projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vlan: make a new function vlan_dev_vlan_proto() and export
[deliverable/linux.git] / drivers / net / bonding / bond_main.c
CommitLineData
1da177e4
LT		 1/*
2 * originally based on the dummy device.
3 *
4 * Copyright 1999, Thomas Davis, tadavis@lbl.gov.
5 * Licensed under the GPL. Based on dummy.c, and eql.c devices.
6 *
7 * bonding.c: an Ethernet Bonding driver
8 *
9 * This is useful to talk to a Cisco EtherChannel compatible equipment:
10 * Cisco 5500
11 * Sun Trunking (Solaris)
12 * Alteon AceDirector Trunks
13 * Linux Bonding
14 * and probably many L2 switches ...
15 *
16 * How it works:
17 * ifconfig bond0 ipaddress netmask up
18 * will setup a network device, with an ip address. No mac address
19 * will be assigned at this time. The hw mac address will come from
20 * the first slave bonded to the channel. All slaves will then use
21 * this hw mac address.
22 *
23 * ifconfig bond0 down
24 * will release all slaves, marking them as down.
25 *
26 * ifenslave bond0 eth0
27 * will attach eth0 to bond0 as a slave. eth0 hw mac address will either
28 * a: be used as initial mac address
29 * b: if a hw mac address already is there, eth0's hw mac address
30 * will then be set from bond0.
31 *
1da177e4
LT		 32 */
33
a4aee5c8
JP		 34#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
35
1da177e4
LT		 36#include <linux/kernel.h>
37#include <linux/module.h>
1da177e4
LT		 38#include <linux/types.h>
39#include <linux/fcntl.h>
40#include <linux/interrupt.h>
41#include <linux/ptrace.h>
42#include <linux/ioport.h>
43#include <linux/in.h>
169a3e66 44#include <net/ip.h>
1da177e4 45#include <linux/ip.h>
169a3e66
JV		 46#include <linux/tcp.h>
47#include <linux/udp.h>
1da177e4
LT		 48#include <linux/slab.h>
49#include <linux/string.h>
50#include <linux/init.h>
51#include <linux/timer.h>
52#include <linux/socket.h>
53#include <linux/ctype.h>
54#include <linux/inet.h>
55#include <linux/bitops.h>
3d632c3f 56#include <linux/io.h>
1da177e4 57#include <asm/dma.h>
3d632c3f 58#include <linux/uaccess.h>
1da177e4
LT		 59#include <linux/errno.h>
60#include <linux/netdevice.h>
61#include <linux/inetdevice.h>
a816c7c7 62#include <linux/igmp.h>
1da177e4
LT		 63#include <linux/etherdevice.h>
64#include <linux/skbuff.h>
65#include <net/sock.h>
66#include <linux/rtnetlink.h>
1da177e4
LT		 67#include <linux/smp.h>
68#include <linux/if_ether.h>
69#include <net/arp.h>
70#include <linux/mii.h>
71#include <linux/ethtool.h>
72#include <linux/if_vlan.h>
73#include <linux/if_bonding.h>
b63bb739 74#include <linux/jiffies.h>
e843fa50 75#include <linux/preempt.h>
c3ade5ca 76#include <net/route.h>
457c4cbc 77#include <net/net_namespace.h>
ec87fd3b 78#include <net/netns/generic.h>
5ee31c68 79#include <net/pkt_sched.h>
278b2083 80#include <linux/rculist.h>
32819dc1 81#include <net/flow_keys.h>
1da177e4
LT		 82#include "bonding.h"
83#include "bond_3ad.h"
84#include "bond_alb.h"
85
86/*---------------------------- Module parameters ----------------------------*/
87
88/* monitor all links that often (in milliseconds). <=0 disables monitoring */
1da177e4
LT		 89
90static int max_bonds = BOND_DEFAULT_MAX_BONDS;
bb1d9123 91static int tx_queues = BOND_DEFAULT_TX_QUEUES;
ad246c99 92static int num_peer_notif = 1;
b98d9c66 93static int miimon;
3d632c3f
SH		 94static int updelay;
95static int downdelay;
1da177e4 96static int use_carrier = 1;
3d632c3f
SH		 97static char *mode;
98static char *primary;
a549952a 99static char *primary_reselect;
3d632c3f 100static char *lacp_rate;
655f8919 101static int min_links;
3d632c3f
SH		 102static char *ad_select;
103static char *xmit_hash_policy;
7bdb04ed 104static int arp_interval;
3d632c3f
SH		 105static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
106static char *arp_validate;
8599b52e 107static char *arp_all_targets;
3d632c3f 108static char *fail_over_mac;
b07ea07b 109static int all_slaves_active;
d2991f75 110static struct bond_params bonding_defaults;
c2952c31 111static int resend_igmp = BOND_DEFAULT_RESEND_IGMP;
73958329 112static int packets_per_slave = 1;
3a7129e5 113static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
1da177e4
LT		 114
115module_param(max_bonds, int, 0);
116MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
bb1d9123
AG		 117module_param(tx_queues, int, 0);
118MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)");
ad246c99 119module_param_named(num_grat_arp, num_peer_notif, int, 0644);
90e62474
AG		 120MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on "
121 "failover event (alias of num_unsol_na)");
ad246c99 122module_param_named(num_unsol_na, num_peer_notif, int, 0644);
90e62474
AG		 123MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on "
124 "failover event (alias of num_grat_arp)");
1da177e4
LT		 125module_param(miimon, int, 0);
126MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
127module_param(updelay, int, 0);
128MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
129module_param(downdelay, int, 0);
2ac47660
MW		 130MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
131 "in milliseconds");
1da177e4 132module_param(use_carrier, int, 0);
2ac47660
MW		 133MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
134 "0 for off, 1 for on (default)");
1da177e4 135module_param(mode, charp, 0);
90e62474 136MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, "
2ac47660
MW		 137 "1 for active-backup, 2 for balance-xor, "
138 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
139 "6 for balance-alb");
1da177e4
LT		 140module_param(primary, charp, 0);
141MODULE_PARM_DESC(primary, "Primary network device to use");
a549952a
JP		 142module_param(primary_reselect, charp, 0);
143MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
144 "once it comes up; "
145 "0 for always (default), "
146 "1 for only if speed of primary is "
147 "better, "
148 "2 for only on active slave "
149 "failure");
1da177e4 150module_param(lacp_rate, charp, 0);
90e62474
AG		 151MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; "
152 "0 for slow, 1 for fast");
fd989c83 153module_param(ad_select, charp, 0);
90e62474
AG		 154MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; "
155 "0 for stable (default), 1 for bandwidth, "
156 "2 for count");
655f8919 157module_param(min_links, int, 0);
158MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier");
159
169a3e66 160module_param(xmit_hash_policy, charp, 0);
90e62474
AG		 161MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; "
162 "0 for layer 2 (default), 1 for layer 3+4, "
32819dc1
NA		 163 "2 for layer 2+3, 3 for encap layer 2+3, "
164 "4 for encap layer 3+4");
1da177e4
LT		 165module_param(arp_interval, int, 0);
166MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
167module_param_array(arp_ip_target, charp, NULL, 0);
168MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
f5b2b966 169module_param(arp_validate, charp, 0);
90e62474
AG		 170MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; "
171 "0 for none (default), 1 for active, "
172 "2 for backup, 3 for all");
8599b52e
VF		 173module_param(arp_all_targets, charp, 0);
174MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all");
3915c1e8 175module_param(fail_over_mac, charp, 0);
90e62474
AG		 176MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to "
177 "the same MAC; 0 for none (default), "
178 "1 for active, 2 for follow");
ebd8e497
AG		 179module_param(all_slaves_active, int, 0);
180MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface"
90e62474 181 "by setting active flag for all slaves; "
ebd8e497 182 "0 for never (default), 1 for always.");
c2952c31 183module_param(resend_igmp, int, 0);
90e62474
AG		 184MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on "
185 "link failure");
73958329
NA		 186module_param(packets_per_slave, int, 0);
187MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr "
188 "mode; 0 for a random slave, 1 packet per "
189 "slave (default), >1 packets per slave.");
3a7129e5 190module_param(lp_interval, uint, 0);
191MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where "
192 "the bonding driver sends learning packets to "
193 "each slaves peer switch. The default is 1.");
1da177e4
LT		 194
195/*----------------------------- Global variables ----------------------------*/
196
e843fa50 197#ifdef CONFIG_NET_POLL_CONTROLLER
fb4fa76a 198atomic_t netpoll_block_tx = ATOMIC_INIT(0);
e843fa50
NH		 199#endif
200
f99189b1 201int bond_net_id __read_mostly;
1da177e4 202
3d632c3f
SH		 203static __be32 arp_target[BOND_MAX_ARP_TARGETS];
204static int arp_ip_count;
1da177e4 205static int bond_mode = BOND_MODE_ROUNDROBIN;
3d632c3f
SH		 206static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
207static int lacp_fast;
1da177e4 208
1da177e4
LT		 209/*-------------------------- Forward declarations ---------------------------*/
210
181470fc 211static int bond_init(struct net_device *bond_dev);
c67dfb29 212static void bond_uninit(struct net_device *bond_dev);
1da177e4
LT		 213
214/*---------------------------- General routines -----------------------------*/
215
bd33acc3 216const char *bond_mode_name(int mode)
1da177e4 217{
77afc92b
HE		 218 static const char *names[] = {
219 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
220 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
221 [BOND_MODE_XOR] = "load balancing (xor)",
222 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
3d632c3f 223 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
77afc92b
HE		 224 [BOND_MODE_TLB] = "transmit load balancing",
225 [BOND_MODE_ALB] = "adaptive load balancing",
226 };
227
f5280948 228 if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB)
1da177e4 229 return "unknown";
77afc92b
HE		 230
231 return names[mode];
1da177e4
LT		 232}
233
234/*---------------------------------- VLAN -----------------------------------*/
235
1da177e4
LT		 236/**
237 * bond_dev_queue_xmit - Prepare skb for xmit.
3d632c3f 238 *
1da177e4
LT		 239 * @bond: bond device that got this skb for tx.
240 * @skb: hw accel VLAN tagged skb to transmit
241 * @slave_dev: slave that is supposed to xmit this skbuff
1da177e4 242 */
d316dedd 243void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
3d632c3f 244 struct net_device *slave_dev)
1da177e4 245{
8387451e 246 skb->dev = slave_dev;
374eeb5a 247
5ee31c68 248 BUILD_BUG_ON(sizeof(skb->queue_mapping) !=
df4ab5b3
JP		 249 sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping));
250 skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping;
374eeb5a 251
e15c3c22 252 if (unlikely(netpoll_tx_running(bond->dev)))
8a8efa22 253 bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb);
080e4130 254 else
f6dc31a8 255 dev_queue_xmit(skb);
1da177e4
LT		 256}
257
258/*
cc0e4070
JP		 259 * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid,
260 * We don't protect the slave list iteration with a lock because:
1da177e4
LT		 261 * a. This operation is performed in IOCTL context,
262 * b. The operation is protected by the RTNL semaphore in the 8021q code,
263 * c. Holding a lock with BH disabled while directly calling a base driver
264 * entry point is generally a BAD idea.
3d632c3f 265 *
1da177e4
LT		 266 * The design of synchronization/protection for this operation in the 8021q
267 * module is good for one or more VLAN devices over a single physical device
268 * and cannot be extended for a teaming solution like bonding, so there is a
269 * potential race condition here where a net device from the vlan group might
270 * be referenced (either by a base driver or the 8021q code) while it is being
271 * removed from the system. However, it turns out we're not making matters
272 * worse, and if it works for regular VLAN usage it will work here too.
273*/
274
1da177e4
LT		 275/**
276 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
277 * @bond_dev: bonding net device that got called
278 * @vid: vlan id being added
279 */
80d5c368
PM		 280static int bond_vlan_rx_add_vid(struct net_device *bond_dev,
281 __be16 proto, u16 vid)
1da177e4 282{
454d7c9b 283 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 284 struct slave *slave, *rollback_slave;
9caff1e7 285 struct list_head *iter;
dec1e90e 286 int res;
1da177e4 287
9caff1e7 288 bond_for_each_slave(bond, slave, iter) {
80d5c368 289 res = vlan_vid_add(slave->dev, proto, vid);
87002b03
JP		 290 if (res)
291 goto unwind;
1da177e4
LT		 292 }
293
8e586137 294 return 0;
87002b03
JP		 295
296unwind:
81f23b13 297 /* unwind to the slave that failed */
9caff1e7 298 bond_for_each_slave(bond, rollback_slave, iter) {
81f23b13
VF		 299 if (rollback_slave == slave)
300 break;
301
302 vlan_vid_del(rollback_slave->dev, proto, vid);
303 }
87002b03
JP		 304
305 return res;
1da177e4
LT		 306}
307
308/**
309 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
310 * @bond_dev: bonding net device that got called
311 * @vid: vlan id being removed
312 */
80d5c368
PM		 313static int bond_vlan_rx_kill_vid(struct net_device *bond_dev,
314 __be16 proto, u16 vid)
1da177e4 315{
454d7c9b 316 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 317 struct list_head *iter;
1da177e4 318 struct slave *slave;
1da177e4 319
9caff1e7 320 bond_for_each_slave(bond, slave, iter)
80d5c368 321 vlan_vid_del(slave->dev, proto, vid);
1da177e4 322
e868b0c9
VF		 323 if (bond_is_lb(bond))
324 bond_alb_clear_vlan(bond, vid);
8e586137
JP		 325
326 return 0;
1da177e4
LT		 327}
328
1da177e4
LT		 329/*------------------------------- Link status -------------------------------*/
330
ff59c456
JV		 331/*
332 * Set the carrier state for the master according to the state of its
333 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
334 * do special 802.3ad magic.
335 *
336 * Returns zero if carrier state does not change, nonzero if it does.
337 */
338static int bond_set_carrier(struct bonding *bond)
339{
9caff1e7 340 struct list_head *iter;
ff59c456 341 struct slave *slave;
ff59c456 342
0965a1f3 343 if (!bond_has_slaves(bond))
ff59c456
JV		 344 goto down;
345
346 if (bond->params.mode == BOND_MODE_8023AD)
347 return bond_3ad_set_carrier(bond);
348
9caff1e7 349 bond_for_each_slave(bond, slave, iter) {
ff59c456
JV		 350 if (slave->link == BOND_LINK_UP) {
351 if (!netif_carrier_ok(bond->dev)) {
352 netif_carrier_on(bond->dev);
353 return 1;
354 }
355 return 0;
356 }
357 }
358
359down:
360 if (netif_carrier_ok(bond->dev)) {
361 netif_carrier_off(bond->dev);
362 return 1;
363 }
364 return 0;
365}
366
1da177e4
LT		 367/*
368 * Get link speed and duplex from the slave's base driver
369 * using ethtool. If for some reason the call fails or the
98f41f69 370 * values are invalid, set speed and duplex to -1,
13b95fb7 371 * and return.
1da177e4 372 */
13b95fb7 373static void bond_update_speed_duplex(struct slave *slave)
1da177e4
LT		 374{
375 struct net_device *slave_dev = slave->dev;
4bc71cb9 376 struct ethtool_cmd ecmd;
5d30530e 377 u32 slave_speed;
61a44b9c 378 int res;
1da177e4 379
589665f5
DC		 380 slave->speed = SPEED_UNKNOWN;
381 slave->duplex = DUPLEX_UNKNOWN;
1da177e4 382
4bc71cb9 383 res = __ethtool_get_settings(slave_dev, &ecmd);
61a44b9c 384 if (res < 0)
13b95fb7 385 return;
1da177e4 386
4bc71cb9 387 slave_speed = ethtool_cmd_speed(&ecmd);
6f92c66f 388 if (slave_speed == 0 || slave_speed == ((__u32) -1))
13b95fb7 389 return;
1da177e4 390
4bc71cb9 391 switch (ecmd.duplex) {
1da177e4
LT		 392 case DUPLEX_FULL:
393 case DUPLEX_HALF:
394 break;
395 default:
13b95fb7 396 return;
1da177e4
LT		 397 }
398
5d30530e 399 slave->speed = slave_speed;
4bc71cb9 400 slave->duplex = ecmd.duplex;
1da177e4 401
13b95fb7 402 return;
1da177e4
LT		 403}
404
07699f9a 405const char *bond_slave_link_status(s8 link)
406{
407 switch (link) {
408 case BOND_LINK_UP:
409 return "up";
410 case BOND_LINK_FAIL:
411 return "going down";
412 case BOND_LINK_DOWN:
413 return "down";
414 case BOND_LINK_BACK:
415 return "going back";
416 default:
417 return "unknown";
418 }
419}
420
1da177e4
LT		 421/*
422 * if <dev> supports MII link status reporting, check its link status.
423 *
424 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
3d632c3f 425 * depending upon the setting of the use_carrier parameter.
1da177e4
LT		 426 *
427 * Return either BMSR_LSTATUS, meaning that the link is up (or we
428 * can't tell and just pretend it is), or 0, meaning that the link is
429 * down.
430 *
431 * If reporting is non-zero, instead of faking link up, return -1 if
432 * both ETHTOOL and MII ioctls fail (meaning the device does not
433 * support them). If use_carrier is set, return whatever it says.
434 * It'd be nice if there was a good way to tell if a driver supports
435 * netif_carrier, but there really isn't.
436 */
3d632c3f
SH		 437static int bond_check_dev_link(struct bonding *bond,
438 struct net_device *slave_dev, int reporting)
1da177e4 439{
eb7cc59a 440 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
d9d52832 441 int (*ioctl)(struct net_device *, struct ifreq *, int);
1da177e4
LT		 442 struct ifreq ifr;
443 struct mii_ioctl_data *mii;
1da177e4 444
6c988853
PG		 445 if (!reporting && !netif_running(slave_dev))
446 return 0;
447
eb7cc59a 448 if (bond->params.use_carrier)
1da177e4 449 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
1da177e4 450
29112f4e 451 /* Try to get link status using Ethtool first. */
c772dde3
BH		 452 if (slave_dev->ethtool_ops->get_link)
453 return slave_dev->ethtool_ops->get_link(slave_dev) ?
454 BMSR_LSTATUS : 0;
29112f4e 455
3d632c3f 456 /* Ethtool can't be used, fallback to MII ioctls. */
eb7cc59a 457 ioctl = slave_ops->ndo_do_ioctl;
1da177e4
LT		 458 if (ioctl) {
459 /* TODO: set pointer to correct ioctl on a per team member */
460 /* bases to make this more efficient. that is, once */
461 /* we determine the correct ioctl, we will always */
462 /* call it and not the others for that team */
463 /* member. */
464
465 /*
466 * We cannot assume that SIOCGMIIPHY will also read a
467 * register; not all network drivers (e.g., e100)
468 * support that.
469 */
470
471 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
472 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
473 mii = if_mii(&ifr);
474 if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
475 mii->reg_num = MII_BMSR;
3d632c3f
SH		 476 if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0)
477 return mii->val_out & BMSR_LSTATUS;
1da177e4
LT		 478 }
479 }
480
1da177e4
LT		 481 /*
482 * If reporting, report that either there's no dev->do_ioctl,
61a44b9c 483 * or both SIOCGMIIREG and get_link failed (meaning that we
1da177e4
LT		 484 * cannot report link status). If not reporting, pretend
485 * we're ok.
486 */
3d632c3f 487 return reporting ? -1 : BMSR_LSTATUS;
1da177e4
LT		 488}
489
490/*----------------------------- Multicast list ------------------------------*/
491
1da177e4
LT		 492/*
493 * Push the promiscuity flag down to appropriate slaves
494 */
7e1a1ac1 495static int bond_set_promiscuity(struct bonding *bond, int inc)
1da177e4 496{
9caff1e7 497 struct list_head *iter;
7e1a1ac1 498 int err = 0;
9caff1e7 499
1da177e4
LT		 500 if (USES_PRIMARY(bond->params.mode)) {
501 /* write lock already acquired */
502 if (bond->curr_active_slave) {
7e1a1ac1
WC		 503 err = dev_set_promiscuity(bond->curr_active_slave->dev,
504 inc);
1da177e4
LT		 505 }
506 } else {
507 struct slave *slave;
dec1e90e 508
9caff1e7 509 bond_for_each_slave(bond, slave, iter) {
7e1a1ac1
WC		 510 err = dev_set_promiscuity(slave->dev, inc);
511 if (err)
512 return err;
1da177e4
LT		 513 }
514 }
7e1a1ac1 515 return err;
1da177e4
LT		 516}
517
518/*
519 * Push the allmulti flag down to all slaves
520 */
7e1a1ac1 521static int bond_set_allmulti(struct bonding *bond, int inc)
1da177e4 522{
9caff1e7 523 struct list_head *iter;
7e1a1ac1 524 int err = 0;
9caff1e7 525
1da177e4
LT		 526 if (USES_PRIMARY(bond->params.mode)) {
527 /* write lock already acquired */
528 if (bond->curr_active_slave) {
7e1a1ac1
WC		 529 err = dev_set_allmulti(bond->curr_active_slave->dev,
530 inc);
1da177e4
LT		 531 }
532 } else {
533 struct slave *slave;
dec1e90e 534
9caff1e7 535 bond_for_each_slave(bond, slave, iter) {
7e1a1ac1
WC		 536 err = dev_set_allmulti(slave->dev, inc);
537 if (err)
538 return err;
1da177e4
LT		 539 }
540 }
7e1a1ac1 541 return err;
1da177e4
LT		 542}
543
5a37e8ca
FL		 544/*
545 * Retrieve the list of registered multicast addresses for the bonding
546 * device and retransmit an IGMP JOIN request to the current active
547 * slave.
548 */
f2369109 549static void bond_resend_igmp_join_requests_delayed(struct work_struct *work)
5a37e8ca 550{
f2369109 551 struct bonding *bond = container_of(work, struct bonding,
552 mcast_work.work);
553
4aa5dee4 554 if (!rtnl_trylock()) {
4beac029 555 queue_delayed_work(bond->wq, &bond->mcast_work, 1);
4aa5dee4 556 return;
5a37e8ca 557 }
4aa5dee4 558 call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev);
5a37e8ca 559
4f5474e7
NA		 560 if (bond->igmp_retrans > 1) {
561 bond->igmp_retrans--;
c2952c31 562 queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5);
4f5474e7 563 }
f2369109 564 rtnl_unlock();
5a37e8ca
FL		 565}
566
303d1cbf 567/* Flush bond's hardware addresses from slave
1da177e4 568 */
303d1cbf 569static void bond_hw_addr_flush(struct net_device *bond_dev,
3d632c3f 570 struct net_device *slave_dev)
1da177e4 571{
454d7c9b 572 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 573
303d1cbf
JV		 574 dev_uc_unsync(slave_dev, bond_dev);
575 dev_mc_unsync(slave_dev, bond_dev);
1da177e4
LT		 576
577 if (bond->params.mode == BOND_MODE_8023AD) {
578 /* del lacpdu mc addr from mc list */
579 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
580
22bedad3 581 dev_mc_del(slave_dev, lacpdu_multicast);
1da177e4
LT		 582 }
583}
584
585/*--------------------------- Active slave change ---------------------------*/
586
303d1cbf
JV		 587/* Update the hardware address list and promisc/allmulti for the new and
588 * old active slaves (if any). Modes that are !USES_PRIMARY keep all
589 * slaves up date at all times; only the USES_PRIMARY modes need to call
590 * this function to swap these settings during a failover.
1da177e4 591 */
303d1cbf
JV		 592static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active,
593 struct slave *old_active)
1da177e4 594{
7864a1ad
VF		 595 ASSERT_RTNL();
596
1da177e4 597 if (old_active) {
3d632c3f 598 if (bond->dev->flags & IFF_PROMISC)
1da177e4 599 dev_set_promiscuity(old_active->dev, -1);
1da177e4 600
3d632c3f 601 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 602 dev_set_allmulti(old_active->dev, -1);
1da177e4 603
303d1cbf 604 bond_hw_addr_flush(bond->dev, old_active->dev);
1da177e4
LT		 605 }
606
607 if (new_active) {
7e1a1ac1 608 /* FIXME: Signal errors upstream. */
3d632c3f 609 if (bond->dev->flags & IFF_PROMISC)
1da177e4 610 dev_set_promiscuity(new_active->dev, 1);
1da177e4 611
3d632c3f 612 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 613 dev_set_allmulti(new_active->dev, 1);
1da177e4 614
d632ce98 615 netif_addr_lock_bh(bond->dev);
303d1cbf
JV		 616 dev_uc_sync(new_active->dev, bond->dev);
617 dev_mc_sync(new_active->dev, bond->dev);
d632ce98 618 netif_addr_unlock_bh(bond->dev);
1da177e4
LT		 619 }
620}
621
ae0d6750 622/**
623 * bond_set_dev_addr - clone slave's address to bond
624 * @bond_dev: bond net device
625 * @slave_dev: slave net device
626 *
627 * Should be called with RTNL held.
628 */
629static void bond_set_dev_addr(struct net_device *bond_dev,
630 struct net_device *slave_dev)
631{
008aebde
NA		 632 pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n",
633 bond_dev, slave_dev, slave_dev->addr_len);
ae0d6750 634 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
635 bond_dev->addr_assign_type = NET_ADDR_STOLEN;
636 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev);
637}
638
3915c1e8
JV		 639/*
640 * bond_do_fail_over_mac
641 *
642 * Perform special MAC address swapping for fail_over_mac settings
643 *
b2e7aceb 644 * Called with RTNL, curr_slave_lock for write_bh.
3915c1e8
JV		 645 */
646static void bond_do_fail_over_mac(struct bonding *bond,
647 struct slave *new_active,
648 struct slave *old_active)
1f78d9f9 649 __releases(&bond->curr_slave_lock)
1f78d9f9 650 __acquires(&bond->curr_slave_lock)
3915c1e8
JV		 651{
652 u8 tmp_mac[ETH_ALEN];
653 struct sockaddr saddr;
654 int rv;
655
656 switch (bond->params.fail_over_mac) {
657 case BOND_FOM_ACTIVE:
7d26bb10 658 if (new_active) {
7d26bb10 659 write_unlock_bh(&bond->curr_slave_lock);
ae0d6750 660 bond_set_dev_addr(bond->dev, new_active->dev);
7d26bb10
WP		 661 write_lock_bh(&bond->curr_slave_lock);
662 }
3915c1e8
JV		 663 break;
664 case BOND_FOM_FOLLOW:
665 /*
666 * if new_active && old_active, swap them
667 * if just old_active, do nothing (going to no active slave)
668 * if just new_active, set new_active to bond's MAC
669 */
670 if (!new_active)
671 return;
672
673 write_unlock_bh(&bond->curr_slave_lock);
3915c1e8
JV		 674
675 if (old_active) {
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;
750 if (slave->link == BOND_LINK_BACK && IS_UP(slave->dev) &&
751 slave->delay < mintime) {
752 mintime = slave->delay;
753 bestslave = slave;
1da177e4
LT		 754 }
755 }
756
757 return bestslave;
758}
759
ad246c99
BH		 760static bool bond_should_notify_peers(struct bonding *bond)
761{
4cb4f97b 762 struct slave *slave;
763
764 rcu_read_lock();
765 slave = rcu_dereference(bond->curr_active_slave);
766 rcu_read_unlock();
ad246c99
BH		 767
768 pr_debug("bond_should_notify_peers: bond %s slave %s\n",
769 bond->dev->name, slave ? slave->dev->name : "NULL");
770
771 if (!slave || !bond->send_peer_notif ||
772 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
773 return false;
774
ad246c99
BH		 775 return true;
776}
777
1da177e4
LT		 778/**
779 * change_active_interface - change the active slave into the specified one
780 * @bond: our bonding struct
781 * @new: the new slave to make the active one
782 *
783 * Set the new slave to the bond's settings and unset them on the old
784 * curr_active_slave.
785 * Setting include flags, mc-list, promiscuity, allmulti, etc.
786 *
787 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
788 * because it is apparently the best available slave we have, even though its
789 * updelay hasn't timed out yet.
790 *
b2e7aceb 791 * If new_active is not NULL, caller must hold curr_slave_lock for write_bh.
1da177e4 792 */
a77b5325 793void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
1da177e4
LT		 794{
795 struct slave *old_active = bond->curr_active_slave;
796
3d632c3f 797 if (old_active == new_active)
1da177e4 798 return;
1da177e4
LT		 799
800 if (new_active) {
8e603460 801 new_active->last_link_up = jiffies;
b2220cad 802
1da177e4
LT		 803 if (new_active->link == BOND_LINK_BACK) {
804 if (USES_PRIMARY(bond->params.mode)) {
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
3d632c3f 813 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 814 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
1da177e4 815
58402054 816 if (bond_is_lb(bond))
1da177e4 817 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
1da177e4
LT		 818 } else {
819 if (USES_PRIMARY(bond->params.mode)) {
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
3d632c3f 826 if (USES_PRIMARY(bond->params.mode))
303d1cbf 827 bond_hw_addr_swap(bond, new_active, old_active);
1da177e4 828
58402054 829 if (bond_is_lb(bond)) {
1da177e4 830 bond_alb_handle_active_change(bond, new_active);
8f903c70 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
JV		 840
841 if (bond->params.mode == 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) &&
879 ((USES_PRIMARY(bond->params.mode) && new_active) ||
880 bond->params.mode == 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
47be03a2 925 np = kzalloc(sizeof(*np), GFP_ATOMIC);
8a8efa22
AW		 926 err = -ENOMEM;
927 if (!np)
928 goto out;
929
47be03a2 930 err = __netpoll_setup(np, slave->dev, GFP_ATOMIC);
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)
8a8efa22
AW		 961 if (IS_UP(slave->dev))
962 slave_disable_netpoll(slave);
f6dc31a8 963}
8a8efa22 964
47be03a2 965static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp)
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
0965a1f3 1004 if (!bond_has_slaves(bond)) {
b2a103e6
MM		 1005 /* Disable adding VLANs to empty bond. But why? --mq */
1006 features |= NETIF_F_VLAN_CHALLENGED;
9b7b165a 1007 return features;
b2a103e6 1008 }
b63365a2 1009
b2a103e6 1010 mask = features;
b63365a2 1011 features &= ~NETIF_F_ONE_FOR_ALL;
b2a103e6 1012 features |= NETIF_F_ALL_FOR_ALL;
7f353bf2 1013
9caff1e7 1014 bond_for_each_slave(bond, slave, iter) {
b63365a2
HX		 1015 features = netdev_increment_features(features,
1016 slave->dev->features,
b2a103e6
MM		 1017 mask);
1018 }
b0ce3508 1019 features = netdev_add_tso_features(features, mask);
b2a103e6 1020
b2a103e6
MM		 1021 return features;
1022}
1023
62f2a3a4
MM		 1024#define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \
1025 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
1026 NETIF_F_HIGHDMA | NETIF_F_LRO)
b2a103e6
MM		 1027
1028static void bond_compute_features(struct bonding *bond)
1029{
c4826861 1030 unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE;
c8f44aff 1031 netdev_features_t vlan_features = BOND_VLAN_FEATURES;
9caff1e7
VF		 1032 struct net_device *bond_dev = bond->dev;
1033 struct list_head *iter;
1034 struct slave *slave;
b2a103e6 1035 unsigned short max_hard_header_len = ETH_HLEN;
0e376bd0
SB		 1036 unsigned int gso_max_size = GSO_MAX_SIZE;
1037 u16 gso_max_segs = GSO_MAX_SEGS;
b2a103e6 1038
0965a1f3 1039 if (!bond_has_slaves(bond))
b2a103e6
MM		 1040 goto done;
1041
9caff1e7 1042 bond_for_each_slave(bond, slave, iter) {
278339a4 1043 vlan_features = netdev_increment_features(vlan_features,
b2a103e6
MM		 1044 slave->dev->vlan_features, BOND_VLAN_FEATURES);
1045
b6fe83e9 1046 dst_release_flag &= slave->dev->priv_flags;
54ef3137
JV		 1047 if (slave->dev->hard_header_len > max_hard_header_len)
1048 max_hard_header_len = slave->dev->hard_header_len;
0e376bd0
SB		 1049
1050 gso_max_size = min(gso_max_size, slave->dev->gso_max_size);
1051 gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs);
54ef3137 1052 }
8531c5ff 1053
b63365a2 1054done:
b2a103e6 1055 bond_dev->vlan_features = vlan_features;
54ef3137 1056 bond_dev->hard_header_len = max_hard_header_len;
0e376bd0
SB		 1057 bond_dev->gso_max_segs = gso_max_segs;
1058 netif_set_gso_max_size(bond_dev, gso_max_size);
8531c5ff 1059
b6fe83e9
ED		 1060 flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE;
1061 bond_dev->priv_flags = flags | dst_release_flag;
1062
b2a103e6 1063 netdev_change_features(bond_dev);
8531c5ff
AK		 1064}
1065
872254dd
MS		 1066static void bond_setup_by_slave(struct net_device *bond_dev,
1067 struct net_device *slave_dev)
1068{
00829823 1069 bond_dev->header_ops = slave_dev->header_ops;
872254dd
MS		 1070
1071 bond_dev->type = slave_dev->type;
1072 bond_dev->hard_header_len = slave_dev->hard_header_len;
1073 bond_dev->addr_len = slave_dev->addr_len;
1074
1075 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1076 slave_dev->addr_len);
1077}
1078
5b2c4dd2 1079/* On bonding slaves other than the currently active slave, suppress
3aba891d 1080 * duplicates except for alb non-mcast/bcast.
5b2c4dd2
JP		 1081 */
1082static bool bond_should_deliver_exact_match(struct sk_buff *skb,
0bd80dad
JP		 1083 struct slave *slave,
1084 struct bonding *bond)
5b2c4dd2 1085{
2d7011ca 1086 if (bond_is_slave_inactive(slave)) {
0bd80dad 1087 if (bond->params.mode == BOND_MODE_ALB &&
5b2c4dd2
JP		 1088 skb->pkt_type != PACKET_BROADCAST &&
1089 skb->pkt_type != PACKET_MULTICAST)
5b2c4dd2 1090 return false;
5b2c4dd2
JP		 1091 return true;
1092 }
1093 return false;
1094}
1095
8a4eb573 1096static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb)
5b2c4dd2 1097{
8a4eb573 1098 struct sk_buff *skb = *pskb;
f1c1775a 1099 struct slave *slave;
0bd80dad 1100 struct bonding *bond;
de063b70
ED		 1101 int (*recv_probe)(const struct sk_buff *, struct bonding *,
1102 struct slave *);
13a8e0c8 1103 int ret = RX_HANDLER_ANOTHER;
5b2c4dd2 1104
8a4eb573
JP		 1105 skb = skb_share_check(skb, GFP_ATOMIC);
1106 if (unlikely(!skb))
1107 return RX_HANDLER_CONSUMED;
1108
1109 *pskb = skb;
5b2c4dd2 1110
35d48903
JP		 1111 slave = bond_slave_get_rcu(skb->dev);
1112 bond = slave->bond;
0bd80dad 1113
4d97480b
MH		 1114 recv_probe = ACCESS_ONCE(bond->recv_probe);
1115 if (recv_probe) {
de063b70
ED		 1116 ret = recv_probe(skb, bond, slave);
1117 if (ret == RX_HANDLER_CONSUMED) {
1118 consume_skb(skb);
1119 return ret;
3aba891d
JP		 1120 }
1121 }
1122
0bd80dad 1123 if (bond_should_deliver_exact_match(skb, slave, bond)) {
8a4eb573 1124 return RX_HANDLER_EXACT;
5b2c4dd2
JP		 1125 }
1126
35d48903 1127 skb->dev = bond->dev;
5b2c4dd2 1128
0bd80dad 1129 if (bond->params.mode == BOND_MODE_ALB &&
35d48903 1130 bond->dev->priv_flags & IFF_BRIDGE_PORT &&
5b2c4dd2 1131 skb->pkt_type == PACKET_HOST) {
5b2c4dd2 1132
541ac7c9
CG		 1133 if (unlikely(skb_cow_head(skb,
1134 skb->data - skb_mac_header(skb)))) {
1135 kfree_skb(skb);
8a4eb573 1136 return RX_HANDLER_CONSUMED;
541ac7c9 1137 }
2a7c183b 1138 ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr);
5b2c4dd2
JP		 1139 }
1140
13a8e0c8 1141 return ret;
5b2c4dd2
JP		 1142}
1143
471cb5a3 1144static int bond_master_upper_dev_link(struct net_device *bond_dev,
1f718f0f
VF		 1145 struct net_device *slave_dev,
1146 struct slave *slave)
471cb5a3
JP		 1147{
1148 int err;
1149
1f718f0f 1150 err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave);
471cb5a3
JP		 1151 if (err)
1152 return err;
1153 slave_dev->flags |= IFF_SLAVE;
7f294054 1154 rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP		 1155 return 0;
1156}
1157
1158static void bond_upper_dev_unlink(struct net_device *bond_dev,
1159 struct net_device *slave_dev)
1160{
1161 netdev_upper_dev_unlink(slave_dev, bond_dev);
1162 slave_dev->flags &= ~IFF_SLAVE;
7f294054 1163 rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL);
471cb5a3
JP		 1164}
1165
1da177e4 1166/* enslave device <slave> to bond device <master> */
a77b5325 1167int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1168{
454d7c9b 1169 struct bonding *bond = netdev_priv(bond_dev);
eb7cc59a 1170 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
c8c23903 1171 struct slave *new_slave = NULL, *prev_slave;
1da177e4
LT		 1172 struct sockaddr addr;
1173 int link_reporting;
8599b52e 1174 int res = 0, i;
1da177e4 1175
c772dde3
BH		 1176 if (!bond->params.use_carrier &&
1177 slave_dev->ethtool_ops->get_link == NULL &&
1178 slave_ops->ndo_do_ioctl == NULL) {
91565ebb
JP		 1179 pr_warn("%s: Warning: no link monitoring support for %s\n",
1180 bond_dev->name, slave_dev->name);
1da177e4
LT		 1181 }
1182
1da177e4
LT		 1183 /* already enslaved */
1184 if (slave_dev->flags & IFF_SLAVE) {
90194264 1185 pr_debug("Error: Device was already enslaved\n");
1da177e4
LT		 1186 return -EBUSY;
1187 }
1188
09a89c21
JB		 1189 if (bond_dev == slave_dev) {
1190 pr_err("%s: cannot enslave bond to itself.\n", bond_dev->name);
1191 return -EPERM;
1192 }
1193
1da177e4
LT		 1194 /* vlan challenged mutual exclusion */
1195 /* no need to lock since we're protected by rtnl_lock */
1196 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 1197 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
55462cf3 1198 if (vlan_uses_dev(bond_dev)) {
a4aee5c8
JP		 1199 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1200 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1201 return -EPERM;
1202 } else {
91565ebb
JP		 1203 pr_warn("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1204 bond_dev->name, slave_dev->name,
1205 slave_dev->name, bond_dev->name);
1da177e4
LT		 1206 }
1207 } else {
5a03cdb7 1208 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4
LT		 1209 }
1210
217df670
JV		 1211 /*
1212 * Old ifenslave binaries are no longer supported. These can
3d632c3f 1213 * be identified with moderate accuracy by the state of the slave:
217df670
JV		 1214 * the current ifenslave will set the interface down prior to
1215 * enslaving it; the old ifenslave will not.
1216 */
1217 if ((slave_dev->flags & IFF_UP)) {
90194264 1218 pr_err("%s is up - this may be due to an out of date ifenslave\n",
217df670
JV		 1219 slave_dev->name);
1220 res = -EPERM;
1221 goto err_undo_flags;
1222 }
1da177e4 1223
872254dd
MS		 1224 /* set bonding device ether type by slave - bonding netdevices are
1225 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1226 * there is a need to override some of the type dependent attribs/funcs.
1227 *
1228 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1229 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1230 */
0965a1f3 1231 if (!bond_has_slaves(bond)) {
e36b9d16 1232 if (bond_dev->type != slave_dev->type) {
e36b9d16 1233 pr_debug("%s: change device type from %d to %d\n",
a4aee5c8
JP		 1234 bond_dev->name,
1235 bond_dev->type, slave_dev->type);
75c78500 1236
b7bc2a5b
AW		 1237 res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE,
1238 bond_dev);
3ca5b404
JP		 1239 res = notifier_to_errno(res);
1240 if (res) {
1241 pr_err("%s: refused to change device type\n",
1242 bond_dev->name);
1243 res = -EBUSY;
1244 goto err_undo_flags;
1245 }
75c78500 1246
32a806c1 1247 /* Flush unicast and multicast addresses */
a748ee24 1248 dev_uc_flush(bond_dev);
22bedad3 1249 dev_mc_flush(bond_dev);
32a806c1 1250
e36b9d16
MS		 1251 if (slave_dev->type != ARPHRD_ETHER)
1252 bond_setup_by_slave(bond_dev, slave_dev);
550fd08c 1253 else {
e36b9d16 1254 ether_setup(bond_dev);
550fd08c
NH		 1255 bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1256 }
75c78500 1257
b7bc2a5b
AW		 1258 call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE,
1259 bond_dev);
e36b9d16 1260 }
872254dd 1261 } else if (bond_dev->type != slave_dev->type) {
90194264
JP		 1262 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it\n",
1263 slave_dev->name, slave_dev->type, bond_dev->type);
a4aee5c8
JP		 1264 res = -EINVAL;
1265 goto err_undo_flags;
872254dd
MS		 1266 }
1267
eb7cc59a 1268 if (slave_ops->ndo_set_mac_address == NULL) {
0965a1f3 1269 if (!bond_has_slaves(bond)) {
90194264 1270 pr_warn("%s: Warning: The first slave device specified does not support setting the MAC address\n",
00503b6f 1271 bond_dev->name);
1272 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
1273 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
90194264 1274 pr_warn("%s: Setting fail_over_mac to active for active-backup mode\n",
00503b6f 1275 bond_dev->name);
1276 }
3915c1e8 1277 } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
90194264 1278 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 1279 bond_dev->name);
2ab82852
MS		 1280 res = -EOPNOTSUPP;
1281 goto err_undo_flags;
1282 }
1da177e4
LT		 1283 }
1284
8d8fc29d
AW		 1285 call_netdevice_notifiers(NETDEV_JOIN, slave_dev);
1286
c20811a7
JP		 1287 /* If this is the first slave, then we need to set the master's hardware
1288 * address to be the same as the slave's. */
0965a1f3 1289 if (!bond_has_slaves(bond) &&
dec1e90e 1290 bond->dev->addr_assign_type == NET_ADDR_RANDOM)
409cc1f8 1291 bond_set_dev_addr(bond->dev, slave_dev);
c20811a7 1292
243cb4e5 1293 new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL);
1da177e4
LT		 1294 if (!new_slave) {
1295 res = -ENOMEM;
1296 goto err_undo_flags;
1297 }
bb1d9123
AG		 1298 /*
1299 * Set the new_slave's queue_id to be zero. Queue ID mapping
1300 * is set via sysfs or module option if desired.
1301 */
1302 new_slave->queue_id = 0;
1303
b15ba0fb
JP		 1304 /* Save slave's original mtu and then set it to match the bond */
1305 new_slave->original_mtu = slave_dev->mtu;
1306 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1307 if (res) {
1308 pr_debug("Error %d calling dev_set_mtu\n", res);
1309 goto err_free;
1310 }
1311
217df670
JV		 1312 /*
1313 * Save slave's original ("permanent") mac address for modes
1314 * that need it, and for restoring it upon release, and then
1315 * set it to the master's address
1316 */
ada0f863 1317 ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr);
1da177e4 1318
00503b6f 1319 if (!bond->params.fail_over_mac ||
1320 bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2ab82852
MS		 1321 /*
1322 * Set slave to master's mac address. The application already
1323 * set the master's mac address to that of the first slave
1324 */
1325 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1326 addr.sa_family = slave_dev->type;
1327 res = dev_set_mac_address(slave_dev, &addr);
1328 if (res) {
5a03cdb7 1329 pr_debug("Error %d calling set_mac_address\n", res);
b15ba0fb 1330 goto err_restore_mtu;
2ab82852 1331 }
217df670 1332 }
1da177e4 1333
217df670
JV		 1334 /* open the slave since the application closed it */
1335 res = dev_open(slave_dev);
1336 if (res) {
3d632c3f 1337 pr_debug("Opening slave %s failed\n", slave_dev->name);
1f718f0f 1338 goto err_restore_mac;
1da177e4
LT		 1339 }
1340
35d48903 1341 new_slave->bond = bond;
1da177e4 1342 new_slave->dev = slave_dev;
0b680e75 1343 slave_dev->priv_flags |= IFF_BONDING;
1da177e4 1344
58402054 1345 if (bond_is_lb(bond)) {
1da177e4
LT		 1346 /* bond_alb_init_slave() must be called before all other stages since
1347 * it might fail and we do not want to have to undo everything
1348 */
1349 res = bond_alb_init_slave(bond, new_slave);
3d632c3f 1350 if (res)
569f0c4d 1351 goto err_close;
1da177e4
LT		 1352 }
1353
303d1cbf
JV		 1354 /* If the mode USES_PRIMARY, then the following is handled by
1355 * bond_change_active_slave().
1da177e4
LT		 1356 */
1357 if (!USES_PRIMARY(bond->params.mode)) {
1358 /* set promiscuity level to new slave */
1359 if (bond_dev->flags & IFF_PROMISC) {
7e1a1ac1
WC		 1360 res = dev_set_promiscuity(slave_dev, 1);
1361 if (res)
1362 goto err_close;
1da177e4
LT		 1363 }
1364
1365 /* set allmulti level to new slave */
1366 if (bond_dev->flags & IFF_ALLMULTI) {
7e1a1ac1
WC		 1367 res = dev_set_allmulti(slave_dev, 1);
1368 if (res)
1369 goto err_close;
1da177e4
LT		 1370 }
1371
b9e40857 1372 netif_addr_lock_bh(bond_dev);
303d1cbf
JV		 1373
1374 dev_mc_sync_multiple(slave_dev, bond_dev);
1375 dev_uc_sync_multiple(slave_dev, bond_dev);
1376
b9e40857 1377 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1378 }
1379
1380 if (bond->params.mode == BOND_MODE_8023AD) {
1381 /* add lacpdu mc addr to mc list */
1382 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1383
22bedad3 1384 dev_mc_add(slave_dev, lacpdu_multicast);
1da177e4
LT		 1385 }
1386
b8e2fde4
WY		 1387 res = vlan_vids_add_by_dev(slave_dev, bond_dev);
1388 if (res) {
1ff412ad 1389 pr_err("%s: Error: Couldn't add bond vlan ids to %s\n",
1390 bond_dev->name, slave_dev->name);
1391 goto err_close;
1392 }
1da177e4 1393
c8c23903 1394 prev_slave = bond_last_slave(bond);
1da177e4
LT		 1395
1396 new_slave->delay = 0;
1397 new_slave->link_failure_count = 0;
1398
876254ae
VF		 1399 bond_update_speed_duplex(new_slave);
1400
49f17de7 1401 new_slave->last_rx = jiffies -
f31c7937 1402 (msecs_to_jiffies(bond->params.arp_interval) + 1);
8599b52e 1403 for (i = 0; i < BOND_MAX_ARP_TARGETS; i++)
49f17de7 1404 new_slave->target_last_arp_rx[i] = new_slave->last_rx;
f5b2b966 1405
1da177e4
LT		 1406 if (bond->params.miimon && !bond->params.use_carrier) {
1407 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1408
1409 if ((link_reporting == -1) && !bond->params.arp_interval) {
1410 /*
1411 * miimon is set but a bonded network driver
1412 * does not support ETHTOOL/MII and
1413 * arp_interval is not set. Note: if
1414 * use_carrier is enabled, we will never go
1415 * here (because netif_carrier is always
1416 * supported); thus, we don't need to change
1417 * the messages for netif_carrier.
1418 */
91565ebb
JP		 1419 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",
1420 bond_dev->name, slave_dev->name);
1da177e4
LT		 1421 } else if (link_reporting == -1) {
1422 /* unable get link status using mii/ethtool */
91565ebb
JP		 1423 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",
1424 bond_dev->name, slave_dev->name);
1da177e4
LT		 1425 }
1426 }
1427
1428 /* check for initial state */
f31c7937
MK		 1429 if (bond->params.miimon) {
1430 if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) {
1431 if (bond->params.updelay) {
1432 new_slave->link = BOND_LINK_BACK;
1433 new_slave->delay = bond->params.updelay;
1434 } else {
1435 new_slave->link = BOND_LINK_UP;
1436 }
1da177e4 1437 } else {
f31c7937 1438 new_slave->link = BOND_LINK_DOWN;
1da177e4 1439 }
f31c7937
MK		 1440 } else if (bond->params.arp_interval) {
1441 new_slave->link = (netif_carrier_ok(slave_dev) ?
1442 BOND_LINK_UP : BOND_LINK_DOWN);
1da177e4 1443 } else {
f31c7937 1444 new_slave->link = BOND_LINK_UP;
1da177e4
LT		 1445 }
1446
f31c7937 1447 if (new_slave->link != BOND_LINK_DOWN)
8e603460 1448 new_slave->last_link_up = jiffies;
f31c7937 1449 pr_debug("Initial state of slave_dev is BOND_LINK_%s\n",
90194264
JP		 1450 new_slave->link == BOND_LINK_DOWN ? "DOWN" :
1451 (new_slave->link == BOND_LINK_UP ? "UP" : "BACK"));
f31c7937 1452
1da177e4
LT		 1453 if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) {
1454 /* if there is a primary slave, remember it */
a549952a 1455 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1da177e4 1456 bond->primary_slave = new_slave;
a549952a
JP		 1457 bond->force_primary = true;
1458 }
1da177e4
LT		 1459 }
1460
1461 switch (bond->params.mode) {
1462 case BOND_MODE_ACTIVEBACKUP:
5e5b0665 1463 bond_set_slave_inactive_flags(new_slave,
1464 BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT		 1465 break;
1466 case BOND_MODE_8023AD:
1467 /* in 802.3ad mode, the internal mechanism
1468 * will activate the slaves in the selected
1469 * aggregator
1470 */
5e5b0665 1471 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1da177e4 1472 /* if this is the first slave */
23c147e0 1473 if (!prev_slave) {
1da177e4
LT		 1474 SLAVE_AD_INFO(new_slave).id = 1;
1475 /* Initialize AD with the number of times that the AD timer is called in 1 second
1476 * can be called only after the mac address of the bond is set
1477 */
56d00c67 1478 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL);
1da177e4
LT		 1479 } else {
1480 SLAVE_AD_INFO(new_slave).id =
dec1e90e 1481 SLAVE_AD_INFO(prev_slave).id + 1;
1da177e4
LT		 1482 }
1483
1484 bond_3ad_bind_slave(new_slave);
1485 break;
1486 case BOND_MODE_TLB:
1487 case BOND_MODE_ALB:
e30bc066 1488 bond_set_active_slave(new_slave);
5e5b0665 1489 bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW);
1da177e4
LT		 1490 break;
1491 default:
5a03cdb7 1492 pr_debug("This slave is always active in trunk mode\n");
1da177e4
LT		 1493
1494 /* always active in trunk mode */
e30bc066 1495 bond_set_active_slave(new_slave);
1da177e4
LT		 1496
1497 /* In trunking mode there is little meaning to curr_active_slave
1498 * anyway (it holds no special properties of the bond device),
1499 * so we can change it without calling change_active_interface()
1500 */
4e591b93 1501 if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP)
278b2083 1502 rcu_assign_pointer(bond->curr_active_slave, new_slave);
3d632c3f 1503
1da177e4
LT		 1504 break;
1505 } /* switch(bond_mode) */
1506
f6dc31a8 1507#ifdef CONFIG_NET_POLL_CONTROLLER
10eccb46 1508 slave_dev->npinfo = bond->dev->npinfo;
8a8efa22
AW		 1509 if (slave_dev->npinfo) {
1510 if (slave_enable_netpoll(new_slave)) {
90194264
JP		 1511 pr_info("Error, %s: master_dev is using netpoll, but new slave device does not support netpoll\n",
1512 bond_dev->name);
8a8efa22 1513 res = -EBUSY;
f7d9821a 1514 goto err_detach;
8a8efa22 1515 }
f6dc31a8
WC		 1516 }
1517#endif
8a8efa22 1518
35d48903
JP		 1519 res = netdev_rx_handler_register(slave_dev, bond_handle_frame,
1520 new_slave);
1521 if (res) {
1522 pr_debug("Error %d calling netdev_rx_handler_register\n", res);
5831d66e 1523 goto err_detach;
35d48903
JP		 1524 }
1525
1f718f0f
VF		 1526 res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave);
1527 if (res) {
1528 pr_debug("Error %d calling bond_master_upper_dev_link\n", res);
1529 goto err_unregister;
1530 }
1531
07699f9a 1532 res = bond_sysfs_slave_add(new_slave);
1533 if (res) {
1534 pr_debug("Error %d calling bond_sysfs_slave_add\n", res);
1535 goto err_upper_unlink;
1536 }
1537
5378c2e6
VF		 1538 bond->slave_cnt++;
1539 bond_compute_features(bond);
1540 bond_set_carrier(bond);
1541
1542 if (USES_PRIMARY(bond->params.mode)) {
f80889a5 1543 block_netpoll_tx();
5378c2e6
VF		 1544 write_lock_bh(&bond->curr_slave_lock);
1545 bond_select_active_slave(bond);
1546 write_unlock_bh(&bond->curr_slave_lock);
f80889a5 1547 unblock_netpoll_tx();
5378c2e6 1548 }
1f718f0f 1549
90194264 1550 pr_info("%s: Enslaving %s as %s interface with %s link\n",
a4aee5c8 1551 bond_dev->name, slave_dev->name,
90194264
JP		 1552 bond_is_active_slave(new_slave) ? "an active" : "a backup",
1553 new_slave->link != BOND_LINK_DOWN ? "an up" : "a down");
1da177e4
LT		 1554
1555 /* enslave is successful */
1556 return 0;
1557
1558/* Undo stages on error */
07699f9a 1559err_upper_unlink:
1560 bond_upper_dev_unlink(bond_dev, slave_dev);
1561
1f718f0f
VF		 1562err_unregister:
1563 netdev_rx_handler_unregister(slave_dev);
1564
f7d9821a 1565err_detach:
303d1cbf
JV		 1566 if (!USES_PRIMARY(bond->params.mode))
1567 bond_hw_addr_flush(bond_dev, slave_dev);
1568
1ff412ad 1569 vlan_vids_del_by_dev(slave_dev, bond_dev);
3c5913b5 1570 if (bond->primary_slave == new_slave)
1571 bond->primary_slave = NULL;
3c5913b5 1572 if (bond->curr_active_slave == new_slave) {
f80889a5 1573 block_netpoll_tx();
3c5913b5 1574 write_lock_bh(&bond->curr_slave_lock);
c8517035 1575 bond_change_active_slave(bond, NULL);
3c5913b5 1576 bond_select_active_slave(bond);
1577 write_unlock_bh(&bond->curr_slave_lock);
f80889a5 1578 unblock_netpoll_tx();
3c5913b5 1579 }
fc7a72ac 1580 slave_disable_netpoll(new_slave);
f7d9821a 1581
1da177e4 1582err_close:
b6a5a7b9 1583 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4
LT		 1584 dev_close(slave_dev);
1585
1586err_restore_mac:
00503b6f 1587 if (!bond->params.fail_over_mac ||
1588 bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
3915c1e8
JV		 1589 /* XXX TODO - fom follow mode needs to change master's
1590 * MAC if this slave's MAC is in use by the bond, or at
1591 * least print a warning.
1592 */
ada0f863 1593 ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr);
2ab82852
MS		 1594 addr.sa_family = slave_dev->type;
1595 dev_set_mac_address(slave_dev, &addr);
1596 }
1da177e4 1597
b15ba0fb
JP		 1598err_restore_mtu:
1599 dev_set_mtu(slave_dev, new_slave->original_mtu);
1600
1da177e4
LT		 1601err_free:
1602 kfree(new_slave);
1603
1604err_undo_flags:
b8fad459 1605 /* Enslave of first slave has failed and we need to fix master's mac */
0965a1f3 1606 if (!bond_has_slaves(bond) &&
844223ab 1607 ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr))
b8fad459 1608 eth_hw_addr_random(bond_dev);
3d632c3f 1609
1da177e4
LT		 1610 return res;
1611}
1612
1613/*
1614 * Try to release the slave device <slave> from the bond device <master>
1615 * It is legal to access curr_active_slave without a lock because all the function
0896341a 1616 * is write-locked. If "all" is true it means that the function is being called
1617 * while destroying a bond interface and all slaves are being released.
1da177e4
LT		 1618 *
1619 * The rules for slave state should be:
1620 * for Active/Backup:
1621 * Active stays on all backups go down
1622 * for Bonded connections:
1623 * The first up interface should be left on and all others downed.
1624 */
0896341a 1625static int __bond_release_one(struct net_device *bond_dev,
1626 struct net_device *slave_dev,
1627 bool all)
1da177e4 1628{
454d7c9b 1629 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1630 struct slave *slave, *oldcurrent;
1631 struct sockaddr addr;
5a0068de 1632 int old_flags = bond_dev->flags;
c8f44aff 1633 netdev_features_t old_features = bond_dev->features;
1da177e4
LT		 1634
1635 /* slave is not a slave or master is not master of this slave */
1636 if (!(slave_dev->flags & IFF_SLAVE) ||
471cb5a3 1637 !netdev_has_upper_dev(slave_dev, bond_dev)) {
90194264 1638 pr_err("%s: Error: cannot release %s\n",
1da177e4
LT		 1639 bond_dev->name, slave_dev->name);
1640 return -EINVAL;
1641 }
1642
e843fa50 1643 block_netpoll_tx();
1da177e4
LT		 1644
1645 slave = bond_get_slave_by_dev(bond, slave_dev);
1646 if (!slave) {
1647 /* not a slave of this bond */
a4aee5c8
JP		 1648 pr_info("%s: %s not enslaved\n",
1649 bond_dev->name, slave_dev->name);
e843fa50 1650 unblock_netpoll_tx();
1da177e4
LT		 1651 return -EINVAL;
1652 }
1653
07699f9a 1654 bond_sysfs_slave_del(slave);
1655
1f718f0f 1656 bond_upper_dev_unlink(bond_dev, slave_dev);
35d48903
JP		 1657 /* unregister rx_handler early so bond_handle_frame wouldn't be called
1658 * for this slave anymore.
1659 */
1660 netdev_rx_handler_unregister(slave_dev);
35d48903
JP		 1661 write_lock_bh(&bond->lock);
1662
1da177e4 1663 /* Inform AD package of unbinding of slave. */
be79bd04 1664 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 1665 bond_3ad_unbind_slave(slave);
be79bd04 1666
c8517035 1667 write_unlock_bh(&bond->lock);
1da177e4 1668
90194264 1669 pr_info("%s: Releasing %s interface %s\n",
a4aee5c8 1670 bond_dev->name,
e30bc066 1671 bond_is_active_slave(slave) ? "active" : "backup",
a4aee5c8 1672 slave_dev->name);
1da177e4
LT		 1673
1674 oldcurrent = bond->curr_active_slave;
1675
1676 bond->current_arp_slave = NULL;
1677
00503b6f 1678 if (!all && (!bond->params.fail_over_mac ||
1679 bond->params.mode != BOND_MODE_ACTIVEBACKUP)) {
844223ab 1680 if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) &&
0965a1f3 1681 bond_has_slaves(bond))
90194264
JP		 1682 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",
1683 bond_dev->name, slave_dev->name,
1684 slave->perm_hwaddr,
1685 bond_dev->name, slave_dev->name);
dec1e90e 1686 }
1687
3d632c3f 1688 if (bond->primary_slave == slave)
1da177e4 1689 bond->primary_slave = NULL;
1da177e4 1690
c8517035 1691 if (oldcurrent == slave) {
1692 write_lock_bh(&bond->curr_slave_lock);
1da177e4 1693 bond_change_active_slave(bond, NULL);
c8517035 1694 write_unlock_bh(&bond->curr_slave_lock);
1695 }
1da177e4 1696
58402054 1697 if (bond_is_lb(bond)) {
1da177e4
LT		 1698 /* Must be called only after the slave has been
1699 * detached from the list and the curr_active_slave
1700 * has been cleared (if our_slave == old_current),
1701 * but before a new active slave is selected.
1702 */
1703 bond_alb_deinit_slave(bond, slave);
1704 }
1705
0896341a 1706 if (all) {
36708b89 1707 RCU_INIT_POINTER(bond->curr_active_slave, NULL);
0896341a 1708 } else if (oldcurrent == slave) {
059fe7a5
JV		 1709 /*
1710 * Note that we hold RTNL over this sequence, so there
1711 * is no concern that another slave add/remove event
1712 * will interfere.
1713 */
059fe7a5
JV		 1714 write_lock_bh(&bond->curr_slave_lock);
1715
1da177e4
LT		 1716 bond_select_active_slave(bond);
1717
059fe7a5 1718 write_unlock_bh(&bond->curr_slave_lock);
059fe7a5
JV		 1719 }
1720
0965a1f3 1721 if (!bond_has_slaves(bond)) {
ff59c456 1722 bond_set_carrier(bond);
409cc1f8 1723 eth_hw_addr_random(bond_dev);
1da177e4 1724
a59d3d21 1725 if (vlan_uses_dev(bond_dev)) {
91565ebb
JP		 1726 pr_warn("%s: Warning: clearing HW address of %s while it still has VLANs\n",
1727 bond_dev->name, bond_dev->name);
1728 pr_warn("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs\n",
1729 bond_dev->name);
1da177e4 1730 }
1da177e4
LT		 1731 }
1732
e843fa50 1733 unblock_netpoll_tx();
278b2083 1734 synchronize_rcu();
ee6154e1 1735 bond->slave_cnt--;
1da177e4 1736
0965a1f3 1737 if (!bond_has_slaves(bond)) {
2af73d4b 1738 call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev);
80028ea1
VF		 1739 call_netdevice_notifiers(NETDEV_RELEASE, bond->dev);
1740 }
2af73d4b 1741
b2a103e6
MM		 1742 bond_compute_features(bond);
1743 if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1744 (old_features & NETIF_F_VLAN_CHALLENGED))
90194264 1745 pr_info("%s: last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n",
b2a103e6
MM		 1746 bond_dev->name, slave_dev->name, bond_dev->name);
1747
b76cdba9 1748 /* must do this from outside any spinlocks */
1ff412ad 1749 vlan_vids_del_by_dev(slave_dev, bond_dev);
1da177e4 1750
303d1cbf
JV		 1751 /* If the mode USES_PRIMARY, then this cases was handled above by
1752 * bond_change_active_slave(..., NULL)
1da177e4
LT		 1753 */
1754 if (!USES_PRIMARY(bond->params.mode)) {
5a0068de
NH		 1755 /* unset promiscuity level from slave
1756 * NOTE: The NETDEV_CHANGEADDR call above may change the value
1757 * of the IFF_PROMISC flag in the bond_dev, but we need the
1758 * value of that flag before that change, as that was the value
1759 * when this slave was attached, so we cache at the start of the
1760 * function and use it here. Same goes for ALLMULTI below
1761 */
1762 if (old_flags & IFF_PROMISC)
1da177e4 1763 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 1764
1765 /* unset allmulti level from slave */
5a0068de 1766 if (old_flags & IFF_ALLMULTI)
1da177e4 1767 dev_set_allmulti(slave_dev, -1);
1da177e4 1768
303d1cbf 1769 bond_hw_addr_flush(bond_dev, slave_dev);
1da177e4
LT		 1770 }
1771
8a8efa22 1772 slave_disable_netpoll(slave);
f6dc31a8 1773
1da177e4
LT		 1774 /* close slave before restoring its mac address */
1775 dev_close(slave_dev);
1776
00503b6f 1777 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE ||
1778 bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2ab82852 1779 /* restore original ("permanent") mac address */
ada0f863 1780 ether_addr_copy(addr.sa_data, slave->perm_hwaddr);
2ab82852
MS		 1781 addr.sa_family = slave_dev->type;
1782 dev_set_mac_address(slave_dev, &addr);
1783 }
1da177e4 1784
b15ba0fb
JP		 1785 dev_set_mtu(slave_dev, slave->original_mtu);
1786
2d7011ca 1787 slave_dev->priv_flags &= ~IFF_BONDING;
1da177e4
LT		 1788
1789 kfree(slave);
1790
1791 return 0; /* deletion OK */
1792}
1793
0896341a 1794/* A wrapper used because of ndo_del_link */
1795int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1796{
1797 return __bond_release_one(bond_dev, slave_dev, false);
1798}
1799
d90a162a 1800/*
dadaa10b 1801* First release a slave and then destroy the bond if no more slaves are left.
d90a162a
MS		 1802* Must be under rtnl_lock when this function is called.
1803*/
26d8ee75 1804static int bond_release_and_destroy(struct net_device *bond_dev,
1805 struct net_device *slave_dev)
d90a162a 1806{
454d7c9b 1807 struct bonding *bond = netdev_priv(bond_dev);
d90a162a
MS		 1808 int ret;
1809
1810 ret = bond_release(bond_dev, slave_dev);
0965a1f3 1811 if (ret == 0 && !bond_has_slaves(bond)) {
8a8efa22 1812 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
90194264 1813 pr_info("%s: Destroying bond %s\n",
a4aee5c8 1814 bond_dev->name, bond_dev->name);
9e71626c 1815 unregister_netdevice(bond_dev);
d90a162a
MS		 1816 }
1817 return ret;
1818}
1819
1da177e4
LT		 1820static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
1821{
454d7c9b 1822 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1823
1824 info->bond_mode = bond->params.mode;
1825 info->miimon = bond->params.miimon;
1826
1da177e4 1827 info->num_slaves = bond->slave_cnt;
1da177e4
LT		 1828
1829 return 0;
1830}
1831
1832static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
1833{
454d7c9b 1834 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 1835 struct list_head *iter;
dec1e90e 1836 int i = 0, res = -ENODEV;
1da177e4 1837 struct slave *slave;
1da177e4 1838
9caff1e7 1839 bond_for_each_slave(bond, slave, iter) {
dec1e90e 1840 if (i++ == (int)info->slave_id) {
689c96cc
ED		 1841 res = 0;
1842 strcpy(info->slave_name, slave->dev->name);
1843 info->link = slave->link;
e30bc066 1844 info->state = bond_slave_state(slave);
689c96cc 1845 info->link_failure_count = slave->link_failure_count;
1da177e4
LT		 1846 break;
1847 }
1848 }
1da177e4 1849
689c96cc 1850 return res;
1da177e4
LT		 1851}
1852
1853/*-------------------------------- Monitoring -------------------------------*/
1854
1da177e4 1855
f0c76d61
JV		 1856static int bond_miimon_inspect(struct bonding *bond)
1857{
dec1e90e 1858 int link_state, commit = 0;
9caff1e7 1859 struct list_head *iter;
f0c76d61 1860 struct slave *slave;
41f89100
JP		 1861 bool ignore_updelay;
1862
1863 ignore_updelay = !bond->curr_active_slave ? true : false;
1da177e4 1864
4cb4f97b 1865 bond_for_each_slave_rcu(bond, slave, iter) {
f0c76d61 1866 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 1867
f0c76d61 1868 link_state = bond_check_dev_link(bond, slave->dev, 0);
1da177e4
LT		 1869
1870 switch (slave->link) {
f0c76d61
JV		 1871 case BOND_LINK_UP:
1872 if (link_state)
1873 continue;
1da177e4 1874
f0c76d61
JV		 1875 slave->link = BOND_LINK_FAIL;
1876 slave->delay = bond->params.downdelay;
1877 if (slave->delay) {
90194264 1878 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms\n",
a4aee5c8
JP		 1879 bond->dev->name,
1880 (bond->params.mode ==
1881 BOND_MODE_ACTIVEBACKUP) ?
e30bc066 1882 (bond_is_active_slave(slave) ?
a4aee5c8
JP		 1883 "active " : "backup ") : "",
1884 slave->dev->name,
1885 bond->params.downdelay * bond->params.miimon);
1da177e4 1886 }
f0c76d61
JV		 1887 /*FALLTHRU*/
1888 case BOND_LINK_FAIL:
1889 if (link_state) {
1890 /*
1891 * recovered before downdelay expired
1892 */
1893 slave->link = BOND_LINK_UP;
8e603460 1894 slave->last_link_up = jiffies;
90194264 1895 pr_info("%s: link status up again after %d ms for interface %s\n",
a4aee5c8
JP		 1896 bond->dev->name,
1897 (bond->params.downdelay - slave->delay) *
1898 bond->params.miimon,
1899 slave->dev->name);
f0c76d61 1900 continue;
1da177e4 1901 }
f0c76d61
JV		 1902
1903 if (slave->delay <= 0) {
1904 slave->new_link = BOND_LINK_DOWN;
1905 commit++;
1906 continue;
1da177e4 1907 }
1da177e4 1908
f0c76d61
JV		 1909 slave->delay--;
1910 break;
1911
1912 case BOND_LINK_DOWN:
1913 if (!link_state)
1914 continue;
1915
1916 slave->link = BOND_LINK_BACK;
1917 slave->delay = bond->params.updelay;
1918
1919 if (slave->delay) {
90194264 1920 pr_info("%s: link status up for interface %s, enabling it in %d ms\n",
a4aee5c8
JP		 1921 bond->dev->name, slave->dev->name,
1922 ignore_updelay ? 0 :
1923 bond->params.updelay *
1924 bond->params.miimon);
f0c76d61
JV		 1925 }
1926 /*FALLTHRU*/
1927 case BOND_LINK_BACK:
1928 if (!link_state) {
1929 slave->link = BOND_LINK_DOWN;
90194264 1930 pr_info("%s: link status down again after %d ms for interface %s\n",
a4aee5c8
JP		 1931 bond->dev->name,
1932 (bond->params.updelay - slave->delay) *
1933 bond->params.miimon,
1934 slave->dev->name);
f0c76d61
JV		 1935
1936 continue;
1937 }
1938
41f89100
JP		 1939 if (ignore_updelay)
1940 slave->delay = 0;
1941
f0c76d61
JV		 1942 if (slave->delay <= 0) {
1943 slave->new_link = BOND_LINK_UP;
1944 commit++;
41f89100 1945 ignore_updelay = false;
f0c76d61 1946 continue;
1da177e4 1947 }
f0c76d61
JV		 1948
1949 slave->delay--;
1da177e4 1950 break;
f0c76d61
JV		 1951 }
1952 }
1da177e4 1953
f0c76d61
JV		 1954 return commit;
1955}
1da177e4 1956
f0c76d61
JV		 1957static void bond_miimon_commit(struct bonding *bond)
1958{
9caff1e7 1959 struct list_head *iter;
f0c76d61 1960 struct slave *slave;
f0c76d61 1961
9caff1e7 1962 bond_for_each_slave(bond, slave, iter) {
f0c76d61
JV		 1963 switch (slave->new_link) {
1964 case BOND_LINK_NOCHANGE:
1965 continue;
1da177e4 1966
f0c76d61
JV		 1967 case BOND_LINK_UP:
1968 slave->link = BOND_LINK_UP;
8e603460 1969 slave->last_link_up = jiffies;
f0c76d61
JV		 1970
1971 if (bond->params.mode == BOND_MODE_8023AD) {
1972 /* prevent it from being the active one */
e30bc066 1973 bond_set_backup_slave(slave);
f0c76d61
JV		 1974 } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
1975 /* make it immediately active */
e30bc066 1976 bond_set_active_slave(slave);
f0c76d61
JV		 1977 } else if (slave != bond->primary_slave) {
1978 /* prevent it from being the active one */
e30bc066 1979 bond_set_backup_slave(slave);
1da177e4 1980 }
1da177e4 1981
90194264 1982 pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex\n",
700c2a77 1983 bond->dev->name, slave->dev->name,
db4e9b2b
NA		 1984 slave->speed == SPEED_UNKNOWN ? 0 : slave->speed,
1985 slave->duplex ? "full" : "half");
1da177e4 1986
f0c76d61
JV		 1987 /* notify ad that the link status has changed */
1988 if (bond->params.mode == BOND_MODE_8023AD)
1989 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
059fe7a5 1990
58402054 1991 if (bond_is_lb(bond))
f0c76d61
JV		 1992 bond_alb_handle_link_change(bond, slave,
1993 BOND_LINK_UP);
1da177e4 1994
f0c76d61
JV		 1995 if (!bond->curr_active_slave ||
1996 (slave == bond->primary_slave))
1997 goto do_failover;
1da177e4 1998
f0c76d61 1999 continue;
059fe7a5 2000
f0c76d61 2001 case BOND_LINK_DOWN:
fba4acda
JV		 2002 if (slave->link_failure_count < UINT_MAX)
2003 slave->link_failure_count++;
2004
f0c76d61 2005 slave->link = BOND_LINK_DOWN;
1da177e4 2006
f0c76d61
JV		 2007 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP ||
2008 bond->params.mode == BOND_MODE_8023AD)
5e5b0665 2009 bond_set_slave_inactive_flags(slave,
2010 BOND_SLAVE_NOTIFY_NOW);
f0c76d61 2011
a4aee5c8
JP		 2012 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2013 bond->dev->name, slave->dev->name);
f0c76d61
JV		 2014
2015 if (bond->params.mode == BOND_MODE_8023AD)
2016 bond_3ad_handle_link_change(slave,
2017 BOND_LINK_DOWN);
2018
ae63e808 2019 if (bond_is_lb(bond))
f0c76d61
JV		 2020 bond_alb_handle_link_change(bond, slave,
2021 BOND_LINK_DOWN);
2022
2023 if (slave == bond->curr_active_slave)
2024 goto do_failover;
2025
2026 continue;
2027
2028 default:
a4aee5c8 2029 pr_err("%s: invalid new link %d on slave %s\n",
f0c76d61
JV		 2030 bond->dev->name, slave->new_link,
2031 slave->dev->name);
2032 slave->new_link = BOND_LINK_NOCHANGE;
2033
2034 continue;
2035 }
2036
2037do_failover:
2038 ASSERT_RTNL();
e843fa50 2039 block_netpoll_tx();
f0c76d61
JV		 2040 write_lock_bh(&bond->curr_slave_lock);
2041 bond_select_active_slave(bond);
2042 write_unlock_bh(&bond->curr_slave_lock);
e843fa50 2043 unblock_netpoll_tx();
f0c76d61
JV		 2044 }
2045
2046 bond_set_carrier(bond);
1da177e4
LT		 2047}
2048
0b0eef66
JV		 2049/*
2050 * bond_mii_monitor
2051 *
2052 * Really a wrapper that splits the mii monitor into two phases: an
f0c76d61
JV		 2053 * inspection, then (if inspection indicates something needs to be done)
2054 * an acquisition of appropriate locks followed by a commit phase to
2055 * implement whatever link state changes are indicated.
0b0eef66 2056 */
6da67d26 2057static void bond_mii_monitor(struct work_struct *work)
0b0eef66
JV		 2058{
2059 struct bonding *bond = container_of(work, struct bonding,
2060 mii_work.work);
ad246c99 2061 bool should_notify_peers = false;
1f2cd845 2062 unsigned long delay;
0b0eef66 2063
1f2cd845
DM		 2064 delay = msecs_to_jiffies(bond->params.miimon);
2065
2066 if (!bond_has_slaves(bond))
f0c76d61 2067 goto re_arm;
b59f9f74 2068
4cb4f97b 2069 rcu_read_lock();
2070
ad246c99
BH		 2071 should_notify_peers = bond_should_notify_peers(bond);
2072
1f2cd845 2073 if (bond_miimon_inspect(bond)) {
4cb4f97b 2074 rcu_read_unlock();
f0c76d61 2075
1f2cd845
DM		 2076 /* Race avoidance with bond_close cancel of workqueue */
2077 if (!rtnl_trylock()) {
1f2cd845
DM		 2078 delay = 1;
2079 should_notify_peers = false;
2080 goto re_arm;
2081 }
6b6c5261 2082
1f2cd845
DM		 2083 bond_miimon_commit(bond);
2084
1f2cd845 2085 rtnl_unlock(); /* might sleep, hold no other locks */
4cb4f97b 2086 } else
2087 rcu_read_unlock();
0b0eef66 2088
f0c76d61 2089re_arm:
e6d265e8 2090 if (bond->params.miimon)
1f2cd845
DM		 2091 queue_delayed_work(bond->wq, &bond->mii_work, delay);
2092
1f2cd845
DM		 2093 if (should_notify_peers) {
2094 if (!rtnl_trylock())
2095 return;
2096 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev);
2097 rtnl_unlock();
2098 }
0b0eef66 2099}
c3ade5ca 2100
50223ce4 2101static bool bond_has_this_ip(struct bonding *bond, __be32 ip)
f5b2b966 2102{
50223ce4
VF		 2103 struct net_device *upper;
2104 struct list_head *iter;
2105 bool ret = false;
f5b2b966 2106
eaddcd76 2107 if (ip == bond_confirm_addr(bond->dev, 0, ip))
50223ce4 2108 return true;
f5b2b966 2109
50223ce4 2110 rcu_read_lock();
2f268f12 2111 netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) {
50223ce4
VF		 2112 if (ip == bond_confirm_addr(upper, 0, ip)) {
2113 ret = true;
2114 break;
2115 }
f5b2b966 2116 }
50223ce4 2117 rcu_read_unlock();
f5b2b966 2118
50223ce4 2119 return ret;
f5b2b966
JV		 2120}
2121
c3ade5ca
JV		 2122/*
2123 * We go to the (large) trouble of VLAN tagging ARP frames because
2124 * switches in VLAN mode (especially if ports are configured as
2125 * "native" to a VLAN) might not pass non-tagged frames.
2126 */
d3bb52b0 2127static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id)
c3ade5ca
JV		 2128{
2129 struct sk_buff *skb;
2130
90194264
JP		 2131 pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n",
2132 arp_op, slave_dev->name, &dest_ip, &src_ip, vlan_id);
3d632c3f 2133
c3ade5ca
JV		 2134 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2135 NULL, slave_dev->dev_addr, NULL);
2136
2137 if (!skb) {
a4aee5c8 2138 pr_err("ARP packet allocation failed\n");
c3ade5ca
JV		 2139 return;
2140 }
2141 if (vlan_id) {
1fd9b1fc 2142 skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id);
c3ade5ca 2143 if (!skb) {
a4aee5c8 2144 pr_err("failed to insert VLAN tag\n");
c3ade5ca
JV		 2145 return;
2146 }
2147 }
2148 arp_xmit(skb);
2149}
2150
2151
1da177e4
LT		 2152static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2153{
27bc11e6
VF		 2154 struct net_device *upper, *vlan_upper;
2155 struct list_head *iter, *vlan_iter;
c3ade5ca 2156 struct rtable *rt;
27bc11e6
VF		 2157 __be32 *targets = bond->params.arp_targets, addr;
2158 int i, vlan_id;
1da177e4 2159
27bc11e6 2160 for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) {
acca2674 2161 pr_debug("basa: target %pI4\n", &targets[i]);
c3ade5ca 2162
27bc11e6 2163 /* Find out through which dev should the packet go */
78fbfd8a
DM		 2164 rt = ip_route_output(dev_net(bond->dev), targets[i], 0,
2165 RTO_ONLINK, 0);
b23dd4fe 2166 if (IS_ERR(rt)) {
28572760
VF		 2167 /* there's no route to target - try to send arp
2168 * probe to generate any traffic (arp_validate=0)
2169 */
2170 if (bond->params.arp_validate && net_ratelimit())
2171 pr_warn("%s: no route to arp_ip_target %pI4 and arp_validate is set\n",
2172 bond->dev->name, &targets[i]);
2173 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], 0, 0);
c3ade5ca
JV		 2174 continue;
2175 }
2176
c3ade5ca 2177 vlan_id = 0;
27bc11e6
VF		 2178
2179 /* bond device itself */
2180 if (rt->dst.dev == bond->dev)
2181 goto found;
2182
2183 rcu_read_lock();
2184 /* first we search only for vlan devices. for every vlan
2185 * found we verify its upper dev list, searching for the
2186 * rt->dst.dev. If found we save the tag of the vlan and
2187 * proceed to send the packet.
2188 *
2189 * TODO: QinQ?
2190 */
2f268f12
VF		 2191 netdev_for_each_all_upper_dev_rcu(bond->dev, vlan_upper,
2192 vlan_iter) {
27bc11e6
VF		 2193 if (!is_vlan_dev(vlan_upper))
2194 continue;
2f268f12
VF		 2195 netdev_for_each_all_upper_dev_rcu(vlan_upper, upper,
2196 iter) {
27bc11e6
VF		 2197 if (upper == rt->dst.dev) {
2198 vlan_id = vlan_dev_vlan_id(vlan_upper);
2199 rcu_read_unlock();
2200 goto found;
2201 }
c3ade5ca
JV		 2202 }
2203 }
2204
27bc11e6
VF		 2205 /* if the device we're looking for is not on top of any of
2206 * our upper vlans, then just search for any dev that
2207 * matches, and in case it's a vlan - save the id
2208 */
2f268f12 2209 netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) {
27bc11e6
VF		 2210 if (upper == rt->dst.dev) {
2211 /* if it's a vlan - get its VID */
2212 if (is_vlan_dev(upper))
2213 vlan_id = vlan_dev_vlan_id(upper);
2214
2215 rcu_read_unlock();
2216 goto found;
2217 }
c3ade5ca 2218 }
27bc11e6 2219 rcu_read_unlock();
c3ade5ca 2220
27bc11e6 2221 /* Not our device - skip */
3e32582f
VF		 2222 pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2223 bond->dev->name, &targets[i],
2224 rt->dst.dev ? rt->dst.dev->name : "NULL");
2225
ed4b9f80 2226 ip_rt_put(rt);
27bc11e6
VF		 2227 continue;
2228
2229found:
2230 addr = bond_confirm_addr(rt->dst.dev, targets[i], 0);
2231 ip_rt_put(rt);
2232 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2233 addr, vlan_id);
c3ade5ca
JV		 2234 }
2235}
2236
d3bb52b0 2237static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
f5b2b966 2238{
8599b52e
VF		 2239 int i;
2240
87a7b84b
VF		 2241 if (!sip || !bond_has_this_ip(bond, tip)) {
2242 pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip);
2243 return;
2244 }
f5b2b966 2245
8599b52e
VF		 2246 i = bond_get_targets_ip(bond->params.arp_targets, sip);
2247 if (i == -1) {
87a7b84b
VF		 2248 pr_debug("bva: sip %pI4 not found in targets\n", &sip);
2249 return;
f5b2b966 2250 }
49f17de7 2251 slave->last_rx = jiffies;
8599b52e 2252 slave->target_last_arp_rx[i] = jiffies;
f5b2b966
JV		 2253}
2254
5bb9e0b5 2255int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond,
2256 struct slave *slave)
f5b2b966 2257{
de063b70 2258 struct arphdr *arp = (struct arphdr *)skb->data;
010d3c39 2259 struct slave *curr_active_slave;
f5b2b966 2260 unsigned char *arp_ptr;
d3bb52b0 2261 __be32 sip, tip;
f2cb691a 2262 int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP);
f5b2b966 2263
f2cb691a
VF		 2264 if (!slave_do_arp_validate(bond, slave)) {
2265 if ((slave_do_arp_validate_only(bond, slave) && is_arp) ||
2266 !slave_do_arp_validate_only(bond, slave))
49f17de7 2267 slave->last_rx = jiffies;
13a8e0c8 2268 return RX_HANDLER_ANOTHER;
f2cb691a
VF		 2269 } else if (!is_arp) {
2270 return RX_HANDLER_ANOTHER;
2271 }
2c146102 2272
de063b70 2273 alen = arp_hdr_len(bond->dev);
f5b2b966 2274
3aba891d
JP		 2275 pr_debug("bond_arp_rcv: bond %s skb->dev %s\n",
2276 bond->dev->name, skb->dev->name);
f5b2b966 2277
de063b70
ED		 2278 if (alen > skb_headlen(skb)) {
2279 arp = kmalloc(alen, GFP_ATOMIC);
2280 if (!arp)
2281 goto out_unlock;
2282 if (skb_copy_bits(skb, 0, arp, alen) < 0)
2283 goto out_unlock;
2284 }
f5b2b966 2285
3aba891d 2286 if (arp->ar_hln != bond->dev->addr_len ||
f5b2b966
JV		 2287 skb->pkt_type == PACKET_OTHERHOST ||
2288 skb->pkt_type == PACKET_LOOPBACK ||
2289 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2290 arp->ar_pro != htons(ETH_P_IP) ||
2291 arp->ar_pln != 4)
2292 goto out_unlock;
2293
2294 arp_ptr = (unsigned char *)(arp + 1);
3aba891d 2295 arp_ptr += bond->dev->addr_len;
f5b2b966 2296 memcpy(&sip, arp_ptr, 4);
3aba891d 2297 arp_ptr += 4 + bond->dev->addr_len;
f5b2b966
JV		 2298 memcpy(&tip, arp_ptr, 4);
2299
5a03cdb7 2300 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
e30bc066 2301 bond->dev->name, slave->dev->name, bond_slave_state(slave),
a4aee5c8
JP		 2302 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2303 &sip, &tip);
f5b2b966 2304
010d3c39
VF		 2305 curr_active_slave = rcu_dereference(bond->curr_active_slave);
2306
f5b2b966
JV		 2307 /*
2308 * Backup slaves won't see the ARP reply, but do come through
2309 * here for each ARP probe (so we swap the sip/tip to validate
2310 * the probe). In a "redundant switch, common router" type of
2311 * configuration, the ARP probe will (hopefully) travel from
2312 * the active, through one switch, the router, then the other
2313 * switch before reaching the backup.
aeea64ac
VF		 2314 *
2315 * We 'trust' the arp requests if there is an active slave and
2316 * it received valid arp reply(s) after it became active. This
2317 * is done to avoid endless looping when we can't reach the
2318 * arp_ip_target and fool ourselves with our own arp requests.
f5b2b966 2319 */
010d3c39 2320
e30bc066 2321 if (bond_is_active_slave(slave))
f5b2b966 2322 bond_validate_arp(bond, slave, sip, tip);
010d3c39
VF		 2323 else if (curr_active_slave &&
2324 time_after(slave_last_rx(bond, curr_active_slave),
2325 curr_active_slave->last_link_up))
f5b2b966
JV		 2326 bond_validate_arp(bond, slave, tip, sip);
2327
2328out_unlock:
de063b70
ED		 2329 if (arp != (struct arphdr *)skb->data)
2330 kfree(arp);
13a8e0c8 2331 return RX_HANDLER_ANOTHER;
f5b2b966
JV		 2332}
2333
e7f63f1d
VF		 2334/* function to verify if we're in the arp_interval timeslice, returns true if
2335 * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval +
2336 * arp_interval/2) . the arp_interval/2 is needed for really fast networks.
2337 */
2338static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act,
2339 int mod)
2340{
2341 int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2342
2343 return time_in_range(jiffies,
2344 last_act - delta_in_ticks,
2345 last_act + mod * delta_in_ticks + delta_in_ticks/2);
2346}
2347
1da177e4
LT		 2348/*
2349 * this function is called regularly to monitor each slave's link
2350 * ensuring that traffic is being sent and received when arp monitoring
2351 * is used in load-balancing mode. if the adapter has been dormant, then an
2352 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2353 * arp monitoring in active backup mode.
2354 */
6da67d26 2355static void bond_loadbalance_arp_mon(struct work_struct *work)
1da177e4 2356{
1b76b316
JV		 2357 struct bonding *bond = container_of(work, struct bonding,
2358 arp_work.work);
1da177e4 2359 struct slave *slave, *oldcurrent;
9caff1e7 2360 struct list_head *iter;
6fde8f03 2361 int do_failover = 0, slave_state_changed = 0;
1da177e4 2362
1f2cd845 2363 if (!bond_has_slaves(bond))
1da177e4 2364 goto re_arm;
1da177e4 2365
2e52f4fe 2366 rcu_read_lock();
2367
2368 oldcurrent = ACCESS_ONCE(bond->curr_active_slave);
1da177e4
LT		 2369 /* see if any of the previous devices are up now (i.e. they have
2370 * xmt and rcv traffic). the curr_active_slave does not come into
8e603460
VF		 2371 * the picture unless it is null. also, slave->last_link_up is not
2372 * needed here because we send an arp on each slave and give a slave
2373 * as long as it needs to get the tx/rx within the delta.
1da177e4
LT		 2374 * TODO: what about up/down delay in arp mode? it wasn't here before
2375 * so it can wait
2376 */
2e52f4fe 2377 bond_for_each_slave_rcu(bond, slave, iter) {
cb32f2a0
JB		 2378 unsigned long trans_start = dev_trans_start(slave->dev);
2379
1da177e4 2380 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2381 if (bond_time_in_interval(bond, trans_start, 1) &&
49f17de7 2382 bond_time_in_interval(bond, slave->last_rx, 1)) {
1da177e4
LT		 2383
2384 slave->link = BOND_LINK_UP;
6fde8f03 2385 slave_state_changed = 1;
1da177e4
LT		 2386
2387 /* primary_slave has no meaning in round-robin
2388 * mode. the window of a slave being up and
2389 * curr_active_slave being null after enslaving
2390 * is closed.
2391 */
2392 if (!oldcurrent) {
90194264 2393 pr_info("%s: link status definitely up for interface %s\n",
a4aee5c8
JP		 2394 bond->dev->name,
2395 slave->dev->name);
1da177e4
LT		 2396 do_failover = 1;
2397 } else {
a4aee5c8
JP		 2398 pr_info("%s: interface %s is now up\n",
2399 bond->dev->name,
2400 slave->dev->name);
1da177e4
LT		 2401 }
2402 }
2403 } else {
2404 /* slave->link == BOND_LINK_UP */
2405
2406 /* not all switches will respond to an arp request
2407 * when the source ip is 0, so don't take the link down
2408 * if we don't know our ip yet
2409 */
e7f63f1d 2410 if (!bond_time_in_interval(bond, trans_start, 2) ||
49f17de7 2411 !bond_time_in_interval(bond, slave->last_rx, 2)) {
1da177e4
LT		 2412
2413 slave->link = BOND_LINK_DOWN;
6fde8f03 2414 slave_state_changed = 1;
1da177e4 2415
3d632c3f 2416 if (slave->link_failure_count < UINT_MAX)
1da177e4 2417 slave->link_failure_count++;
1da177e4 2418
90194264
JP		 2419 pr_info("%s: interface %s is now down\n",
2420 bond->dev->name, slave->dev->name);
1da177e4 2421
3d632c3f 2422 if (slave == oldcurrent)
1da177e4 2423 do_failover = 1;
1da177e4
LT		 2424 }
2425 }
2426
2427 /* note: if switch is in round-robin mode, all links
2428 * must tx arp to ensure all links rx an arp - otherwise
2429 * links may oscillate or not come up at all; if switch is
2430 * in something like xor mode, there is nothing we can
2431 * do - all replies will be rx'ed on same link causing slaves
2432 * to be unstable during low/no traffic periods
2433 */
3d632c3f 2434 if (IS_UP(slave->dev))
1da177e4 2435 bond_arp_send_all(bond, slave);
1da177e4
LT		 2436 }
2437
2e52f4fe 2438 rcu_read_unlock();
2439
6fde8f03 2440 if (do_failover || slave_state_changed) {
2e52f4fe 2441 if (!rtnl_trylock())
2442 goto re_arm;
1da177e4 2443
6fde8f03
DT		 2444 if (slave_state_changed) {
2445 bond_slave_state_change(bond);
2446 } else if (do_failover) {
2447 /* the bond_select_active_slave must hold RTNL
2448 * and curr_slave_lock for write.
2449 */
2450 block_netpoll_tx();
2451 write_lock_bh(&bond->curr_slave_lock);
1da177e4 2452
6fde8f03
DT		 2453 bond_select_active_slave(bond);
2454
2455 write_unlock_bh(&bond->curr_slave_lock);
2456 unblock_netpoll_tx();
2457 }
2e52f4fe 2458 rtnl_unlock();
1da177e4
LT		 2459 }
2460
2461re_arm:
e6d265e8 2462 if (bond->params.arp_interval)
e7f63f1d
VF		 2463 queue_delayed_work(bond->wq, &bond->arp_work,
2464 msecs_to_jiffies(bond->params.arp_interval));
1da177e4
LT		 2465}
2466
2467/*
b2220cad
JV		 2468 * Called to inspect slaves for active-backup mode ARP monitor link state
2469 * changes. Sets new_link in slaves to specify what action should take
2470 * place for the slave. Returns 0 if no changes are found, >0 if changes
2471 * to link states must be committed.
2472 *
eb9fa4b0 2473 * Called with rcu_read_lock hold.
1da177e4 2474 */
e7f63f1d 2475static int bond_ab_arp_inspect(struct bonding *bond)
1da177e4 2476{
def4460c 2477 unsigned long trans_start, last_rx;
9caff1e7 2478 struct list_head *iter;
dec1e90e 2479 struct slave *slave;
dec1e90e 2480 int commit = 0;
da210f55 2481
eb9fa4b0 2482 bond_for_each_slave_rcu(bond, slave, iter) {
b2220cad 2483 slave->new_link = BOND_LINK_NOCHANGE;
def4460c 2484 last_rx = slave_last_rx(bond, slave);
1da177e4 2485
b2220cad 2486 if (slave->link != BOND_LINK_UP) {
e7f63f1d 2487 if (bond_time_in_interval(bond, last_rx, 1)) {
b2220cad
JV		 2488 slave->new_link = BOND_LINK_UP;
2489 commit++;
2490 }
b2220cad
JV		 2491 continue;
2492 }
1da177e4 2493
b2220cad
JV		 2494 /*
2495 * Give slaves 2*delta after being enslaved or made
2496 * active. This avoids bouncing, as the last receive
2497 * times need a full ARP monitor cycle to be updated.
2498 */
8e603460 2499 if (bond_time_in_interval(bond, slave->last_link_up, 2))
b2220cad
JV		 2500 continue;
2501
2502 /*
2503 * Backup slave is down if:
2504 * - No current_arp_slave AND
2505 * - more than 3*delta since last receive AND
2506 * - the bond has an IP address
2507 *
2508 * Note: a non-null current_arp_slave indicates
2509 * the curr_active_slave went down and we are
2510 * searching for a new one; under this condition
2511 * we only take the curr_active_slave down - this
2512 * gives each slave a chance to tx/rx traffic
2513 * before being taken out
2514 */
e30bc066 2515 if (!bond_is_active_slave(slave) &&
b2220cad 2516 !bond->current_arp_slave &&
e7f63f1d 2517 !bond_time_in_interval(bond, last_rx, 3)) {
b2220cad
JV		 2518 slave->new_link = BOND_LINK_DOWN;
2519 commit++;
2520 }
2521
2522 /*
2523 * Active slave is down if:
2524 * - more than 2*delta since transmitting OR
2525 * - (more than 2*delta since receive AND
2526 * the bond has an IP address)
2527 */
cb32f2a0 2528 trans_start = dev_trans_start(slave->dev);
e30bc066 2529 if (bond_is_active_slave(slave) &&
e7f63f1d
VF		 2530 (!bond_time_in_interval(bond, trans_start, 2) ||
2531 !bond_time_in_interval(bond, last_rx, 2))) {
b2220cad
JV		 2532 slave->new_link = BOND_LINK_DOWN;
2533 commit++;
2534 }
1da177e4
LT		 2535 }
2536
b2220cad
JV		 2537 return commit;
2538}
1da177e4 2539
b2220cad
JV		 2540/*
2541 * Called to commit link state changes noted by inspection step of
2542 * active-backup mode ARP monitor.
2543 *
eb9fa4b0 2544 * Called with RTNL hold.
b2220cad 2545 */
e7f63f1d 2546static void bond_ab_arp_commit(struct bonding *bond)
b2220cad 2547{
cb32f2a0 2548 unsigned long trans_start;
9caff1e7 2549 struct list_head *iter;
dec1e90e 2550 struct slave *slave;
1da177e4 2551
9caff1e7 2552 bond_for_each_slave(bond, slave, iter) {
b2220cad
JV		 2553 switch (slave->new_link) {
2554 case BOND_LINK_NOCHANGE:
2555 continue;
ff59c456 2556
b2220cad 2557 case BOND_LINK_UP:
cb32f2a0 2558 trans_start = dev_trans_start(slave->dev);
e7f63f1d
VF		 2559 if (bond->curr_active_slave != slave ||
2560 (!bond->curr_active_slave &&
2561 bond_time_in_interval(bond, trans_start, 1))) {
b2220cad 2562 slave->link = BOND_LINK_UP;
5a430974
VF		 2563 if (bond->current_arp_slave) {
2564 bond_set_slave_inactive_flags(
5e5b0665 2565 bond->current_arp_slave,
2566 BOND_SLAVE_NOTIFY_NOW);
5a430974
VF		 2567 bond->current_arp_slave = NULL;
2568 }
b2220cad 2569
90194264 2570 pr_info("%s: link status definitely up for interface %s\n",
b9f60253 2571 bond->dev->name, slave->dev->name);
b2220cad 2572
b9f60253
JP		 2573 if (!bond->curr_active_slave ||
2574 (slave == bond->primary_slave))
2575 goto do_failover;
1da177e4 2576
b9f60253 2577 }
1da177e4 2578
b9f60253 2579 continue;
1da177e4 2580
b2220cad
JV		 2581 case BOND_LINK_DOWN:
2582 if (slave->link_failure_count < UINT_MAX)
2583 slave->link_failure_count++;
2584
2585 slave->link = BOND_LINK_DOWN;
5e5b0665 2586 bond_set_slave_inactive_flags(slave,
2587 BOND_SLAVE_NOTIFY_NOW);
b2220cad 2588
a4aee5c8 2589 pr_info("%s: link status definitely down for interface %s, disabling it\n",
b9f60253 2590 bond->dev->name, slave->dev->name);
b2220cad 2591
b9f60253 2592 if (slave == bond->curr_active_slave) {
b2220cad 2593 bond->current_arp_slave = NULL;
b9f60253 2594 goto do_failover;
1da177e4 2595 }
b9f60253
JP		 2596
2597 continue;
b2220cad
JV		 2598
2599 default:
a4aee5c8 2600 pr_err("%s: impossible: new_link %d on slave %s\n",
b2220cad
JV		 2601 bond->dev->name, slave->new_link,
2602 slave->dev->name);
b9f60253 2603 continue;
1da177e4 2604 }
1da177e4 2605
b9f60253
JP		 2606do_failover:
2607 ASSERT_RTNL();
e843fa50 2608 block_netpoll_tx();
b2220cad 2609 write_lock_bh(&bond->curr_slave_lock);
b9f60253 2610 bond_select_active_slave(bond);
b2220cad 2611 write_unlock_bh(&bond->curr_slave_lock);
e843fa50 2612 unblock_netpoll_tx();
b2220cad 2613 }
1da177e4 2614
b2220cad
JV		 2615 bond_set_carrier(bond);
2616}
1da177e4 2617
b2220cad
JV		 2618/*
2619 * Send ARP probes for active-backup mode ARP monitor.
b0929915 2620 *
2621 * Called with rcu_read_lock hold.
b2220cad 2622 */
f2ebd477 2623static bool bond_ab_arp_probe(struct bonding *bond)
b2220cad 2624{
eb9fa4b0 2625 struct slave *slave, *before = NULL, *new_slave = NULL,
b0929915 2626 *curr_arp_slave = rcu_dereference(bond->current_arp_slave),
2627 *curr_active_slave = rcu_dereference(bond->curr_active_slave);
4087df87
VF		 2628 struct list_head *iter;
2629 bool found = false;
b0929915 2630 bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER;
f2ebd477 2631
98b90f26 2632 if (curr_arp_slave && curr_active_slave)
a4aee5c8 2633 pr_info("PROBE: c_arp %s && cas %s BAD\n",
eb9fa4b0 2634 curr_arp_slave->dev->name,
98b90f26 2635 curr_active_slave->dev->name);
1da177e4 2636
98b90f26
VF		 2637 if (curr_active_slave) {
2638 bond_arp_send_all(bond, curr_active_slave);
b0929915 2639 return should_notify_rtnl;
b2220cad 2640 }
1da177e4 2641
b2220cad
JV		 2642 /* if we don't have a curr_active_slave, search for the next available
2643 * backup slave from the current_arp_slave and make it the candidate
2644 * for becoming the curr_active_slave
2645 */
1da177e4 2646
eb9fa4b0 2647 if (!curr_arp_slave) {
b0929915 2648 curr_arp_slave = bond_first_slave_rcu(bond);
2649 if (!curr_arp_slave)
2650 return should_notify_rtnl;
b2220cad 2651 }
1da177e4 2652
b0929915 2653 bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER);
059fe7a5 2654
b0929915 2655 bond_for_each_slave_rcu(bond, slave, iter) {
4087df87
VF		 2656 if (!found && !before && IS_UP(slave->dev))
2657 before = slave;
1da177e4 2658
4087df87
VF		 2659 if (found && !new_slave && IS_UP(slave->dev))
2660 new_slave = slave;
b2220cad
JV		 2661 /* if the link state is up at this point, we
2662 * mark it down - this can happen if we have
2663 * simultaneous link failures and
2664 * reselect_active_interface doesn't make this
2665 * one the current slave so it is still marked
2666 * up when it is actually down
1da177e4 2667 */
4087df87 2668 if (!IS_UP(slave->dev) && slave->link == BOND_LINK_UP) {
b2220cad
JV		 2669 slave->link = BOND_LINK_DOWN;
2670 if (slave->link_failure_count < UINT_MAX)
2671 slave->link_failure_count++;
1da177e4 2672
5e5b0665 2673 bond_set_slave_inactive_flags(slave,
b0929915 2674 BOND_SLAVE_NOTIFY_LATER);
b2220cad 2675
90194264 2676 pr_info("%s: backup interface %s is now down\n",
a4aee5c8 2677 bond->dev->name, slave->dev->name);
1da177e4 2678 }
eb9fa4b0 2679 if (slave == curr_arp_slave)
4087df87 2680 found = true;
b2220cad 2681 }
4087df87
VF		 2682
2683 if (!new_slave && before)
2684 new_slave = before;
2685
b0929915 2686 if (!new_slave)
2687 goto check_state;
4087df87
VF		 2688
2689 new_slave->link = BOND_LINK_BACK;
b0929915 2690 bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER);
4087df87 2691 bond_arp_send_all(bond, new_slave);
8e603460 2692 new_slave->last_link_up = jiffies;
eb9fa4b0 2693 rcu_assign_pointer(bond->current_arp_slave, new_slave);
f2ebd477 2694
b0929915 2695check_state:
2696 bond_for_each_slave_rcu(bond, slave, iter) {
2697 if (slave->should_notify) {
2698 should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW;
2699 break;
2700 }
2701 }
2702 return should_notify_rtnl;
b2220cad 2703}
1da177e4 2704
6da67d26 2705static void bond_activebackup_arp_mon(struct work_struct *work)
b2220cad
JV		 2706{
2707 struct bonding *bond = container_of(work, struct bonding,
2708 arp_work.work);
b0929915 2709 bool should_notify_peers = false;
2710 bool should_notify_rtnl = false;
1f2cd845 2711 int delta_in_ticks;
1da177e4 2712
1f2cd845
DM		 2713 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
2714
2715 if (!bond_has_slaves(bond))
b2220cad
JV		 2716 goto re_arm;
2717
eb9fa4b0 2718 rcu_read_lock();
b0929915 2719
ad246c99
BH		 2720 should_notify_peers = bond_should_notify_peers(bond);
2721
b0929915 2722 if (bond_ab_arp_inspect(bond)) {
2723 rcu_read_unlock();
2724
1f2cd845
DM		 2725 /* Race avoidance with bond_close flush of workqueue */
2726 if (!rtnl_trylock()) {
1f2cd845
DM		 2727 delta_in_ticks = 1;
2728 should_notify_peers = false;
2729 goto re_arm;
2730 }
b2220cad 2731
1f2cd845 2732 bond_ab_arp_commit(bond);
b0929915 2733
1f2cd845 2734 rtnl_unlock();
b0929915 2735 rcu_read_lock();
1f2cd845
DM		 2736 }
2737
b0929915 2738 should_notify_rtnl = bond_ab_arp_probe(bond);
2739 rcu_read_unlock();
ad246c99 2740
80b9d236 2741re_arm:
2742 if (bond->params.arp_interval)
1f2cd845
DM		 2743 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
2744
b0929915 2745 if (should_notify_peers || should_notify_rtnl) {
1f2cd845
DM		 2746 if (!rtnl_trylock())
2747 return;
b0929915 2748
2749 if (should_notify_peers)
2750 call_netdevice_notifiers(NETDEV_NOTIFY_PEERS,
2751 bond->dev);
2752 if (should_notify_rtnl)
2753 bond_slave_state_notify(bond);
2754
1f2cd845
DM		 2755 rtnl_unlock();
2756 }
1da177e4
LT		 2757}
2758
1da177e4
LT		 2759/*-------------------------- netdev event handling --------------------------*/
2760
2761/*
2762 * Change device name
2763 */
2764static int bond_event_changename(struct bonding *bond)
2765{
1da177e4
LT		 2766 bond_remove_proc_entry(bond);
2767 bond_create_proc_entry(bond);
7e083840 2768
f073c7ca
TI		 2769 bond_debug_reregister(bond);
2770
1da177e4
LT		 2771 return NOTIFY_DONE;
2772}
2773
3d632c3f
SH		 2774static int bond_master_netdev_event(unsigned long event,
2775 struct net_device *bond_dev)
1da177e4 2776{
454d7c9b 2777 struct bonding *event_bond = netdev_priv(bond_dev);
1da177e4
LT		 2778
2779 switch (event) {
2780 case NETDEV_CHANGENAME:
2781 return bond_event_changename(event_bond);
a64d49c3
EB		 2782 case NETDEV_UNREGISTER:
2783 bond_remove_proc_entry(event_bond);
2784 break;
2785 case NETDEV_REGISTER:
2786 bond_create_proc_entry(event_bond);
2787 break;
6c8d23f7 2788 case NETDEV_NOTIFY_PEERS:
2789 if (event_bond->send_peer_notif)
2790 event_bond->send_peer_notif--;
2791 break;
1da177e4
LT		 2792 default:
2793 break;
2794 }
2795
2796 return NOTIFY_DONE;
2797}
2798
3d632c3f
SH		 2799static int bond_slave_netdev_event(unsigned long event,
2800 struct net_device *slave_dev)
1da177e4 2801{
471cb5a3 2802 struct slave *slave = bond_slave_get_rtnl(slave_dev);
6101391d 2803 struct bonding *bond;
2804 struct net_device *bond_dev;
471cb5a3
JP		 2805 u32 old_speed;
2806 u8 old_duplex;
1da177e4 2807
6101391d 2808 /* A netdev event can be generated while enslaving a device
2809 * before netdev_rx_handler_register is called in which case
2810 * slave will be NULL
2811 */
2812 if (!slave)
2813 return NOTIFY_DONE;
2814 bond_dev = slave->bond->dev;
2815 bond = slave->bond;
2816
1da177e4
LT		 2817 switch (event) {
2818 case NETDEV_UNREGISTER:
8d2ada77 2819 if (bond_dev->type != ARPHRD_ETHER)
471cb5a3
JP		 2820 bond_release_and_destroy(bond_dev, slave_dev);
2821 else
2822 bond_release(bond_dev, slave_dev);
1da177e4 2823 break;
98f41f69 2824 case NETDEV_UP:
1da177e4 2825 case NETDEV_CHANGE:
471cb5a3
JP		 2826 old_speed = slave->speed;
2827 old_duplex = slave->duplex;
17d04500 2828
471cb5a3 2829 bond_update_speed_duplex(slave);
17d04500 2830
471cb5a3
JP		 2831 if (bond->params.mode == BOND_MODE_8023AD) {
2832 if (old_speed != slave->speed)
2833 bond_3ad_adapter_speed_changed(slave);
2834 if (old_duplex != slave->duplex)
2835 bond_3ad_adapter_duplex_changed(slave);
17d04500 2836 }
1da177e4
LT		 2837 break;
2838 case NETDEV_DOWN:
2839 /*
2840 * ... Or is it this?
2841 */
2842 break;
2843 case NETDEV_CHANGEMTU:
2844 /*
2845 * TODO: Should slaves be allowed to
2846 * independently alter their MTU? For
2847 * an active-backup bond, slaves need
2848 * not be the same type of device, so
2849 * MTUs may vary. For other modes,
2850 * slaves arguably should have the
2851 * same MTUs. To do this, we'd need to
2852 * take over the slave's change_mtu
2853 * function for the duration of their
2854 * servitude.
2855 */
2856 break;
2857 case NETDEV_CHANGENAME:
3ec775b9
VF		 2858 /* we don't care if we don't have primary set */
2859 if (!USES_PRIMARY(bond->params.mode) ||
2860 !bond->params.primary[0])
2861 break;
2862
2863 if (slave == bond->primary_slave) {
2864 /* slave's name changed - he's no longer primary */
2865 bond->primary_slave = NULL;
2866 } else if (!strcmp(slave_dev->name, bond->params.primary)) {
2867 /* we have a new primary slave */
2868 bond->primary_slave = slave;
2869 } else { /* we didn't change primary - exit */
2870 break;
2871 }
2872
90194264
JP		 2873 pr_info("%s: Primary slave changed to %s, reselecting active slave\n",
2874 bond->dev->name,
2875 bond->primary_slave ? slave_dev->name : "none");
f80889a5 2876
2877 block_netpoll_tx();
3ec775b9
VF		 2878 write_lock_bh(&bond->curr_slave_lock);
2879 bond_select_active_slave(bond);
2880 write_unlock_bh(&bond->curr_slave_lock);
f80889a5 2881 unblock_netpoll_tx();
1da177e4 2882 break;
8531c5ff
AK		 2883 case NETDEV_FEAT_CHANGE:
2884 bond_compute_features(bond);
2885 break;
4aa5dee4
JP		 2886 case NETDEV_RESEND_IGMP:
2887 /* Propagate to master device */
2888 call_netdevice_notifiers(event, slave->bond->dev);
2889 break;
1da177e4
LT		 2890 default:
2891 break;
2892 }
2893
2894 return NOTIFY_DONE;
2895}
2896
2897/*
2898 * bond_netdev_event: handle netdev notifier chain events.
2899 *
2900 * This function receives events for the netdev chain. The caller (an
e041c683 2901 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
1da177e4
LT		 2902 * locks for us to safely manipulate the slave devices (RTNL lock,
2903 * dev_probe_lock).
2904 */
3d632c3f
SH		 2905static int bond_netdev_event(struct notifier_block *this,
2906 unsigned long event, void *ptr)
1da177e4 2907{
351638e7 2908 struct net_device *event_dev = netdev_notifier_info_to_dev(ptr);
1da177e4 2909
5a03cdb7 2910 pr_debug("event_dev: %s, event: %lx\n",
90194264 2911 event_dev ? event_dev->name : "None", event);
1da177e4 2912
0b680e75
JV		 2913 if (!(event_dev->priv_flags & IFF_BONDING))
2914 return NOTIFY_DONE;
2915
1da177e4 2916 if (event_dev->flags & IFF_MASTER) {
5a03cdb7 2917 pr_debug("IFF_MASTER\n");
1da177e4
LT		 2918 return bond_master_netdev_event(event, event_dev);
2919 }
2920
2921 if (event_dev->flags & IFF_SLAVE) {
5a03cdb7 2922 pr_debug("IFF_SLAVE\n");
1da177e4
LT		 2923 return bond_slave_netdev_event(event, event_dev);
2924 }
2925
2926 return NOTIFY_DONE;
2927}
2928
2929static struct notifier_block bond_netdev_notifier = {
2930 .notifier_call = bond_netdev_event,
2931};
2932
169a3e66
JV		 2933/*---------------------------- Hashing Policies -----------------------------*/
2934
32819dc1
NA		 2935/* L2 hash helper */
2936static inline u32 bond_eth_hash(struct sk_buff *skb)
6b923cb7
JE		 2937{
2938 struct ethhdr *data = (struct ethhdr *)skb->data;
2939
2940 if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto))
32819dc1 2941 return data->h_dest[5] ^ data->h_source[5];
6b923cb7
JE		 2942
2943 return 0;
2944}
2945
32819dc1
NA		 2946/* Extract the appropriate headers based on bond's xmit policy */
2947static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
2948 struct flow_keys *fk)
6f6652be 2949{
32819dc1 2950 const struct ipv6hdr *iph6;
4394542c 2951 const struct iphdr *iph;
32819dc1 2952 int noff, proto = -1;
6b923cb7 2953
32819dc1
NA		 2954 if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
2955 return skb_flow_dissect(skb, fk);
2956
2957 fk->ports = 0;
2958 noff = skb_network_offset(skb);
2959 if (skb->protocol == htons(ETH_P_IP)) {
054bb880 2960 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
32819dc1 2961 return false;
6b923cb7 2962 iph = ip_hdr(skb);
32819dc1
NA		 2963 fk->src = iph->saddr;
2964 fk->dst = iph->daddr;
2965 noff += iph->ihl << 2;
2966 if (!ip_is_fragment(iph))
2967 proto = iph->protocol;
2968 } else if (skb->protocol == htons(ETH_P_IPV6)) {
054bb880 2969 if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
32819dc1
NA		 2970 return false;
2971 iph6 = ipv6_hdr(skb);
2972 fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr);
2973 fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr);
2974 noff += sizeof(*iph6);
2975 proto = iph6->nexthdr;
2976 } else {
2977 return false;
6f6652be 2978 }
32819dc1
NA		 2979 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
2980 fk->ports = skb_flow_get_ports(skb, noff, proto);
6f6652be 2981
32819dc1 2982 return true;
6f6652be
JV		 2983}
2984
32819dc1
NA		 2985/**
2986 * bond_xmit_hash - generate a hash value based on the xmit policy
2987 * @bond: bonding device
2988 * @skb: buffer to use for headers
2989 * @count: modulo value
2990 *
2991 * This function will extract the necessary headers from the skb buffer and use
2992 * them to generate a hash based on the xmit_policy set in the bonding device
2993 * which will be reduced modulo count before returning.
169a3e66 2994 */
32819dc1 2995int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count)
169a3e66 2996{
32819dc1
NA		 2997 struct flow_keys flow;
2998 u32 hash;
4394542c 2999
32819dc1
NA		 3000 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 ||
3001 !bond_flow_dissect(bond, skb, &flow))
3002 return bond_eth_hash(skb) % count;
169a3e66 3003
32819dc1
NA		 3004 if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 ||
3005 bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
3006 hash = bond_eth_hash(skb);
3007 else
3008 hash = (__force u32)flow.ports;
3009 hash ^= (__force u32)flow.dst ^ (__force u32)flow.src;
3010 hash ^= (hash >> 16);
3011 hash ^= (hash >> 8);
3012
3013 return hash % count;
169a3e66
JV		 3014}
3015
1da177e4
LT		 3016/*-------------------------- Device entry points ----------------------------*/
3017
fbb0c41b 3018static void bond_work_init_all(struct bonding *bond)
3019{
3020 INIT_DELAYED_WORK(&bond->mcast_work,
3021 bond_resend_igmp_join_requests_delayed);
3022 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3023 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
3024 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3025 INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon);
3026 else
3027 INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon);
3028 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
3029}
3030
3031static void bond_work_cancel_all(struct bonding *bond)
3032{
3033 cancel_delayed_work_sync(&bond->mii_work);
3034 cancel_delayed_work_sync(&bond->arp_work);
3035 cancel_delayed_work_sync(&bond->alb_work);
3036 cancel_delayed_work_sync(&bond->ad_work);
3037 cancel_delayed_work_sync(&bond->mcast_work);
3038}
3039
1da177e4
LT		 3040static int bond_open(struct net_device *bond_dev)
3041{
454d7c9b 3042 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3043 struct list_head *iter;
ba3211cc 3044 struct slave *slave;
1da177e4 3045
ba3211cc
PP		 3046 /* reset slave->backup and slave->inactive */
3047 read_lock(&bond->lock);
0965a1f3 3048 if (bond_has_slaves(bond)) {
ba3211cc 3049 read_lock(&bond->curr_slave_lock);
9caff1e7 3050 bond_for_each_slave(bond, slave, iter) {
ba3211cc
PP		 3051 if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3052 && (slave != bond->curr_active_slave)) {
5e5b0665 3053 bond_set_slave_inactive_flags(slave,
3054 BOND_SLAVE_NOTIFY_NOW);
ba3211cc 3055 } else {
5e5b0665 3056 bond_set_slave_active_flags(slave,
3057 BOND_SLAVE_NOTIFY_NOW);
ba3211cc
PP		 3058 }
3059 }
3060 read_unlock(&bond->curr_slave_lock);
3061 }
3062 read_unlock(&bond->lock);
3063
fbb0c41b 3064 bond_work_init_all(bond);
5a37e8ca 3065
58402054 3066 if (bond_is_lb(bond)) {
1da177e4
LT		 3067 /* bond_alb_initialize must be called before the timer
3068 * is started.
3069 */
fbb0c41b 3070 if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB)))
b473946a 3071 return -ENOMEM;
1b76b316 3072 queue_delayed_work(bond->wq, &bond->alb_work, 0);
1da177e4
LT		 3073 }
3074
fbb0c41b 3075 if (bond->params.miimon) /* link check interval, in milliseconds. */
1b76b316 3076 queue_delayed_work(bond->wq, &bond->mii_work, 0);
1da177e4
LT		 3077
3078 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316 3079 queue_delayed_work(bond->wq, &bond->arp_work, 0);
3fe68df9 3080 bond->recv_probe = bond_arp_rcv;
1da177e4
LT		 3081 }
3082
3083 if (bond->params.mode == BOND_MODE_8023AD) {
1b76b316 3084 queue_delayed_work(bond->wq, &bond->ad_work, 0);
1da177e4 3085 /* register to receive LACPDUs */
3aba891d 3086 bond->recv_probe = bond_3ad_lacpdu_recv;
fd989c83 3087 bond_3ad_initiate_agg_selection(bond, 1);
1da177e4
LT		 3088 }
3089
3090 return 0;
3091}
3092
3093static int bond_close(struct net_device *bond_dev)
3094{
454d7c9b 3095 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3096
fbb0c41b 3097 bond_work_cancel_all(bond);
6c8d23f7 3098 bond->send_peer_notif = 0;
ee8487c0 3099 if (bond_is_lb(bond))
1da177e4 3100 bond_alb_deinitialize(bond);
3aba891d 3101 bond->recv_probe = NULL;
1da177e4
LT		 3102
3103 return 0;
3104}
3105
28172739
ED		 3106static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
3107 struct rtnl_link_stats64 *stats)
1da177e4 3108{
454d7c9b 3109 struct bonding *bond = netdev_priv(bond_dev);
28172739 3110 struct rtnl_link_stats64 temp;
9caff1e7 3111 struct list_head *iter;
1da177e4 3112 struct slave *slave;
1da177e4 3113
28172739 3114 memset(stats, 0, sizeof(*stats));
1da177e4
LT		 3115
3116 read_lock_bh(&bond->lock);
9caff1e7 3117 bond_for_each_slave(bond, slave, iter) {
be1f3c2c 3118 const struct rtnl_link_stats64 *sstats =
28172739
ED		 3119 dev_get_stats(slave->dev, &temp);
3120
3121 stats->rx_packets += sstats->rx_packets;
3122 stats->rx_bytes += sstats->rx_bytes;
3123 stats->rx_errors += sstats->rx_errors;
3124 stats->rx_dropped += sstats->rx_dropped;
3125
3126 stats->tx_packets += sstats->tx_packets;
3127 stats->tx_bytes += sstats->tx_bytes;
3128 stats->tx_errors += sstats->tx_errors;
3129 stats->tx_dropped += sstats->tx_dropped;
3130
3131 stats->multicast += sstats->multicast;
3132 stats->collisions += sstats->collisions;
3133
3134 stats->rx_length_errors += sstats->rx_length_errors;
3135 stats->rx_over_errors += sstats->rx_over_errors;
3136 stats->rx_crc_errors += sstats->rx_crc_errors;
3137 stats->rx_frame_errors += sstats->rx_frame_errors;
3138 stats->rx_fifo_errors += sstats->rx_fifo_errors;
3139 stats->rx_missed_errors += sstats->rx_missed_errors;
3140
3141 stats->tx_aborted_errors += sstats->tx_aborted_errors;
3142 stats->tx_carrier_errors += sstats->tx_carrier_errors;
3143 stats->tx_fifo_errors += sstats->tx_fifo_errors;
3144 stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
3145 stats->tx_window_errors += sstats->tx_window_errors;
2439f9eb 3146 }
1da177e4
LT		 3147 read_unlock_bh(&bond->lock);
3148
3149 return stats;
3150}
3151
3152static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3153{
080a06e1 3154 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3155 struct net_device *slave_dev = NULL;
3156 struct ifbond k_binfo;
3157 struct ifbond __user *u_binfo = NULL;
3158 struct ifslave k_sinfo;
3159 struct ifslave __user *u_sinfo = NULL;
3160 struct mii_ioctl_data *mii = NULL;
d1fbd3ed 3161 struct bond_opt_value newval;
387ff911 3162 struct net *net;
1da177e4
LT		 3163 int res = 0;
3164
a4aee5c8 3165 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd);
1da177e4
LT		 3166
3167 switch (cmd) {
1da177e4
LT		 3168 case SIOCGMIIPHY:
3169 mii = if_mii(ifr);
3d632c3f 3170 if (!mii)
1da177e4 3171 return -EINVAL;
3d632c3f 3172
1da177e4
LT		 3173 mii->phy_id = 0;
3174 /* Fall Through */
3175 case SIOCGMIIREG:
3176 /*
3177 * We do this again just in case we were called by SIOCGMIIREG
3178 * instead of SIOCGMIIPHY.
3179 */
3180 mii = if_mii(ifr);
3d632c3f 3181 if (!mii)
1da177e4 3182 return -EINVAL;
3d632c3f 3183
1da177e4
LT		 3184
3185 if (mii->reg_num == 1) {
1da177e4 3186 mii->val_out = 0;
6603a6f2 3187 read_lock(&bond->lock);
1da177e4 3188 read_lock(&bond->curr_slave_lock);
3d632c3f 3189 if (netif_carrier_ok(bond->dev))
1da177e4 3190 mii->val_out = BMSR_LSTATUS;
3d632c3f 3191
1da177e4 3192 read_unlock(&bond->curr_slave_lock);
6603a6f2 3193 read_unlock(&bond->lock);
1da177e4
LT		 3194 }
3195
3196 return 0;
3197 case BOND_INFO_QUERY_OLD:
3198 case SIOCBONDINFOQUERY:
3199 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3200
3d632c3f 3201 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
1da177e4 3202 return -EFAULT;
1da177e4
LT		 3203
3204 res = bond_info_query(bond_dev, &k_binfo);
3d632c3f
SH		 3205 if (res == 0 &&
3206 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3207 return -EFAULT;
1da177e4
LT		 3208
3209 return res;
3210 case BOND_SLAVE_INFO_QUERY_OLD:
3211 case SIOCBONDSLAVEINFOQUERY:
3212 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3213
3d632c3f 3214 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
1da177e4 3215 return -EFAULT;
1da177e4
LT		 3216
3217 res = bond_slave_info_query(bond_dev, &k_sinfo);
3d632c3f
SH		 3218 if (res == 0 &&
3219 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3220 return -EFAULT;
1da177e4
LT		 3221
3222 return res;
3223 default:
3224 /* Go on */
3225 break;
3226 }
3227
387ff911
G		 3228 net = dev_net(bond_dev);
3229
3230 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1da177e4 3231 return -EPERM;
1da177e4 3232
0917b933 3233 slave_dev = __dev_get_by_name(net, ifr->ifr_slave);
1da177e4 3234
a4aee5c8 3235 pr_debug("slave_dev=%p:\n", slave_dev);
1da177e4 3236
3d632c3f 3237 if (!slave_dev)
0917b933 3238 return -ENODEV;
1da177e4 3239
0917b933
YX		 3240 pr_debug("slave_dev->name=%s:\n", slave_dev->name);
3241 switch (cmd) {
3242 case BOND_ENSLAVE_OLD:
3243 case SIOCBONDENSLAVE:
3244 res = bond_enslave(bond_dev, slave_dev);
3245 break;
3246 case BOND_RELEASE_OLD:
3247 case SIOCBONDRELEASE:
3248 res = bond_release(bond_dev, slave_dev);
3249 break;
3250 case BOND_SETHWADDR_OLD:
3251 case SIOCBONDSETHWADDR:
3252 bond_set_dev_addr(bond_dev, slave_dev);
3253 res = 0;
3254 break;
3255 case BOND_CHANGE_ACTIVE_OLD:
3256 case SIOCBONDCHANGEACTIVE:
d1fbd3ed
NA		 3257 bond_opt_initstr(&newval, slave_dev->name);
3258 res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval);
0917b933
YX		 3259 break;
3260 default:
3261 res = -EOPNOTSUPP;
1da177e4
LT		 3262 }
3263
1da177e4
LT		 3264 return res;
3265}
3266
d03462b9 3267static void bond_change_rx_flags(struct net_device *bond_dev, int change)
1da177e4 3268{
454d7c9b 3269 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3270
d03462b9
JP		 3271 if (change & IFF_PROMISC)
3272 bond_set_promiscuity(bond,
3273 bond_dev->flags & IFF_PROMISC ? 1 : -1);
3d632c3f 3274
d03462b9
JP		 3275 if (change & IFF_ALLMULTI)
3276 bond_set_allmulti(bond,
3277 bond_dev->flags & IFF_ALLMULTI ? 1 : -1);
3278}
1da177e4 3279
303d1cbf 3280static void bond_set_rx_mode(struct net_device *bond_dev)
d03462b9
JP		 3281{
3282 struct bonding *bond = netdev_priv(bond_dev);
9caff1e7 3283 struct list_head *iter;
303d1cbf 3284 struct slave *slave;
1da177e4 3285
80ee5ad2 3286
b3241870 3287 rcu_read_lock();
303d1cbf 3288 if (USES_PRIMARY(bond->params.mode)) {
b3241870 3289 slave = rcu_dereference(bond->curr_active_slave);
303d1cbf
JV		 3290 if (slave) {
3291 dev_uc_sync(slave->dev, bond_dev);
3292 dev_mc_sync(slave->dev, bond_dev);
3293 }
303d1cbf 3294 } else {
b3241870 3295 bond_for_each_slave_rcu(bond, slave, iter) {
303d1cbf
JV		 3296 dev_uc_sync_multiple(slave->dev, bond_dev);
3297 dev_mc_sync_multiple(slave->dev, bond_dev);
3298 }
1da177e4 3299 }
b3241870 3300 rcu_read_unlock();
1da177e4
LT		 3301}
3302
234bcf8a 3303static int bond_neigh_init(struct neighbour *n)
00829823 3304{
234bcf8a 3305 struct bonding *bond = netdev_priv(n->dev);
234bcf8a
SP		 3306 const struct net_device_ops *slave_ops;
3307 struct neigh_parms parms;
dec1e90e 3308 struct slave *slave;
234bcf8a
SP		 3309 int ret;
3310
dec1e90e 3311 slave = bond_first_slave(bond);
234bcf8a
SP		 3312 if (!slave)
3313 return 0;
234bcf8a 3314 slave_ops = slave->dev->netdev_ops;
234bcf8a
SP		 3315 if (!slave_ops->ndo_neigh_setup)
3316 return 0;
3317
3318 parms.neigh_setup = NULL;
3319 parms.neigh_cleanup = NULL;
3320 ret = slave_ops->ndo_neigh_setup(slave->dev, &parms);
3321 if (ret)
3322 return ret;
3323
3324 /*
3325 * Assign slave's neigh_cleanup to neighbour in case cleanup is called
3326 * after the last slave has been detached. Assumes that all slaves
3327 * utilize the same neigh_cleanup (true at this writing as only user
3328 * is ipoib).
3329 */
3330 n->parms->neigh_cleanup = parms.neigh_cleanup;
3331
3332 if (!parms.neigh_setup)
3333 return 0;
3334
3335 return parms.neigh_setup(n);
3336}
3337
3338/*
3339 * The bonding ndo_neigh_setup is called at init time beofre any
3340 * slave exists. So we must declare proxy setup function which will
3341 * be used at run time to resolve the actual slave neigh param setup.
9918d5bf
VF		 3342 *
3343 * It's also called by master devices (such as vlans) to setup their
3344 * underlying devices. In that case - do nothing, we're already set up from
3345 * our init.
234bcf8a
SP		 3346 */
3347static int bond_neigh_setup(struct net_device *dev,
3348 struct neigh_parms *parms)
3349{
9918d5bf
VF		 3350 /* modify only our neigh_parms */
3351 if (parms->dev == dev)
3352 parms->neigh_setup = bond_neigh_init;
00829823 3353
00829823
SH		 3354 return 0;
3355}
3356
1da177e4
LT		 3357/*
3358 * Change the MTU of all of a master's slaves to match the master
3359 */
3360static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
3361{
454d7c9b 3362 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3363 struct slave *slave, *rollback_slave;
9caff1e7 3364 struct list_head *iter;
1da177e4 3365 int res = 0;
1da177e4 3366
90194264
JP		 3367 pr_debug("bond=%p, name=%s, new_mtu=%d\n",
3368 bond, bond_dev ? bond_dev->name : "None", new_mtu);
1da177e4
LT		 3369
3370 /* Can't hold bond->lock with bh disabled here since
3371 * some base drivers panic. On the other hand we can't
3372 * hold bond->lock without bh disabled because we'll
3373 * deadlock. The only solution is to rely on the fact
3374 * that we're under rtnl_lock here, and the slaves
3375 * list won't change. This doesn't solve the problem
3376 * of setting the slave's MTU while it is
3377 * transmitting, but the assumption is that the base
3378 * driver can handle that.
3379 *
3380 * TODO: figure out a way to safely iterate the slaves
3381 * list, but without holding a lock around the actual
3382 * call to the base driver.
3383 */
3384
9caff1e7 3385 bond_for_each_slave(bond, slave, iter) {
c8c23903 3386 pr_debug("s %p c_m %p\n",
90194264 3387 slave, slave->dev->netdev_ops->ndo_change_mtu);
e944ef79 3388
1da177e4
LT		 3389 res = dev_set_mtu(slave->dev, new_mtu);
3390
3391 if (res) {
3392 /* If we failed to set the slave's mtu to the new value
3393 * we must abort the operation even in ACTIVE_BACKUP
3394 * mode, because if we allow the backup slaves to have
3395 * different mtu values than the active slave we'll
3396 * need to change their mtu when doing a failover. That
3397 * means changing their mtu from timer context, which
3398 * is probably not a good idea.
3399 */
5a03cdb7 3400 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 3401 goto unwind;
3402 }
3403 }
3404
3405 bond_dev->mtu = new_mtu;
3406
3407 return 0;
3408
3409unwind:
3410 /* unwind from head to the slave that failed */
9caff1e7 3411 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT		 3412 int tmp_res;
3413
81f23b13
VF		 3414 if (rollback_slave == slave)
3415 break;
3416
3417 tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu);
1da177e4 3418 if (tmp_res) {
a4aee5c8 3419 pr_debug("unwind err %d dev %s\n",
81f23b13 3420 tmp_res, rollback_slave->dev->name);
1da177e4
LT		 3421 }
3422 }
3423
3424 return res;
3425}
3426
3427/*
3428 * Change HW address
3429 *
3430 * Note that many devices must be down to change the HW address, and
3431 * downing the master releases all slaves. We can make bonds full of
3432 * bonding devices to test this, however.
3433 */
3434static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
3435{
454d7c9b 3436 struct bonding *bond = netdev_priv(bond_dev);
81f23b13 3437 struct slave *slave, *rollback_slave;
1da177e4 3438 struct sockaddr *sa = addr, tmp_sa;
9caff1e7 3439 struct list_head *iter;
1da177e4 3440 int res = 0;
1da177e4 3441
eb7cc59a
SH		 3442 if (bond->params.mode == BOND_MODE_ALB)
3443 return bond_alb_set_mac_address(bond_dev, addr);
3444
3445
a4aee5c8
JP		 3446 pr_debug("bond=%p, name=%s\n",
3447 bond, bond_dev ? bond_dev->name : "None");
1da177e4 3448
1b5acd29
JV		 3449 /* If fail_over_mac is enabled, do nothing and return success.
3450 * Returning an error causes ifenslave to fail.
dd957c57 3451 */
cc689aaa 3452 if (bond->params.fail_over_mac &&
3453 bond->params.mode == BOND_MODE_ACTIVEBACKUP)
dd957c57 3454 return 0;
2ab82852 3455
3d632c3f 3456 if (!is_valid_ether_addr(sa->sa_data))
1da177e4 3457 return -EADDRNOTAVAIL;
1da177e4
LT		 3458
3459 /* Can't hold bond->lock with bh disabled here since
3460 * some base drivers panic. On the other hand we can't
3461 * hold bond->lock without bh disabled because we'll
3462 * deadlock. The only solution is to rely on the fact
3463 * that we're under rtnl_lock here, and the slaves
3464 * list won't change. This doesn't solve the problem
3465 * of setting the slave's hw address while it is
3466 * transmitting, but the assumption is that the base
3467 * driver can handle that.
3468 *
3469 * TODO: figure out a way to safely iterate the slaves
3470 * list, but without holding a lock around the actual
3471 * call to the base driver.
3472 */
3473
9caff1e7 3474 bond_for_each_slave(bond, slave, iter) {
5a03cdb7 3475 pr_debug("slave %p %s\n", slave, slave->dev->name);
1da177e4
LT		 3476 res = dev_set_mac_address(slave->dev, addr);
3477 if (res) {
3478 /* TODO: consider downing the slave
3479 * and retry ?
3480 * User should expect communications
3481 * breakage anyway until ARP finish
3482 * updating, so...
3483 */
5a03cdb7 3484 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 3485 goto unwind;
3486 }
3487 }
3488
3489 /* success */
3490 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
3491 return 0;
3492
3493unwind:
3494 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
3495 tmp_sa.sa_family = bond_dev->type;
3496
3497 /* unwind from head to the slave that failed */
9caff1e7 3498 bond_for_each_slave(bond, rollback_slave, iter) {
1da177e4
LT		 3499 int tmp_res;
3500
81f23b13
VF		 3501 if (rollback_slave == slave)
3502 break;
3503
3504 tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa);
1da177e4 3505 if (tmp_res) {
a4aee5c8 3506 pr_debug("unwind err %d dev %s\n",
81f23b13 3507 tmp_res, rollback_slave->dev->name);
1da177e4
LT		 3508 }
3509 }
3510
3511 return res;
3512}
3513
15077228
NA		 3514/**
3515 * bond_xmit_slave_id - transmit skb through slave with slave_id
3516 * @bond: bonding device that is transmitting
3517 * @skb: buffer to transmit
3518 * @slave_id: slave id up to slave_cnt-1 through which to transmit
3519 *
3520 * This function tries to transmit through slave with slave_id but in case
3521 * it fails, it tries to find the first available slave for transmission.
3522 * The skb is consumed in all cases, thus the function is void.
3523 */
da131ddb 3524static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id)
15077228 3525{
9caff1e7 3526 struct list_head *iter;
15077228
NA		 3527 struct slave *slave;
3528 int i = slave_id;
3529
3530 /* Here we start from the slave with slave_id */
9caff1e7 3531 bond_for_each_slave_rcu(bond, slave, iter) {
15077228
NA		 3532 if (--i < 0) {
3533 if (slave_can_tx(slave)) {
3534 bond_dev_queue_xmit(bond, skb, slave->dev);
3535 return;
3536 }
3537 }
3538 }
3539
3540 /* Here we start from the first slave up to slave_id */
3541 i = slave_id;
9caff1e7 3542 bond_for_each_slave_rcu(bond, slave, iter) {
15077228
NA		 3543 if (--i < 0)
3544 break;
3545 if (slave_can_tx(slave)) {
3546 bond_dev_queue_xmit(bond, skb, slave->dev);
3547 return;
3548 }
3549 }
3550 /* no slave that can tx has been found */
2bb77ab4 3551 dev_kfree_skb_any(skb);
15077228
NA		 3552}
3553
73958329
NA		 3554/**
3555 * bond_rr_gen_slave_id - generate slave id based on packets_per_slave
3556 * @bond: bonding device to use
3557 *
3558 * Based on the value of the bonding device's packets_per_slave parameter
3559 * this function generates a slave id, which is usually used as the next
3560 * slave to transmit through.
3561 */
3562static u32 bond_rr_gen_slave_id(struct bonding *bond)
3563{
73958329 3564 u32 slave_id;
809fa972
HFS		 3565 struct reciprocal_value reciprocal_packets_per_slave;
3566 int packets_per_slave = bond->params.packets_per_slave;
73958329
NA		 3567
3568 switch (packets_per_slave) {
3569 case 0:
3570 slave_id = prandom_u32();
3571 break;
3572 case 1:
3573 slave_id = bond->rr_tx_counter;
3574 break;
3575 default:
809fa972
HFS		 3576 reciprocal_packets_per_slave =
3577 bond->params.reciprocal_packets_per_slave;
73958329 3578 slave_id = reciprocal_divide(bond->rr_tx_counter,
809fa972 3579 reciprocal_packets_per_slave);
73958329
NA		 3580 break;
3581 }
3582 bond->rr_tx_counter++;
3583
3584 return slave_id;
3585}
3586
1da177e4
LT		 3587static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
3588{
454d7c9b 3589 struct bonding *bond = netdev_priv(bond_dev);
a2fd940f 3590 struct iphdr *iph = ip_hdr(skb);
15077228 3591 struct slave *slave;
73958329 3592 u32 slave_id;
1da177e4 3593
73958329 3594 /* Start with the curr_active_slave that joined the bond as the
a2fd940f
AG		 3595 * default for sending IGMP traffic. For failover purposes one
3596 * needs to maintain some consistency for the interface that will
3597 * send the join/membership reports. The curr_active_slave found
3598 * will send all of this type of traffic.
cf5f9044 3599 */
15077228 3600 if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) {
278b2083 3601 slave = rcu_dereference(bond->curr_active_slave);
15077228
NA		 3602 if (slave && slave_can_tx(slave))
3603 bond_dev_queue_xmit(bond, skb, slave->dev);
3604 else
3605 bond_xmit_slave_id(bond, skb, 0);
a2fd940f 3606 } else {
73958329
NA		 3607 slave_id = bond_rr_gen_slave_id(bond);
3608 bond_xmit_slave_id(bond, skb, slave_id % bond->slave_cnt);
1da177e4 3609 }
0693e88e 3610
ec634fe3 3611 return NETDEV_TX_OK;
1da177e4
LT		 3612}
3613
3614/*
3615 * in active-backup mode, we know that bond->curr_active_slave is always valid if
3616 * the bond has a usable interface.
3617 */
3618static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
3619{
454d7c9b 3620 struct bonding *bond = netdev_priv(bond_dev);
71bc3b2d 3621 struct slave *slave;
1da177e4 3622
278b2083 3623 slave = rcu_dereference(bond->curr_active_slave);
71bc3b2d 3624 if (slave)
15077228
NA		 3625 bond_dev_queue_xmit(bond, skb, slave->dev);
3626 else
2bb77ab4 3627 dev_kfree_skb_any(skb);
0693e88e 3628
ec634fe3 3629 return NETDEV_TX_OK;
1da177e4
LT		 3630}
3631
32819dc1 3632/* In bond_xmit_xor() , we determine the output device by using a pre-
169a3e66
JV		 3633 * determined xmit_hash_policy(), If the selected device is not enabled,
3634 * find the next active slave.
1da177e4
LT		 3635 */
3636static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
3637{
454d7c9b 3638 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3639
32819dc1 3640 bond_xmit_slave_id(bond, skb, bond_xmit_hash(bond, skb, bond->slave_cnt));
0693e88e 3641
ec634fe3 3642 return NETDEV_TX_OK;
1da177e4
LT		 3643}
3644
78a646ce 3645/* in broadcast mode, we send everything to all usable interfaces. */
1da177e4
LT		 3646static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
3647{
454d7c9b 3648 struct bonding *bond = netdev_priv(bond_dev);
78a646ce 3649 struct slave *slave = NULL;
9caff1e7 3650 struct list_head *iter;
1da177e4 3651
9caff1e7 3652 bond_for_each_slave_rcu(bond, slave, iter) {
78a646ce
NA		 3653 if (bond_is_last_slave(bond, slave))
3654 break;
3655 if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) {
3656 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1da177e4 3657
78a646ce 3658 if (!skb2) {
9152e26d 3659 net_err_ratelimited("%s: Error: %s: skb_clone() failed\n",
3660 bond_dev->name, __func__);
78a646ce 3661 continue;
1da177e4 3662 }
78a646ce
NA		 3663 /* bond_dev_queue_xmit always returns 0 */
3664 bond_dev_queue_xmit(bond, skb2, slave->dev);
1da177e4
LT		 3665 }
3666 }
78a646ce
NA		 3667 if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP)
3668 bond_dev_queue_xmit(bond, skb, slave->dev);
3669 else
2bb77ab4 3670 dev_kfree_skb_any(skb);
3d632c3f 3671
ec634fe3 3672 return NETDEV_TX_OK;
1da177e4
LT		 3673}
3674
3675/*------------------------- Device initialization ---------------------------*/
3676
bb1d9123
AG		 3677/*
3678 * Lookup the slave that corresponds to a qid
3679 */
3680static inline int bond_slave_override(struct bonding *bond,
3681 struct sk_buff *skb)
3682{
bb1d9123 3683 struct slave *slave = NULL;
9caff1e7 3684 struct list_head *iter;
bb1d9123 3685
0693e88e
MM		 3686 if (!skb->queue_mapping)
3687 return 1;
bb1d9123
AG		 3688
3689 /* Find out if any slaves have the same mapping as this skb. */
3900f290 3690 bond_for_each_slave_rcu(bond, slave, iter) {
3691 if (slave->queue_id == skb->queue_mapping) {
3692 if (slave_can_tx(slave)) {
3693 bond_dev_queue_xmit(bond, skb, slave->dev);
3694 return 0;
3695 }
3696 /* If the slave isn't UP, use default transmit policy. */
bb1d9123
AG		 3697 break;
3698 }
3699 }
3700
3900f290 3701 return 1;
bb1d9123
AG		 3702}
3703
374eeb5a 3704
f663dd9a 3705static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb,
99932d4f 3706 void *accel_priv, select_queue_fallback_t fallback)
bb1d9123
AG		 3707{
3708 /*
3709 * This helper function exists to help dev_pick_tx get the correct
fd0e435b 3710 * destination queue. Using a helper function skips a call to
bb1d9123
AG		 3711 * skb_tx_hash and will put the skbs in the queue we expect on their
3712 * way down to the bonding driver.
3713 */
fd0e435b
PO		 3714 u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0;
3715
374eeb5a
NH		 3716 /*
3717 * Save the original txq to restore before passing to the driver
3718 */
df4ab5b3 3719 qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping;
374eeb5a 3720
fd0e435b 3721 if (unlikely(txq >= dev->real_num_tx_queues)) {
d30ee670 3722 do {
fd0e435b 3723 txq -= dev->real_num_tx_queues;
d30ee670 3724 } while (txq >= dev->real_num_tx_queues);
fd0e435b
PO		 3725 }
3726 return txq;
bb1d9123
AG		 3727}
3728
0693e88e 3729static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
00829823 3730{
bb1d9123
AG		 3731 struct bonding *bond = netdev_priv(dev);
3732
3733 if (TX_QUEUE_OVERRIDE(bond->params.mode)) {
3734 if (!bond_slave_override(bond, skb))
3735 return NETDEV_TX_OK;
3736 }
00829823
SH		 3737
3738 switch (bond->params.mode) {
3739 case BOND_MODE_ROUNDROBIN:
3740 return bond_xmit_roundrobin(skb, dev);
3741 case BOND_MODE_ACTIVEBACKUP:
3742 return bond_xmit_activebackup(skb, dev);
3743 case BOND_MODE_XOR:
3744 return bond_xmit_xor(skb, dev);
3745 case BOND_MODE_BROADCAST:
3746 return bond_xmit_broadcast(skb, dev);
3747 case BOND_MODE_8023AD:
3748 return bond_3ad_xmit_xor(skb, dev);
3749 case BOND_MODE_ALB:
3750 case BOND_MODE_TLB:
3751 return bond_alb_xmit(skb, dev);
3752 default:
3753 /* Should never happen, mode already checked */
a4aee5c8
JP		 3754 pr_err("%s: Error: Unknown bonding mode %d\n",
3755 dev->name, bond->params.mode);
00829823 3756 WARN_ON_ONCE(1);
2bb77ab4 3757 dev_kfree_skb_any(skb);
00829823
SH		 3758 return NETDEV_TX_OK;
3759 }
3760}
3761
0693e88e
MM		 3762static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
3763{
3764 struct bonding *bond = netdev_priv(dev);
3765 netdev_tx_t ret = NETDEV_TX_OK;
3766
3767 /*
3768 * If we risk deadlock from transmitting this in the
3769 * netpoll path, tell netpoll to queue the frame for later tx
3770 */
054bb880 3771 if (unlikely(is_netpoll_tx_blocked(dev)))
0693e88e
MM		 3772 return NETDEV_TX_BUSY;
3773
278b2083 3774 rcu_read_lock();
0965a1f3 3775 if (bond_has_slaves(bond))
0693e88e
MM		 3776 ret = __bond_start_xmit(skb, dev);
3777 else
2bb77ab4 3778 dev_kfree_skb_any(skb);
278b2083 3779 rcu_read_unlock();
0693e88e
MM		 3780
3781 return ret;
3782}
00829823 3783
bb5b052f
AG		 3784static int bond_ethtool_get_settings(struct net_device *bond_dev,
3785 struct ethtool_cmd *ecmd)
3786{
3787 struct bonding *bond = netdev_priv(bond_dev);
bb5b052f 3788 unsigned long speed = 0;
9caff1e7 3789 struct list_head *iter;
dec1e90e 3790 struct slave *slave;
bb5b052f
AG		 3791
3792 ecmd->duplex = DUPLEX_UNKNOWN;
3793 ecmd->port = PORT_OTHER;
3794
3795 /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we
3796 * do not need to check mode. Though link speed might not represent
3797 * the true receive or transmit bandwidth (not all modes are symmetric)
3798 * this is an accurate maximum.
3799 */
3800 read_lock(&bond->lock);
9caff1e7 3801 bond_for_each_slave(bond, slave, iter) {
bb5b052f
AG		 3802 if (SLAVE_IS_OK(slave)) {
3803 if (slave->speed != SPEED_UNKNOWN)
3804 speed += slave->speed;
3805 if (ecmd->duplex == DUPLEX_UNKNOWN &&
3806 slave->duplex != DUPLEX_UNKNOWN)
3807 ecmd->duplex = slave->duplex;
3808 }
3809 }
3810 ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN);
3811 read_unlock(&bond->lock);
dec1e90e 3812
bb5b052f
AG		 3813 return 0;
3814}
3815
217df670 3816static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
7826d43f 3817 struct ethtool_drvinfo *drvinfo)
217df670 3818{
7826d43f
JP		 3819 strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver));
3820 strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version));
3821 snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d",
3822 BOND_ABI_VERSION);
217df670
JV		 3823}
3824
7282d491 3825static const struct ethtool_ops bond_ethtool_ops = {
217df670 3826 .get_drvinfo = bond_ethtool_get_drvinfo,
bb5b052f 3827 .get_settings = bond_ethtool_get_settings,
fa53ebac 3828 .get_link = ethtool_op_get_link,
8531c5ff
AK		 3829};
3830
eb7cc59a 3831static const struct net_device_ops bond_netdev_ops = {
181470fc 3832 .ndo_init = bond_init,
9e71626c 3833 .ndo_uninit = bond_uninit,
eb7cc59a
SH		 3834 .ndo_open = bond_open,
3835 .ndo_stop = bond_close,
00829823 3836 .ndo_start_xmit = bond_start_xmit,
bb1d9123 3837 .ndo_select_queue = bond_select_queue,
be1f3c2c 3838 .ndo_get_stats64 = bond_get_stats,
eb7cc59a 3839 .ndo_do_ioctl = bond_do_ioctl,
d03462b9 3840 .ndo_change_rx_flags = bond_change_rx_flags,
303d1cbf 3841 .ndo_set_rx_mode = bond_set_rx_mode,
eb7cc59a 3842 .ndo_change_mtu = bond_change_mtu,
cc0e4070 3843 .ndo_set_mac_address = bond_set_mac_address,
00829823 3844 .ndo_neigh_setup = bond_neigh_setup,
cc0e4070 3845 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
eb7cc59a 3846 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
f6dc31a8 3847#ifdef CONFIG_NET_POLL_CONTROLLER
8a8efa22 3848 .ndo_netpoll_setup = bond_netpoll_setup,
f6dc31a8
WC		 3849 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
3850 .ndo_poll_controller = bond_poll_controller,
3851#endif
9232ecca
JP		 3852 .ndo_add_slave = bond_enslave,
3853 .ndo_del_slave = bond_release,
b2a103e6 3854 .ndo_fix_features = bond_fix_features,
eb7cc59a
SH		 3855};
3856
b3f92b63
DG		 3857static const struct device_type bond_type = {
3858 .name = "bond",
3859};
3860
9e2e61fb
AW		 3861static void bond_destructor(struct net_device *bond_dev)
3862{
3863 struct bonding *bond = netdev_priv(bond_dev);
3864 if (bond->wq)
3865 destroy_workqueue(bond->wq);
3866 free_netdev(bond_dev);
3867}
3868
0a2a78c4 3869void bond_setup(struct net_device *bond_dev)
1da177e4 3870{
454d7c9b 3871 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3872
1da177e4
LT		 3873 /* initialize rwlocks */
3874 rwlock_init(&bond->lock);
3875 rwlock_init(&bond->curr_slave_lock);
d2991f75 3876 bond->params = bonding_defaults;
1da177e4
LT		 3877
3878 /* Initialize pointers */
1da177e4 3879 bond->dev = bond_dev;
1da177e4
LT		 3880
3881 /* Initialize the device entry points */
181470fc 3882 ether_setup(bond_dev);
eb7cc59a 3883 bond_dev->netdev_ops = &bond_netdev_ops;
8531c5ff 3884 bond_dev->ethtool_ops = &bond_ethtool_ops;
1da177e4 3885
9e2e61fb 3886 bond_dev->destructor = bond_destructor;
1da177e4 3887
b3f92b63
DG		 3888 SET_NETDEV_DEVTYPE(bond_dev, &bond_type);
3889
1da177e4
LT		 3890 /* Initialize the device options */
3891 bond_dev->tx_queue_len = 0;
3892 bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
0b680e75 3893 bond_dev->priv_flags |= IFF_BONDING;
550fd08c 3894 bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
181470fc 3895
1da177e4
LT		 3896 /* At first, we block adding VLANs. That's the only way to
3897 * prevent problems that occur when adding VLANs over an
3898 * empty bond. The block will be removed once non-challenged
3899 * slaves are enslaved.
3900 */
3901 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
3902
932ff279 3903 /* don't acquire bond device's netif_tx_lock when
1da177e4
LT		 3904 * transmitting */
3905 bond_dev->features |= NETIF_F_LLTX;
3906
3907 /* By default, we declare the bond to be fully
3908 * VLAN hardware accelerated capable. Special
3909 * care is taken in the various xmit functions
3910 * when there are slaves that are not hw accel
3911 * capable
3912 */
1da177e4 3913
f9399814
WC		 3914 /* Don't allow bond devices to change network namespaces. */
3915 bond_dev->features |= NETIF_F_NETNS_LOCAL;
3916
b2a103e6 3917 bond_dev->hw_features = BOND_VLAN_FEATURES |
f646968f
PM		 3918 NETIF_F_HW_VLAN_CTAG_TX |
3919 NETIF_F_HW_VLAN_CTAG_RX |
3920 NETIF_F_HW_VLAN_CTAG_FILTER;
b2a103e6 3921
34324dc2 3922 bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM);
b2a103e6 3923 bond_dev->features |= bond_dev->hw_features;
1da177e4
LT		 3924}
3925
c67dfb29
EB		 3926/*
3927* Destroy a bonding device.
3928* Must be under rtnl_lock when this function is called.
3929*/
3930static void bond_uninit(struct net_device *bond_dev)
a434e43f 3931{
454d7c9b 3932 struct bonding *bond = netdev_priv(bond_dev);
544a028e
VF		 3933 struct list_head *iter;
3934 struct slave *slave;
a434e43f 3935
f6dc31a8
WC		 3936 bond_netpoll_cleanup(bond_dev);
3937
c67dfb29 3938 /* Release the bonded slaves */
544a028e 3939 bond_for_each_slave(bond, slave, iter)
dec1e90e 3940 __bond_release_one(bond_dev, slave->dev, true);
90194264 3941 pr_info("%s: Released all slaves\n", bond_dev->name);
c67dfb29 3942
a434e43f
JV		 3943 list_del(&bond->bond_list);
3944
f073c7ca 3945 bond_debug_unregister(bond);
a434e43f
JV		 3946}
3947
1da177e4
LT		 3948/*------------------------- Module initialization ---------------------------*/
3949
1da177e4
LT		 3950static int bond_check_params(struct bond_params *params)
3951{
5a5c5fd4 3952 int arp_validate_value, fail_over_mac_value, primary_reselect_value, i;
f3253339 3953 struct bond_opt_value newval;
3954 const struct bond_opt_value *valptr;
8599b52e 3955 int arp_all_targets_value;
f5b2b966 3956
1da177e4
LT		 3957 /*
3958 * Convert string parameters.
3959 */
3960 if (mode) {
2b3798d5
NA		 3961 bond_opt_initstr(&newval, mode);
3962 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval);
3963 if (!valptr) {
3964 pr_err("Error: Invalid bonding mode \"%s\"\n", mode);
1da177e4
LT		 3965 return -EINVAL;
3966 }
2b3798d5 3967 bond_mode = valptr->value;
1da177e4
LT		 3968 }
3969
169a3e66
JV		 3970 if (xmit_hash_policy) {
3971 if ((bond_mode != BOND_MODE_XOR) &&
3972 (bond_mode != BOND_MODE_8023AD)) {
a4aee5c8 3973 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
90194264 3974 bond_mode_name(bond_mode));
169a3e66 3975 } else {
a4b32ce7
NA		 3976 bond_opt_initstr(&newval, xmit_hash_policy);
3977 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH),
3978 &newval);
3979 if (!valptr) {
a4aee5c8 3980 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
169a3e66
JV		 3981 xmit_hash_policy);
3982 return -EINVAL;
3983 }
a4b32ce7 3984 xmit_hashtype = valptr->value;
169a3e66
JV		 3985 }
3986 }
3987
1da177e4
LT		 3988 if (lacp_rate) {
3989 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8
JP		 3990 pr_info("lacp_rate param is irrelevant in mode %s\n",
3991 bond_mode_name(bond_mode));
1da177e4 3992 } else {
d3131de7
NA		 3993 bond_opt_initstr(&newval, lacp_rate);
3994 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE),
3995 &newval);
3996 if (!valptr) {
a4aee5c8 3997 pr_err("Error: Invalid lacp rate \"%s\"\n",
d3131de7 3998 lacp_rate);
1da177e4
LT		 3999 return -EINVAL;
4000 }
d3131de7 4001 lacp_fast = valptr->value;
1da177e4
LT		 4002 }
4003 }
4004
fd989c83 4005 if (ad_select) {
9e5f5eeb
NA		 4006 bond_opt_initstr(&newval, lacp_rate);
4007 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT),
4008 &newval);
4009 if (!valptr) {
4010 pr_err("Error: Invalid ad_select \"%s\"\n", ad_select);
fd989c83
JV		 4011 return -EINVAL;
4012 }
9e5f5eeb
NA		 4013 params->ad_select = valptr->value;
4014 if (bond_mode != BOND_MODE_8023AD)
91565ebb 4015 pr_warn("ad_select param only affects 802.3ad mode\n");
fd989c83
JV		 4016 } else {
4017 params->ad_select = BOND_AD_STABLE;
4018 }
4019
f5841306 4020 if (max_bonds < 0) {
91565ebb
JP		 4021 pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4022 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
1da177e4
LT		 4023 max_bonds = BOND_DEFAULT_MAX_BONDS;
4024 }
4025
4026 if (miimon < 0) {
91565ebb
JP		 4027 pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4028 miimon, INT_MAX);
b98d9c66 4029 miimon = 0;
1da177e4
LT		 4030 }
4031
4032 if (updelay < 0) {
91565ebb
JP		 4033 pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4034 updelay, INT_MAX);
1da177e4
LT		 4035 updelay = 0;
4036 }
4037
4038 if (downdelay < 0) {
91565ebb
JP		 4039 pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4040 downdelay, INT_MAX);
1da177e4
LT		 4041 downdelay = 0;
4042 }
4043
4044 if ((use_carrier != 0) && (use_carrier != 1)) {
91565ebb
JP		 4045 pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4046 use_carrier);
1da177e4
LT		 4047 use_carrier = 1;
4048 }
4049
ad246c99 4050 if (num_peer_notif < 0 || num_peer_notif > 255) {
91565ebb
JP		 4051 pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
4052 num_peer_notif);
ad246c99
BH		 4053 num_peer_notif = 1;
4054 }
4055
834db4bc 4056 /* reset values for 802.3ad/TLB/ALB */
4057 if (BOND_NO_USES_ARP(bond_mode)) {
1da177e4 4058 if (!miimon) {
91565ebb
JP		 4059 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");
4060 pr_warn("Forcing miimon to 100msec\n");
fe9d04af 4061 miimon = BOND_DEFAULT_MIIMON;
1da177e4
LT		 4062 }
4063 }
4064
bb1d9123 4065 if (tx_queues < 1 || tx_queues > 255) {
91565ebb
JP		 4066 pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n",
4067 tx_queues, BOND_DEFAULT_TX_QUEUES);
bb1d9123
AG		 4068 tx_queues = BOND_DEFAULT_TX_QUEUES;
4069 }
4070
ebd8e497 4071 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
91565ebb
JP		 4072 pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n",
4073 all_slaves_active);
ebd8e497
AG		 4074 all_slaves_active = 0;
4075 }
4076
c2952c31 4077 if (resend_igmp < 0 || resend_igmp > 255) {
91565ebb
JP		 4078 pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n",
4079 resend_igmp, BOND_DEFAULT_RESEND_IGMP);
c2952c31
FL		 4080 resend_igmp = BOND_DEFAULT_RESEND_IGMP;
4081 }
4082
aa59d851
NA		 4083 bond_opt_initval(&newval, packets_per_slave);
4084 if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) {
73958329
NA		 4085 pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n",
4086 packets_per_slave, USHRT_MAX);
4087 packets_per_slave = 1;
4088 }
4089
1da177e4 4090 if (bond_mode == BOND_MODE_ALB) {
a4aee5c8
JP		 4091 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",
4092 updelay);
1da177e4
LT		 4093 }
4094
4095 if (!miimon) {
4096 if (updelay || downdelay) {
4097 /* just warn the user the up/down delay will have
4098 * no effect since miimon is zero...
4099 */
91565ebb
JP		 4100 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",
4101 updelay, downdelay);
1da177e4
LT		 4102 }
4103 } else {
4104 /* don't allow arp monitoring */
4105 if (arp_interval) {
91565ebb
JP		 4106 pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4107 miimon, arp_interval);
1da177e4
LT		 4108 arp_interval = 0;
4109 }
4110
4111 if ((updelay % miimon) != 0) {
91565ebb
JP		 4112 pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4113 updelay, miimon, (updelay / miimon) * miimon);
1da177e4
LT		 4114 }
4115
4116 updelay /= miimon;
4117
4118 if ((downdelay % miimon) != 0) {
91565ebb
JP		 4119 pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4120 downdelay, miimon,
4121 (downdelay / miimon) * miimon);
1da177e4
LT		 4122 }
4123
4124 downdelay /= miimon;
4125 }
4126
4127 if (arp_interval < 0) {
91565ebb
JP		 4128 pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4129 arp_interval, INT_MAX);
7bdb04ed 4130 arp_interval = 0;
1da177e4
LT		 4131 }
4132
5a5c5fd4 4133 for (arp_ip_count = 0, i = 0;
4134 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) {
1da177e4
LT		 4135 /* not complete check, but should be good enough to
4136 catch mistakes */
89015c18 4137 __be32 ip;
4138 if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) ||
4139 IS_IP_TARGET_UNUSABLE_ADDRESS(ip)) {
91565ebb
JP		 4140 pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4141 arp_ip_target[i]);
1da177e4
LT		 4142 arp_interval = 0;
4143 } else {
0afee4e8
VF		 4144 if (bond_get_targets_ip(arp_target, ip) == -1)
4145 arp_target[arp_ip_count++] = ip;
4146 else
91565ebb
JP		 4147 pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n",
4148 &ip);
1da177e4
LT		 4149 }
4150 }
4151
4152 if (arp_interval && !arp_ip_count) {
4153 /* don't allow arping if no arp_ip_target given... */
91565ebb
JP		 4154 pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4155 arp_interval);
1da177e4
LT		 4156 arp_interval = 0;
4157 }
4158
f5b2b966 4159 if (arp_validate) {
f5b2b966 4160 if (!arp_interval) {
a4aee5c8 4161 pr_err("arp_validate requires arp_interval\n");
f5b2b966
JV		 4162 return -EINVAL;
4163 }
4164
16228881
NA		 4165 bond_opt_initstr(&newval, arp_validate);
4166 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE),
4167 &newval);
4168 if (!valptr) {
a4aee5c8 4169 pr_err("Error: invalid arp_validate \"%s\"\n",
16228881 4170 arp_validate);
f5b2b966
JV		 4171 return -EINVAL;
4172 }
16228881
NA		 4173 arp_validate_value = valptr->value;
4174 } else {
f5b2b966 4175 arp_validate_value = 0;
16228881 4176 }
f5b2b966 4177
8599b52e
VF		 4178 arp_all_targets_value = 0;
4179 if (arp_all_targets) {
edf36b24
NA		 4180 bond_opt_initstr(&newval, arp_all_targets);
4181 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS),
4182 &newval);
4183 if (!valptr) {
8599b52e
VF		 4184 pr_err("Error: invalid arp_all_targets_value \"%s\"\n",
4185 arp_all_targets);
4186 arp_all_targets_value = 0;
edf36b24
NA		 4187 } else {
4188 arp_all_targets_value = valptr->value;
8599b52e
VF		 4189 }
4190 }
4191
1da177e4 4192 if (miimon) {
a4aee5c8 4193 pr_info("MII link monitoring set to %d ms\n", miimon);
1da177e4 4194 } else if (arp_interval) {
16228881
NA		 4195 valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE,
4196 arp_validate_value);
a4aee5c8 4197 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
16228881 4198 arp_interval, valptr->string, arp_ip_count);
1da177e4
LT		 4199
4200 for (i = 0; i < arp_ip_count; i++)
90194264 4201 pr_cont(" %s", arp_ip_target[i]);
1da177e4 4202
90194264 4203 pr_cont("\n");
1da177e4 4204
b8a9787e 4205 } else if (max_bonds) {
1da177e4
LT		 4206 /* miimon and arp_interval not set, we need one so things
4207 * work as expected, see bonding.txt for details
4208 */
90194264 4209 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		 4210 }
4211
4212 if (primary && !USES_PRIMARY(bond_mode)) {
4213 /* currently, using a primary only makes sense
4214 * in active backup, TLB or ALB modes
4215 */
91565ebb
JP		 4216 pr_warn("Warning: %s primary device specified but has no effect in %s mode\n",
4217 primary, bond_mode_name(bond_mode));
1da177e4
LT		 4218 primary = NULL;
4219 }
4220
a549952a 4221 if (primary && primary_reselect) {
388d3a6d
NA		 4222 bond_opt_initstr(&newval, primary_reselect);
4223 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT),
4224 &newval);
4225 if (!valptr) {
a4aee5c8 4226 pr_err("Error: Invalid primary_reselect \"%s\"\n",
388d3a6d 4227 primary_reselect);
a549952a
JP		 4228 return -EINVAL;
4229 }
388d3a6d 4230 primary_reselect_value = valptr->value;
a549952a
JP		 4231 } else {
4232 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
4233 }
4234
3915c1e8 4235 if (fail_over_mac) {
1df6b6aa
NA		 4236 bond_opt_initstr(&newval, fail_over_mac);
4237 valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC),
4238 &newval);
4239 if (!valptr) {
a4aee5c8 4240 pr_err("Error: invalid fail_over_mac \"%s\"\n",
1df6b6aa 4241 fail_over_mac);
3915c1e8
JV		 4242 return -EINVAL;
4243 }
1df6b6aa 4244 fail_over_mac_value = valptr->value;
3915c1e8 4245 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
91565ebb 4246 pr_warn("Warning: fail_over_mac only affects active-backup mode\n");
3915c1e8
JV		 4247 } else {
4248 fail_over_mac_value = BOND_FOM_NONE;
4249 }
dd957c57 4250
3a7129e5 4251 if (lp_interval == 0) {
91565ebb
JP		 4252 pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n",
4253 INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL);
3a7129e5 4254 lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL;
4255 }
4256
1da177e4
LT		 4257 /* fill params struct with the proper values */
4258 params->mode = bond_mode;
169a3e66 4259 params->xmit_policy = xmit_hashtype;
1da177e4 4260 params->miimon = miimon;
ad246c99 4261 params->num_peer_notif = num_peer_notif;
1da177e4 4262 params->arp_interval = arp_interval;
f5b2b966 4263 params->arp_validate = arp_validate_value;
8599b52e 4264 params->arp_all_targets = arp_all_targets_value;
1da177e4
LT		 4265 params->updelay = updelay;
4266 params->downdelay = downdelay;
4267 params->use_carrier = use_carrier;
4268 params->lacp_fast = lacp_fast;
4269 params->primary[0] = 0;
a549952a 4270 params->primary_reselect = primary_reselect_value;
3915c1e8 4271 params->fail_over_mac = fail_over_mac_value;
bb1d9123 4272 params->tx_queues = tx_queues;
ebd8e497 4273 params->all_slaves_active = all_slaves_active;
c2952c31 4274 params->resend_igmp = resend_igmp;
655f8919 4275 params->min_links = min_links;
3a7129e5 4276 params->lp_interval = lp_interval;
809fa972
HFS		 4277 params->packets_per_slave = packets_per_slave;
4278 if (packets_per_slave > 0) {
4279 params->reciprocal_packets_per_slave =
4280 reciprocal_value(packets_per_slave);
4281 } else {
4282 /* reciprocal_packets_per_slave is unused if
4283 * packets_per_slave is 0 or 1, just initialize it
4284 */
4285 params->reciprocal_packets_per_slave =
4286 (struct reciprocal_value) { 0 };
4287 }
4288
1da177e4
LT		 4289 if (primary) {
4290 strncpy(params->primary, primary, IFNAMSIZ);
4291 params->primary[IFNAMSIZ - 1] = 0;
4292 }
4293
4294 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
4295
4296 return 0;
4297}
4298
0daa2303 4299static struct lock_class_key bonding_netdev_xmit_lock_key;
cf508b12 4300static struct lock_class_key bonding_netdev_addr_lock_key;
49ee4920 4301static struct lock_class_key bonding_tx_busylock_key;
0daa2303 4302
e8a0464c
DM		 4303static void bond_set_lockdep_class_one(struct net_device *dev,
4304 struct netdev_queue *txq,
4305 void *_unused)
c773e847
DM		 4306{
4307 lockdep_set_class(&txq->_xmit_lock,
4308 &bonding_netdev_xmit_lock_key);
4309}
4310
4311static void bond_set_lockdep_class(struct net_device *dev)
4312{
cf508b12
DM		 4313 lockdep_set_class(&dev->addr_list_lock,
4314 &bonding_netdev_addr_lock_key);
e8a0464c 4315 netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL);
49ee4920 4316 dev->qdisc_tx_busylock = &bonding_tx_busylock_key;
c773e847
DM		 4317}
4318
181470fc
SH		 4319/*
4320 * Called from registration process
4321 */
4322static int bond_init(struct net_device *bond_dev)
4323{
4324 struct bonding *bond = netdev_priv(bond_dev);
ec87fd3b 4325 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
9fe0617d 4326 struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond));
181470fc
SH		 4327
4328 pr_debug("Begin bond_init for %s\n", bond_dev->name);
4329
9fe0617d
NH		 4330 /*
4331 * Initialize locks that may be required during
4332 * en/deslave operations. All of the bond_open work
4333 * (of which this is part) should really be moved to
4334 * a phase prior to dev_open
4335 */
4336 spin_lock_init(&(bond_info->tx_hashtbl_lock));
4337 spin_lock_init(&(bond_info->rx_hashtbl_lock));
4338
181470fc
SH		 4339 bond->wq = create_singlethread_workqueue(bond_dev->name);
4340 if (!bond->wq)
4341 return -ENOMEM;
4342
4343 bond_set_lockdep_class(bond_dev);
4344
ec87fd3b 4345 list_add_tail(&bond->bond_list, &bn->dev_list);
181470fc 4346
6151b3d4 4347 bond_prepare_sysfs_group(bond);
22bedad3 4348
f073c7ca
TI		 4349 bond_debug_register(bond);
4350
409cc1f8
JP		 4351 /* Ensure valid dev_addr */
4352 if (is_zero_ether_addr(bond_dev->dev_addr) &&
97a1e639 4353 bond_dev->addr_assign_type == NET_ADDR_PERM)
409cc1f8 4354 eth_hw_addr_random(bond_dev);
409cc1f8 4355
181470fc
SH		 4356 return 0;
4357}
4358
0a2a78c4 4359unsigned int bond_get_num_tx_queues(void)
d5da4510 4360{
efacb309 4361 return tx_queues;
d5da4510
JP		 4362}
4363
dfe60397 4364/* Create a new bond based on the specified name and bonding parameters.
e4b91c48 4365 * If name is NULL, obtain a suitable "bond%d" name for us.
dfe60397
MW		 4366 * Caller must NOT hold rtnl_lock; we need to release it here before we
4367 * set up our sysfs entries.
4368 */
ec87fd3b 4369int bond_create(struct net *net, const char *name)
dfe60397
MW		 4370{
4371 struct net_device *bond_dev;
4372 int res;
4373
4374 rtnl_lock();
027ea041 4375
1c5cae81
JP		 4376 bond_dev = alloc_netdev_mq(sizeof(struct bonding),
4377 name ? name : "bond%d",
4378 bond_setup, tx_queues);
dfe60397 4379 if (!bond_dev) {
a4aee5c8 4380 pr_err("%s: eek! can't alloc netdev!\n", name);
9e2e61fb
AW		 4381 rtnl_unlock();
4382 return -ENOMEM;
dfe60397
MW		 4383 }
4384
ec87fd3b 4385 dev_net_set(bond_dev, net);
88ead977
EB		 4386 bond_dev->rtnl_link_ops = &bond_link_ops;
4387
dfe60397 4388 res = register_netdevice(bond_dev);
0daa2303 4389
e826eafa
PO		 4390 netif_carrier_off(bond_dev);
4391
7e083840 4392 rtnl_unlock();
9e2e61fb
AW		 4393 if (res < 0)
4394 bond_destructor(bond_dev);
30c15ba9 4395 return res;
dfe60397
MW		 4396}
4397
2c8c1e72 4398static int __net_init bond_net_init(struct net *net)
ec87fd3b 4399{
15449745 4400 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 4401
4402 bn->net = net;
4403 INIT_LIST_HEAD(&bn->dev_list);
4404
ec87fd3b 4405 bond_create_proc_dir(bn);
4c22400a 4406 bond_create_sysfs(bn);
87a7b84b 4407
15449745 4408 return 0;
ec87fd3b
EB		 4409}
4410
2c8c1e72 4411static void __net_exit bond_net_exit(struct net *net)
ec87fd3b 4412{
15449745 4413 struct bond_net *bn = net_generic(net, bond_net_id);
69b0216a 4414 struct bonding *bond, *tmp_bond;
4415 LIST_HEAD(list);
ec87fd3b 4416
4c22400a 4417 bond_destroy_sysfs(bn);
ec87fd3b 4418 bond_destroy_proc_dir(bn);
69b0216a 4419
4420 /* Kill off any bonds created after unregistering bond rtnl ops */
4421 rtnl_lock();
4422 list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list)
4423 unregister_netdevice_queue(bond->dev, &list);
4424 unregister_netdevice_many(&list);
4425 rtnl_unlock();
ec87fd3b
EB		 4426}
4427
4428static struct pernet_operations bond_net_ops = {
4429 .init = bond_net_init,
4430 .exit = bond_net_exit,
15449745
EB		 4431 .id = &bond_net_id,
4432 .size = sizeof(struct bond_net),
ec87fd3b
EB		 4433};
4434
1da177e4
LT		 4435static int __init bonding_init(void)
4436{
1da177e4
LT		 4437 int i;
4438 int res;
4439
bd33acc3 4440 pr_info("%s", bond_version);
1da177e4 4441
dfe60397 4442 res = bond_check_params(&bonding_defaults);
3d632c3f 4443 if (res)
dfe60397 4444 goto out;
1da177e4 4445
15449745 4446 res = register_pernet_subsys(&bond_net_ops);
ec87fd3b
EB		 4447 if (res)
4448 goto out;
027ea041 4449
0a2a78c4 4450 res = bond_netlink_init();
88ead977 4451 if (res)
6639104b 4452 goto err_link;
88ead977 4453
f073c7ca
TI		 4454 bond_create_debugfs();
4455
1da177e4 4456 for (i = 0; i < max_bonds; i++) {
ec87fd3b 4457 res = bond_create(&init_net, NULL);
dfe60397
MW		 4458 if (res)
4459 goto err;
1da177e4
LT		 4460 }
4461
1da177e4 4462 register_netdevice_notifier(&bond_netdev_notifier);
dfe60397 4463out:
1da177e4 4464 return res;
88ead977 4465err:
0a2a78c4 4466 bond_netlink_fini();
6639104b 4467err_link:
15449745 4468 unregister_pernet_subsys(&bond_net_ops);
88ead977 4469 goto out;
dfe60397 4470
1da177e4
LT		 4471}
4472
4473static void __exit bonding_exit(void)
4474{
4475 unregister_netdevice_notifier(&bond_netdev_notifier);
4476
f073c7ca 4477 bond_destroy_debugfs();
ae68c398 4478
0a2a78c4 4479 bond_netlink_fini();
ffcdedb6 4480 unregister_pernet_subsys(&bond_net_ops);
e843fa50
NH		 4481
4482#ifdef CONFIG_NET_POLL_CONTROLLER
fb4fa76a
NH		 4483 /*
4484 * Make sure we don't have an imbalance on our netpoll blocking
4485 */
4486 WARN_ON(atomic_read(&netpoll_block_tx));
e843fa50 4487#endif
1da177e4
LT		 4488}
4489
4490module_init(bonding_init);
4491module_exit(bonding_exit);
4492MODULE_LICENSE("GPL");
4493MODULE_VERSION(DRV_VERSION);
4494MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
4495MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
Linux kernel - Deliverables
This page took 1.673282 seconds and 5 git commands to generate.