projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
bonding: Convert pr_warning to pr_warn, neatening
[deliverable/linux.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
a4aee5c8
JP		 34#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
35
1da177e4
LT		 36#include <linux/kernel.h>
37#include <linux/module.h>
1da177e4
LT		 38#include <linux/types.h>
39#include <linux/fcntl.h>
40#include <linux/interrupt.h>
41#include <linux/ptrace.h>
42#include <linux/ioport.h>
43#include <linux/in.h>
169a3e66 44#include <net/ip.h>
1da177e4 45#include <linux/ip.h>
169a3e66
JV		 46#include <linux/tcp.h>
47#include <linux/udp.h>
1da177e4
LT		 48#include <linux/slab.h>
49#include <linux/string.h>
50#include <linux/init.h>
51#include <linux/timer.h>
52#include <linux/socket.h>
53#include <linux/ctype.h>
54#include <linux/inet.h>
55#include <linux/bitops.h>
3d632c3f 56#include <linux/io.h>
1da177e4 57#include <asm/dma.h>
3d632c3f 58#include <linux/uaccess.h>
1da177e4
LT		 59#include <linux/errno.h>
60#include <linux/netdevice.h>
61#include <linux/inetdevice.h>
a816c7c7 62#include <linux/igmp.h>
1da177e4
LT		 63#include <linux/etherdevice.h>
64#include <linux/skbuff.h>
65#include <net/sock.h>
66#include <linux/rtnetlink.h>
1da177e4
LT		 67#include <linux/smp.h>
68#include <linux/if_ether.h>
69#include <net/arp.h>
70#include <linux/mii.h>
71#include <linux/ethtool.h>
72#include <linux/if_vlan.h>
73#include <linux/if_bonding.h>
b63bb739 74#include <linux/jiffies.h>
e843fa50 75#include <linux/preempt.h>
c3ade5ca 76#include <net/route.h>
457c4cbc 77#include <net/net_namespace.h>
ec87fd3b 78#include <net/netns/generic.h>
5ee31c68 79#include <net/pkt_sched.h>
278b2083 80#include <linux/rculist.h>
32819dc1 81#include <net/flow_keys.h>
1da177e4
LT		 82#include "bonding.h"
83#include "bond_3ad.h"
84#include "bond_alb.h"
85
86/*---------------------------- Module parameters ----------------------------*/
87
88/* monitor all links that often (in milliseconds). <=0 disables monitoring */
1da177e4
LT		 89
90static int max_bonds = BOND_DEFAULT_MAX_BONDS;
bb1d9123 91static int tx_queues = BOND_DEFAULT_TX_QUEUES;
ad246c99 92static int num_peer_notif = 1;
b98d9c66 93static int miimon;
3d632c3f
SH		 94static int updelay;
95static int downdelay;
1da177e4 96static int use_carrier = 1;
3d632c3f
SH		 97static char *mode;
98static char *primary;
a549952a 99static char *primary_reselect;
3d632c3f 100static char *lacp_rate;
655f8919 101static int min_links;
3d632c3f
SH		 102static char *ad_select;
103static char *xmit_hash_policy;
7bdb04ed 104static int arp_interval;
3d632c3f
SH		 105static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
106static char *arp_validate;
8599b52e 107static char *arp_all_targets;
3d632c3f 108static char *fail_over_mac;
b07ea07b 109static int all_slaves_active;
d2991f75 110static struct bond_params bonding_defaults;
c2952c31 111static int resend_igmp = BOND_DEFAULT_RESEND_IGMP;
73958329 112static int packets_per_slave = 1;
3a7129e5 113static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
1da177e4
LT		 114
115module_param(max_bonds, int, 0);
116MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
bb1d9123
AG		 117module_param(tx_queues, int, 0);
118MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)");
ad246c99 119module_param_named(num_grat_arp, num_peer_notif, int, 0644);
90e62474
AG		 120MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on "
121 "failover event (alias of num_unsol_na)");
ad246c99 122module_param_named(num_unsol_na, num_peer_notif, int, 0644);
90e62474
AG		 123MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on "
124 "failover event (alias of num_grat_arp)");
1da177e4
LT		 125module_param(miimon, int, 0);
126MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
127module_param(updelay, int, 0);
128MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
129module_param(downdelay, int, 0);
2ac47660
MW		 130MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
131 "in milliseconds");
1da177e4 132module_param(use_carrier, int, 0);
2ac47660
MW		 133MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
134 "0 for off, 1 for on (default)");
1da177e4 135module_param(mode, charp, 0);
90e62474 136MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, "
2ac47660
MW		 137 "1 for active-backup, 2 for balance-xor, "
138 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
139 "6 for balance-alb");
1da177e4
LT		 140module_param(primary, charp, 0);
141MODULE_PARM_DESC(primary, "Primary network device to use");
a549952a
JP		 142module_param(primary_reselect, charp, 0);
143MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
144 "once it comes up; "
145 "0 for always (default), "
146 "1 for only if speed of primary is "
147 "better, "
148 "2 for only on active slave "
149 "failure");
1da177e4 150module_param(lacp_rate, charp, 0);
90e62474
AG		 151MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; "
152 "0 for slow, 1 for fast");
fd989c83 153module_param(ad_select, charp, 0);
90e62474
AG		 154MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; "
155 "0 for stable (default), 1 for bandwidth, "
156 "2 for count");
655f8919 157module_param(min_links, int, 0);
158MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier");
159
169a3e66 160module_param(xmit_hash_policy, charp, 0);
90e62474
AG		 161MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; "
162 "0 for layer 2 (default), 1 for layer 3+4, "
32819dc1
NA		 163 "2 for layer 2+3, 3 for encap layer 2+3, "
164 "4 for encap layer 3+4");
1da177e4
LT		 165module_param(arp_interval, int, 0);
166MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
167module_param_array(arp_ip_target, charp, NULL, 0);
168MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
f5b2b966 169module_param(arp_validate, charp, 0);
90e62474
AG		 170MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; "
171 "0 for none (default), 1 for active, "
172 "2 for backup, 3 for all");
8599b52e
VF		 173module_param(arp_all_targets, charp, 0);
174MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all");
3915c1e8 175module_param(fail_over_mac, charp, 0);
90e62474
AG		 176MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to "
177 "the same MAC; 0 for none (default), "
178 "1 for active, 2 for follow");
ebd8e497
AG		 179module_param(all_slaves_active, int, 0);
180MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface"
90e62474 181 "by setting active flag for all slaves; "
ebd8e497 182 "0 for never (default), 1 for always.");
c2952c31 183module_param(resend_igmp, int, 0);
90e62474
AG		 184MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on "
185 "link failure");
73958329
NA		 186module_param(packets_per_slave, int, 0);
187MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr "
188 "mode; 0 for a random slave, 1 packet per "
189 "slave (default), >1 packets per slave.");
3a7129e5 190module_param(lp_interval, uint, 0);
191MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where "
192 "the bonding driver sends learning packets to "
193 "each slaves peer switch. The default is 1.");
1da177e4
LT		 194
195/*----------------------------- Global variables ----------------------------*/
196
e843fa50 197#ifdef CONFIG_NET_POLL_CONTROLLER
fb4fa76a 198atomic_t netpoll_block_tx = ATOMIC_INIT(0);
e843fa50
NH		 199#endif
200
f99189b1 201int bond_net_id __read_mostly;
1da177e4 202
3d632c3f
SH		 203static __be32 arp_target[BOND_MAX_ARP_TARGETS];
204static int arp_ip_count;
1da177e4 205static int bond_mode = BOND_MODE_ROUNDROBIN;
3d632c3f
SH		 206static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
207static int lacp_fast;
1da177e4 208
1da177e4
LT		 209/*-------------------------- Forward declarations ---------------------------*/
210
181470fc 211static int bond_init(struct net_device *bond_dev);
c67dfb29 212static void bond_uninit(struct net_device *bond_dev);
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
258/*
cc0e4070
JP		 259 * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid,
260 * We don't protect the slave list iteration with a lock because:
1da177e4
LT		 261 * a. This operation is performed in IOCTL context,
262 * b. The operation is protected by the RTNL semaphore in the 8021q code,
263 * c. Holding a lock with BH disabled while directly calling a base driver
264 * entry point is generally a BAD idea.
3d632c3f 265 *
1da177e4
LT		 266 * The design of synchronization/protection for this operation in the 8021q
267 * module is good for one or more VLAN devices over a single physical device
268 * and cannot be extended for a teaming solution like bonding, so there is a
269 * potential race condition here where a net device from the vlan group might
270 * be referenced (either by a base driver or the 8021q code) while it is being
271 * removed from the system. However, it turns out we're not making matters
272 * worse, and if it works for regular VLAN usage it will work here too.
273*/
274
1da177e4
LT		 275/**
276 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
277 * @bond_dev: bonding net device that got called
278 * @vid: vlan id being added
279 */
80d5c368
PM		 280static int bond_vlan_rx_add_vid(struct net_device *bond_dev,
281 __be16 proto, u16 vid)
1da177e4 282{
454d7c9b 283 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 284 struct slave *slave, *rollback_slave;
9caff1e7 285 struct list_head *iter;
dec1e90e 286 int res;
1da177e4 287
9caff1e7 288 bond_for_each_slave(bond, slave, iter) {
80d5c368 289 res = vlan_vid_add(slave->dev, proto, vid);
87002b03
JP		 290 if (res)
291 goto unwind;
1da177e4
LT		 292 }
293
8e586137 294 return 0;
87002b03
JP		 295
296unwind:
81f23b13 297 /* unwind to the slave that failed */
9caff1e7 298 bond_for_each_slave(bond, rollback_slave, iter) {
81f23b13
VF		 299 if (rollback_slave == slave)
300 break;
301
302 vlan_vid_del(rollback_slave->dev, proto, vid);
303 }
87002b03
JP		 304
305 return res;
1da177e4
LT		 306}
307
308/**
309 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
310 * @bond_dev: bonding net device that got called
311 * @vid: vlan id being removed
312 */
80d5c368
PM		 313static int bond_vlan_rx_kill_vid(struct net_device *bond_dev,
314 __be16 proto, u16 vid)
1da177e4 315{
454d7c9b 316 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 317 struct list_head *iter;
1da177e4 318 struct slave *slave;
1da177e4 319
9caff1e7 320 bond_for_each_slave(bond, slave, iter)
80d5c368 321 vlan_vid_del(slave->dev, proto, vid);
1da177e4 322
e868b0c9
VF		 323 if (bond_is_lb(bond))
324 bond_alb_clear_vlan(bond, vid);
8e586137
JP		 325
326 return 0;
1da177e4
LT		 327}
328
1da177e4
LT		 329/*------------------------------- Link status -------------------------------*/
330
ff59c456
JV		 331/*
332 * Set the carrier state for the master according to the state of its
333 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
334 * do special 802.3ad magic.
335 *
336 * Returns zero if carrier state does not change, nonzero if it does.
337 */
338static int bond_set_carrier(struct bonding *bond)
339{
9caff1e7 340 struct list_head *iter;
ff59c456 341 struct slave *slave;
ff59c456 342
0965a1f3 343 if (!bond_has_slaves(bond))
ff59c456
JV		 344 goto down;
345
346 if (bond->params.mode == BOND_MODE_8023AD)
347 return bond_3ad_set_carrier(bond);
348
9caff1e7 349 bond_for_each_slave(bond, slave, iter) {
ff59c456
JV		 350 if (slave->link == BOND_LINK_UP) {
351 if (!netif_carrier_ok(bond->dev)) {
352 netif_carrier_on(bond->dev);
353 return 1;
354 }
355 return 0;
356 }
357 }
358
359down:
360 if (netif_carrier_ok(bond->dev)) {
361 netif_carrier_off(bond->dev);
362 return 1;
363 }
364 return 0;
365}
366
1da177e4
LT		 367/*
368 * Get link speed and duplex from the slave's base driver
369 * using ethtool. If for some reason the call fails or the
98f41f69 370 * values are invalid, set speed and duplex to -1,
13b95fb7 371 * and return.
1da177e4 372 */
13b95fb7 373static void bond_update_speed_duplex(struct slave *slave)
1da177e4
LT		 374{
375 struct net_device *slave_dev = slave->dev;
4bc71cb9 376 struct ethtool_cmd ecmd;
5d30530e 377 u32 slave_speed;
61a44b9c 378 int res;
1da177e4 379
589665f5
DC		 380 slave->speed = SPEED_UNKNOWN;
381 slave->duplex = DUPLEX_UNKNOWN;
1da177e4 382
4bc71cb9 383 res = __ethtool_get_settings(slave_dev, &ecmd);
61a44b9c 384 if (res < 0)
13b95fb7 385 return;
1da177e4 386
4bc71cb9 387 slave_speed = ethtool_cmd_speed(&ecmd);
6f92c66f 388 if (slave_speed == 0 || slave_speed == ((__u32) -1))
13b95fb7 389 return;
1da177e4 390
4bc71cb9 391 switch (ecmd.duplex) {
1da177e4
LT		 392 case DUPLEX_FULL:
393 case DUPLEX_HALF:
394 break;
395 default:
13b95fb7 396 return;
1da177e4
LT		 397 }
398
5d30530e 399 slave->speed = slave_speed;
4bc71cb9 400 slave->duplex = ecmd.duplex;
1da177e4 401
13b95fb7 402 return;
1da177e4
LT		 403}
404
07699f9a 405const char *bond_slave_link_status(s8 link)
406{
407 switch (link) {
408 case BOND_LINK_UP:
409 return "up";
410 case BOND_LINK_FAIL:
411 return "going down";
412 case BOND_LINK_DOWN:
413 return "down";
414 case BOND_LINK_BACK:
415 return "going back";
416 default:
417 return "unknown";
418 }
419}
420
1da177e4
LT		 421/*
422 * if <dev> supports MII link status reporting, check its link status.
423 *
424 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
3d632c3f 425 * depending upon the setting of the use_carrier parameter.
1da177e4
LT		 426 *
427 * Return either BMSR_LSTATUS, meaning that the link is up (or we
428 * can't tell and just pretend it is), or 0, meaning that the link is
429 * down.
430 *
431 * If reporting is non-zero, instead of faking link up, return -1 if
432 * both ETHTOOL and MII ioctls fail (meaning the device does not
433 * support them). If use_carrier is set, return whatever it says.
434 * It'd be nice if there was a good way to tell if a driver supports
435 * netif_carrier, but there really isn't.
436 */
3d632c3f
SH		 437static int bond_check_dev_link(struct bonding *bond,
438 struct net_device *slave_dev, int reporting)
1da177e4 439{
eb7cc59a 440 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
d9d52832 441 int (*ioctl)(struct net_device *, struct ifreq *, int);
1da177e4
LT		 442 struct ifreq ifr;
443 struct mii_ioctl_data *mii;
1da177e4 444
6c988853
PG		 445 if (!reporting && !netif_running(slave_dev))
446 return 0;
447
eb7cc59a 448 if (bond->params.use_carrier)
1da177e4 449 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
1da177e4 450
29112f4e 451 /* Try to get link status using Ethtool first. */
c772dde3
BH		 452 if (slave_dev->ethtool_ops->get_link)
453 return slave_dev->ethtool_ops->get_link(slave_dev) ?
454 BMSR_LSTATUS : 0;
29112f4e 455
3d632c3f 456 /* Ethtool can't be used, fallback to MII ioctls. */
eb7cc59a 457 ioctl = slave_ops->ndo_do_ioctl;
1da177e4
LT		 458 if (ioctl) {
459 /* TODO: set pointer to correct ioctl on a per team member */
460 /* bases to make this more efficient. that is, once */
461 /* we determine the correct ioctl, we will always */
462 /* call it and not the others for that team */
463 /* member. */
464
465 /*
466 * We cannot assume that SIOCGMIIPHY will also read a
467 * register; not all network drivers (e.g., e100)
468 * support that.
469 */
470
471 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
472 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
473 mii = if_mii(&ifr);
474 if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
475 mii->reg_num = MII_BMSR;
3d632c3f
SH		 476 if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0)
477 return mii->val_out & BMSR_LSTATUS;
1da177e4
LT		 478 }
479 }
480
1da177e4
LT		 481 /*
482 * If reporting, report that either there's no dev->do_ioctl,
61a44b9c 483 * or both SIOCGMIIREG and get_link failed (meaning that we
1da177e4
LT		 484 * cannot report link status). If not reporting, pretend
485 * we're ok.
486 */
3d632c3f 487 return reporting ? -1 : BMSR_LSTATUS;
1da177e4
LT		 488}
489
490/*----------------------------- Multicast list ------------------------------*/
491
1da177e4
LT		 492/*
493 * Push the promiscuity flag down to appropriate slaves
494 */
7e1a1ac1 495static int bond_set_promiscuity(struct bonding *bond, int inc)
1da177e4 496{
9caff1e7 497 struct list_head *iter;
7e1a1ac1 498 int err = 0;
9caff1e7 499
1da177e4
LT		 500 if (USES_PRIMARY(bond->params.mode)) {
501 /* write lock already acquired */
502 if (bond->curr_active_slave) {
7e1a1ac1
WC		 503 err = dev_set_promiscuity(bond->curr_active_slave->dev,
504 inc);
1da177e4
LT		 505 }
506 } else {
507 struct slave *slave;
dec1e90e 508
9caff1e7 509 bond_for_each_slave(bond, slave, iter) {
7e1a1ac1
WC		 510 err = dev_set_promiscuity(slave->dev, inc);
511 if (err)
512 return err;
1da177e4
LT		 513 }
514 }
7e1a1ac1 515 return err;
1da177e4
LT		 516}
517
518/*
519 * Push the allmulti flag down to all slaves
520 */
7e1a1ac1 521static int bond_set_allmulti(struct bonding *bond, int inc)
1da177e4 522{
9caff1e7 523 struct list_head *iter;
7e1a1ac1 524 int err = 0;
9caff1e7 525
1da177e4
LT		 526 if (USES_PRIMARY(bond->params.mode)) {
527 /* write lock already acquired */
528 if (bond->curr_active_slave) {
7e1a1ac1
WC		 529 err = dev_set_allmulti(bond->curr_active_slave->dev,
530 inc);
1da177e4
LT		 531 }
532 } else {
533 struct slave *slave;
dec1e90e 534
9caff1e7 535 bond_for_each_slave(bond, slave, iter) {
7e1a1ac1
WC		 536 err = dev_set_allmulti(slave->dev, inc);
537 if (err)
538 return err;
1da177e4
LT		 539 }
540 }
7e1a1ac1 541 return err;
1da177e4
LT		 542}
543
5a37e8ca
FL		 544/*
545 * Retrieve the list of registered multicast addresses for the bonding
546 * device and retransmit an IGMP JOIN request to the current active
547 * slave.
548 */
f2369109 549static void bond_resend_igmp_join_requests_delayed(struct work_struct *work)
5a37e8ca 550{
f2369109 551 struct bonding *bond = container_of(work, struct bonding,
552 mcast_work.work);
553
4aa5dee4 554 if (!rtnl_trylock()) {
4beac029 555 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
4aa5dee4 556 return;
5a37e8ca 557 }
4aa5dee4 558 call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev);
5a37e8ca 559
4f5474e7
NA		 560 if (bond->igmp_retrans > 1) {
561 bond->igmp_retrans--;
c2952c31 562 queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5);
4f5474e7 563 }
f2369109 564 rtnl_unlock();
5a37e8ca
FL		 565}
566
303d1cbf 567/* Flush bond's hardware addresses from slave
1da177e4 568 */
303d1cbf 569static void bond_hw_addr_flush(struct net_device *bond_dev,
3d632c3f 570 struct net_device *slave_dev)
1da177e4 571{
454d7c9b 572 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 573
303d1cbf
JV		 574 dev_uc_unsync(slave_dev, bond_dev);
575 dev_mc_unsync(slave_dev, bond_dev);
1da177e4
LT		 576
577 if (bond->params.mode == BOND_MODE_8023AD) {
578 /* del lacpdu mc addr from mc list */
579 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
580
22bedad3 581 dev_mc_del(slave_dev, lacpdu_multicast);
1da177e4
LT		 582 }
583}
584
585/*--------------------------- Active slave change ---------------------------*/
586
303d1cbf
JV		 587/* Update the hardware address list and promisc/allmulti for the new and
588 * old active slaves (if any). Modes that are !USES_PRIMARY keep all
589 * slaves up date at all times; only the USES_PRIMARY modes need to call
590 * this function to swap these settings during a failover.
1da177e4 591 */
303d1cbf
JV		 592static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active,
593 struct slave *old_active)
1da177e4 594{
7864a1ad
VF		 595 ASSERT_RTNL();
596
1da177e4 597 if (old_active) {
3d632c3f 598 if (bond->dev->flags & IFF_PROMISC)
1da177e4 599 dev_set_promiscuity(old_active->dev, -1);
1da177e4 600
3d632c3f 601 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 602 dev_set_allmulti(old_active->dev, -1);
1da177e4 603
303d1cbf 604 bond_hw_addr_flush(bond->dev, old_active->dev);
1da177e4
LT		 605 }
606
607 if (new_active) {
7e1a1ac1 608 /* FIXME: Signal errors upstream. */
3d632c3f 609 if (bond->dev->flags & IFF_PROMISC)
1da177e4 610 dev_set_promiscuity(new_active->dev, 1);
1da177e4 611
3d632c3f 612 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 613 dev_set_allmulti(new_active->dev, 1);
1da177e4 614
d632ce98 615 netif_addr_lock_bh(bond->dev);
303d1cbf
JV		 616 dev_uc_sync(new_active->dev, bond->dev);
617 dev_mc_sync(new_active->dev, bond->dev);
d632ce98 618 netif_addr_unlock_bh(bond->dev);
1da177e4
LT		 619 }
620}
621
ae0d6750 622/**
623 * bond_set_dev_addr - clone slave's address to bond
624 * @bond_dev: bond net device
625 * @slave_dev: slave net device
626 *
627 * Should be called with RTNL held.
628 */
629static void bond_set_dev_addr(struct net_device *bond_dev,
630 struct net_device *slave_dev)
631{
008aebde
NA		 632 pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n",
633 bond_dev, slave_dev, slave_dev->addr_len);
ae0d6750 634 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
635 bond_dev->addr_assign_type = NET_ADDR_STOLEN;
636 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev);
637}
638
3915c1e8
JV		 639/*
640 * bond_do_fail_over_mac
641 *
642 * Perform special MAC address swapping for fail_over_mac settings
643 *
b2e7aceb 644 * Called with RTNL, curr_slave_lock for write_bh.
3915c1e8
JV		 645 */
646static void bond_do_fail_over_mac(struct bonding *bond,
647 struct slave *new_active,
648 struct slave *old_active)
1f78d9f9 649 __releases(&bond->curr_slave_lock)
1f78d9f9 650 __acquires(&bond->curr_slave_lock)
3915c1e8
JV		 651{
652 u8 tmp_mac[ETH_ALEN];
653 struct sockaddr saddr;
654 int rv;
655
656 switch (bond->params.fail_over_mac) {
657 case BOND_FOM_ACTIVE:
7d26bb10 658 if (new_active) {
7d26bb10 659 write_unlock_bh(&bond->curr_slave_lock);
ae0d6750 660 bond_set_dev_addr(bond->dev, new_active->dev);
7d26bb10
WP		 661 write_lock_bh(&bond->curr_slave_lock);
662 }
3915c1e8
JV		 663 break;
664 case BOND_FOM_FOLLOW:
665 /*
666 * if new_active && old_active, swap them
667 * if just old_active, do nothing (going to no active slave)
668 * if just new_active, set new_active to bond's MAC
669 */
670 if (!new_active)
671 return;
672
673 write_unlock_bh(&bond->curr_slave_lock);
3915c1e8
JV		 674
675 if (old_active) {
676 memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN);
677 memcpy(saddr.sa_data, old_active->dev->dev_addr,
678 ETH_ALEN);
679 saddr.sa_family = new_active->dev->type;
680 } else {
681 memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN);
682 saddr.sa_family = bond->dev->type;
683 }
684
685 rv = dev_set_mac_address(new_active->dev, &saddr);
686 if (rv) {
a4aee5c8 687 pr_err("%s: Error %d setting MAC of slave %s\n",
3915c1e8
JV		 688 bond->dev->name, -rv, new_active->dev->name);
689 goto out;
690 }
691
692 if (!old_active)
693 goto out;
694
695 memcpy(saddr.sa_data, tmp_mac, ETH_ALEN);
696 saddr.sa_family = old_active->dev->type;
697
698 rv = dev_set_mac_address(old_active->dev, &saddr);
699 if (rv)
a4aee5c8 700 pr_err("%s: Error %d setting MAC of slave %s\n",
3915c1e8
JV		 701 bond->dev->name, -rv, new_active->dev->name);
702out:
3915c1e8
JV		 703 write_lock_bh(&bond->curr_slave_lock);
704 break;
705 default:
a4aee5c8 706 pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n",
3915c1e8
JV		 707 bond->dev->name, bond->params.fail_over_mac);
708 break;
709 }
710
711}
712
a549952a
JP		 713static bool bond_should_change_active(struct bonding *bond)
714{
715 struct slave *prim = bond->primary_slave;
716 struct slave *curr = bond->curr_active_slave;
717
718 if (!prim || !curr || curr->link != BOND_LINK_UP)
719 return true;
720 if (bond->force_primary) {
721 bond->force_primary = false;
722 return true;
723 }
724 if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER &&
725 (prim->speed < curr->speed ||
726 (prim->speed == curr->speed && prim->duplex <= curr->duplex)))
727 return false;
728 if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE)
729 return false;
730 return true;
731}
3915c1e8 732
1da177e4
LT		 733/**
734 * find_best_interface - select the best available slave to be the active one
735 * @bond: our bonding struct
1da177e4
LT		 736 */
737static struct slave *bond_find_best_slave(struct bonding *bond)
738{
77140d29
VF		 739 struct slave *slave, *bestslave = NULL;
740 struct list_head *iter;
1da177e4 741 int mintime = bond->params.updelay;
1da177e4 742
77140d29
VF		 743 if (bond->primary_slave && bond->primary_slave->link == BOND_LINK_UP &&
744 bond_should_change_active(bond))
745 return bond->primary_slave;
1da177e4 746
77140d29
VF		 747 bond_for_each_slave(bond, slave, iter) {
748 if (slave->link == BOND_LINK_UP)
749 return slave;
750 if (slave->link == BOND_LINK_BACK && IS_UP(slave->dev) &&
751 slave->delay < mintime) {
752 mintime = slave->delay;
753 bestslave = slave;
1da177e4
LT		 754 }
755 }
756
757 return bestslave;
758}
759
ad246c99
BH		 760static bool bond_should_notify_peers(struct bonding *bond)
761{
4cb4f97b 762 struct slave *slave;
763
764 rcu_read_lock();
765 slave = rcu_dereference(bond->curr_active_slave);
766 rcu_read_unlock();
ad246c99
BH		 767
768 pr_debug("bond_should_notify_peers: bond %s slave %s\n",
769 bond->dev->name, slave ? slave->dev->name : "NULL");
770
771 if (!slave || !bond->send_peer_notif ||
772 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
773 return false;
774
ad246c99
BH		 775 return true;
776}
777
1da177e4
LT		 778/**
779 * change_active_interface - change the active slave into the specified one
780 * @bond: our bonding struct
781 * @new: the new slave to make the active one
782 *
783 * Set the new slave to the bond's settings and unset them on the old
784 * curr_active_slave.
785 * Setting include flags, mc-list, promiscuity, allmulti, etc.
786 *
787 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
788 * because it is apparently the best available slave we have, even though its
789 * updelay hasn't timed out yet.
790 *
b2e7aceb 791 * If new_active is not NULL, caller must hold curr_slave_lock for write_bh.
1da177e4 792 */
a77b5325 793void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
1da177e4
LT		 794{
795 struct slave *old_active = bond->curr_active_slave;
796
3d632c3f 797 if (old_active == new_active)
1da177e4 798 return;
1da177e4
LT		 799
800 if (new_active) {
b2220cad
JV		 801 new_active->jiffies = jiffies;
802
1da177e4
LT		 803 if (new_active->link == BOND_LINK_BACK) {
804 if (USES_PRIMARY(bond->params.mode)) {
a4aee5c8
JP		 805 pr_info("%s: making interface %s the new active one %d ms earlier.\n",
806 bond->dev->name, new_active->dev->name,
807 (bond->params.updelay - new_active->delay) * bond->params.miimon);
1da177e4
LT		 808 }
809
810 new_active->delay = 0;
811 new_active->link = BOND_LINK_UP;
1da177e4 812
3d632c3f 813 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 814 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
1da177e4 815
58402054 816 if (bond_is_lb(bond))
1da177e4 817 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
1da177e4
LT		 818 } else {
819 if (USES_PRIMARY(bond->params.mode)) {
a4aee5c8
JP		 820 pr_info("%s: making interface %s the new active one.\n",
821 bond->dev->name, new_active->dev->name);
1da177e4
LT		 822 }
823 }
824 }
825
3d632c3f 826 if (USES_PRIMARY(bond->params.mode))
303d1cbf 827 bond_hw_addr_swap(bond, new_active, old_active);
1da177e4 828
58402054 829 if (bond_is_lb(bond)) {
1da177e4 830 bond_alb_handle_active_change(bond, new_active);
8f903c70
JV		 831 if (old_active)
832 bond_set_slave_inactive_flags(old_active);
833 if (new_active)
834 bond_set_slave_active_flags(new_active);
1da177e4 835 } else {
278b2083 836 rcu_assign_pointer(bond->curr_active_slave, new_active);
1da177e4 837 }
c3ade5ca
JV		 838
839 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
3d632c3f 840 if (old_active)
c3ade5ca 841 bond_set_slave_inactive_flags(old_active);
c3ade5ca
JV		 842
843 if (new_active) {
ad246c99
BH		 844 bool should_notify_peers = false;
845
c3ade5ca 846 bond_set_slave_active_flags(new_active);
2ab82852 847
709f8a45
OG		 848 if (bond->params.fail_over_mac)
849 bond_do_fail_over_mac(bond, new_active,
850 old_active);
3915c1e8 851
ad246c99
BH		 852 if (netif_running(bond->dev)) {
853 bond->send_peer_notif =
854 bond->params.num_peer_notif;
855 should_notify_peers =
856 bond_should_notify_peers(bond);
857 }
858
01f3109d 859 write_unlock_bh(&bond->curr_slave_lock);
01f3109d 860
b7bc2a5b 861 call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
ad246c99 862 if (should_notify_peers)
b7bc2a5b
AW		 863 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
864 bond->dev);
01f3109d 865
01f3109d 866 write_lock_bh(&bond->curr_slave_lock);
7893b249 867 }
c3ade5ca 868 }
a2fd940f 869
5a37e8ca 870 /* resend IGMP joins since active slave has changed or
94265cf5
FL		 871 * all were sent on curr_active_slave.
872 * resend only if bond is brought up with the affected
873 * bonding modes and the retransmission is enabled */
874 if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) &&
875 ((USES_PRIMARY(bond->params.mode) && new_active) ||
876 bond->params.mode == BOND_MODE_ROUNDROBIN)) {
c2952c31 877 bond->igmp_retrans = bond->params.resend_igmp;
4beac029 878 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
a2fd940f 879 }
1da177e4
LT		 880}
881
882/**
883 * bond_select_active_slave - select a new active slave, if needed
884 * @bond: our bonding struct
885 *
3d632c3f 886 * This functions should be called when one of the following occurs:
1da177e4
LT		 887 * - The old curr_active_slave has been released or lost its link.
888 * - The primary_slave has got its link back.
889 * - A slave has got its link back and there's no old curr_active_slave.
890 *
b2e7aceb 891 * Caller must hold curr_slave_lock for write_bh.
1da177e4 892 */
a77b5325 893void bond_select_active_slave(struct bonding *bond)
1da177e4
LT		 894{
895 struct slave *best_slave;
ff59c456 896 int rv;
1da177e4
LT		 897
898 best_slave = bond_find_best_slave(bond);
899 if (best_slave != bond->curr_active_slave) {
900 bond_change_active_slave(bond, best_slave);
ff59c456
JV		 901 rv = bond_set_carrier(bond);
902 if (!rv)
903 return;
904
905 if (netif_carrier_ok(bond->dev)) {
a4aee5c8
JP		 906 pr_info("%s: first active interface up!\n",
907 bond->dev->name);
ff59c456 908 } else {
a4aee5c8
JP		 909 pr_info("%s: now running without any active interface !\n",
910 bond->dev->name);
ff59c456 911 }
1da177e4
LT		 912 }
913}
914
f6dc31a8 915#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 916static inline int slave_enable_netpoll(struct slave *slave)
f6dc31a8 917{
8a8efa22
AW		 918 struct netpoll *np;
919 int err = 0;
f6dc31a8 920
47be03a2 921 np = kzalloc(sizeof(*np), GFP_ATOMIC);
8a8efa22
AW		 922 err = -ENOMEM;
923 if (!np)
924 goto out;
925
47be03a2 926 err = __netpoll_setup(np, slave->dev, GFP_ATOMIC);
8a8efa22
AW		 927 if (err) {
928 kfree(np);
929 goto out;
f6dc31a8 930 }
8a8efa22
AW		 931 slave->np = np;
932out:
933 return err;
934}
935static inline void slave_disable_netpoll(struct slave *slave)
936{
937 struct netpoll *np = slave->np;
938
939 if (!np)
940 return;
941
942 slave->np = NULL;
2cde6acd 943 __netpoll_free_async(np);
8a8efa22
AW		 944}
945static inline bool slave_dev_support_netpoll(struct net_device *slave_dev)
946{
947 if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL)
948 return false;
949 if (!slave_dev->netdev_ops->ndo_poll_controller)
950 return false;
951 return true;
f6dc31a8
WC		 952}
953
954static void bond_poll_controller(struct net_device *bond_dev)
955{
8a8efa22
AW		 956}
957
c4cdef9b 958static void bond_netpoll_cleanup(struct net_device *bond_dev)
8a8efa22 959{
c4cdef9b 960 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 961 struct list_head *iter;
c2355e1a 962 struct slave *slave;
c2355e1a 963
9caff1e7 964 bond_for_each_slave(bond, slave, iter)
8a8efa22
AW		 965 if (IS_UP(slave->dev))
966 slave_disable_netpoll(slave);
f6dc31a8 967}
8a8efa22 968
47be03a2 969static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp)
8a8efa22
AW		 970{
971 struct bonding *bond = netdev_priv(dev);
9caff1e7 972 struct list_head *iter;
f6dc31a8 973 struct slave *slave;
dec1e90e 974 int err = 0;
f6dc31a8 975
9caff1e7 976 bond_for_each_slave(bond, slave, iter) {
8a8efa22
AW		 977 err = slave_enable_netpoll(slave);
978 if (err) {
c4cdef9b 979 bond_netpoll_cleanup(dev);
8a8efa22 980 break;
f6dc31a8
WC		 981 }
982 }
8a8efa22 983 return err;
f6dc31a8 984}
8a8efa22
AW		 985#else
986static inline int slave_enable_netpoll(struct slave *slave)
987{
988 return 0;
989}
990static inline void slave_disable_netpoll(struct slave *slave)
991{
992}
f6dc31a8
WC		 993static void bond_netpoll_cleanup(struct net_device *bond_dev)
994{
995}
f6dc31a8
WC		 996#endif
997
1da177e4
LT		 998/*---------------------------------- IOCTL ----------------------------------*/
999
c8f44aff 1000static netdev_features_t bond_fix_features(struct net_device *dev,
9b7b165a 1001 netdev_features_t features)
8531c5ff 1002{
b2a103e6 1003 struct bonding *bond = netdev_priv(dev);
9caff1e7 1004 struct list_head *iter;
c8f44aff 1005 netdev_features_t mask;
9b7b165a 1006 struct slave *slave;
b63365a2 1007
0965a1f3 1008 if (!bond_has_slaves(bond)) {
b2a103e6
MM		 1009 /* Disable adding VLANs to empty bond. But why? --mq */
1010 features |= NETIF_F_VLAN_CHALLENGED;
9b7b165a 1011 return features;
b2a103e6 1012 }
b63365a2 1013
b2a103e6 1014 mask = features;
b63365a2 1015 features &= ~NETIF_F_ONE_FOR_ALL;
b2a103e6 1016 features |= NETIF_F_ALL_FOR_ALL;
7f353bf2 1017
9caff1e7 1018 bond_for_each_slave(bond, slave, iter) {
b63365a2
HX		 1019 features = netdev_increment_features(features,
1020 slave->dev->features,
b2a103e6
MM		 1021 mask);
1022 }
b0ce3508 1023 features = netdev_add_tso_features(features, mask);
b2a103e6 1024
b2a103e6
MM		 1025 return features;
1026}
1027
62f2a3a4
MM		 1028#define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \
1029 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
1030 NETIF_F_HIGHDMA | NETIF_F_LRO)
b2a103e6
MM		 1031
1032static void bond_compute_features(struct bonding *bond)
1033{
c4826861 1034 unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE;
c8f44aff 1035 netdev_features_t vlan_features = BOND_VLAN_FEATURES;
9caff1e7
VF		 1036 struct net_device *bond_dev = bond->dev;
1037 struct list_head *iter;
1038 struct slave *slave;
b2a103e6 1039 unsigned short max_hard_header_len = ETH_HLEN;
0e376bd0
SB		 1040 unsigned int gso_max_size = GSO_MAX_SIZE;
1041 u16 gso_max_segs = GSO_MAX_SEGS;
b2a103e6 1042
0965a1f3 1043 if (!bond_has_slaves(bond))
b2a103e6
MM		 1044 goto done;
1045
9caff1e7 1046 bond_for_each_slave(bond, slave, iter) {
278339a4 1047 vlan_features = netdev_increment_features(vlan_features,
b2a103e6
MM		 1048 slave->dev->vlan_features, BOND_VLAN_FEATURES);
1049
b6fe83e9 1050 dst_release_flag &= slave->dev->priv_flags;
54ef3137
JV		 1051 if (slave->dev->hard_header_len > max_hard_header_len)
1052 max_hard_header_len = slave->dev->hard_header_len;
0e376bd0
SB		 1053
1054 gso_max_size = min(gso_max_size, slave->dev->gso_max_size);
1055 gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs);
54ef3137 1056 }
8531c5ff 1057
b63365a2 1058done:
b2a103e6 1059 bond_dev->vlan_features = vlan_features;
54ef3137 1060 bond_dev->hard_header_len = max_hard_header_len;
0e376bd0
SB		 1061 bond_dev->gso_max_segs = gso_max_segs;
1062 netif_set_gso_max_size(bond_dev, gso_max_size);
8531c5ff 1063
b6fe83e9
ED		 1064 flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE;
1065 bond_dev->priv_flags = flags | dst_release_flag;
1066
b2a103e6 1067 netdev_change_features(bond_dev);
8531c5ff
AK		 1068}
1069
872254dd
MS		 1070static void bond_setup_by_slave(struct net_device *bond_dev,
1071 struct net_device *slave_dev)
1072{
00829823 1073 bond_dev->header_ops = slave_dev->header_ops;
872254dd
MS		 1074
1075 bond_dev->type = slave_dev->type;
1076 bond_dev->hard_header_len = slave_dev->hard_header_len;
1077 bond_dev->addr_len = slave_dev->addr_len;
1078
1079 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1080 slave_dev->addr_len);
1081}
1082
5b2c4dd2 1083/* On bonding slaves other than the currently active slave, suppress
3aba891d 1084 * duplicates except for alb non-mcast/bcast.
5b2c4dd2
JP		 1085 */
1086static bool bond_should_deliver_exact_match(struct sk_buff *skb,
0bd80dad
JP		 1087 struct slave *slave,
1088 struct bonding *bond)
5b2c4dd2 1089{
2d7011ca 1090 if (bond_is_slave_inactive(slave)) {
0bd80dad 1091 if (bond->params.mode == BOND_MODE_ALB &&
5b2c4dd2
JP		 1092 skb->pkt_type != PACKET_BROADCAST &&
1093 skb->pkt_type != PACKET_MULTICAST)
5b2c4dd2 1094 return false;
5b2c4dd2
JP		 1095 return true;
1096 }
1097 return false;
1098}
1099
8a4eb573 1100static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
5b2c4dd2 1101{
8a4eb573 1102 struct sk_buff *skb = *pskb;
f1c1775a 1103 struct slave *slave;
0bd80dad 1104 struct bonding *bond;
de063b70
ED		 1105 int (*recv_probe)(const struct sk_buff *, struct bonding *,
1106 struct slave *);
13a8e0c8 1107 int ret = RX_HANDLER_ANOTHER;
5b2c4dd2 1108
8a4eb573
JP		 1109 skb = skb_share_check(skb, GFP_ATOMIC);
1110 if (unlikely(!skb))
1111 return RX_HANDLER_CONSUMED;
1112
1113 *pskb = skb;
5b2c4dd2 1114
35d48903
JP		 1115 slave = bond_slave_get_rcu(skb->dev);
1116 bond = slave->bond;
0bd80dad
JP		 1117
1118 if (bond->params.arp_interval)
f1c1775a 1119 slave->dev->last_rx = jiffies;
5b2c4dd2 1120
4d97480b
MH		 1121 recv_probe = ACCESS_ONCE(bond->recv_probe);
1122 if (recv_probe) {
de063b70
ED		 1123 ret = recv_probe(skb, bond, slave);
1124 if (ret == RX_HANDLER_CONSUMED) {
1125 consume_skb(skb);
1126 return ret;
3aba891d
JP		 1127 }
1128 }
1129
0bd80dad 1130 if (bond_should_deliver_exact_match(skb, slave, bond)) {
8a4eb573 1131 return RX_HANDLER_EXACT;
5b2c4dd2
JP		 1132 }
1133
35d48903 1134 skb->dev = bond->dev;
5b2c4dd2 1135
0bd80dad 1136 if (bond->params.mode == BOND_MODE_ALB &&
35d48903 1137 bond->dev->priv_flags & IFF_BRIDGE_PORT &&
5b2c4dd2 1138 skb->pkt_type == PACKET_HOST) {
5b2c4dd2 1139
541ac7c9
CG		 1140 if (unlikely(skb_cow_head(skb,
1141 skb->data - skb_mac_header(skb)))) {
1142 kfree_skb(skb);
8a4eb573 1143 return RX_HANDLER_CONSUMED;
541ac7c9 1144 }
35d48903 1145 memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN);
5b2c4dd2
JP		 1146 }
1147
13a8e0c8 1148 return ret;
5b2c4dd2
JP		 1149}
1150
471cb5a3 1151static int bond_master_upper_dev_link(struct net_device *bond_dev,
1f718f0f
VF		 1152 struct net_device *slave_dev,
1153 struct slave *slave)
471cb5a3
JP		 1154{
1155 int err;
1156
1f718f0f 1157 err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave);
471cb5a3
JP		 1158 if (err)
1159 return err;
1160 slave_dev->flags |= IFF_SLAVE;
7f294054 1161 rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP		 1162 return 0;
1163}
1164
1165static void bond_upper_dev_unlink(struct net_device *bond_dev,
1166 struct net_device *slave_dev)
1167{
1168 netdev_upper_dev_unlink(slave_dev, bond_dev);
1169 slave_dev->flags &= ~IFF_SLAVE;
7f294054 1170 rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP		 1171}
1172
1da177e4 1173/* enslave device <slave> to bond device <master> */
a77b5325 1174int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1175{
454d7c9b 1176 struct bonding *bond = netdev_priv(bond_dev);
eb7cc59a 1177 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
c8c23903 1178 struct slave *new_slave = NULL, *prev_slave;
1da177e4
LT		 1179 struct sockaddr addr;
1180 int link_reporting;
8599b52e 1181 int res = 0, i;
1da177e4 1182
c772dde3
BH		 1183 if (!bond->params.use_carrier &&
1184 slave_dev->ethtool_ops->get_link == NULL &&
1185 slave_ops->ndo_do_ioctl == NULL) {
91565ebb
JP		 1186 pr_warn("%s: Warning: no link monitoring support for %s\n",
1187 bond_dev->name, slave_dev->name);
1da177e4
LT		 1188 }
1189
1da177e4
LT		 1190 /* already enslaved */
1191 if (slave_dev->flags & IFF_SLAVE) {
5a03cdb7 1192 pr_debug("Error, Device was already enslaved\n");
1da177e4
LT		 1193 return -EBUSY;
1194 }
1195
1196 /* vlan challenged mutual exclusion */
1197 /* no need to lock since we're protected by rtnl_lock */
1198 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 1199 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
55462cf3 1200 if (vlan_uses_dev(bond_dev)) {
a4aee5c8
JP		 1201 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1202 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1203 return -EPERM;
1204 } else {
91565ebb
JP		 1205 pr_warn("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1206 bond_dev->name, slave_dev->name,
1207 slave_dev->name, bond_dev->name);
1da177e4
LT		 1208 }
1209 } else {
5a03cdb7 1210 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4
LT		 1211 }
1212
217df670
JV		 1213 /*
1214 * Old ifenslave binaries are no longer supported. These can
3d632c3f 1215 * be identified with moderate accuracy by the state of the slave:
217df670
JV		 1216 * the current ifenslave will set the interface down prior to
1217 * enslaving it; the old ifenslave will not.
1218 */
1219 if ((slave_dev->flags & IFF_UP)) {
a4aee5c8 1220 pr_err("%s is up. This may be due to an out of date ifenslave.\n",
217df670
JV		 1221 slave_dev->name);
1222 res = -EPERM;
1223 goto err_undo_flags;
1224 }
1da177e4 1225
872254dd
MS		 1226 /* set bonding device ether type by slave - bonding netdevices are
1227 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1228 * there is a need to override some of the type dependent attribs/funcs.
1229 *
1230 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1231 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1232 */
0965a1f3 1233 if (!bond_has_slaves(bond)) {
e36b9d16 1234 if (bond_dev->type != slave_dev->type) {
e36b9d16 1235 pr_debug("%s: change device type from %d to %d\n",
a4aee5c8
JP		 1236 bond_dev->name,
1237 bond_dev->type, slave_dev->type);
75c78500 1238
b7bc2a5b
AW		 1239 res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE,
1240 bond_dev);
3ca5b404
JP		 1241 res = notifier_to_errno(res);
1242 if (res) {
1243 pr_err("%s: refused to change device type\n",
1244 bond_dev->name);
1245 res = -EBUSY;
1246 goto err_undo_flags;
1247 }
75c78500 1248
32a806c1 1249 /* Flush unicast and multicast addresses */
a748ee24 1250 dev_uc_flush(bond_dev);
22bedad3 1251 dev_mc_flush(bond_dev);
32a806c1 1252
e36b9d16
MS		 1253 if (slave_dev->type != ARPHRD_ETHER)
1254 bond_setup_by_slave(bond_dev, slave_dev);
550fd08c 1255 else {
e36b9d16 1256 ether_setup(bond_dev);
550fd08c
NH		 1257 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1258 }
75c78500 1259
b7bc2a5b
AW		 1260 call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE,
1261 bond_dev);
e36b9d16 1262 }
872254dd 1263 } else if (bond_dev->type != slave_dev->type) {
a4aee5c8
JP		 1264 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n",
1265 slave_dev->name,
1266 slave_dev->type, bond_dev->type);
1267 res = -EINVAL;
1268 goto err_undo_flags;
872254dd
MS		 1269 }
1270
eb7cc59a 1271 if (slave_ops->ndo_set_mac_address == NULL) {
0965a1f3 1272 if (!bond_has_slaves(bond)) {
00503b6f 1273 pr_warn("%s: Warning: The first slave device specified does not support setting the MAC address.\n",
1274 bond_dev->name);
1275 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
1276 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
1277 pr_warn("%s: Setting fail_over_mac to active for active-backup mode.\n",
1278 bond_dev->name);
1279 }
3915c1e8 1280 } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
a4aee5c8
JP		 1281 pr_err("%s: Error: The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active.\n",
1282 bond_dev->name);
2ab82852
MS		 1283 res = -EOPNOTSUPP;
1284 goto err_undo_flags;
1285 }
1da177e4
LT		 1286 }
1287
8d8fc29d
AW		 1288 call_netdevice_notifiers(NETDEV_JOIN, slave_dev);
1289
c20811a7
JP		 1290 /* If this is the first slave, then we need to set the master's hardware
1291 * address to be the same as the slave's. */
0965a1f3 1292 if (!bond_has_slaves(bond) &&
dec1e90e 1293 bond->dev->addr_assign_type == NET_ADDR_RANDOM)
409cc1f8 1294 bond_set_dev_addr(bond->dev, slave_dev);
c20811a7 1295
243cb4e5 1296 new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL);
1da177e4
LT		 1297 if (!new_slave) {
1298 res = -ENOMEM;
1299 goto err_undo_flags;
1300 }
bb1d9123
AG		 1301 /*
1302 * Set the new_slave's queue_id to be zero. Queue ID mapping
1303 * is set via sysfs or module option if desired.
1304 */
1305 new_slave->queue_id = 0;
1306
b15ba0fb
JP		 1307 /* Save slave's original mtu and then set it to match the bond */
1308 new_slave->original_mtu = slave_dev->mtu;
1309 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1310 if (res) {
1311 pr_debug("Error %d calling dev_set_mtu\n", res);
1312 goto err_free;
1313 }
1314
217df670
JV		 1315 /*
1316 * Save slave's original ("permanent") mac address for modes
1317 * that need it, and for restoring it upon release, and then
1318 * set it to the master's address
1319 */
1320 memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN);
1da177e4 1321
00503b6f 1322 if (!bond->params.fail_over_mac ||
1323 bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2ab82852
MS		 1324 /*
1325 * Set slave to master's mac address. The application already
1326 * set the master's mac address to that of the first slave
1327 */
1328 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1329 addr.sa_family = slave_dev->type;
1330 res = dev_set_mac_address(slave_dev, &addr);
1331 if (res) {
5a03cdb7 1332 pr_debug("Error %d calling set_mac_address\n", res);
b15ba0fb 1333 goto err_restore_mtu;
2ab82852 1334 }
217df670 1335 }
1da177e4 1336
217df670
JV		 1337 /* open the slave since the application closed it */
1338 res = dev_open(slave_dev);
1339 if (res) {
3d632c3f 1340 pr_debug("Opening slave %s failed\n", slave_dev->name);
1f718f0f 1341 goto err_restore_mac;
1da177e4
LT		 1342 }
1343
35d48903 1344 new_slave->bond = bond;
1da177e4 1345 new_slave->dev = slave_dev;
0b680e75 1346 slave_dev->priv_flags |= IFF_BONDING;
1da177e4 1347
58402054 1348 if (bond_is_lb(bond)) {
1da177e4
LT		 1349 /* bond_alb_init_slave() must be called before all other stages since
1350 * it might fail and we do not want to have to undo everything
1351 */
1352 res = bond_alb_init_slave(bond, new_slave);
3d632c3f 1353 if (res)
569f0c4d 1354 goto err_close;
1da177e4
LT		 1355 }
1356
303d1cbf
JV		 1357 /* If the mode USES_PRIMARY, then the following is handled by
1358 * bond_change_active_slave().
1da177e4
LT		 1359 */
1360 if (!USES_PRIMARY(bond->params.mode)) {
1361 /* set promiscuity level to new slave */
1362 if (bond_dev->flags & IFF_PROMISC) {
7e1a1ac1
WC		 1363 res = dev_set_promiscuity(slave_dev, 1);
1364 if (res)
1365 goto err_close;
1da177e4
LT		 1366 }
1367
1368 /* set allmulti level to new slave */
1369 if (bond_dev->flags & IFF_ALLMULTI) {
7e1a1ac1
WC		 1370 res = dev_set_allmulti(slave_dev, 1);
1371 if (res)
1372 goto err_close;
1da177e4
LT		 1373 }
1374
b9e40857 1375 netif_addr_lock_bh(bond_dev);
303d1cbf
JV		 1376
1377 dev_mc_sync_multiple(slave_dev, bond_dev);
1378 dev_uc_sync_multiple(slave_dev, bond_dev);
1379
b9e40857 1380 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1381 }
1382
1383 if (bond->params.mode == BOND_MODE_8023AD) {
1384 /* add lacpdu mc addr to mc list */
1385 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1386
22bedad3 1387 dev_mc_add(slave_dev, lacpdu_multicast);
1da177e4
LT		 1388 }
1389
b8e2fde4
WY		 1390 res = vlan_vids_add_by_dev(slave_dev, bond_dev);
1391 if (res) {
1ff412ad 1392 pr_err("%s: Error: Couldn't add bond vlan ids to %s\n",
1393 bond_dev->name, slave_dev->name);
1394 goto err_close;
1395 }
1da177e4 1396
c8c23903 1397 prev_slave = bond_last_slave(bond);
1da177e4
LT		 1398
1399 new_slave->delay = 0;
1400 new_slave->link_failure_count = 0;
1401
876254ae
VF		 1402 bond_update_speed_duplex(new_slave);
1403
f31c7937
MK		 1404 new_slave->last_arp_rx = jiffies -
1405 (msecs_to_jiffies(bond->params.arp_interval) + 1);
8599b52e
VF		 1406 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
1407 new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx;
f5b2b966 1408
1da177e4
LT		 1409 if (bond->params.miimon && !bond->params.use_carrier) {
1410 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1411
1412 if ((link_reporting == -1) && !bond->params.arp_interval) {
1413 /*
1414 * miimon is set but a bonded network driver
1415 * does not support ETHTOOL/MII and
1416 * arp_interval is not set. Note: if
1417 * use_carrier is enabled, we will never go
1418 * here (because netif_carrier is always
1419 * supported); thus, we don't need to change
1420 * the messages for netif_carrier.
1421 */
91565ebb
JP		 1422 pr_warn("%s: Warning: MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n",
1423 bond_dev->name, slave_dev->name);
1da177e4
LT		 1424 } else if (link_reporting == -1) {
1425 /* unable get link status using mii/ethtool */
91565ebb
JP		 1426 pr_warn("%s: Warning: can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface\n",
1427 bond_dev->name, slave_dev->name);
1da177e4
LT		 1428 }
1429 }
1430
1431 /* check for initial state */
f31c7937
MK		 1432 if (bond->params.miimon) {
1433 if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) {
1434 if (bond->params.updelay) {
1435 new_slave->link = BOND_LINK_BACK;
1436 new_slave->delay = bond->params.updelay;
1437 } else {
1438 new_slave->link = BOND_LINK_UP;
1439 }
1da177e4 1440 } else {
f31c7937 1441 new_slave->link = BOND_LINK_DOWN;
1da177e4 1442 }
f31c7937
MK		 1443 } else if (bond->params.arp_interval) {
1444 new_slave->link = (netif_carrier_ok(slave_dev) ?
1445 BOND_LINK_UP : BOND_LINK_DOWN);
1da177e4 1446 } else {
f31c7937 1447 new_slave->link = BOND_LINK_UP;
1da177e4
LT		 1448 }
1449
f31c7937
MK		 1450 if (new_slave->link != BOND_LINK_DOWN)
1451 new_slave->jiffies = jiffies;
1452 pr_debug("Initial state of slave_dev is BOND_LINK_%s\n",
1453 new_slave->link == BOND_LINK_DOWN ? "DOWN" :
1454 (new_slave->link == BOND_LINK_UP ? "UP" : "BACK"));
1455
1da177e4
LT		 1456 if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) {
1457 /* if there is a primary slave, remember it */
a549952a 1458 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1da177e4 1459 bond->primary_slave = new_slave;
a549952a
JP		 1460 bond->force_primary = true;
1461 }
1da177e4
LT		 1462 }
1463
1464 switch (bond->params.mode) {
1465 case BOND_MODE_ACTIVEBACKUP:
8a8e447b 1466 bond_set_slave_inactive_flags(new_slave);
1da177e4
LT		 1467 break;
1468 case BOND_MODE_8023AD:
1469 /* in 802.3ad mode, the internal mechanism
1470 * will activate the slaves in the selected
1471 * aggregator
1472 */
1473 bond_set_slave_inactive_flags(new_slave);
1474 /* if this is the first slave */
23c147e0 1475 if (!prev_slave) {
1da177e4
LT		 1476 SLAVE_AD_INFO(new_slave).id = 1;
1477 /* Initialize AD with the number of times that the AD timer is called in 1 second
1478 * can be called only after the mac address of the bond is set
1479 */
56d00c67 1480 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL);
1da177e4
LT		 1481 } else {
1482 SLAVE_AD_INFO(new_slave).id =
dec1e90e 1483 SLAVE_AD_INFO(prev_slave).id + 1;
1da177e4
LT		 1484 }
1485
1486 bond_3ad_bind_slave(new_slave);
1487 break;
1488 case BOND_MODE_TLB:
1489 case BOND_MODE_ALB:
e30bc066 1490 bond_set_active_slave(new_slave);
059fe7a5 1491 bond_set_slave_inactive_flags(new_slave);
1da177e4
LT		 1492 break;
1493 default:
5a03cdb7 1494 pr_debug("This slave is always active in trunk mode\n");
1da177e4
LT		 1495
1496 /* always active in trunk mode */
e30bc066 1497 bond_set_active_slave(new_slave);
1da177e4
LT		 1498
1499 /* In trunking mode there is little meaning to curr_active_slave
1500 * anyway (it holds no special properties of the bond device),
1501 * so we can change it without calling change_active_interface()
1502 */
4e591b93 1503 if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP)
278b2083 1504 rcu_assign_pointer(bond->curr_active_slave, new_slave);
3d632c3f 1505
1da177e4
LT		 1506 break;
1507 } /* switch(bond_mode) */
1508
f6dc31a8 1509#ifdef CONFIG_NET_POLL_CONTROLLER
10eccb46 1510 slave_dev->npinfo = bond->dev->npinfo;
8a8efa22
AW		 1511 if (slave_dev->npinfo) {
1512 if (slave_enable_netpoll(new_slave)) {
8a8efa22
AW		 1513 pr_info("Error, %s: master_dev is using netpoll, "
1514 "but new slave device does not support netpoll.\n",
1515 bond_dev->name);
1516 res = -EBUSY;
f7d9821a 1517 goto err_detach;
8a8efa22 1518 }
f6dc31a8
WC		 1519 }
1520#endif
8a8efa22 1521
35d48903
JP		 1522 res = netdev_rx_handler_register(slave_dev, bond_handle_frame,
1523 new_slave);
1524 if (res) {
1525 pr_debug("Error %d calling netdev_rx_handler_register\n", res);
5831d66e 1526 goto err_detach;
35d48903
JP		 1527 }
1528
1f718f0f
VF		 1529 res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave);
1530 if (res) {
1531 pr_debug("Error %d calling bond_master_upper_dev_link\n", res);
1532 goto err_unregister;
1533 }
1534
07699f9a 1535 res = bond_sysfs_slave_add(new_slave);
1536 if (res) {
1537 pr_debug("Error %d calling bond_sysfs_slave_add\n", res);
1538 goto err_upper_unlink;
1539 }
1540
5378c2e6
VF		 1541 bond->slave_cnt++;
1542 bond_compute_features(bond);
1543 bond_set_carrier(bond);
1544
1545 if (USES_PRIMARY(bond->params.mode)) {
5378c2e6
VF		 1546 write_lock_bh(&bond->curr_slave_lock);
1547 bond_select_active_slave(bond);
1548 write_unlock_bh(&bond->curr_slave_lock);
5378c2e6 1549 }
1f718f0f 1550
a4aee5c8
JP		 1551 pr_info("%s: enslaving %s as a%s interface with a%s link.\n",
1552 bond_dev->name, slave_dev->name,
e30bc066 1553 bond_is_active_slave(new_slave) ? "n active" : " backup",
a4aee5c8 1554 new_slave->link != BOND_LINK_DOWN ? "n up" : " down");
1da177e4
LT		 1555
1556 /* enslave is successful */
1557 return 0;
1558
1559/* Undo stages on error */
07699f9a 1560err_upper_unlink:
1561 bond_upper_dev_unlink(bond_dev, slave_dev);
1562
1f718f0f
VF		 1563err_unregister:
1564 netdev_rx_handler_unregister(slave_dev);
1565
f7d9821a 1566err_detach:
303d1cbf
JV		 1567 if (!USES_PRIMARY(bond->params.mode))
1568 bond_hw_addr_flush(bond_dev, slave_dev);
1569
1ff412ad 1570 vlan_vids_del_by_dev(slave_dev, bond_dev);
3c5913b5 1571 if (bond->primary_slave == new_slave)
1572 bond->primary_slave = NULL;
3c5913b5 1573 if (bond->curr_active_slave == new_slave) {
3c5913b5 1574 write_lock_bh(&bond->curr_slave_lock);
c8517035 1575 bond_change_active_slave(bond, NULL);
3c5913b5 1576 bond_select_active_slave(bond);
1577 write_unlock_bh(&bond->curr_slave_lock);
3c5913b5 1578 }
fc7a72ac 1579 slave_disable_netpoll(new_slave);
f7d9821a 1580
1da177e4 1581err_close:
b6a5a7b9 1582 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4
LT		 1583 dev_close(slave_dev);
1584
1585err_restore_mac:
00503b6f 1586 if (!bond->params.fail_over_mac ||
1587 bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
3915c1e8
JV		 1588 /* XXX TODO - fom follow mode needs to change master's
1589 * MAC if this slave's MAC is in use by the bond, or at
1590 * least print a warning.
1591 */
2ab82852
MS		 1592 memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN);
1593 addr.sa_family = slave_dev->type;
1594 dev_set_mac_address(slave_dev, &addr);
1595 }
1da177e4 1596
b15ba0fb
JP		 1597err_restore_mtu:
1598 dev_set_mtu(slave_dev, new_slave->original_mtu);
1599
1da177e4
LT		 1600err_free:
1601 kfree(new_slave);
1602
1603err_undo_flags:
b8fad459 1604 /* Enslave of first slave has failed and we need to fix master's mac */
0965a1f3 1605 if (!bond_has_slaves(bond) &&
844223ab 1606 ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr))
b8fad459 1607 eth_hw_addr_random(bond_dev);
3d632c3f 1608
1da177e4
LT		 1609 return res;
1610}
1611
1612/*
1613 * Try to release the slave device <slave> from the bond device <master>
1614 * It is legal to access curr_active_slave without a lock because all the function
0896341a 1615 * is write-locked. If "all" is true it means that the function is being called
1616 * while destroying a bond interface and all slaves are being released.
1da177e4
LT		 1617 *
1618 * The rules for slave state should be:
1619 * for Active/Backup:
1620 * Active stays on all backups go down
1621 * for Bonded connections:
1622 * The first up interface should be left on and all others downed.
1623 */
0896341a 1624static int __bond_release_one(struct net_device *bond_dev,
1625 struct net_device *slave_dev,
1626 bool all)
1da177e4 1627{
454d7c9b 1628 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1629 struct slave *slave, *oldcurrent;
1630 struct sockaddr addr;
5a0068de 1631 int old_flags = bond_dev->flags;
c8f44aff 1632 netdev_features_t old_features = bond_dev->features;
1da177e4
LT		 1633
1634 /* slave is not a slave or master is not master of this slave */
1635 if (!(slave_dev->flags & IFF_SLAVE) ||
471cb5a3 1636 !netdev_has_upper_dev(slave_dev, bond_dev)) {
a4aee5c8 1637 pr_err("%s: Error: cannot release %s.\n",
1da177e4
LT		 1638 bond_dev->name, slave_dev->name);
1639 return -EINVAL;
1640 }
1641
e843fa50 1642 block_netpoll_tx();
1da177e4
LT		 1643
1644 slave = bond_get_slave_by_dev(bond, slave_dev);
1645 if (!slave) {
1646 /* not a slave of this bond */
a4aee5c8
JP		 1647 pr_info("%s: %s not enslaved\n",
1648 bond_dev->name, slave_dev->name);
e843fa50 1649 unblock_netpoll_tx();
1da177e4
LT		 1650 return -EINVAL;
1651 }
1652
5378c2e6
VF		 1653 /* release the slave from its bond */
1654 bond->slave_cnt--;
1655
07699f9a 1656 bond_sysfs_slave_del(slave);
1657
1f718f0f 1658 bond_upper_dev_unlink(bond_dev, slave_dev);
35d48903
JP		 1659 /* unregister rx_handler early so bond_handle_frame wouldn't be called
1660 * for this slave anymore.
1661 */
1662 netdev_rx_handler_unregister(slave_dev);
35d48903
JP		 1663 write_lock_bh(&bond->lock);
1664
1da177e4 1665 /* Inform AD package of unbinding of slave. */
be79bd04 1666 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 1667 bond_3ad_unbind_slave(slave);
be79bd04 1668
c8517035 1669 write_unlock_bh(&bond->lock);
1da177e4 1670
a4aee5c8
JP		 1671 pr_info("%s: releasing %s interface %s\n",
1672 bond_dev->name,
e30bc066 1673 bond_is_active_slave(slave) ? "active" : "backup",
a4aee5c8 1674 slave_dev->name);
1da177e4
LT		 1675
1676 oldcurrent = bond->curr_active_slave;
1677
1678 bond->current_arp_slave = NULL;
1679
00503b6f 1680 if (!all && (!bond->params.fail_over_mac ||
1681 bond->params.mode != BOND_MODE_ACTIVEBACKUP)) {
844223ab 1682 if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) &&
0965a1f3 1683 bond_has_slaves(bond))
dec1e90e 1684 pr_warn("%s: Warning: the permanent HWaddr of %s - %pM - is still in use by %s. Set the HWaddr of %s to a different address to avoid conflicts.\n",
1685 bond_dev->name, slave_dev->name,
1686 slave->perm_hwaddr,
1687 bond_dev->name, slave_dev->name);
1688 }
1689
3d632c3f 1690 if (bond->primary_slave == slave)
1da177e4 1691 bond->primary_slave = NULL;
1da177e4 1692
c8517035 1693 if (oldcurrent == slave) {
1694 write_lock_bh(&bond->curr_slave_lock);
1da177e4 1695 bond_change_active_slave(bond, NULL);
c8517035 1696 write_unlock_bh(&bond->curr_slave_lock);
1697 }
1da177e4 1698
58402054 1699 if (bond_is_lb(bond)) {
1da177e4
LT		 1700 /* Must be called only after the slave has been
1701 * detached from the list and the curr_active_slave
1702 * has been cleared (if our_slave == old_current),
1703 * but before a new active slave is selected.
1704 */
1705 bond_alb_deinit_slave(bond, slave);
1706 }
1707
0896341a 1708 if (all) {
36708b89 1709 RCU_INIT_POINTER(bond->curr_active_slave, NULL);
0896341a 1710 } else if (oldcurrent == slave) {
059fe7a5
JV		 1711 /*
1712 * Note that we hold RTNL over this sequence, so there
1713 * is no concern that another slave add/remove event
1714 * will interfere.
1715 */
059fe7a5
JV		 1716 write_lock_bh(&bond->curr_slave_lock);
1717
1da177e4
LT		 1718 bond_select_active_slave(bond);
1719
059fe7a5 1720 write_unlock_bh(&bond->curr_slave_lock);
059fe7a5
JV		 1721 }
1722
0965a1f3 1723 if (!bond_has_slaves(bond)) {
ff59c456 1724 bond_set_carrier(bond);
409cc1f8 1725 eth_hw_addr_random(bond_dev);
1da177e4 1726
a59d3d21 1727 if (vlan_uses_dev(bond_dev)) {
91565ebb
JP		 1728 pr_warn("%s: Warning: clearing HW address of %s while it still has VLANs\n",
1729 bond_dev->name, bond_dev->name);
1730 pr_warn("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs\n",
1731 bond_dev->name);
1da177e4 1732 }
1da177e4
LT		 1733 }
1734
e843fa50 1735 unblock_netpoll_tx();
278b2083 1736 synchronize_rcu();
1da177e4 1737
0965a1f3 1738 if (!bond_has_slaves(bond)) {
2af73d4b 1739 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev);
80028ea1
VF		 1740 call_netdevice_notifiers(NETDEV_RELEASE, bond->dev);
1741 }
2af73d4b 1742
b2a103e6
MM		 1743 bond_compute_features(bond);
1744 if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1745 (old_features & NETIF_F_VLAN_CHALLENGED))
1746 pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n",
1747 bond_dev->name, slave_dev->name, bond_dev->name);
1748
b76cdba9 1749 /* must do this from outside any spinlocks */
1ff412ad 1750 vlan_vids_del_by_dev(slave_dev, bond_dev);
1da177e4 1751
303d1cbf
JV		 1752 /* If the mode USES_PRIMARY, then this cases was handled above by
1753 * bond_change_active_slave(..., NULL)
1da177e4
LT		 1754 */
1755 if (!USES_PRIMARY(bond->params.mode)) {
5a0068de
NH		 1756 /* unset promiscuity level from slave
1757 * NOTE: The NETDEV_CHANGEADDR call above may change the value
1758 * of the IFF_PROMISC flag in the bond_dev, but we need the
1759 * value of that flag before that change, as that was the value
1760 * when this slave was attached, so we cache at the start of the
1761 * function and use it here. Same goes for ALLMULTI below
1762 */
1763 if (old_flags & IFF_PROMISC)
1da177e4 1764 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 1765
1766 /* unset allmulti level from slave */
5a0068de 1767 if (old_flags & IFF_ALLMULTI)
1da177e4 1768 dev_set_allmulti(slave_dev, -1);
1da177e4 1769
303d1cbf 1770 bond_hw_addr_flush(bond_dev, slave_dev);
1da177e4
LT		 1771 }
1772
8a8efa22 1773 slave_disable_netpoll(slave);
f6dc31a8 1774
1da177e4
LT		 1775 /* close slave before restoring its mac address */
1776 dev_close(slave_dev);
1777
00503b6f 1778 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE ||
1779 bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2ab82852
MS		 1780 /* restore original ("permanent") mac address */
1781 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
1782 addr.sa_family = slave_dev->type;
1783 dev_set_mac_address(slave_dev, &addr);
1784 }
1da177e4 1785
b15ba0fb
JP		 1786 dev_set_mtu(slave_dev, slave->original_mtu);
1787
2d7011ca 1788 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4
LT		 1789
1790 kfree(slave);
1791
1792 return 0; /* deletion OK */
1793}
1794
0896341a 1795/* A wrapper used because of ndo_del_link */
1796int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1797{
1798 return __bond_release_one(bond_dev, slave_dev, false);
1799}
1800
d90a162a 1801/*
dadaa10b 1802* First release a slave and then destroy the bond if no more slaves are left.
d90a162a
MS		 1803* Must be under rtnl_lock when this function is called.
1804*/
26d8ee75 1805static int bond_release_and_destroy(struct net_device *bond_dev,
1806 struct net_device *slave_dev)
d90a162a 1807{
454d7c9b 1808 struct bonding *bond = netdev_priv(bond_dev);
d90a162a
MS		 1809 int ret;
1810
1811 ret = bond_release(bond_dev, slave_dev);
0965a1f3 1812 if (ret == 0 && !bond_has_slaves(bond)) {
8a8efa22 1813 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
a4aee5c8
JP		 1814 pr_info("%s: destroying bond %s.\n",
1815 bond_dev->name, bond_dev->name);
9e71626c 1816 unregister_netdevice(bond_dev);
d90a162a
MS		 1817 }
1818 return ret;
1819}
1820
1da177e4
LT		 1821static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
1822{
454d7c9b 1823 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1824
1825 info->bond_mode = bond->params.mode;
1826 info->miimon = bond->params.miimon;
1827
6603a6f2 1828 read_lock(&bond->lock);
1da177e4 1829 info->num_slaves = bond->slave_cnt;
6603a6f2 1830 read_unlock(&bond->lock);
1da177e4
LT		 1831
1832 return 0;
1833}
1834
1835static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
1836{
454d7c9b 1837 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 1838 struct list_head *iter;
dec1e90e 1839 int i = 0, res = -ENODEV;
1da177e4 1840 struct slave *slave;
1da177e4 1841
6603a6f2 1842 read_lock(&bond->lock);
9caff1e7 1843 bond_for_each_slave(bond, slave, iter) {
dec1e90e 1844 if (i++ == (int)info->slave_id) {
689c96cc
ED		 1845 res = 0;
1846 strcpy(info->slave_name, slave->dev->name);
1847 info->link = slave->link;
e30bc066 1848 info->state = bond_slave_state(slave);
689c96cc 1849 info->link_failure_count = slave->link_failure_count;
1da177e4
LT		 1850 break;
1851 }
1852 }
6603a6f2 1853 read_unlock(&bond->lock);
1da177e4 1854
689c96cc 1855 return res;
1da177e4
LT		 1856}
1857
1858/*-------------------------------- Monitoring -------------------------------*/
1859
1da177e4 1860
f0c76d61
JV		 1861static int bond_miimon_inspect(struct bonding *bond)
1862{
dec1e90e 1863 int link_state, commit = 0;
9caff1e7 1864 struct list_head *iter;
f0c76d61 1865 struct slave *slave;
41f89100
JP		 1866 bool ignore_updelay;
1867
1868 ignore_updelay = !bond->curr_active_slave ? true : false;
1da177e4 1869
4cb4f97b 1870 bond_for_each_slave_rcu(bond, slave, iter) {
f0c76d61 1871 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 1872
f0c76d61 1873 link_state = bond_check_dev_link(bond, slave->dev, 0);
1da177e4
LT		 1874
1875 switch (slave->link) {
f0c76d61
JV		 1876 case BOND_LINK_UP:
1877 if (link_state)
1878 continue;
1da177e4 1879
f0c76d61
JV		 1880 slave->link = BOND_LINK_FAIL;
1881 slave->delay = bond->params.downdelay;
1882 if (slave->delay) {
a4aee5c8
JP		 1883 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n",
1884 bond->dev->name,
1885 (bond->params.mode ==
1886 BOND_MODE_ACTIVEBACKUP) ?
e30bc066 1887 (bond_is_active_slave(slave) ?
a4aee5c8
JP		 1888 "active " : "backup ") : "",
1889 slave->dev->name,
1890 bond->params.downdelay * bond->params.miimon);
1da177e4 1891 }
f0c76d61
JV		 1892 /*FALLTHRU*/
1893 case BOND_LINK_FAIL:
1894 if (link_state) {
1895 /*
1896 * recovered before downdelay expired
1897 */
1898 slave->link = BOND_LINK_UP;
1da177e4 1899 slave->jiffies = jiffies;
a4aee5c8
JP		 1900 pr_info("%s: link status up again after %d ms for interface %s.\n",
1901 bond->dev->name,
1902 (bond->params.downdelay - slave->delay) *
1903 bond->params.miimon,
1904 slave->dev->name);
f0c76d61 1905 continue;
1da177e4 1906 }
f0c76d61
JV		 1907
1908 if (slave->delay <= 0) {
1909 slave->new_link = BOND_LINK_DOWN;
1910 commit++;
1911 continue;
1da177e4 1912 }
1da177e4 1913
f0c76d61
JV		 1914 slave->delay--;
1915 break;
1916
1917 case BOND_LINK_DOWN:
1918 if (!link_state)
1919 continue;
1920
1921 slave->link = BOND_LINK_BACK;
1922 slave->delay = bond->params.updelay;
1923
1924 if (slave->delay) {
a4aee5c8
JP		 1925 pr_info("%s: link status up for interface %s, enabling it in %d ms.\n",
1926 bond->dev->name, slave->dev->name,
1927 ignore_updelay ? 0 :
1928 bond->params.updelay *
1929 bond->params.miimon);
f0c76d61
JV		 1930 }
1931 /*FALLTHRU*/
1932 case BOND_LINK_BACK:
1933 if (!link_state) {
1934 slave->link = BOND_LINK_DOWN;
a4aee5c8
JP		 1935 pr_info("%s: link status down again after %d ms for interface %s.\n",
1936 bond->dev->name,
1937 (bond->params.updelay - slave->delay) *
1938 bond->params.miimon,
1939 slave->dev->name);
f0c76d61
JV		 1940
1941 continue;
1942 }
1943
41f89100
JP		 1944 if (ignore_updelay)
1945 slave->delay = 0;
1946
f0c76d61
JV		 1947 if (slave->delay <= 0) {
1948 slave->new_link = BOND_LINK_UP;
1949 commit++;
41f89100 1950 ignore_updelay = false;
f0c76d61 1951 continue;
1da177e4 1952 }
f0c76d61
JV		 1953
1954 slave->delay--;
1da177e4 1955 break;
f0c76d61
JV		 1956 }
1957 }
1da177e4 1958
f0c76d61
JV		 1959 return commit;
1960}
1da177e4 1961
f0c76d61
JV		 1962static void bond_miimon_commit(struct bonding *bond)
1963{
9caff1e7 1964 struct list_head *iter;
f0c76d61 1965 struct slave *slave;
f0c76d61 1966
9caff1e7 1967 bond_for_each_slave(bond, slave, iter) {
f0c76d61
JV		 1968 switch (slave->new_link) {
1969 case BOND_LINK_NOCHANGE:
1970 continue;
1da177e4 1971
f0c76d61
JV		 1972 case BOND_LINK_UP:
1973 slave->link = BOND_LINK_UP;
1974 slave->jiffies = jiffies;
1975
1976 if (bond->params.mode == BOND_MODE_8023AD) {
1977 /* prevent it from being the active one */
e30bc066 1978 bond_set_backup_slave(slave);
f0c76d61
JV		 1979 } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
1980 /* make it immediately active */
e30bc066 1981 bond_set_active_slave(slave);
f0c76d61
JV		 1982 } else if (slave != bond->primary_slave) {
1983 /* prevent it from being the active one */
e30bc066 1984 bond_set_backup_slave(slave);
1da177e4 1985 }
1da177e4 1986
5d30530e 1987 pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n",
700c2a77 1988 bond->dev->name, slave->dev->name,
db4e9b2b
NA		 1989 slave->speed == SPEED_UNKNOWN ? 0 : slave->speed,
1990 slave->duplex ? "full" : "half");
1da177e4 1991
f0c76d61
JV		 1992 /* notify ad that the link status has changed */
1993 if (bond->params.mode == BOND_MODE_8023AD)
1994 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
059fe7a5 1995
58402054 1996 if (bond_is_lb(bond))
f0c76d61
JV		 1997 bond_alb_handle_link_change(bond, slave,
1998 BOND_LINK_UP);
1da177e4 1999
f0c76d61
JV		 2000 if (!bond->curr_active_slave ||
2001 (slave == bond->primary_slave))
2002 goto do_failover;
1da177e4 2003
f0c76d61 2004 continue;
059fe7a5 2005
f0c76d61 2006 case BOND_LINK_DOWN:
fba4acda
JV		 2007 if (slave->link_failure_count < UINT_MAX)
2008 slave->link_failure_count++;
2009
f0c76d61 2010 slave->link = BOND_LINK_DOWN;
1da177e4 2011
f0c76d61
JV		 2012 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP ||
2013 bond->params.mode == BOND_MODE_8023AD)
2014 bond_set_slave_inactive_flags(slave);
2015
a4aee5c8
JP		 2016 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2017 bond->dev->name, slave->dev->name);
f0c76d61
JV		 2018
2019 if (bond->params.mode == BOND_MODE_8023AD)
2020 bond_3ad_handle_link_change(slave,
2021 BOND_LINK_DOWN);
2022
ae63e808 2023 if (bond_is_lb(bond))
f0c76d61
JV		 2024 bond_alb_handle_link_change(bond, slave,
2025 BOND_LINK_DOWN);
2026
2027 if (slave == bond->curr_active_slave)
2028 goto do_failover;
2029
2030 continue;
2031
2032 default:
a4aee5c8 2033 pr_err("%s: invalid new link %d on slave %s\n",
f0c76d61
JV		 2034 bond->dev->name, slave->new_link,
2035 slave->dev->name);
2036 slave->new_link = BOND_LINK_NOCHANGE;
2037
2038 continue;
2039 }
2040
2041do_failover:
2042 ASSERT_RTNL();
e843fa50 2043 block_netpoll_tx();
f0c76d61
JV		 2044 write_lock_bh(&bond->curr_slave_lock);
2045 bond_select_active_slave(bond);
2046 write_unlock_bh(&bond->curr_slave_lock);
e843fa50 2047 unblock_netpoll_tx();
f0c76d61
JV		 2048 }
2049
2050 bond_set_carrier(bond);
1da177e4
LT		 2051}
2052
0b0eef66
JV		 2053/*
2054 * bond_mii_monitor
2055 *
2056 * Really a wrapper that splits the mii monitor into two phases: an
f0c76d61
JV		 2057 * inspection, then (if inspection indicates something needs to be done)
2058 * an acquisition of appropriate locks followed by a commit phase to
2059 * implement whatever link state changes are indicated.
0b0eef66 2060 */
6da67d26 2061static void bond_mii_monitor(struct work_struct *work)
0b0eef66
JV		 2062{
2063 struct bonding *bond = container_of(work, struct bonding,
2064 mii_work.work);
ad246c99 2065 bool should_notify_peers = false;
1f2cd845 2066 unsigned long delay;
0b0eef66 2067
1f2cd845
DM		 2068 delay = msecs_to_jiffies(bond->params.miimon);
2069
2070 if (!bond_has_slaves(bond))
f0c76d61 2071 goto re_arm;
b59f9f74 2072
4cb4f97b 2073 rcu_read_lock();
2074
ad246c99
BH		 2075 should_notify_peers = bond_should_notify_peers(bond);
2076
1f2cd845 2077 if (bond_miimon_inspect(bond)) {
4cb4f97b 2078 rcu_read_unlock();
f0c76d61 2079
1f2cd845
DM		 2080 /* Race avoidance with bond_close cancel of workqueue */
2081 if (!rtnl_trylock()) {
1f2cd845
DM		 2082 delay = 1;
2083 should_notify_peers = false;
2084 goto re_arm;
2085 }
6b6c5261 2086
1f2cd845
DM		 2087 bond_miimon_commit(bond);
2088
1f2cd845 2089 rtnl_unlock(); /* might sleep, hold no other locks */
4cb4f97b 2090 } else
2091 rcu_read_unlock();
0b0eef66 2092
f0c76d61 2093re_arm:
e6d265e8 2094 if (bond->params.miimon)
1f2cd845
DM		 2095 queue_delayed_work(bond->wq, &bond->mii_work, delay);
2096
1f2cd845
DM		 2097 if (should_notify_peers) {
2098 if (!rtnl_trylock())
2099 return;
2100 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2101 rtnl_unlock();
2102 }
0b0eef66 2103}
c3ade5ca 2104
50223ce4 2105static bool bond_has_this_ip(struct bonding *bond, __be32 ip)
f5b2b966 2106{
50223ce4
VF		 2107 struct net_device *upper;
2108 struct list_head *iter;
2109 bool ret = false;
f5b2b966 2110
eaddcd76 2111 if (ip == bond_confirm_addr(bond->dev, 0, ip))
50223ce4 2112 return true;
f5b2b966 2113
50223ce4 2114 rcu_read_lock();
2f268f12 2115 netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) {
50223ce4
VF		 2116 if (ip == bond_confirm_addr(upper, 0, ip)) {
2117 ret = true;
2118 break;
2119 }
f5b2b966 2120 }
50223ce4 2121 rcu_read_unlock();
f5b2b966 2122
50223ce4 2123 return ret;
f5b2b966
JV		 2124}
2125
c3ade5ca
JV		 2126/*
2127 * We go to the (large) trouble of VLAN tagging ARP frames because
2128 * switches in VLAN mode (especially if ports are configured as
2129 * "native" to a VLAN) might not pass non-tagged frames.
2130 */
d3bb52b0 2131static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id)
c3ade5ca
JV		 2132{
2133 struct sk_buff *skb;
2134
acca2674 2135 pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op,
2136 slave_dev->name, &dest_ip, &src_ip, vlan_id);
3d632c3f 2137
c3ade5ca
JV		 2138 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2139 NULL, slave_dev->dev_addr, NULL);
2140
2141 if (!skb) {
a4aee5c8 2142 pr_err("ARP packet allocation failed\n");
c3ade5ca
JV		 2143 return;
2144 }
2145 if (vlan_id) {
1fd9b1fc 2146 skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id);
c3ade5ca 2147 if (!skb) {
a4aee5c8 2148 pr_err("failed to insert VLAN tag\n");
c3ade5ca
JV		 2149 return;
2150 }
2151 }
2152 arp_xmit(skb);
2153}
2154
2155
1da177e4
LT		 2156static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2157{
27bc11e6
VF		 2158 struct net_device *upper, *vlan_upper;
2159 struct list_head *iter, *vlan_iter;
c3ade5ca 2160 struct rtable *rt;
27bc11e6
VF		 2161 __be32 *targets = bond->params.arp_targets, addr;
2162 int i, vlan_id;
1da177e4 2163
27bc11e6 2164 for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) {
acca2674 2165 pr_debug("basa: target %pI4\n", &targets[i]);
c3ade5ca 2166
27bc11e6 2167 /* Find out through which dev should the packet go */
78fbfd8a
DM		 2168 rt = ip_route_output(dev_net(bond->dev), targets[i], 0,
2169 RTO_ONLINK, 0);
b23dd4fe 2170 if (IS_ERR(rt)) {
3e32582f
VF		 2171 pr_debug("%s: no route to arp_ip_target %pI4\n",
2172 bond->dev->name, &targets[i]);
c3ade5ca
JV		 2173 continue;
2174 }
2175
c3ade5ca 2176 vlan_id = 0;
27bc11e6
VF		 2177
2178 /* bond device itself */
2179 if (rt->dst.dev == bond->dev)
2180 goto found;
2181
2182 rcu_read_lock();
2183 /* first we search only for vlan devices. for every vlan
2184 * found we verify its upper dev list, searching for the
2185 * rt->dst.dev. If found we save the tag of the vlan and
2186 * proceed to send the packet.
2187 *
2188 * TODO: QinQ?
2189 */
2f268f12
VF		 2190 netdev_for_each_all_upper_dev_rcu(bond->dev, vlan_upper,
2191 vlan_iter) {
27bc11e6
VF		 2192 if (!is_vlan_dev(vlan_upper))
2193 continue;
2f268f12
VF		 2194 netdev_for_each_all_upper_dev_rcu(vlan_upper, upper,
2195 iter) {
27bc11e6
VF		 2196 if (upper == rt->dst.dev) {
2197 vlan_id = vlan_dev_vlan_id(vlan_upper);
2198 rcu_read_unlock();
2199 goto found;
2200 }
c3ade5ca
JV		 2201 }
2202 }
2203
27bc11e6
VF		 2204 /* if the device we're looking for is not on top of any of
2205 * our upper vlans, then just search for any dev that
2206 * matches, and in case it's a vlan - save the id
2207 */
2f268f12 2208 netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) {
27bc11e6
VF		 2209 if (upper == rt->dst.dev) {
2210 /* if it's a vlan - get its VID */
2211 if (is_vlan_dev(upper))
2212 vlan_id = vlan_dev_vlan_id(upper);
2213
2214 rcu_read_unlock();
2215 goto found;
2216 }
c3ade5ca 2217 }
27bc11e6 2218 rcu_read_unlock();
c3ade5ca 2219
27bc11e6 2220 /* Not our device - skip */
3e32582f
VF		 2221 pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2222 bond->dev->name, &targets[i],
2223 rt->dst.dev ? rt->dst.dev->name : "NULL");
2224
ed4b9f80 2225 ip_rt_put(rt);
27bc11e6
VF		 2226 continue;
2227
2228found:
2229 addr = bond_confirm_addr(rt->dst.dev, targets[i], 0);
2230 ip_rt_put(rt);
2231 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2232 addr, vlan_id);
c3ade5ca
JV		 2233 }
2234}
2235
d3bb52b0 2236static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
f5b2b966 2237{
8599b52e
VF		 2238 int i;
2239
87a7b84b
VF		 2240 if (!sip || !bond_has_this_ip(bond, tip)) {
2241 pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip);
2242 return;
2243 }
f5b2b966 2244
8599b52e
VF		 2245 i = bond_get_targets_ip(bond->params.arp_targets, sip);
2246 if (i == -1) {
87a7b84b
VF		 2247 pr_debug("bva: sip %pI4 not found in targets\n", &sip);
2248 return;
f5b2b966 2249 }
87a7b84b 2250 slave->last_arp_rx = jiffies;
8599b52e 2251 slave->target_last_arp_rx[i] = jiffies;
f5b2b966
JV		 2252}
2253
5bb9e0b5 2254int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
2255 struct slave *slave)
f5b2b966 2256{
de063b70 2257 struct arphdr *arp = (struct arphdr *)skb->data;
f5b2b966 2258 unsigned char *arp_ptr;
d3bb52b0 2259 __be32 sip, tip;
de063b70 2260 int alen;
f5b2b966 2261
3aba891d 2262 if (skb->protocol != __cpu_to_be16(ETH_P_ARP))
13a8e0c8 2263 return RX_HANDLER_ANOTHER;
f5b2b966 2264
f5b2b966 2265 read_lock(&bond->lock);
2c146102
VF		 2266
2267 if (!slave_do_arp_validate(bond, slave))
2268 goto out_unlock;
2269
de063b70 2270 alen = arp_hdr_len(bond->dev);
f5b2b966 2271
3aba891d
JP		 2272 pr_debug("bond_arp_rcv: bond %s skb->dev %s\n",
2273 bond->dev->name, skb->dev->name);
f5b2b966 2274
de063b70
ED		 2275 if (alen > skb_headlen(skb)) {
2276 arp = kmalloc(alen, GFP_ATOMIC);
2277 if (!arp)
2278 goto out_unlock;
2279 if (skb_copy_bits(skb, 0, arp, alen) < 0)
2280 goto out_unlock;
2281 }
f5b2b966 2282
3aba891d 2283 if (arp->ar_hln != bond->dev->addr_len ||
f5b2b966
JV		 2284 skb->pkt_type == PACKET_OTHERHOST ||
2285 skb->pkt_type == PACKET_LOOPBACK ||
2286 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2287 arp->ar_pro != htons(ETH_P_IP) ||
2288 arp->ar_pln != 4)
2289 goto out_unlock;
2290
2291 arp_ptr = (unsigned char *)(arp + 1);
3aba891d 2292 arp_ptr += bond->dev->addr_len;
f5b2b966 2293 memcpy(&sip, arp_ptr, 4);
3aba891d 2294 arp_ptr += 4 + bond->dev->addr_len;
f5b2b966
JV		 2295 memcpy(&tip, arp_ptr, 4);
2296
5a03cdb7 2297 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
e30bc066 2298 bond->dev->name, slave->dev->name, bond_slave_state(slave),
a4aee5c8
JP		 2299 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2300 &sip, &tip);
f5b2b966
JV		 2301
2302 /*
2303 * Backup slaves won't see the ARP reply, but do come through
2304 * here for each ARP probe (so we swap the sip/tip to validate
2305 * the probe). In a "redundant switch, common router" type of
2306 * configuration, the ARP probe will (hopefully) travel from
2307 * the active, through one switch, the router, then the other
2308 * switch before reaching the backup.
aeea64ac
VF		 2309 *
2310 * We 'trust' the arp requests if there is an active slave and
2311 * it received valid arp reply(s) after it became active. This
2312 * is done to avoid endless looping when we can't reach the
2313 * arp_ip_target and fool ourselves with our own arp requests.
f5b2b966 2314 */
e30bc066 2315 if (bond_is_active_slave(slave))
f5b2b966 2316 bond_validate_arp(bond, slave, sip, tip);
aeea64ac
VF		 2317 else if (bond->curr_active_slave &&
2318 time_after(slave_last_rx(bond, bond->curr_active_slave),
2319 bond->curr_active_slave->jiffies))
f5b2b966
JV		 2320 bond_validate_arp(bond, slave, tip, sip);
2321
2322out_unlock:
2323 read_unlock(&bond->lock);
de063b70
ED		 2324 if (arp != (struct arphdr *)skb->data)
2325 kfree(arp);
13a8e0c8 2326 return RX_HANDLER_ANOTHER;
f5b2b966
JV		 2327}
2328
e7f63f1d
VF		 2329/* function to verify if we're in the arp_interval timeslice, returns true if
2330 * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval +
2331 * arp_interval/2) . the arp_interval/2 is needed for really fast networks.
2332 */
2333static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
2334 int mod)
2335{
2336 int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2337
2338 return time_in_range(jiffies,
2339 last_act - delta_in_ticks,
2340 last_act + mod * delta_in_ticks + delta_in_ticks/2);
2341}
2342
1da177e4
LT		 2343/*
2344 * this function is called regularly to monitor each slave's link
2345 * ensuring that traffic is being sent and received when arp monitoring
2346 * is used in load-balancing mode. if the adapter has been dormant, then an
2347 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2348 * arp monitoring in active backup mode.
2349 */
6da67d26 2350static void bond_loadbalance_arp_mon(struct work_struct *work)
1da177e4 2351{
1b76b316
JV		 2352 struct bonding *bond = container_of(work, struct bonding,
2353 arp_work.work);
1da177e4 2354 struct slave *slave, *oldcurrent;
9caff1e7 2355 struct list_head *iter;
6fde8f03 2356 int do_failover = 0, slave_state_changed = 0;
1da177e4 2357
1f2cd845 2358 if (!bond_has_slaves(bond))
1da177e4 2359 goto re_arm;
1da177e4 2360
2e52f4fe 2361 rcu_read_lock();
2362
2363 oldcurrent = ACCESS_ONCE(bond->curr_active_slave);
1da177e4
LT		 2364 /* see if any of the previous devices are up now (i.e. they have
2365 * xmt and rcv traffic). the curr_active_slave does not come into
2366 * the picture unless it is null. also, slave->jiffies is not needed
2367 * here because we send an arp on each slave and give a slave as
2368 * long as it needs to get the tx/rx within the delta.
2369 * TODO: what about up/down delay in arp mode? it wasn't here before
2370 * so it can wait
2371 */
2e52f4fe 2372 bond_for_each_slave_rcu(bond, slave, iter) {
cb32f2a0
JB		 2373 unsigned long trans_start = dev_trans_start(slave->dev);
2374
1da177e4 2375 if (slave->link != BOND_LINK_UP) {
e7f63f1d
VF		 2376 if (bond_time_in_interval(bond, trans_start, 1) &&
2377 bond_time_in_interval(bond, slave->dev->last_rx, 1)) {
1da177e4
LT		 2378
2379 slave->link = BOND_LINK_UP;
6fde8f03 2380 slave_state_changed = 1;
1da177e4
LT		 2381
2382 /* primary_slave has no meaning in round-robin
2383 * mode. the window of a slave being up and
2384 * curr_active_slave being null after enslaving
2385 * is closed.
2386 */
2387 if (!oldcurrent) {
a4aee5c8
JP		 2388 pr_info("%s: link status definitely up for interface %s, ",
2389 bond->dev->name,
2390 slave->dev->name);
1da177e4
LT		 2391 do_failover = 1;
2392 } else {
a4aee5c8
JP		 2393 pr_info("%s: interface %s is now up\n",
2394 bond->dev->name,
2395 slave->dev->name);
1da177e4
LT		 2396 }
2397 }
2398 } else {
2399 /* slave->link == BOND_LINK_UP */
2400
2401 /* not all switches will respond to an arp request
2402 * when the source ip is 0, so don't take the link down
2403 * if we don't know our ip yet
2404 */
e7f63f1d
VF		 2405 if (!bond_time_in_interval(bond, trans_start, 2) ||
2406 !bond_time_in_interval(bond, slave->dev->last_rx, 2)) {
1da177e4
LT		 2407
2408 slave->link = BOND_LINK_DOWN;
6fde8f03 2409 slave_state_changed = 1;
1da177e4 2410
3d632c3f 2411 if (slave->link_failure_count < UINT_MAX)
1da177e4 2412 slave->link_failure_count++;
1da177e4 2413
a4aee5c8
JP		 2414 pr_info("%s: interface %s is now down.\n",
2415 bond->dev->name,
2416 slave->dev->name);
1da177e4 2417
3d632c3f 2418 if (slave == oldcurrent)
1da177e4 2419 do_failover = 1;
1da177e4
LT		 2420 }
2421 }
2422
2423 /* note: if switch is in round-robin mode, all links
2424 * must tx arp to ensure all links rx an arp - otherwise
2425 * links may oscillate or not come up at all; if switch is
2426 * in something like xor mode, there is nothing we can
2427 * do - all replies will be rx'ed on same link causing slaves
2428 * to be unstable during low/no traffic periods
2429 */
3d632c3f 2430 if (IS_UP(slave->dev))
1da177e4 2431 bond_arp_send_all(bond, slave);
1da177e4
LT		 2432 }
2433
2e52f4fe 2434 rcu_read_unlock();
2435
6fde8f03 2436 if (do_failover || slave_state_changed) {
2e52f4fe 2437 if (!rtnl_trylock())
2438 goto re_arm;
1da177e4 2439
6fde8f03
DT		 2440 if (slave_state_changed) {
2441 bond_slave_state_change(bond);
2442 } else if (do_failover) {
2443 /* the bond_select_active_slave must hold RTNL
2444 * and curr_slave_lock for write.
2445 */
2446 block_netpoll_tx();
2447 write_lock_bh(&bond->curr_slave_lock);
1da177e4 2448
6fde8f03
DT		 2449 bond_select_active_slave(bond);
2450
2451 write_unlock_bh(&bond->curr_slave_lock);
2452 unblock_netpoll_tx();
2453 }
2e52f4fe 2454 rtnl_unlock();
1da177e4
LT		 2455 }
2456
2457re_arm:
e6d265e8 2458 if (bond->params.arp_interval)
e7f63f1d
VF		 2459 queue_delayed_work(bond->wq, &bond->arp_work,
2460 msecs_to_jiffies(bond->params.arp_interval));
1da177e4
LT		 2461}
2462
2463/*
b2220cad
JV		 2464 * Called to inspect slaves for active-backup mode ARP monitor link state
2465 * changes. Sets new_link in slaves to specify what action should take
2466 * place for the slave. Returns 0 if no changes are found, >0 if changes
2467 * to link states must be committed.
2468 *
eb9fa4b0 2469 * Called with rcu_read_lock hold.
1da177e4 2470 */
e7f63f1d 2471static int bond_ab_arp_inspect(struct bonding *bond)
1da177e4 2472{
def4460c 2473 unsigned long trans_start, last_rx;
9caff1e7 2474 struct list_head *iter;
dec1e90e 2475 struct slave *slave;
dec1e90e 2476 int commit = 0;
da210f55 2477
eb9fa4b0 2478 bond_for_each_slave_rcu(bond, slave, iter) {
b2220cad 2479 slave->new_link = BOND_LINK_NOCHANGE;
def4460c 2480 last_rx = slave_last_rx(bond, slave);
1da177e4 2481
b2220cad 2482 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2483 if (bond_time_in_interval(bond, last_rx, 1)) {
b2220cad
JV		 2484 slave->new_link = BOND_LINK_UP;
2485 commit++;
2486 }
b2220cad
JV		 2487 continue;
2488 }
1da177e4 2489
b2220cad
JV		 2490 /*
2491 * Give slaves 2*delta after being enslaved or made
2492 * active. This avoids bouncing, as the last receive
2493 * times need a full ARP monitor cycle to be updated.
2494 */
e7f63f1d 2495 if (bond_time_in_interval(bond, slave->jiffies, 2))
b2220cad
JV		 2496 continue;
2497
2498 /*
2499 * Backup slave is down if:
2500 * - No current_arp_slave AND
2501 * - more than 3*delta since last receive AND
2502 * - the bond has an IP address
2503 *
2504 * Note: a non-null current_arp_slave indicates
2505 * the curr_active_slave went down and we are
2506 * searching for a new one; under this condition
2507 * we only take the curr_active_slave down - this
2508 * gives each slave a chance to tx/rx traffic
2509 * before being taken out
2510 */
e30bc066 2511 if (!bond_is_active_slave(slave) &&
b2220cad 2512 !bond->current_arp_slave &&
e7f63f1d 2513 !bond_time_in_interval(bond, last_rx, 3)) {
b2220cad
JV		 2514 slave->new_link = BOND_LINK_DOWN;
2515 commit++;
2516 }
2517
2518 /*
2519 * Active slave is down if:
2520 * - more than 2*delta since transmitting OR
2521 * - (more than 2*delta since receive AND
2522 * the bond has an IP address)
2523 */
cb32f2a0 2524 trans_start = dev_trans_start(slave->dev);
e30bc066 2525 if (bond_is_active_slave(slave) &&
e7f63f1d
VF		 2526 (!bond_time_in_interval(bond, trans_start, 2) ||
2527 !bond_time_in_interval(bond, last_rx, 2))) {
b2220cad
JV		 2528 slave->new_link = BOND_LINK_DOWN;
2529 commit++;
2530 }
1da177e4
LT		 2531 }
2532
b2220cad
JV		 2533 return commit;
2534}
1da177e4 2535
b2220cad
JV		 2536/*
2537 * Called to commit link state changes noted by inspection step of
2538 * active-backup mode ARP monitor.
2539 *
eb9fa4b0 2540 * Called with RTNL hold.
b2220cad 2541 */
e7f63f1d 2542static void bond_ab_arp_commit(struct bonding *bond)
b2220cad 2543{
cb32f2a0 2544 unsigned long trans_start;
9caff1e7 2545 struct list_head *iter;
dec1e90e 2546 struct slave *slave;
1da177e4 2547
9caff1e7 2548 bond_for_each_slave(bond, slave, iter) {
b2220cad
JV		 2549 switch (slave->new_link) {
2550 case BOND_LINK_NOCHANGE:
2551 continue;
ff59c456 2552
b2220cad 2553 case BOND_LINK_UP:
cb32f2a0 2554 trans_start = dev_trans_start(slave->dev);
e7f63f1d
VF		 2555 if (bond->curr_active_slave != slave ||
2556 (!bond->curr_active_slave &&
2557 bond_time_in_interval(bond, trans_start, 1))) {
b2220cad 2558 slave->link = BOND_LINK_UP;
5a430974
VF		 2559 if (bond->current_arp_slave) {
2560 bond_set_slave_inactive_flags(
2561 bond->current_arp_slave);
2562 bond->current_arp_slave = NULL;
2563 }
b2220cad 2564
a4aee5c8 2565 pr_info("%s: link status definitely up for interface %s.\n",
b9f60253 2566 bond->dev->name, slave->dev->name);
b2220cad 2567
b9f60253
JP		 2568 if (!bond->curr_active_slave ||
2569 (slave == bond->primary_slave))
2570 goto do_failover;
1da177e4 2571
b9f60253 2572 }
1da177e4 2573
b9f60253 2574 continue;
1da177e4 2575
b2220cad
JV		 2576 case BOND_LINK_DOWN:
2577 if (slave->link_failure_count < UINT_MAX)
2578 slave->link_failure_count++;
2579
2580 slave->link = BOND_LINK_DOWN;
b9f60253 2581 bond_set_slave_inactive_flags(slave);
b2220cad 2582
a4aee5c8 2583 pr_info("%s: link status definitely down for interface %s, disabling it\n",
b9f60253 2584 bond->dev->name, slave->dev->name);
b2220cad 2585
b9f60253 2586 if (slave == bond->curr_active_slave) {
b2220cad 2587 bond->current_arp_slave = NULL;
b9f60253 2588 goto do_failover;
1da177e4 2589 }
b9f60253
JP		 2590
2591 continue;
b2220cad
JV		 2592
2593 default:
a4aee5c8 2594 pr_err("%s: impossible: new_link %d on slave %s\n",
b2220cad
JV		 2595 bond->dev->name, slave->new_link,
2596 slave->dev->name);
b9f60253 2597 continue;
1da177e4 2598 }
1da177e4 2599
b9f60253
JP		 2600do_failover:
2601 ASSERT_RTNL();
e843fa50 2602 block_netpoll_tx();
b2220cad 2603 write_lock_bh(&bond->curr_slave_lock);
b9f60253 2604 bond_select_active_slave(bond);
b2220cad 2605 write_unlock_bh(&bond->curr_slave_lock);
e843fa50 2606 unblock_netpoll_tx();
b2220cad 2607 }
1da177e4 2608
b2220cad
JV		 2609 bond_set_carrier(bond);
2610}
1da177e4 2611
b2220cad
JV		 2612/*
2613 * Send ARP probes for active-backup mode ARP monitor.
b2220cad 2614 */
f2ebd477 2615static bool bond_ab_arp_probe(struct bonding *bond)
b2220cad 2616{
eb9fa4b0 2617 struct slave *slave, *before = NULL, *new_slave = NULL,
f2ebd477 2618 *curr_arp_slave, *curr_active_slave;
4087df87
VF		 2619 struct list_head *iter;
2620 bool found = false;
1da177e4 2621
f2ebd477
VF		 2622 rcu_read_lock();
2623 curr_arp_slave = rcu_dereference(bond->current_arp_slave);
2624 curr_active_slave = rcu_dereference(bond->curr_active_slave);
2625
98b90f26 2626 if (curr_arp_slave && curr_active_slave)
a4aee5c8 2627 pr_info("PROBE: c_arp %s && cas %s BAD\n",
eb9fa4b0 2628 curr_arp_slave->dev->name,
98b90f26 2629 curr_active_slave->dev->name);
1da177e4 2630
98b90f26
VF		 2631 if (curr_active_slave) {
2632 bond_arp_send_all(bond, curr_active_slave);
f2ebd477
VF		 2633 rcu_read_unlock();
2634 return true;
b2220cad 2635 }
f2ebd477 2636 rcu_read_unlock();
1da177e4 2637
b2220cad
JV		 2638 /* if we don't have a curr_active_slave, search for the next available
2639 * backup slave from the current_arp_slave and make it the candidate
2640 * for becoming the curr_active_slave
2641 */
1da177e4 2642
f2ebd477
VF		 2643 if (!rtnl_trylock())
2644 return false;
2645 /* curr_arp_slave might have gone away */
2646 curr_arp_slave = ACCESS_ONCE(bond->current_arp_slave);
2647
eb9fa4b0 2648 if (!curr_arp_slave) {
f2ebd477
VF		 2649 curr_arp_slave = bond_first_slave(bond);
2650 if (!curr_arp_slave) {
2651 rtnl_unlock();
2652 return true;
2653 }
b2220cad 2654 }
1da177e4 2655
eb9fa4b0 2656 bond_set_slave_inactive_flags(curr_arp_slave);
059fe7a5 2657
f2ebd477 2658 bond_for_each_slave(bond, slave, iter) {
4087df87
VF		 2659 if (!found && !before && IS_UP(slave->dev))
2660 before = slave;
1da177e4 2661
4087df87
VF		 2662 if (found && !new_slave && IS_UP(slave->dev))
2663 new_slave = slave;
b2220cad
JV		 2664 /* if the link state is up at this point, we
2665 * mark it down - this can happen if we have
2666 * simultaneous link failures and
2667 * reselect_active_interface doesn't make this
2668 * one the current slave so it is still marked
2669 * up when it is actually down
1da177e4 2670 */
4087df87 2671 if (!IS_UP(slave->dev) && slave->link == BOND_LINK_UP) {
b2220cad
JV		 2672 slave->link = BOND_LINK_DOWN;
2673 if (slave->link_failure_count < UINT_MAX)
2674 slave->link_failure_count++;
1da177e4 2675
b2220cad
JV		 2676 bond_set_slave_inactive_flags(slave);
2677
a4aee5c8
JP		 2678 pr_info("%s: backup interface %s is now down.\n",
2679 bond->dev->name, slave->dev->name);
1da177e4 2680 }
eb9fa4b0 2681 if (slave == curr_arp_slave)
4087df87 2682 found = true;
b2220cad 2683 }
4087df87
VF		 2684
2685 if (!new_slave && before)
2686 new_slave = before;
2687
f2ebd477
VF		 2688 if (!new_slave) {
2689 rtnl_unlock();
2690 return true;
2691 }
4087df87
VF		 2692
2693 new_slave->link = BOND_LINK_BACK;
2694 bond_set_slave_active_flags(new_slave);
2695 bond_arp_send_all(bond, new_slave);
2696 new_slave->jiffies = jiffies;
eb9fa4b0 2697 rcu_assign_pointer(bond->current_arp_slave, new_slave);
f2ebd477
VF		 2698 rtnl_unlock();
2699
2700 return true;
b2220cad 2701}
1da177e4 2702
6da67d26 2703static void bond_activebackup_arp_mon(struct work_struct *work)
b2220cad
JV		 2704{
2705 struct bonding *bond = container_of(work, struct bonding,
2706 arp_work.work);
f2ebd477 2707 bool should_notify_peers = false, should_commit = false;
1f2cd845 2708 int delta_in_ticks;
1da177e4 2709
1f2cd845
DM		 2710 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2711
2712 if (!bond_has_slaves(bond))
b2220cad
JV		 2713 goto re_arm;
2714
eb9fa4b0 2715 rcu_read_lock();
ad246c99 2716 should_notify_peers = bond_should_notify_peers(bond);
f2ebd477
VF		 2717 should_commit = bond_ab_arp_inspect(bond);
2718 rcu_read_unlock();
ad246c99 2719
f2ebd477 2720 if (should_commit) {
1f2cd845
DM		 2721 /* Race avoidance with bond_close flush of workqueue */
2722 if (!rtnl_trylock()) {
1f2cd845
DM		 2723 delta_in_ticks = 1;
2724 should_notify_peers = false;
2725 goto re_arm;
2726 }
b2220cad 2727
1f2cd845 2728 bond_ab_arp_commit(bond);
1f2cd845 2729 rtnl_unlock();
1f2cd845
DM		 2730 }
2731
f2ebd477
VF		 2732 if (!bond_ab_arp_probe(bond)) {
2733 /* rtnl locking failed, re-arm */
2734 delta_in_ticks = 1;
2735 should_notify_peers = false;
2736 }
ad246c99 2737
80b9d236 2738re_arm:
2739 if (bond->params.arp_interval)
1f2cd845
DM		 2740 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
2741
1f2cd845
DM		 2742 if (should_notify_peers) {
2743 if (!rtnl_trylock())
2744 return;
2745 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2746 rtnl_unlock();
2747 }
1da177e4
LT		 2748}
2749
1da177e4
LT		 2750/*-------------------------- netdev event handling --------------------------*/
2751
2752/*
2753 * Change device name
2754 */
2755static int bond_event_changename(struct bonding *bond)
2756{
1da177e4
LT		 2757 bond_remove_proc_entry(bond);
2758 bond_create_proc_entry(bond);
7e083840 2759
f073c7ca
TI		 2760 bond_debug_reregister(bond);
2761
1da177e4
LT		 2762 return NOTIFY_DONE;
2763}
2764
3d632c3f
SH		 2765static int bond_master_netdev_event(unsigned long event,
2766 struct net_device *bond_dev)
1da177e4 2767{
454d7c9b 2768 struct bonding *event_bond = netdev_priv(bond_dev);
1da177e4
LT		 2769
2770 switch (event) {
2771 case NETDEV_CHANGENAME:
2772 return bond_event_changename(event_bond);
a64d49c3
EB		 2773 case NETDEV_UNREGISTER:
2774 bond_remove_proc_entry(event_bond);
2775 break;
2776 case NETDEV_REGISTER:
2777 bond_create_proc_entry(event_bond);
2778 break;
6c8d23f7 2779 case NETDEV_NOTIFY_PEERS:
2780 if (event_bond->send_peer_notif)
2781 event_bond->send_peer_notif--;
2782 break;
1da177e4
LT		 2783 default:
2784 break;
2785 }
2786
2787 return NOTIFY_DONE;
2788}
2789
3d632c3f
SH		 2790static int bond_slave_netdev_event(unsigned long event,
2791 struct net_device *slave_dev)
1da177e4 2792{
471cb5a3 2793 struct slave *slave = bond_slave_get_rtnl(slave_dev);
6101391d 2794 struct bonding *bond;
2795 struct net_device *bond_dev;
471cb5a3
JP		 2796 u32 old_speed;
2797 u8 old_duplex;
1da177e4 2798
6101391d 2799 /* A netdev event can be generated while enslaving a device
2800 * before netdev_rx_handler_register is called in which case
2801 * slave will be NULL
2802 */
2803 if (!slave)
2804 return NOTIFY_DONE;
2805 bond_dev = slave->bond->dev;
2806 bond = slave->bond;
2807
1da177e4
LT		 2808 switch (event) {
2809 case NETDEV_UNREGISTER:
8d2ada77 2810 if (bond_dev->type != ARPHRD_ETHER)
471cb5a3
JP		 2811 bond_release_and_destroy(bond_dev, slave_dev);
2812 else
2813 bond_release(bond_dev, slave_dev);
1da177e4 2814 break;
98f41f69 2815 case NETDEV_UP:
1da177e4 2816 case NETDEV_CHANGE:
471cb5a3
JP		 2817 old_speed = slave->speed;
2818 old_duplex = slave->duplex;
17d04500 2819
471cb5a3 2820 bond_update_speed_duplex(slave);
17d04500 2821
471cb5a3
JP		 2822 if (bond->params.mode == BOND_MODE_8023AD) {
2823 if (old_speed != slave->speed)
2824 bond_3ad_adapter_speed_changed(slave);
2825 if (old_duplex != slave->duplex)
2826 bond_3ad_adapter_duplex_changed(slave);
17d04500 2827 }
1da177e4
LT		 2828 break;
2829 case NETDEV_DOWN:
2830 /*
2831 * ... Or is it this?
2832 */
2833 break;
2834 case NETDEV_CHANGEMTU:
2835 /*
2836 * TODO: Should slaves be allowed to
2837 * independently alter their MTU? For
2838 * an active-backup bond, slaves need
2839 * not be the same type of device, so
2840 * MTUs may vary. For other modes,
2841 * slaves arguably should have the
2842 * same MTUs. To do this, we'd need to
2843 * take over the slave's change_mtu
2844 * function for the duration of their
2845 * servitude.
2846 */
2847 break;
2848 case NETDEV_CHANGENAME:
3ec775b9
VF		 2849 /* we don't care if we don't have primary set */
2850 if (!USES_PRIMARY(bond->params.mode) ||
2851 !bond->params.primary[0])
2852 break;
2853
2854 if (slave == bond->primary_slave) {
2855 /* slave's name changed - he's no longer primary */
2856 bond->primary_slave = NULL;
2857 } else if (!strcmp(slave_dev->name, bond->params.primary)) {
2858 /* we have a new primary slave */
2859 bond->primary_slave = slave;
2860 } else { /* we didn't change primary - exit */
2861 break;
2862 }
2863
2864 pr_info("%s: Primary slave changed to %s, reselecting active slave.\n",
2865 bond->dev->name, bond->primary_slave ? slave_dev->name :
2866 "none");
2867 write_lock_bh(&bond->curr_slave_lock);
2868 bond_select_active_slave(bond);
2869 write_unlock_bh(&bond->curr_slave_lock);
1da177e4 2870 break;
8531c5ff
AK		 2871 case NETDEV_FEAT_CHANGE:
2872 bond_compute_features(bond);
2873 break;
4aa5dee4
JP		 2874 case NETDEV_RESEND_IGMP:
2875 /* Propagate to master device */
2876 call_netdevice_notifiers(event, slave->bond->dev);
2877 break;
1da177e4
LT		 2878 default:
2879 break;
2880 }
2881
2882 return NOTIFY_DONE;
2883}
2884
2885/*
2886 * bond_netdev_event: handle netdev notifier chain events.
2887 *
2888 * This function receives events for the netdev chain. The caller (an
e041c683 2889 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
1da177e4
LT		 2890 * locks for us to safely manipulate the slave devices (RTNL lock,
2891 * dev_probe_lock).
2892 */
3d632c3f
SH		 2893static int bond_netdev_event(struct notifier_block *this,
2894 unsigned long event, void *ptr)
1da177e4 2895{
351638e7 2896 struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
1da177e4 2897
5a03cdb7 2898 pr_debug("event_dev: %s, event: %lx\n",
a4aee5c8
JP		 2899 event_dev ? event_dev->name : "None",
2900 event);
1da177e4 2901
0b680e75
JV		 2902 if (!(event_dev->priv_flags & IFF_BONDING))
2903 return NOTIFY_DONE;
2904
1da177e4 2905 if (event_dev->flags & IFF_MASTER) {
5a03cdb7 2906 pr_debug("IFF_MASTER\n");
1da177e4
LT		 2907 return bond_master_netdev_event(event, event_dev);
2908 }
2909
2910 if (event_dev->flags & IFF_SLAVE) {
5a03cdb7 2911 pr_debug("IFF_SLAVE\n");
1da177e4
LT		 2912 return bond_slave_netdev_event(event, event_dev);
2913 }
2914
2915 return NOTIFY_DONE;
2916}
2917
2918static struct notifier_block bond_netdev_notifier = {
2919 .notifier_call = bond_netdev_event,
2920};
2921
169a3e66
JV		 2922/*---------------------------- Hashing Policies -----------------------------*/
2923
32819dc1
NA		 2924/* L2 hash helper */
2925static inline u32 bond_eth_hash(struct sk_buff *skb)
6b923cb7
JE		 2926{
2927 struct ethhdr *data = (struct ethhdr *)skb->data;
2928
2929 if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto))
32819dc1 2930 return data->h_dest[5] ^ data->h_source[5];
6b923cb7
JE		 2931
2932 return 0;
2933}
2934
32819dc1
NA		 2935/* Extract the appropriate headers based on bond's xmit policy */
2936static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
2937 struct flow_keys *fk)
6f6652be 2938{
32819dc1 2939 const struct ipv6hdr *iph6;
4394542c 2940 const struct iphdr *iph;
32819dc1 2941 int noff, proto = -1;
6b923cb7 2942
32819dc1
NA		 2943 if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
2944 return skb_flow_dissect(skb, fk);
2945
2946 fk->ports = 0;
2947 noff = skb_network_offset(skb);
2948 if (skb->protocol == htons(ETH_P_IP)) {
2949 if (!pskb_may_pull(skb, noff + sizeof(*iph)))
2950 return false;
6b923cb7 2951 iph = ip_hdr(skb);
32819dc1
NA		 2952 fk->src = iph->saddr;
2953 fk->dst = iph->daddr;
2954 noff += iph->ihl << 2;
2955 if (!ip_is_fragment(iph))
2956 proto = iph->protocol;
2957 } else if (skb->protocol == htons(ETH_P_IPV6)) {
2958 if (!pskb_may_pull(skb, noff + sizeof(*iph6)))
2959 return false;
2960 iph6 = ipv6_hdr(skb);
2961 fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr);
2962 fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr);
2963 noff += sizeof(*iph6);
2964 proto = iph6->nexthdr;
2965 } else {
2966 return false;
6f6652be 2967 }
32819dc1
NA		 2968 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
2969 fk->ports = skb_flow_get_ports(skb, noff, proto);
6f6652be 2970
32819dc1 2971 return true;
6f6652be
JV		 2972}
2973
32819dc1
NA		 2974/**
2975 * bond_xmit_hash - generate a hash value based on the xmit policy
2976 * @bond: bonding device
2977 * @skb: buffer to use for headers
2978 * @count: modulo value
2979 *
2980 * This function will extract the necessary headers from the skb buffer and use
2981 * them to generate a hash based on the xmit_policy set in the bonding device
2982 * which will be reduced modulo count before returning.
169a3e66 2983 */
32819dc1 2984int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count)
169a3e66 2985{
32819dc1
NA		 2986 struct flow_keys flow;
2987 u32 hash;
4394542c 2988
32819dc1
NA		 2989 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 ||
2990 !bond_flow_dissect(bond, skb, &flow))
2991 return bond_eth_hash(skb) % count;
169a3e66 2992
32819dc1
NA		 2993 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 ||
2994 bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
2995 hash = bond_eth_hash(skb);
2996 else
2997 hash = (__force u32)flow.ports;
2998 hash ^= (__force u32)flow.dst ^ (__force u32)flow.src;
2999 hash ^= (hash >> 16);
3000 hash ^= (hash >> 8);
3001
3002 return hash % count;
169a3e66
JV		 3003}
3004
1da177e4
LT		 3005/*-------------------------- Device entry points ----------------------------*/
3006
fbb0c41b 3007static void bond_work_init_all(struct bonding *bond)
3008{
3009 INIT_DELAYED_WORK(&bond->mcast_work,
3010 bond_resend_igmp_join_requests_delayed);
3011 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3012 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
3013 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3014 INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon);
3015 else
3016 INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon);
3017 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
3018}
3019
3020static void bond_work_cancel_all(struct bonding *bond)
3021{
3022 cancel_delayed_work_sync(&bond->mii_work);
3023 cancel_delayed_work_sync(&bond->arp_work);
3024 cancel_delayed_work_sync(&bond->alb_work);
3025 cancel_delayed_work_sync(&bond->ad_work);
3026 cancel_delayed_work_sync(&bond->mcast_work);
3027}
3028
1da177e4
LT		 3029static int bond_open(struct net_device *bond_dev)
3030{
454d7c9b 3031 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3032 struct list_head *iter;
ba3211cc 3033 struct slave *slave;
1da177e4 3034
ba3211cc
PP		 3035 /* reset slave->backup and slave->inactive */
3036 read_lock(&bond->lock);
0965a1f3 3037 if (bond_has_slaves(bond)) {
ba3211cc 3038 read_lock(&bond->curr_slave_lock);
9caff1e7 3039 bond_for_each_slave(bond, slave, iter) {
ba3211cc
PP		 3040 if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3041 && (slave != bond->curr_active_slave)) {
3042 bond_set_slave_inactive_flags(slave);
3043 } else {
3044 bond_set_slave_active_flags(slave);
3045 }
3046 }
3047 read_unlock(&bond->curr_slave_lock);
3048 }
3049 read_unlock(&bond->lock);
3050
fbb0c41b 3051 bond_work_init_all(bond);
5a37e8ca 3052
58402054 3053 if (bond_is_lb(bond)) {
1da177e4
LT		 3054 /* bond_alb_initialize must be called before the timer
3055 * is started.
3056 */
fbb0c41b 3057 if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB)))
b473946a 3058 return -ENOMEM;
1b76b316 3059 queue_delayed_work(bond->wq, &bond->alb_work, 0);
1da177e4
LT		 3060 }
3061
fbb0c41b 3062 if (bond->params.miimon) /* link check interval, in milliseconds. */
1b76b316 3063 queue_delayed_work(bond->wq, &bond->mii_work, 0);
1da177e4
LT		 3064
3065 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316 3066 queue_delayed_work(bond->wq, &bond->arp_work, 0);
f5b2b966 3067 if (bond->params.arp_validate)
3aba891d 3068 bond->recv_probe = bond_arp_rcv;
1da177e4
LT		 3069 }
3070
3071 if (bond->params.mode == BOND_MODE_8023AD) {
1b76b316 3072 queue_delayed_work(bond->wq, &bond->ad_work, 0);
1da177e4 3073 /* register to receive LACPDUs */
3aba891d 3074 bond->recv_probe = bond_3ad_lacpdu_recv;
fd989c83 3075 bond_3ad_initiate_agg_selection(bond, 1);
1da177e4
LT		 3076 }
3077
3078 return 0;
3079}
3080
3081static int bond_close(struct net_device *bond_dev)
3082{
454d7c9b 3083 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3084
fbb0c41b 3085 bond_work_cancel_all(bond);
6c8d23f7 3086 bond->send_peer_notif = 0;
ee8487c0 3087 if (bond_is_lb(bond))
1da177e4 3088 bond_alb_deinitialize(bond);
3aba891d 3089 bond->recv_probe = NULL;
1da177e4
LT		 3090
3091 return 0;
3092}
3093
28172739
ED		 3094static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
3095 struct rtnl_link_stats64 *stats)
1da177e4 3096{
454d7c9b 3097 struct bonding *bond = netdev_priv(bond_dev);
28172739 3098 struct rtnl_link_stats64 temp;
9caff1e7 3099 struct list_head *iter;
1da177e4 3100 struct slave *slave;
1da177e4 3101
28172739 3102 memset(stats, 0, sizeof(*stats));
1da177e4
LT		 3103
3104 read_lock_bh(&bond->lock);
9caff1e7 3105 bond_for_each_slave(bond, slave, iter) {
be1f3c2c 3106 const struct rtnl_link_stats64 *sstats =
28172739
ED		 3107 dev_get_stats(slave->dev, &temp);
3108
3109 stats->rx_packets += sstats->rx_packets;
3110 stats->rx_bytes += sstats->rx_bytes;
3111 stats->rx_errors += sstats->rx_errors;
3112 stats->rx_dropped += sstats->rx_dropped;
3113
3114 stats->tx_packets += sstats->tx_packets;
3115 stats->tx_bytes += sstats->tx_bytes;
3116 stats->tx_errors += sstats->tx_errors;
3117 stats->tx_dropped += sstats->tx_dropped;
3118
3119 stats->multicast += sstats->multicast;
3120 stats->collisions += sstats->collisions;
3121
3122 stats->rx_length_errors += sstats->rx_length_errors;
3123 stats->rx_over_errors += sstats->rx_over_errors;
3124 stats->rx_crc_errors += sstats->rx_crc_errors;
3125 stats->rx_frame_errors += sstats->rx_frame_errors;
3126 stats->rx_fifo_errors += sstats->rx_fifo_errors;
3127 stats->rx_missed_errors += sstats->rx_missed_errors;
3128
3129 stats->tx_aborted_errors += sstats->tx_aborted_errors;
3130 stats->tx_carrier_errors += sstats->tx_carrier_errors;
3131 stats->tx_fifo_errors += sstats->tx_fifo_errors;
3132 stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
3133 stats->tx_window_errors += sstats->tx_window_errors;
2439f9eb 3134 }
1da177e4
LT		 3135 read_unlock_bh(&bond->lock);
3136
3137 return stats;
3138}
3139
3140static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3141{
080a06e1 3142 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3143 struct net_device *slave_dev = NULL;
3144 struct ifbond k_binfo;
3145 struct ifbond __user *u_binfo = NULL;
3146 struct ifslave k_sinfo;
3147 struct ifslave __user *u_sinfo = NULL;
3148 struct mii_ioctl_data *mii = NULL;
d1fbd3ed 3149 struct bond_opt_value newval;
387ff911 3150 struct net *net;
1da177e4
LT		 3151 int res = 0;
3152
a4aee5c8 3153 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd);
1da177e4
LT		 3154
3155 switch (cmd) {
1da177e4
LT		 3156 case SIOCGMIIPHY:
3157 mii = if_mii(ifr);
3d632c3f 3158 if (!mii)
1da177e4 3159 return -EINVAL;
3d632c3f 3160
1da177e4
LT		 3161 mii->phy_id = 0;
3162 /* Fall Through */
3163 case SIOCGMIIREG:
3164 /*
3165 * We do this again just in case we were called by SIOCGMIIREG
3166 * instead of SIOCGMIIPHY.
3167 */
3168 mii = if_mii(ifr);
3d632c3f 3169 if (!mii)
1da177e4 3170 return -EINVAL;
3d632c3f 3171
1da177e4
LT		 3172
3173 if (mii->reg_num == 1) {
1da177e4 3174 mii->val_out = 0;
6603a6f2 3175 read_lock(&bond->lock);
1da177e4 3176 read_lock(&bond->curr_slave_lock);
3d632c3f 3177 if (netif_carrier_ok(bond->dev))
1da177e4 3178 mii->val_out = BMSR_LSTATUS;
3d632c3f 3179
1da177e4 3180 read_unlock(&bond->curr_slave_lock);
6603a6f2 3181 read_unlock(&bond->lock);
1da177e4
LT		 3182 }
3183
3184 return 0;
3185 case BOND_INFO_QUERY_OLD:
3186 case SIOCBONDINFOQUERY:
3187 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3188
3d632c3f 3189 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
1da177e4 3190 return -EFAULT;
1da177e4
LT		 3191
3192 res = bond_info_query(bond_dev, &k_binfo);
3d632c3f
SH		 3193 if (res == 0 &&
3194 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3195 return -EFAULT;
1da177e4
LT		 3196
3197 return res;
3198 case BOND_SLAVE_INFO_QUERY_OLD:
3199 case SIOCBONDSLAVEINFOQUERY:
3200 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3201
3d632c3f 3202 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
1da177e4 3203 return -EFAULT;
1da177e4
LT		 3204
3205 res = bond_slave_info_query(bond_dev, &k_sinfo);
3d632c3f
SH		 3206 if (res == 0 &&
3207 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3208 return -EFAULT;
1da177e4
LT		 3209
3210 return res;
3211 default:
3212 /* Go on */
3213 break;
3214 }
3215
387ff911
G		 3216 net = dev_net(bond_dev);
3217
3218 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4 3219 return -EPERM;
1da177e4 3220
0917b933 3221 slave_dev = __dev_get_by_name(net, ifr->ifr_slave);
1da177e4 3222
a4aee5c8 3223 pr_debug("slave_dev=%p:\n", slave_dev);
1da177e4 3224
3d632c3f 3225 if (!slave_dev)
0917b933 3226 return -ENODEV;
1da177e4 3227
0917b933
YX		 3228 pr_debug("slave_dev->name=%s:\n", slave_dev->name);
3229 switch (cmd) {
3230 case BOND_ENSLAVE_OLD:
3231 case SIOCBONDENSLAVE:
3232 res = bond_enslave(bond_dev, slave_dev);
3233 break;
3234 case BOND_RELEASE_OLD:
3235 case SIOCBONDRELEASE:
3236 res = bond_release(bond_dev, slave_dev);
3237 break;
3238 case BOND_SETHWADDR_OLD:
3239 case SIOCBONDSETHWADDR:
3240 bond_set_dev_addr(bond_dev, slave_dev);
3241 res = 0;
3242 break;
3243 case BOND_CHANGE_ACTIVE_OLD:
3244 case SIOCBONDCHANGEACTIVE:
d1fbd3ed
NA		 3245 bond_opt_initstr(&newval, slave_dev->name);
3246 res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval);
0917b933
YX		 3247 break;
3248 default:
3249 res = -EOPNOTSUPP;
1da177e4
LT		 3250 }
3251
1da177e4
LT		 3252 return res;
3253}
3254
d03462b9 3255static void bond_change_rx_flags(struct net_device *bond_dev, int change)
1da177e4 3256{
454d7c9b 3257 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3258
d03462b9
JP		 3259 if (change & IFF_PROMISC)
3260 bond_set_promiscuity(bond,
3261 bond_dev->flags & IFF_PROMISC ? 1 : -1);
3d632c3f 3262
d03462b9
JP		 3263 if (change & IFF_ALLMULTI)
3264 bond_set_allmulti(bond,
3265 bond_dev->flags & IFF_ALLMULTI ? 1 : -1);
3266}
1da177e4 3267
303d1cbf 3268static void bond_set_rx_mode(struct net_device *bond_dev)
d03462b9
JP		 3269{
3270 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3271 struct list_head *iter;
303d1cbf 3272 struct slave *slave;
1da177e4 3273
80ee5ad2 3274
b3241870 3275 rcu_read_lock();
303d1cbf 3276 if (USES_PRIMARY(bond->params.mode)) {
b3241870 3277 slave = rcu_dereference(bond->curr_active_slave);
303d1cbf
JV		 3278 if (slave) {
3279 dev_uc_sync(slave->dev, bond_dev);
3280 dev_mc_sync(slave->dev, bond_dev);
3281 }
303d1cbf 3282 } else {
b3241870 3283 bond_for_each_slave_rcu(bond, slave, iter) {
303d1cbf
JV		 3284 dev_uc_sync_multiple(slave->dev, bond_dev);
3285 dev_mc_sync_multiple(slave->dev, bond_dev);
3286 }
1da177e4 3287 }
b3241870 3288 rcu_read_unlock();
1da177e4
LT		 3289}
3290
234bcf8a 3291static int bond_neigh_init(struct neighbour *n)
00829823 3292{
234bcf8a 3293 struct bonding *bond = netdev_priv(n->dev);
234bcf8a
SP		 3294 const struct net_device_ops *slave_ops;
3295 struct neigh_parms parms;
dec1e90e 3296 struct slave *slave;
234bcf8a
SP		 3297 int ret;
3298
dec1e90e 3299 slave = bond_first_slave(bond);
234bcf8a
SP		 3300 if (!slave)
3301 return 0;
234bcf8a 3302 slave_ops = slave->dev->netdev_ops;
234bcf8a
SP		 3303 if (!slave_ops->ndo_neigh_setup)
3304 return 0;
3305
3306 parms.neigh_setup = NULL;
3307 parms.neigh_cleanup = NULL;
3308 ret = slave_ops->ndo_neigh_setup(slave->dev, &parms);
3309 if (ret)
3310 return ret;
3311
3312 /*
3313 * Assign slave's neigh_cleanup to neighbour in case cleanup is called
3314 * after the last slave has been detached. Assumes that all slaves
3315 * utilize the same neigh_cleanup (true at this writing as only user
3316 * is ipoib).
3317 */
3318 n->parms->neigh_cleanup = parms.neigh_cleanup;
3319
3320 if (!parms.neigh_setup)
3321 return 0;
3322
3323 return parms.neigh_setup(n);
3324}
3325
3326/*
3327 * The bonding ndo_neigh_setup is called at init time beofre any
3328 * slave exists. So we must declare proxy setup function which will
3329 * be used at run time to resolve the actual slave neigh param setup.
9918d5bf
VF		 3330 *
3331 * It's also called by master devices (such as vlans) to setup their
3332 * underlying devices. In that case - do nothing, we're already set up from
3333 * our init.
234bcf8a
SP		 3334 */
3335static int bond_neigh_setup(struct net_device *dev,
3336 struct neigh_parms *parms)
3337{
9918d5bf
VF		 3338 /* modify only our neigh_parms */
3339 if (parms->dev == dev)
3340 parms->neigh_setup = bond_neigh_init;
00829823 3341
00829823
SH		 3342 return 0;
3343}
3344
1da177e4
LT		 3345/*
3346 * Change the MTU of all of a master's slaves to match the master
3347 */
3348static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3349{
454d7c9b 3350 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3351 struct slave *slave, *rollback_slave;
9caff1e7 3352 struct list_head *iter;
1da177e4 3353 int res = 0;
1da177e4 3354
5a03cdb7 3355 pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond,
a4aee5c8 3356 (bond_dev ? bond_dev->name : "None"), new_mtu);
1da177e4
LT		 3357
3358 /* Can't hold bond->lock with bh disabled here since
3359 * some base drivers panic. On the other hand we can't
3360 * hold bond->lock without bh disabled because we'll
3361 * deadlock. The only solution is to rely on the fact
3362 * that we're under rtnl_lock here, and the slaves
3363 * list won't change. This doesn't solve the problem
3364 * of setting the slave's MTU while it is
3365 * transmitting, but the assumption is that the base
3366 * driver can handle that.
3367 *
3368 * TODO: figure out a way to safely iterate the slaves
3369 * list, but without holding a lock around the actual
3370 * call to the base driver.
3371 */
3372
9caff1e7 3373 bond_for_each_slave(bond, slave, iter) {
c8c23903 3374 pr_debug("s %p c_m %p\n",
a4aee5c8 3375 slave,
a4aee5c8 3376 slave->dev->netdev_ops->ndo_change_mtu);
e944ef79 3377
1da177e4
LT		 3378 res = dev_set_mtu(slave->dev, new_mtu);
3379
3380 if (res) {
3381 /* If we failed to set the slave's mtu to the new value
3382 * we must abort the operation even in ACTIVE_BACKUP
3383 * mode, because if we allow the backup slaves to have
3384 * different mtu values than the active slave we'll
3385 * need to change their mtu when doing a failover. That
3386 * means changing their mtu from timer context, which
3387 * is probably not a good idea.
3388 */
5a03cdb7 3389 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 3390 goto unwind;
3391 }
3392 }
3393
3394 bond_dev->mtu = new_mtu;
3395
3396 return 0;
3397
3398unwind:
3399 /* unwind from head to the slave that failed */
9caff1e7 3400 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT		 3401 int tmp_res;
3402
81f23b13
VF		 3403 if (rollback_slave == slave)
3404 break;
3405
3406 tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu);
1da177e4 3407 if (tmp_res) {
a4aee5c8 3408 pr_debug("unwind err %d dev %s\n",
81f23b13 3409 tmp_res, rollback_slave->dev->name);
1da177e4
LT		 3410 }
3411 }
3412
3413 return res;
3414}
3415
3416/*
3417 * Change HW address
3418 *
3419 * Note that many devices must be down to change the HW address, and
3420 * downing the master releases all slaves. We can make bonds full of
3421 * bonding devices to test this, however.
3422 */
3423static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
3424{
454d7c9b 3425 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3426 struct slave *slave, *rollback_slave;
1da177e4 3427 struct sockaddr *sa = addr, tmp_sa;
9caff1e7 3428 struct list_head *iter;
1da177e4 3429 int res = 0;
1da177e4 3430
eb7cc59a
SH		 3431 if (bond->params.mode == BOND_MODE_ALB)
3432 return bond_alb_set_mac_address(bond_dev, addr);
3433
3434
a4aee5c8
JP		 3435 pr_debug("bond=%p, name=%s\n",
3436 bond, bond_dev ? bond_dev->name : "None");
1da177e4 3437
1b5acd29
JV		 3438 /* If fail_over_mac is enabled, do nothing and return success.
3439 * Returning an error causes ifenslave to fail.
dd957c57 3440 */
cc689aaa 3441 if (bond->params.fail_over_mac &&
3442 bond->params.mode == BOND_MODE_ACTIVEBACKUP)
dd957c57 3443 return 0;
2ab82852 3444
3d632c3f 3445 if (!is_valid_ether_addr(sa->sa_data))
1da177e4 3446 return -EADDRNOTAVAIL;
1da177e4
LT		 3447
3448 /* Can't hold bond->lock with bh disabled here since
3449 * some base drivers panic. On the other hand we can't
3450 * hold bond->lock without bh disabled because we'll
3451 * deadlock. The only solution is to rely on the fact
3452 * that we're under rtnl_lock here, and the slaves
3453 * list won't change. This doesn't solve the problem
3454 * of setting the slave's hw address while it is
3455 * transmitting, but the assumption is that the base
3456 * driver can handle that.
3457 *
3458 * TODO: figure out a way to safely iterate the slaves
3459 * list, but without holding a lock around the actual
3460 * call to the base driver.
3461 */
3462
9caff1e7 3463 bond_for_each_slave(bond, slave, iter) {
5a03cdb7 3464 pr_debug("slave %p %s\n", slave, slave->dev->name);
1da177e4
LT		 3465 res = dev_set_mac_address(slave->dev, addr);
3466 if (res) {
3467 /* TODO: consider downing the slave
3468 * and retry ?
3469 * User should expect communications
3470 * breakage anyway until ARP finish
3471 * updating, so...
3472 */
5a03cdb7 3473 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 3474 goto unwind;
3475 }
3476 }
3477
3478 /* success */
3479 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
3480 return 0;
3481
3482unwind:
3483 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
3484 tmp_sa.sa_family = bond_dev->type;
3485
3486 /* unwind from head to the slave that failed */
9caff1e7 3487 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT		 3488 int tmp_res;
3489
81f23b13
VF		 3490 if (rollback_slave == slave)
3491 break;
3492
3493 tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa);
1da177e4 3494 if (tmp_res) {
a4aee5c8 3495 pr_debug("unwind err %d dev %s\n",
81f23b13 3496 tmp_res, rollback_slave->dev->name);
1da177e4
LT		 3497 }
3498 }
3499
3500 return res;
3501}
3502
15077228
NA		 3503/**
3504 * bond_xmit_slave_id - transmit skb through slave with slave_id
3505 * @bond: bonding device that is transmitting
3506 * @skb: buffer to transmit
3507 * @slave_id: slave id up to slave_cnt-1 through which to transmit
3508 *
3509 * This function tries to transmit through slave with slave_id but in case
3510 * it fails, it tries to find the first available slave for transmission.
3511 * The skb is consumed in all cases, thus the function is void.
3512 */
da131ddb 3513static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id)
15077228 3514{
9caff1e7 3515 struct list_head *iter;
15077228
NA		 3516 struct slave *slave;
3517 int i = slave_id;
3518
3519 /* Here we start from the slave with slave_id */
9caff1e7 3520 bond_for_each_slave_rcu(bond, slave, iter) {
15077228
NA		 3521 if (--i < 0) {
3522 if (slave_can_tx(slave)) {
3523 bond_dev_queue_xmit(bond, skb, slave->dev);
3524 return;
3525 }
3526 }
3527 }
3528
3529 /* Here we start from the first slave up to slave_id */
3530 i = slave_id;
9caff1e7 3531 bond_for_each_slave_rcu(bond, slave, iter) {
15077228
NA		 3532 if (--i < 0)
3533 break;
3534 if (slave_can_tx(slave)) {
3535 bond_dev_queue_xmit(bond, skb, slave->dev);
3536 return;
3537 }
3538 }
3539 /* no slave that can tx has been found */
3540 kfree_skb(skb);
3541}
3542
73958329
NA		 3543/**
3544 * bond_rr_gen_slave_id - generate slave id based on packets_per_slave
3545 * @bond: bonding device to use
3546 *
3547 * Based on the value of the bonding device's packets_per_slave parameter
3548 * this function generates a slave id, which is usually used as the next
3549 * slave to transmit through.
3550 */
3551static u32 bond_rr_gen_slave_id(struct bonding *bond)
3552{
73958329 3553 u32 slave_id;
809fa972
HFS		 3554 struct reciprocal_value reciprocal_packets_per_slave;
3555 int packets_per_slave = bond->params.packets_per_slave;
73958329
NA		 3556
3557 switch (packets_per_slave) {
3558 case 0:
3559 slave_id = prandom_u32();
3560 break;
3561 case 1:
3562 slave_id = bond->rr_tx_counter;
3563 break;
3564 default:
809fa972
HFS		 3565 reciprocal_packets_per_slave =
3566 bond->params.reciprocal_packets_per_slave;
73958329 3567 slave_id = reciprocal_divide(bond->rr_tx_counter,
809fa972 3568 reciprocal_packets_per_slave);
73958329
NA		 3569 break;
3570 }
3571 bond->rr_tx_counter++;
3572
3573 return slave_id;
3574}
3575
1da177e4
LT		 3576static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
3577{
454d7c9b 3578 struct bonding *bond = netdev_priv(bond_dev);
a2fd940f 3579 struct iphdr *iph = ip_hdr(skb);
15077228 3580 struct slave *slave;
73958329 3581 u32 slave_id;
1da177e4 3582
73958329 3583 /* Start with the curr_active_slave that joined the bond as the
a2fd940f
AG		 3584 * default for sending IGMP traffic. For failover purposes one
3585 * needs to maintain some consistency for the interface that will
3586 * send the join/membership reports. The curr_active_slave found
3587 * will send all of this type of traffic.
cf5f9044 3588 */
15077228 3589 if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) {
278b2083 3590 slave = rcu_dereference(bond->curr_active_slave);
15077228
NA		 3591 if (slave && slave_can_tx(slave))
3592 bond_dev_queue_xmit(bond, skb, slave->dev);
3593 else
3594 bond_xmit_slave_id(bond, skb, 0);
a2fd940f 3595 } else {
73958329
NA		 3596 slave_id = bond_rr_gen_slave_id(bond);
3597 bond_xmit_slave_id(bond, skb, slave_id % bond->slave_cnt);
1da177e4 3598 }
0693e88e 3599
ec634fe3 3600 return NETDEV_TX_OK;
1da177e4
LT		 3601}
3602
3603/*
3604 * in active-backup mode, we know that bond->curr_active_slave is always valid if
3605 * the bond has a usable interface.
3606 */
3607static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
3608{
454d7c9b 3609 struct bonding *bond = netdev_priv(bond_dev);
71bc3b2d 3610 struct slave *slave;
1da177e4 3611
278b2083 3612 slave = rcu_dereference(bond->curr_active_slave);
71bc3b2d 3613 if (slave)
15077228
NA		 3614 bond_dev_queue_xmit(bond, skb, slave->dev);
3615 else
04502430 3616 kfree_skb(skb);
0693e88e 3617
ec634fe3 3618 return NETDEV_TX_OK;
1da177e4
LT		 3619}
3620
32819dc1 3621/* In bond_xmit_xor() , we determine the output device by using a pre-
169a3e66
JV		 3622 * determined xmit_hash_policy(), If the selected device is not enabled,
3623 * find the next active slave.
1da177e4
LT		 3624 */
3625static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
3626{
454d7c9b 3627 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3628
32819dc1 3629 bond_xmit_slave_id(bond, skb, bond_xmit_hash(bond, skb, bond->slave_cnt));
0693e88e 3630
ec634fe3 3631 return NETDEV_TX_OK;
1da177e4
LT		 3632}
3633
78a646ce 3634/* in broadcast mode, we send everything to all usable interfaces. */
1da177e4
LT		 3635static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
3636{
454d7c9b 3637 struct bonding *bond = netdev_priv(bond_dev);
78a646ce 3638 struct slave *slave = NULL;
9caff1e7 3639 struct list_head *iter;
1da177e4 3640
9caff1e7 3641 bond_for_each_slave_rcu(bond, slave, iter) {
78a646ce
NA		 3642 if (bond_is_last_slave(bond, slave))
3643 break;
3644 if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) {
3645 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1da177e4 3646
78a646ce
NA		 3647 if (!skb2) {
3648 pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n",
3649 bond_dev->name);
3650 continue;
1da177e4 3651 }
78a646ce
NA		 3652 /* bond_dev_queue_xmit always returns 0 */
3653 bond_dev_queue_xmit(bond, skb2, slave->dev);
1da177e4
LT		 3654 }
3655 }
78a646ce
NA		 3656 if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP)
3657 bond_dev_queue_xmit(bond, skb, slave->dev);
3658 else
04502430 3659 kfree_skb(skb);
3d632c3f 3660
ec634fe3 3661 return NETDEV_TX_OK;
1da177e4
LT		 3662}
3663
3664/*------------------------- Device initialization ---------------------------*/
3665
bb1d9123
AG		 3666/*
3667 * Lookup the slave that corresponds to a qid
3668 */
3669static inline int bond_slave_override(struct bonding *bond,
3670 struct sk_buff *skb)
3671{
bb1d9123 3672 struct slave *slave = NULL;
9caff1e7 3673 struct list_head *iter;
bb1d9123 3674
0693e88e
MM		 3675 if (!skb->queue_mapping)
3676 return 1;
bb1d9123
AG		 3677
3678 /* Find out if any slaves have the same mapping as this skb. */
3900f290 3679 bond_for_each_slave_rcu(bond, slave, iter) {
3680 if (slave->queue_id == skb->queue_mapping) {
3681 if (slave_can_tx(slave)) {
3682 bond_dev_queue_xmit(bond, skb, slave->dev);
3683 return 0;
3684 }
3685 /* If the slave isn't UP, use default transmit policy. */
bb1d9123
AG		 3686 break;
3687 }
3688 }
3689
3900f290 3690 return 1;
bb1d9123
AG		 3691}
3692
374eeb5a 3693
f663dd9a
JW		 3694static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb,
3695 void *accel_priv)
bb1d9123
AG		 3696{
3697 /*
3698 * This helper function exists to help dev_pick_tx get the correct
fd0e435b 3699 * destination queue. Using a helper function skips a call to
bb1d9123
AG		 3700 * skb_tx_hash and will put the skbs in the queue we expect on their
3701 * way down to the bonding driver.
3702 */
fd0e435b
PO		 3703 u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0;
3704
374eeb5a
NH		 3705 /*
3706 * Save the original txq to restore before passing to the driver
3707 */
df4ab5b3 3708 qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping;
374eeb5a 3709
fd0e435b 3710 if (unlikely(txq >= dev->real_num_tx_queues)) {
d30ee670 3711 do {
fd0e435b 3712 txq -= dev->real_num_tx_queues;
d30ee670 3713 } while (txq >= dev->real_num_tx_queues);
fd0e435b
PO		 3714 }
3715 return txq;
bb1d9123
AG		 3716}
3717
0693e88e 3718static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
00829823 3719{
bb1d9123
AG		 3720 struct bonding *bond = netdev_priv(dev);
3721
3722 if (TX_QUEUE_OVERRIDE(bond->params.mode)) {
3723 if (!bond_slave_override(bond, skb))
3724 return NETDEV_TX_OK;
3725 }
00829823
SH		 3726
3727 switch (bond->params.mode) {
3728 case BOND_MODE_ROUNDROBIN:
3729 return bond_xmit_roundrobin(skb, dev);
3730 case BOND_MODE_ACTIVEBACKUP:
3731 return bond_xmit_activebackup(skb, dev);
3732 case BOND_MODE_XOR:
3733 return bond_xmit_xor(skb, dev);
3734 case BOND_MODE_BROADCAST:
3735 return bond_xmit_broadcast(skb, dev);
3736 case BOND_MODE_8023AD:
3737 return bond_3ad_xmit_xor(skb, dev);
3738 case BOND_MODE_ALB:
3739 case BOND_MODE_TLB:
3740 return bond_alb_xmit(skb, dev);
3741 default:
3742 /* Should never happen, mode already checked */
a4aee5c8
JP		 3743 pr_err("%s: Error: Unknown bonding mode %d\n",
3744 dev->name, bond->params.mode);
00829823 3745 WARN_ON_ONCE(1);
04502430 3746 kfree_skb(skb);
00829823
SH		 3747 return NETDEV_TX_OK;
3748 }
3749}
3750
0693e88e
MM		 3751static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
3752{
3753 struct bonding *bond = netdev_priv(dev);
3754 netdev_tx_t ret = NETDEV_TX_OK;
3755
3756 /*
3757 * If we risk deadlock from transmitting this in the
3758 * netpoll path, tell netpoll to queue the frame for later tx
3759 */
3760 if (is_netpoll_tx_blocked(dev))
3761 return NETDEV_TX_BUSY;
3762
278b2083 3763 rcu_read_lock();
0965a1f3 3764 if (bond_has_slaves(bond))
0693e88e
MM		 3765 ret = __bond_start_xmit(skb, dev);
3766 else
04502430 3767 kfree_skb(skb);
278b2083 3768 rcu_read_unlock();
0693e88e
MM		 3769
3770 return ret;
3771}
00829823 3772
bb5b052f
AG		 3773static int bond_ethtool_get_settings(struct net_device *bond_dev,
3774 struct ethtool_cmd *ecmd)
3775{
3776 struct bonding *bond = netdev_priv(bond_dev);
bb5b052f 3777 unsigned long speed = 0;
9caff1e7 3778 struct list_head *iter;
dec1e90e 3779 struct slave *slave;
bb5b052f
AG		 3780
3781 ecmd->duplex = DUPLEX_UNKNOWN;
3782 ecmd->port = PORT_OTHER;
3783
3784 /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we
3785 * do not need to check mode. Though link speed might not represent
3786 * the true receive or transmit bandwidth (not all modes are symmetric)
3787 * this is an accurate maximum.
3788 */
3789 read_lock(&bond->lock);
9caff1e7 3790 bond_for_each_slave(bond, slave, iter) {
bb5b052f
AG		 3791 if (SLAVE_IS_OK(slave)) {
3792 if (slave->speed != SPEED_UNKNOWN)
3793 speed += slave->speed;
3794 if (ecmd->duplex == DUPLEX_UNKNOWN &&
3795 slave->duplex != DUPLEX_UNKNOWN)
3796 ecmd->duplex = slave->duplex;
3797 }
3798 }
3799 ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN);
3800 read_unlock(&bond->lock);
dec1e90e 3801
bb5b052f
AG		 3802 return 0;
3803}
3804
217df670 3805static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
7826d43f 3806 struct ethtool_drvinfo *drvinfo)
217df670 3807{
7826d43f
JP		 3808 strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
3809 strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
3810 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d",
3811 BOND_ABI_VERSION);
217df670
JV		 3812}
3813
7282d491 3814static const struct ethtool_ops bond_ethtool_ops = {
217df670 3815 .get_drvinfo = bond_ethtool_get_drvinfo,
bb5b052f 3816 .get_settings = bond_ethtool_get_settings,
fa53ebac 3817 .get_link = ethtool_op_get_link,
8531c5ff
AK		 3818};
3819
eb7cc59a 3820static const struct net_device_ops bond_netdev_ops = {
181470fc 3821 .ndo_init = bond_init,
9e71626c 3822 .ndo_uninit = bond_uninit,
eb7cc59a
SH		 3823 .ndo_open = bond_open,
3824 .ndo_stop = bond_close,
00829823 3825 .ndo_start_xmit = bond_start_xmit,
bb1d9123 3826 .ndo_select_queue = bond_select_queue,
be1f3c2c 3827 .ndo_get_stats64 = bond_get_stats,
eb7cc59a 3828 .ndo_do_ioctl = bond_do_ioctl,
d03462b9 3829 .ndo_change_rx_flags = bond_change_rx_flags,
303d1cbf 3830 .ndo_set_rx_mode = bond_set_rx_mode,
eb7cc59a 3831 .ndo_change_mtu = bond_change_mtu,
cc0e4070 3832 .ndo_set_mac_address = bond_set_mac_address,
00829823 3833 .ndo_neigh_setup = bond_neigh_setup,
cc0e4070 3834 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
eb7cc59a 3835 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
f6dc31a8 3836#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 3837 .ndo_netpoll_setup = bond_netpoll_setup,
f6dc31a8
WC		 3838 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
3839 .ndo_poll_controller = bond_poll_controller,
3840#endif
9232ecca
JP		 3841 .ndo_add_slave = bond_enslave,
3842 .ndo_del_slave = bond_release,
b2a103e6 3843 .ndo_fix_features = bond_fix_features,
eb7cc59a
SH		 3844};
3845
b3f92b63
DG		 3846static const struct device_type bond_type = {
3847 .name = "bond",
3848};
3849
9e2e61fb
AW		 3850static void bond_destructor(struct net_device *bond_dev)
3851{
3852 struct bonding *bond = netdev_priv(bond_dev);
3853 if (bond->wq)
3854 destroy_workqueue(bond->wq);
3855 free_netdev(bond_dev);
3856}
3857
0a2a78c4 3858void bond_setup(struct net_device *bond_dev)
1da177e4 3859{
454d7c9b 3860 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3861
1da177e4
LT		 3862 /* initialize rwlocks */
3863 rwlock_init(&bond->lock);
3864 rwlock_init(&bond->curr_slave_lock);
d2991f75 3865 bond->params = bonding_defaults;
1da177e4
LT		 3866
3867 /* Initialize pointers */
1da177e4 3868 bond->dev = bond_dev;
1da177e4
LT		 3869
3870 /* Initialize the device entry points */
181470fc 3871 ether_setup(bond_dev);
eb7cc59a 3872 bond_dev->netdev_ops = &bond_netdev_ops;
8531c5ff 3873 bond_dev->ethtool_ops = &bond_ethtool_ops;
1da177e4 3874
9e2e61fb 3875 bond_dev->destructor = bond_destructor;
1da177e4 3876
b3f92b63
DG		 3877 SET_NETDEV_DEVTYPE(bond_dev, &bond_type);
3878
1da177e4
LT		 3879 /* Initialize the device options */
3880 bond_dev->tx_queue_len = 0;
3881 bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
0b680e75 3882 bond_dev->priv_flags |= IFF_BONDING;
550fd08c 3883 bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
181470fc 3884
1da177e4
LT		 3885 /* At first, we block adding VLANs. That's the only way to
3886 * prevent problems that occur when adding VLANs over an
3887 * empty bond. The block will be removed once non-challenged
3888 * slaves are enslaved.
3889 */
3890 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
3891
932ff279 3892 /* don't acquire bond device's netif_tx_lock when
1da177e4
LT		 3893 * transmitting */
3894 bond_dev->features |= NETIF_F_LLTX;
3895
3896 /* By default, we declare the bond to be fully
3897 * VLAN hardware accelerated capable. Special
3898 * care is taken in the various xmit functions
3899 * when there are slaves that are not hw accel
3900 * capable
3901 */
1da177e4 3902
f9399814
WC		 3903 /* Don't allow bond devices to change network namespaces. */
3904 bond_dev->features |= NETIF_F_NETNS_LOCAL;
3905
b2a103e6 3906 bond_dev->hw_features = BOND_VLAN_FEATURES |
f646968f
PM		 3907 NETIF_F_HW_VLAN_CTAG_TX |
3908 NETIF_F_HW_VLAN_CTAG_RX |
3909 NETIF_F_HW_VLAN_CTAG_FILTER;
b2a103e6 3910
34324dc2 3911 bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM);
b2a103e6 3912 bond_dev->features |= bond_dev->hw_features;
1da177e4
LT		 3913}
3914
c67dfb29
EB		 3915/*
3916* Destroy a bonding device.
3917* Must be under rtnl_lock when this function is called.
3918*/
3919static void bond_uninit(struct net_device *bond_dev)
a434e43f 3920{
454d7c9b 3921 struct bonding *bond = netdev_priv(bond_dev);
544a028e
VF		 3922 struct list_head *iter;
3923 struct slave *slave;
a434e43f 3924
f6dc31a8
WC		 3925 bond_netpoll_cleanup(bond_dev);
3926
c67dfb29 3927 /* Release the bonded slaves */
544a028e 3928 bond_for_each_slave(bond, slave, iter)
dec1e90e 3929 __bond_release_one(bond_dev, slave->dev, true);
0896341a 3930 pr_info("%s: released all slaves\n", bond_dev->name);
c67dfb29 3931
a434e43f
JV		 3932 list_del(&bond->bond_list);
3933
f073c7ca 3934 bond_debug_unregister(bond);
a434e43f
JV		 3935}
3936
1da177e4
LT		 3937/*------------------------- Module initialization ---------------------------*/
3938
ec029fac 3939int bond_parm_tbl_lookup(int mode, const struct bond_parm_tbl *tbl)
3940{
3941 int i;
3942
3943 for (i = 0; tbl[i].modename; i++)
3944 if (mode == tbl[i].mode)
3945 return tbl[i].mode;
3946
3947 return -1;
3948}
3949
3950static int bond_parm_tbl_lookup_name(const char *modename,
3951 const struct bond_parm_tbl *tbl)
3952{
3953 int i;
3954
3955 for (i = 0; tbl[i].modename; i++)
3956 if (strcmp(modename, tbl[i].modename) == 0)
3957 return tbl[i].mode;
3958
3959 return -1;
3960}
3961
1da177e4
LT		 3962/*
3963 * Convert string input module parms. Accept either the
ece95f7f
JV		 3964 * number of the mode or its string name. A bit complicated because
3965 * some mode names are substrings of other names, and calls from sysfs
3966 * may have whitespace in the name (trailing newlines, for example).
1da177e4 3967 */
325dcf7a 3968int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl)
1da177e4 3969{
ec029fac 3970 int modeint;
3971 char *p, modestr[BOND_MAX_MODENAME_LEN + 1];
ece95f7f 3972
a42e534f
JV		 3973 for (p = (char *)buf; *p; p++)
3974 if (!(isdigit(*p) || isspace(*p)))
3975 break;
3976
ec029fac 3977 if (*p && sscanf(buf, "%20s", modestr) != 0)
3978 return bond_parm_tbl_lookup_name(modestr, tbl);
3979 else if (sscanf(buf, "%d", &modeint) != 0)
3980 return bond_parm_tbl_lookup(modeint, tbl);
1da177e4
LT		 3981
3982 return -1;
3983}
3984
3985static int bond_check_params(struct bond_params *params)
3986{
5a5c5fd4 3987 int arp_validate_value, fail_over_mac_value, primary_reselect_value, i;
2b3798d5 3988 struct bond_opt_value newval, *valptr;
8599b52e 3989 int arp_all_targets_value;
f5b2b966 3990
1da177e4
LT		 3991 /*
3992 * Convert string parameters.
3993 */
3994 if (mode) {
2b3798d5
NA		 3995 bond_opt_initstr(&newval, mode);
3996 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval);
3997 if (!valptr) {
3998 pr_err("Error: Invalid bonding mode \"%s\"\n", mode);
1da177e4
LT		 3999 return -EINVAL;
4000 }
2b3798d5 4001 bond_mode = valptr->value;
1da177e4
LT		 4002 }
4003
169a3e66
JV		 4004 if (xmit_hash_policy) {
4005 if ((bond_mode != BOND_MODE_XOR) &&
4006 (bond_mode != BOND_MODE_8023AD)) {
a4aee5c8 4007 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
169a3e66
JV		 4008 bond_mode_name(bond_mode));
4009 } else {
a4b32ce7
NA		 4010 bond_opt_initstr(&newval, xmit_hash_policy);
4011 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH),
4012 &newval);
4013 if (!valptr) {
a4aee5c8 4014 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
169a3e66
JV		 4015 xmit_hash_policy);
4016 return -EINVAL;
4017 }
a4b32ce7 4018 xmit_hashtype = valptr->value;
169a3e66
JV		 4019 }
4020 }
4021
1da177e4
LT		 4022 if (lacp_rate) {
4023 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8
JP		 4024 pr_info("lacp_rate param is irrelevant in mode %s\n",
4025 bond_mode_name(bond_mode));
1da177e4 4026 } else {
d3131de7
NA		 4027 bond_opt_initstr(&newval, lacp_rate);
4028 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE),
4029 &newval);
4030 if (!valptr) {
a4aee5c8 4031 pr_err("Error: Invalid lacp rate \"%s\"\n",
d3131de7 4032 lacp_rate);
1da177e4
LT		 4033 return -EINVAL;
4034 }
d3131de7 4035 lacp_fast = valptr->value;
1da177e4
LT		 4036 }
4037 }
4038
fd989c83 4039 if (ad_select) {
9e5f5eeb
NA		 4040 bond_opt_initstr(&newval, lacp_rate);
4041 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT),
4042 &newval);
4043 if (!valptr) {
4044 pr_err("Error: Invalid ad_select \"%s\"\n", ad_select);
fd989c83
JV		 4045 return -EINVAL;
4046 }
9e5f5eeb
NA		 4047 params->ad_select = valptr->value;
4048 if (bond_mode != BOND_MODE_8023AD)
91565ebb 4049 pr_warn("ad_select param only affects 802.3ad mode\n");
fd989c83
JV		 4050 } else {
4051 params->ad_select = BOND_AD_STABLE;
4052 }
4053
f5841306 4054 if (max_bonds < 0) {
91565ebb
JP		 4055 pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4056 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
1da177e4
LT		 4057 max_bonds = BOND_DEFAULT_MAX_BONDS;
4058 }
4059
4060 if (miimon < 0) {
91565ebb
JP		 4061 pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4062 miimon, INT_MAX);
b98d9c66 4063 miimon = 0;
1da177e4
LT		 4064 }
4065
4066 if (updelay < 0) {
91565ebb
JP		 4067 pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4068 updelay, INT_MAX);
1da177e4
LT		 4069 updelay = 0;
4070 }
4071
4072 if (downdelay < 0) {
91565ebb
JP		 4073 pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4074 downdelay, INT_MAX);
1da177e4
LT		 4075 downdelay = 0;
4076 }
4077
4078 if ((use_carrier != 0) && (use_carrier != 1)) {
91565ebb
JP		 4079 pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4080 use_carrier);
1da177e4
LT		 4081 use_carrier = 1;
4082 }
4083
ad246c99 4084 if (num_peer_notif < 0 || num_peer_notif > 255) {
91565ebb
JP		 4085 pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4086 num_peer_notif);
ad246c99
BH		 4087 num_peer_notif = 1;
4088 }
4089
834db4bc 4090 /* reset values for 802.3ad/TLB/ALB */
4091 if (BOND_NO_USES_ARP(bond_mode)) {
1da177e4 4092 if (!miimon) {
91565ebb
JP		 4093 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");
4094 pr_warn("Forcing miimon to 100msec\n");
fe9d04af 4095 miimon = BOND_DEFAULT_MIIMON;
1da177e4
LT		 4096 }
4097 }
4098
bb1d9123 4099 if (tx_queues < 1 || tx_queues > 255) {
91565ebb
JP		 4100 pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n",
4101 tx_queues, BOND_DEFAULT_TX_QUEUES);
bb1d9123
AG		 4102 tx_queues = BOND_DEFAULT_TX_QUEUES;
4103 }
4104
ebd8e497 4105 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
91565ebb
JP		 4106 pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n",
4107 all_slaves_active);
ebd8e497
AG		 4108 all_slaves_active = 0;
4109 }
4110
c2952c31 4111 if (resend_igmp < 0 || resend_igmp > 255) {
91565ebb
JP		 4112 pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n",
4113 resend_igmp, BOND_DEFAULT_RESEND_IGMP);
c2952c31
FL		 4114 resend_igmp = BOND_DEFAULT_RESEND_IGMP;
4115 }
4116
aa59d851
NA		 4117 bond_opt_initval(&newval, packets_per_slave);
4118 if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) {
73958329
NA		 4119 pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n",
4120 packets_per_slave, USHRT_MAX);
4121 packets_per_slave = 1;
4122 }
4123
1da177e4 4124 if (bond_mode == BOND_MODE_ALB) {
a4aee5c8
JP		 4125 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",
4126 updelay);
1da177e4
LT		 4127 }
4128
4129 if (!miimon) {
4130 if (updelay || downdelay) {
4131 /* just warn the user the up/down delay will have
4132 * no effect since miimon is zero...
4133 */
91565ebb
JP		 4134 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",
4135 updelay, downdelay);
1da177e4
LT		 4136 }
4137 } else {
4138 /* don't allow arp monitoring */
4139 if (arp_interval) {
91565ebb
JP		 4140 pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4141 miimon, arp_interval);
1da177e4
LT		 4142 arp_interval = 0;
4143 }
4144
4145 if ((updelay % miimon) != 0) {
91565ebb
JP		 4146 pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4147 updelay, miimon, (updelay / miimon) * miimon);
1da177e4
LT		 4148 }
4149
4150 updelay /= miimon;
4151
4152 if ((downdelay % miimon) != 0) {
91565ebb
JP		 4153 pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4154 downdelay, miimon,
4155 (downdelay / miimon) * miimon);
1da177e4
LT		 4156 }
4157
4158 downdelay /= miimon;
4159 }
4160
4161 if (arp_interval < 0) {
91565ebb
JP		 4162 pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4163 arp_interval, INT_MAX);
7bdb04ed 4164 arp_interval = 0;
1da177e4
LT		 4165 }
4166
5a5c5fd4 4167 for (arp_ip_count = 0, i = 0;
4168 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) {
1da177e4
LT		 4169 /* not complete check, but should be good enough to
4170 catch mistakes */
89015c18 4171 __be32 ip;
4172 if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) ||
4173 IS_IP_TARGET_UNUSABLE_ADDRESS(ip)) {
91565ebb
JP		 4174 pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4175 arp_ip_target[i]);
1da177e4
LT		 4176 arp_interval = 0;
4177 } else {
0afee4e8
VF		 4178 if (bond_get_targets_ip(arp_target, ip) == -1)
4179 arp_target[arp_ip_count++] = ip;
4180 else
91565ebb
JP		 4181 pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n",
4182 &ip);
1da177e4
LT		 4183 }
4184 }
4185
4186 if (arp_interval && !arp_ip_count) {
4187 /* don't allow arping if no arp_ip_target given... */
91565ebb
JP		 4188 pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4189 arp_interval);
1da177e4
LT		 4190 arp_interval = 0;
4191 }
4192
f5b2b966
JV		 4193 if (arp_validate) {
4194 if (bond_mode != BOND_MODE_ACTIVEBACKUP) {
a4aee5c8 4195 pr_err("arp_validate only supported in active-backup mode\n");
f5b2b966
JV		 4196 return -EINVAL;
4197 }
4198 if (!arp_interval) {
a4aee5c8 4199 pr_err("arp_validate requires arp_interval\n");
f5b2b966
JV		 4200 return -EINVAL;
4201 }
4202
16228881
NA		 4203 bond_opt_initstr(&newval, arp_validate);
4204 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE),
4205 &newval);
4206 if (!valptr) {
a4aee5c8 4207 pr_err("Error: invalid arp_validate \"%s\"\n",
16228881 4208 arp_validate);
f5b2b966
JV		 4209 return -EINVAL;
4210 }
16228881
NA		 4211 arp_validate_value = valptr->value;
4212 } else {
f5b2b966 4213 arp_validate_value = 0;
16228881 4214 }
f5b2b966 4215
8599b52e
VF		 4216 arp_all_targets_value = 0;
4217 if (arp_all_targets) {
edf36b24
NA		 4218 bond_opt_initstr(&newval, arp_all_targets);
4219 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS),
4220 &newval);
4221 if (!valptr) {
8599b52e
VF		 4222 pr_err("Error: invalid arp_all_targets_value \"%s\"\n",
4223 arp_all_targets);
4224 arp_all_targets_value = 0;
edf36b24
NA		 4225 } else {
4226 arp_all_targets_value = valptr->value;
8599b52e
VF		 4227 }
4228 }
4229
1da177e4 4230 if (miimon) {
a4aee5c8 4231 pr_info("MII link monitoring set to %d ms\n", miimon);
1da177e4 4232 } else if (arp_interval) {
16228881
NA		 4233 valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE,
4234 arp_validate_value);
a4aee5c8 4235 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
16228881 4236 arp_interval, valptr->string, arp_ip_count);
1da177e4
LT		 4237
4238 for (i = 0; i < arp_ip_count; i++)
e5e2a8fd 4239 pr_info(" %s", arp_ip_target[i]);
1da177e4 4240
e5e2a8fd 4241 pr_info("\n");
1da177e4 4242
b8a9787e 4243 } else if (max_bonds) {
1da177e4
LT		 4244 /* miimon and arp_interval not set, we need one so things
4245 * work as expected, see bonding.txt for details
4246 */
b2730f4f 4247 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		 4248 }
4249
4250 if (primary && !USES_PRIMARY(bond_mode)) {
4251 /* currently, using a primary only makes sense
4252 * in active backup, TLB or ALB modes
4253 */
91565ebb
JP		 4254 pr_warn("Warning: %s primary device specified but has no effect in %s mode\n",
4255 primary, bond_mode_name(bond_mode));
1da177e4
LT		 4256 primary = NULL;
4257 }
4258
a549952a 4259 if (primary && primary_reselect) {
388d3a6d
NA		 4260 bond_opt_initstr(&newval, primary_reselect);
4261 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT),
4262 &newval);
4263 if (!valptr) {
a4aee5c8 4264 pr_err("Error: Invalid primary_reselect \"%s\"\n",
388d3a6d 4265 primary_reselect);
a549952a
JP		 4266 return -EINVAL;
4267 }
388d3a6d 4268 primary_reselect_value = valptr->value;
a549952a
JP		 4269 } else {
4270 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4271 }
4272
3915c1e8 4273 if (fail_over_mac) {
1df6b6aa
NA		 4274 bond_opt_initstr(&newval, fail_over_mac);
4275 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC),
4276 &newval);
4277 if (!valptr) {
a4aee5c8 4278 pr_err("Error: invalid fail_over_mac \"%s\"\n",
1df6b6aa 4279 fail_over_mac);
3915c1e8
JV		 4280 return -EINVAL;
4281 }
1df6b6aa 4282 fail_over_mac_value = valptr->value;
3915c1e8 4283 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
91565ebb 4284 pr_warn("Warning: fail_over_mac only affects active-backup mode\n");
3915c1e8
JV		 4285 } else {
4286 fail_over_mac_value = BOND_FOM_NONE;
4287 }
dd957c57 4288
3a7129e5 4289 if (lp_interval == 0) {
91565ebb
JP		 4290 pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n",
4291 INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL);
3a7129e5 4292 lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
4293 }
4294
1da177e4
LT		 4295 /* fill params struct with the proper values */
4296 params->mode = bond_mode;
169a3e66 4297 params->xmit_policy = xmit_hashtype;
1da177e4 4298 params->miimon = miimon;
ad246c99 4299 params->num_peer_notif = num_peer_notif;
1da177e4 4300 params->arp_interval = arp_interval;
f5b2b966 4301 params->arp_validate = arp_validate_value;
8599b52e 4302 params->arp_all_targets = arp_all_targets_value;
1da177e4
LT		 4303 params->updelay = updelay;
4304 params->downdelay = downdelay;
4305 params->use_carrier = use_carrier;
4306 params->lacp_fast = lacp_fast;
4307 params->primary[0] = 0;
a549952a 4308 params->primary_reselect = primary_reselect_value;
3915c1e8 4309 params->fail_over_mac = fail_over_mac_value;
bb1d9123 4310 params->tx_queues = tx_queues;
ebd8e497 4311 params->all_slaves_active = all_slaves_active;
c2952c31 4312 params->resend_igmp = resend_igmp;
655f8919 4313 params->min_links = min_links;
3a7129e5 4314 params->lp_interval = lp_interval;
809fa972
HFS		 4315 params->packets_per_slave = packets_per_slave;
4316 if (packets_per_slave > 0) {
4317 params->reciprocal_packets_per_slave =
4318 reciprocal_value(packets_per_slave);
4319 } else {
4320 /* reciprocal_packets_per_slave is unused if
4321 * packets_per_slave is 0 or 1, just initialize it
4322 */
4323 params->reciprocal_packets_per_slave =
4324 (struct reciprocal_value) { 0 };
4325 }
4326
1da177e4
LT		 4327 if (primary) {
4328 strncpy(params->primary, primary, IFNAMSIZ);
4329 params->primary[IFNAMSIZ - 1] = 0;
4330 }
4331
4332 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4333
4334 return 0;
4335}
4336
0daa2303 4337static struct lock_class_key bonding_netdev_xmit_lock_key;
cf508b12 4338static struct lock_class_key bonding_netdev_addr_lock_key;
49ee4920 4339static struct lock_class_key bonding_tx_busylock_key;
0daa2303 4340
e8a0464c
DM		 4341static void bond_set_lockdep_class_one(struct net_device *dev,
4342 struct netdev_queue *txq,
4343 void *_unused)
c773e847
DM		 4344{
4345 lockdep_set_class(&txq->_xmit_lock,
4346 &bonding_netdev_xmit_lock_key);
4347}
4348
4349static void bond_set_lockdep_class(struct net_device *dev)
4350{
cf508b12
DM		 4351 lockdep_set_class(&dev->addr_list_lock,
4352 &bonding_netdev_addr_lock_key);
e8a0464c 4353 netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL);
49ee4920 4354 dev->qdisc_tx_busylock = &bonding_tx_busylock_key;
c773e847
DM		 4355}
4356
181470fc
SH		 4357/*
4358 * Called from registration process
4359 */
4360static int bond_init(struct net_device *bond_dev)
4361{
4362 struct bonding *bond = netdev_priv(bond_dev);
ec87fd3b 4363 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
9fe0617d 4364 struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond));
181470fc
SH		 4365
4366 pr_debug("Begin bond_init for %s\n", bond_dev->name);
4367
9fe0617d
NH		 4368 /*
4369 * Initialize locks that may be required during
4370 * en/deslave operations. All of the bond_open work
4371 * (of which this is part) should really be moved to
4372 * a phase prior to dev_open
4373 */
4374 spin_lock_init(&(bond_info->tx_hashtbl_lock));
4375 spin_lock_init(&(bond_info->rx_hashtbl_lock));
4376
181470fc
SH		 4377 bond->wq = create_singlethread_workqueue(bond_dev->name);
4378 if (!bond->wq)
4379 return -ENOMEM;
4380
4381 bond_set_lockdep_class(bond_dev);
4382
ec87fd3b 4383 list_add_tail(&bond->bond_list, &bn->dev_list);
181470fc 4384
6151b3d4 4385 bond_prepare_sysfs_group(bond);
22bedad3 4386
f073c7ca
TI		 4387 bond_debug_register(bond);
4388
409cc1f8
JP		 4389 /* Ensure valid dev_addr */
4390 if (is_zero_ether_addr(bond_dev->dev_addr) &&
97a1e639 4391 bond_dev->addr_assign_type == NET_ADDR_PERM)
409cc1f8 4392 eth_hw_addr_random(bond_dev);
409cc1f8 4393
181470fc
SH		 4394 return 0;
4395}
4396
0a2a78c4 4397unsigned int bond_get_num_tx_queues(void)
d5da4510 4398{
efacb309 4399 return tx_queues;
d5da4510
JP		 4400}
4401
dfe60397 4402/* Create a new bond based on the specified name and bonding parameters.
e4b91c48 4403 * If name is NULL, obtain a suitable "bond%d" name for us.
dfe60397
MW		 4404 * Caller must NOT hold rtnl_lock; we need to release it here before we
4405 * set up our sysfs entries.
4406 */
ec87fd3b 4407int bond_create(struct net *net, const char *name)
dfe60397
MW		 4408{
4409 struct net_device *bond_dev;
4410 int res;
4411
4412 rtnl_lock();
027ea041 4413
1c5cae81
JP		 4414 bond_dev = alloc_netdev_mq(sizeof(struct bonding),
4415 name ? name : "bond%d",
4416 bond_setup, tx_queues);
dfe60397 4417 if (!bond_dev) {
a4aee5c8 4418 pr_err("%s: eek! can't alloc netdev!\n", name);
9e2e61fb
AW		 4419 rtnl_unlock();
4420 return -ENOMEM;
dfe60397
MW		 4421 }
4422
ec87fd3b 4423 dev_net_set(bond_dev, net);
88ead977
EB		 4424 bond_dev->rtnl_link_ops = &bond_link_ops;
4425
dfe60397 4426 res = register_netdevice(bond_dev);
0daa2303 4427
e826eafa
PO		 4428 netif_carrier_off(bond_dev);
4429
7e083840 4430 rtnl_unlock();
9e2e61fb
AW		 4431 if (res < 0)
4432 bond_destructor(bond_dev);
30c15ba9 4433 return res;
dfe60397
MW		 4434}
4435
2c8c1e72 4436static int __net_init bond_net_init(struct net *net)
ec87fd3b 4437{
15449745 4438 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 4439
4440 bn->net = net;
4441 INIT_LIST_HEAD(&bn->dev_list);
4442
ec87fd3b 4443 bond_create_proc_dir(bn);
4c22400a 4444 bond_create_sysfs(bn);
87a7b84b 4445
15449745 4446 return 0;
ec87fd3b
EB		 4447}
4448
2c8c1e72 4449static void __net_exit bond_net_exit(struct net *net)
ec87fd3b 4450{
15449745 4451 struct bond_net *bn = net_generic(net, bond_net_id);
69b0216a 4452 struct bonding *bond, *tmp_bond;
4453 LIST_HEAD(list);
ec87fd3b 4454
4c22400a 4455 bond_destroy_sysfs(bn);
ec87fd3b 4456 bond_destroy_proc_dir(bn);
69b0216a 4457
4458 /* Kill off any bonds created after unregistering bond rtnl ops */
4459 rtnl_lock();
4460 list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list)
4461 unregister_netdevice_queue(bond->dev, &list);
4462 unregister_netdevice_many(&list);
4463 rtnl_unlock();
ec87fd3b
EB		 4464}
4465
4466static struct pernet_operations bond_net_ops = {
4467 .init = bond_net_init,
4468 .exit = bond_net_exit,
15449745
EB		 4469 .id = &bond_net_id,
4470 .size = sizeof(struct bond_net),
ec87fd3b
EB		 4471};
4472
1da177e4
LT		 4473static int __init bonding_init(void)
4474{
1da177e4
LT		 4475 int i;
4476 int res;
4477
bd33acc3 4478 pr_info("%s", bond_version);
1da177e4 4479
dfe60397 4480 res = bond_check_params(&bonding_defaults);
3d632c3f 4481 if (res)
dfe60397 4482 goto out;
1da177e4 4483
15449745 4484 res = register_pernet_subsys(&bond_net_ops);
ec87fd3b
EB		 4485 if (res)
4486 goto out;
027ea041 4487
0a2a78c4 4488 res = bond_netlink_init();
88ead977 4489 if (res)
6639104b 4490 goto err_link;
88ead977 4491
f073c7ca
TI		 4492 bond_create_debugfs();
4493
1da177e4 4494 for (i = 0; i < max_bonds; i++) {
ec87fd3b 4495 res = bond_create(&init_net, NULL);
dfe60397
MW		 4496 if (res)
4497 goto err;
1da177e4
LT		 4498 }
4499
1da177e4 4500 register_netdevice_notifier(&bond_netdev_notifier);
dfe60397 4501out:
1da177e4 4502 return res;
88ead977 4503err:
0a2a78c4 4504 bond_netlink_fini();
6639104b 4505err_link:
15449745 4506 unregister_pernet_subsys(&bond_net_ops);
88ead977 4507 goto out;
dfe60397 4508
1da177e4
LT		 4509}
4510
4511static void __exit bonding_exit(void)
4512{
4513 unregister_netdevice_notifier(&bond_netdev_notifier);
4514
f073c7ca 4515 bond_destroy_debugfs();
ae68c398 4516
0a2a78c4 4517 bond_netlink_fini();
ffcdedb6 4518 unregister_pernet_subsys(&bond_net_ops);
e843fa50
NH		 4519
4520#ifdef CONFIG_NET_POLL_CONTROLLER
fb4fa76a
NH		 4521 /*
4522 * Make sure we don't have an imbalance on our netpoll blocking
4523 */
4524 WARN_ON(atomic_read(&netpoll_block_tx));
e843fa50 4525#endif
1da177e4
LT		 4526}
4527
4528module_init(bonding_init);
4529module_exit(bonding_exit);
4530MODULE_LICENSE("GPL");
4531MODULE_VERSION(DRV_VERSION);
4532MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
4533MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
Linux kernel - Deliverables
This page took 2.791869 seconds and 5 git commands to generate.