projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
bonding: use rcu_access_pointer() in bonding_show_mii_status()
[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
01844098 346 if (BOND_MODE(bond) == BOND_MODE_8023AD)
ff59c456
JV		 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
ec0865a9 500 if (bond_uses_primary(bond)) {
1da177e4
LT		 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
ec0865a9 526 if (bond_uses_primary(bond)) {
1da177e4
LT		 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 576
01844098 577 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1da177e4
LT		 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 587/* Update the hardware address list and promisc/allmulti for the new and
ec0865a9
VF		 588 * old active slaves (if any). Modes that are not using primary keep all
589 * slaves up date at all times; only the modes that use primary need to call
303d1cbf 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) {
ada0f863 676 ether_addr_copy(tmp_mac, new_active->dev->dev_addr);
2a7c183b
JP		 677 ether_addr_copy(saddr.sa_data,
678 old_active->dev->dev_addr);
3915c1e8
JV		 679 saddr.sa_family = new_active->dev->type;
680 } else {
ada0f863 681 ether_addr_copy(saddr.sa_data, bond->dev->dev_addr);
3915c1e8
JV		 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
ada0f863 695 ether_addr_copy(saddr.sa_data, tmp_mac);
3915c1e8
JV		 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;
b6adc610 750 if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) &&
77140d29
VF		 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) {
8e603460 801 new_active->last_link_up = jiffies;
b2220cad 802
1da177e4 803 if (new_active->link == BOND_LINK_BACK) {
ec0865a9 804 if (bond_uses_primary(bond)) {
90194264 805 pr_info("%s: making interface %s the new active one %d ms earlier\n",
a4aee5c8
JP		 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
01844098 813 if (BOND_MODE(bond) == 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 818 } else {
ec0865a9 819 if (bond_uses_primary(bond)) {
90194264 820 pr_info("%s: making interface %s the new active one\n",
a4aee5c8 821 bond->dev->name, new_active->dev->name);
1da177e4
LT		 822 }
823 }
824 }
825
ec0865a9 826 if (bond_uses_primary(bond))
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 831 if (old_active)
5e5b0665 832 bond_set_slave_inactive_flags(old_active,
833 BOND_SLAVE_NOTIFY_NOW);
8f903c70 834 if (new_active)
5e5b0665 835 bond_set_slave_active_flags(new_active,
836 BOND_SLAVE_NOTIFY_NOW);
1da177e4 837 } else {
278b2083 838 rcu_assign_pointer(bond->curr_active_slave, new_active);
1da177e4 839 }
c3ade5ca 840
01844098 841 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) {
3d632c3f 842 if (old_active)
5e5b0665 843 bond_set_slave_inactive_flags(old_active,
844 BOND_SLAVE_NOTIFY_NOW);
c3ade5ca
JV		 845
846 if (new_active) {
ad246c99
BH		 847 bool should_notify_peers = false;
848
5e5b0665 849 bond_set_slave_active_flags(new_active,
850 BOND_SLAVE_NOTIFY_NOW);
2ab82852 851
709f8a45
OG		 852 if (bond->params.fail_over_mac)
853 bond_do_fail_over_mac(bond, new_active,
854 old_active);
3915c1e8 855
ad246c99
BH		 856 if (netif_running(bond->dev)) {
857 bond->send_peer_notif =
858 bond->params.num_peer_notif;
859 should_notify_peers =
860 bond_should_notify_peers(bond);
861 }
862
01f3109d 863 write_unlock_bh(&bond->curr_slave_lock);
01f3109d 864
b7bc2a5b 865 call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev);
ad246c99 866 if (should_notify_peers)
b7bc2a5b
AW		 867 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
868 bond->dev);
01f3109d 869
01f3109d 870 write_lock_bh(&bond->curr_slave_lock);
7893b249 871 }
c3ade5ca 872 }
a2fd940f 873
5a37e8ca 874 /* resend IGMP joins since active slave has changed or
94265cf5
FL		 875 * all were sent on curr_active_slave.
876 * resend only if bond is brought up with the affected
877 * bonding modes and the retransmission is enabled */
878 if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) &&
ec0865a9 879 ((bond_uses_primary(bond) && new_active) ||
01844098 880 BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) {
c2952c31 881 bond->igmp_retrans = bond->params.resend_igmp;
4beac029 882 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
a2fd940f 883 }
1da177e4
LT		 884}
885
886/**
887 * bond_select_active_slave - select a new active slave, if needed
888 * @bond: our bonding struct
889 *
3d632c3f 890 * This functions should be called when one of the following occurs:
1da177e4
LT		 891 * - The old curr_active_slave has been released or lost its link.
892 * - The primary_slave has got its link back.
893 * - A slave has got its link back and there's no old curr_active_slave.
894 *
b2e7aceb 895 * Caller must hold curr_slave_lock for write_bh.
1da177e4 896 */
a77b5325 897void bond_select_active_slave(struct bonding *bond)
1da177e4
LT		 898{
899 struct slave *best_slave;
ff59c456 900 int rv;
1da177e4
LT		 901
902 best_slave = bond_find_best_slave(bond);
903 if (best_slave != bond->curr_active_slave) {
904 bond_change_active_slave(bond, best_slave);
ff59c456
JV		 905 rv = bond_set_carrier(bond);
906 if (!rv)
907 return;
908
909 if (netif_carrier_ok(bond->dev)) {
a4aee5c8
JP		 910 pr_info("%s: first active interface up!\n",
911 bond->dev->name);
ff59c456 912 } else {
90194264 913 pr_info("%s: now running without any active interface!\n",
a4aee5c8 914 bond->dev->name);
ff59c456 915 }
1da177e4
LT		 916 }
917}
918
f6dc31a8 919#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 920static inline int slave_enable_netpoll(struct slave *slave)
f6dc31a8 921{
8a8efa22
AW		 922 struct netpoll *np;
923 int err = 0;
f6dc31a8 924
a8779ec1 925 np = kzalloc(sizeof(*np), GFP_KERNEL);
8a8efa22
AW		 926 err = -ENOMEM;
927 if (!np)
928 goto out;
929
a8779ec1 930 err = __netpoll_setup(np, slave->dev);
8a8efa22
AW		 931 if (err) {
932 kfree(np);
933 goto out;
f6dc31a8 934 }
8a8efa22
AW		 935 slave->np = np;
936out:
937 return err;
938}
939static inline void slave_disable_netpoll(struct slave *slave)
940{
941 struct netpoll *np = slave->np;
942
943 if (!np)
944 return;
945
946 slave->np = NULL;
2cde6acd 947 __netpoll_free_async(np);
8a8efa22 948}
f6dc31a8
WC		 949
950static void bond_poll_controller(struct net_device *bond_dev)
951{
8a8efa22
AW		 952}
953
c4cdef9b 954static void bond_netpoll_cleanup(struct net_device *bond_dev)
8a8efa22 955{
c4cdef9b 956 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 957 struct list_head *iter;
c2355e1a 958 struct slave *slave;
c2355e1a 959
9caff1e7 960 bond_for_each_slave(bond, slave, iter)
b6adc610 961 if (bond_slave_is_up(slave))
8a8efa22 962 slave_disable_netpoll(slave);
f6dc31a8 963}
8a8efa22 964
a8779ec1 965static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni)
8a8efa22
AW		 966{
967 struct bonding *bond = netdev_priv(dev);
9caff1e7 968 struct list_head *iter;
f6dc31a8 969 struct slave *slave;
dec1e90e 970 int err = 0;
f6dc31a8 971
9caff1e7 972 bond_for_each_slave(bond, slave, iter) {
8a8efa22
AW		 973 err = slave_enable_netpoll(slave);
974 if (err) {
c4cdef9b 975 bond_netpoll_cleanup(dev);
8a8efa22 976 break;
f6dc31a8
WC		 977 }
978 }
8a8efa22 979 return err;
f6dc31a8 980}
8a8efa22
AW		 981#else
982static inline int slave_enable_netpoll(struct slave *slave)
983{
984 return 0;
985}
986static inline void slave_disable_netpoll(struct slave *slave)
987{
988}
f6dc31a8
WC		 989static void bond_netpoll_cleanup(struct net_device *bond_dev)
990{
991}
f6dc31a8
WC		 992#endif
993
1da177e4
LT		 994/*---------------------------------- IOCTL ----------------------------------*/
995
c8f44aff 996static netdev_features_t bond_fix_features(struct net_device *dev,
9b7b165a 997 netdev_features_t features)
8531c5ff 998{
b2a103e6 999 struct bonding *bond = netdev_priv(dev);
9caff1e7 1000 struct list_head *iter;
c8f44aff 1001 netdev_features_t mask;
9b7b165a 1002 struct slave *slave;
b63365a2 1003
b2a103e6 1004 mask = features;
b63365a2 1005 features &= ~NETIF_F_ONE_FOR_ALL;
b2a103e6 1006 features |= NETIF_F_ALL_FOR_ALL;
7f353bf2 1007
9caff1e7 1008 bond_for_each_slave(bond, slave, iter) {
b63365a2
HX		 1009 features = netdev_increment_features(features,
1010 slave->dev->features,
b2a103e6
MM		 1011 mask);
1012 }
b0ce3508 1013 features = netdev_add_tso_features(features, mask);
b2a103e6 1014
b2a103e6
MM		 1015 return features;
1016}
1017
62f2a3a4
MM		 1018#define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \
1019 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
1020 NETIF_F_HIGHDMA | NETIF_F_LRO)
b2a103e6 1021
5a7baa78
OG		 1022#define BOND_ENC_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | NETIF_F_RXCSUM |\
1023 NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL)
1024
b2a103e6
MM		 1025static void bond_compute_features(struct bonding *bond)
1026{
c4826861 1027 unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE;
c8f44aff 1028 netdev_features_t vlan_features = BOND_VLAN_FEATURES;
5a7baa78 1029 netdev_features_t enc_features = BOND_ENC_FEATURES;
9caff1e7
VF		 1030 struct net_device *bond_dev = bond->dev;
1031 struct list_head *iter;
1032 struct slave *slave;
b2a103e6 1033 unsigned short max_hard_header_len = ETH_HLEN;
0e376bd0
SB		 1034 unsigned int gso_max_size = GSO_MAX_SIZE;
1035 u16 gso_max_segs = GSO_MAX_SEGS;
b2a103e6 1036
0965a1f3 1037 if (!bond_has_slaves(bond))
b2a103e6 1038 goto done;
a9b3ace4 1039 vlan_features &= NETIF_F_ALL_FOR_ALL;
b2a103e6 1040
9caff1e7 1041 bond_for_each_slave(bond, slave, iter) {
278339a4 1042 vlan_features = netdev_increment_features(vlan_features,
b2a103e6
MM		 1043 slave->dev->vlan_features, BOND_VLAN_FEATURES);
1044
5a7baa78
OG		 1045 enc_features = netdev_increment_features(enc_features,
1046 slave->dev->hw_enc_features,
1047 BOND_ENC_FEATURES);
b6fe83e9 1048 dst_release_flag &= slave->dev->priv_flags;
54ef3137
JV		 1049 if (slave->dev->hard_header_len > max_hard_header_len)
1050 max_hard_header_len = slave->dev->hard_header_len;
0e376bd0
SB		 1051
1052 gso_max_size = min(gso_max_size, slave->dev->gso_max_size);
1053 gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs);
54ef3137 1054 }
8531c5ff 1055
b63365a2 1056done:
b2a103e6 1057 bond_dev->vlan_features = vlan_features;
5a7baa78 1058 bond_dev->hw_enc_features = enc_features;
54ef3137 1059 bond_dev->hard_header_len = max_hard_header_len;
0e376bd0
SB		 1060 bond_dev->gso_max_segs = gso_max_segs;
1061 netif_set_gso_max_size(bond_dev, gso_max_size);
8531c5ff 1062
b6fe83e9
ED		 1063 flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE;
1064 bond_dev->priv_flags = flags | dst_release_flag;
1065
b2a103e6 1066 netdev_change_features(bond_dev);
8531c5ff
AK		 1067}
1068
872254dd
MS		 1069static void bond_setup_by_slave(struct net_device *bond_dev,
1070 struct net_device *slave_dev)
1071{
00829823 1072 bond_dev->header_ops = slave_dev->header_ops;
872254dd
MS		 1073
1074 bond_dev->type = slave_dev->type;
1075 bond_dev->hard_header_len = slave_dev->hard_header_len;
1076 bond_dev->addr_len = slave_dev->addr_len;
1077
1078 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1079 slave_dev->addr_len);
1080}
1081
5b2c4dd2 1082/* On bonding slaves other than the currently active slave, suppress
3aba891d 1083 * duplicates except for alb non-mcast/bcast.
5b2c4dd2
JP		 1084 */
1085static bool bond_should_deliver_exact_match(struct sk_buff *skb,
0bd80dad
JP		 1086 struct slave *slave,
1087 struct bonding *bond)
5b2c4dd2 1088{
2d7011ca 1089 if (bond_is_slave_inactive(slave)) {
01844098 1090 if (BOND_MODE(bond) == BOND_MODE_ALB &&
5b2c4dd2
JP		 1091 skb->pkt_type != PACKET_BROADCAST &&
1092 skb->pkt_type != PACKET_MULTICAST)
5b2c4dd2 1093 return false;
5b2c4dd2
JP		 1094 return true;
1095 }
1096 return false;
1097}
1098
8a4eb573 1099static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
5b2c4dd2 1100{
8a4eb573 1101 struct sk_buff *skb = *pskb;
f1c1775a 1102 struct slave *slave;
0bd80dad 1103 struct bonding *bond;
de063b70
ED		 1104 int (*recv_probe)(const struct sk_buff *, struct bonding *,
1105 struct slave *);
13a8e0c8 1106 int ret = RX_HANDLER_ANOTHER;
5b2c4dd2 1107
8a4eb573
JP		 1108 skb = skb_share_check(skb, GFP_ATOMIC);
1109 if (unlikely(!skb))
1110 return RX_HANDLER_CONSUMED;
1111
1112 *pskb = skb;
5b2c4dd2 1113
35d48903
JP		 1114 slave = bond_slave_get_rcu(skb->dev);
1115 bond = slave->bond;
0bd80dad 1116
4d97480b
MH		 1117 recv_probe = ACCESS_ONCE(bond->recv_probe);
1118 if (recv_probe) {
de063b70
ED		 1119 ret = recv_probe(skb, bond, slave);
1120 if (ret == RX_HANDLER_CONSUMED) {
1121 consume_skb(skb);
1122 return ret;
3aba891d
JP		 1123 }
1124 }
1125
0bd80dad 1126 if (bond_should_deliver_exact_match(skb, slave, bond)) {
8a4eb573 1127 return RX_HANDLER_EXACT;
5b2c4dd2
JP		 1128 }
1129
35d48903 1130 skb->dev = bond->dev;
5b2c4dd2 1131
01844098 1132 if (BOND_MODE(bond) == BOND_MODE_ALB &&
35d48903 1133 bond->dev->priv_flags & IFF_BRIDGE_PORT &&
5b2c4dd2 1134 skb->pkt_type == PACKET_HOST) {
5b2c4dd2 1135
541ac7c9
CG		 1136 if (unlikely(skb_cow_head(skb,
1137 skb->data - skb_mac_header(skb)))) {
1138 kfree_skb(skb);
8a4eb573 1139 return RX_HANDLER_CONSUMED;
541ac7c9 1140 }
2a7c183b 1141 ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr);
5b2c4dd2
JP		 1142 }
1143
13a8e0c8 1144 return ret;
5b2c4dd2
JP		 1145}
1146
471cb5a3 1147static int bond_master_upper_dev_link(struct net_device *bond_dev,
1f718f0f
VF		 1148 struct net_device *slave_dev,
1149 struct slave *slave)
471cb5a3
JP		 1150{
1151 int err;
1152
1f718f0f 1153 err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave);
471cb5a3
JP		 1154 if (err)
1155 return err;
1156 slave_dev->flags |= IFF_SLAVE;
7f294054 1157 rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP		 1158 return 0;
1159}
1160
1161static void bond_upper_dev_unlink(struct net_device *bond_dev,
1162 struct net_device *slave_dev)
1163{
1164 netdev_upper_dev_unlink(slave_dev, bond_dev);
1165 slave_dev->flags &= ~IFF_SLAVE;
7f294054 1166 rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP		 1167}
1168
3fdddd85 1169static struct slave *bond_alloc_slave(struct bonding *bond)
1170{
1171 struct slave *slave = NULL;
1172
1173 slave = kzalloc(sizeof(struct slave), GFP_KERNEL);
1174 if (!slave)
1175 return NULL;
1176
01844098 1177 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
3fdddd85 1178 SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info),
1179 GFP_KERNEL);
1180 if (!SLAVE_AD_INFO(slave)) {
1181 kfree(slave);
1182 return NULL;
1183 }
1184 }
1185 return slave;
1186}
1187
1188static void bond_free_slave(struct slave *slave)
1189{
1190 struct bonding *bond = bond_get_bond_by_slave(slave);
1191
01844098 1192 if (BOND_MODE(bond) == BOND_MODE_8023AD)
3fdddd85 1193 kfree(SLAVE_AD_INFO(slave));
1194
1195 kfree(slave);
1196}
1197
1da177e4 1198/* enslave device <slave> to bond device <master> */
a77b5325 1199int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1200{
454d7c9b 1201 struct bonding *bond = netdev_priv(bond_dev);
eb7cc59a 1202 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
c8c23903 1203 struct slave *new_slave = NULL, *prev_slave;
1da177e4
LT		 1204 struct sockaddr addr;
1205 int link_reporting;
8599b52e 1206 int res = 0, i;
1da177e4 1207
c772dde3
BH		 1208 if (!bond->params.use_carrier &&
1209 slave_dev->ethtool_ops->get_link == NULL &&
1210 slave_ops->ndo_do_ioctl == NULL) {
91565ebb
JP		 1211 pr_warn("%s: Warning: no link monitoring support for %s\n",
1212 bond_dev->name, slave_dev->name);
1da177e4
LT		 1213 }
1214
1da177e4
LT		 1215 /* already enslaved */
1216 if (slave_dev->flags & IFF_SLAVE) {
90194264 1217 pr_debug("Error: Device was already enslaved\n");
1da177e4
LT		 1218 return -EBUSY;
1219 }
1220
09a89c21
JB		 1221 if (bond_dev == slave_dev) {
1222 pr_err("%s: cannot enslave bond to itself.\n", bond_dev->name);
1223 return -EPERM;
1224 }
1225
1da177e4
LT		 1226 /* vlan challenged mutual exclusion */
1227 /* no need to lock since we're protected by rtnl_lock */
1228 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 1229 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
55462cf3 1230 if (vlan_uses_dev(bond_dev)) {
a4aee5c8
JP		 1231 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1232 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1233 return -EPERM;
1234 } else {
91565ebb
JP		 1235 pr_warn("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1236 bond_dev->name, slave_dev->name,
1237 slave_dev->name, bond_dev->name);
1da177e4
LT		 1238 }
1239 } else {
5a03cdb7 1240 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4
LT		 1241 }
1242
217df670
JV		 1243 /*
1244 * Old ifenslave binaries are no longer supported. These can
3d632c3f 1245 * be identified with moderate accuracy by the state of the slave:
217df670
JV		 1246 * the current ifenslave will set the interface down prior to
1247 * enslaving it; the old ifenslave will not.
1248 */
1249 if ((slave_dev->flags & IFF_UP)) {
90194264 1250 pr_err("%s is up - this may be due to an out of date ifenslave\n",
217df670
JV		 1251 slave_dev->name);
1252 res = -EPERM;
1253 goto err_undo_flags;
1254 }
1da177e4 1255
872254dd
MS		 1256 /* set bonding device ether type by slave - bonding netdevices are
1257 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1258 * there is a need to override some of the type dependent attribs/funcs.
1259 *
1260 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1261 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1262 */
0965a1f3 1263 if (!bond_has_slaves(bond)) {
e36b9d16 1264 if (bond_dev->type != slave_dev->type) {
e36b9d16 1265 pr_debug("%s: change device type from %d to %d\n",
a4aee5c8
JP		 1266 bond_dev->name,
1267 bond_dev->type, slave_dev->type);
75c78500 1268
b7bc2a5b
AW		 1269 res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE,
1270 bond_dev);
3ca5b404
JP		 1271 res = notifier_to_errno(res);
1272 if (res) {
1273 pr_err("%s: refused to change device type\n",
1274 bond_dev->name);
1275 res = -EBUSY;
1276 goto err_undo_flags;
1277 }
75c78500 1278
32a806c1 1279 /* Flush unicast and multicast addresses */
a748ee24 1280 dev_uc_flush(bond_dev);
22bedad3 1281 dev_mc_flush(bond_dev);
32a806c1 1282
e36b9d16
MS		 1283 if (slave_dev->type != ARPHRD_ETHER)
1284 bond_setup_by_slave(bond_dev, slave_dev);
550fd08c 1285 else {
e36b9d16 1286 ether_setup(bond_dev);
550fd08c
NH		 1287 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1288 }
75c78500 1289
b7bc2a5b
AW		 1290 call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE,
1291 bond_dev);
e36b9d16 1292 }
872254dd 1293 } else if (bond_dev->type != slave_dev->type) {
90194264
JP		 1294 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it\n",
1295 slave_dev->name, slave_dev->type, bond_dev->type);
a4aee5c8
JP		 1296 res = -EINVAL;
1297 goto err_undo_flags;
872254dd
MS		 1298 }
1299
eb7cc59a 1300 if (slave_ops->ndo_set_mac_address == NULL) {
0965a1f3 1301 if (!bond_has_slaves(bond)) {
90194264 1302 pr_warn("%s: Warning: The first slave device specified does not support setting the MAC address\n",
00503b6f 1303 bond_dev->name);
01844098 1304 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) {
00503b6f 1305 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
90194264 1306 pr_warn("%s: Setting fail_over_mac to active for active-backup mode\n",
00503b6f 1307 bond_dev->name);
1308 }
3915c1e8 1309 } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
90194264 1310 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",
a4aee5c8 1311 bond_dev->name);
2ab82852
MS		 1312 res = -EOPNOTSUPP;
1313 goto err_undo_flags;
1314 }
1da177e4
LT		 1315 }
1316
8d8fc29d
AW		 1317 call_netdevice_notifiers(NETDEV_JOIN, slave_dev);
1318
c20811a7
JP		 1319 /* If this is the first slave, then we need to set the master's hardware
1320 * address to be the same as the slave's. */
0965a1f3 1321 if (!bond_has_slaves(bond) &&
dec1e90e 1322 bond->dev->addr_assign_type == NET_ADDR_RANDOM)
409cc1f8 1323 bond_set_dev_addr(bond->dev, slave_dev);
c20811a7 1324
3fdddd85 1325 new_slave = bond_alloc_slave(bond);
1da177e4
LT		 1326 if (!new_slave) {
1327 res = -ENOMEM;
1328 goto err_undo_flags;
1329 }
3fdddd85 1330
dc73c41f
VF		 1331 new_slave->bond = bond;
1332 new_slave->dev = slave_dev;
bb1d9123
AG		 1333 /*
1334 * Set the new_slave's queue_id to be zero. Queue ID mapping
1335 * is set via sysfs or module option if desired.
1336 */
1337 new_slave->queue_id = 0;
1338
b15ba0fb
JP		 1339 /* Save slave's original mtu and then set it to match the bond */
1340 new_slave->original_mtu = slave_dev->mtu;
1341 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1342 if (res) {
1343 pr_debug("Error %d calling dev_set_mtu\n", res);
1344 goto err_free;
1345 }
1346
217df670
JV		 1347 /*
1348 * Save slave's original ("permanent") mac address for modes
1349 * that need it, and for restoring it upon release, and then
1350 * set it to the master's address
1351 */
ada0f863 1352 ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr);
1da177e4 1353
00503b6f 1354 if (!bond->params.fail_over_mac ||
01844098 1355 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
2ab82852
MS		 1356 /*
1357 * Set slave to master's mac address. The application already
1358 * set the master's mac address to that of the first slave
1359 */
1360 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1361 addr.sa_family = slave_dev->type;
1362 res = dev_set_mac_address(slave_dev, &addr);
1363 if (res) {
5a03cdb7 1364 pr_debug("Error %d calling set_mac_address\n", res);
b15ba0fb 1365 goto err_restore_mtu;
2ab82852 1366 }
217df670 1367 }
1da177e4 1368
217df670
JV		 1369 /* open the slave since the application closed it */
1370 res = dev_open(slave_dev);
1371 if (res) {
3d632c3f 1372 pr_debug("Opening slave %s failed\n", slave_dev->name);
1f718f0f 1373 goto err_restore_mac;
1da177e4
LT		 1374 }
1375
0b680e75 1376 slave_dev->priv_flags |= IFF_BONDING;
1da177e4 1377
58402054 1378 if (bond_is_lb(bond)) {
1da177e4
LT		 1379 /* bond_alb_init_slave() must be called before all other stages since
1380 * it might fail and we do not want to have to undo everything
1381 */
1382 res = bond_alb_init_slave(bond, new_slave);
3d632c3f 1383 if (res)
569f0c4d 1384 goto err_close;
1da177e4
LT		 1385 }
1386
ec0865a9 1387 /* If the mode uses primary, then the following is handled by
303d1cbf 1388 * bond_change_active_slave().
1da177e4 1389 */
ec0865a9 1390 if (!bond_uses_primary(bond)) {
1da177e4
LT		 1391 /* set promiscuity level to new slave */
1392 if (bond_dev->flags & IFF_PROMISC) {
7e1a1ac1
WC		 1393 res = dev_set_promiscuity(slave_dev, 1);
1394 if (res)
1395 goto err_close;
1da177e4
LT		 1396 }
1397
1398 /* set allmulti level to new slave */
1399 if (bond_dev->flags & IFF_ALLMULTI) {
7e1a1ac1
WC		 1400 res = dev_set_allmulti(slave_dev, 1);
1401 if (res)
1402 goto err_close;
1da177e4
LT		 1403 }
1404
b9e40857 1405 netif_addr_lock_bh(bond_dev);
303d1cbf
JV		 1406
1407 dev_mc_sync_multiple(slave_dev, bond_dev);
1408 dev_uc_sync_multiple(slave_dev, bond_dev);
1409
b9e40857 1410 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1411 }
1412
01844098 1413 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1da177e4
LT		 1414 /* add lacpdu mc addr to mc list */
1415 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1416
22bedad3 1417 dev_mc_add(slave_dev, lacpdu_multicast);
1da177e4
LT		 1418 }
1419
b8e2fde4
WY		 1420 res = vlan_vids_add_by_dev(slave_dev, bond_dev);
1421 if (res) {
1ff412ad 1422 pr_err("%s: Error: Couldn't add bond vlan ids to %s\n",
1423 bond_dev->name, slave_dev->name);
1424 goto err_close;
1425 }
1da177e4 1426
c8c23903 1427 prev_slave = bond_last_slave(bond);
1da177e4
LT		 1428
1429 new_slave->delay = 0;
1430 new_slave->link_failure_count = 0;
1431
876254ae
VF		 1432 bond_update_speed_duplex(new_slave);
1433
49f17de7 1434 new_slave->last_rx = jiffies -
f31c7937 1435 (msecs_to_jiffies(bond->params.arp_interval) + 1);
8599b52e 1436 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
49f17de7 1437 new_slave->target_last_arp_rx[i] = new_slave->last_rx;
f5b2b966 1438
1da177e4
LT		 1439 if (bond->params.miimon && !bond->params.use_carrier) {
1440 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1441
1442 if ((link_reporting == -1) && !bond->params.arp_interval) {
1443 /*
1444 * miimon is set but a bonded network driver
1445 * does not support ETHTOOL/MII and
1446 * arp_interval is not set. Note: if
1447 * use_carrier is enabled, we will never go
1448 * here (because netif_carrier is always
1449 * supported); thus, we don't need to change
1450 * the messages for netif_carrier.
1451 */
91565ebb
JP		 1452 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",
1453 bond_dev->name, slave_dev->name);
1da177e4
LT		 1454 } else if (link_reporting == -1) {
1455 /* unable get link status using mii/ethtool */
91565ebb
JP		 1456 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",
1457 bond_dev->name, slave_dev->name);
1da177e4
LT		 1458 }
1459 }
1460
1461 /* check for initial state */
f31c7937
MK		 1462 if (bond->params.miimon) {
1463 if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) {
1464 if (bond->params.updelay) {
1465 new_slave->link = BOND_LINK_BACK;
1466 new_slave->delay = bond->params.updelay;
1467 } else {
1468 new_slave->link = BOND_LINK_UP;
1469 }
1da177e4 1470 } else {
f31c7937 1471 new_slave->link = BOND_LINK_DOWN;
1da177e4 1472 }
f31c7937
MK		 1473 } else if (bond->params.arp_interval) {
1474 new_slave->link = (netif_carrier_ok(slave_dev) ?
1475 BOND_LINK_UP : BOND_LINK_DOWN);
1da177e4 1476 } else {
f31c7937 1477 new_slave->link = BOND_LINK_UP;
1da177e4
LT		 1478 }
1479
f31c7937 1480 if (new_slave->link != BOND_LINK_DOWN)
8e603460 1481 new_slave->last_link_up = jiffies;
f31c7937 1482 pr_debug("Initial state of slave_dev is BOND_LINK_%s\n",
90194264
JP		 1483 new_slave->link == BOND_LINK_DOWN ? "DOWN" :
1484 (new_slave->link == BOND_LINK_UP ? "UP" : "BACK"));
f31c7937 1485
ec0865a9 1486 if (bond_uses_primary(bond) && bond->params.primary[0]) {
1da177e4 1487 /* if there is a primary slave, remember it */
a549952a 1488 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1da177e4 1489 bond->primary_slave = new_slave;
a549952a
JP		 1490 bond->force_primary = true;
1491 }
1da177e4
LT		 1492 }
1493
01844098 1494 switch (BOND_MODE(bond)) {
1da177e4 1495 case BOND_MODE_ACTIVEBACKUP:
5e5b0665 1496 bond_set_slave_inactive_flags(new_slave,
1497 BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT		 1498 break;
1499 case BOND_MODE_8023AD:
1500 /* in 802.3ad mode, the internal mechanism
1501 * will activate the slaves in the selected
1502 * aggregator
1503 */
5e5b0665 1504 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1da177e4 1505 /* if this is the first slave */
23c147e0 1506 if (!prev_slave) {
3fdddd85 1507 SLAVE_AD_INFO(new_slave)->id = 1;
1da177e4
LT		 1508 /* Initialize AD with the number of times that the AD timer is called in 1 second
1509 * can be called only after the mac address of the bond is set
1510 */
56d00c67 1511 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL);
1da177e4 1512 } else {
3fdddd85 1513 SLAVE_AD_INFO(new_slave)->id =
1514 SLAVE_AD_INFO(prev_slave)->id + 1;
1da177e4
LT		 1515 }
1516
1517 bond_3ad_bind_slave(new_slave);
1518 break;
1519 case BOND_MODE_TLB:
1520 case BOND_MODE_ALB:
e30bc066 1521 bond_set_active_slave(new_slave);
5e5b0665 1522 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT		 1523 break;
1524 default:
5a03cdb7 1525 pr_debug("This slave is always active in trunk mode\n");
1da177e4
LT		 1526
1527 /* always active in trunk mode */
e30bc066 1528 bond_set_active_slave(new_slave);
1da177e4
LT		 1529
1530 /* In trunking mode there is little meaning to curr_active_slave
1531 * anyway (it holds no special properties of the bond device),
1532 * so we can change it without calling change_active_interface()
1533 */
4e591b93 1534 if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP)
278b2083 1535 rcu_assign_pointer(bond->curr_active_slave, new_slave);
3d632c3f 1536
1da177e4
LT		 1537 break;
1538 } /* switch(bond_mode) */
1539
f6dc31a8 1540#ifdef CONFIG_NET_POLL_CONTROLLER
10eccb46 1541 slave_dev->npinfo = bond->dev->npinfo;
8a8efa22
AW		 1542 if (slave_dev->npinfo) {
1543 if (slave_enable_netpoll(new_slave)) {
90194264
JP		 1544 pr_info("Error, %s: master_dev is using netpoll, but new slave device does not support netpoll\n",
1545 bond_dev->name);
8a8efa22 1546 res = -EBUSY;
f7d9821a 1547 goto err_detach;
8a8efa22 1548 }
f6dc31a8
WC		 1549 }
1550#endif
8a8efa22 1551
35d48903
JP		 1552 res = netdev_rx_handler_register(slave_dev, bond_handle_frame,
1553 new_slave);
1554 if (res) {
1555 pr_debug("Error %d calling netdev_rx_handler_register\n", res);
5831d66e 1556 goto err_detach;
35d48903
JP		 1557 }
1558
1f718f0f
VF		 1559 res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave);
1560 if (res) {
1561 pr_debug("Error %d calling bond_master_upper_dev_link\n", res);
1562 goto err_unregister;
1563 }
1564
07699f9a 1565 res = bond_sysfs_slave_add(new_slave);
1566 if (res) {
1567 pr_debug("Error %d calling bond_sysfs_slave_add\n", res);
1568 goto err_upper_unlink;
1569 }
1570
5378c2e6
VF		 1571 bond->slave_cnt++;
1572 bond_compute_features(bond);
1573 bond_set_carrier(bond);
1574
ec0865a9 1575 if (bond_uses_primary(bond)) {
f80889a5 1576 block_netpoll_tx();
5378c2e6
VF		 1577 write_lock_bh(&bond->curr_slave_lock);
1578 bond_select_active_slave(bond);
1579 write_unlock_bh(&bond->curr_slave_lock);
f80889a5 1580 unblock_netpoll_tx();
5378c2e6 1581 }
1f718f0f 1582
90194264 1583 pr_info("%s: Enslaving %s as %s interface with %s link\n",
a4aee5c8 1584 bond_dev->name, slave_dev->name,
90194264
JP		 1585 bond_is_active_slave(new_slave) ? "an active" : "a backup",
1586 new_slave->link != BOND_LINK_DOWN ? "an up" : "a down");
1da177e4
LT		 1587
1588 /* enslave is successful */
1589 return 0;
1590
1591/* Undo stages on error */
07699f9a 1592err_upper_unlink:
1593 bond_upper_dev_unlink(bond_dev, slave_dev);
1594
1f718f0f
VF		 1595err_unregister:
1596 netdev_rx_handler_unregister(slave_dev);
1597
f7d9821a 1598err_detach:
ec0865a9 1599 if (!bond_uses_primary(bond))
303d1cbf
JV		 1600 bond_hw_addr_flush(bond_dev, slave_dev);
1601
1ff412ad 1602 vlan_vids_del_by_dev(slave_dev, bond_dev);
3c5913b5 1603 if (bond->primary_slave == new_slave)
1604 bond->primary_slave = NULL;
3c5913b5 1605 if (bond->curr_active_slave == new_slave) {
f80889a5 1606 block_netpoll_tx();
3c5913b5 1607 write_lock_bh(&bond->curr_slave_lock);
c8517035 1608 bond_change_active_slave(bond, NULL);
3c5913b5 1609 bond_select_active_slave(bond);
1610 write_unlock_bh(&bond->curr_slave_lock);
f80889a5 1611 unblock_netpoll_tx();
3c5913b5 1612 }
fc7a72ac 1613 slave_disable_netpoll(new_slave);
f7d9821a 1614
1da177e4 1615err_close:
b6a5a7b9 1616 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4
LT		 1617 dev_close(slave_dev);
1618
1619err_restore_mac:
00503b6f 1620 if (!bond->params.fail_over_mac ||
01844098 1621 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
3915c1e8
JV		 1622 /* XXX TODO - fom follow mode needs to change master's
1623 * MAC if this slave's MAC is in use by the bond, or at
1624 * least print a warning.
1625 */
ada0f863 1626 ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr);
2ab82852
MS		 1627 addr.sa_family = slave_dev->type;
1628 dev_set_mac_address(slave_dev, &addr);
1629 }
1da177e4 1630
b15ba0fb
JP		 1631err_restore_mtu:
1632 dev_set_mtu(slave_dev, new_slave->original_mtu);
1633
1da177e4 1634err_free:
3fdddd85 1635 bond_free_slave(new_slave);
1da177e4
LT		 1636
1637err_undo_flags:
b8fad459 1638 /* Enslave of first slave has failed and we need to fix master's mac */
0965a1f3 1639 if (!bond_has_slaves(bond) &&
844223ab 1640 ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr))
b8fad459 1641 eth_hw_addr_random(bond_dev);
3d632c3f 1642
1da177e4
LT		 1643 return res;
1644}
1645
1646/*
1647 * Try to release the slave device <slave> from the bond device <master>
1648 * It is legal to access curr_active_slave without a lock because all the function
0896341a 1649 * is write-locked. If "all" is true it means that the function is being called
1650 * while destroying a bond interface and all slaves are being released.
1da177e4
LT		 1651 *
1652 * The rules for slave state should be:
1653 * for Active/Backup:
1654 * Active stays on all backups go down
1655 * for Bonded connections:
1656 * The first up interface should be left on and all others downed.
1657 */
0896341a 1658static int __bond_release_one(struct net_device *bond_dev,
1659 struct net_device *slave_dev,
1660 bool all)
1da177e4 1661{
454d7c9b 1662 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1663 struct slave *slave, *oldcurrent;
1664 struct sockaddr addr;
5a0068de 1665 int old_flags = bond_dev->flags;
c8f44aff 1666 netdev_features_t old_features = bond_dev->features;
1da177e4
LT		 1667
1668 /* slave is not a slave or master is not master of this slave */
1669 if (!(slave_dev->flags & IFF_SLAVE) ||
471cb5a3 1670 !netdev_has_upper_dev(slave_dev, bond_dev)) {
90194264 1671 pr_err("%s: Error: cannot release %s\n",
1da177e4
LT		 1672 bond_dev->name, slave_dev->name);
1673 return -EINVAL;
1674 }
1675
e843fa50 1676 block_netpoll_tx();
1da177e4
LT		 1677
1678 slave = bond_get_slave_by_dev(bond, slave_dev);
1679 if (!slave) {
1680 /* not a slave of this bond */
a4aee5c8
JP		 1681 pr_info("%s: %s not enslaved\n",
1682 bond_dev->name, slave_dev->name);
e843fa50 1683 unblock_netpoll_tx();
1da177e4
LT		 1684 return -EINVAL;
1685 }
1686
07699f9a 1687 bond_sysfs_slave_del(slave);
1688
1f718f0f 1689 bond_upper_dev_unlink(bond_dev, slave_dev);
35d48903
JP		 1690 /* unregister rx_handler early so bond_handle_frame wouldn't be called
1691 * for this slave anymore.
1692 */
1693 netdev_rx_handler_unregister(slave_dev);
35d48903
JP		 1694 write_lock_bh(&bond->lock);
1695
1da177e4 1696 /* Inform AD package of unbinding of slave. */
01844098 1697 if (BOND_MODE(bond) == BOND_MODE_8023AD)
1da177e4 1698 bond_3ad_unbind_slave(slave);
be79bd04 1699
c8517035 1700 write_unlock_bh(&bond->lock);
1da177e4 1701
90194264 1702 pr_info("%s: Releasing %s interface %s\n",
a4aee5c8 1703 bond_dev->name,
e30bc066 1704 bond_is_active_slave(slave) ? "active" : "backup",
a4aee5c8 1705 slave_dev->name);
1da177e4
LT		 1706
1707 oldcurrent = bond->curr_active_slave;
1708
1709 bond->current_arp_slave = NULL;
1710
00503b6f 1711 if (!all && (!bond->params.fail_over_mac ||
01844098 1712 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) {
844223ab 1713 if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) &&
0965a1f3 1714 bond_has_slaves(bond))
90194264
JP		 1715 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",
1716 bond_dev->name, slave_dev->name,
1717 slave->perm_hwaddr,
1718 bond_dev->name, slave_dev->name);
dec1e90e 1719 }
1720
3d632c3f 1721 if (bond->primary_slave == slave)
1da177e4 1722 bond->primary_slave = NULL;
1da177e4 1723
c8517035 1724 if (oldcurrent == slave) {
1725 write_lock_bh(&bond->curr_slave_lock);
1da177e4 1726 bond_change_active_slave(bond, NULL);
c8517035 1727 write_unlock_bh(&bond->curr_slave_lock);
1728 }
1da177e4 1729
58402054 1730 if (bond_is_lb(bond)) {
1da177e4
LT		 1731 /* Must be called only after the slave has been
1732 * detached from the list and the curr_active_slave
1733 * has been cleared (if our_slave == old_current),
1734 * but before a new active slave is selected.
1735 */
1736 bond_alb_deinit_slave(bond, slave);
1737 }
1738
0896341a 1739 if (all) {
36708b89 1740 RCU_INIT_POINTER(bond->curr_active_slave, NULL);
0896341a 1741 } else if (oldcurrent == slave) {
059fe7a5
JV		 1742 /*
1743 * Note that we hold RTNL over this sequence, so there
1744 * is no concern that another slave add/remove event
1745 * will interfere.
1746 */
059fe7a5
JV		 1747 write_lock_bh(&bond->curr_slave_lock);
1748
1da177e4
LT		 1749 bond_select_active_slave(bond);
1750
059fe7a5 1751 write_unlock_bh(&bond->curr_slave_lock);
059fe7a5
JV		 1752 }
1753
0965a1f3 1754 if (!bond_has_slaves(bond)) {
ff59c456 1755 bond_set_carrier(bond);
409cc1f8 1756 eth_hw_addr_random(bond_dev);
1da177e4
LT		 1757 }
1758
e843fa50 1759 unblock_netpoll_tx();
278b2083 1760 synchronize_rcu();
ee6154e1 1761 bond->slave_cnt--;
1da177e4 1762
0965a1f3 1763 if (!bond_has_slaves(bond)) {
2af73d4b 1764 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev);
80028ea1
VF		 1765 call_netdevice_notifiers(NETDEV_RELEASE, bond->dev);
1766 }
2af73d4b 1767
b2a103e6
MM		 1768 bond_compute_features(bond);
1769 if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1770 (old_features & NETIF_F_VLAN_CHALLENGED))
90194264 1771 pr_info("%s: last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n",
b2a103e6
MM		 1772 bond_dev->name, slave_dev->name, bond_dev->name);
1773
b76cdba9 1774 /* must do this from outside any spinlocks */
1ff412ad 1775 vlan_vids_del_by_dev(slave_dev, bond_dev);
1da177e4 1776
ec0865a9 1777 /* If the mode uses primary, then this cases was handled above by
303d1cbf 1778 * bond_change_active_slave(..., NULL)
1da177e4 1779 */
ec0865a9 1780 if (!bond_uses_primary(bond)) {
5a0068de
NH		 1781 /* unset promiscuity level from slave
1782 * NOTE: The NETDEV_CHANGEADDR call above may change the value
1783 * of the IFF_PROMISC flag in the bond_dev, but we need the
1784 * value of that flag before that change, as that was the value
1785 * when this slave was attached, so we cache at the start of the
1786 * function and use it here. Same goes for ALLMULTI below
1787 */
1788 if (old_flags & IFF_PROMISC)
1da177e4 1789 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 1790
1791 /* unset allmulti level from slave */
5a0068de 1792 if (old_flags & IFF_ALLMULTI)
1da177e4 1793 dev_set_allmulti(slave_dev, -1);
1da177e4 1794
303d1cbf 1795 bond_hw_addr_flush(bond_dev, slave_dev);
1da177e4
LT		 1796 }
1797
8a8efa22 1798 slave_disable_netpoll(slave);
f6dc31a8 1799
1da177e4
LT		 1800 /* close slave before restoring its mac address */
1801 dev_close(slave_dev);
1802
00503b6f 1803 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE ||
01844098 1804 BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
2ab82852 1805 /* restore original ("permanent") mac address */
ada0f863 1806 ether_addr_copy(addr.sa_data, slave->perm_hwaddr);
2ab82852
MS		 1807 addr.sa_family = slave_dev->type;
1808 dev_set_mac_address(slave_dev, &addr);
1809 }
1da177e4 1810
b15ba0fb
JP		 1811 dev_set_mtu(slave_dev, slave->original_mtu);
1812
2d7011ca 1813 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4 1814
3fdddd85 1815 bond_free_slave(slave);
1da177e4
LT		 1816
1817 return 0; /* deletion OK */
1818}
1819
0896341a 1820/* A wrapper used because of ndo_del_link */
1821int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1822{
1823 return __bond_release_one(bond_dev, slave_dev, false);
1824}
1825
d90a162a 1826/*
dadaa10b 1827* First release a slave and then destroy the bond if no more slaves are left.
d90a162a
MS		 1828* Must be under rtnl_lock when this function is called.
1829*/
26d8ee75 1830static int bond_release_and_destroy(struct net_device *bond_dev,
1831 struct net_device *slave_dev)
d90a162a 1832{
454d7c9b 1833 struct bonding *bond = netdev_priv(bond_dev);
d90a162a
MS		 1834 int ret;
1835
1836 ret = bond_release(bond_dev, slave_dev);
0965a1f3 1837 if (ret == 0 && !bond_has_slaves(bond)) {
8a8efa22 1838 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
90194264 1839 pr_info("%s: Destroying bond %s\n",
a4aee5c8 1840 bond_dev->name, bond_dev->name);
9e71626c 1841 unregister_netdevice(bond_dev);
d90a162a
MS		 1842 }
1843 return ret;
1844}
1845
1da177e4
LT		 1846static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
1847{
454d7c9b 1848 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 1849
01844098 1850 info->bond_mode = BOND_MODE(bond);
1da177e4
LT		 1851 info->miimon = bond->params.miimon;
1852
1da177e4 1853 info->num_slaves = bond->slave_cnt;
1da177e4
LT		 1854
1855 return 0;
1856}
1857
1858static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
1859{
454d7c9b 1860 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 1861 struct list_head *iter;
dec1e90e 1862 int i = 0, res = -ENODEV;
1da177e4 1863 struct slave *slave;
1da177e4 1864
9caff1e7 1865 bond_for_each_slave(bond, slave, iter) {
dec1e90e 1866 if (i++ == (int)info->slave_id) {
689c96cc
ED		 1867 res = 0;
1868 strcpy(info->slave_name, slave->dev->name);
1869 info->link = slave->link;
e30bc066 1870 info->state = bond_slave_state(slave);
689c96cc 1871 info->link_failure_count = slave->link_failure_count;
1da177e4
LT		 1872 break;
1873 }
1874 }
1da177e4 1875
689c96cc 1876 return res;
1da177e4
LT		 1877}
1878
1879/*-------------------------------- Monitoring -------------------------------*/
1880
1da177e4 1881
f0c76d61
JV		 1882static int bond_miimon_inspect(struct bonding *bond)
1883{
dec1e90e 1884 int link_state, commit = 0;
9caff1e7 1885 struct list_head *iter;
f0c76d61 1886 struct slave *slave;
41f89100
JP		 1887 bool ignore_updelay;
1888
1889 ignore_updelay = !bond->curr_active_slave ? true : false;
1da177e4 1890
4cb4f97b 1891 bond_for_each_slave_rcu(bond, slave, iter) {
f0c76d61 1892 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 1893
f0c76d61 1894 link_state = bond_check_dev_link(bond, slave->dev, 0);
1da177e4
LT		 1895
1896 switch (slave->link) {
f0c76d61
JV		 1897 case BOND_LINK_UP:
1898 if (link_state)
1899 continue;
1da177e4 1900
f0c76d61
JV		 1901 slave->link = BOND_LINK_FAIL;
1902 slave->delay = bond->params.downdelay;
1903 if (slave->delay) {
90194264 1904 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms\n",
a4aee5c8 1905 bond->dev->name,
01844098 1906 (BOND_MODE(bond) ==
a4aee5c8 1907 BOND_MODE_ACTIVEBACKUP) ?
e30bc066 1908 (bond_is_active_slave(slave) ?
a4aee5c8
JP		 1909 "active " : "backup ") : "",
1910 slave->dev->name,
1911 bond->params.downdelay * bond->params.miimon);
1da177e4 1912 }
f0c76d61
JV		 1913 /*FALLTHRU*/
1914 case BOND_LINK_FAIL:
1915 if (link_state) {
1916 /*
1917 * recovered before downdelay expired
1918 */
1919 slave->link = BOND_LINK_UP;
8e603460 1920 slave->last_link_up = jiffies;
90194264 1921 pr_info("%s: link status up again after %d ms for interface %s\n",
a4aee5c8
JP		 1922 bond->dev->name,
1923 (bond->params.downdelay - slave->delay) *
1924 bond->params.miimon,
1925 slave->dev->name);
f0c76d61 1926 continue;
1da177e4 1927 }
f0c76d61
JV		 1928
1929 if (slave->delay <= 0) {
1930 slave->new_link = BOND_LINK_DOWN;
1931 commit++;
1932 continue;
1da177e4 1933 }
1da177e4 1934
f0c76d61
JV		 1935 slave->delay--;
1936 break;
1937
1938 case BOND_LINK_DOWN:
1939 if (!link_state)
1940 continue;
1941
1942 slave->link = BOND_LINK_BACK;
1943 slave->delay = bond->params.updelay;
1944
1945 if (slave->delay) {
90194264 1946 pr_info("%s: link status up for interface %s, enabling it in %d ms\n",
a4aee5c8
JP		 1947 bond->dev->name, slave->dev->name,
1948 ignore_updelay ? 0 :
1949 bond->params.updelay *
1950 bond->params.miimon);
f0c76d61
JV		 1951 }
1952 /*FALLTHRU*/
1953 case BOND_LINK_BACK:
1954 if (!link_state) {
1955 slave->link = BOND_LINK_DOWN;
90194264 1956 pr_info("%s: link status down again after %d ms for interface %s\n",
a4aee5c8
JP		 1957 bond->dev->name,
1958 (bond->params.updelay - slave->delay) *
1959 bond->params.miimon,
1960 slave->dev->name);
f0c76d61
JV		 1961
1962 continue;
1963 }
1964
41f89100
JP		 1965 if (ignore_updelay)
1966 slave->delay = 0;
1967
f0c76d61
JV		 1968 if (slave->delay <= 0) {
1969 slave->new_link = BOND_LINK_UP;
1970 commit++;
41f89100 1971 ignore_updelay = false;
f0c76d61 1972 continue;
1da177e4 1973 }
f0c76d61
JV		 1974
1975 slave->delay--;
1da177e4 1976 break;
f0c76d61
JV		 1977 }
1978 }
1da177e4 1979
f0c76d61
JV		 1980 return commit;
1981}
1da177e4 1982
f0c76d61
JV		 1983static void bond_miimon_commit(struct bonding *bond)
1984{
9caff1e7 1985 struct list_head *iter;
f0c76d61 1986 struct slave *slave;
f0c76d61 1987
9caff1e7 1988 bond_for_each_slave(bond, slave, iter) {
f0c76d61
JV		 1989 switch (slave->new_link) {
1990 case BOND_LINK_NOCHANGE:
1991 continue;
1da177e4 1992
f0c76d61
JV		 1993 case BOND_LINK_UP:
1994 slave->link = BOND_LINK_UP;
8e603460 1995 slave->last_link_up = jiffies;
f0c76d61 1996
01844098 1997 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
f0c76d61 1998 /* prevent it from being the active one */
e30bc066 1999 bond_set_backup_slave(slave);
01844098 2000 } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) {
f0c76d61 2001 /* make it immediately active */
e30bc066 2002 bond_set_active_slave(slave);
f0c76d61
JV		 2003 } else if (slave != bond->primary_slave) {
2004 /* prevent it from being the active one */
e30bc066 2005 bond_set_backup_slave(slave);
1da177e4 2006 }
1da177e4 2007
90194264 2008 pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex\n",
700c2a77 2009 bond->dev->name, slave->dev->name,
db4e9b2b
NA		 2010 slave->speed == SPEED_UNKNOWN ? 0 : slave->speed,
2011 slave->duplex ? "full" : "half");
1da177e4 2012
f0c76d61 2013 /* notify ad that the link status has changed */
01844098 2014 if (BOND_MODE(bond) == BOND_MODE_8023AD)
f0c76d61 2015 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
059fe7a5 2016
58402054 2017 if (bond_is_lb(bond))
f0c76d61
JV		 2018 bond_alb_handle_link_change(bond, slave,
2019 BOND_LINK_UP);
1da177e4 2020
f0c76d61
JV		 2021 if (!bond->curr_active_slave ||
2022 (slave == bond->primary_slave))
2023 goto do_failover;
1da177e4 2024
f0c76d61 2025 continue;
059fe7a5 2026
f0c76d61 2027 case BOND_LINK_DOWN:
fba4acda
JV		 2028 if (slave->link_failure_count < UINT_MAX)
2029 slave->link_failure_count++;
2030
f0c76d61 2031 slave->link = BOND_LINK_DOWN;
1da177e4 2032
01844098
VF		 2033 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP ||
2034 BOND_MODE(bond) == BOND_MODE_8023AD)
5e5b0665 2035 bond_set_slave_inactive_flags(slave,
2036 BOND_SLAVE_NOTIFY_NOW);
f0c76d61 2037
a4aee5c8
JP		 2038 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2039 bond->dev->name, slave->dev->name);
f0c76d61 2040
01844098 2041 if (BOND_MODE(bond) == BOND_MODE_8023AD)
f0c76d61
JV		 2042 bond_3ad_handle_link_change(slave,
2043 BOND_LINK_DOWN);
2044
ae63e808 2045 if (bond_is_lb(bond))
f0c76d61
JV		 2046 bond_alb_handle_link_change(bond, slave,
2047 BOND_LINK_DOWN);
2048
2049 if (slave == bond->curr_active_slave)
2050 goto do_failover;
2051
2052 continue;
2053
2054 default:
a4aee5c8 2055 pr_err("%s: invalid new link %d on slave %s\n",
f0c76d61
JV		 2056 bond->dev->name, slave->new_link,
2057 slave->dev->name);
2058 slave->new_link = BOND_LINK_NOCHANGE;
2059
2060 continue;
2061 }
2062
2063do_failover:
2064 ASSERT_RTNL();
e843fa50 2065 block_netpoll_tx();
f0c76d61
JV		 2066 write_lock_bh(&bond->curr_slave_lock);
2067 bond_select_active_slave(bond);
2068 write_unlock_bh(&bond->curr_slave_lock);
e843fa50 2069 unblock_netpoll_tx();
f0c76d61
JV		 2070 }
2071
2072 bond_set_carrier(bond);
1da177e4
LT		 2073}
2074
0b0eef66
JV		 2075/*
2076 * bond_mii_monitor
2077 *
2078 * Really a wrapper that splits the mii monitor into two phases: an
f0c76d61
JV		 2079 * inspection, then (if inspection indicates something needs to be done)
2080 * an acquisition of appropriate locks followed by a commit phase to
2081 * implement whatever link state changes are indicated.
0b0eef66 2082 */
6da67d26 2083static void bond_mii_monitor(struct work_struct *work)
0b0eef66
JV		 2084{
2085 struct bonding *bond = container_of(work, struct bonding,
2086 mii_work.work);
ad246c99 2087 bool should_notify_peers = false;
1f2cd845 2088 unsigned long delay;
0b0eef66 2089
1f2cd845
DM		 2090 delay = msecs_to_jiffies(bond->params.miimon);
2091
2092 if (!bond_has_slaves(bond))
f0c76d61 2093 goto re_arm;
b59f9f74 2094
4cb4f97b 2095 rcu_read_lock();
2096
ad246c99
BH		 2097 should_notify_peers = bond_should_notify_peers(bond);
2098
1f2cd845 2099 if (bond_miimon_inspect(bond)) {
4cb4f97b 2100 rcu_read_unlock();
f0c76d61 2101
1f2cd845
DM		 2102 /* Race avoidance with bond_close cancel of workqueue */
2103 if (!rtnl_trylock()) {
1f2cd845
DM		 2104 delay = 1;
2105 should_notify_peers = false;
2106 goto re_arm;
2107 }
6b6c5261 2108
1f2cd845
DM		 2109 bond_miimon_commit(bond);
2110
1f2cd845 2111 rtnl_unlock(); /* might sleep, hold no other locks */
4cb4f97b 2112 } else
2113 rcu_read_unlock();
0b0eef66 2114
f0c76d61 2115re_arm:
e6d265e8 2116 if (bond->params.miimon)
1f2cd845
DM		 2117 queue_delayed_work(bond->wq, &bond->mii_work, delay);
2118
1f2cd845
DM		 2119 if (should_notify_peers) {
2120 if (!rtnl_trylock())
2121 return;
2122 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2123 rtnl_unlock();
2124 }
0b0eef66 2125}
c3ade5ca 2126
50223ce4 2127static bool bond_has_this_ip(struct bonding *bond, __be32 ip)
f5b2b966 2128{
50223ce4
VF		 2129 struct net_device *upper;
2130 struct list_head *iter;
2131 bool ret = false;
f5b2b966 2132
eaddcd76 2133 if (ip == bond_confirm_addr(bond->dev, 0, ip))
50223ce4 2134 return true;
f5b2b966 2135
50223ce4 2136 rcu_read_lock();
2f268f12 2137 netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) {
50223ce4
VF		 2138 if (ip == bond_confirm_addr(upper, 0, ip)) {
2139 ret = true;
2140 break;
2141 }
f5b2b966 2142 }
50223ce4 2143 rcu_read_unlock();
f5b2b966 2144
50223ce4 2145 return ret;
f5b2b966
JV		 2146}
2147
c3ade5ca
JV		 2148/*
2149 * We go to the (large) trouble of VLAN tagging ARP frames because
2150 * switches in VLAN mode (especially if ports are configured as
2151 * "native" to a VLAN) might not pass non-tagged frames.
2152 */
fbd929f2 2153static void bond_arp_send(struct net_device *slave_dev, int arp_op,
2154 __be32 dest_ip, __be32 src_ip,
44a40855 2155 struct bond_vlan_tag *tags)
c3ade5ca
JV		 2156{
2157 struct sk_buff *skb;
44a40855 2158 int i;
c3ade5ca 2159
fbd929f2 2160 pr_debug("arp %d on slave %s: dst %pI4 src %pI4\n",
2161 arp_op, slave_dev->name, &dest_ip, &src_ip);
3d632c3f 2162
c3ade5ca
JV		 2163 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2164 NULL, slave_dev->dev_addr, NULL);
2165
2166 if (!skb) {
4873ac3c 2167 net_err_ratelimited("ARP packet allocation failed\n");
c3ade5ca
JV		 2168 return;
2169 }
fbd929f2 2170
44a40855
VY		 2171 /* Go through all the tags backwards and add them to the packet */
2172 for (i = BOND_MAX_VLAN_ENCAP - 1; i > 0; i--) {
2173 if (!tags[i].vlan_id)
2174 continue;
2175
2176 pr_debug("inner tag: proto %X vid %X\n",
2177 ntohs(tags[i].vlan_proto), tags[i].vlan_id);
2178 skb = __vlan_put_tag(skb, tags[i].vlan_proto,
2179 tags[i].vlan_id);
2180 if (!skb) {
2181 net_err_ratelimited("failed to insert inner VLAN tag\n");
2182 return;
2183 }
2184 }
2185 /* Set the outer tag */
2186 if (tags[0].vlan_id) {
2187 pr_debug("outer tag: proto %X vid %X\n",
2188 ntohs(tags[0].vlan_proto), tags[0].vlan_id);
2189 skb = vlan_put_tag(skb, tags[0].vlan_proto, tags[0].vlan_id);
c3ade5ca 2190 if (!skb) {
4873ac3c 2191 net_err_ratelimited("failed to insert outer VLAN tag\n");
c3ade5ca
JV		 2192 return;
2193 }
2194 }
2195 arp_xmit(skb);
2196}
2197
44a40855
VY		 2198/* Validate the device path between the @start_dev and the @end_dev.
2199 * The path is valid if the @end_dev is reachable through device
2200 * stacking.
2201 * When the path is validated, collect any vlan information in the
2202 * path.
2203 */
14af9963
VY		 2204bool bond_verify_device_path(struct net_device *start_dev,
2205 struct net_device *end_dev,
2206 struct bond_vlan_tag *tags)
44a40855
VY		 2207{
2208 struct net_device *upper;
2209 struct list_head *iter;
2210 int idx;
2211
2212 if (start_dev == end_dev)
2213 return true;
2214
2215 netdev_for_each_upper_dev_rcu(start_dev, upper, iter) {
2216 if (bond_verify_device_path(upper, end_dev, tags)) {
2217 if (is_vlan_dev(upper)) {
2218 idx = vlan_get_encap_level(upper);
2219 if (idx >= BOND_MAX_VLAN_ENCAP)
2220 return false;
2221
2222 tags[idx].vlan_proto =
2223 vlan_dev_vlan_proto(upper);
2224 tags[idx].vlan_id = vlan_dev_vlan_id(upper);
2225 }
2226 return true;
2227 }
2228 }
2229
2230 return false;
2231}
c3ade5ca 2232
1da177e4
LT		 2233static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2234{
c3ade5ca 2235 struct rtable *rt;
44a40855 2236 struct bond_vlan_tag tags[BOND_MAX_VLAN_ENCAP];
27bc11e6 2237 __be32 *targets = bond->params.arp_targets, addr;
fbd929f2 2238 int i;
44a40855 2239 bool ret;
1da177e4 2240
27bc11e6 2241 for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) {
acca2674 2242 pr_debug("basa: target %pI4\n", &targets[i]);
44a40855 2243 memset(tags, 0, sizeof(tags));
c3ade5ca 2244
27bc11e6 2245 /* Find out through which dev should the packet go */
78fbfd8a
DM		 2246 rt = ip_route_output(dev_net(bond->dev), targets[i], 0,
2247 RTO_ONLINK, 0);
b23dd4fe 2248 if (IS_ERR(rt)) {
28572760
VF		 2249 /* there's no route to target - try to send arp
2250 * probe to generate any traffic (arp_validate=0)
2251 */
4873ac3c 2252 if (bond->params.arp_validate)
2253 net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n",
2254 bond->dev->name,
2255 &targets[i]);
44a40855
VY		 2256 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2257 0, tags);
c3ade5ca
JV		 2258 continue;
2259 }
2260
27bc11e6
VF		 2261 /* bond device itself */
2262 if (rt->dst.dev == bond->dev)
2263 goto found;
2264
2265 rcu_read_lock();
44a40855 2266 ret = bond_verify_device_path(bond->dev, rt->dst.dev, tags);
27bc11e6 2267 rcu_read_unlock();
c3ade5ca 2268
44a40855
VY		 2269 if (ret)
2270 goto found;
2271
27bc11e6 2272 /* Not our device - skip */
3e32582f
VF		 2273 pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2274 bond->dev->name, &targets[i],
2275 rt->dst.dev ? rt->dst.dev->name : "NULL");
2276
ed4b9f80 2277 ip_rt_put(rt);
27bc11e6
VF		 2278 continue;
2279
2280found:
2281 addr = bond_confirm_addr(rt->dst.dev, targets[i], 0);
2282 ip_rt_put(rt);
2283 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
44a40855 2284 addr, tags);
c3ade5ca
JV		 2285 }
2286}
2287
d3bb52b0 2288static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
f5b2b966 2289{
8599b52e
VF		 2290 int i;
2291
87a7b84b
VF		 2292 if (!sip || !bond_has_this_ip(bond, tip)) {
2293 pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip);
2294 return;
2295 }
f5b2b966 2296
8599b52e
VF		 2297 i = bond_get_targets_ip(bond->params.arp_targets, sip);
2298 if (i == -1) {
87a7b84b
VF		 2299 pr_debug("bva: sip %pI4 not found in targets\n", &sip);
2300 return;
f5b2b966 2301 }
49f17de7 2302 slave->last_rx = jiffies;
8599b52e 2303 slave->target_last_arp_rx[i] = jiffies;
f5b2b966
JV		 2304}
2305
5bb9e0b5 2306int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
2307 struct slave *slave)
f5b2b966 2308{
de063b70 2309 struct arphdr *arp = (struct arphdr *)skb->data;
010d3c39 2310 struct slave *curr_active_slave;
f5b2b966 2311 unsigned char *arp_ptr;
d3bb52b0 2312 __be32 sip, tip;
f2cb691a 2313 int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
f5b2b966 2314
f2cb691a 2315 if (!slave_do_arp_validate(bond, slave)) {
bedabf90 2316 if ((slave_do_arp_validate_only(bond) && is_arp) ||
2317 !slave_do_arp_validate_only(bond))
49f17de7 2318 slave->last_rx = jiffies;
13a8e0c8 2319 return RX_HANDLER_ANOTHER;
f2cb691a
VF		 2320 } else if (!is_arp) {
2321 return RX_HANDLER_ANOTHER;
2322 }
2c146102 2323
de063b70 2324 alen = arp_hdr_len(bond->dev);
f5b2b966 2325
3aba891d
JP		 2326 pr_debug("bond_arp_rcv: bond %s skb->dev %s\n",
2327 bond->dev->name, skb->dev->name);
f5b2b966 2328
de063b70
ED		 2329 if (alen > skb_headlen(skb)) {
2330 arp = kmalloc(alen, GFP_ATOMIC);
2331 if (!arp)
2332 goto out_unlock;
2333 if (skb_copy_bits(skb, 0, arp, alen) < 0)
2334 goto out_unlock;
2335 }
f5b2b966 2336
3aba891d 2337 if (arp->ar_hln != bond->dev->addr_len ||
f5b2b966
JV		 2338 skb->pkt_type == PACKET_OTHERHOST ||
2339 skb->pkt_type == PACKET_LOOPBACK ||
2340 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2341 arp->ar_pro != htons(ETH_P_IP) ||
2342 arp->ar_pln != 4)
2343 goto out_unlock;
2344
2345 arp_ptr = (unsigned char *)(arp + 1);
3aba891d 2346 arp_ptr += bond->dev->addr_len;
f5b2b966 2347 memcpy(&sip, arp_ptr, 4);
3aba891d 2348 arp_ptr += 4 + bond->dev->addr_len;
f5b2b966
JV		 2349 memcpy(&tip, arp_ptr, 4);
2350
5a03cdb7 2351 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
e30bc066 2352 bond->dev->name, slave->dev->name, bond_slave_state(slave),
a4aee5c8
JP		 2353 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2354 &sip, &tip);
f5b2b966 2355
010d3c39
VF		 2356 curr_active_slave = rcu_dereference(bond->curr_active_slave);
2357
f5b2b966
JV		 2358 /*
2359 * Backup slaves won't see the ARP reply, but do come through
2360 * here for each ARP probe (so we swap the sip/tip to validate
2361 * the probe). In a "redundant switch, common router" type of
2362 * configuration, the ARP probe will (hopefully) travel from
2363 * the active, through one switch, the router, then the other
2364 * switch before reaching the backup.
aeea64ac
VF		 2365 *
2366 * We 'trust' the arp requests if there is an active slave and
2367 * it received valid arp reply(s) after it became active. This
2368 * is done to avoid endless looping when we can't reach the
2369 * arp_ip_target and fool ourselves with our own arp requests.
f5b2b966 2370 */
010d3c39 2371
e30bc066 2372 if (bond_is_active_slave(slave))
f5b2b966 2373 bond_validate_arp(bond, slave, sip, tip);
010d3c39
VF		 2374 else if (curr_active_slave &&
2375 time_after(slave_last_rx(bond, curr_active_slave),
2376 curr_active_slave->last_link_up))
f5b2b966
JV		 2377 bond_validate_arp(bond, slave, tip, sip);
2378
2379out_unlock:
de063b70
ED		 2380 if (arp != (struct arphdr *)skb->data)
2381 kfree(arp);
13a8e0c8 2382 return RX_HANDLER_ANOTHER;
f5b2b966
JV		 2383}
2384
e7f63f1d
VF		 2385/* function to verify if we're in the arp_interval timeslice, returns true if
2386 * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval +
2387 * arp_interval/2) . the arp_interval/2 is needed for really fast networks.
2388 */
2389static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
2390 int mod)
2391{
2392 int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2393
2394 return time_in_range(jiffies,
2395 last_act - delta_in_ticks,
2396 last_act + mod * delta_in_ticks + delta_in_ticks/2);
2397}
2398
1da177e4
LT		 2399/*
2400 * this function is called regularly to monitor each slave's link
2401 * ensuring that traffic is being sent and received when arp monitoring
2402 * is used in load-balancing mode. if the adapter has been dormant, then an
2403 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2404 * arp monitoring in active backup mode.
2405 */
6da67d26 2406static void bond_loadbalance_arp_mon(struct work_struct *work)
1da177e4 2407{
1b76b316
JV		 2408 struct bonding *bond = container_of(work, struct bonding,
2409 arp_work.work);
1da177e4 2410 struct slave *slave, *oldcurrent;
9caff1e7 2411 struct list_head *iter;
6fde8f03 2412 int do_failover = 0, slave_state_changed = 0;
1da177e4 2413
1f2cd845 2414 if (!bond_has_slaves(bond))
1da177e4 2415 goto re_arm;
1da177e4 2416
2e52f4fe 2417 rcu_read_lock();
2418
2419 oldcurrent = ACCESS_ONCE(bond->curr_active_slave);
1da177e4
LT		 2420 /* see if any of the previous devices are up now (i.e. they have
2421 * xmt and rcv traffic). the curr_active_slave does not come into
8e603460
VF		 2422 * the picture unless it is null. also, slave->last_link_up is not
2423 * needed here because we send an arp on each slave and give a slave
2424 * as long as it needs to get the tx/rx within the delta.
1da177e4
LT		 2425 * TODO: what about up/down delay in arp mode? it wasn't here before
2426 * so it can wait
2427 */
2e52f4fe 2428 bond_for_each_slave_rcu(bond, slave, iter) {
cb32f2a0
JB		 2429 unsigned long trans_start = dev_trans_start(slave->dev);
2430
1da177e4 2431 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2432 if (bond_time_in_interval(bond, trans_start, 1) &&
49f17de7 2433 bond_time_in_interval(bond, slave->last_rx, 1)) {
1da177e4
LT		 2434
2435 slave->link = BOND_LINK_UP;
6fde8f03 2436 slave_state_changed = 1;
1da177e4
LT		 2437
2438 /* primary_slave has no meaning in round-robin
2439 * mode. the window of a slave being up and
2440 * curr_active_slave being null after enslaving
2441 * is closed.
2442 */
2443 if (!oldcurrent) {
90194264 2444 pr_info("%s: link status definitely up for interface %s\n",
a4aee5c8
JP		 2445 bond->dev->name,
2446 slave->dev->name);
1da177e4
LT		 2447 do_failover = 1;
2448 } else {
a4aee5c8
JP		 2449 pr_info("%s: interface %s is now up\n",
2450 bond->dev->name,
2451 slave->dev->name);
1da177e4
LT		 2452 }
2453 }
2454 } else {
2455 /* slave->link == BOND_LINK_UP */
2456
2457 /* not all switches will respond to an arp request
2458 * when the source ip is 0, so don't take the link down
2459 * if we don't know our ip yet
2460 */
e7f63f1d 2461 if (!bond_time_in_interval(bond, trans_start, 2) ||
49f17de7 2462 !bond_time_in_interval(bond, slave->last_rx, 2)) {
1da177e4
LT		 2463
2464 slave->link = BOND_LINK_DOWN;
6fde8f03 2465 slave_state_changed = 1;
1da177e4 2466
3d632c3f 2467 if (slave->link_failure_count < UINT_MAX)
1da177e4 2468 slave->link_failure_count++;
1da177e4 2469
90194264
JP		 2470 pr_info("%s: interface %s is now down\n",
2471 bond->dev->name, slave->dev->name);
1da177e4 2472
3d632c3f 2473 if (slave == oldcurrent)
1da177e4 2474 do_failover = 1;
1da177e4
LT		 2475 }
2476 }
2477
2478 /* note: if switch is in round-robin mode, all links
2479 * must tx arp to ensure all links rx an arp - otherwise
2480 * links may oscillate or not come up at all; if switch is
2481 * in something like xor mode, there is nothing we can
2482 * do - all replies will be rx'ed on same link causing slaves
2483 * to be unstable during low/no traffic periods
2484 */
b6adc610 2485 if (bond_slave_is_up(slave))
1da177e4 2486 bond_arp_send_all(bond, slave);
1da177e4
LT		 2487 }
2488
2e52f4fe 2489 rcu_read_unlock();
2490
6fde8f03 2491 if (do_failover || slave_state_changed) {
2e52f4fe 2492 if (!rtnl_trylock())
2493 goto re_arm;
1da177e4 2494
6fde8f03
DT		 2495 if (slave_state_changed) {
2496 bond_slave_state_change(bond);
2497 } else if (do_failover) {
2498 /* the bond_select_active_slave must hold RTNL
2499 * and curr_slave_lock for write.
2500 */
2501 block_netpoll_tx();
2502 write_lock_bh(&bond->curr_slave_lock);
1da177e4 2503
6fde8f03
DT		 2504 bond_select_active_slave(bond);
2505
2506 write_unlock_bh(&bond->curr_slave_lock);
2507 unblock_netpoll_tx();
2508 }
2e52f4fe 2509 rtnl_unlock();
1da177e4
LT		 2510 }
2511
2512re_arm:
e6d265e8 2513 if (bond->params.arp_interval)
e7f63f1d
VF		 2514 queue_delayed_work(bond->wq, &bond->arp_work,
2515 msecs_to_jiffies(bond->params.arp_interval));
1da177e4
LT		 2516}
2517
2518/*
b2220cad
JV		 2519 * Called to inspect slaves for active-backup mode ARP monitor link state
2520 * changes. Sets new_link in slaves to specify what action should take
2521 * place for the slave. Returns 0 if no changes are found, >0 if changes
2522 * to link states must be committed.
2523 *
eb9fa4b0 2524 * Called with rcu_read_lock hold.
1da177e4 2525 */
e7f63f1d 2526static int bond_ab_arp_inspect(struct bonding *bond)
1da177e4 2527{
def4460c 2528 unsigned long trans_start, last_rx;
9caff1e7 2529 struct list_head *iter;
dec1e90e 2530 struct slave *slave;
dec1e90e 2531 int commit = 0;
da210f55 2532
eb9fa4b0 2533 bond_for_each_slave_rcu(bond, slave, iter) {
b2220cad 2534 slave->new_link = BOND_LINK_NOCHANGE;
def4460c 2535 last_rx = slave_last_rx(bond, slave);
1da177e4 2536
b2220cad 2537 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2538 if (bond_time_in_interval(bond, last_rx, 1)) {
b2220cad
JV		 2539 slave->new_link = BOND_LINK_UP;
2540 commit++;
2541 }
b2220cad
JV		 2542 continue;
2543 }
1da177e4 2544
b2220cad
JV		 2545 /*
2546 * Give slaves 2*delta after being enslaved or made
2547 * active. This avoids bouncing, as the last receive
2548 * times need a full ARP monitor cycle to be updated.
2549 */
8e603460 2550 if (bond_time_in_interval(bond, slave->last_link_up, 2))
b2220cad
JV		 2551 continue;
2552
2553 /*
2554 * Backup slave is down if:
2555 * - No current_arp_slave AND
2556 * - more than 3*delta since last receive AND
2557 * - the bond has an IP address
2558 *
2559 * Note: a non-null current_arp_slave indicates
2560 * the curr_active_slave went down and we are
2561 * searching for a new one; under this condition
2562 * we only take the curr_active_slave down - this
2563 * gives each slave a chance to tx/rx traffic
2564 * before being taken out
2565 */
e30bc066 2566 if (!bond_is_active_slave(slave) &&
b2220cad 2567 !bond->current_arp_slave &&
e7f63f1d 2568 !bond_time_in_interval(bond, last_rx, 3)) {
b2220cad
JV		 2569 slave->new_link = BOND_LINK_DOWN;
2570 commit++;
2571 }
2572
2573 /*
2574 * Active slave is down if:
2575 * - more than 2*delta since transmitting OR
2576 * - (more than 2*delta since receive AND
2577 * the bond has an IP address)
2578 */
cb32f2a0 2579 trans_start = dev_trans_start(slave->dev);
e30bc066 2580 if (bond_is_active_slave(slave) &&
e7f63f1d
VF		 2581 (!bond_time_in_interval(bond, trans_start, 2) ||
2582 !bond_time_in_interval(bond, last_rx, 2))) {
b2220cad
JV		 2583 slave->new_link = BOND_LINK_DOWN;
2584 commit++;
2585 }
1da177e4
LT		 2586 }
2587
b2220cad
JV		 2588 return commit;
2589}
1da177e4 2590
b2220cad
JV		 2591/*
2592 * Called to commit link state changes noted by inspection step of
2593 * active-backup mode ARP monitor.
2594 *
eb9fa4b0 2595 * Called with RTNL hold.
b2220cad 2596 */
e7f63f1d 2597static void bond_ab_arp_commit(struct bonding *bond)
b2220cad 2598{
cb32f2a0 2599 unsigned long trans_start;
9caff1e7 2600 struct list_head *iter;
dec1e90e 2601 struct slave *slave;
1da177e4 2602
9caff1e7 2603 bond_for_each_slave(bond, slave, iter) {
b2220cad
JV		 2604 switch (slave->new_link) {
2605 case BOND_LINK_NOCHANGE:
2606 continue;
ff59c456 2607
b2220cad 2608 case BOND_LINK_UP:
cb32f2a0 2609 trans_start = dev_trans_start(slave->dev);
e7f63f1d
VF		 2610 if (bond->curr_active_slave != slave ||
2611 (!bond->curr_active_slave &&
2612 bond_time_in_interval(bond, trans_start, 1))) {
b2220cad 2613 slave->link = BOND_LINK_UP;
5a430974
VF		 2614 if (bond->current_arp_slave) {
2615 bond_set_slave_inactive_flags(
5e5b0665 2616 bond->current_arp_slave,
2617 BOND_SLAVE_NOTIFY_NOW);
5a430974
VF		 2618 bond->current_arp_slave = NULL;
2619 }
b2220cad 2620
90194264 2621 pr_info("%s: link status definitely up for interface %s\n",
b9f60253 2622 bond->dev->name, slave->dev->name);
b2220cad 2623
b9f60253
JP		 2624 if (!bond->curr_active_slave ||
2625 (slave == bond->primary_slave))
2626 goto do_failover;
1da177e4 2627
b9f60253 2628 }
1da177e4 2629
b9f60253 2630 continue;
1da177e4 2631
b2220cad
JV		 2632 case BOND_LINK_DOWN:
2633 if (slave->link_failure_count < UINT_MAX)
2634 slave->link_failure_count++;
2635
2636 slave->link = BOND_LINK_DOWN;
5e5b0665 2637 bond_set_slave_inactive_flags(slave,
2638 BOND_SLAVE_NOTIFY_NOW);
b2220cad 2639
a4aee5c8 2640 pr_info("%s: link status definitely down for interface %s, disabling it\n",
b9f60253 2641 bond->dev->name, slave->dev->name);
b2220cad 2642
b9f60253 2643 if (slave == bond->curr_active_slave) {
b2220cad 2644 bond->current_arp_slave = NULL;
b9f60253 2645 goto do_failover;
1da177e4 2646 }
b9f60253
JP		 2647
2648 continue;
b2220cad
JV		 2649
2650 default:
a4aee5c8 2651 pr_err("%s: impossible: new_link %d on slave %s\n",
b2220cad
JV		 2652 bond->dev->name, slave->new_link,
2653 slave->dev->name);
b9f60253 2654 continue;
1da177e4 2655 }
1da177e4 2656
b9f60253
JP		 2657do_failover:
2658 ASSERT_RTNL();
e843fa50 2659 block_netpoll_tx();
b2220cad 2660 write_lock_bh(&bond->curr_slave_lock);
b9f60253 2661 bond_select_active_slave(bond);
b2220cad 2662 write_unlock_bh(&bond->curr_slave_lock);
e843fa50 2663 unblock_netpoll_tx();
b2220cad 2664 }
1da177e4 2665
b2220cad
JV		 2666 bond_set_carrier(bond);
2667}
1da177e4 2668
b2220cad
JV		 2669/*
2670 * Send ARP probes for active-backup mode ARP monitor.
b0929915 2671 *
2672 * Called with rcu_read_lock hold.
b2220cad 2673 */
f2ebd477 2674static bool bond_ab_arp_probe(struct bonding *bond)
b2220cad 2675{
eb9fa4b0 2676 struct slave *slave, *before = NULL, *new_slave = NULL,
b0929915 2677 *curr_arp_slave = rcu_dereference(bond->current_arp_slave),
2678 *curr_active_slave = rcu_dereference(bond->curr_active_slave);
4087df87
VF		 2679 struct list_head *iter;
2680 bool found = false;
b0929915 2681 bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER;
f2ebd477 2682
98b90f26 2683 if (curr_arp_slave && curr_active_slave)
a4aee5c8 2684 pr_info("PROBE: c_arp %s && cas %s BAD\n",
eb9fa4b0 2685 curr_arp_slave->dev->name,
98b90f26 2686 curr_active_slave->dev->name);
1da177e4 2687
98b90f26
VF		 2688 if (curr_active_slave) {
2689 bond_arp_send_all(bond, curr_active_slave);
b0929915 2690 return should_notify_rtnl;
b2220cad 2691 }
1da177e4 2692
b2220cad
JV		 2693 /* if we don't have a curr_active_slave, search for the next available
2694 * backup slave from the current_arp_slave and make it the candidate
2695 * for becoming the curr_active_slave
2696 */
1da177e4 2697
eb9fa4b0 2698 if (!curr_arp_slave) {
b0929915 2699 curr_arp_slave = bond_first_slave_rcu(bond);
2700 if (!curr_arp_slave)
2701 return should_notify_rtnl;
b2220cad 2702 }
1da177e4 2703
b0929915 2704 bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER);
059fe7a5 2705
b0929915 2706 bond_for_each_slave_rcu(bond, slave, iter) {
b6adc610 2707 if (!found && !before && bond_slave_is_up(slave))
4087df87 2708 before = slave;
1da177e4 2709
b6adc610 2710 if (found && !new_slave && bond_slave_is_up(slave))
4087df87 2711 new_slave = slave;
b2220cad
JV		 2712 /* if the link state is up at this point, we
2713 * mark it down - this can happen if we have
2714 * simultaneous link failures and
2715 * reselect_active_interface doesn't make this
2716 * one the current slave so it is still marked
2717 * up when it is actually down
1da177e4 2718 */
b6adc610 2719 if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
b2220cad
JV		 2720 slave->link = BOND_LINK_DOWN;
2721 if (slave->link_failure_count < UINT_MAX)
2722 slave->link_failure_count++;
1da177e4 2723
5e5b0665 2724 bond_set_slave_inactive_flags(slave,
b0929915 2725 BOND_SLAVE_NOTIFY_LATER);
b2220cad 2726
90194264 2727 pr_info("%s: backup interface %s is now down\n",
a4aee5c8 2728 bond->dev->name, slave->dev->name);
1da177e4 2729 }
eb9fa4b0 2730 if (slave == curr_arp_slave)
4087df87 2731 found = true;
b2220cad 2732 }
4087df87
VF		 2733
2734 if (!new_slave && before)
2735 new_slave = before;
2736
b0929915 2737 if (!new_slave)
2738 goto check_state;
4087df87
VF		 2739
2740 new_slave->link = BOND_LINK_BACK;
b0929915 2741 bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER);
4087df87 2742 bond_arp_send_all(bond, new_slave);
8e603460 2743 new_slave->last_link_up = jiffies;
eb9fa4b0 2744 rcu_assign_pointer(bond->current_arp_slave, new_slave);
f2ebd477 2745
b0929915 2746check_state:
2747 bond_for_each_slave_rcu(bond, slave, iter) {
2748 if (slave->should_notify) {
2749 should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW;
2750 break;
2751 }
2752 }
2753 return should_notify_rtnl;
b2220cad 2754}
1da177e4 2755
6da67d26 2756static void bond_activebackup_arp_mon(struct work_struct *work)
b2220cad
JV		 2757{
2758 struct bonding *bond = container_of(work, struct bonding,
2759 arp_work.work);
b0929915 2760 bool should_notify_peers = false;
2761 bool should_notify_rtnl = false;
1f2cd845 2762 int delta_in_ticks;
1da177e4 2763
1f2cd845
DM		 2764 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2765
2766 if (!bond_has_slaves(bond))
b2220cad
JV		 2767 goto re_arm;
2768
eb9fa4b0 2769 rcu_read_lock();
b0929915 2770
ad246c99
BH		 2771 should_notify_peers = bond_should_notify_peers(bond);
2772
b0929915 2773 if (bond_ab_arp_inspect(bond)) {
2774 rcu_read_unlock();
2775
1f2cd845
DM		 2776 /* Race avoidance with bond_close flush of workqueue */
2777 if (!rtnl_trylock()) {
1f2cd845
DM		 2778 delta_in_ticks = 1;
2779 should_notify_peers = false;
2780 goto re_arm;
2781 }
b2220cad 2782
1f2cd845 2783 bond_ab_arp_commit(bond);
b0929915 2784
1f2cd845 2785 rtnl_unlock();
b0929915 2786 rcu_read_lock();
1f2cd845
DM		 2787 }
2788
b0929915 2789 should_notify_rtnl = bond_ab_arp_probe(bond);
2790 rcu_read_unlock();
ad246c99 2791
80b9d236 2792re_arm:
2793 if (bond->params.arp_interval)
1f2cd845
DM		 2794 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
2795
b0929915 2796 if (should_notify_peers || should_notify_rtnl) {
1f2cd845
DM		 2797 if (!rtnl_trylock())
2798 return;
b0929915 2799
2800 if (should_notify_peers)
2801 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
2802 bond->dev);
2803 if (should_notify_rtnl)
2804 bond_slave_state_notify(bond);
2805
1f2cd845
DM		 2806 rtnl_unlock();
2807 }
1da177e4
LT		 2808}
2809
1da177e4
LT		 2810/*-------------------------- netdev event handling --------------------------*/
2811
2812/*
2813 * Change device name
2814 */
2815static int bond_event_changename(struct bonding *bond)
2816{
1da177e4
LT		 2817 bond_remove_proc_entry(bond);
2818 bond_create_proc_entry(bond);
7e083840 2819
f073c7ca
TI		 2820 bond_debug_reregister(bond);
2821
1da177e4
LT		 2822 return NOTIFY_DONE;
2823}
2824
3d632c3f
SH		 2825static int bond_master_netdev_event(unsigned long event,
2826 struct net_device *bond_dev)
1da177e4 2827{
454d7c9b 2828 struct bonding *event_bond = netdev_priv(bond_dev);
1da177e4
LT		 2829
2830 switch (event) {
2831 case NETDEV_CHANGENAME:
2832 return bond_event_changename(event_bond);
a64d49c3
EB		 2833 case NETDEV_UNREGISTER:
2834 bond_remove_proc_entry(event_bond);
2835 break;
2836 case NETDEV_REGISTER:
2837 bond_create_proc_entry(event_bond);
2838 break;
6c8d23f7 2839 case NETDEV_NOTIFY_PEERS:
2840 if (event_bond->send_peer_notif)
2841 event_bond->send_peer_notif--;
2842 break;
1da177e4
LT		 2843 default:
2844 break;
2845 }
2846
2847 return NOTIFY_DONE;
2848}
2849
3d632c3f
SH		 2850static int bond_slave_netdev_event(unsigned long event,
2851 struct net_device *slave_dev)
1da177e4 2852{
471cb5a3 2853 struct slave *slave = bond_slave_get_rtnl(slave_dev);
6101391d 2854 struct bonding *bond;
2855 struct net_device *bond_dev;
471cb5a3
JP		 2856 u32 old_speed;
2857 u8 old_duplex;
1da177e4 2858
6101391d 2859 /* A netdev event can be generated while enslaving a device
2860 * before netdev_rx_handler_register is called in which case
2861 * slave will be NULL
2862 */
2863 if (!slave)
2864 return NOTIFY_DONE;
2865 bond_dev = slave->bond->dev;
2866 bond = slave->bond;
2867
1da177e4
LT		 2868 switch (event) {
2869 case NETDEV_UNREGISTER:
8d2ada77 2870 if (bond_dev->type != ARPHRD_ETHER)
471cb5a3
JP		 2871 bond_release_and_destroy(bond_dev, slave_dev);
2872 else
2873 bond_release(bond_dev, slave_dev);
1da177e4 2874 break;
98f41f69 2875 case NETDEV_UP:
1da177e4 2876 case NETDEV_CHANGE:
471cb5a3
JP		 2877 old_speed = slave->speed;
2878 old_duplex = slave->duplex;
17d04500 2879
471cb5a3 2880 bond_update_speed_duplex(slave);
17d04500 2881
01844098 2882 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
471cb5a3
JP		 2883 if (old_speed != slave->speed)
2884 bond_3ad_adapter_speed_changed(slave);
2885 if (old_duplex != slave->duplex)
2886 bond_3ad_adapter_duplex_changed(slave);
17d04500 2887 }
1da177e4
LT		 2888 break;
2889 case NETDEV_DOWN:
2890 /*
2891 * ... Or is it this?
2892 */
2893 break;
2894 case NETDEV_CHANGEMTU:
2895 /*
2896 * TODO: Should slaves be allowed to
2897 * independently alter their MTU? For
2898 * an active-backup bond, slaves need
2899 * not be the same type of device, so
2900 * MTUs may vary. For other modes,
2901 * slaves arguably should have the
2902 * same MTUs. To do this, we'd need to
2903 * take over the slave's change_mtu
2904 * function for the duration of their
2905 * servitude.
2906 */
2907 break;
2908 case NETDEV_CHANGENAME:
3ec775b9 2909 /* we don't care if we don't have primary set */
ec0865a9 2910 if (!bond_uses_primary(bond) ||
3ec775b9
VF		 2911 !bond->params.primary[0])
2912 break;
2913
2914 if (slave == bond->primary_slave) {
2915 /* slave's name changed - he's no longer primary */
2916 bond->primary_slave = NULL;
2917 } else if (!strcmp(slave_dev->name, bond->params.primary)) {
2918 /* we have a new primary slave */
2919 bond->primary_slave = slave;
2920 } else { /* we didn't change primary - exit */
2921 break;
2922 }
2923
90194264
JP		 2924 pr_info("%s: Primary slave changed to %s, reselecting active slave\n",
2925 bond->dev->name,
2926 bond->primary_slave ? slave_dev->name : "none");
f80889a5 2927
2928 block_netpoll_tx();
3ec775b9
VF		 2929 write_lock_bh(&bond->curr_slave_lock);
2930 bond_select_active_slave(bond);
2931 write_unlock_bh(&bond->curr_slave_lock);
f80889a5 2932 unblock_netpoll_tx();
1da177e4 2933 break;
8531c5ff
AK		 2934 case NETDEV_FEAT_CHANGE:
2935 bond_compute_features(bond);
2936 break;
4aa5dee4
JP		 2937 case NETDEV_RESEND_IGMP:
2938 /* Propagate to master device */
2939 call_netdevice_notifiers(event, slave->bond->dev);
2940 break;
1da177e4
LT		 2941 default:
2942 break;
2943 }
2944
2945 return NOTIFY_DONE;
2946}
2947
2948/*
2949 * bond_netdev_event: handle netdev notifier chain events.
2950 *
2951 * This function receives events for the netdev chain. The caller (an
e041c683 2952 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
1da177e4
LT		 2953 * locks for us to safely manipulate the slave devices (RTNL lock,
2954 * dev_probe_lock).
2955 */
3d632c3f
SH		 2956static int bond_netdev_event(struct notifier_block *this,
2957 unsigned long event, void *ptr)
1da177e4 2958{
351638e7 2959 struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
1da177e4 2960
5a03cdb7 2961 pr_debug("event_dev: %s, event: %lx\n",
90194264 2962 event_dev ? event_dev->name : "None", event);
1da177e4 2963
0b680e75
JV		 2964 if (!(event_dev->priv_flags & IFF_BONDING))
2965 return NOTIFY_DONE;
2966
1da177e4 2967 if (event_dev->flags & IFF_MASTER) {
5a03cdb7 2968 pr_debug("IFF_MASTER\n");
1da177e4
LT		 2969 return bond_master_netdev_event(event, event_dev);
2970 }
2971
2972 if (event_dev->flags & IFF_SLAVE) {
5a03cdb7 2973 pr_debug("IFF_SLAVE\n");
1da177e4
LT		 2974 return bond_slave_netdev_event(event, event_dev);
2975 }
2976
2977 return NOTIFY_DONE;
2978}
2979
2980static struct notifier_block bond_netdev_notifier = {
2981 .notifier_call = bond_netdev_event,
2982};
2983
169a3e66
JV		 2984/*---------------------------- Hashing Policies -----------------------------*/
2985
32819dc1
NA		 2986/* L2 hash helper */
2987static inline u32 bond_eth_hash(struct sk_buff *skb)
6b923cb7
JE		 2988{
2989 struct ethhdr *data = (struct ethhdr *)skb->data;
2990
2991 if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto))
32819dc1 2992 return data->h_dest[5] ^ data->h_source[5];
6b923cb7
JE		 2993
2994 return 0;
2995}
2996
32819dc1
NA		 2997/* Extract the appropriate headers based on bond's xmit policy */
2998static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
2999 struct flow_keys *fk)
6f6652be 3000{
32819dc1 3001 const struct ipv6hdr *iph6;
4394542c 3002 const struct iphdr *iph;
32819dc1 3003 int noff, proto = -1;
6b923cb7 3004
32819dc1
NA		 3005 if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
3006 return skb_flow_dissect(skb, fk);
3007
3008 fk->ports = 0;
3009 noff = skb_network_offset(skb);
3010 if (skb->protocol == htons(ETH_P_IP)) {
054bb880 3011 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
32819dc1 3012 return false;
6b923cb7 3013 iph = ip_hdr(skb);
32819dc1
NA		 3014 fk->src = iph->saddr;
3015 fk->dst = iph->daddr;
3016 noff += iph->ihl << 2;
3017 if (!ip_is_fragment(iph))
3018 proto = iph->protocol;
3019 } else if (skb->protocol == htons(ETH_P_IPV6)) {
054bb880 3020 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
32819dc1
NA		 3021 return false;
3022 iph6 = ipv6_hdr(skb);
3023 fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr);
3024 fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr);
3025 noff += sizeof(*iph6);
3026 proto = iph6->nexthdr;
3027 } else {
3028 return false;
6f6652be 3029 }
32819dc1
NA		 3030 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
3031 fk->ports = skb_flow_get_ports(skb, noff, proto);
6f6652be 3032
32819dc1 3033 return true;
6f6652be
JV		 3034}
3035
32819dc1
NA		 3036/**
3037 * bond_xmit_hash - generate a hash value based on the xmit policy
3038 * @bond: bonding device
3039 * @skb: buffer to use for headers
32819dc1
NA		 3040 *
3041 * This function will extract the necessary headers from the skb buffer and use
3042 * them to generate a hash based on the xmit_policy set in the bonding device
169a3e66 3043 */
ee62e868 3044u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb)
169a3e66 3045{
32819dc1
NA		 3046 struct flow_keys flow;
3047 u32 hash;
4394542c 3048
32819dc1
NA		 3049 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 ||
3050 !bond_flow_dissect(bond, skb, &flow))
ee62e868 3051 return bond_eth_hash(skb);
169a3e66 3052
32819dc1
NA		 3053 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 ||
3054 bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
3055 hash = bond_eth_hash(skb);
3056 else
3057 hash = (__force u32)flow.ports;
3058 hash ^= (__force u32)flow.dst ^ (__force u32)flow.src;
3059 hash ^= (hash >> 16);
3060 hash ^= (hash >> 8);
3061
ee62e868 3062 return hash;
169a3e66
JV		 3063}
3064
1da177e4
LT		 3065/*-------------------------- Device entry points ----------------------------*/
3066
fbb0c41b 3067static void bond_work_init_all(struct bonding *bond)
3068{
3069 INIT_DELAYED_WORK(&bond->mcast_work,
3070 bond_resend_igmp_join_requests_delayed);
3071 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3072 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
01844098 3073 if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
fbb0c41b 3074 INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon);
3075 else
3076 INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon);
3077 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
3078}
3079
3080static void bond_work_cancel_all(struct bonding *bond)
3081{
3082 cancel_delayed_work_sync(&bond->mii_work);
3083 cancel_delayed_work_sync(&bond->arp_work);
3084 cancel_delayed_work_sync(&bond->alb_work);
3085 cancel_delayed_work_sync(&bond->ad_work);
3086 cancel_delayed_work_sync(&bond->mcast_work);
3087}
3088
1da177e4
LT		 3089static int bond_open(struct net_device *bond_dev)
3090{
454d7c9b 3091 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3092 struct list_head *iter;
ba3211cc 3093 struct slave *slave;
1da177e4 3094
ba3211cc
PP		 3095 /* reset slave->backup and slave->inactive */
3096 read_lock(&bond->lock);
0965a1f3 3097 if (bond_has_slaves(bond)) {
ba3211cc 3098 read_lock(&bond->curr_slave_lock);
9caff1e7 3099 bond_for_each_slave(bond, slave, iter) {
ec0865a9 3100 if (bond_uses_primary(bond)
ba3211cc 3101 && (slave != bond->curr_active_slave)) {
5e5b0665 3102 bond_set_slave_inactive_flags(slave,
3103 BOND_SLAVE_NOTIFY_NOW);
ba3211cc 3104 } else {
5e5b0665 3105 bond_set_slave_active_flags(slave,
3106 BOND_SLAVE_NOTIFY_NOW);
ba3211cc
PP		 3107 }
3108 }
3109 read_unlock(&bond->curr_slave_lock);
3110 }
3111 read_unlock(&bond->lock);
3112
fbb0c41b 3113 bond_work_init_all(bond);
5a37e8ca 3114
58402054 3115 if (bond_is_lb(bond)) {
1da177e4
LT		 3116 /* bond_alb_initialize must be called before the timer
3117 * is started.
3118 */
01844098 3119 if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB)))
b473946a 3120 return -ENOMEM;
e9f0fb88
MB		 3121 if (bond->params.tlb_dynamic_lb)
3122 queue_delayed_work(bond->wq, &bond->alb_work, 0);
1da177e4
LT		 3123 }
3124
fbb0c41b 3125 if (bond->params.miimon) /* link check interval, in milliseconds. */
1b76b316 3126 queue_delayed_work(bond->wq, &bond->mii_work, 0);
1da177e4
LT		 3127
3128 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316 3129 queue_delayed_work(bond->wq, &bond->arp_work, 0);
3fe68df9 3130 bond->recv_probe = bond_arp_rcv;
1da177e4
LT		 3131 }
3132
01844098 3133 if (BOND_MODE(bond) == BOND_MODE_8023AD) {
1b76b316 3134 queue_delayed_work(bond->wq, &bond->ad_work, 0);
1da177e4 3135 /* register to receive LACPDUs */
3aba891d 3136 bond->recv_probe = bond_3ad_lacpdu_recv;
fd989c83 3137 bond_3ad_initiate_agg_selection(bond, 1);
1da177e4
LT		 3138 }
3139
3140 return 0;
3141}
3142
3143static int bond_close(struct net_device *bond_dev)
3144{
454d7c9b 3145 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3146
fbb0c41b 3147 bond_work_cancel_all(bond);
6c8d23f7 3148 bond->send_peer_notif = 0;
ee8487c0 3149 if (bond_is_lb(bond))
1da177e4 3150 bond_alb_deinitialize(bond);
3aba891d 3151 bond->recv_probe = NULL;
1da177e4
LT		 3152
3153 return 0;
3154}
3155
28172739
ED		 3156static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
3157 struct rtnl_link_stats64 *stats)
1da177e4 3158{
454d7c9b 3159 struct bonding *bond = netdev_priv(bond_dev);
28172739 3160 struct rtnl_link_stats64 temp;
9caff1e7 3161 struct list_head *iter;
1da177e4 3162 struct slave *slave;
1da177e4 3163
28172739 3164 memset(stats, 0, sizeof(*stats));
1da177e4
LT		 3165
3166 read_lock_bh(&bond->lock);
9caff1e7 3167 bond_for_each_slave(bond, slave, iter) {
be1f3c2c 3168 const struct rtnl_link_stats64 *sstats =
28172739
ED		 3169 dev_get_stats(slave->dev, &temp);
3170
3171 stats->rx_packets += sstats->rx_packets;
3172 stats->rx_bytes += sstats->rx_bytes;
3173 stats->rx_errors += sstats->rx_errors;
3174 stats->rx_dropped += sstats->rx_dropped;
3175
3176 stats->tx_packets += sstats->tx_packets;
3177 stats->tx_bytes += sstats->tx_bytes;
3178 stats->tx_errors += sstats->tx_errors;
3179 stats->tx_dropped += sstats->tx_dropped;
3180
3181 stats->multicast += sstats->multicast;
3182 stats->collisions += sstats->collisions;
3183
3184 stats->rx_length_errors += sstats->rx_length_errors;
3185 stats->rx_over_errors += sstats->rx_over_errors;
3186 stats->rx_crc_errors += sstats->rx_crc_errors;
3187 stats->rx_frame_errors += sstats->rx_frame_errors;
3188 stats->rx_fifo_errors += sstats->rx_fifo_errors;
3189 stats->rx_missed_errors += sstats->rx_missed_errors;
3190
3191 stats->tx_aborted_errors += sstats->tx_aborted_errors;
3192 stats->tx_carrier_errors += sstats->tx_carrier_errors;
3193 stats->tx_fifo_errors += sstats->tx_fifo_errors;
3194 stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
3195 stats->tx_window_errors += sstats->tx_window_errors;
2439f9eb 3196 }
1da177e4
LT		 3197 read_unlock_bh(&bond->lock);
3198
3199 return stats;
3200}
3201
3202static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3203{
080a06e1 3204 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3205 struct net_device *slave_dev = NULL;
3206 struct ifbond k_binfo;
3207 struct ifbond __user *u_binfo = NULL;
3208 struct ifslave k_sinfo;
3209 struct ifslave __user *u_sinfo = NULL;
3210 struct mii_ioctl_data *mii = NULL;
d1fbd3ed 3211 struct bond_opt_value newval;
387ff911 3212 struct net *net;
1da177e4
LT		 3213 int res = 0;
3214
a4aee5c8 3215 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd);
1da177e4
LT		 3216
3217 switch (cmd) {
1da177e4
LT		 3218 case SIOCGMIIPHY:
3219 mii = if_mii(ifr);
3d632c3f 3220 if (!mii)
1da177e4 3221 return -EINVAL;
3d632c3f 3222
1da177e4
LT		 3223 mii->phy_id = 0;
3224 /* Fall Through */
3225 case SIOCGMIIREG:
3226 /*
3227 * We do this again just in case we were called by SIOCGMIIREG
3228 * instead of SIOCGMIIPHY.
3229 */
3230 mii = if_mii(ifr);
3d632c3f 3231 if (!mii)
1da177e4 3232 return -EINVAL;
3d632c3f 3233
1da177e4
LT		 3234
3235 if (mii->reg_num == 1) {
1da177e4 3236 mii->val_out = 0;
6603a6f2 3237 read_lock(&bond->lock);
1da177e4 3238 read_lock(&bond->curr_slave_lock);
3d632c3f 3239 if (netif_carrier_ok(bond->dev))
1da177e4 3240 mii->val_out = BMSR_LSTATUS;
3d632c3f 3241
1da177e4 3242 read_unlock(&bond->curr_slave_lock);
6603a6f2 3243 read_unlock(&bond->lock);
1da177e4
LT		 3244 }
3245
3246 return 0;
3247 case BOND_INFO_QUERY_OLD:
3248 case SIOCBONDINFOQUERY:
3249 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3250
3d632c3f 3251 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
1da177e4 3252 return -EFAULT;
1da177e4
LT		 3253
3254 res = bond_info_query(bond_dev, &k_binfo);
3d632c3f
SH		 3255 if (res == 0 &&
3256 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3257 return -EFAULT;
1da177e4
LT		 3258
3259 return res;
3260 case BOND_SLAVE_INFO_QUERY_OLD:
3261 case SIOCBONDSLAVEINFOQUERY:
3262 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3263
3d632c3f 3264 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
1da177e4 3265 return -EFAULT;
1da177e4
LT		 3266
3267 res = bond_slave_info_query(bond_dev, &k_sinfo);
3d632c3f
SH		 3268 if (res == 0 &&
3269 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3270 return -EFAULT;
1da177e4
LT		 3271
3272 return res;
3273 default:
3274 /* Go on */
3275 break;
3276 }
3277
387ff911
G		 3278 net = dev_net(bond_dev);
3279
3280 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4 3281 return -EPERM;
1da177e4 3282
0917b933 3283 slave_dev = __dev_get_by_name(net, ifr->ifr_slave);
1da177e4 3284
a4aee5c8 3285 pr_debug("slave_dev=%p:\n", slave_dev);
1da177e4 3286
3d632c3f 3287 if (!slave_dev)
0917b933 3288 return -ENODEV;
1da177e4 3289
0917b933
YX		 3290 pr_debug("slave_dev->name=%s:\n", slave_dev->name);
3291 switch (cmd) {
3292 case BOND_ENSLAVE_OLD:
3293 case SIOCBONDENSLAVE:
3294 res = bond_enslave(bond_dev, slave_dev);
3295 break;
3296 case BOND_RELEASE_OLD:
3297 case SIOCBONDRELEASE:
3298 res = bond_release(bond_dev, slave_dev);
3299 break;
3300 case BOND_SETHWADDR_OLD:
3301 case SIOCBONDSETHWADDR:
3302 bond_set_dev_addr(bond_dev, slave_dev);
3303 res = 0;
3304 break;
3305 case BOND_CHANGE_ACTIVE_OLD:
3306 case SIOCBONDCHANGEACTIVE:
d1fbd3ed
NA		 3307 bond_opt_initstr(&newval, slave_dev->name);
3308 res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval);
0917b933
YX		 3309 break;
3310 default:
3311 res = -EOPNOTSUPP;
1da177e4
LT		 3312 }
3313
1da177e4
LT		 3314 return res;
3315}
3316
d03462b9 3317static void bond_change_rx_flags(struct net_device *bond_dev, int change)
1da177e4 3318{
454d7c9b 3319 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3320
d03462b9
JP		 3321 if (change & IFF_PROMISC)
3322 bond_set_promiscuity(bond,
3323 bond_dev->flags & IFF_PROMISC ? 1 : -1);
3d632c3f 3324
d03462b9
JP		 3325 if (change & IFF_ALLMULTI)
3326 bond_set_allmulti(bond,
3327 bond_dev->flags & IFF_ALLMULTI ? 1 : -1);
3328}
1da177e4 3329
303d1cbf 3330static void bond_set_rx_mode(struct net_device *bond_dev)
d03462b9
JP		 3331{
3332 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3333 struct list_head *iter;
303d1cbf 3334 struct slave *slave;
1da177e4 3335
80ee5ad2 3336
b3241870 3337 rcu_read_lock();
ec0865a9 3338 if (bond_uses_primary(bond)) {
b3241870 3339 slave = rcu_dereference(bond->curr_active_slave);
303d1cbf
JV		 3340 if (slave) {
3341 dev_uc_sync(slave->dev, bond_dev);
3342 dev_mc_sync(slave->dev, bond_dev);
3343 }
303d1cbf 3344 } else {
b3241870 3345 bond_for_each_slave_rcu(bond, slave, iter) {
303d1cbf
JV		 3346 dev_uc_sync_multiple(slave->dev, bond_dev);
3347 dev_mc_sync_multiple(slave->dev, bond_dev);
3348 }
1da177e4 3349 }
b3241870 3350 rcu_read_unlock();
1da177e4
LT		 3351}
3352
234bcf8a 3353static int bond_neigh_init(struct neighbour *n)
00829823 3354{
234bcf8a 3355 struct bonding *bond = netdev_priv(n->dev);
234bcf8a
SP		 3356 const struct net_device_ops *slave_ops;
3357 struct neigh_parms parms;
dec1e90e 3358 struct slave *slave;
234bcf8a
SP		 3359 int ret;
3360
dec1e90e 3361 slave = bond_first_slave(bond);
234bcf8a
SP		 3362 if (!slave)
3363 return 0;
234bcf8a 3364 slave_ops = slave->dev->netdev_ops;
234bcf8a
SP		 3365 if (!slave_ops->ndo_neigh_setup)
3366 return 0;
3367
3368 parms.neigh_setup = NULL;
3369 parms.neigh_cleanup = NULL;
3370 ret = slave_ops->ndo_neigh_setup(slave->dev, &parms);
3371 if (ret)
3372 return ret;
3373
3374 /*
3375 * Assign slave's neigh_cleanup to neighbour in case cleanup is called
3376 * after the last slave has been detached. Assumes that all slaves
3377 * utilize the same neigh_cleanup (true at this writing as only user
3378 * is ipoib).
3379 */
3380 n->parms->neigh_cleanup = parms.neigh_cleanup;
3381
3382 if (!parms.neigh_setup)
3383 return 0;
3384
3385 return parms.neigh_setup(n);
3386}
3387
3388/*
3389 * The bonding ndo_neigh_setup is called at init time beofre any
3390 * slave exists. So we must declare proxy setup function which will
3391 * be used at run time to resolve the actual slave neigh param setup.
9918d5bf
VF		 3392 *
3393 * It's also called by master devices (such as vlans) to setup their
3394 * underlying devices. In that case - do nothing, we're already set up from
3395 * our init.
234bcf8a
SP		 3396 */
3397static int bond_neigh_setup(struct net_device *dev,
3398 struct neigh_parms *parms)
3399{
9918d5bf
VF		 3400 /* modify only our neigh_parms */
3401 if (parms->dev == dev)
3402 parms->neigh_setup = bond_neigh_init;
00829823 3403
00829823
SH		 3404 return 0;
3405}
3406
1da177e4
LT		 3407/*
3408 * Change the MTU of all of a master's slaves to match the master
3409 */
3410static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3411{
454d7c9b 3412 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3413 struct slave *slave, *rollback_slave;
9caff1e7 3414 struct list_head *iter;
1da177e4 3415 int res = 0;
1da177e4 3416
90194264
JP		 3417 pr_debug("bond=%p, name=%s, new_mtu=%d\n",
3418 bond, bond_dev ? bond_dev->name : "None", new_mtu);
1da177e4
LT		 3419
3420 /* Can't hold bond->lock with bh disabled here since
3421 * some base drivers panic. On the other hand we can't
3422 * hold bond->lock without bh disabled because we'll
3423 * deadlock. The only solution is to rely on the fact
3424 * that we're under rtnl_lock here, and the slaves
3425 * list won't change. This doesn't solve the problem
3426 * of setting the slave's MTU while it is
3427 * transmitting, but the assumption is that the base
3428 * driver can handle that.
3429 *
3430 * TODO: figure out a way to safely iterate the slaves
3431 * list, but without holding a lock around the actual
3432 * call to the base driver.
3433 */
3434
9caff1e7 3435 bond_for_each_slave(bond, slave, iter) {
c8c23903 3436 pr_debug("s %p c_m %p\n",
90194264 3437 slave, slave->dev->netdev_ops->ndo_change_mtu);
e944ef79 3438
1da177e4
LT		 3439 res = dev_set_mtu(slave->dev, new_mtu);
3440
3441 if (res) {
3442 /* If we failed to set the slave's mtu to the new value
3443 * we must abort the operation even in ACTIVE_BACKUP
3444 * mode, because if we allow the backup slaves to have
3445 * different mtu values than the active slave we'll
3446 * need to change their mtu when doing a failover. That
3447 * means changing their mtu from timer context, which
3448 * is probably not a good idea.
3449 */
5a03cdb7 3450 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 3451 goto unwind;
3452 }
3453 }
3454
3455 bond_dev->mtu = new_mtu;
3456
3457 return 0;
3458
3459unwind:
3460 /* unwind from head to the slave that failed */
9caff1e7 3461 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT		 3462 int tmp_res;
3463
81f23b13
VF		 3464 if (rollback_slave == slave)
3465 break;
3466
3467 tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu);
1da177e4 3468 if (tmp_res) {
a4aee5c8 3469 pr_debug("unwind err %d dev %s\n",
81f23b13 3470 tmp_res, rollback_slave->dev->name);
1da177e4
LT		 3471 }
3472 }
3473
3474 return res;
3475}
3476
3477/*
3478 * Change HW address
3479 *
3480 * Note that many devices must be down to change the HW address, and
3481 * downing the master releases all slaves. We can make bonds full of
3482 * bonding devices to test this, however.
3483 */
3484static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
3485{
454d7c9b 3486 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3487 struct slave *slave, *rollback_slave;
1da177e4 3488 struct sockaddr *sa = addr, tmp_sa;
9caff1e7 3489 struct list_head *iter;
1da177e4 3490 int res = 0;
1da177e4 3491
01844098 3492 if (BOND_MODE(bond) == BOND_MODE_ALB)
eb7cc59a
SH		 3493 return bond_alb_set_mac_address(bond_dev, addr);
3494
3495
a4aee5c8
JP		 3496 pr_debug("bond=%p, name=%s\n",
3497 bond, bond_dev ? bond_dev->name : "None");
1da177e4 3498
1b5acd29
JV		 3499 /* If fail_over_mac is enabled, do nothing and return success.
3500 * Returning an error causes ifenslave to fail.
dd957c57 3501 */
cc689aaa 3502 if (bond->params.fail_over_mac &&
01844098 3503 BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP)
dd957c57 3504 return 0;
2ab82852 3505
3d632c3f 3506 if (!is_valid_ether_addr(sa->sa_data))
1da177e4 3507 return -EADDRNOTAVAIL;
1da177e4
LT		 3508
3509 /* Can't hold bond->lock with bh disabled here since
3510 * some base drivers panic. On the other hand we can't
3511 * hold bond->lock without bh disabled because we'll
3512 * deadlock. The only solution is to rely on the fact
3513 * that we're under rtnl_lock here, and the slaves
3514 * list won't change. This doesn't solve the problem
3515 * of setting the slave's hw address while it is
3516 * transmitting, but the assumption is that the base
3517 * driver can handle that.
3518 *
3519 * TODO: figure out a way to safely iterate the slaves
3520 * list, but without holding a lock around the actual
3521 * call to the base driver.
3522 */
3523
9caff1e7 3524 bond_for_each_slave(bond, slave, iter) {
5a03cdb7 3525 pr_debug("slave %p %s\n", slave, slave->dev->name);
1da177e4
LT		 3526 res = dev_set_mac_address(slave->dev, addr);
3527 if (res) {
3528 /* TODO: consider downing the slave
3529 * and retry ?
3530 * User should expect communications
3531 * breakage anyway until ARP finish
3532 * updating, so...
3533 */
5a03cdb7 3534 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 3535 goto unwind;
3536 }
3537 }
3538
3539 /* success */
3540 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
3541 return 0;
3542
3543unwind:
3544 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
3545 tmp_sa.sa_family = bond_dev->type;
3546
3547 /* unwind from head to the slave that failed */
9caff1e7 3548 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT		 3549 int tmp_res;
3550
81f23b13
VF		 3551 if (rollback_slave == slave)
3552 break;
3553
3554 tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa);
1da177e4 3555 if (tmp_res) {
a4aee5c8 3556 pr_debug("unwind err %d dev %s\n",
81f23b13 3557 tmp_res, rollback_slave->dev->name);
1da177e4
LT		 3558 }
3559 }
3560
3561 return res;
3562}
3563
15077228
NA		 3564/**
3565 * bond_xmit_slave_id - transmit skb through slave with slave_id
3566 * @bond: bonding device that is transmitting
3567 * @skb: buffer to transmit
3568 * @slave_id: slave id up to slave_cnt-1 through which to transmit
3569 *
3570 * This function tries to transmit through slave with slave_id but in case
3571 * it fails, it tries to find the first available slave for transmission.
3572 * The skb is consumed in all cases, thus the function is void.
3573 */
da131ddb 3574static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id)
15077228 3575{
9caff1e7 3576 struct list_head *iter;
15077228
NA		 3577 struct slave *slave;
3578 int i = slave_id;
3579
3580 /* Here we start from the slave with slave_id */
9caff1e7 3581 bond_for_each_slave_rcu(bond, slave, iter) {
15077228 3582 if (--i < 0) {
891ab54d 3583 if (bond_slave_can_tx(slave)) {
15077228
NA		 3584 bond_dev_queue_xmit(bond, skb, slave->dev);
3585 return;
3586 }
3587 }
3588 }
3589
3590 /* Here we start from the first slave up to slave_id */
3591 i = slave_id;
9caff1e7 3592 bond_for_each_slave_rcu(bond, slave, iter) {
15077228
NA		 3593 if (--i < 0)
3594 break;
891ab54d 3595 if (bond_slave_can_tx(slave)) {
15077228
NA		 3596 bond_dev_queue_xmit(bond, skb, slave->dev);
3597 return;
3598 }
3599 }
3600 /* no slave that can tx has been found */
2bb77ab4 3601 dev_kfree_skb_any(skb);
15077228
NA		 3602}
3603
73958329
NA		 3604/**
3605 * bond_rr_gen_slave_id - generate slave id based on packets_per_slave
3606 * @bond: bonding device to use
3607 *
3608 * Based on the value of the bonding device's packets_per_slave parameter
3609 * this function generates a slave id, which is usually used as the next
3610 * slave to transmit through.
3611 */
3612static u32 bond_rr_gen_slave_id(struct bonding *bond)
3613{
73958329 3614 u32 slave_id;
809fa972
HFS		 3615 struct reciprocal_value reciprocal_packets_per_slave;
3616 int packets_per_slave = bond->params.packets_per_slave;
73958329
NA		 3617
3618 switch (packets_per_slave) {
3619 case 0:
3620 slave_id = prandom_u32();
3621 break;
3622 case 1:
3623 slave_id = bond->rr_tx_counter;
3624 break;
3625 default:
809fa972
HFS		 3626 reciprocal_packets_per_slave =
3627 bond->params.reciprocal_packets_per_slave;
73958329 3628 slave_id = reciprocal_divide(bond->rr_tx_counter,
809fa972 3629 reciprocal_packets_per_slave);
73958329
NA		 3630 break;
3631 }
3632 bond->rr_tx_counter++;
3633
3634 return slave_id;
3635}
3636
1da177e4
LT		 3637static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
3638{
454d7c9b 3639 struct bonding *bond = netdev_priv(bond_dev);
a2fd940f 3640 struct iphdr *iph = ip_hdr(skb);
15077228 3641 struct slave *slave;
73958329 3642 u32 slave_id;
1da177e4 3643
73958329 3644 /* Start with the curr_active_slave that joined the bond as the
a2fd940f
AG		 3645 * default for sending IGMP traffic. For failover purposes one
3646 * needs to maintain some consistency for the interface that will
3647 * send the join/membership reports. The curr_active_slave found
3648 * will send all of this type of traffic.
cf5f9044 3649 */
15077228 3650 if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) {
278b2083 3651 slave = rcu_dereference(bond->curr_active_slave);
891ab54d 3652 if (slave && bond_slave_can_tx(slave))
15077228
NA		 3653 bond_dev_queue_xmit(bond, skb, slave->dev);
3654 else
3655 bond_xmit_slave_id(bond, skb, 0);
a2fd940f 3656 } else {
73958329
NA		 3657 slave_id = bond_rr_gen_slave_id(bond);
3658 bond_xmit_slave_id(bond, skb, slave_id % bond->slave_cnt);
1da177e4 3659 }
0693e88e 3660
ec634fe3 3661 return NETDEV_TX_OK;
1da177e4
LT		 3662}
3663
3664/*
3665 * in active-backup mode, we know that bond->curr_active_slave is always valid if
3666 * the bond has a usable interface.
3667 */
3668static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
3669{
454d7c9b 3670 struct bonding *bond = netdev_priv(bond_dev);
71bc3b2d 3671 struct slave *slave;
1da177e4 3672
278b2083 3673 slave = rcu_dereference(bond->curr_active_slave);
71bc3b2d 3674 if (slave)
15077228
NA		 3675 bond_dev_queue_xmit(bond, skb, slave->dev);
3676 else
2bb77ab4 3677 dev_kfree_skb_any(skb);
0693e88e 3678
ec634fe3 3679 return NETDEV_TX_OK;
1da177e4
LT		 3680}
3681
32819dc1 3682/* In bond_xmit_xor() , we determine the output device by using a pre-
169a3e66
JV		 3683 * determined xmit_hash_policy(), If the selected device is not enabled,
3684 * find the next active slave.
1da177e4
LT		 3685 */
3686static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
3687{
454d7c9b 3688 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3689
ee62e868 3690 bond_xmit_slave_id(bond, skb, bond_xmit_hash(bond, skb) % bond->slave_cnt);
0693e88e 3691
ec634fe3 3692 return NETDEV_TX_OK;
1da177e4
LT		 3693}
3694
78a646ce 3695/* in broadcast mode, we send everything to all usable interfaces. */
1da177e4
LT		 3696static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
3697{
454d7c9b 3698 struct bonding *bond = netdev_priv(bond_dev);
78a646ce 3699 struct slave *slave = NULL;
9caff1e7 3700 struct list_head *iter;
1da177e4 3701
9caff1e7 3702 bond_for_each_slave_rcu(bond, slave, iter) {
78a646ce
NA		 3703 if (bond_is_last_slave(bond, slave))
3704 break;
b6adc610 3705 if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) {
78a646ce 3706 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1da177e4 3707
78a646ce 3708 if (!skb2) {
9152e26d 3709 net_err_ratelimited("%s: Error: %s: skb_clone() failed\n",
3710 bond_dev->name, __func__);
78a646ce 3711 continue;
1da177e4 3712 }
78a646ce
NA		 3713 /* bond_dev_queue_xmit always returns 0 */
3714 bond_dev_queue_xmit(bond, skb2, slave->dev);
1da177e4
LT		 3715 }
3716 }
b6adc610 3717 if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP)
78a646ce
NA		 3718 bond_dev_queue_xmit(bond, skb, slave->dev);
3719 else
2bb77ab4 3720 dev_kfree_skb_any(skb);
3d632c3f 3721
ec634fe3 3722 return NETDEV_TX_OK;
1da177e4
LT		 3723}
3724
3725/*------------------------- Device initialization ---------------------------*/
3726
bb1d9123
AG		 3727/*
3728 * Lookup the slave that corresponds to a qid
3729 */
3730static inline int bond_slave_override(struct bonding *bond,
3731 struct sk_buff *skb)
3732{
bb1d9123 3733 struct slave *slave = NULL;
9caff1e7 3734 struct list_head *iter;
bb1d9123 3735
0693e88e
MM		 3736 if (!skb->queue_mapping)
3737 return 1;
bb1d9123
AG		 3738
3739 /* Find out if any slaves have the same mapping as this skb. */
3900f290 3740 bond_for_each_slave_rcu(bond, slave, iter) {
3741 if (slave->queue_id == skb->queue_mapping) {
891ab54d 3742 if (bond_slave_can_tx(slave)) {
3900f290 3743 bond_dev_queue_xmit(bond, skb, slave->dev);
3744 return 0;
3745 }
3746 /* If the slave isn't UP, use default transmit policy. */
bb1d9123
AG		 3747 break;
3748 }
3749 }
3750
3900f290 3751 return 1;
bb1d9123
AG		 3752}
3753
374eeb5a 3754
f663dd9a 3755static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb,
99932d4f 3756 void *accel_priv, select_queue_fallback_t fallback)
bb1d9123
AG		 3757{
3758 /*
3759 * This helper function exists to help dev_pick_tx get the correct
fd0e435b 3760 * destination queue. Using a helper function skips a call to
bb1d9123
AG		 3761 * skb_tx_hash and will put the skbs in the queue we expect on their
3762 * way down to the bonding driver.
3763 */
fd0e435b
PO		 3764 u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0;
3765
374eeb5a
NH		 3766 /*
3767 * Save the original txq to restore before passing to the driver
3768 */
df4ab5b3 3769 qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping;
374eeb5a 3770
fd0e435b 3771 if (unlikely(txq >= dev->real_num_tx_queues)) {
d30ee670 3772 do {
fd0e435b 3773 txq -= dev->real_num_tx_queues;
d30ee670 3774 } while (txq >= dev->real_num_tx_queues);
fd0e435b
PO		 3775 }
3776 return txq;
bb1d9123
AG		 3777}
3778
0693e88e 3779static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
00829823 3780{
bb1d9123
AG		 3781 struct bonding *bond = netdev_priv(dev);
3782
d1e2e5cd
VF		 3783 if (bond_should_override_tx_queue(bond) &&
3784 !bond_slave_override(bond, skb))
3785 return NETDEV_TX_OK;
00829823 3786
01844098 3787 switch (BOND_MODE(bond)) {
00829823
SH		 3788 case BOND_MODE_ROUNDROBIN:
3789 return bond_xmit_roundrobin(skb, dev);
3790 case BOND_MODE_ACTIVEBACKUP:
3791 return bond_xmit_activebackup(skb, dev);
3792 case BOND_MODE_XOR:
3793 return bond_xmit_xor(skb, dev);
3794 case BOND_MODE_BROADCAST:
3795 return bond_xmit_broadcast(skb, dev);
3796 case BOND_MODE_8023AD:
3797 return bond_3ad_xmit_xor(skb, dev);
3798 case BOND_MODE_ALB:
00829823 3799 return bond_alb_xmit(skb, dev);
f05b42ea
MB		 3800 case BOND_MODE_TLB:
3801 return bond_tlb_xmit(skb, dev);
00829823
SH		 3802 default:
3803 /* Should never happen, mode already checked */
a4aee5c8 3804 pr_err("%s: Error: Unknown bonding mode %d\n",
01844098 3805 dev->name, BOND_MODE(bond));
00829823 3806 WARN_ON_ONCE(1);
2bb77ab4 3807 dev_kfree_skb_any(skb);
00829823
SH		 3808 return NETDEV_TX_OK;
3809 }
3810}
3811
0693e88e
MM		 3812static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
3813{
3814 struct bonding *bond = netdev_priv(dev);
3815 netdev_tx_t ret = NETDEV_TX_OK;
3816
3817 /*
3818 * If we risk deadlock from transmitting this in the
3819 * netpoll path, tell netpoll to queue the frame for later tx
3820 */
054bb880 3821 if (unlikely(is_netpoll_tx_blocked(dev)))
0693e88e
MM		 3822 return NETDEV_TX_BUSY;
3823
278b2083 3824 rcu_read_lock();
0965a1f3 3825 if (bond_has_slaves(bond))
0693e88e
MM		 3826 ret = __bond_start_xmit(skb, dev);
3827 else
2bb77ab4 3828 dev_kfree_skb_any(skb);
278b2083 3829 rcu_read_unlock();
0693e88e
MM		 3830
3831 return ret;
3832}
00829823 3833
bb5b052f
AG		 3834static int bond_ethtool_get_settings(struct net_device *bond_dev,
3835 struct ethtool_cmd *ecmd)
3836{
3837 struct bonding *bond = netdev_priv(bond_dev);
bb5b052f 3838 unsigned long speed = 0;
9caff1e7 3839 struct list_head *iter;
dec1e90e 3840 struct slave *slave;
bb5b052f
AG		 3841
3842 ecmd->duplex = DUPLEX_UNKNOWN;
3843 ecmd->port = PORT_OTHER;
3844
8557cd74 3845 /* Since bond_slave_can_tx returns false for all inactive or down slaves, we
bb5b052f
AG		 3846 * do not need to check mode. Though link speed might not represent
3847 * the true receive or transmit bandwidth (not all modes are symmetric)
3848 * this is an accurate maximum.
3849 */
3850 read_lock(&bond->lock);
9caff1e7 3851 bond_for_each_slave(bond, slave, iter) {
8557cd74 3852 if (bond_slave_can_tx(slave)) {
bb5b052f
AG		 3853 if (slave->speed != SPEED_UNKNOWN)
3854 speed += slave->speed;
3855 if (ecmd->duplex == DUPLEX_UNKNOWN &&
3856 slave->duplex != DUPLEX_UNKNOWN)
3857 ecmd->duplex = slave->duplex;
3858 }
3859 }
3860 ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN);
3861 read_unlock(&bond->lock);
dec1e90e 3862
bb5b052f
AG		 3863 return 0;
3864}
3865
217df670 3866static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
7826d43f 3867 struct ethtool_drvinfo *drvinfo)
217df670 3868{
7826d43f
JP		 3869 strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
3870 strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
3871 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d",
3872 BOND_ABI_VERSION);
217df670
JV		 3873}
3874
7282d491 3875static const struct ethtool_ops bond_ethtool_ops = {
217df670 3876 .get_drvinfo = bond_ethtool_get_drvinfo,
bb5b052f 3877 .get_settings = bond_ethtool_get_settings,
fa53ebac 3878 .get_link = ethtool_op_get_link,
8531c5ff
AK		 3879};
3880
eb7cc59a 3881static const struct net_device_ops bond_netdev_ops = {
181470fc 3882 .ndo_init = bond_init,
9e71626c 3883 .ndo_uninit = bond_uninit,
eb7cc59a
SH		 3884 .ndo_open = bond_open,
3885 .ndo_stop = bond_close,
00829823 3886 .ndo_start_xmit = bond_start_xmit,
bb1d9123 3887 .ndo_select_queue = bond_select_queue,
be1f3c2c 3888 .ndo_get_stats64 = bond_get_stats,
eb7cc59a 3889 .ndo_do_ioctl = bond_do_ioctl,
d03462b9 3890 .ndo_change_rx_flags = bond_change_rx_flags,
303d1cbf 3891 .ndo_set_rx_mode = bond_set_rx_mode,
eb7cc59a 3892 .ndo_change_mtu = bond_change_mtu,
cc0e4070 3893 .ndo_set_mac_address = bond_set_mac_address,
00829823 3894 .ndo_neigh_setup = bond_neigh_setup,
cc0e4070 3895 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
eb7cc59a 3896 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
f6dc31a8 3897#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 3898 .ndo_netpoll_setup = bond_netpoll_setup,
f6dc31a8
WC		 3899 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
3900 .ndo_poll_controller = bond_poll_controller,
3901#endif
9232ecca
JP		 3902 .ndo_add_slave = bond_enslave,
3903 .ndo_del_slave = bond_release,
b2a103e6 3904 .ndo_fix_features = bond_fix_features,
eb7cc59a
SH		 3905};
3906
b3f92b63
DG		 3907static const struct device_type bond_type = {
3908 .name = "bond",
3909};
3910
9e2e61fb
AW		 3911static void bond_destructor(struct net_device *bond_dev)
3912{
3913 struct bonding *bond = netdev_priv(bond_dev);
3914 if (bond->wq)
3915 destroy_workqueue(bond->wq);
3916 free_netdev(bond_dev);
3917}
3918
0a2a78c4 3919void bond_setup(struct net_device *bond_dev)
1da177e4 3920{
454d7c9b 3921 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3922
1da177e4
LT		 3923 /* initialize rwlocks */
3924 rwlock_init(&bond->lock);
3925 rwlock_init(&bond->curr_slave_lock);
d2991f75 3926 bond->params = bonding_defaults;
1da177e4
LT		 3927
3928 /* Initialize pointers */
1da177e4 3929 bond->dev = bond_dev;
1da177e4
LT		 3930
3931 /* Initialize the device entry points */
181470fc 3932 ether_setup(bond_dev);
eb7cc59a 3933 bond_dev->netdev_ops = &bond_netdev_ops;
8531c5ff 3934 bond_dev->ethtool_ops = &bond_ethtool_ops;
1da177e4 3935
9e2e61fb 3936 bond_dev->destructor = bond_destructor;
1da177e4 3937
b3f92b63
DG		 3938 SET_NETDEV_DEVTYPE(bond_dev, &bond_type);
3939
1da177e4
LT		 3940 /* Initialize the device options */
3941 bond_dev->tx_queue_len = 0;
3942 bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
c565b488 3943 bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT;
550fd08c 3944 bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
181470fc 3945
932ff279 3946 /* don't acquire bond device's netif_tx_lock when
1da177e4
LT		 3947 * transmitting */
3948 bond_dev->features |= NETIF_F_LLTX;
3949
3950 /* By default, we declare the bond to be fully
3951 * VLAN hardware accelerated capable. Special
3952 * care is taken in the various xmit functions
3953 * when there are slaves that are not hw accel
3954 * capable
3955 */
1da177e4 3956
f9399814
WC		 3957 /* Don't allow bond devices to change network namespaces. */
3958 bond_dev->features |= NETIF_F_NETNS_LOCAL;
3959
b2a103e6 3960 bond_dev->hw_features = BOND_VLAN_FEATURES |
f646968f
PM		 3961 NETIF_F_HW_VLAN_CTAG_TX |
3962 NETIF_F_HW_VLAN_CTAG_RX |
3963 NETIF_F_HW_VLAN_CTAG_FILTER;
b2a103e6 3964
34324dc2 3965 bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM);
5a7baa78 3966 bond_dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
b2a103e6 3967 bond_dev->features |= bond_dev->hw_features;
1da177e4
LT		 3968}
3969
c67dfb29
EB		 3970/*
3971* Destroy a bonding device.
3972* Must be under rtnl_lock when this function is called.
3973*/
3974static void bond_uninit(struct net_device *bond_dev)
a434e43f 3975{
454d7c9b 3976 struct bonding *bond = netdev_priv(bond_dev);
544a028e
VF		 3977 struct list_head *iter;
3978 struct slave *slave;
a434e43f 3979
f6dc31a8
WC		 3980 bond_netpoll_cleanup(bond_dev);
3981
c67dfb29 3982 /* Release the bonded slaves */
544a028e 3983 bond_for_each_slave(bond, slave, iter)
dec1e90e 3984 __bond_release_one(bond_dev, slave->dev, true);
90194264 3985 pr_info("%s: Released all slaves\n", bond_dev->name);
c67dfb29 3986
a434e43f
JV		 3987 list_del(&bond->bond_list);
3988
f073c7ca 3989 bond_debug_unregister(bond);
a434e43f
JV		 3990}
3991
1da177e4
LT		 3992/*------------------------- Module initialization ---------------------------*/
3993
1da177e4
LT		 3994static int bond_check_params(struct bond_params *params)
3995{
5a5c5fd4 3996 int arp_validate_value, fail_over_mac_value, primary_reselect_value, i;
f3253339 3997 struct bond_opt_value newval;
3998 const struct bond_opt_value *valptr;
8599b52e 3999 int arp_all_targets_value;
f5b2b966 4000
1da177e4
LT		 4001 /*
4002 * Convert string parameters.
4003 */
4004 if (mode) {
2b3798d5
NA		 4005 bond_opt_initstr(&newval, mode);
4006 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval);
4007 if (!valptr) {
4008 pr_err("Error: Invalid bonding mode \"%s\"\n", mode);
1da177e4
LT		 4009 return -EINVAL;
4010 }
2b3798d5 4011 bond_mode = valptr->value;
1da177e4
LT		 4012 }
4013
169a3e66
JV		 4014 if (xmit_hash_policy) {
4015 if ((bond_mode != BOND_MODE_XOR) &&
f05b42ea
MB		 4016 (bond_mode != BOND_MODE_8023AD) &&
4017 (bond_mode != BOND_MODE_TLB)) {
a4aee5c8 4018 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
90194264 4019 bond_mode_name(bond_mode));
169a3e66 4020 } else {
a4b32ce7
NA		 4021 bond_opt_initstr(&newval, xmit_hash_policy);
4022 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH),
4023 &newval);
4024 if (!valptr) {
a4aee5c8 4025 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
169a3e66
JV		 4026 xmit_hash_policy);
4027 return -EINVAL;
4028 }
a4b32ce7 4029 xmit_hashtype = valptr->value;
169a3e66
JV		 4030 }
4031 }
4032
1da177e4
LT		 4033 if (lacp_rate) {
4034 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8
JP		 4035 pr_info("lacp_rate param is irrelevant in mode %s\n",
4036 bond_mode_name(bond_mode));
1da177e4 4037 } else {
d3131de7
NA		 4038 bond_opt_initstr(&newval, lacp_rate);
4039 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE),
4040 &newval);
4041 if (!valptr) {
a4aee5c8 4042 pr_err("Error: Invalid lacp rate \"%s\"\n",
d3131de7 4043 lacp_rate);
1da177e4
LT		 4044 return -EINVAL;
4045 }
d3131de7 4046 lacp_fast = valptr->value;
1da177e4
LT		 4047 }
4048 }
4049
fd989c83 4050 if (ad_select) {
9e5f5eeb
NA		 4051 bond_opt_initstr(&newval, lacp_rate);
4052 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT),
4053 &newval);
4054 if (!valptr) {
4055 pr_err("Error: Invalid ad_select \"%s\"\n", ad_select);
fd989c83
JV		 4056 return -EINVAL;
4057 }
9e5f5eeb
NA		 4058 params->ad_select = valptr->value;
4059 if (bond_mode != BOND_MODE_8023AD)
91565ebb 4060 pr_warn("ad_select param only affects 802.3ad mode\n");
fd989c83
JV		 4061 } else {
4062 params->ad_select = BOND_AD_STABLE;
4063 }
4064
f5841306 4065 if (max_bonds < 0) {
91565ebb
JP		 4066 pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4067 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
1da177e4
LT		 4068 max_bonds = BOND_DEFAULT_MAX_BONDS;
4069 }
4070
4071 if (miimon < 0) {
91565ebb
JP		 4072 pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4073 miimon, INT_MAX);
b98d9c66 4074 miimon = 0;
1da177e4
LT		 4075 }
4076
4077 if (updelay < 0) {
91565ebb
JP		 4078 pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4079 updelay, INT_MAX);
1da177e4
LT		 4080 updelay = 0;
4081 }
4082
4083 if (downdelay < 0) {
91565ebb
JP		 4084 pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4085 downdelay, INT_MAX);
1da177e4
LT		 4086 downdelay = 0;
4087 }
4088
4089 if ((use_carrier != 0) && (use_carrier != 1)) {
91565ebb
JP		 4090 pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4091 use_carrier);
1da177e4
LT		 4092 use_carrier = 1;
4093 }
4094
ad246c99 4095 if (num_peer_notif < 0 || num_peer_notif > 255) {
91565ebb
JP		 4096 pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4097 num_peer_notif);
ad246c99
BH		 4098 num_peer_notif = 1;
4099 }
4100
834db4bc 4101 /* reset values for 802.3ad/TLB/ALB */
267bed77 4102 if (!bond_mode_uses_arp(bond_mode)) {
1da177e4 4103 if (!miimon) {
91565ebb
JP		 4104 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");
4105 pr_warn("Forcing miimon to 100msec\n");
fe9d04af 4106 miimon = BOND_DEFAULT_MIIMON;
1da177e4
LT		 4107 }
4108 }
4109
bb1d9123 4110 if (tx_queues < 1 || tx_queues > 255) {
91565ebb
JP		 4111 pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n",
4112 tx_queues, BOND_DEFAULT_TX_QUEUES);
bb1d9123
AG		 4113 tx_queues = BOND_DEFAULT_TX_QUEUES;
4114 }
4115
ebd8e497 4116 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
91565ebb
JP		 4117 pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n",
4118 all_slaves_active);
ebd8e497
AG		 4119 all_slaves_active = 0;
4120 }
4121
c2952c31 4122 if (resend_igmp < 0 || resend_igmp > 255) {
91565ebb
JP		 4123 pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n",
4124 resend_igmp, BOND_DEFAULT_RESEND_IGMP);
c2952c31
FL		 4125 resend_igmp = BOND_DEFAULT_RESEND_IGMP;
4126 }
4127
aa59d851
NA		 4128 bond_opt_initval(&newval, packets_per_slave);
4129 if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) {
73958329
NA		 4130 pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n",
4131 packets_per_slave, USHRT_MAX);
4132 packets_per_slave = 1;
4133 }
4134
1da177e4 4135 if (bond_mode == BOND_MODE_ALB) {
a4aee5c8
JP		 4136 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",
4137 updelay);
1da177e4
LT		 4138 }
4139
4140 if (!miimon) {
4141 if (updelay || downdelay) {
4142 /* just warn the user the up/down delay will have
4143 * no effect since miimon is zero...
4144 */
91565ebb
JP		 4145 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",
4146 updelay, downdelay);
1da177e4
LT		 4147 }
4148 } else {
4149 /* don't allow arp monitoring */
4150 if (arp_interval) {
91565ebb
JP		 4151 pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4152 miimon, arp_interval);
1da177e4
LT		 4153 arp_interval = 0;
4154 }
4155
4156 if ((updelay % miimon) != 0) {
91565ebb
JP		 4157 pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4158 updelay, miimon, (updelay / miimon) * miimon);
1da177e4
LT		 4159 }
4160
4161 updelay /= miimon;
4162
4163 if ((downdelay % miimon) != 0) {
91565ebb
JP		 4164 pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4165 downdelay, miimon,
4166 (downdelay / miimon) * miimon);
1da177e4
LT		 4167 }
4168
4169 downdelay /= miimon;
4170 }
4171
4172 if (arp_interval < 0) {
91565ebb
JP		 4173 pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4174 arp_interval, INT_MAX);
7bdb04ed 4175 arp_interval = 0;
1da177e4
LT		 4176 }
4177
5a5c5fd4 4178 for (arp_ip_count = 0, i = 0;
4179 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) {
1da177e4
LT		 4180 /* not complete check, but should be good enough to
4181 catch mistakes */
89015c18 4182 __be32 ip;
4183 if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) ||
2807a9fe 4184 !bond_is_ip_target_ok(ip)) {
91565ebb
JP		 4185 pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4186 arp_ip_target[i]);
1da177e4
LT		 4187 arp_interval = 0;
4188 } else {
0afee4e8
VF		 4189 if (bond_get_targets_ip(arp_target, ip) == -1)
4190 arp_target[arp_ip_count++] = ip;
4191 else
91565ebb
JP		 4192 pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n",
4193 &ip);
1da177e4
LT		 4194 }
4195 }
4196
4197 if (arp_interval && !arp_ip_count) {
4198 /* don't allow arping if no arp_ip_target given... */
91565ebb
JP		 4199 pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4200 arp_interval);
1da177e4
LT		 4201 arp_interval = 0;
4202 }
4203
f5b2b966 4204 if (arp_validate) {
f5b2b966 4205 if (!arp_interval) {
a4aee5c8 4206 pr_err("arp_validate requires arp_interval\n");
f5b2b966
JV		 4207 return -EINVAL;
4208 }
4209
16228881
NA		 4210 bond_opt_initstr(&newval, arp_validate);
4211 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE),
4212 &newval);
4213 if (!valptr) {
a4aee5c8 4214 pr_err("Error: invalid arp_validate \"%s\"\n",
16228881 4215 arp_validate);
f5b2b966
JV		 4216 return -EINVAL;
4217 }
16228881
NA		 4218 arp_validate_value = valptr->value;
4219 } else {
f5b2b966 4220 arp_validate_value = 0;
16228881 4221 }
f5b2b966 4222
8599b52e
VF		 4223 arp_all_targets_value = 0;
4224 if (arp_all_targets) {
edf36b24
NA		 4225 bond_opt_initstr(&newval, arp_all_targets);
4226 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS),
4227 &newval);
4228 if (!valptr) {
8599b52e
VF		 4229 pr_err("Error: invalid arp_all_targets_value \"%s\"\n",
4230 arp_all_targets);
4231 arp_all_targets_value = 0;
edf36b24
NA		 4232 } else {
4233 arp_all_targets_value = valptr->value;
8599b52e
VF		 4234 }
4235 }
4236
1da177e4 4237 if (miimon) {
a4aee5c8 4238 pr_info("MII link monitoring set to %d ms\n", miimon);
1da177e4 4239 } else if (arp_interval) {
16228881
NA		 4240 valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE,
4241 arp_validate_value);
a4aee5c8 4242 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
16228881 4243 arp_interval, valptr->string, arp_ip_count);
1da177e4
LT		 4244
4245 for (i = 0; i < arp_ip_count; i++)
90194264 4246 pr_cont(" %s", arp_ip_target[i]);
1da177e4 4247
90194264 4248 pr_cont("\n");
1da177e4 4249
b8a9787e 4250 } else if (max_bonds) {
1da177e4
LT		 4251 /* miimon and arp_interval not set, we need one so things
4252 * work as expected, see bonding.txt for details
4253 */
90194264 4254 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		 4255 }
4256
ec0865a9 4257 if (primary && !bond_mode_uses_primary(bond_mode)) {
1da177e4
LT		 4258 /* currently, using a primary only makes sense
4259 * in active backup, TLB or ALB modes
4260 */
91565ebb
JP		 4261 pr_warn("Warning: %s primary device specified but has no effect in %s mode\n",
4262 primary, bond_mode_name(bond_mode));
1da177e4
LT		 4263 primary = NULL;
4264 }
4265
a549952a 4266 if (primary && primary_reselect) {
388d3a6d
NA		 4267 bond_opt_initstr(&newval, primary_reselect);
4268 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT),
4269 &newval);
4270 if (!valptr) {
a4aee5c8 4271 pr_err("Error: Invalid primary_reselect \"%s\"\n",
388d3a6d 4272 primary_reselect);
a549952a
JP		 4273 return -EINVAL;
4274 }
388d3a6d 4275 primary_reselect_value = valptr->value;
a549952a
JP		 4276 } else {
4277 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4278 }
4279
3915c1e8 4280 if (fail_over_mac) {
1df6b6aa
NA		 4281 bond_opt_initstr(&newval, fail_over_mac);
4282 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC),
4283 &newval);
4284 if (!valptr) {
a4aee5c8 4285 pr_err("Error: invalid fail_over_mac \"%s\"\n",
1df6b6aa 4286 fail_over_mac);
3915c1e8
JV		 4287 return -EINVAL;
4288 }
1df6b6aa 4289 fail_over_mac_value = valptr->value;
3915c1e8 4290 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
91565ebb 4291 pr_warn("Warning: fail_over_mac only affects active-backup mode\n");
3915c1e8
JV		 4292 } else {
4293 fail_over_mac_value = BOND_FOM_NONE;
4294 }
dd957c57 4295
3a7129e5 4296 if (lp_interval == 0) {
91565ebb
JP		 4297 pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n",
4298 INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL);
3a7129e5 4299 lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
4300 }
4301
1da177e4
LT		 4302 /* fill params struct with the proper values */
4303 params->mode = bond_mode;
169a3e66 4304 params->xmit_policy = xmit_hashtype;
1da177e4 4305 params->miimon = miimon;
ad246c99 4306 params->num_peer_notif = num_peer_notif;
1da177e4 4307 params->arp_interval = arp_interval;
f5b2b966 4308 params->arp_validate = arp_validate_value;
8599b52e 4309 params->arp_all_targets = arp_all_targets_value;
1da177e4
LT		 4310 params->updelay = updelay;
4311 params->downdelay = downdelay;
4312 params->use_carrier = use_carrier;
4313 params->lacp_fast = lacp_fast;
4314 params->primary[0] = 0;
a549952a 4315 params->primary_reselect = primary_reselect_value;
3915c1e8 4316 params->fail_over_mac = fail_over_mac_value;
bb1d9123 4317 params->tx_queues = tx_queues;
ebd8e497 4318 params->all_slaves_active = all_slaves_active;
c2952c31 4319 params->resend_igmp = resend_igmp;
655f8919 4320 params->min_links = min_links;
3a7129e5 4321 params->lp_interval = lp_interval;
809fa972 4322 params->packets_per_slave = packets_per_slave;
e9f0fb88 4323 params->tlb_dynamic_lb = 1; /* Default value */
809fa972
HFS		 4324 if (packets_per_slave > 0) {
4325 params->reciprocal_packets_per_slave =
4326 reciprocal_value(packets_per_slave);
4327 } else {
4328 /* reciprocal_packets_per_slave is unused if
4329 * packets_per_slave is 0 or 1, just initialize it
4330 */
4331 params->reciprocal_packets_per_slave =
4332 (struct reciprocal_value) { 0 };
4333 }
4334
1da177e4
LT		 4335 if (primary) {
4336 strncpy(params->primary, primary, IFNAMSIZ);
4337 params->primary[IFNAMSIZ - 1] = 0;
4338 }
4339
4340 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4341
4342 return 0;
4343}
4344
0daa2303 4345static struct lock_class_key bonding_netdev_xmit_lock_key;
cf508b12 4346static struct lock_class_key bonding_netdev_addr_lock_key;
49ee4920 4347static struct lock_class_key bonding_tx_busylock_key;
0daa2303 4348
e8a0464c
DM		 4349static void bond_set_lockdep_class_one(struct net_device *dev,
4350 struct netdev_queue *txq,
4351 void *_unused)
c773e847
DM		 4352{
4353 lockdep_set_class(&txq->_xmit_lock,
4354 &bonding_netdev_xmit_lock_key);
4355}
4356
4357static void bond_set_lockdep_class(struct net_device *dev)
4358{
cf508b12
DM		 4359 lockdep_set_class(&dev->addr_list_lock,
4360 &bonding_netdev_addr_lock_key);
e8a0464c 4361 netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL);
49ee4920 4362 dev->qdisc_tx_busylock = &bonding_tx_busylock_key;
c773e847
DM		 4363}
4364
181470fc
SH		 4365/*
4366 * Called from registration process
4367 */
4368static int bond_init(struct net_device *bond_dev)
4369{
4370 struct bonding *bond = netdev_priv(bond_dev);
ec87fd3b 4371 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
9fe0617d 4372 struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond));
181470fc
SH		 4373
4374 pr_debug("Begin bond_init for %s\n", bond_dev->name);
4375
9fe0617d
NH		 4376 /*
4377 * Initialize locks that may be required during
4378 * en/deslave operations. All of the bond_open work
4379 * (of which this is part) should really be moved to
4380 * a phase prior to dev_open
4381 */
4382 spin_lock_init(&(bond_info->tx_hashtbl_lock));
4383 spin_lock_init(&(bond_info->rx_hashtbl_lock));
4384
181470fc
SH		 4385 bond->wq = create_singlethread_workqueue(bond_dev->name);
4386 if (!bond->wq)
4387 return -ENOMEM;
4388
4389 bond_set_lockdep_class(bond_dev);
4390
ec87fd3b 4391 list_add_tail(&bond->bond_list, &bn->dev_list);
181470fc 4392
6151b3d4 4393 bond_prepare_sysfs_group(bond);
22bedad3 4394
f073c7ca
TI		 4395 bond_debug_register(bond);
4396
409cc1f8
JP		 4397 /* Ensure valid dev_addr */
4398 if (is_zero_ether_addr(bond_dev->dev_addr) &&
97a1e639 4399 bond_dev->addr_assign_type == NET_ADDR_PERM)
409cc1f8 4400 eth_hw_addr_random(bond_dev);
409cc1f8 4401
181470fc
SH		 4402 return 0;
4403}
4404
0a2a78c4 4405unsigned int bond_get_num_tx_queues(void)
d5da4510 4406{
efacb309 4407 return tx_queues;
d5da4510
JP		 4408}
4409
dfe60397 4410/* Create a new bond based on the specified name and bonding parameters.
e4b91c48 4411 * If name is NULL, obtain a suitable "bond%d" name for us.
dfe60397
MW		 4412 * Caller must NOT hold rtnl_lock; we need to release it here before we
4413 * set up our sysfs entries.
4414 */
ec87fd3b 4415int bond_create(struct net *net, const char *name)
dfe60397
MW		 4416{
4417 struct net_device *bond_dev;
4418 int res;
4419
4420 rtnl_lock();
027ea041 4421
1c5cae81 4422 bond_dev = alloc_netdev_mq(sizeof(struct bonding),
c835a677 4423 name ? name : "bond%d", NET_NAME_UNKNOWN,
1c5cae81 4424 bond_setup, tx_queues);
dfe60397 4425 if (!bond_dev) {
a4aee5c8 4426 pr_err("%s: eek! can't alloc netdev!\n", name);
9e2e61fb
AW		 4427 rtnl_unlock();
4428 return -ENOMEM;
dfe60397
MW		 4429 }
4430
ec87fd3b 4431 dev_net_set(bond_dev, net);
88ead977
EB		 4432 bond_dev->rtnl_link_ops = &bond_link_ops;
4433
dfe60397 4434 res = register_netdevice(bond_dev);
0daa2303 4435
e826eafa
PO		 4436 netif_carrier_off(bond_dev);
4437
7e083840 4438 rtnl_unlock();
9e2e61fb
AW		 4439 if (res < 0)
4440 bond_destructor(bond_dev);
30c15ba9 4441 return res;
dfe60397
MW		 4442}
4443
2c8c1e72 4444static int __net_init bond_net_init(struct net *net)
ec87fd3b 4445{
15449745 4446 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 4447
4448 bn->net = net;
4449 INIT_LIST_HEAD(&bn->dev_list);
4450
ec87fd3b 4451 bond_create_proc_dir(bn);
4c22400a 4452 bond_create_sysfs(bn);
87a7b84b 4453
15449745 4454 return 0;
ec87fd3b
EB		 4455}
4456
2c8c1e72 4457static void __net_exit bond_net_exit(struct net *net)
ec87fd3b 4458{
15449745 4459 struct bond_net *bn = net_generic(net, bond_net_id);
69b0216a 4460 struct bonding *bond, *tmp_bond;
4461 LIST_HEAD(list);
ec87fd3b 4462
4c22400a 4463 bond_destroy_sysfs(bn);
ec87fd3b 4464 bond_destroy_proc_dir(bn);
69b0216a 4465
4466 /* Kill off any bonds created after unregistering bond rtnl ops */
4467 rtnl_lock();
4468 list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list)
4469 unregister_netdevice_queue(bond->dev, &list);
4470 unregister_netdevice_many(&list);
4471 rtnl_unlock();
ec87fd3b
EB		 4472}
4473
4474static struct pernet_operations bond_net_ops = {
4475 .init = bond_net_init,
4476 .exit = bond_net_exit,
15449745
EB		 4477 .id = &bond_net_id,
4478 .size = sizeof(struct bond_net),
ec87fd3b
EB		 4479};
4480
1da177e4
LT		 4481static int __init bonding_init(void)
4482{
1da177e4
LT		 4483 int i;
4484 int res;
4485
bd33acc3 4486 pr_info("%s", bond_version);
1da177e4 4487
dfe60397 4488 res = bond_check_params(&bonding_defaults);
3d632c3f 4489 if (res)
dfe60397 4490 goto out;
1da177e4 4491
15449745 4492 res = register_pernet_subsys(&bond_net_ops);
ec87fd3b
EB		 4493 if (res)
4494 goto out;
027ea041 4495
0a2a78c4 4496 res = bond_netlink_init();
88ead977 4497 if (res)
6639104b 4498 goto err_link;
88ead977 4499
f073c7ca
TI		 4500 bond_create_debugfs();
4501
1da177e4 4502 for (i = 0; i < max_bonds; i++) {
ec87fd3b 4503 res = bond_create(&init_net, NULL);
dfe60397
MW		 4504 if (res)
4505 goto err;
1da177e4
LT		 4506 }
4507
1da177e4 4508 register_netdevice_notifier(&bond_netdev_notifier);
dfe60397 4509out:
1da177e4 4510 return res;
88ead977 4511err:
db298686 4512 bond_destroy_debugfs();
0a2a78c4 4513 bond_netlink_fini();
6639104b 4514err_link:
15449745 4515 unregister_pernet_subsys(&bond_net_ops);
88ead977 4516 goto out;
dfe60397 4517
1da177e4
LT		 4518}
4519
4520static void __exit bonding_exit(void)
4521{
4522 unregister_netdevice_notifier(&bond_netdev_notifier);
4523
f073c7ca 4524 bond_destroy_debugfs();
ae68c398 4525
0a2a78c4 4526 bond_netlink_fini();
ffcdedb6 4527 unregister_pernet_subsys(&bond_net_ops);
e843fa50
NH		 4528
4529#ifdef CONFIG_NET_POLL_CONTROLLER
fb4fa76a
NH		 4530 /*
4531 * Make sure we don't have an imbalance on our netpoll blocking
4532 */
4533 WARN_ON(atomic_read(&netpoll_block_tx));
e843fa50 4534#endif
1da177e4
LT		 4535}
4536
4537module_init(bonding_init);
4538module_exit(bonding_exit);
4539MODULE_LICENSE("GPL");
4540MODULE_VERSION(DRV_VERSION);
4541MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
4542MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
Linux kernel - Deliverables
This page took 1.612481 seconds and 5 git commands to generate.