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 | 1086 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1087 | netdev_features_t features) |
8531c5ff | 1088 | { |
b2a103e6 | 1089 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1090 | netdev_features_t mask; |
9b7b165a | 1091 | struct slave *slave; |
b63365a2 | 1092 | |
dec1e90e | 1093 | if (list_empty(&bond->slave_list)) { |
b2a103e6 MM |
1094 | /* Disable adding VLANs to empty bond. But why? --mq */ |
1095 | features |= NETIF_F_VLAN_CHALLENGED; | |
9b7b165a | 1096 | return features; |
b2a103e6 | 1097 | } |
b63365a2 | 1098 | |
b2a103e6 | 1099 | mask = features; |
b63365a2 | 1100 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1101 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1102 | |
dec1e90e | 1103 | bond_for_each_slave(bond, slave) { |
b63365a2 HX |
1104 | features = netdev_increment_features(features, |
1105 | slave->dev->features, | |
b2a103e6 MM |
1106 | mask); |
1107 | } | |
b0ce3508 | 1108 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1109 | |
b2a103e6 MM |
1110 | return features; |
1111 | } | |
1112 | ||
62f2a3a4 MM |
1113 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1114 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1115 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1116 | |
1117 | static void bond_compute_features(struct bonding *bond) | |
1118 | { | |
c4826861 | 1119 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
c8f44aff | 1120 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1121 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 | 1122 | unsigned int gso_max_size = GSO_MAX_SIZE; |
c4826861 | 1123 | struct net_device *bond_dev = bond->dev; |
0e376bd0 | 1124 | u16 gso_max_segs = GSO_MAX_SEGS; |
c4826861 | 1125 | struct slave *slave; |
b2a103e6 | 1126 | |
dec1e90e | 1127 | if (list_empty(&bond->slave_list)) |
b2a103e6 MM |
1128 | goto done; |
1129 | ||
dec1e90e | 1130 | bond_for_each_slave(bond, slave) { |
278339a4 | 1131 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1132 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1133 | ||
b6fe83e9 | 1134 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1135 | if (slave->dev->hard_header_len > max_hard_header_len) |
1136 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1137 | |
1138 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1139 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1140 | } |
8531c5ff | 1141 | |
b63365a2 | 1142 | done: |
b2a103e6 | 1143 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1144 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1145 | bond_dev->gso_max_segs = gso_max_segs; |
1146 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1147 | |
b6fe83e9 ED |
1148 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1149 | bond_dev->priv_flags = flags | dst_release_flag; | |
1150 | ||
b2a103e6 | 1151 | netdev_change_features(bond_dev); |
8531c5ff AK |
1152 | } |
1153 | ||
872254dd MS |
1154 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1155 | struct net_device *slave_dev) | |
1156 | { | |
00829823 | 1157 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1158 | |
1159 | bond_dev->type = slave_dev->type; | |
1160 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1161 | bond_dev->addr_len = slave_dev->addr_len; | |
1162 | ||
1163 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1164 | slave_dev->addr_len); | |
1165 | } | |
1166 | ||
5b2c4dd2 | 1167 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1168 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1169 | */ |
1170 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1171 | struct slave *slave, |
1172 | struct bonding *bond) | |
5b2c4dd2 | 1173 | { |
2d7011ca | 1174 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1175 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1176 | skb->pkt_type != PACKET_BROADCAST && |
1177 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1178 | return false; |
5b2c4dd2 JP |
1179 | return true; |
1180 | } | |
1181 | return false; | |
1182 | } | |
1183 | ||
8a4eb573 | 1184 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1185 | { |
8a4eb573 | 1186 | struct sk_buff *skb = *pskb; |
f1c1775a | 1187 | struct slave *slave; |
0bd80dad | 1188 | struct bonding *bond; |
de063b70 ED |
1189 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1190 | struct slave *); | |
13a8e0c8 | 1191 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1192 | |
8a4eb573 JP |
1193 | skb = skb_share_check(skb, GFP_ATOMIC); |
1194 | if (unlikely(!skb)) | |
1195 | return RX_HANDLER_CONSUMED; | |
1196 | ||
1197 | *pskb = skb; | |
5b2c4dd2 | 1198 | |
35d48903 JP |
1199 | slave = bond_slave_get_rcu(skb->dev); |
1200 | bond = slave->bond; | |
0bd80dad JP |
1201 | |
1202 | if (bond->params.arp_interval) | |
f1c1775a | 1203 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1204 | |
4d97480b MH |
1205 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1206 | if (recv_probe) { | |
de063b70 ED |
1207 | ret = recv_probe(skb, bond, slave); |
1208 | if (ret == RX_HANDLER_CONSUMED) { | |
1209 | consume_skb(skb); | |
1210 | return ret; | |
3aba891d JP |
1211 | } |
1212 | } | |
1213 | ||
0bd80dad | 1214 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1215 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1216 | } |
1217 | ||
35d48903 | 1218 | skb->dev = bond->dev; |
5b2c4dd2 | 1219 | |
0bd80dad | 1220 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1221 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1222 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1223 | |
541ac7c9 CG |
1224 | if (unlikely(skb_cow_head(skb, |
1225 | skb->data - skb_mac_header(skb)))) { | |
1226 | kfree_skb(skb); | |
8a4eb573 | 1227 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1228 | } |
35d48903 | 1229 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1230 | } |
1231 | ||
13a8e0c8 | 1232 | return ret; |
5b2c4dd2 JP |
1233 | } |
1234 | ||
471cb5a3 | 1235 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1f718f0f VF |
1236 | struct net_device *slave_dev, |
1237 | struct slave *slave) | |
471cb5a3 JP |
1238 | { |
1239 | int err; | |
1240 | ||
1f718f0f | 1241 | err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave); |
471cb5a3 JP |
1242 | if (err) |
1243 | return err; | |
1244 | slave_dev->flags |= IFF_SLAVE; | |
1245 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1246 | return 0; | |
1247 | } | |
1248 | ||
1249 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1250 | struct net_device *slave_dev) | |
1251 | { | |
1252 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1253 | slave_dev->flags &= ~IFF_SLAVE; | |
1254 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1255 | } | |
1256 | ||
1da177e4 | 1257 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1258 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1259 | { |
454d7c9b | 1260 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1261 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1262 | struct slave *new_slave = NULL; |
1da177e4 LT |
1263 | struct sockaddr addr; |
1264 | int link_reporting; | |
8599b52e | 1265 | int res = 0, i; |
1da177e4 | 1266 | |
c772dde3 BH |
1267 | if (!bond->params.use_carrier && |
1268 | slave_dev->ethtool_ops->get_link == NULL && | |
1269 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1270 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1271 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1272 | } |
1273 | ||
1da177e4 LT |
1274 | /* already enslaved */ |
1275 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1276 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1277 | return -EBUSY; |
1278 | } | |
1279 | ||
1280 | /* vlan challenged mutual exclusion */ | |
1281 | /* no need to lock since we're protected by rtnl_lock */ | |
1282 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1283 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1284 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1285 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1286 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1287 | return -EPERM; |
1288 | } else { | |
a4aee5c8 JP |
1289 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1290 | bond_dev->name, slave_dev->name, | |
1291 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1292 | } |
1293 | } else { | |
5a03cdb7 | 1294 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1295 | } |
1296 | ||
217df670 JV |
1297 | /* |
1298 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1299 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1300 | * the current ifenslave will set the interface down prior to |
1301 | * enslaving it; the old ifenslave will not. | |
1302 | */ | |
1303 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1304 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1305 | slave_dev->name); |
1306 | res = -EPERM; | |
1307 | goto err_undo_flags; | |
1308 | } | |
1da177e4 | 1309 | |
872254dd MS |
1310 | /* set bonding device ether type by slave - bonding netdevices are |
1311 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1312 | * there is a need to override some of the type dependent attribs/funcs. | |
1313 | * | |
1314 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1315 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1316 | */ | |
dec1e90e | 1317 | if (list_empty(&bond->slave_list)) { |
e36b9d16 | 1318 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1319 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1320 | bond_dev->name, |
1321 | bond_dev->type, slave_dev->type); | |
75c78500 | 1322 | |
b7bc2a5b AW |
1323 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1324 | bond_dev); | |
3ca5b404 JP |
1325 | res = notifier_to_errno(res); |
1326 | if (res) { | |
1327 | pr_err("%s: refused to change device type\n", | |
1328 | bond_dev->name); | |
1329 | res = -EBUSY; | |
1330 | goto err_undo_flags; | |
1331 | } | |
75c78500 | 1332 | |
32a806c1 | 1333 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1334 | dev_uc_flush(bond_dev); |
22bedad3 | 1335 | dev_mc_flush(bond_dev); |
32a806c1 | 1336 | |
e36b9d16 MS |
1337 | if (slave_dev->type != ARPHRD_ETHER) |
1338 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1339 | else { |
e36b9d16 | 1340 | ether_setup(bond_dev); |
550fd08c NH |
1341 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1342 | } | |
75c78500 | 1343 | |
b7bc2a5b AW |
1344 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1345 | bond_dev); | |
e36b9d16 | 1346 | } |
872254dd | 1347 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1348 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1349 | slave_dev->name, | |
1350 | slave_dev->type, bond_dev->type); | |
1351 | res = -EINVAL; | |
1352 | goto err_undo_flags; | |
872254dd MS |
1353 | } |
1354 | ||
eb7cc59a | 1355 | if (slave_ops->ndo_set_mac_address == NULL) { |
dec1e90e | 1356 | if (list_empty(&bond->slave_list)) { |
a4aee5c8 JP |
1357 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1358 | bond_dev->name); | |
3915c1e8 JV |
1359 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1360 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1361 | 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", |
1362 | bond_dev->name); | |
2ab82852 MS |
1363 | res = -EOPNOTSUPP; |
1364 | goto err_undo_flags; | |
1365 | } | |
1da177e4 LT |
1366 | } |
1367 | ||
8d8fc29d AW |
1368 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1369 | ||
c20811a7 JP |
1370 | /* If this is the first slave, then we need to set the master's hardware |
1371 | * address to be the same as the slave's. */ | |
dec1e90e | 1372 | if (list_empty(&bond->slave_list) && |
1373 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) | |
409cc1f8 | 1374 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1375 | |
243cb4e5 | 1376 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1377 | if (!new_slave) { |
1378 | res = -ENOMEM; | |
1379 | goto err_undo_flags; | |
1380 | } | |
dec1e90e | 1381 | INIT_LIST_HEAD(&new_slave->list); |
bb1d9123 AG |
1382 | /* |
1383 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1384 | * is set via sysfs or module option if desired. | |
1385 | */ | |
1386 | new_slave->queue_id = 0; | |
1387 | ||
b15ba0fb JP |
1388 | /* Save slave's original mtu and then set it to match the bond */ |
1389 | new_slave->original_mtu = slave_dev->mtu; | |
1390 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1391 | if (res) { | |
1392 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1393 | goto err_free; | |
1394 | } | |
1395 | ||
217df670 JV |
1396 | /* |
1397 | * Save slave's original ("permanent") mac address for modes | |
1398 | * that need it, and for restoring it upon release, and then | |
1399 | * set it to the master's address | |
1400 | */ | |
1401 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1402 | |
dd957c57 | 1403 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1404 | /* |
1405 | * Set slave to master's mac address. The application already | |
1406 | * set the master's mac address to that of the first slave | |
1407 | */ | |
1408 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1409 | addr.sa_family = slave_dev->type; | |
1410 | res = dev_set_mac_address(slave_dev, &addr); | |
1411 | if (res) { | |
5a03cdb7 | 1412 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1413 | goto err_restore_mtu; |
2ab82852 | 1414 | } |
217df670 | 1415 | } |
1da177e4 | 1416 | |
217df670 JV |
1417 | /* open the slave since the application closed it */ |
1418 | res = dev_open(slave_dev); | |
1419 | if (res) { | |
3d632c3f | 1420 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1421 | goto err_restore_mac; |
1da177e4 LT |
1422 | } |
1423 | ||
35d48903 | 1424 | new_slave->bond = bond; |
1da177e4 | 1425 | new_slave->dev = slave_dev; |
0b680e75 | 1426 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1427 | |
58402054 | 1428 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1429 | /* bond_alb_init_slave() must be called before all other stages since |
1430 | * it might fail and we do not want to have to undo everything | |
1431 | */ | |
1432 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1433 | if (res) |
569f0c4d | 1434 | goto err_close; |
1da177e4 LT |
1435 | } |
1436 | ||
303d1cbf JV |
1437 | /* If the mode USES_PRIMARY, then the following is handled by |
1438 | * bond_change_active_slave(). | |
1da177e4 LT |
1439 | */ |
1440 | if (!USES_PRIMARY(bond->params.mode)) { | |
1441 | /* set promiscuity level to new slave */ | |
1442 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1443 | res = dev_set_promiscuity(slave_dev, 1); |
1444 | if (res) | |
1445 | goto err_close; | |
1da177e4 LT |
1446 | } |
1447 | ||
1448 | /* set allmulti level to new slave */ | |
1449 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1450 | res = dev_set_allmulti(slave_dev, 1); |
1451 | if (res) | |
1452 | goto err_close; | |
1da177e4 LT |
1453 | } |
1454 | ||
b9e40857 | 1455 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1456 | |
1457 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1458 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1459 | ||
b9e40857 | 1460 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1461 | } |
1462 | ||
1463 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1464 | /* add lacpdu mc addr to mc list */ | |
1465 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1466 | ||
22bedad3 | 1467 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1468 | } |
1469 | ||
b8e2fde4 WY |
1470 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1471 | if (res) { | |
1ff412ad | 1472 | pr_err("%s: Error: Couldn't add bond vlan ids to %s\n", |
1473 | bond_dev->name, slave_dev->name); | |
1474 | goto err_close; | |
1475 | } | |
1da177e4 LT |
1476 | |
1477 | write_lock_bh(&bond->lock); | |
1478 | ||
1479 | bond_attach_slave(bond, new_slave); | |
1480 | ||
1481 | new_slave->delay = 0; | |
1482 | new_slave->link_failure_count = 0; | |
1483 | ||
3915c1e8 JV |
1484 | write_unlock_bh(&bond->lock); |
1485 | ||
b2a103e6 MM |
1486 | bond_compute_features(bond); |
1487 | ||
876254ae VF |
1488 | bond_update_speed_duplex(new_slave); |
1489 | ||
3915c1e8 JV |
1490 | read_lock(&bond->lock); |
1491 | ||
f31c7937 MK |
1492 | new_slave->last_arp_rx = jiffies - |
1493 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
8599b52e VF |
1494 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
1495 | new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx; | |
f5b2b966 | 1496 | |
1da177e4 LT |
1497 | if (bond->params.miimon && !bond->params.use_carrier) { |
1498 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1499 | ||
1500 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1501 | /* | |
1502 | * miimon is set but a bonded network driver | |
1503 | * does not support ETHTOOL/MII and | |
1504 | * arp_interval is not set. Note: if | |
1505 | * use_carrier is enabled, we will never go | |
1506 | * here (because netif_carrier is always | |
1507 | * supported); thus, we don't need to change | |
1508 | * the messages for netif_carrier. | |
1509 | */ | |
a4aee5c8 | 1510 | 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 | 1511 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1512 | } else if (link_reporting == -1) { |
1513 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1514 | 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", |
1515 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1516 | } |
1517 | } | |
1518 | ||
1519 | /* check for initial state */ | |
f31c7937 MK |
1520 | if (bond->params.miimon) { |
1521 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1522 | if (bond->params.updelay) { | |
1523 | new_slave->link = BOND_LINK_BACK; | |
1524 | new_slave->delay = bond->params.updelay; | |
1525 | } else { | |
1526 | new_slave->link = BOND_LINK_UP; | |
1527 | } | |
1da177e4 | 1528 | } else { |
f31c7937 | 1529 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1530 | } |
f31c7937 MK |
1531 | } else if (bond->params.arp_interval) { |
1532 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1533 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1534 | } else { |
f31c7937 | 1535 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1536 | } |
1537 | ||
f31c7937 MK |
1538 | if (new_slave->link != BOND_LINK_DOWN) |
1539 | new_slave->jiffies = jiffies; | |
1540 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1541 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1542 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1543 | ||
1da177e4 LT |
1544 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1545 | /* if there is a primary slave, remember it */ | |
a549952a | 1546 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1547 | bond->primary_slave = new_slave; |
a549952a JP |
1548 | bond->force_primary = true; |
1549 | } | |
1da177e4 LT |
1550 | } |
1551 | ||
3915c1e8 JV |
1552 | write_lock_bh(&bond->curr_slave_lock); |
1553 | ||
1da177e4 LT |
1554 | switch (bond->params.mode) { |
1555 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1556 | bond_set_slave_inactive_flags(new_slave); |
1557 | bond_select_active_slave(bond); | |
1da177e4 LT |
1558 | break; |
1559 | case BOND_MODE_8023AD: | |
1560 | /* in 802.3ad mode, the internal mechanism | |
1561 | * will activate the slaves in the selected | |
1562 | * aggregator | |
1563 | */ | |
1564 | bond_set_slave_inactive_flags(new_slave); | |
1565 | /* if this is the first slave */ | |
dec1e90e | 1566 | if (bond_first_slave(bond) == new_slave) { |
1da177e4 LT |
1567 | SLAVE_AD_INFO(new_slave).id = 1; |
1568 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1569 | * can be called only after the mac address of the bond is set | |
1570 | */ | |
56d00c67 | 1571 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1572 | } else { |
dec1e90e | 1573 | struct slave *prev_slave; |
1574 | ||
1575 | prev_slave = bond_prev_slave(bond, new_slave); | |
1da177e4 | 1576 | SLAVE_AD_INFO(new_slave).id = |
dec1e90e | 1577 | SLAVE_AD_INFO(prev_slave).id + 1; |
1da177e4 LT |
1578 | } |
1579 | ||
1580 | bond_3ad_bind_slave(new_slave); | |
1581 | break; | |
1582 | case BOND_MODE_TLB: | |
1583 | case BOND_MODE_ALB: | |
e30bc066 | 1584 | bond_set_active_slave(new_slave); |
059fe7a5 | 1585 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1586 | bond_select_active_slave(bond); |
1da177e4 LT |
1587 | break; |
1588 | default: | |
5a03cdb7 | 1589 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1590 | |
1591 | /* always active in trunk mode */ | |
e30bc066 | 1592 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1593 | |
1594 | /* In trunking mode there is little meaning to curr_active_slave | |
1595 | * anyway (it holds no special properties of the bond device), | |
1596 | * so we can change it without calling change_active_interface() | |
1597 | */ | |
4e591b93 | 1598 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
278b2083 | 1599 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1600 | |
1da177e4 LT |
1601 | break; |
1602 | } /* switch(bond_mode) */ | |
1603 | ||
3915c1e8 JV |
1604 | write_unlock_bh(&bond->curr_slave_lock); |
1605 | ||
ff59c456 JV |
1606 | bond_set_carrier(bond); |
1607 | ||
f6dc31a8 | 1608 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1609 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1610 | if (slave_dev->npinfo) { |
1611 | if (slave_enable_netpoll(new_slave)) { | |
1612 | read_unlock(&bond->lock); | |
1613 | pr_info("Error, %s: master_dev is using netpoll, " | |
1614 | "but new slave device does not support netpoll.\n", | |
1615 | bond_dev->name); | |
1616 | res = -EBUSY; | |
f7d9821a | 1617 | goto err_detach; |
8a8efa22 | 1618 | } |
f6dc31a8 WC |
1619 | } |
1620 | #endif | |
8a8efa22 | 1621 | |
3915c1e8 | 1622 | read_unlock(&bond->lock); |
1da177e4 | 1623 | |
b76cdba9 MW |
1624 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1625 | if (res) | |
f7d9821a | 1626 | goto err_detach; |
b76cdba9 | 1627 | |
35d48903 JP |
1628 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1629 | new_slave); | |
1630 | if (res) { | |
1631 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1632 | goto err_dest_symlinks; | |
1633 | } | |
1634 | ||
1f718f0f VF |
1635 | res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave); |
1636 | if (res) { | |
1637 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); | |
1638 | goto err_unregister; | |
1639 | } | |
1640 | ||
1641 | ||
a4aee5c8 JP |
1642 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1643 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1644 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1645 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1646 | |
1647 | /* enslave is successful */ | |
1648 | return 0; | |
1649 | ||
1650 | /* Undo stages on error */ | |
1f718f0f VF |
1651 | err_unregister: |
1652 | netdev_rx_handler_unregister(slave_dev); | |
1653 | ||
35d48903 JP |
1654 | err_dest_symlinks: |
1655 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1656 | ||
f7d9821a | 1657 | err_detach: |
303d1cbf JV |
1658 | if (!USES_PRIMARY(bond->params.mode)) |
1659 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1660 | ||
1ff412ad | 1661 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
f7d9821a | 1662 | write_lock_bh(&bond->lock); |
1663 | bond_detach_slave(bond, new_slave); | |
3c5913b5 | 1664 | if (bond->primary_slave == new_slave) |
1665 | bond->primary_slave = NULL; | |
3c5913b5 | 1666 | if (bond->curr_active_slave == new_slave) { |
c6cdcf6d | 1667 | bond_change_active_slave(bond, NULL); |
1668 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1669 | read_lock(&bond->lock); |
1670 | write_lock_bh(&bond->curr_slave_lock); | |
3c5913b5 | 1671 | bond_select_active_slave(bond); |
1672 | write_unlock_bh(&bond->curr_slave_lock); | |
1673 | read_unlock(&bond->lock); | |
c6cdcf6d | 1674 | } else { |
1675 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1676 | } |
fc7a72ac | 1677 | slave_disable_netpoll(new_slave); |
f7d9821a | 1678 | |
1da177e4 | 1679 | err_close: |
b6a5a7b9 | 1680 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1681 | dev_close(slave_dev); |
1682 | ||
1683 | err_restore_mac: | |
dd957c57 | 1684 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1685 | /* XXX TODO - fom follow mode needs to change master's |
1686 | * MAC if this slave's MAC is in use by the bond, or at | |
1687 | * least print a warning. | |
1688 | */ | |
2ab82852 MS |
1689 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1690 | addr.sa_family = slave_dev->type; | |
1691 | dev_set_mac_address(slave_dev, &addr); | |
1692 | } | |
1da177e4 | 1693 | |
b15ba0fb JP |
1694 | err_restore_mtu: |
1695 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1696 | ||
1da177e4 LT |
1697 | err_free: |
1698 | kfree(new_slave); | |
1699 | ||
1700 | err_undo_flags: | |
b2a103e6 | 1701 | bond_compute_features(bond); |
b8fad459 | 1702 | /* Enslave of first slave has failed and we need to fix master's mac */ |
dec1e90e | 1703 | if (list_empty(&bond->slave_list) && |
b8fad459 NA |
1704 | ether_addr_equal(bond_dev->dev_addr, slave_dev->dev_addr)) |
1705 | eth_hw_addr_random(bond_dev); | |
3d632c3f | 1706 | |
1da177e4 LT |
1707 | return res; |
1708 | } | |
1709 | ||
1710 | /* | |
1711 | * Try to release the slave device <slave> from the bond device <master> | |
1712 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1713 | * is write-locked. If "all" is true it means that the function is being called |
1714 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1715 | * |
1716 | * The rules for slave state should be: | |
1717 | * for Active/Backup: | |
1718 | * Active stays on all backups go down | |
1719 | * for Bonded connections: | |
1720 | * The first up interface should be left on and all others downed. | |
1721 | */ | |
0896341a | 1722 | static int __bond_release_one(struct net_device *bond_dev, |
1723 | struct net_device *slave_dev, | |
1724 | bool all) | |
1da177e4 | 1725 | { |
454d7c9b | 1726 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1727 | struct slave *slave, *oldcurrent; |
1728 | struct sockaddr addr; | |
c8f44aff | 1729 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1730 | |
1731 | /* slave is not a slave or master is not master of this slave */ | |
1732 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1733 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1734 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1735 | bond_dev->name, slave_dev->name); |
1736 | return -EINVAL; | |
1737 | } | |
1738 | ||
e843fa50 | 1739 | block_netpoll_tx(); |
1da177e4 LT |
1740 | write_lock_bh(&bond->lock); |
1741 | ||
1742 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1743 | if (!slave) { | |
1744 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1745 | pr_info("%s: %s not enslaved\n", |
1746 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1747 | write_unlock_bh(&bond->lock); |
e843fa50 | 1748 | unblock_netpoll_tx(); |
1da177e4 LT |
1749 | return -EINVAL; |
1750 | } | |
1751 | ||
fcd99434 | 1752 | write_unlock_bh(&bond->lock); |
1f718f0f VF |
1753 | |
1754 | bond_upper_dev_unlink(bond_dev, slave_dev); | |
35d48903 JP |
1755 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1756 | * for this slave anymore. | |
1757 | */ | |
1758 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1759 | write_lock_bh(&bond->lock); |
1760 | ||
1da177e4 LT |
1761 | /* Inform AD package of unbinding of slave. */ |
1762 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1763 | /* must be called before the slave is | |
1764 | * detached from the list | |
1765 | */ | |
1766 | bond_3ad_unbind_slave(slave); | |
1767 | } | |
1768 | ||
a4aee5c8 JP |
1769 | pr_info("%s: releasing %s interface %s\n", |
1770 | bond_dev->name, | |
e30bc066 | 1771 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1772 | slave_dev->name); |
1da177e4 LT |
1773 | |
1774 | oldcurrent = bond->curr_active_slave; | |
1775 | ||
1776 | bond->current_arp_slave = NULL; | |
1777 | ||
1778 | /* release the slave from its bond */ | |
1779 | bond_detach_slave(bond, slave); | |
1780 | ||
dec1e90e | 1781 | if (!all && !bond->params.fail_over_mac) { |
1782 | if (ether_addr_equal(bond_dev->dev_addr, slave->perm_hwaddr) && | |
1783 | !list_empty(&bond->slave_list)) | |
1784 | 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", | |
1785 | bond_dev->name, slave_dev->name, | |
1786 | slave->perm_hwaddr, | |
1787 | bond_dev->name, slave_dev->name); | |
1788 | } | |
1789 | ||
3d632c3f | 1790 | if (bond->primary_slave == slave) |
1da177e4 | 1791 | bond->primary_slave = NULL; |
1da177e4 | 1792 | |
3d632c3f | 1793 | if (oldcurrent == slave) |
1da177e4 | 1794 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1795 | |
58402054 | 1796 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1797 | /* Must be called only after the slave has been |
1798 | * detached from the list and the curr_active_slave | |
1799 | * has been cleared (if our_slave == old_current), | |
1800 | * but before a new active slave is selected. | |
1801 | */ | |
2543331d | 1802 | write_unlock_bh(&bond->lock); |
1da177e4 | 1803 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1804 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1805 | } |
1806 | ||
0896341a | 1807 | if (all) { |
278b2083 | 1808 | rcu_assign_pointer(bond->curr_active_slave, NULL); |
0896341a | 1809 | } else if (oldcurrent == slave) { |
059fe7a5 JV |
1810 | /* |
1811 | * Note that we hold RTNL over this sequence, so there | |
1812 | * is no concern that another slave add/remove event | |
1813 | * will interfere. | |
1814 | */ | |
1815 | write_unlock_bh(&bond->lock); | |
1816 | read_lock(&bond->lock); | |
1817 | write_lock_bh(&bond->curr_slave_lock); | |
1818 | ||
1da177e4 LT |
1819 | bond_select_active_slave(bond); |
1820 | ||
059fe7a5 JV |
1821 | write_unlock_bh(&bond->curr_slave_lock); |
1822 | read_unlock(&bond->lock); | |
1823 | write_lock_bh(&bond->lock); | |
1824 | } | |
1825 | ||
dec1e90e | 1826 | if (list_empty(&bond->slave_list)) { |
ff59c456 | 1827 | bond_set_carrier(bond); |
409cc1f8 | 1828 | eth_hw_addr_random(bond_dev); |
1da177e4 | 1829 | |
a59d3d21 | 1830 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1831 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
1832 | bond_dev->name, bond_dev->name); | |
1833 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
1834 | bond_dev->name); | |
1da177e4 | 1835 | } |
1da177e4 LT |
1836 | } |
1837 | ||
1838 | write_unlock_bh(&bond->lock); | |
e843fa50 | 1839 | unblock_netpoll_tx(); |
278b2083 | 1840 | synchronize_rcu(); |
1da177e4 | 1841 | |
dec1e90e | 1842 | if (list_empty(&bond->slave_list)) { |
2af73d4b | 1843 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1844 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1845 | } | |
2af73d4b | 1846 | |
b2a103e6 MM |
1847 | bond_compute_features(bond); |
1848 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1849 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
1850 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
1851 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1852 | ||
b76cdba9 MW |
1853 | /* must do this from outside any spinlocks */ |
1854 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1855 | ||
1ff412ad | 1856 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1857 | |
303d1cbf JV |
1858 | /* If the mode USES_PRIMARY, then this cases was handled above by |
1859 | * bond_change_active_slave(..., NULL) | |
1da177e4 LT |
1860 | */ |
1861 | if (!USES_PRIMARY(bond->params.mode)) { | |
1862 | /* unset promiscuity level from slave */ | |
3d632c3f | 1863 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 1864 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1865 | |
1866 | /* unset allmulti level from slave */ | |
3d632c3f | 1867 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 1868 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1869 | |
303d1cbf | 1870 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1871 | } |
1872 | ||
8a8efa22 | 1873 | slave_disable_netpoll(slave); |
f6dc31a8 | 1874 | |
1da177e4 LT |
1875 | /* close slave before restoring its mac address */ |
1876 | dev_close(slave_dev); | |
1877 | ||
3915c1e8 | 1878 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1879 | /* restore original ("permanent") mac address */ |
1880 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1881 | addr.sa_family = slave_dev->type; | |
1882 | dev_set_mac_address(slave_dev, &addr); | |
1883 | } | |
1da177e4 | 1884 | |
b15ba0fb JP |
1885 | dev_set_mtu(slave_dev, slave->original_mtu); |
1886 | ||
2d7011ca | 1887 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1888 | |
1889 | kfree(slave); | |
1890 | ||
1891 | return 0; /* deletion OK */ | |
1892 | } | |
1893 | ||
0896341a | 1894 | /* A wrapper used because of ndo_del_link */ |
1895 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1896 | { | |
1897 | return __bond_release_one(bond_dev, slave_dev, false); | |
1898 | } | |
1899 | ||
d90a162a | 1900 | /* |
dadaa10b | 1901 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
1902 | * Must be under rtnl_lock when this function is called. |
1903 | */ | |
26d8ee75 | 1904 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1905 | struct net_device *slave_dev) | |
d90a162a | 1906 | { |
454d7c9b | 1907 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1908 | int ret; |
1909 | ||
1910 | ret = bond_release(bond_dev, slave_dev); | |
dec1e90e | 1911 | if (ret == 0 && list_empty(&bond->slave_list)) { |
8a8efa22 | 1912 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
1913 | pr_info("%s: destroying bond %s.\n", |
1914 | bond_dev->name, bond_dev->name); | |
9e71626c | 1915 | unregister_netdevice(bond_dev); |
d90a162a MS |
1916 | } |
1917 | return ret; | |
1918 | } | |
1919 | ||
1da177e4 LT |
1920 | /* |
1921 | * This function changes the active slave to slave <slave_dev>. | |
1922 | * It returns -EINVAL in the following cases. | |
1923 | * - <slave_dev> is not found in the list. | |
1924 | * - There is not active slave now. | |
1925 | * - <slave_dev> is already active. | |
1926 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
1927 | * - <slave_dev> is not running. | |
3d632c3f SH |
1928 | * In these cases, this function does nothing. |
1929 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
1930 | */ |
1931 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
1932 | { | |
454d7c9b | 1933 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1934 | struct slave *old_active = NULL; |
1935 | struct slave *new_active = NULL; | |
1936 | int res = 0; | |
1937 | ||
3d632c3f | 1938 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1939 | return -EINVAL; |
1da177e4 | 1940 | |
471cb5a3 JP |
1941 | /* Verify that bond_dev is indeed the master of slave_dev */ |
1942 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1943 | !netdev_has_upper_dev(slave_dev, bond_dev)) | |
1da177e4 | 1944 | return -EINVAL; |
1da177e4 | 1945 | |
059fe7a5 | 1946 | read_lock(&bond->lock); |
1da177e4 LT |
1947 | |
1948 | old_active = bond->curr_active_slave; | |
1949 | new_active = bond_get_slave_by_dev(bond, slave_dev); | |
1da177e4 LT |
1950 | /* |
1951 | * Changing to the current active: do nothing; return success. | |
1952 | */ | |
71bc3b2d | 1953 | if (new_active && new_active == old_active) { |
059fe7a5 | 1954 | read_unlock(&bond->lock); |
1da177e4 LT |
1955 | return 0; |
1956 | } | |
1957 | ||
71bc3b2d | 1958 | if (new_active && |
1959 | old_active && | |
1960 | new_active->link == BOND_LINK_UP && | |
1da177e4 | 1961 | IS_UP(new_active->dev)) { |
e843fa50 | 1962 | block_netpoll_tx(); |
059fe7a5 | 1963 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 1964 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 1965 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 1966 | unblock_netpoll_tx(); |
3d632c3f | 1967 | } else |
1da177e4 | 1968 | res = -EINVAL; |
1da177e4 | 1969 | |
059fe7a5 | 1970 | read_unlock(&bond->lock); |
1da177e4 LT |
1971 | |
1972 | return res; | |
1973 | } | |
1974 | ||
1da177e4 LT |
1975 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1976 | { | |
454d7c9b | 1977 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1978 | |
1979 | info->bond_mode = bond->params.mode; | |
1980 | info->miimon = bond->params.miimon; | |
1981 | ||
6603a6f2 | 1982 | read_lock(&bond->lock); |
1da177e4 | 1983 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 1984 | read_unlock(&bond->lock); |
1da177e4 LT |
1985 | |
1986 | return 0; | |
1987 | } | |
1988 | ||
1989 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1990 | { | |
454d7c9b | 1991 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 1992 | int i = 0, res = -ENODEV; |
1da177e4 | 1993 | struct slave *slave; |
1da177e4 | 1994 | |
6603a6f2 | 1995 | read_lock(&bond->lock); |
dec1e90e | 1996 | bond_for_each_slave(bond, slave) { |
1997 | if (i++ == (int)info->slave_id) { | |
689c96cc ED |
1998 | res = 0; |
1999 | strcpy(info->slave_name, slave->dev->name); | |
2000 | info->link = slave->link; | |
e30bc066 | 2001 | info->state = bond_slave_state(slave); |
689c96cc | 2002 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2003 | break; |
2004 | } | |
2005 | } | |
6603a6f2 | 2006 | read_unlock(&bond->lock); |
1da177e4 | 2007 | |
689c96cc | 2008 | return res; |
1da177e4 LT |
2009 | } |
2010 | ||
2011 | /*-------------------------------- Monitoring -------------------------------*/ | |
2012 | ||
1da177e4 | 2013 | |
f0c76d61 JV |
2014 | static int bond_miimon_inspect(struct bonding *bond) |
2015 | { | |
dec1e90e | 2016 | int link_state, commit = 0; |
f0c76d61 | 2017 | struct slave *slave; |
41f89100 JP |
2018 | bool ignore_updelay; |
2019 | ||
2020 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 | 2021 | |
dec1e90e | 2022 | bond_for_each_slave(bond, slave) { |
f0c76d61 | 2023 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2024 | |
f0c76d61 | 2025 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2026 | |
2027 | switch (slave->link) { | |
f0c76d61 JV |
2028 | case BOND_LINK_UP: |
2029 | if (link_state) | |
2030 | continue; | |
1da177e4 | 2031 | |
f0c76d61 JV |
2032 | slave->link = BOND_LINK_FAIL; |
2033 | slave->delay = bond->params.downdelay; | |
2034 | if (slave->delay) { | |
a4aee5c8 JP |
2035 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2036 | bond->dev->name, | |
2037 | (bond->params.mode == | |
2038 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2039 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2040 | "active " : "backup ") : "", |
2041 | slave->dev->name, | |
2042 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2043 | } |
f0c76d61 JV |
2044 | /*FALLTHRU*/ |
2045 | case BOND_LINK_FAIL: | |
2046 | if (link_state) { | |
2047 | /* | |
2048 | * recovered before downdelay expired | |
2049 | */ | |
2050 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2051 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2052 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2053 | bond->dev->name, | |
2054 | (bond->params.downdelay - slave->delay) * | |
2055 | bond->params.miimon, | |
2056 | slave->dev->name); | |
f0c76d61 | 2057 | continue; |
1da177e4 | 2058 | } |
f0c76d61 JV |
2059 | |
2060 | if (slave->delay <= 0) { | |
2061 | slave->new_link = BOND_LINK_DOWN; | |
2062 | commit++; | |
2063 | continue; | |
1da177e4 | 2064 | } |
1da177e4 | 2065 | |
f0c76d61 JV |
2066 | slave->delay--; |
2067 | break; | |
2068 | ||
2069 | case BOND_LINK_DOWN: | |
2070 | if (!link_state) | |
2071 | continue; | |
2072 | ||
2073 | slave->link = BOND_LINK_BACK; | |
2074 | slave->delay = bond->params.updelay; | |
2075 | ||
2076 | if (slave->delay) { | |
a4aee5c8 JP |
2077 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2078 | bond->dev->name, slave->dev->name, | |
2079 | ignore_updelay ? 0 : | |
2080 | bond->params.updelay * | |
2081 | bond->params.miimon); | |
f0c76d61 JV |
2082 | } |
2083 | /*FALLTHRU*/ | |
2084 | case BOND_LINK_BACK: | |
2085 | if (!link_state) { | |
2086 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2087 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2088 | bond->dev->name, | |
2089 | (bond->params.updelay - slave->delay) * | |
2090 | bond->params.miimon, | |
2091 | slave->dev->name); | |
f0c76d61 JV |
2092 | |
2093 | continue; | |
2094 | } | |
2095 | ||
41f89100 JP |
2096 | if (ignore_updelay) |
2097 | slave->delay = 0; | |
2098 | ||
f0c76d61 JV |
2099 | if (slave->delay <= 0) { |
2100 | slave->new_link = BOND_LINK_UP; | |
2101 | commit++; | |
41f89100 | 2102 | ignore_updelay = false; |
f0c76d61 | 2103 | continue; |
1da177e4 | 2104 | } |
f0c76d61 JV |
2105 | |
2106 | slave->delay--; | |
1da177e4 | 2107 | break; |
f0c76d61 JV |
2108 | } |
2109 | } | |
1da177e4 | 2110 | |
f0c76d61 JV |
2111 | return commit; |
2112 | } | |
1da177e4 | 2113 | |
f0c76d61 JV |
2114 | static void bond_miimon_commit(struct bonding *bond) |
2115 | { | |
2116 | struct slave *slave; | |
f0c76d61 | 2117 | |
dec1e90e | 2118 | bond_for_each_slave(bond, slave) { |
f0c76d61 JV |
2119 | switch (slave->new_link) { |
2120 | case BOND_LINK_NOCHANGE: | |
2121 | continue; | |
1da177e4 | 2122 | |
f0c76d61 JV |
2123 | case BOND_LINK_UP: |
2124 | slave->link = BOND_LINK_UP; | |
2125 | slave->jiffies = jiffies; | |
2126 | ||
2127 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2128 | /* prevent it from being the active one */ | |
e30bc066 | 2129 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2130 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2131 | /* make it immediately active */ | |
e30bc066 | 2132 | bond_set_active_slave(slave); |
f0c76d61 JV |
2133 | } else if (slave != bond->primary_slave) { |
2134 | /* prevent it from being the active one */ | |
e30bc066 | 2135 | bond_set_backup_slave(slave); |
1da177e4 | 2136 | } |
1da177e4 | 2137 | |
5d30530e | 2138 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 | 2139 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
2140 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
2141 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2142 | |
f0c76d61 JV |
2143 | /* notify ad that the link status has changed */ |
2144 | if (bond->params.mode == BOND_MODE_8023AD) | |
2145 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2146 | |
58402054 | 2147 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2148 | bond_alb_handle_link_change(bond, slave, |
2149 | BOND_LINK_UP); | |
1da177e4 | 2150 | |
f0c76d61 JV |
2151 | if (!bond->curr_active_slave || |
2152 | (slave == bond->primary_slave)) | |
2153 | goto do_failover; | |
1da177e4 | 2154 | |
f0c76d61 | 2155 | continue; |
059fe7a5 | 2156 | |
f0c76d61 | 2157 | case BOND_LINK_DOWN: |
fba4acda JV |
2158 | if (slave->link_failure_count < UINT_MAX) |
2159 | slave->link_failure_count++; | |
2160 | ||
f0c76d61 | 2161 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2162 | |
f0c76d61 JV |
2163 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2164 | bond->params.mode == BOND_MODE_8023AD) | |
2165 | bond_set_slave_inactive_flags(slave); | |
2166 | ||
a4aee5c8 JP |
2167 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2168 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2169 | |
2170 | if (bond->params.mode == BOND_MODE_8023AD) | |
2171 | bond_3ad_handle_link_change(slave, | |
2172 | BOND_LINK_DOWN); | |
2173 | ||
ae63e808 | 2174 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2175 | bond_alb_handle_link_change(bond, slave, |
2176 | BOND_LINK_DOWN); | |
2177 | ||
2178 | if (slave == bond->curr_active_slave) | |
2179 | goto do_failover; | |
2180 | ||
2181 | continue; | |
2182 | ||
2183 | default: | |
a4aee5c8 | 2184 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2185 | bond->dev->name, slave->new_link, |
2186 | slave->dev->name); | |
2187 | slave->new_link = BOND_LINK_NOCHANGE; | |
2188 | ||
2189 | continue; | |
2190 | } | |
2191 | ||
2192 | do_failover: | |
2193 | ASSERT_RTNL(); | |
e843fa50 | 2194 | block_netpoll_tx(); |
f0c76d61 JV |
2195 | write_lock_bh(&bond->curr_slave_lock); |
2196 | bond_select_active_slave(bond); | |
2197 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2198 | unblock_netpoll_tx(); |
f0c76d61 JV |
2199 | } |
2200 | ||
2201 | bond_set_carrier(bond); | |
1da177e4 LT |
2202 | } |
2203 | ||
0b0eef66 JV |
2204 | /* |
2205 | * bond_mii_monitor | |
2206 | * | |
2207 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2208 | * inspection, then (if inspection indicates something needs to be done) |
2209 | * an acquisition of appropriate locks followed by a commit phase to | |
2210 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2211 | */ |
2212 | void bond_mii_monitor(struct work_struct *work) | |
2213 | { | |
2214 | struct bonding *bond = container_of(work, struct bonding, | |
2215 | mii_work.work); | |
ad246c99 | 2216 | bool should_notify_peers = false; |
e6d265e8 | 2217 | unsigned long delay; |
0b0eef66 JV |
2218 | |
2219 | read_lock(&bond->lock); | |
e6d265e8 JV |
2220 | |
2221 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 | 2222 | |
dec1e90e | 2223 | if (list_empty(&bond->slave_list)) |
f0c76d61 | 2224 | goto re_arm; |
b59f9f74 | 2225 | |
ad246c99 BH |
2226 | should_notify_peers = bond_should_notify_peers(bond); |
2227 | ||
f0c76d61 | 2228 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2229 | read_unlock(&bond->lock); |
e6d265e8 JV |
2230 | |
2231 | /* Race avoidance with bond_close cancel of workqueue */ | |
2232 | if (!rtnl_trylock()) { | |
2233 | read_lock(&bond->lock); | |
2234 | delay = 1; | |
2235 | should_notify_peers = false; | |
2236 | goto re_arm; | |
2237 | } | |
2238 | ||
0b0eef66 | 2239 | read_lock(&bond->lock); |
f0c76d61 JV |
2240 | |
2241 | bond_miimon_commit(bond); | |
2242 | ||
5655662d JV |
2243 | read_unlock(&bond->lock); |
2244 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2245 | read_lock(&bond->lock); | |
0b0eef66 JV |
2246 | } |
2247 | ||
f0c76d61 | 2248 | re_arm: |
e6d265e8 JV |
2249 | if (bond->params.miimon) |
2250 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2251 | ||
0b0eef66 | 2252 | read_unlock(&bond->lock); |
ad246c99 BH |
2253 | |
2254 | if (should_notify_peers) { | |
6c8d23f7 | 2255 | if (!rtnl_trylock()) |
e6d265e8 | 2256 | return; |
b7bc2a5b | 2257 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2258 | rtnl_unlock(); |
2259 | } | |
0b0eef66 | 2260 | } |
c3ade5ca | 2261 | |
50223ce4 | 2262 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2263 | { |
50223ce4 VF |
2264 | struct net_device *upper; |
2265 | struct list_head *iter; | |
2266 | bool ret = false; | |
f5b2b966 | 2267 | |
eaddcd76 | 2268 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2269 | return true; |
f5b2b966 | 2270 | |
50223ce4 | 2271 | rcu_read_lock(); |
2f268f12 | 2272 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2273 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2274 | ret = true; | |
2275 | break; | |
2276 | } | |
f5b2b966 | 2277 | } |
50223ce4 | 2278 | rcu_read_unlock(); |
f5b2b966 | 2279 | |
50223ce4 | 2280 | return ret; |
f5b2b966 JV |
2281 | } |
2282 | ||
c3ade5ca JV |
2283 | /* |
2284 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2285 | * switches in VLAN mode (especially if ports are configured as | |
2286 | * "native" to a VLAN) might not pass non-tagged frames. | |
2287 | */ | |
d3bb52b0 | 2288 | 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 |
2289 | { |
2290 | struct sk_buff *skb; | |
2291 | ||
acca2674 | 2292 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op, |
2293 | slave_dev->name, &dest_ip, &src_ip, vlan_id); | |
3d632c3f | 2294 | |
c3ade5ca JV |
2295 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2296 | NULL, slave_dev->dev_addr, NULL); | |
2297 | ||
2298 | if (!skb) { | |
a4aee5c8 | 2299 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2300 | return; |
2301 | } | |
2302 | if (vlan_id) { | |
1fd9b1fc | 2303 | skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id); |
c3ade5ca | 2304 | if (!skb) { |
a4aee5c8 | 2305 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2306 | return; |
2307 | } | |
2308 | } | |
2309 | arp_xmit(skb); | |
2310 | } | |
2311 | ||
2312 | ||
1da177e4 LT |
2313 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2314 | { | |
27bc11e6 VF |
2315 | struct net_device *upper, *vlan_upper; |
2316 | struct list_head *iter, *vlan_iter; | |
c3ade5ca | 2317 | struct rtable *rt; |
27bc11e6 VF |
2318 | __be32 *targets = bond->params.arp_targets, addr; |
2319 | int i, vlan_id; | |
1da177e4 | 2320 | |
27bc11e6 | 2321 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
acca2674 | 2322 | pr_debug("basa: target %pI4\n", &targets[i]); |
c3ade5ca | 2323 | |
27bc11e6 | 2324 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2325 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2326 | RTO_ONLINK, 0); | |
b23dd4fe | 2327 | if (IS_ERR(rt)) { |
3e32582f VF |
2328 | pr_debug("%s: no route to arp_ip_target %pI4\n", |
2329 | bond->dev->name, &targets[i]); | |
c3ade5ca JV |
2330 | continue; |
2331 | } | |
2332 | ||
c3ade5ca | 2333 | vlan_id = 0; |
27bc11e6 VF |
2334 | |
2335 | /* bond device itself */ | |
2336 | if (rt->dst.dev == bond->dev) | |
2337 | goto found; | |
2338 | ||
2339 | rcu_read_lock(); | |
2340 | /* first we search only for vlan devices. for every vlan | |
2341 | * found we verify its upper dev list, searching for the | |
2342 | * rt->dst.dev. If found we save the tag of the vlan and | |
2343 | * proceed to send the packet. | |
2344 | * | |
2345 | * TODO: QinQ? | |
2346 | */ | |
2f268f12 VF |
2347 | netdev_for_each_all_upper_dev_rcu(bond->dev, vlan_upper, |
2348 | vlan_iter) { | |
27bc11e6 VF |
2349 | if (!is_vlan_dev(vlan_upper)) |
2350 | continue; | |
2f268f12 VF |
2351 | netdev_for_each_all_upper_dev_rcu(vlan_upper, upper, |
2352 | iter) { | |
27bc11e6 VF |
2353 | if (upper == rt->dst.dev) { |
2354 | vlan_id = vlan_dev_vlan_id(vlan_upper); | |
2355 | rcu_read_unlock(); | |
2356 | goto found; | |
2357 | } | |
c3ade5ca JV |
2358 | } |
2359 | } | |
2360 | ||
27bc11e6 VF |
2361 | /* if the device we're looking for is not on top of any of |
2362 | * our upper vlans, then just search for any dev that | |
2363 | * matches, and in case it's a vlan - save the id | |
2364 | */ | |
2f268f12 | 2365 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
27bc11e6 VF |
2366 | if (upper == rt->dst.dev) { |
2367 | /* if it's a vlan - get its VID */ | |
2368 | if (is_vlan_dev(upper)) | |
2369 | vlan_id = vlan_dev_vlan_id(upper); | |
2370 | ||
2371 | rcu_read_unlock(); | |
2372 | goto found; | |
2373 | } | |
c3ade5ca | 2374 | } |
27bc11e6 | 2375 | rcu_read_unlock(); |
c3ade5ca | 2376 | |
27bc11e6 | 2377 | /* Not our device - skip */ |
3e32582f VF |
2378 | pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2379 | bond->dev->name, &targets[i], | |
2380 | rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
2381 | ||
ed4b9f80 | 2382 | ip_rt_put(rt); |
27bc11e6 VF |
2383 | continue; |
2384 | ||
2385 | found: | |
2386 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2387 | ip_rt_put(rt); | |
2388 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
2389 | addr, vlan_id); | |
c3ade5ca JV |
2390 | } |
2391 | } | |
2392 | ||
d3bb52b0 | 2393 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2394 | { |
8599b52e VF |
2395 | int i; |
2396 | ||
87a7b84b VF |
2397 | if (!sip || !bond_has_this_ip(bond, tip)) { |
2398 | pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip); | |
2399 | return; | |
2400 | } | |
f5b2b966 | 2401 | |
8599b52e VF |
2402 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2403 | if (i == -1) { | |
87a7b84b VF |
2404 | pr_debug("bva: sip %pI4 not found in targets\n", &sip); |
2405 | return; | |
f5b2b966 | 2406 | } |
87a7b84b | 2407 | slave->last_arp_rx = jiffies; |
8599b52e | 2408 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2409 | } |
2410 | ||
5bb9e0b5 | 2411 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2412 | struct slave *slave) | |
f5b2b966 | 2413 | { |
de063b70 | 2414 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2415 | unsigned char *arp_ptr; |
d3bb52b0 | 2416 | __be32 sip, tip; |
de063b70 | 2417 | int alen; |
f5b2b966 | 2418 | |
3aba891d | 2419 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2420 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2421 | |
f5b2b966 | 2422 | read_lock(&bond->lock); |
2c146102 VF |
2423 | |
2424 | if (!slave_do_arp_validate(bond, slave)) | |
2425 | goto out_unlock; | |
2426 | ||
de063b70 | 2427 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2428 | |
3aba891d JP |
2429 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2430 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2431 | |
de063b70 ED |
2432 | if (alen > skb_headlen(skb)) { |
2433 | arp = kmalloc(alen, GFP_ATOMIC); | |
2434 | if (!arp) | |
2435 | goto out_unlock; | |
2436 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2437 | goto out_unlock; | |
2438 | } | |
f5b2b966 | 2439 | |
3aba891d | 2440 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2441 | skb->pkt_type == PACKET_OTHERHOST || |
2442 | skb->pkt_type == PACKET_LOOPBACK || | |
2443 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2444 | arp->ar_pro != htons(ETH_P_IP) || | |
2445 | arp->ar_pln != 4) | |
2446 | goto out_unlock; | |
2447 | ||
2448 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2449 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2450 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2451 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2452 | memcpy(&tip, arp_ptr, 4); |
2453 | ||
5a03cdb7 | 2454 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2455 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2456 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2457 | &sip, &tip); | |
f5b2b966 JV |
2458 | |
2459 | /* | |
2460 | * Backup slaves won't see the ARP reply, but do come through | |
2461 | * here for each ARP probe (so we swap the sip/tip to validate | |
2462 | * the probe). In a "redundant switch, common router" type of | |
2463 | * configuration, the ARP probe will (hopefully) travel from | |
2464 | * the active, through one switch, the router, then the other | |
2465 | * switch before reaching the backup. | |
aeea64ac VF |
2466 | * |
2467 | * We 'trust' the arp requests if there is an active slave and | |
2468 | * it received valid arp reply(s) after it became active. This | |
2469 | * is done to avoid endless looping when we can't reach the | |
2470 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2471 | */ |
e30bc066 | 2472 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2473 | bond_validate_arp(bond, slave, sip, tip); |
aeea64ac VF |
2474 | else if (bond->curr_active_slave && |
2475 | time_after(slave_last_rx(bond, bond->curr_active_slave), | |
2476 | bond->curr_active_slave->jiffies)) | |
f5b2b966 JV |
2477 | bond_validate_arp(bond, slave, tip, sip); |
2478 | ||
2479 | out_unlock: | |
2480 | read_unlock(&bond->lock); | |
de063b70 ED |
2481 | if (arp != (struct arphdr *)skb->data) |
2482 | kfree(arp); | |
13a8e0c8 | 2483 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2484 | } |
2485 | ||
e7f63f1d VF |
2486 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2487 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2488 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2489 | */ | |
2490 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2491 | int mod) | |
2492 | { | |
2493 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2494 | ||
2495 | return time_in_range(jiffies, | |
2496 | last_act - delta_in_ticks, | |
2497 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2498 | } | |
2499 | ||
1da177e4 LT |
2500 | /* |
2501 | * this function is called regularly to monitor each slave's link | |
2502 | * ensuring that traffic is being sent and received when arp monitoring | |
2503 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2504 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2505 | * arp monitoring in active backup mode. | |
2506 | */ | |
1b76b316 | 2507 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2508 | { |
1b76b316 JV |
2509 | struct bonding *bond = container_of(work, struct bonding, |
2510 | arp_work.work); | |
1da177e4 LT |
2511 | struct slave *slave, *oldcurrent; |
2512 | int do_failover = 0; | |
1da177e4 LT |
2513 | |
2514 | read_lock(&bond->lock); | |
2515 | ||
dec1e90e | 2516 | if (list_empty(&bond->slave_list)) |
1da177e4 | 2517 | goto re_arm; |
1da177e4 | 2518 | |
1da177e4 | 2519 | oldcurrent = bond->curr_active_slave; |
1da177e4 LT |
2520 | /* see if any of the previous devices are up now (i.e. they have |
2521 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2522 | * the picture unless it is null. also, slave->jiffies is not needed | |
2523 | * here because we send an arp on each slave and give a slave as | |
2524 | * long as it needs to get the tx/rx within the delta. | |
2525 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2526 | * so it can wait | |
2527 | */ | |
dec1e90e | 2528 | bond_for_each_slave(bond, slave) { |
cb32f2a0 JB |
2529 | unsigned long trans_start = dev_trans_start(slave->dev); |
2530 | ||
1da177e4 | 2531 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d VF |
2532 | if (bond_time_in_interval(bond, trans_start, 1) && |
2533 | bond_time_in_interval(bond, slave->dev->last_rx, 1)) { | |
1da177e4 LT |
2534 | |
2535 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2536 | bond_set_active_slave(slave); |
1da177e4 LT |
2537 | |
2538 | /* primary_slave has no meaning in round-robin | |
2539 | * mode. the window of a slave being up and | |
2540 | * curr_active_slave being null after enslaving | |
2541 | * is closed. | |
2542 | */ | |
2543 | if (!oldcurrent) { | |
a4aee5c8 JP |
2544 | pr_info("%s: link status definitely up for interface %s, ", |
2545 | bond->dev->name, | |
2546 | slave->dev->name); | |
1da177e4 LT |
2547 | do_failover = 1; |
2548 | } else { | |
a4aee5c8 JP |
2549 | pr_info("%s: interface %s is now up\n", |
2550 | bond->dev->name, | |
2551 | slave->dev->name); | |
1da177e4 LT |
2552 | } |
2553 | } | |
2554 | } else { | |
2555 | /* slave->link == BOND_LINK_UP */ | |
2556 | ||
2557 | /* not all switches will respond to an arp request | |
2558 | * when the source ip is 0, so don't take the link down | |
2559 | * if we don't know our ip yet | |
2560 | */ | |
e7f63f1d VF |
2561 | if (!bond_time_in_interval(bond, trans_start, 2) || |
2562 | !bond_time_in_interval(bond, slave->dev->last_rx, 2)) { | |
1da177e4 LT |
2563 | |
2564 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2565 | bond_set_backup_slave(slave); |
1da177e4 | 2566 | |
3d632c3f | 2567 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2568 | slave->link_failure_count++; |
1da177e4 | 2569 | |
a4aee5c8 JP |
2570 | pr_info("%s: interface %s is now down.\n", |
2571 | bond->dev->name, | |
2572 | slave->dev->name); | |
1da177e4 | 2573 | |
3d632c3f | 2574 | if (slave == oldcurrent) |
1da177e4 | 2575 | do_failover = 1; |
1da177e4 LT |
2576 | } |
2577 | } | |
2578 | ||
2579 | /* note: if switch is in round-robin mode, all links | |
2580 | * must tx arp to ensure all links rx an arp - otherwise | |
2581 | * links may oscillate or not come up at all; if switch is | |
2582 | * in something like xor mode, there is nothing we can | |
2583 | * do - all replies will be rx'ed on same link causing slaves | |
2584 | * to be unstable during low/no traffic periods | |
2585 | */ | |
3d632c3f | 2586 | if (IS_UP(slave->dev)) |
1da177e4 | 2587 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2588 | } |
2589 | ||
2590 | if (do_failover) { | |
e843fa50 | 2591 | block_netpoll_tx(); |
059fe7a5 | 2592 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2593 | |
2594 | bond_select_active_slave(bond); | |
2595 | ||
059fe7a5 | 2596 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2597 | unblock_netpoll_tx(); |
1da177e4 LT |
2598 | } |
2599 | ||
2600 | re_arm: | |
e6d265e8 | 2601 | if (bond->params.arp_interval) |
e7f63f1d VF |
2602 | queue_delayed_work(bond->wq, &bond->arp_work, |
2603 | msecs_to_jiffies(bond->params.arp_interval)); | |
e6d265e8 | 2604 | |
1da177e4 LT |
2605 | read_unlock(&bond->lock); |
2606 | } | |
2607 | ||
2608 | /* | |
b2220cad JV |
2609 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2610 | * changes. Sets new_link in slaves to specify what action should take | |
2611 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2612 | * to link states must be committed. | |
2613 | * | |
2614 | * Called with bond->lock held for read. | |
1da177e4 | 2615 | */ |
e7f63f1d | 2616 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2617 | { |
def4460c | 2618 | unsigned long trans_start, last_rx; |
dec1e90e | 2619 | struct slave *slave; |
dec1e90e | 2620 | int commit = 0; |
da210f55 | 2621 | |
dec1e90e | 2622 | bond_for_each_slave(bond, slave) { |
b2220cad | 2623 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2624 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2625 | |
b2220cad | 2626 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2627 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2628 | slave->new_link = BOND_LINK_UP; |
2629 | commit++; | |
2630 | } | |
b2220cad JV |
2631 | continue; |
2632 | } | |
1da177e4 | 2633 | |
b2220cad JV |
2634 | /* |
2635 | * Give slaves 2*delta after being enslaved or made | |
2636 | * active. This avoids bouncing, as the last receive | |
2637 | * times need a full ARP monitor cycle to be updated. | |
2638 | */ | |
e7f63f1d | 2639 | if (bond_time_in_interval(bond, slave->jiffies, 2)) |
b2220cad JV |
2640 | continue; |
2641 | ||
2642 | /* | |
2643 | * Backup slave is down if: | |
2644 | * - No current_arp_slave AND | |
2645 | * - more than 3*delta since last receive AND | |
2646 | * - the bond has an IP address | |
2647 | * | |
2648 | * Note: a non-null current_arp_slave indicates | |
2649 | * the curr_active_slave went down and we are | |
2650 | * searching for a new one; under this condition | |
2651 | * we only take the curr_active_slave down - this | |
2652 | * gives each slave a chance to tx/rx traffic | |
2653 | * before being taken out | |
2654 | */ | |
e30bc066 | 2655 | if (!bond_is_active_slave(slave) && |
b2220cad | 2656 | !bond->current_arp_slave && |
e7f63f1d | 2657 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2658 | slave->new_link = BOND_LINK_DOWN; |
2659 | commit++; | |
2660 | } | |
2661 | ||
2662 | /* | |
2663 | * Active slave is down if: | |
2664 | * - more than 2*delta since transmitting OR | |
2665 | * - (more than 2*delta since receive AND | |
2666 | * the bond has an IP address) | |
2667 | */ | |
cb32f2a0 | 2668 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2669 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2670 | (!bond_time_in_interval(bond, trans_start, 2) || |
2671 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2672 | slave->new_link = BOND_LINK_DOWN; |
2673 | commit++; | |
2674 | } | |
1da177e4 LT |
2675 | } |
2676 | ||
b2220cad JV |
2677 | return commit; |
2678 | } | |
1da177e4 | 2679 | |
b2220cad JV |
2680 | /* |
2681 | * Called to commit link state changes noted by inspection step of | |
2682 | * active-backup mode ARP monitor. | |
2683 | * | |
2684 | * Called with RTNL and bond->lock for read. | |
2685 | */ | |
e7f63f1d | 2686 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2687 | { |
cb32f2a0 | 2688 | unsigned long trans_start; |
dec1e90e | 2689 | struct slave *slave; |
1da177e4 | 2690 | |
dec1e90e | 2691 | bond_for_each_slave(bond, slave) { |
b2220cad JV |
2692 | switch (slave->new_link) { |
2693 | case BOND_LINK_NOCHANGE: | |
2694 | continue; | |
ff59c456 | 2695 | |
b2220cad | 2696 | case BOND_LINK_UP: |
cb32f2a0 | 2697 | trans_start = dev_trans_start(slave->dev); |
e7f63f1d VF |
2698 | if (bond->curr_active_slave != slave || |
2699 | (!bond->curr_active_slave && | |
2700 | bond_time_in_interval(bond, trans_start, 1))) { | |
b2220cad | 2701 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2702 | if (bond->current_arp_slave) { |
2703 | bond_set_slave_inactive_flags( | |
2704 | bond->current_arp_slave); | |
2705 | bond->current_arp_slave = NULL; | |
2706 | } | |
b2220cad | 2707 | |
a4aee5c8 | 2708 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2709 | bond->dev->name, slave->dev->name); |
b2220cad | 2710 | |
b9f60253 JP |
2711 | if (!bond->curr_active_slave || |
2712 | (slave == bond->primary_slave)) | |
2713 | goto do_failover; | |
1da177e4 | 2714 | |
b9f60253 | 2715 | } |
1da177e4 | 2716 | |
b9f60253 | 2717 | continue; |
1da177e4 | 2718 | |
b2220cad JV |
2719 | case BOND_LINK_DOWN: |
2720 | if (slave->link_failure_count < UINT_MAX) | |
2721 | slave->link_failure_count++; | |
2722 | ||
2723 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2724 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2725 | |
a4aee5c8 | 2726 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2727 | bond->dev->name, slave->dev->name); |
b2220cad | 2728 | |
b9f60253 | 2729 | if (slave == bond->curr_active_slave) { |
b2220cad | 2730 | bond->current_arp_slave = NULL; |
b9f60253 | 2731 | goto do_failover; |
1da177e4 | 2732 | } |
b9f60253 JP |
2733 | |
2734 | continue; | |
b2220cad JV |
2735 | |
2736 | default: | |
a4aee5c8 | 2737 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2738 | bond->dev->name, slave->new_link, |
2739 | slave->dev->name); | |
b9f60253 | 2740 | continue; |
1da177e4 | 2741 | } |
1da177e4 | 2742 | |
b9f60253 JP |
2743 | do_failover: |
2744 | ASSERT_RTNL(); | |
e843fa50 | 2745 | block_netpoll_tx(); |
b2220cad | 2746 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2747 | bond_select_active_slave(bond); |
b2220cad | 2748 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2749 | unblock_netpoll_tx(); |
b2220cad | 2750 | } |
1da177e4 | 2751 | |
b2220cad JV |
2752 | bond_set_carrier(bond); |
2753 | } | |
1da177e4 | 2754 | |
b2220cad JV |
2755 | /* |
2756 | * Send ARP probes for active-backup mode ARP monitor. | |
2757 | * | |
2758 | * Called with bond->lock held for read. | |
2759 | */ | |
2760 | static void bond_ab_arp_probe(struct bonding *bond) | |
2761 | { | |
dec1e90e | 2762 | struct slave *slave, *next_slave; |
b2220cad | 2763 | int i; |
1da177e4 | 2764 | |
b2220cad | 2765 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2766 | |
b2220cad | 2767 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
2768 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
2769 | bond->current_arp_slave->dev->name, | |
2770 | bond->curr_active_slave->dev->name); | |
1da177e4 | 2771 | |
b2220cad JV |
2772 | if (bond->curr_active_slave) { |
2773 | bond_arp_send_all(bond, bond->curr_active_slave); | |
2774 | read_unlock(&bond->curr_slave_lock); | |
2775 | return; | |
2776 | } | |
1da177e4 | 2777 | |
b2220cad | 2778 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 2779 | |
b2220cad JV |
2780 | /* if we don't have a curr_active_slave, search for the next available |
2781 | * backup slave from the current_arp_slave and make it the candidate | |
2782 | * for becoming the curr_active_slave | |
2783 | */ | |
1da177e4 | 2784 | |
b2220cad | 2785 | if (!bond->current_arp_slave) { |
dec1e90e | 2786 | bond->current_arp_slave = bond_first_slave(bond); |
b2220cad JV |
2787 | if (!bond->current_arp_slave) |
2788 | return; | |
2789 | } | |
1da177e4 | 2790 | |
b2220cad | 2791 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 2792 | |
b2220cad | 2793 | /* search for next candidate */ |
dec1e90e | 2794 | next_slave = bond_next_slave(bond, bond->current_arp_slave); |
2795 | bond_for_each_slave_from(bond, slave, i, next_slave) { | |
b2220cad JV |
2796 | if (IS_UP(slave->dev)) { |
2797 | slave->link = BOND_LINK_BACK; | |
2798 | bond_set_slave_active_flags(slave); | |
2799 | bond_arp_send_all(bond, slave); | |
1da177e4 | 2800 | slave->jiffies = jiffies; |
b2220cad JV |
2801 | bond->current_arp_slave = slave; |
2802 | break; | |
1da177e4 LT |
2803 | } |
2804 | ||
b2220cad JV |
2805 | /* if the link state is up at this point, we |
2806 | * mark it down - this can happen if we have | |
2807 | * simultaneous link failures and | |
2808 | * reselect_active_interface doesn't make this | |
2809 | * one the current slave so it is still marked | |
2810 | * up when it is actually down | |
1da177e4 | 2811 | */ |
b2220cad JV |
2812 | if (slave->link == BOND_LINK_UP) { |
2813 | slave->link = BOND_LINK_DOWN; | |
2814 | if (slave->link_failure_count < UINT_MAX) | |
2815 | slave->link_failure_count++; | |
1da177e4 | 2816 | |
b2220cad JV |
2817 | bond_set_slave_inactive_flags(slave); |
2818 | ||
a4aee5c8 JP |
2819 | pr_info("%s: backup interface %s is now down.\n", |
2820 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2821 | } |
b2220cad JV |
2822 | } |
2823 | } | |
1da177e4 | 2824 | |
b2220cad JV |
2825 | void bond_activebackup_arp_mon(struct work_struct *work) |
2826 | { | |
2827 | struct bonding *bond = container_of(work, struct bonding, | |
2828 | arp_work.work); | |
ad246c99 | 2829 | bool should_notify_peers = false; |
b2220cad | 2830 | int delta_in_ticks; |
1da177e4 | 2831 | |
b2220cad | 2832 | read_lock(&bond->lock); |
1da177e4 | 2833 | |
b2220cad | 2834 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2835 | |
dec1e90e | 2836 | if (list_empty(&bond->slave_list)) |
b2220cad JV |
2837 | goto re_arm; |
2838 | ||
ad246c99 BH |
2839 | should_notify_peers = bond_should_notify_peers(bond); |
2840 | ||
e7f63f1d | 2841 | if (bond_ab_arp_inspect(bond)) { |
b2220cad | 2842 | read_unlock(&bond->lock); |
e6d265e8 JV |
2843 | |
2844 | /* Race avoidance with bond_close flush of workqueue */ | |
2845 | if (!rtnl_trylock()) { | |
2846 | read_lock(&bond->lock); | |
2847 | delta_in_ticks = 1; | |
2848 | should_notify_peers = false; | |
2849 | goto re_arm; | |
2850 | } | |
2851 | ||
b2220cad JV |
2852 | read_lock(&bond->lock); |
2853 | ||
e7f63f1d | 2854 | bond_ab_arp_commit(bond); |
b2220cad JV |
2855 | |
2856 | read_unlock(&bond->lock); | |
2857 | rtnl_unlock(); | |
2858 | read_lock(&bond->lock); | |
1da177e4 LT |
2859 | } |
2860 | ||
b2220cad JV |
2861 | bond_ab_arp_probe(bond); |
2862 | ||
1da177e4 | 2863 | re_arm: |
e6d265e8 | 2864 | if (bond->params.arp_interval) |
1b76b316 | 2865 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 2866 | |
1da177e4 | 2867 | read_unlock(&bond->lock); |
ad246c99 BH |
2868 | |
2869 | if (should_notify_peers) { | |
6c8d23f7 | 2870 | if (!rtnl_trylock()) |
e6d265e8 | 2871 | return; |
b7bc2a5b | 2872 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2873 | rtnl_unlock(); |
2874 | } | |
1da177e4 LT |
2875 | } |
2876 | ||
1da177e4 LT |
2877 | /*-------------------------- netdev event handling --------------------------*/ |
2878 | ||
2879 | /* | |
2880 | * Change device name | |
2881 | */ | |
2882 | static int bond_event_changename(struct bonding *bond) | |
2883 | { | |
1da177e4 LT |
2884 | bond_remove_proc_entry(bond); |
2885 | bond_create_proc_entry(bond); | |
7e083840 | 2886 | |
f073c7ca TI |
2887 | bond_debug_reregister(bond); |
2888 | ||
1da177e4 LT |
2889 | return NOTIFY_DONE; |
2890 | } | |
2891 | ||
3d632c3f SH |
2892 | static int bond_master_netdev_event(unsigned long event, |
2893 | struct net_device *bond_dev) | |
1da177e4 | 2894 | { |
454d7c9b | 2895 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2896 | |
2897 | switch (event) { | |
2898 | case NETDEV_CHANGENAME: | |
2899 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2900 | case NETDEV_UNREGISTER: |
2901 | bond_remove_proc_entry(event_bond); | |
2902 | break; | |
2903 | case NETDEV_REGISTER: | |
2904 | bond_create_proc_entry(event_bond); | |
2905 | break; | |
6c8d23f7 | 2906 | case NETDEV_NOTIFY_PEERS: |
2907 | if (event_bond->send_peer_notif) | |
2908 | event_bond->send_peer_notif--; | |
2909 | break; | |
1da177e4 LT |
2910 | default: |
2911 | break; | |
2912 | } | |
2913 | ||
2914 | return NOTIFY_DONE; | |
2915 | } | |
2916 | ||
3d632c3f SH |
2917 | static int bond_slave_netdev_event(unsigned long event, |
2918 | struct net_device *slave_dev) | |
1da177e4 | 2919 | { |
471cb5a3 | 2920 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
6101391d | 2921 | struct bonding *bond; |
2922 | struct net_device *bond_dev; | |
471cb5a3 JP |
2923 | u32 old_speed; |
2924 | u8 old_duplex; | |
1da177e4 | 2925 | |
6101391d | 2926 | /* A netdev event can be generated while enslaving a device |
2927 | * before netdev_rx_handler_register is called in which case | |
2928 | * slave will be NULL | |
2929 | */ | |
2930 | if (!slave) | |
2931 | return NOTIFY_DONE; | |
2932 | bond_dev = slave->bond->dev; | |
2933 | bond = slave->bond; | |
2934 | ||
1da177e4 LT |
2935 | switch (event) { |
2936 | case NETDEV_UNREGISTER: | |
8d2ada77 | 2937 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
2938 | bond_release_and_destroy(bond_dev, slave_dev); |
2939 | else | |
2940 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 2941 | break; |
98f41f69 | 2942 | case NETDEV_UP: |
1da177e4 | 2943 | case NETDEV_CHANGE: |
471cb5a3 JP |
2944 | old_speed = slave->speed; |
2945 | old_duplex = slave->duplex; | |
17d04500 | 2946 | |
471cb5a3 | 2947 | bond_update_speed_duplex(slave); |
17d04500 | 2948 | |
471cb5a3 JP |
2949 | if (bond->params.mode == BOND_MODE_8023AD) { |
2950 | if (old_speed != slave->speed) | |
2951 | bond_3ad_adapter_speed_changed(slave); | |
2952 | if (old_duplex != slave->duplex) | |
2953 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 2954 | } |
1da177e4 LT |
2955 | break; |
2956 | case NETDEV_DOWN: | |
2957 | /* | |
2958 | * ... Or is it this? | |
2959 | */ | |
2960 | break; | |
2961 | case NETDEV_CHANGEMTU: | |
2962 | /* | |
2963 | * TODO: Should slaves be allowed to | |
2964 | * independently alter their MTU? For | |
2965 | * an active-backup bond, slaves need | |
2966 | * not be the same type of device, so | |
2967 | * MTUs may vary. For other modes, | |
2968 | * slaves arguably should have the | |
2969 | * same MTUs. To do this, we'd need to | |
2970 | * take over the slave's change_mtu | |
2971 | * function for the duration of their | |
2972 | * servitude. | |
2973 | */ | |
2974 | break; | |
2975 | case NETDEV_CHANGENAME: | |
2976 | /* | |
2977 | * TODO: handle changing the primary's name | |
2978 | */ | |
2979 | break; | |
8531c5ff AK |
2980 | case NETDEV_FEAT_CHANGE: |
2981 | bond_compute_features(bond); | |
2982 | break; | |
4aa5dee4 JP |
2983 | case NETDEV_RESEND_IGMP: |
2984 | /* Propagate to master device */ | |
2985 | call_netdevice_notifiers(event, slave->bond->dev); | |
2986 | break; | |
1da177e4 LT |
2987 | default: |
2988 | break; | |
2989 | } | |
2990 | ||
2991 | return NOTIFY_DONE; | |
2992 | } | |
2993 | ||
2994 | /* | |
2995 | * bond_netdev_event: handle netdev notifier chain events. | |
2996 | * | |
2997 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 2998 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
2999 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3000 | * dev_probe_lock). | |
3001 | */ | |
3d632c3f SH |
3002 | static int bond_netdev_event(struct notifier_block *this, |
3003 | unsigned long event, void *ptr) | |
1da177e4 | 3004 | { |
351638e7 | 3005 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3006 | |
5a03cdb7 | 3007 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3008 | event_dev ? event_dev->name : "None", |
3009 | event); | |
1da177e4 | 3010 | |
0b680e75 JV |
3011 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3012 | return NOTIFY_DONE; | |
3013 | ||
1da177e4 | 3014 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3015 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3016 | return bond_master_netdev_event(event, event_dev); |
3017 | } | |
3018 | ||
3019 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3020 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3021 | return bond_slave_netdev_event(event, event_dev); |
3022 | } | |
3023 | ||
3024 | return NOTIFY_DONE; | |
3025 | } | |
3026 | ||
3027 | static struct notifier_block bond_netdev_notifier = { | |
3028 | .notifier_call = bond_netdev_event, | |
3029 | }; | |
3030 | ||
169a3e66 JV |
3031 | /*---------------------------- Hashing Policies -----------------------------*/ |
3032 | ||
6b923cb7 JE |
3033 | /* |
3034 | * Hash for the output device based upon layer 2 data | |
3035 | */ | |
3036 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) | |
3037 | { | |
3038 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3039 | ||
3040 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
3041 | return (data->h_dest[5] ^ data->h_source[5]) % count; | |
3042 | ||
3043 | return 0; | |
3044 | } | |
3045 | ||
6f6652be JV |
3046 | /* |
3047 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
6b923cb7 | 3048 | * the packet is not IP, fall back on bond_xmit_hash_policy_l2() |
6f6652be | 3049 | */ |
a361c83c | 3050 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be | 3051 | { |
4394542c ED |
3052 | const struct ethhdr *data; |
3053 | const struct iphdr *iph; | |
3054 | const struct ipv6hdr *ipv6h; | |
6b923cb7 | 3055 | u32 v6hash; |
4394542c | 3056 | const __be32 *s, *d; |
6b923cb7 JE |
3057 | |
3058 | if (skb->protocol == htons(ETH_P_IP) && | |
4394542c | 3059 | pskb_network_may_pull(skb, sizeof(*iph))) { |
6b923cb7 | 3060 | iph = ip_hdr(skb); |
4394542c | 3061 | data = (struct ethhdr *)skb->data; |
6f6652be | 3062 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3063 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6b923cb7 | 3064 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3065 | pskb_network_may_pull(skb, sizeof(*ipv6h))) { |
6b923cb7 | 3066 | ipv6h = ipv6_hdr(skb); |
4394542c | 3067 | data = (struct ethhdr *)skb->data; |
6b923cb7 JE |
3068 | s = &ipv6h->saddr.s6_addr32[0]; |
3069 | d = &ipv6h->daddr.s6_addr32[0]; | |
3070 | v6hash = (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3071 | v6hash ^= (v6hash >> 24) ^ (v6hash >> 16) ^ (v6hash >> 8); | |
3072 | return (v6hash ^ data->h_dest[5] ^ data->h_source[5]) % count; | |
6f6652be JV |
3073 | } |
3074 | ||
6b923cb7 | 3075 | return bond_xmit_hash_policy_l2(skb, count); |
6f6652be JV |
3076 | } |
3077 | ||
169a3e66 | 3078 | /* |
59c51591 | 3079 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 | 3080 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
6b923cb7 | 3081 | * altogether not IP, fall back on bond_xmit_hash_policy_l2() |
169a3e66 | 3082 | */ |
a361c83c | 3083 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 | 3084 | { |
6b923cb7 | 3085 | u32 layer4_xor = 0; |
4394542c ED |
3086 | const struct iphdr *iph; |
3087 | const struct ipv6hdr *ipv6h; | |
3088 | const __be32 *s, *d; | |
3089 | const __be16 *l4 = NULL; | |
3090 | __be16 _l4[2]; | |
3091 | int noff = skb_network_offset(skb); | |
3092 | int poff; | |
169a3e66 | 3093 | |
6b923cb7 | 3094 | if (skb->protocol == htons(ETH_P_IP) && |
4394542c | 3095 | pskb_may_pull(skb, noff + sizeof(*iph))) { |
6b923cb7 | 3096 | iph = ip_hdr(skb); |
4394542c ED |
3097 | poff = proto_ports_offset(iph->protocol); |
3098 | ||
3099 | if (!ip_is_fragment(iph) && poff >= 0) { | |
3100 | l4 = skb_header_pointer(skb, noff + (iph->ihl << 2) + poff, | |
3101 | sizeof(_l4), &_l4); | |
3102 | if (l4) | |
3103 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
169a3e66 JV |
3104 | } |
3105 | return (layer4_xor ^ | |
3106 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
6b923cb7 | 3107 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3108 | pskb_may_pull(skb, noff + sizeof(*ipv6h))) { |
6b923cb7 | 3109 | ipv6h = ipv6_hdr(skb); |
4394542c ED |
3110 | poff = proto_ports_offset(ipv6h->nexthdr); |
3111 | if (poff >= 0) { | |
3112 | l4 = skb_header_pointer(skb, noff + sizeof(*ipv6h) + poff, | |
3113 | sizeof(_l4), &_l4); | |
3114 | if (l4) | |
3115 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
6b923cb7 JE |
3116 | } |
3117 | s = &ipv6h->saddr.s6_addr32[0]; | |
3118 | d = &ipv6h->daddr.s6_addr32[0]; | |
3119 | layer4_xor ^= (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3120 | layer4_xor ^= (layer4_xor >> 24) ^ (layer4_xor >> 16) ^ | |
3121 | (layer4_xor >> 8); | |
3122 | return layer4_xor % count; | |
169a3e66 JV |
3123 | } |
3124 | ||
6b923cb7 | 3125 | return bond_xmit_hash_policy_l2(skb, count); |
169a3e66 JV |
3126 | } |
3127 | ||
1da177e4 LT |
3128 | /*-------------------------- Device entry points ----------------------------*/ |
3129 | ||
fbb0c41b | 3130 | static void bond_work_init_all(struct bonding *bond) |
3131 | { | |
3132 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3133 | bond_resend_igmp_join_requests_delayed); | |
3134 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3135 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3136 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3137 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3138 | else | |
3139 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3140 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3141 | } | |
3142 | ||
3143 | static void bond_work_cancel_all(struct bonding *bond) | |
3144 | { | |
3145 | cancel_delayed_work_sync(&bond->mii_work); | |
3146 | cancel_delayed_work_sync(&bond->arp_work); | |
3147 | cancel_delayed_work_sync(&bond->alb_work); | |
3148 | cancel_delayed_work_sync(&bond->ad_work); | |
3149 | cancel_delayed_work_sync(&bond->mcast_work); | |
3150 | } | |
3151 | ||
1da177e4 LT |
3152 | static int bond_open(struct net_device *bond_dev) |
3153 | { | |
454d7c9b | 3154 | struct bonding *bond = netdev_priv(bond_dev); |
ba3211cc | 3155 | struct slave *slave; |
1da177e4 | 3156 | |
ba3211cc PP |
3157 | /* reset slave->backup and slave->inactive */ |
3158 | read_lock(&bond->lock); | |
dec1e90e | 3159 | if (!list_empty(&bond->slave_list)) { |
ba3211cc | 3160 | read_lock(&bond->curr_slave_lock); |
dec1e90e | 3161 | bond_for_each_slave(bond, slave) { |
ba3211cc PP |
3162 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3163 | && (slave != bond->curr_active_slave)) { | |
3164 | bond_set_slave_inactive_flags(slave); | |
3165 | } else { | |
3166 | bond_set_slave_active_flags(slave); | |
3167 | } | |
3168 | } | |
3169 | read_unlock(&bond->curr_slave_lock); | |
3170 | } | |
3171 | read_unlock(&bond->lock); | |
3172 | ||
fbb0c41b | 3173 | bond_work_init_all(bond); |
5a37e8ca | 3174 | |
58402054 | 3175 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3176 | /* bond_alb_initialize must be called before the timer |
3177 | * is started. | |
3178 | */ | |
fbb0c41b | 3179 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3180 | return -ENOMEM; |
1b76b316 | 3181 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3182 | } |
3183 | ||
fbb0c41b | 3184 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3185 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3186 | |
3187 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3188 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
f5b2b966 | 3189 | if (bond->params.arp_validate) |
3aba891d | 3190 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3191 | } |
3192 | ||
3193 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3194 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3195 | /* register to receive LACPDUs */ |
3aba891d | 3196 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3197 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3198 | } |
3199 | ||
3200 | return 0; | |
3201 | } | |
3202 | ||
3203 | static int bond_close(struct net_device *bond_dev) | |
3204 | { | |
454d7c9b | 3205 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3206 | |
fbb0c41b | 3207 | bond_work_cancel_all(bond); |
6c8d23f7 | 3208 | bond->send_peer_notif = 0; |
ee8487c0 | 3209 | if (bond_is_lb(bond)) |
1da177e4 | 3210 | bond_alb_deinitialize(bond); |
3aba891d | 3211 | bond->recv_probe = NULL; |
1da177e4 LT |
3212 | |
3213 | return 0; | |
3214 | } | |
3215 | ||
28172739 ED |
3216 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3217 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3218 | { |
454d7c9b | 3219 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3220 | struct rtnl_link_stats64 temp; |
1da177e4 | 3221 | struct slave *slave; |
1da177e4 | 3222 | |
28172739 | 3223 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3224 | |
3225 | read_lock_bh(&bond->lock); | |
dec1e90e | 3226 | bond_for_each_slave(bond, slave) { |
be1f3c2c | 3227 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3228 | dev_get_stats(slave->dev, &temp); |
3229 | ||
3230 | stats->rx_packets += sstats->rx_packets; | |
3231 | stats->rx_bytes += sstats->rx_bytes; | |
3232 | stats->rx_errors += sstats->rx_errors; | |
3233 | stats->rx_dropped += sstats->rx_dropped; | |
3234 | ||
3235 | stats->tx_packets += sstats->tx_packets; | |
3236 | stats->tx_bytes += sstats->tx_bytes; | |
3237 | stats->tx_errors += sstats->tx_errors; | |
3238 | stats->tx_dropped += sstats->tx_dropped; | |
3239 | ||
3240 | stats->multicast += sstats->multicast; | |
3241 | stats->collisions += sstats->collisions; | |
3242 | ||
3243 | stats->rx_length_errors += sstats->rx_length_errors; | |
3244 | stats->rx_over_errors += sstats->rx_over_errors; | |
3245 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3246 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3247 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3248 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3249 | ||
3250 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3251 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3252 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3253 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3254 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb | 3255 | } |
1da177e4 LT |
3256 | read_unlock_bh(&bond->lock); |
3257 | ||
3258 | return stats; | |
3259 | } | |
3260 | ||
3261 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3262 | { | |
3263 | struct net_device *slave_dev = NULL; | |
3264 | struct ifbond k_binfo; | |
3265 | struct ifbond __user *u_binfo = NULL; | |
3266 | struct ifslave k_sinfo; | |
3267 | struct ifslave __user *u_sinfo = NULL; | |
3268 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3269 | struct net *net; |
1da177e4 LT |
3270 | int res = 0; |
3271 | ||
a4aee5c8 | 3272 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3273 | |
3274 | switch (cmd) { | |
1da177e4 LT |
3275 | case SIOCGMIIPHY: |
3276 | mii = if_mii(ifr); | |
3d632c3f | 3277 | if (!mii) |
1da177e4 | 3278 | return -EINVAL; |
3d632c3f | 3279 | |
1da177e4 LT |
3280 | mii->phy_id = 0; |
3281 | /* Fall Through */ | |
3282 | case SIOCGMIIREG: | |
3283 | /* | |
3284 | * We do this again just in case we were called by SIOCGMIIREG | |
3285 | * instead of SIOCGMIIPHY. | |
3286 | */ | |
3287 | mii = if_mii(ifr); | |
3d632c3f | 3288 | if (!mii) |
1da177e4 | 3289 | return -EINVAL; |
3d632c3f | 3290 | |
1da177e4 LT |
3291 | |
3292 | if (mii->reg_num == 1) { | |
454d7c9b | 3293 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3294 | mii->val_out = 0; |
6603a6f2 | 3295 | read_lock(&bond->lock); |
1da177e4 | 3296 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3297 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3298 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3299 | |
1da177e4 | 3300 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3301 | read_unlock(&bond->lock); |
1da177e4 LT |
3302 | } |
3303 | ||
3304 | return 0; | |
3305 | case BOND_INFO_QUERY_OLD: | |
3306 | case SIOCBONDINFOQUERY: | |
3307 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3308 | ||
3d632c3f | 3309 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3310 | return -EFAULT; |
1da177e4 LT |
3311 | |
3312 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3313 | if (res == 0 && |
3314 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3315 | return -EFAULT; | |
1da177e4 LT |
3316 | |
3317 | return res; | |
3318 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3319 | case SIOCBONDSLAVEINFOQUERY: | |
3320 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3321 | ||
3d632c3f | 3322 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3323 | return -EFAULT; |
1da177e4 LT |
3324 | |
3325 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3326 | if (res == 0 && |
3327 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3328 | return -EFAULT; | |
1da177e4 LT |
3329 | |
3330 | return res; | |
3331 | default: | |
3332 | /* Go on */ | |
3333 | break; | |
3334 | } | |
3335 | ||
387ff911 G |
3336 | net = dev_net(bond_dev); |
3337 | ||
3338 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3339 | return -EPERM; |
1da177e4 | 3340 | |
387ff911 | 3341 | slave_dev = dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3342 | |
a4aee5c8 | 3343 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3344 | |
3d632c3f | 3345 | if (!slave_dev) |
1da177e4 | 3346 | res = -ENODEV; |
3d632c3f | 3347 | else { |
a4aee5c8 | 3348 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3349 | switch (cmd) { |
3350 | case BOND_ENSLAVE_OLD: | |
3351 | case SIOCBONDENSLAVE: | |
3352 | res = bond_enslave(bond_dev, slave_dev); | |
3353 | break; | |
3354 | case BOND_RELEASE_OLD: | |
3355 | case SIOCBONDRELEASE: | |
3356 | res = bond_release(bond_dev, slave_dev); | |
3357 | break; | |
3358 | case BOND_SETHWADDR_OLD: | |
3359 | case SIOCBONDSETHWADDR: | |
409cc1f8 JP |
3360 | bond_set_dev_addr(bond_dev, slave_dev); |
3361 | res = 0; | |
1da177e4 LT |
3362 | break; |
3363 | case BOND_CHANGE_ACTIVE_OLD: | |
3364 | case SIOCBONDCHANGEACTIVE: | |
3365 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3366 | break; | |
3367 | default: | |
3368 | res = -EOPNOTSUPP; | |
3369 | } | |
3370 | ||
3371 | dev_put(slave_dev); | |
3372 | } | |
3373 | ||
1da177e4 LT |
3374 | return res; |
3375 | } | |
3376 | ||
d03462b9 | 3377 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3378 | { |
454d7c9b | 3379 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3380 | |
d03462b9 JP |
3381 | if (change & IFF_PROMISC) |
3382 | bond_set_promiscuity(bond, | |
3383 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3384 | |
d03462b9 JP |
3385 | if (change & IFF_ALLMULTI) |
3386 | bond_set_allmulti(bond, | |
3387 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3388 | } | |
1da177e4 | 3389 | |
303d1cbf | 3390 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3391 | { |
3392 | struct bonding *bond = netdev_priv(bond_dev); | |
303d1cbf | 3393 | struct slave *slave; |
1da177e4 | 3394 | |
7864a1ad | 3395 | ASSERT_RTNL(); |
80ee5ad2 | 3396 | |
303d1cbf | 3397 | if (USES_PRIMARY(bond->params.mode)) { |
7864a1ad | 3398 | slave = rtnl_dereference(bond->curr_active_slave); |
303d1cbf JV |
3399 | if (slave) { |
3400 | dev_uc_sync(slave->dev, bond_dev); | |
3401 | dev_mc_sync(slave->dev, bond_dev); | |
3402 | } | |
303d1cbf | 3403 | } else { |
dec1e90e | 3404 | bond_for_each_slave(bond, slave) { |
303d1cbf JV |
3405 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3406 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3407 | } | |
1da177e4 | 3408 | } |
1da177e4 LT |
3409 | } |
3410 | ||
234bcf8a | 3411 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3412 | { |
234bcf8a | 3413 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3414 | const struct net_device_ops *slave_ops; |
3415 | struct neigh_parms parms; | |
dec1e90e | 3416 | struct slave *slave; |
234bcf8a SP |
3417 | int ret; |
3418 | ||
dec1e90e | 3419 | slave = bond_first_slave(bond); |
234bcf8a SP |
3420 | if (!slave) |
3421 | return 0; | |
234bcf8a | 3422 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3423 | if (!slave_ops->ndo_neigh_setup) |
3424 | return 0; | |
3425 | ||
3426 | parms.neigh_setup = NULL; | |
3427 | parms.neigh_cleanup = NULL; | |
3428 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3429 | if (ret) | |
3430 | return ret; | |
3431 | ||
3432 | /* | |
3433 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3434 | * after the last slave has been detached. Assumes that all slaves | |
3435 | * utilize the same neigh_cleanup (true at this writing as only user | |
3436 | * is ipoib). | |
3437 | */ | |
3438 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3439 | ||
3440 | if (!parms.neigh_setup) | |
3441 | return 0; | |
3442 | ||
3443 | return parms.neigh_setup(n); | |
3444 | } | |
3445 | ||
3446 | /* | |
3447 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3448 | * slave exists. So we must declare proxy setup function which will | |
3449 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3450 | * |
3451 | * It's also called by master devices (such as vlans) to setup their | |
3452 | * underlying devices. In that case - do nothing, we're already set up from | |
3453 | * our init. | |
234bcf8a SP |
3454 | */ |
3455 | static int bond_neigh_setup(struct net_device *dev, | |
3456 | struct neigh_parms *parms) | |
3457 | { | |
9918d5bf VF |
3458 | /* modify only our neigh_parms */ |
3459 | if (parms->dev == dev) | |
3460 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3461 | |
00829823 SH |
3462 | return 0; |
3463 | } | |
3464 | ||
1da177e4 LT |
3465 | /* |
3466 | * Change the MTU of all of a master's slaves to match the master | |
3467 | */ | |
3468 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3469 | { | |
454d7c9b | 3470 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 3471 | struct slave *slave; |
1da177e4 | 3472 | int res = 0; |
1da177e4 | 3473 | |
5a03cdb7 | 3474 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3475 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3476 | |
3477 | /* Can't hold bond->lock with bh disabled here since | |
3478 | * some base drivers panic. On the other hand we can't | |
3479 | * hold bond->lock without bh disabled because we'll | |
3480 | * deadlock. The only solution is to rely on the fact | |
3481 | * that we're under rtnl_lock here, and the slaves | |
3482 | * list won't change. This doesn't solve the problem | |
3483 | * of setting the slave's MTU while it is | |
3484 | * transmitting, but the assumption is that the base | |
3485 | * driver can handle that. | |
3486 | * | |
3487 | * TODO: figure out a way to safely iterate the slaves | |
3488 | * list, but without holding a lock around the actual | |
3489 | * call to the base driver. | |
3490 | */ | |
3491 | ||
dec1e90e | 3492 | bond_for_each_slave(bond, slave) { |
a4aee5c8 JP |
3493 | pr_debug("s %p s->p %p c_m %p\n", |
3494 | slave, | |
dec1e90e | 3495 | bond_prev_slave(bond, slave), |
a4aee5c8 | 3496 | slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 3497 | |
1da177e4 LT |
3498 | res = dev_set_mtu(slave->dev, new_mtu); |
3499 | ||
3500 | if (res) { | |
3501 | /* If we failed to set the slave's mtu to the new value | |
3502 | * we must abort the operation even in ACTIVE_BACKUP | |
3503 | * mode, because if we allow the backup slaves to have | |
3504 | * different mtu values than the active slave we'll | |
3505 | * need to change their mtu when doing a failover. That | |
3506 | * means changing their mtu from timer context, which | |
3507 | * is probably not a good idea. | |
3508 | */ | |
5a03cdb7 | 3509 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3510 | goto unwind; |
3511 | } | |
3512 | } | |
3513 | ||
3514 | bond_dev->mtu = new_mtu; | |
3515 | ||
3516 | return 0; | |
3517 | ||
3518 | unwind: | |
3519 | /* unwind from head to the slave that failed */ | |
dec1e90e | 3520 | bond_for_each_slave_continue_reverse(bond, slave) { |
1da177e4 LT |
3521 | int tmp_res; |
3522 | ||
3523 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3524 | if (tmp_res) { | |
a4aee5c8 JP |
3525 | pr_debug("unwind err %d dev %s\n", |
3526 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3527 | } |
3528 | } | |
3529 | ||
3530 | return res; | |
3531 | } | |
3532 | ||
3533 | /* | |
3534 | * Change HW address | |
3535 | * | |
3536 | * Note that many devices must be down to change the HW address, and | |
3537 | * downing the master releases all slaves. We can make bonds full of | |
3538 | * bonding devices to test this, however. | |
3539 | */ | |
3540 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3541 | { | |
454d7c9b | 3542 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3543 | struct sockaddr *sa = addr, tmp_sa; |
dec1e90e | 3544 | struct slave *slave; |
1da177e4 | 3545 | int res = 0; |
1da177e4 | 3546 | |
eb7cc59a SH |
3547 | if (bond->params.mode == BOND_MODE_ALB) |
3548 | return bond_alb_set_mac_address(bond_dev, addr); | |
3549 | ||
3550 | ||
a4aee5c8 JP |
3551 | pr_debug("bond=%p, name=%s\n", |
3552 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3553 | |
1b5acd29 JV |
3554 | /* If fail_over_mac is enabled, do nothing and return success. |
3555 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3556 | */ |
1b5acd29 | 3557 | if (bond->params.fail_over_mac) |
dd957c57 | 3558 | return 0; |
2ab82852 | 3559 | |
3d632c3f | 3560 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3561 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3562 | |
3563 | /* Can't hold bond->lock with bh disabled here since | |
3564 | * some base drivers panic. On the other hand we can't | |
3565 | * hold bond->lock without bh disabled because we'll | |
3566 | * deadlock. The only solution is to rely on the fact | |
3567 | * that we're under rtnl_lock here, and the slaves | |
3568 | * list won't change. This doesn't solve the problem | |
3569 | * of setting the slave's hw address while it is | |
3570 | * transmitting, but the assumption is that the base | |
3571 | * driver can handle that. | |
3572 | * | |
3573 | * TODO: figure out a way to safely iterate the slaves | |
3574 | * list, but without holding a lock around the actual | |
3575 | * call to the base driver. | |
3576 | */ | |
3577 | ||
dec1e90e | 3578 | bond_for_each_slave(bond, slave) { |
eb7cc59a | 3579 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3580 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3581 | |
eb7cc59a | 3582 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3583 | res = -EOPNOTSUPP; |
5a03cdb7 | 3584 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3585 | goto unwind; |
3586 | } | |
3587 | ||
3588 | res = dev_set_mac_address(slave->dev, addr); | |
3589 | if (res) { | |
3590 | /* TODO: consider downing the slave | |
3591 | * and retry ? | |
3592 | * User should expect communications | |
3593 | * breakage anyway until ARP finish | |
3594 | * updating, so... | |
3595 | */ | |
5a03cdb7 | 3596 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3597 | goto unwind; |
3598 | } | |
3599 | } | |
3600 | ||
3601 | /* success */ | |
3602 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3603 | return 0; | |
3604 | ||
3605 | unwind: | |
3606 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3607 | tmp_sa.sa_family = bond_dev->type; | |
3608 | ||
3609 | /* unwind from head to the slave that failed */ | |
dec1e90e | 3610 | bond_for_each_slave_continue_reverse(bond, slave) { |
1da177e4 LT |
3611 | int tmp_res; |
3612 | ||
3613 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
3614 | if (tmp_res) { | |
a4aee5c8 JP |
3615 | pr_debug("unwind err %d dev %s\n", |
3616 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3617 | } |
3618 | } | |
3619 | ||
3620 | return res; | |
3621 | } | |
3622 | ||
15077228 NA |
3623 | /** |
3624 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3625 | * @bond: bonding device that is transmitting | |
3626 | * @skb: buffer to transmit | |
3627 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3628 | * | |
3629 | * This function tries to transmit through slave with slave_id but in case | |
3630 | * it fails, it tries to find the first available slave for transmission. | |
3631 | * The skb is consumed in all cases, thus the function is void. | |
3632 | */ | |
3633 | void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) | |
3634 | { | |
3635 | struct slave *slave; | |
3636 | int i = slave_id; | |
3637 | ||
3638 | /* Here we start from the slave with slave_id */ | |
278b2083 | 3639 | bond_for_each_slave_rcu(bond, slave) { |
15077228 NA |
3640 | if (--i < 0) { |
3641 | if (slave_can_tx(slave)) { | |
3642 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3643 | return; | |
3644 | } | |
3645 | } | |
3646 | } | |
3647 | ||
3648 | /* Here we start from the first slave up to slave_id */ | |
3649 | i = slave_id; | |
278b2083 | 3650 | bond_for_each_slave_rcu(bond, slave) { |
15077228 NA |
3651 | if (--i < 0) |
3652 | break; | |
3653 | if (slave_can_tx(slave)) { | |
3654 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3655 | return; | |
3656 | } | |
3657 | } | |
3658 | /* no slave that can tx has been found */ | |
3659 | kfree_skb(skb); | |
3660 | } | |
3661 | ||
1da177e4 LT |
3662 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3663 | { | |
454d7c9b | 3664 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3665 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3666 | struct slave *slave; |
1da177e4 | 3667 | |
cf5f9044 | 3668 | /* |
a2fd940f AG |
3669 | * Start with the curr_active_slave that joined the bond as the |
3670 | * default for sending IGMP traffic. For failover purposes one | |
3671 | * needs to maintain some consistency for the interface that will | |
3672 | * send the join/membership reports. The curr_active_slave found | |
3673 | * will send all of this type of traffic. | |
cf5f9044 | 3674 | */ |
15077228 | 3675 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3676 | slave = rcu_dereference(bond->curr_active_slave); |
15077228 NA |
3677 | if (slave && slave_can_tx(slave)) |
3678 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3679 | else | |
3680 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3681 | } else { |
15077228 NA |
3682 | bond_xmit_slave_id(bond, skb, |
3683 | bond->rr_tx_counter++ % bond->slave_cnt); | |
1da177e4 | 3684 | } |
0693e88e | 3685 | |
ec634fe3 | 3686 | return NETDEV_TX_OK; |
1da177e4 LT |
3687 | } |
3688 | ||
3689 | /* | |
3690 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3691 | * the bond has a usable interface. | |
3692 | */ | |
3693 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3694 | { | |
454d7c9b | 3695 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3696 | struct slave *slave; |
1da177e4 | 3697 | |
278b2083 | 3698 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3699 | if (slave) |
15077228 NA |
3700 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3701 | else | |
04502430 | 3702 | kfree_skb(skb); |
0693e88e | 3703 | |
ec634fe3 | 3704 | return NETDEV_TX_OK; |
1da177e4 LT |
3705 | } |
3706 | ||
3707 | /* | |
169a3e66 JV |
3708 | * In bond_xmit_xor() , we determine the output device by using a pre- |
3709 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
3710 | * find the next active slave. | |
1da177e4 LT |
3711 | */ |
3712 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3713 | { | |
454d7c9b | 3714 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3715 | |
15077228 NA |
3716 | bond_xmit_slave_id(bond, skb, |
3717 | bond->xmit_hash_policy(skb, bond->slave_cnt)); | |
0693e88e | 3718 | |
ec634fe3 | 3719 | return NETDEV_TX_OK; |
1da177e4 LT |
3720 | } |
3721 | ||
78a646ce | 3722 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3723 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3724 | { | |
454d7c9b | 3725 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3726 | struct slave *slave = NULL; |
1da177e4 | 3727 | |
278b2083 | 3728 | bond_for_each_slave_rcu(bond, slave) { |
78a646ce NA |
3729 | if (bond_is_last_slave(bond, slave)) |
3730 | break; | |
3731 | if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { | |
3732 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1da177e4 | 3733 | |
78a646ce NA |
3734 | if (!skb2) { |
3735 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", | |
3736 | bond_dev->name); | |
3737 | continue; | |
1da177e4 | 3738 | } |
78a646ce NA |
3739 | /* bond_dev_queue_xmit always returns 0 */ |
3740 | bond_dev_queue_xmit(bond, skb2, slave->dev); | |
1da177e4 LT |
3741 | } |
3742 | } | |
78a646ce NA |
3743 | if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP) |
3744 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3745 | else | |
04502430 | 3746 | kfree_skb(skb); |
3d632c3f | 3747 | |
ec634fe3 | 3748 | return NETDEV_TX_OK; |
1da177e4 LT |
3749 | } |
3750 | ||
3751 | /*------------------------- Device initialization ---------------------------*/ | |
3752 | ||
6f6652be JV |
3753 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
3754 | { | |
3755 | switch (bond->params.xmit_policy) { | |
3756 | case BOND_XMIT_POLICY_LAYER23: | |
3757 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
3758 | break; | |
3759 | case BOND_XMIT_POLICY_LAYER34: | |
3760 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
3761 | break; | |
3762 | case BOND_XMIT_POLICY_LAYER2: | |
3763 | default: | |
3764 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
3765 | break; | |
3766 | } | |
3767 | } | |
3768 | ||
bb1d9123 AG |
3769 | /* |
3770 | * Lookup the slave that corresponds to a qid | |
3771 | */ | |
3772 | static inline int bond_slave_override(struct bonding *bond, | |
3773 | struct sk_buff *skb) | |
3774 | { | |
bb1d9123 AG |
3775 | struct slave *slave = NULL; |
3776 | struct slave *check_slave; | |
dec1e90e | 3777 | int res = 1; |
bb1d9123 | 3778 | |
0693e88e MM |
3779 | if (!skb->queue_mapping) |
3780 | return 1; | |
bb1d9123 AG |
3781 | |
3782 | /* Find out if any slaves have the same mapping as this skb. */ | |
278b2083 | 3783 | bond_for_each_slave_rcu(bond, check_slave) { |
bb1d9123 AG |
3784 | if (check_slave->queue_id == skb->queue_mapping) { |
3785 | slave = check_slave; | |
3786 | break; | |
3787 | } | |
3788 | } | |
3789 | ||
3790 | /* If the slave isn't UP, use default transmit policy. */ | |
3791 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
3792 | (slave->link == BOND_LINK_UP)) { | |
3793 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
3794 | } | |
3795 | ||
bb1d9123 AG |
3796 | return res; |
3797 | } | |
3798 | ||
374eeb5a | 3799 | |
bb1d9123 AG |
3800 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
3801 | { | |
3802 | /* | |
3803 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 3804 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3805 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3806 | * way down to the bonding driver. | |
3807 | */ | |
fd0e435b PO |
3808 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3809 | ||
374eeb5a NH |
3810 | /* |
3811 | * Save the original txq to restore before passing to the driver | |
3812 | */ | |
df4ab5b3 | 3813 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3814 | |
fd0e435b | 3815 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3816 | do { |
fd0e435b | 3817 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3818 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3819 | } |
3820 | return txq; | |
bb1d9123 AG |
3821 | } |
3822 | ||
0693e88e | 3823 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3824 | { |
bb1d9123 AG |
3825 | struct bonding *bond = netdev_priv(dev); |
3826 | ||
3827 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
3828 | if (!bond_slave_override(bond, skb)) | |
3829 | return NETDEV_TX_OK; | |
3830 | } | |
00829823 SH |
3831 | |
3832 | switch (bond->params.mode) { | |
3833 | case BOND_MODE_ROUNDROBIN: | |
3834 | return bond_xmit_roundrobin(skb, dev); | |
3835 | case BOND_MODE_ACTIVEBACKUP: | |
3836 | return bond_xmit_activebackup(skb, dev); | |
3837 | case BOND_MODE_XOR: | |
3838 | return bond_xmit_xor(skb, dev); | |
3839 | case BOND_MODE_BROADCAST: | |
3840 | return bond_xmit_broadcast(skb, dev); | |
3841 | case BOND_MODE_8023AD: | |
3842 | return bond_3ad_xmit_xor(skb, dev); | |
3843 | case BOND_MODE_ALB: | |
3844 | case BOND_MODE_TLB: | |
3845 | return bond_alb_xmit(skb, dev); | |
3846 | default: | |
3847 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3848 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3849 | dev->name, bond->params.mode); | |
00829823 | 3850 | WARN_ON_ONCE(1); |
04502430 | 3851 | kfree_skb(skb); |
00829823 SH |
3852 | return NETDEV_TX_OK; |
3853 | } | |
3854 | } | |
3855 | ||
0693e88e MM |
3856 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3857 | { | |
3858 | struct bonding *bond = netdev_priv(dev); | |
3859 | netdev_tx_t ret = NETDEV_TX_OK; | |
3860 | ||
3861 | /* | |
3862 | * If we risk deadlock from transmitting this in the | |
3863 | * netpoll path, tell netpoll to queue the frame for later tx | |
3864 | */ | |
3865 | if (is_netpoll_tx_blocked(dev)) | |
3866 | return NETDEV_TX_BUSY; | |
3867 | ||
278b2083 | 3868 | rcu_read_lock(); |
dec1e90e | 3869 | if (!list_empty(&bond->slave_list)) |
0693e88e MM |
3870 | ret = __bond_start_xmit(skb, dev); |
3871 | else | |
04502430 | 3872 | kfree_skb(skb); |
278b2083 | 3873 | rcu_read_unlock(); |
0693e88e MM |
3874 | |
3875 | return ret; | |
3876 | } | |
00829823 | 3877 | |
1da177e4 LT |
3878 | /* |
3879 | * set bond mode specific net device operations | |
3880 | */ | |
a77b5325 | 3881 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 3882 | { |
169a3e66 JV |
3883 | struct net_device *bond_dev = bond->dev; |
3884 | ||
1da177e4 LT |
3885 | switch (mode) { |
3886 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
3887 | break; |
3888 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
3889 | break; |
3890 | case BOND_MODE_XOR: | |
6f6652be | 3891 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
3892 | break; |
3893 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
3894 | break; |
3895 | case BOND_MODE_8023AD: | |
6f6652be | 3896 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 3897 | break; |
1da177e4 | 3898 | case BOND_MODE_ALB: |
8f903c70 JV |
3899 | /* FALLTHRU */ |
3900 | case BOND_MODE_TLB: | |
1da177e4 LT |
3901 | break; |
3902 | default: | |
3903 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3904 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3905 | bond_dev->name, mode); | |
1da177e4 LT |
3906 | break; |
3907 | } | |
3908 | } | |
3909 | ||
bb5b052f AG |
3910 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3911 | struct ethtool_cmd *ecmd) | |
3912 | { | |
3913 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3914 | unsigned long speed = 0; |
dec1e90e | 3915 | struct slave *slave; |
bb5b052f AG |
3916 | |
3917 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3918 | ecmd->port = PORT_OTHER; | |
3919 | ||
3920 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
3921 | * do not need to check mode. Though link speed might not represent | |
3922 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3923 | * this is an accurate maximum. | |
3924 | */ | |
3925 | read_lock(&bond->lock); | |
dec1e90e | 3926 | bond_for_each_slave(bond, slave) { |
bb5b052f AG |
3927 | if (SLAVE_IS_OK(slave)) { |
3928 | if (slave->speed != SPEED_UNKNOWN) | |
3929 | speed += slave->speed; | |
3930 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3931 | slave->duplex != DUPLEX_UNKNOWN) | |
3932 | ecmd->duplex = slave->duplex; | |
3933 | } | |
3934 | } | |
3935 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
3936 | read_unlock(&bond->lock); | |
dec1e90e | 3937 | |
bb5b052f AG |
3938 | return 0; |
3939 | } | |
3940 | ||
217df670 | 3941 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3942 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3943 | { |
7826d43f JP |
3944 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3945 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3946 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3947 | BOND_ABI_VERSION); | |
217df670 JV |
3948 | } |
3949 | ||
7282d491 | 3950 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3951 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3952 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3953 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3954 | }; |
3955 | ||
eb7cc59a | 3956 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3957 | .ndo_init = bond_init, |
9e71626c | 3958 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3959 | .ndo_open = bond_open, |
3960 | .ndo_stop = bond_close, | |
00829823 | 3961 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3962 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3963 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3964 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3965 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 3966 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 3967 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 3968 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 3969 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 3970 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 3971 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 3972 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 3973 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
3974 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
3975 | .ndo_poll_controller = bond_poll_controller, | |
3976 | #endif | |
9232ecca JP |
3977 | .ndo_add_slave = bond_enslave, |
3978 | .ndo_del_slave = bond_release, | |
b2a103e6 | 3979 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
3980 | }; |
3981 | ||
b3f92b63 DG |
3982 | static const struct device_type bond_type = { |
3983 | .name = "bond", | |
3984 | }; | |
3985 | ||
9e2e61fb AW |
3986 | static void bond_destructor(struct net_device *bond_dev) |
3987 | { | |
3988 | struct bonding *bond = netdev_priv(bond_dev); | |
3989 | if (bond->wq) | |
3990 | destroy_workqueue(bond->wq); | |
3991 | free_netdev(bond_dev); | |
3992 | } | |
3993 | ||
181470fc | 3994 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 3995 | { |
454d7c9b | 3996 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3997 | |
1da177e4 LT |
3998 | /* initialize rwlocks */ |
3999 | rwlock_init(&bond->lock); | |
4000 | rwlock_init(&bond->curr_slave_lock); | |
dec1e90e | 4001 | INIT_LIST_HEAD(&bond->slave_list); |
d2991f75 | 4002 | bond->params = bonding_defaults; |
1da177e4 LT |
4003 | |
4004 | /* Initialize pointers */ | |
1da177e4 | 4005 | bond->dev = bond_dev; |
1da177e4 LT |
4006 | |
4007 | /* Initialize the device entry points */ | |
181470fc | 4008 | ether_setup(bond_dev); |
eb7cc59a | 4009 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4010 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4011 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4012 | |
9e2e61fb | 4013 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4014 | |
b3f92b63 DG |
4015 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4016 | ||
1da177e4 LT |
4017 | /* Initialize the device options */ |
4018 | bond_dev->tx_queue_len = 0; | |
4019 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4020 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4021 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4022 | |
1da177e4 LT |
4023 | /* At first, we block adding VLANs. That's the only way to |
4024 | * prevent problems that occur when adding VLANs over an | |
4025 | * empty bond. The block will be removed once non-challenged | |
4026 | * slaves are enslaved. | |
4027 | */ | |
4028 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4029 | ||
932ff279 | 4030 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4031 | * transmitting */ |
4032 | bond_dev->features |= NETIF_F_LLTX; | |
4033 | ||
4034 | /* By default, we declare the bond to be fully | |
4035 | * VLAN hardware accelerated capable. Special | |
4036 | * care is taken in the various xmit functions | |
4037 | * when there are slaves that are not hw accel | |
4038 | * capable | |
4039 | */ | |
1da177e4 | 4040 | |
b2a103e6 | 4041 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4042 | NETIF_F_HW_VLAN_CTAG_TX | |
4043 | NETIF_F_HW_VLAN_CTAG_RX | | |
4044 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4045 | |
34324dc2 | 4046 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4047 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4048 | } |
4049 | ||
c67dfb29 EB |
4050 | /* |
4051 | * Destroy a bonding device. | |
4052 | * Must be under rtnl_lock when this function is called. | |
4053 | */ | |
4054 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4055 | { |
454d7c9b | 4056 | struct bonding *bond = netdev_priv(bond_dev); |
dec1e90e | 4057 | struct slave *slave, *tmp_slave; |
a434e43f | 4058 | |
f6dc31a8 WC |
4059 | bond_netpoll_cleanup(bond_dev); |
4060 | ||
c67dfb29 | 4061 | /* Release the bonded slaves */ |
dec1e90e | 4062 | list_for_each_entry_safe(slave, tmp_slave, &bond->slave_list, list) |
4063 | __bond_release_one(bond_dev, slave->dev, true); | |
0896341a | 4064 | pr_info("%s: released all slaves\n", bond_dev->name); |
c67dfb29 | 4065 | |
a434e43f JV |
4066 | list_del(&bond->bond_list); |
4067 | ||
f073c7ca | 4068 | bond_debug_unregister(bond); |
a434e43f JV |
4069 | } |
4070 | ||
1da177e4 LT |
4071 | /*------------------------- Module initialization ---------------------------*/ |
4072 | ||
4073 | /* | |
4074 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4075 | * number of the mode or its string name. A bit complicated because |
4076 | * some mode names are substrings of other names, and calls from sysfs | |
4077 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4078 | */ |
325dcf7a | 4079 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4080 | { |
54b87323 | 4081 | int modeint = -1, i, rv; |
a42e534f | 4082 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4083 | |
a42e534f JV |
4084 | for (p = (char *)buf; *p; p++) |
4085 | if (!(isdigit(*p) || isspace(*p))) | |
4086 | break; | |
4087 | ||
4088 | if (*p) | |
ece95f7f | 4089 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4090 | else |
54b87323 | 4091 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4092 | |
4093 | if (!rv) | |
4094 | return -1; | |
1da177e4 LT |
4095 | |
4096 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4097 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4098 | return tbl[i].mode; |
4099 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4100 | return tbl[i].mode; |
1da177e4 LT |
4101 | } |
4102 | ||
4103 | return -1; | |
4104 | } | |
4105 | ||
4106 | static int bond_check_params(struct bond_params *params) | |
4107 | { | |
5a5c5fd4 | 4108 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
8599b52e | 4109 | int arp_all_targets_value; |
f5b2b966 | 4110 | |
1da177e4 LT |
4111 | /* |
4112 | * Convert string parameters. | |
4113 | */ | |
4114 | if (mode) { | |
4115 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4116 | if (bond_mode == -1) { | |
a4aee5c8 | 4117 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4118 | mode == NULL ? "NULL" : mode); |
4119 | return -EINVAL; | |
4120 | } | |
4121 | } | |
4122 | ||
169a3e66 JV |
4123 | if (xmit_hash_policy) { |
4124 | if ((bond_mode != BOND_MODE_XOR) && | |
4125 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4126 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4127 | bond_mode_name(bond_mode)); |
4128 | } else { | |
4129 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4130 | xmit_hashtype_tbl); | |
4131 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4132 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4133 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4134 | xmit_hash_policy); |
4135 | return -EINVAL; | |
4136 | } | |
4137 | } | |
4138 | } | |
4139 | ||
1da177e4 LT |
4140 | if (lacp_rate) { |
4141 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4142 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4143 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4144 | } else { |
4145 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4146 | if (lacp_fast == -1) { | |
a4aee5c8 | 4147 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4148 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4149 | return -EINVAL; | |
4150 | } | |
4151 | } | |
4152 | } | |
4153 | ||
fd989c83 JV |
4154 | if (ad_select) { |
4155 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4156 | if (params->ad_select == -1) { | |
a4aee5c8 | 4157 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4158 | ad_select == NULL ? "NULL" : ad_select); |
4159 | return -EINVAL; | |
4160 | } | |
4161 | ||
4162 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4163 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4164 | } |
4165 | } else { | |
4166 | params->ad_select = BOND_AD_STABLE; | |
4167 | } | |
4168 | ||
f5841306 | 4169 | if (max_bonds < 0) { |
a4aee5c8 JP |
4170 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4171 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4172 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4173 | } | |
4174 | ||
4175 | if (miimon < 0) { | |
a4aee5c8 JP |
4176 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4177 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4178 | miimon = BOND_LINK_MON_INTERV; |
4179 | } | |
4180 | ||
4181 | if (updelay < 0) { | |
a4aee5c8 JP |
4182 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4183 | updelay, INT_MAX); | |
1da177e4 LT |
4184 | updelay = 0; |
4185 | } | |
4186 | ||
4187 | if (downdelay < 0) { | |
a4aee5c8 JP |
4188 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4189 | downdelay, INT_MAX); | |
1da177e4 LT |
4190 | downdelay = 0; |
4191 | } | |
4192 | ||
4193 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4194 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4195 | use_carrier); | |
1da177e4 LT |
4196 | use_carrier = 1; |
4197 | } | |
4198 | ||
ad246c99 BH |
4199 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4200 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4201 | num_peer_notif); | |
4202 | num_peer_notif = 1; | |
4203 | } | |
4204 | ||
1da177e4 LT |
4205 | /* reset values for 802.3ad */ |
4206 | if (bond_mode == BOND_MODE_8023AD) { | |
4207 | if (!miimon) { | |
a4aee5c8 | 4208 | 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 | 4209 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4210 | miimon = 100; |
4211 | } | |
4212 | } | |
4213 | ||
bb1d9123 AG |
4214 | if (tx_queues < 1 || tx_queues > 255) { |
4215 | pr_warning("Warning: tx_queues (%d) should be between " | |
4216 | "1 and 255, resetting to %d\n", | |
4217 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4218 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4219 | } | |
4220 | ||
ebd8e497 AG |
4221 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4222 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4223 | "not of valid value (0/1), so it was set to " | |
4224 | "0\n", all_slaves_active); | |
4225 | all_slaves_active = 0; | |
4226 | } | |
4227 | ||
c2952c31 FL |
4228 | if (resend_igmp < 0 || resend_igmp > 255) { |
4229 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4230 | "0 and 255, resetting to %d\n", | |
4231 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4232 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4233 | } | |
4234 | ||
1da177e4 LT |
4235 | /* reset values for TLB/ALB */ |
4236 | if ((bond_mode == BOND_MODE_TLB) || | |
4237 | (bond_mode == BOND_MODE_ALB)) { | |
4238 | if (!miimon) { | |
a4aee5c8 | 4239 | 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 | 4240 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4241 | miimon = 100; |
4242 | } | |
4243 | } | |
4244 | ||
4245 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4246 | 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", |
4247 | updelay); | |
1da177e4 LT |
4248 | } |
4249 | ||
4250 | if (!miimon) { | |
4251 | if (updelay || downdelay) { | |
4252 | /* just warn the user the up/down delay will have | |
4253 | * no effect since miimon is zero... | |
4254 | */ | |
a4aee5c8 JP |
4255 | 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", |
4256 | updelay, downdelay); | |
1da177e4 LT |
4257 | } |
4258 | } else { | |
4259 | /* don't allow arp monitoring */ | |
4260 | if (arp_interval) { | |
a4aee5c8 JP |
4261 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4262 | miimon, arp_interval); | |
1da177e4 LT |
4263 | arp_interval = 0; |
4264 | } | |
4265 | ||
4266 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4267 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4268 | updelay, miimon, | |
4269 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4270 | } |
4271 | ||
4272 | updelay /= miimon; | |
4273 | ||
4274 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4275 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4276 | downdelay, miimon, | |
4277 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4278 | } |
4279 | ||
4280 | downdelay /= miimon; | |
4281 | } | |
4282 | ||
4283 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4284 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4285 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4286 | arp_interval = BOND_LINK_ARP_INTERV; |
4287 | } | |
4288 | ||
5a5c5fd4 | 4289 | for (arp_ip_count = 0, i = 0; |
4290 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4291 | /* not complete check, but should be good enough to |
4292 | catch mistakes */ | |
5a5c5fd4 | 4293 | __be32 ip = in_aton(arp_ip_target[i]); |
4294 | if (!isdigit(arp_ip_target[i][0]) || ip == 0 || | |
4295 | ip == htonl(INADDR_BROADCAST)) { | |
a4aee5c8 | 4296 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5a5c5fd4 | 4297 | arp_ip_target[i]); |
1da177e4 LT |
4298 | arp_interval = 0; |
4299 | } else { | |
0afee4e8 VF |
4300 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4301 | arp_target[arp_ip_count++] = ip; | |
4302 | else | |
4303 | pr_warning("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", | |
4304 | &ip); | |
1da177e4 LT |
4305 | } |
4306 | } | |
4307 | ||
4308 | if (arp_interval && !arp_ip_count) { | |
4309 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4310 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4311 | arp_interval); | |
1da177e4 LT |
4312 | arp_interval = 0; |
4313 | } | |
4314 | ||
f5b2b966 JV |
4315 | if (arp_validate) { |
4316 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4317 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4318 | return -EINVAL; |
4319 | } | |
4320 | if (!arp_interval) { | |
a4aee5c8 | 4321 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4322 | return -EINVAL; |
4323 | } | |
4324 | ||
4325 | arp_validate_value = bond_parse_parm(arp_validate, | |
4326 | arp_validate_tbl); | |
4327 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4328 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4329 | arp_validate == NULL ? "NULL" : arp_validate); |
4330 | return -EINVAL; | |
4331 | } | |
4332 | } else | |
4333 | arp_validate_value = 0; | |
4334 | ||
8599b52e VF |
4335 | arp_all_targets_value = 0; |
4336 | if (arp_all_targets) { | |
4337 | arp_all_targets_value = bond_parse_parm(arp_all_targets, | |
4338 | arp_all_targets_tbl); | |
4339 | ||
4340 | if (arp_all_targets_value == -1) { | |
4341 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", | |
4342 | arp_all_targets); | |
4343 | arp_all_targets_value = 0; | |
4344 | } | |
4345 | } | |
4346 | ||
1da177e4 | 4347 | if (miimon) { |
a4aee5c8 | 4348 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4349 | } else if (arp_interval) { |
a4aee5c8 JP |
4350 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4351 | arp_interval, | |
4352 | arp_validate_tbl[arp_validate_value].modename, | |
4353 | arp_ip_count); | |
1da177e4 LT |
4354 | |
4355 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4356 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4357 | |
e5e2a8fd | 4358 | pr_info("\n"); |
1da177e4 | 4359 | |
b8a9787e | 4360 | } else if (max_bonds) { |
1da177e4 LT |
4361 | /* miimon and arp_interval not set, we need one so things |
4362 | * work as expected, see bonding.txt for details | |
4363 | */ | |
b2730f4f | 4364 | 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 |
4365 | } |
4366 | ||
4367 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4368 | /* currently, using a primary only makes sense | |
4369 | * in active backup, TLB or ALB modes | |
4370 | */ | |
a4aee5c8 JP |
4371 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4372 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4373 | primary = NULL; |
4374 | } | |
4375 | ||
a549952a JP |
4376 | if (primary && primary_reselect) { |
4377 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4378 | pri_reselect_tbl); | |
4379 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4380 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4381 | primary_reselect == |
4382 | NULL ? "NULL" : primary_reselect); | |
4383 | return -EINVAL; | |
4384 | } | |
4385 | } else { | |
4386 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4387 | } | |
4388 | ||
3915c1e8 JV |
4389 | if (fail_over_mac) { |
4390 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4391 | fail_over_mac_tbl); | |
4392 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4393 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4394 | arp_validate == NULL ? "NULL" : arp_validate); |
4395 | return -EINVAL; | |
4396 | } | |
4397 | ||
4398 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4399 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4400 | } else { |
4401 | fail_over_mac_value = BOND_FOM_NONE; | |
4402 | } | |
dd957c57 | 4403 | |
1da177e4 LT |
4404 | /* fill params struct with the proper values */ |
4405 | params->mode = bond_mode; | |
169a3e66 | 4406 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4407 | params->miimon = miimon; |
ad246c99 | 4408 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4409 | params->arp_interval = arp_interval; |
f5b2b966 | 4410 | params->arp_validate = arp_validate_value; |
8599b52e | 4411 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4412 | params->updelay = updelay; |
4413 | params->downdelay = downdelay; | |
4414 | params->use_carrier = use_carrier; | |
4415 | params->lacp_fast = lacp_fast; | |
4416 | params->primary[0] = 0; | |
a549952a | 4417 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4418 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4419 | params->tx_queues = tx_queues; |
ebd8e497 | 4420 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4421 | params->resend_igmp = resend_igmp; |
655f8919 | 4422 | params->min_links = min_links; |
7eacd038 | 4423 | params->lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
4424 | |
4425 | if (primary) { | |
4426 | strncpy(params->primary, primary, IFNAMSIZ); | |
4427 | params->primary[IFNAMSIZ - 1] = 0; | |
4428 | } | |
4429 | ||
4430 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4431 | ||
4432 | return 0; | |
4433 | } | |
4434 | ||
0daa2303 | 4435 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4436 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4437 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4438 | |
e8a0464c DM |
4439 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4440 | struct netdev_queue *txq, | |
4441 | void *_unused) | |
c773e847 DM |
4442 | { |
4443 | lockdep_set_class(&txq->_xmit_lock, | |
4444 | &bonding_netdev_xmit_lock_key); | |
4445 | } | |
4446 | ||
4447 | static void bond_set_lockdep_class(struct net_device *dev) | |
4448 | { | |
cf508b12 DM |
4449 | lockdep_set_class(&dev->addr_list_lock, |
4450 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4451 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4452 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4453 | } |
4454 | ||
181470fc SH |
4455 | /* |
4456 | * Called from registration process | |
4457 | */ | |
4458 | static int bond_init(struct net_device *bond_dev) | |
4459 | { | |
4460 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4461 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4462 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4463 | |
4464 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4465 | ||
9fe0617d NH |
4466 | /* |
4467 | * Initialize locks that may be required during | |
4468 | * en/deslave operations. All of the bond_open work | |
4469 | * (of which this is part) should really be moved to | |
4470 | * a phase prior to dev_open | |
4471 | */ | |
4472 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4473 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4474 | ||
181470fc SH |
4475 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4476 | if (!bond->wq) | |
4477 | return -ENOMEM; | |
4478 | ||
4479 | bond_set_lockdep_class(bond_dev); | |
4480 | ||
ec87fd3b | 4481 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4482 | |
6151b3d4 | 4483 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4484 | |
f073c7ca TI |
4485 | bond_debug_register(bond); |
4486 | ||
409cc1f8 JP |
4487 | /* Ensure valid dev_addr */ |
4488 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4489 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4490 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4491 | |
181470fc SH |
4492 | return 0; |
4493 | } | |
4494 | ||
88ead977 EB |
4495 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4496 | { | |
4497 | if (tb[IFLA_ADDRESS]) { | |
4498 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4499 | return -EINVAL; | |
4500 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4501 | return -EADDRNOTAVAIL; | |
4502 | } | |
4503 | return 0; | |
4504 | } | |
4505 | ||
d40156aa | 4506 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4507 | { |
efacb309 | 4508 | return tx_queues; |
d5da4510 JP |
4509 | } |
4510 | ||
88ead977 | 4511 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4512 | .kind = "bond", |
4513 | .priv_size = sizeof(struct bonding), | |
4514 | .setup = bond_setup, | |
4515 | .validate = bond_validate, | |
4516 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4517 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4518 | as for TX queues */ | |
88ead977 EB |
4519 | }; |
4520 | ||
dfe60397 | 4521 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4522 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4523 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4524 | * set up our sysfs entries. | |
4525 | */ | |
ec87fd3b | 4526 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4527 | { |
4528 | struct net_device *bond_dev; | |
4529 | int res; | |
4530 | ||
4531 | rtnl_lock(); | |
027ea041 | 4532 | |
1c5cae81 JP |
4533 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4534 | name ? name : "bond%d", | |
4535 | bond_setup, tx_queues); | |
dfe60397 | 4536 | if (!bond_dev) { |
a4aee5c8 | 4537 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4538 | rtnl_unlock(); |
4539 | return -ENOMEM; | |
dfe60397 MW |
4540 | } |
4541 | ||
ec87fd3b | 4542 | dev_net_set(bond_dev, net); |
88ead977 EB |
4543 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4544 | ||
dfe60397 | 4545 | res = register_netdevice(bond_dev); |
0daa2303 | 4546 | |
e826eafa PO |
4547 | netif_carrier_off(bond_dev); |
4548 | ||
7e083840 | 4549 | rtnl_unlock(); |
9e2e61fb AW |
4550 | if (res < 0) |
4551 | bond_destructor(bond_dev); | |
30c15ba9 | 4552 | return res; |
dfe60397 MW |
4553 | } |
4554 | ||
2c8c1e72 | 4555 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4556 | { |
15449745 | 4557 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4558 | |
4559 | bn->net = net; | |
4560 | INIT_LIST_HEAD(&bn->dev_list); | |
4561 | ||
ec87fd3b | 4562 | bond_create_proc_dir(bn); |
4c22400a | 4563 | bond_create_sysfs(bn); |
87a7b84b | 4564 | |
15449745 | 4565 | return 0; |
ec87fd3b EB |
4566 | } |
4567 | ||
2c8c1e72 | 4568 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4569 | { |
15449745 | 4570 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4571 | struct bonding *bond, *tmp_bond; |
4572 | LIST_HEAD(list); | |
ec87fd3b | 4573 | |
4c22400a | 4574 | bond_destroy_sysfs(bn); |
ec87fd3b | 4575 | bond_destroy_proc_dir(bn); |
69b0216a | 4576 | |
4577 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4578 | rtnl_lock(); | |
4579 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4580 | unregister_netdevice_queue(bond->dev, &list); | |
4581 | unregister_netdevice_many(&list); | |
4582 | rtnl_unlock(); | |
ec87fd3b EB |
4583 | } |
4584 | ||
4585 | static struct pernet_operations bond_net_ops = { | |
4586 | .init = bond_net_init, | |
4587 | .exit = bond_net_exit, | |
15449745 EB |
4588 | .id = &bond_net_id, |
4589 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4590 | }; |
4591 | ||
1da177e4 LT |
4592 | static int __init bonding_init(void) |
4593 | { | |
1da177e4 LT |
4594 | int i; |
4595 | int res; | |
4596 | ||
bd33acc3 | 4597 | pr_info("%s", bond_version); |
1da177e4 | 4598 | |
dfe60397 | 4599 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4600 | if (res) |
dfe60397 | 4601 | goto out; |
1da177e4 | 4602 | |
15449745 | 4603 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4604 | if (res) |
4605 | goto out; | |
027ea041 | 4606 | |
88ead977 EB |
4607 | res = rtnl_link_register(&bond_link_ops); |
4608 | if (res) | |
6639104b | 4609 | goto err_link; |
88ead977 | 4610 | |
f073c7ca TI |
4611 | bond_create_debugfs(); |
4612 | ||
1da177e4 | 4613 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4614 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4615 | if (res) |
4616 | goto err; | |
1da177e4 LT |
4617 | } |
4618 | ||
1da177e4 | 4619 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4620 | out: |
1da177e4 | 4621 | return res; |
88ead977 EB |
4622 | err: |
4623 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 4624 | err_link: |
15449745 | 4625 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4626 | goto out; |
dfe60397 | 4627 | |
1da177e4 LT |
4628 | } |
4629 | ||
4630 | static void __exit bonding_exit(void) | |
4631 | { | |
4632 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4633 | ||
f073c7ca | 4634 | bond_destroy_debugfs(); |
ae68c398 | 4635 | |
4de79c73 | 4636 | rtnl_link_unregister(&bond_link_ops); |
ffcdedb6 | 4637 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4638 | |
4639 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4640 | /* |
4641 | * Make sure we don't have an imbalance on our netpoll blocking | |
4642 | */ | |
4643 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4644 | #endif |
1da177e4 LT |
4645 | } |
4646 | ||
4647 | module_init(bonding_init); | |
4648 | module_exit(bonding_exit); | |
4649 | MODULE_LICENSE("GPL"); | |
4650 | MODULE_VERSION(DRV_VERSION); | |
4651 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4652 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 4653 | MODULE_ALIAS_RTNL_LINK("bond"); |