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); |
dec1e90e | 335 | struct slave *slave; |
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: | |
b20903f2 | 347 | /* unwind from the slave that failed */ |
dec1e90e | 348 | bond_for_each_slave_continue_reverse(bond, slave) |
80d5c368 | 349 | vlan_vid_del(slave->dev, proto, vid); |
87002b03 JP |
350 | |
351 | return res; | |
1da177e4 LT |
352 | } |
353 | ||
354 | /** | |
355 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
356 | * @bond_dev: bonding net device that got called | |
357 | * @vid: vlan id being removed | |
358 | */ | |
80d5c368 PM |
359 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
360 | __be16 proto, u16 vid) | |
1da177e4 | 361 | { |
454d7c9b | 362 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 363 | struct slave *slave; |
1da177e4 | 364 | |
dec1e90e | 365 | bond_for_each_slave(bond, slave) |
80d5c368 | 366 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 367 | |
e868b0c9 VF |
368 | if (bond_is_lb(bond)) |
369 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
370 | |
371 | return 0; | |
1da177e4 LT |
372 | } |
373 | ||
1da177e4 LT |
374 | /*------------------------------- Link status -------------------------------*/ |
375 | ||
ff59c456 JV |
376 | /* |
377 | * Set the carrier state for the master according to the state of its | |
378 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
379 | * do special 802.3ad magic. | |
380 | * | |
381 | * Returns zero if carrier state does not change, nonzero if it does. | |
382 | */ | |
383 | static int bond_set_carrier(struct bonding *bond) | |
384 | { | |
385 | struct slave *slave; | |
ff59c456 | 386 | |
dec1e90e | 387 | if (list_empty(&bond->slave_list)) |
ff59c456 JV |
388 | goto down; |
389 | ||
390 | if (bond->params.mode == BOND_MODE_8023AD) | |
391 | return bond_3ad_set_carrier(bond); | |
392 | ||
dec1e90e | 393 | bond_for_each_slave(bond, slave) { |
ff59c456 JV |
394 | if (slave->link == BOND_LINK_UP) { |
395 | if (!netif_carrier_ok(bond->dev)) { | |
396 | netif_carrier_on(bond->dev); | |
397 | return 1; | |
398 | } | |
399 | return 0; | |
400 | } | |
401 | } | |
402 | ||
403 | down: | |
404 | if (netif_carrier_ok(bond->dev)) { | |
405 | netif_carrier_off(bond->dev); | |
406 | return 1; | |
407 | } | |
408 | return 0; | |
409 | } | |
410 | ||
1da177e4 LT |
411 | /* |
412 | * Get link speed and duplex from the slave's base driver | |
413 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 414 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 415 | * and return. |
1da177e4 | 416 | */ |
13b95fb7 | 417 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
418 | { |
419 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 420 | struct ethtool_cmd ecmd; |
5d30530e | 421 | u32 slave_speed; |
61a44b9c | 422 | int res; |
1da177e4 | 423 | |
589665f5 DC |
424 | slave->speed = SPEED_UNKNOWN; |
425 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 426 | |
4bc71cb9 | 427 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 428 | if (res < 0) |
13b95fb7 | 429 | return; |
1da177e4 | 430 | |
4bc71cb9 | 431 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 432 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 433 | return; |
1da177e4 | 434 | |
4bc71cb9 | 435 | switch (ecmd.duplex) { |
1da177e4 LT |
436 | case DUPLEX_FULL: |
437 | case DUPLEX_HALF: | |
438 | break; | |
439 | default: | |
13b95fb7 | 440 | return; |
1da177e4 LT |
441 | } |
442 | ||
5d30530e | 443 | slave->speed = slave_speed; |
4bc71cb9 | 444 | slave->duplex = ecmd.duplex; |
1da177e4 | 445 | |
13b95fb7 | 446 | return; |
1da177e4 LT |
447 | } |
448 | ||
449 | /* | |
450 | * if <dev> supports MII link status reporting, check its link status. | |
451 | * | |
452 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 453 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
454 | * |
455 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
456 | * can't tell and just pretend it is), or 0, meaning that the link is | |
457 | * down. | |
458 | * | |
459 | * If reporting is non-zero, instead of faking link up, return -1 if | |
460 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
461 | * support them). If use_carrier is set, return whatever it says. | |
462 | * It'd be nice if there was a good way to tell if a driver supports | |
463 | * netif_carrier, but there really isn't. | |
464 | */ | |
3d632c3f SH |
465 | static int bond_check_dev_link(struct bonding *bond, |
466 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 467 | { |
eb7cc59a | 468 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 469 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
470 | struct ifreq ifr; |
471 | struct mii_ioctl_data *mii; | |
1da177e4 | 472 | |
6c988853 PG |
473 | if (!reporting && !netif_running(slave_dev)) |
474 | return 0; | |
475 | ||
eb7cc59a | 476 | if (bond->params.use_carrier) |
1da177e4 | 477 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 478 | |
29112f4e | 479 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
480 | if (slave_dev->ethtool_ops->get_link) |
481 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
482 | BMSR_LSTATUS : 0; | |
29112f4e | 483 | |
3d632c3f | 484 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 485 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
486 | if (ioctl) { |
487 | /* TODO: set pointer to correct ioctl on a per team member */ | |
488 | /* bases to make this more efficient. that is, once */ | |
489 | /* we determine the correct ioctl, we will always */ | |
490 | /* call it and not the others for that team */ | |
491 | /* member. */ | |
492 | ||
493 | /* | |
494 | * We cannot assume that SIOCGMIIPHY will also read a | |
495 | * register; not all network drivers (e.g., e100) | |
496 | * support that. | |
497 | */ | |
498 | ||
499 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
500 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
501 | mii = if_mii(&ifr); | |
502 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
503 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
504 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
505 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
506 | } |
507 | } | |
508 | ||
1da177e4 LT |
509 | /* |
510 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 511 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
512 | * cannot report link status). If not reporting, pretend |
513 | * we're ok. | |
514 | */ | |
3d632c3f | 515 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
516 | } |
517 | ||
518 | /*----------------------------- Multicast list ------------------------------*/ | |
519 | ||
1da177e4 LT |
520 | /* |
521 | * Push the promiscuity flag down to appropriate slaves | |
522 | */ | |
7e1a1ac1 | 523 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 524 | { |
7e1a1ac1 | 525 | int err = 0; |
1da177e4 LT |
526 | if (USES_PRIMARY(bond->params.mode)) { |
527 | /* write lock already acquired */ | |
528 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
529 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
530 | inc); | |
1da177e4 LT |
531 | } |
532 | } else { | |
533 | struct slave *slave; | |
dec1e90e | 534 | |
535 | bond_for_each_slave(bond, slave) { | |
7e1a1ac1 WC |
536 | err = dev_set_promiscuity(slave->dev, inc); |
537 | if (err) | |
538 | return err; | |
1da177e4 LT |
539 | } |
540 | } | |
7e1a1ac1 | 541 | return err; |
1da177e4 LT |
542 | } |
543 | ||
544 | /* | |
545 | * Push the allmulti flag down to all slaves | |
546 | */ | |
7e1a1ac1 | 547 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 548 | { |
7e1a1ac1 | 549 | int err = 0; |
1da177e4 LT |
550 | if (USES_PRIMARY(bond->params.mode)) { |
551 | /* write lock already acquired */ | |
552 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
553 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
554 | inc); | |
1da177e4 LT |
555 | } |
556 | } else { | |
557 | struct slave *slave; | |
dec1e90e | 558 | |
559 | bond_for_each_slave(bond, slave) { | |
7e1a1ac1 WC |
560 | err = dev_set_allmulti(slave->dev, inc); |
561 | if (err) | |
562 | return err; | |
1da177e4 LT |
563 | } |
564 | } | |
7e1a1ac1 | 565 | return err; |
1da177e4 LT |
566 | } |
567 | ||
5a37e8ca FL |
568 | /* |
569 | * Retrieve the list of registered multicast addresses for the bonding | |
570 | * device and retransmit an IGMP JOIN request to the current active | |
571 | * slave. | |
572 | */ | |
573 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
574 | { | |
4aa5dee4 | 575 | if (!rtnl_trylock()) { |
4beac029 | 576 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 577 | return; |
5a37e8ca | 578 | } |
4aa5dee4 JP |
579 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
580 | rtnl_unlock(); | |
5a37e8ca | 581 | |
4f5474e7 NA |
582 | /* We use curr_slave_lock to protect against concurrent access to |
583 | * igmp_retrans from multiple running instances of this function and | |
584 | * bond_change_active_slave | |
585 | */ | |
586 | write_lock_bh(&bond->curr_slave_lock); | |
587 | if (bond->igmp_retrans > 1) { | |
588 | bond->igmp_retrans--; | |
c2952c31 | 589 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 NA |
590 | } |
591 | write_unlock_bh(&bond->curr_slave_lock); | |
5a37e8ca FL |
592 | } |
593 | ||
379b7383 | 594 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
595 | { |
596 | struct bonding *bond = container_of(work, struct bonding, | |
94265cf5 | 597 | mcast_work.work); |
ad999eee | 598 | |
5a37e8ca FL |
599 | bond_resend_igmp_join_requests(bond); |
600 | } | |
601 | ||
303d1cbf | 602 | /* Flush bond's hardware addresses from slave |
1da177e4 | 603 | */ |
303d1cbf | 604 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 605 | struct net_device *slave_dev) |
1da177e4 | 606 | { |
454d7c9b | 607 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 608 | |
303d1cbf JV |
609 | dev_uc_unsync(slave_dev, bond_dev); |
610 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 LT |
611 | |
612 | if (bond->params.mode == BOND_MODE_8023AD) { | |
613 | /* del lacpdu mc addr from mc list */ | |
614 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
615 | ||
22bedad3 | 616 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
617 | } |
618 | } | |
619 | ||
620 | /*--------------------------- Active slave change ---------------------------*/ | |
621 | ||
303d1cbf JV |
622 | /* Update the hardware address list and promisc/allmulti for the new and |
623 | * old active slaves (if any). Modes that are !USES_PRIMARY keep all | |
624 | * slaves up date at all times; only the USES_PRIMARY modes need to call | |
625 | * this function to swap these settings during a failover. | |
1da177e4 | 626 | */ |
303d1cbf JV |
627 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
628 | struct slave *old_active) | |
1da177e4 | 629 | { |
7864a1ad VF |
630 | ASSERT_RTNL(); |
631 | ||
1da177e4 | 632 | if (old_active) { |
3d632c3f | 633 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 634 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 635 | |
3d632c3f | 636 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 637 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 638 | |
303d1cbf | 639 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
640 | } |
641 | ||
642 | if (new_active) { | |
7e1a1ac1 | 643 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 644 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 645 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 646 | |
3d632c3f | 647 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 648 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 649 | |
d632ce98 | 650 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
651 | dev_uc_sync(new_active->dev, bond->dev); |
652 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 653 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
654 | } |
655 | } | |
656 | ||
ae0d6750 | 657 | /** |
658 | * bond_set_dev_addr - clone slave's address to bond | |
659 | * @bond_dev: bond net device | |
660 | * @slave_dev: slave net device | |
661 | * | |
662 | * Should be called with RTNL held. | |
663 | */ | |
664 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
665 | struct net_device *slave_dev) | |
666 | { | |
008aebde NA |
667 | pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
668 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 669 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
670 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
671 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
672 | } | |
673 | ||
3915c1e8 JV |
674 | /* |
675 | * bond_do_fail_over_mac | |
676 | * | |
677 | * Perform special MAC address swapping for fail_over_mac settings | |
678 | * | |
679 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
680 | */ | |
681 | static void bond_do_fail_over_mac(struct bonding *bond, | |
682 | struct slave *new_active, | |
683 | struct slave *old_active) | |
1f78d9f9 HE |
684 | __releases(&bond->curr_slave_lock) |
685 | __releases(&bond->lock) | |
686 | __acquires(&bond->lock) | |
687 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
688 | { |
689 | u8 tmp_mac[ETH_ALEN]; | |
690 | struct sockaddr saddr; | |
691 | int rv; | |
692 | ||
693 | switch (bond->params.fail_over_mac) { | |
694 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 695 | if (new_active) { |
7d26bb10 WP |
696 | write_unlock_bh(&bond->curr_slave_lock); |
697 | read_unlock(&bond->lock); | |
ae0d6750 | 698 | bond_set_dev_addr(bond->dev, new_active->dev); |
7d26bb10 WP |
699 | read_lock(&bond->lock); |
700 | write_lock_bh(&bond->curr_slave_lock); | |
701 | } | |
3915c1e8 JV |
702 | break; |
703 | case BOND_FOM_FOLLOW: | |
704 | /* | |
705 | * if new_active && old_active, swap them | |
706 | * if just old_active, do nothing (going to no active slave) | |
707 | * if just new_active, set new_active to bond's MAC | |
708 | */ | |
709 | if (!new_active) | |
710 | return; | |
711 | ||
712 | write_unlock_bh(&bond->curr_slave_lock); | |
713 | read_unlock(&bond->lock); | |
714 | ||
715 | if (old_active) { | |
716 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
717 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
718 | ETH_ALEN); | |
719 | saddr.sa_family = new_active->dev->type; | |
720 | } else { | |
721 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
722 | saddr.sa_family = bond->dev->type; | |
723 | } | |
724 | ||
725 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
726 | if (rv) { | |
a4aee5c8 | 727 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
728 | bond->dev->name, -rv, new_active->dev->name); |
729 | goto out; | |
730 | } | |
731 | ||
732 | if (!old_active) | |
733 | goto out; | |
734 | ||
735 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
736 | saddr.sa_family = old_active->dev->type; | |
737 | ||
738 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
739 | if (rv) | |
a4aee5c8 | 740 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
741 | bond->dev->name, -rv, new_active->dev->name); |
742 | out: | |
743 | read_lock(&bond->lock); | |
744 | write_lock_bh(&bond->curr_slave_lock); | |
745 | break; | |
746 | default: | |
a4aee5c8 | 747 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
748 | bond->dev->name, bond->params.fail_over_mac); |
749 | break; | |
750 | } | |
751 | ||
752 | } | |
753 | ||
a549952a JP |
754 | static bool bond_should_change_active(struct bonding *bond) |
755 | { | |
756 | struct slave *prim = bond->primary_slave; | |
757 | struct slave *curr = bond->curr_active_slave; | |
758 | ||
759 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
760 | return true; | |
761 | if (bond->force_primary) { | |
762 | bond->force_primary = false; | |
763 | return true; | |
764 | } | |
765 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
766 | (prim->speed < curr->speed || | |
767 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
768 | return false; | |
769 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
770 | return false; | |
771 | return true; | |
772 | } | |
3915c1e8 | 773 | |
1da177e4 LT |
774 | /** |
775 | * find_best_interface - select the best available slave to be the active one | |
776 | * @bond: our bonding struct | |
777 | * | |
778 | * Warning: Caller must hold curr_slave_lock for writing. | |
779 | */ | |
780 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
781 | { | |
782 | struct slave *new_active, *old_active; | |
783 | struct slave *bestslave = NULL; | |
784 | int mintime = bond->params.updelay; | |
785 | int i; | |
786 | ||
49b4ad92 | 787 | new_active = bond->curr_active_slave; |
1da177e4 LT |
788 | |
789 | if (!new_active) { /* there were no active slaves left */ | |
dec1e90e | 790 | new_active = bond_first_slave(bond); |
791 | if (!new_active) | |
1da177e4 | 792 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
793 | } |
794 | ||
1da177e4 | 795 | if ((bond->primary_slave) && |
a549952a JP |
796 | bond->primary_slave->link == BOND_LINK_UP && |
797 | bond_should_change_active(bond)) { | |
1da177e4 LT |
798 | new_active = bond->primary_slave; |
799 | } | |
800 | ||
801 | /* remember where to stop iterating over the slaves */ | |
802 | old_active = new_active; | |
803 | ||
804 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
805 | if (new_active->link == BOND_LINK_UP) { |
806 | return new_active; | |
807 | } else if (new_active->link == BOND_LINK_BACK && | |
808 | IS_UP(new_active->dev)) { | |
809 | /* link up, but waiting for stabilization */ | |
810 | if (new_active->delay < mintime) { | |
811 | mintime = new_active->delay; | |
812 | bestslave = new_active; | |
1da177e4 LT |
813 | } |
814 | } | |
815 | } | |
816 | ||
817 | return bestslave; | |
818 | } | |
819 | ||
ad246c99 BH |
820 | static bool bond_should_notify_peers(struct bonding *bond) |
821 | { | |
822 | struct slave *slave = bond->curr_active_slave; | |
823 | ||
824 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
825 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
826 | ||
827 | if (!slave || !bond->send_peer_notif || | |
828 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
829 | return false; | |
830 | ||
ad246c99 BH |
831 | return true; |
832 | } | |
833 | ||
1da177e4 LT |
834 | /** |
835 | * change_active_interface - change the active slave into the specified one | |
836 | * @bond: our bonding struct | |
837 | * @new: the new slave to make the active one | |
838 | * | |
839 | * Set the new slave to the bond's settings and unset them on the old | |
840 | * curr_active_slave. | |
841 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
842 | * | |
843 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
844 | * because it is apparently the best available slave we have, even though its | |
845 | * updelay hasn't timed out yet. | |
846 | * | |
3915c1e8 JV |
847 | * If new_active is not NULL, caller must hold bond->lock for read and |
848 | * curr_slave_lock for write_bh. | |
1da177e4 | 849 | */ |
a77b5325 | 850 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
851 | { |
852 | struct slave *old_active = bond->curr_active_slave; | |
853 | ||
3d632c3f | 854 | if (old_active == new_active) |
1da177e4 | 855 | return; |
1da177e4 LT |
856 | |
857 | if (new_active) { | |
b2220cad JV |
858 | new_active->jiffies = jiffies; |
859 | ||
1da177e4 LT |
860 | if (new_active->link == BOND_LINK_BACK) { |
861 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
862 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
863 | bond->dev->name, new_active->dev->name, | |
864 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
865 | } |
866 | ||
867 | new_active->delay = 0; | |
868 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 869 | |
3d632c3f | 870 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 871 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 872 | |
58402054 | 873 | if (bond_is_lb(bond)) |
1da177e4 | 874 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
875 | } else { |
876 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
877 | pr_info("%s: making interface %s the new active one.\n", |
878 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
879 | } |
880 | } | |
881 | } | |
882 | ||
3d632c3f | 883 | if (USES_PRIMARY(bond->params.mode)) |
303d1cbf | 884 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 885 | |
58402054 | 886 | if (bond_is_lb(bond)) { |
1da177e4 | 887 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
888 | if (old_active) |
889 | bond_set_slave_inactive_flags(old_active); | |
890 | if (new_active) | |
891 | bond_set_slave_active_flags(new_active); | |
1da177e4 | 892 | } else { |
278b2083 | 893 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 894 | } |
c3ade5ca JV |
895 | |
896 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 897 | if (old_active) |
c3ade5ca | 898 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
899 | |
900 | if (new_active) { | |
ad246c99 BH |
901 | bool should_notify_peers = false; |
902 | ||
c3ade5ca | 903 | bond_set_slave_active_flags(new_active); |
2ab82852 | 904 | |
709f8a45 OG |
905 | if (bond->params.fail_over_mac) |
906 | bond_do_fail_over_mac(bond, new_active, | |
907 | old_active); | |
3915c1e8 | 908 | |
ad246c99 BH |
909 | if (netif_running(bond->dev)) { |
910 | bond->send_peer_notif = | |
911 | bond->params.num_peer_notif; | |
912 | should_notify_peers = | |
913 | bond_should_notify_peers(bond); | |
914 | } | |
915 | ||
01f3109d OG |
916 | write_unlock_bh(&bond->curr_slave_lock); |
917 | read_unlock(&bond->lock); | |
918 | ||
b7bc2a5b | 919 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 920 | if (should_notify_peers) |
b7bc2a5b AW |
921 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
922 | bond->dev); | |
01f3109d OG |
923 | |
924 | read_lock(&bond->lock); | |
925 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 926 | } |
c3ade5ca | 927 | } |
a2fd940f | 928 | |
5a37e8ca | 929 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
930 | * all were sent on curr_active_slave. |
931 | * resend only if bond is brought up with the affected | |
932 | * bonding modes and the retransmission is enabled */ | |
933 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
934 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
935 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 936 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 937 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 938 | } |
1da177e4 LT |
939 | } |
940 | ||
941 | /** | |
942 | * bond_select_active_slave - select a new active slave, if needed | |
943 | * @bond: our bonding struct | |
944 | * | |
3d632c3f | 945 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
946 | * - The old curr_active_slave has been released or lost its link. |
947 | * - The primary_slave has got its link back. | |
948 | * - A slave has got its link back and there's no old curr_active_slave. | |
949 | * | |
3915c1e8 | 950 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 951 | */ |
a77b5325 | 952 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
953 | { |
954 | struct slave *best_slave; | |
ff59c456 | 955 | int rv; |
1da177e4 LT |
956 | |
957 | best_slave = bond_find_best_slave(bond); | |
958 | if (best_slave != bond->curr_active_slave) { | |
959 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
960 | rv = bond_set_carrier(bond); |
961 | if (!rv) | |
962 | return; | |
963 | ||
964 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
965 | pr_info("%s: first active interface up!\n", |
966 | bond->dev->name); | |
ff59c456 | 967 | } else { |
a4aee5c8 JP |
968 | pr_info("%s: now running without any active interface !\n", |
969 | bond->dev->name); | |
ff59c456 | 970 | } |
1da177e4 LT |
971 | } |
972 | } | |
973 | ||
974 | /*--------------------------- slave list handling ---------------------------*/ | |
975 | ||
976 | /* | |
977 | * This function attaches the slave to the end of list. | |
978 | * | |
979 | * bond->lock held for writing by caller. | |
980 | */ | |
981 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
982 | { | |
278b2083 | 983 | list_add_tail_rcu(&new_slave->list, &bond->slave_list); |
1da177e4 LT |
984 | bond->slave_cnt++; |
985 | } | |
986 | ||
987 | /* | |
988 | * This function detaches the slave from the list. | |
989 | * WARNING: no check is made to verify if the slave effectively | |
990 | * belongs to <bond>. | |
991 | * Nothing is freed on return, structures are just unchained. | |
992 | * If any slave pointer in bond was pointing to <slave>, | |
993 | * it should be changed by the calling function. | |
994 | * | |
995 | * bond->lock held for writing by caller. | |
996 | */ | |
997 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
998 | { | |
278b2083 | 999 | list_del_rcu(&slave->list); |
1da177e4 LT |
1000 | bond->slave_cnt--; |
1001 | } | |
1002 | ||
f6dc31a8 | 1003 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1004 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1005 | { |
8a8efa22 AW |
1006 | struct netpoll *np; |
1007 | int err = 0; | |
f6dc31a8 | 1008 | |
47be03a2 | 1009 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
1010 | err = -ENOMEM; |
1011 | if (!np) | |
1012 | goto out; | |
1013 | ||
47be03a2 | 1014 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
1015 | if (err) { |
1016 | kfree(np); | |
1017 | goto out; | |
f6dc31a8 | 1018 | } |
8a8efa22 AW |
1019 | slave->np = np; |
1020 | out: | |
1021 | return err; | |
1022 | } | |
1023 | static inline void slave_disable_netpoll(struct slave *slave) | |
1024 | { | |
1025 | struct netpoll *np = slave->np; | |
1026 | ||
1027 | if (!np) | |
1028 | return; | |
1029 | ||
1030 | slave->np = NULL; | |
2cde6acd | 1031 | __netpoll_free_async(np); |
8a8efa22 AW |
1032 | } |
1033 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1034 | { | |
1035 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1036 | return false; | |
1037 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1038 | return false; | |
1039 | return true; | |
f6dc31a8 WC |
1040 | } |
1041 | ||
1042 | static void bond_poll_controller(struct net_device *bond_dev) | |
1043 | { | |
8a8efa22 AW |
1044 | } |
1045 | ||
c4cdef9b | 1046 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1047 | { |
c4cdef9b | 1048 | struct bonding *bond = netdev_priv(bond_dev); |
c2355e1a | 1049 | struct slave *slave; |
c2355e1a | 1050 | |
dec1e90e | 1051 | bond_for_each_slave(bond, slave) |
8a8efa22 AW |
1052 | if (IS_UP(slave->dev)) |
1053 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1054 | } |
8a8efa22 | 1055 | |
47be03a2 | 1056 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1057 | { |
1058 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1059 | struct slave *slave; |
dec1e90e | 1060 | int err = 0; |
f6dc31a8 | 1061 | |
dec1e90e | 1062 | bond_for_each_slave(bond, slave) { |
8a8efa22 AW |
1063 | err = slave_enable_netpoll(slave); |
1064 | if (err) { | |
c4cdef9b | 1065 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1066 | break; |
f6dc31a8 WC |
1067 | } |
1068 | } | |
8a8efa22 | 1069 | return err; |
f6dc31a8 | 1070 | } |
8a8efa22 AW |
1071 | #else |
1072 | static inline int slave_enable_netpoll(struct slave *slave) | |
1073 | { | |
1074 | return 0; | |
1075 | } | |
1076 | static inline void slave_disable_netpoll(struct slave *slave) | |
1077 | { | |
1078 | } | |
f6dc31a8 WC |
1079 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1080 | { | |
1081 | } | |
f6dc31a8 WC |
1082 | #endif |
1083 | ||
1da177e4 LT |
1084 | /*---------------------------------- IOCTL ----------------------------------*/ |
1085 | ||
c8f44aff MM |
1086 | static netdev_features_t bond_fix_features(struct net_device *dev, |
1087 | netdev_features_t features) | |
8531c5ff | 1088 | { |
8531c5ff | 1089 | struct slave *slave; |
b2a103e6 | 1090 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1091 | netdev_features_t mask; |
8531c5ff | 1092 | |
b2a103e6 | 1093 | read_lock(&bond->lock); |
b63365a2 | 1094 | |
dec1e90e | 1095 | if (list_empty(&bond->slave_list)) { |
b2a103e6 MM |
1096 | /* Disable adding VLANs to empty bond. But why? --mq */ |
1097 | features |= NETIF_F_VLAN_CHALLENGED; | |
1098 | goto out; | |
1099 | } | |
b63365a2 | 1100 | |
b2a103e6 | 1101 | mask = features; |
b63365a2 | 1102 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1103 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1104 | |
dec1e90e | 1105 | bond_for_each_slave(bond, slave) { |
b63365a2 HX |
1106 | features = netdev_increment_features(features, |
1107 | slave->dev->features, | |
b2a103e6 MM |
1108 | mask); |
1109 | } | |
b0ce3508 | 1110 | features = netdev_add_tso_features(features, mask); |
b2a103e6 MM |
1111 | |
1112 | out: | |
1113 | read_unlock(&bond->lock); | |
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 | { | |
1123 | struct slave *slave; | |
1124 | struct net_device *bond_dev = bond->dev; | |
c8f44aff | 1125 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1126 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1127 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1128 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b6fe83e9 | 1129 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
b2a103e6 MM |
1130 | |
1131 | read_lock(&bond->lock); | |
1132 | ||
dec1e90e | 1133 | if (list_empty(&bond->slave_list)) |
b2a103e6 MM |
1134 | goto done; |
1135 | ||
dec1e90e | 1136 | bond_for_each_slave(bond, slave) { |
278339a4 | 1137 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1138 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1139 | ||
b6fe83e9 | 1140 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1141 | if (slave->dev->hard_header_len > max_hard_header_len) |
1142 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1143 | |
1144 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1145 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1146 | } |
8531c5ff | 1147 | |
b63365a2 | 1148 | done: |
b2a103e6 | 1149 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1150 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1151 | bond_dev->gso_max_segs = gso_max_segs; |
1152 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1153 | |
b6fe83e9 ED |
1154 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1155 | bond_dev->priv_flags = flags | dst_release_flag; | |
1156 | ||
b2a103e6 MM |
1157 | read_unlock(&bond->lock); |
1158 | ||
1159 | netdev_change_features(bond_dev); | |
8531c5ff AK |
1160 | } |
1161 | ||
872254dd MS |
1162 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1163 | struct net_device *slave_dev) | |
1164 | { | |
00829823 | 1165 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1166 | |
1167 | bond_dev->type = slave_dev->type; | |
1168 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1169 | bond_dev->addr_len = slave_dev->addr_len; | |
1170 | ||
1171 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1172 | slave_dev->addr_len); | |
1173 | } | |
1174 | ||
5b2c4dd2 | 1175 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1176 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1177 | */ |
1178 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1179 | struct slave *slave, |
1180 | struct bonding *bond) | |
5b2c4dd2 | 1181 | { |
2d7011ca | 1182 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1183 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1184 | skb->pkt_type != PACKET_BROADCAST && |
1185 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1186 | return false; |
5b2c4dd2 JP |
1187 | return true; |
1188 | } | |
1189 | return false; | |
1190 | } | |
1191 | ||
8a4eb573 | 1192 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1193 | { |
8a4eb573 | 1194 | struct sk_buff *skb = *pskb; |
f1c1775a | 1195 | struct slave *slave; |
0bd80dad | 1196 | struct bonding *bond; |
de063b70 ED |
1197 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1198 | struct slave *); | |
13a8e0c8 | 1199 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1200 | |
8a4eb573 JP |
1201 | skb = skb_share_check(skb, GFP_ATOMIC); |
1202 | if (unlikely(!skb)) | |
1203 | return RX_HANDLER_CONSUMED; | |
1204 | ||
1205 | *pskb = skb; | |
5b2c4dd2 | 1206 | |
35d48903 JP |
1207 | slave = bond_slave_get_rcu(skb->dev); |
1208 | bond = slave->bond; | |
0bd80dad JP |
1209 | |
1210 | if (bond->params.arp_interval) | |
f1c1775a | 1211 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1212 | |
4d97480b MH |
1213 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1214 | if (recv_probe) { | |
de063b70 ED |
1215 | ret = recv_probe(skb, bond, slave); |
1216 | if (ret == RX_HANDLER_CONSUMED) { | |
1217 | consume_skb(skb); | |
1218 | return ret; | |
3aba891d JP |
1219 | } |
1220 | } | |
1221 | ||
0bd80dad | 1222 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1223 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1224 | } |
1225 | ||
35d48903 | 1226 | skb->dev = bond->dev; |
5b2c4dd2 | 1227 | |
0bd80dad | 1228 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1229 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1230 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1231 | |
541ac7c9 CG |
1232 | if (unlikely(skb_cow_head(skb, |
1233 | skb->data - skb_mac_header(skb)))) { | |
1234 | kfree_skb(skb); | |
8a4eb573 | 1235 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1236 | } |
35d48903 | 1237 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1238 | } |
1239 | ||
13a8e0c8 | 1240 | return ret; |
5b2c4dd2 JP |
1241 | } |
1242 | ||
471cb5a3 JP |
1243 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1244 | struct net_device *slave_dev) | |
1245 | { | |
1246 | int err; | |
1247 | ||
1248 | err = netdev_master_upper_dev_link(slave_dev, bond_dev); | |
1249 | if (err) | |
1250 | return err; | |
1251 | slave_dev->flags |= IFF_SLAVE; | |
1252 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1253 | return 0; | |
1254 | } | |
1255 | ||
1256 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1257 | struct net_device *slave_dev) | |
1258 | { | |
1259 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1260 | slave_dev->flags &= ~IFF_SLAVE; | |
1261 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1262 | } | |
1263 | ||
1da177e4 | 1264 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1265 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1266 | { |
454d7c9b | 1267 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1268 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1269 | struct slave *new_slave = NULL; |
1da177e4 LT |
1270 | struct sockaddr addr; |
1271 | int link_reporting; | |
8599b52e | 1272 | int res = 0, i; |
1da177e4 | 1273 | |
c772dde3 BH |
1274 | if (!bond->params.use_carrier && |
1275 | slave_dev->ethtool_ops->get_link == NULL && | |
1276 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1277 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1278 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1279 | } |
1280 | ||
1da177e4 LT |
1281 | /* already enslaved */ |
1282 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1283 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1284 | return -EBUSY; |
1285 | } | |
1286 | ||
1287 | /* vlan challenged mutual exclusion */ | |
1288 | /* no need to lock since we're protected by rtnl_lock */ | |
1289 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1290 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1291 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1292 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1293 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1294 | return -EPERM; |
1295 | } else { | |
a4aee5c8 JP |
1296 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1297 | bond_dev->name, slave_dev->name, | |
1298 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1299 | } |
1300 | } else { | |
5a03cdb7 | 1301 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1302 | } |
1303 | ||
217df670 JV |
1304 | /* |
1305 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1306 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1307 | * the current ifenslave will set the interface down prior to |
1308 | * enslaving it; the old ifenslave will not. | |
1309 | */ | |
1310 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1311 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1312 | slave_dev->name); |
1313 | res = -EPERM; | |
1314 | goto err_undo_flags; | |
1315 | } | |
1da177e4 | 1316 | |
872254dd MS |
1317 | /* set bonding device ether type by slave - bonding netdevices are |
1318 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1319 | * there is a need to override some of the type dependent attribs/funcs. | |
1320 | * | |
1321 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1322 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1323 | */ | |
dec1e90e | 1324 | if (list_empty(&bond->slave_list)) { |
e36b9d16 | 1325 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1326 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1327 | bond_dev->name, |
1328 | bond_dev->type, slave_dev->type); | |
75c78500 | 1329 | |
b7bc2a5b AW |
1330 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1331 | bond_dev); | |
3ca5b404 JP |
1332 | res = notifier_to_errno(res); |
1333 | if (res) { | |
1334 | pr_err("%s: refused to change device type\n", | |
1335 | bond_dev->name); | |
1336 | res = -EBUSY; | |
1337 | goto err_undo_flags; | |
1338 | } | |
75c78500 | 1339 | |
32a806c1 | 1340 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1341 | dev_uc_flush(bond_dev); |
22bedad3 | 1342 | dev_mc_flush(bond_dev); |
32a806c1 | 1343 | |
e36b9d16 MS |
1344 | if (slave_dev->type != ARPHRD_ETHER) |
1345 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1346 | else { |
e36b9d16 | 1347 | ether_setup(bond_dev); |
550fd08c NH |
1348 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1349 | } | |
75c78500 | 1350 | |
b7bc2a5b AW |
1351 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1352 | bond_dev); | |
e36b9d16 | 1353 | } |
872254dd | 1354 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1355 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1356 | slave_dev->name, | |
1357 | slave_dev->type, bond_dev->type); | |
1358 | res = -EINVAL; | |
1359 | goto err_undo_flags; | |
872254dd MS |
1360 | } |
1361 | ||
eb7cc59a | 1362 | if (slave_ops->ndo_set_mac_address == NULL) { |
dec1e90e | 1363 | if (list_empty(&bond->slave_list)) { |
a4aee5c8 JP |
1364 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1365 | bond_dev->name); | |
3915c1e8 JV |
1366 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1367 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1368 | 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", |
1369 | bond_dev->name); | |
2ab82852 MS |
1370 | res = -EOPNOTSUPP; |
1371 | goto err_undo_flags; | |
1372 | } | |
1da177e4 LT |
1373 | } |
1374 | ||
8d8fc29d AW |
1375 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1376 | ||
c20811a7 JP |
1377 | /* If this is the first slave, then we need to set the master's hardware |
1378 | * address to be the same as the slave's. */ | |
dec1e90e | 1379 | if (list_empty(&bond->slave_list) && |
1380 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) | |
409cc1f8 | 1381 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1382 | |
243cb4e5 | 1383 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1384 | if (!new_slave) { |
1385 | res = -ENOMEM; | |
1386 | goto err_undo_flags; | |
1387 | } | |
dec1e90e | 1388 | INIT_LIST_HEAD(&new_slave->list); |
bb1d9123 AG |
1389 | /* |
1390 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1391 | * is set via sysfs or module option if desired. | |
1392 | */ | |
1393 | new_slave->queue_id = 0; | |
1394 | ||
b15ba0fb JP |
1395 | /* Save slave's original mtu and then set it to match the bond */ |
1396 | new_slave->original_mtu = slave_dev->mtu; | |
1397 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1398 | if (res) { | |
1399 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1400 | goto err_free; | |
1401 | } | |
1402 | ||
217df670 JV |
1403 | /* |
1404 | * Save slave's original ("permanent") mac address for modes | |
1405 | * that need it, and for restoring it upon release, and then | |
1406 | * set it to the master's address | |
1407 | */ | |
1408 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1409 | |
dd957c57 | 1410 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1411 | /* |
1412 | * Set slave to master's mac address. The application already | |
1413 | * set the master's mac address to that of the first slave | |
1414 | */ | |
1415 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1416 | addr.sa_family = slave_dev->type; | |
1417 | res = dev_set_mac_address(slave_dev, &addr); | |
1418 | if (res) { | |
5a03cdb7 | 1419 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1420 | goto err_restore_mtu; |
2ab82852 | 1421 | } |
217df670 | 1422 | } |
1da177e4 | 1423 | |
471cb5a3 | 1424 | res = bond_master_upper_dev_link(bond_dev, slave_dev); |
c2edacf8 | 1425 | if (res) { |
471cb5a3 | 1426 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); |
569f0c4d | 1427 | goto err_restore_mac; |
c2edacf8 | 1428 | } |
5b2c4dd2 | 1429 | |
217df670 JV |
1430 | /* open the slave since the application closed it */ |
1431 | res = dev_open(slave_dev); | |
1432 | if (res) { | |
3d632c3f | 1433 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1434 | goto err_unset_master; |
1da177e4 LT |
1435 | } |
1436 | ||
35d48903 | 1437 | new_slave->bond = bond; |
1da177e4 | 1438 | new_slave->dev = slave_dev; |
0b680e75 | 1439 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1440 | |
58402054 | 1441 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1442 | /* bond_alb_init_slave() must be called before all other stages since |
1443 | * it might fail and we do not want to have to undo everything | |
1444 | */ | |
1445 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1446 | if (res) |
569f0c4d | 1447 | goto err_close; |
1da177e4 LT |
1448 | } |
1449 | ||
303d1cbf JV |
1450 | /* If the mode USES_PRIMARY, then the following is handled by |
1451 | * bond_change_active_slave(). | |
1da177e4 LT |
1452 | */ |
1453 | if (!USES_PRIMARY(bond->params.mode)) { | |
1454 | /* set promiscuity level to new slave */ | |
1455 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1456 | res = dev_set_promiscuity(slave_dev, 1); |
1457 | if (res) | |
1458 | goto err_close; | |
1da177e4 LT |
1459 | } |
1460 | ||
1461 | /* set allmulti level to new slave */ | |
1462 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1463 | res = dev_set_allmulti(slave_dev, 1); |
1464 | if (res) | |
1465 | goto err_close; | |
1da177e4 LT |
1466 | } |
1467 | ||
b9e40857 | 1468 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1469 | |
1470 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1471 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1472 | ||
b9e40857 | 1473 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1474 | } |
1475 | ||
1476 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1477 | /* add lacpdu mc addr to mc list */ | |
1478 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1479 | ||
22bedad3 | 1480 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1481 | } |
1482 | ||
b8e2fde4 WY |
1483 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1484 | if (res) { | |
1ff412ad | 1485 | pr_err("%s: Error: Couldn't add bond vlan ids to %s\n", |
1486 | bond_dev->name, slave_dev->name); | |
1487 | goto err_close; | |
1488 | } | |
1da177e4 LT |
1489 | |
1490 | write_lock_bh(&bond->lock); | |
1491 | ||
1492 | bond_attach_slave(bond, new_slave); | |
1493 | ||
1494 | new_slave->delay = 0; | |
1495 | new_slave->link_failure_count = 0; | |
1496 | ||
3915c1e8 JV |
1497 | write_unlock_bh(&bond->lock); |
1498 | ||
b2a103e6 MM |
1499 | bond_compute_features(bond); |
1500 | ||
876254ae VF |
1501 | bond_update_speed_duplex(new_slave); |
1502 | ||
3915c1e8 JV |
1503 | read_lock(&bond->lock); |
1504 | ||
f31c7937 MK |
1505 | new_slave->last_arp_rx = jiffies - |
1506 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
8599b52e VF |
1507 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
1508 | new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx; | |
f5b2b966 | 1509 | |
1da177e4 LT |
1510 | if (bond->params.miimon && !bond->params.use_carrier) { |
1511 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1512 | ||
1513 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1514 | /* | |
1515 | * miimon is set but a bonded network driver | |
1516 | * does not support ETHTOOL/MII and | |
1517 | * arp_interval is not set. Note: if | |
1518 | * use_carrier is enabled, we will never go | |
1519 | * here (because netif_carrier is always | |
1520 | * supported); thus, we don't need to change | |
1521 | * the messages for netif_carrier. | |
1522 | */ | |
a4aee5c8 | 1523 | 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 | 1524 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1525 | } else if (link_reporting == -1) { |
1526 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1527 | 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", |
1528 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1529 | } |
1530 | } | |
1531 | ||
1532 | /* check for initial state */ | |
f31c7937 MK |
1533 | if (bond->params.miimon) { |
1534 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1535 | if (bond->params.updelay) { | |
1536 | new_slave->link = BOND_LINK_BACK; | |
1537 | new_slave->delay = bond->params.updelay; | |
1538 | } else { | |
1539 | new_slave->link = BOND_LINK_UP; | |
1540 | } | |
1da177e4 | 1541 | } else { |
f31c7937 | 1542 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1543 | } |
f31c7937 MK |
1544 | } else if (bond->params.arp_interval) { |
1545 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1546 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1547 | } else { |
f31c7937 | 1548 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1549 | } |
1550 | ||
f31c7937 MK |
1551 | if (new_slave->link != BOND_LINK_DOWN) |
1552 | new_slave->jiffies = jiffies; | |
1553 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1554 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1555 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1556 | ||
1da177e4 LT |
1557 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1558 | /* if there is a primary slave, remember it */ | |
a549952a | 1559 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1560 | bond->primary_slave = new_slave; |
a549952a JP |
1561 | bond->force_primary = true; |
1562 | } | |
1da177e4 LT |
1563 | } |
1564 | ||
3915c1e8 JV |
1565 | write_lock_bh(&bond->curr_slave_lock); |
1566 | ||
1da177e4 LT |
1567 | switch (bond->params.mode) { |
1568 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1569 | bond_set_slave_inactive_flags(new_slave); |
1570 | bond_select_active_slave(bond); | |
1da177e4 LT |
1571 | break; |
1572 | case BOND_MODE_8023AD: | |
1573 | /* in 802.3ad mode, the internal mechanism | |
1574 | * will activate the slaves in the selected | |
1575 | * aggregator | |
1576 | */ | |
1577 | bond_set_slave_inactive_flags(new_slave); | |
1578 | /* if this is the first slave */ | |
dec1e90e | 1579 | if (bond_first_slave(bond) == new_slave) { |
1da177e4 LT |
1580 | SLAVE_AD_INFO(new_slave).id = 1; |
1581 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1582 | * can be called only after the mac address of the bond is set | |
1583 | */ | |
56d00c67 | 1584 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1585 | } else { |
dec1e90e | 1586 | struct slave *prev_slave; |
1587 | ||
1588 | prev_slave = bond_prev_slave(bond, new_slave); | |
1da177e4 | 1589 | SLAVE_AD_INFO(new_slave).id = |
dec1e90e | 1590 | SLAVE_AD_INFO(prev_slave).id + 1; |
1da177e4 LT |
1591 | } |
1592 | ||
1593 | bond_3ad_bind_slave(new_slave); | |
1594 | break; | |
1595 | case BOND_MODE_TLB: | |
1596 | case BOND_MODE_ALB: | |
e30bc066 | 1597 | bond_set_active_slave(new_slave); |
059fe7a5 | 1598 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1599 | bond_select_active_slave(bond); |
1da177e4 LT |
1600 | break; |
1601 | default: | |
5a03cdb7 | 1602 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1603 | |
1604 | /* always active in trunk mode */ | |
e30bc066 | 1605 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1606 | |
1607 | /* In trunking mode there is little meaning to curr_active_slave | |
1608 | * anyway (it holds no special properties of the bond device), | |
1609 | * so we can change it without calling change_active_interface() | |
1610 | */ | |
4e591b93 | 1611 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
278b2083 | 1612 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1613 | |
1da177e4 LT |
1614 | break; |
1615 | } /* switch(bond_mode) */ | |
1616 | ||
3915c1e8 JV |
1617 | write_unlock_bh(&bond->curr_slave_lock); |
1618 | ||
ff59c456 JV |
1619 | bond_set_carrier(bond); |
1620 | ||
f6dc31a8 | 1621 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1622 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1623 | if (slave_dev->npinfo) { |
1624 | if (slave_enable_netpoll(new_slave)) { | |
1625 | read_unlock(&bond->lock); | |
1626 | pr_info("Error, %s: master_dev is using netpoll, " | |
1627 | "but new slave device does not support netpoll.\n", | |
1628 | bond_dev->name); | |
1629 | res = -EBUSY; | |
f7d9821a | 1630 | goto err_detach; |
8a8efa22 | 1631 | } |
f6dc31a8 WC |
1632 | } |
1633 | #endif | |
8a8efa22 | 1634 | |
3915c1e8 | 1635 | read_unlock(&bond->lock); |
1da177e4 | 1636 | |
b76cdba9 MW |
1637 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1638 | if (res) | |
f7d9821a | 1639 | goto err_detach; |
b76cdba9 | 1640 | |
35d48903 JP |
1641 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1642 | new_slave); | |
1643 | if (res) { | |
1644 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1645 | goto err_dest_symlinks; | |
1646 | } | |
1647 | ||
a4aee5c8 JP |
1648 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1649 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1650 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1651 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1652 | |
1653 | /* enslave is successful */ | |
1654 | return 0; | |
1655 | ||
1656 | /* Undo stages on error */ | |
35d48903 JP |
1657 | err_dest_symlinks: |
1658 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1659 | ||
f7d9821a | 1660 | err_detach: |
303d1cbf JV |
1661 | if (!USES_PRIMARY(bond->params.mode)) |
1662 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1663 | ||
1ff412ad | 1664 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
f7d9821a | 1665 | write_lock_bh(&bond->lock); |
1666 | bond_detach_slave(bond, new_slave); | |
3c5913b5 | 1667 | if (bond->primary_slave == new_slave) |
1668 | bond->primary_slave = NULL; | |
3c5913b5 | 1669 | if (bond->curr_active_slave == new_slave) { |
c6cdcf6d | 1670 | bond_change_active_slave(bond, NULL); |
1671 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1672 | read_lock(&bond->lock); |
1673 | write_lock_bh(&bond->curr_slave_lock); | |
3c5913b5 | 1674 | bond_select_active_slave(bond); |
1675 | write_unlock_bh(&bond->curr_slave_lock); | |
1676 | read_unlock(&bond->lock); | |
c6cdcf6d | 1677 | } else { |
1678 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1679 | } |
fc7a72ac | 1680 | slave_disable_netpoll(new_slave); |
f7d9821a | 1681 | |
1da177e4 | 1682 | err_close: |
b6a5a7b9 | 1683 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1684 | dev_close(slave_dev); |
1685 | ||
569f0c4d | 1686 | err_unset_master: |
471cb5a3 | 1687 | bond_upper_dev_unlink(bond_dev, slave_dev); |
569f0c4d | 1688 | |
1da177e4 | 1689 | err_restore_mac: |
dd957c57 | 1690 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1691 | /* XXX TODO - fom follow mode needs to change master's |
1692 | * MAC if this slave's MAC is in use by the bond, or at | |
1693 | * least print a warning. | |
1694 | */ | |
2ab82852 MS |
1695 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1696 | addr.sa_family = slave_dev->type; | |
1697 | dev_set_mac_address(slave_dev, &addr); | |
1698 | } | |
1da177e4 | 1699 | |
b15ba0fb JP |
1700 | err_restore_mtu: |
1701 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1702 | ||
1da177e4 LT |
1703 | err_free: |
1704 | kfree(new_slave); | |
1705 | ||
1706 | err_undo_flags: | |
b2a103e6 | 1707 | bond_compute_features(bond); |
b8fad459 | 1708 | /* Enslave of first slave has failed and we need to fix master's mac */ |
dec1e90e | 1709 | if (list_empty(&bond->slave_list) && |
b8fad459 NA |
1710 | ether_addr_equal(bond_dev->dev_addr, slave_dev->dev_addr)) |
1711 | eth_hw_addr_random(bond_dev); | |
3d632c3f | 1712 | |
1da177e4 LT |
1713 | return res; |
1714 | } | |
1715 | ||
1716 | /* | |
1717 | * Try to release the slave device <slave> from the bond device <master> | |
1718 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1719 | * is write-locked. If "all" is true it means that the function is being called |
1720 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1721 | * |
1722 | * The rules for slave state should be: | |
1723 | * for Active/Backup: | |
1724 | * Active stays on all backups go down | |
1725 | * for Bonded connections: | |
1726 | * The first up interface should be left on and all others downed. | |
1727 | */ | |
0896341a | 1728 | static int __bond_release_one(struct net_device *bond_dev, |
1729 | struct net_device *slave_dev, | |
1730 | bool all) | |
1da177e4 | 1731 | { |
454d7c9b | 1732 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1733 | struct slave *slave, *oldcurrent; |
1734 | struct sockaddr addr; | |
c8f44aff | 1735 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1736 | |
1737 | /* slave is not a slave or master is not master of this slave */ | |
1738 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1739 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1740 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1741 | bond_dev->name, slave_dev->name); |
1742 | return -EINVAL; | |
1743 | } | |
1744 | ||
e843fa50 | 1745 | block_netpoll_tx(); |
1da177e4 LT |
1746 | write_lock_bh(&bond->lock); |
1747 | ||
1748 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1749 | if (!slave) { | |
1750 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1751 | pr_info("%s: %s not enslaved\n", |
1752 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1753 | write_unlock_bh(&bond->lock); |
e843fa50 | 1754 | unblock_netpoll_tx(); |
1da177e4 LT |
1755 | return -EINVAL; |
1756 | } | |
1757 | ||
fcd99434 | 1758 | write_unlock_bh(&bond->lock); |
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 | ||
471cb5a3 | 1877 | bond_upper_dev_unlink(bond_dev, slave_dev); |
1da177e4 | 1878 | |
8a8efa22 | 1879 | slave_disable_netpoll(slave); |
f6dc31a8 | 1880 | |
1da177e4 LT |
1881 | /* close slave before restoring its mac address */ |
1882 | dev_close(slave_dev); | |
1883 | ||
3915c1e8 | 1884 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1885 | /* restore original ("permanent") mac address */ |
1886 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1887 | addr.sa_family = slave_dev->type; | |
1888 | dev_set_mac_address(slave_dev, &addr); | |
1889 | } | |
1da177e4 | 1890 | |
b15ba0fb JP |
1891 | dev_set_mtu(slave_dev, slave->original_mtu); |
1892 | ||
2d7011ca | 1893 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1894 | |
1895 | kfree(slave); | |
1896 | ||
1897 | return 0; /* deletion OK */ | |
1898 | } | |
1899 | ||
0896341a | 1900 | /* A wrapper used because of ndo_del_link */ |
1901 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1902 | { | |
1903 | return __bond_release_one(bond_dev, slave_dev, false); | |
1904 | } | |
1905 | ||
d90a162a | 1906 | /* |
dadaa10b | 1907 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
1908 | * Must be under rtnl_lock when this function is called. |
1909 | */ | |
26d8ee75 | 1910 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1911 | struct net_device *slave_dev) | |
d90a162a | 1912 | { |
454d7c9b | 1913 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1914 | int ret; |
1915 | ||
1916 | ret = bond_release(bond_dev, slave_dev); | |
dec1e90e | 1917 | if (ret == 0 && list_empty(&bond->slave_list)) { |
8a8efa22 | 1918 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
1919 | pr_info("%s: destroying bond %s.\n", |
1920 | bond_dev->name, bond_dev->name); | |
9e71626c | 1921 | unregister_netdevice(bond_dev); |
d90a162a MS |
1922 | } |
1923 | return ret; | |
1924 | } | |
1925 | ||
1da177e4 LT |
1926 | /* |
1927 | * This function changes the active slave to slave <slave_dev>. | |
1928 | * It returns -EINVAL in the following cases. | |
1929 | * - <slave_dev> is not found in the list. | |
1930 | * - There is not active slave now. | |
1931 | * - <slave_dev> is already active. | |
1932 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
1933 | * - <slave_dev> is not running. | |
3d632c3f SH |
1934 | * In these cases, this function does nothing. |
1935 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
1936 | */ |
1937 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
1938 | { | |
454d7c9b | 1939 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1940 | struct slave *old_active = NULL; |
1941 | struct slave *new_active = NULL; | |
1942 | int res = 0; | |
1943 | ||
3d632c3f | 1944 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1945 | return -EINVAL; |
1da177e4 | 1946 | |
471cb5a3 JP |
1947 | /* Verify that bond_dev is indeed the master of slave_dev */ |
1948 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1949 | !netdev_has_upper_dev(slave_dev, bond_dev)) | |
1da177e4 | 1950 | return -EINVAL; |
1da177e4 | 1951 | |
059fe7a5 | 1952 | read_lock(&bond->lock); |
1da177e4 LT |
1953 | |
1954 | old_active = bond->curr_active_slave; | |
1955 | new_active = bond_get_slave_by_dev(bond, slave_dev); | |
1da177e4 LT |
1956 | /* |
1957 | * Changing to the current active: do nothing; return success. | |
1958 | */ | |
71bc3b2d | 1959 | if (new_active && new_active == old_active) { |
059fe7a5 | 1960 | read_unlock(&bond->lock); |
1da177e4 LT |
1961 | return 0; |
1962 | } | |
1963 | ||
71bc3b2d | 1964 | if (new_active && |
1965 | old_active && | |
1966 | new_active->link == BOND_LINK_UP && | |
1da177e4 | 1967 | IS_UP(new_active->dev)) { |
e843fa50 | 1968 | block_netpoll_tx(); |
059fe7a5 | 1969 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 1970 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 1971 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 1972 | unblock_netpoll_tx(); |
3d632c3f | 1973 | } else |
1da177e4 | 1974 | res = -EINVAL; |
1da177e4 | 1975 | |
059fe7a5 | 1976 | read_unlock(&bond->lock); |
1da177e4 LT |
1977 | |
1978 | return res; | |
1979 | } | |
1980 | ||
1da177e4 LT |
1981 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1982 | { | |
454d7c9b | 1983 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1984 | |
1985 | info->bond_mode = bond->params.mode; | |
1986 | info->miimon = bond->params.miimon; | |
1987 | ||
6603a6f2 | 1988 | read_lock(&bond->lock); |
1da177e4 | 1989 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 1990 | read_unlock(&bond->lock); |
1da177e4 LT |
1991 | |
1992 | return 0; | |
1993 | } | |
1994 | ||
1995 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1996 | { | |
454d7c9b | 1997 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 1998 | int i = 0, res = -ENODEV; |
1da177e4 | 1999 | struct slave *slave; |
1da177e4 | 2000 | |
6603a6f2 | 2001 | read_lock(&bond->lock); |
dec1e90e | 2002 | bond_for_each_slave(bond, slave) { |
2003 | if (i++ == (int)info->slave_id) { | |
689c96cc ED |
2004 | res = 0; |
2005 | strcpy(info->slave_name, slave->dev->name); | |
2006 | info->link = slave->link; | |
e30bc066 | 2007 | info->state = bond_slave_state(slave); |
689c96cc | 2008 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2009 | break; |
2010 | } | |
2011 | } | |
6603a6f2 | 2012 | read_unlock(&bond->lock); |
1da177e4 | 2013 | |
689c96cc | 2014 | return res; |
1da177e4 LT |
2015 | } |
2016 | ||
2017 | /*-------------------------------- Monitoring -------------------------------*/ | |
2018 | ||
1da177e4 | 2019 | |
f0c76d61 JV |
2020 | static int bond_miimon_inspect(struct bonding *bond) |
2021 | { | |
dec1e90e | 2022 | int link_state, commit = 0; |
f0c76d61 | 2023 | struct slave *slave; |
41f89100 JP |
2024 | bool ignore_updelay; |
2025 | ||
2026 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 | 2027 | |
dec1e90e | 2028 | bond_for_each_slave(bond, slave) { |
f0c76d61 | 2029 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2030 | |
f0c76d61 | 2031 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2032 | |
2033 | switch (slave->link) { | |
f0c76d61 JV |
2034 | case BOND_LINK_UP: |
2035 | if (link_state) | |
2036 | continue; | |
1da177e4 | 2037 | |
f0c76d61 JV |
2038 | slave->link = BOND_LINK_FAIL; |
2039 | slave->delay = bond->params.downdelay; | |
2040 | if (slave->delay) { | |
a4aee5c8 JP |
2041 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2042 | bond->dev->name, | |
2043 | (bond->params.mode == | |
2044 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2045 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2046 | "active " : "backup ") : "", |
2047 | slave->dev->name, | |
2048 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2049 | } |
f0c76d61 JV |
2050 | /*FALLTHRU*/ |
2051 | case BOND_LINK_FAIL: | |
2052 | if (link_state) { | |
2053 | /* | |
2054 | * recovered before downdelay expired | |
2055 | */ | |
2056 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2057 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2058 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2059 | bond->dev->name, | |
2060 | (bond->params.downdelay - slave->delay) * | |
2061 | bond->params.miimon, | |
2062 | slave->dev->name); | |
f0c76d61 | 2063 | continue; |
1da177e4 | 2064 | } |
f0c76d61 JV |
2065 | |
2066 | if (slave->delay <= 0) { | |
2067 | slave->new_link = BOND_LINK_DOWN; | |
2068 | commit++; | |
2069 | continue; | |
1da177e4 | 2070 | } |
1da177e4 | 2071 | |
f0c76d61 JV |
2072 | slave->delay--; |
2073 | break; | |
2074 | ||
2075 | case BOND_LINK_DOWN: | |
2076 | if (!link_state) | |
2077 | continue; | |
2078 | ||
2079 | slave->link = BOND_LINK_BACK; | |
2080 | slave->delay = bond->params.updelay; | |
2081 | ||
2082 | if (slave->delay) { | |
a4aee5c8 JP |
2083 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2084 | bond->dev->name, slave->dev->name, | |
2085 | ignore_updelay ? 0 : | |
2086 | bond->params.updelay * | |
2087 | bond->params.miimon); | |
f0c76d61 JV |
2088 | } |
2089 | /*FALLTHRU*/ | |
2090 | case BOND_LINK_BACK: | |
2091 | if (!link_state) { | |
2092 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2093 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2094 | bond->dev->name, | |
2095 | (bond->params.updelay - slave->delay) * | |
2096 | bond->params.miimon, | |
2097 | slave->dev->name); | |
f0c76d61 JV |
2098 | |
2099 | continue; | |
2100 | } | |
2101 | ||
41f89100 JP |
2102 | if (ignore_updelay) |
2103 | slave->delay = 0; | |
2104 | ||
f0c76d61 JV |
2105 | if (slave->delay <= 0) { |
2106 | slave->new_link = BOND_LINK_UP; | |
2107 | commit++; | |
41f89100 | 2108 | ignore_updelay = false; |
f0c76d61 | 2109 | continue; |
1da177e4 | 2110 | } |
f0c76d61 JV |
2111 | |
2112 | slave->delay--; | |
1da177e4 | 2113 | break; |
f0c76d61 JV |
2114 | } |
2115 | } | |
1da177e4 | 2116 | |
f0c76d61 JV |
2117 | return commit; |
2118 | } | |
1da177e4 | 2119 | |
f0c76d61 JV |
2120 | static void bond_miimon_commit(struct bonding *bond) |
2121 | { | |
2122 | struct slave *slave; | |
f0c76d61 | 2123 | |
dec1e90e | 2124 | bond_for_each_slave(bond, slave) { |
f0c76d61 JV |
2125 | switch (slave->new_link) { |
2126 | case BOND_LINK_NOCHANGE: | |
2127 | continue; | |
1da177e4 | 2128 | |
f0c76d61 JV |
2129 | case BOND_LINK_UP: |
2130 | slave->link = BOND_LINK_UP; | |
2131 | slave->jiffies = jiffies; | |
2132 | ||
2133 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2134 | /* prevent it from being the active one */ | |
e30bc066 | 2135 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2136 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2137 | /* make it immediately active */ | |
e30bc066 | 2138 | bond_set_active_slave(slave); |
f0c76d61 JV |
2139 | } else if (slave != bond->primary_slave) { |
2140 | /* prevent it from being the active one */ | |
e30bc066 | 2141 | bond_set_backup_slave(slave); |
1da177e4 | 2142 | } |
1da177e4 | 2143 | |
5d30530e | 2144 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 | 2145 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
2146 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
2147 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2148 | |
f0c76d61 JV |
2149 | /* notify ad that the link status has changed */ |
2150 | if (bond->params.mode == BOND_MODE_8023AD) | |
2151 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2152 | |
58402054 | 2153 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2154 | bond_alb_handle_link_change(bond, slave, |
2155 | BOND_LINK_UP); | |
1da177e4 | 2156 | |
f0c76d61 JV |
2157 | if (!bond->curr_active_slave || |
2158 | (slave == bond->primary_slave)) | |
2159 | goto do_failover; | |
1da177e4 | 2160 | |
f0c76d61 | 2161 | continue; |
059fe7a5 | 2162 | |
f0c76d61 | 2163 | case BOND_LINK_DOWN: |
fba4acda JV |
2164 | if (slave->link_failure_count < UINT_MAX) |
2165 | slave->link_failure_count++; | |
2166 | ||
f0c76d61 | 2167 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2168 | |
f0c76d61 JV |
2169 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2170 | bond->params.mode == BOND_MODE_8023AD) | |
2171 | bond_set_slave_inactive_flags(slave); | |
2172 | ||
a4aee5c8 JP |
2173 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2174 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2175 | |
2176 | if (bond->params.mode == BOND_MODE_8023AD) | |
2177 | bond_3ad_handle_link_change(slave, | |
2178 | BOND_LINK_DOWN); | |
2179 | ||
ae63e808 | 2180 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2181 | bond_alb_handle_link_change(bond, slave, |
2182 | BOND_LINK_DOWN); | |
2183 | ||
2184 | if (slave == bond->curr_active_slave) | |
2185 | goto do_failover; | |
2186 | ||
2187 | continue; | |
2188 | ||
2189 | default: | |
a4aee5c8 | 2190 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2191 | bond->dev->name, slave->new_link, |
2192 | slave->dev->name); | |
2193 | slave->new_link = BOND_LINK_NOCHANGE; | |
2194 | ||
2195 | continue; | |
2196 | } | |
2197 | ||
2198 | do_failover: | |
2199 | ASSERT_RTNL(); | |
e843fa50 | 2200 | block_netpoll_tx(); |
f0c76d61 JV |
2201 | write_lock_bh(&bond->curr_slave_lock); |
2202 | bond_select_active_slave(bond); | |
2203 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2204 | unblock_netpoll_tx(); |
f0c76d61 JV |
2205 | } |
2206 | ||
2207 | bond_set_carrier(bond); | |
1da177e4 LT |
2208 | } |
2209 | ||
0b0eef66 JV |
2210 | /* |
2211 | * bond_mii_monitor | |
2212 | * | |
2213 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2214 | * inspection, then (if inspection indicates something needs to be done) |
2215 | * an acquisition of appropriate locks followed by a commit phase to | |
2216 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2217 | */ |
2218 | void bond_mii_monitor(struct work_struct *work) | |
2219 | { | |
2220 | struct bonding *bond = container_of(work, struct bonding, | |
2221 | mii_work.work); | |
ad246c99 | 2222 | bool should_notify_peers = false; |
e6d265e8 | 2223 | unsigned long delay; |
0b0eef66 JV |
2224 | |
2225 | read_lock(&bond->lock); | |
e6d265e8 JV |
2226 | |
2227 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 | 2228 | |
dec1e90e | 2229 | if (list_empty(&bond->slave_list)) |
f0c76d61 | 2230 | goto re_arm; |
b59f9f74 | 2231 | |
ad246c99 BH |
2232 | should_notify_peers = bond_should_notify_peers(bond); |
2233 | ||
f0c76d61 | 2234 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2235 | read_unlock(&bond->lock); |
e6d265e8 JV |
2236 | |
2237 | /* Race avoidance with bond_close cancel of workqueue */ | |
2238 | if (!rtnl_trylock()) { | |
2239 | read_lock(&bond->lock); | |
2240 | delay = 1; | |
2241 | should_notify_peers = false; | |
2242 | goto re_arm; | |
2243 | } | |
2244 | ||
0b0eef66 | 2245 | read_lock(&bond->lock); |
f0c76d61 JV |
2246 | |
2247 | bond_miimon_commit(bond); | |
2248 | ||
5655662d JV |
2249 | read_unlock(&bond->lock); |
2250 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2251 | read_lock(&bond->lock); | |
0b0eef66 JV |
2252 | } |
2253 | ||
f0c76d61 | 2254 | re_arm: |
e6d265e8 JV |
2255 | if (bond->params.miimon) |
2256 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2257 | ||
0b0eef66 | 2258 | read_unlock(&bond->lock); |
ad246c99 BH |
2259 | |
2260 | if (should_notify_peers) { | |
6c8d23f7 | 2261 | if (!rtnl_trylock()) |
e6d265e8 | 2262 | return; |
b7bc2a5b | 2263 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2264 | rtnl_unlock(); |
2265 | } | |
0b0eef66 | 2266 | } |
c3ade5ca | 2267 | |
50223ce4 | 2268 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2269 | { |
50223ce4 VF |
2270 | struct net_device *upper; |
2271 | struct list_head *iter; | |
2272 | bool ret = false; | |
f5b2b966 | 2273 | |
eaddcd76 | 2274 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2275 | return true; |
f5b2b966 | 2276 | |
50223ce4 VF |
2277 | rcu_read_lock(); |
2278 | netdev_for_each_upper_dev_rcu(bond->dev, upper, iter) { | |
2279 | if (ip == bond_confirm_addr(upper, 0, ip)) { | |
2280 | ret = true; | |
2281 | break; | |
2282 | } | |
f5b2b966 | 2283 | } |
50223ce4 | 2284 | rcu_read_unlock(); |
f5b2b966 | 2285 | |
50223ce4 | 2286 | return ret; |
f5b2b966 JV |
2287 | } |
2288 | ||
c3ade5ca JV |
2289 | /* |
2290 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2291 | * switches in VLAN mode (especially if ports are configured as | |
2292 | * "native" to a VLAN) might not pass non-tagged frames. | |
2293 | */ | |
d3bb52b0 | 2294 | 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 |
2295 | { |
2296 | struct sk_buff *skb; | |
2297 | ||
acca2674 | 2298 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op, |
2299 | slave_dev->name, &dest_ip, &src_ip, vlan_id); | |
3d632c3f | 2300 | |
c3ade5ca JV |
2301 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2302 | NULL, slave_dev->dev_addr, NULL); | |
2303 | ||
2304 | if (!skb) { | |
a4aee5c8 | 2305 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2306 | return; |
2307 | } | |
2308 | if (vlan_id) { | |
1fd9b1fc | 2309 | skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id); |
c3ade5ca | 2310 | if (!skb) { |
a4aee5c8 | 2311 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2312 | return; |
2313 | } | |
2314 | } | |
2315 | arp_xmit(skb); | |
2316 | } | |
2317 | ||
2318 | ||
1da177e4 LT |
2319 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2320 | { | |
27bc11e6 VF |
2321 | struct net_device *upper, *vlan_upper; |
2322 | struct list_head *iter, *vlan_iter; | |
c3ade5ca | 2323 | struct rtable *rt; |
27bc11e6 VF |
2324 | __be32 *targets = bond->params.arp_targets, addr; |
2325 | int i, vlan_id; | |
1da177e4 | 2326 | |
27bc11e6 | 2327 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
acca2674 | 2328 | pr_debug("basa: target %pI4\n", &targets[i]); |
c3ade5ca | 2329 | |
27bc11e6 | 2330 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2331 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2332 | RTO_ONLINK, 0); | |
b23dd4fe | 2333 | if (IS_ERR(rt)) { |
3e32582f VF |
2334 | pr_debug("%s: no route to arp_ip_target %pI4\n", |
2335 | bond->dev->name, &targets[i]); | |
c3ade5ca JV |
2336 | continue; |
2337 | } | |
2338 | ||
c3ade5ca | 2339 | vlan_id = 0; |
27bc11e6 VF |
2340 | |
2341 | /* bond device itself */ | |
2342 | if (rt->dst.dev == bond->dev) | |
2343 | goto found; | |
2344 | ||
2345 | rcu_read_lock(); | |
2346 | /* first we search only for vlan devices. for every vlan | |
2347 | * found we verify its upper dev list, searching for the | |
2348 | * rt->dst.dev. If found we save the tag of the vlan and | |
2349 | * proceed to send the packet. | |
2350 | * | |
2351 | * TODO: QinQ? | |
2352 | */ | |
2353 | netdev_for_each_upper_dev_rcu(bond->dev, vlan_upper, vlan_iter) { | |
2354 | if (!is_vlan_dev(vlan_upper)) | |
2355 | continue; | |
2356 | netdev_for_each_upper_dev_rcu(vlan_upper, upper, iter) { | |
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 | */ | |
2369 | netdev_for_each_upper_dev_rcu(bond->dev, upper, iter) { | |
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 | ||
de063b70 ED |
2415 | static 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; |
58402054 | 3213 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3214 | /* Must be called only after all |
3215 | * slaves have been released | |
3216 | */ | |
3217 | bond_alb_deinitialize(bond); | |
3218 | } | |
3aba891d | 3219 | bond->recv_probe = NULL; |
1da177e4 LT |
3220 | |
3221 | return 0; | |
3222 | } | |
3223 | ||
28172739 ED |
3224 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3225 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3226 | { |
454d7c9b | 3227 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3228 | struct rtnl_link_stats64 temp; |
1da177e4 | 3229 | struct slave *slave; |
1da177e4 | 3230 | |
28172739 | 3231 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3232 | |
3233 | read_lock_bh(&bond->lock); | |
dec1e90e | 3234 | bond_for_each_slave(bond, slave) { |
be1f3c2c | 3235 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3236 | dev_get_stats(slave->dev, &temp); |
3237 | ||
3238 | stats->rx_packets += sstats->rx_packets; | |
3239 | stats->rx_bytes += sstats->rx_bytes; | |
3240 | stats->rx_errors += sstats->rx_errors; | |
3241 | stats->rx_dropped += sstats->rx_dropped; | |
3242 | ||
3243 | stats->tx_packets += sstats->tx_packets; | |
3244 | stats->tx_bytes += sstats->tx_bytes; | |
3245 | stats->tx_errors += sstats->tx_errors; | |
3246 | stats->tx_dropped += sstats->tx_dropped; | |
3247 | ||
3248 | stats->multicast += sstats->multicast; | |
3249 | stats->collisions += sstats->collisions; | |
3250 | ||
3251 | stats->rx_length_errors += sstats->rx_length_errors; | |
3252 | stats->rx_over_errors += sstats->rx_over_errors; | |
3253 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3254 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3255 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3256 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3257 | ||
3258 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3259 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3260 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3261 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3262 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb | 3263 | } |
1da177e4 LT |
3264 | read_unlock_bh(&bond->lock); |
3265 | ||
3266 | return stats; | |
3267 | } | |
3268 | ||
3269 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3270 | { | |
3271 | struct net_device *slave_dev = NULL; | |
3272 | struct ifbond k_binfo; | |
3273 | struct ifbond __user *u_binfo = NULL; | |
3274 | struct ifslave k_sinfo; | |
3275 | struct ifslave __user *u_sinfo = NULL; | |
3276 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3277 | struct net *net; |
1da177e4 LT |
3278 | int res = 0; |
3279 | ||
a4aee5c8 | 3280 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3281 | |
3282 | switch (cmd) { | |
1da177e4 LT |
3283 | case SIOCGMIIPHY: |
3284 | mii = if_mii(ifr); | |
3d632c3f | 3285 | if (!mii) |
1da177e4 | 3286 | return -EINVAL; |
3d632c3f | 3287 | |
1da177e4 LT |
3288 | mii->phy_id = 0; |
3289 | /* Fall Through */ | |
3290 | case SIOCGMIIREG: | |
3291 | /* | |
3292 | * We do this again just in case we were called by SIOCGMIIREG | |
3293 | * instead of SIOCGMIIPHY. | |
3294 | */ | |
3295 | mii = if_mii(ifr); | |
3d632c3f | 3296 | if (!mii) |
1da177e4 | 3297 | return -EINVAL; |
3d632c3f | 3298 | |
1da177e4 LT |
3299 | |
3300 | if (mii->reg_num == 1) { | |
454d7c9b | 3301 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3302 | mii->val_out = 0; |
6603a6f2 | 3303 | read_lock(&bond->lock); |
1da177e4 | 3304 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3305 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3306 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3307 | |
1da177e4 | 3308 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3309 | read_unlock(&bond->lock); |
1da177e4 LT |
3310 | } |
3311 | ||
3312 | return 0; | |
3313 | case BOND_INFO_QUERY_OLD: | |
3314 | case SIOCBONDINFOQUERY: | |
3315 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3316 | ||
3d632c3f | 3317 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3318 | return -EFAULT; |
1da177e4 LT |
3319 | |
3320 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3321 | if (res == 0 && |
3322 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3323 | return -EFAULT; | |
1da177e4 LT |
3324 | |
3325 | return res; | |
3326 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3327 | case SIOCBONDSLAVEINFOQUERY: | |
3328 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3329 | ||
3d632c3f | 3330 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3331 | return -EFAULT; |
1da177e4 LT |
3332 | |
3333 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3334 | if (res == 0 && |
3335 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3336 | return -EFAULT; | |
1da177e4 LT |
3337 | |
3338 | return res; | |
3339 | default: | |
3340 | /* Go on */ | |
3341 | break; | |
3342 | } | |
3343 | ||
387ff911 G |
3344 | net = dev_net(bond_dev); |
3345 | ||
3346 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3347 | return -EPERM; |
1da177e4 | 3348 | |
387ff911 | 3349 | slave_dev = dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3350 | |
a4aee5c8 | 3351 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3352 | |
3d632c3f | 3353 | if (!slave_dev) |
1da177e4 | 3354 | res = -ENODEV; |
3d632c3f | 3355 | else { |
a4aee5c8 | 3356 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3357 | switch (cmd) { |
3358 | case BOND_ENSLAVE_OLD: | |
3359 | case SIOCBONDENSLAVE: | |
3360 | res = bond_enslave(bond_dev, slave_dev); | |
3361 | break; | |
3362 | case BOND_RELEASE_OLD: | |
3363 | case SIOCBONDRELEASE: | |
3364 | res = bond_release(bond_dev, slave_dev); | |
3365 | break; | |
3366 | case BOND_SETHWADDR_OLD: | |
3367 | case SIOCBONDSETHWADDR: | |
409cc1f8 JP |
3368 | bond_set_dev_addr(bond_dev, slave_dev); |
3369 | res = 0; | |
1da177e4 LT |
3370 | break; |
3371 | case BOND_CHANGE_ACTIVE_OLD: | |
3372 | case SIOCBONDCHANGEACTIVE: | |
3373 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3374 | break; | |
3375 | default: | |
3376 | res = -EOPNOTSUPP; | |
3377 | } | |
3378 | ||
3379 | dev_put(slave_dev); | |
3380 | } | |
3381 | ||
1da177e4 LT |
3382 | return res; |
3383 | } | |
3384 | ||
d03462b9 | 3385 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3386 | { |
454d7c9b | 3387 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3388 | |
d03462b9 JP |
3389 | if (change & IFF_PROMISC) |
3390 | bond_set_promiscuity(bond, | |
3391 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3392 | |
d03462b9 JP |
3393 | if (change & IFF_ALLMULTI) |
3394 | bond_set_allmulti(bond, | |
3395 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3396 | } | |
1da177e4 | 3397 | |
303d1cbf | 3398 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3399 | { |
3400 | struct bonding *bond = netdev_priv(bond_dev); | |
303d1cbf | 3401 | struct slave *slave; |
1da177e4 | 3402 | |
7864a1ad | 3403 | ASSERT_RTNL(); |
80ee5ad2 | 3404 | |
303d1cbf | 3405 | if (USES_PRIMARY(bond->params.mode)) { |
7864a1ad | 3406 | slave = rtnl_dereference(bond->curr_active_slave); |
303d1cbf JV |
3407 | if (slave) { |
3408 | dev_uc_sync(slave->dev, bond_dev); | |
3409 | dev_mc_sync(slave->dev, bond_dev); | |
3410 | } | |
303d1cbf | 3411 | } else { |
dec1e90e | 3412 | bond_for_each_slave(bond, slave) { |
303d1cbf JV |
3413 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3414 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3415 | } | |
1da177e4 | 3416 | } |
1da177e4 LT |
3417 | } |
3418 | ||
234bcf8a | 3419 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3420 | { |
234bcf8a | 3421 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3422 | const struct net_device_ops *slave_ops; |
3423 | struct neigh_parms parms; | |
dec1e90e | 3424 | struct slave *slave; |
234bcf8a SP |
3425 | int ret; |
3426 | ||
dec1e90e | 3427 | slave = bond_first_slave(bond); |
234bcf8a SP |
3428 | if (!slave) |
3429 | return 0; | |
234bcf8a | 3430 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3431 | if (!slave_ops->ndo_neigh_setup) |
3432 | return 0; | |
3433 | ||
3434 | parms.neigh_setup = NULL; | |
3435 | parms.neigh_cleanup = NULL; | |
3436 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3437 | if (ret) | |
3438 | return ret; | |
3439 | ||
3440 | /* | |
3441 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3442 | * after the last slave has been detached. Assumes that all slaves | |
3443 | * utilize the same neigh_cleanup (true at this writing as only user | |
3444 | * is ipoib). | |
3445 | */ | |
3446 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3447 | ||
3448 | if (!parms.neigh_setup) | |
3449 | return 0; | |
3450 | ||
3451 | return parms.neigh_setup(n); | |
3452 | } | |
3453 | ||
3454 | /* | |
3455 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3456 | * slave exists. So we must declare proxy setup function which will | |
3457 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3458 | * |
3459 | * It's also called by master devices (such as vlans) to setup their | |
3460 | * underlying devices. In that case - do nothing, we're already set up from | |
3461 | * our init. | |
234bcf8a SP |
3462 | */ |
3463 | static int bond_neigh_setup(struct net_device *dev, | |
3464 | struct neigh_parms *parms) | |
3465 | { | |
9918d5bf VF |
3466 | /* modify only our neigh_parms */ |
3467 | if (parms->dev == dev) | |
3468 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3469 | |
00829823 SH |
3470 | return 0; |
3471 | } | |
3472 | ||
1da177e4 LT |
3473 | /* |
3474 | * Change the MTU of all of a master's slaves to match the master | |
3475 | */ | |
3476 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3477 | { | |
454d7c9b | 3478 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 3479 | struct slave *slave; |
1da177e4 | 3480 | int res = 0; |
1da177e4 | 3481 | |
5a03cdb7 | 3482 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3483 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3484 | |
3485 | /* Can't hold bond->lock with bh disabled here since | |
3486 | * some base drivers panic. On the other hand we can't | |
3487 | * hold bond->lock without bh disabled because we'll | |
3488 | * deadlock. The only solution is to rely on the fact | |
3489 | * that we're under rtnl_lock here, and the slaves | |
3490 | * list won't change. This doesn't solve the problem | |
3491 | * of setting the slave's MTU while it is | |
3492 | * transmitting, but the assumption is that the base | |
3493 | * driver can handle that. | |
3494 | * | |
3495 | * TODO: figure out a way to safely iterate the slaves | |
3496 | * list, but without holding a lock around the actual | |
3497 | * call to the base driver. | |
3498 | */ | |
3499 | ||
dec1e90e | 3500 | bond_for_each_slave(bond, slave) { |
a4aee5c8 JP |
3501 | pr_debug("s %p s->p %p c_m %p\n", |
3502 | slave, | |
dec1e90e | 3503 | bond_prev_slave(bond, slave), |
a4aee5c8 | 3504 | slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 3505 | |
1da177e4 LT |
3506 | res = dev_set_mtu(slave->dev, new_mtu); |
3507 | ||
3508 | if (res) { | |
3509 | /* If we failed to set the slave's mtu to the new value | |
3510 | * we must abort the operation even in ACTIVE_BACKUP | |
3511 | * mode, because if we allow the backup slaves to have | |
3512 | * different mtu values than the active slave we'll | |
3513 | * need to change their mtu when doing a failover. That | |
3514 | * means changing their mtu from timer context, which | |
3515 | * is probably not a good idea. | |
3516 | */ | |
5a03cdb7 | 3517 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3518 | goto unwind; |
3519 | } | |
3520 | } | |
3521 | ||
3522 | bond_dev->mtu = new_mtu; | |
3523 | ||
3524 | return 0; | |
3525 | ||
3526 | unwind: | |
3527 | /* unwind from head to the slave that failed */ | |
dec1e90e | 3528 | bond_for_each_slave_continue_reverse(bond, slave) { |
1da177e4 LT |
3529 | int tmp_res; |
3530 | ||
3531 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3532 | if (tmp_res) { | |
a4aee5c8 JP |
3533 | pr_debug("unwind err %d dev %s\n", |
3534 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3535 | } |
3536 | } | |
3537 | ||
3538 | return res; | |
3539 | } | |
3540 | ||
3541 | /* | |
3542 | * Change HW address | |
3543 | * | |
3544 | * Note that many devices must be down to change the HW address, and | |
3545 | * downing the master releases all slaves. We can make bonds full of | |
3546 | * bonding devices to test this, however. | |
3547 | */ | |
3548 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3549 | { | |
454d7c9b | 3550 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3551 | struct sockaddr *sa = addr, tmp_sa; |
dec1e90e | 3552 | struct slave *slave; |
1da177e4 | 3553 | int res = 0; |
1da177e4 | 3554 | |
eb7cc59a SH |
3555 | if (bond->params.mode == BOND_MODE_ALB) |
3556 | return bond_alb_set_mac_address(bond_dev, addr); | |
3557 | ||
3558 | ||
a4aee5c8 JP |
3559 | pr_debug("bond=%p, name=%s\n", |
3560 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3561 | |
1b5acd29 JV |
3562 | /* If fail_over_mac is enabled, do nothing and return success. |
3563 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3564 | */ |
1b5acd29 | 3565 | if (bond->params.fail_over_mac) |
dd957c57 | 3566 | return 0; |
2ab82852 | 3567 | |
3d632c3f | 3568 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3569 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3570 | |
3571 | /* Can't hold bond->lock with bh disabled here since | |
3572 | * some base drivers panic. On the other hand we can't | |
3573 | * hold bond->lock without bh disabled because we'll | |
3574 | * deadlock. The only solution is to rely on the fact | |
3575 | * that we're under rtnl_lock here, and the slaves | |
3576 | * list won't change. This doesn't solve the problem | |
3577 | * of setting the slave's hw address while it is | |
3578 | * transmitting, but the assumption is that the base | |
3579 | * driver can handle that. | |
3580 | * | |
3581 | * TODO: figure out a way to safely iterate the slaves | |
3582 | * list, but without holding a lock around the actual | |
3583 | * call to the base driver. | |
3584 | */ | |
3585 | ||
dec1e90e | 3586 | bond_for_each_slave(bond, slave) { |
eb7cc59a | 3587 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3588 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3589 | |
eb7cc59a | 3590 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3591 | res = -EOPNOTSUPP; |
5a03cdb7 | 3592 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3593 | goto unwind; |
3594 | } | |
3595 | ||
3596 | res = dev_set_mac_address(slave->dev, addr); | |
3597 | if (res) { | |
3598 | /* TODO: consider downing the slave | |
3599 | * and retry ? | |
3600 | * User should expect communications | |
3601 | * breakage anyway until ARP finish | |
3602 | * updating, so... | |
3603 | */ | |
5a03cdb7 | 3604 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3605 | goto unwind; |
3606 | } | |
3607 | } | |
3608 | ||
3609 | /* success */ | |
3610 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3611 | return 0; | |
3612 | ||
3613 | unwind: | |
3614 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3615 | tmp_sa.sa_family = bond_dev->type; | |
3616 | ||
3617 | /* unwind from head to the slave that failed */ | |
dec1e90e | 3618 | bond_for_each_slave_continue_reverse(bond, slave) { |
1da177e4 LT |
3619 | int tmp_res; |
3620 | ||
3621 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
3622 | if (tmp_res) { | |
a4aee5c8 JP |
3623 | pr_debug("unwind err %d dev %s\n", |
3624 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3625 | } |
3626 | } | |
3627 | ||
3628 | return res; | |
3629 | } | |
3630 | ||
15077228 NA |
3631 | /** |
3632 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3633 | * @bond: bonding device that is transmitting | |
3634 | * @skb: buffer to transmit | |
3635 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3636 | * | |
3637 | * This function tries to transmit through slave with slave_id but in case | |
3638 | * it fails, it tries to find the first available slave for transmission. | |
3639 | * The skb is consumed in all cases, thus the function is void. | |
3640 | */ | |
3641 | void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) | |
3642 | { | |
3643 | struct slave *slave; | |
3644 | int i = slave_id; | |
3645 | ||
3646 | /* Here we start from the slave with slave_id */ | |
278b2083 | 3647 | bond_for_each_slave_rcu(bond, slave) { |
15077228 NA |
3648 | if (--i < 0) { |
3649 | if (slave_can_tx(slave)) { | |
3650 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3651 | return; | |
3652 | } | |
3653 | } | |
3654 | } | |
3655 | ||
3656 | /* Here we start from the first slave up to slave_id */ | |
3657 | i = slave_id; | |
278b2083 | 3658 | bond_for_each_slave_rcu(bond, slave) { |
15077228 NA |
3659 | if (--i < 0) |
3660 | break; | |
3661 | if (slave_can_tx(slave)) { | |
3662 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3663 | return; | |
3664 | } | |
3665 | } | |
3666 | /* no slave that can tx has been found */ | |
3667 | kfree_skb(skb); | |
3668 | } | |
3669 | ||
1da177e4 LT |
3670 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3671 | { | |
454d7c9b | 3672 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3673 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3674 | struct slave *slave; |
1da177e4 | 3675 | |
cf5f9044 | 3676 | /* |
a2fd940f AG |
3677 | * Start with the curr_active_slave that joined the bond as the |
3678 | * default for sending IGMP traffic. For failover purposes one | |
3679 | * needs to maintain some consistency for the interface that will | |
3680 | * send the join/membership reports. The curr_active_slave found | |
3681 | * will send all of this type of traffic. | |
cf5f9044 | 3682 | */ |
15077228 | 3683 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3684 | slave = rcu_dereference(bond->curr_active_slave); |
15077228 NA |
3685 | if (slave && slave_can_tx(slave)) |
3686 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3687 | else | |
3688 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3689 | } else { |
15077228 NA |
3690 | bond_xmit_slave_id(bond, skb, |
3691 | bond->rr_tx_counter++ % bond->slave_cnt); | |
1da177e4 | 3692 | } |
0693e88e | 3693 | |
ec634fe3 | 3694 | return NETDEV_TX_OK; |
1da177e4 LT |
3695 | } |
3696 | ||
3697 | /* | |
3698 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3699 | * the bond has a usable interface. | |
3700 | */ | |
3701 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3702 | { | |
454d7c9b | 3703 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3704 | struct slave *slave; |
1da177e4 | 3705 | |
278b2083 | 3706 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3707 | if (slave) |
15077228 NA |
3708 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3709 | else | |
04502430 | 3710 | kfree_skb(skb); |
0693e88e | 3711 | |
ec634fe3 | 3712 | return NETDEV_TX_OK; |
1da177e4 LT |
3713 | } |
3714 | ||
3715 | /* | |
169a3e66 JV |
3716 | * In bond_xmit_xor() , we determine the output device by using a pre- |
3717 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
3718 | * find the next active slave. | |
1da177e4 LT |
3719 | */ |
3720 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3721 | { | |
454d7c9b | 3722 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3723 | |
15077228 NA |
3724 | bond_xmit_slave_id(bond, skb, |
3725 | bond->xmit_hash_policy(skb, bond->slave_cnt)); | |
0693e88e | 3726 | |
ec634fe3 | 3727 | return NETDEV_TX_OK; |
1da177e4 LT |
3728 | } |
3729 | ||
78a646ce | 3730 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3731 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3732 | { | |
454d7c9b | 3733 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3734 | struct slave *slave = NULL; |
1da177e4 | 3735 | |
278b2083 | 3736 | bond_for_each_slave_rcu(bond, slave) { |
78a646ce NA |
3737 | if (bond_is_last_slave(bond, slave)) |
3738 | break; | |
3739 | if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { | |
3740 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1da177e4 | 3741 | |
78a646ce NA |
3742 | if (!skb2) { |
3743 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", | |
3744 | bond_dev->name); | |
3745 | continue; | |
1da177e4 | 3746 | } |
78a646ce NA |
3747 | /* bond_dev_queue_xmit always returns 0 */ |
3748 | bond_dev_queue_xmit(bond, skb2, slave->dev); | |
1da177e4 LT |
3749 | } |
3750 | } | |
78a646ce NA |
3751 | if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP) |
3752 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3753 | else | |
04502430 | 3754 | kfree_skb(skb); |
3d632c3f | 3755 | |
ec634fe3 | 3756 | return NETDEV_TX_OK; |
1da177e4 LT |
3757 | } |
3758 | ||
3759 | /*------------------------- Device initialization ---------------------------*/ | |
3760 | ||
6f6652be JV |
3761 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
3762 | { | |
3763 | switch (bond->params.xmit_policy) { | |
3764 | case BOND_XMIT_POLICY_LAYER23: | |
3765 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
3766 | break; | |
3767 | case BOND_XMIT_POLICY_LAYER34: | |
3768 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
3769 | break; | |
3770 | case BOND_XMIT_POLICY_LAYER2: | |
3771 | default: | |
3772 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
3773 | break; | |
3774 | } | |
3775 | } | |
3776 | ||
bb1d9123 AG |
3777 | /* |
3778 | * Lookup the slave that corresponds to a qid | |
3779 | */ | |
3780 | static inline int bond_slave_override(struct bonding *bond, | |
3781 | struct sk_buff *skb) | |
3782 | { | |
bb1d9123 AG |
3783 | struct slave *slave = NULL; |
3784 | struct slave *check_slave; | |
dec1e90e | 3785 | int res = 1; |
bb1d9123 | 3786 | |
0693e88e MM |
3787 | if (!skb->queue_mapping) |
3788 | return 1; | |
bb1d9123 AG |
3789 | |
3790 | /* Find out if any slaves have the same mapping as this skb. */ | |
278b2083 | 3791 | bond_for_each_slave_rcu(bond, check_slave) { |
bb1d9123 AG |
3792 | if (check_slave->queue_id == skb->queue_mapping) { |
3793 | slave = check_slave; | |
3794 | break; | |
3795 | } | |
3796 | } | |
3797 | ||
3798 | /* If the slave isn't UP, use default transmit policy. */ | |
3799 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
3800 | (slave->link == BOND_LINK_UP)) { | |
3801 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
3802 | } | |
3803 | ||
bb1d9123 AG |
3804 | return res; |
3805 | } | |
3806 | ||
374eeb5a | 3807 | |
bb1d9123 AG |
3808 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
3809 | { | |
3810 | /* | |
3811 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 3812 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3813 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3814 | * way down to the bonding driver. | |
3815 | */ | |
fd0e435b PO |
3816 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3817 | ||
374eeb5a NH |
3818 | /* |
3819 | * Save the original txq to restore before passing to the driver | |
3820 | */ | |
df4ab5b3 | 3821 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3822 | |
fd0e435b | 3823 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3824 | do { |
fd0e435b | 3825 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3826 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3827 | } |
3828 | return txq; | |
bb1d9123 AG |
3829 | } |
3830 | ||
0693e88e | 3831 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3832 | { |
bb1d9123 AG |
3833 | struct bonding *bond = netdev_priv(dev); |
3834 | ||
3835 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
3836 | if (!bond_slave_override(bond, skb)) | |
3837 | return NETDEV_TX_OK; | |
3838 | } | |
00829823 SH |
3839 | |
3840 | switch (bond->params.mode) { | |
3841 | case BOND_MODE_ROUNDROBIN: | |
3842 | return bond_xmit_roundrobin(skb, dev); | |
3843 | case BOND_MODE_ACTIVEBACKUP: | |
3844 | return bond_xmit_activebackup(skb, dev); | |
3845 | case BOND_MODE_XOR: | |
3846 | return bond_xmit_xor(skb, dev); | |
3847 | case BOND_MODE_BROADCAST: | |
3848 | return bond_xmit_broadcast(skb, dev); | |
3849 | case BOND_MODE_8023AD: | |
3850 | return bond_3ad_xmit_xor(skb, dev); | |
3851 | case BOND_MODE_ALB: | |
3852 | case BOND_MODE_TLB: | |
3853 | return bond_alb_xmit(skb, dev); | |
3854 | default: | |
3855 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3856 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3857 | dev->name, bond->params.mode); | |
00829823 | 3858 | WARN_ON_ONCE(1); |
04502430 | 3859 | kfree_skb(skb); |
00829823 SH |
3860 | return NETDEV_TX_OK; |
3861 | } | |
3862 | } | |
3863 | ||
0693e88e MM |
3864 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3865 | { | |
3866 | struct bonding *bond = netdev_priv(dev); | |
3867 | netdev_tx_t ret = NETDEV_TX_OK; | |
3868 | ||
3869 | /* | |
3870 | * If we risk deadlock from transmitting this in the | |
3871 | * netpoll path, tell netpoll to queue the frame for later tx | |
3872 | */ | |
3873 | if (is_netpoll_tx_blocked(dev)) | |
3874 | return NETDEV_TX_BUSY; | |
3875 | ||
278b2083 | 3876 | rcu_read_lock(); |
dec1e90e | 3877 | if (!list_empty(&bond->slave_list)) |
0693e88e MM |
3878 | ret = __bond_start_xmit(skb, dev); |
3879 | else | |
04502430 | 3880 | kfree_skb(skb); |
278b2083 | 3881 | rcu_read_unlock(); |
0693e88e MM |
3882 | |
3883 | return ret; | |
3884 | } | |
00829823 | 3885 | |
1da177e4 LT |
3886 | /* |
3887 | * set bond mode specific net device operations | |
3888 | */ | |
a77b5325 | 3889 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 3890 | { |
169a3e66 JV |
3891 | struct net_device *bond_dev = bond->dev; |
3892 | ||
1da177e4 LT |
3893 | switch (mode) { |
3894 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
3895 | break; |
3896 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
3897 | break; |
3898 | case BOND_MODE_XOR: | |
6f6652be | 3899 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
3900 | break; |
3901 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
3902 | break; |
3903 | case BOND_MODE_8023AD: | |
6f6652be | 3904 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 3905 | break; |
1da177e4 | 3906 | case BOND_MODE_ALB: |
8f903c70 JV |
3907 | /* FALLTHRU */ |
3908 | case BOND_MODE_TLB: | |
1da177e4 LT |
3909 | break; |
3910 | default: | |
3911 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3912 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3913 | bond_dev->name, mode); | |
1da177e4 LT |
3914 | break; |
3915 | } | |
3916 | } | |
3917 | ||
bb5b052f AG |
3918 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3919 | struct ethtool_cmd *ecmd) | |
3920 | { | |
3921 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3922 | unsigned long speed = 0; |
dec1e90e | 3923 | struct slave *slave; |
bb5b052f AG |
3924 | |
3925 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3926 | ecmd->port = PORT_OTHER; | |
3927 | ||
3928 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
3929 | * do not need to check mode. Though link speed might not represent | |
3930 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3931 | * this is an accurate maximum. | |
3932 | */ | |
3933 | read_lock(&bond->lock); | |
dec1e90e | 3934 | bond_for_each_slave(bond, slave) { |
bb5b052f AG |
3935 | if (SLAVE_IS_OK(slave)) { |
3936 | if (slave->speed != SPEED_UNKNOWN) | |
3937 | speed += slave->speed; | |
3938 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3939 | slave->duplex != DUPLEX_UNKNOWN) | |
3940 | ecmd->duplex = slave->duplex; | |
3941 | } | |
3942 | } | |
3943 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
3944 | read_unlock(&bond->lock); | |
dec1e90e | 3945 | |
bb5b052f AG |
3946 | return 0; |
3947 | } | |
3948 | ||
217df670 | 3949 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3950 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3951 | { |
7826d43f JP |
3952 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3953 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3954 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3955 | BOND_ABI_VERSION); | |
217df670 JV |
3956 | } |
3957 | ||
7282d491 | 3958 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3959 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3960 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3961 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3962 | }; |
3963 | ||
eb7cc59a | 3964 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3965 | .ndo_init = bond_init, |
9e71626c | 3966 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3967 | .ndo_open = bond_open, |
3968 | .ndo_stop = bond_close, | |
00829823 | 3969 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3970 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3971 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3972 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3973 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 3974 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 3975 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 3976 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 3977 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 3978 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 3979 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 3980 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 3981 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
3982 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
3983 | .ndo_poll_controller = bond_poll_controller, | |
3984 | #endif | |
9232ecca JP |
3985 | .ndo_add_slave = bond_enslave, |
3986 | .ndo_del_slave = bond_release, | |
b2a103e6 | 3987 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
3988 | }; |
3989 | ||
b3f92b63 DG |
3990 | static const struct device_type bond_type = { |
3991 | .name = "bond", | |
3992 | }; | |
3993 | ||
9e2e61fb AW |
3994 | static void bond_destructor(struct net_device *bond_dev) |
3995 | { | |
3996 | struct bonding *bond = netdev_priv(bond_dev); | |
3997 | if (bond->wq) | |
3998 | destroy_workqueue(bond->wq); | |
3999 | free_netdev(bond_dev); | |
4000 | } | |
4001 | ||
181470fc | 4002 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4003 | { |
454d7c9b | 4004 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4005 | |
1da177e4 LT |
4006 | /* initialize rwlocks */ |
4007 | rwlock_init(&bond->lock); | |
4008 | rwlock_init(&bond->curr_slave_lock); | |
dec1e90e | 4009 | INIT_LIST_HEAD(&bond->slave_list); |
d2991f75 | 4010 | bond->params = bonding_defaults; |
1da177e4 LT |
4011 | |
4012 | /* Initialize pointers */ | |
1da177e4 | 4013 | bond->dev = bond_dev; |
1da177e4 LT |
4014 | |
4015 | /* Initialize the device entry points */ | |
181470fc | 4016 | ether_setup(bond_dev); |
eb7cc59a | 4017 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4018 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4019 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4020 | |
9e2e61fb | 4021 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4022 | |
b3f92b63 DG |
4023 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4024 | ||
1da177e4 LT |
4025 | /* Initialize the device options */ |
4026 | bond_dev->tx_queue_len = 0; | |
4027 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4028 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4029 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4030 | |
1da177e4 LT |
4031 | /* At first, we block adding VLANs. That's the only way to |
4032 | * prevent problems that occur when adding VLANs over an | |
4033 | * empty bond. The block will be removed once non-challenged | |
4034 | * slaves are enslaved. | |
4035 | */ | |
4036 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4037 | ||
932ff279 | 4038 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4039 | * transmitting */ |
4040 | bond_dev->features |= NETIF_F_LLTX; | |
4041 | ||
4042 | /* By default, we declare the bond to be fully | |
4043 | * VLAN hardware accelerated capable. Special | |
4044 | * care is taken in the various xmit functions | |
4045 | * when there are slaves that are not hw accel | |
4046 | * capable | |
4047 | */ | |
1da177e4 | 4048 | |
b2a103e6 | 4049 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4050 | NETIF_F_HW_VLAN_CTAG_TX | |
4051 | NETIF_F_HW_VLAN_CTAG_RX | | |
4052 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4053 | |
34324dc2 | 4054 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4055 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4056 | } |
4057 | ||
c67dfb29 EB |
4058 | /* |
4059 | * Destroy a bonding device. | |
4060 | * Must be under rtnl_lock when this function is called. | |
4061 | */ | |
4062 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4063 | { |
454d7c9b | 4064 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 4065 | struct slave *slave, *tmp_slave; |
a434e43f | 4066 | |
f6dc31a8 WC |
4067 | bond_netpoll_cleanup(bond_dev); |
4068 | ||
c67dfb29 | 4069 | /* Release the bonded slaves */ |
dec1e90e | 4070 | list_for_each_entry_safe(slave, tmp_slave, &bond->slave_list, list) |
4071 | __bond_release_one(bond_dev, slave->dev, true); | |
0896341a | 4072 | pr_info("%s: released all slaves\n", bond_dev->name); |
c67dfb29 | 4073 | |
a434e43f JV |
4074 | list_del(&bond->bond_list); |
4075 | ||
f073c7ca | 4076 | bond_debug_unregister(bond); |
a434e43f JV |
4077 | } |
4078 | ||
1da177e4 LT |
4079 | /*------------------------- Module initialization ---------------------------*/ |
4080 | ||
4081 | /* | |
4082 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4083 | * number of the mode or its string name. A bit complicated because |
4084 | * some mode names are substrings of other names, and calls from sysfs | |
4085 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4086 | */ |
325dcf7a | 4087 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4088 | { |
54b87323 | 4089 | int modeint = -1, i, rv; |
a42e534f | 4090 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4091 | |
a42e534f JV |
4092 | for (p = (char *)buf; *p; p++) |
4093 | if (!(isdigit(*p) || isspace(*p))) | |
4094 | break; | |
4095 | ||
4096 | if (*p) | |
ece95f7f | 4097 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4098 | else |
54b87323 | 4099 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4100 | |
4101 | if (!rv) | |
4102 | return -1; | |
1da177e4 LT |
4103 | |
4104 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4105 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4106 | return tbl[i].mode; |
4107 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4108 | return tbl[i].mode; |
1da177e4 LT |
4109 | } |
4110 | ||
4111 | return -1; | |
4112 | } | |
4113 | ||
4114 | static int bond_check_params(struct bond_params *params) | |
4115 | { | |
5a5c5fd4 | 4116 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
8599b52e | 4117 | int arp_all_targets_value; |
f5b2b966 | 4118 | |
1da177e4 LT |
4119 | /* |
4120 | * Convert string parameters. | |
4121 | */ | |
4122 | if (mode) { | |
4123 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4124 | if (bond_mode == -1) { | |
a4aee5c8 | 4125 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4126 | mode == NULL ? "NULL" : mode); |
4127 | return -EINVAL; | |
4128 | } | |
4129 | } | |
4130 | ||
169a3e66 JV |
4131 | if (xmit_hash_policy) { |
4132 | if ((bond_mode != BOND_MODE_XOR) && | |
4133 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4134 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4135 | bond_mode_name(bond_mode)); |
4136 | } else { | |
4137 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4138 | xmit_hashtype_tbl); | |
4139 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4140 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4141 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4142 | xmit_hash_policy); |
4143 | return -EINVAL; | |
4144 | } | |
4145 | } | |
4146 | } | |
4147 | ||
1da177e4 LT |
4148 | if (lacp_rate) { |
4149 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4150 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4151 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4152 | } else { |
4153 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4154 | if (lacp_fast == -1) { | |
a4aee5c8 | 4155 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4156 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4157 | return -EINVAL; | |
4158 | } | |
4159 | } | |
4160 | } | |
4161 | ||
fd989c83 JV |
4162 | if (ad_select) { |
4163 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4164 | if (params->ad_select == -1) { | |
a4aee5c8 | 4165 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4166 | ad_select == NULL ? "NULL" : ad_select); |
4167 | return -EINVAL; | |
4168 | } | |
4169 | ||
4170 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4171 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4172 | } |
4173 | } else { | |
4174 | params->ad_select = BOND_AD_STABLE; | |
4175 | } | |
4176 | ||
f5841306 | 4177 | if (max_bonds < 0) { |
a4aee5c8 JP |
4178 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4179 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4180 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4181 | } | |
4182 | ||
4183 | if (miimon < 0) { | |
a4aee5c8 JP |
4184 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4185 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4186 | miimon = BOND_LINK_MON_INTERV; |
4187 | } | |
4188 | ||
4189 | if (updelay < 0) { | |
a4aee5c8 JP |
4190 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4191 | updelay, INT_MAX); | |
1da177e4 LT |
4192 | updelay = 0; |
4193 | } | |
4194 | ||
4195 | if (downdelay < 0) { | |
a4aee5c8 JP |
4196 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4197 | downdelay, INT_MAX); | |
1da177e4 LT |
4198 | downdelay = 0; |
4199 | } | |
4200 | ||
4201 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4202 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4203 | use_carrier); | |
1da177e4 LT |
4204 | use_carrier = 1; |
4205 | } | |
4206 | ||
ad246c99 BH |
4207 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4208 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4209 | num_peer_notif); | |
4210 | num_peer_notif = 1; | |
4211 | } | |
4212 | ||
1da177e4 LT |
4213 | /* reset values for 802.3ad */ |
4214 | if (bond_mode == BOND_MODE_8023AD) { | |
4215 | if (!miimon) { | |
a4aee5c8 | 4216 | 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 | 4217 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4218 | miimon = 100; |
4219 | } | |
4220 | } | |
4221 | ||
bb1d9123 AG |
4222 | if (tx_queues < 1 || tx_queues > 255) { |
4223 | pr_warning("Warning: tx_queues (%d) should be between " | |
4224 | "1 and 255, resetting to %d\n", | |
4225 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4226 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4227 | } | |
4228 | ||
ebd8e497 AG |
4229 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4230 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4231 | "not of valid value (0/1), so it was set to " | |
4232 | "0\n", all_slaves_active); | |
4233 | all_slaves_active = 0; | |
4234 | } | |
4235 | ||
c2952c31 FL |
4236 | if (resend_igmp < 0 || resend_igmp > 255) { |
4237 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4238 | "0 and 255, resetting to %d\n", | |
4239 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4240 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4241 | } | |
4242 | ||
1da177e4 LT |
4243 | /* reset values for TLB/ALB */ |
4244 | if ((bond_mode == BOND_MODE_TLB) || | |
4245 | (bond_mode == BOND_MODE_ALB)) { | |
4246 | if (!miimon) { | |
a4aee5c8 | 4247 | 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 | 4248 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4249 | miimon = 100; |
4250 | } | |
4251 | } | |
4252 | ||
4253 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4254 | 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", |
4255 | updelay); | |
1da177e4 LT |
4256 | } |
4257 | ||
4258 | if (!miimon) { | |
4259 | if (updelay || downdelay) { | |
4260 | /* just warn the user the up/down delay will have | |
4261 | * no effect since miimon is zero... | |
4262 | */ | |
a4aee5c8 JP |
4263 | 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", |
4264 | updelay, downdelay); | |
1da177e4 LT |
4265 | } |
4266 | } else { | |
4267 | /* don't allow arp monitoring */ | |
4268 | if (arp_interval) { | |
a4aee5c8 JP |
4269 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4270 | miimon, arp_interval); | |
1da177e4 LT |
4271 | arp_interval = 0; |
4272 | } | |
4273 | ||
4274 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4275 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4276 | updelay, miimon, | |
4277 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4278 | } |
4279 | ||
4280 | updelay /= miimon; | |
4281 | ||
4282 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4283 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4284 | downdelay, miimon, | |
4285 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4286 | } |
4287 | ||
4288 | downdelay /= miimon; | |
4289 | } | |
4290 | ||
4291 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4292 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4293 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4294 | arp_interval = BOND_LINK_ARP_INTERV; |
4295 | } | |
4296 | ||
5a5c5fd4 | 4297 | for (arp_ip_count = 0, i = 0; |
4298 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4299 | /* not complete check, but should be good enough to |
4300 | catch mistakes */ | |
5a5c5fd4 | 4301 | __be32 ip = in_aton(arp_ip_target[i]); |
4302 | if (!isdigit(arp_ip_target[i][0]) || ip == 0 || | |
4303 | ip == htonl(INADDR_BROADCAST)) { | |
a4aee5c8 | 4304 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5a5c5fd4 | 4305 | arp_ip_target[i]); |
1da177e4 LT |
4306 | arp_interval = 0; |
4307 | } else { | |
0afee4e8 VF |
4308 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4309 | arp_target[arp_ip_count++] = ip; | |
4310 | else | |
4311 | pr_warning("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", | |
4312 | &ip); | |
1da177e4 LT |
4313 | } |
4314 | } | |
4315 | ||
4316 | if (arp_interval && !arp_ip_count) { | |
4317 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4318 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4319 | arp_interval); | |
1da177e4 LT |
4320 | arp_interval = 0; |
4321 | } | |
4322 | ||
f5b2b966 JV |
4323 | if (arp_validate) { |
4324 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4325 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4326 | return -EINVAL; |
4327 | } | |
4328 | if (!arp_interval) { | |
a4aee5c8 | 4329 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4330 | return -EINVAL; |
4331 | } | |
4332 | ||
4333 | arp_validate_value = bond_parse_parm(arp_validate, | |
4334 | arp_validate_tbl); | |
4335 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4336 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4337 | arp_validate == NULL ? "NULL" : arp_validate); |
4338 | return -EINVAL; | |
4339 | } | |
4340 | } else | |
4341 | arp_validate_value = 0; | |
4342 | ||
8599b52e VF |
4343 | arp_all_targets_value = 0; |
4344 | if (arp_all_targets) { | |
4345 | arp_all_targets_value = bond_parse_parm(arp_all_targets, | |
4346 | arp_all_targets_tbl); | |
4347 | ||
4348 | if (arp_all_targets_value == -1) { | |
4349 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", | |
4350 | arp_all_targets); | |
4351 | arp_all_targets_value = 0; | |
4352 | } | |
4353 | } | |
4354 | ||
1da177e4 | 4355 | if (miimon) { |
a4aee5c8 | 4356 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4357 | } else if (arp_interval) { |
a4aee5c8 JP |
4358 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4359 | arp_interval, | |
4360 | arp_validate_tbl[arp_validate_value].modename, | |
4361 | arp_ip_count); | |
1da177e4 LT |
4362 | |
4363 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4364 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4365 | |
e5e2a8fd | 4366 | pr_info("\n"); |
1da177e4 | 4367 | |
b8a9787e | 4368 | } else if (max_bonds) { |
1da177e4 LT |
4369 | /* miimon and arp_interval not set, we need one so things |
4370 | * work as expected, see bonding.txt for details | |
4371 | */ | |
b2730f4f | 4372 | 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 |
4373 | } |
4374 | ||
4375 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4376 | /* currently, using a primary only makes sense | |
4377 | * in active backup, TLB or ALB modes | |
4378 | */ | |
a4aee5c8 JP |
4379 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4380 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4381 | primary = NULL; |
4382 | } | |
4383 | ||
a549952a JP |
4384 | if (primary && primary_reselect) { |
4385 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4386 | pri_reselect_tbl); | |
4387 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4388 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4389 | primary_reselect == |
4390 | NULL ? "NULL" : primary_reselect); | |
4391 | return -EINVAL; | |
4392 | } | |
4393 | } else { | |
4394 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4395 | } | |
4396 | ||
3915c1e8 JV |
4397 | if (fail_over_mac) { |
4398 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4399 | fail_over_mac_tbl); | |
4400 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4401 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4402 | arp_validate == NULL ? "NULL" : arp_validate); |
4403 | return -EINVAL; | |
4404 | } | |
4405 | ||
4406 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4407 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4408 | } else { |
4409 | fail_over_mac_value = BOND_FOM_NONE; | |
4410 | } | |
dd957c57 | 4411 | |
1da177e4 LT |
4412 | /* fill params struct with the proper values */ |
4413 | params->mode = bond_mode; | |
169a3e66 | 4414 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4415 | params->miimon = miimon; |
ad246c99 | 4416 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4417 | params->arp_interval = arp_interval; |
f5b2b966 | 4418 | params->arp_validate = arp_validate_value; |
8599b52e | 4419 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4420 | params->updelay = updelay; |
4421 | params->downdelay = downdelay; | |
4422 | params->use_carrier = use_carrier; | |
4423 | params->lacp_fast = lacp_fast; | |
4424 | params->primary[0] = 0; | |
a549952a | 4425 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4426 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4427 | params->tx_queues = tx_queues; |
ebd8e497 | 4428 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4429 | params->resend_igmp = resend_igmp; |
655f8919 | 4430 | params->min_links = min_links; |
1da177e4 LT |
4431 | |
4432 | if (primary) { | |
4433 | strncpy(params->primary, primary, IFNAMSIZ); | |
4434 | params->primary[IFNAMSIZ - 1] = 0; | |
4435 | } | |
4436 | ||
4437 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4438 | ||
4439 | return 0; | |
4440 | } | |
4441 | ||
0daa2303 | 4442 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4443 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4444 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4445 | |
e8a0464c DM |
4446 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4447 | struct netdev_queue *txq, | |
4448 | void *_unused) | |
c773e847 DM |
4449 | { |
4450 | lockdep_set_class(&txq->_xmit_lock, | |
4451 | &bonding_netdev_xmit_lock_key); | |
4452 | } | |
4453 | ||
4454 | static void bond_set_lockdep_class(struct net_device *dev) | |
4455 | { | |
cf508b12 DM |
4456 | lockdep_set_class(&dev->addr_list_lock, |
4457 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4458 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4459 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4460 | } |
4461 | ||
181470fc SH |
4462 | /* |
4463 | * Called from registration process | |
4464 | */ | |
4465 | static int bond_init(struct net_device *bond_dev) | |
4466 | { | |
4467 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4468 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4469 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4470 | |
4471 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4472 | ||
9fe0617d NH |
4473 | /* |
4474 | * Initialize locks that may be required during | |
4475 | * en/deslave operations. All of the bond_open work | |
4476 | * (of which this is part) should really be moved to | |
4477 | * a phase prior to dev_open | |
4478 | */ | |
4479 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4480 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4481 | ||
181470fc SH |
4482 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4483 | if (!bond->wq) | |
4484 | return -ENOMEM; | |
4485 | ||
4486 | bond_set_lockdep_class(bond_dev); | |
4487 | ||
ec87fd3b | 4488 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4489 | |
6151b3d4 | 4490 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4491 | |
f073c7ca TI |
4492 | bond_debug_register(bond); |
4493 | ||
409cc1f8 JP |
4494 | /* Ensure valid dev_addr */ |
4495 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4496 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4497 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4498 | |
181470fc SH |
4499 | return 0; |
4500 | } | |
4501 | ||
88ead977 EB |
4502 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4503 | { | |
4504 | if (tb[IFLA_ADDRESS]) { | |
4505 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4506 | return -EINVAL; | |
4507 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4508 | return -EADDRNOTAVAIL; | |
4509 | } | |
4510 | return 0; | |
4511 | } | |
4512 | ||
d40156aa | 4513 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4514 | { |
efacb309 | 4515 | return tx_queues; |
d5da4510 JP |
4516 | } |
4517 | ||
88ead977 | 4518 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4519 | .kind = "bond", |
4520 | .priv_size = sizeof(struct bonding), | |
4521 | .setup = bond_setup, | |
4522 | .validate = bond_validate, | |
4523 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4524 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4525 | as for TX queues */ | |
88ead977 EB |
4526 | }; |
4527 | ||
dfe60397 | 4528 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4529 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4530 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4531 | * set up our sysfs entries. | |
4532 | */ | |
ec87fd3b | 4533 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4534 | { |
4535 | struct net_device *bond_dev; | |
4536 | int res; | |
4537 | ||
4538 | rtnl_lock(); | |
027ea041 | 4539 | |
1c5cae81 JP |
4540 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4541 | name ? name : "bond%d", | |
4542 | bond_setup, tx_queues); | |
dfe60397 | 4543 | if (!bond_dev) { |
a4aee5c8 | 4544 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4545 | rtnl_unlock(); |
4546 | return -ENOMEM; | |
dfe60397 MW |
4547 | } |
4548 | ||
ec87fd3b | 4549 | dev_net_set(bond_dev, net); |
88ead977 EB |
4550 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4551 | ||
dfe60397 | 4552 | res = register_netdevice(bond_dev); |
0daa2303 | 4553 | |
e826eafa PO |
4554 | netif_carrier_off(bond_dev); |
4555 | ||
7e083840 | 4556 | rtnl_unlock(); |
9e2e61fb AW |
4557 | if (res < 0) |
4558 | bond_destructor(bond_dev); | |
30c15ba9 | 4559 | return res; |
dfe60397 MW |
4560 | } |
4561 | ||
2c8c1e72 | 4562 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4563 | { |
15449745 | 4564 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4565 | |
4566 | bn->net = net; | |
4567 | INIT_LIST_HEAD(&bn->dev_list); | |
4568 | ||
ec87fd3b | 4569 | bond_create_proc_dir(bn); |
4c22400a | 4570 | bond_create_sysfs(bn); |
87a7b84b | 4571 | |
15449745 | 4572 | return 0; |
ec87fd3b EB |
4573 | } |
4574 | ||
2c8c1e72 | 4575 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4576 | { |
15449745 | 4577 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4578 | struct bonding *bond, *tmp_bond; |
4579 | LIST_HEAD(list); | |
ec87fd3b | 4580 | |
4c22400a | 4581 | bond_destroy_sysfs(bn); |
ec87fd3b | 4582 | bond_destroy_proc_dir(bn); |
69b0216a | 4583 | |
4584 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4585 | rtnl_lock(); | |
4586 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4587 | unregister_netdevice_queue(bond->dev, &list); | |
4588 | unregister_netdevice_many(&list); | |
4589 | rtnl_unlock(); | |
ec87fd3b EB |
4590 | } |
4591 | ||
4592 | static struct pernet_operations bond_net_ops = { | |
4593 | .init = bond_net_init, | |
4594 | .exit = bond_net_exit, | |
15449745 EB |
4595 | .id = &bond_net_id, |
4596 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4597 | }; |
4598 | ||
1da177e4 LT |
4599 | static int __init bonding_init(void) |
4600 | { | |
1da177e4 LT |
4601 | int i; |
4602 | int res; | |
4603 | ||
bd33acc3 | 4604 | pr_info("%s", bond_version); |
1da177e4 | 4605 | |
dfe60397 | 4606 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4607 | if (res) |
dfe60397 | 4608 | goto out; |
1da177e4 | 4609 | |
15449745 | 4610 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4611 | if (res) |
4612 | goto out; | |
027ea041 | 4613 | |
88ead977 EB |
4614 | res = rtnl_link_register(&bond_link_ops); |
4615 | if (res) | |
6639104b | 4616 | goto err_link; |
88ead977 | 4617 | |
f073c7ca TI |
4618 | bond_create_debugfs(); |
4619 | ||
1da177e4 | 4620 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4621 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4622 | if (res) |
4623 | goto err; | |
1da177e4 LT |
4624 | } |
4625 | ||
1da177e4 | 4626 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4627 | out: |
1da177e4 | 4628 | return res; |
88ead977 EB |
4629 | err: |
4630 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 4631 | err_link: |
15449745 | 4632 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4633 | goto out; |
dfe60397 | 4634 | |
1da177e4 LT |
4635 | } |
4636 | ||
4637 | static void __exit bonding_exit(void) | |
4638 | { | |
4639 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4640 | ||
f073c7ca | 4641 | bond_destroy_debugfs(); |
ae68c398 | 4642 | |
4de79c73 | 4643 | rtnl_link_unregister(&bond_link_ops); |
ffcdedb6 | 4644 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4645 | |
4646 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4647 | /* |
4648 | * Make sure we don't have an imbalance on our netpoll blocking | |
4649 | */ | |
4650 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4651 | #endif |
1da177e4 LT |
4652 | } |
4653 | ||
4654 | module_init(bonding_init); | |
4655 | module_exit(bonding_exit); | |
4656 | MODULE_LICENSE("GPL"); | |
4657 | MODULE_VERSION(DRV_VERSION); | |
4658 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4659 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 4660 | MODULE_ALIAS_RTNL_LINK("bond"); |