Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * originally based on the dummy device. | |
3 | * | |
4 | * Copyright 1999, Thomas Davis, tadavis@lbl.gov. | |
5 | * Licensed under the GPL. Based on dummy.c, and eql.c devices. | |
6 | * | |
7 | * bonding.c: an Ethernet Bonding driver | |
8 | * | |
9 | * This is useful to talk to a Cisco EtherChannel compatible equipment: | |
10 | * Cisco 5500 | |
11 | * Sun Trunking (Solaris) | |
12 | * Alteon AceDirector Trunks | |
13 | * Linux Bonding | |
14 | * and probably many L2 switches ... | |
15 | * | |
16 | * How it works: | |
17 | * ifconfig bond0 ipaddress netmask up | |
18 | * will setup a network device, with an ip address. No mac address | |
19 | * will be assigned at this time. The hw mac address will come from | |
20 | * the first slave bonded to the channel. All slaves will then use | |
21 | * this hw mac address. | |
22 | * | |
23 | * ifconfig bond0 down | |
24 | * will release all slaves, marking them as down. | |
25 | * | |
26 | * ifenslave bond0 eth0 | |
27 | * will attach eth0 to bond0 as a slave. eth0 hw mac address will either | |
28 | * a: be used as initial mac address | |
29 | * b: if a hw mac address already is there, eth0's hw mac address | |
30 | * will then be set from bond0. | |
31 | * | |
1da177e4 LT |
32 | */ |
33 | ||
a4aee5c8 JP |
34 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
35 | ||
1da177e4 LT |
36 | #include <linux/kernel.h> |
37 | #include <linux/module.h> | |
1da177e4 LT |
38 | #include <linux/types.h> |
39 | #include <linux/fcntl.h> | |
40 | #include <linux/interrupt.h> | |
41 | #include <linux/ptrace.h> | |
42 | #include <linux/ioport.h> | |
43 | #include <linux/in.h> | |
169a3e66 | 44 | #include <net/ip.h> |
1da177e4 | 45 | #include <linux/ip.h> |
169a3e66 JV |
46 | #include <linux/tcp.h> |
47 | #include <linux/udp.h> | |
1da177e4 LT |
48 | #include <linux/slab.h> |
49 | #include <linux/string.h> | |
50 | #include <linux/init.h> | |
51 | #include <linux/timer.h> | |
52 | #include <linux/socket.h> | |
53 | #include <linux/ctype.h> | |
54 | #include <linux/inet.h> | |
55 | #include <linux/bitops.h> | |
3d632c3f | 56 | #include <linux/io.h> |
1da177e4 | 57 | #include <asm/dma.h> |
3d632c3f | 58 | #include <linux/uaccess.h> |
1da177e4 LT |
59 | #include <linux/errno.h> |
60 | #include <linux/netdevice.h> | |
61 | #include <linux/inetdevice.h> | |
a816c7c7 | 62 | #include <linux/igmp.h> |
1da177e4 LT |
63 | #include <linux/etherdevice.h> |
64 | #include <linux/skbuff.h> | |
65 | #include <net/sock.h> | |
66 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
67 | #include <linux/smp.h> |
68 | #include <linux/if_ether.h> | |
69 | #include <net/arp.h> | |
70 | #include <linux/mii.h> | |
71 | #include <linux/ethtool.h> | |
72 | #include <linux/if_vlan.h> | |
73 | #include <linux/if_bonding.h> | |
b63bb739 | 74 | #include <linux/jiffies.h> |
e843fa50 | 75 | #include <linux/preempt.h> |
c3ade5ca | 76 | #include <net/route.h> |
457c4cbc | 77 | #include <net/net_namespace.h> |
ec87fd3b | 78 | #include <net/netns/generic.h> |
5ee31c68 | 79 | #include <net/pkt_sched.h> |
278b2083 | 80 | #include <linux/rculist.h> |
1da177e4 LT |
81 | #include "bonding.h" |
82 | #include "bond_3ad.h" | |
83 | #include "bond_alb.h" | |
84 | ||
85 | /*---------------------------- Module parameters ----------------------------*/ | |
86 | ||
87 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
88 | #define BOND_LINK_MON_INTERV 0 | |
89 | #define BOND_LINK_ARP_INTERV 0 | |
90 | ||
91 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 92 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 93 | static int num_peer_notif = 1; |
1da177e4 | 94 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
95 | static int updelay; |
96 | static int downdelay; | |
1da177e4 | 97 | static int use_carrier = 1; |
3d632c3f SH |
98 | static char *mode; |
99 | static char *primary; | |
a549952a | 100 | static char *primary_reselect; |
3d632c3f | 101 | static char *lacp_rate; |
655f8919 | 102 | static int min_links; |
3d632c3f SH |
103 | static char *ad_select; |
104 | static char *xmit_hash_policy; | |
1da177e4 | 105 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
106 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
107 | static char *arp_validate; | |
8599b52e | 108 | static char *arp_all_targets; |
3d632c3f | 109 | static char *fail_over_mac; |
b07ea07b | 110 | static int all_slaves_active; |
d2991f75 | 111 | static struct bond_params bonding_defaults; |
c2952c31 | 112 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
1da177e4 LT |
113 | |
114 | module_param(max_bonds, int, 0); | |
115 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
116 | module_param(tx_queues, int, 0); |
117 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 118 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
119 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
120 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 121 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
122 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
123 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
124 | module_param(miimon, int, 0); |
125 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
126 | module_param(updelay, int, 0); | |
127 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
128 | module_param(downdelay, int, 0); | |
2ac47660 MW |
129 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
130 | "in milliseconds"); | |
1da177e4 | 131 | module_param(use_carrier, int, 0); |
2ac47660 MW |
132 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
133 | "0 for off, 1 for on (default)"); | |
1da177e4 | 134 | module_param(mode, charp, 0); |
90e62474 | 135 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
136 | "1 for active-backup, 2 for balance-xor, " |
137 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
138 | "6 for balance-alb"); | |
1da177e4 LT |
139 | module_param(primary, charp, 0); |
140 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
141 | module_param(primary_reselect, charp, 0); |
142 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
143 | "once it comes up; " | |
144 | "0 for always (default), " | |
145 | "1 for only if speed of primary is " | |
146 | "better, " | |
147 | "2 for only on active slave " | |
148 | "failure"); | |
1da177e4 | 149 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
150 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
151 | "0 for slow, 1 for fast"); | |
fd989c83 | 152 | module_param(ad_select, charp, 0); |
90e62474 AG |
153 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
154 | "0 for stable (default), 1 for bandwidth, " | |
155 | "2 for count"); | |
655f8919 | 156 | module_param(min_links, int, 0); |
157 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
158 | ||
169a3e66 | 159 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
160 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
161 | "0 for layer 2 (default), 1 for layer 3+4, " | |
162 | "2 for layer 2+3"); | |
1da177e4 LT |
163 | module_param(arp_interval, int, 0); |
164 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
165 | module_param_array(arp_ip_target, charp, NULL, 0); | |
166 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 167 | module_param(arp_validate, charp, 0); |
90e62474 AG |
168 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
169 | "0 for none (default), 1 for active, " | |
170 | "2 for backup, 3 for all"); | |
8599b52e VF |
171 | module_param(arp_all_targets, charp, 0); |
172 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 173 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
174 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
175 | "the same MAC; 0 for none (default), " | |
176 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
177 | module_param(all_slaves_active, int, 0); |
178 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 179 | "by setting active flag for all slaves; " |
ebd8e497 | 180 | "0 for never (default), 1 for always."); |
c2952c31 | 181 | module_param(resend_igmp, int, 0); |
90e62474 AG |
182 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
183 | "link failure"); | |
1da177e4 LT |
184 | |
185 | /*----------------------------- Global variables ----------------------------*/ | |
186 | ||
e843fa50 | 187 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 188 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
189 | #endif |
190 | ||
f99189b1 | 191 | int bond_net_id __read_mostly; |
1da177e4 | 192 | |
3d632c3f SH |
193 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
194 | static int arp_ip_count; | |
1da177e4 | 195 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
196 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
197 | static int lacp_fast; | |
1da177e4 | 198 | |
e97fd7c6 | 199 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
200 | { "slow", AD_LACP_SLOW}, |
201 | { "fast", AD_LACP_FAST}, | |
202 | { NULL, -1}, | |
203 | }; | |
204 | ||
e97fd7c6 | 205 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
206 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
207 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
208 | { "balance-xor", BOND_MODE_XOR}, | |
209 | { "broadcast", BOND_MODE_BROADCAST}, | |
210 | { "802.3ad", BOND_MODE_8023AD}, | |
211 | { "balance-tlb", BOND_MODE_TLB}, | |
212 | { "balance-alb", BOND_MODE_ALB}, | |
213 | { NULL, -1}, | |
214 | }; | |
215 | ||
e97fd7c6 | 216 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
217 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
218 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 219 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
220 | { NULL, -1}, |
221 | }; | |
222 | ||
8599b52e VF |
223 | const struct bond_parm_tbl arp_all_targets_tbl[] = { |
224 | { "any", BOND_ARP_TARGETS_ANY}, | |
225 | { "all", BOND_ARP_TARGETS_ALL}, | |
226 | { NULL, -1}, | |
227 | }; | |
228 | ||
e97fd7c6 | 229 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
230 | { "none", BOND_ARP_VALIDATE_NONE}, |
231 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
232 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
233 | { "all", BOND_ARP_VALIDATE_ALL}, | |
234 | { NULL, -1}, | |
235 | }; | |
236 | ||
e97fd7c6 | 237 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
238 | { "none", BOND_FOM_NONE}, |
239 | { "active", BOND_FOM_ACTIVE}, | |
240 | { "follow", BOND_FOM_FOLLOW}, | |
241 | { NULL, -1}, | |
242 | }; | |
243 | ||
a549952a JP |
244 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
245 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
246 | { "better", BOND_PRI_RESELECT_BETTER}, | |
247 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
248 | { NULL, -1}, | |
249 | }; | |
250 | ||
fd989c83 JV |
251 | struct bond_parm_tbl ad_select_tbl[] = { |
252 | { "stable", BOND_AD_STABLE}, | |
253 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
254 | { "count", BOND_AD_COUNT}, | |
255 | { NULL, -1}, | |
256 | }; | |
257 | ||
1da177e4 LT |
258 | /*-------------------------- Forward declarations ---------------------------*/ |
259 | ||
181470fc | 260 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 261 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
262 | |
263 | /*---------------------------- General routines -----------------------------*/ | |
264 | ||
bd33acc3 | 265 | const char *bond_mode_name(int mode) |
1da177e4 | 266 | { |
77afc92b HE |
267 | static const char *names[] = { |
268 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
269 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
270 | [BOND_MODE_XOR] = "load balancing (xor)", | |
271 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 272 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
273 | [BOND_MODE_TLB] = "transmit load balancing", |
274 | [BOND_MODE_ALB] = "adaptive load balancing", | |
275 | }; | |
276 | ||
f5280948 | 277 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 278 | return "unknown"; |
77afc92b HE |
279 | |
280 | return names[mode]; | |
1da177e4 LT |
281 | } |
282 | ||
283 | /*---------------------------------- VLAN -----------------------------------*/ | |
284 | ||
1da177e4 LT |
285 | /** |
286 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 287 | * |
1da177e4 LT |
288 | * @bond: bond device that got this skb for tx. |
289 | * @skb: hw accel VLAN tagged skb to transmit | |
290 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 291 | */ |
3d632c3f SH |
292 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
293 | struct net_device *slave_dev) | |
1da177e4 | 294 | { |
8387451e | 295 | skb->dev = slave_dev; |
374eeb5a | 296 | |
5ee31c68 | 297 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
298 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
299 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 300 | |
e15c3c22 | 301 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 302 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 303 | else |
f6dc31a8 | 304 | dev_queue_xmit(skb); |
1da177e4 LT |
305 | |
306 | return 0; | |
307 | } | |
308 | ||
309 | /* | |
cc0e4070 JP |
310 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
311 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
312 | * a. This operation is performed in IOCTL context, |
313 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
314 | * c. Holding a lock with BH disabled while directly calling a base driver | |
315 | * entry point is generally a BAD idea. | |
3d632c3f | 316 | * |
1da177e4 LT |
317 | * The design of synchronization/protection for this operation in the 8021q |
318 | * module is good for one or more VLAN devices over a single physical device | |
319 | * and cannot be extended for a teaming solution like bonding, so there is a | |
320 | * potential race condition here where a net device from the vlan group might | |
321 | * be referenced (either by a base driver or the 8021q code) while it is being | |
322 | * removed from the system. However, it turns out we're not making matters | |
323 | * worse, and if it works for regular VLAN usage it will work here too. | |
324 | */ | |
325 | ||
1da177e4 LT |
326 | /** |
327 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
328 | * @bond_dev: bonding net device that got called | |
329 | * @vid: vlan id being added | |
330 | */ | |
80d5c368 PM |
331 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
332 | __be16 proto, u16 vid) | |
1da177e4 | 333 | { |
454d7c9b | 334 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 335 | struct slave *slave, *rollback_slave; |
dec1e90e | 336 | int res; |
1da177e4 | 337 | |
dec1e90e | 338 | bond_for_each_slave(bond, slave) { |
80d5c368 | 339 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
340 | if (res) |
341 | goto unwind; | |
1da177e4 LT |
342 | } |
343 | ||
8e586137 | 344 | return 0; |
87002b03 JP |
345 | |
346 | unwind: | |
81f23b13 VF |
347 | /* unwind to the slave that failed */ |
348 | bond_for_each_slave(bond, rollback_slave) { | |
349 | if (rollback_slave == slave) | |
350 | break; | |
351 | ||
352 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
353 | } | |
87002b03 JP |
354 | |
355 | return res; | |
1da177e4 LT |
356 | } |
357 | ||
358 | /** | |
359 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
360 | * @bond_dev: bonding net device that got called | |
361 | * @vid: vlan id being removed | |
362 | */ | |
80d5c368 PM |
363 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
364 | __be16 proto, u16 vid) | |
1da177e4 | 365 | { |
454d7c9b | 366 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 367 | struct slave *slave; |
1da177e4 | 368 | |
dec1e90e | 369 | bond_for_each_slave(bond, slave) |
80d5c368 | 370 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 371 | |
e868b0c9 VF |
372 | if (bond_is_lb(bond)) |
373 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
374 | |
375 | return 0; | |
1da177e4 LT |
376 | } |
377 | ||
1da177e4 LT |
378 | /*------------------------------- Link status -------------------------------*/ |
379 | ||
ff59c456 JV |
380 | /* |
381 | * Set the carrier state for the master according to the state of its | |
382 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
383 | * do special 802.3ad magic. | |
384 | * | |
385 | * Returns zero if carrier state does not change, nonzero if it does. | |
386 | */ | |
387 | static int bond_set_carrier(struct bonding *bond) | |
388 | { | |
389 | struct slave *slave; | |
ff59c456 | 390 | |
dec1e90e | 391 | if (list_empty(&bond->slave_list)) |
ff59c456 JV |
392 | goto down; |
393 | ||
394 | if (bond->params.mode == BOND_MODE_8023AD) | |
395 | return bond_3ad_set_carrier(bond); | |
396 | ||
dec1e90e | 397 | bond_for_each_slave(bond, slave) { |
ff59c456 JV |
398 | if (slave->link == BOND_LINK_UP) { |
399 | if (!netif_carrier_ok(bond->dev)) { | |
400 | netif_carrier_on(bond->dev); | |
401 | return 1; | |
402 | } | |
403 | return 0; | |
404 | } | |
405 | } | |
406 | ||
407 | down: | |
408 | if (netif_carrier_ok(bond->dev)) { | |
409 | netif_carrier_off(bond->dev); | |
410 | return 1; | |
411 | } | |
412 | return 0; | |
413 | } | |
414 | ||
1da177e4 LT |
415 | /* |
416 | * Get link speed and duplex from the slave's base driver | |
417 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 418 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 419 | * and return. |
1da177e4 | 420 | */ |
13b95fb7 | 421 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
422 | { |
423 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 424 | struct ethtool_cmd ecmd; |
5d30530e | 425 | u32 slave_speed; |
61a44b9c | 426 | int res; |
1da177e4 | 427 | |
589665f5 DC |
428 | slave->speed = SPEED_UNKNOWN; |
429 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 430 | |
4bc71cb9 | 431 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 432 | if (res < 0) |
13b95fb7 | 433 | return; |
1da177e4 | 434 | |
4bc71cb9 | 435 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 436 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 437 | return; |
1da177e4 | 438 | |
4bc71cb9 | 439 | switch (ecmd.duplex) { |
1da177e4 LT |
440 | case DUPLEX_FULL: |
441 | case DUPLEX_HALF: | |
442 | break; | |
443 | default: | |
13b95fb7 | 444 | return; |
1da177e4 LT |
445 | } |
446 | ||
5d30530e | 447 | slave->speed = slave_speed; |
4bc71cb9 | 448 | slave->duplex = ecmd.duplex; |
1da177e4 | 449 | |
13b95fb7 | 450 | return; |
1da177e4 LT |
451 | } |
452 | ||
453 | /* | |
454 | * if <dev> supports MII link status reporting, check its link status. | |
455 | * | |
456 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 457 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
458 | * |
459 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
460 | * can't tell and just pretend it is), or 0, meaning that the link is | |
461 | * down. | |
462 | * | |
463 | * If reporting is non-zero, instead of faking link up, return -1 if | |
464 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
465 | * support them). If use_carrier is set, return whatever it says. | |
466 | * It'd be nice if there was a good way to tell if a driver supports | |
467 | * netif_carrier, but there really isn't. | |
468 | */ | |
3d632c3f SH |
469 | static int bond_check_dev_link(struct bonding *bond, |
470 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 471 | { |
eb7cc59a | 472 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 473 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
474 | struct ifreq ifr; |
475 | struct mii_ioctl_data *mii; | |
1da177e4 | 476 | |
6c988853 PG |
477 | if (!reporting && !netif_running(slave_dev)) |
478 | return 0; | |
479 | ||
eb7cc59a | 480 | if (bond->params.use_carrier) |
1da177e4 | 481 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 482 | |
29112f4e | 483 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
484 | if (slave_dev->ethtool_ops->get_link) |
485 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
486 | BMSR_LSTATUS : 0; | |
29112f4e | 487 | |
3d632c3f | 488 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 489 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
490 | if (ioctl) { |
491 | /* TODO: set pointer to correct ioctl on a per team member */ | |
492 | /* bases to make this more efficient. that is, once */ | |
493 | /* we determine the correct ioctl, we will always */ | |
494 | /* call it and not the others for that team */ | |
495 | /* member. */ | |
496 | ||
497 | /* | |
498 | * We cannot assume that SIOCGMIIPHY will also read a | |
499 | * register; not all network drivers (e.g., e100) | |
500 | * support that. | |
501 | */ | |
502 | ||
503 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
504 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
505 | mii = if_mii(&ifr); | |
506 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
507 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
508 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
509 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
510 | } |
511 | } | |
512 | ||
1da177e4 LT |
513 | /* |
514 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 515 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
516 | * cannot report link status). If not reporting, pretend |
517 | * we're ok. | |
518 | */ | |
3d632c3f | 519 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
520 | } |
521 | ||
522 | /*----------------------------- Multicast list ------------------------------*/ | |
523 | ||
1da177e4 LT |
524 | /* |
525 | * Push the promiscuity flag down to appropriate slaves | |
526 | */ | |
7e1a1ac1 | 527 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 528 | { |
7e1a1ac1 | 529 | int err = 0; |
1da177e4 LT |
530 | if (USES_PRIMARY(bond->params.mode)) { |
531 | /* write lock already acquired */ | |
532 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
533 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
534 | inc); | |
1da177e4 LT |
535 | } |
536 | } else { | |
537 | struct slave *slave; | |
dec1e90e | 538 | |
539 | bond_for_each_slave(bond, slave) { | |
7e1a1ac1 WC |
540 | err = dev_set_promiscuity(slave->dev, inc); |
541 | if (err) | |
542 | return err; | |
1da177e4 LT |
543 | } |
544 | } | |
7e1a1ac1 | 545 | return err; |
1da177e4 LT |
546 | } |
547 | ||
548 | /* | |
549 | * Push the allmulti flag down to all slaves | |
550 | */ | |
7e1a1ac1 | 551 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 552 | { |
7e1a1ac1 | 553 | int err = 0; |
1da177e4 LT |
554 | if (USES_PRIMARY(bond->params.mode)) { |
555 | /* write lock already acquired */ | |
556 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
557 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
558 | inc); | |
1da177e4 LT |
559 | } |
560 | } else { | |
561 | struct slave *slave; | |
dec1e90e | 562 | |
563 | bond_for_each_slave(bond, slave) { | |
7e1a1ac1 WC |
564 | err = dev_set_allmulti(slave->dev, inc); |
565 | if (err) | |
566 | return err; | |
1da177e4 LT |
567 | } |
568 | } | |
7e1a1ac1 | 569 | return err; |
1da177e4 LT |
570 | } |
571 | ||
5a37e8ca FL |
572 | /* |
573 | * Retrieve the list of registered multicast addresses for the bonding | |
574 | * device and retransmit an IGMP JOIN request to the current active | |
575 | * slave. | |
576 | */ | |
577 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
578 | { | |
4aa5dee4 | 579 | if (!rtnl_trylock()) { |
4beac029 | 580 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 581 | return; |
5a37e8ca | 582 | } |
4aa5dee4 JP |
583 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
584 | rtnl_unlock(); | |
5a37e8ca | 585 | |
4f5474e7 NA |
586 | /* We use curr_slave_lock to protect against concurrent access to |
587 | * igmp_retrans from multiple running instances of this function and | |
588 | * bond_change_active_slave | |
589 | */ | |
590 | write_lock_bh(&bond->curr_slave_lock); | |
591 | if (bond->igmp_retrans > 1) { | |
592 | bond->igmp_retrans--; | |
c2952c31 | 593 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 NA |
594 | } |
595 | write_unlock_bh(&bond->curr_slave_lock); | |
5a37e8ca FL |
596 | } |
597 | ||
379b7383 | 598 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
599 | { |
600 | struct bonding *bond = container_of(work, struct bonding, | |
94265cf5 | 601 | mcast_work.work); |
ad999eee | 602 | |
5a37e8ca FL |
603 | bond_resend_igmp_join_requests(bond); |
604 | } | |
605 | ||
303d1cbf | 606 | /* Flush bond's hardware addresses from slave |
1da177e4 | 607 | */ |
303d1cbf | 608 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 609 | struct net_device *slave_dev) |
1da177e4 | 610 | { |
454d7c9b | 611 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 612 | |
303d1cbf JV |
613 | dev_uc_unsync(slave_dev, bond_dev); |
614 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 LT |
615 | |
616 | if (bond->params.mode == BOND_MODE_8023AD) { | |
617 | /* del lacpdu mc addr from mc list */ | |
618 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
619 | ||
22bedad3 | 620 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
621 | } |
622 | } | |
623 | ||
624 | /*--------------------------- Active slave change ---------------------------*/ | |
625 | ||
303d1cbf JV |
626 | /* Update the hardware address list and promisc/allmulti for the new and |
627 | * old active slaves (if any). Modes that are !USES_PRIMARY keep all | |
628 | * slaves up date at all times; only the USES_PRIMARY modes need to call | |
629 | * this function to swap these settings during a failover. | |
1da177e4 | 630 | */ |
303d1cbf JV |
631 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
632 | struct slave *old_active) | |
1da177e4 | 633 | { |
7864a1ad VF |
634 | ASSERT_RTNL(); |
635 | ||
1da177e4 | 636 | if (old_active) { |
3d632c3f | 637 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 638 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 639 | |
3d632c3f | 640 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 641 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 642 | |
303d1cbf | 643 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
644 | } |
645 | ||
646 | if (new_active) { | |
7e1a1ac1 | 647 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 648 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 649 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 650 | |
3d632c3f | 651 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 652 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 653 | |
d632ce98 | 654 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
655 | dev_uc_sync(new_active->dev, bond->dev); |
656 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 657 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
658 | } |
659 | } | |
660 | ||
ae0d6750 | 661 | /** |
662 | * bond_set_dev_addr - clone slave's address to bond | |
663 | * @bond_dev: bond net device | |
664 | * @slave_dev: slave net device | |
665 | * | |
666 | * Should be called with RTNL held. | |
667 | */ | |
668 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
669 | struct net_device *slave_dev) | |
670 | { | |
008aebde NA |
671 | pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
672 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 673 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
674 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
675 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
676 | } | |
677 | ||
3915c1e8 JV |
678 | /* |
679 | * bond_do_fail_over_mac | |
680 | * | |
681 | * Perform special MAC address swapping for fail_over_mac settings | |
682 | * | |
683 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
684 | */ | |
685 | static void bond_do_fail_over_mac(struct bonding *bond, | |
686 | struct slave *new_active, | |
687 | struct slave *old_active) | |
1f78d9f9 HE |
688 | __releases(&bond->curr_slave_lock) |
689 | __releases(&bond->lock) | |
690 | __acquires(&bond->lock) | |
691 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
692 | { |
693 | u8 tmp_mac[ETH_ALEN]; | |
694 | struct sockaddr saddr; | |
695 | int rv; | |
696 | ||
697 | switch (bond->params.fail_over_mac) { | |
698 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 699 | if (new_active) { |
7d26bb10 WP |
700 | write_unlock_bh(&bond->curr_slave_lock); |
701 | read_unlock(&bond->lock); | |
ae0d6750 | 702 | bond_set_dev_addr(bond->dev, new_active->dev); |
7d26bb10 WP |
703 | read_lock(&bond->lock); |
704 | write_lock_bh(&bond->curr_slave_lock); | |
705 | } | |
3915c1e8 JV |
706 | break; |
707 | case BOND_FOM_FOLLOW: | |
708 | /* | |
709 | * if new_active && old_active, swap them | |
710 | * if just old_active, do nothing (going to no active slave) | |
711 | * if just new_active, set new_active to bond's MAC | |
712 | */ | |
713 | if (!new_active) | |
714 | return; | |
715 | ||
716 | write_unlock_bh(&bond->curr_slave_lock); | |
717 | read_unlock(&bond->lock); | |
718 | ||
719 | if (old_active) { | |
720 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
721 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
722 | ETH_ALEN); | |
723 | saddr.sa_family = new_active->dev->type; | |
724 | } else { | |
725 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
726 | saddr.sa_family = bond->dev->type; | |
727 | } | |
728 | ||
729 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
730 | if (rv) { | |
a4aee5c8 | 731 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
732 | bond->dev->name, -rv, new_active->dev->name); |
733 | goto out; | |
734 | } | |
735 | ||
736 | if (!old_active) | |
737 | goto out; | |
738 | ||
739 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
740 | saddr.sa_family = old_active->dev->type; | |
741 | ||
742 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
743 | if (rv) | |
a4aee5c8 | 744 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
745 | bond->dev->name, -rv, new_active->dev->name); |
746 | out: | |
747 | read_lock(&bond->lock); | |
748 | write_lock_bh(&bond->curr_slave_lock); | |
749 | break; | |
750 | default: | |
a4aee5c8 | 751 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
752 | bond->dev->name, bond->params.fail_over_mac); |
753 | break; | |
754 | } | |
755 | ||
756 | } | |
757 | ||
a549952a JP |
758 | static bool bond_should_change_active(struct bonding *bond) |
759 | { | |
760 | struct slave *prim = bond->primary_slave; | |
761 | struct slave *curr = bond->curr_active_slave; | |
762 | ||
763 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
764 | return true; | |
765 | if (bond->force_primary) { | |
766 | bond->force_primary = false; | |
767 | return true; | |
768 | } | |
769 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
770 | (prim->speed < curr->speed || | |
771 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
772 | return false; | |
773 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
774 | return false; | |
775 | return true; | |
776 | } | |
3915c1e8 | 777 | |
1da177e4 LT |
778 | /** |
779 | * find_best_interface - select the best available slave to be the active one | |
780 | * @bond: our bonding struct | |
781 | * | |
782 | * Warning: Caller must hold curr_slave_lock for writing. | |
783 | */ | |
784 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
785 | { | |
786 | struct slave *new_active, *old_active; | |
787 | struct slave *bestslave = NULL; | |
788 | int mintime = bond->params.updelay; | |
789 | int i; | |
790 | ||
49b4ad92 | 791 | new_active = bond->curr_active_slave; |
1da177e4 LT |
792 | |
793 | if (!new_active) { /* there were no active slaves left */ | |
dec1e90e | 794 | new_active = bond_first_slave(bond); |
795 | if (!new_active) | |
1da177e4 | 796 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
797 | } |
798 | ||
1da177e4 | 799 | if ((bond->primary_slave) && |
a549952a JP |
800 | bond->primary_slave->link == BOND_LINK_UP && |
801 | bond_should_change_active(bond)) { | |
1da177e4 LT |
802 | new_active = bond->primary_slave; |
803 | } | |
804 | ||
805 | /* remember where to stop iterating over the slaves */ | |
806 | old_active = new_active; | |
807 | ||
808 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
809 | if (new_active->link == BOND_LINK_UP) { |
810 | return new_active; | |
811 | } else if (new_active->link == BOND_LINK_BACK && | |
812 | IS_UP(new_active->dev)) { | |
813 | /* link up, but waiting for stabilization */ | |
814 | if (new_active->delay < mintime) { | |
815 | mintime = new_active->delay; | |
816 | bestslave = new_active; | |
1da177e4 LT |
817 | } |
818 | } | |
819 | } | |
820 | ||
821 | return bestslave; | |
822 | } | |
823 | ||
ad246c99 BH |
824 | static bool bond_should_notify_peers(struct bonding *bond) |
825 | { | |
826 | struct slave *slave = bond->curr_active_slave; | |
827 | ||
828 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
829 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
830 | ||
831 | if (!slave || !bond->send_peer_notif || | |
832 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
833 | return false; | |
834 | ||
ad246c99 BH |
835 | return true; |
836 | } | |
837 | ||
1da177e4 LT |
838 | /** |
839 | * change_active_interface - change the active slave into the specified one | |
840 | * @bond: our bonding struct | |
841 | * @new: the new slave to make the active one | |
842 | * | |
843 | * Set the new slave to the bond's settings and unset them on the old | |
844 | * curr_active_slave. | |
845 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
846 | * | |
847 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
848 | * because it is apparently the best available slave we have, even though its | |
849 | * updelay hasn't timed out yet. | |
850 | * | |
3915c1e8 JV |
851 | * If new_active is not NULL, caller must hold bond->lock for read and |
852 | * curr_slave_lock for write_bh. | |
1da177e4 | 853 | */ |
a77b5325 | 854 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
855 | { |
856 | struct slave *old_active = bond->curr_active_slave; | |
857 | ||
3d632c3f | 858 | if (old_active == new_active) |
1da177e4 | 859 | return; |
1da177e4 LT |
860 | |
861 | if (new_active) { | |
b2220cad JV |
862 | new_active->jiffies = jiffies; |
863 | ||
1da177e4 LT |
864 | if (new_active->link == BOND_LINK_BACK) { |
865 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
866 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
867 | bond->dev->name, new_active->dev->name, | |
868 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
869 | } |
870 | ||
871 | new_active->delay = 0; | |
872 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 873 | |
3d632c3f | 874 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 875 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 876 | |
58402054 | 877 | if (bond_is_lb(bond)) |
1da177e4 | 878 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
879 | } else { |
880 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
881 | pr_info("%s: making interface %s the new active one.\n", |
882 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
883 | } |
884 | } | |
885 | } | |
886 | ||
3d632c3f | 887 | if (USES_PRIMARY(bond->params.mode)) |
303d1cbf | 888 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 889 | |
58402054 | 890 | if (bond_is_lb(bond)) { |
1da177e4 | 891 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
892 | if (old_active) |
893 | bond_set_slave_inactive_flags(old_active); | |
894 | if (new_active) | |
895 | bond_set_slave_active_flags(new_active); | |
1da177e4 | 896 | } else { |
278b2083 | 897 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 898 | } |
c3ade5ca JV |
899 | |
900 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 901 | if (old_active) |
c3ade5ca | 902 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
903 | |
904 | if (new_active) { | |
ad246c99 BH |
905 | bool should_notify_peers = false; |
906 | ||
c3ade5ca | 907 | bond_set_slave_active_flags(new_active); |
2ab82852 | 908 | |
709f8a45 OG |
909 | if (bond->params.fail_over_mac) |
910 | bond_do_fail_over_mac(bond, new_active, | |
911 | old_active); | |
3915c1e8 | 912 | |
ad246c99 BH |
913 | if (netif_running(bond->dev)) { |
914 | bond->send_peer_notif = | |
915 | bond->params.num_peer_notif; | |
916 | should_notify_peers = | |
917 | bond_should_notify_peers(bond); | |
918 | } | |
919 | ||
01f3109d OG |
920 | write_unlock_bh(&bond->curr_slave_lock); |
921 | read_unlock(&bond->lock); | |
922 | ||
b7bc2a5b | 923 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 924 | if (should_notify_peers) |
b7bc2a5b AW |
925 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
926 | bond->dev); | |
01f3109d OG |
927 | |
928 | read_lock(&bond->lock); | |
929 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 930 | } |
c3ade5ca | 931 | } |
a2fd940f | 932 | |
5a37e8ca | 933 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
934 | * all were sent on curr_active_slave. |
935 | * resend only if bond is brought up with the affected | |
936 | * bonding modes and the retransmission is enabled */ | |
937 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
938 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
939 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 940 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 941 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 942 | } |
1da177e4 LT |
943 | } |
944 | ||
945 | /** | |
946 | * bond_select_active_slave - select a new active slave, if needed | |
947 | * @bond: our bonding struct | |
948 | * | |
3d632c3f | 949 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
950 | * - The old curr_active_slave has been released or lost its link. |
951 | * - The primary_slave has got its link back. | |
952 | * - A slave has got its link back and there's no old curr_active_slave. | |
953 | * | |
3915c1e8 | 954 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 955 | */ |
a77b5325 | 956 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
957 | { |
958 | struct slave *best_slave; | |
ff59c456 | 959 | int rv; |
1da177e4 LT |
960 | |
961 | best_slave = bond_find_best_slave(bond); | |
962 | if (best_slave != bond->curr_active_slave) { | |
963 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
964 | rv = bond_set_carrier(bond); |
965 | if (!rv) | |
966 | return; | |
967 | ||
968 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
969 | pr_info("%s: first active interface up!\n", |
970 | bond->dev->name); | |
ff59c456 | 971 | } else { |
a4aee5c8 JP |
972 | pr_info("%s: now running without any active interface !\n", |
973 | bond->dev->name); | |
ff59c456 | 974 | } |
1da177e4 LT |
975 | } |
976 | } | |
977 | ||
978 | /*--------------------------- slave list handling ---------------------------*/ | |
979 | ||
980 | /* | |
981 | * This function attaches the slave to the end of list. | |
982 | * | |
983 | * bond->lock held for writing by caller. | |
984 | */ | |
985 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
986 | { | |
278b2083 | 987 | list_add_tail_rcu(&new_slave->list, &bond->slave_list); |
1da177e4 LT |
988 | bond->slave_cnt++; |
989 | } | |
990 | ||
991 | /* | |
992 | * This function detaches the slave from the list. | |
993 | * WARNING: no check is made to verify if the slave effectively | |
994 | * belongs to <bond>. | |
995 | * Nothing is freed on return, structures are just unchained. | |
996 | * If any slave pointer in bond was pointing to <slave>, | |
997 | * it should be changed by the calling function. | |
998 | * | |
999 | * bond->lock held for writing by caller. | |
1000 | */ | |
1001 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1002 | { | |
278b2083 | 1003 | list_del_rcu(&slave->list); |
1da177e4 LT |
1004 | bond->slave_cnt--; |
1005 | } | |
1006 | ||
f6dc31a8 | 1007 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1008 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1009 | { |
8a8efa22 AW |
1010 | struct netpoll *np; |
1011 | int err = 0; | |
f6dc31a8 | 1012 | |
47be03a2 | 1013 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
1014 | err = -ENOMEM; |
1015 | if (!np) | |
1016 | goto out; | |
1017 | ||
47be03a2 | 1018 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
1019 | if (err) { |
1020 | kfree(np); | |
1021 | goto out; | |
f6dc31a8 | 1022 | } |
8a8efa22 AW |
1023 | slave->np = np; |
1024 | out: | |
1025 | return err; | |
1026 | } | |
1027 | static inline void slave_disable_netpoll(struct slave *slave) | |
1028 | { | |
1029 | struct netpoll *np = slave->np; | |
1030 | ||
1031 | if (!np) | |
1032 | return; | |
1033 | ||
1034 | slave->np = NULL; | |
2cde6acd | 1035 | __netpoll_free_async(np); |
8a8efa22 AW |
1036 | } |
1037 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1038 | { | |
1039 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1040 | return false; | |
1041 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1042 | return false; | |
1043 | return true; | |
f6dc31a8 WC |
1044 | } |
1045 | ||
1046 | static void bond_poll_controller(struct net_device *bond_dev) | |
1047 | { | |
8a8efa22 AW |
1048 | } |
1049 | ||
c4cdef9b | 1050 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1051 | { |
c4cdef9b | 1052 | struct bonding *bond = netdev_priv(bond_dev); |
c2355e1a | 1053 | struct slave *slave; |
c2355e1a | 1054 | |
dec1e90e | 1055 | bond_for_each_slave(bond, slave) |
8a8efa22 AW |
1056 | if (IS_UP(slave->dev)) |
1057 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1058 | } |
8a8efa22 | 1059 | |
47be03a2 | 1060 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1061 | { |
1062 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1063 | struct slave *slave; |
dec1e90e | 1064 | int err = 0; |
f6dc31a8 | 1065 | |
dec1e90e | 1066 | bond_for_each_slave(bond, slave) { |
8a8efa22 AW |
1067 | err = slave_enable_netpoll(slave); |
1068 | if (err) { | |
c4cdef9b | 1069 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1070 | break; |
f6dc31a8 WC |
1071 | } |
1072 | } | |
8a8efa22 | 1073 | return err; |
f6dc31a8 | 1074 | } |
8a8efa22 AW |
1075 | #else |
1076 | static inline int slave_enable_netpoll(struct slave *slave) | |
1077 | { | |
1078 | return 0; | |
1079 | } | |
1080 | static inline void slave_disable_netpoll(struct slave *slave) | |
1081 | { | |
1082 | } | |
f6dc31a8 WC |
1083 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1084 | { | |
1085 | } | |
f6dc31a8 WC |
1086 | #endif |
1087 | ||
1da177e4 LT |
1088 | /*---------------------------------- IOCTL ----------------------------------*/ |
1089 | ||
c8f44aff | 1090 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1091 | netdev_features_t features) |
8531c5ff | 1092 | { |
b2a103e6 | 1093 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1094 | netdev_features_t mask; |
9b7b165a | 1095 | struct slave *slave; |
b63365a2 | 1096 | |
dec1e90e | 1097 | if (list_empty(&bond->slave_list)) { |
b2a103e6 MM |
1098 | /* Disable adding VLANs to empty bond. But why? --mq */ |
1099 | features |= NETIF_F_VLAN_CHALLENGED; | |
9b7b165a | 1100 | return features; |
b2a103e6 | 1101 | } |
b63365a2 | 1102 | |
b2a103e6 | 1103 | mask = features; |
b63365a2 | 1104 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1105 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1106 | |
dec1e90e | 1107 | bond_for_each_slave(bond, slave) { |
b63365a2 HX |
1108 | features = netdev_increment_features(features, |
1109 | slave->dev->features, | |
b2a103e6 MM |
1110 | mask); |
1111 | } | |
b0ce3508 | 1112 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1113 | |
b2a103e6 MM |
1114 | return features; |
1115 | } | |
1116 | ||
62f2a3a4 MM |
1117 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1118 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1119 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1120 | |
1121 | static void bond_compute_features(struct bonding *bond) | |
1122 | { | |
c4826861 | 1123 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
c8f44aff | 1124 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1125 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 | 1126 | unsigned int gso_max_size = GSO_MAX_SIZE; |
c4826861 | 1127 | struct net_device *bond_dev = bond->dev; |
0e376bd0 | 1128 | u16 gso_max_segs = GSO_MAX_SEGS; |
c4826861 | 1129 | struct slave *slave; |
b2a103e6 | 1130 | |
dec1e90e | 1131 | if (list_empty(&bond->slave_list)) |
b2a103e6 MM |
1132 | goto done; |
1133 | ||
dec1e90e | 1134 | bond_for_each_slave(bond, slave) { |
278339a4 | 1135 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1136 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1137 | ||
b6fe83e9 | 1138 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1139 | if (slave->dev->hard_header_len > max_hard_header_len) |
1140 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1141 | |
1142 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1143 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1144 | } |
8531c5ff | 1145 | |
b63365a2 | 1146 | done: |
b2a103e6 | 1147 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1148 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1149 | bond_dev->gso_max_segs = gso_max_segs; |
1150 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1151 | |
b6fe83e9 ED |
1152 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1153 | bond_dev->priv_flags = flags | dst_release_flag; | |
1154 | ||
b2a103e6 | 1155 | netdev_change_features(bond_dev); |
8531c5ff AK |
1156 | } |
1157 | ||
872254dd MS |
1158 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1159 | struct net_device *slave_dev) | |
1160 | { | |
00829823 | 1161 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1162 | |
1163 | bond_dev->type = slave_dev->type; | |
1164 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1165 | bond_dev->addr_len = slave_dev->addr_len; | |
1166 | ||
1167 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1168 | slave_dev->addr_len); | |
1169 | } | |
1170 | ||
5b2c4dd2 | 1171 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1172 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1173 | */ |
1174 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1175 | struct slave *slave, |
1176 | struct bonding *bond) | |
5b2c4dd2 | 1177 | { |
2d7011ca | 1178 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1179 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1180 | skb->pkt_type != PACKET_BROADCAST && |
1181 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1182 | return false; |
5b2c4dd2 JP |
1183 | return true; |
1184 | } | |
1185 | return false; | |
1186 | } | |
1187 | ||
8a4eb573 | 1188 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1189 | { |
8a4eb573 | 1190 | struct sk_buff *skb = *pskb; |
f1c1775a | 1191 | struct slave *slave; |
0bd80dad | 1192 | struct bonding *bond; |
de063b70 ED |
1193 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1194 | struct slave *); | |
13a8e0c8 | 1195 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1196 | |
8a4eb573 JP |
1197 | skb = skb_share_check(skb, GFP_ATOMIC); |
1198 | if (unlikely(!skb)) | |
1199 | return RX_HANDLER_CONSUMED; | |
1200 | ||
1201 | *pskb = skb; | |
5b2c4dd2 | 1202 | |
35d48903 JP |
1203 | slave = bond_slave_get_rcu(skb->dev); |
1204 | bond = slave->bond; | |
0bd80dad JP |
1205 | |
1206 | if (bond->params.arp_interval) | |
f1c1775a | 1207 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1208 | |
4d97480b MH |
1209 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1210 | if (recv_probe) { | |
de063b70 ED |
1211 | ret = recv_probe(skb, bond, slave); |
1212 | if (ret == RX_HANDLER_CONSUMED) { | |
1213 | consume_skb(skb); | |
1214 | return ret; | |
3aba891d JP |
1215 | } |
1216 | } | |
1217 | ||
0bd80dad | 1218 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1219 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1220 | } |
1221 | ||
35d48903 | 1222 | skb->dev = bond->dev; |
5b2c4dd2 | 1223 | |
0bd80dad | 1224 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1225 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1226 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1227 | |
541ac7c9 CG |
1228 | if (unlikely(skb_cow_head(skb, |
1229 | skb->data - skb_mac_header(skb)))) { | |
1230 | kfree_skb(skb); | |
8a4eb573 | 1231 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1232 | } |
35d48903 | 1233 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1234 | } |
1235 | ||
13a8e0c8 | 1236 | return ret; |
5b2c4dd2 JP |
1237 | } |
1238 | ||
471cb5a3 | 1239 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1f718f0f VF |
1240 | struct net_device *slave_dev, |
1241 | struct slave *slave) | |
471cb5a3 JP |
1242 | { |
1243 | int err; | |
1244 | ||
1f718f0f | 1245 | err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave); |
471cb5a3 JP |
1246 | if (err) |
1247 | return err; | |
1248 | slave_dev->flags |= IFF_SLAVE; | |
1249 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1250 | return 0; | |
1251 | } | |
1252 | ||
1253 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1254 | struct net_device *slave_dev) | |
1255 | { | |
1256 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1257 | slave_dev->flags &= ~IFF_SLAVE; | |
1258 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1259 | } | |
1260 | ||
1da177e4 | 1261 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1262 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1263 | { |
454d7c9b | 1264 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1265 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1266 | struct slave *new_slave = NULL; |
1da177e4 LT |
1267 | struct sockaddr addr; |
1268 | int link_reporting; | |
8599b52e | 1269 | int res = 0, i; |
1da177e4 | 1270 | |
c772dde3 BH |
1271 | if (!bond->params.use_carrier && |
1272 | slave_dev->ethtool_ops->get_link == NULL && | |
1273 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1274 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1275 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1276 | } |
1277 | ||
1da177e4 LT |
1278 | /* already enslaved */ |
1279 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1280 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1281 | return -EBUSY; |
1282 | } | |
1283 | ||
1284 | /* vlan challenged mutual exclusion */ | |
1285 | /* no need to lock since we're protected by rtnl_lock */ | |
1286 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1287 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1288 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1289 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1290 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1291 | return -EPERM; |
1292 | } else { | |
a4aee5c8 JP |
1293 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1294 | bond_dev->name, slave_dev->name, | |
1295 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1296 | } |
1297 | } else { | |
5a03cdb7 | 1298 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1299 | } |
1300 | ||
217df670 JV |
1301 | /* |
1302 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1303 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1304 | * the current ifenslave will set the interface down prior to |
1305 | * enslaving it; the old ifenslave will not. | |
1306 | */ | |
1307 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1308 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1309 | slave_dev->name); |
1310 | res = -EPERM; | |
1311 | goto err_undo_flags; | |
1312 | } | |
1da177e4 | 1313 | |
872254dd MS |
1314 | /* set bonding device ether type by slave - bonding netdevices are |
1315 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1316 | * there is a need to override some of the type dependent attribs/funcs. | |
1317 | * | |
1318 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1319 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1320 | */ | |
dec1e90e | 1321 | if (list_empty(&bond->slave_list)) { |
e36b9d16 | 1322 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1323 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1324 | bond_dev->name, |
1325 | bond_dev->type, slave_dev->type); | |
75c78500 | 1326 | |
b7bc2a5b AW |
1327 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1328 | bond_dev); | |
3ca5b404 JP |
1329 | res = notifier_to_errno(res); |
1330 | if (res) { | |
1331 | pr_err("%s: refused to change device type\n", | |
1332 | bond_dev->name); | |
1333 | res = -EBUSY; | |
1334 | goto err_undo_flags; | |
1335 | } | |
75c78500 | 1336 | |
32a806c1 | 1337 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1338 | dev_uc_flush(bond_dev); |
22bedad3 | 1339 | dev_mc_flush(bond_dev); |
32a806c1 | 1340 | |
e36b9d16 MS |
1341 | if (slave_dev->type != ARPHRD_ETHER) |
1342 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1343 | else { |
e36b9d16 | 1344 | ether_setup(bond_dev); |
550fd08c NH |
1345 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1346 | } | |
75c78500 | 1347 | |
b7bc2a5b AW |
1348 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1349 | bond_dev); | |
e36b9d16 | 1350 | } |
872254dd | 1351 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1352 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1353 | slave_dev->name, | |
1354 | slave_dev->type, bond_dev->type); | |
1355 | res = -EINVAL; | |
1356 | goto err_undo_flags; | |
872254dd MS |
1357 | } |
1358 | ||
eb7cc59a | 1359 | if (slave_ops->ndo_set_mac_address == NULL) { |
dec1e90e | 1360 | if (list_empty(&bond->slave_list)) { |
a4aee5c8 JP |
1361 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1362 | bond_dev->name); | |
3915c1e8 JV |
1363 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1364 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1365 | 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", |
1366 | bond_dev->name); | |
2ab82852 MS |
1367 | res = -EOPNOTSUPP; |
1368 | goto err_undo_flags; | |
1369 | } | |
1da177e4 LT |
1370 | } |
1371 | ||
8d8fc29d AW |
1372 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1373 | ||
c20811a7 JP |
1374 | /* If this is the first slave, then we need to set the master's hardware |
1375 | * address to be the same as the slave's. */ | |
dec1e90e | 1376 | if (list_empty(&bond->slave_list) && |
1377 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) | |
409cc1f8 | 1378 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1379 | |
243cb4e5 | 1380 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1381 | if (!new_slave) { |
1382 | res = -ENOMEM; | |
1383 | goto err_undo_flags; | |
1384 | } | |
dec1e90e | 1385 | INIT_LIST_HEAD(&new_slave->list); |
bb1d9123 AG |
1386 | /* |
1387 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1388 | * is set via sysfs or module option if desired. | |
1389 | */ | |
1390 | new_slave->queue_id = 0; | |
1391 | ||
b15ba0fb JP |
1392 | /* Save slave's original mtu and then set it to match the bond */ |
1393 | new_slave->original_mtu = slave_dev->mtu; | |
1394 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1395 | if (res) { | |
1396 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1397 | goto err_free; | |
1398 | } | |
1399 | ||
217df670 JV |
1400 | /* |
1401 | * Save slave's original ("permanent") mac address for modes | |
1402 | * that need it, and for restoring it upon release, and then | |
1403 | * set it to the master's address | |
1404 | */ | |
1405 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1406 | |
dd957c57 | 1407 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1408 | /* |
1409 | * Set slave to master's mac address. The application already | |
1410 | * set the master's mac address to that of the first slave | |
1411 | */ | |
1412 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1413 | addr.sa_family = slave_dev->type; | |
1414 | res = dev_set_mac_address(slave_dev, &addr); | |
1415 | if (res) { | |
5a03cdb7 | 1416 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1417 | goto err_restore_mtu; |
2ab82852 | 1418 | } |
217df670 | 1419 | } |
1da177e4 | 1420 | |
217df670 JV |
1421 | /* open the slave since the application closed it */ |
1422 | res = dev_open(slave_dev); | |
1423 | if (res) { | |
3d632c3f | 1424 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1425 | goto err_restore_mac; |
1da177e4 LT |
1426 | } |
1427 | ||
35d48903 | 1428 | new_slave->bond = bond; |
1da177e4 | 1429 | new_slave->dev = slave_dev; |
0b680e75 | 1430 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1431 | |
58402054 | 1432 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1433 | /* bond_alb_init_slave() must be called before all other stages since |
1434 | * it might fail and we do not want to have to undo everything | |
1435 | */ | |
1436 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1437 | if (res) |
569f0c4d | 1438 | goto err_close; |
1da177e4 LT |
1439 | } |
1440 | ||
303d1cbf JV |
1441 | /* If the mode USES_PRIMARY, then the following is handled by |
1442 | * bond_change_active_slave(). | |
1da177e4 LT |
1443 | */ |
1444 | if (!USES_PRIMARY(bond->params.mode)) { | |
1445 | /* set promiscuity level to new slave */ | |
1446 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1447 | res = dev_set_promiscuity(slave_dev, 1); |
1448 | if (res) | |
1449 | goto err_close; | |
1da177e4 LT |
1450 | } |
1451 | ||
1452 | /* set allmulti level to new slave */ | |
1453 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1454 | res = dev_set_allmulti(slave_dev, 1); |
1455 | if (res) | |
1456 | goto err_close; | |
1da177e4 LT |
1457 | } |
1458 | ||
b9e40857 | 1459 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1460 | |
1461 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1462 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1463 | ||
b9e40857 | 1464 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1465 | } |
1466 | ||
1467 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1468 | /* add lacpdu mc addr to mc list */ | |
1469 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1470 | ||
22bedad3 | 1471 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1472 | } |
1473 | ||
b8e2fde4 WY |
1474 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1475 | if (res) { | |
1ff412ad | 1476 | pr_err("%s: Error: Couldn't add bond vlan ids to %s\n", |
1477 | bond_dev->name, slave_dev->name); | |
1478 | goto err_close; | |
1479 | } | |
1da177e4 LT |
1480 | |
1481 | write_lock_bh(&bond->lock); | |
1482 | ||
1483 | bond_attach_slave(bond, new_slave); | |
1484 | ||
1485 | new_slave->delay = 0; | |
1486 | new_slave->link_failure_count = 0; | |
1487 | ||
3915c1e8 JV |
1488 | write_unlock_bh(&bond->lock); |
1489 | ||
b2a103e6 MM |
1490 | bond_compute_features(bond); |
1491 | ||
876254ae VF |
1492 | bond_update_speed_duplex(new_slave); |
1493 | ||
3915c1e8 JV |
1494 | read_lock(&bond->lock); |
1495 | ||
f31c7937 MK |
1496 | new_slave->last_arp_rx = jiffies - |
1497 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
8599b52e VF |
1498 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
1499 | new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx; | |
f5b2b966 | 1500 | |
1da177e4 LT |
1501 | if (bond->params.miimon && !bond->params.use_carrier) { |
1502 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1503 | ||
1504 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1505 | /* | |
1506 | * miimon is set but a bonded network driver | |
1507 | * does not support ETHTOOL/MII and | |
1508 | * arp_interval is not set. Note: if | |
1509 | * use_carrier is enabled, we will never go | |
1510 | * here (because netif_carrier is always | |
1511 | * supported); thus, we don't need to change | |
1512 | * the messages for netif_carrier. | |
1513 | */ | |
a4aee5c8 | 1514 | 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 | 1515 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1516 | } else if (link_reporting == -1) { |
1517 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1518 | 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", |
1519 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1520 | } |
1521 | } | |
1522 | ||
1523 | /* check for initial state */ | |
f31c7937 MK |
1524 | if (bond->params.miimon) { |
1525 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1526 | if (bond->params.updelay) { | |
1527 | new_slave->link = BOND_LINK_BACK; | |
1528 | new_slave->delay = bond->params.updelay; | |
1529 | } else { | |
1530 | new_slave->link = BOND_LINK_UP; | |
1531 | } | |
1da177e4 | 1532 | } else { |
f31c7937 | 1533 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1534 | } |
f31c7937 MK |
1535 | } else if (bond->params.arp_interval) { |
1536 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1537 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1538 | } else { |
f31c7937 | 1539 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1540 | } |
1541 | ||
f31c7937 MK |
1542 | if (new_slave->link != BOND_LINK_DOWN) |
1543 | new_slave->jiffies = jiffies; | |
1544 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1545 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1546 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1547 | ||
1da177e4 LT |
1548 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1549 | /* if there is a primary slave, remember it */ | |
a549952a | 1550 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1551 | bond->primary_slave = new_slave; |
a549952a JP |
1552 | bond->force_primary = true; |
1553 | } | |
1da177e4 LT |
1554 | } |
1555 | ||
3915c1e8 JV |
1556 | write_lock_bh(&bond->curr_slave_lock); |
1557 | ||
1da177e4 LT |
1558 | switch (bond->params.mode) { |
1559 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1560 | bond_set_slave_inactive_flags(new_slave); |
1561 | bond_select_active_slave(bond); | |
1da177e4 LT |
1562 | break; |
1563 | case BOND_MODE_8023AD: | |
1564 | /* in 802.3ad mode, the internal mechanism | |
1565 | * will activate the slaves in the selected | |
1566 | * aggregator | |
1567 | */ | |
1568 | bond_set_slave_inactive_flags(new_slave); | |
1569 | /* if this is the first slave */ | |
dec1e90e | 1570 | if (bond_first_slave(bond) == new_slave) { |
1da177e4 LT |
1571 | SLAVE_AD_INFO(new_slave).id = 1; |
1572 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1573 | * can be called only after the mac address of the bond is set | |
1574 | */ | |
56d00c67 | 1575 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1576 | } else { |
dec1e90e | 1577 | struct slave *prev_slave; |
1578 | ||
1579 | prev_slave = bond_prev_slave(bond, new_slave); | |
1da177e4 | 1580 | SLAVE_AD_INFO(new_slave).id = |
dec1e90e | 1581 | SLAVE_AD_INFO(prev_slave).id + 1; |
1da177e4 LT |
1582 | } |
1583 | ||
1584 | bond_3ad_bind_slave(new_slave); | |
1585 | break; | |
1586 | case BOND_MODE_TLB: | |
1587 | case BOND_MODE_ALB: | |
e30bc066 | 1588 | bond_set_active_slave(new_slave); |
059fe7a5 | 1589 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1590 | bond_select_active_slave(bond); |
1da177e4 LT |
1591 | break; |
1592 | default: | |
5a03cdb7 | 1593 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1594 | |
1595 | /* always active in trunk mode */ | |
e30bc066 | 1596 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1597 | |
1598 | /* In trunking mode there is little meaning to curr_active_slave | |
1599 | * anyway (it holds no special properties of the bond device), | |
1600 | * so we can change it without calling change_active_interface() | |
1601 | */ | |
4e591b93 | 1602 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
278b2083 | 1603 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1604 | |
1da177e4 LT |
1605 | break; |
1606 | } /* switch(bond_mode) */ | |
1607 | ||
3915c1e8 JV |
1608 | write_unlock_bh(&bond->curr_slave_lock); |
1609 | ||
ff59c456 JV |
1610 | bond_set_carrier(bond); |
1611 | ||
f6dc31a8 | 1612 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1613 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1614 | if (slave_dev->npinfo) { |
1615 | if (slave_enable_netpoll(new_slave)) { | |
1616 | read_unlock(&bond->lock); | |
1617 | pr_info("Error, %s: master_dev is using netpoll, " | |
1618 | "but new slave device does not support netpoll.\n", | |
1619 | bond_dev->name); | |
1620 | res = -EBUSY; | |
f7d9821a | 1621 | goto err_detach; |
8a8efa22 | 1622 | } |
f6dc31a8 WC |
1623 | } |
1624 | #endif | |
8a8efa22 | 1625 | |
3915c1e8 | 1626 | read_unlock(&bond->lock); |
1da177e4 | 1627 | |
b76cdba9 MW |
1628 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1629 | if (res) | |
f7d9821a | 1630 | goto err_detach; |
b76cdba9 | 1631 | |
35d48903 JP |
1632 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1633 | new_slave); | |
1634 | if (res) { | |
1635 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1636 | goto err_dest_symlinks; | |
1637 | } | |
1638 | ||
1f718f0f VF |
1639 | res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave); |
1640 | if (res) { | |
1641 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); | |
1642 | goto err_unregister; | |
1643 | } | |
1644 | ||
1645 | ||
a4aee5c8 JP |
1646 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1647 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1648 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1649 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1650 | |
1651 | /* enslave is successful */ | |
1652 | return 0; | |
1653 | ||
1654 | /* Undo stages on error */ | |
1f718f0f VF |
1655 | err_unregister: |
1656 | netdev_rx_handler_unregister(slave_dev); | |
1657 | ||
35d48903 JP |
1658 | err_dest_symlinks: |
1659 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1660 | ||
f7d9821a | 1661 | err_detach: |
303d1cbf JV |
1662 | if (!USES_PRIMARY(bond->params.mode)) |
1663 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1664 | ||
1ff412ad | 1665 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
f7d9821a | 1666 | write_lock_bh(&bond->lock); |
1667 | bond_detach_slave(bond, new_slave); | |
3c5913b5 | 1668 | if (bond->primary_slave == new_slave) |
1669 | bond->primary_slave = NULL; | |
3c5913b5 | 1670 | if (bond->curr_active_slave == new_slave) { |
c6cdcf6d | 1671 | bond_change_active_slave(bond, NULL); |
1672 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1673 | read_lock(&bond->lock); |
1674 | write_lock_bh(&bond->curr_slave_lock); | |
3c5913b5 | 1675 | bond_select_active_slave(bond); |
1676 | write_unlock_bh(&bond->curr_slave_lock); | |
1677 | read_unlock(&bond->lock); | |
c6cdcf6d | 1678 | } else { |
1679 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1680 | } |
fc7a72ac | 1681 | slave_disable_netpoll(new_slave); |
f7d9821a | 1682 | |
1da177e4 | 1683 | err_close: |
b6a5a7b9 | 1684 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1685 | dev_close(slave_dev); |
1686 | ||
1687 | err_restore_mac: | |
dd957c57 | 1688 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1689 | /* XXX TODO - fom follow mode needs to change master's |
1690 | * MAC if this slave's MAC is in use by the bond, or at | |
1691 | * least print a warning. | |
1692 | */ | |
2ab82852 MS |
1693 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1694 | addr.sa_family = slave_dev->type; | |
1695 | dev_set_mac_address(slave_dev, &addr); | |
1696 | } | |
1da177e4 | 1697 | |
b15ba0fb JP |
1698 | err_restore_mtu: |
1699 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1700 | ||
1da177e4 LT |
1701 | err_free: |
1702 | kfree(new_slave); | |
1703 | ||
1704 | err_undo_flags: | |
b2a103e6 | 1705 | bond_compute_features(bond); |
b8fad459 | 1706 | /* Enslave of first slave has failed and we need to fix master's mac */ |
dec1e90e | 1707 | if (list_empty(&bond->slave_list) && |
b8fad459 NA |
1708 | ether_addr_equal(bond_dev->dev_addr, slave_dev->dev_addr)) |
1709 | eth_hw_addr_random(bond_dev); | |
3d632c3f | 1710 | |
1da177e4 LT |
1711 | return res; |
1712 | } | |
1713 | ||
1714 | /* | |
1715 | * Try to release the slave device <slave> from the bond device <master> | |
1716 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1717 | * is write-locked. If "all" is true it means that the function is being called |
1718 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1719 | * |
1720 | * The rules for slave state should be: | |
1721 | * for Active/Backup: | |
1722 | * Active stays on all backups go down | |
1723 | * for Bonded connections: | |
1724 | * The first up interface should be left on and all others downed. | |
1725 | */ | |
0896341a | 1726 | static int __bond_release_one(struct net_device *bond_dev, |
1727 | struct net_device *slave_dev, | |
1728 | bool all) | |
1da177e4 | 1729 | { |
454d7c9b | 1730 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1731 | struct slave *slave, *oldcurrent; |
1732 | struct sockaddr addr; | |
c8f44aff | 1733 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1734 | |
1735 | /* slave is not a slave or master is not master of this slave */ | |
1736 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1737 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1738 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1739 | bond_dev->name, slave_dev->name); |
1740 | return -EINVAL; | |
1741 | } | |
1742 | ||
e843fa50 | 1743 | block_netpoll_tx(); |
1da177e4 LT |
1744 | write_lock_bh(&bond->lock); |
1745 | ||
1746 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1747 | if (!slave) { | |
1748 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1749 | pr_info("%s: %s not enslaved\n", |
1750 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1751 | write_unlock_bh(&bond->lock); |
e843fa50 | 1752 | unblock_netpoll_tx(); |
1da177e4 LT |
1753 | return -EINVAL; |
1754 | } | |
1755 | ||
fcd99434 | 1756 | write_unlock_bh(&bond->lock); |
1f718f0f VF |
1757 | |
1758 | bond_upper_dev_unlink(bond_dev, slave_dev); | |
35d48903 JP |
1759 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1760 | * for this slave anymore. | |
1761 | */ | |
1762 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1763 | write_lock_bh(&bond->lock); |
1764 | ||
1da177e4 LT |
1765 | /* Inform AD package of unbinding of slave. */ |
1766 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1767 | /* must be called before the slave is | |
1768 | * detached from the list | |
1769 | */ | |
1770 | bond_3ad_unbind_slave(slave); | |
1771 | } | |
1772 | ||
a4aee5c8 JP |
1773 | pr_info("%s: releasing %s interface %s\n", |
1774 | bond_dev->name, | |
e30bc066 | 1775 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1776 | slave_dev->name); |
1da177e4 LT |
1777 | |
1778 | oldcurrent = bond->curr_active_slave; | |
1779 | ||
1780 | bond->current_arp_slave = NULL; | |
1781 | ||
1782 | /* release the slave from its bond */ | |
1783 | bond_detach_slave(bond, slave); | |
1784 | ||
dec1e90e | 1785 | if (!all && !bond->params.fail_over_mac) { |
1786 | if (ether_addr_equal(bond_dev->dev_addr, slave->perm_hwaddr) && | |
1787 | !list_empty(&bond->slave_list)) | |
1788 | pr_warn("%s: Warning: the permanent HWaddr of %s - %pM - is still in use by %s. Set the HWaddr of %s to a different address to avoid conflicts.\n", | |
1789 | bond_dev->name, slave_dev->name, | |
1790 | slave->perm_hwaddr, | |
1791 | bond_dev->name, slave_dev->name); | |
1792 | } | |
1793 | ||
3d632c3f | 1794 | if (bond->primary_slave == slave) |
1da177e4 | 1795 | bond->primary_slave = NULL; |
1da177e4 | 1796 | |
3d632c3f | 1797 | if (oldcurrent == slave) |
1da177e4 | 1798 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1799 | |
58402054 | 1800 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1801 | /* Must be called only after the slave has been |
1802 | * detached from the list and the curr_active_slave | |
1803 | * has been cleared (if our_slave == old_current), | |
1804 | * but before a new active slave is selected. | |
1805 | */ | |
2543331d | 1806 | write_unlock_bh(&bond->lock); |
1da177e4 | 1807 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1808 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1809 | } |
1810 | ||
0896341a | 1811 | if (all) { |
278b2083 | 1812 | rcu_assign_pointer(bond->curr_active_slave, NULL); |
0896341a | 1813 | } else if (oldcurrent == slave) { |
059fe7a5 JV |
1814 | /* |
1815 | * Note that we hold RTNL over this sequence, so there | |
1816 | * is no concern that another slave add/remove event | |
1817 | * will interfere. | |
1818 | */ | |
1819 | write_unlock_bh(&bond->lock); | |
1820 | read_lock(&bond->lock); | |
1821 | write_lock_bh(&bond->curr_slave_lock); | |
1822 | ||
1da177e4 LT |
1823 | bond_select_active_slave(bond); |
1824 | ||
059fe7a5 JV |
1825 | write_unlock_bh(&bond->curr_slave_lock); |
1826 | read_unlock(&bond->lock); | |
1827 | write_lock_bh(&bond->lock); | |
1828 | } | |
1829 | ||
dec1e90e | 1830 | if (list_empty(&bond->slave_list)) { |
ff59c456 | 1831 | bond_set_carrier(bond); |
409cc1f8 | 1832 | eth_hw_addr_random(bond_dev); |
1da177e4 | 1833 | |
a59d3d21 | 1834 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1835 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
1836 | bond_dev->name, bond_dev->name); | |
1837 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
1838 | bond_dev->name); | |
1da177e4 | 1839 | } |
1da177e4 LT |
1840 | } |
1841 | ||
1842 | write_unlock_bh(&bond->lock); | |
e843fa50 | 1843 | unblock_netpoll_tx(); |
278b2083 | 1844 | synchronize_rcu(); |
1da177e4 | 1845 | |
dec1e90e | 1846 | if (list_empty(&bond->slave_list)) { |
2af73d4b | 1847 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1848 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1849 | } | |
2af73d4b | 1850 | |
b2a103e6 MM |
1851 | bond_compute_features(bond); |
1852 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1853 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
1854 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
1855 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1856 | ||
b76cdba9 MW |
1857 | /* must do this from outside any spinlocks */ |
1858 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1859 | ||
1ff412ad | 1860 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1861 | |
303d1cbf JV |
1862 | /* If the mode USES_PRIMARY, then this cases was handled above by |
1863 | * bond_change_active_slave(..., NULL) | |
1da177e4 LT |
1864 | */ |
1865 | if (!USES_PRIMARY(bond->params.mode)) { | |
1866 | /* unset promiscuity level from slave */ | |
3d632c3f | 1867 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 1868 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1869 | |
1870 | /* unset allmulti level from slave */ | |
3d632c3f | 1871 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 1872 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1873 | |
303d1cbf | 1874 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1875 | } |
1876 | ||
8a8efa22 | 1877 | slave_disable_netpoll(slave); |
f6dc31a8 | 1878 | |
1da177e4 LT |
1879 | /* close slave before restoring its mac address */ |
1880 | dev_close(slave_dev); | |
1881 | ||
3915c1e8 | 1882 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1883 | /* restore original ("permanent") mac address */ |
1884 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1885 | addr.sa_family = slave_dev->type; | |
1886 | dev_set_mac_address(slave_dev, &addr); | |
1887 | } | |
1da177e4 | 1888 | |
b15ba0fb JP |
1889 | dev_set_mtu(slave_dev, slave->original_mtu); |
1890 | ||
2d7011ca | 1891 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1892 | |
1893 | kfree(slave); | |
1894 | ||
1895 | return 0; /* deletion OK */ | |
1896 | } | |
1897 | ||
0896341a | 1898 | /* A wrapper used because of ndo_del_link */ |
1899 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1900 | { | |
1901 | return __bond_release_one(bond_dev, slave_dev, false); | |
1902 | } | |
1903 | ||
d90a162a | 1904 | /* |
dadaa10b | 1905 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
1906 | * Must be under rtnl_lock when this function is called. |
1907 | */ | |
26d8ee75 | 1908 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1909 | struct net_device *slave_dev) | |
d90a162a | 1910 | { |
454d7c9b | 1911 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1912 | int ret; |
1913 | ||
1914 | ret = bond_release(bond_dev, slave_dev); | |
dec1e90e | 1915 | if (ret == 0 && list_empty(&bond->slave_list)) { |
8a8efa22 | 1916 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
1917 | pr_info("%s: destroying bond %s.\n", |
1918 | bond_dev->name, bond_dev->name); | |
9e71626c | 1919 | unregister_netdevice(bond_dev); |
d90a162a MS |
1920 | } |
1921 | return ret; | |
1922 | } | |
1923 | ||
1da177e4 LT |
1924 | /* |
1925 | * This function changes the active slave to slave <slave_dev>. | |
1926 | * It returns -EINVAL in the following cases. | |
1927 | * - <slave_dev> is not found in the list. | |
1928 | * - There is not active slave now. | |
1929 | * - <slave_dev> is already active. | |
1930 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
1931 | * - <slave_dev> is not running. | |
3d632c3f SH |
1932 | * In these cases, this function does nothing. |
1933 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
1934 | */ |
1935 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
1936 | { | |
454d7c9b | 1937 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1938 | struct slave *old_active = NULL; |
1939 | struct slave *new_active = NULL; | |
1940 | int res = 0; | |
1941 | ||
3d632c3f | 1942 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1943 | return -EINVAL; |
1da177e4 | 1944 | |
471cb5a3 JP |
1945 | /* Verify that bond_dev is indeed the master of slave_dev */ |
1946 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1947 | !netdev_has_upper_dev(slave_dev, bond_dev)) | |
1da177e4 | 1948 | return -EINVAL; |
1da177e4 | 1949 | |
059fe7a5 | 1950 | read_lock(&bond->lock); |
1da177e4 LT |
1951 | |
1952 | old_active = bond->curr_active_slave; | |
1953 | new_active = bond_get_slave_by_dev(bond, slave_dev); | |
1da177e4 LT |
1954 | /* |
1955 | * Changing to the current active: do nothing; return success. | |
1956 | */ | |
71bc3b2d | 1957 | if (new_active && new_active == old_active) { |
059fe7a5 | 1958 | read_unlock(&bond->lock); |
1da177e4 LT |
1959 | return 0; |
1960 | } | |
1961 | ||
71bc3b2d | 1962 | if (new_active && |
1963 | old_active && | |
1964 | new_active->link == BOND_LINK_UP && | |
1da177e4 | 1965 | IS_UP(new_active->dev)) { |
e843fa50 | 1966 | block_netpoll_tx(); |
059fe7a5 | 1967 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 1968 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 1969 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 1970 | unblock_netpoll_tx(); |
3d632c3f | 1971 | } else |
1da177e4 | 1972 | res = -EINVAL; |
1da177e4 | 1973 | |
059fe7a5 | 1974 | read_unlock(&bond->lock); |
1da177e4 LT |
1975 | |
1976 | return res; | |
1977 | } | |
1978 | ||
1da177e4 LT |
1979 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1980 | { | |
454d7c9b | 1981 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1982 | |
1983 | info->bond_mode = bond->params.mode; | |
1984 | info->miimon = bond->params.miimon; | |
1985 | ||
6603a6f2 | 1986 | read_lock(&bond->lock); |
1da177e4 | 1987 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 1988 | read_unlock(&bond->lock); |
1da177e4 LT |
1989 | |
1990 | return 0; | |
1991 | } | |
1992 | ||
1993 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1994 | { | |
454d7c9b | 1995 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 1996 | int i = 0, res = -ENODEV; |
1da177e4 | 1997 | struct slave *slave; |
1da177e4 | 1998 | |
6603a6f2 | 1999 | read_lock(&bond->lock); |
dec1e90e | 2000 | bond_for_each_slave(bond, slave) { |
2001 | if (i++ == (int)info->slave_id) { | |
689c96cc ED |
2002 | res = 0; |
2003 | strcpy(info->slave_name, slave->dev->name); | |
2004 | info->link = slave->link; | |
e30bc066 | 2005 | info->state = bond_slave_state(slave); |
689c96cc | 2006 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2007 | break; |
2008 | } | |
2009 | } | |
6603a6f2 | 2010 | read_unlock(&bond->lock); |
1da177e4 | 2011 | |
689c96cc | 2012 | return res; |
1da177e4 LT |
2013 | } |
2014 | ||
2015 | /*-------------------------------- Monitoring -------------------------------*/ | |
2016 | ||
1da177e4 | 2017 | |
f0c76d61 JV |
2018 | static int bond_miimon_inspect(struct bonding *bond) |
2019 | { | |
dec1e90e | 2020 | int link_state, commit = 0; |
f0c76d61 | 2021 | struct slave *slave; |
41f89100 JP |
2022 | bool ignore_updelay; |
2023 | ||
2024 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 | 2025 | |
dec1e90e | 2026 | bond_for_each_slave(bond, slave) { |
f0c76d61 | 2027 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2028 | |
f0c76d61 | 2029 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2030 | |
2031 | switch (slave->link) { | |
f0c76d61 JV |
2032 | case BOND_LINK_UP: |
2033 | if (link_state) | |
2034 | continue; | |
1da177e4 | 2035 | |
f0c76d61 JV |
2036 | slave->link = BOND_LINK_FAIL; |
2037 | slave->delay = bond->params.downdelay; | |
2038 | if (slave->delay) { | |
a4aee5c8 JP |
2039 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2040 | bond->dev->name, | |
2041 | (bond->params.mode == | |
2042 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2043 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2044 | "active " : "backup ") : "", |
2045 | slave->dev->name, | |
2046 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2047 | } |
f0c76d61 JV |
2048 | /*FALLTHRU*/ |
2049 | case BOND_LINK_FAIL: | |
2050 | if (link_state) { | |
2051 | /* | |
2052 | * recovered before downdelay expired | |
2053 | */ | |
2054 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2055 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2056 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2057 | bond->dev->name, | |
2058 | (bond->params.downdelay - slave->delay) * | |
2059 | bond->params.miimon, | |
2060 | slave->dev->name); | |
f0c76d61 | 2061 | continue; |
1da177e4 | 2062 | } |
f0c76d61 JV |
2063 | |
2064 | if (slave->delay <= 0) { | |
2065 | slave->new_link = BOND_LINK_DOWN; | |
2066 | commit++; | |
2067 | continue; | |
1da177e4 | 2068 | } |
1da177e4 | 2069 | |
f0c76d61 JV |
2070 | slave->delay--; |
2071 | break; | |
2072 | ||
2073 | case BOND_LINK_DOWN: | |
2074 | if (!link_state) | |
2075 | continue; | |
2076 | ||
2077 | slave->link = BOND_LINK_BACK; | |
2078 | slave->delay = bond->params.updelay; | |
2079 | ||
2080 | if (slave->delay) { | |
a4aee5c8 JP |
2081 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2082 | bond->dev->name, slave->dev->name, | |
2083 | ignore_updelay ? 0 : | |
2084 | bond->params.updelay * | |
2085 | bond->params.miimon); | |
f0c76d61 JV |
2086 | } |
2087 | /*FALLTHRU*/ | |
2088 | case BOND_LINK_BACK: | |
2089 | if (!link_state) { | |
2090 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2091 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2092 | bond->dev->name, | |
2093 | (bond->params.updelay - slave->delay) * | |
2094 | bond->params.miimon, | |
2095 | slave->dev->name); | |
f0c76d61 JV |
2096 | |
2097 | continue; | |
2098 | } | |
2099 | ||
41f89100 JP |
2100 | if (ignore_updelay) |
2101 | slave->delay = 0; | |
2102 | ||
f0c76d61 JV |
2103 | if (slave->delay <= 0) { |
2104 | slave->new_link = BOND_LINK_UP; | |
2105 | commit++; | |
41f89100 | 2106 | ignore_updelay = false; |
f0c76d61 | 2107 | continue; |
1da177e4 | 2108 | } |
f0c76d61 JV |
2109 | |
2110 | slave->delay--; | |
1da177e4 | 2111 | break; |
f0c76d61 JV |
2112 | } |
2113 | } | |
1da177e4 | 2114 | |
f0c76d61 JV |
2115 | return commit; |
2116 | } | |
1da177e4 | 2117 | |
f0c76d61 JV |
2118 | static void bond_miimon_commit(struct bonding *bond) |
2119 | { | |
2120 | struct slave *slave; | |
f0c76d61 | 2121 | |
dec1e90e | 2122 | bond_for_each_slave(bond, slave) { |
f0c76d61 JV |
2123 | switch (slave->new_link) { |
2124 | case BOND_LINK_NOCHANGE: | |
2125 | continue; | |
1da177e4 | 2126 | |
f0c76d61 JV |
2127 | case BOND_LINK_UP: |
2128 | slave->link = BOND_LINK_UP; | |
2129 | slave->jiffies = jiffies; | |
2130 | ||
2131 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2132 | /* prevent it from being the active one */ | |
e30bc066 | 2133 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2134 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2135 | /* make it immediately active */ | |
e30bc066 | 2136 | bond_set_active_slave(slave); |
f0c76d61 JV |
2137 | } else if (slave != bond->primary_slave) { |
2138 | /* prevent it from being the active one */ | |
e30bc066 | 2139 | bond_set_backup_slave(slave); |
1da177e4 | 2140 | } |
1da177e4 | 2141 | |
5d30530e | 2142 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 | 2143 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
2144 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
2145 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2146 | |
f0c76d61 JV |
2147 | /* notify ad that the link status has changed */ |
2148 | if (bond->params.mode == BOND_MODE_8023AD) | |
2149 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2150 | |
58402054 | 2151 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2152 | bond_alb_handle_link_change(bond, slave, |
2153 | BOND_LINK_UP); | |
1da177e4 | 2154 | |
f0c76d61 JV |
2155 | if (!bond->curr_active_slave || |
2156 | (slave == bond->primary_slave)) | |
2157 | goto do_failover; | |
1da177e4 | 2158 | |
f0c76d61 | 2159 | continue; |
059fe7a5 | 2160 | |
f0c76d61 | 2161 | case BOND_LINK_DOWN: |
fba4acda JV |
2162 | if (slave->link_failure_count < UINT_MAX) |
2163 | slave->link_failure_count++; | |
2164 | ||
f0c76d61 | 2165 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2166 | |
f0c76d61 JV |
2167 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2168 | bond->params.mode == BOND_MODE_8023AD) | |
2169 | bond_set_slave_inactive_flags(slave); | |
2170 | ||
a4aee5c8 JP |
2171 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2172 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2173 | |
2174 | if (bond->params.mode == BOND_MODE_8023AD) | |
2175 | bond_3ad_handle_link_change(slave, | |
2176 | BOND_LINK_DOWN); | |
2177 | ||
ae63e808 | 2178 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2179 | bond_alb_handle_link_change(bond, slave, |
2180 | BOND_LINK_DOWN); | |
2181 | ||
2182 | if (slave == bond->curr_active_slave) | |
2183 | goto do_failover; | |
2184 | ||
2185 | continue; | |
2186 | ||
2187 | default: | |
a4aee5c8 | 2188 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2189 | bond->dev->name, slave->new_link, |
2190 | slave->dev->name); | |
2191 | slave->new_link = BOND_LINK_NOCHANGE; | |
2192 | ||
2193 | continue; | |
2194 | } | |
2195 | ||
2196 | do_failover: | |
2197 | ASSERT_RTNL(); | |
e843fa50 | 2198 | block_netpoll_tx(); |
f0c76d61 JV |
2199 | write_lock_bh(&bond->curr_slave_lock); |
2200 | bond_select_active_slave(bond); | |
2201 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2202 | unblock_netpoll_tx(); |
f0c76d61 JV |
2203 | } |
2204 | ||
2205 | bond_set_carrier(bond); | |
1da177e4 LT |
2206 | } |
2207 | ||
0b0eef66 JV |
2208 | /* |
2209 | * bond_mii_monitor | |
2210 | * | |
2211 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2212 | * inspection, then (if inspection indicates something needs to be done) |
2213 | * an acquisition of appropriate locks followed by a commit phase to | |
2214 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2215 | */ |
2216 | void bond_mii_monitor(struct work_struct *work) | |
2217 | { | |
2218 | struct bonding *bond = container_of(work, struct bonding, | |
2219 | mii_work.work); | |
ad246c99 | 2220 | bool should_notify_peers = false; |
e6d265e8 | 2221 | unsigned long delay; |
0b0eef66 JV |
2222 | |
2223 | read_lock(&bond->lock); | |
e6d265e8 JV |
2224 | |
2225 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 | 2226 | |
dec1e90e | 2227 | if (list_empty(&bond->slave_list)) |
f0c76d61 | 2228 | goto re_arm; |
b59f9f74 | 2229 | |
ad246c99 BH |
2230 | should_notify_peers = bond_should_notify_peers(bond); |
2231 | ||
f0c76d61 | 2232 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2233 | read_unlock(&bond->lock); |
e6d265e8 JV |
2234 | |
2235 | /* Race avoidance with bond_close cancel of workqueue */ | |
2236 | if (!rtnl_trylock()) { | |
2237 | read_lock(&bond->lock); | |
2238 | delay = 1; | |
2239 | should_notify_peers = false; | |
2240 | goto re_arm; | |
2241 | } | |
2242 | ||
0b0eef66 | 2243 | read_lock(&bond->lock); |
f0c76d61 JV |
2244 | |
2245 | bond_miimon_commit(bond); | |
2246 | ||
5655662d JV |
2247 | read_unlock(&bond->lock); |
2248 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2249 | read_lock(&bond->lock); | |
0b0eef66 JV |
2250 | } |
2251 | ||
f0c76d61 | 2252 | re_arm: |
e6d265e8 JV |
2253 | if (bond->params.miimon) |
2254 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2255 | ||
0b0eef66 | 2256 | read_unlock(&bond->lock); |
ad246c99 BH |
2257 | |
2258 | if (should_notify_peers) { | |
6c8d23f7 | 2259 | if (!rtnl_trylock()) |
e6d265e8 | 2260 | return; |
b7bc2a5b | 2261 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2262 | rtnl_unlock(); |
2263 | } | |
0b0eef66 | 2264 | } |
c3ade5ca | 2265 | |
50223ce4 | 2266 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2267 | { |
50223ce4 VF |
2268 | struct net_device *upper; |
2269 | struct list_head *iter; | |
2270 | bool ret = false; | |
f5b2b966 | 2271 | |
eaddcd76 | 2272 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2273 | return true; |
f5b2b966 | 2274 | |
50223ce4 | 2275 | rcu_read_lock(); |
2f268f12 | 2276 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2277 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2278 | ret = true; | |
2279 | break; | |
2280 | } | |
f5b2b966 | 2281 | } |
50223ce4 | 2282 | rcu_read_unlock(); |
f5b2b966 | 2283 | |
50223ce4 | 2284 | return ret; |
f5b2b966 JV |
2285 | } |
2286 | ||
c3ade5ca JV |
2287 | /* |
2288 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2289 | * switches in VLAN mode (especially if ports are configured as | |
2290 | * "native" to a VLAN) might not pass non-tagged frames. | |
2291 | */ | |
d3bb52b0 | 2292 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id) |
c3ade5ca JV |
2293 | { |
2294 | struct sk_buff *skb; | |
2295 | ||
acca2674 | 2296 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op, |
2297 | slave_dev->name, &dest_ip, &src_ip, vlan_id); | |
3d632c3f | 2298 | |
c3ade5ca JV |
2299 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2300 | NULL, slave_dev->dev_addr, NULL); | |
2301 | ||
2302 | if (!skb) { | |
a4aee5c8 | 2303 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2304 | return; |
2305 | } | |
2306 | if (vlan_id) { | |
1fd9b1fc | 2307 | skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id); |
c3ade5ca | 2308 | if (!skb) { |
a4aee5c8 | 2309 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2310 | return; |
2311 | } | |
2312 | } | |
2313 | arp_xmit(skb); | |
2314 | } | |
2315 | ||
2316 | ||
1da177e4 LT |
2317 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2318 | { | |
27bc11e6 VF |
2319 | struct net_device *upper, *vlan_upper; |
2320 | struct list_head *iter, *vlan_iter; | |
c3ade5ca | 2321 | struct rtable *rt; |
27bc11e6 VF |
2322 | __be32 *targets = bond->params.arp_targets, addr; |
2323 | int i, vlan_id; | |
1da177e4 | 2324 | |
27bc11e6 | 2325 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
acca2674 | 2326 | pr_debug("basa: target %pI4\n", &targets[i]); |
c3ade5ca | 2327 | |
27bc11e6 | 2328 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2329 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2330 | RTO_ONLINK, 0); | |
b23dd4fe | 2331 | if (IS_ERR(rt)) { |
3e32582f VF |
2332 | pr_debug("%s: no route to arp_ip_target %pI4\n", |
2333 | bond->dev->name, &targets[i]); | |
c3ade5ca JV |
2334 | continue; |
2335 | } | |
2336 | ||
c3ade5ca | 2337 | vlan_id = 0; |
27bc11e6 VF |
2338 | |
2339 | /* bond device itself */ | |
2340 | if (rt->dst.dev == bond->dev) | |
2341 | goto found; | |
2342 | ||
2343 | rcu_read_lock(); | |
2344 | /* first we search only for vlan devices. for every vlan | |
2345 | * found we verify its upper dev list, searching for the | |
2346 | * rt->dst.dev. If found we save the tag of the vlan and | |
2347 | * proceed to send the packet. | |
2348 | * | |
2349 | * TODO: QinQ? | |
2350 | */ | |
2f268f12 VF |
2351 | netdev_for_each_all_upper_dev_rcu(bond->dev, vlan_upper, |
2352 | vlan_iter) { | |
27bc11e6 VF |
2353 | if (!is_vlan_dev(vlan_upper)) |
2354 | continue; | |
2f268f12 VF |
2355 | netdev_for_each_all_upper_dev_rcu(vlan_upper, upper, |
2356 | iter) { | |
27bc11e6 VF |
2357 | if (upper == rt->dst.dev) { |
2358 | vlan_id = vlan_dev_vlan_id(vlan_upper); | |
2359 | rcu_read_unlock(); | |
2360 | goto found; | |
2361 | } | |
c3ade5ca JV |
2362 | } |
2363 | } | |
2364 | ||
27bc11e6 VF |
2365 | /* if the device we're looking for is not on top of any of |
2366 | * our upper vlans, then just search for any dev that | |
2367 | * matches, and in case it's a vlan - save the id | |
2368 | */ | |
2f268f12 | 2369 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
27bc11e6 VF |
2370 | if (upper == rt->dst.dev) { |
2371 | /* if it's a vlan - get its VID */ | |
2372 | if (is_vlan_dev(upper)) | |
2373 | vlan_id = vlan_dev_vlan_id(upper); | |
2374 | ||
2375 | rcu_read_unlock(); | |
2376 | goto found; | |
2377 | } | |
c3ade5ca | 2378 | } |
27bc11e6 | 2379 | rcu_read_unlock(); |
c3ade5ca | 2380 | |
27bc11e6 | 2381 | /* Not our device - skip */ |
3e32582f VF |
2382 | pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2383 | bond->dev->name, &targets[i], | |
2384 | rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
2385 | ||
ed4b9f80 | 2386 | ip_rt_put(rt); |
27bc11e6 VF |
2387 | continue; |
2388 | ||
2389 | found: | |
2390 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2391 | ip_rt_put(rt); | |
2392 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
2393 | addr, vlan_id); | |
c3ade5ca JV |
2394 | } |
2395 | } | |
2396 | ||
d3bb52b0 | 2397 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2398 | { |
8599b52e VF |
2399 | int i; |
2400 | ||
87a7b84b VF |
2401 | if (!sip || !bond_has_this_ip(bond, tip)) { |
2402 | pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip); | |
2403 | return; | |
2404 | } | |
f5b2b966 | 2405 | |
8599b52e VF |
2406 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2407 | if (i == -1) { | |
87a7b84b VF |
2408 | pr_debug("bva: sip %pI4 not found in targets\n", &sip); |
2409 | return; | |
f5b2b966 | 2410 | } |
87a7b84b | 2411 | slave->last_arp_rx = jiffies; |
8599b52e | 2412 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2413 | } |
2414 | ||
5bb9e0b5 | 2415 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2416 | struct slave *slave) | |
f5b2b966 | 2417 | { |
de063b70 | 2418 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2419 | unsigned char *arp_ptr; |
d3bb52b0 | 2420 | __be32 sip, tip; |
de063b70 | 2421 | int alen; |
f5b2b966 | 2422 | |
3aba891d | 2423 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2424 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2425 | |
f5b2b966 | 2426 | read_lock(&bond->lock); |
2c146102 VF |
2427 | |
2428 | if (!slave_do_arp_validate(bond, slave)) | |
2429 | goto out_unlock; | |
2430 | ||
de063b70 | 2431 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2432 | |
3aba891d JP |
2433 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2434 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2435 | |
de063b70 ED |
2436 | if (alen > skb_headlen(skb)) { |
2437 | arp = kmalloc(alen, GFP_ATOMIC); | |
2438 | if (!arp) | |
2439 | goto out_unlock; | |
2440 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2441 | goto out_unlock; | |
2442 | } | |
f5b2b966 | 2443 | |
3aba891d | 2444 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2445 | skb->pkt_type == PACKET_OTHERHOST || |
2446 | skb->pkt_type == PACKET_LOOPBACK || | |
2447 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2448 | arp->ar_pro != htons(ETH_P_IP) || | |
2449 | arp->ar_pln != 4) | |
2450 | goto out_unlock; | |
2451 | ||
2452 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2453 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2454 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2455 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2456 | memcpy(&tip, arp_ptr, 4); |
2457 | ||
5a03cdb7 | 2458 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2459 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2460 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2461 | &sip, &tip); | |
f5b2b966 JV |
2462 | |
2463 | /* | |
2464 | * Backup slaves won't see the ARP reply, but do come through | |
2465 | * here for each ARP probe (so we swap the sip/tip to validate | |
2466 | * the probe). In a "redundant switch, common router" type of | |
2467 | * configuration, the ARP probe will (hopefully) travel from | |
2468 | * the active, through one switch, the router, then the other | |
2469 | * switch before reaching the backup. | |
aeea64ac VF |
2470 | * |
2471 | * We 'trust' the arp requests if there is an active slave and | |
2472 | * it received valid arp reply(s) after it became active. This | |
2473 | * is done to avoid endless looping when we can't reach the | |
2474 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2475 | */ |
e30bc066 | 2476 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2477 | bond_validate_arp(bond, slave, sip, tip); |
aeea64ac VF |
2478 | else if (bond->curr_active_slave && |
2479 | time_after(slave_last_rx(bond, bond->curr_active_slave), | |
2480 | bond->curr_active_slave->jiffies)) | |
f5b2b966 JV |
2481 | bond_validate_arp(bond, slave, tip, sip); |
2482 | ||
2483 | out_unlock: | |
2484 | read_unlock(&bond->lock); | |
de063b70 ED |
2485 | if (arp != (struct arphdr *)skb->data) |
2486 | kfree(arp); | |
13a8e0c8 | 2487 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2488 | } |
2489 | ||
e7f63f1d VF |
2490 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2491 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2492 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2493 | */ | |
2494 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2495 | int mod) | |
2496 | { | |
2497 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2498 | ||
2499 | return time_in_range(jiffies, | |
2500 | last_act - delta_in_ticks, | |
2501 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2502 | } | |
2503 | ||
1da177e4 LT |
2504 | /* |
2505 | * this function is called regularly to monitor each slave's link | |
2506 | * ensuring that traffic is being sent and received when arp monitoring | |
2507 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2508 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2509 | * arp monitoring in active backup mode. | |
2510 | */ | |
1b76b316 | 2511 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2512 | { |
1b76b316 JV |
2513 | struct bonding *bond = container_of(work, struct bonding, |
2514 | arp_work.work); | |
1da177e4 LT |
2515 | struct slave *slave, *oldcurrent; |
2516 | int do_failover = 0; | |
1da177e4 LT |
2517 | |
2518 | read_lock(&bond->lock); | |
2519 | ||
dec1e90e | 2520 | if (list_empty(&bond->slave_list)) |
1da177e4 | 2521 | goto re_arm; |
1da177e4 | 2522 | |
1da177e4 | 2523 | oldcurrent = bond->curr_active_slave; |
1da177e4 LT |
2524 | /* see if any of the previous devices are up now (i.e. they have |
2525 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2526 | * the picture unless it is null. also, slave->jiffies is not needed | |
2527 | * here because we send an arp on each slave and give a slave as | |
2528 | * long as it needs to get the tx/rx within the delta. | |
2529 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2530 | * so it can wait | |
2531 | */ | |
dec1e90e | 2532 | bond_for_each_slave(bond, slave) { |
cb32f2a0 JB |
2533 | unsigned long trans_start = dev_trans_start(slave->dev); |
2534 | ||
1da177e4 | 2535 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d VF |
2536 | if (bond_time_in_interval(bond, trans_start, 1) && |
2537 | bond_time_in_interval(bond, slave->dev->last_rx, 1)) { | |
1da177e4 LT |
2538 | |
2539 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2540 | bond_set_active_slave(slave); |
1da177e4 LT |
2541 | |
2542 | /* primary_slave has no meaning in round-robin | |
2543 | * mode. the window of a slave being up and | |
2544 | * curr_active_slave being null after enslaving | |
2545 | * is closed. | |
2546 | */ | |
2547 | if (!oldcurrent) { | |
a4aee5c8 JP |
2548 | pr_info("%s: link status definitely up for interface %s, ", |
2549 | bond->dev->name, | |
2550 | slave->dev->name); | |
1da177e4 LT |
2551 | do_failover = 1; |
2552 | } else { | |
a4aee5c8 JP |
2553 | pr_info("%s: interface %s is now up\n", |
2554 | bond->dev->name, | |
2555 | slave->dev->name); | |
1da177e4 LT |
2556 | } |
2557 | } | |
2558 | } else { | |
2559 | /* slave->link == BOND_LINK_UP */ | |
2560 | ||
2561 | /* not all switches will respond to an arp request | |
2562 | * when the source ip is 0, so don't take the link down | |
2563 | * if we don't know our ip yet | |
2564 | */ | |
e7f63f1d VF |
2565 | if (!bond_time_in_interval(bond, trans_start, 2) || |
2566 | !bond_time_in_interval(bond, slave->dev->last_rx, 2)) { | |
1da177e4 LT |
2567 | |
2568 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2569 | bond_set_backup_slave(slave); |
1da177e4 | 2570 | |
3d632c3f | 2571 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2572 | slave->link_failure_count++; |
1da177e4 | 2573 | |
a4aee5c8 JP |
2574 | pr_info("%s: interface %s is now down.\n", |
2575 | bond->dev->name, | |
2576 | slave->dev->name); | |
1da177e4 | 2577 | |
3d632c3f | 2578 | if (slave == oldcurrent) |
1da177e4 | 2579 | do_failover = 1; |
1da177e4 LT |
2580 | } |
2581 | } | |
2582 | ||
2583 | /* note: if switch is in round-robin mode, all links | |
2584 | * must tx arp to ensure all links rx an arp - otherwise | |
2585 | * links may oscillate or not come up at all; if switch is | |
2586 | * in something like xor mode, there is nothing we can | |
2587 | * do - all replies will be rx'ed on same link causing slaves | |
2588 | * to be unstable during low/no traffic periods | |
2589 | */ | |
3d632c3f | 2590 | if (IS_UP(slave->dev)) |
1da177e4 | 2591 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2592 | } |
2593 | ||
2594 | if (do_failover) { | |
e843fa50 | 2595 | block_netpoll_tx(); |
059fe7a5 | 2596 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2597 | |
2598 | bond_select_active_slave(bond); | |
2599 | ||
059fe7a5 | 2600 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2601 | unblock_netpoll_tx(); |
1da177e4 LT |
2602 | } |
2603 | ||
2604 | re_arm: | |
e6d265e8 | 2605 | if (bond->params.arp_interval) |
e7f63f1d VF |
2606 | queue_delayed_work(bond->wq, &bond->arp_work, |
2607 | msecs_to_jiffies(bond->params.arp_interval)); | |
e6d265e8 | 2608 | |
1da177e4 LT |
2609 | read_unlock(&bond->lock); |
2610 | } | |
2611 | ||
2612 | /* | |
b2220cad JV |
2613 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2614 | * changes. Sets new_link in slaves to specify what action should take | |
2615 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2616 | * to link states must be committed. | |
2617 | * | |
2618 | * Called with bond->lock held for read. | |
1da177e4 | 2619 | */ |
e7f63f1d | 2620 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2621 | { |
def4460c | 2622 | unsigned long trans_start, last_rx; |
dec1e90e | 2623 | struct slave *slave; |
dec1e90e | 2624 | int commit = 0; |
da210f55 | 2625 | |
dec1e90e | 2626 | bond_for_each_slave(bond, slave) { |
b2220cad | 2627 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2628 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2629 | |
b2220cad | 2630 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2631 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2632 | slave->new_link = BOND_LINK_UP; |
2633 | commit++; | |
2634 | } | |
b2220cad JV |
2635 | continue; |
2636 | } | |
1da177e4 | 2637 | |
b2220cad JV |
2638 | /* |
2639 | * Give slaves 2*delta after being enslaved or made | |
2640 | * active. This avoids bouncing, as the last receive | |
2641 | * times need a full ARP monitor cycle to be updated. | |
2642 | */ | |
e7f63f1d | 2643 | if (bond_time_in_interval(bond, slave->jiffies, 2)) |
b2220cad JV |
2644 | continue; |
2645 | ||
2646 | /* | |
2647 | * Backup slave is down if: | |
2648 | * - No current_arp_slave AND | |
2649 | * - more than 3*delta since last receive AND | |
2650 | * - the bond has an IP address | |
2651 | * | |
2652 | * Note: a non-null current_arp_slave indicates | |
2653 | * the curr_active_slave went down and we are | |
2654 | * searching for a new one; under this condition | |
2655 | * we only take the curr_active_slave down - this | |
2656 | * gives each slave a chance to tx/rx traffic | |
2657 | * before being taken out | |
2658 | */ | |
e30bc066 | 2659 | if (!bond_is_active_slave(slave) && |
b2220cad | 2660 | !bond->current_arp_slave && |
e7f63f1d | 2661 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2662 | slave->new_link = BOND_LINK_DOWN; |
2663 | commit++; | |
2664 | } | |
2665 | ||
2666 | /* | |
2667 | * Active slave is down if: | |
2668 | * - more than 2*delta since transmitting OR | |
2669 | * - (more than 2*delta since receive AND | |
2670 | * the bond has an IP address) | |
2671 | */ | |
cb32f2a0 | 2672 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2673 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2674 | (!bond_time_in_interval(bond, trans_start, 2) || |
2675 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2676 | slave->new_link = BOND_LINK_DOWN; |
2677 | commit++; | |
2678 | } | |
1da177e4 LT |
2679 | } |
2680 | ||
b2220cad JV |
2681 | return commit; |
2682 | } | |
1da177e4 | 2683 | |
b2220cad JV |
2684 | /* |
2685 | * Called to commit link state changes noted by inspection step of | |
2686 | * active-backup mode ARP monitor. | |
2687 | * | |
2688 | * Called with RTNL and bond->lock for read. | |
2689 | */ | |
e7f63f1d | 2690 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2691 | { |
cb32f2a0 | 2692 | unsigned long trans_start; |
dec1e90e | 2693 | struct slave *slave; |
1da177e4 | 2694 | |
dec1e90e | 2695 | bond_for_each_slave(bond, slave) { |
b2220cad JV |
2696 | switch (slave->new_link) { |
2697 | case BOND_LINK_NOCHANGE: | |
2698 | continue; | |
ff59c456 | 2699 | |
b2220cad | 2700 | case BOND_LINK_UP: |
cb32f2a0 | 2701 | trans_start = dev_trans_start(slave->dev); |
e7f63f1d VF |
2702 | if (bond->curr_active_slave != slave || |
2703 | (!bond->curr_active_slave && | |
2704 | bond_time_in_interval(bond, trans_start, 1))) { | |
b2220cad | 2705 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2706 | if (bond->current_arp_slave) { |
2707 | bond_set_slave_inactive_flags( | |
2708 | bond->current_arp_slave); | |
2709 | bond->current_arp_slave = NULL; | |
2710 | } | |
b2220cad | 2711 | |
a4aee5c8 | 2712 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2713 | bond->dev->name, slave->dev->name); |
b2220cad | 2714 | |
b9f60253 JP |
2715 | if (!bond->curr_active_slave || |
2716 | (slave == bond->primary_slave)) | |
2717 | goto do_failover; | |
1da177e4 | 2718 | |
b9f60253 | 2719 | } |
1da177e4 | 2720 | |
b9f60253 | 2721 | continue; |
1da177e4 | 2722 | |
b2220cad JV |
2723 | case BOND_LINK_DOWN: |
2724 | if (slave->link_failure_count < UINT_MAX) | |
2725 | slave->link_failure_count++; | |
2726 | ||
2727 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2728 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2729 | |
a4aee5c8 | 2730 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2731 | bond->dev->name, slave->dev->name); |
b2220cad | 2732 | |
b9f60253 | 2733 | if (slave == bond->curr_active_slave) { |
b2220cad | 2734 | bond->current_arp_slave = NULL; |
b9f60253 | 2735 | goto do_failover; |
1da177e4 | 2736 | } |
b9f60253 JP |
2737 | |
2738 | continue; | |
b2220cad JV |
2739 | |
2740 | default: | |
a4aee5c8 | 2741 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2742 | bond->dev->name, slave->new_link, |
2743 | slave->dev->name); | |
b9f60253 | 2744 | continue; |
1da177e4 | 2745 | } |
1da177e4 | 2746 | |
b9f60253 JP |
2747 | do_failover: |
2748 | ASSERT_RTNL(); | |
e843fa50 | 2749 | block_netpoll_tx(); |
b2220cad | 2750 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2751 | bond_select_active_slave(bond); |
b2220cad | 2752 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2753 | unblock_netpoll_tx(); |
b2220cad | 2754 | } |
1da177e4 | 2755 | |
b2220cad JV |
2756 | bond_set_carrier(bond); |
2757 | } | |
1da177e4 | 2758 | |
b2220cad JV |
2759 | /* |
2760 | * Send ARP probes for active-backup mode ARP monitor. | |
2761 | * | |
2762 | * Called with bond->lock held for read. | |
2763 | */ | |
2764 | static void bond_ab_arp_probe(struct bonding *bond) | |
2765 | { | |
dec1e90e | 2766 | struct slave *slave, *next_slave; |
b2220cad | 2767 | int i; |
1da177e4 | 2768 | |
b2220cad | 2769 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2770 | |
b2220cad | 2771 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
2772 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
2773 | bond->current_arp_slave->dev->name, | |
2774 | bond->curr_active_slave->dev->name); | |
1da177e4 | 2775 | |
b2220cad JV |
2776 | if (bond->curr_active_slave) { |
2777 | bond_arp_send_all(bond, bond->curr_active_slave); | |
2778 | read_unlock(&bond->curr_slave_lock); | |
2779 | return; | |
2780 | } | |
1da177e4 | 2781 | |
b2220cad | 2782 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 2783 | |
b2220cad JV |
2784 | /* if we don't have a curr_active_slave, search for the next available |
2785 | * backup slave from the current_arp_slave and make it the candidate | |
2786 | * for becoming the curr_active_slave | |
2787 | */ | |
1da177e4 | 2788 | |
b2220cad | 2789 | if (!bond->current_arp_slave) { |
dec1e90e | 2790 | bond->current_arp_slave = bond_first_slave(bond); |
b2220cad JV |
2791 | if (!bond->current_arp_slave) |
2792 | return; | |
2793 | } | |
1da177e4 | 2794 | |
b2220cad | 2795 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 2796 | |
b2220cad | 2797 | /* search for next candidate */ |
dec1e90e | 2798 | next_slave = bond_next_slave(bond, bond->current_arp_slave); |
2799 | bond_for_each_slave_from(bond, slave, i, next_slave) { | |
b2220cad JV |
2800 | if (IS_UP(slave->dev)) { |
2801 | slave->link = BOND_LINK_BACK; | |
2802 | bond_set_slave_active_flags(slave); | |
2803 | bond_arp_send_all(bond, slave); | |
1da177e4 | 2804 | slave->jiffies = jiffies; |
b2220cad JV |
2805 | bond->current_arp_slave = slave; |
2806 | break; | |
1da177e4 LT |
2807 | } |
2808 | ||
b2220cad JV |
2809 | /* if the link state is up at this point, we |
2810 | * mark it down - this can happen if we have | |
2811 | * simultaneous link failures and | |
2812 | * reselect_active_interface doesn't make this | |
2813 | * one the current slave so it is still marked | |
2814 | * up when it is actually down | |
1da177e4 | 2815 | */ |
b2220cad JV |
2816 | if (slave->link == BOND_LINK_UP) { |
2817 | slave->link = BOND_LINK_DOWN; | |
2818 | if (slave->link_failure_count < UINT_MAX) | |
2819 | slave->link_failure_count++; | |
1da177e4 | 2820 | |
b2220cad JV |
2821 | bond_set_slave_inactive_flags(slave); |
2822 | ||
a4aee5c8 JP |
2823 | pr_info("%s: backup interface %s is now down.\n", |
2824 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2825 | } |
b2220cad JV |
2826 | } |
2827 | } | |
1da177e4 | 2828 | |
b2220cad JV |
2829 | void bond_activebackup_arp_mon(struct work_struct *work) |
2830 | { | |
2831 | struct bonding *bond = container_of(work, struct bonding, | |
2832 | arp_work.work); | |
ad246c99 | 2833 | bool should_notify_peers = false; |
b2220cad | 2834 | int delta_in_ticks; |
1da177e4 | 2835 | |
b2220cad | 2836 | read_lock(&bond->lock); |
1da177e4 | 2837 | |
b2220cad | 2838 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2839 | |
dec1e90e | 2840 | if (list_empty(&bond->slave_list)) |
b2220cad JV |
2841 | goto re_arm; |
2842 | ||
ad246c99 BH |
2843 | should_notify_peers = bond_should_notify_peers(bond); |
2844 | ||
e7f63f1d | 2845 | if (bond_ab_arp_inspect(bond)) { |
b2220cad | 2846 | read_unlock(&bond->lock); |
e6d265e8 JV |
2847 | |
2848 | /* Race avoidance with bond_close flush of workqueue */ | |
2849 | if (!rtnl_trylock()) { | |
2850 | read_lock(&bond->lock); | |
2851 | delta_in_ticks = 1; | |
2852 | should_notify_peers = false; | |
2853 | goto re_arm; | |
2854 | } | |
2855 | ||
b2220cad JV |
2856 | read_lock(&bond->lock); |
2857 | ||
e7f63f1d | 2858 | bond_ab_arp_commit(bond); |
b2220cad JV |
2859 | |
2860 | read_unlock(&bond->lock); | |
2861 | rtnl_unlock(); | |
2862 | read_lock(&bond->lock); | |
1da177e4 LT |
2863 | } |
2864 | ||
b2220cad JV |
2865 | bond_ab_arp_probe(bond); |
2866 | ||
1da177e4 | 2867 | re_arm: |
e6d265e8 | 2868 | if (bond->params.arp_interval) |
1b76b316 | 2869 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 2870 | |
1da177e4 | 2871 | read_unlock(&bond->lock); |
ad246c99 BH |
2872 | |
2873 | if (should_notify_peers) { | |
6c8d23f7 | 2874 | if (!rtnl_trylock()) |
e6d265e8 | 2875 | return; |
b7bc2a5b | 2876 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2877 | rtnl_unlock(); |
2878 | } | |
1da177e4 LT |
2879 | } |
2880 | ||
1da177e4 LT |
2881 | /*-------------------------- netdev event handling --------------------------*/ |
2882 | ||
2883 | /* | |
2884 | * Change device name | |
2885 | */ | |
2886 | static int bond_event_changename(struct bonding *bond) | |
2887 | { | |
1da177e4 LT |
2888 | bond_remove_proc_entry(bond); |
2889 | bond_create_proc_entry(bond); | |
7e083840 | 2890 | |
f073c7ca TI |
2891 | bond_debug_reregister(bond); |
2892 | ||
1da177e4 LT |
2893 | return NOTIFY_DONE; |
2894 | } | |
2895 | ||
3d632c3f SH |
2896 | static int bond_master_netdev_event(unsigned long event, |
2897 | struct net_device *bond_dev) | |
1da177e4 | 2898 | { |
454d7c9b | 2899 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2900 | |
2901 | switch (event) { | |
2902 | case NETDEV_CHANGENAME: | |
2903 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2904 | case NETDEV_UNREGISTER: |
2905 | bond_remove_proc_entry(event_bond); | |
2906 | break; | |
2907 | case NETDEV_REGISTER: | |
2908 | bond_create_proc_entry(event_bond); | |
2909 | break; | |
6c8d23f7 | 2910 | case NETDEV_NOTIFY_PEERS: |
2911 | if (event_bond->send_peer_notif) | |
2912 | event_bond->send_peer_notif--; | |
2913 | break; | |
1da177e4 LT |
2914 | default: |
2915 | break; | |
2916 | } | |
2917 | ||
2918 | return NOTIFY_DONE; | |
2919 | } | |
2920 | ||
3d632c3f SH |
2921 | static int bond_slave_netdev_event(unsigned long event, |
2922 | struct net_device *slave_dev) | |
1da177e4 | 2923 | { |
471cb5a3 | 2924 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
6101391d | 2925 | struct bonding *bond; |
2926 | struct net_device *bond_dev; | |
471cb5a3 JP |
2927 | u32 old_speed; |
2928 | u8 old_duplex; | |
1da177e4 | 2929 | |
6101391d | 2930 | /* A netdev event can be generated while enslaving a device |
2931 | * before netdev_rx_handler_register is called in which case | |
2932 | * slave will be NULL | |
2933 | */ | |
2934 | if (!slave) | |
2935 | return NOTIFY_DONE; | |
2936 | bond_dev = slave->bond->dev; | |
2937 | bond = slave->bond; | |
2938 | ||
1da177e4 LT |
2939 | switch (event) { |
2940 | case NETDEV_UNREGISTER: | |
8d2ada77 | 2941 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
2942 | bond_release_and_destroy(bond_dev, slave_dev); |
2943 | else | |
2944 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 2945 | break; |
98f41f69 | 2946 | case NETDEV_UP: |
1da177e4 | 2947 | case NETDEV_CHANGE: |
471cb5a3 JP |
2948 | old_speed = slave->speed; |
2949 | old_duplex = slave->duplex; | |
17d04500 | 2950 | |
471cb5a3 | 2951 | bond_update_speed_duplex(slave); |
17d04500 | 2952 | |
471cb5a3 JP |
2953 | if (bond->params.mode == BOND_MODE_8023AD) { |
2954 | if (old_speed != slave->speed) | |
2955 | bond_3ad_adapter_speed_changed(slave); | |
2956 | if (old_duplex != slave->duplex) | |
2957 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 2958 | } |
1da177e4 LT |
2959 | break; |
2960 | case NETDEV_DOWN: | |
2961 | /* | |
2962 | * ... Or is it this? | |
2963 | */ | |
2964 | break; | |
2965 | case NETDEV_CHANGEMTU: | |
2966 | /* | |
2967 | * TODO: Should slaves be allowed to | |
2968 | * independently alter their MTU? For | |
2969 | * an active-backup bond, slaves need | |
2970 | * not be the same type of device, so | |
2971 | * MTUs may vary. For other modes, | |
2972 | * slaves arguably should have the | |
2973 | * same MTUs. To do this, we'd need to | |
2974 | * take over the slave's change_mtu | |
2975 | * function for the duration of their | |
2976 | * servitude. | |
2977 | */ | |
2978 | break; | |
2979 | case NETDEV_CHANGENAME: | |
2980 | /* | |
2981 | * TODO: handle changing the primary's name | |
2982 | */ | |
2983 | break; | |
8531c5ff AK |
2984 | case NETDEV_FEAT_CHANGE: |
2985 | bond_compute_features(bond); | |
2986 | break; | |
4aa5dee4 JP |
2987 | case NETDEV_RESEND_IGMP: |
2988 | /* Propagate to master device */ | |
2989 | call_netdevice_notifiers(event, slave->bond->dev); | |
2990 | break; | |
1da177e4 LT |
2991 | default: |
2992 | break; | |
2993 | } | |
2994 | ||
2995 | return NOTIFY_DONE; | |
2996 | } | |
2997 | ||
2998 | /* | |
2999 | * bond_netdev_event: handle netdev notifier chain events. | |
3000 | * | |
3001 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3002 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3003 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3004 | * dev_probe_lock). | |
3005 | */ | |
3d632c3f SH |
3006 | static int bond_netdev_event(struct notifier_block *this, |
3007 | unsigned long event, void *ptr) | |
1da177e4 | 3008 | { |
351638e7 | 3009 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3010 | |
5a03cdb7 | 3011 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3012 | event_dev ? event_dev->name : "None", |
3013 | event); | |
1da177e4 | 3014 | |
0b680e75 JV |
3015 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3016 | return NOTIFY_DONE; | |
3017 | ||
1da177e4 | 3018 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3019 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3020 | return bond_master_netdev_event(event, event_dev); |
3021 | } | |
3022 | ||
3023 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3024 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3025 | return bond_slave_netdev_event(event, event_dev); |
3026 | } | |
3027 | ||
3028 | return NOTIFY_DONE; | |
3029 | } | |
3030 | ||
3031 | static struct notifier_block bond_netdev_notifier = { | |
3032 | .notifier_call = bond_netdev_event, | |
3033 | }; | |
3034 | ||
169a3e66 JV |
3035 | /*---------------------------- Hashing Policies -----------------------------*/ |
3036 | ||
6b923cb7 JE |
3037 | /* |
3038 | * Hash for the output device based upon layer 2 data | |
3039 | */ | |
3040 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) | |
3041 | { | |
3042 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3043 | ||
3044 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
3045 | return (data->h_dest[5] ^ data->h_source[5]) % count; | |
3046 | ||
3047 | return 0; | |
3048 | } | |
3049 | ||
6f6652be JV |
3050 | /* |
3051 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
6b923cb7 | 3052 | * the packet is not IP, fall back on bond_xmit_hash_policy_l2() |
6f6652be | 3053 | */ |
a361c83c | 3054 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be | 3055 | { |
4394542c ED |
3056 | const struct ethhdr *data; |
3057 | const struct iphdr *iph; | |
3058 | const struct ipv6hdr *ipv6h; | |
6b923cb7 | 3059 | u32 v6hash; |
4394542c | 3060 | const __be32 *s, *d; |
6b923cb7 JE |
3061 | |
3062 | if (skb->protocol == htons(ETH_P_IP) && | |
4394542c | 3063 | pskb_network_may_pull(skb, sizeof(*iph))) { |
6b923cb7 | 3064 | iph = ip_hdr(skb); |
4394542c | 3065 | data = (struct ethhdr *)skb->data; |
6f6652be | 3066 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3067 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6b923cb7 | 3068 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3069 | pskb_network_may_pull(skb, sizeof(*ipv6h))) { |
6b923cb7 | 3070 | ipv6h = ipv6_hdr(skb); |
4394542c | 3071 | data = (struct ethhdr *)skb->data; |
6b923cb7 JE |
3072 | s = &ipv6h->saddr.s6_addr32[0]; |
3073 | d = &ipv6h->daddr.s6_addr32[0]; | |
3074 | v6hash = (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3075 | v6hash ^= (v6hash >> 24) ^ (v6hash >> 16) ^ (v6hash >> 8); | |
3076 | return (v6hash ^ data->h_dest[5] ^ data->h_source[5]) % count; | |
6f6652be JV |
3077 | } |
3078 | ||
6b923cb7 | 3079 | return bond_xmit_hash_policy_l2(skb, count); |
6f6652be JV |
3080 | } |
3081 | ||
169a3e66 | 3082 | /* |
59c51591 | 3083 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 | 3084 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
6b923cb7 | 3085 | * altogether not IP, fall back on bond_xmit_hash_policy_l2() |
169a3e66 | 3086 | */ |
a361c83c | 3087 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 | 3088 | { |
6b923cb7 | 3089 | u32 layer4_xor = 0; |
4394542c ED |
3090 | const struct iphdr *iph; |
3091 | const struct ipv6hdr *ipv6h; | |
3092 | const __be32 *s, *d; | |
3093 | const __be16 *l4 = NULL; | |
3094 | __be16 _l4[2]; | |
3095 | int noff = skb_network_offset(skb); | |
3096 | int poff; | |
169a3e66 | 3097 | |
6b923cb7 | 3098 | if (skb->protocol == htons(ETH_P_IP) && |
4394542c | 3099 | pskb_may_pull(skb, noff + sizeof(*iph))) { |
6b923cb7 | 3100 | iph = ip_hdr(skb); |
4394542c ED |
3101 | poff = proto_ports_offset(iph->protocol); |
3102 | ||
3103 | if (!ip_is_fragment(iph) && poff >= 0) { | |
3104 | l4 = skb_header_pointer(skb, noff + (iph->ihl << 2) + poff, | |
3105 | sizeof(_l4), &_l4); | |
3106 | if (l4) | |
3107 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
169a3e66 JV |
3108 | } |
3109 | return (layer4_xor ^ | |
3110 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
6b923cb7 | 3111 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3112 | pskb_may_pull(skb, noff + sizeof(*ipv6h))) { |
6b923cb7 | 3113 | ipv6h = ipv6_hdr(skb); |
4394542c ED |
3114 | poff = proto_ports_offset(ipv6h->nexthdr); |
3115 | if (poff >= 0) { | |
3116 | l4 = skb_header_pointer(skb, noff + sizeof(*ipv6h) + poff, | |
3117 | sizeof(_l4), &_l4); | |
3118 | if (l4) | |
3119 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
6b923cb7 JE |
3120 | } |
3121 | s = &ipv6h->saddr.s6_addr32[0]; | |
3122 | d = &ipv6h->daddr.s6_addr32[0]; | |
3123 | layer4_xor ^= (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3124 | layer4_xor ^= (layer4_xor >> 24) ^ (layer4_xor >> 16) ^ | |
3125 | (layer4_xor >> 8); | |
3126 | return layer4_xor % count; | |
169a3e66 JV |
3127 | } |
3128 | ||
6b923cb7 | 3129 | return bond_xmit_hash_policy_l2(skb, count); |
169a3e66 JV |
3130 | } |
3131 | ||
1da177e4 LT |
3132 | /*-------------------------- Device entry points ----------------------------*/ |
3133 | ||
fbb0c41b | 3134 | static void bond_work_init_all(struct bonding *bond) |
3135 | { | |
3136 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3137 | bond_resend_igmp_join_requests_delayed); | |
3138 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3139 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3140 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3141 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3142 | else | |
3143 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3144 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3145 | } | |
3146 | ||
3147 | static void bond_work_cancel_all(struct bonding *bond) | |
3148 | { | |
3149 | cancel_delayed_work_sync(&bond->mii_work); | |
3150 | cancel_delayed_work_sync(&bond->arp_work); | |
3151 | cancel_delayed_work_sync(&bond->alb_work); | |
3152 | cancel_delayed_work_sync(&bond->ad_work); | |
3153 | cancel_delayed_work_sync(&bond->mcast_work); | |
3154 | } | |
3155 | ||
1da177e4 LT |
3156 | static int bond_open(struct net_device *bond_dev) |
3157 | { | |
454d7c9b | 3158 | struct bonding *bond = netdev_priv(bond_dev); |
ba3211cc | 3159 | struct slave *slave; |
1da177e4 | 3160 | |
ba3211cc PP |
3161 | /* reset slave->backup and slave->inactive */ |
3162 | read_lock(&bond->lock); | |
dec1e90e | 3163 | if (!list_empty(&bond->slave_list)) { |
ba3211cc | 3164 | read_lock(&bond->curr_slave_lock); |
dec1e90e | 3165 | bond_for_each_slave(bond, slave) { |
ba3211cc PP |
3166 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3167 | && (slave != bond->curr_active_slave)) { | |
3168 | bond_set_slave_inactive_flags(slave); | |
3169 | } else { | |
3170 | bond_set_slave_active_flags(slave); | |
3171 | } | |
3172 | } | |
3173 | read_unlock(&bond->curr_slave_lock); | |
3174 | } | |
3175 | read_unlock(&bond->lock); | |
3176 | ||
fbb0c41b | 3177 | bond_work_init_all(bond); |
5a37e8ca | 3178 | |
58402054 | 3179 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3180 | /* bond_alb_initialize must be called before the timer |
3181 | * is started. | |
3182 | */ | |
fbb0c41b | 3183 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3184 | return -ENOMEM; |
1b76b316 | 3185 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3186 | } |
3187 | ||
fbb0c41b | 3188 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3189 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3190 | |
3191 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3192 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
f5b2b966 | 3193 | if (bond->params.arp_validate) |
3aba891d | 3194 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3195 | } |
3196 | ||
3197 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3198 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3199 | /* register to receive LACPDUs */ |
3aba891d | 3200 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3201 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3202 | } |
3203 | ||
3204 | return 0; | |
3205 | } | |
3206 | ||
3207 | static int bond_close(struct net_device *bond_dev) | |
3208 | { | |
454d7c9b | 3209 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3210 | |
fbb0c41b | 3211 | bond_work_cancel_all(bond); |
6c8d23f7 | 3212 | bond->send_peer_notif = 0; |
ee8487c0 | 3213 | if (bond_is_lb(bond)) |
1da177e4 | 3214 | bond_alb_deinitialize(bond); |
3aba891d | 3215 | bond->recv_probe = NULL; |
1da177e4 LT |
3216 | |
3217 | return 0; | |
3218 | } | |
3219 | ||
28172739 ED |
3220 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3221 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3222 | { |
454d7c9b | 3223 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3224 | struct rtnl_link_stats64 temp; |
1da177e4 | 3225 | struct slave *slave; |
1da177e4 | 3226 | |
28172739 | 3227 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3228 | |
3229 | read_lock_bh(&bond->lock); | |
dec1e90e | 3230 | bond_for_each_slave(bond, slave) { |
be1f3c2c | 3231 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3232 | dev_get_stats(slave->dev, &temp); |
3233 | ||
3234 | stats->rx_packets += sstats->rx_packets; | |
3235 | stats->rx_bytes += sstats->rx_bytes; | |
3236 | stats->rx_errors += sstats->rx_errors; | |
3237 | stats->rx_dropped += sstats->rx_dropped; | |
3238 | ||
3239 | stats->tx_packets += sstats->tx_packets; | |
3240 | stats->tx_bytes += sstats->tx_bytes; | |
3241 | stats->tx_errors += sstats->tx_errors; | |
3242 | stats->tx_dropped += sstats->tx_dropped; | |
3243 | ||
3244 | stats->multicast += sstats->multicast; | |
3245 | stats->collisions += sstats->collisions; | |
3246 | ||
3247 | stats->rx_length_errors += sstats->rx_length_errors; | |
3248 | stats->rx_over_errors += sstats->rx_over_errors; | |
3249 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3250 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3251 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3252 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3253 | ||
3254 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3255 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3256 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3257 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3258 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb | 3259 | } |
1da177e4 LT |
3260 | read_unlock_bh(&bond->lock); |
3261 | ||
3262 | return stats; | |
3263 | } | |
3264 | ||
3265 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3266 | { | |
3267 | struct net_device *slave_dev = NULL; | |
3268 | struct ifbond k_binfo; | |
3269 | struct ifbond __user *u_binfo = NULL; | |
3270 | struct ifslave k_sinfo; | |
3271 | struct ifslave __user *u_sinfo = NULL; | |
3272 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3273 | struct net *net; |
1da177e4 LT |
3274 | int res = 0; |
3275 | ||
a4aee5c8 | 3276 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3277 | |
3278 | switch (cmd) { | |
1da177e4 LT |
3279 | case SIOCGMIIPHY: |
3280 | mii = if_mii(ifr); | |
3d632c3f | 3281 | if (!mii) |
1da177e4 | 3282 | return -EINVAL; |
3d632c3f | 3283 | |
1da177e4 LT |
3284 | mii->phy_id = 0; |
3285 | /* Fall Through */ | |
3286 | case SIOCGMIIREG: | |
3287 | /* | |
3288 | * We do this again just in case we were called by SIOCGMIIREG | |
3289 | * instead of SIOCGMIIPHY. | |
3290 | */ | |
3291 | mii = if_mii(ifr); | |
3d632c3f | 3292 | if (!mii) |
1da177e4 | 3293 | return -EINVAL; |
3d632c3f | 3294 | |
1da177e4 LT |
3295 | |
3296 | if (mii->reg_num == 1) { | |
454d7c9b | 3297 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3298 | mii->val_out = 0; |
6603a6f2 | 3299 | read_lock(&bond->lock); |
1da177e4 | 3300 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3301 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3302 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3303 | |
1da177e4 | 3304 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3305 | read_unlock(&bond->lock); |
1da177e4 LT |
3306 | } |
3307 | ||
3308 | return 0; | |
3309 | case BOND_INFO_QUERY_OLD: | |
3310 | case SIOCBONDINFOQUERY: | |
3311 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3312 | ||
3d632c3f | 3313 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3314 | return -EFAULT; |
1da177e4 LT |
3315 | |
3316 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3317 | if (res == 0 && |
3318 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3319 | return -EFAULT; | |
1da177e4 LT |
3320 | |
3321 | return res; | |
3322 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3323 | case SIOCBONDSLAVEINFOQUERY: | |
3324 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3325 | ||
3d632c3f | 3326 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3327 | return -EFAULT; |
1da177e4 LT |
3328 | |
3329 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3330 | if (res == 0 && |
3331 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3332 | return -EFAULT; | |
1da177e4 LT |
3333 | |
3334 | return res; | |
3335 | default: | |
3336 | /* Go on */ | |
3337 | break; | |
3338 | } | |
3339 | ||
387ff911 G |
3340 | net = dev_net(bond_dev); |
3341 | ||
3342 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3343 | return -EPERM; |
1da177e4 | 3344 | |
387ff911 | 3345 | slave_dev = dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3346 | |
a4aee5c8 | 3347 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3348 | |
3d632c3f | 3349 | if (!slave_dev) |
1da177e4 | 3350 | res = -ENODEV; |
3d632c3f | 3351 | else { |
a4aee5c8 | 3352 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3353 | switch (cmd) { |
3354 | case BOND_ENSLAVE_OLD: | |
3355 | case SIOCBONDENSLAVE: | |
3356 | res = bond_enslave(bond_dev, slave_dev); | |
3357 | break; | |
3358 | case BOND_RELEASE_OLD: | |
3359 | case SIOCBONDRELEASE: | |
3360 | res = bond_release(bond_dev, slave_dev); | |
3361 | break; | |
3362 | case BOND_SETHWADDR_OLD: | |
3363 | case SIOCBONDSETHWADDR: | |
409cc1f8 JP |
3364 | bond_set_dev_addr(bond_dev, slave_dev); |
3365 | res = 0; | |
1da177e4 LT |
3366 | break; |
3367 | case BOND_CHANGE_ACTIVE_OLD: | |
3368 | case SIOCBONDCHANGEACTIVE: | |
3369 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3370 | break; | |
3371 | default: | |
3372 | res = -EOPNOTSUPP; | |
3373 | } | |
3374 | ||
3375 | dev_put(slave_dev); | |
3376 | } | |
3377 | ||
1da177e4 LT |
3378 | return res; |
3379 | } | |
3380 | ||
d03462b9 | 3381 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3382 | { |
454d7c9b | 3383 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3384 | |
d03462b9 JP |
3385 | if (change & IFF_PROMISC) |
3386 | bond_set_promiscuity(bond, | |
3387 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3388 | |
d03462b9 JP |
3389 | if (change & IFF_ALLMULTI) |
3390 | bond_set_allmulti(bond, | |
3391 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3392 | } | |
1da177e4 | 3393 | |
303d1cbf | 3394 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3395 | { |
3396 | struct bonding *bond = netdev_priv(bond_dev); | |
303d1cbf | 3397 | struct slave *slave; |
1da177e4 | 3398 | |
7864a1ad | 3399 | ASSERT_RTNL(); |
80ee5ad2 | 3400 | |
303d1cbf | 3401 | if (USES_PRIMARY(bond->params.mode)) { |
7864a1ad | 3402 | slave = rtnl_dereference(bond->curr_active_slave); |
303d1cbf JV |
3403 | if (slave) { |
3404 | dev_uc_sync(slave->dev, bond_dev); | |
3405 | dev_mc_sync(slave->dev, bond_dev); | |
3406 | } | |
303d1cbf | 3407 | } else { |
dec1e90e | 3408 | bond_for_each_slave(bond, slave) { |
303d1cbf JV |
3409 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3410 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3411 | } | |
1da177e4 | 3412 | } |
1da177e4 LT |
3413 | } |
3414 | ||
234bcf8a | 3415 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3416 | { |
234bcf8a | 3417 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3418 | const struct net_device_ops *slave_ops; |
3419 | struct neigh_parms parms; | |
dec1e90e | 3420 | struct slave *slave; |
234bcf8a SP |
3421 | int ret; |
3422 | ||
dec1e90e | 3423 | slave = bond_first_slave(bond); |
234bcf8a SP |
3424 | if (!slave) |
3425 | return 0; | |
234bcf8a | 3426 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3427 | if (!slave_ops->ndo_neigh_setup) |
3428 | return 0; | |
3429 | ||
3430 | parms.neigh_setup = NULL; | |
3431 | parms.neigh_cleanup = NULL; | |
3432 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3433 | if (ret) | |
3434 | return ret; | |
3435 | ||
3436 | /* | |
3437 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3438 | * after the last slave has been detached. Assumes that all slaves | |
3439 | * utilize the same neigh_cleanup (true at this writing as only user | |
3440 | * is ipoib). | |
3441 | */ | |
3442 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3443 | ||
3444 | if (!parms.neigh_setup) | |
3445 | return 0; | |
3446 | ||
3447 | return parms.neigh_setup(n); | |
3448 | } | |
3449 | ||
3450 | /* | |
3451 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3452 | * slave exists. So we must declare proxy setup function which will | |
3453 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3454 | * |
3455 | * It's also called by master devices (such as vlans) to setup their | |
3456 | * underlying devices. In that case - do nothing, we're already set up from | |
3457 | * our init. | |
234bcf8a SP |
3458 | */ |
3459 | static int bond_neigh_setup(struct net_device *dev, | |
3460 | struct neigh_parms *parms) | |
3461 | { | |
9918d5bf VF |
3462 | /* modify only our neigh_parms */ |
3463 | if (parms->dev == dev) | |
3464 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3465 | |
00829823 SH |
3466 | return 0; |
3467 | } | |
3468 | ||
1da177e4 LT |
3469 | /* |
3470 | * Change the MTU of all of a master's slaves to match the master | |
3471 | */ | |
3472 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3473 | { | |
454d7c9b | 3474 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3475 | struct slave *slave, *rollback_slave; |
1da177e4 | 3476 | int res = 0; |
1da177e4 | 3477 | |
5a03cdb7 | 3478 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3479 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3480 | |
3481 | /* Can't hold bond->lock with bh disabled here since | |
3482 | * some base drivers panic. On the other hand we can't | |
3483 | * hold bond->lock without bh disabled because we'll | |
3484 | * deadlock. The only solution is to rely on the fact | |
3485 | * that we're under rtnl_lock here, and the slaves | |
3486 | * list won't change. This doesn't solve the problem | |
3487 | * of setting the slave's MTU while it is | |
3488 | * transmitting, but the assumption is that the base | |
3489 | * driver can handle that. | |
3490 | * | |
3491 | * TODO: figure out a way to safely iterate the slaves | |
3492 | * list, but without holding a lock around the actual | |
3493 | * call to the base driver. | |
3494 | */ | |
3495 | ||
dec1e90e | 3496 | bond_for_each_slave(bond, slave) { |
a4aee5c8 JP |
3497 | pr_debug("s %p s->p %p c_m %p\n", |
3498 | slave, | |
dec1e90e | 3499 | bond_prev_slave(bond, slave), |
a4aee5c8 | 3500 | slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 3501 | |
1da177e4 LT |
3502 | res = dev_set_mtu(slave->dev, new_mtu); |
3503 | ||
3504 | if (res) { | |
3505 | /* If we failed to set the slave's mtu to the new value | |
3506 | * we must abort the operation even in ACTIVE_BACKUP | |
3507 | * mode, because if we allow the backup slaves to have | |
3508 | * different mtu values than the active slave we'll | |
3509 | * need to change their mtu when doing a failover. That | |
3510 | * means changing their mtu from timer context, which | |
3511 | * is probably not a good idea. | |
3512 | */ | |
5a03cdb7 | 3513 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3514 | goto unwind; |
3515 | } | |
3516 | } | |
3517 | ||
3518 | bond_dev->mtu = new_mtu; | |
3519 | ||
3520 | return 0; | |
3521 | ||
3522 | unwind: | |
3523 | /* unwind from head to the slave that failed */ | |
81f23b13 | 3524 | bond_for_each_slave(bond, rollback_slave) { |
1da177e4 LT |
3525 | int tmp_res; |
3526 | ||
81f23b13 VF |
3527 | if (rollback_slave == slave) |
3528 | break; | |
3529 | ||
3530 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3531 | if (tmp_res) { |
a4aee5c8 | 3532 | pr_debug("unwind err %d dev %s\n", |
81f23b13 | 3533 | tmp_res, rollback_slave->dev->name); |
1da177e4 LT |
3534 | } |
3535 | } | |
3536 | ||
3537 | return res; | |
3538 | } | |
3539 | ||
3540 | /* | |
3541 | * Change HW address | |
3542 | * | |
3543 | * Note that many devices must be down to change the HW address, and | |
3544 | * downing the master releases all slaves. We can make bonds full of | |
3545 | * bonding devices to test this, however. | |
3546 | */ | |
3547 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3548 | { | |
454d7c9b | 3549 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3550 | struct slave *slave, *rollback_slave; |
1da177e4 | 3551 | struct sockaddr *sa = addr, tmp_sa; |
1da177e4 | 3552 | int res = 0; |
1da177e4 | 3553 | |
eb7cc59a SH |
3554 | if (bond->params.mode == BOND_MODE_ALB) |
3555 | return bond_alb_set_mac_address(bond_dev, addr); | |
3556 | ||
3557 | ||
a4aee5c8 JP |
3558 | pr_debug("bond=%p, name=%s\n", |
3559 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3560 | |
1b5acd29 JV |
3561 | /* If fail_over_mac is enabled, do nothing and return success. |
3562 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3563 | */ |
1b5acd29 | 3564 | if (bond->params.fail_over_mac) |
dd957c57 | 3565 | return 0; |
2ab82852 | 3566 | |
3d632c3f | 3567 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3568 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3569 | |
3570 | /* Can't hold bond->lock with bh disabled here since | |
3571 | * some base drivers panic. On the other hand we can't | |
3572 | * hold bond->lock without bh disabled because we'll | |
3573 | * deadlock. The only solution is to rely on the fact | |
3574 | * that we're under rtnl_lock here, and the slaves | |
3575 | * list won't change. This doesn't solve the problem | |
3576 | * of setting the slave's hw address while it is | |
3577 | * transmitting, but the assumption is that the base | |
3578 | * driver can handle that. | |
3579 | * | |
3580 | * TODO: figure out a way to safely iterate the slaves | |
3581 | * list, but without holding a lock around the actual | |
3582 | * call to the base driver. | |
3583 | */ | |
3584 | ||
dec1e90e | 3585 | bond_for_each_slave(bond, slave) { |
eb7cc59a | 3586 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3587 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3588 | |
eb7cc59a | 3589 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3590 | res = -EOPNOTSUPP; |
5a03cdb7 | 3591 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3592 | goto unwind; |
3593 | } | |
3594 | ||
3595 | res = dev_set_mac_address(slave->dev, addr); | |
3596 | if (res) { | |
3597 | /* TODO: consider downing the slave | |
3598 | * and retry ? | |
3599 | * User should expect communications | |
3600 | * breakage anyway until ARP finish | |
3601 | * updating, so... | |
3602 | */ | |
5a03cdb7 | 3603 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3604 | goto unwind; |
3605 | } | |
3606 | } | |
3607 | ||
3608 | /* success */ | |
3609 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3610 | return 0; | |
3611 | ||
3612 | unwind: | |
3613 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3614 | tmp_sa.sa_family = bond_dev->type; | |
3615 | ||
3616 | /* unwind from head to the slave that failed */ | |
81f23b13 | 3617 | bond_for_each_slave(bond, rollback_slave) { |
1da177e4 LT |
3618 | int tmp_res; |
3619 | ||
81f23b13 VF |
3620 | if (rollback_slave == slave) |
3621 | break; | |
3622 | ||
3623 | tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa); | |
1da177e4 | 3624 | if (tmp_res) { |
a4aee5c8 | 3625 | pr_debug("unwind err %d dev %s\n", |
81f23b13 | 3626 | tmp_res, rollback_slave->dev->name); |
1da177e4 LT |
3627 | } |
3628 | } | |
3629 | ||
3630 | return res; | |
3631 | } | |
3632 | ||
15077228 NA |
3633 | /** |
3634 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3635 | * @bond: bonding device that is transmitting | |
3636 | * @skb: buffer to transmit | |
3637 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3638 | * | |
3639 | * This function tries to transmit through slave with slave_id but in case | |
3640 | * it fails, it tries to find the first available slave for transmission. | |
3641 | * The skb is consumed in all cases, thus the function is void. | |
3642 | */ | |
3643 | void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) | |
3644 | { | |
3645 | struct slave *slave; | |
3646 | int i = slave_id; | |
3647 | ||
3648 | /* Here we start from the slave with slave_id */ | |
278b2083 | 3649 | bond_for_each_slave_rcu(bond, slave) { |
15077228 NA |
3650 | if (--i < 0) { |
3651 | if (slave_can_tx(slave)) { | |
3652 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3653 | return; | |
3654 | } | |
3655 | } | |
3656 | } | |
3657 | ||
3658 | /* Here we start from the first slave up to slave_id */ | |
3659 | i = slave_id; | |
278b2083 | 3660 | bond_for_each_slave_rcu(bond, slave) { |
15077228 NA |
3661 | if (--i < 0) |
3662 | break; | |
3663 | if (slave_can_tx(slave)) { | |
3664 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3665 | return; | |
3666 | } | |
3667 | } | |
3668 | /* no slave that can tx has been found */ | |
3669 | kfree_skb(skb); | |
3670 | } | |
3671 | ||
1da177e4 LT |
3672 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3673 | { | |
454d7c9b | 3674 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3675 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3676 | struct slave *slave; |
1da177e4 | 3677 | |
cf5f9044 | 3678 | /* |
a2fd940f AG |
3679 | * Start with the curr_active_slave that joined the bond as the |
3680 | * default for sending IGMP traffic. For failover purposes one | |
3681 | * needs to maintain some consistency for the interface that will | |
3682 | * send the join/membership reports. The curr_active_slave found | |
3683 | * will send all of this type of traffic. | |
cf5f9044 | 3684 | */ |
15077228 | 3685 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3686 | slave = rcu_dereference(bond->curr_active_slave); |
15077228 NA |
3687 | if (slave && slave_can_tx(slave)) |
3688 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3689 | else | |
3690 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3691 | } else { |
15077228 NA |
3692 | bond_xmit_slave_id(bond, skb, |
3693 | bond->rr_tx_counter++ % bond->slave_cnt); | |
1da177e4 | 3694 | } |
0693e88e | 3695 | |
ec634fe3 | 3696 | return NETDEV_TX_OK; |
1da177e4 LT |
3697 | } |
3698 | ||
3699 | /* | |
3700 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3701 | * the bond has a usable interface. | |
3702 | */ | |
3703 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3704 | { | |
454d7c9b | 3705 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3706 | struct slave *slave; |
1da177e4 | 3707 | |
278b2083 | 3708 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3709 | if (slave) |
15077228 NA |
3710 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3711 | else | |
04502430 | 3712 | kfree_skb(skb); |
0693e88e | 3713 | |
ec634fe3 | 3714 | return NETDEV_TX_OK; |
1da177e4 LT |
3715 | } |
3716 | ||
3717 | /* | |
169a3e66 JV |
3718 | * In bond_xmit_xor() , we determine the output device by using a pre- |
3719 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
3720 | * find the next active slave. | |
1da177e4 LT |
3721 | */ |
3722 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3723 | { | |
454d7c9b | 3724 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3725 | |
15077228 NA |
3726 | bond_xmit_slave_id(bond, skb, |
3727 | bond->xmit_hash_policy(skb, bond->slave_cnt)); | |
0693e88e | 3728 | |
ec634fe3 | 3729 | return NETDEV_TX_OK; |
1da177e4 LT |
3730 | } |
3731 | ||
78a646ce | 3732 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3733 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3734 | { | |
454d7c9b | 3735 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3736 | struct slave *slave = NULL; |
1da177e4 | 3737 | |
278b2083 | 3738 | bond_for_each_slave_rcu(bond, slave) { |
78a646ce NA |
3739 | if (bond_is_last_slave(bond, slave)) |
3740 | break; | |
3741 | if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { | |
3742 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1da177e4 | 3743 | |
78a646ce NA |
3744 | if (!skb2) { |
3745 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", | |
3746 | bond_dev->name); | |
3747 | continue; | |
1da177e4 | 3748 | } |
78a646ce NA |
3749 | /* bond_dev_queue_xmit always returns 0 */ |
3750 | bond_dev_queue_xmit(bond, skb2, slave->dev); | |
1da177e4 LT |
3751 | } |
3752 | } | |
78a646ce NA |
3753 | if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP) |
3754 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3755 | else | |
04502430 | 3756 | kfree_skb(skb); |
3d632c3f | 3757 | |
ec634fe3 | 3758 | return NETDEV_TX_OK; |
1da177e4 LT |
3759 | } |
3760 | ||
3761 | /*------------------------- Device initialization ---------------------------*/ | |
3762 | ||
6f6652be JV |
3763 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
3764 | { | |
3765 | switch (bond->params.xmit_policy) { | |
3766 | case BOND_XMIT_POLICY_LAYER23: | |
3767 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
3768 | break; | |
3769 | case BOND_XMIT_POLICY_LAYER34: | |
3770 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
3771 | break; | |
3772 | case BOND_XMIT_POLICY_LAYER2: | |
3773 | default: | |
3774 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
3775 | break; | |
3776 | } | |
3777 | } | |
3778 | ||
bb1d9123 AG |
3779 | /* |
3780 | * Lookup the slave that corresponds to a qid | |
3781 | */ | |
3782 | static inline int bond_slave_override(struct bonding *bond, | |
3783 | struct sk_buff *skb) | |
3784 | { | |
bb1d9123 AG |
3785 | struct slave *slave = NULL; |
3786 | struct slave *check_slave; | |
dec1e90e | 3787 | int res = 1; |
bb1d9123 | 3788 | |
0693e88e MM |
3789 | if (!skb->queue_mapping) |
3790 | return 1; | |
bb1d9123 AG |
3791 | |
3792 | /* Find out if any slaves have the same mapping as this skb. */ | |
278b2083 | 3793 | bond_for_each_slave_rcu(bond, check_slave) { |
bb1d9123 AG |
3794 | if (check_slave->queue_id == skb->queue_mapping) { |
3795 | slave = check_slave; | |
3796 | break; | |
3797 | } | |
3798 | } | |
3799 | ||
3800 | /* If the slave isn't UP, use default transmit policy. */ | |
3801 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
3802 | (slave->link == BOND_LINK_UP)) { | |
3803 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
3804 | } | |
3805 | ||
bb1d9123 AG |
3806 | return res; |
3807 | } | |
3808 | ||
374eeb5a | 3809 | |
bb1d9123 AG |
3810 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
3811 | { | |
3812 | /* | |
3813 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 3814 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3815 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3816 | * way down to the bonding driver. | |
3817 | */ | |
fd0e435b PO |
3818 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3819 | ||
374eeb5a NH |
3820 | /* |
3821 | * Save the original txq to restore before passing to the driver | |
3822 | */ | |
df4ab5b3 | 3823 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3824 | |
fd0e435b | 3825 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3826 | do { |
fd0e435b | 3827 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3828 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3829 | } |
3830 | return txq; | |
bb1d9123 AG |
3831 | } |
3832 | ||
0693e88e | 3833 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3834 | { |
bb1d9123 AG |
3835 | struct bonding *bond = netdev_priv(dev); |
3836 | ||
3837 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
3838 | if (!bond_slave_override(bond, skb)) | |
3839 | return NETDEV_TX_OK; | |
3840 | } | |
00829823 SH |
3841 | |
3842 | switch (bond->params.mode) { | |
3843 | case BOND_MODE_ROUNDROBIN: | |
3844 | return bond_xmit_roundrobin(skb, dev); | |
3845 | case BOND_MODE_ACTIVEBACKUP: | |
3846 | return bond_xmit_activebackup(skb, dev); | |
3847 | case BOND_MODE_XOR: | |
3848 | return bond_xmit_xor(skb, dev); | |
3849 | case BOND_MODE_BROADCAST: | |
3850 | return bond_xmit_broadcast(skb, dev); | |
3851 | case BOND_MODE_8023AD: | |
3852 | return bond_3ad_xmit_xor(skb, dev); | |
3853 | case BOND_MODE_ALB: | |
3854 | case BOND_MODE_TLB: | |
3855 | return bond_alb_xmit(skb, dev); | |
3856 | default: | |
3857 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3858 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3859 | dev->name, bond->params.mode); | |
00829823 | 3860 | WARN_ON_ONCE(1); |
04502430 | 3861 | kfree_skb(skb); |
00829823 SH |
3862 | return NETDEV_TX_OK; |
3863 | } | |
3864 | } | |
3865 | ||
0693e88e MM |
3866 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3867 | { | |
3868 | struct bonding *bond = netdev_priv(dev); | |
3869 | netdev_tx_t ret = NETDEV_TX_OK; | |
3870 | ||
3871 | /* | |
3872 | * If we risk deadlock from transmitting this in the | |
3873 | * netpoll path, tell netpoll to queue the frame for later tx | |
3874 | */ | |
3875 | if (is_netpoll_tx_blocked(dev)) | |
3876 | return NETDEV_TX_BUSY; | |
3877 | ||
278b2083 | 3878 | rcu_read_lock(); |
dec1e90e | 3879 | if (!list_empty(&bond->slave_list)) |
0693e88e MM |
3880 | ret = __bond_start_xmit(skb, dev); |
3881 | else | |
04502430 | 3882 | kfree_skb(skb); |
278b2083 | 3883 | rcu_read_unlock(); |
0693e88e MM |
3884 | |
3885 | return ret; | |
3886 | } | |
00829823 | 3887 | |
1da177e4 LT |
3888 | /* |
3889 | * set bond mode specific net device operations | |
3890 | */ | |
a77b5325 | 3891 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 3892 | { |
169a3e66 JV |
3893 | struct net_device *bond_dev = bond->dev; |
3894 | ||
1da177e4 LT |
3895 | switch (mode) { |
3896 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
3897 | break; |
3898 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
3899 | break; |
3900 | case BOND_MODE_XOR: | |
6f6652be | 3901 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
3902 | break; |
3903 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
3904 | break; |
3905 | case BOND_MODE_8023AD: | |
6f6652be | 3906 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 3907 | break; |
1da177e4 | 3908 | case BOND_MODE_ALB: |
8f903c70 JV |
3909 | /* FALLTHRU */ |
3910 | case BOND_MODE_TLB: | |
1da177e4 LT |
3911 | break; |
3912 | default: | |
3913 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3914 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3915 | bond_dev->name, mode); | |
1da177e4 LT |
3916 | break; |
3917 | } | |
3918 | } | |
3919 | ||
bb5b052f AG |
3920 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3921 | struct ethtool_cmd *ecmd) | |
3922 | { | |
3923 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3924 | unsigned long speed = 0; |
dec1e90e | 3925 | struct slave *slave; |
bb5b052f AG |
3926 | |
3927 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3928 | ecmd->port = PORT_OTHER; | |
3929 | ||
3930 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
3931 | * do not need to check mode. Though link speed might not represent | |
3932 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3933 | * this is an accurate maximum. | |
3934 | */ | |
3935 | read_lock(&bond->lock); | |
dec1e90e | 3936 | bond_for_each_slave(bond, slave) { |
bb5b052f AG |
3937 | if (SLAVE_IS_OK(slave)) { |
3938 | if (slave->speed != SPEED_UNKNOWN) | |
3939 | speed += slave->speed; | |
3940 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3941 | slave->duplex != DUPLEX_UNKNOWN) | |
3942 | ecmd->duplex = slave->duplex; | |
3943 | } | |
3944 | } | |
3945 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
3946 | read_unlock(&bond->lock); | |
dec1e90e | 3947 | |
bb5b052f AG |
3948 | return 0; |
3949 | } | |
3950 | ||
217df670 | 3951 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3952 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3953 | { |
7826d43f JP |
3954 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3955 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3956 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3957 | BOND_ABI_VERSION); | |
217df670 JV |
3958 | } |
3959 | ||
7282d491 | 3960 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3961 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3962 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3963 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3964 | }; |
3965 | ||
eb7cc59a | 3966 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3967 | .ndo_init = bond_init, |
9e71626c | 3968 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3969 | .ndo_open = bond_open, |
3970 | .ndo_stop = bond_close, | |
00829823 | 3971 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3972 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3973 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3974 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3975 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 3976 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 3977 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 3978 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 3979 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 3980 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 3981 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 3982 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 3983 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
3984 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
3985 | .ndo_poll_controller = bond_poll_controller, | |
3986 | #endif | |
9232ecca JP |
3987 | .ndo_add_slave = bond_enslave, |
3988 | .ndo_del_slave = bond_release, | |
b2a103e6 | 3989 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
3990 | }; |
3991 | ||
b3f92b63 DG |
3992 | static const struct device_type bond_type = { |
3993 | .name = "bond", | |
3994 | }; | |
3995 | ||
9e2e61fb AW |
3996 | static void bond_destructor(struct net_device *bond_dev) |
3997 | { | |
3998 | struct bonding *bond = netdev_priv(bond_dev); | |
3999 | if (bond->wq) | |
4000 | destroy_workqueue(bond->wq); | |
4001 | free_netdev(bond_dev); | |
4002 | } | |
4003 | ||
181470fc | 4004 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4005 | { |
454d7c9b | 4006 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4007 | |
1da177e4 LT |
4008 | /* initialize rwlocks */ |
4009 | rwlock_init(&bond->lock); | |
4010 | rwlock_init(&bond->curr_slave_lock); | |
dec1e90e | 4011 | INIT_LIST_HEAD(&bond->slave_list); |
d2991f75 | 4012 | bond->params = bonding_defaults; |
1da177e4 LT |
4013 | |
4014 | /* Initialize pointers */ | |
1da177e4 | 4015 | bond->dev = bond_dev; |
1da177e4 LT |
4016 | |
4017 | /* Initialize the device entry points */ | |
181470fc | 4018 | ether_setup(bond_dev); |
eb7cc59a | 4019 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4020 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4021 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4022 | |
9e2e61fb | 4023 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4024 | |
b3f92b63 DG |
4025 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4026 | ||
1da177e4 LT |
4027 | /* Initialize the device options */ |
4028 | bond_dev->tx_queue_len = 0; | |
4029 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4030 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4031 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4032 | |
1da177e4 LT |
4033 | /* At first, we block adding VLANs. That's the only way to |
4034 | * prevent problems that occur when adding VLANs over an | |
4035 | * empty bond. The block will be removed once non-challenged | |
4036 | * slaves are enslaved. | |
4037 | */ | |
4038 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4039 | ||
932ff279 | 4040 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4041 | * transmitting */ |
4042 | bond_dev->features |= NETIF_F_LLTX; | |
4043 | ||
4044 | /* By default, we declare the bond to be fully | |
4045 | * VLAN hardware accelerated capable. Special | |
4046 | * care is taken in the various xmit functions | |
4047 | * when there are slaves that are not hw accel | |
4048 | * capable | |
4049 | */ | |
1da177e4 | 4050 | |
b2a103e6 | 4051 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4052 | NETIF_F_HW_VLAN_CTAG_TX | |
4053 | NETIF_F_HW_VLAN_CTAG_RX | | |
4054 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4055 | |
34324dc2 | 4056 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4057 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4058 | } |
4059 | ||
c67dfb29 EB |
4060 | /* |
4061 | * Destroy a bonding device. | |
4062 | * Must be under rtnl_lock when this function is called. | |
4063 | */ | |
4064 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4065 | { |
454d7c9b | 4066 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 4067 | struct slave *slave, *tmp_slave; |
a434e43f | 4068 | |
f6dc31a8 WC |
4069 | bond_netpoll_cleanup(bond_dev); |
4070 | ||
c67dfb29 | 4071 | /* Release the bonded slaves */ |
dec1e90e | 4072 | list_for_each_entry_safe(slave, tmp_slave, &bond->slave_list, list) |
4073 | __bond_release_one(bond_dev, slave->dev, true); | |
0896341a | 4074 | pr_info("%s: released all slaves\n", bond_dev->name); |
c67dfb29 | 4075 | |
a434e43f JV |
4076 | list_del(&bond->bond_list); |
4077 | ||
f073c7ca | 4078 | bond_debug_unregister(bond); |
a434e43f JV |
4079 | } |
4080 | ||
1da177e4 LT |
4081 | /*------------------------- Module initialization ---------------------------*/ |
4082 | ||
4083 | /* | |
4084 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4085 | * number of the mode or its string name. A bit complicated because |
4086 | * some mode names are substrings of other names, and calls from sysfs | |
4087 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4088 | */ |
325dcf7a | 4089 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4090 | { |
54b87323 | 4091 | int modeint = -1, i, rv; |
a42e534f | 4092 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4093 | |
a42e534f JV |
4094 | for (p = (char *)buf; *p; p++) |
4095 | if (!(isdigit(*p) || isspace(*p))) | |
4096 | break; | |
4097 | ||
4098 | if (*p) | |
ece95f7f | 4099 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4100 | else |
54b87323 | 4101 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4102 | |
4103 | if (!rv) | |
4104 | return -1; | |
1da177e4 LT |
4105 | |
4106 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4107 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4108 | return tbl[i].mode; |
4109 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4110 | return tbl[i].mode; |
1da177e4 LT |
4111 | } |
4112 | ||
4113 | return -1; | |
4114 | } | |
4115 | ||
4116 | static int bond_check_params(struct bond_params *params) | |
4117 | { | |
5a5c5fd4 | 4118 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
8599b52e | 4119 | int arp_all_targets_value; |
f5b2b966 | 4120 | |
1da177e4 LT |
4121 | /* |
4122 | * Convert string parameters. | |
4123 | */ | |
4124 | if (mode) { | |
4125 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4126 | if (bond_mode == -1) { | |
a4aee5c8 | 4127 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4128 | mode == NULL ? "NULL" : mode); |
4129 | return -EINVAL; | |
4130 | } | |
4131 | } | |
4132 | ||
169a3e66 JV |
4133 | if (xmit_hash_policy) { |
4134 | if ((bond_mode != BOND_MODE_XOR) && | |
4135 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4136 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4137 | bond_mode_name(bond_mode)); |
4138 | } else { | |
4139 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4140 | xmit_hashtype_tbl); | |
4141 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4142 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4143 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4144 | xmit_hash_policy); |
4145 | return -EINVAL; | |
4146 | } | |
4147 | } | |
4148 | } | |
4149 | ||
1da177e4 LT |
4150 | if (lacp_rate) { |
4151 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4152 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4153 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4154 | } else { |
4155 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4156 | if (lacp_fast == -1) { | |
a4aee5c8 | 4157 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4158 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4159 | return -EINVAL; | |
4160 | } | |
4161 | } | |
4162 | } | |
4163 | ||
fd989c83 JV |
4164 | if (ad_select) { |
4165 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4166 | if (params->ad_select == -1) { | |
a4aee5c8 | 4167 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4168 | ad_select == NULL ? "NULL" : ad_select); |
4169 | return -EINVAL; | |
4170 | } | |
4171 | ||
4172 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4173 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4174 | } |
4175 | } else { | |
4176 | params->ad_select = BOND_AD_STABLE; | |
4177 | } | |
4178 | ||
f5841306 | 4179 | if (max_bonds < 0) { |
a4aee5c8 JP |
4180 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4181 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4182 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4183 | } | |
4184 | ||
4185 | if (miimon < 0) { | |
a4aee5c8 JP |
4186 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4187 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4188 | miimon = BOND_LINK_MON_INTERV; |
4189 | } | |
4190 | ||
4191 | if (updelay < 0) { | |
a4aee5c8 JP |
4192 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4193 | updelay, INT_MAX); | |
1da177e4 LT |
4194 | updelay = 0; |
4195 | } | |
4196 | ||
4197 | if (downdelay < 0) { | |
a4aee5c8 JP |
4198 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4199 | downdelay, INT_MAX); | |
1da177e4 LT |
4200 | downdelay = 0; |
4201 | } | |
4202 | ||
4203 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4204 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4205 | use_carrier); | |
1da177e4 LT |
4206 | use_carrier = 1; |
4207 | } | |
4208 | ||
ad246c99 BH |
4209 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4210 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4211 | num_peer_notif); | |
4212 | num_peer_notif = 1; | |
4213 | } | |
4214 | ||
1da177e4 LT |
4215 | /* reset values for 802.3ad */ |
4216 | if (bond_mode == BOND_MODE_8023AD) { | |
4217 | if (!miimon) { | |
a4aee5c8 | 4218 | 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 | 4219 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4220 | miimon = 100; |
4221 | } | |
4222 | } | |
4223 | ||
bb1d9123 AG |
4224 | if (tx_queues < 1 || tx_queues > 255) { |
4225 | pr_warning("Warning: tx_queues (%d) should be between " | |
4226 | "1 and 255, resetting to %d\n", | |
4227 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4228 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4229 | } | |
4230 | ||
ebd8e497 AG |
4231 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4232 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4233 | "not of valid value (0/1), so it was set to " | |
4234 | "0\n", all_slaves_active); | |
4235 | all_slaves_active = 0; | |
4236 | } | |
4237 | ||
c2952c31 FL |
4238 | if (resend_igmp < 0 || resend_igmp > 255) { |
4239 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4240 | "0 and 255, resetting to %d\n", | |
4241 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4242 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4243 | } | |
4244 | ||
1da177e4 LT |
4245 | /* reset values for TLB/ALB */ |
4246 | if ((bond_mode == BOND_MODE_TLB) || | |
4247 | (bond_mode == BOND_MODE_ALB)) { | |
4248 | if (!miimon) { | |
a4aee5c8 | 4249 | 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 | 4250 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4251 | miimon = 100; |
4252 | } | |
4253 | } | |
4254 | ||
4255 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4256 | 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", |
4257 | updelay); | |
1da177e4 LT |
4258 | } |
4259 | ||
4260 | if (!miimon) { | |
4261 | if (updelay || downdelay) { | |
4262 | /* just warn the user the up/down delay will have | |
4263 | * no effect since miimon is zero... | |
4264 | */ | |
a4aee5c8 JP |
4265 | 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", |
4266 | updelay, downdelay); | |
1da177e4 LT |
4267 | } |
4268 | } else { | |
4269 | /* don't allow arp monitoring */ | |
4270 | if (arp_interval) { | |
a4aee5c8 JP |
4271 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4272 | miimon, arp_interval); | |
1da177e4 LT |
4273 | arp_interval = 0; |
4274 | } | |
4275 | ||
4276 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4277 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4278 | updelay, miimon, | |
4279 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4280 | } |
4281 | ||
4282 | updelay /= miimon; | |
4283 | ||
4284 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4285 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4286 | downdelay, miimon, | |
4287 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4288 | } |
4289 | ||
4290 | downdelay /= miimon; | |
4291 | } | |
4292 | ||
4293 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4294 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4295 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4296 | arp_interval = BOND_LINK_ARP_INTERV; |
4297 | } | |
4298 | ||
5a5c5fd4 | 4299 | for (arp_ip_count = 0, i = 0; |
4300 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4301 | /* not complete check, but should be good enough to |
4302 | catch mistakes */ | |
5a5c5fd4 | 4303 | __be32 ip = in_aton(arp_ip_target[i]); |
4304 | if (!isdigit(arp_ip_target[i][0]) || ip == 0 || | |
4305 | ip == htonl(INADDR_BROADCAST)) { | |
a4aee5c8 | 4306 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5a5c5fd4 | 4307 | arp_ip_target[i]); |
1da177e4 LT |
4308 | arp_interval = 0; |
4309 | } else { | |
0afee4e8 VF |
4310 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4311 | arp_target[arp_ip_count++] = ip; | |
4312 | else | |
4313 | pr_warning("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", | |
4314 | &ip); | |
1da177e4 LT |
4315 | } |
4316 | } | |
4317 | ||
4318 | if (arp_interval && !arp_ip_count) { | |
4319 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4320 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4321 | arp_interval); | |
1da177e4 LT |
4322 | arp_interval = 0; |
4323 | } | |
4324 | ||
f5b2b966 JV |
4325 | if (arp_validate) { |
4326 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4327 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4328 | return -EINVAL; |
4329 | } | |
4330 | if (!arp_interval) { | |
a4aee5c8 | 4331 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4332 | return -EINVAL; |
4333 | } | |
4334 | ||
4335 | arp_validate_value = bond_parse_parm(arp_validate, | |
4336 | arp_validate_tbl); | |
4337 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4338 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4339 | arp_validate == NULL ? "NULL" : arp_validate); |
4340 | return -EINVAL; | |
4341 | } | |
4342 | } else | |
4343 | arp_validate_value = 0; | |
4344 | ||
8599b52e VF |
4345 | arp_all_targets_value = 0; |
4346 | if (arp_all_targets) { | |
4347 | arp_all_targets_value = bond_parse_parm(arp_all_targets, | |
4348 | arp_all_targets_tbl); | |
4349 | ||
4350 | if (arp_all_targets_value == -1) { | |
4351 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", | |
4352 | arp_all_targets); | |
4353 | arp_all_targets_value = 0; | |
4354 | } | |
4355 | } | |
4356 | ||
1da177e4 | 4357 | if (miimon) { |
a4aee5c8 | 4358 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4359 | } else if (arp_interval) { |
a4aee5c8 JP |
4360 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4361 | arp_interval, | |
4362 | arp_validate_tbl[arp_validate_value].modename, | |
4363 | arp_ip_count); | |
1da177e4 LT |
4364 | |
4365 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4366 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4367 | |
e5e2a8fd | 4368 | pr_info("\n"); |
1da177e4 | 4369 | |
b8a9787e | 4370 | } else if (max_bonds) { |
1da177e4 LT |
4371 | /* miimon and arp_interval not set, we need one so things |
4372 | * work as expected, see bonding.txt for details | |
4373 | */ | |
b2730f4f | 4374 | pr_debug("Warning: either miimon or arp_interval and arp_ip_target module parameters must be specified, otherwise bonding will not detect link failures! see bonding.txt for details.\n"); |
1da177e4 LT |
4375 | } |
4376 | ||
4377 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4378 | /* currently, using a primary only makes sense | |
4379 | * in active backup, TLB or ALB modes | |
4380 | */ | |
a4aee5c8 JP |
4381 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4382 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4383 | primary = NULL; |
4384 | } | |
4385 | ||
a549952a JP |
4386 | if (primary && primary_reselect) { |
4387 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4388 | pri_reselect_tbl); | |
4389 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4390 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4391 | primary_reselect == |
4392 | NULL ? "NULL" : primary_reselect); | |
4393 | return -EINVAL; | |
4394 | } | |
4395 | } else { | |
4396 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4397 | } | |
4398 | ||
3915c1e8 JV |
4399 | if (fail_over_mac) { |
4400 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4401 | fail_over_mac_tbl); | |
4402 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4403 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4404 | arp_validate == NULL ? "NULL" : arp_validate); |
4405 | return -EINVAL; | |
4406 | } | |
4407 | ||
4408 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4409 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4410 | } else { |
4411 | fail_over_mac_value = BOND_FOM_NONE; | |
4412 | } | |
dd957c57 | 4413 | |
1da177e4 LT |
4414 | /* fill params struct with the proper values */ |
4415 | params->mode = bond_mode; | |
169a3e66 | 4416 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4417 | params->miimon = miimon; |
ad246c99 | 4418 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4419 | params->arp_interval = arp_interval; |
f5b2b966 | 4420 | params->arp_validate = arp_validate_value; |
8599b52e | 4421 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4422 | params->updelay = updelay; |
4423 | params->downdelay = downdelay; | |
4424 | params->use_carrier = use_carrier; | |
4425 | params->lacp_fast = lacp_fast; | |
4426 | params->primary[0] = 0; | |
a549952a | 4427 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4428 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4429 | params->tx_queues = tx_queues; |
ebd8e497 | 4430 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4431 | params->resend_igmp = resend_igmp; |
655f8919 | 4432 | params->min_links = min_links; |
7eacd038 | 4433 | params->lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
4434 | |
4435 | if (primary) { | |
4436 | strncpy(params->primary, primary, IFNAMSIZ); | |
4437 | params->primary[IFNAMSIZ - 1] = 0; | |
4438 | } | |
4439 | ||
4440 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4441 | ||
4442 | return 0; | |
4443 | } | |
4444 | ||
0daa2303 | 4445 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4446 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4447 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4448 | |
e8a0464c DM |
4449 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4450 | struct netdev_queue *txq, | |
4451 | void *_unused) | |
c773e847 DM |
4452 | { |
4453 | lockdep_set_class(&txq->_xmit_lock, | |
4454 | &bonding_netdev_xmit_lock_key); | |
4455 | } | |
4456 | ||
4457 | static void bond_set_lockdep_class(struct net_device *dev) | |
4458 | { | |
cf508b12 DM |
4459 | lockdep_set_class(&dev->addr_list_lock, |
4460 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4461 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4462 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4463 | } |
4464 | ||
181470fc SH |
4465 | /* |
4466 | * Called from registration process | |
4467 | */ | |
4468 | static int bond_init(struct net_device *bond_dev) | |
4469 | { | |
4470 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4471 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4472 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4473 | |
4474 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4475 | ||
9fe0617d NH |
4476 | /* |
4477 | * Initialize locks that may be required during | |
4478 | * en/deslave operations. All of the bond_open work | |
4479 | * (of which this is part) should really be moved to | |
4480 | * a phase prior to dev_open | |
4481 | */ | |
4482 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4483 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4484 | ||
181470fc SH |
4485 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4486 | if (!bond->wq) | |
4487 | return -ENOMEM; | |
4488 | ||
4489 | bond_set_lockdep_class(bond_dev); | |
4490 | ||
ec87fd3b | 4491 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4492 | |
6151b3d4 | 4493 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4494 | |
f073c7ca TI |
4495 | bond_debug_register(bond); |
4496 | ||
409cc1f8 JP |
4497 | /* Ensure valid dev_addr */ |
4498 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4499 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4500 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4501 | |
181470fc SH |
4502 | return 0; |
4503 | } | |
4504 | ||
88ead977 EB |
4505 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4506 | { | |
4507 | if (tb[IFLA_ADDRESS]) { | |
4508 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4509 | return -EINVAL; | |
4510 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4511 | return -EADDRNOTAVAIL; | |
4512 | } | |
4513 | return 0; | |
4514 | } | |
4515 | ||
d40156aa | 4516 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4517 | { |
efacb309 | 4518 | return tx_queues; |
d5da4510 JP |
4519 | } |
4520 | ||
88ead977 | 4521 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4522 | .kind = "bond", |
4523 | .priv_size = sizeof(struct bonding), | |
4524 | .setup = bond_setup, | |
4525 | .validate = bond_validate, | |
4526 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4527 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4528 | as for TX queues */ | |
88ead977 EB |
4529 | }; |
4530 | ||
dfe60397 | 4531 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4532 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4533 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4534 | * set up our sysfs entries. | |
4535 | */ | |
ec87fd3b | 4536 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4537 | { |
4538 | struct net_device *bond_dev; | |
4539 | int res; | |
4540 | ||
4541 | rtnl_lock(); | |
027ea041 | 4542 | |
1c5cae81 JP |
4543 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4544 | name ? name : "bond%d", | |
4545 | bond_setup, tx_queues); | |
dfe60397 | 4546 | if (!bond_dev) { |
a4aee5c8 | 4547 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4548 | rtnl_unlock(); |
4549 | return -ENOMEM; | |
dfe60397 MW |
4550 | } |
4551 | ||
ec87fd3b | 4552 | dev_net_set(bond_dev, net); |
88ead977 EB |
4553 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4554 | ||
dfe60397 | 4555 | res = register_netdevice(bond_dev); |
0daa2303 | 4556 | |
e826eafa PO |
4557 | netif_carrier_off(bond_dev); |
4558 | ||
7e083840 | 4559 | rtnl_unlock(); |
9e2e61fb AW |
4560 | if (res < 0) |
4561 | bond_destructor(bond_dev); | |
30c15ba9 | 4562 | return res; |
dfe60397 MW |
4563 | } |
4564 | ||
2c8c1e72 | 4565 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4566 | { |
15449745 | 4567 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4568 | |
4569 | bn->net = net; | |
4570 | INIT_LIST_HEAD(&bn->dev_list); | |
4571 | ||
ec87fd3b | 4572 | bond_create_proc_dir(bn); |
4c22400a | 4573 | bond_create_sysfs(bn); |
87a7b84b | 4574 | |
15449745 | 4575 | return 0; |
ec87fd3b EB |
4576 | } |
4577 | ||
2c8c1e72 | 4578 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4579 | { |
15449745 | 4580 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4581 | struct bonding *bond, *tmp_bond; |
4582 | LIST_HEAD(list); | |
ec87fd3b | 4583 | |
4c22400a | 4584 | bond_destroy_sysfs(bn); |
ec87fd3b | 4585 | bond_destroy_proc_dir(bn); |
69b0216a | 4586 | |
4587 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4588 | rtnl_lock(); | |
4589 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4590 | unregister_netdevice_queue(bond->dev, &list); | |
4591 | unregister_netdevice_many(&list); | |
4592 | rtnl_unlock(); | |
ec87fd3b EB |
4593 | } |
4594 | ||
4595 | static struct pernet_operations bond_net_ops = { | |
4596 | .init = bond_net_init, | |
4597 | .exit = bond_net_exit, | |
15449745 EB |
4598 | .id = &bond_net_id, |
4599 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4600 | }; |
4601 | ||
1da177e4 LT |
4602 | static int __init bonding_init(void) |
4603 | { | |
1da177e4 LT |
4604 | int i; |
4605 | int res; | |
4606 | ||
bd33acc3 | 4607 | pr_info("%s", bond_version); |
1da177e4 | 4608 | |
dfe60397 | 4609 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4610 | if (res) |
dfe60397 | 4611 | goto out; |
1da177e4 | 4612 | |
15449745 | 4613 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4614 | if (res) |
4615 | goto out; | |
027ea041 | 4616 | |
88ead977 EB |
4617 | res = rtnl_link_register(&bond_link_ops); |
4618 | if (res) | |
6639104b | 4619 | goto err_link; |
88ead977 | 4620 | |
f073c7ca TI |
4621 | bond_create_debugfs(); |
4622 | ||
1da177e4 | 4623 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4624 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4625 | if (res) |
4626 | goto err; | |
1da177e4 LT |
4627 | } |
4628 | ||
1da177e4 | 4629 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4630 | out: |
1da177e4 | 4631 | return res; |
88ead977 EB |
4632 | err: |
4633 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 4634 | err_link: |
15449745 | 4635 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4636 | goto out; |
dfe60397 | 4637 | |
1da177e4 LT |
4638 | } |
4639 | ||
4640 | static void __exit bonding_exit(void) | |
4641 | { | |
4642 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4643 | ||
f073c7ca | 4644 | bond_destroy_debugfs(); |
ae68c398 | 4645 | |
4de79c73 | 4646 | rtnl_link_unregister(&bond_link_ops); |
ffcdedb6 | 4647 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4648 | |
4649 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4650 | /* |
4651 | * Make sure we don't have an imbalance on our netpoll blocking | |
4652 | */ | |
4653 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4654 | #endif |
1da177e4 LT |
4655 | } |
4656 | ||
4657 | module_init(bonding_init); | |
4658 | module_exit(bonding_exit); | |
4659 | MODULE_LICENSE("GPL"); | |
4660 | MODULE_VERSION(DRV_VERSION); | |
4661 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4662 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 4663 | MODULE_ALIAS_RTNL_LINK("bond"); |