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 | ||
1da177e4 LT |
34 | #include <linux/kernel.h> |
35 | #include <linux/module.h> | |
1da177e4 LT |
36 | #include <linux/types.h> |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/interrupt.h> | |
39 | #include <linux/ptrace.h> | |
40 | #include <linux/ioport.h> | |
41 | #include <linux/in.h> | |
169a3e66 | 42 | #include <net/ip.h> |
1da177e4 | 43 | #include <linux/ip.h> |
169a3e66 JV |
44 | #include <linux/tcp.h> |
45 | #include <linux/udp.h> | |
1da177e4 LT |
46 | #include <linux/slab.h> |
47 | #include <linux/string.h> | |
48 | #include <linux/init.h> | |
49 | #include <linux/timer.h> | |
50 | #include <linux/socket.h> | |
51 | #include <linux/ctype.h> | |
52 | #include <linux/inet.h> | |
53 | #include <linux/bitops.h> | |
3d632c3f | 54 | #include <linux/io.h> |
1da177e4 | 55 | #include <asm/dma.h> |
3d632c3f | 56 | #include <linux/uaccess.h> |
1da177e4 LT |
57 | #include <linux/errno.h> |
58 | #include <linux/netdevice.h> | |
59 | #include <linux/inetdevice.h> | |
a816c7c7 | 60 | #include <linux/igmp.h> |
1da177e4 LT |
61 | #include <linux/etherdevice.h> |
62 | #include <linux/skbuff.h> | |
63 | #include <net/sock.h> | |
64 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
65 | #include <linux/smp.h> |
66 | #include <linux/if_ether.h> | |
67 | #include <net/arp.h> | |
68 | #include <linux/mii.h> | |
69 | #include <linux/ethtool.h> | |
70 | #include <linux/if_vlan.h> | |
71 | #include <linux/if_bonding.h> | |
b63bb739 | 72 | #include <linux/jiffies.h> |
e843fa50 | 73 | #include <linux/preempt.h> |
c3ade5ca | 74 | #include <net/route.h> |
457c4cbc | 75 | #include <net/net_namespace.h> |
ec87fd3b | 76 | #include <net/netns/generic.h> |
5ee31c68 | 77 | #include <net/pkt_sched.h> |
278b2083 | 78 | #include <linux/rculist.h> |
1bd758eb | 79 | #include <net/flow_dissector.h> |
c158cba3 | 80 | #include <net/switchdev.h> |
1ef8019b DM |
81 | #include <net/bonding.h> |
82 | #include <net/bond_3ad.h> | |
83 | #include <net/bond_alb.h> | |
1da177e4 | 84 | |
73b5a6f2 MB |
85 | #include "bonding_priv.h" |
86 | ||
1da177e4 LT |
87 | /*---------------------------- Module parameters ----------------------------*/ |
88 | ||
89 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
1da177e4 LT |
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; |
b98d9c66 | 94 | static int miimon; |
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; | |
7bdb04ed | 105 | static int arp_interval; |
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; |
73958329 | 113 | static int packets_per_slave = 1; |
3a7129e5 | 114 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
115 | |
116 | module_param(max_bonds, int, 0); | |
117 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
118 | module_param(tx_queues, int, 0); |
119 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 120 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
121 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
122 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 123 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
124 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
125 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
126 | module_param(miimon, int, 0); |
127 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
128 | module_param(updelay, int, 0); | |
129 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
130 | module_param(downdelay, int, 0); | |
2ac47660 MW |
131 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
132 | "in milliseconds"); | |
1da177e4 | 133 | module_param(use_carrier, int, 0); |
2ac47660 MW |
134 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
135 | "0 for off, 1 for on (default)"); | |
1da177e4 | 136 | module_param(mode, charp, 0); |
90e62474 | 137 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
138 | "1 for active-backup, 2 for balance-xor, " |
139 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
140 | "6 for balance-alb"); | |
1da177e4 LT |
141 | module_param(primary, charp, 0); |
142 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
143 | module_param(primary_reselect, charp, 0); |
144 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
145 | "once it comes up; " | |
146 | "0 for always (default), " | |
147 | "1 for only if speed of primary is " | |
148 | "better, " | |
149 | "2 for only on active slave " | |
150 | "failure"); | |
1da177e4 | 151 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
152 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
153 | "0 for slow, 1 for fast"); | |
fd989c83 | 154 | module_param(ad_select, charp, 0); |
90e62474 AG |
155 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
156 | "0 for stable (default), 1 for bandwidth, " | |
157 | "2 for count"); | |
655f8919 | 158 | module_param(min_links, int, 0); |
159 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
160 | ||
169a3e66 | 161 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
162 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
163 | "0 for layer 2 (default), 1 for layer 3+4, " | |
32819dc1 NA |
164 | "2 for layer 2+3, 3 for encap layer 2+3, " |
165 | "4 for encap layer 3+4"); | |
1da177e4 LT |
166 | module_param(arp_interval, int, 0); |
167 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
168 | module_param_array(arp_ip_target, charp, NULL, 0); | |
169 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 170 | module_param(arp_validate, charp, 0); |
90e62474 AG |
171 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
172 | "0 for none (default), 1 for active, " | |
173 | "2 for backup, 3 for all"); | |
8599b52e VF |
174 | module_param(arp_all_targets, charp, 0); |
175 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 176 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
177 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
178 | "the same MAC; 0 for none (default), " | |
179 | "1 for active, 2 for follow"); | |
ebd8e497 | 180 | module_param(all_slaves_active, int, 0); |
37b7021d | 181 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface " |
90e62474 | 182 | "by setting active flag for all slaves; " |
ebd8e497 | 183 | "0 for never (default), 1 for always."); |
c2952c31 | 184 | module_param(resend_igmp, int, 0); |
90e62474 AG |
185 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
186 | "link failure"); | |
73958329 NA |
187 | module_param(packets_per_slave, int, 0); |
188 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
189 | "mode; 0 for a random slave, 1 packet per " | |
190 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 191 | module_param(lp_interval, uint, 0); |
192 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
193 | "the bonding driver sends learning packets to " | |
194 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
195 | |
196 | /*----------------------------- Global variables ----------------------------*/ | |
197 | ||
e843fa50 | 198 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 199 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
200 | #endif |
201 | ||
f99189b1 | 202 | int bond_net_id __read_mostly; |
1da177e4 | 203 | |
3d632c3f SH |
204 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
205 | static int arp_ip_count; | |
1da177e4 | 206 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
207 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
208 | static int lacp_fast; | |
1da177e4 | 209 | |
1da177e4 LT |
210 | /*-------------------------- Forward declarations ---------------------------*/ |
211 | ||
181470fc | 212 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 213 | static void bond_uninit(struct net_device *bond_dev); |
5f0c5f73 AG |
214 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
215 | struct rtnl_link_stats64 *stats); | |
ee637714 | 216 | static void bond_slave_arr_handler(struct work_struct *work); |
21a75f09 JV |
217 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, |
218 | int mod); | |
1da177e4 LT |
219 | |
220 | /*---------------------------- General routines -----------------------------*/ | |
221 | ||
bd33acc3 | 222 | const char *bond_mode_name(int mode) |
1da177e4 | 223 | { |
77afc92b HE |
224 | static const char *names[] = { |
225 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
226 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
227 | [BOND_MODE_XOR] = "load balancing (xor)", | |
228 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 229 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
230 | [BOND_MODE_TLB] = "transmit load balancing", |
231 | [BOND_MODE_ALB] = "adaptive load balancing", | |
232 | }; | |
233 | ||
f5280948 | 234 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 235 | return "unknown"; |
77afc92b HE |
236 | |
237 | return names[mode]; | |
1da177e4 LT |
238 | } |
239 | ||
240 | /*---------------------------------- VLAN -----------------------------------*/ | |
241 | ||
1da177e4 LT |
242 | /** |
243 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 244 | * |
1da177e4 LT |
245 | * @bond: bond device that got this skb for tx. |
246 | * @skb: hw accel VLAN tagged skb to transmit | |
247 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 248 | */ |
d316dedd | 249 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 250 | struct net_device *slave_dev) |
1da177e4 | 251 | { |
8387451e | 252 | skb->dev = slave_dev; |
374eeb5a | 253 | |
5ee31c68 | 254 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
255 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
256 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 257 | |
e15c3c22 | 258 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 259 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 260 | else |
f6dc31a8 | 261 | dev_queue_xmit(skb); |
1da177e4 LT |
262 | } |
263 | ||
547942ca | 264 | /* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
cc0e4070 | 265 | * We don't protect the slave list iteration with a lock because: |
1da177e4 LT |
266 | * a. This operation is performed in IOCTL context, |
267 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
268 | * c. Holding a lock with BH disabled while directly calling a base driver | |
269 | * entry point is generally a BAD idea. | |
3d632c3f | 270 | * |
1da177e4 LT |
271 | * The design of synchronization/protection for this operation in the 8021q |
272 | * module is good for one or more VLAN devices over a single physical device | |
273 | * and cannot be extended for a teaming solution like bonding, so there is a | |
274 | * potential race condition here where a net device from the vlan group might | |
275 | * be referenced (either by a base driver or the 8021q code) while it is being | |
276 | * removed from the system. However, it turns out we're not making matters | |
277 | * worse, and if it works for regular VLAN usage it will work here too. | |
278 | */ | |
279 | ||
1da177e4 LT |
280 | /** |
281 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
282 | * @bond_dev: bonding net device that got called | |
283 | * @vid: vlan id being added | |
284 | */ | |
80d5c368 PM |
285 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
286 | __be16 proto, u16 vid) | |
1da177e4 | 287 | { |
454d7c9b | 288 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 289 | struct slave *slave, *rollback_slave; |
9caff1e7 | 290 | struct list_head *iter; |
dec1e90e | 291 | int res; |
1da177e4 | 292 | |
9caff1e7 | 293 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 294 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
295 | if (res) |
296 | goto unwind; | |
1da177e4 LT |
297 | } |
298 | ||
8e586137 | 299 | return 0; |
87002b03 JP |
300 | |
301 | unwind: | |
81f23b13 | 302 | /* unwind to the slave that failed */ |
9caff1e7 | 303 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
304 | if (rollback_slave == slave) |
305 | break; | |
306 | ||
307 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
308 | } | |
87002b03 JP |
309 | |
310 | return res; | |
1da177e4 LT |
311 | } |
312 | ||
313 | /** | |
314 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
315 | * @bond_dev: bonding net device that got called | |
316 | * @vid: vlan id being removed | |
317 | */ | |
80d5c368 PM |
318 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
319 | __be16 proto, u16 vid) | |
1da177e4 | 320 | { |
454d7c9b | 321 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 322 | struct list_head *iter; |
1da177e4 | 323 | struct slave *slave; |
1da177e4 | 324 | |
9caff1e7 | 325 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 326 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 327 | |
e868b0c9 VF |
328 | if (bond_is_lb(bond)) |
329 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
330 | |
331 | return 0; | |
1da177e4 LT |
332 | } |
333 | ||
1da177e4 LT |
334 | /*------------------------------- Link status -------------------------------*/ |
335 | ||
547942ca | 336 | /* Set the carrier state for the master according to the state of its |
ff59c456 JV |
337 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, |
338 | * do special 802.3ad magic. | |
339 | * | |
340 | * Returns zero if carrier state does not change, nonzero if it does. | |
341 | */ | |
2477bc9a | 342 | int bond_set_carrier(struct bonding *bond) |
ff59c456 | 343 | { |
9caff1e7 | 344 | struct list_head *iter; |
ff59c456 | 345 | struct slave *slave; |
ff59c456 | 346 | |
0965a1f3 | 347 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
348 | goto down; |
349 | ||
01844098 | 350 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
ff59c456 JV |
351 | return bond_3ad_set_carrier(bond); |
352 | ||
9caff1e7 | 353 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
354 | if (slave->link == BOND_LINK_UP) { |
355 | if (!netif_carrier_ok(bond->dev)) { | |
356 | netif_carrier_on(bond->dev); | |
357 | return 1; | |
358 | } | |
359 | return 0; | |
360 | } | |
361 | } | |
362 | ||
363 | down: | |
364 | if (netif_carrier_ok(bond->dev)) { | |
365 | netif_carrier_off(bond->dev); | |
366 | return 1; | |
367 | } | |
368 | return 0; | |
369 | } | |
370 | ||
547942ca | 371 | /* Get link speed and duplex from the slave's base driver |
1da177e4 | 372 | * using ethtool. If for some reason the call fails or the |
98f41f69 | 373 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 374 | * and return. |
1da177e4 | 375 | */ |
13b95fb7 | 376 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
377 | { |
378 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 379 | struct ethtool_cmd ecmd; |
5d30530e | 380 | u32 slave_speed; |
61a44b9c | 381 | int res; |
1da177e4 | 382 | |
589665f5 DC |
383 | slave->speed = SPEED_UNKNOWN; |
384 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 385 | |
4bc71cb9 | 386 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 387 | if (res < 0) |
13b95fb7 | 388 | return; |
1da177e4 | 389 | |
4bc71cb9 | 390 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 391 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 392 | return; |
1da177e4 | 393 | |
4bc71cb9 | 394 | switch (ecmd.duplex) { |
1da177e4 LT |
395 | case DUPLEX_FULL: |
396 | case DUPLEX_HALF: | |
397 | break; | |
398 | default: | |
13b95fb7 | 399 | return; |
1da177e4 LT |
400 | } |
401 | ||
5d30530e | 402 | slave->speed = slave_speed; |
4bc71cb9 | 403 | slave->duplex = ecmd.duplex; |
1da177e4 | 404 | |
13b95fb7 | 405 | return; |
1da177e4 LT |
406 | } |
407 | ||
07699f9a | 408 | const char *bond_slave_link_status(s8 link) |
409 | { | |
410 | switch (link) { | |
411 | case BOND_LINK_UP: | |
412 | return "up"; | |
413 | case BOND_LINK_FAIL: | |
414 | return "going down"; | |
415 | case BOND_LINK_DOWN: | |
416 | return "down"; | |
417 | case BOND_LINK_BACK: | |
418 | return "going back"; | |
419 | default: | |
420 | return "unknown"; | |
421 | } | |
422 | } | |
423 | ||
547942ca | 424 | /* if <dev> supports MII link status reporting, check its link status. |
1da177e4 LT |
425 | * |
426 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 427 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
428 | * |
429 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
430 | * can't tell and just pretend it is), or 0, meaning that the link is | |
431 | * down. | |
432 | * | |
433 | * If reporting is non-zero, instead of faking link up, return -1 if | |
434 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
435 | * support them). If use_carrier is set, return whatever it says. | |
436 | * It'd be nice if there was a good way to tell if a driver supports | |
437 | * netif_carrier, but there really isn't. | |
438 | */ | |
3d632c3f SH |
439 | static int bond_check_dev_link(struct bonding *bond, |
440 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 441 | { |
eb7cc59a | 442 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 443 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
444 | struct ifreq ifr; |
445 | struct mii_ioctl_data *mii; | |
1da177e4 | 446 | |
6c988853 PG |
447 | if (!reporting && !netif_running(slave_dev)) |
448 | return 0; | |
449 | ||
eb7cc59a | 450 | if (bond->params.use_carrier) |
1da177e4 | 451 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 452 | |
29112f4e | 453 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
454 | if (slave_dev->ethtool_ops->get_link) |
455 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
456 | BMSR_LSTATUS : 0; | |
29112f4e | 457 | |
3d632c3f | 458 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 459 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 | 460 | if (ioctl) { |
547942ca NA |
461 | /* TODO: set pointer to correct ioctl on a per team member |
462 | * bases to make this more efficient. that is, once | |
463 | * we determine the correct ioctl, we will always | |
464 | * call it and not the others for that team | |
465 | * member. | |
466 | */ | |
467 | ||
468 | /* We cannot assume that SIOCGMIIPHY will also read a | |
1da177e4 LT |
469 | * register; not all network drivers (e.g., e100) |
470 | * support that. | |
471 | */ | |
472 | ||
473 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
474 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
475 | mii = if_mii(&ifr); | |
476 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
477 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
478 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
479 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
480 | } |
481 | } | |
482 | ||
547942ca | 483 | /* If reporting, report that either there's no dev->do_ioctl, |
61a44b9c | 484 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
485 | * cannot report link status). If not reporting, pretend |
486 | * we're ok. | |
487 | */ | |
3d632c3f | 488 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
489 | } |
490 | ||
491 | /*----------------------------- Multicast list ------------------------------*/ | |
492 | ||
547942ca | 493 | /* Push the promiscuity flag down to appropriate slaves */ |
7e1a1ac1 | 494 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 495 | { |
9caff1e7 | 496 | struct list_head *iter; |
7e1a1ac1 | 497 | int err = 0; |
9caff1e7 | 498 | |
ec0865a9 | 499 | if (bond_uses_primary(bond)) { |
14056e79 | 500 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 501 | |
4740d638 ED |
502 | if (curr_active) |
503 | err = dev_set_promiscuity(curr_active->dev, inc); | |
1da177e4 LT |
504 | } else { |
505 | struct slave *slave; | |
dec1e90e | 506 | |
9caff1e7 | 507 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
508 | err = dev_set_promiscuity(slave->dev, inc); |
509 | if (err) | |
510 | return err; | |
1da177e4 LT |
511 | } |
512 | } | |
7e1a1ac1 | 513 | return err; |
1da177e4 LT |
514 | } |
515 | ||
547942ca | 516 | /* Push the allmulti flag down to all slaves */ |
7e1a1ac1 | 517 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 518 | { |
9caff1e7 | 519 | struct list_head *iter; |
7e1a1ac1 | 520 | int err = 0; |
9caff1e7 | 521 | |
ec0865a9 | 522 | if (bond_uses_primary(bond)) { |
14056e79 | 523 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 524 | |
4740d638 ED |
525 | if (curr_active) |
526 | err = dev_set_allmulti(curr_active->dev, inc); | |
1da177e4 LT |
527 | } else { |
528 | struct slave *slave; | |
dec1e90e | 529 | |
9caff1e7 | 530 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
531 | err = dev_set_allmulti(slave->dev, inc); |
532 | if (err) | |
533 | return err; | |
1da177e4 LT |
534 | } |
535 | } | |
7e1a1ac1 | 536 | return err; |
1da177e4 LT |
537 | } |
538 | ||
547942ca | 539 | /* Retrieve the list of registered multicast addresses for the bonding |
5a37e8ca FL |
540 | * device and retransmit an IGMP JOIN request to the current active |
541 | * slave. | |
542 | */ | |
f2369109 | 543 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 544 | { |
f2369109 | 545 | struct bonding *bond = container_of(work, struct bonding, |
546 | mcast_work.work); | |
547 | ||
4aa5dee4 | 548 | if (!rtnl_trylock()) { |
4beac029 | 549 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 550 | return; |
5a37e8ca | 551 | } |
4aa5dee4 | 552 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 553 | |
4f5474e7 NA |
554 | if (bond->igmp_retrans > 1) { |
555 | bond->igmp_retrans--; | |
c2952c31 | 556 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 557 | } |
f2369109 | 558 | rtnl_unlock(); |
5a37e8ca FL |
559 | } |
560 | ||
547942ca | 561 | /* Flush bond's hardware addresses from slave */ |
303d1cbf | 562 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 563 | struct net_device *slave_dev) |
1da177e4 | 564 | { |
454d7c9b | 565 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 566 | |
303d1cbf JV |
567 | dev_uc_unsync(slave_dev, bond_dev); |
568 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 | 569 | |
01844098 | 570 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
571 | /* del lacpdu mc addr from mc list */ |
572 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
573 | ||
22bedad3 | 574 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
575 | } |
576 | } | |
577 | ||
578 | /*--------------------------- Active slave change ---------------------------*/ | |
579 | ||
303d1cbf | 580 | /* Update the hardware address list and promisc/allmulti for the new and |
ec0865a9 VF |
581 | * old active slaves (if any). Modes that are not using primary keep all |
582 | * slaves up date at all times; only the modes that use primary need to call | |
303d1cbf | 583 | * this function to swap these settings during a failover. |
1da177e4 | 584 | */ |
303d1cbf JV |
585 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
586 | struct slave *old_active) | |
1da177e4 | 587 | { |
1da177e4 | 588 | if (old_active) { |
3d632c3f | 589 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 590 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 591 | |
3d632c3f | 592 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 593 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 594 | |
303d1cbf | 595 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
596 | } |
597 | ||
598 | if (new_active) { | |
7e1a1ac1 | 599 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 600 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 601 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 602 | |
3d632c3f | 603 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 604 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 605 | |
d632ce98 | 606 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
607 | dev_uc_sync(new_active->dev, bond->dev); |
608 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 609 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
610 | } |
611 | } | |
612 | ||
ae0d6750 | 613 | /** |
614 | * bond_set_dev_addr - clone slave's address to bond | |
615 | * @bond_dev: bond net device | |
616 | * @slave_dev: slave net device | |
617 | * | |
618 | * Should be called with RTNL held. | |
619 | */ | |
620 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
621 | struct net_device *slave_dev) | |
622 | { | |
76444f50 VF |
623 | netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
624 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 625 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
626 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
627 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
628 | } | |
629 | ||
a951bc1e | 630 | static struct slave *bond_get_old_active(struct bonding *bond, |
631 | struct slave *new_active) | |
632 | { | |
633 | struct slave *slave; | |
634 | struct list_head *iter; | |
635 | ||
636 | bond_for_each_slave(bond, slave, iter) { | |
637 | if (slave == new_active) | |
638 | continue; | |
639 | ||
640 | if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr)) | |
641 | return slave; | |
642 | } | |
643 | ||
644 | return NULL; | |
645 | } | |
646 | ||
547942ca | 647 | /* bond_do_fail_over_mac |
3915c1e8 JV |
648 | * |
649 | * Perform special MAC address swapping for fail_over_mac settings | |
650 | * | |
1c72cfdc | 651 | * Called with RTNL |
3915c1e8 JV |
652 | */ |
653 | static void bond_do_fail_over_mac(struct bonding *bond, | |
654 | struct slave *new_active, | |
655 | struct slave *old_active) | |
656 | { | |
657 | u8 tmp_mac[ETH_ALEN]; | |
658 | struct sockaddr saddr; | |
659 | int rv; | |
660 | ||
661 | switch (bond->params.fail_over_mac) { | |
662 | case BOND_FOM_ACTIVE: | |
1c72cfdc | 663 | if (new_active) |
ae0d6750 | 664 | bond_set_dev_addr(bond->dev, new_active->dev); |
3915c1e8 JV |
665 | break; |
666 | case BOND_FOM_FOLLOW: | |
547942ca | 667 | /* if new_active && old_active, swap them |
3915c1e8 JV |
668 | * if just old_active, do nothing (going to no active slave) |
669 | * if just new_active, set new_active to bond's MAC | |
670 | */ | |
671 | if (!new_active) | |
672 | return; | |
673 | ||
a951bc1e | 674 | if (!old_active) |
675 | old_active = bond_get_old_active(bond, new_active); | |
676 | ||
3915c1e8 | 677 | if (old_active) { |
ada0f863 | 678 | ether_addr_copy(tmp_mac, new_active->dev->dev_addr); |
2a7c183b JP |
679 | ether_addr_copy(saddr.sa_data, |
680 | old_active->dev->dev_addr); | |
3915c1e8 JV |
681 | saddr.sa_family = new_active->dev->type; |
682 | } else { | |
ada0f863 | 683 | ether_addr_copy(saddr.sa_data, bond->dev->dev_addr); |
3915c1e8 JV |
684 | saddr.sa_family = bond->dev->type; |
685 | } | |
686 | ||
687 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
688 | if (rv) { | |
76444f50 VF |
689 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
690 | -rv, new_active->dev->name); | |
3915c1e8 JV |
691 | goto out; |
692 | } | |
693 | ||
694 | if (!old_active) | |
695 | goto out; | |
696 | ||
ada0f863 | 697 | ether_addr_copy(saddr.sa_data, tmp_mac); |
3915c1e8 JV |
698 | saddr.sa_family = old_active->dev->type; |
699 | ||
700 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
701 | if (rv) | |
76444f50 VF |
702 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
703 | -rv, new_active->dev->name); | |
3915c1e8 | 704 | out: |
3915c1e8 JV |
705 | break; |
706 | default: | |
76444f50 VF |
707 | netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n", |
708 | bond->params.fail_over_mac); | |
3915c1e8 JV |
709 | break; |
710 | } | |
711 | ||
712 | } | |
713 | ||
b5a983f3 | 714 | static struct slave *bond_choose_primary_or_current(struct bonding *bond) |
a549952a | 715 | { |
059b47e8 | 716 | struct slave *prim = rtnl_dereference(bond->primary_slave); |
1c72cfdc | 717 | struct slave *curr = rtnl_dereference(bond->curr_active_slave); |
a549952a | 718 | |
b5a983f3 MR |
719 | if (!prim || prim->link != BOND_LINK_UP) { |
720 | if (!curr || curr->link != BOND_LINK_UP) | |
721 | return NULL; | |
722 | return curr; | |
723 | } | |
724 | ||
a549952a JP |
725 | if (bond->force_primary) { |
726 | bond->force_primary = false; | |
b5a983f3 MR |
727 | return prim; |
728 | } | |
729 | ||
730 | if (!curr || curr->link != BOND_LINK_UP) | |
731 | return prim; | |
732 | ||
733 | /* At this point, prim and curr are both up */ | |
734 | switch (bond->params.primary_reselect) { | |
735 | case BOND_PRI_RESELECT_ALWAYS: | |
736 | return prim; | |
737 | case BOND_PRI_RESELECT_BETTER: | |
738 | if (prim->speed < curr->speed) | |
739 | return curr; | |
740 | if (prim->speed == curr->speed && prim->duplex <= curr->duplex) | |
741 | return curr; | |
742 | return prim; | |
743 | case BOND_PRI_RESELECT_FAILURE: | |
744 | return curr; | |
745 | default: | |
746 | netdev_err(bond->dev, "impossible primary_reselect %d\n", | |
747 | bond->params.primary_reselect); | |
748 | return curr; | |
a549952a | 749 | } |
a549952a | 750 | } |
3915c1e8 | 751 | |
1da177e4 | 752 | /** |
b5a983f3 | 753 | * bond_find_best_slave - select the best available slave to be the active one |
1da177e4 | 754 | * @bond: our bonding struct |
1da177e4 LT |
755 | */ |
756 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
757 | { | |
b5a983f3 | 758 | struct slave *slave, *bestslave = NULL; |
77140d29 | 759 | struct list_head *iter; |
1da177e4 | 760 | int mintime = bond->params.updelay; |
1da177e4 | 761 | |
b5a983f3 MR |
762 | slave = bond_choose_primary_or_current(bond); |
763 | if (slave) | |
764 | return slave; | |
1da177e4 | 765 | |
77140d29 VF |
766 | bond_for_each_slave(bond, slave, iter) { |
767 | if (slave->link == BOND_LINK_UP) | |
768 | return slave; | |
b6adc610 | 769 | if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) && |
77140d29 VF |
770 | slave->delay < mintime) { |
771 | mintime = slave->delay; | |
772 | bestslave = slave; | |
1da177e4 LT |
773 | } |
774 | } | |
775 | ||
776 | return bestslave; | |
777 | } | |
778 | ||
ad246c99 BH |
779 | static bool bond_should_notify_peers(struct bonding *bond) |
780 | { | |
4cb4f97b | 781 | struct slave *slave; |
782 | ||
783 | rcu_read_lock(); | |
784 | slave = rcu_dereference(bond->curr_active_slave); | |
785 | rcu_read_unlock(); | |
ad246c99 | 786 | |
76444f50 VF |
787 | netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n", |
788 | slave ? slave->dev->name : "NULL"); | |
ad246c99 BH |
789 | |
790 | if (!slave || !bond->send_peer_notif || | |
b02e3e94 | 791 | !netif_carrier_ok(bond->dev) || |
ad246c99 BH |
792 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) |
793 | return false; | |
794 | ||
ad246c99 BH |
795 | return true; |
796 | } | |
797 | ||
1da177e4 LT |
798 | /** |
799 | * change_active_interface - change the active slave into the specified one | |
800 | * @bond: our bonding struct | |
801 | * @new: the new slave to make the active one | |
802 | * | |
803 | * Set the new slave to the bond's settings and unset them on the old | |
804 | * curr_active_slave. | |
805 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
806 | * | |
807 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
808 | * because it is apparently the best available slave we have, even though its | |
809 | * updelay hasn't timed out yet. | |
810 | * | |
1c72cfdc | 811 | * Caller must hold RTNL. |
1da177e4 | 812 | */ |
a77b5325 | 813 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 | 814 | { |
4740d638 ED |
815 | struct slave *old_active; |
816 | ||
1c72cfdc NA |
817 | ASSERT_RTNL(); |
818 | ||
819 | old_active = rtnl_dereference(bond->curr_active_slave); | |
1da177e4 | 820 | |
3d632c3f | 821 | if (old_active == new_active) |
1da177e4 | 822 | return; |
1da177e4 LT |
823 | |
824 | if (new_active) { | |
8e603460 | 825 | new_active->last_link_up = jiffies; |
b2220cad | 826 | |
1da177e4 | 827 | if (new_active->link == BOND_LINK_BACK) { |
ec0865a9 | 828 | if (bond_uses_primary(bond)) { |
76444f50 VF |
829 | netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n", |
830 | new_active->dev->name, | |
831 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
832 | } |
833 | ||
834 | new_active->delay = 0; | |
5d397061 JP |
835 | bond_set_slave_link_state(new_active, BOND_LINK_UP, |
836 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 837 | |
01844098 | 838 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 839 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 840 | |
58402054 | 841 | if (bond_is_lb(bond)) |
1da177e4 | 842 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 | 843 | } else { |
ec0865a9 | 844 | if (bond_uses_primary(bond)) { |
76444f50 VF |
845 | netdev_info(bond->dev, "making interface %s the new active one\n", |
846 | new_active->dev->name); | |
1da177e4 LT |
847 | } |
848 | } | |
849 | } | |
850 | ||
ec0865a9 | 851 | if (bond_uses_primary(bond)) |
303d1cbf | 852 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 853 | |
58402054 | 854 | if (bond_is_lb(bond)) { |
1da177e4 | 855 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 856 | if (old_active) |
5e5b0665 | 857 | bond_set_slave_inactive_flags(old_active, |
858 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 859 | if (new_active) |
5e5b0665 | 860 | bond_set_slave_active_flags(new_active, |
861 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 862 | } else { |
278b2083 | 863 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 864 | } |
c3ade5ca | 865 | |
01844098 | 866 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) { |
3d632c3f | 867 | if (old_active) |
5e5b0665 | 868 | bond_set_slave_inactive_flags(old_active, |
869 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
870 | |
871 | if (new_active) { | |
ad246c99 BH |
872 | bool should_notify_peers = false; |
873 | ||
5e5b0665 | 874 | bond_set_slave_active_flags(new_active, |
875 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 876 | |
709f8a45 OG |
877 | if (bond->params.fail_over_mac) |
878 | bond_do_fail_over_mac(bond, new_active, | |
879 | old_active); | |
3915c1e8 | 880 | |
ad246c99 BH |
881 | if (netif_running(bond->dev)) { |
882 | bond->send_peer_notif = | |
883 | bond->params.num_peer_notif; | |
884 | should_notify_peers = | |
885 | bond_should_notify_peers(bond); | |
886 | } | |
887 | ||
b7bc2a5b | 888 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 889 | if (should_notify_peers) |
b7bc2a5b AW |
890 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
891 | bond->dev); | |
7893b249 | 892 | } |
c3ade5ca | 893 | } |
a2fd940f | 894 | |
5a37e8ca | 895 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
896 | * all were sent on curr_active_slave. |
897 | * resend only if bond is brought up with the affected | |
547942ca NA |
898 | * bonding modes and the retransmission is enabled |
899 | */ | |
94265cf5 | 900 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && |
ec0865a9 | 901 | ((bond_uses_primary(bond) && new_active) || |
01844098 | 902 | BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) { |
c2952c31 | 903 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 904 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 905 | } |
1da177e4 LT |
906 | } |
907 | ||
908 | /** | |
909 | * bond_select_active_slave - select a new active slave, if needed | |
910 | * @bond: our bonding struct | |
911 | * | |
3d632c3f | 912 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
913 | * - The old curr_active_slave has been released or lost its link. |
914 | * - The primary_slave has got its link back. | |
915 | * - A slave has got its link back and there's no old curr_active_slave. | |
916 | * | |
1c72cfdc | 917 | * Caller must hold RTNL. |
1da177e4 | 918 | */ |
a77b5325 | 919 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
920 | { |
921 | struct slave *best_slave; | |
ff59c456 | 922 | int rv; |
1da177e4 | 923 | |
e0974585 NA |
924 | ASSERT_RTNL(); |
925 | ||
1da177e4 | 926 | best_slave = bond_find_best_slave(bond); |
1c72cfdc | 927 | if (best_slave != rtnl_dereference(bond->curr_active_slave)) { |
1da177e4 | 928 | bond_change_active_slave(bond, best_slave); |
ff59c456 JV |
929 | rv = bond_set_carrier(bond); |
930 | if (!rv) | |
931 | return; | |
932 | ||
933 | if (netif_carrier_ok(bond->dev)) { | |
76444f50 | 934 | netdev_info(bond->dev, "first active interface up!\n"); |
ff59c456 | 935 | } else { |
76444f50 | 936 | netdev_info(bond->dev, "now running without any active interface!\n"); |
ff59c456 | 937 | } |
1da177e4 LT |
938 | } |
939 | } | |
940 | ||
f6dc31a8 | 941 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 942 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 943 | { |
8a8efa22 AW |
944 | struct netpoll *np; |
945 | int err = 0; | |
f6dc31a8 | 946 | |
a8779ec1 | 947 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
948 | err = -ENOMEM; |
949 | if (!np) | |
950 | goto out; | |
951 | ||
a8779ec1 | 952 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
953 | if (err) { |
954 | kfree(np); | |
955 | goto out; | |
f6dc31a8 | 956 | } |
8a8efa22 AW |
957 | slave->np = np; |
958 | out: | |
959 | return err; | |
960 | } | |
961 | static inline void slave_disable_netpoll(struct slave *slave) | |
962 | { | |
963 | struct netpoll *np = slave->np; | |
964 | ||
965 | if (!np) | |
966 | return; | |
967 | ||
968 | slave->np = NULL; | |
2cde6acd | 969 | __netpoll_free_async(np); |
8a8efa22 | 970 | } |
f6dc31a8 WC |
971 | |
972 | static void bond_poll_controller(struct net_device *bond_dev) | |
973 | { | |
616f4541 MB |
974 | struct bonding *bond = netdev_priv(bond_dev); |
975 | struct slave *slave = NULL; | |
976 | struct list_head *iter; | |
977 | struct ad_info ad_info; | |
978 | struct netpoll_info *ni; | |
979 | const struct net_device_ops *ops; | |
980 | ||
981 | if (BOND_MODE(bond) == BOND_MODE_8023AD) | |
982 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) | |
983 | return; | |
984 | ||
616f4541 MB |
985 | bond_for_each_slave_rcu(bond, slave, iter) { |
986 | ops = slave->dev->netdev_ops; | |
987 | if (!bond_slave_is_up(slave) || !ops->ndo_poll_controller) | |
988 | continue; | |
989 | ||
990 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
991 | struct aggregator *agg = | |
992 | SLAVE_AD_INFO(slave)->port.aggregator; | |
993 | ||
994 | if (agg && | |
995 | agg->aggregator_identifier != ad_info.aggregator_id) | |
996 | continue; | |
997 | } | |
998 | ||
999 | ni = rcu_dereference_bh(slave->dev->npinfo); | |
1000 | if (down_trylock(&ni->dev_lock)) | |
1001 | continue; | |
1002 | ops->ndo_poll_controller(slave->dev); | |
1003 | up(&ni->dev_lock); | |
1004 | } | |
8a8efa22 AW |
1005 | } |
1006 | ||
c4cdef9b | 1007 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1008 | { |
c4cdef9b | 1009 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1010 | struct list_head *iter; |
c2355e1a | 1011 | struct slave *slave; |
c2355e1a | 1012 | |
9caff1e7 | 1013 | bond_for_each_slave(bond, slave, iter) |
b6adc610 | 1014 | if (bond_slave_is_up(slave)) |
8a8efa22 | 1015 | slave_disable_netpoll(slave); |
f6dc31a8 | 1016 | } |
8a8efa22 | 1017 | |
a8779ec1 | 1018 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
1019 | { |
1020 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 1021 | struct list_head *iter; |
f6dc31a8 | 1022 | struct slave *slave; |
dec1e90e | 1023 | int err = 0; |
f6dc31a8 | 1024 | |
9caff1e7 | 1025 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
1026 | err = slave_enable_netpoll(slave); |
1027 | if (err) { | |
c4cdef9b | 1028 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1029 | break; |
f6dc31a8 WC |
1030 | } |
1031 | } | |
8a8efa22 | 1032 | return err; |
f6dc31a8 | 1033 | } |
8a8efa22 AW |
1034 | #else |
1035 | static inline int slave_enable_netpoll(struct slave *slave) | |
1036 | { | |
1037 | return 0; | |
1038 | } | |
1039 | static inline void slave_disable_netpoll(struct slave *slave) | |
1040 | { | |
1041 | } | |
f6dc31a8 WC |
1042 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1043 | { | |
1044 | } | |
f6dc31a8 WC |
1045 | #endif |
1046 | ||
1da177e4 LT |
1047 | /*---------------------------------- IOCTL ----------------------------------*/ |
1048 | ||
c8f44aff | 1049 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1050 | netdev_features_t features) |
8531c5ff | 1051 | { |
b2a103e6 | 1052 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 1053 | struct list_head *iter; |
c8f44aff | 1054 | netdev_features_t mask; |
9b7b165a | 1055 | struct slave *slave; |
b63365a2 | 1056 | |
7889cbee | 1057 | mask = features; |
c158cba3 | 1058 | |
b63365a2 | 1059 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1060 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1061 | |
9caff1e7 | 1062 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
1063 | features = netdev_increment_features(features, |
1064 | slave->dev->features, | |
b2a103e6 MM |
1065 | mask); |
1066 | } | |
b0ce3508 | 1067 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1068 | |
b2a103e6 MM |
1069 | return features; |
1070 | } | |
1071 | ||
a188222b | 1072 | #define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
62f2a3a4 MM |
1073 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ |
1074 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 | 1075 | |
a188222b TH |
1076 | #define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
1077 | NETIF_F_RXCSUM | NETIF_F_ALL_TSO) | |
5a7baa78 | 1078 | |
b2a103e6 MM |
1079 | static void bond_compute_features(struct bonding *bond) |
1080 | { | |
02875878 ED |
1081 | unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | |
1082 | IFF_XMIT_DST_RELEASE_PERM; | |
c8f44aff | 1083 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
5a7baa78 | 1084 | netdev_features_t enc_features = BOND_ENC_FEATURES; |
9caff1e7 VF |
1085 | struct net_device *bond_dev = bond->dev; |
1086 | struct list_head *iter; | |
1087 | struct slave *slave; | |
b2a103e6 | 1088 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1089 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1090 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1091 | |
0965a1f3 | 1092 | if (!bond_has_slaves(bond)) |
b2a103e6 | 1093 | goto done; |
a9b3ace4 | 1094 | vlan_features &= NETIF_F_ALL_FOR_ALL; |
b2a103e6 | 1095 | |
9caff1e7 | 1096 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1097 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1098 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1099 | ||
5a7baa78 OG |
1100 | enc_features = netdev_increment_features(enc_features, |
1101 | slave->dev->hw_enc_features, | |
1102 | BOND_ENC_FEATURES); | |
b6fe83e9 | 1103 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1104 | if (slave->dev->hard_header_len > max_hard_header_len) |
1105 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1106 | |
1107 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1108 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1109 | } |
8531c5ff | 1110 | |
b63365a2 | 1111 | done: |
b2a103e6 | 1112 | bond_dev->vlan_features = vlan_features; |
24f87d4c | 1113 | bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL; |
54ef3137 | 1114 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1115 | bond_dev->gso_max_segs = gso_max_segs; |
1116 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1117 | |
02875878 ED |
1118 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
1119 | if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) && | |
1120 | dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM)) | |
1121 | bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE; | |
b6fe83e9 | 1122 | |
b2a103e6 | 1123 | netdev_change_features(bond_dev); |
8531c5ff AK |
1124 | } |
1125 | ||
872254dd MS |
1126 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1127 | struct net_device *slave_dev) | |
1128 | { | |
00829823 | 1129 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1130 | |
1131 | bond_dev->type = slave_dev->type; | |
1132 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1133 | bond_dev->addr_len = slave_dev->addr_len; | |
1134 | ||
1135 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1136 | slave_dev->addr_len); | |
1137 | } | |
1138 | ||
5b2c4dd2 | 1139 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1140 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1141 | */ |
1142 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1143 | struct slave *slave, |
1144 | struct bonding *bond) | |
5b2c4dd2 | 1145 | { |
2d7011ca | 1146 | if (bond_is_slave_inactive(slave)) { |
01844098 | 1147 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
5b2c4dd2 JP |
1148 | skb->pkt_type != PACKET_BROADCAST && |
1149 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1150 | return false; |
5b2c4dd2 JP |
1151 | return true; |
1152 | } | |
1153 | return false; | |
1154 | } | |
1155 | ||
8a4eb573 | 1156 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1157 | { |
8a4eb573 | 1158 | struct sk_buff *skb = *pskb; |
f1c1775a | 1159 | struct slave *slave; |
0bd80dad | 1160 | struct bonding *bond; |
de063b70 ED |
1161 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1162 | struct slave *); | |
13a8e0c8 | 1163 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1164 | |
8a4eb573 JP |
1165 | skb = skb_share_check(skb, GFP_ATOMIC); |
1166 | if (unlikely(!skb)) | |
1167 | return RX_HANDLER_CONSUMED; | |
1168 | ||
1169 | *pskb = skb; | |
5b2c4dd2 | 1170 | |
35d48903 JP |
1171 | slave = bond_slave_get_rcu(skb->dev); |
1172 | bond = slave->bond; | |
0bd80dad | 1173 | |
4d97480b MH |
1174 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1175 | if (recv_probe) { | |
de063b70 ED |
1176 | ret = recv_probe(skb, bond, slave); |
1177 | if (ret == RX_HANDLER_CONSUMED) { | |
1178 | consume_skb(skb); | |
1179 | return ret; | |
3aba891d JP |
1180 | } |
1181 | } | |
1182 | ||
0bd80dad | 1183 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1184 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1185 | } |
1186 | ||
35d48903 | 1187 | skb->dev = bond->dev; |
5b2c4dd2 | 1188 | |
01844098 | 1189 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
35d48903 | 1190 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1191 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1192 | |
541ac7c9 CG |
1193 | if (unlikely(skb_cow_head(skb, |
1194 | skb->data - skb_mac_header(skb)))) { | |
1195 | kfree_skb(skb); | |
8a4eb573 | 1196 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1197 | } |
2a7c183b | 1198 | ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr); |
5b2c4dd2 JP |
1199 | } |
1200 | ||
13a8e0c8 | 1201 | return ret; |
5b2c4dd2 JP |
1202 | } |
1203 | ||
41f0b049 | 1204 | static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond) |
471cb5a3 | 1205 | { |
41f0b049 JP |
1206 | switch (BOND_MODE(bond)) { |
1207 | case BOND_MODE_ROUNDROBIN: | |
1208 | return NETDEV_LAG_TX_TYPE_ROUNDROBIN; | |
1209 | case BOND_MODE_ACTIVEBACKUP: | |
1210 | return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP; | |
1211 | case BOND_MODE_BROADCAST: | |
1212 | return NETDEV_LAG_TX_TYPE_BROADCAST; | |
1213 | case BOND_MODE_XOR: | |
1214 | case BOND_MODE_8023AD: | |
1215 | return NETDEV_LAG_TX_TYPE_HASH; | |
1216 | default: | |
1217 | return NETDEV_LAG_TX_TYPE_UNKNOWN; | |
1218 | } | |
1219 | } | |
1220 | ||
1221 | static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave) | |
1222 | { | |
1223 | struct netdev_lag_upper_info lag_upper_info; | |
471cb5a3 JP |
1224 | int err; |
1225 | ||
41f0b049 JP |
1226 | lag_upper_info.tx_type = bond_lag_tx_type(bond); |
1227 | err = netdev_master_upper_dev_link(slave->dev, bond->dev, slave, | |
1228 | &lag_upper_info); | |
471cb5a3 JP |
1229 | if (err) |
1230 | return err; | |
41f0b049 | 1231 | rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1232 | return 0; |
1233 | } | |
1234 | ||
41f0b049 | 1235 | static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave) |
471cb5a3 | 1236 | { |
41f0b049 JP |
1237 | netdev_upper_dev_unlink(slave->dev, bond->dev); |
1238 | slave->dev->flags &= ~IFF_SLAVE; | |
1239 | rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL); | |
471cb5a3 JP |
1240 | } |
1241 | ||
3fdddd85 | 1242 | static struct slave *bond_alloc_slave(struct bonding *bond) |
1243 | { | |
1244 | struct slave *slave = NULL; | |
1245 | ||
1246 | slave = kzalloc(sizeof(struct slave), GFP_KERNEL); | |
1247 | if (!slave) | |
1248 | return NULL; | |
1249 | ||
01844098 | 1250 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
3fdddd85 | 1251 | SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info), |
1252 | GFP_KERNEL); | |
1253 | if (!SLAVE_AD_INFO(slave)) { | |
1254 | kfree(slave); | |
1255 | return NULL; | |
1256 | } | |
1257 | } | |
1258 | return slave; | |
1259 | } | |
1260 | ||
1261 | static void bond_free_slave(struct slave *slave) | |
1262 | { | |
1263 | struct bonding *bond = bond_get_bond_by_slave(slave); | |
1264 | ||
01844098 | 1265 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3fdddd85 | 1266 | kfree(SLAVE_AD_INFO(slave)); |
1267 | ||
1268 | kfree(slave); | |
1269 | } | |
1270 | ||
69a2338e MS |
1271 | static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info) |
1272 | { | |
1273 | info->bond_mode = BOND_MODE(bond); | |
1274 | info->miimon = bond->params.miimon; | |
1275 | info->num_slaves = bond->slave_cnt; | |
1276 | } | |
1277 | ||
1278 | static void bond_fill_ifslave(struct slave *slave, struct ifslave *info) | |
1279 | { | |
1280 | strcpy(info->slave_name, slave->dev->name); | |
1281 | info->link = slave->link; | |
1282 | info->state = bond_slave_state(slave); | |
1283 | info->link_failure_count = slave->link_failure_count; | |
1284 | } | |
1285 | ||
92e584fe MS |
1286 | static void bond_netdev_notify(struct net_device *dev, |
1287 | struct netdev_bonding_info *info) | |
69e61133 | 1288 | { |
69e61133 | 1289 | rtnl_lock(); |
92e584fe | 1290 | netdev_bonding_info_change(dev, info); |
69e61133 MS |
1291 | rtnl_unlock(); |
1292 | } | |
1293 | ||
1294 | static void bond_netdev_notify_work(struct work_struct *_work) | |
1295 | { | |
1296 | struct netdev_notify_work *w = | |
1297 | container_of(_work, struct netdev_notify_work, work.work); | |
1298 | ||
92e584fe | 1299 | bond_netdev_notify(w->dev, &w->bonding_info); |
69e61133 | 1300 | dev_put(w->dev); |
92e584fe | 1301 | kfree(w); |
69e61133 MS |
1302 | } |
1303 | ||
1304 | void bond_queue_slave_event(struct slave *slave) | |
1305 | { | |
92e584fe | 1306 | struct bonding *bond = slave->bond; |
69e61133 MS |
1307 | struct netdev_notify_work *nnw = kzalloc(sizeof(*nnw), GFP_ATOMIC); |
1308 | ||
1309 | if (!nnw) | |
1310 | return; | |
1311 | ||
92e584fe | 1312 | dev_hold(slave->dev); |
69e61133 | 1313 | nnw->dev = slave->dev; |
92e584fe MS |
1314 | bond_fill_ifslave(slave, &nnw->bonding_info.slave); |
1315 | bond_fill_ifbond(bond, &nnw->bonding_info.master); | |
1316 | INIT_DELAYED_WORK(&nnw->work, bond_netdev_notify_work); | |
69e61133 | 1317 | |
92e584fe | 1318 | queue_delayed_work(slave->bond->wq, &nnw->work, 0); |
69e61133 MS |
1319 | } |
1320 | ||
f7c7eb7f JP |
1321 | void bond_lower_state_changed(struct slave *slave) |
1322 | { | |
1323 | struct netdev_lag_lower_state_info info; | |
1324 | ||
1325 | info.link_up = slave->link == BOND_LINK_UP || | |
1326 | slave->link == BOND_LINK_FAIL; | |
1327 | info.tx_enabled = bond_is_active_slave(slave); | |
1328 | netdev_lower_state_changed(slave->dev, &info); | |
1329 | } | |
1330 | ||
1da177e4 | 1331 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1332 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1333 | { |
454d7c9b | 1334 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1335 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1336 | struct slave *new_slave = NULL, *prev_slave; |
1da177e4 LT |
1337 | struct sockaddr addr; |
1338 | int link_reporting; | |
8599b52e | 1339 | int res = 0, i; |
1da177e4 | 1340 | |
c772dde3 BH |
1341 | if (!bond->params.use_carrier && |
1342 | slave_dev->ethtool_ops->get_link == NULL && | |
1343 | slave_ops->ndo_do_ioctl == NULL) { | |
76444f50 VF |
1344 | netdev_warn(bond_dev, "no link monitoring support for %s\n", |
1345 | slave_dev->name); | |
1da177e4 LT |
1346 | } |
1347 | ||
1da177e4 LT |
1348 | /* already enslaved */ |
1349 | if (slave_dev->flags & IFF_SLAVE) { | |
76444f50 | 1350 | netdev_dbg(bond_dev, "Error: Device was already enslaved\n"); |
1da177e4 LT |
1351 | return -EBUSY; |
1352 | } | |
1353 | ||
09a89c21 | 1354 | if (bond_dev == slave_dev) { |
76444f50 | 1355 | netdev_err(bond_dev, "cannot enslave bond to itself.\n"); |
09a89c21 JB |
1356 | return -EPERM; |
1357 | } | |
1358 | ||
1da177e4 LT |
1359 | /* vlan challenged mutual exclusion */ |
1360 | /* no need to lock since we're protected by rtnl_lock */ | |
1361 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
76444f50 VF |
1362 | netdev_dbg(bond_dev, "%s is NETIF_F_VLAN_CHALLENGED\n", |
1363 | slave_dev->name); | |
55462cf3 | 1364 | if (vlan_uses_dev(bond_dev)) { |
76444f50 VF |
1365 | netdev_err(bond_dev, "Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1366 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1367 | return -EPERM; |
1368 | } else { | |
76444f50 VF |
1369 | netdev_warn(bond_dev, "enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1370 | slave_dev->name, slave_dev->name, | |
1371 | bond_dev->name); | |
1da177e4 LT |
1372 | } |
1373 | } else { | |
76444f50 VF |
1374 | netdev_dbg(bond_dev, "%s is !NETIF_F_VLAN_CHALLENGED\n", |
1375 | slave_dev->name); | |
1da177e4 LT |
1376 | } |
1377 | ||
547942ca | 1378 | /* Old ifenslave binaries are no longer supported. These can |
3d632c3f | 1379 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1380 | * the current ifenslave will set the interface down prior to |
1381 | * enslaving it; the old ifenslave will not. | |
1382 | */ | |
ce3ea1c7 | 1383 | if (slave_dev->flags & IFF_UP) { |
76444f50 VF |
1384 | netdev_err(bond_dev, "%s is up - this may be due to an out of date ifenslave\n", |
1385 | slave_dev->name); | |
217df670 JV |
1386 | res = -EPERM; |
1387 | goto err_undo_flags; | |
1388 | } | |
1da177e4 | 1389 | |
872254dd MS |
1390 | /* set bonding device ether type by slave - bonding netdevices are |
1391 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1392 | * there is a need to override some of the type dependent attribs/funcs. | |
1393 | * | |
1394 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1395 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1396 | */ | |
0965a1f3 | 1397 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1398 | if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1399 | netdev_dbg(bond_dev, "change device type from %d to %d\n", |
1400 | bond_dev->type, slave_dev->type); | |
75c78500 | 1401 | |
b7bc2a5b AW |
1402 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1403 | bond_dev); | |
3ca5b404 JP |
1404 | res = notifier_to_errno(res); |
1405 | if (res) { | |
76444f50 | 1406 | netdev_err(bond_dev, "refused to change device type\n"); |
3ca5b404 JP |
1407 | res = -EBUSY; |
1408 | goto err_undo_flags; | |
1409 | } | |
75c78500 | 1410 | |
32a806c1 | 1411 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1412 | dev_uc_flush(bond_dev); |
22bedad3 | 1413 | dev_mc_flush(bond_dev); |
32a806c1 | 1414 | |
e36b9d16 MS |
1415 | if (slave_dev->type != ARPHRD_ETHER) |
1416 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1417 | else { |
e36b9d16 | 1418 | ether_setup(bond_dev); |
550fd08c NH |
1419 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1420 | } | |
75c78500 | 1421 | |
b7bc2a5b AW |
1422 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1423 | bond_dev); | |
e36b9d16 | 1424 | } |
872254dd | 1425 | } else if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1426 | netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n", |
1427 | slave_dev->name, slave_dev->type, bond_dev->type); | |
a4aee5c8 JP |
1428 | res = -EINVAL; |
1429 | goto err_undo_flags; | |
872254dd MS |
1430 | } |
1431 | ||
eb7cc59a | 1432 | if (slave_ops->ndo_set_mac_address == NULL) { |
76444f50 | 1433 | netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n"); |
f5442441 VF |
1434 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP && |
1435 | bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
1436 | if (!bond_has_slaves(bond)) { | |
00503b6f | 1437 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
76444f50 | 1438 | netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n"); |
f5442441 | 1439 | } else { |
76444f50 | 1440 | netdev_err(bond_dev, "The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n"); |
f5442441 VF |
1441 | res = -EOPNOTSUPP; |
1442 | goto err_undo_flags; | |
00503b6f | 1443 | } |
2ab82852 | 1444 | } |
1da177e4 LT |
1445 | } |
1446 | ||
8d8fc29d AW |
1447 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1448 | ||
c20811a7 | 1449 | /* If this is the first slave, then we need to set the master's hardware |
547942ca NA |
1450 | * address to be the same as the slave's. |
1451 | */ | |
0965a1f3 | 1452 | if (!bond_has_slaves(bond) && |
dec1e90e | 1453 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1454 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1455 | |
3fdddd85 | 1456 | new_slave = bond_alloc_slave(bond); |
1da177e4 LT |
1457 | if (!new_slave) { |
1458 | res = -ENOMEM; | |
1459 | goto err_undo_flags; | |
1460 | } | |
3fdddd85 | 1461 | |
dc73c41f VF |
1462 | new_slave->bond = bond; |
1463 | new_slave->dev = slave_dev; | |
547942ca | 1464 | /* Set the new_slave's queue_id to be zero. Queue ID mapping |
bb1d9123 AG |
1465 | * is set via sysfs or module option if desired. |
1466 | */ | |
1467 | new_slave->queue_id = 0; | |
1468 | ||
b15ba0fb JP |
1469 | /* Save slave's original mtu and then set it to match the bond */ |
1470 | new_slave->original_mtu = slave_dev->mtu; | |
1471 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1472 | if (res) { | |
76444f50 | 1473 | netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res); |
b15ba0fb JP |
1474 | goto err_free; |
1475 | } | |
1476 | ||
547942ca | 1477 | /* Save slave's original ("permanent") mac address for modes |
217df670 JV |
1478 | * that need it, and for restoring it upon release, and then |
1479 | * set it to the master's address | |
1480 | */ | |
ada0f863 | 1481 | ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr); |
1da177e4 | 1482 | |
00503b6f | 1483 | if (!bond->params.fail_over_mac || |
01844098 | 1484 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
547942ca | 1485 | /* Set slave to master's mac address. The application already |
2ab82852 MS |
1486 | * set the master's mac address to that of the first slave |
1487 | */ | |
1488 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1489 | addr.sa_family = slave_dev->type; | |
1490 | res = dev_set_mac_address(slave_dev, &addr); | |
1491 | if (res) { | |
76444f50 | 1492 | netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res); |
b15ba0fb | 1493 | goto err_restore_mtu; |
2ab82852 | 1494 | } |
217df670 | 1495 | } |
1da177e4 | 1496 | |
03d84a5f KH |
1497 | /* set slave flag before open to prevent IPv6 addrconf */ |
1498 | slave_dev->flags |= IFF_SLAVE; | |
1499 | ||
217df670 JV |
1500 | /* open the slave since the application closed it */ |
1501 | res = dev_open(slave_dev); | |
1502 | if (res) { | |
76444f50 | 1503 | netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1504 | goto err_restore_mac; |
1da177e4 LT |
1505 | } |
1506 | ||
0b680e75 | 1507 | slave_dev->priv_flags |= IFF_BONDING; |
5f0c5f73 AG |
1508 | /* initialize slave stats */ |
1509 | dev_get_stats(new_slave->dev, &new_slave->slave_stats); | |
1da177e4 | 1510 | |
58402054 | 1511 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1512 | /* bond_alb_init_slave() must be called before all other stages since |
1513 | * it might fail and we do not want to have to undo everything | |
1514 | */ | |
1515 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1516 | if (res) |
569f0c4d | 1517 | goto err_close; |
1da177e4 LT |
1518 | } |
1519 | ||
ec0865a9 | 1520 | /* If the mode uses primary, then the following is handled by |
303d1cbf | 1521 | * bond_change_active_slave(). |
1da177e4 | 1522 | */ |
ec0865a9 | 1523 | if (!bond_uses_primary(bond)) { |
1da177e4 LT |
1524 | /* set promiscuity level to new slave */ |
1525 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1526 | res = dev_set_promiscuity(slave_dev, 1); |
1527 | if (res) | |
1528 | goto err_close; | |
1da177e4 LT |
1529 | } |
1530 | ||
1531 | /* set allmulti level to new slave */ | |
1532 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1533 | res = dev_set_allmulti(slave_dev, 1); |
1534 | if (res) | |
1535 | goto err_close; | |
1da177e4 LT |
1536 | } |
1537 | ||
b9e40857 | 1538 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1539 | |
1540 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1541 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1542 | ||
b9e40857 | 1543 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1544 | } |
1545 | ||
01844098 | 1546 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
1547 | /* add lacpdu mc addr to mc list */ |
1548 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1549 | ||
22bedad3 | 1550 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1551 | } |
1552 | ||
b8e2fde4 WY |
1553 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1554 | if (res) { | |
76444f50 VF |
1555 | netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n", |
1556 | slave_dev->name); | |
1ff412ad | 1557 | goto err_close; |
1558 | } | |
1da177e4 | 1559 | |
c8c23903 | 1560 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1561 | |
1562 | new_slave->delay = 0; | |
1563 | new_slave->link_failure_count = 0; | |
1564 | ||
876254ae VF |
1565 | bond_update_speed_duplex(new_slave); |
1566 | ||
49f17de7 | 1567 | new_slave->last_rx = jiffies - |
f31c7937 | 1568 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1569 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1570 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1571 | |
1da177e4 LT |
1572 | if (bond->params.miimon && !bond->params.use_carrier) { |
1573 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1574 | ||
1575 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
547942ca | 1576 | /* miimon is set but a bonded network driver |
1da177e4 LT |
1577 | * does not support ETHTOOL/MII and |
1578 | * arp_interval is not set. Note: if | |
1579 | * use_carrier is enabled, we will never go | |
1580 | * here (because netif_carrier is always | |
1581 | * supported); thus, we don't need to change | |
1582 | * the messages for netif_carrier. | |
1583 | */ | |
76444f50 VF |
1584 | netdev_warn(bond_dev, "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", |
1585 | slave_dev->name); | |
1da177e4 LT |
1586 | } else if (link_reporting == -1) { |
1587 | /* unable get link status using mii/ethtool */ | |
76444f50 VF |
1588 | netdev_warn(bond_dev, "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", |
1589 | slave_dev->name); | |
1da177e4 LT |
1590 | } |
1591 | } | |
1592 | ||
1593 | /* check for initial state */ | |
f31c7937 MK |
1594 | if (bond->params.miimon) { |
1595 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1596 | if (bond->params.updelay) { | |
69a2338e | 1597 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1598 | BOND_LINK_BACK, |
1599 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 MK |
1600 | new_slave->delay = bond->params.updelay; |
1601 | } else { | |
69a2338e | 1602 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1603 | BOND_LINK_UP, |
1604 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 | 1605 | } |
1da177e4 | 1606 | } else { |
5d397061 JP |
1607 | bond_set_slave_link_state(new_slave, BOND_LINK_DOWN, |
1608 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1609 | } |
f31c7937 | 1610 | } else if (bond->params.arp_interval) { |
69a2338e MS |
1611 | bond_set_slave_link_state(new_slave, |
1612 | (netif_carrier_ok(slave_dev) ? | |
5d397061 JP |
1613 | BOND_LINK_UP : BOND_LINK_DOWN), |
1614 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1615 | } else { |
5d397061 JP |
1616 | bond_set_slave_link_state(new_slave, BOND_LINK_UP, |
1617 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1618 | } |
1619 | ||
f31c7937 | 1620 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 1621 | new_slave->last_link_up = jiffies; |
76444f50 VF |
1622 | netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n", |
1623 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1624 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 1625 | |
ec0865a9 | 1626 | if (bond_uses_primary(bond) && bond->params.primary[0]) { |
1da177e4 | 1627 | /* if there is a primary slave, remember it */ |
a549952a | 1628 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
059b47e8 | 1629 | rcu_assign_pointer(bond->primary_slave, new_slave); |
a549952a JP |
1630 | bond->force_primary = true; |
1631 | } | |
1da177e4 LT |
1632 | } |
1633 | ||
01844098 | 1634 | switch (BOND_MODE(bond)) { |
1da177e4 | 1635 | case BOND_MODE_ACTIVEBACKUP: |
5e5b0665 | 1636 | bond_set_slave_inactive_flags(new_slave, |
1637 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1638 | break; |
1639 | case BOND_MODE_8023AD: | |
1640 | /* in 802.3ad mode, the internal mechanism | |
1641 | * will activate the slaves in the selected | |
1642 | * aggregator | |
1643 | */ | |
5e5b0665 | 1644 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 1645 | /* if this is the first slave */ |
23c147e0 | 1646 | if (!prev_slave) { |
3fdddd85 | 1647 | SLAVE_AD_INFO(new_slave)->id = 1; |
1da177e4 LT |
1648 | /* Initialize AD with the number of times that the AD timer is called in 1 second |
1649 | * can be called only after the mac address of the bond is set | |
1650 | */ | |
56d00c67 | 1651 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1652 | } else { |
3fdddd85 | 1653 | SLAVE_AD_INFO(new_slave)->id = |
1654 | SLAVE_AD_INFO(prev_slave)->id + 1; | |
1da177e4 LT |
1655 | } |
1656 | ||
1657 | bond_3ad_bind_slave(new_slave); | |
1658 | break; | |
1659 | case BOND_MODE_TLB: | |
1660 | case BOND_MODE_ALB: | |
e30bc066 | 1661 | bond_set_active_slave(new_slave); |
5e5b0665 | 1662 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
1663 | break; |
1664 | default: | |
76444f50 | 1665 | netdev_dbg(bond_dev, "This slave is always active in trunk mode\n"); |
1da177e4 LT |
1666 | |
1667 | /* always active in trunk mode */ | |
e30bc066 | 1668 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1669 | |
1670 | /* In trunking mode there is little meaning to curr_active_slave | |
1671 | * anyway (it holds no special properties of the bond device), | |
1672 | * so we can change it without calling change_active_interface() | |
1673 | */ | |
4740d638 ED |
1674 | if (!rcu_access_pointer(bond->curr_active_slave) && |
1675 | new_slave->link == BOND_LINK_UP) | |
278b2083 | 1676 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1677 | |
1da177e4 LT |
1678 | break; |
1679 | } /* switch(bond_mode) */ | |
1680 | ||
f6dc31a8 | 1681 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1682 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1683 | if (slave_dev->npinfo) { |
1684 | if (slave_enable_netpoll(new_slave)) { | |
76444f50 | 1685 | netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n"); |
8a8efa22 | 1686 | res = -EBUSY; |
f7d9821a | 1687 | goto err_detach; |
8a8efa22 | 1688 | } |
f6dc31a8 WC |
1689 | } |
1690 | #endif | |
8a8efa22 | 1691 | |
fbe168ba MK |
1692 | if (!(bond_dev->features & NETIF_F_LRO)) |
1693 | dev_disable_lro(slave_dev); | |
1694 | ||
35d48903 JP |
1695 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1696 | new_slave); | |
1697 | if (res) { | |
76444f50 | 1698 | netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res); |
5831d66e | 1699 | goto err_detach; |
35d48903 JP |
1700 | } |
1701 | ||
41f0b049 | 1702 | res = bond_master_upper_dev_link(bond, new_slave); |
1f718f0f | 1703 | if (res) { |
76444f50 | 1704 | netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res); |
1f718f0f VF |
1705 | goto err_unregister; |
1706 | } | |
1707 | ||
07699f9a | 1708 | res = bond_sysfs_slave_add(new_slave); |
1709 | if (res) { | |
76444f50 | 1710 | netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res); |
07699f9a | 1711 | goto err_upper_unlink; |
1712 | } | |
1713 | ||
5378c2e6 VF |
1714 | bond->slave_cnt++; |
1715 | bond_compute_features(bond); | |
1716 | bond_set_carrier(bond); | |
1717 | ||
ec0865a9 | 1718 | if (bond_uses_primary(bond)) { |
f80889a5 | 1719 | block_netpoll_tx(); |
5378c2e6 | 1720 | bond_select_active_slave(bond); |
f80889a5 | 1721 | unblock_netpoll_tx(); |
5378c2e6 | 1722 | } |
1f718f0f | 1723 | |
ee637714 MB |
1724 | if (bond_mode_uses_xmit_hash(bond)) |
1725 | bond_update_slave_arr(bond, NULL); | |
1726 | ||
76444f50 VF |
1727 | netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n", |
1728 | slave_dev->name, | |
1729 | bond_is_active_slave(new_slave) ? "an active" : "a backup", | |
1730 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
1731 | |
1732 | /* enslave is successful */ | |
69e61133 | 1733 | bond_queue_slave_event(new_slave); |
1da177e4 LT |
1734 | return 0; |
1735 | ||
1736 | /* Undo stages on error */ | |
07699f9a | 1737 | err_upper_unlink: |
41f0b049 | 1738 | bond_upper_dev_unlink(bond, new_slave); |
07699f9a | 1739 | |
1f718f0f VF |
1740 | err_unregister: |
1741 | netdev_rx_handler_unregister(slave_dev); | |
1742 | ||
f7d9821a | 1743 | err_detach: |
ec0865a9 | 1744 | if (!bond_uses_primary(bond)) |
303d1cbf JV |
1745 | bond_hw_addr_flush(bond_dev, slave_dev); |
1746 | ||
1ff412ad | 1747 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
059b47e8 NA |
1748 | if (rcu_access_pointer(bond->primary_slave) == new_slave) |
1749 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
4740d638 | 1750 | if (rcu_access_pointer(bond->curr_active_slave) == new_slave) { |
f80889a5 | 1751 | block_netpoll_tx(); |
c8517035 | 1752 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1753 | bond_select_active_slave(bond); |
f80889a5 | 1754 | unblock_netpoll_tx(); |
3c5913b5 | 1755 | } |
059b47e8 NA |
1756 | /* either primary_slave or curr_active_slave might've changed */ |
1757 | synchronize_rcu(); | |
fc7a72ac | 1758 | slave_disable_netpoll(new_slave); |
f7d9821a | 1759 | |
1da177e4 | 1760 | err_close: |
b6a5a7b9 | 1761 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1762 | dev_close(slave_dev); |
1763 | ||
1764 | err_restore_mac: | |
03d84a5f | 1765 | slave_dev->flags &= ~IFF_SLAVE; |
00503b6f | 1766 | if (!bond->params.fail_over_mac || |
01844098 | 1767 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
3915c1e8 JV |
1768 | /* XXX TODO - fom follow mode needs to change master's |
1769 | * MAC if this slave's MAC is in use by the bond, or at | |
1770 | * least print a warning. | |
1771 | */ | |
ada0f863 | 1772 | ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr); |
2ab82852 MS |
1773 | addr.sa_family = slave_dev->type; |
1774 | dev_set_mac_address(slave_dev, &addr); | |
1775 | } | |
1da177e4 | 1776 | |
b15ba0fb JP |
1777 | err_restore_mtu: |
1778 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1779 | ||
1da177e4 | 1780 | err_free: |
3fdddd85 | 1781 | bond_free_slave(new_slave); |
1da177e4 LT |
1782 | |
1783 | err_undo_flags: | |
b8fad459 | 1784 | /* Enslave of first slave has failed and we need to fix master's mac */ |
7d5cd2ce NA |
1785 | if (!bond_has_slaves(bond)) { |
1786 | if (ether_addr_equal_64bits(bond_dev->dev_addr, | |
1787 | slave_dev->dev_addr)) | |
1788 | eth_hw_addr_random(bond_dev); | |
1789 | if (bond_dev->type != ARPHRD_ETHER) { | |
40baec22 | 1790 | dev_close(bond_dev); |
7d5cd2ce NA |
1791 | ether_setup(bond_dev); |
1792 | bond_dev->flags |= IFF_MASTER; | |
1793 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; | |
1794 | } | |
1795 | } | |
3d632c3f | 1796 | |
1da177e4 LT |
1797 | return res; |
1798 | } | |
1799 | ||
547942ca | 1800 | /* Try to release the slave device <slave> from the bond device <master> |
1da177e4 | 1801 | * It is legal to access curr_active_slave without a lock because all the function |
8c0bc550 | 1802 | * is RTNL-locked. If "all" is true it means that the function is being called |
0896341a | 1803 | * while destroying a bond interface and all slaves are being released. |
1da177e4 LT |
1804 | * |
1805 | * The rules for slave state should be: | |
1806 | * for Active/Backup: | |
1807 | * Active stays on all backups go down | |
1808 | * for Bonded connections: | |
1809 | * The first up interface should be left on and all others downed. | |
1810 | */ | |
0896341a | 1811 | static int __bond_release_one(struct net_device *bond_dev, |
1812 | struct net_device *slave_dev, | |
1813 | bool all) | |
1da177e4 | 1814 | { |
454d7c9b | 1815 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1816 | struct slave *slave, *oldcurrent; |
1817 | struct sockaddr addr; | |
5a0068de | 1818 | int old_flags = bond_dev->flags; |
c8f44aff | 1819 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1820 | |
1821 | /* slave is not a slave or master is not master of this slave */ | |
1822 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1823 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a22a9e41 | 1824 | netdev_dbg(bond_dev, "cannot release %s\n", |
76444f50 | 1825 | slave_dev->name); |
1da177e4 LT |
1826 | return -EINVAL; |
1827 | } | |
1828 | ||
e843fa50 | 1829 | block_netpoll_tx(); |
1da177e4 LT |
1830 | |
1831 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1832 | if (!slave) { | |
1833 | /* not a slave of this bond */ | |
76444f50 VF |
1834 | netdev_info(bond_dev, "%s not enslaved\n", |
1835 | slave_dev->name); | |
e843fa50 | 1836 | unblock_netpoll_tx(); |
1da177e4 LT |
1837 | return -EINVAL; |
1838 | } | |
1839 | ||
57beaca8 JP |
1840 | bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW); |
1841 | ||
07699f9a | 1842 | bond_sysfs_slave_del(slave); |
1843 | ||
5f0c5f73 AG |
1844 | /* recompute stats just before removing the slave */ |
1845 | bond_get_stats(bond->dev, &bond->bond_stats); | |
1846 | ||
41f0b049 | 1847 | bond_upper_dev_unlink(bond, slave); |
35d48903 JP |
1848 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1849 | * for this slave anymore. | |
1850 | */ | |
1851 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 | 1852 | |
e470259f | 1853 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 1854 | bond_3ad_unbind_slave(slave); |
1da177e4 | 1855 | |
ee637714 MB |
1856 | if (bond_mode_uses_xmit_hash(bond)) |
1857 | bond_update_slave_arr(bond, slave); | |
1858 | ||
76444f50 VF |
1859 | netdev_info(bond_dev, "Releasing %s interface %s\n", |
1860 | bond_is_active_slave(slave) ? "active" : "backup", | |
1861 | slave_dev->name); | |
1da177e4 | 1862 | |
4740d638 | 1863 | oldcurrent = rcu_access_pointer(bond->curr_active_slave); |
1da177e4 | 1864 | |
85741718 | 1865 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
1da177e4 | 1866 | |
00503b6f | 1867 | if (!all && (!bond->params.fail_over_mac || |
01844098 | 1868 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) { |
844223ab | 1869 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1870 | bond_has_slaves(bond)) |
76444f50 VF |
1871 | netdev_warn(bond_dev, "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", |
1872 | slave_dev->name, slave->perm_hwaddr, | |
1873 | bond_dev->name, slave_dev->name); | |
dec1e90e | 1874 | } |
1875 | ||
059b47e8 NA |
1876 | if (rtnl_dereference(bond->primary_slave) == slave) |
1877 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
1da177e4 | 1878 | |
1c72cfdc | 1879 | if (oldcurrent == slave) |
1da177e4 | 1880 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1881 | |
58402054 | 1882 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1883 | /* Must be called only after the slave has been |
1884 | * detached from the list and the curr_active_slave | |
1885 | * has been cleared (if our_slave == old_current), | |
1886 | * but before a new active slave is selected. | |
1887 | */ | |
1888 | bond_alb_deinit_slave(bond, slave); | |
1889 | } | |
1890 | ||
0896341a | 1891 | if (all) { |
36708b89 | 1892 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 1893 | } else if (oldcurrent == slave) { |
547942ca | 1894 | /* Note that we hold RTNL over this sequence, so there |
059fe7a5 JV |
1895 | * is no concern that another slave add/remove event |
1896 | * will interfere. | |
1897 | */ | |
1da177e4 | 1898 | bond_select_active_slave(bond); |
059fe7a5 JV |
1899 | } |
1900 | ||
0965a1f3 | 1901 | if (!bond_has_slaves(bond)) { |
ff59c456 | 1902 | bond_set_carrier(bond); |
409cc1f8 | 1903 | eth_hw_addr_random(bond_dev); |
1da177e4 LT |
1904 | } |
1905 | ||
e843fa50 | 1906 | unblock_netpoll_tx(); |
278b2083 | 1907 | synchronize_rcu(); |
ee6154e1 | 1908 | bond->slave_cnt--; |
1da177e4 | 1909 | |
0965a1f3 | 1910 | if (!bond_has_slaves(bond)) { |
2af73d4b | 1911 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1912 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1913 | } | |
2af73d4b | 1914 | |
b2a103e6 MM |
1915 | bond_compute_features(bond); |
1916 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1917 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
76444f50 VF |
1918 | netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n", |
1919 | slave_dev->name, bond_dev->name); | |
b2a103e6 | 1920 | |
1ff412ad | 1921 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1922 | |
547942ca | 1923 | /* If the mode uses primary, then this case was handled above by |
303d1cbf | 1924 | * bond_change_active_slave(..., NULL) |
1da177e4 | 1925 | */ |
ec0865a9 | 1926 | if (!bond_uses_primary(bond)) { |
5a0068de NH |
1927 | /* unset promiscuity level from slave |
1928 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
1929 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
1930 | * value of that flag before that change, as that was the value | |
1931 | * when this slave was attached, so we cache at the start of the | |
1932 | * function and use it here. Same goes for ALLMULTI below | |
1933 | */ | |
1934 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 1935 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1936 | |
1937 | /* unset allmulti level from slave */ | |
5a0068de | 1938 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 1939 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1940 | |
303d1cbf | 1941 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1942 | } |
1943 | ||
8a8efa22 | 1944 | slave_disable_netpoll(slave); |
f6dc31a8 | 1945 | |
1da177e4 LT |
1946 | /* close slave before restoring its mac address */ |
1947 | dev_close(slave_dev); | |
1948 | ||
00503b6f | 1949 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
01844098 | 1950 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 | 1951 | /* restore original ("permanent") mac address */ |
ada0f863 | 1952 | ether_addr_copy(addr.sa_data, slave->perm_hwaddr); |
2ab82852 MS |
1953 | addr.sa_family = slave_dev->type; |
1954 | dev_set_mac_address(slave_dev, &addr); | |
1955 | } | |
1da177e4 | 1956 | |
b15ba0fb JP |
1957 | dev_set_mtu(slave_dev, slave->original_mtu); |
1958 | ||
2d7011ca | 1959 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 | 1960 | |
3fdddd85 | 1961 | bond_free_slave(slave); |
1da177e4 | 1962 | |
547942ca | 1963 | return 0; |
1da177e4 LT |
1964 | } |
1965 | ||
0896341a | 1966 | /* A wrapper used because of ndo_del_link */ |
1967 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1968 | { | |
1969 | return __bond_release_one(bond_dev, slave_dev, false); | |
1970 | } | |
1971 | ||
547942ca NA |
1972 | /* First release a slave and then destroy the bond if no more slaves are left. |
1973 | * Must be under rtnl_lock when this function is called. | |
1974 | */ | |
26d8ee75 | 1975 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1976 | struct net_device *slave_dev) | |
d90a162a | 1977 | { |
454d7c9b | 1978 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1979 | int ret; |
1980 | ||
1981 | ret = bond_release(bond_dev, slave_dev); | |
0965a1f3 | 1982 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 1983 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
76444f50 VF |
1984 | netdev_info(bond_dev, "Destroying bond %s\n", |
1985 | bond_dev->name); | |
06f6d109 | 1986 | bond_remove_proc_entry(bond); |
9e71626c | 1987 | unregister_netdevice(bond_dev); |
d90a162a MS |
1988 | } |
1989 | return ret; | |
1990 | } | |
1991 | ||
1da177e4 LT |
1992 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1993 | { | |
454d7c9b | 1994 | struct bonding *bond = netdev_priv(bond_dev); |
69a2338e | 1995 | bond_fill_ifbond(bond, info); |
1da177e4 LT |
1996 | return 0; |
1997 | } | |
1998 | ||
1999 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2000 | { | |
454d7c9b | 2001 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 2002 | struct list_head *iter; |
dec1e90e | 2003 | int i = 0, res = -ENODEV; |
1da177e4 | 2004 | struct slave *slave; |
1da177e4 | 2005 | |
9caff1e7 | 2006 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 2007 | if (i++ == (int)info->slave_id) { |
689c96cc | 2008 | res = 0; |
69a2338e | 2009 | bond_fill_ifslave(slave, info); |
1da177e4 LT |
2010 | break; |
2011 | } | |
2012 | } | |
1da177e4 | 2013 | |
689c96cc | 2014 | return res; |
1da177e4 LT |
2015 | } |
2016 | ||
2017 | /*-------------------------------- Monitoring -------------------------------*/ | |
2018 | ||
4740d638 | 2019 | /* called with rcu_read_lock() */ |
f0c76d61 JV |
2020 | static int bond_miimon_inspect(struct bonding *bond) |
2021 | { | |
dec1e90e | 2022 | int link_state, commit = 0; |
9caff1e7 | 2023 | struct list_head *iter; |
f0c76d61 | 2024 | struct slave *slave; |
41f89100 JP |
2025 | bool ignore_updelay; |
2026 | ||
4740d638 | 2027 | ignore_updelay = !rcu_dereference(bond->curr_active_slave); |
1da177e4 | 2028 | |
4cb4f97b | 2029 | bond_for_each_slave_rcu(bond, slave, iter) { |
f0c76d61 | 2030 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2031 | |
f0c76d61 | 2032 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2033 | |
2034 | switch (slave->link) { | |
f0c76d61 JV |
2035 | case BOND_LINK_UP: |
2036 | if (link_state) | |
2037 | continue; | |
1da177e4 | 2038 | |
5d397061 JP |
2039 | bond_set_slave_link_state(slave, BOND_LINK_FAIL, |
2040 | BOND_SLAVE_NOTIFY_LATER); | |
f0c76d61 JV |
2041 | slave->delay = bond->params.downdelay; |
2042 | if (slave->delay) { | |
76444f50 VF |
2043 | netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n", |
2044 | (BOND_MODE(bond) == | |
2045 | BOND_MODE_ACTIVEBACKUP) ? | |
2046 | (bond_is_active_slave(slave) ? | |
2047 | "active " : "backup ") : "", | |
2048 | slave->dev->name, | |
2049 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2050 | } |
f0c76d61 JV |
2051 | /*FALLTHRU*/ |
2052 | case BOND_LINK_FAIL: | |
2053 | if (link_state) { | |
547942ca | 2054 | /* recovered before downdelay expired */ |
5d397061 JP |
2055 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2056 | BOND_SLAVE_NOTIFY_LATER); | |
8e603460 | 2057 | slave->last_link_up = jiffies; |
76444f50 VF |
2058 | netdev_info(bond->dev, "link status up again after %d ms for interface %s\n", |
2059 | (bond->params.downdelay - slave->delay) * | |
2060 | bond->params.miimon, | |
2061 | slave->dev->name); | |
f0c76d61 | 2062 | continue; |
1da177e4 | 2063 | } |
f0c76d61 JV |
2064 | |
2065 | if (slave->delay <= 0) { | |
2066 | slave->new_link = BOND_LINK_DOWN; | |
2067 | commit++; | |
2068 | continue; | |
1da177e4 | 2069 | } |
1da177e4 | 2070 | |
f0c76d61 JV |
2071 | slave->delay--; |
2072 | break; | |
2073 | ||
2074 | case BOND_LINK_DOWN: | |
2075 | if (!link_state) | |
2076 | continue; | |
2077 | ||
5d397061 JP |
2078 | bond_set_slave_link_state(slave, BOND_LINK_BACK, |
2079 | BOND_SLAVE_NOTIFY_LATER); | |
f0c76d61 JV |
2080 | slave->delay = bond->params.updelay; |
2081 | ||
2082 | if (slave->delay) { | |
76444f50 VF |
2083 | netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n", |
2084 | slave->dev->name, | |
2085 | ignore_updelay ? 0 : | |
2086 | bond->params.updelay * | |
2087 | bond->params.miimon); | |
f0c76d61 JV |
2088 | } |
2089 | /*FALLTHRU*/ | |
2090 | case BOND_LINK_BACK: | |
2091 | if (!link_state) { | |
69a2338e | 2092 | bond_set_slave_link_state(slave, |
5d397061 JP |
2093 | BOND_LINK_DOWN, |
2094 | BOND_SLAVE_NOTIFY_LATER); | |
76444f50 VF |
2095 | netdev_info(bond->dev, "link status down again after %d ms for interface %s\n", |
2096 | (bond->params.updelay - slave->delay) * | |
2097 | bond->params.miimon, | |
2098 | slave->dev->name); | |
f0c76d61 JV |
2099 | |
2100 | continue; | |
2101 | } | |
2102 | ||
41f89100 JP |
2103 | if (ignore_updelay) |
2104 | slave->delay = 0; | |
2105 | ||
f0c76d61 JV |
2106 | if (slave->delay <= 0) { |
2107 | slave->new_link = BOND_LINK_UP; | |
2108 | commit++; | |
41f89100 | 2109 | ignore_updelay = false; |
f0c76d61 | 2110 | continue; |
1da177e4 | 2111 | } |
f0c76d61 JV |
2112 | |
2113 | slave->delay--; | |
1da177e4 | 2114 | break; |
f0c76d61 JV |
2115 | } |
2116 | } | |
1da177e4 | 2117 | |
f0c76d61 JV |
2118 | return commit; |
2119 | } | |
1da177e4 | 2120 | |
f0c76d61 JV |
2121 | static void bond_miimon_commit(struct bonding *bond) |
2122 | { | |
9caff1e7 | 2123 | struct list_head *iter; |
059b47e8 | 2124 | struct slave *slave, *primary; |
f0c76d61 | 2125 | |
9caff1e7 | 2126 | bond_for_each_slave(bond, slave, iter) { |
f0c76d61 JV |
2127 | switch (slave->new_link) { |
2128 | case BOND_LINK_NOCHANGE: | |
2129 | continue; | |
1da177e4 | 2130 | |
f0c76d61 | 2131 | case BOND_LINK_UP: |
266b495f | 2132 | bond_update_speed_duplex(slave); |
5d397061 JP |
2133 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2134 | BOND_SLAVE_NOTIFY_NOW); | |
8e603460 | 2135 | slave->last_link_up = jiffies; |
f0c76d61 | 2136 | |
059b47e8 | 2137 | primary = rtnl_dereference(bond->primary_slave); |
01844098 | 2138 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
f0c76d61 | 2139 | /* prevent it from being the active one */ |
e30bc066 | 2140 | bond_set_backup_slave(slave); |
01844098 | 2141 | } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
f0c76d61 | 2142 | /* make it immediately active */ |
e30bc066 | 2143 | bond_set_active_slave(slave); |
059b47e8 | 2144 | } else if (slave != primary) { |
f0c76d61 | 2145 | /* prevent it from being the active one */ |
e30bc066 | 2146 | bond_set_backup_slave(slave); |
1da177e4 | 2147 | } |
1da177e4 | 2148 | |
76444f50 VF |
2149 | netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n", |
2150 | slave->dev->name, | |
2151 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, | |
2152 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2153 | |
f0c76d61 | 2154 | /* notify ad that the link status has changed */ |
01844098 | 2155 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 | 2156 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); |
059fe7a5 | 2157 | |
58402054 | 2158 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2159 | bond_alb_handle_link_change(bond, slave, |
2160 | BOND_LINK_UP); | |
1da177e4 | 2161 | |
ee637714 MB |
2162 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2163 | bond_update_slave_arr(bond, NULL); | |
2164 | ||
059b47e8 | 2165 | if (!bond->curr_active_slave || slave == primary) |
f0c76d61 | 2166 | goto do_failover; |
1da177e4 | 2167 | |
f0c76d61 | 2168 | continue; |
059fe7a5 | 2169 | |
f0c76d61 | 2170 | case BOND_LINK_DOWN: |
fba4acda JV |
2171 | if (slave->link_failure_count < UINT_MAX) |
2172 | slave->link_failure_count++; | |
2173 | ||
5d397061 JP |
2174 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2175 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 2176 | |
01844098 VF |
2177 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || |
2178 | BOND_MODE(bond) == BOND_MODE_8023AD) | |
5e5b0665 | 2179 | bond_set_slave_inactive_flags(slave, |
2180 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2181 | |
76444f50 VF |
2182 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2183 | slave->dev->name); | |
f0c76d61 | 2184 | |
01844098 | 2185 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 JV |
2186 | bond_3ad_handle_link_change(slave, |
2187 | BOND_LINK_DOWN); | |
2188 | ||
ae63e808 | 2189 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2190 | bond_alb_handle_link_change(bond, slave, |
2191 | BOND_LINK_DOWN); | |
2192 | ||
ee637714 MB |
2193 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2194 | bond_update_slave_arr(bond, NULL); | |
2195 | ||
4740d638 | 2196 | if (slave == rcu_access_pointer(bond->curr_active_slave)) |
f0c76d61 JV |
2197 | goto do_failover; |
2198 | ||
2199 | continue; | |
2200 | ||
2201 | default: | |
76444f50 VF |
2202 | netdev_err(bond->dev, "invalid new link %d on slave %s\n", |
2203 | slave->new_link, slave->dev->name); | |
f0c76d61 JV |
2204 | slave->new_link = BOND_LINK_NOCHANGE; |
2205 | ||
2206 | continue; | |
2207 | } | |
2208 | ||
2209 | do_failover: | |
e843fa50 | 2210 | block_netpoll_tx(); |
f0c76d61 | 2211 | bond_select_active_slave(bond); |
e843fa50 | 2212 | unblock_netpoll_tx(); |
f0c76d61 JV |
2213 | } |
2214 | ||
2215 | bond_set_carrier(bond); | |
1da177e4 LT |
2216 | } |
2217 | ||
547942ca | 2218 | /* bond_mii_monitor |
0b0eef66 JV |
2219 | * |
2220 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2221 | * inspection, then (if inspection indicates something needs to be done) |
2222 | * an acquisition of appropriate locks followed by a commit phase to | |
2223 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2224 | */ |
6da67d26 | 2225 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2226 | { |
2227 | struct bonding *bond = container_of(work, struct bonding, | |
2228 | mii_work.work); | |
ad246c99 | 2229 | bool should_notify_peers = false; |
1f2cd845 | 2230 | unsigned long delay; |
0b0eef66 | 2231 | |
1f2cd845 DM |
2232 | delay = msecs_to_jiffies(bond->params.miimon); |
2233 | ||
2234 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2235 | goto re_arm; |
b59f9f74 | 2236 | |
4cb4f97b | 2237 | rcu_read_lock(); |
2238 | ||
ad246c99 BH |
2239 | should_notify_peers = bond_should_notify_peers(bond); |
2240 | ||
1f2cd845 | 2241 | if (bond_miimon_inspect(bond)) { |
4cb4f97b | 2242 | rcu_read_unlock(); |
f0c76d61 | 2243 | |
1f2cd845 DM |
2244 | /* Race avoidance with bond_close cancel of workqueue */ |
2245 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2246 | delay = 1; |
2247 | should_notify_peers = false; | |
2248 | goto re_arm; | |
2249 | } | |
6b6c5261 | 2250 | |
1f2cd845 DM |
2251 | bond_miimon_commit(bond); |
2252 | ||
1f2cd845 | 2253 | rtnl_unlock(); /* might sleep, hold no other locks */ |
4cb4f97b | 2254 | } else |
2255 | rcu_read_unlock(); | |
0b0eef66 | 2256 | |
f0c76d61 | 2257 | re_arm: |
e6d265e8 | 2258 | if (bond->params.miimon) |
1f2cd845 DM |
2259 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2260 | ||
1f2cd845 DM |
2261 | if (should_notify_peers) { |
2262 | if (!rtnl_trylock()) | |
2263 | return; | |
2264 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2265 | rtnl_unlock(); | |
2266 | } | |
0b0eef66 | 2267 | } |
c3ade5ca | 2268 | |
50223ce4 | 2269 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2270 | { |
50223ce4 VF |
2271 | struct net_device *upper; |
2272 | struct list_head *iter; | |
2273 | bool ret = false; | |
f5b2b966 | 2274 | |
eaddcd76 | 2275 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2276 | return true; |
f5b2b966 | 2277 | |
50223ce4 | 2278 | rcu_read_lock(); |
2f268f12 | 2279 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2280 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2281 | ret = true; | |
2282 | break; | |
2283 | } | |
f5b2b966 | 2284 | } |
50223ce4 | 2285 | rcu_read_unlock(); |
f5b2b966 | 2286 | |
50223ce4 | 2287 | return ret; |
f5b2b966 JV |
2288 | } |
2289 | ||
547942ca | 2290 | /* We go to the (large) trouble of VLAN tagging ARP frames because |
c3ade5ca JV |
2291 | * switches in VLAN mode (especially if ports are configured as |
2292 | * "native" to a VLAN) might not pass non-tagged frames. | |
2293 | */ | |
fbd929f2 | 2294 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, |
2295 | __be32 dest_ip, __be32 src_ip, | |
44a40855 | 2296 | struct bond_vlan_tag *tags) |
c3ade5ca JV |
2297 | { |
2298 | struct sk_buff *skb; | |
3e403a77 | 2299 | struct bond_vlan_tag *outer_tag = tags; |
c3ade5ca | 2300 | |
76444f50 VF |
2301 | netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n", |
2302 | arp_op, slave_dev->name, &dest_ip, &src_ip); | |
3d632c3f | 2303 | |
c3ade5ca JV |
2304 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2305 | NULL, slave_dev->dev_addr, NULL); | |
2306 | ||
2307 | if (!skb) { | |
4873ac3c | 2308 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2309 | return; |
2310 | } | |
fbd929f2 | 2311 | |
3e403a77 VF |
2312 | if (!tags || tags->vlan_proto == VLAN_N_VID) |
2313 | goto xmit; | |
2314 | ||
2315 | tags++; | |
2316 | ||
44a40855 | 2317 | /* Go through all the tags backwards and add them to the packet */ |
3e403a77 VF |
2318 | while (tags->vlan_proto != VLAN_N_VID) { |
2319 | if (!tags->vlan_id) { | |
2320 | tags++; | |
44a40855 | 2321 | continue; |
3e403a77 | 2322 | } |
44a40855 | 2323 | |
76444f50 | 2324 | netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n", |
3e403a77 | 2325 | ntohs(outer_tag->vlan_proto), tags->vlan_id); |
62749e2c JP |
2326 | skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto, |
2327 | tags->vlan_id); | |
44a40855 VY |
2328 | if (!skb) { |
2329 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2330 | return; | |
2331 | } | |
3e403a77 VF |
2332 | |
2333 | tags++; | |
44a40855 VY |
2334 | } |
2335 | /* Set the outer tag */ | |
3e403a77 | 2336 | if (outer_tag->vlan_id) { |
76444f50 | 2337 | netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n", |
3e403a77 | 2338 | ntohs(outer_tag->vlan_proto), outer_tag->vlan_id); |
b4bef1b5 JP |
2339 | __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto, |
2340 | outer_tag->vlan_id); | |
c3ade5ca | 2341 | } |
3e403a77 VF |
2342 | |
2343 | xmit: | |
c3ade5ca JV |
2344 | arp_xmit(skb); |
2345 | } | |
2346 | ||
44a40855 VY |
2347 | /* Validate the device path between the @start_dev and the @end_dev. |
2348 | * The path is valid if the @end_dev is reachable through device | |
2349 | * stacking. | |
2350 | * When the path is validated, collect any vlan information in the | |
2351 | * path. | |
2352 | */ | |
3e403a77 VF |
2353 | struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, |
2354 | struct net_device *end_dev, | |
2355 | int level) | |
44a40855 | 2356 | { |
3e403a77 | 2357 | struct bond_vlan_tag *tags; |
44a40855 VY |
2358 | struct net_device *upper; |
2359 | struct list_head *iter; | |
44a40855 | 2360 | |
3e403a77 VF |
2361 | if (start_dev == end_dev) { |
2362 | tags = kzalloc(sizeof(*tags) * (level + 1), GFP_ATOMIC); | |
2363 | if (!tags) | |
2364 | return ERR_PTR(-ENOMEM); | |
2365 | tags[level].vlan_proto = VLAN_N_VID; | |
2366 | return tags; | |
2367 | } | |
44a40855 VY |
2368 | |
2369 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
3e403a77 VF |
2370 | tags = bond_verify_device_path(upper, end_dev, level + 1); |
2371 | if (IS_ERR_OR_NULL(tags)) { | |
2372 | if (IS_ERR(tags)) | |
2373 | return tags; | |
2374 | continue; | |
44a40855 | 2375 | } |
3e403a77 VF |
2376 | if (is_vlan_dev(upper)) { |
2377 | tags[level].vlan_proto = vlan_dev_vlan_proto(upper); | |
2378 | tags[level].vlan_id = vlan_dev_vlan_id(upper); | |
2379 | } | |
2380 | ||
2381 | return tags; | |
44a40855 VY |
2382 | } |
2383 | ||
3e403a77 | 2384 | return NULL; |
44a40855 | 2385 | } |
c3ade5ca | 2386 | |
1da177e4 LT |
2387 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2388 | { | |
c3ade5ca | 2389 | struct rtable *rt; |
3e403a77 | 2390 | struct bond_vlan_tag *tags; |
27bc11e6 | 2391 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2392 | int i; |
1da177e4 | 2393 | |
27bc11e6 | 2394 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
76444f50 | 2395 | netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]); |
3e403a77 | 2396 | tags = NULL; |
c3ade5ca | 2397 | |
27bc11e6 | 2398 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2399 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2400 | RTO_ONLINK, 0); | |
b23dd4fe | 2401 | if (IS_ERR(rt)) { |
28572760 VF |
2402 | /* there's no route to target - try to send arp |
2403 | * probe to generate any traffic (arp_validate=0) | |
2404 | */ | |
4873ac3c | 2405 | if (bond->params.arp_validate) |
2406 | net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", | |
2407 | bond->dev->name, | |
2408 | &targets[i]); | |
44a40855 VY |
2409 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2410 | 0, tags); | |
c3ade5ca JV |
2411 | continue; |
2412 | } | |
2413 | ||
27bc11e6 VF |
2414 | /* bond device itself */ |
2415 | if (rt->dst.dev == bond->dev) | |
2416 | goto found; | |
2417 | ||
2418 | rcu_read_lock(); | |
3e403a77 | 2419 | tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0); |
27bc11e6 | 2420 | rcu_read_unlock(); |
c3ade5ca | 2421 | |
3e403a77 | 2422 | if (!IS_ERR_OR_NULL(tags)) |
44a40855 VY |
2423 | goto found; |
2424 | ||
27bc11e6 | 2425 | /* Not our device - skip */ |
76444f50 VF |
2426 | netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n", |
2427 | &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
3e32582f | 2428 | |
ed4b9f80 | 2429 | ip_rt_put(rt); |
27bc11e6 VF |
2430 | continue; |
2431 | ||
2432 | found: | |
2433 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2434 | ip_rt_put(rt); | |
2435 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
44a40855 | 2436 | addr, tags); |
a67eed57 | 2437 | kfree(tags); |
c3ade5ca JV |
2438 | } |
2439 | } | |
2440 | ||
d3bb52b0 | 2441 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2442 | { |
8599b52e VF |
2443 | int i; |
2444 | ||
87a7b84b | 2445 | if (!sip || !bond_has_this_ip(bond, tip)) { |
76444f50 VF |
2446 | netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n", |
2447 | &sip, &tip); | |
87a7b84b VF |
2448 | return; |
2449 | } | |
f5b2b966 | 2450 | |
8599b52e VF |
2451 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2452 | if (i == -1) { | |
76444f50 VF |
2453 | netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n", |
2454 | &sip); | |
87a7b84b | 2455 | return; |
f5b2b966 | 2456 | } |
49f17de7 | 2457 | slave->last_rx = jiffies; |
8599b52e | 2458 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2459 | } |
2460 | ||
5bb9e0b5 | 2461 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2462 | struct slave *slave) | |
f5b2b966 | 2463 | { |
de063b70 | 2464 | struct arphdr *arp = (struct arphdr *)skb->data; |
21a75f09 | 2465 | struct slave *curr_active_slave, *curr_arp_slave; |
f5b2b966 | 2466 | unsigned char *arp_ptr; |
d3bb52b0 | 2467 | __be32 sip, tip; |
f2cb691a | 2468 | int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
f5b2b966 | 2469 | |
f2cb691a | 2470 | if (!slave_do_arp_validate(bond, slave)) { |
bedabf90 | 2471 | if ((slave_do_arp_validate_only(bond) && is_arp) || |
2472 | !slave_do_arp_validate_only(bond)) | |
49f17de7 | 2473 | slave->last_rx = jiffies; |
13a8e0c8 | 2474 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2475 | } else if (!is_arp) { |
2476 | return RX_HANDLER_ANOTHER; | |
2477 | } | |
2c146102 | 2478 | |
de063b70 | 2479 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2480 | |
76444f50 VF |
2481 | netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n", |
2482 | skb->dev->name); | |
f5b2b966 | 2483 | |
de063b70 ED |
2484 | if (alen > skb_headlen(skb)) { |
2485 | arp = kmalloc(alen, GFP_ATOMIC); | |
2486 | if (!arp) | |
2487 | goto out_unlock; | |
2488 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2489 | goto out_unlock; | |
2490 | } | |
f5b2b966 | 2491 | |
3aba891d | 2492 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2493 | skb->pkt_type == PACKET_OTHERHOST || |
2494 | skb->pkt_type == PACKET_LOOPBACK || | |
2495 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2496 | arp->ar_pro != htons(ETH_P_IP) || | |
2497 | arp->ar_pln != 4) | |
2498 | goto out_unlock; | |
2499 | ||
2500 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2501 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2502 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2503 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2504 | memcpy(&tip, arp_ptr, 4); |
2505 | ||
76444f50 VF |
2506 | netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
2507 | slave->dev->name, bond_slave_state(slave), | |
2508 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2509 | &sip, &tip); | |
f5b2b966 | 2510 | |
010d3c39 | 2511 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
21a75f09 | 2512 | curr_arp_slave = rcu_dereference(bond->current_arp_slave); |
010d3c39 | 2513 | |
21a75f09 | 2514 | /* We 'trust' the received ARP enough to validate it if: |
aeea64ac | 2515 | * |
21a75f09 JV |
2516 | * (a) the slave receiving the ARP is active (which includes the |
2517 | * current ARP slave, if any), or | |
2518 | * | |
2519 | * (b) the receiving slave isn't active, but there is a currently | |
2520 | * active slave and it received valid arp reply(s) after it became | |
2521 | * the currently active slave, or | |
2522 | * | |
2523 | * (c) there is an ARP slave that sent an ARP during the prior ARP | |
2524 | * interval, and we receive an ARP reply on any slave. We accept | |
2525 | * these because switch FDB update delays may deliver the ARP | |
2526 | * reply to a slave other than the sender of the ARP request. | |
2527 | * | |
2528 | * Note: for (b), backup slaves are receiving the broadcast ARP | |
2529 | * request, not a reply. This request passes from the sending | |
2530 | * slave through the L2 switch(es) to the receiving slave. Since | |
2531 | * this is checking the request, sip/tip are swapped for | |
2532 | * validation. | |
2533 | * | |
2534 | * This is done to avoid endless looping when we can't reach the | |
aeea64ac | 2535 | * arp_ip_target and fool ourselves with our own arp requests. |
f5b2b966 | 2536 | */ |
e30bc066 | 2537 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2538 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
2539 | else if (curr_active_slave && |
2540 | time_after(slave_last_rx(bond, curr_active_slave), | |
2541 | curr_active_slave->last_link_up)) | |
f5b2b966 | 2542 | bond_validate_arp(bond, slave, tip, sip); |
21a75f09 JV |
2543 | else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) && |
2544 | bond_time_in_interval(bond, | |
2545 | dev_trans_start(curr_arp_slave->dev), 1)) | |
2546 | bond_validate_arp(bond, slave, sip, tip); | |
f5b2b966 JV |
2547 | |
2548 | out_unlock: | |
de063b70 ED |
2549 | if (arp != (struct arphdr *)skb->data) |
2550 | kfree(arp); | |
13a8e0c8 | 2551 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2552 | } |
2553 | ||
e7f63f1d VF |
2554 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2555 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2556 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2557 | */ | |
2558 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2559 | int mod) | |
2560 | { | |
2561 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2562 | ||
2563 | return time_in_range(jiffies, | |
2564 | last_act - delta_in_ticks, | |
2565 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2566 | } | |
2567 | ||
547942ca | 2568 | /* This function is called regularly to monitor each slave's link |
1da177e4 LT |
2569 | * ensuring that traffic is being sent and received when arp monitoring |
2570 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2571 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2572 | * arp monitoring in active backup mode. | |
2573 | */ | |
6da67d26 | 2574 | static void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2575 | { |
1b76b316 JV |
2576 | struct bonding *bond = container_of(work, struct bonding, |
2577 | arp_work.work); | |
1da177e4 | 2578 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2579 | struct list_head *iter; |
6fde8f03 | 2580 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 2581 | |
1f2cd845 | 2582 | if (!bond_has_slaves(bond)) |
1da177e4 | 2583 | goto re_arm; |
1da177e4 | 2584 | |
2e52f4fe | 2585 | rcu_read_lock(); |
2586 | ||
4740d638 | 2587 | oldcurrent = rcu_dereference(bond->curr_active_slave); |
1da177e4 LT |
2588 | /* see if any of the previous devices are up now (i.e. they have |
2589 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
2590 | * the picture unless it is null. also, slave->last_link_up is not |
2591 | * needed here because we send an arp on each slave and give a slave | |
2592 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
2593 | * TODO: what about up/down delay in arp mode? it wasn't here before |
2594 | * so it can wait | |
2595 | */ | |
2e52f4fe | 2596 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2597 | unsigned long trans_start = dev_trans_start(slave->dev); |
2598 | ||
1da177e4 | 2599 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2600 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 2601 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 LT |
2602 | |
2603 | slave->link = BOND_LINK_UP; | |
6fde8f03 | 2604 | slave_state_changed = 1; |
1da177e4 LT |
2605 | |
2606 | /* primary_slave has no meaning in round-robin | |
2607 | * mode. the window of a slave being up and | |
2608 | * curr_active_slave being null after enslaving | |
2609 | * is closed. | |
2610 | */ | |
2611 | if (!oldcurrent) { | |
76444f50 VF |
2612 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2613 | slave->dev->name); | |
1da177e4 LT |
2614 | do_failover = 1; |
2615 | } else { | |
76444f50 VF |
2616 | netdev_info(bond->dev, "interface %s is now up\n", |
2617 | slave->dev->name); | |
1da177e4 LT |
2618 | } |
2619 | } | |
2620 | } else { | |
2621 | /* slave->link == BOND_LINK_UP */ | |
2622 | ||
2623 | /* not all switches will respond to an arp request | |
2624 | * when the source ip is 0, so don't take the link down | |
2625 | * if we don't know our ip yet | |
2626 | */ | |
e7f63f1d | 2627 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 2628 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 LT |
2629 | |
2630 | slave->link = BOND_LINK_DOWN; | |
6fde8f03 | 2631 | slave_state_changed = 1; |
1da177e4 | 2632 | |
3d632c3f | 2633 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2634 | slave->link_failure_count++; |
1da177e4 | 2635 | |
76444f50 VF |
2636 | netdev_info(bond->dev, "interface %s is now down\n", |
2637 | slave->dev->name); | |
1da177e4 | 2638 | |
3d632c3f | 2639 | if (slave == oldcurrent) |
1da177e4 | 2640 | do_failover = 1; |
1da177e4 LT |
2641 | } |
2642 | } | |
2643 | ||
2644 | /* note: if switch is in round-robin mode, all links | |
2645 | * must tx arp to ensure all links rx an arp - otherwise | |
2646 | * links may oscillate or not come up at all; if switch is | |
2647 | * in something like xor mode, there is nothing we can | |
2648 | * do - all replies will be rx'ed on same link causing slaves | |
2649 | * to be unstable during low/no traffic periods | |
2650 | */ | |
b6adc610 | 2651 | if (bond_slave_is_up(slave)) |
1da177e4 | 2652 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2653 | } |
2654 | ||
2e52f4fe | 2655 | rcu_read_unlock(); |
2656 | ||
6fde8f03 | 2657 | if (do_failover || slave_state_changed) { |
2e52f4fe | 2658 | if (!rtnl_trylock()) |
2659 | goto re_arm; | |
1da177e4 | 2660 | |
6fde8f03 DT |
2661 | if (slave_state_changed) { |
2662 | bond_slave_state_change(bond); | |
ee637714 MB |
2663 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2664 | bond_update_slave_arr(bond, NULL); | |
b8e4500f NA |
2665 | } |
2666 | if (do_failover) { | |
6fde8f03 | 2667 | block_netpoll_tx(); |
6fde8f03 | 2668 | bond_select_active_slave(bond); |
6fde8f03 DT |
2669 | unblock_netpoll_tx(); |
2670 | } | |
2e52f4fe | 2671 | rtnl_unlock(); |
1da177e4 LT |
2672 | } |
2673 | ||
2674 | re_arm: | |
e6d265e8 | 2675 | if (bond->params.arp_interval) |
e7f63f1d VF |
2676 | queue_delayed_work(bond->wq, &bond->arp_work, |
2677 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2678 | } |
2679 | ||
547942ca | 2680 | /* Called to inspect slaves for active-backup mode ARP monitor link state |
b2220cad JV |
2681 | * changes. Sets new_link in slaves to specify what action should take |
2682 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2683 | * to link states must be committed. | |
2684 | * | |
8c0bc550 | 2685 | * Called with rcu_read_lock held. |
1da177e4 | 2686 | */ |
e7f63f1d | 2687 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2688 | { |
def4460c | 2689 | unsigned long trans_start, last_rx; |
9caff1e7 | 2690 | struct list_head *iter; |
dec1e90e | 2691 | struct slave *slave; |
dec1e90e | 2692 | int commit = 0; |
da210f55 | 2693 | |
eb9fa4b0 | 2694 | bond_for_each_slave_rcu(bond, slave, iter) { |
b2220cad | 2695 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2696 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2697 | |
b2220cad | 2698 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2699 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2700 | slave->new_link = BOND_LINK_UP; |
2701 | commit++; | |
2702 | } | |
b2220cad JV |
2703 | continue; |
2704 | } | |
1da177e4 | 2705 | |
547942ca | 2706 | /* Give slaves 2*delta after being enslaved or made |
b2220cad JV |
2707 | * active. This avoids bouncing, as the last receive |
2708 | * times need a full ARP monitor cycle to be updated. | |
2709 | */ | |
8e603460 | 2710 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
2711 | continue; |
2712 | ||
547942ca | 2713 | /* Backup slave is down if: |
b2220cad JV |
2714 | * - No current_arp_slave AND |
2715 | * - more than 3*delta since last receive AND | |
2716 | * - the bond has an IP address | |
2717 | * | |
2718 | * Note: a non-null current_arp_slave indicates | |
2719 | * the curr_active_slave went down and we are | |
2720 | * searching for a new one; under this condition | |
2721 | * we only take the curr_active_slave down - this | |
2722 | * gives each slave a chance to tx/rx traffic | |
2723 | * before being taken out | |
2724 | */ | |
e30bc066 | 2725 | if (!bond_is_active_slave(slave) && |
85741718 | 2726 | !rcu_access_pointer(bond->current_arp_slave) && |
e7f63f1d | 2727 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2728 | slave->new_link = BOND_LINK_DOWN; |
2729 | commit++; | |
2730 | } | |
2731 | ||
547942ca | 2732 | /* Active slave is down if: |
b2220cad JV |
2733 | * - more than 2*delta since transmitting OR |
2734 | * - (more than 2*delta since receive AND | |
2735 | * the bond has an IP address) | |
2736 | */ | |
cb32f2a0 | 2737 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2738 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2739 | (!bond_time_in_interval(bond, trans_start, 2) || |
2740 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2741 | slave->new_link = BOND_LINK_DOWN; |
2742 | commit++; | |
2743 | } | |
1da177e4 LT |
2744 | } |
2745 | ||
b2220cad JV |
2746 | return commit; |
2747 | } | |
1da177e4 | 2748 | |
547942ca | 2749 | /* Called to commit link state changes noted by inspection step of |
b2220cad JV |
2750 | * active-backup mode ARP monitor. |
2751 | * | |
eb9fa4b0 | 2752 | * Called with RTNL hold. |
b2220cad | 2753 | */ |
e7f63f1d | 2754 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2755 | { |
cb32f2a0 | 2756 | unsigned long trans_start; |
9caff1e7 | 2757 | struct list_head *iter; |
dec1e90e | 2758 | struct slave *slave; |
1da177e4 | 2759 | |
9caff1e7 | 2760 | bond_for_each_slave(bond, slave, iter) { |
b2220cad JV |
2761 | switch (slave->new_link) { |
2762 | case BOND_LINK_NOCHANGE: | |
2763 | continue; | |
ff59c456 | 2764 | |
b2220cad | 2765 | case BOND_LINK_UP: |
cb32f2a0 | 2766 | trans_start = dev_trans_start(slave->dev); |
4740d638 ED |
2767 | if (rtnl_dereference(bond->curr_active_slave) != slave || |
2768 | (!rtnl_dereference(bond->curr_active_slave) && | |
e7f63f1d | 2769 | bond_time_in_interval(bond, trans_start, 1))) { |
85741718 ED |
2770 | struct slave *current_arp_slave; |
2771 | ||
2772 | current_arp_slave = rtnl_dereference(bond->current_arp_slave); | |
5d397061 JP |
2773 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2774 | BOND_SLAVE_NOTIFY_NOW); | |
85741718 | 2775 | if (current_arp_slave) { |
5a430974 | 2776 | bond_set_slave_inactive_flags( |
85741718 | 2777 | current_arp_slave, |
5e5b0665 | 2778 | BOND_SLAVE_NOTIFY_NOW); |
85741718 | 2779 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
5a430974 | 2780 | } |
b2220cad | 2781 | |
76444f50 VF |
2782 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2783 | slave->dev->name); | |
b2220cad | 2784 | |
4740d638 | 2785 | if (!rtnl_dereference(bond->curr_active_slave) || |
059b47e8 | 2786 | slave == rtnl_dereference(bond->primary_slave)) |
b9f60253 | 2787 | goto do_failover; |
1da177e4 | 2788 | |
b9f60253 | 2789 | } |
1da177e4 | 2790 | |
b9f60253 | 2791 | continue; |
1da177e4 | 2792 | |
b2220cad JV |
2793 | case BOND_LINK_DOWN: |
2794 | if (slave->link_failure_count < UINT_MAX) | |
2795 | slave->link_failure_count++; | |
2796 | ||
5d397061 JP |
2797 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2798 | BOND_SLAVE_NOTIFY_NOW); | |
5e5b0665 | 2799 | bond_set_slave_inactive_flags(slave, |
2800 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 2801 | |
76444f50 VF |
2802 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2803 | slave->dev->name); | |
b2220cad | 2804 | |
4740d638 | 2805 | if (slave == rtnl_dereference(bond->curr_active_slave)) { |
85741718 | 2806 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
b9f60253 | 2807 | goto do_failover; |
1da177e4 | 2808 | } |
b9f60253 JP |
2809 | |
2810 | continue; | |
b2220cad JV |
2811 | |
2812 | default: | |
76444f50 VF |
2813 | netdev_err(bond->dev, "impossible: new_link %d on slave %s\n", |
2814 | slave->new_link, slave->dev->name); | |
b9f60253 | 2815 | continue; |
1da177e4 | 2816 | } |
1da177e4 | 2817 | |
b9f60253 | 2818 | do_failover: |
e843fa50 | 2819 | block_netpoll_tx(); |
b9f60253 | 2820 | bond_select_active_slave(bond); |
e843fa50 | 2821 | unblock_netpoll_tx(); |
b2220cad | 2822 | } |
1da177e4 | 2823 | |
b2220cad JV |
2824 | bond_set_carrier(bond); |
2825 | } | |
1da177e4 | 2826 | |
547942ca | 2827 | /* Send ARP probes for active-backup mode ARP monitor. |
b0929915 | 2828 | * |
8c0bc550 | 2829 | * Called with rcu_read_lock held. |
b2220cad | 2830 | */ |
f2ebd477 | 2831 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 2832 | { |
eb9fa4b0 | 2833 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 2834 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
2835 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
2836 | struct list_head *iter; |
2837 | bool found = false; | |
b0929915 | 2838 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 2839 | |
98b90f26 | 2840 | if (curr_arp_slave && curr_active_slave) |
76444f50 VF |
2841 | netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n", |
2842 | curr_arp_slave->dev->name, | |
2843 | curr_active_slave->dev->name); | |
1da177e4 | 2844 | |
98b90f26 VF |
2845 | if (curr_active_slave) { |
2846 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 2847 | return should_notify_rtnl; |
b2220cad | 2848 | } |
1da177e4 | 2849 | |
b2220cad JV |
2850 | /* if we don't have a curr_active_slave, search for the next available |
2851 | * backup slave from the current_arp_slave and make it the candidate | |
2852 | * for becoming the curr_active_slave | |
2853 | */ | |
1da177e4 | 2854 | |
eb9fa4b0 | 2855 | if (!curr_arp_slave) { |
b0929915 | 2856 | curr_arp_slave = bond_first_slave_rcu(bond); |
2857 | if (!curr_arp_slave) | |
2858 | return should_notify_rtnl; | |
b2220cad | 2859 | } |
1da177e4 | 2860 | |
b0929915 | 2861 | bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER); |
059fe7a5 | 2862 | |
b0929915 | 2863 | bond_for_each_slave_rcu(bond, slave, iter) { |
b6adc610 | 2864 | if (!found && !before && bond_slave_is_up(slave)) |
4087df87 | 2865 | before = slave; |
1da177e4 | 2866 | |
b6adc610 | 2867 | if (found && !new_slave && bond_slave_is_up(slave)) |
4087df87 | 2868 | new_slave = slave; |
b2220cad JV |
2869 | /* if the link state is up at this point, we |
2870 | * mark it down - this can happen if we have | |
2871 | * simultaneous link failures and | |
2872 | * reselect_active_interface doesn't make this | |
2873 | * one the current slave so it is still marked | |
2874 | * up when it is actually down | |
1da177e4 | 2875 | */ |
b6adc610 | 2876 | if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
5d397061 JP |
2877 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2878 | BOND_SLAVE_NOTIFY_LATER); | |
b2220cad JV |
2879 | if (slave->link_failure_count < UINT_MAX) |
2880 | slave->link_failure_count++; | |
1da177e4 | 2881 | |
5e5b0665 | 2882 | bond_set_slave_inactive_flags(slave, |
b0929915 | 2883 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 2884 | |
76444f50 VF |
2885 | netdev_info(bond->dev, "backup interface %s is now down\n", |
2886 | slave->dev->name); | |
1da177e4 | 2887 | } |
eb9fa4b0 | 2888 | if (slave == curr_arp_slave) |
4087df87 | 2889 | found = true; |
b2220cad | 2890 | } |
4087df87 VF |
2891 | |
2892 | if (!new_slave && before) | |
2893 | new_slave = before; | |
2894 | ||
b0929915 | 2895 | if (!new_slave) |
2896 | goto check_state; | |
4087df87 | 2897 | |
5d397061 JP |
2898 | bond_set_slave_link_state(new_slave, BOND_LINK_BACK, |
2899 | BOND_SLAVE_NOTIFY_LATER); | |
b0929915 | 2900 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 2901 | bond_arp_send_all(bond, new_slave); |
8e603460 | 2902 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 2903 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 2904 | |
b0929915 | 2905 | check_state: |
2906 | bond_for_each_slave_rcu(bond, slave, iter) { | |
5d397061 | 2907 | if (slave->should_notify || slave->should_notify_link) { |
b0929915 | 2908 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; |
2909 | break; | |
2910 | } | |
2911 | } | |
2912 | return should_notify_rtnl; | |
b2220cad | 2913 | } |
1da177e4 | 2914 | |
6da67d26 | 2915 | static void bond_activebackup_arp_mon(struct work_struct *work) |
b2220cad JV |
2916 | { |
2917 | struct bonding *bond = container_of(work, struct bonding, | |
2918 | arp_work.work); | |
b0929915 | 2919 | bool should_notify_peers = false; |
2920 | bool should_notify_rtnl = false; | |
1f2cd845 | 2921 | int delta_in_ticks; |
1da177e4 | 2922 | |
1f2cd845 DM |
2923 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
2924 | ||
2925 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
2926 | goto re_arm; |
2927 | ||
eb9fa4b0 | 2928 | rcu_read_lock(); |
b0929915 | 2929 | |
ad246c99 BH |
2930 | should_notify_peers = bond_should_notify_peers(bond); |
2931 | ||
b0929915 | 2932 | if (bond_ab_arp_inspect(bond)) { |
2933 | rcu_read_unlock(); | |
2934 | ||
1f2cd845 DM |
2935 | /* Race avoidance with bond_close flush of workqueue */ |
2936 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2937 | delta_in_ticks = 1; |
2938 | should_notify_peers = false; | |
2939 | goto re_arm; | |
2940 | } | |
b2220cad | 2941 | |
1f2cd845 | 2942 | bond_ab_arp_commit(bond); |
b0929915 | 2943 | |
1f2cd845 | 2944 | rtnl_unlock(); |
b0929915 | 2945 | rcu_read_lock(); |
1f2cd845 DM |
2946 | } |
2947 | ||
b0929915 | 2948 | should_notify_rtnl = bond_ab_arp_probe(bond); |
2949 | rcu_read_unlock(); | |
ad246c99 | 2950 | |
80b9d236 | 2951 | re_arm: |
2952 | if (bond->params.arp_interval) | |
1f2cd845 DM |
2953 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
2954 | ||
b0929915 | 2955 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
2956 | if (!rtnl_trylock()) |
2957 | return; | |
b0929915 | 2958 | |
2959 | if (should_notify_peers) | |
2960 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
2961 | bond->dev); | |
5d397061 | 2962 | if (should_notify_rtnl) { |
b0929915 | 2963 | bond_slave_state_notify(bond); |
5d397061 JP |
2964 | bond_slave_link_notify(bond); |
2965 | } | |
b0929915 | 2966 | |
1f2cd845 DM |
2967 | rtnl_unlock(); |
2968 | } | |
1da177e4 LT |
2969 | } |
2970 | ||
1da177e4 LT |
2971 | /*-------------------------- netdev event handling --------------------------*/ |
2972 | ||
547942ca | 2973 | /* Change device name */ |
1da177e4 LT |
2974 | static int bond_event_changename(struct bonding *bond) |
2975 | { | |
1da177e4 LT |
2976 | bond_remove_proc_entry(bond); |
2977 | bond_create_proc_entry(bond); | |
7e083840 | 2978 | |
f073c7ca TI |
2979 | bond_debug_reregister(bond); |
2980 | ||
1da177e4 LT |
2981 | return NOTIFY_DONE; |
2982 | } | |
2983 | ||
3d632c3f SH |
2984 | static int bond_master_netdev_event(unsigned long event, |
2985 | struct net_device *bond_dev) | |
1da177e4 | 2986 | { |
454d7c9b | 2987 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2988 | |
2989 | switch (event) { | |
2990 | case NETDEV_CHANGENAME: | |
2991 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2992 | case NETDEV_UNREGISTER: |
2993 | bond_remove_proc_entry(event_bond); | |
2994 | break; | |
2995 | case NETDEV_REGISTER: | |
2996 | bond_create_proc_entry(event_bond); | |
2997 | break; | |
6c8d23f7 | 2998 | case NETDEV_NOTIFY_PEERS: |
2999 | if (event_bond->send_peer_notif) | |
3000 | event_bond->send_peer_notif--; | |
3001 | break; | |
1da177e4 LT |
3002 | default: |
3003 | break; | |
3004 | } | |
3005 | ||
3006 | return NOTIFY_DONE; | |
3007 | } | |
3008 | ||
3d632c3f SH |
3009 | static int bond_slave_netdev_event(unsigned long event, |
3010 | struct net_device *slave_dev) | |
1da177e4 | 3011 | { |
059b47e8 | 3012 | struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary; |
6101391d | 3013 | struct bonding *bond; |
3014 | struct net_device *bond_dev; | |
1da177e4 | 3015 | |
6101391d | 3016 | /* A netdev event can be generated while enslaving a device |
3017 | * before netdev_rx_handler_register is called in which case | |
3018 | * slave will be NULL | |
3019 | */ | |
3020 | if (!slave) | |
3021 | return NOTIFY_DONE; | |
3022 | bond_dev = slave->bond->dev; | |
3023 | bond = slave->bond; | |
059b47e8 | 3024 | primary = rtnl_dereference(bond->primary_slave); |
6101391d | 3025 | |
1da177e4 LT |
3026 | switch (event) { |
3027 | case NETDEV_UNREGISTER: | |
8d2ada77 | 3028 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
3029 | bond_release_and_destroy(bond_dev, slave_dev); |
3030 | else | |
3031 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 3032 | break; |
98f41f69 | 3033 | case NETDEV_UP: |
1da177e4 | 3034 | case NETDEV_CHANGE: |
471cb5a3 | 3035 | bond_update_speed_duplex(slave); |
52bc6716 MB |
3036 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3037 | bond_3ad_adapter_speed_duplex_changed(slave); | |
950ddcb1 MB |
3038 | /* Fallthrough */ |
3039 | case NETDEV_DOWN: | |
ee637714 MB |
3040 | /* Refresh slave-array if applicable! |
3041 | * If the setup does not use miimon or arpmon (mode-specific!), | |
3042 | * then these events will not cause the slave-array to be | |
3043 | * refreshed. This will cause xmit to use a slave that is not | |
3044 | * usable. Avoid such situation by refeshing the array at these | |
3045 | * events. If these (miimon/arpmon) parameters are configured | |
3046 | * then array gets refreshed twice and that should be fine! | |
3047 | */ | |
3048 | if (bond_mode_uses_xmit_hash(bond)) | |
3049 | bond_update_slave_arr(bond, NULL); | |
1da177e4 | 3050 | break; |
1da177e4 | 3051 | case NETDEV_CHANGEMTU: |
547942ca | 3052 | /* TODO: Should slaves be allowed to |
1da177e4 LT |
3053 | * independently alter their MTU? For |
3054 | * an active-backup bond, slaves need | |
3055 | * not be the same type of device, so | |
3056 | * MTUs may vary. For other modes, | |
3057 | * slaves arguably should have the | |
3058 | * same MTUs. To do this, we'd need to | |
3059 | * take over the slave's change_mtu | |
3060 | * function for the duration of their | |
3061 | * servitude. | |
3062 | */ | |
3063 | break; | |
3064 | case NETDEV_CHANGENAME: | |
3ec775b9 | 3065 | /* we don't care if we don't have primary set */ |
ec0865a9 | 3066 | if (!bond_uses_primary(bond) || |
3ec775b9 VF |
3067 | !bond->params.primary[0]) |
3068 | break; | |
3069 | ||
059b47e8 | 3070 | if (slave == primary) { |
3ec775b9 | 3071 | /* slave's name changed - he's no longer primary */ |
059b47e8 | 3072 | RCU_INIT_POINTER(bond->primary_slave, NULL); |
3ec775b9 VF |
3073 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { |
3074 | /* we have a new primary slave */ | |
059b47e8 | 3075 | rcu_assign_pointer(bond->primary_slave, slave); |
3ec775b9 VF |
3076 | } else { /* we didn't change primary - exit */ |
3077 | break; | |
3078 | } | |
3079 | ||
76444f50 | 3080 | netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n", |
059b47e8 | 3081 | primary ? slave_dev->name : "none"); |
f80889a5 | 3082 | |
3083 | block_netpoll_tx(); | |
3ec775b9 | 3084 | bond_select_active_slave(bond); |
f80889a5 | 3085 | unblock_netpoll_tx(); |
1da177e4 | 3086 | break; |
8531c5ff AK |
3087 | case NETDEV_FEAT_CHANGE: |
3088 | bond_compute_features(bond); | |
3089 | break; | |
4aa5dee4 JP |
3090 | case NETDEV_RESEND_IGMP: |
3091 | /* Propagate to master device */ | |
3092 | call_netdevice_notifiers(event, slave->bond->dev); | |
3093 | break; | |
1da177e4 LT |
3094 | default: |
3095 | break; | |
3096 | } | |
3097 | ||
3098 | return NOTIFY_DONE; | |
3099 | } | |
3100 | ||
547942ca | 3101 | /* bond_netdev_event: handle netdev notifier chain events. |
1da177e4 LT |
3102 | * |
3103 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3104 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3105 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3106 | * dev_probe_lock). | |
3107 | */ | |
3d632c3f SH |
3108 | static int bond_netdev_event(struct notifier_block *this, |
3109 | unsigned long event, void *ptr) | |
1da177e4 | 3110 | { |
351638e7 | 3111 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3112 | |
76444f50 | 3113 | netdev_dbg(event_dev, "event: %lx\n", event); |
1da177e4 | 3114 | |
0b680e75 JV |
3115 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3116 | return NOTIFY_DONE; | |
3117 | ||
1da177e4 | 3118 | if (event_dev->flags & IFF_MASTER) { |
76444f50 | 3119 | netdev_dbg(event_dev, "IFF_MASTER\n"); |
1da177e4 LT |
3120 | return bond_master_netdev_event(event, event_dev); |
3121 | } | |
3122 | ||
3123 | if (event_dev->flags & IFF_SLAVE) { | |
76444f50 | 3124 | netdev_dbg(event_dev, "IFF_SLAVE\n"); |
1da177e4 LT |
3125 | return bond_slave_netdev_event(event, event_dev); |
3126 | } | |
3127 | ||
3128 | return NOTIFY_DONE; | |
3129 | } | |
3130 | ||
3131 | static struct notifier_block bond_netdev_notifier = { | |
3132 | .notifier_call = bond_netdev_event, | |
3133 | }; | |
3134 | ||
169a3e66 JV |
3135 | /*---------------------------- Hashing Policies -----------------------------*/ |
3136 | ||
32819dc1 NA |
3137 | /* L2 hash helper */ |
3138 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 | 3139 | { |
ce04d635 | 3140 | struct ethhdr *ep, hdr_tmp; |
6b923cb7 | 3141 | |
ce04d635 JX |
3142 | ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp); |
3143 | if (ep) | |
3144 | return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto; | |
6b923cb7 JE |
3145 | return 0; |
3146 | } | |
3147 | ||
32819dc1 NA |
3148 | /* Extract the appropriate headers based on bond's xmit policy */ |
3149 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
3150 | struct flow_keys *fk) | |
6f6652be | 3151 | { |
32819dc1 | 3152 | const struct ipv6hdr *iph6; |
4394542c | 3153 | const struct iphdr *iph; |
32819dc1 | 3154 | int noff, proto = -1; |
6b923cb7 | 3155 | |
32819dc1 | 3156 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) |
cd79a238 | 3157 | return skb_flow_dissect_flow_keys(skb, fk, 0); |
32819dc1 | 3158 | |
06635a35 | 3159 | fk->ports.ports = 0; |
32819dc1 NA |
3160 | noff = skb_network_offset(skb); |
3161 | if (skb->protocol == htons(ETH_P_IP)) { | |
054bb880 | 3162 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
32819dc1 | 3163 | return false; |
6b923cb7 | 3164 | iph = ip_hdr(skb); |
c3f83241 | 3165 | iph_to_flow_copy_v4addrs(fk, iph); |
32819dc1 NA |
3166 | noff += iph->ihl << 2; |
3167 | if (!ip_is_fragment(iph)) | |
3168 | proto = iph->protocol; | |
3169 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
054bb880 | 3170 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6)))) |
32819dc1 NA |
3171 | return false; |
3172 | iph6 = ipv6_hdr(skb); | |
c3f83241 | 3173 | iph_to_flow_copy_v6addrs(fk, iph6); |
32819dc1 NA |
3174 | noff += sizeof(*iph6); |
3175 | proto = iph6->nexthdr; | |
3176 | } else { | |
3177 | return false; | |
6f6652be | 3178 | } |
32819dc1 | 3179 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) |
06635a35 | 3180 | fk->ports.ports = skb_flow_get_ports(skb, noff, proto); |
6f6652be | 3181 | |
32819dc1 | 3182 | return true; |
6f6652be JV |
3183 | } |
3184 | ||
32819dc1 NA |
3185 | /** |
3186 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3187 | * @bond: bonding device | |
3188 | * @skb: buffer to use for headers | |
32819dc1 NA |
3189 | * |
3190 | * This function will extract the necessary headers from the skb buffer and use | |
3191 | * them to generate a hash based on the xmit_policy set in the bonding device | |
169a3e66 | 3192 | */ |
ee62e868 | 3193 | u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) |
169a3e66 | 3194 | { |
32819dc1 NA |
3195 | struct flow_keys flow; |
3196 | u32 hash; | |
4394542c | 3197 | |
4b1b865e ED |
3198 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 && |
3199 | skb->l4_hash) | |
3200 | return skb->hash; | |
3201 | ||
32819dc1 NA |
3202 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3203 | !bond_flow_dissect(bond, skb, &flow)) | |
ee62e868 | 3204 | return bond_eth_hash(skb); |
169a3e66 | 3205 | |
32819dc1 NA |
3206 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
3207 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) | |
3208 | hash = bond_eth_hash(skb); | |
3209 | else | |
06635a35 | 3210 | hash = (__force u32)flow.ports.ports; |
c3f83241 TH |
3211 | hash ^= (__force u32)flow_get_u32_dst(&flow) ^ |
3212 | (__force u32)flow_get_u32_src(&flow); | |
32819dc1 NA |
3213 | hash ^= (hash >> 16); |
3214 | hash ^= (hash >> 8); | |
3215 | ||
ee62e868 | 3216 | return hash; |
169a3e66 JV |
3217 | } |
3218 | ||
1da177e4 LT |
3219 | /*-------------------------- Device entry points ----------------------------*/ |
3220 | ||
fbb0c41b | 3221 | static void bond_work_init_all(struct bonding *bond) |
3222 | { | |
3223 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3224 | bond_resend_igmp_join_requests_delayed); | |
3225 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3226 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
01844098 | 3227 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
fbb0c41b | 3228 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); |
3229 | else | |
3230 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3231 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
ee637714 | 3232 | INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler); |
fbb0c41b | 3233 | } |
3234 | ||
3235 | static void bond_work_cancel_all(struct bonding *bond) | |
3236 | { | |
3237 | cancel_delayed_work_sync(&bond->mii_work); | |
3238 | cancel_delayed_work_sync(&bond->arp_work); | |
3239 | cancel_delayed_work_sync(&bond->alb_work); | |
3240 | cancel_delayed_work_sync(&bond->ad_work); | |
3241 | cancel_delayed_work_sync(&bond->mcast_work); | |
ee637714 | 3242 | cancel_delayed_work_sync(&bond->slave_arr_work); |
fbb0c41b | 3243 | } |
3244 | ||
1da177e4 LT |
3245 | static int bond_open(struct net_device *bond_dev) |
3246 | { | |
454d7c9b | 3247 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3248 | struct list_head *iter; |
ba3211cc | 3249 | struct slave *slave; |
1da177e4 | 3250 | |
ba3211cc | 3251 | /* reset slave->backup and slave->inactive */ |
0965a1f3 | 3252 | if (bond_has_slaves(bond)) { |
9caff1e7 | 3253 | bond_for_each_slave(bond, slave, iter) { |
4740d638 ED |
3254 | if (bond_uses_primary(bond) && |
3255 | slave != rcu_access_pointer(bond->curr_active_slave)) { | |
5e5b0665 | 3256 | bond_set_slave_inactive_flags(slave, |
3257 | BOND_SLAVE_NOTIFY_NOW); | |
8bbe71a5 | 3258 | } else if (BOND_MODE(bond) != BOND_MODE_8023AD) { |
5e5b0665 | 3259 | bond_set_slave_active_flags(slave, |
3260 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3261 | } |
3262 | } | |
ba3211cc | 3263 | } |
ba3211cc | 3264 | |
fbb0c41b | 3265 | bond_work_init_all(bond); |
5a37e8ca | 3266 | |
58402054 | 3267 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3268 | /* bond_alb_initialize must be called before the timer |
3269 | * is started. | |
3270 | */ | |
01844098 | 3271 | if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) |
b473946a | 3272 | return -ENOMEM; |
e9f0fb88 MB |
3273 | if (bond->params.tlb_dynamic_lb) |
3274 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3275 | } |
3276 | ||
fbb0c41b | 3277 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3278 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3279 | |
3280 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3281 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3282 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3283 | } |
3284 | ||
01844098 | 3285 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1b76b316 | 3286 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3287 | /* register to receive LACPDUs */ |
3aba891d | 3288 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3289 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3290 | } |
3291 | ||
ee637714 MB |
3292 | if (bond_mode_uses_xmit_hash(bond)) |
3293 | bond_update_slave_arr(bond, NULL); | |
3294 | ||
1da177e4 LT |
3295 | return 0; |
3296 | } | |
3297 | ||
3298 | static int bond_close(struct net_device *bond_dev) | |
3299 | { | |
454d7c9b | 3300 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3301 | |
fbb0c41b | 3302 | bond_work_cancel_all(bond); |
6c8d23f7 | 3303 | bond->send_peer_notif = 0; |
ee8487c0 | 3304 | if (bond_is_lb(bond)) |
1da177e4 | 3305 | bond_alb_deinitialize(bond); |
3aba891d | 3306 | bond->recv_probe = NULL; |
1da177e4 LT |
3307 | |
3308 | return 0; | |
3309 | } | |
3310 | ||
28172739 ED |
3311 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3312 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3313 | { |
454d7c9b | 3314 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3315 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3316 | struct list_head *iter; |
1da177e4 | 3317 | struct slave *slave; |
1da177e4 | 3318 | |
5f0c5f73 | 3319 | memcpy(stats, &bond->bond_stats, sizeof(*stats)); |
1da177e4 | 3320 | |
9caff1e7 | 3321 | bond_for_each_slave(bond, slave, iter) { |
be1f3c2c | 3322 | const struct rtnl_link_stats64 *sstats = |
28172739 | 3323 | dev_get_stats(slave->dev, &temp); |
5f0c5f73 AG |
3324 | struct rtnl_link_stats64 *pstats = &slave->slave_stats; |
3325 | ||
3326 | stats->rx_packets += sstats->rx_packets - pstats->rx_packets; | |
3327 | stats->rx_bytes += sstats->rx_bytes - pstats->rx_bytes; | |
3328 | stats->rx_errors += sstats->rx_errors - pstats->rx_errors; | |
3329 | stats->rx_dropped += sstats->rx_dropped - pstats->rx_dropped; | |
3330 | ||
3331 | stats->tx_packets += sstats->tx_packets - pstats->tx_packets;; | |
3332 | stats->tx_bytes += sstats->tx_bytes - pstats->tx_bytes; | |
3333 | stats->tx_errors += sstats->tx_errors - pstats->tx_errors; | |
3334 | stats->tx_dropped += sstats->tx_dropped - pstats->tx_dropped; | |
3335 | ||
3336 | stats->multicast += sstats->multicast - pstats->multicast; | |
3337 | stats->collisions += sstats->collisions - pstats->collisions; | |
3338 | ||
3339 | stats->rx_length_errors += sstats->rx_length_errors - pstats->rx_length_errors; | |
3340 | stats->rx_over_errors += sstats->rx_over_errors - pstats->rx_over_errors; | |
3341 | stats->rx_crc_errors += sstats->rx_crc_errors - pstats->rx_crc_errors; | |
3342 | stats->rx_frame_errors += sstats->rx_frame_errors - pstats->rx_frame_errors; | |
3343 | stats->rx_fifo_errors += sstats->rx_fifo_errors - pstats->rx_fifo_errors; | |
3344 | stats->rx_missed_errors += sstats->rx_missed_errors - pstats->rx_missed_errors; | |
3345 | ||
3346 | stats->tx_aborted_errors += sstats->tx_aborted_errors - pstats->tx_aborted_errors; | |
3347 | stats->tx_carrier_errors += sstats->tx_carrier_errors - pstats->tx_carrier_errors; | |
3348 | stats->tx_fifo_errors += sstats->tx_fifo_errors - pstats->tx_fifo_errors; | |
3349 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors - pstats->tx_heartbeat_errors; | |
3350 | stats->tx_window_errors += sstats->tx_window_errors - pstats->tx_window_errors; | |
3351 | ||
3352 | /* save off the slave stats for the next run */ | |
3353 | memcpy(pstats, sstats, sizeof(*sstats)); | |
3354 | } | |
3355 | memcpy(&bond->bond_stats, stats, sizeof(*stats)); | |
1da177e4 LT |
3356 | |
3357 | return stats; | |
3358 | } | |
3359 | ||
3360 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3361 | { | |
080a06e1 | 3362 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3363 | struct net_device *slave_dev = NULL; |
3364 | struct ifbond k_binfo; | |
3365 | struct ifbond __user *u_binfo = NULL; | |
3366 | struct ifslave k_sinfo; | |
3367 | struct ifslave __user *u_sinfo = NULL; | |
3368 | struct mii_ioctl_data *mii = NULL; | |
d1fbd3ed | 3369 | struct bond_opt_value newval; |
387ff911 | 3370 | struct net *net; |
1da177e4 LT |
3371 | int res = 0; |
3372 | ||
76444f50 | 3373 | netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd); |
1da177e4 LT |
3374 | |
3375 | switch (cmd) { | |
1da177e4 LT |
3376 | case SIOCGMIIPHY: |
3377 | mii = if_mii(ifr); | |
3d632c3f | 3378 | if (!mii) |
1da177e4 | 3379 | return -EINVAL; |
3d632c3f | 3380 | |
1da177e4 LT |
3381 | mii->phy_id = 0; |
3382 | /* Fall Through */ | |
3383 | case SIOCGMIIREG: | |
547942ca | 3384 | /* We do this again just in case we were called by SIOCGMIIREG |
1da177e4 LT |
3385 | * instead of SIOCGMIIPHY. |
3386 | */ | |
3387 | mii = if_mii(ifr); | |
3d632c3f | 3388 | if (!mii) |
1da177e4 | 3389 | return -EINVAL; |
3d632c3f | 3390 | |
1da177e4 | 3391 | if (mii->reg_num == 1) { |
1da177e4 | 3392 | mii->val_out = 0; |
3d632c3f | 3393 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3394 | mii->val_out = BMSR_LSTATUS; |
1da177e4 LT |
3395 | } |
3396 | ||
3397 | return 0; | |
3398 | case BOND_INFO_QUERY_OLD: | |
3399 | case SIOCBONDINFOQUERY: | |
3400 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3401 | ||
3d632c3f | 3402 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3403 | return -EFAULT; |
1da177e4 LT |
3404 | |
3405 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3406 | if (res == 0 && |
3407 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3408 | return -EFAULT; | |
1da177e4 LT |
3409 | |
3410 | return res; | |
3411 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3412 | case SIOCBONDSLAVEINFOQUERY: | |
3413 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3414 | ||
3d632c3f | 3415 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3416 | return -EFAULT; |
1da177e4 LT |
3417 | |
3418 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3419 | if (res == 0 && |
3420 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3421 | return -EFAULT; | |
1da177e4 LT |
3422 | |
3423 | return res; | |
3424 | default: | |
1da177e4 LT |
3425 | break; |
3426 | } | |
3427 | ||
387ff911 G |
3428 | net = dev_net(bond_dev); |
3429 | ||
3430 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3431 | return -EPERM; |
1da177e4 | 3432 | |
0917b933 | 3433 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3434 | |
76444f50 | 3435 | netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev); |
1da177e4 | 3436 | |
3d632c3f | 3437 | if (!slave_dev) |
0917b933 | 3438 | return -ENODEV; |
1da177e4 | 3439 | |
76444f50 | 3440 | netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name); |
0917b933 YX |
3441 | switch (cmd) { |
3442 | case BOND_ENSLAVE_OLD: | |
3443 | case SIOCBONDENSLAVE: | |
3444 | res = bond_enslave(bond_dev, slave_dev); | |
3445 | break; | |
3446 | case BOND_RELEASE_OLD: | |
3447 | case SIOCBONDRELEASE: | |
3448 | res = bond_release(bond_dev, slave_dev); | |
3449 | break; | |
3450 | case BOND_SETHWADDR_OLD: | |
3451 | case SIOCBONDSETHWADDR: | |
3452 | bond_set_dev_addr(bond_dev, slave_dev); | |
3453 | res = 0; | |
3454 | break; | |
3455 | case BOND_CHANGE_ACTIVE_OLD: | |
3456 | case SIOCBONDCHANGEACTIVE: | |
d1fbd3ed NA |
3457 | bond_opt_initstr(&newval, slave_dev->name); |
3458 | res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval); | |
0917b933 YX |
3459 | break; |
3460 | default: | |
3461 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3462 | } |
3463 | ||
1da177e4 LT |
3464 | return res; |
3465 | } | |
3466 | ||
d03462b9 | 3467 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3468 | { |
454d7c9b | 3469 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3470 | |
d03462b9 JP |
3471 | if (change & IFF_PROMISC) |
3472 | bond_set_promiscuity(bond, | |
3473 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3474 | |
d03462b9 JP |
3475 | if (change & IFF_ALLMULTI) |
3476 | bond_set_allmulti(bond, | |
3477 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3478 | } | |
1da177e4 | 3479 | |
303d1cbf | 3480 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3481 | { |
3482 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3483 | struct list_head *iter; |
303d1cbf | 3484 | struct slave *slave; |
1da177e4 | 3485 | |
b3241870 | 3486 | rcu_read_lock(); |
ec0865a9 | 3487 | if (bond_uses_primary(bond)) { |
b3241870 | 3488 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3489 | if (slave) { |
3490 | dev_uc_sync(slave->dev, bond_dev); | |
3491 | dev_mc_sync(slave->dev, bond_dev); | |
3492 | } | |
303d1cbf | 3493 | } else { |
b3241870 | 3494 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3495 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3496 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3497 | } | |
1da177e4 | 3498 | } |
b3241870 | 3499 | rcu_read_unlock(); |
1da177e4 LT |
3500 | } |
3501 | ||
234bcf8a | 3502 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3503 | { |
234bcf8a | 3504 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3505 | const struct net_device_ops *slave_ops; |
3506 | struct neigh_parms parms; | |
dec1e90e | 3507 | struct slave *slave; |
234bcf8a SP |
3508 | int ret; |
3509 | ||
dec1e90e | 3510 | slave = bond_first_slave(bond); |
234bcf8a SP |
3511 | if (!slave) |
3512 | return 0; | |
234bcf8a | 3513 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3514 | if (!slave_ops->ndo_neigh_setup) |
3515 | return 0; | |
3516 | ||
3517 | parms.neigh_setup = NULL; | |
3518 | parms.neigh_cleanup = NULL; | |
3519 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3520 | if (ret) | |
3521 | return ret; | |
3522 | ||
547942ca | 3523 | /* Assign slave's neigh_cleanup to neighbour in case cleanup is called |
234bcf8a SP |
3524 | * after the last slave has been detached. Assumes that all slaves |
3525 | * utilize the same neigh_cleanup (true at this writing as only user | |
3526 | * is ipoib). | |
3527 | */ | |
3528 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3529 | ||
3530 | if (!parms.neigh_setup) | |
3531 | return 0; | |
3532 | ||
3533 | return parms.neigh_setup(n); | |
3534 | } | |
3535 | ||
547942ca | 3536 | /* The bonding ndo_neigh_setup is called at init time beofre any |
234bcf8a SP |
3537 | * slave exists. So we must declare proxy setup function which will |
3538 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3539 | * |
3540 | * It's also called by master devices (such as vlans) to setup their | |
3541 | * underlying devices. In that case - do nothing, we're already set up from | |
3542 | * our init. | |
234bcf8a SP |
3543 | */ |
3544 | static int bond_neigh_setup(struct net_device *dev, | |
3545 | struct neigh_parms *parms) | |
3546 | { | |
9918d5bf VF |
3547 | /* modify only our neigh_parms */ |
3548 | if (parms->dev == dev) | |
3549 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3550 | |
00829823 SH |
3551 | return 0; |
3552 | } | |
3553 | ||
547942ca | 3554 | /* Change the MTU of all of a master's slaves to match the master */ |
1da177e4 LT |
3555 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) |
3556 | { | |
454d7c9b | 3557 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3558 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3559 | struct list_head *iter; |
1da177e4 | 3560 | int res = 0; |
1da177e4 | 3561 | |
76444f50 | 3562 | netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu); |
1da177e4 | 3563 | |
9caff1e7 | 3564 | bond_for_each_slave(bond, slave, iter) { |
76444f50 VF |
3565 | netdev_dbg(bond_dev, "s %p c_m %p\n", |
3566 | slave, slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3567 | |
1da177e4 LT |
3568 | res = dev_set_mtu(slave->dev, new_mtu); |
3569 | ||
3570 | if (res) { | |
3571 | /* If we failed to set the slave's mtu to the new value | |
3572 | * we must abort the operation even in ACTIVE_BACKUP | |
3573 | * mode, because if we allow the backup slaves to have | |
3574 | * different mtu values than the active slave we'll | |
3575 | * need to change their mtu when doing a failover. That | |
3576 | * means changing their mtu from timer context, which | |
3577 | * is probably not a good idea. | |
3578 | */ | |
76444f50 VF |
3579 | netdev_dbg(bond_dev, "err %d %s\n", res, |
3580 | slave->dev->name); | |
1da177e4 LT |
3581 | goto unwind; |
3582 | } | |
3583 | } | |
3584 | ||
3585 | bond_dev->mtu = new_mtu; | |
3586 | ||
3587 | return 0; | |
3588 | ||
3589 | unwind: | |
3590 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3591 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3592 | int tmp_res; |
3593 | ||
81f23b13 VF |
3594 | if (rollback_slave == slave) |
3595 | break; | |
3596 | ||
3597 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3598 | if (tmp_res) { |
76444f50 VF |
3599 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3600 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3601 | } |
3602 | } | |
3603 | ||
3604 | return res; | |
3605 | } | |
3606 | ||
547942ca | 3607 | /* Change HW address |
1da177e4 LT |
3608 | * |
3609 | * Note that many devices must be down to change the HW address, and | |
3610 | * downing the master releases all slaves. We can make bonds full of | |
3611 | * bonding devices to test this, however. | |
3612 | */ | |
3613 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3614 | { | |
454d7c9b | 3615 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3616 | struct slave *slave, *rollback_slave; |
1da177e4 | 3617 | struct sockaddr *sa = addr, tmp_sa; |
9caff1e7 | 3618 | struct list_head *iter; |
1da177e4 | 3619 | int res = 0; |
1da177e4 | 3620 | |
01844098 | 3621 | if (BOND_MODE(bond) == BOND_MODE_ALB) |
eb7cc59a SH |
3622 | return bond_alb_set_mac_address(bond_dev, addr); |
3623 | ||
3624 | ||
76444f50 | 3625 | netdev_dbg(bond_dev, "bond=%p\n", bond); |
1da177e4 | 3626 | |
1b5acd29 JV |
3627 | /* If fail_over_mac is enabled, do nothing and return success. |
3628 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3629 | */ |
cc689aaa | 3630 | if (bond->params.fail_over_mac && |
01844098 | 3631 | BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
dd957c57 | 3632 | return 0; |
2ab82852 | 3633 | |
3d632c3f | 3634 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3635 | return -EADDRNOTAVAIL; |
1da177e4 | 3636 | |
9caff1e7 | 3637 | bond_for_each_slave(bond, slave, iter) { |
76444f50 | 3638 | netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name); |
1da177e4 LT |
3639 | res = dev_set_mac_address(slave->dev, addr); |
3640 | if (res) { | |
3641 | /* TODO: consider downing the slave | |
3642 | * and retry ? | |
3643 | * User should expect communications | |
3644 | * breakage anyway until ARP finish | |
3645 | * updating, so... | |
3646 | */ | |
76444f50 | 3647 | netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3648 | goto unwind; |
3649 | } | |
3650 | } | |
3651 | ||
3652 | /* success */ | |
3653 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3654 | return 0; | |
3655 | ||
3656 | unwind: | |
3657 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3658 | tmp_sa.sa_family = bond_dev->type; | |
3659 | ||
3660 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3661 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3662 | int tmp_res; |
3663 | ||
81f23b13 VF |
3664 | if (rollback_slave == slave) |
3665 | break; | |
3666 | ||
3667 | tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa); | |
1da177e4 | 3668 | if (tmp_res) { |
76444f50 VF |
3669 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3670 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3671 | } |
3672 | } | |
3673 | ||
3674 | return res; | |
3675 | } | |
3676 | ||
15077228 NA |
3677 | /** |
3678 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3679 | * @bond: bonding device that is transmitting | |
3680 | * @skb: buffer to transmit | |
3681 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3682 | * | |
3683 | * This function tries to transmit through slave with slave_id but in case | |
3684 | * it fails, it tries to find the first available slave for transmission. | |
3685 | * The skb is consumed in all cases, thus the function is void. | |
3686 | */ | |
da131ddb | 3687 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3688 | { |
9caff1e7 | 3689 | struct list_head *iter; |
15077228 NA |
3690 | struct slave *slave; |
3691 | int i = slave_id; | |
3692 | ||
3693 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3694 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 | 3695 | if (--i < 0) { |
891ab54d | 3696 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3697 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3698 | return; | |
3699 | } | |
3700 | } | |
3701 | } | |
3702 | ||
3703 | /* Here we start from the first slave up to slave_id */ | |
3704 | i = slave_id; | |
9caff1e7 | 3705 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3706 | if (--i < 0) |
3707 | break; | |
891ab54d | 3708 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3709 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3710 | return; | |
3711 | } | |
3712 | } | |
3713 | /* no slave that can tx has been found */ | |
31aa860e | 3714 | bond_tx_drop(bond->dev, skb); |
15077228 NA |
3715 | } |
3716 | ||
73958329 NA |
3717 | /** |
3718 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3719 | * @bond: bonding device to use | |
3720 | * | |
3721 | * Based on the value of the bonding device's packets_per_slave parameter | |
3722 | * this function generates a slave id, which is usually used as the next | |
3723 | * slave to transmit through. | |
3724 | */ | |
3725 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3726 | { | |
73958329 | 3727 | u32 slave_id; |
809fa972 HFS |
3728 | struct reciprocal_value reciprocal_packets_per_slave; |
3729 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
3730 | |
3731 | switch (packets_per_slave) { | |
3732 | case 0: | |
3733 | slave_id = prandom_u32(); | |
3734 | break; | |
3735 | case 1: | |
3736 | slave_id = bond->rr_tx_counter; | |
3737 | break; | |
3738 | default: | |
809fa972 HFS |
3739 | reciprocal_packets_per_slave = |
3740 | bond->params.reciprocal_packets_per_slave; | |
73958329 | 3741 | slave_id = reciprocal_divide(bond->rr_tx_counter, |
809fa972 | 3742 | reciprocal_packets_per_slave); |
73958329 NA |
3743 | break; |
3744 | } | |
3745 | bond->rr_tx_counter++; | |
3746 | ||
3747 | return slave_id; | |
3748 | } | |
3749 | ||
1da177e4 LT |
3750 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3751 | { | |
454d7c9b | 3752 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3753 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3754 | struct slave *slave; |
73958329 | 3755 | u32 slave_id; |
1da177e4 | 3756 | |
73958329 | 3757 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3758 | * default for sending IGMP traffic. For failover purposes one |
3759 | * needs to maintain some consistency for the interface that will | |
3760 | * send the join/membership reports. The curr_active_slave found | |
3761 | * will send all of this type of traffic. | |
cf5f9044 | 3762 | */ |
15077228 | 3763 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3764 | slave = rcu_dereference(bond->curr_active_slave); |
a64d044e | 3765 | if (slave) |
15077228 NA |
3766 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3767 | else | |
3768 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3769 | } else { |
9a72c2da NA |
3770 | int slave_cnt = ACCESS_ONCE(bond->slave_cnt); |
3771 | ||
3772 | if (likely(slave_cnt)) { | |
3773 | slave_id = bond_rr_gen_slave_id(bond); | |
3774 | bond_xmit_slave_id(bond, skb, slave_id % slave_cnt); | |
3775 | } else { | |
31aa860e | 3776 | bond_tx_drop(bond_dev, skb); |
9a72c2da | 3777 | } |
1da177e4 | 3778 | } |
0693e88e | 3779 | |
ec634fe3 | 3780 | return NETDEV_TX_OK; |
1da177e4 LT |
3781 | } |
3782 | ||
547942ca | 3783 | /* In active-backup mode, we know that bond->curr_active_slave is always valid if |
1da177e4 LT |
3784 | * the bond has a usable interface. |
3785 | */ | |
3786 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3787 | { | |
454d7c9b | 3788 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3789 | struct slave *slave; |
1da177e4 | 3790 | |
278b2083 | 3791 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3792 | if (slave) |
15077228 NA |
3793 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3794 | else | |
31aa860e | 3795 | bond_tx_drop(bond_dev, skb); |
0693e88e | 3796 | |
ec634fe3 | 3797 | return NETDEV_TX_OK; |
1da177e4 LT |
3798 | } |
3799 | ||
ee637714 MB |
3800 | /* Use this to update slave_array when (a) it's not appropriate to update |
3801 | * slave_array right away (note that update_slave_array() may sleep) | |
3802 | * and / or (b) RTNL is not held. | |
1da177e4 | 3803 | */ |
ee637714 | 3804 | void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay) |
1da177e4 | 3805 | { |
ee637714 MB |
3806 | queue_delayed_work(bond->wq, &bond->slave_arr_work, delay); |
3807 | } | |
1da177e4 | 3808 | |
ee637714 MB |
3809 | /* Slave array work handler. Holds only RTNL */ |
3810 | static void bond_slave_arr_handler(struct work_struct *work) | |
3811 | { | |
3812 | struct bonding *bond = container_of(work, struct bonding, | |
3813 | slave_arr_work.work); | |
3814 | int ret; | |
3815 | ||
3816 | if (!rtnl_trylock()) | |
3817 | goto err; | |
3818 | ||
3819 | ret = bond_update_slave_arr(bond, NULL); | |
3820 | rtnl_unlock(); | |
3821 | if (ret) { | |
3822 | pr_warn_ratelimited("Failed to update slave array from WT\n"); | |
3823 | goto err; | |
3824 | } | |
3825 | return; | |
3826 | ||
3827 | err: | |
3828 | bond_slave_arr_work_rearm(bond, 1); | |
3829 | } | |
3830 | ||
3831 | /* Build the usable slaves array in control path for modes that use xmit-hash | |
3832 | * to determine the slave interface - | |
3833 | * (a) BOND_MODE_8023AD | |
3834 | * (b) BOND_MODE_XOR | |
3835 | * (c) BOND_MODE_TLB && tlb_dynamic_lb == 0 | |
3836 | * | |
3837 | * The caller is expected to hold RTNL only and NO other lock! | |
3838 | */ | |
3839 | int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave) | |
3840 | { | |
3841 | struct slave *slave; | |
3842 | struct list_head *iter; | |
3843 | struct bond_up_slave *new_arr, *old_arr; | |
ee637714 MB |
3844 | int agg_id = 0; |
3845 | int ret = 0; | |
3846 | ||
3847 | #ifdef CONFIG_LOCKDEP | |
3848 | WARN_ON(lockdep_is_held(&bond->mode_lock)); | |
3849 | #endif | |
3850 | ||
3851 | new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]), | |
3852 | GFP_KERNEL); | |
3853 | if (!new_arr) { | |
3854 | ret = -ENOMEM; | |
3855 | pr_err("Failed to build slave-array.\n"); | |
3856 | goto out; | |
3857 | } | |
3858 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
3859 | struct ad_info ad_info; | |
3860 | ||
3861 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3862 | pr_debug("bond_3ad_get_active_agg_info failed\n"); | |
3863 | kfree_rcu(new_arr, rcu); | |
3864 | /* No active aggragator means it's not safe to use | |
3865 | * the previous array. | |
3866 | */ | |
3867 | old_arr = rtnl_dereference(bond->slave_arr); | |
3868 | if (old_arr) { | |
3869 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
3870 | kfree_rcu(old_arr, rcu); | |
3871 | } | |
3872 | goto out; | |
3873 | } | |
ee637714 MB |
3874 | agg_id = ad_info.aggregator_id; |
3875 | } | |
3876 | bond_for_each_slave(bond, slave, iter) { | |
3877 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
3878 | struct aggregator *agg; | |
3879 | ||
3880 | agg = SLAVE_AD_INFO(slave)->port.aggregator; | |
3881 | if (!agg || agg->aggregator_identifier != agg_id) | |
3882 | continue; | |
3883 | } | |
3884 | if (!bond_slave_can_tx(slave)) | |
3885 | continue; | |
3886 | if (skipslave == slave) | |
3887 | continue; | |
3888 | new_arr->arr[new_arr->count++] = slave; | |
3889 | } | |
3890 | ||
3891 | old_arr = rtnl_dereference(bond->slave_arr); | |
3892 | rcu_assign_pointer(bond->slave_arr, new_arr); | |
3893 | if (old_arr) | |
3894 | kfree_rcu(old_arr, rcu); | |
3895 | out: | |
3896 | if (ret != 0 && skipslave) { | |
3897 | int idx; | |
3898 | ||
3899 | /* Rare situation where caller has asked to skip a specific | |
3900 | * slave but allocation failed (most likely!). BTW this is | |
3901 | * only possible when the call is initiated from | |
3902 | * __bond_release_one(). In this situation; overwrite the | |
3903 | * skipslave entry in the array with the last entry from the | |
3904 | * array to avoid a situation where the xmit path may choose | |
3905 | * this to-be-skipped slave to send a packet out. | |
3906 | */ | |
3907 | old_arr = rtnl_dereference(bond->slave_arr); | |
3908 | for (idx = 0; idx < old_arr->count; idx++) { | |
3909 | if (skipslave == old_arr->arr[idx]) { | |
3910 | old_arr->arr[idx] = | |
3911 | old_arr->arr[old_arr->count-1]; | |
3912 | old_arr->count--; | |
3913 | break; | |
3914 | } | |
3915 | } | |
3916 | } | |
3917 | return ret; | |
3918 | } | |
3919 | ||
3920 | /* Use this Xmit function for 3AD as well as XOR modes. The current | |
3921 | * usable slave array is formed in the control path. The xmit function | |
3922 | * just calculates hash and sends the packet out. | |
3923 | */ | |
30369104 | 3924 | static int bond_3ad_xor_xmit(struct sk_buff *skb, struct net_device *dev) |
ee637714 MB |
3925 | { |
3926 | struct bonding *bond = netdev_priv(dev); | |
3927 | struct slave *slave; | |
3928 | struct bond_up_slave *slaves; | |
3929 | unsigned int count; | |
3930 | ||
3931 | slaves = rcu_dereference(bond->slave_arr); | |
3932 | count = slaves ? ACCESS_ONCE(slaves->count) : 0; | |
3933 | if (likely(count)) { | |
3934 | slave = slaves->arr[bond_xmit_hash(bond, skb) % count]; | |
3935 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3936 | } else { | |
31aa860e | 3937 | bond_tx_drop(dev, skb); |
ee637714 | 3938 | } |
0693e88e | 3939 | |
ec634fe3 | 3940 | return NETDEV_TX_OK; |
1da177e4 LT |
3941 | } |
3942 | ||
78a646ce | 3943 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3944 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3945 | { | |
454d7c9b | 3946 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3947 | struct slave *slave = NULL; |
9caff1e7 | 3948 | struct list_head *iter; |
1da177e4 | 3949 | |
9caff1e7 | 3950 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
3951 | if (bond_is_last_slave(bond, slave)) |
3952 | break; | |
b6adc610 | 3953 | if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
78a646ce | 3954 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 | 3955 | |
78a646ce | 3956 | if (!skb2) { |
9152e26d | 3957 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
3958 | bond_dev->name, __func__); | |
78a646ce | 3959 | continue; |
1da177e4 | 3960 | } |
78a646ce | 3961 | bond_dev_queue_xmit(bond, skb2, slave->dev); |
1da177e4 LT |
3962 | } |
3963 | } | |
b6adc610 | 3964 | if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) |
78a646ce NA |
3965 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3966 | else | |
31aa860e | 3967 | bond_tx_drop(bond_dev, skb); |
3d632c3f | 3968 | |
ec634fe3 | 3969 | return NETDEV_TX_OK; |
1da177e4 LT |
3970 | } |
3971 | ||
3972 | /*------------------------- Device initialization ---------------------------*/ | |
3973 | ||
547942ca | 3974 | /* Lookup the slave that corresponds to a qid */ |
bb1d9123 AG |
3975 | static inline int bond_slave_override(struct bonding *bond, |
3976 | struct sk_buff *skb) | |
3977 | { | |
bb1d9123 | 3978 | struct slave *slave = NULL; |
9caff1e7 | 3979 | struct list_head *iter; |
bb1d9123 | 3980 | |
0693e88e MM |
3981 | if (!skb->queue_mapping) |
3982 | return 1; | |
bb1d9123 AG |
3983 | |
3984 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 3985 | bond_for_each_slave_rcu(bond, slave, iter) { |
3986 | if (slave->queue_id == skb->queue_mapping) { | |
f5e2dc5d AN |
3987 | if (bond_slave_is_up(slave) && |
3988 | slave->link == BOND_LINK_UP) { | |
3900f290 | 3989 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3990 | return 0; | |
3991 | } | |
3992 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
3993 | break; |
3994 | } | |
3995 | } | |
3996 | ||
3900f290 | 3997 | return 1; |
bb1d9123 AG |
3998 | } |
3999 | ||
374eeb5a | 4000 | |
f663dd9a | 4001 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 4002 | void *accel_priv, select_queue_fallback_t fallback) |
bb1d9123 | 4003 | { |
547942ca | 4004 | /* This helper function exists to help dev_pick_tx get the correct |
fd0e435b | 4005 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4006 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4007 | * way down to the bonding driver. | |
4008 | */ | |
fd0e435b PO |
4009 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4010 | ||
547942ca | 4011 | /* Save the original txq to restore before passing to the driver */ |
df4ab5b3 | 4012 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 4013 | |
fd0e435b | 4014 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4015 | do { |
fd0e435b | 4016 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4017 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4018 | } |
4019 | return txq; | |
bb1d9123 AG |
4020 | } |
4021 | ||
0693e88e | 4022 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4023 | { |
bb1d9123 AG |
4024 | struct bonding *bond = netdev_priv(dev); |
4025 | ||
d1e2e5cd VF |
4026 | if (bond_should_override_tx_queue(bond) && |
4027 | !bond_slave_override(bond, skb)) | |
4028 | return NETDEV_TX_OK; | |
00829823 | 4029 | |
01844098 | 4030 | switch (BOND_MODE(bond)) { |
00829823 SH |
4031 | case BOND_MODE_ROUNDROBIN: |
4032 | return bond_xmit_roundrobin(skb, dev); | |
4033 | case BOND_MODE_ACTIVEBACKUP: | |
4034 | return bond_xmit_activebackup(skb, dev); | |
ee637714 | 4035 | case BOND_MODE_8023AD: |
00829823 | 4036 | case BOND_MODE_XOR: |
ee637714 | 4037 | return bond_3ad_xor_xmit(skb, dev); |
00829823 SH |
4038 | case BOND_MODE_BROADCAST: |
4039 | return bond_xmit_broadcast(skb, dev); | |
00829823 | 4040 | case BOND_MODE_ALB: |
00829823 | 4041 | return bond_alb_xmit(skb, dev); |
f05b42ea MB |
4042 | case BOND_MODE_TLB: |
4043 | return bond_tlb_xmit(skb, dev); | |
00829823 SH |
4044 | default: |
4045 | /* Should never happen, mode already checked */ | |
76444f50 | 4046 | netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); |
00829823 | 4047 | WARN_ON_ONCE(1); |
31aa860e | 4048 | bond_tx_drop(dev, skb); |
00829823 SH |
4049 | return NETDEV_TX_OK; |
4050 | } | |
4051 | } | |
4052 | ||
0693e88e MM |
4053 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4054 | { | |
4055 | struct bonding *bond = netdev_priv(dev); | |
4056 | netdev_tx_t ret = NETDEV_TX_OK; | |
4057 | ||
547942ca | 4058 | /* If we risk deadlock from transmitting this in the |
0693e88e MM |
4059 | * netpoll path, tell netpoll to queue the frame for later tx |
4060 | */ | |
054bb880 | 4061 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
4062 | return NETDEV_TX_BUSY; |
4063 | ||
278b2083 | 4064 | rcu_read_lock(); |
0965a1f3 | 4065 | if (bond_has_slaves(bond)) |
0693e88e MM |
4066 | ret = __bond_start_xmit(skb, dev); |
4067 | else | |
31aa860e | 4068 | bond_tx_drop(dev, skb); |
278b2083 | 4069 | rcu_read_unlock(); |
0693e88e MM |
4070 | |
4071 | return ret; | |
4072 | } | |
00829823 | 4073 | |
bb5b052f AG |
4074 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
4075 | struct ethtool_cmd *ecmd) | |
4076 | { | |
4077 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 4078 | unsigned long speed = 0; |
9caff1e7 | 4079 | struct list_head *iter; |
dec1e90e | 4080 | struct slave *slave; |
bb5b052f AG |
4081 | |
4082 | ecmd->duplex = DUPLEX_UNKNOWN; | |
4083 | ecmd->port = PORT_OTHER; | |
4084 | ||
8557cd74 | 4085 | /* Since bond_slave_can_tx returns false for all inactive or down slaves, we |
bb5b052f AG |
4086 | * do not need to check mode. Though link speed might not represent |
4087 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
4088 | * this is an accurate maximum. | |
4089 | */ | |
9caff1e7 | 4090 | bond_for_each_slave(bond, slave, iter) { |
8557cd74 | 4091 | if (bond_slave_can_tx(slave)) { |
bb5b052f AG |
4092 | if (slave->speed != SPEED_UNKNOWN) |
4093 | speed += slave->speed; | |
4094 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
4095 | slave->duplex != DUPLEX_UNKNOWN) | |
4096 | ecmd->duplex = slave->duplex; | |
4097 | } | |
4098 | } | |
4099 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
dec1e90e | 4100 | |
bb5b052f AG |
4101 | return 0; |
4102 | } | |
4103 | ||
217df670 | 4104 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 4105 | struct ethtool_drvinfo *drvinfo) |
217df670 | 4106 | { |
7826d43f JP |
4107 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
4108 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
4109 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
4110 | BOND_ABI_VERSION); | |
217df670 JV |
4111 | } |
4112 | ||
7282d491 | 4113 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4114 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 4115 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 4116 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
4117 | }; |
4118 | ||
eb7cc59a | 4119 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4120 | .ndo_init = bond_init, |
9e71626c | 4121 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4122 | .ndo_open = bond_open, |
4123 | .ndo_stop = bond_close, | |
00829823 | 4124 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4125 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4126 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4127 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4128 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 4129 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 4130 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4131 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4132 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4133 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4134 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4135 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4136 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4137 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4138 | .ndo_poll_controller = bond_poll_controller, | |
4139 | #endif | |
9232ecca JP |
4140 | .ndo_add_slave = bond_enslave, |
4141 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4142 | .ndo_fix_features = bond_fix_features, |
fc8f40d8 | 4143 | .ndo_bridge_setlink = switchdev_port_bridge_setlink, |
85fdb956 | 4144 | .ndo_bridge_getlink = switchdev_port_bridge_getlink, |
54ba5a0b | 4145 | .ndo_bridge_dellink = switchdev_port_bridge_dellink, |
45d4122c SS |
4146 | .ndo_fdb_add = switchdev_port_fdb_add, |
4147 | .ndo_fdb_del = switchdev_port_fdb_del, | |
4148 | .ndo_fdb_dump = switchdev_port_fdb_dump, | |
4847f049 | 4149 | .ndo_features_check = passthru_features_check, |
eb7cc59a SH |
4150 | }; |
4151 | ||
b3f92b63 DG |
4152 | static const struct device_type bond_type = { |
4153 | .name = "bond", | |
4154 | }; | |
4155 | ||
9e2e61fb AW |
4156 | static void bond_destructor(struct net_device *bond_dev) |
4157 | { | |
4158 | struct bonding *bond = netdev_priv(bond_dev); | |
4159 | if (bond->wq) | |
4160 | destroy_workqueue(bond->wq); | |
4161 | free_netdev(bond_dev); | |
4162 | } | |
4163 | ||
0a2a78c4 | 4164 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4165 | { |
454d7c9b | 4166 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4167 | |
b7435628 | 4168 | spin_lock_init(&bond->mode_lock); |
d2991f75 | 4169 | bond->params = bonding_defaults; |
1da177e4 LT |
4170 | |
4171 | /* Initialize pointers */ | |
1da177e4 | 4172 | bond->dev = bond_dev; |
1da177e4 LT |
4173 | |
4174 | /* Initialize the device entry points */ | |
181470fc | 4175 | ether_setup(bond_dev); |
eb7cc59a | 4176 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4177 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 4178 | |
9e2e61fb | 4179 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4180 | |
b3f92b63 DG |
4181 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4182 | ||
1da177e4 | 4183 | /* Initialize the device options */ |
1da177e4 | 4184 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; |
1e6f20ca | 4185 | bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE; |
550fd08c | 4186 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4187 | |
547942ca | 4188 | /* don't acquire bond device's netif_tx_lock when transmitting */ |
1da177e4 LT |
4189 | bond_dev->features |= NETIF_F_LLTX; |
4190 | ||
4191 | /* By default, we declare the bond to be fully | |
4192 | * VLAN hardware accelerated capable. Special | |
4193 | * care is taken in the various xmit functions | |
4194 | * when there are slaves that are not hw accel | |
4195 | * capable | |
4196 | */ | |
1da177e4 | 4197 | |
f9399814 WC |
4198 | /* Don't allow bond devices to change network namespaces. */ |
4199 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
4200 | ||
b2a103e6 | 4201 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4202 | NETIF_F_HW_VLAN_CTAG_TX | |
4203 | NETIF_F_HW_VLAN_CTAG_RX | | |
4204 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4205 | |
24f87d4c | 4206 | bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL; |
b2a103e6 | 4207 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4208 | } |
4209 | ||
547942ca NA |
4210 | /* Destroy a bonding device. |
4211 | * Must be under rtnl_lock when this function is called. | |
4212 | */ | |
c67dfb29 | 4213 | static void bond_uninit(struct net_device *bond_dev) |
a434e43f | 4214 | { |
454d7c9b | 4215 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
4216 | struct list_head *iter; |
4217 | struct slave *slave; | |
ee637714 | 4218 | struct bond_up_slave *arr; |
a434e43f | 4219 | |
f6dc31a8 WC |
4220 | bond_netpoll_cleanup(bond_dev); |
4221 | ||
c67dfb29 | 4222 | /* Release the bonded slaves */ |
544a028e | 4223 | bond_for_each_slave(bond, slave, iter) |
dec1e90e | 4224 | __bond_release_one(bond_dev, slave->dev, true); |
76444f50 | 4225 | netdev_info(bond_dev, "Released all slaves\n"); |
c67dfb29 | 4226 | |
ee637714 MB |
4227 | arr = rtnl_dereference(bond->slave_arr); |
4228 | if (arr) { | |
4229 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
4230 | kfree_rcu(arr, rcu); | |
4231 | } | |
4232 | ||
a434e43f JV |
4233 | list_del(&bond->bond_list); |
4234 | ||
f073c7ca | 4235 | bond_debug_unregister(bond); |
a434e43f JV |
4236 | } |
4237 | ||
1da177e4 LT |
4238 | /*------------------------- Module initialization ---------------------------*/ |
4239 | ||
1da177e4 LT |
4240 | static int bond_check_params(struct bond_params *params) |
4241 | { | |
5a5c5fd4 | 4242 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 4243 | struct bond_opt_value newval; |
4244 | const struct bond_opt_value *valptr; | |
8599b52e | 4245 | int arp_all_targets_value; |
6791e466 | 4246 | u16 ad_actor_sys_prio = 0; |
d22a5fc0 | 4247 | u16 ad_user_port_key = 0; |
f5b2b966 | 4248 | |
547942ca | 4249 | /* Convert string parameters. */ |
1da177e4 | 4250 | if (mode) { |
2b3798d5 NA |
4251 | bond_opt_initstr(&newval, mode); |
4252 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
4253 | if (!valptr) { | |
4254 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
4255 | return -EINVAL; |
4256 | } | |
2b3798d5 | 4257 | bond_mode = valptr->value; |
1da177e4 LT |
4258 | } |
4259 | ||
169a3e66 JV |
4260 | if (xmit_hash_policy) { |
4261 | if ((bond_mode != BOND_MODE_XOR) && | |
f05b42ea MB |
4262 | (bond_mode != BOND_MODE_8023AD) && |
4263 | (bond_mode != BOND_MODE_TLB)) { | |
a4aee5c8 | 4264 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 4265 | bond_mode_name(bond_mode)); |
169a3e66 | 4266 | } else { |
a4b32ce7 NA |
4267 | bond_opt_initstr(&newval, xmit_hash_policy); |
4268 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
4269 | &newval); | |
4270 | if (!valptr) { | |
a4aee5c8 | 4271 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
4272 | xmit_hash_policy); |
4273 | return -EINVAL; | |
4274 | } | |
a4b32ce7 | 4275 | xmit_hashtype = valptr->value; |
169a3e66 JV |
4276 | } |
4277 | } | |
4278 | ||
1da177e4 LT |
4279 | if (lacp_rate) { |
4280 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4281 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4282 | bond_mode_name(bond_mode)); | |
1da177e4 | 4283 | } else { |
d3131de7 NA |
4284 | bond_opt_initstr(&newval, lacp_rate); |
4285 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
4286 | &newval); | |
4287 | if (!valptr) { | |
a4aee5c8 | 4288 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 4289 | lacp_rate); |
1da177e4 LT |
4290 | return -EINVAL; |
4291 | } | |
d3131de7 | 4292 | lacp_fast = valptr->value; |
1da177e4 LT |
4293 | } |
4294 | } | |
4295 | ||
fd989c83 | 4296 | if (ad_select) { |
548d28bd | 4297 | bond_opt_initstr(&newval, ad_select); |
9e5f5eeb NA |
4298 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), |
4299 | &newval); | |
4300 | if (!valptr) { | |
4301 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
4302 | return -EINVAL; |
4303 | } | |
9e5f5eeb NA |
4304 | params->ad_select = valptr->value; |
4305 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 4306 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4307 | } else { |
4308 | params->ad_select = BOND_AD_STABLE; | |
4309 | } | |
4310 | ||
f5841306 | 4311 | if (max_bonds < 0) { |
91565ebb JP |
4312 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4313 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4314 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4315 | } | |
4316 | ||
4317 | if (miimon < 0) { | |
91565ebb JP |
4318 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4319 | miimon, INT_MAX); | |
b98d9c66 | 4320 | miimon = 0; |
1da177e4 LT |
4321 | } |
4322 | ||
4323 | if (updelay < 0) { | |
91565ebb JP |
4324 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4325 | updelay, INT_MAX); | |
1da177e4 LT |
4326 | updelay = 0; |
4327 | } | |
4328 | ||
4329 | if (downdelay < 0) { | |
91565ebb JP |
4330 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4331 | downdelay, INT_MAX); | |
1da177e4 LT |
4332 | downdelay = 0; |
4333 | } | |
4334 | ||
4335 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
91565ebb JP |
4336 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4337 | use_carrier); | |
1da177e4 LT |
4338 | use_carrier = 1; |
4339 | } | |
4340 | ||
ad246c99 | 4341 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
4342 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
4343 | num_peer_notif); | |
ad246c99 BH |
4344 | num_peer_notif = 1; |
4345 | } | |
4346 | ||
834db4bc | 4347 | /* reset values for 802.3ad/TLB/ALB */ |
267bed77 | 4348 | if (!bond_mode_uses_arp(bond_mode)) { |
1da177e4 | 4349 | if (!miimon) { |
91565ebb JP |
4350 | pr_warn("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n"); |
4351 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 4352 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4353 | } |
4354 | } | |
4355 | ||
bb1d9123 | 4356 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
4357 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
4358 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
4359 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
4360 | } | |
4361 | ||
ebd8e497 | 4362 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
4363 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
4364 | all_slaves_active); | |
ebd8e497 AG |
4365 | all_slaves_active = 0; |
4366 | } | |
4367 | ||
c2952c31 | 4368 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
4369 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
4370 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
4371 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
4372 | } | |
4373 | ||
aa59d851 NA |
4374 | bond_opt_initval(&newval, packets_per_slave); |
4375 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
4376 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
4377 | packets_per_slave, USHRT_MAX); | |
4378 | packets_per_slave = 1; | |
4379 | } | |
4380 | ||
1da177e4 | 4381 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4382 | 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", |
4383 | updelay); | |
1da177e4 LT |
4384 | } |
4385 | ||
4386 | if (!miimon) { | |
4387 | if (updelay || downdelay) { | |
4388 | /* just warn the user the up/down delay will have | |
4389 | * no effect since miimon is zero... | |
4390 | */ | |
91565ebb JP |
4391 | pr_warn("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", |
4392 | updelay, downdelay); | |
1da177e4 LT |
4393 | } |
4394 | } else { | |
4395 | /* don't allow arp monitoring */ | |
4396 | if (arp_interval) { | |
91565ebb JP |
4397 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4398 | miimon, arp_interval); | |
1da177e4 LT |
4399 | arp_interval = 0; |
4400 | } | |
4401 | ||
4402 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
4403 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4404 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
4405 | } |
4406 | ||
4407 | updelay /= miimon; | |
4408 | ||
4409 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
4410 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4411 | downdelay, miimon, | |
4412 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4413 | } |
4414 | ||
4415 | downdelay /= miimon; | |
4416 | } | |
4417 | ||
4418 | if (arp_interval < 0) { | |
91565ebb JP |
4419 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4420 | arp_interval, INT_MAX); | |
7bdb04ed | 4421 | arp_interval = 0; |
1da177e4 LT |
4422 | } |
4423 | ||
5a5c5fd4 | 4424 | for (arp_ip_count = 0, i = 0; |
4425 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
89015c18 | 4426 | __be32 ip; |
547942ca NA |
4427 | |
4428 | /* not a complete check, but good enough to catch mistakes */ | |
89015c18 | 4429 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || |
2807a9fe | 4430 | !bond_is_ip_target_ok(ip)) { |
91565ebb JP |
4431 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4432 | arp_ip_target[i]); | |
1da177e4 LT |
4433 | arp_interval = 0; |
4434 | } else { | |
0afee4e8 VF |
4435 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4436 | arp_target[arp_ip_count++] = ip; | |
4437 | else | |
91565ebb JP |
4438 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
4439 | &ip); | |
1da177e4 LT |
4440 | } |
4441 | } | |
4442 | ||
4443 | if (arp_interval && !arp_ip_count) { | |
4444 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
4445 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4446 | arp_interval); | |
1da177e4 LT |
4447 | arp_interval = 0; |
4448 | } | |
4449 | ||
f5b2b966 | 4450 | if (arp_validate) { |
f5b2b966 | 4451 | if (!arp_interval) { |
a4aee5c8 | 4452 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4453 | return -EINVAL; |
4454 | } | |
4455 | ||
16228881 NA |
4456 | bond_opt_initstr(&newval, arp_validate); |
4457 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
4458 | &newval); | |
4459 | if (!valptr) { | |
a4aee5c8 | 4460 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 4461 | arp_validate); |
f5b2b966 JV |
4462 | return -EINVAL; |
4463 | } | |
16228881 NA |
4464 | arp_validate_value = valptr->value; |
4465 | } else { | |
f5b2b966 | 4466 | arp_validate_value = 0; |
16228881 | 4467 | } |
f5b2b966 | 4468 | |
8599b52e VF |
4469 | arp_all_targets_value = 0; |
4470 | if (arp_all_targets) { | |
edf36b24 NA |
4471 | bond_opt_initstr(&newval, arp_all_targets); |
4472 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
4473 | &newval); | |
4474 | if (!valptr) { | |
8599b52e VF |
4475 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
4476 | arp_all_targets); | |
4477 | arp_all_targets_value = 0; | |
edf36b24 NA |
4478 | } else { |
4479 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
4480 | } |
4481 | } | |
4482 | ||
1da177e4 | 4483 | if (miimon) { |
a4aee5c8 | 4484 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4485 | } else if (arp_interval) { |
16228881 NA |
4486 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
4487 | arp_validate_value); | |
a4aee5c8 | 4488 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 4489 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
4490 | |
4491 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 4492 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 4493 | |
90194264 | 4494 | pr_cont("\n"); |
1da177e4 | 4495 | |
b8a9787e | 4496 | } else if (max_bonds) { |
1da177e4 LT |
4497 | /* miimon and arp_interval not set, we need one so things |
4498 | * work as expected, see bonding.txt for details | |
4499 | */ | |
90194264 | 4500 | 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 |
4501 | } |
4502 | ||
ec0865a9 | 4503 | if (primary && !bond_mode_uses_primary(bond_mode)) { |
1da177e4 LT |
4504 | /* currently, using a primary only makes sense |
4505 | * in active backup, TLB or ALB modes | |
4506 | */ | |
91565ebb JP |
4507 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
4508 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4509 | primary = NULL; |
4510 | } | |
4511 | ||
a549952a | 4512 | if (primary && primary_reselect) { |
388d3a6d NA |
4513 | bond_opt_initstr(&newval, primary_reselect); |
4514 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
4515 | &newval); | |
4516 | if (!valptr) { | |
a4aee5c8 | 4517 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 4518 | primary_reselect); |
a549952a JP |
4519 | return -EINVAL; |
4520 | } | |
388d3a6d | 4521 | primary_reselect_value = valptr->value; |
a549952a JP |
4522 | } else { |
4523 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4524 | } | |
4525 | ||
3915c1e8 | 4526 | if (fail_over_mac) { |
1df6b6aa NA |
4527 | bond_opt_initstr(&newval, fail_over_mac); |
4528 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
4529 | &newval); | |
4530 | if (!valptr) { | |
a4aee5c8 | 4531 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 4532 | fail_over_mac); |
3915c1e8 JV |
4533 | return -EINVAL; |
4534 | } | |
1df6b6aa | 4535 | fail_over_mac_value = valptr->value; |
3915c1e8 | 4536 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 4537 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
4538 | } else { |
4539 | fail_over_mac_value = BOND_FOM_NONE; | |
4540 | } | |
dd957c57 | 4541 | |
6791e466 MB |
4542 | bond_opt_initstr(&newval, "default"); |
4543 | valptr = bond_opt_parse( | |
4544 | bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO), | |
4545 | &newval); | |
4546 | if (!valptr) { | |
4547 | pr_err("Error: No ad_actor_sys_prio default value"); | |
4548 | return -EINVAL; | |
4549 | } | |
4550 | ad_actor_sys_prio = valptr->value; | |
4551 | ||
d22a5fc0 MB |
4552 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY), |
4553 | &newval); | |
4554 | if (!valptr) { | |
4555 | pr_err("Error: No ad_user_port_key default value"); | |
4556 | return -EINVAL; | |
4557 | } | |
4558 | ad_user_port_key = valptr->value; | |
4559 | ||
3a7129e5 | 4560 | if (lp_interval == 0) { |
91565ebb JP |
4561 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
4562 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 4563 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
4564 | } | |
4565 | ||
1da177e4 LT |
4566 | /* fill params struct with the proper values */ |
4567 | params->mode = bond_mode; | |
169a3e66 | 4568 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4569 | params->miimon = miimon; |
ad246c99 | 4570 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4571 | params->arp_interval = arp_interval; |
f5b2b966 | 4572 | params->arp_validate = arp_validate_value; |
8599b52e | 4573 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4574 | params->updelay = updelay; |
4575 | params->downdelay = downdelay; | |
4576 | params->use_carrier = use_carrier; | |
4577 | params->lacp_fast = lacp_fast; | |
4578 | params->primary[0] = 0; | |
a549952a | 4579 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4580 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4581 | params->tx_queues = tx_queues; |
ebd8e497 | 4582 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4583 | params->resend_igmp = resend_igmp; |
655f8919 | 4584 | params->min_links = min_links; |
3a7129e5 | 4585 | params->lp_interval = lp_interval; |
809fa972 | 4586 | params->packets_per_slave = packets_per_slave; |
e9f0fb88 | 4587 | params->tlb_dynamic_lb = 1; /* Default value */ |
6791e466 | 4588 | params->ad_actor_sys_prio = ad_actor_sys_prio; |
74514957 | 4589 | eth_zero_addr(params->ad_actor_system); |
d22a5fc0 | 4590 | params->ad_user_port_key = ad_user_port_key; |
809fa972 HFS |
4591 | if (packets_per_slave > 0) { |
4592 | params->reciprocal_packets_per_slave = | |
4593 | reciprocal_value(packets_per_slave); | |
4594 | } else { | |
4595 | /* reciprocal_packets_per_slave is unused if | |
4596 | * packets_per_slave is 0 or 1, just initialize it | |
4597 | */ | |
4598 | params->reciprocal_packets_per_slave = | |
4599 | (struct reciprocal_value) { 0 }; | |
4600 | } | |
4601 | ||
1da177e4 LT |
4602 | if (primary) { |
4603 | strncpy(params->primary, primary, IFNAMSIZ); | |
4604 | params->primary[IFNAMSIZ - 1] = 0; | |
4605 | } | |
4606 | ||
4607 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4608 | ||
4609 | return 0; | |
4610 | } | |
4611 | ||
0daa2303 | 4612 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4613 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4614 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4615 | |
e8a0464c DM |
4616 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4617 | struct netdev_queue *txq, | |
4618 | void *_unused) | |
c773e847 DM |
4619 | { |
4620 | lockdep_set_class(&txq->_xmit_lock, | |
4621 | &bonding_netdev_xmit_lock_key); | |
4622 | } | |
4623 | ||
4624 | static void bond_set_lockdep_class(struct net_device *dev) | |
4625 | { | |
cf508b12 DM |
4626 | lockdep_set_class(&dev->addr_list_lock, |
4627 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4628 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4629 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4630 | } |
4631 | ||
547942ca | 4632 | /* Called from registration process */ |
181470fc SH |
4633 | static int bond_init(struct net_device *bond_dev) |
4634 | { | |
4635 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4636 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc | 4637 | |
76444f50 | 4638 | netdev_dbg(bond_dev, "Begin bond_init\n"); |
181470fc SH |
4639 | |
4640 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4641 | if (!bond->wq) | |
4642 | return -ENOMEM; | |
4643 | ||
4644 | bond_set_lockdep_class(bond_dev); | |
4645 | ||
ec87fd3b | 4646 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4647 | |
6151b3d4 | 4648 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4649 | |
f073c7ca TI |
4650 | bond_debug_register(bond); |
4651 | ||
409cc1f8 JP |
4652 | /* Ensure valid dev_addr */ |
4653 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4654 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4655 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4656 | |
181470fc SH |
4657 | return 0; |
4658 | } | |
4659 | ||
0a2a78c4 | 4660 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4661 | { |
efacb309 | 4662 | return tx_queues; |
d5da4510 JP |
4663 | } |
4664 | ||
dfe60397 | 4665 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4666 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4667 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4668 | * set up our sysfs entries. | |
4669 | */ | |
ec87fd3b | 4670 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4671 | { |
4672 | struct net_device *bond_dev; | |
e913fb27 P |
4673 | struct bonding *bond; |
4674 | struct alb_bond_info *bond_info; | |
dfe60397 MW |
4675 | int res; |
4676 | ||
4677 | rtnl_lock(); | |
027ea041 | 4678 | |
1c5cae81 | 4679 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
c835a677 | 4680 | name ? name : "bond%d", NET_NAME_UNKNOWN, |
1c5cae81 | 4681 | bond_setup, tx_queues); |
dfe60397 | 4682 | if (!bond_dev) { |
a4aee5c8 | 4683 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4684 | rtnl_unlock(); |
4685 | return -ENOMEM; | |
dfe60397 MW |
4686 | } |
4687 | ||
e913fb27 P |
4688 | /* |
4689 | * Initialize rx_hashtbl_used_head to RLB_NULL_INDEX. | |
4690 | * It is set to 0 by default which is wrong. | |
4691 | */ | |
4692 | bond = netdev_priv(bond_dev); | |
4693 | bond_info = &(BOND_ALB_INFO(bond)); | |
4694 | bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX; | |
4695 | ||
ec87fd3b | 4696 | dev_net_set(bond_dev, net); |
88ead977 EB |
4697 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4698 | ||
dfe60397 | 4699 | res = register_netdevice(bond_dev); |
0daa2303 | 4700 | |
e826eafa PO |
4701 | netif_carrier_off(bond_dev); |
4702 | ||
7e083840 | 4703 | rtnl_unlock(); |
9e2e61fb AW |
4704 | if (res < 0) |
4705 | bond_destructor(bond_dev); | |
30c15ba9 | 4706 | return res; |
dfe60397 MW |
4707 | } |
4708 | ||
2c8c1e72 | 4709 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4710 | { |
15449745 | 4711 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4712 | |
4713 | bn->net = net; | |
4714 | INIT_LIST_HEAD(&bn->dev_list); | |
4715 | ||
ec87fd3b | 4716 | bond_create_proc_dir(bn); |
4c22400a | 4717 | bond_create_sysfs(bn); |
87a7b84b | 4718 | |
15449745 | 4719 | return 0; |
ec87fd3b EB |
4720 | } |
4721 | ||
2c8c1e72 | 4722 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4723 | { |
15449745 | 4724 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4725 | struct bonding *bond, *tmp_bond; |
4726 | LIST_HEAD(list); | |
ec87fd3b | 4727 | |
4c22400a | 4728 | bond_destroy_sysfs(bn); |
69b0216a | 4729 | |
4730 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4731 | rtnl_lock(); | |
4732 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4733 | unregister_netdevice_queue(bond->dev, &list); | |
4734 | unregister_netdevice_many(&list); | |
4735 | rtnl_unlock(); | |
23fa5c2c VF |
4736 | |
4737 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
4738 | } |
4739 | ||
4740 | static struct pernet_operations bond_net_ops = { | |
4741 | .init = bond_net_init, | |
4742 | .exit = bond_net_exit, | |
15449745 EB |
4743 | .id = &bond_net_id, |
4744 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4745 | }; |
4746 | ||
1da177e4 LT |
4747 | static int __init bonding_init(void) |
4748 | { | |
1da177e4 LT |
4749 | int i; |
4750 | int res; | |
4751 | ||
bd33acc3 | 4752 | pr_info("%s", bond_version); |
1da177e4 | 4753 | |
dfe60397 | 4754 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4755 | if (res) |
dfe60397 | 4756 | goto out; |
1da177e4 | 4757 | |
15449745 | 4758 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4759 | if (res) |
4760 | goto out; | |
027ea041 | 4761 | |
0a2a78c4 | 4762 | res = bond_netlink_init(); |
88ead977 | 4763 | if (res) |
6639104b | 4764 | goto err_link; |
88ead977 | 4765 | |
f073c7ca TI |
4766 | bond_create_debugfs(); |
4767 | ||
1da177e4 | 4768 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4769 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4770 | if (res) |
4771 | goto err; | |
1da177e4 LT |
4772 | } |
4773 | ||
1da177e4 | 4774 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4775 | out: |
1da177e4 | 4776 | return res; |
88ead977 | 4777 | err: |
db298686 | 4778 | bond_destroy_debugfs(); |
0a2a78c4 | 4779 | bond_netlink_fini(); |
6639104b | 4780 | err_link: |
15449745 | 4781 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4782 | goto out; |
dfe60397 | 4783 | |
1da177e4 LT |
4784 | } |
4785 | ||
4786 | static void __exit bonding_exit(void) | |
4787 | { | |
4788 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4789 | ||
f073c7ca | 4790 | bond_destroy_debugfs(); |
ae68c398 | 4791 | |
0a2a78c4 | 4792 | bond_netlink_fini(); |
ffcdedb6 | 4793 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4794 | |
4795 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
547942ca | 4796 | /* Make sure we don't have an imbalance on our netpoll blocking */ |
fb4fa76a | 4797 | WARN_ON(atomic_read(&netpoll_block_tx)); |
e843fa50 | 4798 | #endif |
1da177e4 LT |
4799 | } |
4800 | ||
4801 | module_init(bonding_init); | |
4802 | module_exit(bonding_exit); | |
4803 | MODULE_LICENSE("GPL"); | |
4804 | MODULE_VERSION(DRV_VERSION); | |
4805 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4806 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |