]> bbs.cooldavid.org Git - net-next-2.6.git/blame - drivers/net/bonding/bond_main.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net/fec: restore interrupt mask after software-reset in fec_stop()
[net-next-2.6.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;
90e1795b 179#ifdef CONFIG_NET_POLL_CONTROLLER
c22d7ac8 180static int disable_netpoll = 1;
90e1795b 181#endif
1da177e4 182
e97fd7c6 183const struct bond_parm_tbl bond_lacp_tbl[] = {
1da177e4
LT		 184{ "slow", AD_LACP_SLOW},
185{ "fast", AD_LACP_FAST},
186{ NULL, -1},
187};
188
e97fd7c6 189const struct bond_parm_tbl bond_mode_tbl[] = {
1da177e4
LT		 190{ "balance-rr", BOND_MODE_ROUNDROBIN},
191{ "active-backup", BOND_MODE_ACTIVEBACKUP},
192{ "balance-xor", BOND_MODE_XOR},
193{ "broadcast", BOND_MODE_BROADCAST},
194{ "802.3ad", BOND_MODE_8023AD},
195{ "balance-tlb", BOND_MODE_TLB},
196{ "balance-alb", BOND_MODE_ALB},
197{ NULL, -1},
198};
199
e97fd7c6 200const struct bond_parm_tbl xmit_hashtype_tbl[] = {
169a3e66
JV		 201{ "layer2", BOND_XMIT_POLICY_LAYER2},
202{ "layer3+4", BOND_XMIT_POLICY_LAYER34},
6f6652be 203{ "layer2+3", BOND_XMIT_POLICY_LAYER23},
169a3e66
JV		 204{ NULL, -1},
205};
206
e97fd7c6 207const struct bond_parm_tbl arp_validate_tbl[] = {
f5b2b966
JV		 208{ "none", BOND_ARP_VALIDATE_NONE},
209{ "active", BOND_ARP_VALIDATE_ACTIVE},
210{ "backup", BOND_ARP_VALIDATE_BACKUP},
211{ "all", BOND_ARP_VALIDATE_ALL},
212{ NULL, -1},
213};
214
e97fd7c6 215const struct bond_parm_tbl fail_over_mac_tbl[] = {
3915c1e8
JV		 216{ "none", BOND_FOM_NONE},
217{ "active", BOND_FOM_ACTIVE},
218{ "follow", BOND_FOM_FOLLOW},
219{ NULL, -1},
220};
221
a549952a
JP		 222const struct bond_parm_tbl pri_reselect_tbl[] = {
223{ "always", BOND_PRI_RESELECT_ALWAYS},
224{ "better", BOND_PRI_RESELECT_BETTER},
225{ "failure", BOND_PRI_RESELECT_FAILURE},
226{ NULL, -1},
227};
228
fd989c83
JV		 229struct bond_parm_tbl ad_select_tbl[] = {
230{ "stable", BOND_AD_STABLE},
231{ "bandwidth", BOND_AD_BANDWIDTH},
232{ "count", BOND_AD_COUNT},
233{ NULL, -1},
234};
235
1da177e4
LT		 236/*-------------------------- Forward declarations ---------------------------*/
237
c3ade5ca 238static void bond_send_gratuitous_arp(struct bonding *bond);
181470fc 239static int bond_init(struct net_device *bond_dev);
c67dfb29 240static void bond_uninit(struct net_device *bond_dev);
1da177e4
LT		 241
242/*---------------------------- General routines -----------------------------*/
243
4ad072c9 244static const char *bond_mode_name(int mode)
1da177e4 245{
77afc92b
HE		 246 static const char *names[] = {
247 [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)",
248 [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)",
249 [BOND_MODE_XOR] = "load balancing (xor)",
250 [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)",
3d632c3f 251 [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation",
77afc92b
HE		 252 [BOND_MODE_TLB] = "transmit load balancing",
253 [BOND_MODE_ALB] = "adaptive load balancing",
254 };
255
256 if (mode < 0 || mode > BOND_MODE_ALB)
1da177e4 257 return "unknown";
77afc92b
HE		 258
259 return names[mode];
1da177e4
LT		 260}
261
262/*---------------------------------- VLAN -----------------------------------*/
263
264/**
265 * bond_add_vlan - add a new vlan id on bond
266 * @bond: bond that got the notification
267 * @vlan_id: the vlan id to add
268 *
269 * Returns -ENOMEM if allocation failed.
270 */
271static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id)
272{
273 struct vlan_entry *vlan;
274
5a03cdb7 275 pr_debug("bond: %s, vlan id %d\n",
a4aee5c8 276 (bond ? bond->dev->name : "None"), vlan_id);
1da177e4 277
305d552a 278 vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL);
3d632c3f 279 if (!vlan)
1da177e4 280 return -ENOMEM;
1da177e4
LT		 281
282 INIT_LIST_HEAD(&vlan->vlan_list);
283 vlan->vlan_id = vlan_id;
284
285 write_lock_bh(&bond->lock);
286
287 list_add_tail(&vlan->vlan_list, &bond->vlan_list);
288
289 write_unlock_bh(&bond->lock);
290
5a03cdb7 291 pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name);
1da177e4
LT		 292
293 return 0;
294}
295
296/**
297 * bond_del_vlan - delete a vlan id from bond
298 * @bond: bond that got the notification
299 * @vlan_id: the vlan id to delete
300 *
301 * returns -ENODEV if @vlan_id was not found in @bond.
302 */
303static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id)
304{
0883beca 305 struct vlan_entry *vlan;
1da177e4
LT		 306 int res = -ENODEV;
307
5a03cdb7 308 pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id);
1da177e4
LT		 309
310 write_lock_bh(&bond->lock);
311
0883beca 312 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
1da177e4
LT		 313 if (vlan->vlan_id == vlan_id) {
314 list_del(&vlan->vlan_list);
315
58402054 316 if (bond_is_lb(bond))
1da177e4 317 bond_alb_clear_vlan(bond, vlan_id);
1da177e4 318
a4aee5c8
JP		 319 pr_debug("removed VLAN ID %d from bond %s\n",
320 vlan_id, bond->dev->name);
1da177e4
LT		 321
322 kfree(vlan);
323
324 if (list_empty(&bond->vlan_list) &&
325 (bond->slave_cnt == 0)) {
326 /* Last VLAN removed and no slaves, so
327 * restore block on adding VLANs. This will
328 * be removed once new slaves that are not
329 * VLAN challenged will be added.
330 */
331 bond->dev->features |= NETIF_F_VLAN_CHALLENGED;
332 }
333
334 res = 0;
335 goto out;
336 }
337 }
338
a4aee5c8
JP		 339 pr_debug("couldn't find VLAN ID %d in bond %s\n",
340 vlan_id, bond->dev->name);
1da177e4
LT		 341
342out:
343 write_unlock_bh(&bond->lock);
344 return res;
345}
346
347/**
348 * bond_has_challenged_slaves
349 * @bond: the bond we're working on
350 *
351 * Searches the slave list. Returns 1 if a vlan challenged slave
352 * was found, 0 otherwise.
353 *
354 * Assumes bond->lock is held.
355 */
356static int bond_has_challenged_slaves(struct bonding *bond)
357{
358 struct slave *slave;
359 int i;
360
361 bond_for_each_slave(bond, slave, i) {
362 if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 363 pr_debug("found VLAN challenged slave - %s\n",
a4aee5c8 364 slave->dev->name);
1da177e4
LT		 365 return 1;
366 }
367 }
368
5a03cdb7 369 pr_debug("no VLAN challenged slaves found\n");
1da177e4
LT		 370 return 0;
371}
372
373/**
374 * bond_next_vlan - safely skip to the next item in the vlans list.
375 * @bond: the bond we're working on
376 * @curr: item we're advancing from
377 *
378 * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL,
379 * or @curr->next otherwise (even if it is @curr itself again).
3d632c3f 380 *
1da177e4
LT		 381 * Caller must hold bond->lock
382 */
383struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr)
384{
385 struct vlan_entry *next, *last;
386
3d632c3f 387 if (list_empty(&bond->vlan_list))
1da177e4 388 return NULL;
1da177e4
LT		 389
390 if (!curr) {
391 next = list_entry(bond->vlan_list.next,
392 struct vlan_entry, vlan_list);
393 } else {
394 last = list_entry(bond->vlan_list.prev,
395 struct vlan_entry, vlan_list);
396 if (last == curr) {
397 next = list_entry(bond->vlan_list.next,
398 struct vlan_entry, vlan_list);
399 } else {
400 next = list_entry(curr->vlan_list.next,
401 struct vlan_entry, vlan_list);
402 }
403 }
404
405 return next;
406}
407
408/**
409 * bond_dev_queue_xmit - Prepare skb for xmit.
3d632c3f 410 *
1da177e4
LT		 411 * @bond: bond device that got this skb for tx.
412 * @skb: hw accel VLAN tagged skb to transmit
413 * @slave_dev: slave that is supposed to xmit this skbuff
3d632c3f 414 *
1da177e4
LT		 415 * When the bond gets an skb to transmit that is
416 * already hardware accelerated VLAN tagged, and it
417 * needs to relay this skb to a slave that is not
418 * hw accel capable, the skb needs to be "unaccelerated",
419 * i.e. strip the hwaccel tag and re-insert it as part
420 * of the payload.
421 */
3d632c3f
SH		 422int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb,
423 struct net_device *slave_dev)
1da177e4 424{
966bc6f4 425 unsigned short uninitialized_var(vlan_id);
1da177e4
LT		 426
427 if (!list_empty(&bond->vlan_list) &&
428 !(slave_dev->features & NETIF_F_HW_VLAN_TX) &&
429 vlan_get_tag(skb, &vlan_id) == 0) {
430 skb->dev = slave_dev;
431 skb = vlan_put_tag(skb, vlan_id);
432 if (!skb) {
433 /* vlan_put_tag() frees the skb in case of error,
434 * so return success here so the calling functions
435 * won't attempt to free is again.
436 */
437 return 0;
438 }
439 } else {
440 skb->dev = slave_dev;
441 }
442
443 skb->priority = 1;
f6dc31a8
WC		 444#ifdef CONFIG_NET_POLL_CONTROLLER
445 if (unlikely(bond->dev->priv_flags & IFF_IN_NETPOLL)) {
446 struct netpoll *np = bond->dev->npinfo->netpoll;
447 slave_dev->npinfo = bond->dev->npinfo;
448 np->real_dev = np->dev = skb->dev;
449 slave_dev->priv_flags |= IFF_IN_NETPOLL;
450 netpoll_send_skb(np, skb);
451 slave_dev->priv_flags &= ~IFF_IN_NETPOLL;
452 np->dev = bond->dev;
453 } else
454#endif
455 dev_queue_xmit(skb);
1da177e4
LT		 456
457 return 0;
458}
459
460/*
461 * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid
462 * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a
463 * lock because:
464 * a. This operation is performed in IOCTL context,
465 * b. The operation is protected by the RTNL semaphore in the 8021q code,
466 * c. Holding a lock with BH disabled while directly calling a base driver
467 * entry point is generally a BAD idea.
3d632c3f 468 *
1da177e4
LT		 469 * The design of synchronization/protection for this operation in the 8021q
470 * module is good for one or more VLAN devices over a single physical device
471 * and cannot be extended for a teaming solution like bonding, so there is a
472 * potential race condition here where a net device from the vlan group might
473 * be referenced (either by a base driver or the 8021q code) while it is being
474 * removed from the system. However, it turns out we're not making matters
475 * worse, and if it works for regular VLAN usage it will work here too.
476*/
477
478/**
479 * bond_vlan_rx_register - Propagates registration to slaves
480 * @bond_dev: bonding net device that got called
481 * @grp: vlan group being registered
482 */
3d632c3f
SH		 483static void bond_vlan_rx_register(struct net_device *bond_dev,
484 struct vlan_group *grp)
1da177e4 485{
454d7c9b 486 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 487 struct slave *slave;
488 int i;
489
490 bond->vlgrp = grp;
491
492 bond_for_each_slave(bond, slave, i) {
493 struct net_device *slave_dev = slave->dev;
eb7cc59a 494 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 495
496 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
eb7cc59a
SH		 497 slave_ops->ndo_vlan_rx_register) {
498 slave_ops->ndo_vlan_rx_register(slave_dev, grp);
1da177e4
LT		 499 }
500 }
501}
502
503/**
504 * bond_vlan_rx_add_vid - Propagates adding an id to slaves
505 * @bond_dev: bonding net device that got called
506 * @vid: vlan id being added
507 */
508static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid)
509{
454d7c9b 510 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 511 struct slave *slave;
512 int i, res;
513
514 bond_for_each_slave(bond, slave, i) {
515 struct net_device *slave_dev = slave->dev;
eb7cc59a 516 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 517
518 if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
eb7cc59a
SH		 519 slave_ops->ndo_vlan_rx_add_vid) {
520 slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid);
1da177e4
LT		 521 }
522 }
523
524 res = bond_add_vlan(bond, vid);
525 if (res) {
a4aee5c8 526 pr_err("%s: Error: Failed to add vlan id %d\n",
1da177e4
LT		 527 bond_dev->name, vid);
528 }
529}
530
531/**
532 * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
533 * @bond_dev: bonding net device that got called
534 * @vid: vlan id being removed
535 */
536static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid)
537{
454d7c9b 538 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 539 struct slave *slave;
540 struct net_device *vlan_dev;
541 int i, res;
542
543 bond_for_each_slave(bond, slave, i) {
544 struct net_device *slave_dev = slave->dev;
eb7cc59a 545 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 546
547 if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
eb7cc59a 548 slave_ops->ndo_vlan_rx_kill_vid) {
1da177e4
LT		 549 /* Save and then restore vlan_dev in the grp array,
550 * since the slave's driver might clear it.
551 */
5c15bdec 552 vlan_dev = vlan_group_get_device(bond->vlgrp, vid);
eb7cc59a 553 slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid);
5c15bdec 554 vlan_group_set_device(bond->vlgrp, vid, vlan_dev);
1da177e4
LT		 555 }
556 }
557
558 res = bond_del_vlan(bond, vid);
559 if (res) {
a4aee5c8 560 pr_err("%s: Error: Failed to remove vlan id %d\n",
1da177e4
LT		 561 bond_dev->name, vid);
562 }
563}
564
565static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev)
566{
567 struct vlan_entry *vlan;
eb7cc59a 568 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 569
570 write_lock_bh(&bond->lock);
571
eb7cc59a 572 if (list_empty(&bond->vlan_list))
1da177e4 573 goto out;
1da177e4
LT		 574
575 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
eb7cc59a
SH		 576 slave_ops->ndo_vlan_rx_register)
577 slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp);
1da177e4
LT		 578
579 if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
eb7cc59a 580 !(slave_ops->ndo_vlan_rx_add_vid))
1da177e4 581 goto out;
1da177e4 582
eb7cc59a
SH		 583 list_for_each_entry(vlan, &bond->vlan_list, vlan_list)
584 slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id);
1da177e4
LT		 585
586out:
587 write_unlock_bh(&bond->lock);
588}
589
3d632c3f
SH		 590static void bond_del_vlans_from_slave(struct bonding *bond,
591 struct net_device *slave_dev)
1da177e4 592{
eb7cc59a 593 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4
LT		 594 struct vlan_entry *vlan;
595 struct net_device *vlan_dev;
596
597 write_lock_bh(&bond->lock);
598
eb7cc59a 599 if (list_empty(&bond->vlan_list))
1da177e4 600 goto out;
1da177e4
LT		 601
602 if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
eb7cc59a 603 !(slave_ops->ndo_vlan_rx_kill_vid))
1da177e4 604 goto unreg;
1da177e4
LT		 605
606 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
607 /* Save and then restore vlan_dev in the grp array,
608 * since the slave's driver might clear it.
609 */
5c15bdec 610 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
eb7cc59a 611 slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id);
5c15bdec 612 vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev);
1da177e4
LT		 613 }
614
615unreg:
616 if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
eb7cc59a
SH		 617 slave_ops->ndo_vlan_rx_register)
618 slave_ops->ndo_vlan_rx_register(slave_dev, NULL);
1da177e4
LT		 619
620out:
621 write_unlock_bh(&bond->lock);
622}
623
624/*------------------------------- Link status -------------------------------*/
625
ff59c456
JV		 626/*
627 * Set the carrier state for the master according to the state of its
628 * slaves. If any slaves are up, the master is up. In 802.3ad mode,
629 * do special 802.3ad magic.
630 *
631 * Returns zero if carrier state does not change, nonzero if it does.
632 */
633static int bond_set_carrier(struct bonding *bond)
634{
635 struct slave *slave;
636 int i;
637
638 if (bond->slave_cnt == 0)
639 goto down;
640
641 if (bond->params.mode == BOND_MODE_8023AD)
642 return bond_3ad_set_carrier(bond);
643
644 bond_for_each_slave(bond, slave, i) {
645 if (slave->link == BOND_LINK_UP) {
646 if (!netif_carrier_ok(bond->dev)) {
647 netif_carrier_on(bond->dev);
648 return 1;
649 }
650 return 0;
651 }
652 }
653
654down:
655 if (netif_carrier_ok(bond->dev)) {
656 netif_carrier_off(bond->dev);
657 return 1;
658 }
659 return 0;
660}
661
1da177e4
LT		 662/*
663 * Get link speed and duplex from the slave's base driver
664 * using ethtool. If for some reason the call fails or the
665 * values are invalid, fake speed and duplex to 100/Full
666 * and return error.
667 */
668static int bond_update_speed_duplex(struct slave *slave)
669{
670 struct net_device *slave_dev = slave->dev;
1da177e4 671 struct ethtool_cmd etool;
61a44b9c 672 int res;
1da177e4
LT		 673
674 /* Fake speed and duplex */
675 slave->speed = SPEED_100;
676 slave->duplex = DUPLEX_FULL;
677
61a44b9c
MW		 678 if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings)
679 return -1;
1da177e4 680
61a44b9c
MW		 681 res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool);
682 if (res < 0)
1da177e4 683 return -1;
1da177e4 684
1da177e4
LT		 685 switch (etool.speed) {
686 case SPEED_10:
687 case SPEED_100:
688 case SPEED_1000:
94dbffd5 689 case SPEED_10000:
1da177e4
LT		 690 break;
691 default:
692 return -1;
693 }
694
695 switch (etool.duplex) {
696 case DUPLEX_FULL:
697 case DUPLEX_HALF:
698 break;
699 default:
700 return -1;
701 }
702
703 slave->speed = etool.speed;
704 slave->duplex = etool.duplex;
705
706 return 0;
707}
708
709/*
710 * if <dev> supports MII link status reporting, check its link status.
711 *
712 * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
3d632c3f 713 * depending upon the setting of the use_carrier parameter.
1da177e4
LT		 714 *
715 * Return either BMSR_LSTATUS, meaning that the link is up (or we
716 * can't tell and just pretend it is), or 0, meaning that the link is
717 * down.
718 *
719 * If reporting is non-zero, instead of faking link up, return -1 if
720 * both ETHTOOL and MII ioctls fail (meaning the device does not
721 * support them). If use_carrier is set, return whatever it says.
722 * It'd be nice if there was a good way to tell if a driver supports
723 * netif_carrier, but there really isn't.
724 */
3d632c3f
SH		 725static int bond_check_dev_link(struct bonding *bond,
726 struct net_device *slave_dev, int reporting)
1da177e4 727{
eb7cc59a 728 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
d9d52832 729 int (*ioctl)(struct net_device *, struct ifreq *, int);
1da177e4
LT		 730 struct ifreq ifr;
731 struct mii_ioctl_data *mii;
1da177e4 732
6c988853
PG		 733 if (!reporting && !netif_running(slave_dev))
734 return 0;
735
eb7cc59a 736 if (bond->params.use_carrier)
1da177e4 737 return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
1da177e4 738
29112f4e
JP		 739 /* Try to get link status using Ethtool first. */
740 if (slave_dev->ethtool_ops) {
741 if (slave_dev->ethtool_ops->get_link) {
742 u32 link;
743
744 link = slave_dev->ethtool_ops->get_link(slave_dev);
745
746 return link ? BMSR_LSTATUS : 0;
747 }
748 }
749
3d632c3f 750 /* Ethtool can't be used, fallback to MII ioctls. */
eb7cc59a 751 ioctl = slave_ops->ndo_do_ioctl;
1da177e4
LT		 752 if (ioctl) {
753 /* TODO: set pointer to correct ioctl on a per team member */
754 /* bases to make this more efficient. that is, once */
755 /* we determine the correct ioctl, we will always */
756 /* call it and not the others for that team */
757 /* member. */
758
759 /*
760 * We cannot assume that SIOCGMIIPHY will also read a
761 * register; not all network drivers (e.g., e100)
762 * support that.
763 */
764
765 /* Yes, the mii is overlaid on the ifreq.ifr_ifru */
766 strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
767 mii = if_mii(&ifr);
768 if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
769 mii->reg_num = MII_BMSR;
3d632c3f
SH		 770 if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0)
771 return mii->val_out & BMSR_LSTATUS;
1da177e4
LT		 772 }
773 }
774
1da177e4
LT		 775 /*
776 * If reporting, report that either there's no dev->do_ioctl,
61a44b9c 777 * or both SIOCGMIIREG and get_link failed (meaning that we
1da177e4
LT		 778 * cannot report link status). If not reporting, pretend
779 * we're ok.
780 */
3d632c3f 781 return reporting ? -1 : BMSR_LSTATUS;
1da177e4
LT		 782}
783
784/*----------------------------- Multicast list ------------------------------*/
785
1da177e4
LT		 786/*
787 * Push the promiscuity flag down to appropriate slaves
788 */
7e1a1ac1 789static int bond_set_promiscuity(struct bonding *bond, int inc)
1da177e4 790{
7e1a1ac1 791 int err = 0;
1da177e4
LT		 792 if (USES_PRIMARY(bond->params.mode)) {
793 /* write lock already acquired */
794 if (bond->curr_active_slave) {
7e1a1ac1
WC		 795 err = dev_set_promiscuity(bond->curr_active_slave->dev,
796 inc);
1da177e4
LT		 797 }
798 } else {
799 struct slave *slave;
800 int i;
801 bond_for_each_slave(bond, slave, i) {
7e1a1ac1
WC		 802 err = dev_set_promiscuity(slave->dev, inc);
803 if (err)
804 return err;
1da177e4
LT		 805 }
806 }
7e1a1ac1 807 return err;
1da177e4
LT		 808}
809
810/*
811 * Push the allmulti flag down to all slaves
812 */
7e1a1ac1 813static int bond_set_allmulti(struct bonding *bond, int inc)
1da177e4 814{
7e1a1ac1 815 int err = 0;
1da177e4
LT		 816 if (USES_PRIMARY(bond->params.mode)) {
817 /* write lock already acquired */
818 if (bond->curr_active_slave) {
7e1a1ac1
WC		 819 err = dev_set_allmulti(bond->curr_active_slave->dev,
820 inc);
1da177e4
LT		 821 }
822 } else {
823 struct slave *slave;
824 int i;
825 bond_for_each_slave(bond, slave, i) {
7e1a1ac1
WC		 826 err = dev_set_allmulti(slave->dev, inc);
827 if (err)
828 return err;
1da177e4
LT		 829 }
830 }
7e1a1ac1 831 return err;
1da177e4
LT		 832}
833
834/*
835 * Add a Multicast address to slaves
836 * according to mode
837 */
22bedad3 838static void bond_mc_add(struct bonding *bond, void *addr)
1da177e4
LT		 839{
840 if (USES_PRIMARY(bond->params.mode)) {
841 /* write lock already acquired */
3d632c3f 842 if (bond->curr_active_slave)
22bedad3 843 dev_mc_add(bond->curr_active_slave->dev, addr);
1da177e4
LT		 844 } else {
845 struct slave *slave;
846 int i;
3d632c3f
SH		 847
848 bond_for_each_slave(bond, slave, i)
22bedad3 849 dev_mc_add(slave->dev, addr);
1da177e4
LT		 850 }
851}
852
853/*
854 * Remove a multicast address from slave
855 * according to mode
856 */
22bedad3 857static void bond_mc_del(struct bonding *bond, void *addr)
1da177e4
LT		 858{
859 if (USES_PRIMARY(bond->params.mode)) {
860 /* write lock already acquired */
3d632c3f 861 if (bond->curr_active_slave)
22bedad3 862 dev_mc_del(bond->curr_active_slave->dev, addr);
1da177e4
LT		 863 } else {
864 struct slave *slave;
865 int i;
866 bond_for_each_slave(bond, slave, i) {
22bedad3 867 dev_mc_del(slave->dev, addr);
1da177e4
LT		 868 }
869 }
870}
871
a816c7c7
JV		 872
873/*
874 * Retrieve the list of registered multicast addresses for the bonding
875 * device and retransmit an IGMP JOIN request to the current active
876 * slave.
877 */
878static void bond_resend_igmp_join_requests(struct bonding *bond)
879{
880 struct in_device *in_dev;
881 struct ip_mc_list *im;
882
883 rcu_read_lock();
884 in_dev = __in_dev_get_rcu(bond->dev);
885 if (in_dev) {
3d632c3f 886 for (im = in_dev->mc_list; im; im = im->next)
a816c7c7 887 ip_mc_rejoin_group(im);
a816c7c7
JV		 888 }
889
890 rcu_read_unlock();
891}
892
1da177e4
LT		 893/*
894 * flush all members of flush->mc_list from device dev->mc_list
895 */
3d632c3f
SH		 896static void bond_mc_list_flush(struct net_device *bond_dev,
897 struct net_device *slave_dev)
1da177e4 898{
454d7c9b 899 struct bonding *bond = netdev_priv(bond_dev);
22bedad3 900 struct netdev_hw_addr *ha;
1da177e4 901
22bedad3
JP		 902 netdev_for_each_mc_addr(ha, bond_dev)
903 dev_mc_del(slave_dev, ha->addr);
1da177e4
LT		 904
905 if (bond->params.mode == BOND_MODE_8023AD) {
906 /* del lacpdu mc addr from mc list */
907 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
908
22bedad3 909 dev_mc_del(slave_dev, lacpdu_multicast);
1da177e4
LT		 910 }
911}
912
913/*--------------------------- Active slave change ---------------------------*/
914
915/*
916 * Update the mc list and multicast-related flags for the new and
917 * old active slaves (if any) according to the multicast mode, and
918 * promiscuous flags unconditionally.
919 */
3d632c3f
SH		 920static void bond_mc_swap(struct bonding *bond, struct slave *new_active,
921 struct slave *old_active)
1da177e4 922{
22bedad3 923 struct netdev_hw_addr *ha;
1da177e4 924
3d632c3f 925 if (!USES_PRIMARY(bond->params.mode))
1da177e4
LT		 926 /* nothing to do - mc list is already up-to-date on
927 * all slaves
928 */
929 return;
1da177e4
LT		 930
931 if (old_active) {
3d632c3f 932 if (bond->dev->flags & IFF_PROMISC)
1da177e4 933 dev_set_promiscuity(old_active->dev, -1);
1da177e4 934
3d632c3f 935 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 936 dev_set_allmulti(old_active->dev, -1);
1da177e4 937
22bedad3
JP		 938 netdev_for_each_mc_addr(ha, bond->dev)
939 dev_mc_del(old_active->dev, ha->addr);
1da177e4
LT		 940 }
941
942 if (new_active) {
7e1a1ac1 943 /* FIXME: Signal errors upstream. */
3d632c3f 944 if (bond->dev->flags & IFF_PROMISC)
1da177e4 945 dev_set_promiscuity(new_active->dev, 1);
1da177e4 946
3d632c3f 947 if (bond->dev->flags & IFF_ALLMULTI)
1da177e4 948 dev_set_allmulti(new_active->dev, 1);
1da177e4 949
22bedad3
JP		 950 netdev_for_each_mc_addr(ha, bond->dev)
951 dev_mc_add(new_active->dev, ha->addr);
a816c7c7 952 bond_resend_igmp_join_requests(bond);
1da177e4
LT		 953 }
954}
955
3915c1e8
JV		 956/*
957 * bond_do_fail_over_mac
958 *
959 * Perform special MAC address swapping for fail_over_mac settings
960 *
961 * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh.
962 */
963static void bond_do_fail_over_mac(struct bonding *bond,
964 struct slave *new_active,
965 struct slave *old_active)
1f78d9f9
HE		 966 __releases(&bond->curr_slave_lock)
967 __releases(&bond->lock)
968 __acquires(&bond->lock)
969 __acquires(&bond->curr_slave_lock)
3915c1e8
JV		 970{
971 u8 tmp_mac[ETH_ALEN];
972 struct sockaddr saddr;
973 int rv;
974
975 switch (bond->params.fail_over_mac) {
976 case BOND_FOM_ACTIVE:
977 if (new_active)
978 memcpy(bond->dev->dev_addr, new_active->dev->dev_addr,
979 new_active->dev->addr_len);
980 break;
981 case BOND_FOM_FOLLOW:
982 /*
983 * if new_active && old_active, swap them
984 * if just old_active, do nothing (going to no active slave)
985 * if just new_active, set new_active to bond's MAC
986 */
987 if (!new_active)
988 return;
989
990 write_unlock_bh(&bond->curr_slave_lock);
991 read_unlock(&bond->lock);
992
993 if (old_active) {
994 memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN);
995 memcpy(saddr.sa_data, old_active->dev->dev_addr,
996 ETH_ALEN);
997 saddr.sa_family = new_active->dev->type;
998 } else {
999 memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN);
1000 saddr.sa_family = bond->dev->type;
1001 }
1002
1003 rv = dev_set_mac_address(new_active->dev, &saddr);
1004 if (rv) {
a4aee5c8 1005 pr_err("%s: Error %d setting MAC of slave %s\n",
3915c1e8
JV		 1006 bond->dev->name, -rv, new_active->dev->name);
1007 goto out;
1008 }
1009
1010 if (!old_active)
1011 goto out;
1012
1013 memcpy(saddr.sa_data, tmp_mac, ETH_ALEN);
1014 saddr.sa_family = old_active->dev->type;
1015
1016 rv = dev_set_mac_address(old_active->dev, &saddr);
1017 if (rv)
a4aee5c8 1018 pr_err("%s: Error %d setting MAC of slave %s\n",
3915c1e8
JV		 1019 bond->dev->name, -rv, new_active->dev->name);
1020out:
1021 read_lock(&bond->lock);
1022 write_lock_bh(&bond->curr_slave_lock);
1023 break;
1024 default:
a4aee5c8 1025 pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n",
3915c1e8
JV		 1026 bond->dev->name, bond->params.fail_over_mac);
1027 break;
1028 }
1029
1030}
1031
a549952a
JP		 1032static bool bond_should_change_active(struct bonding *bond)
1033{
1034 struct slave *prim = bond->primary_slave;
1035 struct slave *curr = bond->curr_active_slave;
1036
1037 if (!prim || !curr || curr->link != BOND_LINK_UP)
1038 return true;
1039 if (bond->force_primary) {
1040 bond->force_primary = false;
1041 return true;
1042 }
1043 if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER &&
1044 (prim->speed < curr->speed ||
1045 (prim->speed == curr->speed && prim->duplex <= curr->duplex)))
1046 return false;
1047 if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE)
1048 return false;
1049 return true;
1050}
3915c1e8 1051
1da177e4
LT		 1052/**
1053 * find_best_interface - select the best available slave to be the active one
1054 * @bond: our bonding struct
1055 *
1056 * Warning: Caller must hold curr_slave_lock for writing.
1057 */
1058static struct slave *bond_find_best_slave(struct bonding *bond)
1059{
1060 struct slave *new_active, *old_active;
1061 struct slave *bestslave = NULL;
1062 int mintime = bond->params.updelay;
1063 int i;
1064
49b4ad92 1065 new_active = bond->curr_active_slave;
1da177e4
LT		 1066
1067 if (!new_active) { /* there were no active slaves left */
3d632c3f 1068 if (bond->slave_cnt > 0) /* found one slave */
1da177e4 1069 new_active = bond->first_slave;
3d632c3f 1070 else
1da177e4 1071 return NULL; /* still no slave, return NULL */
1da177e4
LT		 1072 }
1073
1da177e4 1074 if ((bond->primary_slave) &&
a549952a
JP		 1075 bond->primary_slave->link == BOND_LINK_UP &&
1076 bond_should_change_active(bond)) {
1da177e4
LT		 1077 new_active = bond->primary_slave;
1078 }
1079
1080 /* remember where to stop iterating over the slaves */
1081 old_active = new_active;
1082
1083 bond_for_each_slave_from(bond, new_active, i, old_active) {
b9f60253
JP		 1084 if (new_active->link == BOND_LINK_UP) {
1085 return new_active;
1086 } else if (new_active->link == BOND_LINK_BACK &&
1087 IS_UP(new_active->dev)) {
1088 /* link up, but waiting for stabilization */
1089 if (new_active->delay < mintime) {
1090 mintime = new_active->delay;
1091 bestslave = new_active;
1da177e4
LT		 1092 }
1093 }
1094 }
1095
1096 return bestslave;
1097}
1098
1099/**
1100 * change_active_interface - change the active slave into the specified one
1101 * @bond: our bonding struct
1102 * @new: the new slave to make the active one
1103 *
1104 * Set the new slave to the bond's settings and unset them on the old
1105 * curr_active_slave.
1106 * Setting include flags, mc-list, promiscuity, allmulti, etc.
1107 *
1108 * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
1109 * because it is apparently the best available slave we have, even though its
1110 * updelay hasn't timed out yet.
1111 *
3915c1e8
JV		 1112 * If new_active is not NULL, caller must hold bond->lock for read and
1113 * curr_slave_lock for write_bh.
1da177e4 1114 */
a77b5325 1115void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
1da177e4
LT		 1116{
1117 struct slave *old_active = bond->curr_active_slave;
1118
3d632c3f 1119 if (old_active == new_active)
1da177e4 1120 return;
1da177e4
LT		 1121
1122 if (new_active) {
b2220cad
JV		 1123 new_active->jiffies = jiffies;
1124
1da177e4
LT		 1125 if (new_active->link == BOND_LINK_BACK) {
1126 if (USES_PRIMARY(bond->params.mode)) {
a4aee5c8
JP		 1127 pr_info("%s: making interface %s the new active one %d ms earlier.\n",
1128 bond->dev->name, new_active->dev->name,
1129 (bond->params.updelay - new_active->delay) * bond->params.miimon);
1da177e4
LT		 1130 }
1131
1132 new_active->delay = 0;
1133 new_active->link = BOND_LINK_UP;
1da177e4 1134
3d632c3f 1135 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 1136 bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
1da177e4 1137
58402054 1138 if (bond_is_lb(bond))
1da177e4 1139 bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
1da177e4
LT		 1140 } else {
1141 if (USES_PRIMARY(bond->params.mode)) {
a4aee5c8
JP		 1142 pr_info("%s: making interface %s the new active one.\n",
1143 bond->dev->name, new_active->dev->name);
1da177e4
LT		 1144 }
1145 }
1146 }
1147
3d632c3f 1148 if (USES_PRIMARY(bond->params.mode))
1da177e4 1149 bond_mc_swap(bond, new_active, old_active);
1da177e4 1150
58402054 1151 if (bond_is_lb(bond)) {
1da177e4 1152 bond_alb_handle_active_change(bond, new_active);
8f903c70
JV		 1153 if (old_active)
1154 bond_set_slave_inactive_flags(old_active);
1155 if (new_active)
1156 bond_set_slave_active_flags(new_active);
1da177e4
LT		 1157 } else {
1158 bond->curr_active_slave = new_active;
1159 }
c3ade5ca
JV		 1160
1161 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
3d632c3f 1162 if (old_active)
c3ade5ca 1163 bond_set_slave_inactive_flags(old_active);
c3ade5ca
JV		 1164
1165 if (new_active) {
1166 bond_set_slave_active_flags(new_active);
2ab82852 1167
709f8a45
OG		 1168 if (bond->params.fail_over_mac)
1169 bond_do_fail_over_mac(bond, new_active,
1170 old_active);
3915c1e8 1171
709f8a45 1172 bond->send_grat_arp = bond->params.num_grat_arp;
b59f9f74 1173 bond_send_gratuitous_arp(bond);
01f3109d 1174
305d552a
BH		 1175 bond->send_unsol_na = bond->params.num_unsol_na;
1176 bond_send_unsolicited_na(bond);
1177
01f3109d
OG		 1178 write_unlock_bh(&bond->curr_slave_lock);
1179 read_unlock(&bond->lock);
1180
75c78500 1181 netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER);
01f3109d
OG		 1182
1183 read_lock(&bond->lock);
1184 write_lock_bh(&bond->curr_slave_lock);
7893b249 1185 }
c3ade5ca 1186 }
a2fd940f
AG		 1187
1188 /* resend IGMP joins since all were sent on curr_active_slave */
1189 if (bond->params.mode == BOND_MODE_ROUNDROBIN) {
1190 bond_resend_igmp_join_requests(bond);
1191 }
1da177e4
LT		 1192}
1193
1194/**
1195 * bond_select_active_slave - select a new active slave, if needed
1196 * @bond: our bonding struct
1197 *
3d632c3f 1198 * This functions should be called when one of the following occurs:
1da177e4
LT		 1199 * - The old curr_active_slave has been released or lost its link.
1200 * - The primary_slave has got its link back.
1201 * - A slave has got its link back and there's no old curr_active_slave.
1202 *
3915c1e8 1203 * Caller must hold bond->lock for read and curr_slave_lock for write_bh.
1da177e4 1204 */
a77b5325 1205void bond_select_active_slave(struct bonding *bond)
1da177e4
LT		 1206{
1207 struct slave *best_slave;
ff59c456 1208 int rv;
1da177e4
LT		 1209
1210 best_slave = bond_find_best_slave(bond);
1211 if (best_slave != bond->curr_active_slave) {
1212 bond_change_active_slave(bond, best_slave);
ff59c456
JV		 1213 rv = bond_set_carrier(bond);
1214 if (!rv)
1215 return;
1216
1217 if (netif_carrier_ok(bond->dev)) {
a4aee5c8
JP		 1218 pr_info("%s: first active interface up!\n",
1219 bond->dev->name);
ff59c456 1220 } else {
a4aee5c8
JP		 1221 pr_info("%s: now running without any active interface !\n",
1222 bond->dev->name);
ff59c456 1223 }
1da177e4
LT		 1224 }
1225}
1226
1227/*--------------------------- slave list handling ---------------------------*/
1228
1229/*
1230 * This function attaches the slave to the end of list.
1231 *
1232 * bond->lock held for writing by caller.
1233 */
1234static void bond_attach_slave(struct bonding *bond, struct slave *new_slave)
1235{
1236 if (bond->first_slave == NULL) { /* attaching the first slave */
1237 new_slave->next = new_slave;
1238 new_slave->prev = new_slave;
1239 bond->first_slave = new_slave;
1240 } else {
1241 new_slave->next = bond->first_slave;
1242 new_slave->prev = bond->first_slave->prev;
1243 new_slave->next->prev = new_slave;
1244 new_slave->prev->next = new_slave;
1245 }
1246
1247 bond->slave_cnt++;
1248}
1249
1250/*
1251 * This function detaches the slave from the list.
1252 * WARNING: no check is made to verify if the slave effectively
1253 * belongs to <bond>.
1254 * Nothing is freed on return, structures are just unchained.
1255 * If any slave pointer in bond was pointing to <slave>,
1256 * it should be changed by the calling function.
1257 *
1258 * bond->lock held for writing by caller.
1259 */
1260static void bond_detach_slave(struct bonding *bond, struct slave *slave)
1261{
3d632c3f 1262 if (slave->next)
1da177e4 1263 slave->next->prev = slave->prev;
1da177e4 1264
3d632c3f 1265 if (slave->prev)
1da177e4 1266 slave->prev->next = slave->next;
1da177e4
LT		 1267
1268 if (bond->first_slave == slave) { /* slave is the first slave */
1269 if (bond->slave_cnt > 1) { /* there are more slave */
1270 bond->first_slave = slave->next;
1271 } else {
1272 bond->first_slave = NULL; /* slave was the last one */
1273 }
1274 }
1275
1276 slave->next = NULL;
1277 slave->prev = NULL;
1278 bond->slave_cnt--;
1279}
1280
f6dc31a8
WC		 1281#ifdef CONFIG_NET_POLL_CONTROLLER
1282/*
1283 * You must hold read lock on bond->lock before calling this.
1284 */
1285static bool slaves_support_netpoll(struct net_device *bond_dev)
1286{
1287 struct bonding *bond = netdev_priv(bond_dev);
1288 struct slave *slave;
1289 int i = 0;
1290 bool ret = true;
1291
1292 bond_for_each_slave(bond, slave, i) {
1293 if ((slave->dev->priv_flags & IFF_DISABLE_NETPOLL) ||
1294 !slave->dev->netdev_ops->ndo_poll_controller)
1295 ret = false;
1296 }
1297 return i != 0 && ret;
1298}
1299
1300static void bond_poll_controller(struct net_device *bond_dev)
1301{
1302 struct net_device *dev = bond_dev->npinfo->netpoll->real_dev;
1303 if (dev != bond_dev)
1304 netpoll_poll_dev(dev);
1305}
1306
1307static void bond_netpoll_cleanup(struct net_device *bond_dev)
1308{
1309 struct bonding *bond = netdev_priv(bond_dev);
1310 struct slave *slave;
1311 const struct net_device_ops *ops;
1312 int i;
1313
1314 read_lock(&bond->lock);
1315 bond_dev->npinfo = NULL;
1316 bond_for_each_slave(bond, slave, i) {
1317 if (slave->dev) {
1318 ops = slave->dev->netdev_ops;
1319 if (ops->ndo_netpoll_cleanup)
1320 ops->ndo_netpoll_cleanup(slave->dev);
1321 else
1322 slave->dev->npinfo = NULL;
1323 }
1324 }
1325 read_unlock(&bond->lock);
1326}
1327
1328#else
1329
1330static void bond_netpoll_cleanup(struct net_device *bond_dev)
1331{
1332}
1333
1334#endif
1335
1da177e4
LT		 1336/*---------------------------------- IOCTL ----------------------------------*/
1337
4ad072c9
AB		 1338static int bond_sethwaddr(struct net_device *bond_dev,
1339 struct net_device *slave_dev)
1da177e4 1340{
5a03cdb7
HE		 1341 pr_debug("bond_dev=%p\n", bond_dev);
1342 pr_debug("slave_dev=%p\n", slave_dev);
1343 pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len);
1da177e4
LT		 1344 memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
1345 return 0;
1346}
1347
7f353bf2
HX		 1348#define BOND_VLAN_FEATURES \
1349 (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \
1350 NETIF_F_HW_VLAN_FILTER)
8531c5ff 1351
3d632c3f 1352/*
8e3babcd 1353 * Compute the common dev->feature set available to all slaves. Some
7f353bf2
HX		 1354 * feature bits are managed elsewhere, so preserve those feature bits
1355 * on the master device.
8531c5ff
AK		 1356 */
1357static int bond_compute_features(struct bonding *bond)
1358{
8531c5ff
AK		 1359 struct slave *slave;
1360 struct net_device *bond_dev = bond->dev;
7f353bf2 1361 unsigned long features = bond_dev->features;
278339a4 1362 unsigned long vlan_features = 0;
3158bf7d
MS		 1363 unsigned short max_hard_header_len = max((u16)ETH_HLEN,
1364 bond_dev->hard_header_len);
8e3babcd 1365 int i;
8531c5ff 1366
7f353bf2 1367 features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES);
b63365a2
HX		 1368 features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM;
1369
1370 if (!bond->first_slave)
1371 goto done;
1372
1373 features &= ~NETIF_F_ONE_FOR_ALL;
7f353bf2 1374
278339a4 1375 vlan_features = bond->first_slave->dev->vlan_features;
54ef3137 1376 bond_for_each_slave(bond, slave, i) {
b63365a2
HX		 1377 features = netdev_increment_features(features,
1378 slave->dev->features,
1379 NETIF_F_ONE_FOR_ALL);
278339a4
JV		 1380 vlan_features = netdev_increment_features(vlan_features,
1381 slave->dev->vlan_features,
1382 NETIF_F_ONE_FOR_ALL);
54ef3137
JV		 1383 if (slave->dev->hard_header_len > max_hard_header_len)
1384 max_hard_header_len = slave->dev->hard_header_len;
1385 }
8531c5ff 1386
b63365a2 1387done:
7f353bf2 1388 features |= (bond_dev->features & BOND_VLAN_FEATURES);
b63365a2 1389 bond_dev->features = netdev_fix_features(features, NULL);
278339a4 1390 bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL);
54ef3137 1391 bond_dev->hard_header_len = max_hard_header_len;
8531c5ff
AK		 1392
1393 return 0;
1394}
1395
872254dd
MS		 1396static void bond_setup_by_slave(struct net_device *bond_dev,
1397 struct net_device *slave_dev)
1398{
454d7c9b 1399 struct bonding *bond = netdev_priv(bond_dev);
d90a162a 1400
00829823 1401 bond_dev->header_ops = slave_dev->header_ops;
872254dd
MS		 1402
1403 bond_dev->type = slave_dev->type;
1404 bond_dev->hard_header_len = slave_dev->hard_header_len;
1405 bond_dev->addr_len = slave_dev->addr_len;
1406
1407 memcpy(bond_dev->broadcast, slave_dev->broadcast,
1408 slave_dev->addr_len);
d90a162a 1409 bond->setup_by_slave = 1;
872254dd
MS		 1410}
1411
1da177e4 1412/* enslave device <slave> to bond device <master> */
a77b5325 1413int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1414{
454d7c9b 1415 struct bonding *bond = netdev_priv(bond_dev);
eb7cc59a 1416 const struct net_device_ops *slave_ops = slave_dev->netdev_ops;
1da177e4 1417 struct slave *new_slave = NULL;
22bedad3 1418 struct netdev_hw_addr *ha;
1da177e4
LT		 1419 struct sockaddr addr;
1420 int link_reporting;
1421 int old_features = bond_dev->features;
1422 int res = 0;
1423
552709d5 1424 if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL &&
eb7cc59a 1425 slave_ops->ndo_do_ioctl == NULL) {
a4aee5c8
JP		 1426 pr_warning("%s: Warning: no link monitoring support for %s\n",
1427 bond_dev->name, slave_dev->name);
1da177e4
LT		 1428 }
1429
1430 /* bond must be initialized by bond_open() before enslaving */
1431 if (!(bond_dev->flags & IFF_UP)) {
a4aee5c8
JP		 1432 pr_warning("%s: master_dev is not up in bond_enslave\n",
1433 bond_dev->name);
1da177e4
LT		 1434 }
1435
1436 /* already enslaved */
1437 if (slave_dev->flags & IFF_SLAVE) {
5a03cdb7 1438 pr_debug("Error, Device was already enslaved\n");
1da177e4
LT		 1439 return -EBUSY;
1440 }
1441
1442 /* vlan challenged mutual exclusion */
1443 /* no need to lock since we're protected by rtnl_lock */
1444 if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
5a03cdb7 1445 pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4 1446 if (!list_empty(&bond->vlan_list)) {
a4aee5c8
JP		 1447 pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n",
1448 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1449 return -EPERM;
1450 } else {
a4aee5c8
JP		 1451 pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n",
1452 bond_dev->name, slave_dev->name,
1453 slave_dev->name, bond_dev->name);
1da177e4
LT		 1454 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
1455 }
1456 } else {
5a03cdb7 1457 pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
1da177e4
LT		 1458 if (bond->slave_cnt == 0) {
1459 /* First slave, and it is not VLAN challenged,
1460 * so remove the block of adding VLANs over the bond.
1461 */
1462 bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
1463 }
1464 }
1465
217df670
JV		 1466 /*
1467 * Old ifenslave binaries are no longer supported. These can
3d632c3f 1468 * be identified with moderate accuracy by the state of the slave:
217df670
JV		 1469 * the current ifenslave will set the interface down prior to
1470 * enslaving it; the old ifenslave will not.
1471 */
1472 if ((slave_dev->flags & IFF_UP)) {
a4aee5c8 1473 pr_err("%s is up. This may be due to an out of date ifenslave.\n",
217df670
JV		 1474 slave_dev->name);
1475 res = -EPERM;
1476 goto err_undo_flags;
1477 }
1da177e4 1478
872254dd
MS		 1479 /* set bonding device ether type by slave - bonding netdevices are
1480 * created with ether_setup, so when the slave type is not ARPHRD_ETHER
1481 * there is a need to override some of the type dependent attribs/funcs.
1482 *
1483 * bond ether type mutual exclusion - don't allow slaves of dissimilar
1484 * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond
1485 */
1486 if (bond->slave_cnt == 0) {
e36b9d16 1487 if (bond_dev->type != slave_dev->type) {
e36b9d16 1488 pr_debug("%s: change device type from %d to %d\n",
a4aee5c8
JP		 1489 bond_dev->name,
1490 bond_dev->type, slave_dev->type);
75c78500 1491
3ca5b404
JP		 1492 res = netdev_bonding_change(bond_dev,
1493 NETDEV_PRE_TYPE_CHANGE);
1494 res = notifier_to_errno(res);
1495 if (res) {
1496 pr_err("%s: refused to change device type\n",
1497 bond_dev->name);
1498 res = -EBUSY;
1499 goto err_undo_flags;
1500 }
75c78500 1501
32a806c1 1502 /* Flush unicast and multicast addresses */
a748ee24 1503 dev_uc_flush(bond_dev);
22bedad3 1504 dev_mc_flush(bond_dev);
32a806c1 1505
e36b9d16
MS		 1506 if (slave_dev->type != ARPHRD_ETHER)
1507 bond_setup_by_slave(bond_dev, slave_dev);
1508 else
1509 ether_setup(bond_dev);
75c78500 1510
93d9b7d7
JP		 1511 netdev_bonding_change(bond_dev,
1512 NETDEV_POST_TYPE_CHANGE);
e36b9d16 1513 }
872254dd 1514 } else if (bond_dev->type != slave_dev->type) {
a4aee5c8
JP		 1515 pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n",
1516 slave_dev->name,
1517 slave_dev->type, bond_dev->type);
1518 res = -EINVAL;
1519 goto err_undo_flags;
872254dd
MS		 1520 }
1521
eb7cc59a 1522 if (slave_ops->ndo_set_mac_address == NULL) {
2ab82852 1523 if (bond->slave_cnt == 0) {
a4aee5c8
JP		 1524 pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.",
1525 bond_dev->name);
3915c1e8
JV		 1526 bond->params.fail_over_mac = BOND_FOM_ACTIVE;
1527 } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
a4aee5c8
JP		 1528 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",
1529 bond_dev->name);
2ab82852
MS		 1530 res = -EOPNOTSUPP;
1531 goto err_undo_flags;
1532 }
1da177e4
LT		 1533 }
1534
c20811a7
JP		 1535 /* If this is the first slave, then we need to set the master's hardware
1536 * address to be the same as the slave's. */
1537 if (bond->slave_cnt == 0)
1538 memcpy(bond->dev->dev_addr, slave_dev->dev_addr,
1539 slave_dev->addr_len);
1540
1541
243cb4e5 1542 new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL);
1da177e4
LT		 1543 if (!new_slave) {
1544 res = -ENOMEM;
1545 goto err_undo_flags;
1546 }
1547
bb1d9123
AG		 1548 /*
1549 * Set the new_slave's queue_id to be zero. Queue ID mapping
1550 * is set via sysfs or module option if desired.
1551 */
1552 new_slave->queue_id = 0;
1553
b15ba0fb
JP		 1554 /* Save slave's original mtu and then set it to match the bond */
1555 new_slave->original_mtu = slave_dev->mtu;
1556 res = dev_set_mtu(slave_dev, bond->dev->mtu);
1557 if (res) {
1558 pr_debug("Error %d calling dev_set_mtu\n", res);
1559 goto err_free;
1560 }
1561
217df670
JV		 1562 /*
1563 * Save slave's original ("permanent") mac address for modes
1564 * that need it, and for restoring it upon release, and then
1565 * set it to the master's address
1566 */
1567 memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN);
1da177e4 1568
dd957c57 1569 if (!bond->params.fail_over_mac) {
2ab82852
MS		 1570 /*
1571 * Set slave to master's mac address. The application already
1572 * set the master's mac address to that of the first slave
1573 */
1574 memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
1575 addr.sa_family = slave_dev->type;
1576 res = dev_set_mac_address(slave_dev, &addr);
1577 if (res) {
5a03cdb7 1578 pr_debug("Error %d calling set_mac_address\n", res);
b15ba0fb 1579 goto err_restore_mtu;
2ab82852 1580 }
217df670 1581 }
1da177e4 1582
c2edacf8
JV		 1583 res = netdev_set_master(slave_dev, bond_dev);
1584 if (res) {
5a03cdb7 1585 pr_debug("Error %d calling netdev_set_master\n", res);
569f0c4d 1586 goto err_restore_mac;
c2edacf8 1587 }
217df670
JV		 1588 /* open the slave since the application closed it */
1589 res = dev_open(slave_dev);
1590 if (res) {
3d632c3f 1591 pr_debug("Opening slave %s failed\n", slave_dev->name);
569f0c4d 1592 goto err_unset_master;
1da177e4
LT		 1593 }
1594
1da177e4 1595 new_slave->dev = slave_dev;
0b680e75 1596 slave_dev->priv_flags |= IFF_BONDING;
1da177e4 1597
58402054 1598 if (bond_is_lb(bond)) {
1da177e4
LT		 1599 /* bond_alb_init_slave() must be called before all other stages since
1600 * it might fail and we do not want to have to undo everything
1601 */
1602 res = bond_alb_init_slave(bond, new_slave);
3d632c3f 1603 if (res)
569f0c4d 1604 goto err_close;
1da177e4
LT		 1605 }
1606
1607 /* If the mode USES_PRIMARY, then the new slave gets the
1608 * master's promisc (and mc) settings only if it becomes the
1609 * curr_active_slave, and that is taken care of later when calling
1610 * bond_change_active()
1611 */
1612 if (!USES_PRIMARY(bond->params.mode)) {
1613 /* set promiscuity level to new slave */
1614 if (bond_dev->flags & IFF_PROMISC) {
7e1a1ac1
WC		 1615 res = dev_set_promiscuity(slave_dev, 1);
1616 if (res)
1617 goto err_close;
1da177e4
LT		 1618 }
1619
1620 /* set allmulti level to new slave */
1621 if (bond_dev->flags & IFF_ALLMULTI) {
7e1a1ac1
WC		 1622 res = dev_set_allmulti(slave_dev, 1);
1623 if (res)
1624 goto err_close;
1da177e4
LT		 1625 }
1626
b9e40857 1627 netif_addr_lock_bh(bond_dev);
1da177e4 1628 /* upload master's mc_list to new slave */
22bedad3
JP		 1629 netdev_for_each_mc_addr(ha, bond_dev)
1630 dev_mc_add(slave_dev, ha->addr);
b9e40857 1631 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1632 }
1633
1634 if (bond->params.mode == BOND_MODE_8023AD) {
1635 /* add lacpdu mc addr to mc list */
1636 u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
1637
22bedad3 1638 dev_mc_add(slave_dev, lacpdu_multicast);
1da177e4
LT		 1639 }
1640
1641 bond_add_vlans_on_slave(bond, slave_dev);
1642
1643 write_lock_bh(&bond->lock);
1644
1645 bond_attach_slave(bond, new_slave);
1646
1647 new_slave->delay = 0;
1648 new_slave->link_failure_count = 0;
1649
8531c5ff
AK		 1650 bond_compute_features(bond);
1651
3915c1e8
JV		 1652 write_unlock_bh(&bond->lock);
1653
1654 read_lock(&bond->lock);
1655
f5b2b966
JV		 1656 new_slave->last_arp_rx = jiffies;
1657
1da177e4
LT		 1658 if (bond->params.miimon && !bond->params.use_carrier) {
1659 link_reporting = bond_check_dev_link(bond, slave_dev, 1);
1660
1661 if ((link_reporting == -1) && !bond->params.arp_interval) {
1662 /*
1663 * miimon is set but a bonded network driver
1664 * does not support ETHTOOL/MII and
1665 * arp_interval is not set. Note: if
1666 * use_carrier is enabled, we will never go
1667 * here (because netif_carrier is always
1668 * supported); thus, we don't need to change
1669 * the messages for netif_carrier.
1670 */
a4aee5c8 1671 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 1672 bond_dev->name, slave_dev->name);
1da177e4
LT		 1673 } else if (link_reporting == -1) {
1674 /* unable get link status using mii/ethtool */
a4aee5c8
JP		 1675 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",
1676 bond_dev->name, slave_dev->name);
1da177e4
LT		 1677 }
1678 }
1679
1680 /* check for initial state */
1681 if (!bond->params.miimon ||
1682 (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) {
1683 if (bond->params.updelay) {
a4aee5c8 1684 pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n");
1da177e4
LT		 1685 new_slave->link = BOND_LINK_BACK;
1686 new_slave->delay = bond->params.updelay;
1687 } else {
a4aee5c8 1688 pr_debug("Initial state of slave_dev is BOND_LINK_UP\n");
1da177e4
LT		 1689 new_slave->link = BOND_LINK_UP;
1690 }
1691 new_slave->jiffies = jiffies;
1692 } else {
a4aee5c8 1693 pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n");
1da177e4
LT		 1694 new_slave->link = BOND_LINK_DOWN;
1695 }
1696
1697 if (bond_update_speed_duplex(new_slave) &&
1698 (new_slave->link != BOND_LINK_DOWN)) {
a4aee5c8
JP		 1699 pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n",
1700 bond_dev->name, new_slave->dev->name);
1da177e4
LT		 1701
1702 if (bond->params.mode == BOND_MODE_8023AD) {
a4aee5c8
JP		 1703 pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n",
1704 bond_dev->name);
1da177e4
LT		 1705 }
1706 }
1707
1708 if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) {
1709 /* if there is a primary slave, remember it */
a549952a 1710 if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
1da177e4 1711 bond->primary_slave = new_slave;
a549952a
JP		 1712 bond->force_primary = true;
1713 }
1da177e4
LT		 1714 }
1715
3915c1e8
JV		 1716 write_lock_bh(&bond->curr_slave_lock);
1717
1da177e4
LT		 1718 switch (bond->params.mode) {
1719 case BOND_MODE_ACTIVEBACKUP:
8a8e447b
JV		 1720 bond_set_slave_inactive_flags(new_slave);
1721 bond_select_active_slave(bond);
1da177e4
LT		 1722 break;
1723 case BOND_MODE_8023AD:
1724 /* in 802.3ad mode, the internal mechanism
1725 * will activate the slaves in the selected
1726 * aggregator
1727 */
1728 bond_set_slave_inactive_flags(new_slave);
1729 /* if this is the first slave */
1730 if (bond->slave_cnt == 1) {
1731 SLAVE_AD_INFO(new_slave).id = 1;
1732 /* Initialize AD with the number of times that the AD timer is called in 1 second
1733 * can be called only after the mac address of the bond is set
1734 */
1735 bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL,
1736 bond->params.lacp_fast);
1737 } else {
1738 SLAVE_AD_INFO(new_slave).id =
1739 SLAVE_AD_INFO(new_slave->prev).id + 1;
1740 }
1741
1742 bond_3ad_bind_slave(new_slave);
1743 break;
1744 case BOND_MODE_TLB:
1745 case BOND_MODE_ALB:
1746 new_slave->state = BOND_STATE_ACTIVE;
059fe7a5 1747 bond_set_slave_inactive_flags(new_slave);
5a29f789 1748 bond_select_active_slave(bond);
1da177e4
LT		 1749 break;
1750 default:
5a03cdb7 1751 pr_debug("This slave is always active in trunk mode\n");
1da177e4
LT		 1752
1753 /* always active in trunk mode */
1754 new_slave->state = BOND_STATE_ACTIVE;
1755
1756 /* In trunking mode there is little meaning to curr_active_slave
1757 * anyway (it holds no special properties of the bond device),
1758 * so we can change it without calling change_active_interface()
1759 */
3d632c3f 1760 if (!bond->curr_active_slave)
1da177e4 1761 bond->curr_active_slave = new_slave;
3d632c3f 1762
1da177e4
LT		 1763 break;
1764 } /* switch(bond_mode) */
1765
3915c1e8
JV		 1766 write_unlock_bh(&bond->curr_slave_lock);
1767
ff59c456
JV		 1768 bond_set_carrier(bond);
1769
f6dc31a8 1770#ifdef CONFIG_NET_POLL_CONTROLLER
c22d7ac8
AG		 1771 /*
1772 * Netpoll and bonding is broken, make sure it is not initialized
1773 * until it is fixed.
1774 */
1775 if (disable_netpoll) {
f6dc31a8 1776 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
c22d7ac8
AG		 1777 } else {
1778 if (slaves_support_netpoll(bond_dev)) {
1779 bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL;
1780 if (bond_dev->npinfo)
1781 slave_dev->npinfo = bond_dev->npinfo;
1782 } else if (!(bond_dev->priv_flags & IFF_DISABLE_NETPOLL)) {
1783 bond_dev->priv_flags |= IFF_DISABLE_NETPOLL;
1784 pr_info("New slave device %s does not support netpoll\n",
1785 slave_dev->name);
1786 pr_info("Disabling netpoll support for %s\n", bond_dev->name);
1787 }
f6dc31a8
WC		 1788 }
1789#endif
3915c1e8 1790 read_unlock(&bond->lock);
1da177e4 1791
b76cdba9
MW		 1792 res = bond_create_slave_symlinks(bond_dev, slave_dev);
1793 if (res)
569f0c4d 1794 goto err_close;
b76cdba9 1795
a4aee5c8
JP		 1796 pr_info("%s: enslaving %s as a%s interface with a%s link.\n",
1797 bond_dev->name, slave_dev->name,
1798 new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup",
1799 new_slave->link != BOND_LINK_DOWN ? "n up" : " down");
1da177e4
LT		 1800
1801 /* enslave is successful */
1802 return 0;
1803
1804/* Undo stages on error */
1da177e4
LT		 1805err_close:
1806 dev_close(slave_dev);
1807
569f0c4d
JV		 1808err_unset_master:
1809 netdev_set_master(slave_dev, NULL);
1810
1da177e4 1811err_restore_mac:
dd957c57 1812 if (!bond->params.fail_over_mac) {
3915c1e8
JV		 1813 /* XXX TODO - fom follow mode needs to change master's
1814 * MAC if this slave's MAC is in use by the bond, or at
1815 * least print a warning.
1816 */
2ab82852
MS		 1817 memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN);
1818 addr.sa_family = slave_dev->type;
1819 dev_set_mac_address(slave_dev, &addr);
1820 }
1da177e4 1821
b15ba0fb
JP		 1822err_restore_mtu:
1823 dev_set_mtu(slave_dev, new_slave->original_mtu);
1824
1da177e4
LT		 1825err_free:
1826 kfree(new_slave);
1827
1828err_undo_flags:
1829 bond_dev->features = old_features;
3d632c3f 1830
1da177e4
LT		 1831 return res;
1832}
1833
1834/*
1835 * Try to release the slave device <slave> from the bond device <master>
1836 * It is legal to access curr_active_slave without a lock because all the function
1837 * is write-locked.
1838 *
1839 * The rules for slave state should be:
1840 * for Active/Backup:
1841 * Active stays on all backups go down
1842 * for Bonded connections:
1843 * The first up interface should be left on and all others downed.
1844 */
a77b5325 1845int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
1da177e4 1846{
454d7c9b 1847 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 1848 struct slave *slave, *oldcurrent;
1849 struct sockaddr addr;
1da177e4
LT		 1850
1851 /* slave is not a slave or master is not master of this slave */
1852 if (!(slave_dev->flags & IFF_SLAVE) ||
1853 (slave_dev->master != bond_dev)) {
a4aee5c8 1854 pr_err("%s: Error: cannot release %s.\n",
1da177e4
LT		 1855 bond_dev->name, slave_dev->name);
1856 return -EINVAL;
1857 }
1858
f6dc31a8 1859 netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE);
1da177e4
LT		 1860 write_lock_bh(&bond->lock);
1861
1862 slave = bond_get_slave_by_dev(bond, slave_dev);
1863 if (!slave) {
1864 /* not a slave of this bond */
a4aee5c8
JP		 1865 pr_info("%s: %s not enslaved\n",
1866 bond_dev->name, slave_dev->name);
f5e2a7b2 1867 write_unlock_bh(&bond->lock);
1da177e4
LT		 1868 return -EINVAL;
1869 }
1870
3915c1e8 1871 if (!bond->params.fail_over_mac) {
8e95a202
JP		 1872 if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) &&
1873 bond->slave_cnt > 1)
a4aee5c8
JP		 1874 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",
1875 bond_dev->name, slave_dev->name,
1876 slave->perm_hwaddr,
1877 bond_dev->name, slave_dev->name);
1da177e4
LT		 1878 }
1879
1880 /* Inform AD package of unbinding of slave. */
1881 if (bond->params.mode == BOND_MODE_8023AD) {
1882 /* must be called before the slave is
1883 * detached from the list
1884 */
1885 bond_3ad_unbind_slave(slave);
1886 }
1887
a4aee5c8
JP		 1888 pr_info("%s: releasing %s interface %s\n",
1889 bond_dev->name,
1890 (slave->state == BOND_STATE_ACTIVE) ? "active" : "backup",
1891 slave_dev->name);
1da177e4
LT		 1892
1893 oldcurrent = bond->curr_active_slave;
1894
1895 bond->current_arp_slave = NULL;
1896
1897 /* release the slave from its bond */
1898 bond_detach_slave(bond, slave);
1899
8531c5ff
AK		 1900 bond_compute_features(bond);
1901
3d632c3f 1902 if (bond->primary_slave == slave)
1da177e4 1903 bond->primary_slave = NULL;
1da177e4 1904
3d632c3f 1905 if (oldcurrent == slave)
1da177e4 1906 bond_change_active_slave(bond, NULL);
1da177e4 1907
58402054 1908 if (bond_is_lb(bond)) {
1da177e4
LT		 1909 /* Must be called only after the slave has been
1910 * detached from the list and the curr_active_slave
1911 * has been cleared (if our_slave == old_current),
1912 * but before a new active slave is selected.
1913 */
2543331d 1914 write_unlock_bh(&bond->lock);
1da177e4 1915 bond_alb_deinit_slave(bond, slave);
2543331d 1916 write_lock_bh(&bond->lock);
1da177e4
LT		 1917 }
1918
059fe7a5
JV		 1919 if (oldcurrent == slave) {
1920 /*
1921 * Note that we hold RTNL over this sequence, so there
1922 * is no concern that another slave add/remove event
1923 * will interfere.
1924 */
1925 write_unlock_bh(&bond->lock);
1926 read_lock(&bond->lock);
1927 write_lock_bh(&bond->curr_slave_lock);
1928
1da177e4
LT		 1929 bond_select_active_slave(bond);
1930
059fe7a5
JV		 1931 write_unlock_bh(&bond->curr_slave_lock);
1932 read_unlock(&bond->lock);
1933 write_lock_bh(&bond->lock);
1934 }
1935
1da177e4 1936 if (bond->slave_cnt == 0) {
ff59c456
JV		 1937 bond_set_carrier(bond);
1938
1da177e4
LT		 1939 /* if the last slave was removed, zero the mac address
1940 * of the master so it will be set by the application
1941 * to the mac address of the first slave
1942 */
1943 memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
1944
1945 if (list_empty(&bond->vlan_list)) {
1946 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
1947 } else {
a4aee5c8
JP		 1948 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
1949 bond_dev->name, bond_dev->name);
1950 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
1951 bond_dev->name);
1da177e4
LT		 1952 }
1953 } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
1954 !bond_has_challenged_slaves(bond)) {
a4aee5c8
JP		 1955 pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n",
1956 bond_dev->name, slave_dev->name, bond_dev->name);
1da177e4
LT		 1957 bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
1958 }
1959
1960 write_unlock_bh(&bond->lock);
1961
b76cdba9
MW		 1962 /* must do this from outside any spinlocks */
1963 bond_destroy_slave_symlinks(bond_dev, slave_dev);
1964
1da177e4
LT		 1965 bond_del_vlans_from_slave(bond, slave_dev);
1966
1967 /* If the mode USES_PRIMARY, then we should only remove its
1968 * promisc and mc settings if it was the curr_active_slave, but that was
1969 * already taken care of above when we detached the slave
1970 */
1971 if (!USES_PRIMARY(bond->params.mode)) {
1972 /* unset promiscuity level from slave */
3d632c3f 1973 if (bond_dev->flags & IFF_PROMISC)
1da177e4 1974 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 1975
1976 /* unset allmulti level from slave */
3d632c3f 1977 if (bond_dev->flags & IFF_ALLMULTI)
1da177e4 1978 dev_set_allmulti(slave_dev, -1);
1da177e4
LT		 1979
1980 /* flush master's mc_list from slave */
b9e40857 1981 netif_addr_lock_bh(bond_dev);
1da177e4 1982 bond_mc_list_flush(bond_dev, slave_dev);
b9e40857 1983 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 1984 }
1985
1986 netdev_set_master(slave_dev, NULL);
1987
f6dc31a8
WC		 1988#ifdef CONFIG_NET_POLL_CONTROLLER
1989 read_lock_bh(&bond->lock);
c22d7ac8
AG		 1990
1991 /* Make sure netpoll over stays disabled until fixed. */
1992 if (!disable_netpoll)
1993 if (slaves_support_netpoll(bond_dev))
1994 bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL;
f6dc31a8
WC		 1995 read_unlock_bh(&bond->lock);
1996 if (slave_dev->netdev_ops->ndo_netpoll_cleanup)
1997 slave_dev->netdev_ops->ndo_netpoll_cleanup(slave_dev);
1998 else
1999 slave_dev->npinfo = NULL;
2000#endif
2001
1da177e4
LT		 2002 /* close slave before restoring its mac address */
2003 dev_close(slave_dev);
2004
3915c1e8 2005 if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) {
2ab82852
MS		 2006 /* restore original ("permanent") mac address */
2007 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
2008 addr.sa_family = slave_dev->type;
2009 dev_set_mac_address(slave_dev, &addr);
2010 }
1da177e4 2011
b15ba0fb
JP		 2012 dev_set_mtu(slave_dev, slave->original_mtu);
2013
8f903c70 2014 slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB |
f5b2b966
JV		 2015 IFF_SLAVE_INACTIVE | IFF_BONDING |
2016 IFF_SLAVE_NEEDARP);
1da177e4
LT		 2017
2018 kfree(slave);
2019
2020 return 0; /* deletion OK */
2021}
2022
d90a162a 2023/*
3d632c3f 2024* First release a slave and than destroy the bond if no more slaves are left.
d90a162a
MS		 2025* Must be under rtnl_lock when this function is called.
2026*/
3d632c3f
SH		 2027int bond_release_and_destroy(struct net_device *bond_dev,
2028 struct net_device *slave_dev)
d90a162a 2029{
454d7c9b 2030 struct bonding *bond = netdev_priv(bond_dev);
d90a162a
MS		 2031 int ret;
2032
2033 ret = bond_release(bond_dev, slave_dev);
2034 if ((ret == 0) && (bond->slave_cnt == 0)) {
a4aee5c8
JP		 2035 pr_info("%s: destroying bond %s.\n",
2036 bond_dev->name, bond_dev->name);
9e71626c 2037 unregister_netdevice(bond_dev);
d90a162a
MS		 2038 }
2039 return ret;
2040}
2041
1da177e4
LT		 2042/*
2043 * This function releases all slaves.
2044 */
2045static int bond_release_all(struct net_device *bond_dev)
2046{
454d7c9b 2047 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 2048 struct slave *slave;
2049 struct net_device *slave_dev;
2050 struct sockaddr addr;
2051
2052 write_lock_bh(&bond->lock);
2053
ff59c456
JV		 2054 netif_carrier_off(bond_dev);
2055
3d632c3f 2056 if (bond->slave_cnt == 0)
1da177e4 2057 goto out;
1da177e4
LT		 2058
2059 bond->current_arp_slave = NULL;
2060 bond->primary_slave = NULL;
2061 bond_change_active_slave(bond, NULL);
2062
2063 while ((slave = bond->first_slave) != NULL) {
2064 /* Inform AD package of unbinding of slave
2065 * before slave is detached from the list.
2066 */
3d632c3f 2067 if (bond->params.mode == BOND_MODE_8023AD)
1da177e4 2068 bond_3ad_unbind_slave(slave);
1da177e4
LT		 2069
2070 slave_dev = slave->dev;
2071 bond_detach_slave(bond, slave);
2072
2543331d
JV		 2073 /* now that the slave is detached, unlock and perform
2074 * all the undo steps that should not be called from
2075 * within a lock.
2076 */
2077 write_unlock_bh(&bond->lock);
2078
58402054 2079 if (bond_is_lb(bond)) {
1da177e4
LT		 2080 /* must be called only after the slave
2081 * has been detached from the list
2082 */
2083 bond_alb_deinit_slave(bond, slave);
2084 }
2085
8531c5ff
AK		 2086 bond_compute_features(bond);
2087
b76cdba9 2088 bond_destroy_slave_symlinks(bond_dev, slave_dev);
1da177e4
LT		 2089 bond_del_vlans_from_slave(bond, slave_dev);
2090
2091 /* If the mode USES_PRIMARY, then we should only remove its
2092 * promisc and mc settings if it was the curr_active_slave, but that was
2093 * already taken care of above when we detached the slave
2094 */
2095 if (!USES_PRIMARY(bond->params.mode)) {
2096 /* unset promiscuity level from slave */
3d632c3f 2097 if (bond_dev->flags & IFF_PROMISC)
1da177e4 2098 dev_set_promiscuity(slave_dev, -1);
1da177e4
LT		 2099
2100 /* unset allmulti level from slave */
3d632c3f 2101 if (bond_dev->flags & IFF_ALLMULTI)
1da177e4 2102 dev_set_allmulti(slave_dev, -1);
1da177e4
LT		 2103
2104 /* flush master's mc_list from slave */
b9e40857 2105 netif_addr_lock_bh(bond_dev);
1da177e4 2106 bond_mc_list_flush(bond_dev, slave_dev);
b9e40857 2107 netif_addr_unlock_bh(bond_dev);
1da177e4
LT		 2108 }
2109
2110 netdev_set_master(slave_dev, NULL);
2111
2112 /* close slave before restoring its mac address */
2113 dev_close(slave_dev);
2114
dd957c57 2115 if (!bond->params.fail_over_mac) {
2ab82852
MS		 2116 /* restore original ("permanent") mac address*/
2117 memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
2118 addr.sa_family = slave_dev->type;
2119 dev_set_mac_address(slave_dev, &addr);
2120 }
1da177e4 2121
8f903c70
JV		 2122 slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB |
2123 IFF_SLAVE_INACTIVE);
1da177e4
LT		 2124
2125 kfree(slave);
2126
2127 /* re-acquire the lock before getting the next slave */
2128 write_lock_bh(&bond->lock);
2129 }
2130
2131 /* zero the mac address of the master so it will be
2132 * set by the application to the mac address of the
2133 * first slave
2134 */
2135 memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
2136
3d632c3f 2137 if (list_empty(&bond->vlan_list))
1da177e4 2138 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
3d632c3f 2139 else {
a4aee5c8
JP		 2140 pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n",
2141 bond_dev->name, bond_dev->name);
2142 pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n",
2143 bond_dev->name);
1da177e4
LT		 2144 }
2145
a4aee5c8 2146 pr_info("%s: released all slaves\n", bond_dev->name);
1da177e4
LT		 2147
2148out:
2149 write_unlock_bh(&bond->lock);
2150
2151 return 0;
2152}
2153
2154/*
2155 * This function changes the active slave to slave <slave_dev>.
2156 * It returns -EINVAL in the following cases.
2157 * - <slave_dev> is not found in the list.
2158 * - There is not active slave now.
2159 * - <slave_dev> is already active.
2160 * - The link state of <slave_dev> is not BOND_LINK_UP.
2161 * - <slave_dev> is not running.
3d632c3f
SH		 2162 * In these cases, this function does nothing.
2163 * In the other cases, current_slave pointer is changed and 0 is returned.
1da177e4
LT		 2164 */
2165static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev)
2166{
454d7c9b 2167 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 2168 struct slave *old_active = NULL;
2169 struct slave *new_active = NULL;
2170 int res = 0;
2171
3d632c3f 2172 if (!USES_PRIMARY(bond->params.mode))
1da177e4 2173 return -EINVAL;
1da177e4
LT		 2174
2175 /* Verify that master_dev is indeed the master of slave_dev */
3d632c3f 2176 if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev))
1da177e4 2177 return -EINVAL;
1da177e4 2178
059fe7a5 2179 read_lock(&bond->lock);
1da177e4 2180
059fe7a5 2181 read_lock(&bond->curr_slave_lock);
1da177e4 2182 old_active = bond->curr_active_slave;
059fe7a5
JV		 2183 read_unlock(&bond->curr_slave_lock);
2184
1da177e4
LT		 2185 new_active = bond_get_slave_by_dev(bond, slave_dev);
2186
2187 /*
2188 * Changing to the current active: do nothing; return success.
2189 */
2190 if (new_active && (new_active == old_active)) {
059fe7a5 2191 read_unlock(&bond->lock);
1da177e4
LT		 2192 return 0;
2193 }
2194
2195 if ((new_active) &&
2196 (old_active) &&
2197 (new_active->link == BOND_LINK_UP) &&
2198 IS_UP(new_active->dev)) {
059fe7a5 2199 write_lock_bh(&bond->curr_slave_lock);
1da177e4 2200 bond_change_active_slave(bond, new_active);
059fe7a5 2201 write_unlock_bh(&bond->curr_slave_lock);
3d632c3f 2202 } else
1da177e4 2203 res = -EINVAL;
1da177e4 2204
059fe7a5 2205 read_unlock(&bond->lock);
1da177e4
LT		 2206
2207 return res;
2208}
2209
1da177e4
LT		 2210static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
2211{
454d7c9b 2212 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 2213
2214 info->bond_mode = bond->params.mode;
2215 info->miimon = bond->params.miimon;
2216
6603a6f2 2217 read_lock(&bond->lock);
1da177e4 2218 info->num_slaves = bond->slave_cnt;
6603a6f2 2219 read_unlock(&bond->lock);
1da177e4
LT		 2220
2221 return 0;
2222}
2223
2224static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
2225{
454d7c9b 2226 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 2227 struct slave *slave;
689c96cc 2228 int i, res = -ENODEV;
1da177e4 2229
6603a6f2 2230 read_lock(&bond->lock);
1da177e4
LT		 2231
2232 bond_for_each_slave(bond, slave, i) {
2233 if (i == (int)info->slave_id) {
689c96cc
ED		 2234 res = 0;
2235 strcpy(info->slave_name, slave->dev->name);
2236 info->link = slave->link;
2237 info->state = slave->state;
2238 info->link_failure_count = slave->link_failure_count;
1da177e4
LT		 2239 break;
2240 }
2241 }
2242
6603a6f2 2243 read_unlock(&bond->lock);
1da177e4 2244
689c96cc 2245 return res;
1da177e4
LT		 2246}
2247
2248/*-------------------------------- Monitoring -------------------------------*/
2249
1da177e4 2250
f0c76d61
JV		 2251static int bond_miimon_inspect(struct bonding *bond)
2252{
2253 struct slave *slave;
2254 int i, link_state, commit = 0;
41f89100
JP		 2255 bool ignore_updelay;
2256
2257 ignore_updelay = !bond->curr_active_slave ? true : false;
1da177e4
LT		 2258
2259 bond_for_each_slave(bond, slave, i) {
f0c76d61 2260 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 2261
f0c76d61 2262 link_state = bond_check_dev_link(bond, slave->dev, 0);
1da177e4
LT		 2263
2264 switch (slave->link) {
f0c76d61
JV		 2265 case BOND_LINK_UP:
2266 if (link_state)
2267 continue;
1da177e4 2268
f0c76d61
JV		 2269 slave->link = BOND_LINK_FAIL;
2270 slave->delay = bond->params.downdelay;
2271 if (slave->delay) {
a4aee5c8
JP		 2272 pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n",
2273 bond->dev->name,
2274 (bond->params.mode ==
2275 BOND_MODE_ACTIVEBACKUP) ?
2276 ((slave->state == BOND_STATE_ACTIVE) ?
2277 "active " : "backup ") : "",
2278 slave->dev->name,
2279 bond->params.downdelay * bond->params.miimon);
1da177e4 2280 }
f0c76d61
JV		 2281 /*FALLTHRU*/
2282 case BOND_LINK_FAIL:
2283 if (link_state) {
2284 /*
2285 * recovered before downdelay expired
2286 */
2287 slave->link = BOND_LINK_UP;
1da177e4 2288 slave->jiffies = jiffies;
a4aee5c8
JP		 2289 pr_info("%s: link status up again after %d ms for interface %s.\n",
2290 bond->dev->name,
2291 (bond->params.downdelay - slave->delay) *
2292 bond->params.miimon,
2293 slave->dev->name);
f0c76d61 2294 continue;
1da177e4 2295 }
f0c76d61
JV		 2296
2297 if (slave->delay <= 0) {
2298 slave->new_link = BOND_LINK_DOWN;
2299 commit++;
2300 continue;
1da177e4 2301 }
1da177e4 2302
f0c76d61
JV		 2303 slave->delay--;
2304 break;
2305
2306 case BOND_LINK_DOWN:
2307 if (!link_state)
2308 continue;
2309
2310 slave->link = BOND_LINK_BACK;
2311 slave->delay = bond->params.updelay;
2312
2313 if (slave->delay) {
a4aee5c8
JP		 2314 pr_info("%s: link status up for interface %s, enabling it in %d ms.\n",
2315 bond->dev->name, slave->dev->name,
2316 ignore_updelay ? 0 :
2317 bond->params.updelay *
2318 bond->params.miimon);
f0c76d61
JV		 2319 }
2320 /*FALLTHRU*/
2321 case BOND_LINK_BACK:
2322 if (!link_state) {
2323 slave->link = BOND_LINK_DOWN;
a4aee5c8
JP		 2324 pr_info("%s: link status down again after %d ms for interface %s.\n",
2325 bond->dev->name,
2326 (bond->params.updelay - slave->delay) *
2327 bond->params.miimon,
2328 slave->dev->name);
f0c76d61
JV		 2329
2330 continue;
2331 }
2332
41f89100
JP		 2333 if (ignore_updelay)
2334 slave->delay = 0;
2335
f0c76d61
JV		 2336 if (slave->delay <= 0) {
2337 slave->new_link = BOND_LINK_UP;
2338 commit++;
41f89100 2339 ignore_updelay = false;
f0c76d61 2340 continue;
1da177e4 2341 }
f0c76d61
JV		 2342
2343 slave->delay--;
1da177e4 2344 break;
f0c76d61
JV		 2345 }
2346 }
1da177e4 2347
f0c76d61
JV		 2348 return commit;
2349}
1da177e4 2350
f0c76d61
JV		 2351static void bond_miimon_commit(struct bonding *bond)
2352{
2353 struct slave *slave;
2354 int i;
2355
2356 bond_for_each_slave(bond, slave, i) {
2357 switch (slave->new_link) {
2358 case BOND_LINK_NOCHANGE:
2359 continue;
1da177e4 2360
f0c76d61
JV		 2361 case BOND_LINK_UP:
2362 slave->link = BOND_LINK_UP;
2363 slave->jiffies = jiffies;
2364
2365 if (bond->params.mode == BOND_MODE_8023AD) {
2366 /* prevent it from being the active one */
2367 slave->state = BOND_STATE_BACKUP;
2368 } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
2369 /* make it immediately active */
2370 slave->state = BOND_STATE_ACTIVE;
2371 } else if (slave != bond->primary_slave) {
2372 /* prevent it from being the active one */
2373 slave->state = BOND_STATE_BACKUP;
1da177e4 2374 }
1da177e4 2375
a4aee5c8
JP		 2376 pr_info("%s: link status definitely up for interface %s.\n",
2377 bond->dev->name, slave->dev->name);
1da177e4 2378
f0c76d61
JV		 2379 /* notify ad that the link status has changed */
2380 if (bond->params.mode == BOND_MODE_8023AD)
2381 bond_3ad_handle_link_change(slave, BOND_LINK_UP);
059fe7a5 2382
58402054 2383 if (bond_is_lb(bond))
f0c76d61
JV		 2384 bond_alb_handle_link_change(bond, slave,
2385 BOND_LINK_UP);
1da177e4 2386
f0c76d61
JV		 2387 if (!bond->curr_active_slave ||
2388 (slave == bond->primary_slave))
2389 goto do_failover;
1da177e4 2390
f0c76d61 2391 continue;
059fe7a5 2392
f0c76d61 2393 case BOND_LINK_DOWN:
fba4acda
JV		 2394 if (slave->link_failure_count < UINT_MAX)
2395 slave->link_failure_count++;
2396
f0c76d61 2397 slave->link = BOND_LINK_DOWN;
1da177e4 2398
f0c76d61
JV		 2399 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP ||
2400 bond->params.mode == BOND_MODE_8023AD)
2401 bond_set_slave_inactive_flags(slave);
2402
a4aee5c8
JP		 2403 pr_info("%s: link status definitely down for interface %s, disabling it\n",
2404 bond->dev->name, slave->dev->name);
f0c76d61
JV		 2405
2406 if (bond->params.mode == BOND_MODE_8023AD)
2407 bond_3ad_handle_link_change(slave,
2408 BOND_LINK_DOWN);
2409
ae63e808 2410 if (bond_is_lb(bond))
f0c76d61
JV		 2411 bond_alb_handle_link_change(bond, slave,
2412 BOND_LINK_DOWN);
2413
2414 if (slave == bond->curr_active_slave)
2415 goto do_failover;
2416
2417 continue;
2418
2419 default:
a4aee5c8 2420 pr_err("%s: invalid new link %d on slave %s\n",
f0c76d61
JV		 2421 bond->dev->name, slave->new_link,
2422 slave->dev->name);
2423 slave->new_link = BOND_LINK_NOCHANGE;
2424
2425 continue;
2426 }
2427
2428do_failover:
2429 ASSERT_RTNL();
2430 write_lock_bh(&bond->curr_slave_lock);
2431 bond_select_active_slave(bond);
2432 write_unlock_bh(&bond->curr_slave_lock);
2433 }
2434
2435 bond_set_carrier(bond);
1da177e4
LT		 2436}
2437
0b0eef66
JV		 2438/*
2439 * bond_mii_monitor
2440 *
2441 * Really a wrapper that splits the mii monitor into two phases: an
f0c76d61
JV		 2442 * inspection, then (if inspection indicates something needs to be done)
2443 * an acquisition of appropriate locks followed by a commit phase to
2444 * implement whatever link state changes are indicated.
0b0eef66
JV		 2445 */
2446void bond_mii_monitor(struct work_struct *work)
2447{
2448 struct bonding *bond = container_of(work, struct bonding,
2449 mii_work.work);
0b0eef66
JV		 2450
2451 read_lock(&bond->lock);
f0c76d61
JV		 2452 if (bond->kill_timers)
2453 goto out;
2454
2455 if (bond->slave_cnt == 0)
2456 goto re_arm;
b59f9f74
JV		 2457
2458 if (bond->send_grat_arp) {
2459 read_lock(&bond->curr_slave_lock);
2460 bond_send_gratuitous_arp(bond);
2461 read_unlock(&bond->curr_slave_lock);
2462 }
2463
305d552a
BH		 2464 if (bond->send_unsol_na) {
2465 read_lock(&bond->curr_slave_lock);
2466 bond_send_unsolicited_na(bond);
2467 read_unlock(&bond->curr_slave_lock);
2468 }
2469
f0c76d61 2470 if (bond_miimon_inspect(bond)) {
0b0eef66
JV		 2471 read_unlock(&bond->lock);
2472 rtnl_lock();
2473 read_lock(&bond->lock);
f0c76d61
JV		 2474
2475 bond_miimon_commit(bond);
2476
5655662d
JV		 2477 read_unlock(&bond->lock);
2478 rtnl_unlock(); /* might sleep, hold no other locks */
2479 read_lock(&bond->lock);
0b0eef66
JV		 2480 }
2481
f0c76d61
JV		 2482re_arm:
2483 if (bond->params.miimon)
2484 queue_delayed_work(bond->wq, &bond->mii_work,
2485 msecs_to_jiffies(bond->params.miimon));
2486out:
0b0eef66 2487 read_unlock(&bond->lock);
0b0eef66 2488}
c3ade5ca 2489
d3bb52b0 2490static __be32 bond_glean_dev_ip(struct net_device *dev)
c3ade5ca
JV		 2491{
2492 struct in_device *idev;
2493 struct in_ifaddr *ifa;
a144ea4b 2494 __be32 addr = 0;
c3ade5ca
JV		 2495
2496 if (!dev)
2497 return 0;
2498
2499 rcu_read_lock();
e5ed6399 2500 idev = __in_dev_get_rcu(dev);
c3ade5ca
JV		 2501 if (!idev)
2502 goto out;
2503
2504 ifa = idev->ifa_list;
2505 if (!ifa)
2506 goto out;
2507
2508 addr = ifa->ifa_local;
2509out:
2510 rcu_read_unlock();
2511 return addr;
2512}
2513
d3bb52b0 2514static int bond_has_this_ip(struct bonding *bond, __be32 ip)
f5b2b966 2515{
0883beca 2516 struct vlan_entry *vlan;
f5b2b966
JV		 2517
2518 if (ip == bond->master_ip)
2519 return 1;
2520
0883beca 2521 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
f5b2b966
JV		 2522 if (ip == vlan->vlan_ip)
2523 return 1;
2524 }
2525
2526 return 0;
2527}
2528
c3ade5ca
JV		 2529/*
2530 * We go to the (large) trouble of VLAN tagging ARP frames because
2531 * switches in VLAN mode (especially if ports are configured as
2532 * "native" to a VLAN) might not pass non-tagged frames.
2533 */
d3bb52b0 2534static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id)
c3ade5ca
JV		 2535{
2536 struct sk_buff *skb;
2537
5a03cdb7 2538 pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op,
a4aee5c8 2539 slave_dev->name, dest_ip, src_ip, vlan_id);
3d632c3f 2540
c3ade5ca
JV		 2541 skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip,
2542 NULL, slave_dev->dev_addr, NULL);
2543
2544 if (!skb) {
a4aee5c8 2545 pr_err("ARP packet allocation failed\n");
c3ade5ca
JV		 2546 return;
2547 }
2548 if (vlan_id) {
2549 skb = vlan_put_tag(skb, vlan_id);
2550 if (!skb) {
a4aee5c8 2551 pr_err("failed to insert VLAN tag\n");
c3ade5ca
JV		 2552 return;
2553 }
2554 }
2555 arp_xmit(skb);
2556}
2557
2558
1da177e4
LT		 2559static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
2560{
c3ade5ca 2561 int i, vlan_id, rv;
d3bb52b0 2562 __be32 *targets = bond->params.arp_targets;
0883beca 2563 struct vlan_entry *vlan;
c3ade5ca
JV		 2564 struct net_device *vlan_dev;
2565 struct flowi fl;
2566 struct rtable *rt;
1da177e4 2567
6b780567
MW		 2568 for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) {
2569 if (!targets[i])
5a31bec0 2570 break;
5a03cdb7 2571 pr_debug("basa: target %x\n", targets[i]);
c3ade5ca 2572 if (list_empty(&bond->vlan_list)) {
5a03cdb7 2573 pr_debug("basa: empty vlan: arp_send\n");
c3ade5ca
JV		 2574 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2575 bond->master_ip, 0);
2576 continue;
2577 }
2578
2579 /*
2580 * If VLANs are configured, we do a route lookup to
2581 * determine which VLAN interface would be used, so we
2582 * can tag the ARP with the proper VLAN tag.
2583 */
2584 memset(&fl, 0, sizeof(fl));
2585 fl.fl4_dst = targets[i];
2586 fl.fl4_tos = RTO_ONLINK;
2587
ec87fd3b 2588 rv = ip_route_output_key(dev_net(bond->dev), &rt, &fl);
c3ade5ca
JV		 2589 if (rv) {
2590 if (net_ratelimit()) {
a4aee5c8
JP		 2591 pr_warning("%s: no route to arp_ip_target %pI4\n",
2592 bond->dev->name, &fl.fl4_dst);
c3ade5ca
JV		 2593 }
2594 continue;
2595 }
2596
2597 /*
2598 * This target is not on a VLAN
2599 */
d8d1f30b 2600 if (rt->dst.dev == bond->dev) {
ed4b9f80 2601 ip_rt_put(rt);
5a03cdb7 2602 pr_debug("basa: rtdev == bond->dev: arp_send\n");
c3ade5ca
JV		 2603 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2604 bond->master_ip, 0);
2605 continue;
2606 }
2607
2608 vlan_id = 0;
0883beca 2609 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
5c15bdec 2610 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
d8d1f30b 2611 if (vlan_dev == rt->dst.dev) {
c3ade5ca 2612 vlan_id = vlan->vlan_id;
5a03cdb7 2613 pr_debug("basa: vlan match on %s %d\n",
c3ade5ca
JV		 2614 vlan_dev->name, vlan_id);
2615 break;
2616 }
2617 }
2618
2619 if (vlan_id) {
ed4b9f80 2620 ip_rt_put(rt);
c3ade5ca
JV		 2621 bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i],
2622 vlan->vlan_ip, vlan_id);
2623 continue;
2624 }
2625
2626 if (net_ratelimit()) {
a4aee5c8
JP		 2627 pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n",
2628 bond->dev->name, &fl.fl4_dst,
d8d1f30b 2629 rt->dst.dev ? rt->dst.dev->name : "NULL");
c3ade5ca 2630 }
ed4b9f80 2631 ip_rt_put(rt);
c3ade5ca
JV		 2632 }
2633}
2634
2635/*
2636 * Kick out a gratuitous ARP for an IP on the bonding master plus one
2637 * for each VLAN above us.
b59f9f74
JV		 2638 *
2639 * Caller must hold curr_slave_lock for read or better
c3ade5ca
JV		 2640 */
2641static void bond_send_gratuitous_arp(struct bonding *bond)
2642{
2643 struct slave *slave = bond->curr_active_slave;
2644 struct vlan_entry *vlan;
2645 struct net_device *vlan_dev;
2646
a4aee5c8
JP		 2647 pr_debug("bond_send_grat_arp: bond %s slave %s\n",
2648 bond->dev->name, slave ? slave->dev->name : "NULL");
b59f9f74
JV		 2649
2650 if (!slave || !bond->send_grat_arp ||
2651 test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state))
c3ade5ca
JV		 2652 return;
2653
b59f9f74
JV		 2654 bond->send_grat_arp--;
2655
c3ade5ca
JV		 2656 if (bond->master_ip) {
2657 bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip,
1053f62c 2658 bond->master_ip, 0);
c3ade5ca
JV		 2659 }
2660
2661 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
5c15bdec 2662 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
c3ade5ca
JV		 2663 if (vlan->vlan_ip) {
2664 bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip,
2665 vlan->vlan_ip, vlan->vlan_id);
2666 }
1da177e4
LT		 2667 }
2668}
2669
d3bb52b0 2670static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip)
f5b2b966
JV		 2671{
2672 int i;
d3bb52b0 2673 __be32 *targets = bond->params.arp_targets;
f5b2b966 2674
f5b2b966 2675 for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) {
5a03cdb7 2676 pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n",
a4aee5c8
JP		 2677 &sip, &tip, i, &targets[i],
2678 bond_has_this_ip(bond, tip));
f5b2b966
JV		 2679 if (sip == targets[i]) {
2680 if (bond_has_this_ip(bond, tip))
2681 slave->last_arp_rx = jiffies;
2682 return;
2683 }
2684 }
2685}
2686
2687static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
2688{
2689 struct arphdr *arp;
2690 struct slave *slave;
2691 struct bonding *bond;
2692 unsigned char *arp_ptr;
d3bb52b0 2693 __be32 sip, tip;
f5b2b966 2694
1f3c8804
AG		 2695 if (dev->priv_flags & IFF_802_1Q_VLAN) {
2696 /*
2697 * When using VLANS and bonding, dev and oriv_dev may be
2698 * incorrect if the physical interface supports VLAN
2699 * acceleration. With this change ARP validation now
2700 * works for hosts only reachable on the VLAN interface.
2701 */
2702 dev = vlan_dev_real_dev(dev);
2703 orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif);
2704 }
2705
f5b2b966
JV		 2706 if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER))
2707 goto out;
2708
454d7c9b 2709 bond = netdev_priv(dev);
f5b2b966
JV		 2710 read_lock(&bond->lock);
2711
5a03cdb7 2712 pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n",
a4aee5c8
JP		 2713 bond->dev->name, skb->dev ? skb->dev->name : "NULL",
2714 orig_dev ? orig_dev->name : "NULL");
f5b2b966
JV		 2715
2716 slave = bond_get_slave_by_dev(bond, orig_dev);
2717 if (!slave || !slave_do_arp_validate(bond, slave))
2718 goto out_unlock;
2719
988b7050 2720 if (!pskb_may_pull(skb, arp_hdr_len(dev)))
f5b2b966
JV		 2721 goto out_unlock;
2722
d0a92be0 2723 arp = arp_hdr(skb);
f5b2b966
JV		 2724 if (arp->ar_hln != dev->addr_len ||
2725 skb->pkt_type == PACKET_OTHERHOST ||
2726 skb->pkt_type == PACKET_LOOPBACK ||
2727 arp->ar_hrd != htons(ARPHRD_ETHER) ||
2728 arp->ar_pro != htons(ETH_P_IP) ||
2729 arp->ar_pln != 4)
2730 goto out_unlock;
2731
2732 arp_ptr = (unsigned char *)(arp + 1);
2733 arp_ptr += dev->addr_len;
2734 memcpy(&sip, arp_ptr, 4);
2735 arp_ptr += 4 + dev->addr_len;
2736 memcpy(&tip, arp_ptr, 4);
2737
5a03cdb7 2738 pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n",
a4aee5c8
JP		 2739 bond->dev->name, slave->dev->name, slave->state,
2740 bond->params.arp_validate, slave_do_arp_validate(bond, slave),
2741 &sip, &tip);
f5b2b966
JV		 2742
2743 /*
2744 * Backup slaves won't see the ARP reply, but do come through
2745 * here for each ARP probe (so we swap the sip/tip to validate
2746 * the probe). In a "redundant switch, common router" type of
2747 * configuration, the ARP probe will (hopefully) travel from
2748 * the active, through one switch, the router, then the other
2749 * switch before reaching the backup.
2750 */
2751 if (slave->state == BOND_STATE_ACTIVE)
2752 bond_validate_arp(bond, slave, sip, tip);
2753 else
2754 bond_validate_arp(bond, slave, tip, sip);
2755
2756out_unlock:
2757 read_unlock(&bond->lock);
2758out:
2759 dev_kfree_skb(skb);
2760 return NET_RX_SUCCESS;
2761}
2762
1da177e4
LT		 2763/*
2764 * this function is called regularly to monitor each slave's link
2765 * ensuring that traffic is being sent and received when arp monitoring
2766 * is used in load-balancing mode. if the adapter has been dormant, then an
2767 * arp is transmitted to generate traffic. see activebackup_arp_monitor for
2768 * arp monitoring in active backup mode.
2769 */
1b76b316 2770void bond_loadbalance_arp_mon(struct work_struct *work)
1da177e4 2771{
1b76b316
JV		 2772 struct bonding *bond = container_of(work, struct bonding,
2773 arp_work.work);
1da177e4
LT		 2774 struct slave *slave, *oldcurrent;
2775 int do_failover = 0;
2776 int delta_in_ticks;
2777 int i;
2778
2779 read_lock(&bond->lock);
2780
5ce0da8f 2781 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
1da177e4 2782
3d632c3f 2783 if (bond->kill_timers)
1da177e4 2784 goto out;
1da177e4 2785
3d632c3f 2786 if (bond->slave_cnt == 0)
1da177e4 2787 goto re_arm;
1da177e4
LT		 2788
2789 read_lock(&bond->curr_slave_lock);
2790 oldcurrent = bond->curr_active_slave;
2791 read_unlock(&bond->curr_slave_lock);
2792
2793 /* see if any of the previous devices are up now (i.e. they have
2794 * xmt and rcv traffic). the curr_active_slave does not come into
2795 * the picture unless it is null. also, slave->jiffies is not needed
2796 * here because we send an arp on each slave and give a slave as
2797 * long as it needs to get the tx/rx within the delta.
2798 * TODO: what about up/down delay in arp mode? it wasn't here before
2799 * so it can wait
2800 */
2801 bond_for_each_slave(bond, slave, i) {
2802 if (slave->link != BOND_LINK_UP) {
9d21493b 2803 if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) &&
b63bb739 2804 time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) {
1da177e4
LT		 2805
2806 slave->link = BOND_LINK_UP;
2807 slave->state = BOND_STATE_ACTIVE;
2808
2809 /* primary_slave has no meaning in round-robin
2810 * mode. the window of a slave being up and
2811 * curr_active_slave being null after enslaving
2812 * is closed.
2813 */
2814 if (!oldcurrent) {
a4aee5c8
JP		 2815 pr_info("%s: link status definitely up for interface %s, ",
2816 bond->dev->name,
2817 slave->dev->name);
1da177e4
LT		 2818 do_failover = 1;
2819 } else {
a4aee5c8
JP		 2820 pr_info("%s: interface %s is now up\n",
2821 bond->dev->name,
2822 slave->dev->name);
1da177e4
LT		 2823 }
2824 }
2825 } else {
2826 /* slave->link == BOND_LINK_UP */
2827
2828 /* not all switches will respond to an arp request
2829 * when the source ip is 0, so don't take the link down
2830 * if we don't know our ip yet
2831 */
9d21493b 2832 if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) ||
4b8a9239 2833 (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) {
1da177e4
LT		 2834
2835 slave->link = BOND_LINK_DOWN;
2836 slave->state = BOND_STATE_BACKUP;
2837
3d632c3f 2838 if (slave->link_failure_count < UINT_MAX)
1da177e4 2839 slave->link_failure_count++;
1da177e4 2840
a4aee5c8
JP		 2841 pr_info("%s: interface %s is now down.\n",
2842 bond->dev->name,
2843 slave->dev->name);
1da177e4 2844
3d632c3f 2845 if (slave == oldcurrent)
1da177e4 2846 do_failover = 1;
1da177e4
LT		 2847 }
2848 }
2849
2850 /* note: if switch is in round-robin mode, all links
2851 * must tx arp to ensure all links rx an arp - otherwise
2852 * links may oscillate or not come up at all; if switch is
2853 * in something like xor mode, there is nothing we can
2854 * do - all replies will be rx'ed on same link causing slaves
2855 * to be unstable during low/no traffic periods
2856 */
3d632c3f 2857 if (IS_UP(slave->dev))
1da177e4 2858 bond_arp_send_all(bond, slave);
1da177e4
LT		 2859 }
2860
2861 if (do_failover) {
059fe7a5 2862 write_lock_bh(&bond->curr_slave_lock);
1da177e4
LT		 2863
2864 bond_select_active_slave(bond);
2865
059fe7a5 2866 write_unlock_bh(&bond->curr_slave_lock);
1da177e4
LT		 2867 }
2868
2869re_arm:
1b76b316
JV		 2870 if (bond->params.arp_interval)
2871 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
1da177e4
LT		 2872out:
2873 read_unlock(&bond->lock);
2874}
2875
2876/*
b2220cad
JV		 2877 * Called to inspect slaves for active-backup mode ARP monitor link state
2878 * changes. Sets new_link in slaves to specify what action should take
2879 * place for the slave. Returns 0 if no changes are found, >0 if changes
2880 * to link states must be committed.
2881 *
2882 * Called with bond->lock held for read.
1da177e4 2883 */
b2220cad 2884static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks)
1da177e4 2885{
1da177e4 2886 struct slave *slave;
b2220cad 2887 int i, commit = 0;
1da177e4 2888
b2220cad
JV		 2889 bond_for_each_slave(bond, slave, i) {
2890 slave->new_link = BOND_LINK_NOCHANGE;
1da177e4 2891
b2220cad
JV		 2892 if (slave->link != BOND_LINK_UP) {
2893 if (time_before_eq(jiffies, slave_last_rx(bond, slave) +
2894 delta_in_ticks)) {
2895 slave->new_link = BOND_LINK_UP;
2896 commit++;
2897 }
1da177e4 2898
b2220cad
JV		 2899 continue;
2900 }
1da177e4 2901
b2220cad
JV		 2902 /*
2903 * Give slaves 2*delta after being enslaved or made
2904 * active. This avoids bouncing, as the last receive
2905 * times need a full ARP monitor cycle to be updated.
2906 */
2907 if (!time_after_eq(jiffies, slave->jiffies +
2908 2 * delta_in_ticks))
2909 continue;
2910
2911 /*
2912 * Backup slave is down if:
2913 * - No current_arp_slave AND
2914 * - more than 3*delta since last receive AND
2915 * - the bond has an IP address
2916 *
2917 * Note: a non-null current_arp_slave indicates
2918 * the curr_active_slave went down and we are
2919 * searching for a new one; under this condition
2920 * we only take the curr_active_slave down - this
2921 * gives each slave a chance to tx/rx traffic
2922 * before being taken out
2923 */
2924 if (slave->state == BOND_STATE_BACKUP &&
2925 !bond->current_arp_slave &&
2926 time_after(jiffies, slave_last_rx(bond, slave) +
2927 3 * delta_in_ticks)) {
2928 slave->new_link = BOND_LINK_DOWN;
2929 commit++;
2930 }
2931
2932 /*
2933 * Active slave is down if:
2934 * - more than 2*delta since transmitting OR
2935 * - (more than 2*delta since receive AND
2936 * the bond has an IP address)
2937 */
2938 if ((slave->state == BOND_STATE_ACTIVE) &&
9d21493b 2939 (time_after_eq(jiffies, dev_trans_start(slave->dev) +
b2220cad
JV		 2940 2 * delta_in_ticks) ||
2941 (time_after_eq(jiffies, slave_last_rx(bond, slave)
2942 + 2 * delta_in_ticks)))) {
2943 slave->new_link = BOND_LINK_DOWN;
2944 commit++;
2945 }
1da177e4
LT		 2946 }
2947
b2220cad
JV		 2948 return commit;
2949}
1da177e4 2950
b2220cad
JV		 2951/*
2952 * Called to commit link state changes noted by inspection step of
2953 * active-backup mode ARP monitor.
2954 *
2955 * Called with RTNL and bond->lock for read.
2956 */
2957static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks)
2958{
2959 struct slave *slave;
2960 int i;
1da177e4 2961
b2220cad
JV		 2962 bond_for_each_slave(bond, slave, i) {
2963 switch (slave->new_link) {
2964 case BOND_LINK_NOCHANGE:
2965 continue;
ff59c456 2966
b2220cad 2967 case BOND_LINK_UP:
b9f60253
JP		 2968 if ((!bond->curr_active_slave &&
2969 time_before_eq(jiffies,
2970 dev_trans_start(slave->dev) +
2971 delta_in_ticks)) ||
2972 bond->curr_active_slave != slave) {
b2220cad 2973 slave->link = BOND_LINK_UP;
b2220cad
JV		 2974 bond->current_arp_slave = NULL;
2975
a4aee5c8 2976 pr_info("%s: link status definitely up for interface %s.\n",
b9f60253 2977 bond->dev->name, slave->dev->name);
b2220cad 2978
b9f60253
JP		 2979 if (!bond->curr_active_slave ||
2980 (slave == bond->primary_slave))
2981 goto do_failover;
1da177e4 2982
b9f60253 2983 }
1da177e4 2984
b9f60253 2985 continue;
1da177e4 2986
b2220cad
JV		 2987 case BOND_LINK_DOWN:
2988 if (slave->link_failure_count < UINT_MAX)
2989 slave->link_failure_count++;
2990
2991 slave->link = BOND_LINK_DOWN;
b9f60253 2992 bond_set_slave_inactive_flags(slave);
b2220cad 2993
a4aee5c8 2994 pr_info("%s: link status definitely down for interface %s, disabling it\n",
b9f60253 2995 bond->dev->name, slave->dev->name);
b2220cad 2996
b9f60253 2997 if (slave == bond->curr_active_slave) {
b2220cad 2998 bond->current_arp_slave = NULL;
b9f60253 2999 goto do_failover;
1da177e4 3000 }
b9f60253
JP		 3001
3002 continue;
b2220cad
JV		 3003
3004 default:
a4aee5c8 3005 pr_err("%s: impossible: new_link %d on slave %s\n",
b2220cad
JV		 3006 bond->dev->name, slave->new_link,
3007 slave->dev->name);
b9f60253 3008 continue;
1da177e4 3009 }
1da177e4 3010
b9f60253
JP		 3011do_failover:
3012 ASSERT_RTNL();
b2220cad 3013 write_lock_bh(&bond->curr_slave_lock);
b9f60253 3014 bond_select_active_slave(bond);
b2220cad
JV		 3015 write_unlock_bh(&bond->curr_slave_lock);
3016 }
1da177e4 3017
b2220cad
JV		 3018 bond_set_carrier(bond);
3019}
1da177e4 3020
b2220cad
JV		 3021/*
3022 * Send ARP probes for active-backup mode ARP monitor.
3023 *
3024 * Called with bond->lock held for read.
3025 */
3026static void bond_ab_arp_probe(struct bonding *bond)
3027{
3028 struct slave *slave;
3029 int i;
1da177e4 3030
b2220cad 3031 read_lock(&bond->curr_slave_lock);
1da177e4 3032
b2220cad 3033 if (bond->current_arp_slave && bond->curr_active_slave)
a4aee5c8
JP		 3034 pr_info("PROBE: c_arp %s && cas %s BAD\n",
3035 bond->current_arp_slave->dev->name,
3036 bond->curr_active_slave->dev->name);
1da177e4 3037
b2220cad
JV		 3038 if (bond->curr_active_slave) {
3039 bond_arp_send_all(bond, bond->curr_active_slave);
3040 read_unlock(&bond->curr_slave_lock);
3041 return;
3042 }
1da177e4 3043
b2220cad 3044 read_unlock(&bond->curr_slave_lock);
059fe7a5 3045
b2220cad
JV		 3046 /* if we don't have a curr_active_slave, search for the next available
3047 * backup slave from the current_arp_slave and make it the candidate
3048 * for becoming the curr_active_slave
3049 */
1da177e4 3050
b2220cad
JV		 3051 if (!bond->current_arp_slave) {
3052 bond->current_arp_slave = bond->first_slave;
3053 if (!bond->current_arp_slave)
3054 return;
3055 }
1da177e4 3056
b2220cad 3057 bond_set_slave_inactive_flags(bond->current_arp_slave);
059fe7a5 3058
b2220cad
JV		 3059 /* search for next candidate */
3060 bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) {
3061 if (IS_UP(slave->dev)) {
3062 slave->link = BOND_LINK_BACK;
3063 bond_set_slave_active_flags(slave);
3064 bond_arp_send_all(bond, slave);
1da177e4 3065 slave->jiffies = jiffies;
b2220cad
JV		 3066 bond->current_arp_slave = slave;
3067 break;
1da177e4
LT		 3068 }
3069
b2220cad
JV		 3070 /* if the link state is up at this point, we
3071 * mark it down - this can happen if we have
3072 * simultaneous link failures and
3073 * reselect_active_interface doesn't make this
3074 * one the current slave so it is still marked
3075 * up when it is actually down
1da177e4 3076 */
b2220cad
JV		 3077 if (slave->link == BOND_LINK_UP) {
3078 slave->link = BOND_LINK_DOWN;
3079 if (slave->link_failure_count < UINT_MAX)
3080 slave->link_failure_count++;
1da177e4 3081
b2220cad
JV		 3082 bond_set_slave_inactive_flags(slave);
3083
a4aee5c8
JP		 3084 pr_info("%s: backup interface %s is now down.\n",
3085 bond->dev->name, slave->dev->name);
1da177e4 3086 }
b2220cad
JV		 3087 }
3088}
1da177e4 3089
b2220cad
JV		 3090void bond_activebackup_arp_mon(struct work_struct *work)
3091{
3092 struct bonding *bond = container_of(work, struct bonding,
3093 arp_work.work);
3094 int delta_in_ticks;
1da177e4 3095
b2220cad 3096 read_lock(&bond->lock);
1da177e4 3097
b2220cad
JV		 3098 if (bond->kill_timers)
3099 goto out;
1da177e4 3100
b2220cad 3101 delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval);
1da177e4 3102
b2220cad
JV		 3103 if (bond->slave_cnt == 0)
3104 goto re_arm;
3105
b59f9f74
JV		 3106 if (bond->send_grat_arp) {
3107 read_lock(&bond->curr_slave_lock);
3108 bond_send_gratuitous_arp(bond);
3109 read_unlock(&bond->curr_slave_lock);
3110 }
3111
305d552a
BH		 3112 if (bond->send_unsol_na) {
3113 read_lock(&bond->curr_slave_lock);
3114 bond_send_unsolicited_na(bond);
3115 read_unlock(&bond->curr_slave_lock);
3116 }
3117
b2220cad
JV		 3118 if (bond_ab_arp_inspect(bond, delta_in_ticks)) {
3119 read_unlock(&bond->lock);
3120 rtnl_lock();
3121 read_lock(&bond->lock);
3122
3123 bond_ab_arp_commit(bond, delta_in_ticks);
3124
3125 read_unlock(&bond->lock);
3126 rtnl_unlock();
3127 read_lock(&bond->lock);
1da177e4
LT		 3128 }
3129
b2220cad
JV		 3130 bond_ab_arp_probe(bond);
3131
1da177e4 3132re_arm:
3d632c3f 3133 if (bond->params.arp_interval)
1b76b316 3134 queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks);
1da177e4
LT		 3135out:
3136 read_unlock(&bond->lock);
3137}
3138
3139/*------------------------------ proc/seq_file-------------------------------*/
3140
3141#ifdef CONFIG_PROC_FS
3142
1da177e4 3143static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos)
1f78d9f9
HE		 3144 __acquires(&dev_base_lock)
3145 __acquires(&bond->lock)
1da177e4
LT		 3146{
3147 struct bonding *bond = seq->private;
3148 loff_t off = 0;
3149 struct slave *slave;
3150 int i;
3151
3152 /* make sure the bond won't be taken away */
3153 read_lock(&dev_base_lock);
6603a6f2 3154 read_lock(&bond->lock);
1da177e4 3155
3d632c3f 3156 if (*pos == 0)
1da177e4 3157 return SEQ_START_TOKEN;
1da177e4
LT		 3158
3159 bond_for_each_slave(bond, slave, i) {
3d632c3f 3160 if (++off == *pos)
1da177e4 3161 return slave;
1da177e4
LT		 3162 }
3163
3164 return NULL;
3165}
3166
3167static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3168{
3169 struct bonding *bond = seq->private;
3170 struct slave *slave = v;
3171
3172 ++*pos;
3d632c3f 3173 if (v == SEQ_START_TOKEN)
1da177e4 3174 return bond->first_slave;
1da177e4
LT		 3175
3176 slave = slave->next;
3177
3178 return (slave == bond->first_slave) ? NULL : slave;
3179}
3180
3181static void bond_info_seq_stop(struct seq_file *seq, void *v)
1f78d9f9
HE		 3182 __releases(&bond->lock)
3183 __releases(&dev_base_lock)
1da177e4
LT		 3184{
3185 struct bonding *bond = seq->private;
3186
6603a6f2 3187 read_unlock(&bond->lock);
1da177e4
LT		 3188 read_unlock(&dev_base_lock);
3189}
3190
3191static void bond_info_show_master(struct seq_file *seq)
3192{
3193 struct bonding *bond = seq->private;
3194 struct slave *curr;
4756b02f 3195 int i;
1da177e4
LT		 3196
3197 read_lock(&bond->curr_slave_lock);
3198 curr = bond->curr_active_slave;
3199 read_unlock(&bond->curr_slave_lock);
3200
dd957c57 3201 seq_printf(seq, "Bonding Mode: %s",
1da177e4
LT		 3202 bond_mode_name(bond->params.mode));
3203
dd957c57
JV		 3204 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP &&
3205 bond->params.fail_over_mac)
3915c1e8
JV		 3206 seq_printf(seq, " (fail_over_mac %s)",
3207 fail_over_mac_tbl[bond->params.fail_over_mac].modename);
dd957c57
JV		 3208
3209 seq_printf(seq, "\n");
3210
c61b75ad
MW		 3211 if (bond->params.mode == BOND_MODE_XOR ||
3212 bond->params.mode == BOND_MODE_8023AD) {
3213 seq_printf(seq, "Transmit Hash Policy: %s (%d)\n",
3214 xmit_hashtype_tbl[bond->params.xmit_policy].modename,
3215 bond->params.xmit_policy);
3216 }
3217
1da177e4 3218 if (USES_PRIMARY(bond->params.mode)) {
a549952a 3219 seq_printf(seq, "Primary Slave: %s",
0f418b2a
MW		 3220 (bond->primary_slave) ?
3221 bond->primary_slave->dev->name : "None");
a549952a
JP		 3222 if (bond->primary_slave)
3223 seq_printf(seq, " (primary_reselect %s)",
3224 pri_reselect_tbl[bond->params.primary_reselect].modename);
1da177e4 3225
a549952a 3226 seq_printf(seq, "\nCurrently Active Slave: %s\n",
1da177e4
LT		 3227 (curr) ? curr->dev->name : "None");
3228 }
3229
ff59c456
JV		 3230 seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ?
3231 "up" : "down");
1da177e4
LT		 3232 seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon);
3233 seq_printf(seq, "Up Delay (ms): %d\n",
3234 bond->params.updelay * bond->params.miimon);
3235 seq_printf(seq, "Down Delay (ms): %d\n",
3236 bond->params.downdelay * bond->params.miimon);
3237
4756b02f
MW		 3238
3239 /* ARP information */
3d632c3f
SH		 3240 if (bond->params.arp_interval > 0) {
3241 int printed = 0;
4756b02f
MW		 3242 seq_printf(seq, "ARP Polling Interval (ms): %d\n",
3243 bond->params.arp_interval);
3244
3245 seq_printf(seq, "ARP IP target/s (n.n.n.n form):");
3246
3d632c3f 3247 for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) {
4756b02f 3248 if (!bond->params.arp_targets[i])
5a31bec0 3249 break;
4756b02f
MW		 3250 if (printed)
3251 seq_printf(seq, ",");
8cf14e38 3252 seq_printf(seq, " %pI4", &bond->params.arp_targets[i]);
4756b02f
MW		 3253 printed = 1;
3254 }
3255 seq_printf(seq, "\n");
3256 }
3257
1da177e4
LT		 3258 if (bond->params.mode == BOND_MODE_8023AD) {
3259 struct ad_info ad_info;
3260
3261 seq_puts(seq, "\n802.3ad info\n");
3262 seq_printf(seq, "LACP rate: %s\n",
3263 (bond->params.lacp_fast) ? "fast" : "slow");
fd989c83
JV		 3264 seq_printf(seq, "Aggregator selection policy (ad_select): %s\n",
3265 ad_select_tbl[bond->params.ad_select].modename);
1da177e4
LT		 3266
3267 if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
3268 seq_printf(seq, "bond %s has no active aggregator\n",
3269 bond->dev->name);
3270 } else {
3271 seq_printf(seq, "Active Aggregator Info:\n");
3272
3273 seq_printf(seq, "\tAggregator ID: %d\n",
3274 ad_info.aggregator_id);
3275 seq_printf(seq, "\tNumber of ports: %d\n",
3276 ad_info.ports);
3277 seq_printf(seq, "\tActor Key: %d\n",
3278 ad_info.actor_key);
3279 seq_printf(seq, "\tPartner Key: %d\n",
3280 ad_info.partner_key);
e174961c
JB		 3281 seq_printf(seq, "\tPartner Mac Address: %pM\n",
3282 ad_info.partner_system);
1da177e4
LT		 3283 }
3284 }
3285}
3286
3d632c3f
SH		 3287static void bond_info_show_slave(struct seq_file *seq,
3288 const struct slave *slave)
1da177e4
LT		 3289{
3290 struct bonding *bond = seq->private;
3291
3292 seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name);
3293 seq_printf(seq, "MII Status: %s\n",
3294 (slave->link == BOND_LINK_UP) ? "up" : "down");
65509645 3295 seq_printf(seq, "Link Failure Count: %u\n",
1da177e4
LT		 3296 slave->link_failure_count);
3297
e174961c 3298 seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr);
1da177e4
LT		 3299
3300 if (bond->params.mode == BOND_MODE_8023AD) {
3301 const struct aggregator *agg
3302 = SLAVE_AD_INFO(slave).port.aggregator;
3303
3d632c3f 3304 if (agg)
1da177e4
LT		 3305 seq_printf(seq, "Aggregator ID: %d\n",
3306 agg->aggregator_identifier);
3d632c3f 3307 else
1da177e4 3308 seq_puts(seq, "Aggregator ID: N/A\n");
1da177e4 3309 }
bb1d9123 3310 seq_printf(seq, "Slave queue ID: %d\n", slave->queue_id);
1da177e4
LT		 3311}
3312
3313static int bond_info_seq_show(struct seq_file *seq, void *v)
3314{
3315 if (v == SEQ_START_TOKEN) {
3316 seq_printf(seq, "%s\n", version);
3317 bond_info_show_master(seq);
3d632c3f 3318 } else
1da177e4 3319 bond_info_show_slave(seq, v);
1da177e4
LT		 3320
3321 return 0;
3322}
3323
4101dec9 3324static const struct seq_operations bond_info_seq_ops = {
1da177e4
LT		 3325 .start = bond_info_seq_start,
3326 .next = bond_info_seq_next,
3327 .stop = bond_info_seq_stop,
3328 .show = bond_info_seq_show,
3329};
3330
3331static int bond_info_open(struct inode *inode, struct file *file)
3332{
3333 struct seq_file *seq;
3334 struct proc_dir_entry *proc;
3335 int res;
3336
3337 res = seq_open(file, &bond_info_seq_ops);
3338 if (!res) {
3339 /* recover the pointer buried in proc_dir_entry data */
3340 seq = file->private_data;
3341 proc = PDE(inode);
3342 seq->private = proc->data;
3343 }
3344
3345 return res;
3346}
3347
d54b1fdb 3348static const struct file_operations bond_info_fops = {
1da177e4
LT		 3349 .owner = THIS_MODULE,
3350 .open = bond_info_open,
3351 .read = seq_read,
3352 .llseek = seq_lseek,
3353 .release = seq_release,
3354};
3355
38fc0026 3356static void bond_create_proc_entry(struct bonding *bond)
1da177e4
LT		 3357{
3358 struct net_device *bond_dev = bond->dev;
ec87fd3b 3359 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
1da177e4 3360
ec87fd3b 3361 if (bn->proc_dir) {
a95609cb 3362 bond->proc_entry = proc_create_data(bond_dev->name,
ec87fd3b 3363 S_IRUGO, bn->proc_dir,
a95609cb 3364 &bond_info_fops, bond);
3d632c3f 3365 if (bond->proc_entry == NULL)
a4aee5c8
JP		 3366 pr_warning("Warning: Cannot create /proc/net/%s/%s\n",
3367 DRV_NAME, bond_dev->name);
3d632c3f 3368 else
1da177e4 3369 memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ);
1da177e4 3370 }
1da177e4
LT		 3371}
3372
3373static void bond_remove_proc_entry(struct bonding *bond)
3374{
ec87fd3b
EB		 3375 struct net_device *bond_dev = bond->dev;
3376 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
3377
3378 if (bn->proc_dir && bond->proc_entry) {
3379 remove_proc_entry(bond->proc_file_name, bn->proc_dir);
1da177e4
LT		 3380 memset(bond->proc_file_name, 0, IFNAMSIZ);
3381 bond->proc_entry = NULL;
3382 }
3383}
3384
3385/* Create the bonding directory under /proc/net, if doesn't exist yet.
3386 * Caller must hold rtnl_lock.
3387 */
2c8c1e72 3388static void __net_init bond_create_proc_dir(struct bond_net *bn)
1da177e4 3389{
ec87fd3b
EB		 3390 if (!bn->proc_dir) {
3391 bn->proc_dir = proc_mkdir(DRV_NAME, bn->net->proc_net);
3392 if (!bn->proc_dir)
a4aee5c8
JP		 3393 pr_warning("Warning: cannot create /proc/net/%s\n",
3394 DRV_NAME);
1da177e4
LT		 3395 }
3396}
3397
3398/* Destroy the bonding directory under /proc/net, if empty.
3399 * Caller must hold rtnl_lock.
3400 */
2c8c1e72 3401static void __net_exit bond_destroy_proc_dir(struct bond_net *bn)
1da177e4 3402{
ec87fd3b
EB		 3403 if (bn->proc_dir) {
3404 remove_proc_entry(DRV_NAME, bn->net->proc_net);
3405 bn->proc_dir = NULL;
1da177e4
LT		 3406 }
3407}
aee64faf
JP		 3408
3409#else /* !CONFIG_PROC_FS */
3410
38fc0026 3411static void bond_create_proc_entry(struct bonding *bond)
aee64faf
JP		 3412{
3413}
3414
3415static void bond_remove_proc_entry(struct bonding *bond)
3416{
3417}
3418
2c8c1e72 3419static inline void bond_create_proc_dir(struct bond_net *bn)
aee64faf
JP		 3420{
3421}
3422
2c8c1e72 3423static inline void bond_destroy_proc_dir(struct bond_net *bn)
aee64faf
JP		 3424{
3425}
3426
1da177e4
LT		 3427#endif /* CONFIG_PROC_FS */
3428
aee64faf 3429
1da177e4
LT		 3430/*-------------------------- netdev event handling --------------------------*/
3431
3432/*
3433 * Change device name
3434 */
3435static int bond_event_changename(struct bonding *bond)
3436{
1da177e4
LT		 3437 bond_remove_proc_entry(bond);
3438 bond_create_proc_entry(bond);
7e083840 3439
1da177e4
LT		 3440 return NOTIFY_DONE;
3441}
3442
3d632c3f
SH		 3443static int bond_master_netdev_event(unsigned long event,
3444 struct net_device *bond_dev)
1da177e4 3445{
454d7c9b 3446 struct bonding *event_bond = netdev_priv(bond_dev);
1da177e4
LT		 3447
3448 switch (event) {
3449 case NETDEV_CHANGENAME:
3450 return bond_event_changename(event_bond);
1da177e4
LT		 3451 default:
3452 break;
3453 }
3454
3455 return NOTIFY_DONE;
3456}
3457
3d632c3f
SH		 3458static int bond_slave_netdev_event(unsigned long event,
3459 struct net_device *slave_dev)
1da177e4
LT		 3460{
3461 struct net_device *bond_dev = slave_dev->master;
454d7c9b 3462 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3463
3464 switch (event) {
3465 case NETDEV_UNREGISTER:
3466 if (bond_dev) {
1284cd3a
JV		 3467 if (bond->setup_by_slave)
3468 bond_release_and_destroy(bond_dev, slave_dev);
3469 else
3470 bond_release(bond_dev, slave_dev);
1da177e4
LT		 3471 }
3472 break;
3473 case NETDEV_CHANGE:
17d04500
JV		 3474 if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) {
3475 struct slave *slave;
3476
3477 slave = bond_get_slave_by_dev(bond, slave_dev);
3478 if (slave) {
3479 u16 old_speed = slave->speed;
3480 u16 old_duplex = slave->duplex;
3481
3482 bond_update_speed_duplex(slave);
3483
3484 if (bond_is_lb(bond))
3485 break;
3486
3487 if (old_speed != slave->speed)
3488 bond_3ad_adapter_speed_changed(slave);
3489 if (old_duplex != slave->duplex)
3490 bond_3ad_adapter_duplex_changed(slave);
3491 }
3492 }
3493
1da177e4
LT		 3494 break;
3495 case NETDEV_DOWN:
3496 /*
3497 * ... Or is it this?
3498 */
3499 break;
3500 case NETDEV_CHANGEMTU:
3501 /*
3502 * TODO: Should slaves be allowed to
3503 * independently alter their MTU? For
3504 * an active-backup bond, slaves need
3505 * not be the same type of device, so
3506 * MTUs may vary. For other modes,
3507 * slaves arguably should have the
3508 * same MTUs. To do this, we'd need to
3509 * take over the slave's change_mtu
3510 * function for the duration of their
3511 * servitude.
3512 */
3513 break;
3514 case NETDEV_CHANGENAME:
3515 /*
3516 * TODO: handle changing the primary's name
3517 */
3518 break;
8531c5ff
AK		 3519 case NETDEV_FEAT_CHANGE:
3520 bond_compute_features(bond);
3521 break;
1da177e4
LT		 3522 default:
3523 break;
3524 }
3525
3526 return NOTIFY_DONE;
3527}
3528
3529/*
3530 * bond_netdev_event: handle netdev notifier chain events.
3531 *
3532 * This function receives events for the netdev chain. The caller (an
e041c683 3533 * ioctl handler calling blocking_notifier_call_chain) holds the necessary
1da177e4
LT		 3534 * locks for us to safely manipulate the slave devices (RTNL lock,
3535 * dev_probe_lock).
3536 */
3d632c3f
SH		 3537static int bond_netdev_event(struct notifier_block *this,
3538 unsigned long event, void *ptr)
1da177e4
LT		 3539{
3540 struct net_device *event_dev = (struct net_device *)ptr;
3541
5a03cdb7 3542 pr_debug("event_dev: %s, event: %lx\n",
a4aee5c8
JP		 3543 event_dev ? event_dev->name : "None",
3544 event);
1da177e4 3545
0b680e75
JV		 3546 if (!(event_dev->priv_flags & IFF_BONDING))
3547 return NOTIFY_DONE;
3548
1da177e4 3549 if (event_dev->flags & IFF_MASTER) {
5a03cdb7 3550 pr_debug("IFF_MASTER\n");
1da177e4
LT		 3551 return bond_master_netdev_event(event, event_dev);
3552 }
3553
3554 if (event_dev->flags & IFF_SLAVE) {
5a03cdb7 3555 pr_debug("IFF_SLAVE\n");
1da177e4
LT		 3556 return bond_slave_netdev_event(event, event_dev);
3557 }
3558
3559 return NOTIFY_DONE;
3560}
3561
c3ade5ca
JV		 3562/*
3563 * bond_inetaddr_event: handle inetaddr notifier chain events.
3564 *
3565 * We keep track of device IPs primarily to use as source addresses in
3566 * ARP monitor probes (rather than spewing out broadcasts all the time).
3567 *
3568 * We track one IP for the main device (if it has one), plus one per VLAN.
3569 */
3570static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr)
3571{
3572 struct in_ifaddr *ifa = ptr;
3573 struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev;
ec87fd3b 3574 struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id);
0883beca
PE		 3575 struct bonding *bond;
3576 struct vlan_entry *vlan;
c3ade5ca 3577
ec87fd3b 3578 list_for_each_entry(bond, &bn->dev_list, bond_list) {
c3ade5ca
JV		 3579 if (bond->dev == event_dev) {
3580 switch (event) {
3581 case NETDEV_UP:
3582 bond->master_ip = ifa->ifa_local;
3583 return NOTIFY_OK;
3584 case NETDEV_DOWN:
3585 bond->master_ip = bond_glean_dev_ip(bond->dev);
3586 return NOTIFY_OK;
3587 default:
3588 return NOTIFY_DONE;
3589 }
3590 }
3591
0883beca 3592 list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
5c15bdec 3593 vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id);
c3ade5ca
JV		 3594 if (vlan_dev == event_dev) {
3595 switch (event) {
3596 case NETDEV_UP:
3597 vlan->vlan_ip = ifa->ifa_local;
3598 return NOTIFY_OK;
3599 case NETDEV_DOWN:
3600 vlan->vlan_ip =
3601 bond_glean_dev_ip(vlan_dev);
3602 return NOTIFY_OK;
3603 default:
3604 return NOTIFY_DONE;
3605 }
3606 }
3607 }
3608 }
3609 return NOTIFY_DONE;
3610}
3611
1da177e4
LT		 3612static struct notifier_block bond_netdev_notifier = {
3613 .notifier_call = bond_netdev_event,
3614};
3615
c3ade5ca
JV		 3616static struct notifier_block bond_inetaddr_notifier = {
3617 .notifier_call = bond_inetaddr_event,
3618};
3619
1da177e4
LT		 3620/*-------------------------- Packet type handling ---------------------------*/
3621
3622/* register to receive lacpdus on a bond */
3623static void bond_register_lacpdu(struct bonding *bond)
3624{
3625 struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type);
3626
3627 /* initialize packet type */
3628 pk_type->type = PKT_TYPE_LACPDU;
3629 pk_type->dev = bond->dev;
3630 pk_type->func = bond_3ad_lacpdu_recv;
3631
3632 dev_add_pack(pk_type);
3633}
3634
3635/* unregister to receive lacpdus on a bond */
3636static void bond_unregister_lacpdu(struct bonding *bond)
3637{
3638 dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type));
3639}
3640
f5b2b966
JV		 3641void bond_register_arp(struct bonding *bond)
3642{
3643 struct packet_type *pt = &bond->arp_mon_pt;
3644
c4f283b1
JV		 3645 if (pt->type)
3646 return;
3647
f5b2b966 3648 pt->type = htons(ETH_P_ARP);
e245cb71 3649 pt->dev = bond->dev;
f5b2b966
JV		 3650 pt->func = bond_arp_rcv;
3651 dev_add_pack(pt);
3652}
3653
3654void bond_unregister_arp(struct bonding *bond)
3655{
c4f283b1
JV		 3656 struct packet_type *pt = &bond->arp_mon_pt;
3657
3658 dev_remove_pack(pt);
3659 pt->type = 0;
f5b2b966
JV		 3660}
3661
169a3e66
JV		 3662/*---------------------------- Hashing Policies -----------------------------*/
3663
6f6652be
JV		 3664/*
3665 * Hash for the output device based upon layer 2 and layer 3 data. If
3666 * the packet is not IP mimic bond_xmit_hash_policy_l2()
3667 */
a361c83c 3668static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count)
6f6652be
JV		 3669{
3670 struct ethhdr *data = (struct ethhdr *)skb->data;
3671 struct iphdr *iph = ip_hdr(skb);
3672
f14c4e4e 3673 if (skb->protocol == htons(ETH_P_IP)) {
6f6652be 3674 return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^
d3da6831 3675 (data->h_dest[5] ^ data->h_source[5])) % count;
6f6652be
JV		 3676 }
3677
d3da6831 3678 return (data->h_dest[5] ^ data->h_source[5]) % count;
6f6652be
JV		 3679}
3680
169a3e66 3681/*
59c51591 3682 * Hash for the output device based upon layer 3 and layer 4 data. If
169a3e66
JV		 3683 * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is
3684 * altogether not IP, mimic bond_xmit_hash_policy_l2()
3685 */
a361c83c 3686static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count)
169a3e66
JV		 3687{
3688 struct ethhdr *data = (struct ethhdr *)skb->data;
eddc9ec5 3689 struct iphdr *iph = ip_hdr(skb);
d3bb52b0 3690 __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl);
169a3e66
JV		 3691 int layer4_xor = 0;
3692
f14c4e4e
BH		 3693 if (skb->protocol == htons(ETH_P_IP)) {
3694 if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) &&
169a3e66
JV		 3695 (iph->protocol == IPPROTO_TCP ||
3696 iph->protocol == IPPROTO_UDP)) {
d3bb52b0 3697 layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1)));
169a3e66
JV		 3698 }
3699 return (layer4_xor ^
3700 ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count;
3701
3702 }
3703
d3da6831 3704 return (data->h_dest[5] ^ data->h_source[5]) % count;
169a3e66
JV		 3705}
3706
3707/*
3708 * Hash for the output device based upon layer 2 data
3709 */
a361c83c 3710static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count)
169a3e66
JV		 3711{
3712 struct ethhdr *data = (struct ethhdr *)skb->data;
3713
d3da6831 3714 return (data->h_dest[5] ^ data->h_source[5]) % count;
169a3e66
JV		 3715}
3716
1da177e4
LT		 3717/*-------------------------- Device entry points ----------------------------*/
3718
3719static int bond_open(struct net_device *bond_dev)
3720{
454d7c9b 3721 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3722
3723 bond->kill_timers = 0;
3724
58402054 3725 if (bond_is_lb(bond)) {
1da177e4
LT		 3726 /* bond_alb_initialize must be called before the timer
3727 * is started.
3728 */
3729 if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) {
3730 /* something went wrong - fail the open operation */
b473946a 3731 return -ENOMEM;
1da177e4
LT		 3732 }
3733
1b76b316
JV		 3734 INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor);
3735 queue_delayed_work(bond->wq, &bond->alb_work, 0);
1da177e4
LT		 3736 }
3737
3738 if (bond->params.miimon) { /* link check interval, in milliseconds. */
1b76b316
JV		 3739 INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor);
3740 queue_delayed_work(bond->wq, &bond->mii_work, 0);
1da177e4
LT		 3741 }
3742
3743 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316
JV		 3744 if (bond->params.mode == BOND_MODE_ACTIVEBACKUP)
3745 INIT_DELAYED_WORK(&bond->arp_work,
3746 bond_activebackup_arp_mon);
3747 else
3748 INIT_DELAYED_WORK(&bond->arp_work,
3749 bond_loadbalance_arp_mon);
3750
3751 queue_delayed_work(bond->wq, &bond->arp_work, 0);
f5b2b966
JV		 3752 if (bond->params.arp_validate)
3753 bond_register_arp(bond);
1da177e4
LT		 3754 }
3755
3756 if (bond->params.mode == BOND_MODE_8023AD) {
a40745f5 3757 INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler);
1b76b316 3758 queue_delayed_work(bond->wq, &bond->ad_work, 0);
1da177e4
LT		 3759 /* register to receive LACPDUs */
3760 bond_register_lacpdu(bond);
fd989c83 3761 bond_3ad_initiate_agg_selection(bond, 1);
1da177e4
LT		 3762 }
3763
3764 return 0;
3765}
3766
3767static int bond_close(struct net_device *bond_dev)
3768{
454d7c9b 3769 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 3770
3771 if (bond->params.mode == BOND_MODE_8023AD) {
3772 /* Unregister the receive of LACPDUs */
3773 bond_unregister_lacpdu(bond);
3774 }
3775
f5b2b966
JV		 3776 if (bond->params.arp_validate)
3777 bond_unregister_arp(bond);
3778
1da177e4
LT		 3779 write_lock_bh(&bond->lock);
3780
b59f9f74 3781 bond->send_grat_arp = 0;
305d552a 3782 bond->send_unsol_na = 0;
1da177e4
LT		 3783
3784 /* signal timers not to re-arm */
3785 bond->kill_timers = 1;
3786
3787 write_unlock_bh(&bond->lock);
3788
1da177e4 3789 if (bond->params.miimon) { /* link check interval, in milliseconds. */
1b76b316 3790 cancel_delayed_work(&bond->mii_work);
1da177e4
LT		 3791 }
3792
3793 if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
1b76b316 3794 cancel_delayed_work(&bond->arp_work);
1da177e4
LT		 3795 }
3796
3797 switch (bond->params.mode) {
3798 case BOND_MODE_8023AD:
1b76b316 3799 cancel_delayed_work(&bond->ad_work);
1da177e4
LT		 3800 break;
3801 case BOND_MODE_TLB:
3802 case BOND_MODE_ALB:
1b76b316 3803 cancel_delayed_work(&bond->alb_work);
1da177e4
LT		 3804 break;
3805 default:
3806 break;
3807 }
3808
1da177e4 3809
58402054 3810 if (bond_is_lb(bond)) {
1da177e4
LT		 3811 /* Must be called only after all
3812 * slaves have been released
3813 */
3814 bond_alb_deinitialize(bond);
3815 }
3816
3817 return 0;
3818}
3819
28172739
ED		 3820static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev,
3821 struct rtnl_link_stats64 *stats)
1da177e4 3822{
454d7c9b 3823 struct bonding *bond = netdev_priv(bond_dev);
28172739 3824 struct rtnl_link_stats64 temp;
1da177e4
LT		 3825 struct slave *slave;
3826 int i;
3827
28172739 3828 memset(stats, 0, sizeof(*stats));
1da177e4
LT		 3829
3830 read_lock_bh(&bond->lock);
3831
3832 bond_for_each_slave(bond, slave, i) {
be1f3c2c 3833 const struct rtnl_link_stats64 *sstats =
28172739
ED		 3834 dev_get_stats(slave->dev, &temp);
3835
3836 stats->rx_packets += sstats->rx_packets;
3837 stats->rx_bytes += sstats->rx_bytes;
3838 stats->rx_errors += sstats->rx_errors;
3839 stats->rx_dropped += sstats->rx_dropped;
3840
3841 stats->tx_packets += sstats->tx_packets;
3842 stats->tx_bytes += sstats->tx_bytes;
3843 stats->tx_errors += sstats->tx_errors;
3844 stats->tx_dropped += sstats->tx_dropped;
3845
3846 stats->multicast += sstats->multicast;
3847 stats->collisions += sstats->collisions;
3848
3849 stats->rx_length_errors += sstats->rx_length_errors;
3850 stats->rx_over_errors += sstats->rx_over_errors;
3851 stats->rx_crc_errors += sstats->rx_crc_errors;
3852 stats->rx_frame_errors += sstats->rx_frame_errors;
3853 stats->rx_fifo_errors += sstats->rx_fifo_errors;
3854 stats->rx_missed_errors += sstats->rx_missed_errors;
3855
3856 stats->tx_aborted_errors += sstats->tx_aborted_errors;
3857 stats->tx_carrier_errors += sstats->tx_carrier_errors;
3858 stats->tx_fifo_errors += sstats->tx_fifo_errors;
3859 stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
3860 stats->tx_window_errors += sstats->tx_window_errors;
2439f9eb
AG		 3861 }
3862
1da177e4
LT		 3863 read_unlock_bh(&bond->lock);
3864
3865 return stats;
3866}
3867
3868static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
3869{
3870 struct net_device *slave_dev = NULL;
3871 struct ifbond k_binfo;
3872 struct ifbond __user *u_binfo = NULL;
3873 struct ifslave k_sinfo;
3874 struct ifslave __user *u_sinfo = NULL;
3875 struct mii_ioctl_data *mii = NULL;
1da177e4
LT		 3876 int res = 0;
3877
a4aee5c8 3878 pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd);
1da177e4
LT		 3879
3880 switch (cmd) {
1da177e4
LT		 3881 case SIOCGMIIPHY:
3882 mii = if_mii(ifr);
3d632c3f 3883 if (!mii)
1da177e4 3884 return -EINVAL;
3d632c3f 3885
1da177e4
LT		 3886 mii->phy_id = 0;
3887 /* Fall Through */
3888 case SIOCGMIIREG:
3889 /*
3890 * We do this again just in case we were called by SIOCGMIIREG
3891 * instead of SIOCGMIIPHY.
3892 */
3893 mii = if_mii(ifr);
3d632c3f 3894 if (!mii)
1da177e4 3895 return -EINVAL;
3d632c3f 3896
1da177e4
LT		 3897
3898 if (mii->reg_num == 1) {
454d7c9b 3899 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 3900 mii->val_out = 0;
6603a6f2 3901 read_lock(&bond->lock);
1da177e4 3902 read_lock(&bond->curr_slave_lock);
3d632c3f 3903 if (netif_carrier_ok(bond->dev))
1da177e4 3904 mii->val_out = BMSR_LSTATUS;
3d632c3f 3905
1da177e4 3906 read_unlock(&bond->curr_slave_lock);
6603a6f2 3907 read_unlock(&bond->lock);
1da177e4
LT		 3908 }
3909
3910 return 0;
3911 case BOND_INFO_QUERY_OLD:
3912 case SIOCBONDINFOQUERY:
3913 u_binfo = (struct ifbond __user *)ifr->ifr_data;
3914
3d632c3f 3915 if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond)))
1da177e4 3916 return -EFAULT;
1da177e4
LT		 3917
3918 res = bond_info_query(bond_dev, &k_binfo);
3d632c3f
SH		 3919 if (res == 0 &&
3920 copy_to_user(u_binfo, &k_binfo, sizeof(ifbond)))
3921 return -EFAULT;
1da177e4
LT		 3922
3923 return res;
3924 case BOND_SLAVE_INFO_QUERY_OLD:
3925 case SIOCBONDSLAVEINFOQUERY:
3926 u_sinfo = (struct ifslave __user *)ifr->ifr_data;
3927
3d632c3f 3928 if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave)))
1da177e4 3929 return -EFAULT;
1da177e4
LT		 3930
3931 res = bond_slave_info_query(bond_dev, &k_sinfo);
3d632c3f
SH		 3932 if (res == 0 &&
3933 copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave)))
3934 return -EFAULT;
1da177e4
LT		 3935
3936 return res;
3937 default:
3938 /* Go on */
3939 break;
3940 }
3941
3d632c3f 3942 if (!capable(CAP_NET_ADMIN))
1da177e4 3943 return -EPERM;
1da177e4 3944
ec87fd3b 3945 slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave);
1da177e4 3946
a4aee5c8 3947 pr_debug("slave_dev=%p:\n", slave_dev);
1da177e4 3948
3d632c3f 3949 if (!slave_dev)
1da177e4 3950 res = -ENODEV;
3d632c3f 3951 else {
a4aee5c8 3952 pr_debug("slave_dev->name=%s:\n", slave_dev->name);
1da177e4
LT		 3953 switch (cmd) {
3954 case BOND_ENSLAVE_OLD:
3955 case SIOCBONDENSLAVE:
3956 res = bond_enslave(bond_dev, slave_dev);
3957 break;
3958 case BOND_RELEASE_OLD:
3959 case SIOCBONDRELEASE:
3960 res = bond_release(bond_dev, slave_dev);
3961 break;
3962 case BOND_SETHWADDR_OLD:
3963 case SIOCBONDSETHWADDR:
3964 res = bond_sethwaddr(bond_dev, slave_dev);
3965 break;
3966 case BOND_CHANGE_ACTIVE_OLD:
3967 case SIOCBONDCHANGEACTIVE:
3968 res = bond_ioctl_change_active(bond_dev, slave_dev);
3969 break;
3970 default:
3971 res = -EOPNOTSUPP;
3972 }
3973
3974 dev_put(slave_dev);
3975 }
3976
1da177e4
LT		 3977 return res;
3978}
3979
22bedad3
JP		 3980static bool bond_addr_in_mc_list(unsigned char *addr,
3981 struct netdev_hw_addr_list *list,
3982 int addrlen)
3983{
3984 struct netdev_hw_addr *ha;
3985
3986 netdev_hw_addr_list_for_each(ha, list)
3987 if (!memcmp(ha->addr, addr, addrlen))
3988 return true;
3989
3990 return false;
3991}
3992
1da177e4
LT		 3993static void bond_set_multicast_list(struct net_device *bond_dev)
3994{
454d7c9b 3995 struct bonding *bond = netdev_priv(bond_dev);
22bedad3
JP		 3996 struct netdev_hw_addr *ha;
3997 bool found;
1da177e4 3998
1da177e4
LT		 3999 /*
4000 * Do promisc before checking multicast_mode
4001 */
3d632c3f 4002 if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC))
7e1a1ac1
WC		 4003 /*
4004 * FIXME: Need to handle the error when one of the multi-slaves
4005 * encounters error.
4006 */
1da177e4 4007 bond_set_promiscuity(bond, 1);
1da177e4 4008
3d632c3f
SH		 4009
4010 if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC))
1da177e4 4011 bond_set_promiscuity(bond, -1);
3d632c3f 4012
1da177e4
LT		 4013
4014 /* set allmulti flag to slaves */
3d632c3f 4015 if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI))
7e1a1ac1
WC		 4016 /*
4017 * FIXME: Need to handle the error when one of the multi-slaves
4018 * encounters error.
4019 */
1da177e4 4020 bond_set_allmulti(bond, 1);
1da177e4 4021
3d632c3f
SH		 4022
4023 if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI))
1da177e4 4024 bond_set_allmulti(bond, -1);
3d632c3f 4025
1da177e4 4026
80ee5ad2
JV		 4027 read_lock(&bond->lock);
4028
1da177e4
LT		 4029 bond->flags = bond_dev->flags;
4030
4031 /* looking for addresses to add to slaves' mc list */
22bedad3
JP		 4032 netdev_for_each_mc_addr(ha, bond_dev) {
4033 found = bond_addr_in_mc_list(ha->addr, &bond->mc_list,
4034 bond_dev->addr_len);
4035 if (!found)
4036 bond_mc_add(bond, ha->addr);
1da177e4
LT		 4037 }
4038
4039 /* looking for addresses to delete from slaves' list */
22bedad3
JP		 4040 netdev_hw_addr_list_for_each(ha, &bond->mc_list) {
4041 found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc,
4042 bond_dev->addr_len);
4043 if (!found)
4044 bond_mc_del(bond, ha->addr);
1da177e4
LT		 4045 }
4046
4047 /* save master's multicast list */
22bedad3
JP		 4048 __hw_addr_flush(&bond->mc_list);
4049 __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc,
4050 bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST);
1da177e4 4051
80ee5ad2 4052 read_unlock(&bond->lock);
1da177e4
LT		 4053}
4054
00829823
SH		 4055static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms)
4056{
4057 struct bonding *bond = netdev_priv(dev);
4058 struct slave *slave = bond->first_slave;
4059
4060 if (slave) {
4061 const struct net_device_ops *slave_ops
4062 = slave->dev->netdev_ops;
4063 if (slave_ops->ndo_neigh_setup)
72e2240f 4064 return slave_ops->ndo_neigh_setup(slave->dev, parms);
00829823
SH		 4065 }
4066 return 0;
4067}
4068
1da177e4
LT		 4069/*
4070 * Change the MTU of all of a master's slaves to match the master
4071 */
4072static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
4073{
454d7c9b 4074 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4075 struct slave *slave, *stop_at;
4076 int res = 0;
4077 int i;
4078
5a03cdb7 4079 pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond,
a4aee5c8 4080 (bond_dev ? bond_dev->name : "None"), new_mtu);
1da177e4
LT		 4081
4082 /* Can't hold bond->lock with bh disabled here since
4083 * some base drivers panic. On the other hand we can't
4084 * hold bond->lock without bh disabled because we'll
4085 * deadlock. The only solution is to rely on the fact
4086 * that we're under rtnl_lock here, and the slaves
4087 * list won't change. This doesn't solve the problem
4088 * of setting the slave's MTU while it is
4089 * transmitting, but the assumption is that the base
4090 * driver can handle that.
4091 *
4092 * TODO: figure out a way to safely iterate the slaves
4093 * list, but without holding a lock around the actual
4094 * call to the base driver.
4095 */
4096
4097 bond_for_each_slave(bond, slave, i) {
a4aee5c8
JP		 4098 pr_debug("s %p s->p %p c_m %p\n",
4099 slave,
4100 slave->prev,
4101 slave->dev->netdev_ops->ndo_change_mtu);
e944ef79 4102
1da177e4
LT		 4103 res = dev_set_mtu(slave->dev, new_mtu);
4104
4105 if (res) {
4106 /* If we failed to set the slave's mtu to the new value
4107 * we must abort the operation even in ACTIVE_BACKUP
4108 * mode, because if we allow the backup slaves to have
4109 * different mtu values than the active slave we'll
4110 * need to change their mtu when doing a failover. That
4111 * means changing their mtu from timer context, which
4112 * is probably not a good idea.
4113 */
5a03cdb7 4114 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 4115 goto unwind;
4116 }
4117 }
4118
4119 bond_dev->mtu = new_mtu;
4120
4121 return 0;
4122
4123unwind:
4124 /* unwind from head to the slave that failed */
4125 stop_at = slave;
4126 bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
4127 int tmp_res;
4128
4129 tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu);
4130 if (tmp_res) {
a4aee5c8
JP		 4131 pr_debug("unwind err %d dev %s\n",
4132 tmp_res, slave->dev->name);
1da177e4
LT		 4133 }
4134 }
4135
4136 return res;
4137}
4138
4139/*
4140 * Change HW address
4141 *
4142 * Note that many devices must be down to change the HW address, and
4143 * downing the master releases all slaves. We can make bonds full of
4144 * bonding devices to test this, however.
4145 */
4146static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
4147{
454d7c9b 4148 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4149 struct sockaddr *sa = addr, tmp_sa;
4150 struct slave *slave, *stop_at;
4151 int res = 0;
4152 int i;
4153
eb7cc59a
SH		 4154 if (bond->params.mode == BOND_MODE_ALB)
4155 return bond_alb_set_mac_address(bond_dev, addr);
4156
4157
a4aee5c8
JP		 4158 pr_debug("bond=%p, name=%s\n",
4159 bond, bond_dev ? bond_dev->name : "None");
1da177e4 4160
dd957c57 4161 /*
3915c1e8
JV		 4162 * If fail_over_mac is set to active, do nothing and return
4163 * success. Returning an error causes ifenslave to fail.
dd957c57 4164 */
3915c1e8 4165 if (bond->params.fail_over_mac == BOND_FOM_ACTIVE)
dd957c57 4166 return 0;
2ab82852 4167
3d632c3f 4168 if (!is_valid_ether_addr(sa->sa_data))
1da177e4 4169 return -EADDRNOTAVAIL;
1da177e4
LT		 4170
4171 /* Can't hold bond->lock with bh disabled here since
4172 * some base drivers panic. On the other hand we can't
4173 * hold bond->lock without bh disabled because we'll
4174 * deadlock. The only solution is to rely on the fact
4175 * that we're under rtnl_lock here, and the slaves
4176 * list won't change. This doesn't solve the problem
4177 * of setting the slave's hw address while it is
4178 * transmitting, but the assumption is that the base
4179 * driver can handle that.
4180 *
4181 * TODO: figure out a way to safely iterate the slaves
4182 * list, but without holding a lock around the actual
4183 * call to the base driver.
4184 */
4185
4186 bond_for_each_slave(bond, slave, i) {
eb7cc59a 4187 const struct net_device_ops *slave_ops = slave->dev->netdev_ops;
5a03cdb7 4188 pr_debug("slave %p %s\n", slave, slave->dev->name);
1da177e4 4189
eb7cc59a 4190 if (slave_ops->ndo_set_mac_address == NULL) {
1da177e4 4191 res = -EOPNOTSUPP;
5a03cdb7 4192 pr_debug("EOPNOTSUPP %s\n", slave->dev->name);
1da177e4
LT		 4193 goto unwind;
4194 }
4195
4196 res = dev_set_mac_address(slave->dev, addr);
4197 if (res) {
4198 /* TODO: consider downing the slave
4199 * and retry ?
4200 * User should expect communications
4201 * breakage anyway until ARP finish
4202 * updating, so...
4203 */
5a03cdb7 4204 pr_debug("err %d %s\n", res, slave->dev->name);
1da177e4
LT		 4205 goto unwind;
4206 }
4207 }
4208
4209 /* success */
4210 memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
4211 return 0;
4212
4213unwind:
4214 memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
4215 tmp_sa.sa_family = bond_dev->type;
4216
4217 /* unwind from head to the slave that failed */
4218 stop_at = slave;
4219 bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
4220 int tmp_res;
4221
4222 tmp_res = dev_set_mac_address(slave->dev, &tmp_sa);
4223 if (tmp_res) {
a4aee5c8
JP		 4224 pr_debug("unwind err %d dev %s\n",
4225 tmp_res, slave->dev->name);
1da177e4
LT		 4226 }
4227 }
4228
4229 return res;
4230}
4231
4232static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
4233{
454d7c9b 4234 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 4235 struct slave *slave, *start_at;
cf5f9044 4236 int i, slave_no, res = 1;
a2fd940f 4237 struct iphdr *iph = ip_hdr(skb);
1da177e4
LT		 4238
4239 read_lock(&bond->lock);
4240
3d632c3f 4241 if (!BOND_IS_OK(bond))
1da177e4 4242 goto out;
cf5f9044 4243 /*
a2fd940f
AG		 4244 * Start with the curr_active_slave that joined the bond as the
4245 * default for sending IGMP traffic. For failover purposes one
4246 * needs to maintain some consistency for the interface that will
4247 * send the join/membership reports. The curr_active_slave found
4248 * will send all of this type of traffic.
cf5f9044 4249 */
00ae7028 4250 if ((iph->protocol == IPPROTO_IGMP) &&
a2fd940f 4251 (skb->protocol == htons(ETH_P_IP))) {
1da177e4 4252
a2fd940f
AG		 4253 read_lock(&bond->curr_slave_lock);
4254 slave = bond->curr_active_slave;
4255 read_unlock(&bond->curr_slave_lock);
4256
4257 if (!slave)
4258 goto out;
4259 } else {
4260 /*
4261 * Concurrent TX may collide on rr_tx_counter; we accept
4262 * that as being rare enough not to justify using an
4263 * atomic op here.
4264 */
4265 slave_no = bond->rr_tx_counter++ % bond->slave_cnt;
4266
4267 bond_for_each_slave(bond, slave, i) {
4268 slave_no--;
4269 if (slave_no < 0)
4270 break;
4271 }
1da177e4
LT		 4272 }
4273
cf5f9044 4274 start_at = slave;
1da177e4
LT		 4275 bond_for_each_slave_from(bond, slave, i, start_at) {
4276 if (IS_UP(slave->dev) &&
4277 (slave->link == BOND_LINK_UP) &&
4278 (slave->state == BOND_STATE_ACTIVE)) {
4279 res = bond_dev_queue_xmit(bond, skb, slave->dev);
1da177e4
LT		 4280 break;
4281 }
4282 }
4283
1da177e4
LT		 4284out:
4285 if (res) {
4286 /* no suitable interface, frame not sent */
4287 dev_kfree_skb(skb);
4288 }
4289 read_unlock(&bond->lock);
ec634fe3 4290 return NETDEV_TX_OK;
1da177e4
LT		 4291}
4292
075897ce 4293
1da177e4
LT		 4294/*
4295 * in active-backup mode, we know that bond->curr_active_slave is always valid if
4296 * the bond has a usable interface.
4297 */
4298static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
4299{
454d7c9b 4300 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4301 int res = 1;
4302
1da177e4
LT		 4303 read_lock(&bond->lock);
4304 read_lock(&bond->curr_slave_lock);
4305
3d632c3f 4306 if (!BOND_IS_OK(bond))
1da177e4 4307 goto out;
1da177e4 4308
075897ce
JL		 4309 if (!bond->curr_active_slave)
4310 goto out;
4311
075897ce
JL		 4312 res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev);
4313
1da177e4 4314out:
3d632c3f 4315 if (res)
1da177e4
LT		 4316 /* no suitable interface, frame not sent */
4317 dev_kfree_skb(skb);
3d632c3f 4318
1da177e4
LT		 4319 read_unlock(&bond->curr_slave_lock);
4320 read_unlock(&bond->lock);
ec634fe3 4321 return NETDEV_TX_OK;
1da177e4
LT		 4322}
4323
4324/*
169a3e66
JV		 4325 * In bond_xmit_xor() , we determine the output device by using a pre-
4326 * determined xmit_hash_policy(), If the selected device is not enabled,
4327 * find the next active slave.
1da177e4
LT		 4328 */
4329static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
4330{
454d7c9b 4331 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4332 struct slave *slave, *start_at;
4333 int slave_no;
4334 int i;
4335 int res = 1;
4336
4337 read_lock(&bond->lock);
4338
3d632c3f 4339 if (!BOND_IS_OK(bond))
1da177e4 4340 goto out;
1da177e4 4341
a361c83c 4342 slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt);
1da177e4
LT		 4343
4344 bond_for_each_slave(bond, slave, i) {
4345 slave_no--;
3d632c3f 4346 if (slave_no < 0)
1da177e4 4347 break;
1da177e4
LT		 4348 }
4349
4350 start_at = slave;
4351
4352 bond_for_each_slave_from(bond, slave, i, start_at) {
4353 if (IS_UP(slave->dev) &&
4354 (slave->link == BOND_LINK_UP) &&
4355 (slave->state == BOND_STATE_ACTIVE)) {
4356 res = bond_dev_queue_xmit(bond, skb, slave->dev);
4357 break;
4358 }
4359 }
4360
4361out:
4362 if (res) {
4363 /* no suitable interface, frame not sent */
4364 dev_kfree_skb(skb);
4365 }
4366 read_unlock(&bond->lock);
ec634fe3 4367 return NETDEV_TX_OK;
1da177e4
LT		 4368}
4369
4370/*
4371 * in broadcast mode, we send everything to all usable interfaces.
4372 */
4373static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
4374{
454d7c9b 4375 struct bonding *bond = netdev_priv(bond_dev);
1da177e4
LT		 4376 struct slave *slave, *start_at;
4377 struct net_device *tx_dev = NULL;
4378 int i;
4379 int res = 1;
4380
4381 read_lock(&bond->lock);
4382
3d632c3f 4383 if (!BOND_IS_OK(bond))
1da177e4 4384 goto out;
1da177e4
LT		 4385
4386 read_lock(&bond->curr_slave_lock);
4387 start_at = bond->curr_active_slave;
4388 read_unlock(&bond->curr_slave_lock);
4389
3d632c3f 4390 if (!start_at)
1da177e4 4391 goto out;
1da177e4
LT		 4392
4393 bond_for_each_slave_from(bond, slave, i, start_at) {
4394 if (IS_UP(slave->dev) &&
4395 (slave->link == BOND_LINK_UP) &&
4396 (slave->state == BOND_STATE_ACTIVE)) {
4397 if (tx_dev) {
4398 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
4399 if (!skb2) {
a4aee5c8 4400 pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n",
4e0952c7 4401 bond_dev->name);
1da177e4
LT		 4402 continue;
4403 }
4404
4405 res = bond_dev_queue_xmit(bond, skb2, tx_dev);
4406 if (res) {
4407 dev_kfree_skb(skb2);
4408 continue;
4409 }
4410 }
4411 tx_dev = slave->dev;
4412 }
4413 }
4414
3d632c3f 4415 if (tx_dev)
1da177e4 4416 res = bond_dev_queue_xmit(bond, skb, tx_dev);
1da177e4
LT		 4417
4418out:
3d632c3f 4419 if (res)
1da177e4
LT		 4420 /* no suitable interface, frame not sent */
4421 dev_kfree_skb(skb);
3d632c3f 4422
1da177e4
LT		 4423 /* frame sent to all suitable interfaces */
4424 read_unlock(&bond->lock);
ec634fe3 4425 return NETDEV_TX_OK;
1da177e4
LT		 4426}
4427
4428/*------------------------- Device initialization ---------------------------*/
4429
6f6652be
JV		 4430static void bond_set_xmit_hash_policy(struct bonding *bond)
4431{
4432 switch (bond->params.xmit_policy) {
4433 case BOND_XMIT_POLICY_LAYER23:
4434 bond->xmit_hash_policy = bond_xmit_hash_policy_l23;
4435 break;
4436 case BOND_XMIT_POLICY_LAYER34:
4437 bond->xmit_hash_policy = bond_xmit_hash_policy_l34;
4438 break;
4439 case BOND_XMIT_POLICY_LAYER2:
4440 default:
4441 bond->xmit_hash_policy = bond_xmit_hash_policy_l2;
4442 break;
4443 }
4444}
4445
bb1d9123
AG		 4446/*
4447 * Lookup the slave that corresponds to a qid
4448 */
4449static inline int bond_slave_override(struct bonding *bond,
4450 struct sk_buff *skb)
4451{
4452 int i, res = 1;
4453 struct slave *slave = NULL;
4454 struct slave *check_slave;
4455
4456 read_lock(&bond->lock);
4457
4458 if (!BOND_IS_OK(bond) || !skb->queue_mapping)
4459 goto out;
4460
4461 /* Find out if any slaves have the same mapping as this skb. */
4462 bond_for_each_slave(bond, check_slave, i) {
4463 if (check_slave->queue_id == skb->queue_mapping) {
4464 slave = check_slave;
4465 break;
4466 }
4467 }
4468
4469 /* If the slave isn't UP, use default transmit policy. */
4470 if (slave && slave->queue_id && IS_UP(slave->dev) &&
4471 (slave->link == BOND_LINK_UP)) {
4472 res = bond_dev_queue_xmit(bond, skb, slave->dev);
4473 }
4474
4475out:
4476 read_unlock(&bond->lock);
4477 return res;
4478}
4479
4480static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb)
4481{
4482 /*
4483 * This helper function exists to help dev_pick_tx get the correct
4484 * destination queue. Using a helper function skips the a call to
4485 * skb_tx_hash and will put the skbs in the queue we expect on their
4486 * way down to the bonding driver.
4487 */
4488 return skb->queue_mapping;
4489}
4490
424efe9c 4491static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)
00829823 4492{
bb1d9123
AG		 4493 struct bonding *bond = netdev_priv(dev);
4494
4495 if (TX_QUEUE_OVERRIDE(bond->params.mode)) {
4496 if (!bond_slave_override(bond, skb))
4497 return NETDEV_TX_OK;
4498 }
00829823
SH		 4499
4500 switch (bond->params.mode) {
4501 case BOND_MODE_ROUNDROBIN:
4502 return bond_xmit_roundrobin(skb, dev);
4503 case BOND_MODE_ACTIVEBACKUP:
4504 return bond_xmit_activebackup(skb, dev);
4505 case BOND_MODE_XOR:
4506 return bond_xmit_xor(skb, dev);
4507 case BOND_MODE_BROADCAST:
4508 return bond_xmit_broadcast(skb, dev);
4509 case BOND_MODE_8023AD:
4510 return bond_3ad_xmit_xor(skb, dev);
4511 case BOND_MODE_ALB:
4512 case BOND_MODE_TLB:
4513 return bond_alb_xmit(skb, dev);
4514 default:
4515 /* Should never happen, mode already checked */
a4aee5c8
JP		 4516 pr_err("%s: Error: Unknown bonding mode %d\n",
4517 dev->name, bond->params.mode);
00829823
SH		 4518 WARN_ON_ONCE(1);
4519 dev_kfree_skb(skb);
4520 return NETDEV_TX_OK;
4521 }
4522}
4523
4524
1da177e4
LT		 4525/*
4526 * set bond mode specific net device operations
4527 */
a77b5325 4528void bond_set_mode_ops(struct bonding *bond, int mode)
1da177e4 4529{
169a3e66
JV		 4530 struct net_device *bond_dev = bond->dev;
4531
1da177e4
LT		 4532 switch (mode) {
4533 case BOND_MODE_ROUNDROBIN:
1da177e4
LT		 4534 break;
4535 case BOND_MODE_ACTIVEBACKUP:
1da177e4
LT		 4536 break;
4537 case BOND_MODE_XOR:
6f6652be 4538 bond_set_xmit_hash_policy(bond);
1da177e4
LT		 4539 break;
4540 case BOND_MODE_BROADCAST:
1da177e4
LT		 4541 break;
4542 case BOND_MODE_8023AD:
8f903c70 4543 bond_set_master_3ad_flags(bond);
6f6652be 4544 bond_set_xmit_hash_policy(bond);
1da177e4 4545 break;
1da177e4 4546 case BOND_MODE_ALB:
8f903c70
JV		 4547 bond_set_master_alb_flags(bond);
4548 /* FALLTHRU */
4549 case BOND_MODE_TLB:
1da177e4
LT		 4550 break;
4551 default:
4552 /* Should never happen, mode already checked */
a4aee5c8
JP		 4553 pr_err("%s: Error: Unknown bonding mode %d\n",
4554 bond_dev->name, mode);
1da177e4
LT		 4555 break;
4556 }
4557}
4558
217df670
JV		 4559static void bond_ethtool_get_drvinfo(struct net_device *bond_dev,
4560 struct ethtool_drvinfo *drvinfo)
4561{
4562 strncpy(drvinfo->driver, DRV_NAME, 32);
4563 strncpy(drvinfo->version, DRV_VERSION, 32);
4564 snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION);
4565}
4566
7282d491 4567static const struct ethtool_ops bond_ethtool_ops = {
217df670 4568 .get_drvinfo = bond_ethtool_get_drvinfo,
fa53ebac
SH		 4569 .get_link = ethtool_op_get_link,
4570 .get_tx_csum = ethtool_op_get_tx_csum,
4571 .get_sg = ethtool_op_get_sg,
4572 .get_tso = ethtool_op_get_tso,
4573 .get_ufo = ethtool_op_get_ufo,
4574 .get_flags = ethtool_op_get_flags,
8531c5ff
AK		 4575};
4576
eb7cc59a 4577static const struct net_device_ops bond_netdev_ops = {
181470fc 4578 .ndo_init = bond_init,
9e71626c 4579 .ndo_uninit = bond_uninit,
eb7cc59a
SH		 4580 .ndo_open = bond_open,
4581 .ndo_stop = bond_close,
00829823 4582 .ndo_start_xmit = bond_start_xmit,
bb1d9123 4583 .ndo_select_queue = bond_select_queue,
be1f3c2c 4584 .ndo_get_stats64 = bond_get_stats,
eb7cc59a
SH		 4585 .ndo_do_ioctl = bond_do_ioctl,
4586 .ndo_set_multicast_list = bond_set_multicast_list,
4587 .ndo_change_mtu = bond_change_mtu,
eb7cc59a 4588 .ndo_set_mac_address = bond_set_mac_address,
00829823 4589 .ndo_neigh_setup = bond_neigh_setup,
eb7cc59a
SH		 4590 .ndo_vlan_rx_register = bond_vlan_rx_register,
4591 .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid,
4592 .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid,
f6dc31a8
WC		 4593#ifdef CONFIG_NET_POLL_CONTROLLER
4594 .ndo_netpoll_cleanup = bond_netpoll_cleanup,
4595 .ndo_poll_controller = bond_poll_controller,
4596#endif
eb7cc59a
SH		 4597};
4598
9e2e61fb
AW		 4599static void bond_destructor(struct net_device *bond_dev)
4600{
4601 struct bonding *bond = netdev_priv(bond_dev);
4602 if (bond->wq)
4603 destroy_workqueue(bond->wq);
4604 free_netdev(bond_dev);
4605}
4606
181470fc 4607static void bond_setup(struct net_device *bond_dev)
1da177e4 4608{
454d7c9b 4609 struct bonding *bond = netdev_priv(bond_dev);
1da177e4 4610
1da177e4
LT		 4611 /* initialize rwlocks */
4612 rwlock_init(&bond->lock);
4613 rwlock_init(&bond->curr_slave_lock);
4614
d2991f75 4615 bond->params = bonding_defaults;
1da177e4
LT		 4616
4617 /* Initialize pointers */
1da177e4
LT		 4618 bond->dev = bond_dev;
4619 INIT_LIST_HEAD(&bond->vlan_list);
4620
4621 /* Initialize the device entry points */
181470fc 4622 ether_setup(bond_dev);
eb7cc59a 4623 bond_dev->netdev_ops = &bond_netdev_ops;
8531c5ff 4624 bond_dev->ethtool_ops = &bond_ethtool_ops;
169a3e66 4625 bond_set_mode_ops(bond, bond->params.mode);
1da177e4 4626
9e2e61fb 4627 bond_dev->destructor = bond_destructor;
1da177e4
LT		 4628
4629 /* Initialize the device options */
4630 bond_dev->tx_queue_len = 0;
4631 bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
0b680e75 4632 bond_dev->priv_flags |= IFF_BONDING;
181470fc
SH		 4633 bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
4634
6cf3f41e
JV		 4635 if (bond->params.arp_interval)
4636 bond_dev->priv_flags |= IFF_MASTER_ARPMON;
1da177e4
LT		 4637
4638 /* At first, we block adding VLANs. That's the only way to
4639 * prevent problems that occur when adding VLANs over an
4640 * empty bond. The block will be removed once non-challenged
4641 * slaves are enslaved.
4642 */
4643 bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
4644
932ff279 4645 /* don't acquire bond device's netif_tx_lock when
1da177e4
LT		 4646 * transmitting */
4647 bond_dev->features |= NETIF_F_LLTX;
4648
4649 /* By default, we declare the bond to be fully
4650 * VLAN hardware accelerated capable. Special
4651 * care is taken in the various xmit functions
4652 * when there are slaves that are not hw accel
4653 * capable
4654 */
1da177e4
LT		 4655 bond_dev->features |= (NETIF_F_HW_VLAN_TX |
4656 NETIF_F_HW_VLAN_RX |
4657 NETIF_F_HW_VLAN_FILTER);
4658
1da177e4
LT		 4659}
4660
fdaea7a9
JV		 4661static void bond_work_cancel_all(struct bonding *bond)
4662{
4663 write_lock_bh(&bond->lock);
4664 bond->kill_timers = 1;
4665 write_unlock_bh(&bond->lock);
4666
4667 if (bond->params.miimon && delayed_work_pending(&bond->mii_work))
4668 cancel_delayed_work(&bond->mii_work);
4669
4670 if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work))
4671 cancel_delayed_work(&bond->arp_work);
4672
4673 if (bond->params.mode == BOND_MODE_ALB &&
4674 delayed_work_pending(&bond->alb_work))
4675 cancel_delayed_work(&bond->alb_work);
4676
4677 if (bond->params.mode == BOND_MODE_8023AD &&
4678 delayed_work_pending(&bond->ad_work))
4679 cancel_delayed_work(&bond->ad_work);
4680}
4681
c67dfb29
EB		 4682/*
4683* Destroy a bonding device.
4684* Must be under rtnl_lock when this function is called.
4685*/
4686static void bond_uninit(struct net_device *bond_dev)
a434e43f 4687{
454d7c9b 4688 struct bonding *bond = netdev_priv(bond_dev);
a434e43f 4689
f6dc31a8
WC		 4690 bond_netpoll_cleanup(bond_dev);
4691
c67dfb29
EB		 4692 /* Release the bonded slaves */
4693 bond_release_all(bond_dev);
4694
a434e43f
JV		 4695 list_del(&bond->bond_list);
4696
4697 bond_work_cancel_all(bond);
4698
a434e43f 4699 bond_remove_proc_entry(bond);
c67dfb29 4700
22bedad3 4701 __hw_addr_flush(&bond->mc_list);
a434e43f
JV		 4702}
4703
1da177e4
LT		 4704/*------------------------- Module initialization ---------------------------*/
4705
4706/*
4707 * Convert string input module parms. Accept either the
ece95f7f
JV		 4708 * number of the mode or its string name. A bit complicated because
4709 * some mode names are substrings of other names, and calls from sysfs
4710 * may have whitespace in the name (trailing newlines, for example).
1da177e4 4711 */
325dcf7a 4712int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl)
1da177e4 4713{
54b87323 4714 int modeint = -1, i, rv;
a42e534f 4715 char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, };
ece95f7f 4716
a42e534f
JV		 4717 for (p = (char *)buf; *p; p++)
4718 if (!(isdigit(*p) || isspace(*p)))
4719 break;
4720
4721 if (*p)
ece95f7f 4722 rv = sscanf(buf, "%20s", modestr);
a42e534f 4723 else
54b87323 4724 rv = sscanf(buf, "%d", &modeint);
a42e534f
JV		 4725
4726 if (!rv)
4727 return -1;
1da177e4
LT		 4728
4729 for (i = 0; tbl[i].modename; i++) {
54b87323 4730 if (modeint == tbl[i].mode)
ece95f7f
JV		 4731 return tbl[i].mode;
4732 if (strcmp(modestr, tbl[i].modename) == 0)
1da177e4 4733 return tbl[i].mode;
1da177e4
LT		 4734 }
4735
4736 return -1;
4737}
4738
4739static int bond_check_params(struct bond_params *params)
4740{
a549952a 4741 int arp_validate_value, fail_over_mac_value, primary_reselect_value;
f5b2b966 4742
1da177e4
LT		 4743 /*
4744 * Convert string parameters.
4745 */
4746 if (mode) {
4747 bond_mode = bond_parse_parm(mode, bond_mode_tbl);
4748 if (bond_mode == -1) {
a4aee5c8 4749 pr_err("Error: Invalid bonding mode \"%s\"\n",
1da177e4
LT		 4750 mode == NULL ? "NULL" : mode);
4751 return -EINVAL;
4752 }
4753 }
4754
169a3e66
JV		 4755 if (xmit_hash_policy) {
4756 if ((bond_mode != BOND_MODE_XOR) &&
4757 (bond_mode != BOND_MODE_8023AD)) {
a4aee5c8 4758 pr_info("xmit_hash_policy param is irrelevant in mode %s\n",
169a3e66
JV		 4759 bond_mode_name(bond_mode));
4760 } else {
4761 xmit_hashtype = bond_parse_parm(xmit_hash_policy,
4762 xmit_hashtype_tbl);
4763 if (xmit_hashtype == -1) {
a4aee5c8 4764 pr_err("Error: Invalid xmit_hash_policy \"%s\"\n",
3d632c3f 4765 xmit_hash_policy == NULL ? "NULL" :
169a3e66
JV		 4766 xmit_hash_policy);
4767 return -EINVAL;
4768 }
4769 }
4770 }
4771
1da177e4
LT		 4772 if (lacp_rate) {
4773 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8
JP		 4774 pr_info("lacp_rate param is irrelevant in mode %s\n",
4775 bond_mode_name(bond_mode));
1da177e4
LT		 4776 } else {
4777 lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl);
4778 if (lacp_fast == -1) {
a4aee5c8 4779 pr_err("Error: Invalid lacp rate \"%s\"\n",
1da177e4
LT		 4780 lacp_rate == NULL ? "NULL" : lacp_rate);
4781 return -EINVAL;
4782 }
4783 }
4784 }
4785
fd989c83
JV		 4786 if (ad_select) {
4787 params->ad_select = bond_parse_parm(ad_select, ad_select_tbl);
4788 if (params->ad_select == -1) {
a4aee5c8 4789 pr_err("Error: Invalid ad_select \"%s\"\n",
fd989c83
JV		 4790 ad_select == NULL ? "NULL" : ad_select);
4791 return -EINVAL;
4792 }
4793
4794 if (bond_mode != BOND_MODE_8023AD) {
a4aee5c8 4795 pr_warning("ad_select param only affects 802.3ad mode\n");
fd989c83
JV		 4796 }
4797 } else {
4798 params->ad_select = BOND_AD_STABLE;
4799 }
4800
f5841306 4801 if (max_bonds < 0) {
a4aee5c8
JP		 4802 pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n",
4803 max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS);
1da177e4
LT		 4804 max_bonds = BOND_DEFAULT_MAX_BONDS;
4805 }
4806
4807 if (miimon < 0) {
a4aee5c8
JP		 4808 pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n",
4809 miimon, INT_MAX, BOND_LINK_MON_INTERV);
1da177e4
LT		 4810 miimon = BOND_LINK_MON_INTERV;
4811 }
4812
4813 if (updelay < 0) {
a4aee5c8
JP		 4814 pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4815 updelay, INT_MAX);
1da177e4
LT		 4816 updelay = 0;
4817 }
4818
4819 if (downdelay < 0) {
a4aee5c8
JP		 4820 pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n",
4821 downdelay, INT_MAX);
1da177e4
LT		 4822 downdelay = 0;
4823 }
4824
4825 if ((use_carrier != 0) && (use_carrier != 1)) {
a4aee5c8
JP		 4826 pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n",
4827 use_carrier);
1da177e4
LT		 4828 use_carrier = 1;
4829 }
4830
7893b249 4831 if (num_grat_arp < 0 || num_grat_arp > 255) {
2381a55c 4832 pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n",
a4aee5c8 4833 num_grat_arp);
7893b249
MS		 4834 num_grat_arp = 1;
4835 }
4836
305d552a 4837 if (num_unsol_na < 0 || num_unsol_na > 255) {
2381a55c 4838 pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n",
a4aee5c8 4839 num_unsol_na);
305d552a
BH		 4840 num_unsol_na = 1;
4841 }
4842
1da177e4
LT		 4843 /* reset values for 802.3ad */
4844 if (bond_mode == BOND_MODE_8023AD) {
4845 if (!miimon) {
a4aee5c8 4846 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 4847 pr_warning("Forcing miimon to 100msec\n");
1da177e4
LT		 4848 miimon = 100;
4849 }
4850 }
4851
bb1d9123
AG		 4852 if (tx_queues < 1 || tx_queues > 255) {
4853 pr_warning("Warning: tx_queues (%d) should be between "
4854 "1 and 255, resetting to %d\n",
4855 tx_queues, BOND_DEFAULT_TX_QUEUES);
4856 tx_queues = BOND_DEFAULT_TX_QUEUES;
4857 }
4858
ebd8e497
AG		 4859 if ((all_slaves_active != 0) && (all_slaves_active != 1)) {
4860 pr_warning("Warning: all_slaves_active module parameter (%d), "
4861 "not of valid value (0/1), so it was set to "
4862 "0\n", all_slaves_active);
4863 all_slaves_active = 0;
4864 }
4865
1da177e4
LT		 4866 /* reset values for TLB/ALB */
4867 if ((bond_mode == BOND_MODE_TLB) ||
4868 (bond_mode == BOND_MODE_ALB)) {
4869 if (!miimon) {
a4aee5c8 4870 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 4871 pr_warning("Forcing miimon to 100msec\n");
1da177e4
LT		 4872 miimon = 100;
4873 }
4874 }
4875
4876 if (bond_mode == BOND_MODE_ALB) {
a4aee5c8
JP		 4877 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",
4878 updelay);
1da177e4
LT		 4879 }
4880
4881 if (!miimon) {
4882 if (updelay || downdelay) {
4883 /* just warn the user the up/down delay will have
4884 * no effect since miimon is zero...
4885 */
a4aee5c8
JP		 4886 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",
4887 updelay, downdelay);
1da177e4
LT		 4888 }
4889 } else {
4890 /* don't allow arp monitoring */
4891 if (arp_interval) {
a4aee5c8
JP		 4892 pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n",
4893 miimon, arp_interval);
1da177e4
LT		 4894 arp_interval = 0;
4895 }
4896
4897 if ((updelay % miimon) != 0) {
a4aee5c8
JP		 4898 pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n",
4899 updelay, miimon,
4900 (updelay / miimon) * miimon);
1da177e4
LT		 4901 }
4902
4903 updelay /= miimon;
4904
4905 if ((downdelay % miimon) != 0) {
a4aee5c8
JP		 4906 pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n",
4907 downdelay, miimon,
4908 (downdelay / miimon) * miimon);
1da177e4
LT		 4909 }
4910
4911 downdelay /= miimon;
4912 }
4913
4914 if (arp_interval < 0) {
a4aee5c8
JP		 4915 pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n",
4916 arp_interval, INT_MAX, BOND_LINK_ARP_INTERV);
1da177e4
LT		 4917 arp_interval = BOND_LINK_ARP_INTERV;
4918 }
4919
4920 for (arp_ip_count = 0;
4921 (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count];
4922 arp_ip_count++) {
4923 /* not complete check, but should be good enough to
4924 catch mistakes */
4925 if (!isdigit(arp_ip_target[arp_ip_count][0])) {
a4aee5c8
JP		 4926 pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n",
4927 arp_ip_target[arp_ip_count]);
1da177e4
LT		 4928 arp_interval = 0;
4929 } else {
d3bb52b0 4930 __be32 ip = in_aton(arp_ip_target[arp_ip_count]);
1da177e4
LT		 4931 arp_target[arp_ip_count] = ip;
4932 }
4933 }
4934
4935 if (arp_interval && !arp_ip_count) {
4936 /* don't allow arping if no arp_ip_target given... */
a4aee5c8
JP		 4937 pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n",
4938 arp_interval);
1da177e4
LT		 4939 arp_interval = 0;
4940 }
4941
f5b2b966
JV		 4942 if (arp_validate) {
4943 if (bond_mode != BOND_MODE_ACTIVEBACKUP) {
a4aee5c8 4944 pr_err("arp_validate only supported in active-backup mode\n");
f5b2b966
JV		 4945 return -EINVAL;
4946 }
4947 if (!arp_interval) {
a4aee5c8 4948 pr_err("arp_validate requires arp_interval\n");
f5b2b966
JV		 4949 return -EINVAL;
4950 }
4951
4952 arp_validate_value = bond_parse_parm(arp_validate,
4953 arp_validate_tbl);
4954 if (arp_validate_value == -1) {
a4aee5c8 4955 pr_err("Error: invalid arp_validate \"%s\"\n",
f5b2b966
JV		 4956 arp_validate == NULL ? "NULL" : arp_validate);
4957 return -EINVAL;
4958 }
4959 } else
4960 arp_validate_value = 0;
4961
1da177e4 4962 if (miimon) {
a4aee5c8 4963 pr_info("MII link monitoring set to %d ms\n", miimon);
1da177e4
LT		 4964 } else if (arp_interval) {
4965 int i;
4966
a4aee5c8
JP		 4967 pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):",
4968 arp_interval,
4969 arp_validate_tbl[arp_validate_value].modename,
4970 arp_ip_count);
1da177e4
LT		 4971
4972 for (i = 0; i < arp_ip_count; i++)
e5e2a8fd 4973 pr_info(" %s", arp_ip_target[i]);
1da177e4 4974
e5e2a8fd 4975 pr_info("\n");
1da177e4 4976
b8a9787e 4977 } else if (max_bonds) {
1da177e4
LT		 4978 /* miimon and arp_interval not set, we need one so things
4979 * work as expected, see bonding.txt for details
4980 */
a4aee5c8 4981 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		 4982 }
4983
4984 if (primary && !USES_PRIMARY(bond_mode)) {
4985 /* currently, using a primary only makes sense
4986 * in active backup, TLB or ALB modes
4987 */
a4aee5c8
JP		 4988 pr_warning("Warning: %s primary device specified but has no effect in %s mode\n",
4989 primary, bond_mode_name(bond_mode));
1da177e4
LT		 4990 primary = NULL;
4991 }
4992
a549952a
JP		 4993 if (primary && primary_reselect) {
4994 primary_reselect_value = bond_parse_parm(primary_reselect,
4995 pri_reselect_tbl);
4996 if (primary_reselect_value == -1) {
a4aee5c8 4997 pr_err("Error: Invalid primary_reselect \"%s\"\n",
a549952a
JP		 4998 primary_reselect ==
4999 NULL ? "NULL" : primary_reselect);
5000 return -EINVAL;
5001 }
5002 } else {
5003 primary_reselect_value = BOND_PRI_RESELECT_ALWAYS;
5004 }
5005
3915c1e8
JV		 5006 if (fail_over_mac) {
5007 fail_over_mac_value = bond_parse_parm(fail_over_mac,
5008 fail_over_mac_tbl);
5009 if (fail_over_mac_value == -1) {
a4aee5c8 5010 pr_err("Error: invalid fail_over_mac \"%s\"\n",
3915c1e8
JV		 5011 arp_validate == NULL ? "NULL" : arp_validate);
5012 return -EINVAL;
5013 }
5014
5015 if (bond_mode != BOND_MODE_ACTIVEBACKUP)
a4aee5c8 5016 pr_warning("Warning: fail_over_mac only affects active-backup mode.\n");
3915c1e8
JV		 5017 } else {
5018 fail_over_mac_value = BOND_FOM_NONE;
5019 }
dd957c57 5020
1da177e4
LT		 5021 /* fill params struct with the proper values */
5022 params->mode = bond_mode;
169a3e66 5023 params->xmit_policy = xmit_hashtype;
1da177e4 5024 params->miimon = miimon;
7893b249 5025 params->num_grat_arp = num_grat_arp;
305d552a 5026 params->num_unsol_na = num_unsol_na;
1da177e4 5027 params->arp_interval = arp_interval;
f5b2b966 5028 params->arp_validate = arp_validate_value;
1da177e4
LT		 5029 params->updelay = updelay;
5030 params->downdelay = downdelay;
5031 params->use_carrier = use_carrier;
5032 params->lacp_fast = lacp_fast;
5033 params->primary[0] = 0;
a549952a 5034 params->primary_reselect = primary_reselect_value;
3915c1e8 5035 params->fail_over_mac = fail_over_mac_value;
bb1d9123 5036 params->tx_queues = tx_queues;
ebd8e497 5037 params->all_slaves_active = all_slaves_active;
1da177e4
LT		 5038
5039 if (primary) {
5040 strncpy(params->primary, primary, IFNAMSIZ);
5041 params->primary[IFNAMSIZ - 1] = 0;
5042 }
5043
5044 memcpy(params->arp_targets, arp_target, sizeof(arp_target));
5045
5046 return 0;
5047}
5048
0daa2303 5049static struct lock_class_key bonding_netdev_xmit_lock_key;
cf508b12 5050static struct lock_class_key bonding_netdev_addr_lock_key;
0daa2303 5051
e8a0464c
DM		 5052static void bond_set_lockdep_class_one(struct net_device *dev,
5053 struct netdev_queue *txq,
5054 void *_unused)
c773e847
DM		 5055{
5056 lockdep_set_class(&txq->_xmit_lock,
5057 &bonding_netdev_xmit_lock_key);
5058}
5059
5060static void bond_set_lockdep_class(struct net_device *dev)
5061{
cf508b12
DM		 5062 lockdep_set_class(&dev->addr_list_lock,
5063 &bonding_netdev_addr_lock_key);
e8a0464c 5064 netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL);
c773e847
DM		 5065}
5066
181470fc
SH		 5067/*
5068 * Called from registration process
5069 */
5070static int bond_init(struct net_device *bond_dev)
5071{
5072 struct bonding *bond = netdev_priv(bond_dev);
ec87fd3b 5073 struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id);
181470fc
SH		 5074
5075 pr_debug("Begin bond_init for %s\n", bond_dev->name);
5076
5077 bond->wq = create_singlethread_workqueue(bond_dev->name);
5078 if (!bond->wq)
5079 return -ENOMEM;
5080
5081 bond_set_lockdep_class(bond_dev);
5082
5083 netif_carrier_off(bond_dev);
5084
5085 bond_create_proc_entry(bond);
ec87fd3b 5086 list_add_tail(&bond->bond_list, &bn->dev_list);
181470fc 5087
6151b3d4 5088 bond_prepare_sysfs_group(bond);
22bedad3
JP		 5089
5090 __hw_addr_init(&bond->mc_list);
181470fc
SH		 5091 return 0;
5092}
5093
88ead977
EB		 5094static int bond_validate(struct nlattr *tb[], struct nlattr *data[])
5095{
5096 if (tb[IFLA_ADDRESS]) {
5097 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
5098 return -EINVAL;
5099 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
5100 return -EADDRNOTAVAIL;
5101 }
5102 return 0;
5103}
5104
5105static struct rtnl_link_ops bond_link_ops __read_mostly = {
5106 .kind = "bond",
6639104b 5107 .priv_size = sizeof(struct bonding),
88ead977
EB		 5108 .setup = bond_setup,
5109 .validate = bond_validate,
5110};
5111
dfe60397 5112/* Create a new bond based on the specified name and bonding parameters.
e4b91c48 5113 * If name is NULL, obtain a suitable "bond%d" name for us.
dfe60397
MW		 5114 * Caller must NOT hold rtnl_lock; we need to release it here before we
5115 * set up our sysfs entries.
5116 */
ec87fd3b 5117int bond_create(struct net *net, const char *name)
dfe60397
MW		 5118{
5119 struct net_device *bond_dev;
5120 int res;
5121
5122 rtnl_lock();
027ea041 5123
bb1d9123
AG		 5124 bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "",
5125 bond_setup, tx_queues);
dfe60397 5126 if (!bond_dev) {
a4aee5c8 5127 pr_err("%s: eek! can't alloc netdev!\n", name);
9e2e61fb
AW		 5128 rtnl_unlock();
5129 return -ENOMEM;
dfe60397
MW		 5130 }
5131
ec87fd3b 5132 dev_net_set(bond_dev, net);
88ead977
EB		 5133 bond_dev->rtnl_link_ops = &bond_link_ops;
5134
e4b91c48
JV		 5135 if (!name) {
5136 res = dev_alloc_name(bond_dev, "bond%d");
5137 if (res < 0)
9e2e61fb 5138 goto out;
e4b91c48
JV		 5139 }
5140
dfe60397 5141 res = register_netdevice(bond_dev);
0daa2303 5142
30c15ba9 5143out:
7e083840 5144 rtnl_unlock();
9e2e61fb
AW		 5145 if (res < 0)
5146 bond_destructor(bond_dev);
30c15ba9 5147 return res;
dfe60397
MW		 5148}
5149
2c8c1e72 5150static int __net_init bond_net_init(struct net *net)
ec87fd3b 5151{
15449745 5152 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 5153
5154 bn->net = net;
5155 INIT_LIST_HEAD(&bn->dev_list);
5156
ec87fd3b 5157 bond_create_proc_dir(bn);
15449745
EB		 5158
5159 return 0;
ec87fd3b
EB		 5160}
5161
2c8c1e72 5162static void __net_exit bond_net_exit(struct net *net)
ec87fd3b 5163{
15449745 5164 struct bond_net *bn = net_generic(net, bond_net_id);
ec87fd3b
EB		 5165
5166 bond_destroy_proc_dir(bn);
ec87fd3b
EB		 5167}
5168
5169static struct pernet_operations bond_net_ops = {
5170 .init = bond_net_init,
5171 .exit = bond_net_exit,
15449745
EB		 5172 .id = &bond_net_id,
5173 .size = sizeof(struct bond_net),
ec87fd3b
EB		 5174};
5175
1da177e4
LT		 5176static int __init bonding_init(void)
5177{
1da177e4
LT		 5178 int i;
5179 int res;
5180
3d632c3f 5181 pr_info("%s", version);
1da177e4 5182
dfe60397 5183 res = bond_check_params(&bonding_defaults);
3d632c3f 5184 if (res)
dfe60397 5185 goto out;
1da177e4 5186
15449745 5187 res = register_pernet_subsys(&bond_net_ops);
ec87fd3b
EB		 5188 if (res)
5189 goto out;
027ea041 5190
88ead977
EB		 5191 res = rtnl_link_register(&bond_link_ops);
5192 if (res)
6639104b 5193 goto err_link;
88ead977 5194
1da177e4 5195 for (i = 0; i < max_bonds; i++) {
ec87fd3b 5196 res = bond_create(&init_net, NULL);
dfe60397
MW		 5197 if (res)
5198 goto err;
1da177e4
LT		 5199 }
5200
b76cdba9
MW		 5201 res = bond_create_sysfs();
5202 if (res)
5203 goto err;
5204
1da177e4 5205 register_netdevice_notifier(&bond_netdev_notifier);
c3ade5ca 5206 register_inetaddr_notifier(&bond_inetaddr_notifier);
305d552a 5207 bond_register_ipv6_notifier();
dfe60397 5208out:
1da177e4 5209 return res;
88ead977
EB		 5210err:
5211 rtnl_link_unregister(&bond_link_ops);
6639104b 5212err_link:
15449745 5213 unregister_pernet_subsys(&bond_net_ops);
88ead977 5214 goto out;
dfe60397 5215
1da177e4
LT		 5216}
5217
5218static void __exit bonding_exit(void)
5219{
5220 unregister_netdevice_notifier(&bond_netdev_notifier);
c3ade5ca 5221 unregister_inetaddr_notifier(&bond_inetaddr_notifier);
305d552a 5222 bond_unregister_ipv6_notifier();
1da177e4 5223
ae68c398
PE		 5224 bond_destroy_sysfs();
5225
88ead977 5226 rtnl_link_unregister(&bond_link_ops);
15449745 5227 unregister_pernet_subsys(&bond_net_ops);
1da177e4
LT		 5228}
5229
5230module_init(bonding_init);
5231module_exit(bonding_exit);
5232MODULE_LICENSE("GPL");
5233MODULE_VERSION(DRV_VERSION);
5234MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
5235MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
88ead977 5236MODULE_ALIAS_RTNL_LINK("bond");
Clone of davem's net-next-2.6 tree