 * (b) the receiving slave isn't active, but there is a currently
2522 * active slave and it received valid arp reply(s) after it became
2523 * the currently active slave, or
2524 *
2525 * (c) there is an ARP slave that sent an ARP during the prior ARP
2526 * interval, and we receive an ARP reply on any slave. We accept
2527 * these because switch FDB update delays may deliver the ARP
2528 * reply to a slave other than the sender of the ARP request.
2529 *
2530 * Note: for (b), backup slaves are receiving the broadcast ARP
2531 * request, not a reply. This request passes from the sending
2532 * slave through the L2 switch(es) to the receiving slave. Since
2533 * this is checking the request, sip/tip are swapped for
2534 * validation.
2535 *
2536 * This is done to avoid endless looping when we can't reach the
aeea64ac 2537 * arp_ip_target and fool ourselves with our own arp requests.
f5b2b966 2538 */
e30bc066 2539 if (bond_is_active_slave(slave))
f5b2b966 2540 bond_validate_arp(bond, slave, sip, tip);
010d3c39
VF
2541 else if (curr_active_slave &&
2542 time_after(slave_last_rx(bond, curr_active_slave),
2543 curr_active_slave->last_link_up))
f5b2b966 2544 bond_validate_arp(bond, slave, tip, sip);
21a75f09
JV
2545 else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) &&
2546 bond_time_in_interval(bond,
2547 dev_trans_start(curr_arp_slave->dev), 1))
2548 bond_validate_arp(bond, slave, sip, tip);
f5b2b966
JV
2549
2550out_unlock:
de063b70
ED
2551 if (arp != (struct arphdr *)skb->data)
2552 kfree(arp);
13a8e0c8 2553 return RX_HANDLER_ANOTHER;
f5b2b966
JV
2554}
2555
e7f63f1d
VF
2556/* function to verify if we're in the arp_interval timeslice, returns true if
2557 * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval +
2558 * arp_interval/2) . the arp_interval/2 is needed for really fast networks.
2559 */
2560static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
2561 int mod)
2562{
2563 int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2564
2565 return time_in_range(jiffies,
2566 last_act - delta_in_ticks,
2567 last_act + mod * delta_in_ticks + delta_in_ticks/2);
2568}
2569
547942ca 2570/* This function is called regularly to monitor each slave's link
1da177e4
LT
2571 * ensuring that traffic is being sent and received when arp monitoring
2572 * is used in load-balancing mode. if the adapter has been dormant, then an
2573 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2574 * arp monitoring in active backup mode.
2575 */
d5e73f7b 2576static void bond_loadbalance_arp_mon(struct bonding *bond)
1da177e4 2577{
1da177e4 2578 struct slave *slave, *oldcurrent;
9caff1e7 2579 struct list_head *iter;
6fde8f03 2580 int do_failover = 0, slave_state_changed = 0;
1da177e4 2581
1f2cd845 2582 if (!bond_has_slaves(bond))
1da177e4 2583 goto re_arm;
1da177e4 2584
2e52f4fe 2585 rcu_read_lock();
2586
4740d638 2587 oldcurrent = rcu_dereference(bond->curr_active_slave);
1da177e4
LT
2588 /* see if any of the previous devices are up now (i.e. they have
2589 * xmt and rcv traffic). the curr_active_slave does not come into
8e603460
VF
2590 * the picture unless it is null. also, slave->last_link_up is not
2591 * needed here because we send an arp on each slave and give a slave
2592 * as long as it needs to get the tx/rx within the delta.
1da177e4
LT
2593 * TODO: what about up/down delay in arp mode? it wasn't here before
2594 * so it can wait
2595 */
2e52f4fe 2596 bond_for_each_slave_rcu(bond, slave, iter) {
cb32f2a0
JB
2597 unsigned long trans_start = dev_trans_start(slave->dev);
2598
1da177e4 2599 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2600 if (bond_time_in_interval(bond, trans_start, 1) &&
49f17de7 2601 bond_time_in_interval(bond, slave->last_rx, 1)) {
1da177e4
LT
2602
2603 slave->link = BOND_LINK_UP;
6fde8f03 2604 slave_state_changed = 1;
1da177e4
LT
2605
2606 /* primary_slave has no meaning in round-robin
2607 * mode. the window of a slave being up and
2608 * curr_active_slave being null after enslaving
2609 * is closed.
2610 */
2611 if (!oldcurrent) {
76444f50
VF
2612 netdev_info(bond->dev, "link status definitely up for interface %s\n",
2613 slave->dev->name);
1da177e4
LT
2614 do_failover = 1;
2615 } else {
76444f50
VF
2616 netdev_info(bond->dev, "interface %s is now up\n",
2617 slave->dev->name);
1da177e4
LT
2618 }
2619 }
2620 } else {
2621 /* slave->link == BOND_LINK_UP */
2622
2623 /* not all switches will respond to an arp request
2624 * when the source ip is 0, so don't take the link down
2625 * if we don't know our ip yet
2626 */
e7f63f1d 2627 if (!bond_time_in_interval(bond, trans_start, 2) ||
49f17de7 2628 !bond_time_in_interval(bond, slave->last_rx, 2)) {
1da177e4
LT
2629
2630 slave->link = BOND_LINK_DOWN;
6fde8f03 2631 slave_state_changed = 1;
1da177e4 2632
3d632c3f 2633 if (slave->link_failure_count < UINT_MAX)
1da177e4 2634 slave->link_failure_count++;
1da177e4 2635
76444f50
VF
2636 netdev_info(bond->dev, "interface %s is now down\n",
2637 slave->dev->name);
1da177e4 2638
3d632c3f 2639 if (slave == oldcurrent)
1da177e4 2640 do_failover = 1;
1da177e4
LT
2641 }
2642 }
2643
2644 /* note: if switch is in round-robin mode, all links
2645 * must tx arp to ensure all links rx an arp - otherwise
2646 * links may oscillate or not come up at all; if switch is
2647 * in something like xor mode, there is nothing we can
2648 * do - all replies will be rx'ed on same link causing slaves
2649 * to be unstable during low/no traffic periods
2650 */
b6adc610 2651 if (bond_slave_is_up(slave))
1da177e4 2652 bond_arp_send_all(bond, slave);
1da177e4
LT
2653 }
2654
2e52f4fe 2655 rcu_read_unlock();
2656
6fde8f03 2657 if (do_failover || slave_state_changed) {
2e52f4fe 2658 if (!rtnl_trylock())
2659 goto re_arm;
1da177e4 2660
6fde8f03
DT
2661 if (slave_state_changed) {
2662 bond_slave_state_change(bond);
ee637714
MB
2663 if (BOND_MODE(bond) == BOND_MODE_XOR)
2664 bond_update_slave_arr(bond, NULL);
b8e4500f
NA
2665 }
2666 if (do_failover) {
6fde8f03 2667 block_netpoll_tx();
6fde8f03 2668 bond_select_active_slave(bond);
6fde8f03
DT
2669 unblock_netpoll_tx();
2670 }
2e52f4fe 2671 rtnl_unlock();
1da177e4
LT
2672 }
2673
2674re_arm:
e6d265e8 2675 if (bond->params.arp_interval)
e7f63f1d
VF
2676 queue_delayed_work(bond->wq, &bond->arp_work,
2677 msecs_to_jiffies(bond->params.arp_interval));
1da177e4
LT
2678}
2679
547942ca 2680/* Called to inspect slaves for active-backup mode ARP monitor link state
b2220cad
JV
2681 * changes. Sets new_link in slaves to specify what action should take
2682 * place for the slave. Returns 0 if no changes are found, >0 if changes
2683 * to link states must be committed.
2684 *
8c0bc550 2685 * Called with rcu_read_lock held.
1da177e4 2686 */
e7f63f1d 2687static int bond_ab_arp_inspect(struct bonding *bond)
1da177e4 2688{
def4460c 2689 unsigned long trans_start, last_rx;
9caff1e7 2690 struct list_head *iter;
dec1e90e 2691 struct slave *slave;
dec1e90e 2692 int commit = 0;
da210f55 2693
eb9fa4b0 2694 bond_for_each_slave_rcu(bond, slave, iter) {
b2220cad 2695 slave->new_link = BOND_LINK_NOCHANGE;
def4460c 2696 last_rx = slave_last_rx(bond, slave);
1da177e4 2697
b2220cad 2698 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2699 if (bond_time_in_interval(bond, last_rx, 1)) {
b2220cad
JV
2700 slave->new_link = BOND_LINK_UP;
2701 commit++;
2702 }
b2220cad
JV
2703 continue;
2704 }
1da177e4 2705
547942ca 2706 /* Give slaves 2*delta after being enslaved or made
b2220cad
JV
2707 * active. This avoids bouncing, as the last receive
2708 * times need a full ARP monitor cycle to be updated.
2709 */
8e603460 2710 if (bond_time_in_interval(bond, slave->last_link_up, 2))
b2220cad
JV
2711 continue;
2712
547942ca 2713 /* Backup slave is down if:
b2220cad
JV
2714 * - No current_arp_slave AND
2715 * - more than 3*delta since last receive AND
2716 * - the bond has an IP address
2717 *
2718 * Note: a non-null current_arp_slave indicates
2719 * the curr_active_slave went down and we are
2720 * searching for a new one; under this condition
2721 * we only take the curr_active_slave down - this
2722 * gives each slave a chance to tx/rx traffic
2723 * before being taken out
2724 */
e30bc066 2725 if (!bond_is_active_slave(slave) &&
85741718 2726 !rcu_access_pointer(bond->current_arp_slave) &&
e7f63f1d 2727 !bond_time_in_interval(bond, last_rx, 3)) {
b2220cad
JV
2728 slave->new_link = BOND_LINK_DOWN;
2729 commit++;
2730 }
2731
547942ca 2732 /* Active slave is down if:
b2220cad
JV
2733 * - more than 2*delta since transmitting OR
2734 * - (more than 2*delta since receive AND
2735 * the bond has an IP address)
2736 */
cb32f2a0 2737 trans_start = dev_trans_start(slave->dev);
e30bc066 2738 if (bond_is_active_slave(slave) &&
e7f63f1d
VF
2739 (!bond_time_in_interval(bond, trans_start, 2) ||
2740 !bond_time_in_interval(bond, last_rx, 2))) {
b2220cad
JV
2741 slave->new_link = BOND_LINK_DOWN;
2742 commit++;
2743 }
1da177e4
LT
2744 }
2745
b2220cad
JV
2746 return commit;
2747}
1da177e4 2748
547942ca 2749/* Called to commit link state changes noted by inspection step of
b2220cad
JV
2750 * active-backup mode ARP monitor.
2751 *
eb9fa4b0 2752 * Called with RTNL hold.
b2220cad 2753 */
e7f63f1d 2754static void bond_ab_arp_commit(struct bonding *bond)
b2220cad 2755{
cb32f2a0 2756 unsigned long trans_start;
9caff1e7 2757 struct list_head *iter;
dec1e90e 2758 struct slave *slave;
1da177e4 2759
9caff1e7 2760 bond_for_each_slave(bond, slave, iter) {
b2220cad
JV
2761 switch (slave->new_link) {
2762 case BOND_LINK_NOCHANGE:
2763 continue;
ff59c456 2764
b2220cad 2765 case BOND_LINK_UP:
cb32f2a0 2766 trans_start = dev_trans_start(slave->dev);
4740d638
ED
2767 if (rtnl_dereference(bond->curr_active_slave) != slave ||
2768 (!rtnl_dereference(bond->curr_active_slave) &&
e7f63f1d 2769 bond_time_in_interval(bond, trans_start, 1))) {
85741718
ED
2770 struct slave *current_arp_slave;
2771
2772 current_arp_slave = rtnl_dereference(bond->current_arp_slave);
5d397061
JP
2773 bond_set_slave_link_state(slave, BOND_LINK_UP,
2774 BOND_SLAVE_NOTIFY_NOW);
85741718 2775 if (current_arp_slave) {
5a430974 2776 bond_set_slave_inactive_flags(
85741718 2777 current_arp_slave,
5e5b0665 2778 BOND_SLAVE_NOTIFY_NOW);
85741718 2779 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
5a430974 2780 }
b2220cad 2781
76444f50
VF
2782 netdev_info(bond->dev, "link status definitely up for interface %s\n",
2783 slave->dev->name);
b2220cad 2784
4740d638 2785 if (!rtnl_dereference(bond->curr_active_slave) ||
059b47e8 2786 slave == rtnl_dereference(bond->primary_slave))
b9f60253 2787 goto do_failover;
1da177e4 2788
b9f60253 2789 }
1da177e4 2790
b9f60253 2791 continue;
1da177e4 2792
b2220cad
JV
2793 case BOND_LINK_DOWN:
2794 if (slave->link_failure_count < UINT_MAX)
2795 slave->link_failure_count++;
2796
5d397061
JP
2797 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2798 BOND_SLAVE_NOTIFY_NOW);
5e5b0665 2799 bond_set_slave_inactive_flags(slave,
2800 BOND_SLAVE_NOTIFY_NOW);
b2220cad 2801
76444f50
VF
2802 netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n",
2803 slave->dev->name);
b2220cad 2804
4740d638 2805 if (slave == rtnl_dereference(bond->curr_active_slave)) {
85741718 2806 RCU_INIT_POINTER(bond->current_arp_slave, NULL);
b9f60253 2807 goto do_failover;
1da177e4 2808 }
b9f60253
JP
2809
2810 continue;
b2220cad
JV
2811
2812 default:
76444f50
VF
2813 netdev_err(bond->dev, "impossible: new_link %d on slave %s\n",
2814 slave->new_link, slave->dev->name);
b9f60253 2815 continue;
1da177e4 2816 }
1da177e4 2817
b9f60253 2818do_failover:
e843fa50 2819 block_netpoll_tx();
b9f60253 2820 bond_select_active_slave(bond);
e843fa50 2821 unblock_netpoll_tx();
b2220cad 2822 }
1da177e4 2823
b2220cad
JV
2824 bond_set_carrier(bond);
2825}
1da177e4 2826
547942ca 2827/* Send ARP probes for active-backup mode ARP monitor.
b0929915 2828 *
8c0bc550 2829 * Called with rcu_read_lock held.
b2220cad 2830 */
f2ebd477 2831static bool bond_ab_arp_probe(struct bonding *bond)
b2220cad 2832{
eb9fa4b0 2833 struct slave *slave, *before = NULL, *new_slave = NULL,
b0929915 2834 *curr_arp_slave = rcu_dereference(bond->current_arp_slave),
2835 *curr_active_slave = rcu_dereference(bond->curr_active_slave);
4087df87
VF
2836 struct list_head *iter;
2837 bool found = false;
b0929915 2838 bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER;
f2ebd477 2839
98b90f26 2840 if (curr_arp_slave && curr_active_slave)
76444f50
VF
2841 netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n",
2842 curr_arp_slave->dev->name,
2843 curr_active_slave->dev->name);
1da177e4 2844
98b90f26
VF
2845 if (curr_active_slave) {
2846 bond_arp_send_all(bond, curr_active_slave);
b0929915 2847 return should_notify_rtnl;
b2220cad 2848 }
1da177e4 2849
b2220cad
JV
2850 /* if we don't have a curr_active_slave, search for the next available
2851 * backup slave from the current_arp_slave and make it the candidate
2852 * for becoming the curr_active_slave
2853 */
1da177e4 2854
eb9fa4b0 2855 if (!curr_arp_slave) {
b0929915 2856 curr_arp_slave = bond_first_slave_rcu(bond);
2857 if (!curr_arp_slave)
2858 return should_notify_rtnl;
b2220cad 2859 }
1da177e4 2860
b0929915 2861 bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER);
059fe7a5 2862
b0929915 2863 bond_for_each_slave_rcu(bond, slave, iter) {
b6adc610 2864 if (!found && !before && bond_slave_is_up(slave))
4087df87 2865 before = slave;
1da177e4 2866
b6adc610 2867 if (found && !new_slave && bond_slave_is_up(slave))
4087df87 2868 new_slave = slave;
b2220cad
JV
2869 /* if the link state is up at this point, we
2870 * mark it down - this can happen if we have
2871 * simultaneous link failures and
2872 * reselect_active_interface doesn't make this
2873 * one the current slave so it is still marked
2874 * up when it is actually down
1da177e4 2875 */
b6adc610 2876 if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
5d397061
JP
2877 bond_set_slave_link_state(slave, BOND_LINK_DOWN,
2878 BOND_SLAVE_NOTIFY_LATER);
b2220cad
JV
2879 if (slave->link_failure_count < UINT_MAX)
2880 slave->link_failure_count++;
1da177e4 2881
5e5b0665 2882 bond_set_slave_inactive_flags(slave,
b0929915 2883 BOND_SLAVE_NOTIFY_LATER);
b2220cad 2884
76444f50
VF
2885 netdev_info(bond->dev, "backup interface %s is now down\n",
2886 slave->dev->name);
1da177e4 2887 }
eb9fa4b0 2888 if (slave == curr_arp_slave)
4087df87 2889 found = true;
b2220cad 2890 }
4087df87
VF
2891
2892 if (!new_slave && before)
2893 new_slave = before;
2894
b0929915 2895 if (!new_slave)
2896 goto check_state;
4087df87 2897
5d397061
JP
2898 bond_set_slave_link_state(new_slave, BOND_LINK_BACK,
2899 BOND_SLAVE_NOTIFY_LATER);
b0929915 2900 bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER);
4087df87 2901 bond_arp_send_all(bond, new_slave);
8e603460 2902 new_slave->last_link_up = jiffies;
eb9fa4b0 2903 rcu_assign_pointer(bond->current_arp_slave, new_slave);
f2ebd477 2904
b0929915 2905check_state:
2906 bond_for_each_slave_rcu(bond, slave, iter) {
5d397061 2907 if (slave->should_notify || slave->should_notify_link) {
b0929915 2908 should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW;
2909 break;
2910 }
2911 }
2912 return should_notify_rtnl;
b2220cad 2913}
1da177e4 2914
d5e73f7b 2915static void bond_activebackup_arp_mon(struct bonding *bond)
b2220cad 2916{
b0929915 2917 bool should_notify_peers = false;
2918 bool should_notify_rtnl = false;
1f2cd845 2919 int delta_in_ticks;
1da177e4 2920
1f2cd845
DM
2921 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2922
2923 if (!bond_has_slaves(bond))
b2220cad
JV
2924 goto re_arm;
2925
eb9fa4b0 2926 rcu_read_lock();
b0929915 2927
ad246c99
BH
2928 should_notify_peers = bond_should_notify_peers(bond);
2929
b0929915 2930 if (bond_ab_arp_inspect(bond)) {
2931 rcu_read_unlock();
2932
1f2cd845
DM
2933 /* Race avoidance with bond_close flush of workqueue */
2934 if (!rtnl_trylock()) {
1f2cd845
DM
2935 delta_in_ticks = 1;
2936 should_notify_peers = false;
2937 goto re_arm;
2938 }
b2220cad 2939
1f2cd845 2940 bond_ab_arp_commit(bond);
b0929915 2941
1f2cd845 2942 rtnl_unlock();
b0929915 2943 rcu_read_lock();
1f2cd845
DM
2944 }
2945
b0929915 2946 should_notify_rtnl = bond_ab_arp_probe(bond);
2947 rcu_read_unlock();
ad246c99 2948
80b9d236 2949re_arm:
2950 if (bond->params.arp_interval)
1f2cd845
DM
2951 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
2952
b0929915 2953 if (should_notify_peers || should_notify_rtnl) {
1f2cd845
DM
2954 if (!rtnl_trylock())
2955 return;
b0929915 2956
2957 if (should_notify_peers)
2958 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
2959 bond->dev);
5d397061 2960 if (should_notify_rtnl) {
b0929915 2961 bond_slave_state_notify(bond);
5d397061
JP
2962 bond_slave_link_notify(bond);
2963 }
b0929915 2964
1f2cd845
DM
2965 rtnl_unlock();
2966 }
1da177e4
LT
2967}
2968
d5e73f7b
MB
2969static void bond_arp_monitor(struct work_struct *work)
2970{
2971 struct bonding *bond = container_of(work, struct bonding,
2972 arp_work.work);
2973
2974 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
2975 bond_activebackup_arp_mon(bond);
2976 else
2977 bond_loadbalance_arp_mon(bond);
2978}
2979
1da177e4
LT
2980/*-------------------------- netdev event handling --------------------------*/
2981
547942ca 2982/* Change device name */
1da177e4
LT
2983static int bond_event_changename(struct bonding *bond)
2984{
1da177e4
LT
2985 bond_remove_proc_entry(bond);
2986 bond_create_proc_entry(bond);
7e083840 2987
f073c7ca
TI
2988 bond_debug_reregister(bond);
2989
1da177e4
LT
2990 return NOTIFY_DONE;
2991}
2992
3d632c3f
SH
2993static int bond_master_netdev_event(unsigned long event,
2994 struct net_device *bond_dev)
1da177e4 2995{
454d7c9b 2996 struct bonding *event_bond = netdev_priv(bond_dev);
1da177e4
LT
2997
2998 switch (event) {
2999 case NETDEV_CHANGENAME:
3000 return bond_event_changename(event_bond);
a64d49c3
EB
3001 case NETDEV_UNREGISTER:
3002 bond_remove_proc_entry(event_bond);
3003 break;
3004 case NETDEV_REGISTER:
3005 bond_create_proc_entry(event_bond);
3006 break;
6c8d23f7 3007 case NETDEV_NOTIFY_PEERS:
3008 if (event_bond->send_peer_notif)
3009 event_bond->send_peer_notif--;
3010 break;
1da177e4
LT
3011 default:
3012 break;
3013 }
3014
3015 return NOTIFY_DONE;
3016}
3017
3d632c3f
SH
3018static int bond_slave_netdev_event(unsigned long event,
3019 struct net_device *slave_dev)
1da177e4 3020{
059b47e8 3021 struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary;
6101391d 3022 struct bonding *bond;
3023 struct net_device *bond_dev;
1da177e4 3024
6101391d 3025 /* A netdev event can be generated while enslaving a device
3026 * before netdev_rx_handler_register is called in which case
3027 * slave will be NULL
3028 */
3029 if (!slave)
3030 return NOTIFY_DONE;
3031 bond_dev = slave->bond->dev;
3032 bond = slave->bond;
059b47e8 3033 primary = rtnl_dereference(bond->primary_slave);
6101391d 3034
1da177e4
LT
3035 switch (event) {
3036 case NETDEV_UNREGISTER:
8d2ada77 3037 if (bond_dev->type != ARPHRD_ETHER)
471cb5a3
JP
3038 bond_release_and_destroy(bond_dev, slave_dev);
3039 else
3040 bond_release(bond_dev, slave_dev);
1da177e4 3041 break;
98f41f69 3042 case NETDEV_UP:
1da177e4 3043 case NETDEV_CHANGE:
471cb5a3 3044 bond_update_speed_duplex(slave);
52bc6716
MB
3045 if (BOND_MODE(bond) == BOND_MODE_8023AD)
3046 bond_3ad_adapter_speed_duplex_changed(slave);
950ddcb1
MB
3047 /* Fallthrough */
3048 case NETDEV_DOWN:
ee637714
MB
3049 /* Refresh slave-array if applicable!
3050 * If the setup does not use miimon or arpmon (mode-specific!),
3051 * then these events will not cause the slave-array to be
3052 * refreshed. This will cause xmit to use a slave that is not
3053 * usable. Avoid such situation by refeshing the array at these
3054 * events. If these (miimon/arpmon) parameters are configured
3055 * then array gets refreshed twice and that should be fine!
3056 */
3057 if (bond_mode_uses_xmit_hash(bond))
3058 bond_update_slave_arr(bond, NULL);
1da177e4 3059 break;
1da177e4 3060 case NETDEV_CHANGEMTU:
547942ca 3061 /* TODO: Should slaves be allowed to
1da177e4
LT
3062 * independently alter their MTU? For
3063 * an active-backup bond, slaves need
3064 * not be the same type of device, so
3065 * MTUs may vary. For other modes,
3066 * slaves arguably should have the
3067 * same MTUs. To do this, we'd need to
3068 * take over the slave's change_mtu
3069 * function for the duration of their
3070 * servitude.
3071 */
3072 break;
3073 case NETDEV_CHANGENAME:
3ec775b9 3074 /* we don't care if we don't have primary set */
ec0865a9 3075 if (!bond_uses_primary(bond) ||
3ec775b9
VF
3076 !bond->params.primary[0])
3077 break;
3078
059b47e8 3079 if (slave == primary) {
3ec775b9 3080 /* slave's name changed - he's no longer primary */
059b47e8 3081 RCU_INIT_POINTER(bond->primary_slave, NULL);
3ec775b9
VF
3082 } else if (!strcmp(slave_dev->name, bond->params.primary)) {
3083 /* we have a new primary slave */
059b47e8 3084 rcu_assign_pointer(bond->primary_slave, slave);
3ec775b9
VF
3085 } else { /* we didn't change primary - exit */
3086 break;
3087 }
3088
76444f50 3089 netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n",
059b47e8 3090 primary ? slave_dev->name : "none");
f80889a5 3091
3092 block_netpoll_tx();
3ec775b9 3093 bond_select_active_slave(bond);
f80889a5 3094 unblock_netpoll_tx();
1da177e4 3095 break;
8531c5ff
AK
3096 case NETDEV_FEAT_CHANGE:
3097 bond_compute_features(bond);
3098 break;
4aa5dee4
JP
3099 case NETDEV_RESEND_IGMP:
3100 /* Propagate to master device */
3101 call_netdevice_notifiers(event, slave->bond->dev);
3102 break;
1da177e4
LT
3103 default:
3104 break;
3105 }
3106
3107 return NOTIFY_DONE;
3108}
3109
547942ca 3110/* bond_netdev_event: handle netdev notifier chain events.
1da177e4
LT
3111 *
3112 * This function receives events for the netdev chain. The caller (an
e041c683 3113 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
1da177e4
LT
3114 * locks for us to safely manipulate the slave devices (RTNL lock,
3115 * dev_probe_lock).
3116 */
3d632c3f
SH
3117static int bond_netdev_event(struct notifier_block *this,
3118 unsigned long event, void *ptr)
1da177e4 3119{
351638e7 3120 struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
1da177e4 3121
76444f50 3122 netdev_dbg(event_dev, "event: %lx\n", event);
1da177e4 3123
0b680e75
JV
3124 if (!(event_dev->priv_flags & IFF_BONDING))
3125 return NOTIFY_DONE;
3126
1da177e4 3127 if (event_dev->flags & IFF_MASTER) {
76444f50 3128 netdev_dbg(event_dev, "IFF_MASTER\n");
1da177e4
LT
3129 return bond_master_netdev_event(event, event_dev);
3130 }
3131
3132 if (event_dev->flags & IFF_SLAVE) {
76444f50 3133 netdev_dbg(event_dev, "IFF_SLAVE\n");
1da177e4
LT
3134 return bond_slave_netdev_event(event, event_dev);
3135 }
3136
3137 return NOTIFY_DONE;
3138}
3139
3140static struct notifier_block bond_netdev_notifier = {
3141 .notifier_call = bond_netdev_event,
3142};
3143
169a3e66
JV
3144/*---------------------------- Hashing Policies -----------------------------*/
3145
32819dc1
NA
3146/* L2 hash helper */
3147static inline u32 bond_eth_hash(struct sk_buff *skb)
6b923cb7 3148{
ce04d635 3149 struct ethhdr *ep, hdr_tmp;
6b923cb7 3150
ce04d635
JX
3151 ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp);
3152 if (ep)
3153 return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto;
6b923cb7
JE
3154 return 0;
3155}
3156
32819dc1
NA
3157/* Extract the appropriate headers based on bond's xmit policy */
3158static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
3159 struct flow_keys *fk)
6f6652be 3160{
32819dc1 3161 const struct ipv6hdr *iph6;
4394542c 3162 const struct iphdr *iph;
32819dc1 3163 int noff, proto = -1;
6b923cb7 3164
32819dc1 3165 if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
cd79a238 3166 return skb_flow_dissect_flow_keys(skb, fk, 0);
32819dc1 3167
06635a35 3168 fk->ports.ports = 0;
32819dc1
NA
3169 noff = skb_network_offset(skb);
3170 if (skb->protocol == htons(ETH_P_IP)) {
054bb880 3171 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
32819dc1 3172 return false;
6b923cb7 3173 iph = ip_hdr(skb);
c3f83241 3174 iph_to_flow_copy_v4addrs(fk, iph);
32819dc1
NA
3175 noff += iph->ihl << 2;
3176 if (!ip_is_fragment(iph))
3177 proto = iph->protocol;
3178 } else if (skb->protocol == htons(ETH_P_IPV6)) {
054bb880 3179 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
32819dc1
NA
3180 return false;
3181 iph6 = ipv6_hdr(skb);
c3f83241 3182 iph_to_flow_copy_v6addrs(fk, iph6);
32819dc1
NA
3183 noff += sizeof(*iph6);
3184 proto = iph6->nexthdr;
3185 } else {
3186 return false;
6f6652be 3187 }
32819dc1 3188 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
06635a35 3189 fk->ports.ports = skb_flow_get_ports(skb, noff, proto);
6f6652be 3190
32819dc1 3191 return true;
6f6652be
JV
3192}
3193
32819dc1
NA
3194/**
3195 * bond_xmit_hash - generate a hash value based on the xmit policy
3196 * @bond: bonding device
3197 * @skb: buffer to use for headers
32819dc1
NA
3198 *
3199 * This function will extract the necessary headers from the skb buffer and use
3200 * them to generate a hash based on the xmit_policy set in the bonding device
169a3e66 3201 */
ee62e868 3202u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb)
169a3e66 3203{
32819dc1
NA
3204 struct flow_keys flow;
3205 u32 hash;
4394542c 3206
4b1b865e
ED
3207 if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 &&
3208 skb->l4_hash)
3209 return skb->hash;
3210
32819dc1
NA
3211 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 ||
3212 !bond_flow_dissect(bond, skb, &flow))
ee62e868 3213 return bond_eth_hash(skb);
169a3e66 3214
32819dc1
NA
3215 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 ||
3216 bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
3217 hash = bond_eth_hash(skb);
3218 else
06635a35 3219 hash = (__force u32)flow.ports.ports;
c3f83241
TH
3220 hash ^= (__force u32)flow_get_u32_dst(&flow) ^
3221 (__force u32)flow_get_u32_src(&flow);
32819dc1
NA
3222 hash ^= (hash >> 16);
3223 hash ^= (hash >> 8);
3224
ee62e868 3225 return hash;
169a3e66
JV
3226}
3227
1da177e4
LT
3228/*-------------------------- Device entry points ----------------------------*/
3229
fbb0c41b 3230static void bond_work_init_all(struct bonding *bond)
3231{
3232 INIT_DELAYED_WORK(&bond->mcast_work,
3233 bond_resend_igmp_join_requests_delayed);
3234 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3235 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
d5e73f7b 3236 INIT_DELAYED_WORK(&bond->arp_work, bond_arp_monitor);
fbb0c41b 3237 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
ee637714 3238 INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler);
fbb0c41b 3239}
3240
3241static void bond_work_cancel_all(struct bonding *bond)
3242{
3243 cancel_delayed_work_sync(&bond->mii_work);
3244 cancel_delayed_work_sync(&bond->arp_work);
3245 cancel_delayed_work_sync(&bond->alb_work);
3246 cancel_delayed_work_sync(&bond->ad_work);
3247 cancel_delayed_work_sync(&bond->mcast_work);
ee637714 3248 cancel_delayed_work_sync(&bond->slave_arr_work);
fbb0c41b 3249}
3250
1da177e4
LT
3251static int bond_open(struct net_device *bond_dev)
3252{
454d7c9b 3253 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3254 struct list_head *iter;
ba3211cc 3255 struct slave *slave;
1da177e4 3256
ba3211cc 3257 /* reset slave->backup and slave->inactive */
0965a1f3 3258 if (bond_has_slaves(bond)) {
9caff1e7 3259 bond_for_each_slave(bond, slave, iter) {
4740d638
ED
3260 if (bond_uses_primary(bond) &&
3261 slave != rcu_access_pointer(bond->curr_active_slave)) {
5e5b0665 3262 bond_set_slave_inactive_flags(slave,
3263 BOND_SLAVE_NOTIFY_NOW);
8bbe71a5 3264 } else if (BOND_MODE(bond) != BOND_MODE_8023AD) {
5e5b0665 3265 bond_set_slave_active_flags(slave,
3266 BOND_SLAVE_NOTIFY_NOW);
ba3211cc
PP
3267 }
3268 }
ba3211cc 3269 }
ba3211cc 3270
58402054 3271 if (bond_is_lb(bond)) {
1da177e4
LT
3272 /* bond_alb_initialize must be called before the timer
3273 * is started.
3274 */
01844098 3275 if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB)))
b473946a 3276 return -ENOMEM;
e9f0fb88
MB
3277 if (bond->params.tlb_dynamic_lb)
3278 queue_delayed_work(bond->wq, &bond->alb_work, 0);
1da177e4
LT
3279 }
3280
fbb0c41b 3281 if (bond->params.miimon) /* link check interval, in milliseconds. */
1b76b316 3282 queue_delayed_work(bond->wq, &bond->mii_work, 0);
1da177e4
LT
3283
3284 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316 3285 queue_delayed_work(bond->wq, &bond->arp_work, 0);
3fe68df9 3286 bond->recv_probe = bond_arp_rcv;
1da177e4
LT
3287 }
3288
01844098 3289 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1b76b316 3290 queue_delayed_work(bond->wq, &bond->ad_work, 0);
1da177e4 3291 /* register to receive LACPDUs */
3aba891d 3292 bond->recv_probe = bond_3ad_lacpdu_recv;
fd989c83 3293 bond_3ad_initiate_agg_selection(bond, 1);
1da177e4
LT
3294 }
3295
ee637714
MB
3296 if (bond_mode_uses_xmit_hash(bond))
3297 bond_update_slave_arr(bond, NULL);
3298
1da177e4
LT
3299 return 0;
3300}
3301
3302static int bond_close(struct net_device *bond_dev)
3303{
454d7c9b 3304 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3305
fbb0c41b 3306 bond_work_cancel_all(bond);
6c8d23f7 3307 bond->send_peer_notif = 0;
ee8487c0 3308 if (bond_is_lb(bond))
1da177e4 3309 bond_alb_deinitialize(bond);
3aba891d 3310 bond->recv_probe = NULL;
1da177e4
LT
3311
3312 return 0;
3313}
3314
fe30937b
ED
3315/* fold stats, assuming all rtnl_link_stats64 fields are u64, but
3316 * that some drivers can provide 32bit values only.
3317 */
3318static void bond_fold_stats(struct rtnl_link_stats64 *_res,
3319 const struct rtnl_link_stats64 *_new,
3320 const struct rtnl_link_stats64 *_old)
3321{
3322 const u64 *new = (const u64 *)_new;
3323 const u64 *old = (const u64 *)_old;
3324 u64 *res = (u64 *)_res;
3325 int i;
3326
3327 for (i = 0; i < sizeof(*_res) / sizeof(u64); i++) {
3328 u64 nv = new[i];
3329 u64 ov = old[i];
3330
3331 /* detects if this particular field is 32bit only */
3332 if (((nv | ov) >> 32) == 0)
3333 res[i] += (u32)nv - (u32)ov;
3334 else
3335 res[i] += nv - ov;
3336 }
3337}
3338
bc1f4470 3339static void bond_get_stats(struct net_device *bond_dev,
3340 struct rtnl_link_stats64 *stats)
1da177e4 3341{
454d7c9b 3342 struct bonding *bond = netdev_priv(bond_dev);
28172739 3343 struct rtnl_link_stats64 temp;
9caff1e7 3344 struct list_head *iter;
1da177e4 3345 struct slave *slave;
1da177e4 3346
fe30937b 3347 spin_lock(&bond->stats_lock);
5f0c5f73 3348 memcpy(stats, &bond->bond_stats, sizeof(*stats));
1da177e4 3349
fe30937b
ED
3350 rcu_read_lock();
3351 bond_for_each_slave_rcu(bond, slave, iter) {
3352 const struct rtnl_link_stats64 *new =
28172739 3353 dev_get_stats(slave->dev, &temp);
fe30937b
ED
3354
3355 bond_fold_stats(stats, new, &slave->slave_stats);
5f0c5f73
AG
3356
3357 /* save off the slave stats for the next run */
fe30937b 3358 memcpy(&slave->slave_stats, new, sizeof(*new));
5f0c5f73 3359 }
fe30937b
ED
3360 rcu_read_unlock();
3361
5f0c5f73 3362 memcpy(&bond->bond_stats, stats, sizeof(*stats));
fe30937b 3363 spin_unlock(&bond->stats_lock);
1da177e4
LT
3364}
3365
3366static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3367{
080a06e1 3368 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT
3369 struct net_device *slave_dev = NULL;
3370 struct ifbond k_binfo;
3371 struct ifbond __user *u_binfo = NULL;
3372 struct ifslave k_sinfo;
3373 struct ifslave __user *u_sinfo = NULL;
3374 struct mii_ioctl_data *mii = NULL;
d1fbd3ed 3375 struct bond_opt_value newval;
387ff911 3376 struct net *net;
1da177e4
LT
3377 int res = 0;
3378
76444f50 3379 netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd);
1da177e4
LT
3380
3381 switch (cmd) {
1da177e4
LT
3382 case SIOCGMIIPHY:
3383 mii = if_mii(ifr);
3d632c3f 3384 if (!mii)
1da177e4 3385 return -EINVAL;
3d632c3f 3386
1da177e4
LT
3387 mii->phy_id = 0;
3388 /* Fall Through */
3389 case SIOCGMIIREG:
547942ca 3390 /* We do this again just in case we were called by SIOCGMIIREG
1da177e4
LT
3391 * instead of SIOCGMIIPHY.
3392 */
3393 mii = if_mii(ifr);
3d632c3f 3394 if (!mii)
1da177e4 3395 return -EINVAL;
3d632c3f 3396
1da177e4 3397 if (mii->reg_num == 1) {
1da177e4 3398 mii->val_out = 0;
3d632c3f 3399 if (netif_carrier_ok(bond->dev))
1da177e4 3400 mii->val_out = BMSR_LSTATUS;
1da177e4
LT
3401 }
3402
3403 return 0;
3404 case BOND_INFO_QUERY_OLD:
3405 case SIOCBONDINFOQUERY:
3406 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3407
3d632c3f 3408 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
1da177e4 3409 return -EFAULT;
1da177e4 3410
3d67576d
ZY
3411 bond_info_query(bond_dev, &k_binfo);
3412 if (copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3d632c3f 3413 return -EFAULT;
1da177e4 3414
3d67576d 3415 return 0;
1da177e4
LT
3416 case BOND_SLAVE_INFO_QUERY_OLD:
3417 case SIOCBONDSLAVEINFOQUERY:
3418 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3419
3d632c3f 3420 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
1da177e4 3421 return -EFAULT;
1da177e4
LT
3422
3423 res = bond_slave_info_query(bond_dev, &k_sinfo);
3d632c3f
SH
3424 if (res == 0 &&
3425 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3426 return -EFAULT;
1da177e4
LT
3427
3428 return res;
3429 default:
1da177e4
LT
3430 break;
3431 }
3432
387ff911
G
3433 net = dev_net(bond_dev);
3434
3435 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4 3436 return -EPERM;
1da177e4 3437
0917b933 3438 slave_dev = __dev_get_by_name(net, ifr->ifr_slave);
1da177e4 3439
76444f50 3440 netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev);
1da177e4 3441
3d632c3f 3442 if (!slave_dev)
0917b933 3443 return -ENODEV;
1da177e4 3444
76444f50 3445 netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name);
0917b933
YX
3446 switch (cmd) {
3447 case BOND_ENSLAVE_OLD:
3448 case SIOCBONDENSLAVE:
3449 res = bond_enslave(bond_dev, slave_dev);
3450 break;
3451 case BOND_RELEASE_OLD:
3452 case SIOCBONDRELEASE:
3453 res = bond_release(bond_dev, slave_dev);
3454 break;
3455 case BOND_SETHWADDR_OLD:
3456 case SIOCBONDSETHWADDR:
3457 bond_set_dev_addr(bond_dev, slave_dev);
3458 res = 0;
3459 break;
3460 case BOND_CHANGE_ACTIVE_OLD:
3461 case SIOCBONDCHANGEACTIVE:
d1fbd3ed
NA
3462 bond_opt_initstr(&newval, slave_dev->name);
3463 res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval);
0917b933
YX
3464 break;
3465 default:
3466 res = -EOPNOTSUPP;
1da177e4
LT
3467 }
3468
1da177e4
LT
3469 return res;
3470}
3471
d03462b9 3472static void bond_change_rx_flags(struct net_device *bond_dev, int change)
1da177e4 3473{
454d7c9b 3474 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3475
d03462b9
JP
3476 if (change & IFF_PROMISC)
3477 bond_set_promiscuity(bond,
3478 bond_dev->flags & IFF_PROMISC ? 1 : -1);
3d632c3f 3479
d03462b9
JP
3480 if (change & IFF_ALLMULTI)
3481 bond_set_allmulti(bond,
3482 bond_dev->flags & IFF_ALLMULTI ? 1 : -1);
3483}
1da177e4 3484
303d1cbf 3485static void bond_set_rx_mode(struct net_device *bond_dev)
d03462b9
JP
3486{
3487 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3488 struct list_head *iter;
303d1cbf 3489 struct slave *slave;
1da177e4 3490
b3241870 3491 rcu_read_lock();
ec0865a9 3492 if (bond_uses_primary(bond)) {
b3241870 3493 slave = rcu_dereference(bond->curr_active_slave);
303d1cbf
JV
3494 if (slave) {
3495 dev_uc_sync(slave->dev, bond_dev);
3496 dev_mc_sync(slave->dev, bond_dev);
3497 }
303d1cbf 3498 } else {
b3241870 3499 bond_for_each_slave_rcu(bond, slave, iter) {
303d1cbf
JV
3500 dev_uc_sync_multiple(slave->dev, bond_dev);
3501 dev_mc_sync_multiple(slave->dev, bond_dev);
3502 }
1da177e4 3503 }
b3241870 3504 rcu_read_unlock();
1da177e4
LT
3505}
3506
234bcf8a 3507static int bond_neigh_init(struct neighbour *n)
00829823 3508{
234bcf8a 3509 struct bonding *bond = netdev_priv(n->dev);
234bcf8a
SP
3510 const struct net_device_ops *slave_ops;
3511 struct neigh_parms parms;
dec1e90e 3512 struct slave *slave;
234bcf8a
SP
3513 int ret;
3514
dec1e90e 3515 slave = bond_first_slave(bond);
234bcf8a
SP
3516 if (!slave)
3517 return 0;
234bcf8a 3518 slave_ops = slave->dev->netdev_ops;
234bcf8a
SP
3519 if (!slave_ops->ndo_neigh_setup)
3520 return 0;
3521
3522 parms.neigh_setup = NULL;
3523 parms.neigh_cleanup = NULL;
3524 ret = slave_ops->ndo_neigh_setup(slave->dev, &parms);
3525 if (ret)
3526 return ret;
3527
547942ca 3528 /* Assign slave's neigh_cleanup to neighbour in case cleanup is called
234bcf8a
SP
3529 * after the last slave has been detached. Assumes that all slaves
3530 * utilize the same neigh_cleanup (true at this writing as only user
3531 * is ipoib).
3532 */
3533 n->parms->neigh_cleanup = parms.neigh_cleanup;
3534
3535 if (!parms.neigh_setup)
3536 return 0;
3537
3538 return parms.neigh_setup(n);
3539}
3540
547942ca 3541/* The bonding ndo_neigh_setup is called at init time beofre any
234bcf8a
SP
3542 * slave exists. So we must declare proxy setup function which will
3543 * be used at run time to resolve the actual slave neigh param setup.
9918d5bf
VF
3544 *
3545 * It's also called by master devices (such as vlans) to setup their
3546 * underlying devices. In that case - do nothing, we're already set up from
3547 * our init.
234bcf8a
SP
3548 */
3549static int bond_neigh_setup(struct net_device *dev,
3550 struct neigh_parms *parms)
3551{
9918d5bf
VF
3552 /* modify only our neigh_parms */
3553 if (parms->dev == dev)
3554 parms->neigh_setup = bond_neigh_init;
00829823 3555
00829823
SH
3556 return 0;
3557}
3558
547942ca 3559/* Change the MTU of all of a master's slaves to match the master */
1da177e4
LT
3560static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3561{
454d7c9b 3562 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3563 struct slave *slave, *rollback_slave;
9caff1e7 3564 struct list_head *iter;
1da177e4 3565 int res = 0;
1da177e4 3566
76444f50 3567 netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu);
1da177e4 3568
9caff1e7 3569 bond_for_each_slave(bond, slave, iter) {
76444f50
VF
3570 netdev_dbg(bond_dev, "s %p c_m %p\n",
3571 slave, slave->dev->netdev_ops->ndo_change_mtu);
e944ef79 3572
1da177e4
LT
3573 res = dev_set_mtu(slave->dev, new_mtu);
3574
3575 if (res) {
3576 /* If we failed to set the slave's mtu to the new value
3577 * we must abort the operation even in ACTIVE_BACKUP
3578 * mode, because if we allow the backup slaves to have
3579 * different mtu values than the active slave we'll
3580 * need to change their mtu when doing a failover. That
3581 * means changing their mtu from timer context, which
3582 * is probably not a good idea.
3583 */
76444f50
VF
3584 netdev_dbg(bond_dev, "err %d %s\n", res,
3585 slave->dev->name);
1da177e4
LT
3586 goto unwind;
3587 }
3588 }
3589
3590 bond_dev->mtu = new_mtu;
3591
3592 return 0;
3593
3594unwind:
3595 /* unwind from head to the slave that failed */
9caff1e7 3596 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT
3597 int tmp_res;
3598
81f23b13
VF
3599 if (rollback_slave == slave)
3600 break;
3601
3602 tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu);
1da177e4 3603 if (tmp_res) {
76444f50
VF
3604 netdev_dbg(bond_dev, "unwind err %d dev %s\n",
3605 tmp_res, rollback_slave->dev->name);
1da177e4
LT
3606 }
3607 }
3608
3609 return res;
3610}
3611
547942ca 3612/* Change HW address
1da177e4
LT
3613 *
3614 * Note that many devices must be down to change the HW address, and
3615 * downing the master releases all slaves. We can make bonds full of
3616 * bonding devices to test this, however.
3617 */
3618static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
3619{
454d7c9b 3620 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3621 struct slave *slave, *rollback_slave;
1da177e4 3622 struct sockaddr *sa = addr, tmp_sa;
9caff1e7 3623 struct list_head *iter;
1da177e4 3624 int res = 0;
1da177e4 3625
01844098 3626 if (BOND_MODE(bond) == BOND_MODE_ALB)
eb7cc59a
SH
3627 return bond_alb_set_mac_address(bond_dev, addr);
3628
3629
76444f50 3630 netdev_dbg(bond_dev, "bond=%p\n", bond);
1da177e4 3631
1b5acd29
JV
3632 /* If fail_over_mac is enabled, do nothing and return success.
3633 * Returning an error causes ifenslave to fail.
dd957c57 3634 */
cc689aaa 3635 if (bond->params.fail_over_mac &&
01844098 3636 BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
dd957c57 3637 return 0;
2ab82852 3638
3d632c3f 3639 if (!is_valid_ether_addr(sa->sa_data))
1da177e4 3640 return -EADDRNOTAVAIL;
1da177e4 3641
9caff1e7 3642 bond_for_each_slave(bond, slave, iter) {
76444f50 3643 netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name);
1da177e4
LT
3644 res = dev_set_mac_address(slave->dev, addr);
3645 if (res) {
3646 /* TODO: consider downing the slave
3647 * and retry ?
3648 * User should expect communications
3649 * breakage anyway until ARP finish
3650 * updating, so...
3651 */
76444f50 3652 netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name);
1da177e4
LT
3653 goto unwind;
3654 }
3655 }
3656
3657 /* success */
3658 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
3659 return 0;
3660
3661unwind:
3662 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
3663 tmp_sa.sa_family = bond_dev->type;
3664
3665 /* unwind from head to the slave that failed */
9caff1e7 3666 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT
3667 int tmp_res;
3668
81f23b13
VF
3669 if (rollback_slave == slave)
3670 break;
3671
3672 tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa);
1da177e4 3673 if (tmp_res) {
76444f50
VF
3674 netdev_dbg(bond_dev, "unwind err %d dev %s\n",
3675 tmp_res, rollback_slave->dev->name);
1da177e4
LT
3676 }
3677 }
3678
3679 return res;
3680}
3681
15077228
NA
3682/**
3683 * bond_xmit_slave_id - transmit skb through slave with slave_id
3684 * @bond: bonding device that is transmitting
3685 * @skb: buffer to transmit
3686 * @slave_id: slave id up to slave_cnt-1 through which to transmit
3687 *
3688 * This function tries to transmit through slave with slave_id but in case
3689 * it fails, it tries to find the first available slave for transmission.
3690 * The skb is consumed in all cases, thus the function is void.
3691 */
da131ddb 3692static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id)
15077228 3693{
9caff1e7 3694 struct list_head *iter;
15077228
NA
3695 struct slave *slave;
3696 int i = slave_id;
3697
3698 /* Here we start from the slave with slave_id */
9caff1e7 3699 bond_for_each_slave_rcu(bond, slave, iter) {
15077228 3700 if (--i < 0) {
891ab54d 3701 if (bond_slave_can_tx(slave)) {
15077228
NA
3702 bond_dev_queue_xmit(bond, skb, slave->dev);
3703 return;
3704 }
3705 }
3706 }
3707
3708 /* Here we start from the first slave up to slave_id */
3709 i = slave_id;
9caff1e7 3710 bond_for_each_slave_rcu(bond, slave, iter) {
15077228
NA
3711 if (--i < 0)
3712 break;
891ab54d 3713 if (bond_slave_can_tx(slave)) {
15077228
NA
3714 bond_dev_queue_xmit(bond, skb, slave->dev);
3715 return;
3716 }
3717 }
3718 /* no slave that can tx has been found */
31aa860e 3719 bond_tx_drop(bond->dev, skb);
15077228
NA
3720}
3721
73958329
NA
3722/**
3723 * bond_rr_gen_slave_id - generate slave id based on packets_per_slave
3724 * @bond: bonding device to use
3725 *
3726 * Based on the value of the bonding device's packets_per_slave parameter
3727 * this function generates a slave id, which is usually used as the next
3728 * slave to transmit through.
3729 */
3730static u32 bond_rr_gen_slave_id(struct bonding *bond)
3731{
73958329 3732 u32 slave_id;
809fa972
HFS
3733 struct reciprocal_value reciprocal_packets_per_slave;
3734 int packets_per_slave = bond->params.packets_per_slave;
73958329
NA
3735
3736 switch (packets_per_slave) {
3737 case 0:
3738 slave_id = prandom_u32();
3739 break;
3740 case 1:
3741 slave_id = bond->rr_tx_counter;
3742 break;
3743 default:
809fa972
HFS
3744 reciprocal_packets_per_slave =
3745 bond->params.reciprocal_packets_per_slave;
73958329 3746 slave_id = reciprocal_divide(bond->rr_tx_counter,
809fa972 3747 reciprocal_packets_per_slave);
73958329
NA
3748 break;
3749 }
3750 bond->rr_tx_counter++;
3751
3752 return slave_id;
3753}
3754
1da177e4
LT
3755static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
3756{
454d7c9b 3757 struct bonding *bond = netdev_priv(bond_dev);
a2fd940f 3758 struct iphdr *iph = ip_hdr(skb);
15077228 3759 struct slave *slave;
73958329 3760 u32 slave_id;
1da177e4 3761
73958329 3762 /* Start with the curr_active_slave that joined the bond as the
a2fd940f
AG
3763 * default for sending IGMP traffic. For failover purposes one
3764 * needs to maintain some consistency for the interface that will
3765 * send the join/membership reports. The curr_active_slave found
3766 * will send all of this type of traffic.
cf5f9044 3767 */
15077228 3768 if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) {
278b2083 3769 slave = rcu_dereference(bond->curr_active_slave);
a64d044e 3770 if (slave)
15077228
NA
3771 bond_dev_queue_xmit(bond, skb, slave->dev);
3772 else
3773 bond_xmit_slave_id(bond, skb, 0);
a2fd940f 3774 } else {
9a72c2da
NA
3775 int slave_cnt = ACCESS_ONCE(bond->slave_cnt);
3776
3777 if (likely(slave_cnt)) {
3778 slave_id = bond_rr_gen_slave_id(bond);
3779 bond_xmit_slave_id(bond, skb, slave_id % slave_cnt);
3780 } else {
31aa860e 3781 bond_tx_drop(bond_dev, skb);
9a72c2da 3782 }
1da177e4 3783 }
0693e88e 3784
ec634fe3 3785 return NETDEV_TX_OK;
1da177e4
LT
3786}
3787
547942ca 3788/* In active-backup mode, we know that bond->curr_active_slave is always valid if
1da177e4
LT
3789 * the bond has a usable interface.
3790 */
3791static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
3792{
454d7c9b 3793 struct bonding *bond = netdev_priv(bond_dev);
71bc3b2d 3794 struct slave *slave;
1da177e4 3795
278b2083 3796 slave = rcu_dereference(bond->curr_active_slave);
71bc3b2d 3797 if (slave)
15077228
NA
3798 bond_dev_queue_xmit(bond, skb, slave->dev);
3799 else
31aa860e 3800 bond_tx_drop(bond_dev, skb);
0693e88e 3801
ec634fe3 3802 return NETDEV_TX_OK;
1da177e4
LT
3803}
3804
ee637714
MB
3805/* Use this to update slave_array when (a) it's not appropriate to update
3806 * slave_array right away (note that update_slave_array() may sleep)
3807 * and / or (b) RTNL is not held.
1da177e4 3808 */
ee637714 3809void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay)
1da177e4 3810{
ee637714
MB
3811 queue_delayed_work(bond->wq, &bond->slave_arr_work, delay);
3812}
1da177e4 3813
ee637714
MB
3814/* Slave array work handler. Holds only RTNL */
3815static void bond_slave_arr_handler(struct work_struct *work)
3816{
3817 struct bonding *bond = container_of(work, struct bonding,
3818 slave_arr_work.work);
3819 int ret;
3820
3821 if (!rtnl_trylock())
3822 goto err;
3823
3824 ret = bond_update_slave_arr(bond, NULL);
3825 rtnl_unlock();
3826 if (ret) {
3827 pr_warn_ratelimited("Failed to update slave array from WT\n");
3828 goto err;
3829 }
3830 return;
3831
3832err:
3833 bond_slave_arr_work_rearm(bond, 1);
3834}
3835
3836/* Build the usable slaves array in control path for modes that use xmit-hash
3837 * to determine the slave interface -
3838 * (a) BOND_MODE_8023AD
3839 * (b) BOND_MODE_XOR
3840 * (c) BOND_MODE_TLB && tlb_dynamic_lb == 0
3841 *
3842 * The caller is expected to hold RTNL only and NO other lock!
3843 */
3844int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave)
3845{
3846 struct slave *slave;
3847 struct list_head *iter;
3848 struct bond_up_slave *new_arr, *old_arr;
ee637714
MB
3849 int agg_id = 0;
3850 int ret = 0;
3851
3852#ifdef CONFIG_LOCKDEP
3853 WARN_ON(lockdep_is_held(&bond->mode_lock));
3854#endif
3855
3856 new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]),
3857 GFP_KERNEL);
3858 if (!new_arr) {
3859 ret = -ENOMEM;
3860 pr_err("Failed to build slave-array.\n");
3861 goto out;
3862 }
3863 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3864 struct ad_info ad_info;
3865
3866 if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
3867 pr_debug("bond_3ad_get_active_agg_info failed\n");
3868 kfree_rcu(new_arr, rcu);
3869 /* No active aggragator means it's not safe to use
3870 * the previous array.
3871 */
3872 old_arr = rtnl_dereference(bond->slave_arr);
3873 if (old_arr) {
3874 RCU_INIT_POINTER(bond->slave_arr, NULL);
3875 kfree_rcu(old_arr, rcu);
3876 }
3877 goto out;
3878 }
ee637714
MB
3879 agg_id = ad_info.aggregator_id;
3880 }
3881 bond_for_each_slave(bond, slave, iter) {
3882 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3883 struct aggregator *agg;
3884
3885 agg = SLAVE_AD_INFO(slave)->port.aggregator;
3886 if (!agg || agg->aggregator_identifier != agg_id)
3887 continue;
3888 }
3889 if (!bond_slave_can_tx(slave))
3890 continue;
3891 if (skipslave == slave)
3892 continue;
3893 new_arr->arr[new_arr->count++] = slave;
3894 }
3895
3896 old_arr = rtnl_dereference(bond->slave_arr);
3897 rcu_assign_pointer(bond->slave_arr, new_arr);
3898 if (old_arr)
3899 kfree_rcu(old_arr, rcu);
3900out:
3901 if (ret != 0 && skipslave) {
3902 int idx;
3903
3904 /* Rare situation where caller has asked to skip a specific
3905 * slave but allocation failed (most likely!). BTW this is
3906 * only possible when the call is initiated from
3907 * __bond_release_one(). In this situation; overwrite the
3908 * skipslave entry in the array with the last entry from the
3909 * array to avoid a situation where the xmit path may choose
3910 * this to-be-skipped slave to send a packet out.
3911 */
3912 old_arr = rtnl_dereference(bond->slave_arr);
3913 for (idx = 0; idx < old_arr->count; idx++) {
3914 if (skipslave == old_arr->arr[idx]) {
3915 old_arr->arr[idx] =
3916 old_arr->arr[old_arr->count-1];
3917 old_arr->count--;
3918 break;
3919 }
3920 }
3921 }
3922 return ret;
3923}
3924
3925/* Use this Xmit function for 3AD as well as XOR modes. The current
3926 * usable slave array is formed in the control path. The xmit function
3927 * just calculates hash and sends the packet out.
3928 */
30369104 3929static int bond_3ad_xor_xmit(struct sk_buff *skb, struct net_device *dev)
ee637714
MB
3930{
3931 struct bonding *bond = netdev_priv(dev);
3932 struct slave *slave;
3933 struct bond_up_slave *slaves;
3934 unsigned int count;
3935
3936 slaves = rcu_dereference(bond->slave_arr);
3937 count = slaves ? ACCESS_ONCE(slaves->count) : 0;
3938 if (likely(count)) {
3939 slave = slaves->arr[bond_xmit_hash(bond, skb) % count];
3940 bond_dev_queue_xmit(bond, skb, slave->dev);
3941 } else {
31aa860e 3942 bond_tx_drop(dev, skb);
ee637714 3943 }
0693e88e 3944
ec634fe3 3945 return NETDEV_TX_OK;
1da177e4
LT
3946}
3947
78a646ce 3948/* in broadcast mode, we send everything to all usable interfaces. */
1da177e4
LT
3949static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
3950{
454d7c9b 3951 struct bonding *bond = netdev_priv(bond_dev);
78a646ce 3952 struct slave *slave = NULL;
9caff1e7 3953 struct list_head *iter;
1da177e4 3954
9caff1e7 3955 bond_for_each_slave_rcu(bond, slave, iter) {
78a646ce
NA
3956 if (bond_is_last_slave(bond, slave))
3957 break;
b6adc610 3958 if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
78a646ce 3959 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1da177e4 3960
78a646ce 3961 if (!skb2) {
9152e26d 3962 net_err_ratelimited("%s: Error: %s: skb_clone() failed\n",
3963 bond_dev->name, __func__);
78a646ce 3964 continue;
1da177e4 3965 }
78a646ce 3966 bond_dev_queue_xmit(bond, skb2, slave->dev);
1da177e4
LT
3967 }
3968 }
b6adc610 3969 if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP)
78a646ce
NA
3970 bond_dev_queue_xmit(bond, skb, slave->dev);
3971 else
31aa860e 3972 bond_tx_drop(bond_dev, skb);
3d632c3f 3973
ec634fe3 3974 return NETDEV_TX_OK;
1da177e4
LT
3975}
3976
3977/*------------------------- Device initialization ---------------------------*/
3978
547942ca 3979/* Lookup the slave that corresponds to a qid */
bb1d9123
AG
3980static inline int bond_slave_override(struct bonding *bond,
3981 struct sk_buff *skb)
3982{
bb1d9123 3983 struct slave *slave = NULL;
9caff1e7 3984 struct list_head *iter;
bb1d9123 3985
0693e88e
MM
3986 if (!skb->queue_mapping)
3987 return 1;
bb1d9123
AG
3988
3989 /* Find out if any slaves have the same mapping as this skb. */
3900f290 3990 bond_for_each_slave_rcu(bond, slave, iter) {
3991 if (slave->queue_id == skb->queue_mapping) {
f5e2dc5d
AN
3992 if (bond_slave_is_up(slave) &&
3993 slave->link == BOND_LINK_UP) {
3900f290 3994 bond_dev_queue_xmit(bond, skb, slave->dev);
3995 return 0;
3996 }
3997 /* If the slave isn't UP, use default transmit policy. */
bb1d9123
AG
3998 break;
3999 }
4000 }
4001
3900f290 4002 return 1;
bb1d9123
AG
4003}
4004
374eeb5a 4005
f663dd9a 4006static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb,
99932d4f 4007 void *accel_priv, select_queue_fallback_t fallback)
bb1d9123 4008{
547942ca 4009 /* This helper function exists to help dev_pick_tx get the correct
fd0e435b 4010 * destination queue. Using a helper function skips a call to
bb1d9123
AG
4011 * skb_tx_hash and will put the skbs in the queue we expect on their
4012 * way down to the bonding driver.
4013 */
fd0e435b
PO
4014 u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0;
4015
547942ca 4016 /* Save the original txq to restore before passing to the driver */
df4ab5b3 4017 qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping;
374eeb5a 4018
fd0e435b 4019 if (unlikely(txq >= dev->real_num_tx_queues)) {
d30ee670 4020 do {
fd0e435b 4021 txq -= dev->real_num_tx_queues;
d30ee670 4022 } while (txq >= dev->real_num_tx_queues);
fd0e435b
PO
4023 }
4024 return txq;
bb1d9123
AG
4025}
4026
0693e88e 4027static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
00829823 4028{
bb1d9123
AG
4029 struct bonding *bond = netdev_priv(dev);
4030
d1e2e5cd
VF
4031 if (bond_should_override_tx_queue(bond) &&
4032 !bond_slave_override(bond, skb))
4033 return NETDEV_TX_OK;
00829823 4034
01844098 4035 switch (BOND_MODE(bond)) {
00829823
SH
4036 case BOND_MODE_ROUNDROBIN:
4037 return bond_xmit_roundrobin(skb, dev);
4038 case BOND_MODE_ACTIVEBACKUP:
4039 return bond_xmit_activebackup(skb, dev);
ee637714 4040 case BOND_MODE_8023AD:
00829823 4041 case BOND_MODE_XOR:
ee637714 4042 return bond_3ad_xor_xmit(skb, dev);
00829823
SH
4043 case BOND_MODE_BROADCAST:
4044 return bond_xmit_broadcast(skb, dev);
00829823 4045 case BOND_MODE_ALB:
00829823 4046 return bond_alb_xmit(skb, dev);
f05b42ea
MB
4047 case BOND_MODE_TLB:
4048 return bond_tlb_xmit(skb, dev);
00829823
SH
4049 default:
4050 /* Should never happen, mode already checked */
76444f50 4051 netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond));
00829823 4052 WARN_ON_ONCE(1);
31aa860e 4053 bond_tx_drop(dev, skb);
00829823
SH
4054 return NETDEV_TX_OK;
4055 }
4056}
4057
0693e88e
MM
4058static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
4059{
4060 struct bonding *bond = netdev_priv(dev);
4061 netdev_tx_t ret = NETDEV_TX_OK;
4062
547942ca 4063 /* If we risk deadlock from transmitting this in the
0693e88e
MM
4064 * netpoll path, tell netpoll to queue the frame for later tx
4065 */
054bb880 4066 if (unlikely(is_netpoll_tx_blocked(dev)))
0693e88e
MM
4067 return NETDEV_TX_BUSY;
4068
278b2083 4069 rcu_read_lock();
0965a1f3 4070 if (bond_has_slaves(bond))
0693e88e
MM
4071 ret = __bond_start_xmit(skb, dev);
4072 else
31aa860e 4073 bond_tx_drop(dev, skb);
278b2083 4074 rcu_read_unlock();
0693e88e
MM
4075
4076 return ret;
4077}
00829823 4078
d46b6349
PR
4079static int bond_ethtool_get_link_ksettings(struct net_device *bond_dev,
4080 struct ethtool_link_ksettings *cmd)
bb5b052f
AG
4081{
4082 struct bonding *bond = netdev_priv(bond_dev);
bb5b052f 4083 unsigned long speed = 0;
9caff1e7 4084 struct list_head *iter;
dec1e90e 4085 struct slave *slave;
bb5b052f 4086
d46b6349
PR
4087 cmd->base.duplex = DUPLEX_UNKNOWN;
4088 cmd->base.port = PORT_OTHER;
bb5b052f 4089
8557cd74 4090 /* Since bond_slave_can_tx returns false for all inactive or down slaves, we
bb5b052f
AG
4091 * do not need to check mode. Though link speed might not represent
4092 * the true receive or transmit bandwidth (not all modes are symmetric)
4093 * this is an accurate maximum.
4094 */
9caff1e7 4095 bond_for_each_slave(bond, slave, iter) {
8557cd74 4096 if (bond_slave_can_tx(slave)) {
bb5b052f
AG
4097 if (slave->speed != SPEED_UNKNOWN)
4098 speed += slave->speed;
d46b6349 4099 if (cmd->base.duplex == DUPLEX_UNKNOWN &&
bb5b052f 4100 slave->duplex != DUPLEX_UNKNOWN)
d46b6349 4101 cmd->base.duplex = slave->duplex;
bb5b052f
AG
4102 }
4103 }
d46b6349 4104 cmd->base.speed = speed ? : SPEED_UNKNOWN;
dec1e90e 4105
bb5b052f
AG
4106 return 0;
4107}
4108
217df670 4109static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
7826d43f 4110 struct ethtool_drvinfo *drvinfo)
217df670 4111{
7826d43f
JP
4112 strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
4113 strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
4114 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d",
4115 BOND_ABI_VERSION);
217df670
JV
4116}
4117
7282d491 4118static const struct ethtool_ops bond_ethtool_ops = {
217df670 4119 .get_drvinfo = bond_ethtool_get_drvinfo,
fa53ebac 4120 .get_link = ethtool_op_get_link,
d46b6349 4121 .get_link_ksettings = bond_ethtool_get_link_ksettings,
8531c5ff
AK
4122};
4123
eb7cc59a 4124static const struct net_device_ops bond_netdev_ops = {
181470fc 4125 .ndo_init = bond_init,
9e71626c 4126 .ndo_uninit = bond_uninit,
eb7cc59a
SH
4127 .ndo_open = bond_open,
4128 .ndo_stop = bond_close,
00829823 4129 .ndo_start_xmit = bond_start_xmit,
bb1d9123 4130 .ndo_select_queue = bond_select_queue,
be1f3c2c 4131 .ndo_get_stats64 = bond_get_stats,
eb7cc59a 4132 .ndo_do_ioctl = bond_do_ioctl,
d03462b9 4133 .ndo_change_rx_flags = bond_change_rx_flags,
303d1cbf 4134 .ndo_set_rx_mode = bond_set_rx_mode,
eb7cc59a 4135 .ndo_change_mtu = bond_change_mtu,
cc0e4070 4136 .ndo_set_mac_address = bond_set_mac_address,
00829823 4137 .ndo_neigh_setup = bond_neigh_setup,
cc0e4070 4138 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
eb7cc59a 4139 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
f6dc31a8 4140#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 4141 .ndo_netpoll_setup = bond_netpoll_setup,
f6dc31a8
WC
4142 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
4143 .ndo_poll_controller = bond_poll_controller,
4144#endif
9232ecca
JP
4145 .ndo_add_slave = bond_enslave,
4146 .ndo_del_slave = bond_release,
b2a103e6 4147 .ndo_fix_features = bond_fix_features,
fc8f40d8 4148 .ndo_bridge_setlink = switchdev_port_bridge_setlink,
85fdb956 4149 .ndo_bridge_getlink = switchdev_port_bridge_getlink,
54ba5a0b 4150 .ndo_bridge_dellink = switchdev_port_bridge_dellink,
45d4122c
SS
4151 .ndo_fdb_add = switchdev_port_fdb_add,
4152 .ndo_fdb_del = switchdev_port_fdb_del,
4153 .ndo_fdb_dump = switchdev_port_fdb_dump,
4847f049 4154 .ndo_features_check = passthru_features_check,
eb7cc59a
SH
4155};
4156
b3f92b63
DG
4157static const struct device_type bond_type = {
4158 .name = "bond",
4159};
4160
9e2e61fb
AW
4161static void bond_destructor(struct net_device *bond_dev)
4162{
4163 struct bonding *bond = netdev_priv(bond_dev);
4164 if (bond->wq)
4165 destroy_workqueue(bond->wq);
4166 free_netdev(bond_dev);
4167}
4168
0a2a78c4 4169void bond_setup(struct net_device *bond_dev)
1da177e4 4170{
454d7c9b 4171 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 4172
b7435628 4173 spin_lock_init(&bond->mode_lock);
fe30937b 4174 spin_lock_init(&bond->stats_lock);
d2991f75 4175 bond->params = bonding_defaults;
1da177e4
LT
4176
4177 /* Initialize pointers */
1da177e4 4178 bond->dev = bond_dev;
1da177e4
LT
4179
4180 /* Initialize the device entry points */
181470fc 4181 ether_setup(bond_dev);
31c05415 4182 bond_dev->max_mtu = ETH_MAX_MTU;
eb7cc59a 4183 bond_dev->netdev_ops = &bond_netdev_ops;
8531c5ff 4184 bond_dev->ethtool_ops = &bond_ethtool_ops;
1da177e4 4185
9e2e61fb 4186 bond_dev->destructor = bond_destructor;
1da177e4 4187
b3f92b63
DG
4188 SET_NETDEV_DEVTYPE(bond_dev, &bond_type);
4189
1da177e4 4190 /* Initialize the device options */
1098cee6 4191 bond_dev->flags |= IFF_MASTER;
1e6f20ca 4192 bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE;
550fd08c 4193 bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
181470fc 4194
547942ca 4195 /* don't acquire bond device's netif_tx_lock when transmitting */
1da177e4
LT
4196 bond_dev->features |= NETIF_F_LLTX;
4197
4198 /* By default, we declare the bond to be fully
4199 * VLAN hardware accelerated capable. Special
4200 * care is taken in the various xmit functions
4201 * when there are slaves that are not hw accel
4202 * capable
4203 */
1da177e4 4204
f9399814
WC
4205 /* Don't allow bond devices to change network namespaces. */
4206 bond_dev->features |= NETIF_F_NETNS_LOCAL;
4207
b2a103e6 4208 bond_dev->hw_features = BOND_VLAN_FEATURES |
f646968f
PM
4209 NETIF_F_HW_VLAN_CTAG_TX |
4210 NETIF_F_HW_VLAN_CTAG_RX |
4211 NETIF_F_HW_VLAN_CTAG_FILTER;
b2a103e6 4212
24f87d4c 4213 bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL;
b2a103e6 4214 bond_dev->features |= bond_dev->hw_features;
1da177e4
LT
4215}
4216
547942ca
NA
4217/* Destroy a bonding device.
4218 * Must be under rtnl_lock when this function is called.
4219 */
c67dfb29 4220static void bond_uninit(struct net_device *bond_dev)
a434e43f 4221{
454d7c9b 4222 struct bonding *bond = netdev_priv(bond_dev);
544a028e
VF
4223 struct list_head *iter;
4224 struct slave *slave;
ee637714 4225 struct bond_up_slave *arr;
a434e43f 4226
f6dc31a8
WC
4227 bond_netpoll_cleanup(bond_dev);
4228
c67dfb29 4229 /* Release the bonded slaves */
544a028e 4230 bond_for_each_slave(bond, slave, iter)
dec1e90e 4231 __bond_release_one(bond_dev, slave->dev, true);
76444f50 4232 netdev_info(bond_dev, "Released all slaves\n");
c67dfb29 4233
ee637714
MB
4234 arr = rtnl_dereference(bond->slave_arr);
4235 if (arr) {
4236 RCU_INIT_POINTER(bond->slave_arr, NULL);
4237 kfree_rcu(arr, rcu);
4238 }
4239
a434e43f
JV
4240 list_del(&bond->bond_list);
4241
f073c7ca 4242 bond_debug_unregister(bond);
a434e43f
JV
4243}
4244
1da177e4
LT
4245/*------------------------- Module initialization ---------------------------*/
4246
1da177e4
LT
4247static int bond_check_params(struct bond_params *params)
4248{
5a5c5fd4 4249 int arp_validate_value, fail_over_mac_value, primary_reselect_value, i;
f3253339 4250 struct bond_opt_value newval;
4251 const struct bond_opt_value *valptr;
8599b52e 4252 int arp_all_targets_value;
6791e466 4253 u16 ad_actor_sys_prio = 0;
d22a5fc0 4254 u16 ad_user_port_key = 0;
dc9c4d0f
MB
4255 __be32 arp_target[BOND_MAX_ARP_TARGETS];
4256 int arp_ip_count;
4257 int bond_mode = BOND_MODE_ROUNDROBIN;
4258 int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
4259 int lacp_fast = 0;
8b426dc5 4260 int tlb_dynamic_lb = 0;
f5b2b966 4261
547942ca 4262 /* Convert string parameters. */
1da177e4 4263 if (mode) {
2b3798d5
NA
4264 bond_opt_initstr(&newval, mode);
4265 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval);
4266 if (!valptr) {
4267 pr_err("Error: Invalid bonding mode \"%s\"\n", mode);
1da177e4
LT
4268 return -EINVAL;
4269 }
2b3798d5 4270 bond_mode = valptr->value;
1da177e4
LT
4271 }
4272
169a3e66
JV
4273 if (xmit_hash_policy) {
4274 if ((bond_mode != BOND_MODE_XOR) &&
f05b42ea
MB
4275 (bond_mode != BOND_MODE_8023AD) &&
4276 (bond_mode != BOND_MODE_TLB)) {
a4aee5c8 4277 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
90194264 4278 bond_mode_name(bond_mode));
169a3e66 4279 } else {
a4b32ce7
NA
4280 bond_opt_initstr(&newval, xmit_hash_policy);
4281 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH),
4282 &newval);
4283 if (!valptr) {
a4aee5c8 4284 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
169a3e66
JV
4285 xmit_hash_policy);
4286 return -EINVAL;
4287 }
a4b32ce7 4288 xmit_hashtype = valptr->value;
169a3e66
JV
4289 }
4290 }
4291
1da177e4
LT
4292 if (lacp_rate) {
4293 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8
JP
4294 pr_info("lacp_rate param is irrelevant in mode %s\n",
4295 bond_mode_name(bond_mode));
1da177e4 4296 } else {
d3131de7
NA
4297 bond_opt_initstr(&newval, lacp_rate);
4298 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE),
4299 &newval);
4300 if (!valptr) {
a4aee5c8 4301 pr_err("Error: Invalid lacp rate \"%s\"\n",
d3131de7 4302 lacp_rate);
1da177e4
LT
4303 return -EINVAL;
4304 }
d3131de7 4305 lacp_fast = valptr->value;
1da177e4
LT
4306 }
4307 }
4308
fd989c83 4309 if (ad_select) {
548d28bd 4310 bond_opt_initstr(&newval, ad_select);
9e5f5eeb
NA
4311 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT),
4312 &newval);
4313 if (!valptr) {
4314 pr_err("Error: Invalid ad_select \"%s\"\n", ad_select);
fd989c83
JV
4315 return -EINVAL;
4316 }
9e5f5eeb
NA
4317 params->ad_select = valptr->value;
4318 if (bond_mode != BOND_MODE_8023AD)
91565ebb 4319 pr_warn("ad_select param only affects 802.3ad mode\n");
fd989c83
JV
4320 } else {
4321 params->ad_select = BOND_AD_STABLE;
4322 }
4323
f5841306 4324 if (max_bonds < 0) {
91565ebb
JP
4325 pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4326 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
1da177e4
LT
4327 max_bonds = BOND_DEFAULT_MAX_BONDS;
4328 }
4329
4330 if (miimon < 0) {
91565ebb
JP
4331 pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4332 miimon, INT_MAX);
b98d9c66 4333 miimon = 0;
1da177e4
LT
4334 }
4335
4336 if (updelay < 0) {
91565ebb
JP
4337 pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4338 updelay, INT_MAX);
1da177e4
LT
4339 updelay = 0;
4340 }
4341
4342 if (downdelay < 0) {
91565ebb
JP
4343 pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4344 downdelay, INT_MAX);
1da177e4
LT
4345 downdelay = 0;
4346 }
4347
4348 if ((use_carrier != 0) && (use_carrier != 1)) {
91565ebb
JP
4349 pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4350 use_carrier);
1da177e4
LT
4351 use_carrier = 1;
4352 }
4353
ad246c99 4354 if (num_peer_notif < 0 || num_peer_notif > 255) {
91565ebb
JP
4355 pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4356 num_peer_notif);
ad246c99
BH
4357 num_peer_notif = 1;
4358 }
4359
834db4bc 4360 /* reset values for 802.3ad/TLB/ALB */
267bed77 4361 if (!bond_mode_uses_arp(bond_mode)) {
1da177e4 4362 if (!miimon) {
91565ebb
JP
4363 pr_warn("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n");
4364 pr_warn("Forcing miimon to 100msec\n");
fe9d04af 4365 miimon = BOND_DEFAULT_MIIMON;
1da177e4
LT
4366 }
4367 }
4368
bb1d9123 4369 if (tx_queues < 1 || tx_queues > 255) {
91565ebb
JP
4370 pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n",
4371 tx_queues, BOND_DEFAULT_TX_QUEUES);
bb1d9123
AG
4372 tx_queues = BOND_DEFAULT_TX_QUEUES;
4373 }
4374
ebd8e497 4375 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
91565ebb
JP
4376 pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n",
4377 all_slaves_active);
ebd8e497
AG
4378 all_slaves_active = 0;
4379 }
4380
c2952c31 4381 if (resend_igmp < 0 || resend_igmp > 255) {
91565ebb
JP
4382 pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n",
4383 resend_igmp, BOND_DEFAULT_RESEND_IGMP);
c2952c31
FL
4384 resend_igmp = BOND_DEFAULT_RESEND_IGMP;
4385 }
4386
aa59d851
NA
4387 bond_opt_initval(&newval, packets_per_slave);
4388 if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) {
73958329
NA
4389 pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n",
4390 packets_per_slave, USHRT_MAX);
4391 packets_per_slave = 1;
4392 }
4393
1da177e4 4394 if (bond_mode == BOND_MODE_ALB) {
a4aee5c8
JP
4395 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",
4396 updelay);
1da177e4
LT
4397 }
4398
4399 if (!miimon) {
4400 if (updelay || downdelay) {
4401 /* just warn the user the up/down delay will have
4402 * no effect since miimon is zero...
4403 */
91565ebb
JP
4404 pr_warn("Warning: miimon module parameter not set and updelay (%d) or downdelay (%d) module parameter is set; updelay and downdelay have no effect unless miimon is set\n",
4405 updelay, downdelay);
1da177e4
LT
4406 }
4407 } else {
4408 /* don't allow arp monitoring */
4409 if (arp_interval) {
91565ebb
JP
4410 pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4411 miimon, arp_interval);
1da177e4
LT
4412 arp_interval = 0;
4413 }
4414
4415 if ((updelay % miimon) != 0) {
91565ebb
JP
4416 pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4417 updelay, miimon, (updelay / miimon) * miimon);
1da177e4
LT
4418 }
4419
4420 updelay /= miimon;
4421
4422 if ((downdelay % miimon) != 0) {
91565ebb
JP
4423 pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4424 downdelay, miimon,
4425 (downdelay / miimon) * miimon);
1da177e4
LT
4426 }
4427
4428 downdelay /= miimon;
4429 }
4430
4431 if (arp_interval < 0) {
91565ebb
JP
4432 pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4433 arp_interval, INT_MAX);
7bdb04ed 4434 arp_interval = 0;
1da177e4
LT
4435 }
4436
5a5c5fd4 4437 for (arp_ip_count = 0, i = 0;
4438 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) {
89015c18 4439 __be32 ip;
547942ca
NA
4440
4441 /* not a complete check, but good enough to catch mistakes */
89015c18 4442 if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) ||
2807a9fe 4443 !bond_is_ip_target_ok(ip)) {
91565ebb
JP
4444 pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4445 arp_ip_target[i]);
1da177e4
LT
4446 arp_interval = 0;
4447 } else {
0afee4e8
VF
4448 if (bond_get_targets_ip(arp_target, ip) == -1)
4449 arp_target[arp_ip_count++] = ip;
4450 else
91565ebb
JP
4451 pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n",
4452 &ip);
1da177e4
LT
4453 }
4454 }
4455
4456 if (arp_interval && !arp_ip_count) {
4457 /* don't allow arping if no arp_ip_target given... */
91565ebb
JP
4458 pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4459 arp_interval);
1da177e4
LT
4460 arp_interval = 0;
4461 }
4462
f5b2b966 4463 if (arp_validate) {
f5b2b966 4464 if (!arp_interval) {
a4aee5c8 4465 pr_err("arp_validate requires arp_interval\n");
f5b2b966
JV
4466 return -EINVAL;
4467 }
4468
16228881
NA
4469 bond_opt_initstr(&newval, arp_validate);
4470 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE),
4471 &newval);
4472 if (!valptr) {
a4aee5c8 4473 pr_err("Error: invalid arp_validate \"%s\"\n",
16228881 4474 arp_validate);
f5b2b966
JV
4475 return -EINVAL;
4476 }
16228881
NA
4477 arp_validate_value = valptr->value;
4478 } else {
f5b2b966 4479 arp_validate_value = 0;
16228881 4480 }
f5b2b966 4481
8599b52e
VF
4482 arp_all_targets_value = 0;
4483 if (arp_all_targets) {
edf36b24
NA
4484 bond_opt_initstr(&newval, arp_all_targets);
4485 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS),
4486 &newval);
4487 if (!valptr) {
8599b52e
VF
4488 pr_err("Error: invalid arp_all_targets_value \"%s\"\n",
4489 arp_all_targets);
4490 arp_all_targets_value = 0;
edf36b24
NA
4491 } else {
4492 arp_all_targets_value = valptr->value;
8599b52e
VF
4493 }
4494 }
4495
1da177e4 4496 if (miimon) {
a4aee5c8 4497 pr_info("MII link monitoring set to %d ms\n", miimon);
1da177e4 4498 } else if (arp_interval) {
16228881
NA
4499 valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE,
4500 arp_validate_value);
a4aee5c8 4501 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
16228881 4502 arp_interval, valptr->string, arp_ip_count);
1da177e4
LT
4503
4504 for (i = 0; i < arp_ip_count; i++)
90194264 4505 pr_cont(" %s", arp_ip_target[i]);
1da177e4 4506
90194264 4507 pr_cont("\n");
1da177e4 4508
b8a9787e 4509 } else if (max_bonds) {
1da177e4
LT
4510 /* miimon and arp_interval not set, we need one so things
4511 * work as expected, see bonding.txt for details
4512 */
90194264 4513 pr_debug("Warning: either miimon or arp_interval and arp_ip_target module parameters must be specified, otherwise bonding will not detect link failures! see bonding.txt for details\n");
1da177e4
LT
4514 }
4515
ec0865a9 4516 if (primary && !bond_mode_uses_primary(bond_mode)) {
1da177e4
LT
4517 /* currently, using a primary only makes sense
4518 * in active backup, TLB or ALB modes
4519 */
91565ebb
JP
4520 pr_warn("Warning: %s primary device specified but has no effect in %s mode\n",
4521 primary, bond_mode_name(bond_mode));
1da177e4
LT
4522 primary = NULL;
4523 }
4524
a549952a 4525 if (primary && primary_reselect) {
388d3a6d
NA
4526 bond_opt_initstr(&newval, primary_reselect);
4527 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT),
4528 &newval);
4529 if (!valptr) {
a4aee5c8 4530 pr_err("Error: Invalid primary_reselect \"%s\"\n",
388d3a6d 4531 primary_reselect);
a549952a
JP
4532 return -EINVAL;
4533 }
388d3a6d 4534 primary_reselect_value = valptr->value;
a549952a
JP
4535 } else {
4536 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4537 }
4538
3915c1e8 4539 if (fail_over_mac) {
1df6b6aa
NA
4540 bond_opt_initstr(&newval, fail_over_mac);
4541 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC),
4542 &newval);
4543 if (!valptr) {
a4aee5c8 4544 pr_err("Error: invalid fail_over_mac \"%s\"\n",
1df6b6aa 4545 fail_over_mac);
3915c1e8
JV
4546 return -EINVAL;
4547 }
1df6b6aa 4548 fail_over_mac_value = valptr->value;
3915c1e8 4549 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
91565ebb 4550 pr_warn("Warning: fail_over_mac only affects active-backup mode\n");
3915c1e8
JV
4551 } else {
4552 fail_over_mac_value = BOND_FOM_NONE;
4553 }
dd957c57 4554
6791e466
MB
4555 bond_opt_initstr(&newval, "default");
4556 valptr = bond_opt_parse(
4557 bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO),
4558 &newval);
4559 if (!valptr) {
4560 pr_err("Error: No ad_actor_sys_prio default value");
4561 return -EINVAL;
4562 }
4563 ad_actor_sys_prio = valptr->value;
4564
d22a5fc0
MB
4565 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY),
4566 &newval);
4567 if (!valptr) {
4568 pr_err("Error: No ad_user_port_key default value");
4569 return -EINVAL;
4570 }
4571 ad_user_port_key = valptr->value;
4572
8b426dc5
MB
4573 if (bond_mode == BOND_MODE_TLB) {
4574 bond_opt_initstr(&newval, "default");
4575 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_TLB_DYNAMIC_LB),
4576 &newval);
4577 if (!valptr) {
4578 pr_err("Error: No tlb_dynamic_lb default value");
4579 return -EINVAL;
4580 }
4581 tlb_dynamic_lb = valptr->value;
4582 }
4583
3a7129e5 4584 if (lp_interval == 0) {
91565ebb
JP
4585 pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n",
4586 INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL);
3a7129e5 4587 lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
4588 }
4589
1da177e4
LT
4590 /* fill params struct with the proper values */
4591 params->mode = bond_mode;
169a3e66 4592 params->xmit_policy = xmit_hashtype;
1da177e4 4593 params->miimon = miimon;
ad246c99 4594 params->num_peer_notif = num_peer_notif;
1da177e4 4595 params->arp_interval = arp_interval;
f5b2b966 4596 params->arp_validate = arp_validate_value;
8599b52e 4597 params->arp_all_targets = arp_all_targets_value;
1da177e4
LT
4598 params->updelay = updelay;
4599 params->downdelay = downdelay;
4600 params->use_carrier = use_carrier;
4601 params->lacp_fast = lacp_fast;
4602 params->primary[0] = 0;
a549952a 4603 params->primary_reselect = primary_reselect_value;
3915c1e8 4604 params->fail_over_mac = fail_over_mac_value;
bb1d9123 4605 params->tx_queues = tx_queues;
ebd8e497 4606 params->all_slaves_active = all_slaves_active;
c2952c31 4607 params->resend_igmp = resend_igmp;
655f8919 4608 params->min_links = min_links;
3a7129e5 4609 params->lp_interval = lp_interval;
809fa972 4610 params->packets_per_slave = packets_per_slave;
8b426dc5 4611 params->tlb_dynamic_lb = tlb_dynamic_lb;
6791e466 4612 params->ad_actor_sys_prio = ad_actor_sys_prio;
74514957 4613 eth_zero_addr(params->ad_actor_system);
d22a5fc0 4614 params->ad_user_port_key = ad_user_port_key;
809fa972
HFS
4615 if (packets_per_slave > 0) {
4616 params->reciprocal_packets_per_slave =
4617 reciprocal_value(packets_per_slave);
4618 } else {
4619 /* reciprocal_packets_per_slave is unused if
4620 * packets_per_slave is 0 or 1, just initialize it
4621 */
4622 params->reciprocal_packets_per_slave =
4623 (struct reciprocal_value) { 0 };
4624 }
4625
1da177e4
LT
4626 if (primary) {
4627 strncpy(params->primary, primary, IFNAMSIZ);
4628 params->primary[IFNAMSIZ - 1] = 0;
4629 }
4630
4631 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4632
4633 return 0;
4634}
4635
547942ca 4636/* Called from registration process */
181470fc
SH
4637static int bond_init(struct net_device *bond_dev)
4638{
4639 struct bonding *bond = netdev_priv(bond_dev);
ec87fd3b 4640 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
181470fc 4641
76444f50 4642 netdev_dbg(bond_dev, "Begin bond_init\n");
181470fc 4643
f9f225eb 4644 bond->wq = alloc_ordered_workqueue(bond_dev->name, WQ_MEM_RECLAIM);
181470fc
SH
4645 if (!bond->wq)
4646 return -ENOMEM;
4647
d3fff6c4 4648 netdev_lockdep_set_classes(bond_dev);
181470fc 4649
ec87fd3b 4650 list_add_tail(&bond->bond_list, &bn->dev_list);
181470fc 4651
6151b3d4 4652 bond_prepare_sysfs_group(bond);
22bedad3 4653
f073c7ca
TI
4654 bond_debug_register(bond);
4655
409cc1f8
JP
4656 /* Ensure valid dev_addr */
4657 if (is_zero_ether_addr(bond_dev->dev_addr) &&
97a1e639 4658 bond_dev->addr_assign_type == NET_ADDR_PERM)
409cc1f8 4659 eth_hw_addr_random(bond_dev);
409cc1f8 4660
181470fc
SH
4661 return 0;
4662}
4663
0a2a78c4 4664unsigned int bond_get_num_tx_queues(void)
d5da4510 4665{
efacb309 4666 return tx_queues;
d5da4510
JP
4667}
4668
dfe60397 4669/* Create a new bond based on the specified name and bonding parameters.
e4b91c48 4670 * If name is NULL, obtain a suitable "bond%d" name for us.
dfe60397
MW
4671 * Caller must NOT hold rtnl_lock; we need to release it here before we
4672 * set up our sysfs entries.
4673 */
ec87fd3b 4674int bond_create(struct net *net, const char *name)
dfe60397
MW
4675{
4676 struct net_device *bond_dev;
e913fb27
P
4677 struct bonding *bond;
4678 struct alb_bond_info *bond_info;
dfe60397
MW
4679 int res;
4680
4681 rtnl_lock();
027ea041 4682
1c5cae81 4683 bond_dev = alloc_netdev_mq(sizeof(struct bonding),
c835a677 4684 name ? name : "bond%d", NET_NAME_UNKNOWN,
1c5cae81 4685 bond_setup, tx_queues);
dfe60397 4686 if (!bond_dev) {
a4aee5c8 4687 pr_err("%s: eek! can't alloc netdev!\n", name);
9e2e61fb
AW
4688 rtnl_unlock();
4689 return -ENOMEM;
dfe60397
MW
4690 }
4691
e913fb27
P
4692 /*
4693 * Initialize rx_hashtbl_used_head to RLB_NULL_INDEX.
4694 * It is set to 0 by default which is wrong.
4695 */
4696 bond = netdev_priv(bond_dev);
4697 bond_info = &(BOND_ALB_INFO(bond));
4698 bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX;
4699
ec87fd3b 4700 dev_net_set(bond_dev, net);
88ead977
EB
4701 bond_dev->rtnl_link_ops = &bond_link_ops;
4702
dfe60397 4703 res = register_netdevice(bond_dev);
0daa2303 4704
e826eafa
PO
4705 netif_carrier_off(bond_dev);
4706
4493b81b
MB
4707 bond_work_init_all(bond);
4708
7e083840 4709 rtnl_unlock();
9e2e61fb
AW
4710 if (res < 0)
4711 bond_destructor(bond_dev);
30c15ba9 4712 return res;
dfe60397
MW
4713}
4714
2c8c1e72 4715static int __net_init bond_net_init(struct net *net)
ec87fd3b 4716{
15449745 4717 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB
4718
4719 bn->net = net;
4720 INIT_LIST_HEAD(&bn->dev_list);
4721
ec87fd3b 4722 bond_create_proc_dir(bn);
4c22400a 4723 bond_create_sysfs(bn);
87a7b84b 4724
15449745 4725 return 0;
ec87fd3b
EB
4726}
4727
2c8c1e72 4728static void __net_exit bond_net_exit(struct net *net)
ec87fd3b 4729{
15449745 4730 struct bond_net *bn = net_generic(net, bond_net_id);
69b0216a 4731 struct bonding *bond, *tmp_bond;
4732 LIST_HEAD(list);
ec87fd3b 4733
4c22400a 4734 bond_destroy_sysfs(bn);
69b0216a 4735
4736 /* Kill off any bonds created after unregistering bond rtnl ops */
4737 rtnl_lock();
4738 list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list)
4739 unregister_netdevice_queue(bond->dev, &list);
4740 unregister_netdevice_many(&list);
4741 rtnl_unlock();
23fa5c2c
VF
4742
4743 bond_destroy_proc_dir(bn);
ec87fd3b
EB
4744}
4745
4746static struct pernet_operations bond_net_ops = {
4747 .init = bond_net_init,
4748 .exit = bond_net_exit,
15449745
EB
4749 .id = &bond_net_id,
4750 .size = sizeof(struct bond_net),
ec87fd3b
EB
4751};
4752
1da177e4
LT
4753static int __init bonding_init(void)
4754{
1da177e4
LT
4755 int i;
4756 int res;
4757
bd33acc3 4758 pr_info("%s", bond_version);
1da177e4 4759
dfe60397 4760 res = bond_check_params(&bonding_defaults);
3d632c3f 4761 if (res)
dfe60397 4762 goto out;
1da177e4 4763
15449745 4764 res = register_pernet_subsys(&bond_net_ops);
ec87fd3b
EB
4765 if (res)
4766 goto out;
027ea041 4767
0a2a78c4 4768 res = bond_netlink_init();
88ead977 4769 if (res)
6639104b 4770 goto err_link;
88ead977 4771
f073c7ca
TI
4772 bond_create_debugfs();
4773
1da177e4 4774 for (i = 0; i < max_bonds; i++) {
ec87fd3b 4775 res = bond_create(&init_net, NULL);
dfe60397
MW
4776 if (res)
4777 goto err;
1da177e4
LT
4778 }
4779
1da177e4 4780 register_netdevice_notifier(&bond_netdev_notifier);
dfe60397 4781out:
1da177e4 4782 return res;
88ead977 4783err:
db298686 4784 bond_destroy_debugfs();
0a2a78c4 4785 bond_netlink_fini();
6639104b 4786err_link:
15449745 4787 unregister_pernet_subsys(&bond_net_ops);
88ead977 4788 goto out;
dfe60397 4789
1da177e4
LT
4790}
4791
4792static void __exit bonding_exit(void)
4793{
4794 unregister_netdevice_notifier(&bond_netdev_notifier);
4795
f073c7ca 4796 bond_destroy_debugfs();
ae68c398 4797
0a2a78c4 4798 bond_netlink_fini();
ffcdedb6 4799 unregister_pernet_subsys(&bond_net_ops);
e843fa50
NH
4800
4801#ifdef CONFIG_NET_POLL_CONTROLLER
547942ca 4802 /* Make sure we don't have an imbalance on our netpoll blocking */
fb4fa76a 4803 WARN_ON(atomic_read(&netpoll_block_tx));
e843fa50 4804#endif
1da177e4
LT
4805}
4806
4807module_init(bonding_init);
4808module_exit(bonding_exit);
4809MODULE_LICENSE("GPL");
4810MODULE_VERSION(DRV_VERSION);
4811MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
4812MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");