projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
bonding: fix bond_inet6addr_event()
[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/system.h>
1da177e4 58#include <asm/dma.h>
3d632c3f 59#include <linux/uaccess.h>
1da177e4
LT		 60#include <linux/errno.h>
61#include <linux/netdevice.h>
f6dc31a8 62#include <linux/netpoll.h>
1da177e4 63#include <linux/inetdevice.h>
a816c7c7 64#include <linux/igmp.h>
1da177e4
LT		 65#include <linux/etherdevice.h>
66#include <linux/skbuff.h>
67#include <net/sock.h>
68#include <linux/rtnetlink.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
71#include <linux/smp.h>
72#include <linux/if_ether.h>
73#include <net/arp.h>
74#include <linux/mii.h>
75#include <linux/ethtool.h>
76#include <linux/if_vlan.h>
77#include <linux/if_bonding.h>
b63bb739 78#include <linux/jiffies.h>
c3ade5ca 79#include <net/route.h>
457c4cbc 80#include <net/net_namespace.h>
ec87fd3b 81#include <net/netns/generic.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 */
89#define BOND_LINK_MON_INTERV 0
90#define BOND_LINK_ARP_INTERV 0
91
92static int max_bonds = BOND_DEFAULT_MAX_BONDS;
bb1d9123 93static int tx_queues = BOND_DEFAULT_TX_QUEUES;
7893b249 94static int num_grat_arp = 1;
305d552a 95static int num_unsol_na = 1;
1da177e4 96static int miimon = BOND_LINK_MON_INTERV;
3d632c3f
SH		 97static int updelay;
98static int downdelay;
1da177e4 99static int use_carrier = 1;
3d632c3f
SH		 100static char *mode;
101static char *primary;
a549952a 102static char *primary_reselect;
3d632c3f
SH		 103static char *lacp_rate;
104static char *ad_select;
105static char *xmit_hash_policy;
1da177e4 106static int arp_interval = BOND_LINK_ARP_INTERV;
3d632c3f
SH		 107static char *arp_ip_target[BOND_MAX_ARP_TARGETS];
108static char *arp_validate;
109static char *fail_over_mac;
ebd8e497 110static int all_slaves_active = 0;
d2991f75 111static struct bond_params bonding_defaults;
1da177e4
LT		 112
113module_param(max_bonds, int, 0);
114MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
bb1d9123
AG		 115module_param(tx_queues, int, 0);
116MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)");
7893b249
MS		 117module_param(num_grat_arp, int, 0644);
118MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packets to send on failover event");
305d552a
BH		 119module_param(num_unsol_na, int, 0644);
120MODULE_PARM_DESC(num_unsol_na, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event");
1da177e4
LT		 121module_param(miimon, int, 0);
122MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
123module_param(updelay, int, 0);
124MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
125module_param(downdelay, int, 0);
2ac47660
MW		 126MODULE_PARM_DESC(downdelay, "Delay before considering link down, "
127 "in milliseconds");
1da177e4 128module_param(use_carrier, int, 0);
2ac47660
MW		 129MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; "
130 "0 for off, 1 for on (default)");
1da177e4 131module_param(mode, charp, 0);
2ac47660
MW		 132MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, "
133 "1 for active-backup, 2 for balance-xor, "
134 "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, "
135 "6 for balance-alb");
1da177e4
LT		 136module_param(primary, charp, 0);
137MODULE_PARM_DESC(primary, "Primary network device to use");
a549952a
JP		 138module_param(primary_reselect, charp, 0);
139MODULE_PARM_DESC(primary_reselect, "Reselect primary slave "
140 "once it comes up; "
141 "0 for always (default), "
142 "1 for only if speed of primary is "
143 "better, "
144 "2 for only on active slave "
145 "failure");
1da177e4 146module_param(lacp_rate, charp, 0);
2ac47660
MW		 147MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner "
148 "(slow/fast)");
fd989c83
JV		 149module_param(ad_select, charp, 0);
150MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)");
169a3e66 151module_param(xmit_hash_policy, charp, 0);
2ac47660
MW		 152MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)"
153 ", 1 for layer 3+4");
1da177e4
LT		 154module_param(arp_interval, int, 0);
155MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
156module_param_array(arp_ip_target, charp, NULL, 0);
157MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
f5b2b966
JV		 158module_param(arp_validate, charp, 0);
159MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all");
3915c1e8
JV		 160module_param(fail_over_mac, charp, 0);
161MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow");
ebd8e497
AG		 162module_param(all_slaves_active, int, 0);
163MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface"
164 "by setting active flag for all slaves. "
165 "0 for never (default), 1 for always.");
1da177e4
LT		 166
167/*----------------------------- Global variables ----------------------------*/
168
f71e1309 169static const char * const version =
1da177e4
LT		 170 DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n";
171
f99189b1 172int bond_net_id __read_mostly;
1da177e4 173
3d632c3f
SH		 174static __be32 arp_target[BOND_MAX_ARP_TARGETS];
175static int arp_ip_count;
1da177e4 176static int bond_mode = BOND_MODE_ROUNDROBIN;
3d632c3f
SH		 177static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2;
178static int lacp_fast;
c22d7ac8 179static int disable_netpoll = 1;
1da177e4 180
e97fd7c6 181const struct bond_parm_tbl bond_lacp_tbl[] = {
1da177e4
LT		 182{ "slow", AD_LACP_SLOW},
183{ "fast", AD_LACP_FAST},
184{ NULL, -1},
185};
186
e97fd7c6 187const struct bond_parm_tbl bond_mode_tbl[] = {
1da177e4
LT		 188{ "balance-rr", BOND_MODE_ROUNDROBIN},
189{ "active-backup", BOND_MODE_ACTIVEBACKUP},
190{ "balance-xor", BOND_MODE_XOR},
191{ "broadcast", BOND_MODE_BROADCAST},
192{ "802.3ad", BOND_MODE_8023AD},
193{ "balance-tlb", BOND_MODE_TLB},
194{ "balance-alb", BOND_MODE_ALB},
195{ NULL, -1},
196};
197
e97fd7c6 198const struct bond_parm_tbl xmit_hashtype_tbl[] = {
169a3e66
JV		 199{ "layer2", BOND_XMIT_POLICY_LAYER2},
200{ "layer3+4", BOND_XMIT_POLICY_LAYER34},
6f6652be 201{ "layer2+3", BOND_XMIT_POLICY_LAYER23},
169a3e66
JV		 202{ NULL, -1},
203};
204
e97fd7c6 205const struct bond_parm_tbl arp_validate_tbl[] = {
f5b2b966
JV		 206{ "none", BOND_ARP_VALIDATE_NONE},
207{ "active", BOND_ARP_VALIDATE_ACTIVE},
208{ "backup", BOND_ARP_VALIDATE_BACKUP},
209{ "all", BOND_ARP_VALIDATE_ALL},
210{ NULL, -1},
211};
212
e97fd7c6 213const struct bond_parm_tbl fail_over_mac_tbl[] = {
3915c1e8
JV		 214{ "none", BOND_FOM_NONE},
215{ "active", BOND_FOM_ACTIVE},
216{ "follow", BOND_FOM_FOLLOW},
217{ NULL, -1},
218};
219
a549952a
JP		 220const struct bond_parm_tbl pri_reselect_tbl[] = {
221{ "always", BOND_PRI_RESELECT_ALWAYS},
222{ "better", BOND_PRI_RESELECT_BETTER},
223{ "failure", BOND_PRI_RESELECT_FAILURE},
224{ NULL, -1},
225};
226
fd989c83
JV		 227struct bond_parm_tbl ad_select_tbl[] = {
228{ "stable", BOND_AD_STABLE},
229{ "bandwidth", BOND_AD_BANDWIDTH},
230{ "count", BOND_AD_COUNT},
231{ NULL, -1},
232};
233
1da177e4
LT		 234/*-------------------------- Forward declarations ---------------------------*/
235
c3ade5ca 236static void bond_send_gratuitous_arp(struct bonding *bond);
181470fc 237static int bond_init(struct net_device *bond_dev);
c67dfb29 238static void bond_uninit(struct net_device *bond_dev);
1da177e4
LT		 239
240/*---------------------------- General routines -----------------------------*/
241
4ad072c9 242static const char *bond_mode_name(int mode)
1da177e4 243{
77afc92b
HE		 244 static const char *names[] = {
245 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
246 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
247 [BOND_MODE_XOR] = "load balancing (xor)",
248 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
3d632c3f 249 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
77afc92b
HE		 250 [BOND_MODE_TLB] = "transmit load balancing",
251 [BOND_MODE_ALB] = "adaptive load balancing",
252 };
253
254 if (mode < 0 || mode > BOND_MODE_ALB)
1da177e4 255 return "unknown";
77afc92b
HE		 256
257 return names[mode];
1da177e4
LT		 258}
259
260/*---------------------------------- VLAN -----------------------------------*/
261
262/**
263 * bond_add_vlan - add a new vlan id on bond
264 * @bond: bond that got the notification
265 * @vlan_id: the vlan id to add
266 *
267 * Returns -ENOMEM if allocation failed.
268 */
269static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id)
270{
271 struct vlan_entry *vlan;
272
5a03cdb7 273 pr_debug("bond: %s, vlan id %d\n",
a4aee5c8 274 (bond ? bond->dev->name : "None"), vlan_id);
1da177e4 275
305d552a 276 vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL);
3d632c3f 277 if (!vlan)
1da177e4 278 return -ENOMEM;
1da177e4
LT		 279
280 INIT_LIST_HEAD(&vlan->vlan_list);
281 vlan->vlan_id = vlan_id;
282
283 write_lock_bh(&bond->lock);
284
285 list_add_tail(&vlan->vlan_list, &bond->vlan_list);
286
287 write_unlock_bh(&bond->lock);
288
5a03cdb7 289 pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name);
1da177e4
LT		 290
291 return 0;
292}
293
294/**
295 * bond_del_vlan - delete a vlan id from bond
296 * @bond: bond that got the notification
297 * @vlan_id: the vlan id to delete
298 *
299 * returns -ENODEV if @vlan_id was not found in @bond.
300 */
301static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id)
302{
0883beca 303 struct vlan_entry *vlan;
1da177e4
LT		 304 int res = -ENODEV;
305
5a03cdb7 306 pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id);
1da177e4
LT		 307
308 write_lock_bh(&bond->lock);
309
0883beca 310 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
1da177e4
LT		 311 if (vlan->vlan_id == vlan_id) {
312 list_del(&vlan->vlan_list);
313
58402054 314 if (bond_is_lb(bond))
1da177e4 315 bond_alb_clear_vlan(bond, vlan_id);
1da177e4 316
a4aee5c8
JP		 317 pr_debug("removed VLAN ID %d from bond %s\n",
318 vlan_id, bond->dev->name);
1da177e4
LT		 319
320 kfree(vlan);
321
322 if (list_empty(&bond->vlan_list) &&
323 (bond->slave_cnt == 0)) {
324 /* Last VLAN removed and no slaves, so
325 * restore block on adding VLANs. This will
326 * be removed once new slaves that are not
327 * VLAN challenged will be added.
328 */
329 bond->dev->features |= NETIF_F_VLAN_CHALLENGED;
330 }
331
332 res = 0;
333 goto out;
334 }
335 }
336
a4aee5c8
JP		 337 pr_debug("couldn't find VLAN ID %d in bond %s\n",
338 vlan_id, bond->dev->name);
1da177e4
LT		 339
340out:
341 write_unlock_bh(&bond->lock);
342 return res;
343}
344
345/**
346 * bond_has_challenged_slaves
347 * @bond: the bond we're working on
348 *
349 * Searches the slave list. Returns 1 if a vlan challenged slave
350 * was found, 0 otherwise.
351 *
352 * Assumes bond->lock is held.
353 */
354static int bond_has_challenged_slaves(struct bonding *bond)
355{
356 struct slave *slave;
357 int i;
358
359 bond_for_each_slave(bond, slave, i) {
360 if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 361 pr_debug("found VLAN challenged slave - %s\n",
a4aee5c8 362 slave->dev->name);
1da177e4
LT		 363 return 1;
364 }
365 }
366
5a03cdb7 367 pr_debug("no VLAN challenged slaves found\n");
1da177e4
LT		 368 return 0;
369}
370
371/**
372 * bond_next_vlan - safely skip to the next item in the vlans list.
373 * @bond: the bond we're working on
374 * @curr: item we're advancing from
375 *
376 * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL,
377 * or @curr->next otherwise (even if it is @curr itself again).
3d632c3f 378 *
1da177e4
LT		 379 * Caller must hold bond->lock
380 */
381struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr)
382{
383 struct vlan_entry *next, *last;
384
3d632c3f 385 if (list_empty(&bond->vlan_list))
1da177e4 386 return NULL;
1da177e4
LT		 387
388 if (!curr) {
389 next = list_entry(bond->vlan_list.next,
390 struct vlan_entry, vlan_list);
391 } else {
392 last = list_entry(bond->vlan_list.prev,
393 struct vlan_entry, vlan_list);
394 if (last == curr) {
395 next = list_entry(bond->vlan_list.next,
396 struct vlan_entry, vlan_list);
397 } else {
398 next = list_entry(curr->vlan_list.next,
399 struct vlan_entry, vlan_list);
400 }
401 }
402
403 return next;
404}
405
406/**
407 * bond_dev_queue_xmit - Prepare skb for xmit.
3d632c3f 408 *
1da177e4
LT		 409 * @bond: bond device that got this skb for tx.
410 * @skb: hw accel VLAN tagged skb to transmit
411 * @slave_dev: slave that is supposed to xmit this skbuff
3d632c3f 412 *
1da177e4
LT		 413 * When the bond gets an skb to transmit that is
414 * already hardware accelerated VLAN tagged, and it
415 * needs to relay this skb to a slave that is not
416 * hw accel capable, the skb needs to be "unaccelerated",
417 * i.e. strip the hwaccel tag and re-insert it as part
418 * of the payload.
419 */
3d632c3f
SH		 420int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
421 struct net_device *slave_dev)
1da177e4 422{
966bc6f4 423 unsigned short uninitialized_var(vlan_id);
1da177e4
LT		 424
425 if (!list_empty(&bond->vlan_list) &&
426 !(slave_dev->features & NETIF_F_HW_VLAN_TX) &&
427 vlan_get_tag(skb, &vlan_id) == 0) {
428 skb->dev = slave_dev;
429 skb = vlan_put_tag(skb, vlan_id);
430 if (!skb) {
431 /* vlan_put_tag() frees the skb in case of error,
432 * so return success here so the calling functions
433 * won't attempt to free is again.
434 */
435 return 0;
436 }
437 } else {
438 skb->dev = slave_dev;
439 }
440
441 skb->priority = 1;
f6dc31a8
WC		 442#ifdef CONFIG_NET_POLL_CONTROLLER
443 if (unlikely(bond->dev->priv_flags & IFF_IN_NETPOLL)) {
444 struct netpoll *np = bond->dev->npinfo->netpoll;
445 slave_dev->npinfo = bond->dev->npinfo;
446 np->real_dev = np->dev = skb->dev;
447 slave_dev->priv_flags |= IFF_IN_NETPOLL;
448 netpoll_send_skb(np, skb);
449 slave_dev->priv_flags &= ~IFF_IN_NETPOLL;
450 np->dev = bond->dev;
451 } else
452#endif
453 dev_queue_xmit(skb);
1da177e4
LT		 454
455 return 0;
456}
457
458/*
459 * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid
460 * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a
461 * lock because:
462 * a. This operation is performed in IOCTL context,
463 * b. The operation is protected by the RTNL semaphore in the 8021q code,
464 * c. Holding a lock with BH disabled while directly calling a base driver
465 * entry point is generally a BAD idea.
3d632c3f 466 *
1da177e4
LT		 467 * The design of synchronization/protection for this operation in the 8021q
468 * module is good for one or more VLAN devices over a single physical device
469 * and cannot be extended for a teaming solution like bonding, so there is a
470 * potential race condition here where a net device from the vlan group might
471 * be referenced (either by a base driver or the 8021q code) while it is being
472 * removed from the system. However, it turns out we're not making matters
473 * worse, and if it works for regular VLAN usage it will work here too.
474*/
475
476/**
477 * bond_vlan_rx_register - Propagates registration to slaves
478 * @bond_dev: bonding net device that got called
479 * @grp: vlan group being registered
480 */
3d632c3f
SH		 481static void bond_vlan_rx_register(struct net_device *bond_dev,
482 struct vlan_group *grp)
1da177e4 483{
454d7c9b 484 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 485 struct slave *slave;
486 int i;
487
488 bond->vlgrp = grp;
489
490 bond_for_each_slave(bond, slave, i) {
491 struct net_device *slave_dev = slave->dev;
eb7cc59a 492 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 493
494 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
eb7cc59a
SH		 495 slave_ops->ndo_vlan_rx_register) {
496 slave_ops->ndo_vlan_rx_register(slave_dev, grp);
1da177e4
LT		 497 }
498 }
499}
500
501/**
502 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
503 * @bond_dev: bonding net device that got called
504 * @vid: vlan id being added
505 */
506static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid)
507{
454d7c9b 508 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 509 struct slave *slave;
510 int i, res;
511
512 bond_for_each_slave(bond, slave, i) {
513 struct net_device *slave_dev = slave->dev;
eb7cc59a 514 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 515
516 if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
eb7cc59a
SH		 517 slave_ops->ndo_vlan_rx_add_vid) {
518 slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid);
1da177e4
LT		 519 }
520 }
521
522 res = bond_add_vlan(bond, vid);
523 if (res) {
a4aee5c8 524 pr_err("%s: Error: Failed to add vlan id %d\n",
1da177e4
LT		 525 bond_dev->name, vid);
526 }
527}
528
529/**
530 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
531 * @bond_dev: bonding net device that got called
532 * @vid: vlan id being removed
533 */
534static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid)
535{
454d7c9b 536 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 537 struct slave *slave;
538 struct net_device *vlan_dev;
539 int i, res;
540
541 bond_for_each_slave(bond, slave, i) {
542 struct net_device *slave_dev = slave->dev;
eb7cc59a 543 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 544
545 if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
eb7cc59a 546 slave_ops->ndo_vlan_rx_kill_vid) {
1da177e4
LT		 547 /* Save and then restore vlan_dev in the grp array,
548 * since the slave's driver might clear it.
549 */
5c15bdec 550 vlan_dev = vlan_group_get_device(bond->vlgrp, vid);
eb7cc59a 551 slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid);
5c15bdec 552 vlan_group_set_device(bond->vlgrp, vid, vlan_dev);
1da177e4
LT		 553 }
554 }
555
556 res = bond_del_vlan(bond, vid);
557 if (res) {
a4aee5c8 558 pr_err("%s: Error: Failed to remove vlan id %d\n",
1da177e4
LT		 559 bond_dev->name, vid);
560 }
561}
562
563static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev)
564{
565 struct vlan_entry *vlan;
eb7cc59a 566 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 567
568 write_lock_bh(&bond->lock);
569
eb7cc59a 570 if (list_empty(&bond->vlan_list))
1da177e4 571 goto out;
1da177e4
LT		 572
573 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
eb7cc59a
SH		 574 slave_ops->ndo_vlan_rx_register)
575 slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp);
1da177e4
LT		 576
577 if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
eb7cc59a 578 !(slave_ops->ndo_vlan_rx_add_vid))
1da177e4 579 goto out;
1da177e4 580
eb7cc59a
SH		 581 list_for_each_entry(vlan, &bond->vlan_list, vlan_list)
582 slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id);
1da177e4
LT		 583
584out:
585 write_unlock_bh(&bond->lock);
586}
587
3d632c3f
SH		 588static void bond_del_vlans_from_slave(struct bonding *bond,
589 struct net_device *slave_dev)
1da177e4 590{
eb7cc59a 591 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 592 struct vlan_entry *vlan;
593 struct net_device *vlan_dev;
594
595 write_lock_bh(&bond->lock);
596
eb7cc59a 597 if (list_empty(&bond->vlan_list))
1da177e4 598 goto out;
1da177e4
LT		 599
600 if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
eb7cc59a 601 !(slave_ops->ndo_vlan_rx_kill_vid))
1da177e4 602 goto unreg;
1da177e4
LT		 603
604 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
605 /* Save and then restore vlan_dev in the grp array,
606 * since the slave's driver might clear it.
607 */
5c15bdec 608 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
eb7cc59a 609 slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id);
5c15bdec 610 vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev);
1da177e4
LT		 611 }
612
613unreg:
614 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
eb7cc59a
SH		 615 slave_ops->ndo_vlan_rx_register)
616 slave_ops->ndo_vlan_rx_register(slave_dev, NULL);
1da177e4
LT		 617
618out:
619 write_unlock_bh(&bond->lock);
620}
621
622/*------------------------------- Link status -------------------------------*/
623
ff59c456
JV		 624/*
625 * Set the carrier state for the master according to the state of its
626 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
627 * do special 802.3ad magic.
628 *
629 * Returns zero if carrier state does not change, nonzero if it does.
630 */
631static int bond_set_carrier(struct bonding *bond)
632{
633 struct slave *slave;
634 int i;
635
636 if (bond->slave_cnt == 0)
637 goto down;
638
639 if (bond->params.mode == BOND_MODE_8023AD)
640 return bond_3ad_set_carrier(bond);
641
642 bond_for_each_slave(bond, slave, i) {
643 if (slave->link == BOND_LINK_UP) {
644 if (!netif_carrier_ok(bond->dev)) {
645 netif_carrier_on(bond->dev);
646 return 1;
647 }
648 return 0;
649 }
650 }
651
652down:
653 if (netif_carrier_ok(bond->dev)) {
654 netif_carrier_off(bond->dev);
655 return 1;
656 }
657 return 0;
658}
659
1da177e4
LT		 660/*
661 * Get link speed and duplex from the slave's base driver
662 * using ethtool. If for some reason the call fails or the
663 * values are invalid, fake speed and duplex to 100/Full
664 * and return error.
665 */
666static int bond_update_speed_duplex(struct slave *slave)
667{
668 struct net_device *slave_dev = slave->dev;
1da177e4 669 struct ethtool_cmd etool;
61a44b9c 670 int res;
1da177e4
LT		 671
672 /* Fake speed and duplex */
673 slave->speed = SPEED_100;
674 slave->duplex = DUPLEX_FULL;
675
61a44b9c
MW		 676 if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings)
677 return -1;
1da177e4 678
61a44b9c
MW		 679 res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool);
680 if (res < 0)
1da177e4 681 return -1;
1da177e4 682
1da177e4
LT		 683 switch (etool.speed) {
684 case SPEED_10:
685 case SPEED_100:
686 case SPEED_1000:
94dbffd5 687 case SPEED_10000:
1da177e4
LT		 688 break;
689 default:
690 return -1;
691 }
692
693 switch (etool.duplex) {
694 case DUPLEX_FULL:
695 case DUPLEX_HALF:
696 break;
697 default:
698 return -1;
699 }
700
701 slave->speed = etool.speed;
702 slave->duplex = etool.duplex;
703
704 return 0;
705}
706
707/*
708 * if <dev> supports MII link status reporting, check its link status.
709 *
710 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
3d632c3f 711 * depending upon the setting of the use_carrier parameter.
1da177e4
LT		 712 *
713 * Return either BMSR_LSTATUS, meaning that the link is up (or we
714 * can't tell and just pretend it is), or 0, meaning that the link is
715 * down.
716 *
717 * If reporting is non-zero, instead of faking link up, return -1 if
718 * both ETHTOOL and MII ioctls fail (meaning the device does not
719 * support them). If use_carrier is set, return whatever it says.
720 * It'd be nice if there was a good way to tell if a driver supports
721 * netif_carrier, but there really isn't.
722 */
3d632c3f
SH		 723static int bond_check_dev_link(struct bonding *bond,
724 struct net_device *slave_dev, int reporting)
1da177e4 725{
eb7cc59a 726 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
d9d52832 727 int (*ioctl)(struct net_device *, struct ifreq *, int);
1da177e4
LT		 728 struct ifreq ifr;
729 struct mii_ioctl_data *mii;
1da177e4 730
6c988853
PG		 731 if (!reporting && !netif_running(slave_dev))
732 return 0;
733
eb7cc59a 734 if (bond->params.use_carrier)
1da177e4 735 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
1da177e4 736
29112f4e
JP		 737 /* Try to get link status using Ethtool first. */
738 if (slave_dev->ethtool_ops) {
739 if (slave_dev->ethtool_ops->get_link) {
740 u32 link;
741
742 link = slave_dev->ethtool_ops->get_link(slave_dev);
743
744 return link ? BMSR_LSTATUS : 0;
745 }
746 }
747
3d632c3f 748 /* Ethtool can't be used, fallback to MII ioctls. */
eb7cc59a 749 ioctl = slave_ops->ndo_do_ioctl;
1da177e4
LT		 750 if (ioctl) {
751 /* TODO: set pointer to correct ioctl on a per team member */
752 /* bases to make this more efficient. that is, once */
753 /* we determine the correct ioctl, we will always */
754 /* call it and not the others for that team */
755 /* member. */
756
757 /*
758 * We cannot assume that SIOCGMIIPHY will also read a
759 * register; not all network drivers (e.g., e100)
760 * support that.
761 */
762
763 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
764 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
765 mii = if_mii(&ifr);
766 if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
767 mii->reg_num = MII_BMSR;
3d632c3f
SH		 768 if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0)
769 return mii->val_out & BMSR_LSTATUS;
1da177e4
LT		 770 }
771 }
772
1da177e4
LT		 773 /*
774 * If reporting, report that either there's no dev->do_ioctl,
61a44b9c 775 * or both SIOCGMIIREG and get_link failed (meaning that we
1da177e4
LT		 776 * cannot report link status). If not reporting, pretend
777 * we're ok.
778 */
3d632c3f 779 return reporting ? -1 : BMSR_LSTATUS;
1da177e4
LT		 780}
781
782/*----------------------------- Multicast list ------------------------------*/
783
1da177e4
LT		 784/*
785 * Push the promiscuity flag down to appropriate slaves
786 */
7e1a1ac1 787static int bond_set_promiscuity(struct bonding *bond, int inc)
1da177e4 788{
7e1a1ac1 789 int err = 0;
1da177e4
LT		 790 if (USES_PRIMARY(bond->params.mode)) {
791 /* write lock already acquired */
792 if (bond->curr_active_slave) {
7e1a1ac1
WC		 793 err = dev_set_promiscuity(bond->curr_active_slave->dev,
794 inc);
1da177e4
LT		 795 }
796 } else {
797 struct slave *slave;
798 int i;
799 bond_for_each_slave(bond, slave, i) {
7e1a1ac1
WC		 800 err = dev_set_promiscuity(slave->dev, inc);
801 if (err)
802 return err;
1da177e4
LT		 803 }
804 }
7e1a1ac1 805 return err;
1da177e4
LT		 806}
807
808/*
809 * Push the allmulti flag down to all slaves
810 */
7e1a1ac1 811static int bond_set_allmulti(struct bonding *bond, int inc)
1da177e4 812{
7e1a1ac1 813 int err = 0;
1da177e4
LT		 814 if (USES_PRIMARY(bond->params.mode)) {
815 /* write lock already acquired */
816 if (bond->curr_active_slave) {
7e1a1ac1
WC		 817 err = dev_set_allmulti(bond->curr_active_slave->dev,
818 inc);
1da177e4
LT		 819 }
820 } else {
821 struct slave *slave;
822 int i;
823 bond_for_each_slave(bond, slave, i) {
7e1a1ac1
WC		 824 err = dev_set_allmulti(slave->dev, inc);
825 if (err)
826 return err;
1da177e4
LT		 827 }
828 }
7e1a1ac1 829 return err;
1da177e4
LT		 830}
831
832/*
833 * Add a Multicast address to slaves
834 * according to mode
835 */
22bedad3 836static void bond_mc_add(struct bonding *bond, void *addr)
1da177e4
LT		 837{
838 if (USES_PRIMARY(bond->params.mode)) {
839 /* write lock already acquired */
3d632c3f 840 if (bond->curr_active_slave)
22bedad3 841 dev_mc_add(bond->curr_active_slave->dev, addr);
1da177e4
LT		 842 } else {
843 struct slave *slave;
844 int i;
3d632c3f
SH		 845
846 bond_for_each_slave(bond, slave, i)
22bedad3 847 dev_mc_add(slave->dev, addr);
1da177e4
LT		 848 }
849}
850
851/*
852 * Remove a multicast address from slave
853 * according to mode
854 */
22bedad3 855static void bond_mc_del(struct bonding *bond, void *addr)
1da177e4
LT		 856{
857 if (USES_PRIMARY(bond->params.mode)) {
858 /* write lock already acquired */
3d632c3f 859 if (bond->curr_active_slave)
22bedad3 860 dev_mc_del(bond->curr_active_slave->dev, addr);
1da177e4
LT		 861 } else {
862 struct slave *slave;
863 int i;
864 bond_for_each_slave(bond, slave, i) {
22bedad3 865 dev_mc_del(slave->dev, addr);
1da177e4
LT		 866 }
867 }
868}
869
a816c7c7
JV		 870
871/*
872 * Retrieve the list of registered multicast addresses for the bonding
873 * device and retransmit an IGMP JOIN request to the current active
874 * slave.
875 */
876static void bond_resend_igmp_join_requests(struct bonding *bond)
877{
878 struct in_device *in_dev;
879 struct ip_mc_list *im;
880
881 rcu_read_lock();
882 in_dev = __in_dev_get_rcu(bond->dev);
883 if (in_dev) {
3d632c3f 884 for (im = in_dev->mc_list; im; im = im->next)
a816c7c7 885 ip_mc_rejoin_group(im);
a816c7c7
JV		 886 }
887
888 rcu_read_unlock();
889}
890
1da177e4
LT		 891/*
892 * flush all members of flush->mc_list from device dev->mc_list
893 */
3d632c3f
SH		 894static void bond_mc_list_flush(struct net_device *bond_dev,
895 struct net_device *slave_dev)
1da177e4 896{
454d7c9b 897 struct bonding *bond = netdev_priv(bond_dev);
22bedad3 898 struct netdev_hw_addr *ha;
1da177e4 899
22bedad3
JP		 900 netdev_for_each_mc_addr(ha, bond_dev)
901 dev_mc_del(slave_dev, ha->addr);
1da177e4
LT		 902
903 if (bond->params.mode == BOND_MODE_8023AD) {
904 /* del lacpdu mc addr from mc list */
905 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
906
22bedad3 907 dev_mc_del(slave_dev, lacpdu_multicast);
1da177e4
LT		 908 }
909}
910
911/*--------------------------- Active slave change ---------------------------*/
912
913/*
914 * Update the mc list and multicast-related flags for the new and
915 * old active slaves (if any) according to the multicast mode, and
916 * promiscuous flags unconditionally.
917 */
3d632c3f
SH		 918static void bond_mc_swap(struct bonding *bond, struct slave *new_active,
919 struct slave *old_active)
1da177e4 920{
22bedad3 921 struct netdev_hw_addr *ha;
1da177e4 922
3d632c3f 923 if (!USES_PRIMARY(bond->params.mode))
1da177e4
LT		 924 /* nothing to do - mc list is already up-to-date on
925 * all slaves
926 */
927 return;
1da177e4
LT		 928
929 if (old_active) {
3d632c3f 930 if (bond->dev->flags & IFF_PROMISC)
1da177e4 931 dev_set_promiscuity(old_active->dev, -1);
1da177e4 932
3d632c3f 933 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 934 dev_set_allmulti(old_active->dev, -1);
1da177e4 935
22bedad3
JP		 936 netdev_for_each_mc_addr(ha, bond->dev)
937 dev_mc_del(old_active->dev, ha->addr);
1da177e4
LT		 938 }
939
940 if (new_active) {
7e1a1ac1 941 /* FIXME: Signal errors upstream. */
3d632c3f 942 if (bond->dev->flags & IFF_PROMISC)
1da177e4 943 dev_set_promiscuity(new_active->dev, 1);
1da177e4 944
3d632c3f 945 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 946 dev_set_allmulti(new_active->dev, 1);
1da177e4 947
22bedad3
JP		 948 netdev_for_each_mc_addr(ha, bond->dev)
949 dev_mc_add(new_active->dev, ha->addr);
a816c7c7 950 bond_resend_igmp_join_requests(bond);
1da177e4
LT		 951 }
952}
953
3915c1e8
JV		 954/*
955 * bond_do_fail_over_mac
956 *
957 * Perform special MAC address swapping for fail_over_mac settings
958 *
959 * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh.
960 */
961static void bond_do_fail_over_mac(struct bonding *bond,
962 struct slave *new_active,
963 struct slave *old_active)
1f78d9f9
HE		 964 __releases(&bond->curr_slave_lock)
965 __releases(&bond->lock)
966 __acquires(&bond->lock)
967 __acquires(&bond->curr_slave_lock)
3915c1e8
JV		 968{
969 u8 tmp_mac[ETH_ALEN];
970 struct sockaddr saddr;
971 int rv;
972
973 switch (bond->params.fail_over_mac) {
974 case BOND_FOM_ACTIVE:
975 if (new_active)
976 memcpy(bond->dev->dev_addr, new_active->dev->dev_addr,
977 new_active->dev->addr_len);
978 break;
979 case BOND_FOM_FOLLOW:
980 /*
981 * if new_active && old_active, swap them
982 * if just old_active, do nothing (going to no active slave)
983 * if just new_active, set new_active to bond's MAC
984 */
985 if (!new_active)
986 return;
987
988 write_unlock_bh(&bond->curr_slave_lock);
989 read_unlock(&bond->lock);
990
991 if (old_active) {
992 memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN);
993 memcpy(saddr.sa_data, old_active->dev->dev_addr,
994 ETH_ALEN);
995 saddr.sa_family = new_active->dev->type;
996 } else {
997 memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN);
998 saddr.sa_family = bond->dev->type;
999 }
1000
1001 rv = dev_set_mac_address(new_active->dev, &saddr);
1002 if (rv) {
a4aee5c8 1003 pr_err("%s: Error %d setting MAC of slave %s\n",
3915c1e8
JV		 1004 bond->dev->name, -rv, new_active->dev->name);
1005 goto out;
1006 }
1007
1008 if (!old_active)
1009 goto out;
1010
1011 memcpy(saddr.sa_data, tmp_mac, ETH_ALEN);
1012 saddr.sa_family = old_active->dev->type;
1013
1014 rv = dev_set_mac_address(old_active->dev, &saddr);
1015 if (rv)
a4aee5c8 1016 pr_err("%s: Error %d setting MAC of slave %s\n",
3915c1e8
JV		 1017 bond->dev->name, -rv, new_active->dev->name);
1018out:
1019 read_lock(&bond->lock);
1020 write_lock_bh(&bond->curr_slave_lock);
1021 break;
1022 default:
a4aee5c8 1023 pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n",
3915c1e8
JV		 1024 bond->dev->name, bond->params.fail_over_mac);
1025 break;
1026 }
1027
1028}
1029
a549952a
JP		 1030static bool bond_should_change_active(struct bonding *bond)
1031{
1032 struct slave *prim = bond->primary_slave;
1033 struct slave *curr = bond->curr_active_slave;
1034
1035 if (!prim || !curr || curr->link != BOND_LINK_UP)
1036 return true;
1037 if (bond->force_primary) {
1038 bond->force_primary = false;
1039 return true;
1040 }
1041 if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER &&
1042 (prim->speed < curr->speed ||
1043 (prim->speed == curr->speed && prim->duplex <= curr->duplex)))
1044 return false;
1045 if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE)
1046 return false;
1047 return true;
1048}
3915c1e8 1049
1da177e4
LT		 1050/**
1051 * find_best_interface - select the best available slave to be the active one
1052 * @bond: our bonding struct
1053 *
1054 * Warning: Caller must hold curr_slave_lock for writing.
1055 */
1056static struct slave *bond_find_best_slave(struct bonding *bond)
1057{
1058 struct slave *new_active, *old_active;
1059 struct slave *bestslave = NULL;
1060 int mintime = bond->params.updelay;
1061 int i;
1062
49b4ad92 1063 new_active = bond->curr_active_slave;
1da177e4
LT		 1064
1065 if (!new_active) { /* there were no active slaves left */
3d632c3f 1066 if (bond->slave_cnt > 0) /* found one slave */
1da177e4 1067 new_active = bond->first_slave;
3d632c3f 1068 else
1da177e4 1069 return NULL; /* still no slave, return NULL */
1da177e4
LT		 1070 }
1071
1da177e4 1072 if ((bond->primary_slave) &&
a549952a
JP		 1073 bond->primary_slave->link == BOND_LINK_UP &&
1074 bond_should_change_active(bond)) {
1da177e4
LT		 1075 new_active = bond->primary_slave;
1076 }
1077
1078 /* remember where to stop iterating over the slaves */
1079 old_active = new_active;
1080
1081 bond_for_each_slave_from(bond, new_active, i, old_active) {
b9f60253
JP		 1082 if (new_active->link == BOND_LINK_UP) {
1083 return new_active;
1084 } else if (new_active->link == BOND_LINK_BACK &&
1085 IS_UP(new_active->dev)) {
1086 /* link up, but waiting for stabilization */
1087 if (new_active->delay < mintime) {
1088 mintime = new_active->delay;
1089 bestslave = new_active;
1da177e4
LT		 1090 }
1091 }
1092 }
1093
1094 return bestslave;
1095}
1096
1097/**
1098 * change_active_interface - change the active slave into the specified one
1099 * @bond: our bonding struct
1100 * @new: the new slave to make the active one
1101 *
1102 * Set the new slave to the bond's settings and unset them on the old
1103 * curr_active_slave.
1104 * Setting include flags, mc-list, promiscuity, allmulti, etc.
1105 *
1106 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
1107 * because it is apparently the best available slave we have, even though its
1108 * updelay hasn't timed out yet.
1109 *
3915c1e8
JV		 1110 * If new_active is not NULL, caller must hold bond->lock for read and
1111 * curr_slave_lock for write_bh.
1da177e4 1112 */
a77b5325 1113void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
1da177e4
LT		 1114{
1115 struct slave *old_active = bond->curr_active_slave;
1116
3d632c3f 1117 if (old_active == new_active)
1da177e4 1118 return;
1da177e4
LT		 1119
1120 if (new_active) {
b2220cad
JV		 1121 new_active->jiffies = jiffies;
1122
1da177e4
LT		 1123 if (new_active->link == BOND_LINK_BACK) {
1124 if (USES_PRIMARY(bond->params.mode)) {
a4aee5c8
JP		 1125 pr_info("%s: making interface %s the new active one %d ms earlier.\n",
1126 bond->dev->name, new_active->dev->name,
1127 (bond->params.updelay - new_active->delay) * bond->params.miimon);
1da177e4
LT		 1128 }
1129
1130 new_active->delay = 0;
1131 new_active->link = BOND_LINK_UP;
1da177e4 1132
3d632c3f 1133 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 1134 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
1da177e4 1135
58402054 1136 if (bond_is_lb(bond))
1da177e4 1137 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
1da177e4
LT		 1138 } else {
1139 if (USES_PRIMARY(bond->params.mode)) {
a4aee5c8
JP		 1140 pr_info("%s: making interface %s the new active one.\n",
1141 bond->dev->name, new_active->dev->name);
1da177e4
LT		 1142 }
1143 }
1144 }
1145
3d632c3f 1146 if (USES_PRIMARY(bond->params.mode))
1da177e4 1147 bond_mc_swap(bond, new_active, old_active);
1da177e4 1148
58402054 1149 if (bond_is_lb(bond)) {
1da177e4 1150 bond_alb_handle_active_change(bond, new_active);
8f903c70
JV		 1151 if (old_active)
1152 bond_set_slave_inactive_flags(old_active);
1153 if (new_active)
1154 bond_set_slave_active_flags(new_active);
1da177e4
LT		 1155 } else {
1156 bond->curr_active_slave = new_active;
1157 }
c3ade5ca
JV		 1158
1159 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
3d632c3f 1160 if (old_active)
c3ade5ca 1161 bond_set_slave_inactive_flags(old_active);
c3ade5ca
JV		 1162
1163 if (new_active) {
1164 bond_set_slave_active_flags(new_active);
2ab82852 1165
709f8a45
OG		 1166 if (bond->params.fail_over_mac)
1167 bond_do_fail_over_mac(bond, new_active,
1168 old_active);
3915c1e8 1169
709f8a45 1170 bond->send_grat_arp = bond->params.num_grat_arp;
b59f9f74 1171 bond_send_gratuitous_arp(bond);
01f3109d 1172
305d552a
BH		 1173 bond->send_unsol_na = bond->params.num_unsol_na;
1174 bond_send_unsolicited_na(bond);
1175
01f3109d
OG		 1176 write_unlock_bh(&bond->curr_slave_lock);
1177 read_unlock(&bond->lock);
1178
75c78500 1179 netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER);
01f3109d
OG		 1180
1181 read_lock(&bond->lock);
1182 write_lock_bh(&bond->curr_slave_lock);
7893b249 1183 }
c3ade5ca 1184 }
a2fd940f
AG		 1185
1186 /* resend IGMP joins since all were sent on curr_active_slave */
1187 if (bond->params.mode == BOND_MODE_ROUNDROBIN) {
1188 bond_resend_igmp_join_requests(bond);
1189 }
1da177e4
LT		 1190}
1191
1192/**
1193 * bond_select_active_slave - select a new active slave, if needed
1194 * @bond: our bonding struct
1195 *
3d632c3f 1196 * This functions should be called when one of the following occurs:
1da177e4
LT		 1197 * - The old curr_active_slave has been released or lost its link.
1198 * - The primary_slave has got its link back.
1199 * - A slave has got its link back and there's no old curr_active_slave.
1200 *
3915c1e8 1201 * Caller must hold bond->lock for read and curr_slave_lock for write_bh.
1da177e4 1202 */
a77b5325 1203void bond_select_active_slave(struct bonding *bond)
1da177e4
LT		 1204{
1205 struct slave *best_slave;
ff59c456 1206 int rv;
1da177e4
LT		 1207
1208 best_slave = bond_find_best_slave(bond);
1209 if (best_slave != bond->curr_active_slave) {
1210 bond_change_active_slave(bond, best_slave);
ff59c456
JV		 1211 rv = bond_set_carrier(bond);
1212 if (!rv)
1213 return;
1214
1215 if (netif_carrier_ok(bond->dev)) {
a4aee5c8
JP		 1216 pr_info("%s: first active interface up!\n",
1217 bond->dev->name);
ff59c456 1218 } else {
a4aee5c8
JP		 1219 pr_info("%s: now running without any active interface !\n",
1220 bond->dev->name);
ff59c456 1221 }
1da177e4
LT		 1222 }
1223}
1224
1225/*--------------------------- slave list handling ---------------------------*/
1226
1227/*
1228 * This function attaches the slave to the end of list.
1229 *
1230 * bond->lock held for writing by caller.
1231 */
1232static void bond_attach_slave(struct bonding *bond, struct slave *new_slave)
1233{
1234 if (bond->first_slave == NULL) { /* attaching the first slave */
1235 new_slave->next = new_slave;
1236 new_slave->prev = new_slave;
1237 bond->first_slave = new_slave;
1238 } else {
1239 new_slave->next = bond->first_slave;
1240 new_slave->prev = bond->first_slave->prev;
1241 new_slave->next->prev = new_slave;
1242 new_slave->prev->next = new_slave;
1243 }
1244
1245 bond->slave_cnt++;
1246}
1247
1248/*
1249 * This function detaches the slave from the list.
1250 * WARNING: no check is made to verify if the slave effectively
1251 * belongs to <bond>.
1252 * Nothing is freed on return, structures are just unchained.
1253 * If any slave pointer in bond was pointing to <slave>,
1254 * it should be changed by the calling function.
1255 *
1256 * bond->lock held for writing by caller.
1257 */
1258static void bond_detach_slave(struct bonding *bond, struct slave *slave)
1259{
3d632c3f 1260 if (slave->next)
1da177e4 1261 slave->next->prev = slave->prev;
1da177e4 1262
3d632c3f 1263 if (slave->prev)
1da177e4 1264 slave->prev->next = slave->next;
1da177e4
LT		 1265
1266 if (bond->first_slave == slave) { /* slave is the first slave */
1267 if (bond->slave_cnt > 1) { /* there are more slave */
1268 bond->first_slave = slave->next;
1269 } else {
1270 bond->first_slave = NULL; /* slave was the last one */
1271 }
1272 }
1273
1274 slave->next = NULL;
1275 slave->prev = NULL;
1276 bond->slave_cnt--;
1277}
1278
f6dc31a8
WC		 1279#ifdef CONFIG_NET_POLL_CONTROLLER
1280/*
1281 * You must hold read lock on bond->lock before calling this.
1282 */
1283static bool slaves_support_netpoll(struct net_device *bond_dev)
1284{
1285 struct bonding *bond = netdev_priv(bond_dev);
1286 struct slave *slave;
1287 int i = 0;
1288 bool ret = true;
1289
1290 bond_for_each_slave(bond, slave, i) {
1291 if ((slave->dev->priv_flags & IFF_DISABLE_NETPOLL) ||
1292 !slave->dev->netdev_ops->ndo_poll_controller)
1293 ret = false;
1294 }
1295 return i != 0 && ret;
1296}
1297
1298static void bond_poll_controller(struct net_device *bond_dev)
1299{
1300 struct net_device *dev = bond_dev->npinfo->netpoll->real_dev;
1301 if (dev != bond_dev)
1302 netpoll_poll_dev(dev);
1303}
1304
1305static void bond_netpoll_cleanup(struct net_device *bond_dev)
1306{
1307 struct bonding *bond = netdev_priv(bond_dev);
1308 struct slave *slave;
1309 const struct net_device_ops *ops;
1310 int i;
1311
1312 read_lock(&bond->lock);
1313 bond_dev->npinfo = NULL;
1314 bond_for_each_slave(bond, slave, i) {
1315 if (slave->dev) {
1316 ops = slave->dev->netdev_ops;
1317 if (ops->ndo_netpoll_cleanup)
1318 ops->ndo_netpoll_cleanup(slave->dev);
1319 else
1320 slave->dev->npinfo = NULL;
1321 }
1322 }
1323 read_unlock(&bond->lock);
1324}
1325
1326#else
1327
1328static void bond_netpoll_cleanup(struct net_device *bond_dev)
1329{
1330}
1331
1332#endif
1333
1da177e4
LT		 1334/*---------------------------------- IOCTL ----------------------------------*/
1335
4ad072c9
AB		 1336static int bond_sethwaddr(struct net_device *bond_dev,
1337 struct net_device *slave_dev)
1da177e4 1338{
5a03cdb7
HE		 1339 pr_debug("bond_dev=%p\n", bond_dev);
1340 pr_debug("slave_dev=%p\n", slave_dev);
1341 pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len);
1da177e4
LT		 1342 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
1343 return 0;
1344}
1345
7f353bf2
HX		 1346#define BOND_VLAN_FEATURES \
1347 (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \
1348 NETIF_F_HW_VLAN_FILTER)
8531c5ff 1349
3d632c3f 1350/*
8e3babcd 1351 * Compute the common dev->feature set available to all slaves. Some
7f353bf2
HX		 1352 * feature bits are managed elsewhere, so preserve those feature bits
1353 * on the master device.
8531c5ff
AK		 1354 */
1355static int bond_compute_features(struct bonding *bond)
1356{
8531c5ff
AK		 1357 struct slave *slave;
1358 struct net_device *bond_dev = bond->dev;
7f353bf2 1359 unsigned long features = bond_dev->features;
278339a4 1360 unsigned long vlan_features = 0;
3158bf7d
MS		 1361 unsigned short max_hard_header_len = max((u16)ETH_HLEN,
1362 bond_dev->hard_header_len);
8e3babcd 1363 int i;
8531c5ff 1364
7f353bf2 1365 features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES);
b63365a2
HX		 1366 features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM;
1367
1368 if (!bond->first_slave)
1369 goto done;
1370
1371 features &= ~NETIF_F_ONE_FOR_ALL;
7f353bf2 1372
278339a4 1373 vlan_features = bond->first_slave->dev->vlan_features;
54ef3137 1374 bond_for_each_slave(bond, slave, i) {
b63365a2
HX		 1375 features = netdev_increment_features(features,
1376 slave->dev->features,
1377 NETIF_F_ONE_FOR_ALL);
278339a4
JV		 1378 vlan_features = netdev_increment_features(vlan_features,
1379 slave->dev->vlan_features,
1380 NETIF_F_ONE_FOR_ALL);
54ef3137
JV		 1381 if (slave->dev->hard_header_len > max_hard_header_len)
1382 max_hard_header_len = slave->dev->hard_header_len;
1383 }
8531c5ff 1384
b63365a2 1385done:
7f353bf2 1386 features |= (bond_dev->features & BOND_VLAN_FEATURES);
b63365a2 1387 bond_dev->features = netdev_fix_features(features, NULL);
278339a4 1388 bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL);
54ef3137 1389 bond_dev->hard_header_len = max_hard_header_len;
8531c5ff
AK		 1390
1391 return 0;
1392}
1393
872254dd
MS		 1394static void bond_setup_by_slave(struct net_device *bond_dev,
1395 struct net_device *slave_dev)
1396{
454d7c9b 1397 struct bonding *bond = netdev_priv(bond_dev);
d90a162a 1398
00829823 1399 bond_dev->header_ops = slave_dev->header_ops;
872254dd
MS		 1400
1401 bond_dev->type = slave_dev->type;
1402 bond_dev->hard_header_len = slave_dev->hard_header_len;
1403 bond_dev->addr_len = slave_dev->addr_len;
1404
1405 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1406 slave_dev->addr_len);
d90a162a 1407 bond->setup_by_slave = 1;
872254dd
MS		 1408}
1409
1da177e4 1410/* enslave device <slave> to bond device <master> */
a77b5325 1411int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1412{
454d7c9b 1413 struct bonding *bond = netdev_priv(bond_dev);
eb7cc59a 1414 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4 1415 struct slave *new_slave = NULL;
22bedad3 1416 struct netdev_hw_addr *ha;
1da177e4
LT		 1417 struct sockaddr addr;
1418 int link_reporting;
1419 int old_features = bond_dev->features;
1420 int res = 0;
1421
552709d5 1422 if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL &&
eb7cc59a 1423 slave_ops->ndo_do_ioctl == NULL) {
a4aee5c8
JP		 1424 pr_warning("%s: Warning: no link monitoring support for %s\n",
1425 bond_dev->name, slave_dev->name);
1da177e4
LT		 1426 }
1427
1428 /* bond must be initialized by bond_open() before enslaving */
1429 if (!(bond_dev->flags & IFF_UP)) {
a4aee5c8
JP		 1430 pr_warning("%s: master_dev is not up in bond_enslave\n",
1431 bond_dev->name);
1da177e4
LT		 1432 }
1433
1434 /* already enslaved */
1435 if (slave_dev->flags & IFF_SLAVE) {
5a03cdb7 1436 pr_debug("Error, Device was already enslaved\n");
1da177e4
LT		 1437 return -EBUSY;
1438 }
1439
1440 /* vlan challenged mutual exclusion */
1441 /* no need to lock since we're protected by rtnl_lock */
1442 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 1443 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4 1444 if (!list_empty(&bond->vlan_list)) {
a4aee5c8
JP		 1445 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1446 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1447 return -EPERM;
1448 } else {
a4aee5c8
JP		 1449 pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1450 bond_dev->name, slave_dev->name,
1451 slave_dev->name, bond_dev->name);
1da177e4
LT		 1452 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
1453 }
1454 } else {
5a03cdb7 1455 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4
LT		 1456 if (bond->slave_cnt == 0) {
1457 /* First slave, and it is not VLAN challenged,
1458 * so remove the block of adding VLANs over the bond.
1459 */
1460 bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
1461 }
1462 }
1463
217df670
JV		 1464 /*
1465 * Old ifenslave binaries are no longer supported. These can
3d632c3f 1466 * be identified with moderate accuracy by the state of the slave:
217df670
JV		 1467 * the current ifenslave will set the interface down prior to
1468 * enslaving it; the old ifenslave will not.
1469 */
1470 if ((slave_dev->flags & IFF_UP)) {
a4aee5c8 1471 pr_err("%s is up. This may be due to an out of date ifenslave.\n",
217df670
JV		 1472 slave_dev->name);
1473 res = -EPERM;
1474 goto err_undo_flags;
1475 }
1da177e4 1476
872254dd
MS		 1477 /* set bonding device ether type by slave - bonding netdevices are
1478 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1479 * there is a need to override some of the type dependent attribs/funcs.
1480 *
1481 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1482 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1483 */
1484 if (bond->slave_cnt == 0) {
e36b9d16 1485 if (bond_dev->type != slave_dev->type) {
e36b9d16 1486 pr_debug("%s: change device type from %d to %d\n",
a4aee5c8
JP		 1487 bond_dev->name,
1488 bond_dev->type, slave_dev->type);
75c78500 1489
3ca5b404
JP		 1490 res = netdev_bonding_change(bond_dev,
1491 NETDEV_PRE_TYPE_CHANGE);
1492 res = notifier_to_errno(res);
1493 if (res) {
1494 pr_err("%s: refused to change device type\n",
1495 bond_dev->name);
1496 res = -EBUSY;
1497 goto err_undo_flags;
1498 }
75c78500 1499
32a806c1 1500 /* Flush unicast and multicast addresses */
a748ee24 1501 dev_uc_flush(bond_dev);
22bedad3 1502 dev_mc_flush(bond_dev);
32a806c1 1503
e36b9d16
MS		 1504 if (slave_dev->type != ARPHRD_ETHER)
1505 bond_setup_by_slave(bond_dev, slave_dev);
1506 else
1507 ether_setup(bond_dev);
75c78500 1508
93d9b7d7
JP		 1509 netdev_bonding_change(bond_dev,
1510 NETDEV_POST_TYPE_CHANGE);
e36b9d16 1511 }
872254dd 1512 } else if (bond_dev->type != slave_dev->type) {
a4aee5c8
JP		 1513 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n",
1514 slave_dev->name,
1515 slave_dev->type, bond_dev->type);
1516 res = -EINVAL;
1517 goto err_undo_flags;
872254dd
MS		 1518 }
1519
eb7cc59a 1520 if (slave_ops->ndo_set_mac_address == NULL) {
2ab82852 1521 if (bond->slave_cnt == 0) {
a4aee5c8
JP		 1522 pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.",
1523 bond_dev->name);
3915c1e8
JV		 1524 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
1525 } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
a4aee5c8
JP		 1526 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",
1527 bond_dev->name);
2ab82852
MS		 1528 res = -EOPNOTSUPP;
1529 goto err_undo_flags;
1530 }
1da177e4
LT		 1531 }
1532
c20811a7
JP		 1533 /* If this is the first slave, then we need to set the master's hardware
1534 * address to be the same as the slave's. */
1535 if (bond->slave_cnt == 0)
1536 memcpy(bond->dev->dev_addr, slave_dev->dev_addr,
1537 slave_dev->addr_len);
1538
1539
243cb4e5 1540 new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL);
1da177e4
LT		 1541 if (!new_slave) {
1542 res = -ENOMEM;
1543 goto err_undo_flags;
1544 }
1545
bb1d9123
AG		 1546 /*
1547 * Set the new_slave's queue_id to be zero. Queue ID mapping
1548 * is set via sysfs or module option if desired.
1549 */
1550 new_slave->queue_id = 0;
1551
b15ba0fb
JP		 1552 /* Save slave's original mtu and then set it to match the bond */
1553 new_slave->original_mtu = slave_dev->mtu;
1554 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1555 if (res) {
1556 pr_debug("Error %d calling dev_set_mtu\n", res);
1557 goto err_free;
1558 }
1559
217df670
JV		 1560 /*
1561 * Save slave's original ("permanent") mac address for modes
1562 * that need it, and for restoring it upon release, and then
1563 * set it to the master's address
1564 */
1565 memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN);
1da177e4 1566
dd957c57 1567 if (!bond->params.fail_over_mac) {
2ab82852
MS		 1568 /*
1569 * Set slave to master's mac address. The application already
1570 * set the master's mac address to that of the first slave
1571 */
1572 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1573 addr.sa_family = slave_dev->type;
1574 res = dev_set_mac_address(slave_dev, &addr);
1575 if (res) {
5a03cdb7 1576 pr_debug("Error %d calling set_mac_address\n", res);
b15ba0fb 1577 goto err_restore_mtu;
2ab82852 1578 }
217df670 1579 }
1da177e4 1580
c2edacf8
JV		 1581 res = netdev_set_master(slave_dev, bond_dev);
1582 if (res) {
5a03cdb7 1583 pr_debug("Error %d calling netdev_set_master\n", res);
569f0c4d 1584 goto err_restore_mac;
c2edacf8 1585 }
217df670
JV		 1586 /* open the slave since the application closed it */
1587 res = dev_open(slave_dev);
1588 if (res) {
3d632c3f 1589 pr_debug("Opening slave %s failed\n", slave_dev->name);
569f0c4d 1590 goto err_unset_master;
1da177e4
LT		 1591 }
1592
1da177e4 1593 new_slave->dev = slave_dev;
0b680e75 1594 slave_dev->priv_flags |= IFF_BONDING;
1da177e4 1595
58402054 1596 if (bond_is_lb(bond)) {
1da177e4
LT		 1597 /* bond_alb_init_slave() must be called before all other stages since
1598 * it might fail and we do not want to have to undo everything
1599 */
1600 res = bond_alb_init_slave(bond, new_slave);
3d632c3f 1601 if (res)
569f0c4d 1602 goto err_close;
1da177e4
LT		 1603 }
1604
1605 /* If the mode USES_PRIMARY, then the new slave gets the
1606 * master's promisc (and mc) settings only if it becomes the
1607 * curr_active_slave, and that is taken care of later when calling
1608 * bond_change_active()
1609 */
1610 if (!USES_PRIMARY(bond->params.mode)) {
1611 /* set promiscuity level to new slave */
1612 if (bond_dev->flags & IFF_PROMISC) {
7e1a1ac1
WC		 1613 res = dev_set_promiscuity(slave_dev, 1);
1614 if (res)
1615 goto err_close;
1da177e4
LT		 1616 }
1617
1618 /* set allmulti level to new slave */
1619 if (bond_dev->flags & IFF_ALLMULTI) {
7e1a1ac1
WC		 1620 res = dev_set_allmulti(slave_dev, 1);
1621 if (res)
1622 goto err_close;
1da177e4
LT		 1623 }
1624
b9e40857 1625 netif_addr_lock_bh(bond_dev);
1da177e4 1626 /* upload master's mc_list to new slave */
22bedad3
JP		 1627 netdev_for_each_mc_addr(ha, bond_dev)
1628 dev_mc_add(slave_dev, ha->addr);
b9e40857 1629 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1630 }
1631
1632 if (bond->params.mode == BOND_MODE_8023AD) {
1633 /* add lacpdu mc addr to mc list */
1634 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1635
22bedad3 1636 dev_mc_add(slave_dev, lacpdu_multicast);
1da177e4
LT		 1637 }
1638
1639 bond_add_vlans_on_slave(bond, slave_dev);
1640
1641 write_lock_bh(&bond->lock);
1642
1643 bond_attach_slave(bond, new_slave);
1644
1645 new_slave->delay = 0;
1646 new_slave->link_failure_count = 0;
1647
8531c5ff
AK		 1648 bond_compute_features(bond);
1649
3915c1e8
JV		 1650 write_unlock_bh(&bond->lock);
1651
1652 read_lock(&bond->lock);
1653
f5b2b966
JV		 1654 new_slave->last_arp_rx = jiffies;
1655
1da177e4
LT		 1656 if (bond->params.miimon && !bond->params.use_carrier) {
1657 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1658
1659 if ((link_reporting == -1) && !bond->params.arp_interval) {
1660 /*
1661 * miimon is set but a bonded network driver
1662 * does not support ETHTOOL/MII and
1663 * arp_interval is not set. Note: if
1664 * use_carrier is enabled, we will never go
1665 * here (because netif_carrier is always
1666 * supported); thus, we don't need to change
1667 * the messages for netif_carrier.
1668 */
a4aee5c8 1669 pr_warning("%s: Warning: MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details.\n",
4e0952c7 1670 bond_dev->name, slave_dev->name);
1da177e4
LT		 1671 } else if (link_reporting == -1) {
1672 /* unable get link status using mii/ethtool */
a4aee5c8
JP		 1673 pr_warning("%s: Warning: can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface.\n",
1674 bond_dev->name, slave_dev->name);
1da177e4
LT		 1675 }
1676 }
1677
1678 /* check for initial state */
1679 if (!bond->params.miimon ||
1680 (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) {
1681 if (bond->params.updelay) {
a4aee5c8 1682 pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n");
1da177e4
LT		 1683 new_slave->link = BOND_LINK_BACK;
1684 new_slave->delay = bond->params.updelay;
1685 } else {
a4aee5c8 1686 pr_debug("Initial state of slave_dev is BOND_LINK_UP\n");
1da177e4
LT		 1687 new_slave->link = BOND_LINK_UP;
1688 }
1689 new_slave->jiffies = jiffies;
1690 } else {
a4aee5c8 1691 pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n");
1da177e4
LT		 1692 new_slave->link = BOND_LINK_DOWN;
1693 }
1694
1695 if (bond_update_speed_duplex(new_slave) &&
1696 (new_slave->link != BOND_LINK_DOWN)) {
a4aee5c8
JP		 1697 pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n",
1698 bond_dev->name, new_slave->dev->name);
1da177e4
LT		 1699
1700 if (bond->params.mode == BOND_MODE_8023AD) {
a4aee5c8
JP		 1701 pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n",
1702 bond_dev->name);
1da177e4
LT		 1703 }
1704 }
1705
1706 if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) {
1707 /* if there is a primary slave, remember it */
a549952a 1708 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1da177e4 1709 bond->primary_slave = new_slave;
a549952a
JP		 1710 bond->force_primary = true;
1711 }
1da177e4
LT		 1712 }
1713
3915c1e8
JV		 1714 write_lock_bh(&bond->curr_slave_lock);
1715
1da177e4
LT		 1716 switch (bond->params.mode) {
1717 case BOND_MODE_ACTIVEBACKUP:
8a8e447b
JV		 1718 bond_set_slave_inactive_flags(new_slave);
1719 bond_select_active_slave(bond);
1da177e4
LT		 1720 break;
1721 case BOND_MODE_8023AD:
1722 /* in 802.3ad mode, the internal mechanism
1723 * will activate the slaves in the selected
1724 * aggregator
1725 */
1726 bond_set_slave_inactive_flags(new_slave);
1727 /* if this is the first slave */
1728 if (bond->slave_cnt == 1) {
1729 SLAVE_AD_INFO(new_slave).id = 1;
1730 /* Initialize AD with the number of times that the AD timer is called in 1 second
1731 * can be called only after the mac address of the bond is set
1732 */
1733 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL,
1734 bond->params.lacp_fast);
1735 } else {
1736 SLAVE_AD_INFO(new_slave).id =
1737 SLAVE_AD_INFO(new_slave->prev).id + 1;
1738 }
1739
1740 bond_3ad_bind_slave(new_slave);
1741 break;
1742 case BOND_MODE_TLB:
1743 case BOND_MODE_ALB:
1744 new_slave->state = BOND_STATE_ACTIVE;
059fe7a5 1745 bond_set_slave_inactive_flags(new_slave);
5a29f789 1746 bond_select_active_slave(bond);
1da177e4
LT		 1747 break;
1748 default:
5a03cdb7 1749 pr_debug("This slave is always active in trunk mode\n");
1da177e4
LT		 1750
1751 /* always active in trunk mode */
1752 new_slave->state = BOND_STATE_ACTIVE;
1753
1754 /* In trunking mode there is little meaning to curr_active_slave
1755 * anyway (it holds no special properties of the bond device),
1756 * so we can change it without calling change_active_interface()
1757 */
3d632c3f 1758 if (!bond->curr_active_slave)
1da177e4 1759 bond->curr_active_slave = new_slave;
3d632c3f 1760
1da177e4
LT		 1761 break;
1762 } /* switch(bond_mode) */
1763
3915c1e8
JV		 1764 write_unlock_bh(&bond->curr_slave_lock);
1765
ff59c456
JV		 1766 bond_set_carrier(bond);
1767
f6dc31a8 1768#ifdef CONFIG_NET_POLL_CONTROLLER
c22d7ac8
AG		 1769 /*
1770 * Netpoll and bonding is broken, make sure it is not initialized
1771 * until it is fixed.
1772 */
1773 if (disable_netpoll) {
f6dc31a8 1774 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
c22d7ac8
AG		 1775 } else {
1776 if (slaves_support_netpoll(bond_dev)) {
1777 bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL;
1778 if (bond_dev->npinfo)
1779 slave_dev->npinfo = bond_dev->npinfo;
1780 } else if (!(bond_dev->priv_flags & IFF_DISABLE_NETPOLL)) {
1781 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
1782 pr_info("New slave device %s does not support netpoll\n",
1783 slave_dev->name);
1784 pr_info("Disabling netpoll support for %s\n", bond_dev->name);
1785 }
f6dc31a8
WC		 1786 }
1787#endif
3915c1e8 1788 read_unlock(&bond->lock);
1da177e4 1789
b76cdba9
MW		 1790 res = bond_create_slave_symlinks(bond_dev, slave_dev);
1791 if (res)
569f0c4d 1792 goto err_close;
b76cdba9 1793
a4aee5c8
JP		 1794 pr_info("%s: enslaving %s as a%s interface with a%s link.\n",
1795 bond_dev->name, slave_dev->name,
1796 new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup",
1797 new_slave->link != BOND_LINK_DOWN ? "n up" : " down");
1da177e4
LT		 1798
1799 /* enslave is successful */
1800 return 0;
1801
1802/* Undo stages on error */
1da177e4
LT		 1803err_close:
1804 dev_close(slave_dev);
1805
569f0c4d
JV		 1806err_unset_master:
1807 netdev_set_master(slave_dev, NULL);
1808
1da177e4 1809err_restore_mac:
dd957c57 1810 if (!bond->params.fail_over_mac) {
3915c1e8
JV		 1811 /* XXX TODO - fom follow mode needs to change master's
1812 * MAC if this slave's MAC is in use by the bond, or at
1813 * least print a warning.
1814 */
2ab82852
MS		 1815 memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN);
1816 addr.sa_family = slave_dev->type;
1817 dev_set_mac_address(slave_dev, &addr);
1818 }
1da177e4 1819
b15ba0fb
JP		 1820err_restore_mtu:
1821 dev_set_mtu(slave_dev, new_slave->original_mtu);
1822
1da177e4
LT		 1823err_free:
1824 kfree(new_slave);
1825
1826err_undo_flags:
1827 bond_dev->features = old_features;
3d632c3f 1828
1da177e4
LT		 1829 return res;
1830}
1831
1832/*
1833 * Try to release the slave device <slave> from the bond device <master>
1834 * It is legal to access curr_active_slave without a lock because all the function
1835 * is write-locked.
1836 *
1837 * The rules for slave state should be:
1838 * for Active/Backup:
1839 * Active stays on all backups go down
1840 * for Bonded connections:
1841 * The first up interface should be left on and all others downed.
1842 */
a77b5325 1843int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1844{
454d7c9b 1845 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1846 struct slave *slave, *oldcurrent;
1847 struct sockaddr addr;
1da177e4
LT		 1848
1849 /* slave is not a slave or master is not master of this slave */
1850 if (!(slave_dev->flags & IFF_SLAVE) ||
1851 (slave_dev->master != bond_dev)) {
a4aee5c8 1852 pr_err("%s: Error: cannot release %s.\n",
1da177e4
LT		 1853 bond_dev->name, slave_dev->name);
1854 return -EINVAL;
1855 }
1856
f6dc31a8 1857 netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE);
1da177e4
LT		 1858 write_lock_bh(&bond->lock);
1859
1860 slave = bond_get_slave_by_dev(bond, slave_dev);
1861 if (!slave) {
1862 /* not a slave of this bond */
a4aee5c8
JP		 1863 pr_info("%s: %s not enslaved\n",
1864 bond_dev->name, slave_dev->name);
f5e2a7b2 1865 write_unlock_bh(&bond->lock);
1da177e4
LT		 1866 return -EINVAL;
1867 }
1868
3915c1e8 1869 if (!bond->params.fail_over_mac) {
8e95a202
JP		 1870 if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) &&
1871 bond->slave_cnt > 1)
a4aee5c8
JP		 1872 pr_warning("%s: Warning: the permanent HWaddr of %s - %pM - is still in use by %s. Set the HWaddr of %s to a different address to avoid conflicts.\n",
1873 bond_dev->name, slave_dev->name,
1874 slave->perm_hwaddr,
1875 bond_dev->name, slave_dev->name);
1da177e4
LT		 1876 }
1877
1878 /* Inform AD package of unbinding of slave. */
1879 if (bond->params.mode == BOND_MODE_8023AD) {
1880 /* must be called before the slave is
1881 * detached from the list
1882 */
1883 bond_3ad_unbind_slave(slave);
1884 }
1885
a4aee5c8
JP		 1886 pr_info("%s: releasing %s interface %s\n",
1887 bond_dev->name,
1888 (slave->state == BOND_STATE_ACTIVE) ? "active" : "backup",
1889 slave_dev->name);
1da177e4
LT		 1890
1891 oldcurrent = bond->curr_active_slave;
1892
1893 bond->current_arp_slave = NULL;
1894
1895 /* release the slave from its bond */
1896 bond_detach_slave(bond, slave);
1897
8531c5ff
AK		 1898 bond_compute_features(bond);
1899
3d632c3f 1900 if (bond->primary_slave == slave)
1da177e4 1901 bond->primary_slave = NULL;
1da177e4 1902
3d632c3f 1903 if (oldcurrent == slave)
1da177e4 1904 bond_change_active_slave(bond, NULL);
1da177e4 1905
58402054 1906 if (bond_is_lb(bond)) {
1da177e4
LT		 1907 /* Must be called only after the slave has been
1908 * detached from the list and the curr_active_slave
1909 * has been cleared (if our_slave == old_current),
1910 * but before a new active slave is selected.
1911 */
2543331d 1912 write_unlock_bh(&bond->lock);
1da177e4 1913 bond_alb_deinit_slave(bond, slave);
2543331d 1914 write_lock_bh(&bond->lock);
1da177e4
LT		 1915 }
1916
059fe7a5
JV		 1917 if (oldcurrent == slave) {
1918 /*
1919 * Note that we hold RTNL over this sequence, so there
1920 * is no concern that another slave add/remove event
1921 * will interfere.
1922 */
1923 write_unlock_bh(&bond->lock);
1924 read_lock(&bond->lock);
1925 write_lock_bh(&bond->curr_slave_lock);
1926
1da177e4
LT		 1927 bond_select_active_slave(bond);
1928
059fe7a5
JV		 1929 write_unlock_bh(&bond->curr_slave_lock);
1930 read_unlock(&bond->lock);
1931 write_lock_bh(&bond->lock);
1932 }
1933
1da177e4 1934 if (bond->slave_cnt == 0) {
ff59c456
JV		 1935 bond_set_carrier(bond);
1936
1da177e4
LT		 1937 /* if the last slave was removed, zero the mac address
1938 * of the master so it will be set by the application
1939 * to the mac address of the first slave
1940 */
1941 memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
1942
1943 if (list_empty(&bond->vlan_list)) {
1944 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
1945 } else {
a4aee5c8
JP		 1946 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
1947 bond_dev->name, bond_dev->name);
1948 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
1949 bond_dev->name);
1da177e4
LT		 1950 }
1951 } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1952 !bond_has_challenged_slaves(bond)) {
a4aee5c8
JP		 1953 pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n",
1954 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1955 bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
1956 }
1957
1958 write_unlock_bh(&bond->lock);
1959
b76cdba9
MW		 1960 /* must do this from outside any spinlocks */
1961 bond_destroy_slave_symlinks(bond_dev, slave_dev);
1962
1da177e4
LT		 1963 bond_del_vlans_from_slave(bond, slave_dev);
1964
1965 /* If the mode USES_PRIMARY, then we should only remove its
1966 * promisc and mc settings if it was the curr_active_slave, but that was
1967 * already taken care of above when we detached the slave
1968 */
1969 if (!USES_PRIMARY(bond->params.mode)) {
1970 /* unset promiscuity level from slave */
3d632c3f 1971 if (bond_dev->flags & IFF_PROMISC)
1da177e4 1972 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 1973
1974 /* unset allmulti level from slave */
3d632c3f 1975 if (bond_dev->flags & IFF_ALLMULTI)
1da177e4 1976 dev_set_allmulti(slave_dev, -1);
1da177e4
LT		 1977
1978 /* flush master's mc_list from slave */
b9e40857 1979 netif_addr_lock_bh(bond_dev);
1da177e4 1980 bond_mc_list_flush(bond_dev, slave_dev);
b9e40857 1981 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1982 }
1983
1984 netdev_set_master(slave_dev, NULL);
1985
f6dc31a8
WC		 1986#ifdef CONFIG_NET_POLL_CONTROLLER
1987 read_lock_bh(&bond->lock);
c22d7ac8
AG		 1988
1989 /* Make sure netpoll over stays disabled until fixed. */
1990 if (!disable_netpoll)
1991 if (slaves_support_netpoll(bond_dev))
1992 bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL;
f6dc31a8
WC		 1993 read_unlock_bh(&bond->lock);
1994 if (slave_dev->netdev_ops->ndo_netpoll_cleanup)
1995 slave_dev->netdev_ops->ndo_netpoll_cleanup(slave_dev);
1996 else
1997 slave_dev->npinfo = NULL;
1998#endif
1999
1da177e4
LT		 2000 /* close slave before restoring its mac address */
2001 dev_close(slave_dev);
2002
3915c1e8 2003 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
2ab82852
MS		 2004 /* restore original ("permanent") mac address */
2005 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
2006 addr.sa_family = slave_dev->type;
2007 dev_set_mac_address(slave_dev, &addr);
2008 }
1da177e4 2009
b15ba0fb
JP		 2010 dev_set_mtu(slave_dev, slave->original_mtu);
2011
8f903c70 2012 slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB |
f5b2b966
JV		 2013 IFF_SLAVE_INACTIVE | IFF_BONDING |
2014 IFF_SLAVE_NEEDARP);
1da177e4
LT		 2015
2016 kfree(slave);
2017
2018 return 0; /* deletion OK */
2019}
2020
d90a162a 2021/*
3d632c3f 2022* First release a slave and than destroy the bond if no more slaves are left.
d90a162a
MS		 2023* Must be under rtnl_lock when this function is called.
2024*/
3d632c3f
SH		 2025int bond_release_and_destroy(struct net_device *bond_dev,
2026 struct net_device *slave_dev)
d90a162a 2027{
454d7c9b 2028 struct bonding *bond = netdev_priv(bond_dev);
d90a162a
MS		 2029 int ret;
2030
2031 ret = bond_release(bond_dev, slave_dev);
2032 if ((ret == 0) && (bond->slave_cnt == 0)) {
a4aee5c8
JP		 2033 pr_info("%s: destroying bond %s.\n",
2034 bond_dev->name, bond_dev->name);
9e71626c 2035 unregister_netdevice(bond_dev);
d90a162a
MS		 2036 }
2037 return ret;
2038}
2039
1da177e4
LT		 2040/*
2041 * This function releases all slaves.
2042 */
2043static int bond_release_all(struct net_device *bond_dev)
2044{
454d7c9b 2045 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 2046 struct slave *slave;
2047 struct net_device *slave_dev;
2048 struct sockaddr addr;
2049
2050 write_lock_bh(&bond->lock);
2051
ff59c456
JV		 2052 netif_carrier_off(bond_dev);
2053
3d632c3f 2054 if (bond->slave_cnt == 0)
1da177e4 2055 goto out;
1da177e4
LT		 2056
2057 bond->current_arp_slave = NULL;
2058 bond->primary_slave = NULL;
2059 bond_change_active_slave(bond, NULL);
2060
2061 while ((slave = bond->first_slave) != NULL) {
2062 /* Inform AD package of unbinding of slave
2063 * before slave is detached from the list.
2064 */
3d632c3f 2065 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 2066 bond_3ad_unbind_slave(slave);
1da177e4
LT		 2067
2068 slave_dev = slave->dev;
2069 bond_detach_slave(bond, slave);
2070
2543331d
JV		 2071 /* now that the slave is detached, unlock and perform
2072 * all the undo steps that should not be called from
2073 * within a lock.
2074 */
2075 write_unlock_bh(&bond->lock);
2076
58402054 2077 if (bond_is_lb(bond)) {
1da177e4
LT		 2078 /* must be called only after the slave
2079 * has been detached from the list
2080 */
2081 bond_alb_deinit_slave(bond, slave);
2082 }
2083
8531c5ff
AK		 2084 bond_compute_features(bond);
2085
b76cdba9 2086 bond_destroy_slave_symlinks(bond_dev, slave_dev);
1da177e4
LT		 2087 bond_del_vlans_from_slave(bond, slave_dev);
2088
2089 /* If the mode USES_PRIMARY, then we should only remove its
2090 * promisc and mc settings if it was the curr_active_slave, but that was
2091 * already taken care of above when we detached the slave
2092 */
2093 if (!USES_PRIMARY(bond->params.mode)) {
2094 /* unset promiscuity level from slave */
3d632c3f 2095 if (bond_dev->flags & IFF_PROMISC)
1da177e4 2096 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 2097
2098 /* unset allmulti level from slave */
3d632c3f 2099 if (bond_dev->flags & IFF_ALLMULTI)
1da177e4 2100 dev_set_allmulti(slave_dev, -1);
1da177e4
LT		 2101
2102 /* flush master's mc_list from slave */
b9e40857 2103 netif_addr_lock_bh(bond_dev);
1da177e4 2104 bond_mc_list_flush(bond_dev, slave_dev);
b9e40857 2105 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 2106 }
2107
2108 netdev_set_master(slave_dev, NULL);
2109
2110 /* close slave before restoring its mac address */
2111 dev_close(slave_dev);
2112
dd957c57 2113 if (!bond->params.fail_over_mac) {
2ab82852
MS		 2114 /* restore original ("permanent") mac address*/
2115 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
2116 addr.sa_family = slave_dev->type;
2117 dev_set_mac_address(slave_dev, &addr);
2118 }
1da177e4 2119
8f903c70
JV		 2120 slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB |
2121 IFF_SLAVE_INACTIVE);
1da177e4
LT		 2122
2123 kfree(slave);
2124
2125 /* re-acquire the lock before getting the next slave */
2126 write_lock_bh(&bond->lock);
2127 }
2128
2129 /* zero the mac address of the master so it will be
2130 * set by the application to the mac address of the
2131 * first slave
2132 */
2133 memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
2134
3d632c3f 2135 if (list_empty(&bond->vlan_list))
1da177e4 2136 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
3d632c3f 2137 else {
a4aee5c8
JP		 2138 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
2139 bond_dev->name, bond_dev->name);
2140 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
2141 bond_dev->name);
1da177e4
LT		 2142 }
2143
a4aee5c8 2144 pr_info("%s: released all slaves\n", bond_dev->name);
1da177e4
LT		 2145
2146out:
2147 write_unlock_bh(&bond->lock);
2148
2149 return 0;
2150}
2151
2152/*
2153 * This function changes the active slave to slave <slave_dev>.
2154 * It returns -EINVAL in the following cases.
2155 * - <slave_dev> is not found in the list.
2156 * - There is not active slave now.
2157 * - <slave_dev> is already active.
2158 * - The link state of <slave_dev> is not BOND_LINK_UP.
2159 * - <slave_dev> is not running.
3d632c3f
SH		 2160 * In these cases, this function does nothing.
2161 * In the other cases, current_slave pointer is changed and 0 is returned.
1da177e4
LT		 2162 */
2163static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev)
2164{
454d7c9b 2165 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 2166 struct slave *old_active = NULL;
2167 struct slave *new_active = NULL;
2168 int res = 0;
2169
3d632c3f 2170 if (!USES_PRIMARY(bond->params.mode))
1da177e4 2171 return -EINVAL;
1da177e4
LT		 2172
2173 /* Verify that master_dev is indeed the master of slave_dev */
3d632c3f 2174 if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev))
1da177e4 2175 return -EINVAL;
1da177e4 2176
059fe7a5 2177 read_lock(&bond->lock);
1da177e4 2178
059fe7a5 2179 read_lock(&bond->curr_slave_lock);
1da177e4 2180 old_active = bond->curr_active_slave;
059fe7a5
JV		 2181 read_unlock(&bond->curr_slave_lock);
2182
1da177e4
LT		 2183 new_active = bond_get_slave_by_dev(bond, slave_dev);
2184
2185 /*
2186 * Changing to the current active: do nothing; return success.
2187 */
2188 if (new_active && (new_active == old_active)) {
059fe7a5 2189 read_unlock(&bond->lock);
1da177e4
LT		 2190 return 0;
2191 }
2192
2193 if ((new_active) &&
2194 (old_active) &&
2195 (new_active->link == BOND_LINK_UP) &&
2196 IS_UP(new_active->dev)) {
059fe7a5 2197 write_lock_bh(&bond->curr_slave_lock);
1da177e4 2198 bond_change_active_slave(bond, new_active);
059fe7a5 2199 write_unlock_bh(&bond->curr_slave_lock);
3d632c3f 2200 } else
1da177e4 2201 res = -EINVAL;
1da177e4 2202
059fe7a5 2203 read_unlock(&bond->lock);
1da177e4
LT		 2204
2205 return res;
2206}
2207
1da177e4
LT		 2208static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
2209{
454d7c9b 2210 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 2211
2212 info->bond_mode = bond->params.mode;
2213 info->miimon = bond->params.miimon;
2214
6603a6f2 2215 read_lock(&bond->lock);
1da177e4 2216 info->num_slaves = bond->slave_cnt;
6603a6f2 2217 read_unlock(&bond->lock);
1da177e4
LT		 2218
2219 return 0;
2220}
2221
2222static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
2223{
454d7c9b 2224 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 2225 struct slave *slave;
689c96cc 2226 int i, res = -ENODEV;
1da177e4 2227
6603a6f2 2228 read_lock(&bond->lock);
1da177e4
LT		 2229
2230 bond_for_each_slave(bond, slave, i) {
2231 if (i == (int)info->slave_id) {
689c96cc
ED		 2232 res = 0;
2233 strcpy(info->slave_name, slave->dev->name);
2234 info->link = slave->link;
2235 info->state = slave->state;
2236 info->link_failure_count = slave->link_failure_count;
1da177e4
LT		 2237 break;
2238 }
2239 }
2240
6603a6f2 2241 read_unlock(&bond->lock);
1da177e4 2242
689c96cc 2243 return res;
1da177e4
LT		 2244}
2245
2246/*-------------------------------- Monitoring -------------------------------*/
2247
1da177e4 2248
f0c76d61
JV		 2249static int bond_miimon_inspect(struct bonding *bond)
2250{
2251 struct slave *slave;
2252 int i, link_state, commit = 0;
41f89100
JP		 2253 bool ignore_updelay;
2254
2255 ignore_updelay = !bond->curr_active_slave ? true : false;
1da177e4
LT		 2256
2257 bond_for_each_slave(bond, slave, i) {
f0c76d61 2258 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 2259
f0c76d61 2260 link_state = bond_check_dev_link(bond, slave->dev, 0);
1da177e4
LT		 2261
2262 switch (slave->link) {
f0c76d61
JV		 2263 case BOND_LINK_UP:
2264 if (link_state)
2265 continue;
1da177e4 2266
f0c76d61
JV		 2267 slave->link = BOND_LINK_FAIL;
2268 slave->delay = bond->params.downdelay;
2269 if (slave->delay) {
a4aee5c8
JP		 2270 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n",
2271 bond->dev->name,
2272 (bond->params.mode ==
2273 BOND_MODE_ACTIVEBACKUP) ?
2274 ((slave->state == BOND_STATE_ACTIVE) ?
2275 "active " : "backup ") : "",
2276 slave->dev->name,
2277 bond->params.downdelay * bond->params.miimon);
1da177e4 2278 }
f0c76d61
JV		 2279 /*FALLTHRU*/
2280 case BOND_LINK_FAIL:
2281 if (link_state) {
2282 /*
2283 * recovered before downdelay expired
2284 */
2285 slave->link = BOND_LINK_UP;
1da177e4 2286 slave->jiffies = jiffies;
a4aee5c8
JP		 2287 pr_info("%s: link status up again after %d ms for interface %s.\n",
2288 bond->dev->name,
2289 (bond->params.downdelay - slave->delay) *
2290 bond->params.miimon,
2291 slave->dev->name);
f0c76d61 2292 continue;
1da177e4 2293 }
f0c76d61
JV		 2294
2295 if (slave->delay <= 0) {
2296 slave->new_link = BOND_LINK_DOWN;
2297 commit++;
2298 continue;
1da177e4 2299 }
1da177e4 2300
f0c76d61
JV		 2301 slave->delay--;
2302 break;
2303
2304 case BOND_LINK_DOWN:
2305 if (!link_state)
2306 continue;
2307
2308 slave->link = BOND_LINK_BACK;
2309 slave->delay = bond->params.updelay;
2310
2311 if (slave->delay) {
a4aee5c8
JP		 2312 pr_info("%s: link status up for interface %s, enabling it in %d ms.\n",
2313 bond->dev->name, slave->dev->name,
2314 ignore_updelay ? 0 :
2315 bond->params.updelay *
2316 bond->params.miimon);
f0c76d61
JV		 2317 }
2318 /*FALLTHRU*/
2319 case BOND_LINK_BACK:
2320 if (!link_state) {
2321 slave->link = BOND_LINK_DOWN;
a4aee5c8
JP		 2322 pr_info("%s: link status down again after %d ms for interface %s.\n",
2323 bond->dev->name,
2324 (bond->params.updelay - slave->delay) *
2325 bond->params.miimon,
2326 slave->dev->name);
f0c76d61
JV		 2327
2328 continue;
2329 }
2330
41f89100
JP		 2331 if (ignore_updelay)
2332 slave->delay = 0;
2333
f0c76d61
JV		 2334 if (slave->delay <= 0) {
2335 slave->new_link = BOND_LINK_UP;
2336 commit++;
41f89100 2337 ignore_updelay = false;
f0c76d61 2338 continue;
1da177e4 2339 }
f0c76d61
JV		 2340
2341 slave->delay--;
1da177e4 2342 break;
f0c76d61
JV		 2343 }
2344 }
1da177e4 2345
f0c76d61
JV		 2346 return commit;
2347}
1da177e4 2348
f0c76d61
JV		 2349static void bond_miimon_commit(struct bonding *bond)
2350{
2351 struct slave *slave;
2352 int i;
2353
2354 bond_for_each_slave(bond, slave, i) {
2355 switch (slave->new_link) {
2356 case BOND_LINK_NOCHANGE:
2357 continue;
1da177e4 2358
f0c76d61
JV		 2359 case BOND_LINK_UP:
2360 slave->link = BOND_LINK_UP;
2361 slave->jiffies = jiffies;
2362
2363 if (bond->params.mode == BOND_MODE_8023AD) {
2364 /* prevent it from being the active one */
2365 slave->state = BOND_STATE_BACKUP;
2366 } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2367 /* make it immediately active */
2368 slave->state = BOND_STATE_ACTIVE;
2369 } else if (slave != bond->primary_slave) {
2370 /* prevent it from being the active one */
2371 slave->state = BOND_STATE_BACKUP;
1da177e4 2372 }
1da177e4 2373
a4aee5c8
JP		 2374 pr_info("%s: link status definitely up for interface %s.\n",
2375 bond->dev->name, slave->dev->name);
1da177e4 2376
f0c76d61
JV		 2377 /* notify ad that the link status has changed */
2378 if (bond->params.mode == BOND_MODE_8023AD)
2379 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
059fe7a5 2380
58402054 2381 if (bond_is_lb(bond))
f0c76d61
JV		 2382 bond_alb_handle_link_change(bond, slave,
2383 BOND_LINK_UP);
1da177e4 2384
f0c76d61
JV		 2385 if (!bond->curr_active_slave ||
2386 (slave == bond->primary_slave))
2387 goto do_failover;
1da177e4 2388
f0c76d61 2389 continue;
059fe7a5 2390
f0c76d61 2391 case BOND_LINK_DOWN:
fba4acda
JV		 2392 if (slave->link_failure_count < UINT_MAX)
2393 slave->link_failure_count++;
2394
f0c76d61 2395 slave->link = BOND_LINK_DOWN;
1da177e4 2396
f0c76d61
JV		 2397 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP ||
2398 bond->params.mode == BOND_MODE_8023AD)
2399 bond_set_slave_inactive_flags(slave);
2400
a4aee5c8
JP		 2401 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2402 bond->dev->name, slave->dev->name);
f0c76d61
JV		 2403
2404 if (bond->params.mode == BOND_MODE_8023AD)
2405 bond_3ad_handle_link_change(slave,
2406 BOND_LINK_DOWN);
2407
ae63e808 2408 if (bond_is_lb(bond))
f0c76d61
JV		 2409 bond_alb_handle_link_change(bond, slave,
2410 BOND_LINK_DOWN);
2411
2412 if (slave == bond->curr_active_slave)
2413 goto do_failover;
2414
2415 continue;
2416
2417 default:
a4aee5c8 2418 pr_err("%s: invalid new link %d on slave %s\n",
f0c76d61
JV		 2419 bond->dev->name, slave->new_link,
2420 slave->dev->name);
2421 slave->new_link = BOND_LINK_NOCHANGE;
2422
2423 continue;
2424 }
2425
2426do_failover:
2427 ASSERT_RTNL();
2428 write_lock_bh(&bond->curr_slave_lock);
2429 bond_select_active_slave(bond);
2430 write_unlock_bh(&bond->curr_slave_lock);
2431 }
2432
2433 bond_set_carrier(bond);
1da177e4
LT		 2434}
2435
0b0eef66
JV		 2436/*
2437 * bond_mii_monitor
2438 *
2439 * Really a wrapper that splits the mii monitor into two phases: an
f0c76d61
JV		 2440 * inspection, then (if inspection indicates something needs to be done)
2441 * an acquisition of appropriate locks followed by a commit phase to
2442 * implement whatever link state changes are indicated.
0b0eef66
JV		 2443 */
2444void bond_mii_monitor(struct work_struct *work)
2445{
2446 struct bonding *bond = container_of(work, struct bonding,
2447 mii_work.work);
0b0eef66
JV		 2448
2449 read_lock(&bond->lock);
f0c76d61
JV		 2450 if (bond->kill_timers)
2451 goto out;
2452
2453 if (bond->slave_cnt == 0)
2454 goto re_arm;
b59f9f74
JV		 2455
2456 if (bond->send_grat_arp) {
2457 read_lock(&bond->curr_slave_lock);
2458 bond_send_gratuitous_arp(bond);
2459 read_unlock(&bond->curr_slave_lock);
2460 }
2461
305d552a
BH		 2462 if (bond->send_unsol_na) {
2463 read_lock(&bond->curr_slave_lock);
2464 bond_send_unsolicited_na(bond);
2465 read_unlock(&bond->curr_slave_lock);
2466 }
2467
f0c76d61 2468 if (bond_miimon_inspect(bond)) {
0b0eef66
JV		 2469 read_unlock(&bond->lock);
2470 rtnl_lock();
2471 read_lock(&bond->lock);
f0c76d61
JV		 2472
2473 bond_miimon_commit(bond);
2474
5655662d
JV		 2475 read_unlock(&bond->lock);
2476 rtnl_unlock(); /* might sleep, hold no other locks */
2477 read_lock(&bond->lock);
0b0eef66
JV		 2478 }
2479
f0c76d61
JV		 2480re_arm:
2481 if (bond->params.miimon)
2482 queue_delayed_work(bond->wq, &bond->mii_work,
2483 msecs_to_jiffies(bond->params.miimon));
2484out:
0b0eef66 2485 read_unlock(&bond->lock);
0b0eef66 2486}
c3ade5ca 2487
d3bb52b0 2488static __be32 bond_glean_dev_ip(struct net_device *dev)
c3ade5ca
JV		 2489{
2490 struct in_device *idev;
2491 struct in_ifaddr *ifa;
a144ea4b 2492 __be32 addr = 0;
c3ade5ca
JV		 2493
2494 if (!dev)
2495 return 0;
2496
2497 rcu_read_lock();
e5ed6399 2498 idev = __in_dev_get_rcu(dev);
c3ade5ca
JV		 2499 if (!idev)
2500 goto out;
2501
2502 ifa = idev->ifa_list;
2503 if (!ifa)
2504 goto out;
2505
2506 addr = ifa->ifa_local;
2507out:
2508 rcu_read_unlock();
2509 return addr;
2510}
2511
d3bb52b0 2512static int bond_has_this_ip(struct bonding *bond, __be32 ip)
f5b2b966 2513{
0883beca 2514 struct vlan_entry *vlan;
f5b2b966
JV		 2515
2516 if (ip == bond->master_ip)
2517 return 1;
2518
0883beca 2519 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
f5b2b966
JV		 2520 if (ip == vlan->vlan_ip)
2521 return 1;
2522 }
2523
2524 return 0;
2525}
2526
c3ade5ca
JV		 2527/*
2528 * We go to the (large) trouble of VLAN tagging ARP frames because
2529 * switches in VLAN mode (especially if ports are configured as
2530 * "native" to a VLAN) might not pass non-tagged frames.
2531 */
d3bb52b0 2532static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id)
c3ade5ca
JV		 2533{
2534 struct sk_buff *skb;
2535
5a03cdb7 2536 pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op,
a4aee5c8 2537 slave_dev->name, dest_ip, src_ip, vlan_id);
3d632c3f 2538
c3ade5ca
JV		 2539 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2540 NULL, slave_dev->dev_addr, NULL);
2541
2542 if (!skb) {
a4aee5c8 2543 pr_err("ARP packet allocation failed\n");
c3ade5ca
JV		 2544 return;
2545 }
2546 if (vlan_id) {
2547 skb = vlan_put_tag(skb, vlan_id);
2548 if (!skb) {
a4aee5c8 2549 pr_err("failed to insert VLAN tag\n");
c3ade5ca
JV		 2550 return;
2551 }
2552 }
2553 arp_xmit(skb);
2554}
2555
2556
1da177e4
LT		 2557static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2558{
c3ade5ca 2559 int i, vlan_id, rv;
d3bb52b0 2560 __be32 *targets = bond->params.arp_targets;
0883beca 2561 struct vlan_entry *vlan;
c3ade5ca
JV		 2562 struct net_device *vlan_dev;
2563 struct flowi fl;
2564 struct rtable *rt;
1da177e4 2565
6b780567
MW		 2566 for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) {
2567 if (!targets[i])
5a31bec0 2568 break;
5a03cdb7 2569 pr_debug("basa: target %x\n", targets[i]);
c3ade5ca 2570 if (list_empty(&bond->vlan_list)) {
5a03cdb7 2571 pr_debug("basa: empty vlan: arp_send\n");
c3ade5ca
JV		 2572 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2573 bond->master_ip, 0);
2574 continue;
2575 }
2576
2577 /*
2578 * If VLANs are configured, we do a route lookup to
2579 * determine which VLAN interface would be used, so we
2580 * can tag the ARP with the proper VLAN tag.
2581 */
2582 memset(&fl, 0, sizeof(fl));
2583 fl.fl4_dst = targets[i];
2584 fl.fl4_tos = RTO_ONLINK;
2585
ec87fd3b 2586 rv = ip_route_output_key(dev_net(bond->dev), &rt, &fl);
c3ade5ca
JV		 2587 if (rv) {
2588 if (net_ratelimit()) {
a4aee5c8
JP		 2589 pr_warning("%s: no route to arp_ip_target %pI4\n",
2590 bond->dev->name, &fl.fl4_dst);
c3ade5ca
JV		 2591 }
2592 continue;
2593 }
2594
2595 /*
2596 * This target is not on a VLAN
2597 */
d8d1f30b 2598 if (rt->dst.dev == bond->dev) {
ed4b9f80 2599 ip_rt_put(rt);
5a03cdb7 2600 pr_debug("basa: rtdev == bond->dev: arp_send\n");
c3ade5ca
JV		 2601 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2602 bond->master_ip, 0);
2603 continue;
2604 }
2605
2606 vlan_id = 0;
0883beca 2607 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
5c15bdec 2608 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
d8d1f30b 2609 if (vlan_dev == rt->dst.dev) {
c3ade5ca 2610 vlan_id = vlan->vlan_id;
5a03cdb7 2611 pr_debug("basa: vlan match on %s %d\n",
c3ade5ca
JV		 2612 vlan_dev->name, vlan_id);
2613 break;
2614 }
2615 }
2616
2617 if (vlan_id) {
ed4b9f80 2618 ip_rt_put(rt);
c3ade5ca
JV		 2619 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2620 vlan->vlan_ip, vlan_id);
2621 continue;
2622 }
2623
2624 if (net_ratelimit()) {
a4aee5c8
JP		 2625 pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2626 bond->dev->name, &fl.fl4_dst,
d8d1f30b 2627 rt->dst.dev ? rt->dst.dev->name : "NULL");
c3ade5ca 2628 }
ed4b9f80 2629 ip_rt_put(rt);
c3ade5ca
JV		 2630 }
2631}
2632
2633/*
2634 * Kick out a gratuitous ARP for an IP on the bonding master plus one
2635 * for each VLAN above us.
b59f9f74
JV		 2636 *
2637 * Caller must hold curr_slave_lock for read or better
c3ade5ca
JV		 2638 */
2639static void bond_send_gratuitous_arp(struct bonding *bond)
2640{
2641 struct slave *slave = bond->curr_active_slave;
2642 struct vlan_entry *vlan;
2643 struct net_device *vlan_dev;
2644
a4aee5c8
JP		 2645 pr_debug("bond_send_grat_arp: bond %s slave %s\n",
2646 bond->dev->name, slave ? slave->dev->name : "NULL");
b59f9f74
JV		 2647
2648 if (!slave || !bond->send_grat_arp ||
2649 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
c3ade5ca
JV		 2650 return;
2651
b59f9f74
JV		 2652 bond->send_grat_arp--;
2653
c3ade5ca
JV		 2654 if (bond->master_ip) {
2655 bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip,
1053f62c 2656 bond->master_ip, 0);
c3ade5ca
JV		 2657 }
2658
2659 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
5c15bdec 2660 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
c3ade5ca
JV		 2661 if (vlan->vlan_ip) {
2662 bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip,
2663 vlan->vlan_ip, vlan->vlan_id);
2664 }
1da177e4
LT		 2665 }
2666}
2667
d3bb52b0 2668static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
f5b2b966
JV		 2669{
2670 int i;
d3bb52b0 2671 __be32 *targets = bond->params.arp_targets;
f5b2b966 2672
f5b2b966 2673 for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) {
5a03cdb7 2674 pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n",
a4aee5c8
JP		 2675 &sip, &tip, i, &targets[i],
2676 bond_has_this_ip(bond, tip));
f5b2b966
JV		 2677 if (sip == targets[i]) {
2678 if (bond_has_this_ip(bond, tip))
2679 slave->last_arp_rx = jiffies;
2680 return;
2681 }
2682 }
2683}
2684
2685static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
2686{
2687 struct arphdr *arp;
2688 struct slave *slave;
2689 struct bonding *bond;
2690 unsigned char *arp_ptr;
d3bb52b0 2691 __be32 sip, tip;
f5b2b966 2692
1f3c8804
AG		 2693 if (dev->priv_flags & IFF_802_1Q_VLAN) {
2694 /*
2695 * When using VLANS and bonding, dev and oriv_dev may be
2696 * incorrect if the physical interface supports VLAN
2697 * acceleration. With this change ARP validation now
2698 * works for hosts only reachable on the VLAN interface.
2699 */
2700 dev = vlan_dev_real_dev(dev);
2701 orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif);
2702 }
2703
f5b2b966
JV		 2704 if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER))
2705 goto out;
2706
454d7c9b 2707 bond = netdev_priv(dev);
f5b2b966
JV		 2708 read_lock(&bond->lock);
2709
5a03cdb7 2710 pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n",
a4aee5c8
JP		 2711 bond->dev->name, skb->dev ? skb->dev->name : "NULL",
2712 orig_dev ? orig_dev->name : "NULL");
f5b2b966
JV		 2713
2714 slave = bond_get_slave_by_dev(bond, orig_dev);
2715 if (!slave || !slave_do_arp_validate(bond, slave))
2716 goto out_unlock;
2717
988b7050 2718 if (!pskb_may_pull(skb, arp_hdr_len(dev)))
f5b2b966
JV		 2719 goto out_unlock;
2720
d0a92be0 2721 arp = arp_hdr(skb);
f5b2b966
JV		 2722 if (arp->ar_hln != dev->addr_len ||
2723 skb->pkt_type == PACKET_OTHERHOST ||
2724 skb->pkt_type == PACKET_LOOPBACK ||
2725 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2726 arp->ar_pro != htons(ETH_P_IP) ||
2727 arp->ar_pln != 4)
2728 goto out_unlock;
2729
2730 arp_ptr = (unsigned char *)(arp + 1);
2731 arp_ptr += dev->addr_len;
2732 memcpy(&sip, arp_ptr, 4);
2733 arp_ptr += 4 + dev->addr_len;
2734 memcpy(&tip, arp_ptr, 4);
2735
5a03cdb7 2736 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
a4aee5c8
JP		 2737 bond->dev->name, slave->dev->name, slave->state,
2738 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2739 &sip, &tip);
f5b2b966
JV		 2740
2741 /*
2742 * Backup slaves won't see the ARP reply, but do come through
2743 * here for each ARP probe (so we swap the sip/tip to validate
2744 * the probe). In a "redundant switch, common router" type of
2745 * configuration, the ARP probe will (hopefully) travel from
2746 * the active, through one switch, the router, then the other
2747 * switch before reaching the backup.
2748 */
2749 if (slave->state == BOND_STATE_ACTIVE)
2750 bond_validate_arp(bond, slave, sip, tip);
2751 else
2752 bond_validate_arp(bond, slave, tip, sip);
2753
2754out_unlock:
2755 read_unlock(&bond->lock);
2756out:
2757 dev_kfree_skb(skb);
2758 return NET_RX_SUCCESS;
2759}
2760
1da177e4
LT		 2761/*
2762 * this function is called regularly to monitor each slave's link
2763 * ensuring that traffic is being sent and received when arp monitoring
2764 * is used in load-balancing mode. if the adapter has been dormant, then an
2765 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2766 * arp monitoring in active backup mode.
2767 */
1b76b316 2768void bond_loadbalance_arp_mon(struct work_struct *work)
1da177e4 2769{
1b76b316
JV		 2770 struct bonding *bond = container_of(work, struct bonding,
2771 arp_work.work);
1da177e4
LT		 2772 struct slave *slave, *oldcurrent;
2773 int do_failover = 0;
2774 int delta_in_ticks;
2775 int i;
2776
2777 read_lock(&bond->lock);
2778
5ce0da8f 2779 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
1da177e4 2780
3d632c3f 2781 if (bond->kill_timers)
1da177e4 2782 goto out;
1da177e4 2783
3d632c3f 2784 if (bond->slave_cnt == 0)
1da177e4 2785 goto re_arm;
1da177e4
LT		 2786
2787 read_lock(&bond->curr_slave_lock);
2788 oldcurrent = bond->curr_active_slave;
2789 read_unlock(&bond->curr_slave_lock);
2790
2791 /* see if any of the previous devices are up now (i.e. they have
2792 * xmt and rcv traffic). the curr_active_slave does not come into
2793 * the picture unless it is null. also, slave->jiffies is not needed
2794 * here because we send an arp on each slave and give a slave as
2795 * long as it needs to get the tx/rx within the delta.
2796 * TODO: what about up/down delay in arp mode? it wasn't here before
2797 * so it can wait
2798 */
2799 bond_for_each_slave(bond, slave, i) {
2800 if (slave->link != BOND_LINK_UP) {
9d21493b 2801 if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) &&
b63bb739 2802 time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) {
1da177e4
LT		 2803
2804 slave->link = BOND_LINK_UP;
2805 slave->state = BOND_STATE_ACTIVE;
2806
2807 /* primary_slave has no meaning in round-robin
2808 * mode. the window of a slave being up and
2809 * curr_active_slave being null after enslaving
2810 * is closed.
2811 */
2812 if (!oldcurrent) {
a4aee5c8
JP		 2813 pr_info("%s: link status definitely up for interface %s, ",
2814 bond->dev->name,
2815 slave->dev->name);
1da177e4
LT		 2816 do_failover = 1;
2817 } else {
a4aee5c8
JP		 2818 pr_info("%s: interface %s is now up\n",
2819 bond->dev->name,
2820 slave->dev->name);
1da177e4
LT		 2821 }
2822 }
2823 } else {
2824 /* slave->link == BOND_LINK_UP */
2825
2826 /* not all switches will respond to an arp request
2827 * when the source ip is 0, so don't take the link down
2828 * if we don't know our ip yet
2829 */
9d21493b 2830 if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) ||
4b8a9239 2831 (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) {
1da177e4
LT		 2832
2833 slave->link = BOND_LINK_DOWN;
2834 slave->state = BOND_STATE_BACKUP;
2835
3d632c3f 2836 if (slave->link_failure_count < UINT_MAX)
1da177e4 2837 slave->link_failure_count++;
1da177e4 2838
a4aee5c8
JP		 2839 pr_info("%s: interface %s is now down.\n",
2840 bond->dev->name,
2841 slave->dev->name);
1da177e4 2842
3d632c3f 2843 if (slave == oldcurrent)
1da177e4 2844 do_failover = 1;
1da177e4
LT		 2845 }
2846 }
2847
2848 /* note: if switch is in round-robin mode, all links
2849 * must tx arp to ensure all links rx an arp - otherwise
2850 * links may oscillate or not come up at all; if switch is
2851 * in something like xor mode, there is nothing we can
2852 * do - all replies will be rx'ed on same link causing slaves
2853 * to be unstable during low/no traffic periods
2854 */
3d632c3f 2855 if (IS_UP(slave->dev))
1da177e4 2856 bond_arp_send_all(bond, slave);
1da177e4
LT		 2857 }
2858
2859 if (do_failover) {
059fe7a5 2860 write_lock_bh(&bond->curr_slave_lock);
1da177e4
LT		 2861
2862 bond_select_active_slave(bond);
2863
059fe7a5 2864 write_unlock_bh(&bond->curr_slave_lock);
1da177e4
LT		 2865 }
2866
2867re_arm:
1b76b316
JV		 2868 if (bond->params.arp_interval)
2869 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
1da177e4
LT		 2870out:
2871 read_unlock(&bond->lock);
2872}
2873
2874/*
b2220cad
JV		 2875 * Called to inspect slaves for active-backup mode ARP monitor link state
2876 * changes. Sets new_link in slaves to specify what action should take
2877 * place for the slave. Returns 0 if no changes are found, >0 if changes
2878 * to link states must be committed.
2879 *
2880 * Called with bond->lock held for read.
1da177e4 2881 */
b2220cad 2882static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks)
1da177e4 2883{
1da177e4 2884 struct slave *slave;
b2220cad 2885 int i, commit = 0;
1da177e4 2886
b2220cad
JV		 2887 bond_for_each_slave(bond, slave, i) {
2888 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 2889
b2220cad
JV		 2890 if (slave->link != BOND_LINK_UP) {
2891 if (time_before_eq(jiffies, slave_last_rx(bond, slave) +
2892 delta_in_ticks)) {
2893 slave->new_link = BOND_LINK_UP;
2894 commit++;
2895 }
1da177e4 2896
b2220cad
JV		 2897 continue;
2898 }
1da177e4 2899
b2220cad
JV		 2900 /*
2901 * Give slaves 2*delta after being enslaved or made
2902 * active. This avoids bouncing, as the last receive
2903 * times need a full ARP monitor cycle to be updated.
2904 */
2905 if (!time_after_eq(jiffies, slave->jiffies +
2906 2 * delta_in_ticks))
2907 continue;
2908
2909 /*
2910 * Backup slave is down if:
2911 * - No current_arp_slave AND
2912 * - more than 3*delta since last receive AND
2913 * - the bond has an IP address
2914 *
2915 * Note: a non-null current_arp_slave indicates
2916 * the curr_active_slave went down and we are
2917 * searching for a new one; under this condition
2918 * we only take the curr_active_slave down - this
2919 * gives each slave a chance to tx/rx traffic
2920 * before being taken out
2921 */
2922 if (slave->state == BOND_STATE_BACKUP &&
2923 !bond->current_arp_slave &&
2924 time_after(jiffies, slave_last_rx(bond, slave) +
2925 3 * delta_in_ticks)) {
2926 slave->new_link = BOND_LINK_DOWN;
2927 commit++;
2928 }
2929
2930 /*
2931 * Active slave is down if:
2932 * - more than 2*delta since transmitting OR
2933 * - (more than 2*delta since receive AND
2934 * the bond has an IP address)
2935 */
2936 if ((slave->state == BOND_STATE_ACTIVE) &&
9d21493b 2937 (time_after_eq(jiffies, dev_trans_start(slave->dev) +
b2220cad
JV		 2938 2 * delta_in_ticks) ||
2939 (time_after_eq(jiffies, slave_last_rx(bond, slave)
2940 + 2 * delta_in_ticks)))) {
2941 slave->new_link = BOND_LINK_DOWN;
2942 commit++;
2943 }
1da177e4
LT		 2944 }
2945
b2220cad
JV		 2946 return commit;
2947}
1da177e4 2948
b2220cad
JV		 2949/*
2950 * Called to commit link state changes noted by inspection step of
2951 * active-backup mode ARP monitor.
2952 *
2953 * Called with RTNL and bond->lock for read.
2954 */
2955static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks)
2956{
2957 struct slave *slave;
2958 int i;
1da177e4 2959
b2220cad
JV		 2960 bond_for_each_slave(bond, slave, i) {
2961 switch (slave->new_link) {
2962 case BOND_LINK_NOCHANGE:
2963 continue;
ff59c456 2964
b2220cad 2965 case BOND_LINK_UP:
b9f60253
JP		 2966 if ((!bond->curr_active_slave &&
2967 time_before_eq(jiffies,
2968 dev_trans_start(slave->dev) +
2969 delta_in_ticks)) ||
2970 bond->curr_active_slave != slave) {
b2220cad 2971 slave->link = BOND_LINK_UP;
b2220cad
JV		 2972 bond->current_arp_slave = NULL;
2973
a4aee5c8 2974 pr_info("%s: link status definitely up for interface %s.\n",
b9f60253 2975 bond->dev->name, slave->dev->name);
b2220cad 2976
b9f60253
JP		 2977 if (!bond->curr_active_slave ||
2978 (slave == bond->primary_slave))
2979 goto do_failover;
1da177e4 2980
b9f60253 2981 }
1da177e4 2982
b9f60253 2983 continue;
1da177e4 2984
b2220cad
JV		 2985 case BOND_LINK_DOWN:
2986 if (slave->link_failure_count < UINT_MAX)
2987 slave->link_failure_count++;
2988
2989 slave->link = BOND_LINK_DOWN;
b9f60253 2990 bond_set_slave_inactive_flags(slave);
b2220cad 2991
a4aee5c8 2992 pr_info("%s: link status definitely down for interface %s, disabling it\n",
b9f60253 2993 bond->dev->name, slave->dev->name);
b2220cad 2994
b9f60253 2995 if (slave == bond->curr_active_slave) {
b2220cad 2996 bond->current_arp_slave = NULL;
b9f60253 2997 goto do_failover;
1da177e4 2998 }
b9f60253
JP		 2999
3000 continue;
b2220cad
JV		 3001
3002 default:
a4aee5c8 3003 pr_err("%s: impossible: new_link %d on slave %s\n",
b2220cad
JV		 3004 bond->dev->name, slave->new_link,
3005 slave->dev->name);
b9f60253 3006 continue;
1da177e4 3007 }
1da177e4 3008
b9f60253
JP		 3009do_failover:
3010 ASSERT_RTNL();
b2220cad 3011 write_lock_bh(&bond->curr_slave_lock);
b9f60253 3012 bond_select_active_slave(bond);
b2220cad
JV		 3013 write_unlock_bh(&bond->curr_slave_lock);
3014 }
1da177e4 3015
b2220cad
JV		 3016 bond_set_carrier(bond);
3017}
1da177e4 3018
b2220cad
JV		 3019/*
3020 * Send ARP probes for active-backup mode ARP monitor.
3021 *
3022 * Called with bond->lock held for read.
3023 */
3024static void bond_ab_arp_probe(struct bonding *bond)
3025{
3026 struct slave *slave;
3027 int i;
1da177e4 3028
b2220cad 3029 read_lock(&bond->curr_slave_lock);
1da177e4 3030
b2220cad 3031 if (bond->current_arp_slave && bond->curr_active_slave)
a4aee5c8
JP		 3032 pr_info("PROBE: c_arp %s && cas %s BAD\n",
3033 bond->current_arp_slave->dev->name,
3034 bond->curr_active_slave->dev->name);
1da177e4 3035
b2220cad
JV		 3036 if (bond->curr_active_slave) {
3037 bond_arp_send_all(bond, bond->curr_active_slave);
3038 read_unlock(&bond->curr_slave_lock);
3039 return;
3040 }
1da177e4 3041
b2220cad 3042 read_unlock(&bond->curr_slave_lock);
059fe7a5 3043
b2220cad
JV		 3044 /* if we don't have a curr_active_slave, search for the next available
3045 * backup slave from the current_arp_slave and make it the candidate
3046 * for becoming the curr_active_slave
3047 */
1da177e4 3048
b2220cad
JV		 3049 if (!bond->current_arp_slave) {
3050 bond->current_arp_slave = bond->first_slave;
3051 if (!bond->current_arp_slave)
3052 return;
3053 }
1da177e4 3054
b2220cad 3055 bond_set_slave_inactive_flags(bond->current_arp_slave);
059fe7a5 3056
b2220cad
JV		 3057 /* search for next candidate */
3058 bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) {
3059 if (IS_UP(slave->dev)) {
3060 slave->link = BOND_LINK_BACK;
3061 bond_set_slave_active_flags(slave);
3062 bond_arp_send_all(bond, slave);
1da177e4 3063 slave->jiffies = jiffies;
b2220cad
JV		 3064 bond->current_arp_slave = slave;
3065 break;
1da177e4
LT		 3066 }
3067
b2220cad
JV		 3068 /* if the link state is up at this point, we
3069 * mark it down - this can happen if we have
3070 * simultaneous link failures and
3071 * reselect_active_interface doesn't make this
3072 * one the current slave so it is still marked
3073 * up when it is actually down
1da177e4 3074 */
b2220cad
JV		 3075 if (slave->link == BOND_LINK_UP) {
3076 slave->link = BOND_LINK_DOWN;
3077 if (slave->link_failure_count < UINT_MAX)
3078 slave->link_failure_count++;
1da177e4 3079
b2220cad
JV		 3080 bond_set_slave_inactive_flags(slave);
3081
a4aee5c8
JP		 3082 pr_info("%s: backup interface %s is now down.\n",
3083 bond->dev->name, slave->dev->name);
1da177e4 3084 }
b2220cad
JV		 3085 }
3086}
1da177e4 3087
b2220cad
JV		 3088void bond_activebackup_arp_mon(struct work_struct *work)
3089{
3090 struct bonding *bond = container_of(work, struct bonding,
3091 arp_work.work);
3092 int delta_in_ticks;
1da177e4 3093
b2220cad 3094 read_lock(&bond->lock);
1da177e4 3095
b2220cad
JV		 3096 if (bond->kill_timers)
3097 goto out;
1da177e4 3098
b2220cad 3099 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
1da177e4 3100
b2220cad
JV		 3101 if (bond->slave_cnt == 0)
3102 goto re_arm;
3103
b59f9f74
JV		 3104 if (bond->send_grat_arp) {
3105 read_lock(&bond->curr_slave_lock);
3106 bond_send_gratuitous_arp(bond);
3107 read_unlock(&bond->curr_slave_lock);
3108 }
3109
305d552a
BH		 3110 if (bond->send_unsol_na) {
3111 read_lock(&bond->curr_slave_lock);
3112 bond_send_unsolicited_na(bond);
3113 read_unlock(&bond->curr_slave_lock);
3114 }
3115
b2220cad
JV		 3116 if (bond_ab_arp_inspect(bond, delta_in_ticks)) {
3117 read_unlock(&bond->lock);
3118 rtnl_lock();
3119 read_lock(&bond->lock);
3120
3121 bond_ab_arp_commit(bond, delta_in_ticks);
3122
3123 read_unlock(&bond->lock);
3124 rtnl_unlock();
3125 read_lock(&bond->lock);
1da177e4
LT		 3126 }
3127
b2220cad
JV		 3128 bond_ab_arp_probe(bond);
3129
1da177e4 3130re_arm:
3d632c3f 3131 if (bond->params.arp_interval)
1b76b316 3132 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
1da177e4
LT		 3133out:
3134 read_unlock(&bond->lock);
3135}
3136
3137/*------------------------------ proc/seq_file-------------------------------*/
3138
3139#ifdef CONFIG_PROC_FS
3140
1da177e4 3141static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos)
1f78d9f9
HE		 3142 __acquires(&dev_base_lock)
3143 __acquires(&bond->lock)
1da177e4
LT		 3144{
3145 struct bonding *bond = seq->private;
3146 loff_t off = 0;
3147 struct slave *slave;
3148 int i;
3149
3150 /* make sure the bond won't be taken away */
3151 read_lock(&dev_base_lock);
6603a6f2 3152 read_lock(&bond->lock);
1da177e4 3153
3d632c3f 3154 if (*pos == 0)
1da177e4 3155 return SEQ_START_TOKEN;
1da177e4
LT		 3156
3157 bond_for_each_slave(bond, slave, i) {
3d632c3f 3158 if (++off == *pos)
1da177e4 3159 return slave;
1da177e4
LT		 3160 }
3161
3162 return NULL;
3163}
3164
3165static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3166{
3167 struct bonding *bond = seq->private;
3168 struct slave *slave = v;
3169
3170 ++*pos;
3d632c3f 3171 if (v == SEQ_START_TOKEN)
1da177e4 3172 return bond->first_slave;
1da177e4
LT		 3173
3174 slave = slave->next;
3175
3176 return (slave == bond->first_slave) ? NULL : slave;
3177}
3178
3179static void bond_info_seq_stop(struct seq_file *seq, void *v)
1f78d9f9
HE		 3180 __releases(&bond->lock)
3181 __releases(&dev_base_lock)
1da177e4
LT		 3182{
3183 struct bonding *bond = seq->private;
3184
6603a6f2 3185 read_unlock(&bond->lock);
1da177e4
LT		 3186 read_unlock(&dev_base_lock);
3187}
3188
3189static void bond_info_show_master(struct seq_file *seq)
3190{
3191 struct bonding *bond = seq->private;
3192 struct slave *curr;
4756b02f 3193 int i;
1da177e4
LT		 3194
3195 read_lock(&bond->curr_slave_lock);
3196 curr = bond->curr_active_slave;
3197 read_unlock(&bond->curr_slave_lock);
3198
dd957c57 3199 seq_printf(seq, "Bonding Mode: %s",
1da177e4
LT		 3200 bond_mode_name(bond->params.mode));
3201
dd957c57
JV		 3202 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP &&
3203 bond->params.fail_over_mac)
3915c1e8
JV		 3204 seq_printf(seq, " (fail_over_mac %s)",
3205 fail_over_mac_tbl[bond->params.fail_over_mac].modename);
dd957c57
JV		 3206
3207 seq_printf(seq, "\n");
3208
c61b75ad
MW		 3209 if (bond->params.mode == BOND_MODE_XOR ||
3210 bond->params.mode == BOND_MODE_8023AD) {
3211 seq_printf(seq, "Transmit Hash Policy: %s (%d)\n",
3212 xmit_hashtype_tbl[bond->params.xmit_policy].modename,
3213 bond->params.xmit_policy);
3214 }
3215
1da177e4 3216 if (USES_PRIMARY(bond->params.mode)) {
a549952a 3217 seq_printf(seq, "Primary Slave: %s",
0f418b2a
MW		 3218 (bond->primary_slave) ?
3219 bond->primary_slave->dev->name : "None");
a549952a
JP		 3220 if (bond->primary_slave)
3221 seq_printf(seq, " (primary_reselect %s)",
3222 pri_reselect_tbl[bond->params.primary_reselect].modename);
1da177e4 3223
a549952a 3224 seq_printf(seq, "\nCurrently Active Slave: %s\n",
1da177e4
LT		 3225 (curr) ? curr->dev->name : "None");
3226 }
3227
ff59c456
JV		 3228 seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ?
3229 "up" : "down");
1da177e4
LT		 3230 seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon);
3231 seq_printf(seq, "Up Delay (ms): %d\n",
3232 bond->params.updelay * bond->params.miimon);
3233 seq_printf(seq, "Down Delay (ms): %d\n",
3234 bond->params.downdelay * bond->params.miimon);
3235
4756b02f
MW		 3236
3237 /* ARP information */
3d632c3f
SH		 3238 if (bond->params.arp_interval > 0) {
3239 int printed = 0;
4756b02f
MW		 3240 seq_printf(seq, "ARP Polling Interval (ms): %d\n",
3241 bond->params.arp_interval);
3242
3243 seq_printf(seq, "ARP IP target/s (n.n.n.n form):");
3244
3d632c3f 3245 for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) {
4756b02f 3246 if (!bond->params.arp_targets[i])
5a31bec0 3247 break;
4756b02f
MW		 3248 if (printed)
3249 seq_printf(seq, ",");
8cf14e38 3250 seq_printf(seq, " %pI4", &bond->params.arp_targets[i]);
4756b02f
MW		 3251 printed = 1;
3252 }
3253 seq_printf(seq, "\n");
3254 }
3255
1da177e4
LT		 3256 if (bond->params.mode == BOND_MODE_8023AD) {
3257 struct ad_info ad_info;
3258
3259 seq_puts(seq, "\n802.3ad info\n");
3260 seq_printf(seq, "LACP rate: %s\n",
3261 (bond->params.lacp_fast) ? "fast" : "slow");
fd989c83
JV		 3262 seq_printf(seq, "Aggregator selection policy (ad_select): %s\n",
3263 ad_select_tbl[bond->params.ad_select].modename);
1da177e4
LT		 3264
3265 if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
3266 seq_printf(seq, "bond %s has no active aggregator\n",
3267 bond->dev->name);
3268 } else {
3269 seq_printf(seq, "Active Aggregator Info:\n");
3270
3271 seq_printf(seq, "\tAggregator ID: %d\n",
3272 ad_info.aggregator_id);
3273 seq_printf(seq, "\tNumber of ports: %d\n",
3274 ad_info.ports);
3275 seq_printf(seq, "\tActor Key: %d\n",
3276 ad_info.actor_key);
3277 seq_printf(seq, "\tPartner Key: %d\n",
3278 ad_info.partner_key);
e174961c
JB		 3279 seq_printf(seq, "\tPartner Mac Address: %pM\n",
3280 ad_info.partner_system);
1da177e4
LT		 3281 }
3282 }
3283}
3284
3d632c3f
SH		 3285static void bond_info_show_slave(struct seq_file *seq,
3286 const struct slave *slave)
1da177e4
LT		 3287{
3288 struct bonding *bond = seq->private;
3289
3290 seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name);
3291 seq_printf(seq, "MII Status: %s\n",
3292 (slave->link == BOND_LINK_UP) ? "up" : "down");
65509645 3293 seq_printf(seq, "Link Failure Count: %u\n",
1da177e4
LT		 3294 slave->link_failure_count);
3295
e174961c 3296 seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr);
1da177e4
LT		 3297
3298 if (bond->params.mode == BOND_MODE_8023AD) {
3299 const struct aggregator *agg
3300 = SLAVE_AD_INFO(slave).port.aggregator;
3301
3d632c3f 3302 if (agg)
1da177e4
LT		 3303 seq_printf(seq, "Aggregator ID: %d\n",
3304 agg->aggregator_identifier);
3d632c3f 3305 else
1da177e4 3306 seq_puts(seq, "Aggregator ID: N/A\n");
1da177e4 3307 }
bb1d9123 3308 seq_printf(seq, "Slave queue ID: %d\n", slave->queue_id);
1da177e4
LT		 3309}
3310
3311static int bond_info_seq_show(struct seq_file *seq, void *v)
3312{
3313 if (v == SEQ_START_TOKEN) {
3314 seq_printf(seq, "%s\n", version);
3315 bond_info_show_master(seq);
3d632c3f 3316 } else
1da177e4 3317 bond_info_show_slave(seq, v);
1da177e4
LT		 3318
3319 return 0;
3320}
3321
4101dec9 3322static const struct seq_operations bond_info_seq_ops = {
1da177e4
LT		 3323 .start = bond_info_seq_start,
3324 .next = bond_info_seq_next,
3325 .stop = bond_info_seq_stop,
3326 .show = bond_info_seq_show,
3327};
3328
3329static int bond_info_open(struct inode *inode, struct file *file)
3330{
3331 struct seq_file *seq;
3332 struct proc_dir_entry *proc;
3333 int res;
3334
3335 res = seq_open(file, &bond_info_seq_ops);
3336 if (!res) {
3337 /* recover the pointer buried in proc_dir_entry data */
3338 seq = file->private_data;
3339 proc = PDE(inode);
3340 seq->private = proc->data;
3341 }
3342
3343 return res;
3344}
3345
d54b1fdb 3346static const struct file_operations bond_info_fops = {
1da177e4
LT		 3347 .owner = THIS_MODULE,
3348 .open = bond_info_open,
3349 .read = seq_read,
3350 .llseek = seq_lseek,
3351 .release = seq_release,
3352};
3353
38fc0026 3354static void bond_create_proc_entry(struct bonding *bond)
1da177e4
LT		 3355{
3356 struct net_device *bond_dev = bond->dev;
ec87fd3b 3357 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
1da177e4 3358
ec87fd3b 3359 if (bn->proc_dir) {
a95609cb 3360 bond->proc_entry = proc_create_data(bond_dev->name,
ec87fd3b 3361 S_IRUGO, bn->proc_dir,
a95609cb 3362 &bond_info_fops, bond);
3d632c3f 3363 if (bond->proc_entry == NULL)
a4aee5c8
JP		 3364 pr_warning("Warning: Cannot create /proc/net/%s/%s\n",
3365 DRV_NAME, bond_dev->name);
3d632c3f 3366 else
1da177e4 3367 memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ);
1da177e4 3368 }
1da177e4
LT		 3369}
3370
3371static void bond_remove_proc_entry(struct bonding *bond)
3372{
ec87fd3b
EB		 3373 struct net_device *bond_dev = bond->dev;
3374 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
3375
3376 if (bn->proc_dir && bond->proc_entry) {
3377 remove_proc_entry(bond->proc_file_name, bn->proc_dir);
1da177e4
LT		 3378 memset(bond->proc_file_name, 0, IFNAMSIZ);
3379 bond->proc_entry = NULL;
3380 }
3381}
3382
3383/* Create the bonding directory under /proc/net, if doesn't exist yet.
3384 * Caller must hold rtnl_lock.
3385 */
2c8c1e72 3386static void __net_init bond_create_proc_dir(struct bond_net *bn)
1da177e4 3387{
ec87fd3b
EB		 3388 if (!bn->proc_dir) {
3389 bn->proc_dir = proc_mkdir(DRV_NAME, bn->net->proc_net);
3390 if (!bn->proc_dir)
a4aee5c8
JP		 3391 pr_warning("Warning: cannot create /proc/net/%s\n",
3392 DRV_NAME);
1da177e4
LT		 3393 }
3394}
3395
3396/* Destroy the bonding directory under /proc/net, if empty.
3397 * Caller must hold rtnl_lock.
3398 */
2c8c1e72 3399static void __net_exit bond_destroy_proc_dir(struct bond_net *bn)
1da177e4 3400{
ec87fd3b
EB		 3401 if (bn->proc_dir) {
3402 remove_proc_entry(DRV_NAME, bn->net->proc_net);
3403 bn->proc_dir = NULL;
1da177e4
LT		 3404 }
3405}
aee64faf
JP		 3406
3407#else /* !CONFIG_PROC_FS */
3408
38fc0026 3409static void bond_create_proc_entry(struct bonding *bond)
aee64faf
JP		 3410{
3411}
3412
3413static void bond_remove_proc_entry(struct bonding *bond)
3414{
3415}
3416
2c8c1e72 3417static inline void bond_create_proc_dir(struct bond_net *bn)
aee64faf
JP		 3418{
3419}
3420
2c8c1e72 3421static inline void bond_destroy_proc_dir(struct bond_net *bn)
aee64faf
JP		 3422{
3423}
3424
1da177e4
LT		 3425#endif /* CONFIG_PROC_FS */
3426
aee64faf 3427
1da177e4
LT		 3428/*-------------------------- netdev event handling --------------------------*/
3429
3430/*
3431 * Change device name
3432 */
3433static int bond_event_changename(struct bonding *bond)
3434{
1da177e4
LT		 3435 bond_remove_proc_entry(bond);
3436 bond_create_proc_entry(bond);
7e083840 3437
1da177e4
LT		 3438 return NOTIFY_DONE;
3439}
3440
3d632c3f
SH		 3441static int bond_master_netdev_event(unsigned long event,
3442 struct net_device *bond_dev)
1da177e4 3443{
454d7c9b 3444 struct bonding *event_bond = netdev_priv(bond_dev);
1da177e4
LT		 3445
3446 switch (event) {
3447 case NETDEV_CHANGENAME:
3448 return bond_event_changename(event_bond);
1da177e4
LT		 3449 default:
3450 break;
3451 }
3452
3453 return NOTIFY_DONE;
3454}
3455
3d632c3f
SH		 3456static int bond_slave_netdev_event(unsigned long event,
3457 struct net_device *slave_dev)
1da177e4
LT		 3458{
3459 struct net_device *bond_dev = slave_dev->master;
454d7c9b 3460 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3461
3462 switch (event) {
3463 case NETDEV_UNREGISTER:
3464 if (bond_dev) {
1284cd3a
JV		 3465 if (bond->setup_by_slave)
3466 bond_release_and_destroy(bond_dev, slave_dev);
3467 else
3468 bond_release(bond_dev, slave_dev);
1da177e4
LT		 3469 }
3470 break;
3471 case NETDEV_CHANGE:
17d04500
JV		 3472 if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) {
3473 struct slave *slave;
3474
3475 slave = bond_get_slave_by_dev(bond, slave_dev);
3476 if (slave) {
3477 u16 old_speed = slave->speed;
3478 u16 old_duplex = slave->duplex;
3479
3480 bond_update_speed_duplex(slave);
3481
3482 if (bond_is_lb(bond))
3483 break;
3484
3485 if (old_speed != slave->speed)
3486 bond_3ad_adapter_speed_changed(slave);
3487 if (old_duplex != slave->duplex)
3488 bond_3ad_adapter_duplex_changed(slave);
3489 }
3490 }
3491
1da177e4
LT		 3492 break;
3493 case NETDEV_DOWN:
3494 /*
3495 * ... Or is it this?
3496 */
3497 break;
3498 case NETDEV_CHANGEMTU:
3499 /*
3500 * TODO: Should slaves be allowed to
3501 * independently alter their MTU? For
3502 * an active-backup bond, slaves need
3503 * not be the same type of device, so
3504 * MTUs may vary. For other modes,
3505 * slaves arguably should have the
3506 * same MTUs. To do this, we'd need to
3507 * take over the slave's change_mtu
3508 * function for the duration of their
3509 * servitude.
3510 */
3511 break;
3512 case NETDEV_CHANGENAME:
3513 /*
3514 * TODO: handle changing the primary's name
3515 */
3516 break;
8531c5ff
AK		 3517 case NETDEV_FEAT_CHANGE:
3518 bond_compute_features(bond);
3519 break;
1da177e4
LT		 3520 default:
3521 break;
3522 }
3523
3524 return NOTIFY_DONE;
3525}
3526
3527/*
3528 * bond_netdev_event: handle netdev notifier chain events.
3529 *
3530 * This function receives events for the netdev chain. The caller (an
e041c683 3531 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
1da177e4
LT		 3532 * locks for us to safely manipulate the slave devices (RTNL lock,
3533 * dev_probe_lock).
3534 */
3d632c3f
SH		 3535static int bond_netdev_event(struct notifier_block *this,
3536 unsigned long event, void *ptr)
1da177e4
LT		 3537{
3538 struct net_device *event_dev = (struct net_device *)ptr;
3539
5a03cdb7 3540 pr_debug("event_dev: %s, event: %lx\n",
a4aee5c8
JP		 3541 event_dev ? event_dev->name : "None",
3542 event);
1da177e4 3543
0b680e75
JV		 3544 if (!(event_dev->priv_flags & IFF_BONDING))
3545 return NOTIFY_DONE;
3546
1da177e4 3547 if (event_dev->flags & IFF_MASTER) {
5a03cdb7 3548 pr_debug("IFF_MASTER\n");
1da177e4
LT		 3549 return bond_master_netdev_event(event, event_dev);
3550 }
3551
3552 if (event_dev->flags & IFF_SLAVE) {
5a03cdb7 3553 pr_debug("IFF_SLAVE\n");
1da177e4
LT		 3554 return bond_slave_netdev_event(event, event_dev);
3555 }
3556
3557 return NOTIFY_DONE;
3558}
3559
c3ade5ca
JV		 3560/*
3561 * bond_inetaddr_event: handle inetaddr notifier chain events.
3562 *
3563 * We keep track of device IPs primarily to use as source addresses in
3564 * ARP monitor probes (rather than spewing out broadcasts all the time).
3565 *
3566 * We track one IP for the main device (if it has one), plus one per VLAN.
3567 */
3568static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
3569{
3570 struct in_ifaddr *ifa = ptr;
3571 struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev;
ec87fd3b 3572 struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id);
0883beca
PE		 3573 struct bonding *bond;
3574 struct vlan_entry *vlan;
c3ade5ca 3575
ec87fd3b 3576 list_for_each_entry(bond, &bn->dev_list, bond_list) {
c3ade5ca
JV		 3577 if (bond->dev == event_dev) {
3578 switch (event) {
3579 case NETDEV_UP:
3580 bond->master_ip = ifa->ifa_local;
3581 return NOTIFY_OK;
3582 case NETDEV_DOWN:
3583 bond->master_ip = bond_glean_dev_ip(bond->dev);
3584 return NOTIFY_OK;
3585 default:
3586 return NOTIFY_DONE;
3587 }
3588 }
3589
0883beca 3590 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
5c15bdec 3591 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
c3ade5ca
JV		 3592 if (vlan_dev == event_dev) {
3593 switch (event) {
3594 case NETDEV_UP:
3595 vlan->vlan_ip = ifa->ifa_local;
3596 return NOTIFY_OK;
3597 case NETDEV_DOWN:
3598 vlan->vlan_ip =
3599 bond_glean_dev_ip(vlan_dev);
3600 return NOTIFY_OK;
3601 default:
3602 return NOTIFY_DONE;
3603 }
3604 }
3605 }
3606 }
3607 return NOTIFY_DONE;
3608}
3609
1da177e4
LT		 3610static struct notifier_block bond_netdev_notifier = {
3611 .notifier_call = bond_netdev_event,
3612};
3613
c3ade5ca
JV		 3614static struct notifier_block bond_inetaddr_notifier = {
3615 .notifier_call = bond_inetaddr_event,
3616};
3617
1da177e4
LT		 3618/*-------------------------- Packet type handling ---------------------------*/
3619
3620/* register to receive lacpdus on a bond */
3621static void bond_register_lacpdu(struct bonding *bond)
3622{
3623 struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type);
3624
3625 /* initialize packet type */
3626 pk_type->type = PKT_TYPE_LACPDU;
3627 pk_type->dev = bond->dev;
3628 pk_type->func = bond_3ad_lacpdu_recv;
3629
3630 dev_add_pack(pk_type);
3631}
3632
3633/* unregister to receive lacpdus on a bond */
3634static void bond_unregister_lacpdu(struct bonding *bond)
3635{
3636 dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type));
3637}
3638
f5b2b966
JV		 3639void bond_register_arp(struct bonding *bond)
3640{
3641 struct packet_type *pt = &bond->arp_mon_pt;
3642
c4f283b1
JV		 3643 if (pt->type)
3644 return;
3645
f5b2b966 3646 pt->type = htons(ETH_P_ARP);
e245cb71 3647 pt->dev = bond->dev;
f5b2b966
JV		 3648 pt->func = bond_arp_rcv;
3649 dev_add_pack(pt);
3650}
3651
3652void bond_unregister_arp(struct bonding *bond)
3653{
c4f283b1
JV		 3654 struct packet_type *pt = &bond->arp_mon_pt;
3655
3656 dev_remove_pack(pt);
3657 pt->type = 0;
f5b2b966
JV		 3658}
3659
169a3e66
JV		 3660/*---------------------------- Hashing Policies -----------------------------*/
3661
6f6652be
JV		 3662/*
3663 * Hash for the output device based upon layer 2 and layer 3 data. If
3664 * the packet is not IP mimic bond_xmit_hash_policy_l2()
3665 */
a361c83c 3666static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count)
6f6652be
JV		 3667{
3668 struct ethhdr *data = (struct ethhdr *)skb->data;
3669 struct iphdr *iph = ip_hdr(skb);
3670
f14c4e4e 3671 if (skb->protocol == htons(ETH_P_IP)) {
6f6652be 3672 return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^
d3da6831 3673 (data->h_dest[5] ^ data->h_source[5])) % count;
6f6652be
JV		 3674 }
3675
d3da6831 3676 return (data->h_dest[5] ^ data->h_source[5]) % count;
6f6652be
JV		 3677}
3678
169a3e66 3679/*
59c51591 3680 * Hash for the output device based upon layer 3 and layer 4 data. If
169a3e66
JV		 3681 * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is
3682 * altogether not IP, mimic bond_xmit_hash_policy_l2()
3683 */
a361c83c 3684static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count)
169a3e66
JV		 3685{
3686 struct ethhdr *data = (struct ethhdr *)skb->data;
eddc9ec5 3687 struct iphdr *iph = ip_hdr(skb);
d3bb52b0 3688 __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl);
169a3e66
JV		 3689 int layer4_xor = 0;
3690
f14c4e4e
BH		 3691 if (skb->protocol == htons(ETH_P_IP)) {
3692 if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) &&
169a3e66
JV		 3693 (iph->protocol == IPPROTO_TCP ||
3694 iph->protocol == IPPROTO_UDP)) {
d3bb52b0 3695 layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1)));
169a3e66
JV		 3696 }
3697 return (layer4_xor ^
3698 ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count;
3699
3700 }
3701
d3da6831 3702 return (data->h_dest[5] ^ data->h_source[5]) % count;
169a3e66
JV		 3703}
3704
3705/*
3706 * Hash for the output device based upon layer 2 data
3707 */
a361c83c 3708static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count)
169a3e66
JV		 3709{
3710 struct ethhdr *data = (struct ethhdr *)skb->data;
3711
d3da6831 3712 return (data->h_dest[5] ^ data->h_source[5]) % count;
169a3e66
JV		 3713}
3714
1da177e4
LT		 3715/*-------------------------- Device entry points ----------------------------*/
3716
3717static int bond_open(struct net_device *bond_dev)
3718{
454d7c9b 3719 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3720
3721 bond->kill_timers = 0;
3722
58402054 3723 if (bond_is_lb(bond)) {
1da177e4
LT		 3724 /* bond_alb_initialize must be called before the timer
3725 * is started.
3726 */
3727 if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) {
3728 /* something went wrong - fail the open operation */
b473946a 3729 return -ENOMEM;
1da177e4
LT		 3730 }
3731
1b76b316
JV		 3732 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3733 queue_delayed_work(bond->wq, &bond->alb_work, 0);
1da177e4
LT		 3734 }
3735
3736 if (bond->params.miimon) { /* link check interval, in milliseconds. */
1b76b316
JV		 3737 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
3738 queue_delayed_work(bond->wq, &bond->mii_work, 0);
1da177e4
LT		 3739 }
3740
3741 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316
JV		 3742 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3743 INIT_DELAYED_WORK(&bond->arp_work,
3744 bond_activebackup_arp_mon);
3745 else
3746 INIT_DELAYED_WORK(&bond->arp_work,
3747 bond_loadbalance_arp_mon);
3748
3749 queue_delayed_work(bond->wq, &bond->arp_work, 0);
f5b2b966
JV		 3750 if (bond->params.arp_validate)
3751 bond_register_arp(bond);
1da177e4
LT		 3752 }
3753
3754 if (bond->params.mode == BOND_MODE_8023AD) {
a40745f5 3755 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
1b76b316 3756 queue_delayed_work(bond->wq, &bond->ad_work, 0);
1da177e4
LT		 3757 /* register to receive LACPDUs */
3758 bond_register_lacpdu(bond);
fd989c83 3759 bond_3ad_initiate_agg_selection(bond, 1);
1da177e4
LT		 3760 }
3761
3762 return 0;
3763}
3764
3765static int bond_close(struct net_device *bond_dev)
3766{
454d7c9b 3767 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3768
3769 if (bond->params.mode == BOND_MODE_8023AD) {
3770 /* Unregister the receive of LACPDUs */
3771 bond_unregister_lacpdu(bond);
3772 }
3773
f5b2b966
JV		 3774 if (bond->params.arp_validate)
3775 bond_unregister_arp(bond);
3776
1da177e4
LT		 3777 write_lock_bh(&bond->lock);
3778
b59f9f74 3779 bond->send_grat_arp = 0;
305d552a 3780 bond->send_unsol_na = 0;
1da177e4
LT		 3781
3782 /* signal timers not to re-arm */
3783 bond->kill_timers = 1;
3784
3785 write_unlock_bh(&bond->lock);
3786
1da177e4 3787 if (bond->params.miimon) { /* link check interval, in milliseconds. */
1b76b316 3788 cancel_delayed_work(&bond->mii_work);
1da177e4
LT		 3789 }
3790
3791 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316 3792 cancel_delayed_work(&bond->arp_work);
1da177e4
LT		 3793 }
3794
3795 switch (bond->params.mode) {
3796 case BOND_MODE_8023AD:
1b76b316 3797 cancel_delayed_work(&bond->ad_work);
1da177e4
LT		 3798 break;
3799 case BOND_MODE_TLB:
3800 case BOND_MODE_ALB:
1b76b316 3801 cancel_delayed_work(&bond->alb_work);
1da177e4
LT		 3802 break;
3803 default:
3804 break;
3805 }
3806
1da177e4 3807
58402054 3808 if (bond_is_lb(bond)) {
1da177e4
LT		 3809 /* Must be called only after all
3810 * slaves have been released
3811 */
3812 bond_alb_deinitialize(bond);
3813 }
3814
3815 return 0;
3816}
3817
28172739
ED		 3818static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
3819 struct rtnl_link_stats64 *stats)
1da177e4 3820{
454d7c9b 3821 struct bonding *bond = netdev_priv(bond_dev);
28172739 3822 struct rtnl_link_stats64 temp;
1da177e4
LT		 3823 struct slave *slave;
3824 int i;
3825
28172739 3826 memset(stats, 0, sizeof(*stats));
1da177e4
LT		 3827
3828 read_lock_bh(&bond->lock);
3829
3830 bond_for_each_slave(bond, slave, i) {
be1f3c2c 3831 const struct rtnl_link_stats64 *sstats =
28172739
ED		 3832 dev_get_stats(slave->dev, &temp);
3833
3834 stats->rx_packets += sstats->rx_packets;
3835 stats->rx_bytes += sstats->rx_bytes;
3836 stats->rx_errors += sstats->rx_errors;
3837 stats->rx_dropped += sstats->rx_dropped;
3838
3839 stats->tx_packets += sstats->tx_packets;
3840 stats->tx_bytes += sstats->tx_bytes;
3841 stats->tx_errors += sstats->tx_errors;
3842 stats->tx_dropped += sstats->tx_dropped;
3843
3844 stats->multicast += sstats->multicast;
3845 stats->collisions += sstats->collisions;
3846
3847 stats->rx_length_errors += sstats->rx_length_errors;
3848 stats->rx_over_errors += sstats->rx_over_errors;
3849 stats->rx_crc_errors += sstats->rx_crc_errors;
3850 stats->rx_frame_errors += sstats->rx_frame_errors;
3851 stats->rx_fifo_errors += sstats->rx_fifo_errors;
3852 stats->rx_missed_errors += sstats->rx_missed_errors;
3853
3854 stats->tx_aborted_errors += sstats->tx_aborted_errors;
3855 stats->tx_carrier_errors += sstats->tx_carrier_errors;
3856 stats->tx_fifo_errors += sstats->tx_fifo_errors;
3857 stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
3858 stats->tx_window_errors += sstats->tx_window_errors;
2439f9eb
AG		 3859 }
3860
1da177e4
LT		 3861 read_unlock_bh(&bond->lock);
3862
3863 return stats;
3864}
3865
3866static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3867{
3868 struct net_device *slave_dev = NULL;
3869 struct ifbond k_binfo;
3870 struct ifbond __user *u_binfo = NULL;
3871 struct ifslave k_sinfo;
3872 struct ifslave __user *u_sinfo = NULL;
3873 struct mii_ioctl_data *mii = NULL;
1da177e4
LT		 3874 int res = 0;
3875
a4aee5c8 3876 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd);
1da177e4
LT		 3877
3878 switch (cmd) {
1da177e4
LT		 3879 case SIOCGMIIPHY:
3880 mii = if_mii(ifr);
3d632c3f 3881 if (!mii)
1da177e4 3882 return -EINVAL;
3d632c3f 3883
1da177e4
LT		 3884 mii->phy_id = 0;
3885 /* Fall Through */
3886 case SIOCGMIIREG:
3887 /*
3888 * We do this again just in case we were called by SIOCGMIIREG
3889 * instead of SIOCGMIIPHY.
3890 */
3891 mii = if_mii(ifr);
3d632c3f 3892 if (!mii)
1da177e4 3893 return -EINVAL;
3d632c3f 3894
1da177e4
LT		 3895
3896 if (mii->reg_num == 1) {
454d7c9b 3897 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3898 mii->val_out = 0;
6603a6f2 3899 read_lock(&bond->lock);
1da177e4 3900 read_lock(&bond->curr_slave_lock);
3d632c3f 3901 if (netif_carrier_ok(bond->dev))
1da177e4 3902 mii->val_out = BMSR_LSTATUS;
3d632c3f 3903
1da177e4 3904 read_unlock(&bond->curr_slave_lock);
6603a6f2 3905 read_unlock(&bond->lock);
1da177e4
LT		 3906 }
3907
3908 return 0;
3909 case BOND_INFO_QUERY_OLD:
3910 case SIOCBONDINFOQUERY:
3911 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3912
3d632c3f 3913 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
1da177e4 3914 return -EFAULT;
1da177e4
LT		 3915
3916 res = bond_info_query(bond_dev, &k_binfo);
3d632c3f
SH		 3917 if (res == 0 &&
3918 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3919 return -EFAULT;
1da177e4
LT		 3920
3921 return res;
3922 case BOND_SLAVE_INFO_QUERY_OLD:
3923 case SIOCBONDSLAVEINFOQUERY:
3924 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3925
3d632c3f 3926 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
1da177e4 3927 return -EFAULT;
1da177e4
LT		 3928
3929 res = bond_slave_info_query(bond_dev, &k_sinfo);
3d632c3f
SH		 3930 if (res == 0 &&
3931 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3932 return -EFAULT;
1da177e4
LT		 3933
3934 return res;
3935 default:
3936 /* Go on */
3937 break;
3938 }
3939
3d632c3f 3940 if (!capable(CAP_NET_ADMIN))
1da177e4 3941 return -EPERM;
1da177e4 3942
ec87fd3b 3943 slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave);
1da177e4 3944
a4aee5c8 3945 pr_debug("slave_dev=%p:\n", slave_dev);
1da177e4 3946
3d632c3f 3947 if (!slave_dev)
1da177e4 3948 res = -ENODEV;
3d632c3f 3949 else {
a4aee5c8 3950 pr_debug("slave_dev->name=%s:\n", slave_dev->name);
1da177e4
LT		 3951 switch (cmd) {
3952 case BOND_ENSLAVE_OLD:
3953 case SIOCBONDENSLAVE:
3954 res = bond_enslave(bond_dev, slave_dev);
3955 break;
3956 case BOND_RELEASE_OLD:
3957 case SIOCBONDRELEASE:
3958 res = bond_release(bond_dev, slave_dev);
3959 break;
3960 case BOND_SETHWADDR_OLD:
3961 case SIOCBONDSETHWADDR:
3962 res = bond_sethwaddr(bond_dev, slave_dev);
3963 break;
3964 case BOND_CHANGE_ACTIVE_OLD:
3965 case SIOCBONDCHANGEACTIVE:
3966 res = bond_ioctl_change_active(bond_dev, slave_dev);
3967 break;
3968 default:
3969 res = -EOPNOTSUPP;
3970 }
3971
3972 dev_put(slave_dev);
3973 }
3974
1da177e4
LT		 3975 return res;
3976}
3977
22bedad3
JP		 3978static bool bond_addr_in_mc_list(unsigned char *addr,
3979 struct netdev_hw_addr_list *list,
3980 int addrlen)
3981{
3982 struct netdev_hw_addr *ha;
3983
3984 netdev_hw_addr_list_for_each(ha, list)
3985 if (!memcmp(ha->addr, addr, addrlen))
3986 return true;
3987
3988 return false;
3989}
3990
1da177e4
LT		 3991static void bond_set_multicast_list(struct net_device *bond_dev)
3992{
454d7c9b 3993 struct bonding *bond = netdev_priv(bond_dev);
22bedad3
JP		 3994 struct netdev_hw_addr *ha;
3995 bool found;
1da177e4 3996
1da177e4
LT		 3997 /*
3998 * Do promisc before checking multicast_mode
3999 */
3d632c3f 4000 if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC))
7e1a1ac1
WC		 4001 /*
4002 * FIXME: Need to handle the error when one of the multi-slaves
4003 * encounters error.
4004 */
1da177e4 4005 bond_set_promiscuity(bond, 1);
1da177e4 4006
3d632c3f
SH		 4007
4008 if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC))
1da177e4 4009 bond_set_promiscuity(bond, -1);
3d632c3f 4010
1da177e4
LT		 4011
4012 /* set allmulti flag to slaves */
3d632c3f 4013 if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI))
7e1a1ac1
WC		 4014 /*
4015 * FIXME: Need to handle the error when one of the multi-slaves
4016 * encounters error.
4017 */
1da177e4 4018 bond_set_allmulti(bond, 1);
1da177e4 4019
3d632c3f
SH		 4020
4021 if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI))
1da177e4 4022 bond_set_allmulti(bond, -1);
3d632c3f 4023
1da177e4 4024
80ee5ad2
JV		 4025 read_lock(&bond->lock);
4026
1da177e4
LT		 4027 bond->flags = bond_dev->flags;
4028
4029 /* looking for addresses to add to slaves' mc list */
22bedad3
JP		 4030 netdev_for_each_mc_addr(ha, bond_dev) {
4031 found = bond_addr_in_mc_list(ha->addr, &bond->mc_list,
4032 bond_dev->addr_len);
4033 if (!found)
4034 bond_mc_add(bond, ha->addr);
1da177e4
LT		 4035 }
4036
4037 /* looking for addresses to delete from slaves' list */
22bedad3
JP		 4038 netdev_hw_addr_list_for_each(ha, &bond->mc_list) {
4039 found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc,
4040 bond_dev->addr_len);
4041 if (!found)
4042 bond_mc_del(bond, ha->addr);
1da177e4
LT		 4043 }
4044
4045 /* save master's multicast list */
22bedad3
JP		 4046 __hw_addr_flush(&bond->mc_list);
4047 __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc,
4048 bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST);
1da177e4 4049
80ee5ad2 4050 read_unlock(&bond->lock);
1da177e4
LT		 4051}
4052
00829823
SH		 4053static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms)
4054{
4055 struct bonding *bond = netdev_priv(dev);
4056 struct slave *slave = bond->first_slave;
4057
4058 if (slave) {
4059 const struct net_device_ops *slave_ops
4060 = slave->dev->netdev_ops;
4061 if (slave_ops->ndo_neigh_setup)
72e2240f 4062 return slave_ops->ndo_neigh_setup(slave->dev, parms);
00829823
SH		 4063 }
4064 return 0;
4065}
4066
1da177e4
LT		 4067/*
4068 * Change the MTU of all of a master's slaves to match the master
4069 */
4070static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
4071{
454d7c9b 4072 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4073 struct slave *slave, *stop_at;
4074 int res = 0;
4075 int i;
4076
5a03cdb7 4077 pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond,
a4aee5c8 4078 (bond_dev ? bond_dev->name : "None"), new_mtu);
1da177e4
LT		 4079
4080 /* Can't hold bond->lock with bh disabled here since
4081 * some base drivers panic. On the other hand we can't
4082 * hold bond->lock without bh disabled because we'll
4083 * deadlock. The only solution is to rely on the fact
4084 * that we're under rtnl_lock here, and the slaves
4085 * list won't change. This doesn't solve the problem
4086 * of setting the slave's MTU while it is
4087 * transmitting, but the assumption is that the base
4088 * driver can handle that.
4089 *
4090 * TODO: figure out a way to safely iterate the slaves
4091 * list, but without holding a lock around the actual
4092 * call to the base driver.
4093 */
4094
4095 bond_for_each_slave(bond, slave, i) {
a4aee5c8
JP		 4096 pr_debug("s %p s->p %p c_m %p\n",
4097 slave,
4098 slave->prev,
4099 slave->dev->netdev_ops->ndo_change_mtu);
e944ef79 4100
1da177e4
LT		 4101 res = dev_set_mtu(slave->dev, new_mtu);
4102
4103 if (res) {
4104 /* If we failed to set the slave's mtu to the new value
4105 * we must abort the operation even in ACTIVE_BACKUP
4106 * mode, because if we allow the backup slaves to have
4107 * different mtu values than the active slave we'll
4108 * need to change their mtu when doing a failover. That
4109 * means changing their mtu from timer context, which
4110 * is probably not a good idea.
4111 */
5a03cdb7 4112 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 4113 goto unwind;
4114 }
4115 }
4116
4117 bond_dev->mtu = new_mtu;
4118
4119 return 0;
4120
4121unwind:
4122 /* unwind from head to the slave that failed */
4123 stop_at = slave;
4124 bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
4125 int tmp_res;
4126
4127 tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu);
4128 if (tmp_res) {
a4aee5c8
JP		 4129 pr_debug("unwind err %d dev %s\n",
4130 tmp_res, slave->dev->name);
1da177e4
LT		 4131 }
4132 }
4133
4134 return res;
4135}
4136
4137/*
4138 * Change HW address
4139 *
4140 * Note that many devices must be down to change the HW address, and
4141 * downing the master releases all slaves. We can make bonds full of
4142 * bonding devices to test this, however.
4143 */
4144static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
4145{
454d7c9b 4146 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4147 struct sockaddr *sa = addr, tmp_sa;
4148 struct slave *slave, *stop_at;
4149 int res = 0;
4150 int i;
4151
eb7cc59a
SH		 4152 if (bond->params.mode == BOND_MODE_ALB)
4153 return bond_alb_set_mac_address(bond_dev, addr);
4154
4155
a4aee5c8
JP		 4156 pr_debug("bond=%p, name=%s\n",
4157 bond, bond_dev ? bond_dev->name : "None");
1da177e4 4158
dd957c57 4159 /*
3915c1e8
JV		 4160 * If fail_over_mac is set to active, do nothing and return
4161 * success. Returning an error causes ifenslave to fail.
dd957c57 4162 */
3915c1e8 4163 if (bond->params.fail_over_mac == BOND_FOM_ACTIVE)
dd957c57 4164 return 0;
2ab82852 4165
3d632c3f 4166 if (!is_valid_ether_addr(sa->sa_data))
1da177e4 4167 return -EADDRNOTAVAIL;
1da177e4
LT		 4168
4169 /* Can't hold bond->lock with bh disabled here since
4170 * some base drivers panic. On the other hand we can't
4171 * hold bond->lock without bh disabled because we'll
4172 * deadlock. The only solution is to rely on the fact
4173 * that we're under rtnl_lock here, and the slaves
4174 * list won't change. This doesn't solve the problem
4175 * of setting the slave's hw address while it is
4176 * transmitting, but the assumption is that the base
4177 * driver can handle that.
4178 *
4179 * TODO: figure out a way to safely iterate the slaves
4180 * list, but without holding a lock around the actual
4181 * call to the base driver.
4182 */
4183
4184 bond_for_each_slave(bond, slave, i) {
eb7cc59a 4185 const struct net_device_ops *slave_ops = slave->dev->netdev_ops;
5a03cdb7 4186 pr_debug("slave %p %s\n", slave, slave->dev->name);
1da177e4 4187
eb7cc59a 4188 if (slave_ops->ndo_set_mac_address == NULL) {
1da177e4 4189 res = -EOPNOTSUPP;
5a03cdb7 4190 pr_debug("EOPNOTSUPP %s\n", slave->dev->name);
1da177e4
LT		 4191 goto unwind;
4192 }
4193
4194 res = dev_set_mac_address(slave->dev, addr);
4195 if (res) {
4196 /* TODO: consider downing the slave
4197 * and retry ?
4198 * User should expect communications
4199 * breakage anyway until ARP finish
4200 * updating, so...
4201 */
5a03cdb7 4202 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 4203 goto unwind;
4204 }
4205 }
4206
4207 /* success */
4208 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
4209 return 0;
4210
4211unwind:
4212 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
4213 tmp_sa.sa_family = bond_dev->type;
4214
4215 /* unwind from head to the slave that failed */
4216 stop_at = slave;
4217 bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
4218 int tmp_res;
4219
4220 tmp_res = dev_set_mac_address(slave->dev, &tmp_sa);
4221 if (tmp_res) {
a4aee5c8
JP		 4222 pr_debug("unwind err %d dev %s\n",
4223 tmp_res, slave->dev->name);
1da177e4
LT		 4224 }
4225 }
4226
4227 return res;
4228}
4229
4230static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
4231{
454d7c9b 4232 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 4233 struct slave *slave, *start_at;
cf5f9044 4234 int i, slave_no, res = 1;
a2fd940f 4235 struct iphdr *iph = ip_hdr(skb);
1da177e4
LT		 4236
4237 read_lock(&bond->lock);
4238
3d632c3f 4239 if (!BOND_IS_OK(bond))
1da177e4 4240 goto out;
cf5f9044 4241 /*
a2fd940f
AG		 4242 * Start with the curr_active_slave that joined the bond as the
4243 * default for sending IGMP traffic. For failover purposes one
4244 * needs to maintain some consistency for the interface that will
4245 * send the join/membership reports. The curr_active_slave found
4246 * will send all of this type of traffic.
cf5f9044 4247 */
00ae7028 4248 if ((iph->protocol == IPPROTO_IGMP) &&
a2fd940f 4249 (skb->protocol == htons(ETH_P_IP))) {
1da177e4 4250
a2fd940f
AG		 4251 read_lock(&bond->curr_slave_lock);
4252 slave = bond->curr_active_slave;
4253 read_unlock(&bond->curr_slave_lock);
4254
4255 if (!slave)
4256 goto out;
4257 } else {
4258 /*
4259 * Concurrent TX may collide on rr_tx_counter; we accept
4260 * that as being rare enough not to justify using an
4261 * atomic op here.
4262 */
4263 slave_no = bond->rr_tx_counter++ % bond->slave_cnt;
4264
4265 bond_for_each_slave(bond, slave, i) {
4266 slave_no--;
4267 if (slave_no < 0)
4268 break;
4269 }
1da177e4
LT		 4270 }
4271
cf5f9044 4272 start_at = slave;
1da177e4
LT		 4273 bond_for_each_slave_from(bond, slave, i, start_at) {
4274 if (IS_UP(slave->dev) &&
4275 (slave->link == BOND_LINK_UP) &&
4276 (slave->state == BOND_STATE_ACTIVE)) {
4277 res = bond_dev_queue_xmit(bond, skb, slave->dev);
1da177e4
LT		 4278 break;
4279 }
4280 }
4281
1da177e4
LT		 4282out:
4283 if (res) {
4284 /* no suitable interface, frame not sent */
4285 dev_kfree_skb(skb);
4286 }
4287 read_unlock(&bond->lock);
ec634fe3 4288 return NETDEV_TX_OK;
1da177e4
LT		 4289}
4290
075897ce 4291
1da177e4
LT		 4292/*
4293 * in active-backup mode, we know that bond->curr_active_slave is always valid if
4294 * the bond has a usable interface.
4295 */
4296static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
4297{
454d7c9b 4298 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4299 int res = 1;
4300
1da177e4
LT		 4301 read_lock(&bond->lock);
4302 read_lock(&bond->curr_slave_lock);
4303
3d632c3f 4304 if (!BOND_IS_OK(bond))
1da177e4 4305 goto out;
1da177e4 4306
075897ce
JL		 4307 if (!bond->curr_active_slave)
4308 goto out;
4309
075897ce
JL		 4310 res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev);
4311
1da177e4 4312out:
3d632c3f 4313 if (res)
1da177e4
LT		 4314 /* no suitable interface, frame not sent */
4315 dev_kfree_skb(skb);
3d632c3f 4316
1da177e4
LT		 4317 read_unlock(&bond->curr_slave_lock);
4318 read_unlock(&bond->lock);
ec634fe3 4319 return NETDEV_TX_OK;
1da177e4
LT		 4320}
4321
4322/*
169a3e66
JV		 4323 * In bond_xmit_xor() , we determine the output device by using a pre-
4324 * determined xmit_hash_policy(), If the selected device is not enabled,
4325 * find the next active slave.
1da177e4
LT		 4326 */
4327static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
4328{
454d7c9b 4329 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4330 struct slave *slave, *start_at;
4331 int slave_no;
4332 int i;
4333 int res = 1;
4334
4335 read_lock(&bond->lock);
4336
3d632c3f 4337 if (!BOND_IS_OK(bond))
1da177e4 4338 goto out;
1da177e4 4339
a361c83c 4340 slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt);
1da177e4
LT		 4341
4342 bond_for_each_slave(bond, slave, i) {
4343 slave_no--;
3d632c3f 4344 if (slave_no < 0)
1da177e4 4345 break;
1da177e4
LT		 4346 }
4347
4348 start_at = slave;
4349
4350 bond_for_each_slave_from(bond, slave, i, start_at) {
4351 if (IS_UP(slave->dev) &&
4352 (slave->link == BOND_LINK_UP) &&
4353 (slave->state == BOND_STATE_ACTIVE)) {
4354 res = bond_dev_queue_xmit(bond, skb, slave->dev);
4355 break;
4356 }
4357 }
4358
4359out:
4360 if (res) {
4361 /* no suitable interface, frame not sent */
4362 dev_kfree_skb(skb);
4363 }
4364 read_unlock(&bond->lock);
ec634fe3 4365 return NETDEV_TX_OK;
1da177e4
LT		 4366}
4367
4368/*
4369 * in broadcast mode, we send everything to all usable interfaces.
4370 */
4371static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
4372{
454d7c9b 4373 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4374 struct slave *slave, *start_at;
4375 struct net_device *tx_dev = NULL;
4376 int i;
4377 int res = 1;
4378
4379 read_lock(&bond->lock);
4380
3d632c3f 4381 if (!BOND_IS_OK(bond))
1da177e4 4382 goto out;
1da177e4
LT		 4383
4384 read_lock(&bond->curr_slave_lock);
4385 start_at = bond->curr_active_slave;
4386 read_unlock(&bond->curr_slave_lock);
4387
3d632c3f 4388 if (!start_at)
1da177e4 4389 goto out;
1da177e4
LT		 4390
4391 bond_for_each_slave_from(bond, slave, i, start_at) {
4392 if (IS_UP(slave->dev) &&
4393 (slave->link == BOND_LINK_UP) &&
4394 (slave->state == BOND_STATE_ACTIVE)) {
4395 if (tx_dev) {
4396 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
4397 if (!skb2) {
a4aee5c8 4398 pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n",
4e0952c7 4399 bond_dev->name);
1da177e4
LT		 4400 continue;
4401 }
4402
4403 res = bond_dev_queue_xmit(bond, skb2, tx_dev);
4404 if (res) {
4405 dev_kfree_skb(skb2);
4406 continue;
4407 }
4408 }
4409 tx_dev = slave->dev;
4410 }
4411 }
4412
3d632c3f 4413 if (tx_dev)
1da177e4 4414 res = bond_dev_queue_xmit(bond, skb, tx_dev);
1da177e4
LT		 4415
4416out:
3d632c3f 4417 if (res)
1da177e4
LT		 4418 /* no suitable interface, frame not sent */
4419 dev_kfree_skb(skb);
3d632c3f 4420
1da177e4
LT		 4421 /* frame sent to all suitable interfaces */
4422 read_unlock(&bond->lock);
ec634fe3 4423 return NETDEV_TX_OK;
1da177e4
LT		 4424}
4425
4426/*------------------------- Device initialization ---------------------------*/
4427
6f6652be
JV		 4428static void bond_set_xmit_hash_policy(struct bonding *bond)
4429{
4430 switch (bond->params.xmit_policy) {
4431 case BOND_XMIT_POLICY_LAYER23:
4432 bond->xmit_hash_policy = bond_xmit_hash_policy_l23;
4433 break;
4434 case BOND_XMIT_POLICY_LAYER34:
4435 bond->xmit_hash_policy = bond_xmit_hash_policy_l34;
4436 break;
4437 case BOND_XMIT_POLICY_LAYER2:
4438 default:
4439 bond->xmit_hash_policy = bond_xmit_hash_policy_l2;
4440 break;
4441 }
4442}
4443
bb1d9123
AG		 4444/*
4445 * Lookup the slave that corresponds to a qid
4446 */
4447static inline int bond_slave_override(struct bonding *bond,
4448 struct sk_buff *skb)
4449{
4450 int i, res = 1;
4451 struct slave *slave = NULL;
4452 struct slave *check_slave;
4453
4454 read_lock(&bond->lock);
4455
4456 if (!BOND_IS_OK(bond) || !skb->queue_mapping)
4457 goto out;
4458
4459 /* Find out if any slaves have the same mapping as this skb. */
4460 bond_for_each_slave(bond, check_slave, i) {
4461 if (check_slave->queue_id == skb->queue_mapping) {
4462 slave = check_slave;
4463 break;
4464 }
4465 }
4466
4467 /* If the slave isn't UP, use default transmit policy. */
4468 if (slave && slave->queue_id && IS_UP(slave->dev) &&
4469 (slave->link == BOND_LINK_UP)) {
4470 res = bond_dev_queue_xmit(bond, skb, slave->dev);
4471 }
4472
4473out:
4474 read_unlock(&bond->lock);
4475 return res;
4476}
4477
4478static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb)
4479{
4480 /*
4481 * This helper function exists to help dev_pick_tx get the correct
4482 * destination queue. Using a helper function skips the a call to
4483 * skb_tx_hash and will put the skbs in the queue we expect on their
4484 * way down to the bonding driver.
4485 */
4486 return skb->queue_mapping;
4487}
4488
424efe9c 4489static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
00829823 4490{
bb1d9123
AG		 4491 struct bonding *bond = netdev_priv(dev);
4492
4493 if (TX_QUEUE_OVERRIDE(bond->params.mode)) {
4494 if (!bond_slave_override(bond, skb))
4495 return NETDEV_TX_OK;
4496 }
00829823
SH		 4497
4498 switch (bond->params.mode) {
4499 case BOND_MODE_ROUNDROBIN:
4500 return bond_xmit_roundrobin(skb, dev);
4501 case BOND_MODE_ACTIVEBACKUP:
4502 return bond_xmit_activebackup(skb, dev);
4503 case BOND_MODE_XOR:
4504 return bond_xmit_xor(skb, dev);
4505 case BOND_MODE_BROADCAST:
4506 return bond_xmit_broadcast(skb, dev);
4507 case BOND_MODE_8023AD:
4508 return bond_3ad_xmit_xor(skb, dev);
4509 case BOND_MODE_ALB:
4510 case BOND_MODE_TLB:
4511 return bond_alb_xmit(skb, dev);
4512 default:
4513 /* Should never happen, mode already checked */
a4aee5c8
JP		 4514 pr_err("%s: Error: Unknown bonding mode %d\n",
4515 dev->name, bond->params.mode);
00829823
SH		 4516 WARN_ON_ONCE(1);
4517 dev_kfree_skb(skb);
4518 return NETDEV_TX_OK;
4519 }
4520}
4521
4522
1da177e4
LT		 4523/*
4524 * set bond mode specific net device operations
4525 */
a77b5325 4526void bond_set_mode_ops(struct bonding *bond, int mode)
1da177e4 4527{
169a3e66
JV		 4528 struct net_device *bond_dev = bond->dev;
4529
1da177e4
LT		 4530 switch (mode) {
4531 case BOND_MODE_ROUNDROBIN:
1da177e4
LT		 4532 break;
4533 case BOND_MODE_ACTIVEBACKUP:
1da177e4
LT		 4534 break;
4535 case BOND_MODE_XOR:
6f6652be 4536 bond_set_xmit_hash_policy(bond);
1da177e4
LT		 4537 break;
4538 case BOND_MODE_BROADCAST:
1da177e4
LT		 4539 break;
4540 case BOND_MODE_8023AD:
8f903c70 4541 bond_set_master_3ad_flags(bond);
6f6652be 4542 bond_set_xmit_hash_policy(bond);
1da177e4 4543 break;
1da177e4 4544 case BOND_MODE_ALB:
8f903c70
JV		 4545 bond_set_master_alb_flags(bond);
4546 /* FALLTHRU */
4547 case BOND_MODE_TLB:
1da177e4
LT		 4548 break;
4549 default:
4550 /* Should never happen, mode already checked */
a4aee5c8
JP		 4551 pr_err("%s: Error: Unknown bonding mode %d\n",
4552 bond_dev->name, mode);
1da177e4
LT		 4553 break;
4554 }
4555}
4556
217df670
JV		 4557static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
4558 struct ethtool_drvinfo *drvinfo)
4559{
4560 strncpy(drvinfo->driver, DRV_NAME, 32);
4561 strncpy(drvinfo->version, DRV_VERSION, 32);
4562 snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION);
4563}
4564
7282d491 4565static const struct ethtool_ops bond_ethtool_ops = {
217df670 4566 .get_drvinfo = bond_ethtool_get_drvinfo,
fa53ebac
SH		 4567 .get_link = ethtool_op_get_link,
4568 .get_tx_csum = ethtool_op_get_tx_csum,
4569 .get_sg = ethtool_op_get_sg,
4570 .get_tso = ethtool_op_get_tso,
4571 .get_ufo = ethtool_op_get_ufo,
4572 .get_flags = ethtool_op_get_flags,
8531c5ff
AK		 4573};
4574
eb7cc59a 4575static const struct net_device_ops bond_netdev_ops = {
181470fc 4576 .ndo_init = bond_init,
9e71626c 4577 .ndo_uninit = bond_uninit,
eb7cc59a
SH		 4578 .ndo_open = bond_open,
4579 .ndo_stop = bond_close,
00829823 4580 .ndo_start_xmit = bond_start_xmit,
bb1d9123 4581 .ndo_select_queue = bond_select_queue,
be1f3c2c 4582 .ndo_get_stats64 = bond_get_stats,
eb7cc59a
SH		 4583 .ndo_do_ioctl = bond_do_ioctl,
4584 .ndo_set_multicast_list = bond_set_multicast_list,
4585 .ndo_change_mtu = bond_change_mtu,
eb7cc59a 4586 .ndo_set_mac_address = bond_set_mac_address,
00829823 4587 .ndo_neigh_setup = bond_neigh_setup,
eb7cc59a
SH		 4588 .ndo_vlan_rx_register = bond_vlan_rx_register,
4589 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
4590 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
f6dc31a8
WC		 4591#ifdef CONFIG_NET_POLL_CONTROLLER
4592 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
4593 .ndo_poll_controller = bond_poll_controller,
4594#endif
eb7cc59a
SH		 4595};
4596
9e2e61fb
AW		 4597static void bond_destructor(struct net_device *bond_dev)
4598{
4599 struct bonding *bond = netdev_priv(bond_dev);
4600 if (bond->wq)
4601 destroy_workqueue(bond->wq);
4602 free_netdev(bond_dev);
4603}
4604
181470fc 4605static void bond_setup(struct net_device *bond_dev)
1da177e4 4606{
454d7c9b 4607 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 4608
1da177e4
LT		 4609 /* initialize rwlocks */
4610 rwlock_init(&bond->lock);
4611 rwlock_init(&bond->curr_slave_lock);
4612
d2991f75 4613 bond->params = bonding_defaults;
1da177e4
LT		 4614
4615 /* Initialize pointers */
1da177e4
LT		 4616 bond->dev = bond_dev;
4617 INIT_LIST_HEAD(&bond->vlan_list);
4618
4619 /* Initialize the device entry points */
181470fc 4620 ether_setup(bond_dev);
eb7cc59a 4621 bond_dev->netdev_ops = &bond_netdev_ops;
8531c5ff 4622 bond_dev->ethtool_ops = &bond_ethtool_ops;
169a3e66 4623 bond_set_mode_ops(bond, bond->params.mode);
1da177e4 4624
9e2e61fb 4625 bond_dev->destructor = bond_destructor;
1da177e4
LT		 4626
4627 /* Initialize the device options */
4628 bond_dev->tx_queue_len = 0;
4629 bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
0b680e75 4630 bond_dev->priv_flags |= IFF_BONDING;
181470fc
SH		 4631 bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
4632
6cf3f41e
JV		 4633 if (bond->params.arp_interval)
4634 bond_dev->priv_flags |= IFF_MASTER_ARPMON;
1da177e4
LT		 4635
4636 /* At first, we block adding VLANs. That's the only way to
4637 * prevent problems that occur when adding VLANs over an
4638 * empty bond. The block will be removed once non-challenged
4639 * slaves are enslaved.
4640 */
4641 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
4642
932ff279 4643 /* don't acquire bond device's netif_tx_lock when
1da177e4
LT		 4644 * transmitting */
4645 bond_dev->features |= NETIF_F_LLTX;
4646
4647 /* By default, we declare the bond to be fully
4648 * VLAN hardware accelerated capable. Special
4649 * care is taken in the various xmit functions
4650 * when there are slaves that are not hw accel
4651 * capable
4652 */
1da177e4
LT		 4653 bond_dev->features |= (NETIF_F_HW_VLAN_TX |
4654 NETIF_F_HW_VLAN_RX |
4655 NETIF_F_HW_VLAN_FILTER);
4656
1da177e4
LT		 4657}
4658
fdaea7a9
JV		 4659static void bond_work_cancel_all(struct bonding *bond)
4660{
4661 write_lock_bh(&bond->lock);
4662 bond->kill_timers = 1;
4663 write_unlock_bh(&bond->lock);
4664
4665 if (bond->params.miimon && delayed_work_pending(&bond->mii_work))
4666 cancel_delayed_work(&bond->mii_work);
4667
4668 if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work))
4669 cancel_delayed_work(&bond->arp_work);
4670
4671 if (bond->params.mode == BOND_MODE_ALB &&
4672 delayed_work_pending(&bond->alb_work))
4673 cancel_delayed_work(&bond->alb_work);
4674
4675 if (bond->params.mode == BOND_MODE_8023AD &&
4676 delayed_work_pending(&bond->ad_work))
4677 cancel_delayed_work(&bond->ad_work);
4678}
4679
c67dfb29
EB		 4680/*
4681* Destroy a bonding device.
4682* Must be under rtnl_lock when this function is called.
4683*/
4684static void bond_uninit(struct net_device *bond_dev)
a434e43f 4685{
454d7c9b 4686 struct bonding *bond = netdev_priv(bond_dev);
a434e43f 4687
f6dc31a8
WC		 4688 bond_netpoll_cleanup(bond_dev);
4689
c67dfb29
EB		 4690 /* Release the bonded slaves */
4691 bond_release_all(bond_dev);
4692
a434e43f
JV		 4693 list_del(&bond->bond_list);
4694
4695 bond_work_cancel_all(bond);
4696
a434e43f 4697 bond_remove_proc_entry(bond);
c67dfb29 4698
22bedad3 4699 __hw_addr_flush(&bond->mc_list);
a434e43f
JV		 4700}
4701
1da177e4
LT		 4702/*------------------------- Module initialization ---------------------------*/
4703
4704/*
4705 * Convert string input module parms. Accept either the
ece95f7f
JV		 4706 * number of the mode or its string name. A bit complicated because
4707 * some mode names are substrings of other names, and calls from sysfs
4708 * may have whitespace in the name (trailing newlines, for example).
1da177e4 4709 */
325dcf7a 4710int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl)
1da177e4 4711{
54b87323 4712 int modeint = -1, i, rv;
a42e534f 4713 char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, };
ece95f7f 4714
a42e534f
JV		 4715 for (p = (char *)buf; *p; p++)
4716 if (!(isdigit(*p) || isspace(*p)))
4717 break;
4718
4719 if (*p)
ece95f7f 4720 rv = sscanf(buf, "%20s", modestr);
a42e534f 4721 else
54b87323 4722 rv = sscanf(buf, "%d", &modeint);
a42e534f
JV		 4723
4724 if (!rv)
4725 return -1;
1da177e4
LT		 4726
4727 for (i = 0; tbl[i].modename; i++) {
54b87323 4728 if (modeint == tbl[i].mode)
ece95f7f
JV		 4729 return tbl[i].mode;
4730 if (strcmp(modestr, tbl[i].modename) == 0)
1da177e4 4731 return tbl[i].mode;
1da177e4
LT		 4732 }
4733
4734 return -1;
4735}
4736
4737static int bond_check_params(struct bond_params *params)
4738{
a549952a 4739 int arp_validate_value, fail_over_mac_value, primary_reselect_value;
f5b2b966 4740
1da177e4
LT		 4741 /*
4742 * Convert string parameters.
4743 */
4744 if (mode) {
4745 bond_mode = bond_parse_parm(mode, bond_mode_tbl);
4746 if (bond_mode == -1) {
a4aee5c8 4747 pr_err("Error: Invalid bonding mode \"%s\"\n",
1da177e4
LT		 4748 mode == NULL ? "NULL" : mode);
4749 return -EINVAL;
4750 }
4751 }
4752
169a3e66
JV		 4753 if (xmit_hash_policy) {
4754 if ((bond_mode != BOND_MODE_XOR) &&
4755 (bond_mode != BOND_MODE_8023AD)) {
a4aee5c8 4756 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
169a3e66
JV		 4757 bond_mode_name(bond_mode));
4758 } else {
4759 xmit_hashtype = bond_parse_parm(xmit_hash_policy,
4760 xmit_hashtype_tbl);
4761 if (xmit_hashtype == -1) {
a4aee5c8 4762 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
3d632c3f 4763 xmit_hash_policy == NULL ? "NULL" :
169a3e66
JV		 4764 xmit_hash_policy);
4765 return -EINVAL;
4766 }
4767 }
4768 }
4769
1da177e4
LT		 4770 if (lacp_rate) {
4771 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8
JP		 4772 pr_info("lacp_rate param is irrelevant in mode %s\n",
4773 bond_mode_name(bond_mode));
1da177e4
LT		 4774 } else {
4775 lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl);
4776 if (lacp_fast == -1) {
a4aee5c8 4777 pr_err("Error: Invalid lacp rate \"%s\"\n",
1da177e4
LT		 4778 lacp_rate == NULL ? "NULL" : lacp_rate);
4779 return -EINVAL;
4780 }
4781 }
4782 }
4783
fd989c83
JV		 4784 if (ad_select) {
4785 params->ad_select = bond_parse_parm(ad_select, ad_select_tbl);
4786 if (params->ad_select == -1) {
a4aee5c8 4787 pr_err("Error: Invalid ad_select \"%s\"\n",
fd989c83
JV		 4788 ad_select == NULL ? "NULL" : ad_select);
4789 return -EINVAL;
4790 }
4791
4792 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8 4793 pr_warning("ad_select param only affects 802.3ad mode\n");
fd989c83
JV		 4794 }
4795 } else {
4796 params->ad_select = BOND_AD_STABLE;
4797 }
4798
f5841306 4799 if (max_bonds < 0) {
a4aee5c8
JP		 4800 pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4801 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
1da177e4
LT		 4802 max_bonds = BOND_DEFAULT_MAX_BONDS;
4803 }
4804
4805 if (miimon < 0) {
a4aee5c8
JP		 4806 pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n",
4807 miimon, INT_MAX, BOND_LINK_MON_INTERV);
1da177e4
LT		 4808 miimon = BOND_LINK_MON_INTERV;
4809 }
4810
4811 if (updelay < 0) {
a4aee5c8
JP		 4812 pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4813 updelay, INT_MAX);
1da177e4
LT		 4814 updelay = 0;
4815 }
4816
4817 if (downdelay < 0) {
a4aee5c8
JP		 4818 pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4819 downdelay, INT_MAX);
1da177e4
LT		 4820 downdelay = 0;
4821 }
4822
4823 if ((use_carrier != 0) && (use_carrier != 1)) {
a4aee5c8
JP		 4824 pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4825 use_carrier);
1da177e4
LT		 4826 use_carrier = 1;
4827 }
4828
7893b249 4829 if (num_grat_arp < 0 || num_grat_arp > 255) {
2381a55c 4830 pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n",
a4aee5c8 4831 num_grat_arp);
7893b249
MS		 4832 num_grat_arp = 1;
4833 }
4834
305d552a 4835 if (num_unsol_na < 0 || num_unsol_na > 255) {
2381a55c 4836 pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
a4aee5c8 4837 num_unsol_na);
305d552a
BH		 4838 num_unsol_na = 1;
4839 }
4840
1da177e4
LT		 4841 /* reset values for 802.3ad */
4842 if (bond_mode == BOND_MODE_8023AD) {
4843 if (!miimon) {
a4aee5c8 4844 pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n");
3d632c3f 4845 pr_warning("Forcing miimon to 100msec\n");
1da177e4
LT		 4846 miimon = 100;
4847 }
4848 }
4849
bb1d9123
AG		 4850 if (tx_queues < 1 || tx_queues > 255) {
4851 pr_warning("Warning: tx_queues (%d) should be between "
4852 "1 and 255, resetting to %d\n",
4853 tx_queues, BOND_DEFAULT_TX_QUEUES);
4854 tx_queues = BOND_DEFAULT_TX_QUEUES;
4855 }
4856
ebd8e497
AG		 4857 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
4858 pr_warning("Warning: all_slaves_active module parameter (%d), "
4859 "not of valid value (0/1), so it was set to "
4860 "0\n", all_slaves_active);
4861 all_slaves_active = 0;
4862 }
4863
1da177e4
LT		 4864 /* reset values for TLB/ALB */
4865 if ((bond_mode == BOND_MODE_TLB) ||
4866 (bond_mode == BOND_MODE_ALB)) {
4867 if (!miimon) {
a4aee5c8 4868 pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure and link speed which are essential for TLB/ALB load balancing\n");
3d632c3f 4869 pr_warning("Forcing miimon to 100msec\n");
1da177e4
LT		 4870 miimon = 100;
4871 }
4872 }
4873
4874 if (bond_mode == BOND_MODE_ALB) {
a4aee5c8
JP		 4875 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",
4876 updelay);
1da177e4
LT		 4877 }
4878
4879 if (!miimon) {
4880 if (updelay || downdelay) {
4881 /* just warn the user the up/down delay will have
4882 * no effect since miimon is zero...
4883 */
a4aee5c8
JP		 4884 pr_warning("Warning: miimon module parameter not set and updelay (%d) or downdelay (%d) module parameter is set; updelay and downdelay have no effect unless miimon is set\n",
4885 updelay, downdelay);
1da177e4
LT		 4886 }
4887 } else {
4888 /* don't allow arp monitoring */
4889 if (arp_interval) {
a4aee5c8
JP		 4890 pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4891 miimon, arp_interval);
1da177e4
LT		 4892 arp_interval = 0;
4893 }
4894
4895 if ((updelay % miimon) != 0) {
a4aee5c8
JP		 4896 pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4897 updelay, miimon,
4898 (updelay / miimon) * miimon);
1da177e4
LT		 4899 }
4900
4901 updelay /= miimon;
4902
4903 if ((downdelay % miimon) != 0) {
a4aee5c8
JP		 4904 pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4905 downdelay, miimon,
4906 (downdelay / miimon) * miimon);
1da177e4
LT		 4907 }
4908
4909 downdelay /= miimon;
4910 }
4911
4912 if (arp_interval < 0) {
a4aee5c8
JP		 4913 pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n",
4914 arp_interval, INT_MAX, BOND_LINK_ARP_INTERV);
1da177e4
LT		 4915 arp_interval = BOND_LINK_ARP_INTERV;
4916 }
4917
4918 for (arp_ip_count = 0;
4919 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count];
4920 arp_ip_count++) {
4921 /* not complete check, but should be good enough to
4922 catch mistakes */
4923 if (!isdigit(arp_ip_target[arp_ip_count][0])) {
a4aee5c8
JP		 4924 pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4925 arp_ip_target[arp_ip_count]);
1da177e4
LT		 4926 arp_interval = 0;
4927 } else {
d3bb52b0 4928 __be32 ip = in_aton(arp_ip_target[arp_ip_count]);
1da177e4
LT		 4929 arp_target[arp_ip_count] = ip;
4930 }
4931 }
4932
4933 if (arp_interval && !arp_ip_count) {
4934 /* don't allow arping if no arp_ip_target given... */
a4aee5c8
JP		 4935 pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4936 arp_interval);
1da177e4
LT		 4937 arp_interval = 0;
4938 }
4939
f5b2b966
JV		 4940 if (arp_validate) {
4941 if (bond_mode != BOND_MODE_ACTIVEBACKUP) {
a4aee5c8 4942 pr_err("arp_validate only supported in active-backup mode\n");
f5b2b966
JV		 4943 return -EINVAL;
4944 }
4945 if (!arp_interval) {
a4aee5c8 4946 pr_err("arp_validate requires arp_interval\n");
f5b2b966
JV		 4947 return -EINVAL;
4948 }
4949
4950 arp_validate_value = bond_parse_parm(arp_validate,
4951 arp_validate_tbl);
4952 if (arp_validate_value == -1) {
a4aee5c8 4953 pr_err("Error: invalid arp_validate \"%s\"\n",
f5b2b966
JV		 4954 arp_validate == NULL ? "NULL" : arp_validate);
4955 return -EINVAL;
4956 }
4957 } else
4958 arp_validate_value = 0;
4959
1da177e4 4960 if (miimon) {
a4aee5c8 4961 pr_info("MII link monitoring set to %d ms\n", miimon);
1da177e4
LT		 4962 } else if (arp_interval) {
4963 int i;
4964
a4aee5c8
JP		 4965 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
4966 arp_interval,
4967 arp_validate_tbl[arp_validate_value].modename,
4968 arp_ip_count);
1da177e4
LT		 4969
4970 for (i = 0; i < arp_ip_count; i++)
e5e2a8fd 4971 pr_info(" %s", arp_ip_target[i]);
1da177e4 4972
e5e2a8fd 4973 pr_info("\n");
1da177e4 4974
b8a9787e 4975 } else if (max_bonds) {
1da177e4
LT		 4976 /* miimon and arp_interval not set, we need one so things
4977 * work as expected, see bonding.txt for details
4978 */
a4aee5c8 4979 pr_warning("Warning: either miimon or arp_interval and arp_ip_target module parameters must be specified, otherwise bonding will not detect link failures! see bonding.txt for details.\n");
1da177e4
LT		 4980 }
4981
4982 if (primary && !USES_PRIMARY(bond_mode)) {
4983 /* currently, using a primary only makes sense
4984 * in active backup, TLB or ALB modes
4985 */
a4aee5c8
JP		 4986 pr_warning("Warning: %s primary device specified but has no effect in %s mode\n",
4987 primary, bond_mode_name(bond_mode));
1da177e4
LT		 4988 primary = NULL;
4989 }
4990
a549952a
JP		 4991 if (primary && primary_reselect) {
4992 primary_reselect_value = bond_parse_parm(primary_reselect,
4993 pri_reselect_tbl);
4994 if (primary_reselect_value == -1) {
a4aee5c8 4995 pr_err("Error: Invalid primary_reselect \"%s\"\n",
a549952a
JP		 4996 primary_reselect ==
4997 NULL ? "NULL" : primary_reselect);
4998 return -EINVAL;
4999 }
5000 } else {
5001 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
5002 }
5003
3915c1e8
JV		 5004 if (fail_over_mac) {
5005 fail_over_mac_value = bond_parse_parm(fail_over_mac,
5006 fail_over_mac_tbl);
5007 if (fail_over_mac_value == -1) {
a4aee5c8 5008 pr_err("Error: invalid fail_over_mac \"%s\"\n",
3915c1e8
JV		 5009 arp_validate == NULL ? "NULL" : arp_validate);
5010 return -EINVAL;
5011 }
5012
5013 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
a4aee5c8 5014 pr_warning("Warning: fail_over_mac only affects active-backup mode.\n");
3915c1e8
JV		 5015 } else {
5016 fail_over_mac_value = BOND_FOM_NONE;
5017 }
dd957c57 5018
1da177e4
LT		 5019 /* fill params struct with the proper values */
5020 params->mode = bond_mode;
169a3e66 5021 params->xmit_policy = xmit_hashtype;
1da177e4 5022 params->miimon = miimon;
7893b249 5023 params->num_grat_arp = num_grat_arp;
305d552a 5024 params->num_unsol_na = num_unsol_na;
1da177e4 5025 params->arp_interval = arp_interval;
f5b2b966 5026 params->arp_validate = arp_validate_value;
1da177e4
LT		 5027 params->updelay = updelay;
5028 params->downdelay = downdelay;
5029 params->use_carrier = use_carrier;
5030 params->lacp_fast = lacp_fast;
5031 params->primary[0] = 0;
a549952a 5032 params->primary_reselect = primary_reselect_value;
3915c1e8 5033 params->fail_over_mac = fail_over_mac_value;
bb1d9123 5034 params->tx_queues = tx_queues;
ebd8e497 5035 params->all_slaves_active = all_slaves_active;
1da177e4
LT		 5036
5037 if (primary) {
5038 strncpy(params->primary, primary, IFNAMSIZ);
5039 params->primary[IFNAMSIZ - 1] = 0;
5040 }
5041
5042 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
5043
5044 return 0;
5045}
5046
0daa2303 5047static struct lock_class_key bonding_netdev_xmit_lock_key;
cf508b12 5048static struct lock_class_key bonding_netdev_addr_lock_key;
0daa2303 5049
e8a0464c
DM		 5050static void bond_set_lockdep_class_one(struct net_device *dev,
5051 struct netdev_queue *txq,
5052 void *_unused)
c773e847
DM		 5053{
5054 lockdep_set_class(&txq->_xmit_lock,
5055 &bonding_netdev_xmit_lock_key);
5056}
5057
5058static void bond_set_lockdep_class(struct net_device *dev)
5059{
cf508b12
DM		 5060 lockdep_set_class(&dev->addr_list_lock,
5061 &bonding_netdev_addr_lock_key);
e8a0464c 5062 netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL);
c773e847
DM		 5063}
5064
181470fc
SH		 5065/*
5066 * Called from registration process
5067 */
5068static int bond_init(struct net_device *bond_dev)
5069{
5070 struct bonding *bond = netdev_priv(bond_dev);
ec87fd3b 5071 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
181470fc
SH		 5072
5073 pr_debug("Begin bond_init for %s\n", bond_dev->name);
5074
5075 bond->wq = create_singlethread_workqueue(bond_dev->name);
5076 if (!bond->wq)
5077 return -ENOMEM;
5078
5079 bond_set_lockdep_class(bond_dev);
5080
5081 netif_carrier_off(bond_dev);
5082
5083 bond_create_proc_entry(bond);
ec87fd3b 5084 list_add_tail(&bond->bond_list, &bn->dev_list);
181470fc 5085
6151b3d4 5086 bond_prepare_sysfs_group(bond);
22bedad3
JP		 5087
5088 __hw_addr_init(&bond->mc_list);
181470fc
SH		 5089 return 0;
5090}
5091
88ead977
EB		 5092static int bond_validate(struct nlattr *tb[], struct nlattr *data[])
5093{
5094 if (tb[IFLA_ADDRESS]) {
5095 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
5096 return -EINVAL;
5097 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
5098 return -EADDRNOTAVAIL;
5099 }
5100 return 0;
5101}
5102
5103static struct rtnl_link_ops bond_link_ops __read_mostly = {
5104 .kind = "bond",
6639104b 5105 .priv_size = sizeof(struct bonding),
88ead977
EB		 5106 .setup = bond_setup,
5107 .validate = bond_validate,
5108};
5109
dfe60397 5110/* Create a new bond based on the specified name and bonding parameters.
e4b91c48 5111 * If name is NULL, obtain a suitable "bond%d" name for us.
dfe60397
MW		 5112 * Caller must NOT hold rtnl_lock; we need to release it here before we
5113 * set up our sysfs entries.
5114 */
ec87fd3b 5115int bond_create(struct net *net, const char *name)
dfe60397
MW		 5116{
5117 struct net_device *bond_dev;
5118 int res;
5119
5120 rtnl_lock();
027ea041 5121
bb1d9123
AG		 5122 bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "",
5123 bond_setup, tx_queues);
dfe60397 5124 if (!bond_dev) {
a4aee5c8 5125 pr_err("%s: eek! can't alloc netdev!\n", name);
9e2e61fb
AW		 5126 rtnl_unlock();
5127 return -ENOMEM;
dfe60397
MW		 5128 }
5129
ec87fd3b 5130 dev_net_set(bond_dev, net);
88ead977
EB		 5131 bond_dev->rtnl_link_ops = &bond_link_ops;
5132
e4b91c48
JV		 5133 if (!name) {
5134 res = dev_alloc_name(bond_dev, "bond%d");
5135 if (res < 0)
9e2e61fb 5136 goto out;
e4b91c48
JV		 5137 }
5138
dfe60397 5139 res = register_netdevice(bond_dev);
0daa2303 5140
30c15ba9 5141out:
7e083840 5142 rtnl_unlock();
9e2e61fb
AW		 5143 if (res < 0)
5144 bond_destructor(bond_dev);
30c15ba9 5145 return res;
dfe60397
MW		 5146}
5147
2c8c1e72 5148static int __net_init bond_net_init(struct net *net)
ec87fd3b 5149{
15449745 5150 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 5151
5152 bn->net = net;
5153 INIT_LIST_HEAD(&bn->dev_list);
5154
ec87fd3b 5155 bond_create_proc_dir(bn);
15449745
EB		 5156
5157 return 0;
ec87fd3b
EB		 5158}
5159
2c8c1e72 5160static void __net_exit bond_net_exit(struct net *net)
ec87fd3b 5161{
15449745 5162 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 5163
5164 bond_destroy_proc_dir(bn);
ec87fd3b
EB		 5165}
5166
5167static struct pernet_operations bond_net_ops = {
5168 .init = bond_net_init,
5169 .exit = bond_net_exit,
15449745
EB		 5170 .id = &bond_net_id,
5171 .size = sizeof(struct bond_net),
ec87fd3b
EB		 5172};
5173
1da177e4
LT		 5174static int __init bonding_init(void)
5175{
1da177e4
LT		 5176 int i;
5177 int res;
5178
3d632c3f 5179 pr_info("%s", version);
1da177e4 5180
dfe60397 5181 res = bond_check_params(&bonding_defaults);
3d632c3f 5182 if (res)
dfe60397 5183 goto out;
1da177e4 5184
15449745 5185 res = register_pernet_subsys(&bond_net_ops);
ec87fd3b
EB		 5186 if (res)
5187 goto out;
027ea041 5188
88ead977
EB		 5189 res = rtnl_link_register(&bond_link_ops);
5190 if (res)
6639104b 5191 goto err_link;
88ead977 5192
1da177e4 5193 for (i = 0; i < max_bonds; i++) {
ec87fd3b 5194 res = bond_create(&init_net, NULL);
dfe60397
MW		 5195 if (res)
5196 goto err;
1da177e4
LT		 5197 }
5198
b76cdba9
MW		 5199 res = bond_create_sysfs();
5200 if (res)
5201 goto err;
5202
1da177e4 5203 register_netdevice_notifier(&bond_netdev_notifier);
c3ade5ca 5204 register_inetaddr_notifier(&bond_inetaddr_notifier);
305d552a 5205 bond_register_ipv6_notifier();
dfe60397 5206out:
1da177e4 5207 return res;
88ead977
EB		 5208err:
5209 rtnl_link_unregister(&bond_link_ops);
6639104b 5210err_link:
15449745 5211 unregister_pernet_subsys(&bond_net_ops);
88ead977 5212 goto out;
dfe60397 5213
1da177e4
LT		 5214}
5215
5216static void __exit bonding_exit(void)
5217{
5218 unregister_netdevice_notifier(&bond_netdev_notifier);
c3ade5ca 5219 unregister_inetaddr_notifier(&bond_inetaddr_notifier);
305d552a 5220 bond_unregister_ipv6_notifier();
1da177e4 5221
ae68c398
PE		 5222 bond_destroy_sysfs();
5223
88ead977 5224 rtnl_link_unregister(&bond_link_ops);
15449745 5225 unregister_pernet_subsys(&bond_net_ops);
1da177e4
LT		 5226}
5227
5228module_init(bonding_init);
5229module_exit(bonding_exit);
5230MODULE_LICENSE("GPL");
5231MODULE_VERSION(DRV_VERSION);
5232MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
5233MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
88ead977 5234MODULE_ALIAS_RTNL_LINK("bond");
Linux kernel - Deliverables
This page took 1.150734 seconds and 5 git commands to generate.