]>
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); |
0d039f33 | 155 | MODULE_PARM_DESC(ad_select, "802.3ad aggregation selection logic; " |
90e62474 AG |
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 | ||
c7d03a00 | 202 | unsigned int bond_net_id __read_mostly; |
1da177e4 | 203 | |
1da177e4 LT |
204 | /*-------------------------- Forward declarations ---------------------------*/ |
205 | ||
181470fc | 206 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 207 | static void bond_uninit(struct net_device *bond_dev); |
bc1f4470 | 208 | static void bond_get_stats(struct net_device *bond_dev, |
209 | struct rtnl_link_stats64 *stats); | |
ee637714 | 210 | static void bond_slave_arr_handler(struct work_struct *work); |
21a75f09 JV |
211 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, |
212 | int mod); | |
1da177e4 LT |
213 | |
214 | /*---------------------------- General routines -----------------------------*/ | |
215 | ||
bd33acc3 | 216 | const char *bond_mode_name(int mode) |
1da177e4 | 217 | { |
77afc92b HE |
218 | static const char *names[] = { |
219 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
220 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
221 | [BOND_MODE_XOR] = "load balancing (xor)", | |
222 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 223 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
224 | [BOND_MODE_TLB] = "transmit load balancing", |
225 | [BOND_MODE_ALB] = "adaptive load balancing", | |
226 | }; | |
227 | ||
f5280948 | 228 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 229 | return "unknown"; |
77afc92b HE |
230 | |
231 | return names[mode]; | |
1da177e4 LT |
232 | } |
233 | ||
234 | /*---------------------------------- VLAN -----------------------------------*/ | |
235 | ||
1da177e4 LT |
236 | /** |
237 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 238 | * |
1da177e4 LT |
239 | * @bond: bond device that got this skb for tx. |
240 | * @skb: hw accel VLAN tagged skb to transmit | |
241 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 242 | */ |
d316dedd | 243 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 244 | struct net_device *slave_dev) |
1da177e4 | 245 | { |
8387451e | 246 | skb->dev = slave_dev; |
374eeb5a | 247 | |
5ee31c68 | 248 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
249 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
250 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 251 | |
e15c3c22 | 252 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 253 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 254 | else |
f6dc31a8 | 255 | dev_queue_xmit(skb); |
1da177e4 LT |
256 | } |
257 | ||
547942ca | 258 | /* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
cc0e4070 | 259 | * We don't protect the slave list iteration with a lock because: |
1da177e4 LT |
260 | * a. This operation is performed in IOCTL context, |
261 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
262 | * c. Holding a lock with BH disabled while directly calling a base driver | |
263 | * entry point is generally a BAD idea. | |
3d632c3f | 264 | * |
1da177e4 LT |
265 | * The design of synchronization/protection for this operation in the 8021q |
266 | * module is good for one or more VLAN devices over a single physical device | |
267 | * and cannot be extended for a teaming solution like bonding, so there is a | |
268 | * potential race condition here where a net device from the vlan group might | |
269 | * be referenced (either by a base driver or the 8021q code) while it is being | |
270 | * removed from the system. However, it turns out we're not making matters | |
271 | * worse, and if it works for regular VLAN usage it will work here too. | |
272 | */ | |
273 | ||
1da177e4 LT |
274 | /** |
275 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
276 | * @bond_dev: bonding net device that got called | |
277 | * @vid: vlan id being added | |
278 | */ | |
80d5c368 PM |
279 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
280 | __be16 proto, u16 vid) | |
1da177e4 | 281 | { |
454d7c9b | 282 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 283 | struct slave *slave, *rollback_slave; |
9caff1e7 | 284 | struct list_head *iter; |
dec1e90e | 285 | int res; |
1da177e4 | 286 | |
9caff1e7 | 287 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 288 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
289 | if (res) |
290 | goto unwind; | |
1da177e4 LT |
291 | } |
292 | ||
8e586137 | 293 | return 0; |
87002b03 JP |
294 | |
295 | unwind: | |
81f23b13 | 296 | /* unwind to the slave that failed */ |
9caff1e7 | 297 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
298 | if (rollback_slave == slave) |
299 | break; | |
300 | ||
301 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
302 | } | |
87002b03 JP |
303 | |
304 | return res; | |
1da177e4 LT |
305 | } |
306 | ||
307 | /** | |
308 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
309 | * @bond_dev: bonding net device that got called | |
310 | * @vid: vlan id being removed | |
311 | */ | |
80d5c368 PM |
312 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
313 | __be16 proto, u16 vid) | |
1da177e4 | 314 | { |
454d7c9b | 315 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 316 | struct list_head *iter; |
1da177e4 | 317 | struct slave *slave; |
1da177e4 | 318 | |
9caff1e7 | 319 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 320 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 321 | |
e868b0c9 VF |
322 | if (bond_is_lb(bond)) |
323 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
324 | |
325 | return 0; | |
1da177e4 LT |
326 | } |
327 | ||
1da177e4 LT |
328 | /*------------------------------- Link status -------------------------------*/ |
329 | ||
547942ca | 330 | /* Set the carrier state for the master according to the state of its |
ff59c456 JV |
331 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, |
332 | * do special 802.3ad magic. | |
333 | * | |
334 | * Returns zero if carrier state does not change, nonzero if it does. | |
335 | */ | |
2477bc9a | 336 | int bond_set_carrier(struct bonding *bond) |
ff59c456 | 337 | { |
9caff1e7 | 338 | struct list_head *iter; |
ff59c456 | 339 | struct slave *slave; |
ff59c456 | 340 | |
0965a1f3 | 341 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
342 | goto down; |
343 | ||
01844098 | 344 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
ff59c456 JV |
345 | return bond_3ad_set_carrier(bond); |
346 | ||
9caff1e7 | 347 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
348 | if (slave->link == BOND_LINK_UP) { |
349 | if (!netif_carrier_ok(bond->dev)) { | |
350 | netif_carrier_on(bond->dev); | |
351 | return 1; | |
352 | } | |
353 | return 0; | |
354 | } | |
355 | } | |
356 | ||
357 | down: | |
358 | if (netif_carrier_ok(bond->dev)) { | |
359 | netif_carrier_off(bond->dev); | |
360 | return 1; | |
361 | } | |
362 | return 0; | |
363 | } | |
364 | ||
547942ca | 365 | /* Get link speed and duplex from the slave's base driver |
1da177e4 | 366 | * using ethtool. If for some reason the call fails or the |
98f41f69 | 367 | * values are invalid, set speed and duplex to -1, |
c4adfc82 MB |
368 | * and return. Return 1 if speed or duplex settings are |
369 | * UNKNOWN; 0 otherwise. | |
1da177e4 | 370 | */ |
c4adfc82 | 371 | static int bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
372 | { |
373 | struct net_device *slave_dev = slave->dev; | |
9856909c | 374 | struct ethtool_link_ksettings ecmd; |
61a44b9c | 375 | int res; |
1da177e4 | 376 | |
589665f5 DC |
377 | slave->speed = SPEED_UNKNOWN; |
378 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 379 | |
9856909c | 380 | res = __ethtool_get_link_ksettings(slave_dev, &ecmd); |
61a44b9c | 381 | if (res < 0) |
c4adfc82 | 382 | return 1; |
9856909c | 383 | if (ecmd.base.speed == 0 || ecmd.base.speed == ((__u32)-1)) |
c4adfc82 | 384 | return 1; |
9856909c | 385 | switch (ecmd.base.duplex) { |
1da177e4 LT |
386 | case DUPLEX_FULL: |
387 | case DUPLEX_HALF: | |
388 | break; | |
389 | default: | |
c4adfc82 | 390 | return 1; |
1da177e4 LT |
391 | } |
392 | ||
9856909c DD |
393 | slave->speed = ecmd.base.speed; |
394 | slave->duplex = ecmd.base.duplex; | |
1da177e4 | 395 | |
c4adfc82 | 396 | return 0; |
1da177e4 LT |
397 | } |
398 | ||
07699f9a | 399 | const char *bond_slave_link_status(s8 link) |
400 | { | |
401 | switch (link) { | |
402 | case BOND_LINK_UP: | |
403 | return "up"; | |
404 | case BOND_LINK_FAIL: | |
405 | return "going down"; | |
406 | case BOND_LINK_DOWN: | |
407 | return "down"; | |
408 | case BOND_LINK_BACK: | |
409 | return "going back"; | |
410 | default: | |
411 | return "unknown"; | |
412 | } | |
413 | } | |
414 | ||
547942ca | 415 | /* if <dev> supports MII link status reporting, check its link status. |
1da177e4 LT |
416 | * |
417 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 418 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
419 | * |
420 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
421 | * can't tell and just pretend it is), or 0, meaning that the link is | |
422 | * down. | |
423 | * | |
424 | * If reporting is non-zero, instead of faking link up, return -1 if | |
425 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
426 | * support them). If use_carrier is set, return whatever it says. | |
427 | * It'd be nice if there was a good way to tell if a driver supports | |
428 | * netif_carrier, but there really isn't. | |
429 | */ | |
3d632c3f SH |
430 | static int bond_check_dev_link(struct bonding *bond, |
431 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 432 | { |
eb7cc59a | 433 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 434 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
435 | struct ifreq ifr; |
436 | struct mii_ioctl_data *mii; | |
1da177e4 | 437 | |
6c988853 PG |
438 | if (!reporting && !netif_running(slave_dev)) |
439 | return 0; | |
440 | ||
eb7cc59a | 441 | if (bond->params.use_carrier) |
1da177e4 | 442 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 443 | |
29112f4e | 444 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
445 | if (slave_dev->ethtool_ops->get_link) |
446 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
447 | BMSR_LSTATUS : 0; | |
29112f4e | 448 | |
3d632c3f | 449 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 450 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 | 451 | if (ioctl) { |
547942ca NA |
452 | /* TODO: set pointer to correct ioctl on a per team member |
453 | * bases to make this more efficient. that is, once | |
454 | * we determine the correct ioctl, we will always | |
455 | * call it and not the others for that team | |
456 | * member. | |
457 | */ | |
458 | ||
459 | /* We cannot assume that SIOCGMIIPHY will also read a | |
1da177e4 LT |
460 | * register; not all network drivers (e.g., e100) |
461 | * support that. | |
462 | */ | |
463 | ||
464 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
465 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
466 | mii = if_mii(&ifr); | |
4ad41c1e | 467 | if (ioctl(slave_dev, &ifr, SIOCGMIIPHY) == 0) { |
1da177e4 | 468 | mii->reg_num = MII_BMSR; |
4ad41c1e | 469 | if (ioctl(slave_dev, &ifr, SIOCGMIIREG) == 0) |
3d632c3f | 470 | return mii->val_out & BMSR_LSTATUS; |
1da177e4 LT |
471 | } |
472 | } | |
473 | ||
547942ca | 474 | /* If reporting, report that either there's no dev->do_ioctl, |
61a44b9c | 475 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
476 | * cannot report link status). If not reporting, pretend |
477 | * we're ok. | |
478 | */ | |
3d632c3f | 479 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
480 | } |
481 | ||
482 | /*----------------------------- Multicast list ------------------------------*/ | |
483 | ||
547942ca | 484 | /* Push the promiscuity flag down to appropriate slaves */ |
7e1a1ac1 | 485 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 486 | { |
9caff1e7 | 487 | struct list_head *iter; |
7e1a1ac1 | 488 | int err = 0; |
9caff1e7 | 489 | |
ec0865a9 | 490 | if (bond_uses_primary(bond)) { |
14056e79 | 491 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 492 | |
4740d638 ED |
493 | if (curr_active) |
494 | err = dev_set_promiscuity(curr_active->dev, inc); | |
1da177e4 LT |
495 | } else { |
496 | struct slave *slave; | |
dec1e90e | 497 | |
9caff1e7 | 498 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
499 | err = dev_set_promiscuity(slave->dev, inc); |
500 | if (err) | |
501 | return err; | |
1da177e4 LT |
502 | } |
503 | } | |
7e1a1ac1 | 504 | return err; |
1da177e4 LT |
505 | } |
506 | ||
547942ca | 507 | /* Push the allmulti flag down to all slaves */ |
7e1a1ac1 | 508 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 509 | { |
9caff1e7 | 510 | struct list_head *iter; |
7e1a1ac1 | 511 | int err = 0; |
9caff1e7 | 512 | |
ec0865a9 | 513 | if (bond_uses_primary(bond)) { |
14056e79 | 514 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 515 | |
4740d638 ED |
516 | if (curr_active) |
517 | err = dev_set_allmulti(curr_active->dev, inc); | |
1da177e4 LT |
518 | } else { |
519 | struct slave *slave; | |
dec1e90e | 520 | |
9caff1e7 | 521 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
522 | err = dev_set_allmulti(slave->dev, inc); |
523 | if (err) | |
524 | return err; | |
1da177e4 LT |
525 | } |
526 | } | |
7e1a1ac1 | 527 | return err; |
1da177e4 LT |
528 | } |
529 | ||
547942ca | 530 | /* Retrieve the list of registered multicast addresses for the bonding |
5a37e8ca FL |
531 | * device and retransmit an IGMP JOIN request to the current active |
532 | * slave. | |
533 | */ | |
f2369109 | 534 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 535 | { |
f2369109 | 536 | struct bonding *bond = container_of(work, struct bonding, |
537 | mcast_work.work); | |
538 | ||
4aa5dee4 | 539 | if (!rtnl_trylock()) { |
4beac029 | 540 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 541 | return; |
5a37e8ca | 542 | } |
4aa5dee4 | 543 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 544 | |
4f5474e7 NA |
545 | if (bond->igmp_retrans > 1) { |
546 | bond->igmp_retrans--; | |
c2952c31 | 547 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 548 | } |
f2369109 | 549 | rtnl_unlock(); |
5a37e8ca FL |
550 | } |
551 | ||
547942ca | 552 | /* Flush bond's hardware addresses from slave */ |
303d1cbf | 553 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 554 | struct net_device *slave_dev) |
1da177e4 | 555 | { |
454d7c9b | 556 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 557 | |
303d1cbf JV |
558 | dev_uc_unsync(slave_dev, bond_dev); |
559 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 | 560 | |
01844098 | 561 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
562 | /* del lacpdu mc addr from mc list */ |
563 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
564 | ||
22bedad3 | 565 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
566 | } |
567 | } | |
568 | ||
569 | /*--------------------------- Active slave change ---------------------------*/ | |
570 | ||
303d1cbf | 571 | /* Update the hardware address list and promisc/allmulti for the new and |
ec0865a9 VF |
572 | * old active slaves (if any). Modes that are not using primary keep all |
573 | * slaves up date at all times; only the modes that use primary need to call | |
303d1cbf | 574 | * this function to swap these settings during a failover. |
1da177e4 | 575 | */ |
303d1cbf JV |
576 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
577 | struct slave *old_active) | |
1da177e4 | 578 | { |
1da177e4 | 579 | if (old_active) { |
3d632c3f | 580 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 581 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 582 | |
3d632c3f | 583 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 584 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 585 | |
303d1cbf | 586 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
587 | } |
588 | ||
589 | if (new_active) { | |
7e1a1ac1 | 590 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 591 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 592 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 593 | |
3d632c3f | 594 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 595 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 596 | |
d632ce98 | 597 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
598 | dev_uc_sync(new_active->dev, bond->dev); |
599 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 600 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
601 | } |
602 | } | |
603 | ||
ae0d6750 | 604 | /** |
605 | * bond_set_dev_addr - clone slave's address to bond | |
606 | * @bond_dev: bond net device | |
607 | * @slave_dev: slave net device | |
608 | * | |
609 | * Should be called with RTNL held. | |
610 | */ | |
611 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
612 | struct net_device *slave_dev) | |
613 | { | |
c6140a29 ZS |
614 | netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->name=%s slave_dev->addr_len=%d\n", |
615 | bond_dev, slave_dev, slave_dev->name, slave_dev->addr_len); | |
ae0d6750 | 616 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
617 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
618 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
619 | } | |
620 | ||
a951bc1e | 621 | static struct slave *bond_get_old_active(struct bonding *bond, |
622 | struct slave *new_active) | |
623 | { | |
624 | struct slave *slave; | |
625 | struct list_head *iter; | |
626 | ||
627 | bond_for_each_slave(bond, slave, iter) { | |
628 | if (slave == new_active) | |
629 | continue; | |
630 | ||
631 | if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr)) | |
632 | return slave; | |
633 | } | |
634 | ||
635 | return NULL; | |
636 | } | |
637 | ||
547942ca | 638 | /* bond_do_fail_over_mac |
3915c1e8 JV |
639 | * |
640 | * Perform special MAC address swapping for fail_over_mac settings | |
641 | * | |
1c72cfdc | 642 | * Called with RTNL |
3915c1e8 JV |
643 | */ |
644 | static void bond_do_fail_over_mac(struct bonding *bond, | |
645 | struct slave *new_active, | |
646 | struct slave *old_active) | |
647 | { | |
faeeb317 JW |
648 | u8 tmp_mac[MAX_ADDR_LEN]; |
649 | struct sockaddr_storage ss; | |
3915c1e8 JV |
650 | int rv; |
651 | ||
652 | switch (bond->params.fail_over_mac) { | |
653 | case BOND_FOM_ACTIVE: | |
1c72cfdc | 654 | if (new_active) |
ae0d6750 | 655 | bond_set_dev_addr(bond->dev, new_active->dev); |
3915c1e8 JV |
656 | break; |
657 | case BOND_FOM_FOLLOW: | |
547942ca | 658 | /* if new_active && old_active, swap them |
3915c1e8 JV |
659 | * if just old_active, do nothing (going to no active slave) |
660 | * if just new_active, set new_active to bond's MAC | |
661 | */ | |
662 | if (!new_active) | |
663 | return; | |
664 | ||
a951bc1e | 665 | if (!old_active) |
666 | old_active = bond_get_old_active(bond, new_active); | |
667 | ||
3915c1e8 | 668 | if (old_active) { |
faeeb317 JW |
669 | bond_hw_addr_copy(tmp_mac, new_active->dev->dev_addr, |
670 | new_active->dev->addr_len); | |
671 | bond_hw_addr_copy(ss.__data, | |
672 | old_active->dev->dev_addr, | |
673 | old_active->dev->addr_len); | |
674 | ss.ss_family = new_active->dev->type; | |
3915c1e8 | 675 | } else { |
faeeb317 JW |
676 | bond_hw_addr_copy(ss.__data, bond->dev->dev_addr, |
677 | bond->dev->addr_len); | |
678 | ss.ss_family = bond->dev->type; | |
3915c1e8 JV |
679 | } |
680 | ||
faeeb317 JW |
681 | rv = dev_set_mac_address(new_active->dev, |
682 | (struct sockaddr *)&ss); | |
3915c1e8 | 683 | if (rv) { |
76444f50 VF |
684 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
685 | -rv, new_active->dev->name); | |
3915c1e8 JV |
686 | goto out; |
687 | } | |
688 | ||
689 | if (!old_active) | |
690 | goto out; | |
691 | ||
faeeb317 JW |
692 | bond_hw_addr_copy(ss.__data, tmp_mac, |
693 | new_active->dev->addr_len); | |
694 | ss.ss_family = old_active->dev->type; | |
3915c1e8 | 695 | |
faeeb317 JW |
696 | rv = dev_set_mac_address(old_active->dev, |
697 | (struct sockaddr *)&ss); | |
3915c1e8 | 698 | if (rv) |
76444f50 VF |
699 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
700 | -rv, new_active->dev->name); | |
3915c1e8 | 701 | out: |
3915c1e8 JV |
702 | break; |
703 | default: | |
76444f50 VF |
704 | netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n", |
705 | bond->params.fail_over_mac); | |
3915c1e8 JV |
706 | break; |
707 | } | |
708 | ||
709 | } | |
710 | ||
b5a983f3 | 711 | static struct slave *bond_choose_primary_or_current(struct bonding *bond) |
a549952a | 712 | { |
059b47e8 | 713 | struct slave *prim = rtnl_dereference(bond->primary_slave); |
1c72cfdc | 714 | struct slave *curr = rtnl_dereference(bond->curr_active_slave); |
a549952a | 715 | |
b5a983f3 MR |
716 | if (!prim || prim->link != BOND_LINK_UP) { |
717 | if (!curr || curr->link != BOND_LINK_UP) | |
718 | return NULL; | |
719 | return curr; | |
720 | } | |
721 | ||
a549952a JP |
722 | if (bond->force_primary) { |
723 | bond->force_primary = false; | |
b5a983f3 MR |
724 | return prim; |
725 | } | |
726 | ||
727 | if (!curr || curr->link != BOND_LINK_UP) | |
728 | return prim; | |
729 | ||
730 | /* At this point, prim and curr are both up */ | |
731 | switch (bond->params.primary_reselect) { | |
732 | case BOND_PRI_RESELECT_ALWAYS: | |
733 | return prim; | |
734 | case BOND_PRI_RESELECT_BETTER: | |
735 | if (prim->speed < curr->speed) | |
736 | return curr; | |
737 | if (prim->speed == curr->speed && prim->duplex <= curr->duplex) | |
738 | return curr; | |
739 | return prim; | |
740 | case BOND_PRI_RESELECT_FAILURE: | |
741 | return curr; | |
742 | default: | |
743 | netdev_err(bond->dev, "impossible primary_reselect %d\n", | |
744 | bond->params.primary_reselect); | |
745 | return curr; | |
a549952a | 746 | } |
a549952a | 747 | } |
3915c1e8 | 748 | |
1da177e4 | 749 | /** |
b5a983f3 | 750 | * bond_find_best_slave - select the best available slave to be the active one |
1da177e4 | 751 | * @bond: our bonding struct |
1da177e4 LT |
752 | */ |
753 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
754 | { | |
b5a983f3 | 755 | struct slave *slave, *bestslave = NULL; |
77140d29 | 756 | struct list_head *iter; |
1da177e4 | 757 | int mintime = bond->params.updelay; |
1da177e4 | 758 | |
b5a983f3 MR |
759 | slave = bond_choose_primary_or_current(bond); |
760 | if (slave) | |
761 | return slave; | |
1da177e4 | 762 | |
77140d29 VF |
763 | bond_for_each_slave(bond, slave, iter) { |
764 | if (slave->link == BOND_LINK_UP) | |
765 | return slave; | |
b6adc610 | 766 | if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) && |
77140d29 VF |
767 | slave->delay < mintime) { |
768 | mintime = slave->delay; | |
769 | bestslave = slave; | |
1da177e4 LT |
770 | } |
771 | } | |
772 | ||
773 | return bestslave; | |
774 | } | |
775 | ||
ad246c99 BH |
776 | static bool bond_should_notify_peers(struct bonding *bond) |
777 | { | |
4cb4f97b | 778 | struct slave *slave; |
779 | ||
780 | rcu_read_lock(); | |
781 | slave = rcu_dereference(bond->curr_active_slave); | |
782 | rcu_read_unlock(); | |
ad246c99 | 783 | |
76444f50 VF |
784 | netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n", |
785 | slave ? slave->dev->name : "NULL"); | |
ad246c99 BH |
786 | |
787 | if (!slave || !bond->send_peer_notif || | |
b02e3e94 | 788 | !netif_carrier_ok(bond->dev) || |
ad246c99 BH |
789 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) |
790 | return false; | |
791 | ||
ad246c99 BH |
792 | return true; |
793 | } | |
794 | ||
1da177e4 LT |
795 | /** |
796 | * change_active_interface - change the active slave into the specified one | |
797 | * @bond: our bonding struct | |
798 | * @new: the new slave to make the active one | |
799 | * | |
800 | * Set the new slave to the bond's settings and unset them on the old | |
801 | * curr_active_slave. | |
802 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
803 | * | |
804 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
805 | * because it is apparently the best available slave we have, even though its | |
806 | * updelay hasn't timed out yet. | |
807 | * | |
1c72cfdc | 808 | * Caller must hold RTNL. |
1da177e4 | 809 | */ |
a77b5325 | 810 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 | 811 | { |
4740d638 ED |
812 | struct slave *old_active; |
813 | ||
1c72cfdc NA |
814 | ASSERT_RTNL(); |
815 | ||
816 | old_active = rtnl_dereference(bond->curr_active_slave); | |
1da177e4 | 817 | |
3d632c3f | 818 | if (old_active == new_active) |
1da177e4 | 819 | return; |
1da177e4 LT |
820 | |
821 | if (new_active) { | |
8e603460 | 822 | new_active->last_link_up = jiffies; |
b2220cad | 823 | |
1da177e4 | 824 | if (new_active->link == BOND_LINK_BACK) { |
ec0865a9 | 825 | if (bond_uses_primary(bond)) { |
76444f50 VF |
826 | netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n", |
827 | new_active->dev->name, | |
828 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
829 | } |
830 | ||
831 | new_active->delay = 0; | |
5d397061 JP |
832 | bond_set_slave_link_state(new_active, BOND_LINK_UP, |
833 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 834 | |
01844098 | 835 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 836 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 837 | |
58402054 | 838 | if (bond_is_lb(bond)) |
1da177e4 | 839 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 | 840 | } else { |
ec0865a9 | 841 | if (bond_uses_primary(bond)) { |
76444f50 VF |
842 | netdev_info(bond->dev, "making interface %s the new active one\n", |
843 | new_active->dev->name); | |
1da177e4 LT |
844 | } |
845 | } | |
846 | } | |
847 | ||
ec0865a9 | 848 | if (bond_uses_primary(bond)) |
303d1cbf | 849 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 850 | |
58402054 | 851 | if (bond_is_lb(bond)) { |
1da177e4 | 852 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 853 | if (old_active) |
5e5b0665 | 854 | bond_set_slave_inactive_flags(old_active, |
855 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 856 | if (new_active) |
5e5b0665 | 857 | bond_set_slave_active_flags(new_active, |
858 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 859 | } else { |
278b2083 | 860 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 861 | } |
c3ade5ca | 862 | |
01844098 | 863 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) { |
3d632c3f | 864 | if (old_active) |
5e5b0665 | 865 | bond_set_slave_inactive_flags(old_active, |
866 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
867 | |
868 | if (new_active) { | |
ad246c99 BH |
869 | bool should_notify_peers = false; |
870 | ||
5e5b0665 | 871 | bond_set_slave_active_flags(new_active, |
872 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 873 | |
709f8a45 OG |
874 | if (bond->params.fail_over_mac) |
875 | bond_do_fail_over_mac(bond, new_active, | |
876 | old_active); | |
3915c1e8 | 877 | |
ad246c99 BH |
878 | if (netif_running(bond->dev)) { |
879 | bond->send_peer_notif = | |
880 | bond->params.num_peer_notif; | |
881 | should_notify_peers = | |
882 | bond_should_notify_peers(bond); | |
883 | } | |
884 | ||
b7bc2a5b | 885 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 886 | if (should_notify_peers) |
b7bc2a5b AW |
887 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
888 | bond->dev); | |
7893b249 | 889 | } |
c3ade5ca | 890 | } |
a2fd940f | 891 | |
5a37e8ca | 892 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
893 | * all were sent on curr_active_slave. |
894 | * resend only if bond is brought up with the affected | |
547942ca NA |
895 | * bonding modes and the retransmission is enabled |
896 | */ | |
94265cf5 | 897 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && |
ec0865a9 | 898 | ((bond_uses_primary(bond) && new_active) || |
01844098 | 899 | BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) { |
c2952c31 | 900 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 901 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 902 | } |
1da177e4 LT |
903 | } |
904 | ||
905 | /** | |
906 | * bond_select_active_slave - select a new active slave, if needed | |
907 | * @bond: our bonding struct | |
908 | * | |
3d632c3f | 909 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
910 | * - The old curr_active_slave has been released or lost its link. |
911 | * - The primary_slave has got its link back. | |
912 | * - A slave has got its link back and there's no old curr_active_slave. | |
913 | * | |
1c72cfdc | 914 | * Caller must hold RTNL. |
1da177e4 | 915 | */ |
a77b5325 | 916 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
917 | { |
918 | struct slave *best_slave; | |
ff59c456 | 919 | int rv; |
1da177e4 | 920 | |
e0974585 NA |
921 | ASSERT_RTNL(); |
922 | ||
1da177e4 | 923 | best_slave = bond_find_best_slave(bond); |
1c72cfdc | 924 | if (best_slave != rtnl_dereference(bond->curr_active_slave)) { |
1da177e4 | 925 | bond_change_active_slave(bond, best_slave); |
ff59c456 JV |
926 | rv = bond_set_carrier(bond); |
927 | if (!rv) | |
928 | return; | |
929 | ||
d66bd905 | 930 | if (netif_carrier_ok(bond->dev)) |
76444f50 | 931 | netdev_info(bond->dev, "first active interface up!\n"); |
d66bd905 | 932 | else |
76444f50 | 933 | netdev_info(bond->dev, "now running without any active interface!\n"); |
1da177e4 LT |
934 | } |
935 | } | |
936 | ||
f6dc31a8 | 937 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 938 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 939 | { |
8a8efa22 AW |
940 | struct netpoll *np; |
941 | int err = 0; | |
f6dc31a8 | 942 | |
a8779ec1 | 943 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
944 | err = -ENOMEM; |
945 | if (!np) | |
946 | goto out; | |
947 | ||
a8779ec1 | 948 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
949 | if (err) { |
950 | kfree(np); | |
951 | goto out; | |
f6dc31a8 | 952 | } |
8a8efa22 AW |
953 | slave->np = np; |
954 | out: | |
955 | return err; | |
956 | } | |
957 | static inline void slave_disable_netpoll(struct slave *slave) | |
958 | { | |
959 | struct netpoll *np = slave->np; | |
960 | ||
961 | if (!np) | |
962 | return; | |
963 | ||
964 | slave->np = NULL; | |
2cde6acd | 965 | __netpoll_free_async(np); |
8a8efa22 | 966 | } |
f6dc31a8 WC |
967 | |
968 | static void bond_poll_controller(struct net_device *bond_dev) | |
969 | { | |
616f4541 MB |
970 | struct bonding *bond = netdev_priv(bond_dev); |
971 | struct slave *slave = NULL; | |
972 | struct list_head *iter; | |
973 | struct ad_info ad_info; | |
974 | struct netpoll_info *ni; | |
975 | const struct net_device_ops *ops; | |
976 | ||
977 | if (BOND_MODE(bond) == BOND_MODE_8023AD) | |
978 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) | |
979 | return; | |
980 | ||
616f4541 MB |
981 | bond_for_each_slave_rcu(bond, slave, iter) { |
982 | ops = slave->dev->netdev_ops; | |
983 | if (!bond_slave_is_up(slave) || !ops->ndo_poll_controller) | |
984 | continue; | |
985 | ||
986 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
987 | struct aggregator *agg = | |
988 | SLAVE_AD_INFO(slave)->port.aggregator; | |
989 | ||
990 | if (agg && | |
991 | agg->aggregator_identifier != ad_info.aggregator_id) | |
992 | continue; | |
993 | } | |
994 | ||
995 | ni = rcu_dereference_bh(slave->dev->npinfo); | |
996 | if (down_trylock(&ni->dev_lock)) | |
997 | continue; | |
998 | ops->ndo_poll_controller(slave->dev); | |
999 | up(&ni->dev_lock); | |
1000 | } | |
8a8efa22 AW |
1001 | } |
1002 | ||
c4cdef9b | 1003 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1004 | { |
c4cdef9b | 1005 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1006 | struct list_head *iter; |
c2355e1a | 1007 | struct slave *slave; |
c2355e1a | 1008 | |
9caff1e7 | 1009 | bond_for_each_slave(bond, slave, iter) |
b6adc610 | 1010 | if (bond_slave_is_up(slave)) |
8a8efa22 | 1011 | slave_disable_netpoll(slave); |
f6dc31a8 | 1012 | } |
8a8efa22 | 1013 | |
a8779ec1 | 1014 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
1015 | { |
1016 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 1017 | struct list_head *iter; |
f6dc31a8 | 1018 | struct slave *slave; |
dec1e90e | 1019 | int err = 0; |
f6dc31a8 | 1020 | |
9caff1e7 | 1021 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
1022 | err = slave_enable_netpoll(slave); |
1023 | if (err) { | |
c4cdef9b | 1024 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1025 | break; |
f6dc31a8 WC |
1026 | } |
1027 | } | |
8a8efa22 | 1028 | return err; |
f6dc31a8 | 1029 | } |
8a8efa22 AW |
1030 | #else |
1031 | static inline int slave_enable_netpoll(struct slave *slave) | |
1032 | { | |
1033 | return 0; | |
1034 | } | |
1035 | static inline void slave_disable_netpoll(struct slave *slave) | |
1036 | { | |
1037 | } | |
f6dc31a8 WC |
1038 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1039 | { | |
1040 | } | |
f6dc31a8 WC |
1041 | #endif |
1042 | ||
1da177e4 LT |
1043 | /*---------------------------------- IOCTL ----------------------------------*/ |
1044 | ||
c8f44aff | 1045 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1046 | netdev_features_t features) |
8531c5ff | 1047 | { |
b2a103e6 | 1048 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 1049 | struct list_head *iter; |
c8f44aff | 1050 | netdev_features_t mask; |
9b7b165a | 1051 | struct slave *slave; |
b63365a2 | 1052 | |
7889cbee | 1053 | mask = features; |
c158cba3 | 1054 | |
b63365a2 | 1055 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1056 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1057 | |
9caff1e7 | 1058 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
1059 | features = netdev_increment_features(features, |
1060 | slave->dev->features, | |
b2a103e6 MM |
1061 | mask); |
1062 | } | |
b0ce3508 | 1063 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1064 | |
b2a103e6 MM |
1065 | return features; |
1066 | } | |
1067 | ||
a188222b | 1068 | #define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
62f2a3a4 MM |
1069 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ |
1070 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 | 1071 | |
a188222b TH |
1072 | #define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
1073 | NETIF_F_RXCSUM | NETIF_F_ALL_TSO) | |
5a7baa78 | 1074 | |
b2a103e6 MM |
1075 | static void bond_compute_features(struct bonding *bond) |
1076 | { | |
02875878 ED |
1077 | unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | |
1078 | IFF_XMIT_DST_RELEASE_PERM; | |
c8f44aff | 1079 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
5a7baa78 | 1080 | netdev_features_t enc_features = BOND_ENC_FEATURES; |
9caff1e7 VF |
1081 | struct net_device *bond_dev = bond->dev; |
1082 | struct list_head *iter; | |
1083 | struct slave *slave; | |
b2a103e6 | 1084 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1085 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1086 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1087 | |
0965a1f3 | 1088 | if (!bond_has_slaves(bond)) |
b2a103e6 | 1089 | goto done; |
a9b3ace4 | 1090 | vlan_features &= NETIF_F_ALL_FOR_ALL; |
b2a103e6 | 1091 | |
9caff1e7 | 1092 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1093 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1094 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1095 | ||
5a7baa78 OG |
1096 | enc_features = netdev_increment_features(enc_features, |
1097 | slave->dev->hw_enc_features, | |
1098 | BOND_ENC_FEATURES); | |
b6fe83e9 | 1099 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1100 | if (slave->dev->hard_header_len > max_hard_header_len) |
1101 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1102 | |
1103 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1104 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1105 | } |
19cdead3 | 1106 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff | 1107 | |
b63365a2 | 1108 | done: |
b2a103e6 | 1109 | bond_dev->vlan_features = vlan_features; |
24f87d4c | 1110 | bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL; |
0e376bd0 SB |
1111 | bond_dev->gso_max_segs = gso_max_segs; |
1112 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1113 | |
02875878 ED |
1114 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
1115 | if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) && | |
1116 | dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM)) | |
1117 | bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE; | |
b6fe83e9 | 1118 | |
b2a103e6 | 1119 | netdev_change_features(bond_dev); |
8531c5ff AK |
1120 | } |
1121 | ||
872254dd MS |
1122 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1123 | struct net_device *slave_dev) | |
1124 | { | |
00829823 | 1125 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1126 | |
1127 | bond_dev->type = slave_dev->type; | |
1128 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1129 | bond_dev->addr_len = slave_dev->addr_len; | |
1130 | ||
1131 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1132 | slave_dev->addr_len); | |
1133 | } | |
1134 | ||
5b2c4dd2 | 1135 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1136 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1137 | */ |
1138 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1139 | struct slave *slave, |
1140 | struct bonding *bond) | |
5b2c4dd2 | 1141 | { |
2d7011ca | 1142 | if (bond_is_slave_inactive(slave)) { |
01844098 | 1143 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
5b2c4dd2 JP |
1144 | skb->pkt_type != PACKET_BROADCAST && |
1145 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1146 | return false; |
5b2c4dd2 JP |
1147 | return true; |
1148 | } | |
1149 | return false; | |
1150 | } | |
1151 | ||
8a4eb573 | 1152 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1153 | { |
8a4eb573 | 1154 | struct sk_buff *skb = *pskb; |
f1c1775a | 1155 | struct slave *slave; |
0bd80dad | 1156 | struct bonding *bond; |
de063b70 ED |
1157 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1158 | struct slave *); | |
13a8e0c8 | 1159 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1160 | |
8a4eb573 JP |
1161 | skb = skb_share_check(skb, GFP_ATOMIC); |
1162 | if (unlikely(!skb)) | |
1163 | return RX_HANDLER_CONSUMED; | |
1164 | ||
1165 | *pskb = skb; | |
5b2c4dd2 | 1166 | |
35d48903 JP |
1167 | slave = bond_slave_get_rcu(skb->dev); |
1168 | bond = slave->bond; | |
0bd80dad | 1169 | |
4d97480b MH |
1170 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1171 | if (recv_probe) { | |
de063b70 ED |
1172 | ret = recv_probe(skb, bond, slave); |
1173 | if (ret == RX_HANDLER_CONSUMED) { | |
1174 | consume_skb(skb); | |
1175 | return ret; | |
3aba891d JP |
1176 | } |
1177 | } | |
1178 | ||
b89f04c6 CL |
1179 | /* don't change skb->dev for link-local packets */ |
1180 | if (is_link_local_ether_addr(eth_hdr(skb)->h_dest)) | |
1181 | return RX_HANDLER_PASS; | |
d66bd905 | 1182 | if (bond_should_deliver_exact_match(skb, slave, bond)) |
8a4eb573 | 1183 | return RX_HANDLER_EXACT; |
5b2c4dd2 | 1184 | |
35d48903 | 1185 | skb->dev = bond->dev; |
5b2c4dd2 | 1186 | |
01844098 | 1187 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
35d48903 | 1188 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1189 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1190 | |
541ac7c9 CG |
1191 | if (unlikely(skb_cow_head(skb, |
1192 | skb->data - skb_mac_header(skb)))) { | |
1193 | kfree_skb(skb); | |
8a4eb573 | 1194 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1195 | } |
faeeb317 JW |
1196 | bond_hw_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, |
1197 | bond->dev->addr_len); | |
5b2c4dd2 JP |
1198 | } |
1199 | ||
13a8e0c8 | 1200 | return ret; |
5b2c4dd2 JP |
1201 | } |
1202 | ||
41f0b049 | 1203 | static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond) |
471cb5a3 | 1204 | { |
41f0b049 JP |
1205 | switch (BOND_MODE(bond)) { |
1206 | case BOND_MODE_ROUNDROBIN: | |
1207 | return NETDEV_LAG_TX_TYPE_ROUNDROBIN; | |
1208 | case BOND_MODE_ACTIVEBACKUP: | |
1209 | return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP; | |
1210 | case BOND_MODE_BROADCAST: | |
1211 | return NETDEV_LAG_TX_TYPE_BROADCAST; | |
1212 | case BOND_MODE_XOR: | |
1213 | case BOND_MODE_8023AD: | |
1214 | return NETDEV_LAG_TX_TYPE_HASH; | |
1215 | default: | |
1216 | return NETDEV_LAG_TX_TYPE_UNKNOWN; | |
1217 | } | |
1218 | } | |
1219 | ||
1220 | static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave) | |
1221 | { | |
1222 | struct netdev_lag_upper_info lag_upper_info; | |
471cb5a3 JP |
1223 | int err; |
1224 | ||
41f0b049 JP |
1225 | lag_upper_info.tx_type = bond_lag_tx_type(bond); |
1226 | err = netdev_master_upper_dev_link(slave->dev, bond->dev, slave, | |
1227 | &lag_upper_info); | |
471cb5a3 JP |
1228 | if (err) |
1229 | return err; | |
41f0b049 | 1230 | rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1231 | return 0; |
1232 | } | |
1233 | ||
41f0b049 | 1234 | static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave) |
471cb5a3 | 1235 | { |
41f0b049 JP |
1236 | netdev_upper_dev_unlink(slave->dev, bond->dev); |
1237 | slave->dev->flags &= ~IFF_SLAVE; | |
1238 | rtmsg_ifinfo(RTM_NEWLINK, slave->dev, IFF_SLAVE, GFP_KERNEL); | |
471cb5a3 JP |
1239 | } |
1240 | ||
3fdddd85 | 1241 | static struct slave *bond_alloc_slave(struct bonding *bond) |
1242 | { | |
1243 | struct slave *slave = NULL; | |
1244 | ||
d66bd905 | 1245 | slave = kzalloc(sizeof(*slave), GFP_KERNEL); |
3fdddd85 | 1246 | if (!slave) |
1247 | return NULL; | |
1248 | ||
01844098 | 1249 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
3fdddd85 | 1250 | SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info), |
1251 | GFP_KERNEL); | |
1252 | if (!SLAVE_AD_INFO(slave)) { | |
1253 | kfree(slave); | |
1254 | return NULL; | |
1255 | } | |
1256 | } | |
1257 | return slave; | |
1258 | } | |
1259 | ||
1260 | static void bond_free_slave(struct slave *slave) | |
1261 | { | |
1262 | struct bonding *bond = bond_get_bond_by_slave(slave); | |
1263 | ||
01844098 | 1264 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3fdddd85 | 1265 | kfree(SLAVE_AD_INFO(slave)); |
1266 | ||
1267 | kfree(slave); | |
1268 | } | |
1269 | ||
69a2338e MS |
1270 | static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info) |
1271 | { | |
1272 | info->bond_mode = BOND_MODE(bond); | |
1273 | info->miimon = bond->params.miimon; | |
1274 | info->num_slaves = bond->slave_cnt; | |
1275 | } | |
1276 | ||
1277 | static void bond_fill_ifslave(struct slave *slave, struct ifslave *info) | |
1278 | { | |
1279 | strcpy(info->slave_name, slave->dev->name); | |
1280 | info->link = slave->link; | |
1281 | info->state = bond_slave_state(slave); | |
1282 | info->link_failure_count = slave->link_failure_count; | |
1283 | } | |
1284 | ||
92e584fe MS |
1285 | static void bond_netdev_notify(struct net_device *dev, |
1286 | struct netdev_bonding_info *info) | |
69e61133 | 1287 | { |
69e61133 | 1288 | rtnl_lock(); |
92e584fe | 1289 | netdev_bonding_info_change(dev, info); |
69e61133 MS |
1290 | rtnl_unlock(); |
1291 | } | |
1292 | ||
1293 | static void bond_netdev_notify_work(struct work_struct *_work) | |
1294 | { | |
1295 | struct netdev_notify_work *w = | |
1296 | container_of(_work, struct netdev_notify_work, work.work); | |
1297 | ||
92e584fe | 1298 | bond_netdev_notify(w->dev, &w->bonding_info); |
69e61133 | 1299 | dev_put(w->dev); |
92e584fe | 1300 | kfree(w); |
69e61133 MS |
1301 | } |
1302 | ||
1303 | void bond_queue_slave_event(struct slave *slave) | |
1304 | { | |
92e584fe | 1305 | struct bonding *bond = slave->bond; |
69e61133 MS |
1306 | struct netdev_notify_work *nnw = kzalloc(sizeof(*nnw), GFP_ATOMIC); |
1307 | ||
1308 | if (!nnw) | |
1309 | return; | |
1310 | ||
92e584fe | 1311 | dev_hold(slave->dev); |
69e61133 | 1312 | nnw->dev = slave->dev; |
92e584fe MS |
1313 | bond_fill_ifslave(slave, &nnw->bonding_info.slave); |
1314 | bond_fill_ifbond(bond, &nnw->bonding_info.master); | |
1315 | INIT_DELAYED_WORK(&nnw->work, bond_netdev_notify_work); | |
69e61133 | 1316 | |
92e584fe | 1317 | queue_delayed_work(slave->bond->wq, &nnw->work, 0); |
69e61133 MS |
1318 | } |
1319 | ||
f7c7eb7f JP |
1320 | void bond_lower_state_changed(struct slave *slave) |
1321 | { | |
1322 | struct netdev_lag_lower_state_info info; | |
1323 | ||
1324 | info.link_up = slave->link == BOND_LINK_UP || | |
1325 | slave->link == BOND_LINK_FAIL; | |
1326 | info.tx_enabled = bond_is_active_slave(slave); | |
1327 | netdev_lower_state_changed(slave->dev, &info); | |
1328 | } | |
1329 | ||
1da177e4 | 1330 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1331 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1332 | { |
454d7c9b | 1333 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1334 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1335 | struct slave *new_slave = NULL, *prev_slave; |
faeeb317 | 1336 | struct sockaddr_storage ss; |
1da177e4 | 1337 | int link_reporting; |
8599b52e | 1338 | int res = 0, i; |
1da177e4 | 1339 | |
c772dde3 BH |
1340 | if (!bond->params.use_carrier && |
1341 | slave_dev->ethtool_ops->get_link == NULL && | |
1342 | slave_ops->ndo_do_ioctl == NULL) { | |
76444f50 VF |
1343 | netdev_warn(bond_dev, "no link monitoring support for %s\n", |
1344 | slave_dev->name); | |
1da177e4 LT |
1345 | } |
1346 | ||
24b27fc4 MB |
1347 | /* already in-use? */ |
1348 | if (netdev_is_rx_handler_busy(slave_dev)) { | |
1349 | netdev_err(bond_dev, | |
1350 | "Error: Device is in use and cannot be 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); | |
1e2a8868 | 1386 | return -EPERM; |
217df670 | 1387 | } |
1da177e4 | 1388 | |
872254dd MS |
1389 | /* set bonding device ether type by slave - bonding netdevices are |
1390 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1391 | * there is a need to override some of the type dependent attribs/funcs. | |
1392 | * | |
1393 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1394 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1395 | */ | |
0965a1f3 | 1396 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1397 | if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1398 | netdev_dbg(bond_dev, "change device type from %d to %d\n", |
1399 | bond_dev->type, slave_dev->type); | |
75c78500 | 1400 | |
b7bc2a5b AW |
1401 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1402 | bond_dev); | |
3ca5b404 JP |
1403 | res = notifier_to_errno(res); |
1404 | if (res) { | |
76444f50 | 1405 | netdev_err(bond_dev, "refused to change device type\n"); |
1e2a8868 | 1406 | return -EBUSY; |
3ca5b404 | 1407 | } |
75c78500 | 1408 | |
32a806c1 | 1409 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1410 | dev_uc_flush(bond_dev); |
22bedad3 | 1411 | dev_mc_flush(bond_dev); |
32a806c1 | 1412 | |
e36b9d16 MS |
1413 | if (slave_dev->type != ARPHRD_ETHER) |
1414 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1415 | else { |
e36b9d16 | 1416 | ether_setup(bond_dev); |
550fd08c NH |
1417 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1418 | } | |
75c78500 | 1419 | |
b7bc2a5b AW |
1420 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1421 | bond_dev); | |
e36b9d16 | 1422 | } |
872254dd | 1423 | } else if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1424 | netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n", |
1425 | slave_dev->name, slave_dev->type, bond_dev->type); | |
1e2a8868 | 1426 | return -EINVAL; |
872254dd MS |
1427 | } |
1428 | ||
1533e773 MB |
1429 | if (slave_dev->type == ARPHRD_INFINIBAND && |
1430 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { | |
1431 | netdev_warn(bond_dev, "Type (%d) supports only active-backup mode\n", | |
1432 | slave_dev->type); | |
1433 | res = -EOPNOTSUPP; | |
1434 | goto err_undo_flags; | |
1435 | } | |
1436 | ||
1437 | if (!slave_ops->ndo_set_mac_address || | |
1438 | slave_dev->type == ARPHRD_INFINIBAND) { | |
76444f50 | 1439 | netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n"); |
f5442441 VF |
1440 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP && |
1441 | bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
1442 | if (!bond_has_slaves(bond)) { | |
00503b6f | 1443 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
76444f50 | 1444 | netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n"); |
f5442441 | 1445 | } else { |
76444f50 | 1446 | 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 |
1447 | res = -EOPNOTSUPP; |
1448 | goto err_undo_flags; | |
00503b6f | 1449 | } |
2ab82852 | 1450 | } |
1da177e4 LT |
1451 | } |
1452 | ||
8d8fc29d AW |
1453 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1454 | ||
c20811a7 | 1455 | /* If this is the first slave, then we need to set the master's hardware |
547942ca NA |
1456 | * address to be the same as the slave's. |
1457 | */ | |
0965a1f3 | 1458 | if (!bond_has_slaves(bond) && |
dec1e90e | 1459 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1460 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1461 | |
3fdddd85 | 1462 | new_slave = bond_alloc_slave(bond); |
1da177e4 LT |
1463 | if (!new_slave) { |
1464 | res = -ENOMEM; | |
1465 | goto err_undo_flags; | |
1466 | } | |
3fdddd85 | 1467 | |
dc73c41f VF |
1468 | new_slave->bond = bond; |
1469 | new_slave->dev = slave_dev; | |
547942ca | 1470 | /* Set the new_slave's queue_id to be zero. Queue ID mapping |
bb1d9123 AG |
1471 | * is set via sysfs or module option if desired. |
1472 | */ | |
1473 | new_slave->queue_id = 0; | |
1474 | ||
b15ba0fb JP |
1475 | /* Save slave's original mtu and then set it to match the bond */ |
1476 | new_slave->original_mtu = slave_dev->mtu; | |
1477 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1478 | if (res) { | |
76444f50 | 1479 | netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res); |
b15ba0fb JP |
1480 | goto err_free; |
1481 | } | |
1482 | ||
547942ca | 1483 | /* Save slave's original ("permanent") mac address for modes |
217df670 JV |
1484 | * that need it, and for restoring it upon release, and then |
1485 | * set it to the master's address | |
1486 | */ | |
faeeb317 JW |
1487 | bond_hw_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr, |
1488 | slave_dev->addr_len); | |
1da177e4 | 1489 | |
00503b6f | 1490 | if (!bond->params.fail_over_mac || |
01844098 | 1491 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
547942ca | 1492 | /* Set slave to master's mac address. The application already |
2ab82852 MS |
1493 | * set the master's mac address to that of the first slave |
1494 | */ | |
faeeb317 JW |
1495 | memcpy(ss.__data, bond_dev->dev_addr, bond_dev->addr_len); |
1496 | ss.ss_family = slave_dev->type; | |
1497 | res = dev_set_mac_address(slave_dev, (struct sockaddr *)&ss); | |
2ab82852 | 1498 | if (res) { |
76444f50 | 1499 | netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res); |
b15ba0fb | 1500 | goto err_restore_mtu; |
2ab82852 | 1501 | } |
217df670 | 1502 | } |
1da177e4 | 1503 | |
03d84a5f KH |
1504 | /* set slave flag before open to prevent IPv6 addrconf */ |
1505 | slave_dev->flags |= IFF_SLAVE; | |
1506 | ||
217df670 JV |
1507 | /* open the slave since the application closed it */ |
1508 | res = dev_open(slave_dev); | |
1509 | if (res) { | |
76444f50 | 1510 | netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1511 | goto err_restore_mac; |
1da177e4 LT |
1512 | } |
1513 | ||
0b680e75 | 1514 | slave_dev->priv_flags |= IFF_BONDING; |
5f0c5f73 AG |
1515 | /* initialize slave stats */ |
1516 | dev_get_stats(new_slave->dev, &new_slave->slave_stats); | |
1da177e4 | 1517 | |
58402054 | 1518 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1519 | /* bond_alb_init_slave() must be called before all other stages since |
1520 | * it might fail and we do not want to have to undo everything | |
1521 | */ | |
1522 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1523 | if (res) |
569f0c4d | 1524 | goto err_close; |
1da177e4 LT |
1525 | } |
1526 | ||
ec0865a9 | 1527 | /* If the mode uses primary, then the following is handled by |
303d1cbf | 1528 | * bond_change_active_slave(). |
1da177e4 | 1529 | */ |
ec0865a9 | 1530 | if (!bond_uses_primary(bond)) { |
1da177e4 LT |
1531 | /* set promiscuity level to new slave */ |
1532 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1533 | res = dev_set_promiscuity(slave_dev, 1); |
1534 | if (res) | |
1535 | goto err_close; | |
1da177e4 LT |
1536 | } |
1537 | ||
1538 | /* set allmulti level to new slave */ | |
1539 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1540 | res = dev_set_allmulti(slave_dev, 1); |
1541 | if (res) | |
1542 | goto err_close; | |
1da177e4 LT |
1543 | } |
1544 | ||
b9e40857 | 1545 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1546 | |
1547 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1548 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1549 | ||
b9e40857 | 1550 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1551 | } |
1552 | ||
01844098 | 1553 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
1554 | /* add lacpdu mc addr to mc list */ |
1555 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1556 | ||
22bedad3 | 1557 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1558 | } |
1559 | ||
b8e2fde4 WY |
1560 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1561 | if (res) { | |
76444f50 VF |
1562 | netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n", |
1563 | slave_dev->name); | |
1ff412ad | 1564 | goto err_close; |
1565 | } | |
1da177e4 | 1566 | |
c8c23903 | 1567 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1568 | |
1569 | new_slave->delay = 0; | |
1570 | new_slave->link_failure_count = 0; | |
1571 | ||
3f3c278c MB |
1572 | if (bond_update_speed_duplex(new_slave)) |
1573 | new_slave->link = BOND_LINK_DOWN; | |
876254ae | 1574 | |
49f17de7 | 1575 | new_slave->last_rx = jiffies - |
f31c7937 | 1576 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1577 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1578 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1579 | |
1da177e4 LT |
1580 | if (bond->params.miimon && !bond->params.use_carrier) { |
1581 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1582 | ||
1583 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
547942ca | 1584 | /* miimon is set but a bonded network driver |
1da177e4 LT |
1585 | * does not support ETHTOOL/MII and |
1586 | * arp_interval is not set. Note: if | |
1587 | * use_carrier is enabled, we will never go | |
1588 | * here (because netif_carrier is always | |
1589 | * supported); thus, we don't need to change | |
1590 | * the messages for netif_carrier. | |
1591 | */ | |
76444f50 VF |
1592 | 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", |
1593 | slave_dev->name); | |
1da177e4 LT |
1594 | } else if (link_reporting == -1) { |
1595 | /* unable get link status using mii/ethtool */ | |
76444f50 VF |
1596 | 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", |
1597 | slave_dev->name); | |
1da177e4 LT |
1598 | } |
1599 | } | |
1600 | ||
1601 | /* check for initial state */ | |
a30b0168 | 1602 | new_slave->link = BOND_LINK_NOCHANGE; |
f31c7937 MK |
1603 | if (bond->params.miimon) { |
1604 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1605 | if (bond->params.updelay) { | |
69a2338e | 1606 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1607 | BOND_LINK_BACK, |
1608 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 MK |
1609 | new_slave->delay = bond->params.updelay; |
1610 | } else { | |
69a2338e | 1611 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1612 | BOND_LINK_UP, |
1613 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 | 1614 | } |
1da177e4 | 1615 | } else { |
5d397061 JP |
1616 | bond_set_slave_link_state(new_slave, BOND_LINK_DOWN, |
1617 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1618 | } |
f31c7937 | 1619 | } else if (bond->params.arp_interval) { |
69a2338e MS |
1620 | bond_set_slave_link_state(new_slave, |
1621 | (netif_carrier_ok(slave_dev) ? | |
5d397061 JP |
1622 | BOND_LINK_UP : BOND_LINK_DOWN), |
1623 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1624 | } else { |
5d397061 JP |
1625 | bond_set_slave_link_state(new_slave, BOND_LINK_UP, |
1626 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1627 | } |
1628 | ||
f31c7937 | 1629 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 1630 | new_slave->last_link_up = jiffies; |
76444f50 VF |
1631 | netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n", |
1632 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1633 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 1634 | |
ec0865a9 | 1635 | if (bond_uses_primary(bond) && bond->params.primary[0]) { |
1da177e4 | 1636 | /* if there is a primary slave, remember it */ |
a549952a | 1637 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
059b47e8 | 1638 | rcu_assign_pointer(bond->primary_slave, new_slave); |
a549952a JP |
1639 | bond->force_primary = true; |
1640 | } | |
1da177e4 LT |
1641 | } |
1642 | ||
01844098 | 1643 | switch (BOND_MODE(bond)) { |
1da177e4 | 1644 | case BOND_MODE_ACTIVEBACKUP: |
5e5b0665 | 1645 | bond_set_slave_inactive_flags(new_slave, |
1646 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1647 | break; |
1648 | case BOND_MODE_8023AD: | |
1649 | /* in 802.3ad mode, the internal mechanism | |
1650 | * will activate the slaves in the selected | |
1651 | * aggregator | |
1652 | */ | |
5e5b0665 | 1653 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 1654 | /* if this is the first slave */ |
23c147e0 | 1655 | if (!prev_slave) { |
3fdddd85 | 1656 | SLAVE_AD_INFO(new_slave)->id = 1; |
1da177e4 LT |
1657 | /* Initialize AD with the number of times that the AD timer is called in 1 second |
1658 | * can be called only after the mac address of the bond is set | |
1659 | */ | |
56d00c67 | 1660 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1661 | } else { |
3fdddd85 | 1662 | SLAVE_AD_INFO(new_slave)->id = |
1663 | SLAVE_AD_INFO(prev_slave)->id + 1; | |
1da177e4 LT |
1664 | } |
1665 | ||
1666 | bond_3ad_bind_slave(new_slave); | |
1667 | break; | |
1668 | case BOND_MODE_TLB: | |
1669 | case BOND_MODE_ALB: | |
e30bc066 | 1670 | bond_set_active_slave(new_slave); |
5e5b0665 | 1671 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
1672 | break; |
1673 | default: | |
76444f50 | 1674 | netdev_dbg(bond_dev, "This slave is always active in trunk mode\n"); |
1da177e4 LT |
1675 | |
1676 | /* always active in trunk mode */ | |
e30bc066 | 1677 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1678 | |
1679 | /* In trunking mode there is little meaning to curr_active_slave | |
1680 | * anyway (it holds no special properties of the bond device), | |
1681 | * so we can change it without calling change_active_interface() | |
1682 | */ | |
4740d638 ED |
1683 | if (!rcu_access_pointer(bond->curr_active_slave) && |
1684 | new_slave->link == BOND_LINK_UP) | |
278b2083 | 1685 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1686 | |
1da177e4 LT |
1687 | break; |
1688 | } /* switch(bond_mode) */ | |
1689 | ||
f6dc31a8 | 1690 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1691 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1692 | if (slave_dev->npinfo) { |
1693 | if (slave_enable_netpoll(new_slave)) { | |
76444f50 | 1694 | netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n"); |
8a8efa22 | 1695 | res = -EBUSY; |
f7d9821a | 1696 | goto err_detach; |
8a8efa22 | 1697 | } |
f6dc31a8 WC |
1698 | } |
1699 | #endif | |
8a8efa22 | 1700 | |
fbe168ba MK |
1701 | if (!(bond_dev->features & NETIF_F_LRO)) |
1702 | dev_disable_lro(slave_dev); | |
1703 | ||
35d48903 JP |
1704 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1705 | new_slave); | |
1706 | if (res) { | |
76444f50 | 1707 | netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res); |
5831d66e | 1708 | goto err_detach; |
35d48903 JP |
1709 | } |
1710 | ||
41f0b049 | 1711 | res = bond_master_upper_dev_link(bond, new_slave); |
1f718f0f | 1712 | if (res) { |
76444f50 | 1713 | netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res); |
1f718f0f VF |
1714 | goto err_unregister; |
1715 | } | |
1716 | ||
07699f9a | 1717 | res = bond_sysfs_slave_add(new_slave); |
1718 | if (res) { | |
76444f50 | 1719 | netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res); |
07699f9a | 1720 | goto err_upper_unlink; |
1721 | } | |
1722 | ||
5378c2e6 VF |
1723 | bond->slave_cnt++; |
1724 | bond_compute_features(bond); | |
1725 | bond_set_carrier(bond); | |
1726 | ||
ec0865a9 | 1727 | if (bond_uses_primary(bond)) { |
f80889a5 | 1728 | block_netpoll_tx(); |
5378c2e6 | 1729 | bond_select_active_slave(bond); |
f80889a5 | 1730 | unblock_netpoll_tx(); |
5378c2e6 | 1731 | } |
1f718f0f | 1732 | |
ee637714 MB |
1733 | if (bond_mode_uses_xmit_hash(bond)) |
1734 | bond_update_slave_arr(bond, NULL); | |
1735 | ||
76444f50 VF |
1736 | netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n", |
1737 | slave_dev->name, | |
1738 | bond_is_active_slave(new_slave) ? "an active" : "a backup", | |
1739 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
1740 | |
1741 | /* enslave is successful */ | |
69e61133 | 1742 | bond_queue_slave_event(new_slave); |
1da177e4 LT |
1743 | return 0; |
1744 | ||
1745 | /* Undo stages on error */ | |
07699f9a | 1746 | err_upper_unlink: |
41f0b049 | 1747 | bond_upper_dev_unlink(bond, new_slave); |
07699f9a | 1748 | |
1f718f0f VF |
1749 | err_unregister: |
1750 | netdev_rx_handler_unregister(slave_dev); | |
1751 | ||
f7d9821a | 1752 | err_detach: |
ec0865a9 | 1753 | if (!bond_uses_primary(bond)) |
303d1cbf JV |
1754 | bond_hw_addr_flush(bond_dev, slave_dev); |
1755 | ||
1ff412ad | 1756 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
059b47e8 NA |
1757 | if (rcu_access_pointer(bond->primary_slave) == new_slave) |
1758 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
4740d638 | 1759 | if (rcu_access_pointer(bond->curr_active_slave) == new_slave) { |
f80889a5 | 1760 | block_netpoll_tx(); |
c8517035 | 1761 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1762 | bond_select_active_slave(bond); |
f80889a5 | 1763 | unblock_netpoll_tx(); |
3c5913b5 | 1764 | } |
059b47e8 NA |
1765 | /* either primary_slave or curr_active_slave might've changed */ |
1766 | synchronize_rcu(); | |
fc7a72ac | 1767 | slave_disable_netpoll(new_slave); |
f7d9821a | 1768 | |
1da177e4 | 1769 | err_close: |
b6a5a7b9 | 1770 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1771 | dev_close(slave_dev); |
1772 | ||
1773 | err_restore_mac: | |
03d84a5f | 1774 | slave_dev->flags &= ~IFF_SLAVE; |
00503b6f | 1775 | if (!bond->params.fail_over_mac || |
01844098 | 1776 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
3915c1e8 JV |
1777 | /* XXX TODO - fom follow mode needs to change master's |
1778 | * MAC if this slave's MAC is in use by the bond, or at | |
1779 | * least print a warning. | |
1780 | */ | |
faeeb317 JW |
1781 | bond_hw_addr_copy(ss.__data, new_slave->perm_hwaddr, |
1782 | new_slave->dev->addr_len); | |
1783 | ss.ss_family = slave_dev->type; | |
1784 | dev_set_mac_address(slave_dev, (struct sockaddr *)&ss); | |
2ab82852 | 1785 | } |
1da177e4 | 1786 | |
b15ba0fb JP |
1787 | err_restore_mtu: |
1788 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1789 | ||
1da177e4 | 1790 | err_free: |
3fdddd85 | 1791 | bond_free_slave(new_slave); |
1da177e4 LT |
1792 | |
1793 | err_undo_flags: | |
b8fad459 | 1794 | /* Enslave of first slave has failed and we need to fix master's mac */ |
7d5cd2ce NA |
1795 | if (!bond_has_slaves(bond)) { |
1796 | if (ether_addr_equal_64bits(bond_dev->dev_addr, | |
1797 | slave_dev->dev_addr)) | |
1798 | eth_hw_addr_random(bond_dev); | |
1799 | if (bond_dev->type != ARPHRD_ETHER) { | |
40baec22 | 1800 | dev_close(bond_dev); |
7d5cd2ce NA |
1801 | ether_setup(bond_dev); |
1802 | bond_dev->flags |= IFF_MASTER; | |
1803 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; | |
1804 | } | |
1805 | } | |
3d632c3f | 1806 | |
1da177e4 LT |
1807 | return res; |
1808 | } | |
1809 | ||
547942ca | 1810 | /* Try to release the slave device <slave> from the bond device <master> |
1da177e4 | 1811 | * It is legal to access curr_active_slave without a lock because all the function |
8c0bc550 | 1812 | * is RTNL-locked. If "all" is true it means that the function is being called |
0896341a | 1813 | * while destroying a bond interface and all slaves are being released. |
1da177e4 LT |
1814 | * |
1815 | * The rules for slave state should be: | |
1816 | * for Active/Backup: | |
1817 | * Active stays on all backups go down | |
1818 | * for Bonded connections: | |
1819 | * The first up interface should be left on and all others downed. | |
1820 | */ | |
0896341a | 1821 | static int __bond_release_one(struct net_device *bond_dev, |
1822 | struct net_device *slave_dev, | |
1823 | bool all) | |
1da177e4 | 1824 | { |
454d7c9b | 1825 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 1826 | struct slave *slave, *oldcurrent; |
faeeb317 | 1827 | struct sockaddr_storage ss; |
5a0068de | 1828 | int old_flags = bond_dev->flags; |
c8f44aff | 1829 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1830 | |
1831 | /* slave is not a slave or master is not master of this slave */ | |
1832 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1833 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a22a9e41 | 1834 | netdev_dbg(bond_dev, "cannot release %s\n", |
76444f50 | 1835 | slave_dev->name); |
1da177e4 LT |
1836 | return -EINVAL; |
1837 | } | |
1838 | ||
e843fa50 | 1839 | block_netpoll_tx(); |
1da177e4 LT |
1840 | |
1841 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1842 | if (!slave) { | |
1843 | /* not a slave of this bond */ | |
76444f50 VF |
1844 | netdev_info(bond_dev, "%s not enslaved\n", |
1845 | slave_dev->name); | |
e843fa50 | 1846 | unblock_netpoll_tx(); |
1da177e4 LT |
1847 | return -EINVAL; |
1848 | } | |
1849 | ||
57beaca8 JP |
1850 | bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW); |
1851 | ||
07699f9a | 1852 | bond_sysfs_slave_del(slave); |
1853 | ||
5f0c5f73 AG |
1854 | /* recompute stats just before removing the slave */ |
1855 | bond_get_stats(bond->dev, &bond->bond_stats); | |
1856 | ||
41f0b049 | 1857 | bond_upper_dev_unlink(bond, slave); |
35d48903 JP |
1858 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1859 | * for this slave anymore. | |
1860 | */ | |
1861 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 | 1862 | |
e470259f | 1863 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 1864 | bond_3ad_unbind_slave(slave); |
1da177e4 | 1865 | |
ee637714 MB |
1866 | if (bond_mode_uses_xmit_hash(bond)) |
1867 | bond_update_slave_arr(bond, slave); | |
1868 | ||
76444f50 VF |
1869 | netdev_info(bond_dev, "Releasing %s interface %s\n", |
1870 | bond_is_active_slave(slave) ? "active" : "backup", | |
1871 | slave_dev->name); | |
1da177e4 | 1872 | |
4740d638 | 1873 | oldcurrent = rcu_access_pointer(bond->curr_active_slave); |
1da177e4 | 1874 | |
85741718 | 1875 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
1da177e4 | 1876 | |
00503b6f | 1877 | if (!all && (!bond->params.fail_over_mac || |
01844098 | 1878 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) { |
844223ab | 1879 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1880 | bond_has_slaves(bond)) |
76444f50 VF |
1881 | 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", |
1882 | slave_dev->name, slave->perm_hwaddr, | |
1883 | bond_dev->name, slave_dev->name); | |
dec1e90e | 1884 | } |
1885 | ||
059b47e8 NA |
1886 | if (rtnl_dereference(bond->primary_slave) == slave) |
1887 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
1da177e4 | 1888 | |
1c72cfdc | 1889 | if (oldcurrent == slave) |
1da177e4 | 1890 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1891 | |
58402054 | 1892 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1893 | /* Must be called only after the slave has been |
1894 | * detached from the list and the curr_active_slave | |
1895 | * has been cleared (if our_slave == old_current), | |
1896 | * but before a new active slave is selected. | |
1897 | */ | |
1898 | bond_alb_deinit_slave(bond, slave); | |
1899 | } | |
1900 | ||
0896341a | 1901 | if (all) { |
36708b89 | 1902 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 1903 | } else if (oldcurrent == slave) { |
547942ca | 1904 | /* Note that we hold RTNL over this sequence, so there |
059fe7a5 JV |
1905 | * is no concern that another slave add/remove event |
1906 | * will interfere. | |
1907 | */ | |
1da177e4 | 1908 | bond_select_active_slave(bond); |
059fe7a5 JV |
1909 | } |
1910 | ||
0965a1f3 | 1911 | if (!bond_has_slaves(bond)) { |
ff59c456 | 1912 | bond_set_carrier(bond); |
409cc1f8 | 1913 | eth_hw_addr_random(bond_dev); |
1da177e4 LT |
1914 | } |
1915 | ||
e843fa50 | 1916 | unblock_netpoll_tx(); |
278b2083 | 1917 | synchronize_rcu(); |
ee6154e1 | 1918 | bond->slave_cnt--; |
1da177e4 | 1919 | |
0965a1f3 | 1920 | if (!bond_has_slaves(bond)) { |
2af73d4b | 1921 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1922 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1923 | } | |
2af73d4b | 1924 | |
b2a103e6 MM |
1925 | bond_compute_features(bond); |
1926 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1927 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
76444f50 VF |
1928 | netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n", |
1929 | slave_dev->name, bond_dev->name); | |
b2a103e6 | 1930 | |
1ff412ad | 1931 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1932 | |
547942ca | 1933 | /* If the mode uses primary, then this case was handled above by |
303d1cbf | 1934 | * bond_change_active_slave(..., NULL) |
1da177e4 | 1935 | */ |
ec0865a9 | 1936 | if (!bond_uses_primary(bond)) { |
5a0068de NH |
1937 | /* unset promiscuity level from slave |
1938 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
1939 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
1940 | * value of that flag before that change, as that was the value | |
1941 | * when this slave was attached, so we cache at the start of the | |
1942 | * function and use it here. Same goes for ALLMULTI below | |
1943 | */ | |
1944 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 1945 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1946 | |
1947 | /* unset allmulti level from slave */ | |
5a0068de | 1948 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 1949 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1950 | |
303d1cbf | 1951 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1952 | } |
1953 | ||
8a8efa22 | 1954 | slave_disable_netpoll(slave); |
f6dc31a8 | 1955 | |
1da177e4 LT |
1956 | /* close slave before restoring its mac address */ |
1957 | dev_close(slave_dev); | |
1958 | ||
00503b6f | 1959 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
01844098 | 1960 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 | 1961 | /* restore original ("permanent") mac address */ |
faeeb317 JW |
1962 | bond_hw_addr_copy(ss.__data, slave->perm_hwaddr, |
1963 | slave->dev->addr_len); | |
1964 | ss.ss_family = slave_dev->type; | |
1965 | dev_set_mac_address(slave_dev, (struct sockaddr *)&ss); | |
2ab82852 | 1966 | } |
1da177e4 | 1967 | |
b15ba0fb JP |
1968 | dev_set_mtu(slave_dev, slave->original_mtu); |
1969 | ||
2d7011ca | 1970 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 | 1971 | |
3fdddd85 | 1972 | bond_free_slave(slave); |
1da177e4 | 1973 | |
547942ca | 1974 | return 0; |
1da177e4 LT |
1975 | } |
1976 | ||
0896341a | 1977 | /* A wrapper used because of ndo_del_link */ |
1978 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1979 | { | |
1980 | return __bond_release_one(bond_dev, slave_dev, false); | |
1981 | } | |
1982 | ||
547942ca NA |
1983 | /* First release a slave and then destroy the bond if no more slaves are left. |
1984 | * Must be under rtnl_lock when this function is called. | |
1985 | */ | |
26d8ee75 | 1986 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1987 | struct net_device *slave_dev) | |
d90a162a | 1988 | { |
454d7c9b | 1989 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1990 | int ret; |
1991 | ||
1992 | ret = bond_release(bond_dev, slave_dev); | |
0965a1f3 | 1993 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 1994 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
76444f50 VF |
1995 | netdev_info(bond_dev, "Destroying bond %s\n", |
1996 | bond_dev->name); | |
06f6d109 | 1997 | bond_remove_proc_entry(bond); |
9e71626c | 1998 | unregister_netdevice(bond_dev); |
d90a162a MS |
1999 | } |
2000 | return ret; | |
2001 | } | |
2002 | ||
3d67576d | 2003 | static void bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1da177e4 | 2004 | { |
454d7c9b | 2005 | struct bonding *bond = netdev_priv(bond_dev); |
69a2338e | 2006 | bond_fill_ifbond(bond, info); |
1da177e4 LT |
2007 | } |
2008 | ||
2009 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2010 | { | |
454d7c9b | 2011 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 2012 | struct list_head *iter; |
dec1e90e | 2013 | int i = 0, res = -ENODEV; |
1da177e4 | 2014 | struct slave *slave; |
1da177e4 | 2015 | |
9caff1e7 | 2016 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 2017 | if (i++ == (int)info->slave_id) { |
689c96cc | 2018 | res = 0; |
69a2338e | 2019 | bond_fill_ifslave(slave, info); |
1da177e4 LT |
2020 | break; |
2021 | } | |
2022 | } | |
1da177e4 | 2023 | |
689c96cc | 2024 | return res; |
1da177e4 LT |
2025 | } |
2026 | ||
2027 | /*-------------------------------- Monitoring -------------------------------*/ | |
2028 | ||
4740d638 | 2029 | /* called with rcu_read_lock() */ |
f0c76d61 JV |
2030 | static int bond_miimon_inspect(struct bonding *bond) |
2031 | { | |
dec1e90e | 2032 | int link_state, commit = 0; |
9caff1e7 | 2033 | struct list_head *iter; |
f0c76d61 | 2034 | struct slave *slave; |
41f89100 JP |
2035 | bool ignore_updelay; |
2036 | ||
4740d638 | 2037 | ignore_updelay = !rcu_dereference(bond->curr_active_slave); |
1da177e4 | 2038 | |
4cb4f97b | 2039 | bond_for_each_slave_rcu(bond, slave, iter) { |
f0c76d61 | 2040 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2041 | |
f0c76d61 | 2042 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2043 | |
2044 | switch (slave->link) { | |
f0c76d61 JV |
2045 | case BOND_LINK_UP: |
2046 | if (link_state) | |
2047 | continue; | |
1da177e4 | 2048 | |
de77ecd4 | 2049 | bond_propose_link_state(slave, BOND_LINK_FAIL); |
f0c76d61 JV |
2050 | slave->delay = bond->params.downdelay; |
2051 | if (slave->delay) { | |
76444f50 VF |
2052 | netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n", |
2053 | (BOND_MODE(bond) == | |
2054 | BOND_MODE_ACTIVEBACKUP) ? | |
2055 | (bond_is_active_slave(slave) ? | |
2056 | "active " : "backup ") : "", | |
2057 | slave->dev->name, | |
2058 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2059 | } |
f0c76d61 JV |
2060 | /*FALLTHRU*/ |
2061 | case BOND_LINK_FAIL: | |
2062 | if (link_state) { | |
547942ca | 2063 | /* recovered before downdelay expired */ |
de77ecd4 | 2064 | bond_propose_link_state(slave, BOND_LINK_UP); |
8e603460 | 2065 | slave->last_link_up = jiffies; |
76444f50 VF |
2066 | netdev_info(bond->dev, "link status up again after %d ms for interface %s\n", |
2067 | (bond->params.downdelay - slave->delay) * | |
2068 | bond->params.miimon, | |
2069 | slave->dev->name); | |
fb9eb899 | 2070 | commit++; |
f0c76d61 | 2071 | continue; |
1da177e4 | 2072 | } |
f0c76d61 JV |
2073 | |
2074 | if (slave->delay <= 0) { | |
2075 | slave->new_link = BOND_LINK_DOWN; | |
2076 | commit++; | |
2077 | continue; | |
1da177e4 | 2078 | } |
1da177e4 | 2079 | |
f0c76d61 JV |
2080 | slave->delay--; |
2081 | break; | |
2082 | ||
2083 | case BOND_LINK_DOWN: | |
2084 | if (!link_state) | |
2085 | continue; | |
2086 | ||
de77ecd4 | 2087 | bond_propose_link_state(slave, BOND_LINK_BACK); |
f0c76d61 JV |
2088 | slave->delay = bond->params.updelay; |
2089 | ||
2090 | if (slave->delay) { | |
76444f50 VF |
2091 | netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n", |
2092 | slave->dev->name, | |
2093 | ignore_updelay ? 0 : | |
2094 | bond->params.updelay * | |
2095 | bond->params.miimon); | |
f0c76d61 JV |
2096 | } |
2097 | /*FALLTHRU*/ | |
2098 | case BOND_LINK_BACK: | |
2099 | if (!link_state) { | |
de77ecd4 | 2100 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
76444f50 VF |
2101 | netdev_info(bond->dev, "link status down again after %d ms for interface %s\n", |
2102 | (bond->params.updelay - slave->delay) * | |
2103 | bond->params.miimon, | |
2104 | slave->dev->name); | |
fb9eb899 | 2105 | commit++; |
f0c76d61 JV |
2106 | continue; |
2107 | } | |
2108 | ||
41f89100 JP |
2109 | if (ignore_updelay) |
2110 | slave->delay = 0; | |
2111 | ||
f0c76d61 JV |
2112 | if (slave->delay <= 0) { |
2113 | slave->new_link = BOND_LINK_UP; | |
2114 | commit++; | |
41f89100 | 2115 | ignore_updelay = false; |
f0c76d61 | 2116 | continue; |
1da177e4 | 2117 | } |
f0c76d61 JV |
2118 | |
2119 | slave->delay--; | |
1da177e4 | 2120 | break; |
f0c76d61 JV |
2121 | } |
2122 | } | |
1da177e4 | 2123 | |
f0c76d61 JV |
2124 | return commit; |
2125 | } | |
1da177e4 | 2126 | |
f0c76d61 JV |
2127 | static void bond_miimon_commit(struct bonding *bond) |
2128 | { | |
9caff1e7 | 2129 | struct list_head *iter; |
059b47e8 | 2130 | struct slave *slave, *primary; |
f0c76d61 | 2131 | |
9caff1e7 | 2132 | bond_for_each_slave(bond, slave, iter) { |
f0c76d61 JV |
2133 | switch (slave->new_link) { |
2134 | case BOND_LINK_NOCHANGE: | |
2135 | continue; | |
1da177e4 | 2136 | |
f0c76d61 | 2137 | case BOND_LINK_UP: |
b5bf0f5b | 2138 | if (bond_update_speed_duplex(slave)) { |
3f3c278c | 2139 | slave->link = BOND_LINK_DOWN; |
b5bf0f5b MB |
2140 | netdev_warn(bond->dev, |
2141 | "failed to get link speed/duplex for %s\n", | |
2142 | slave->dev->name); | |
2143 | continue; | |
2144 | } | |
5d397061 JP |
2145 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2146 | BOND_SLAVE_NOTIFY_NOW); | |
8e603460 | 2147 | slave->last_link_up = jiffies; |
f0c76d61 | 2148 | |
059b47e8 | 2149 | primary = rtnl_dereference(bond->primary_slave); |
01844098 | 2150 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
f0c76d61 | 2151 | /* prevent it from being the active one */ |
e30bc066 | 2152 | bond_set_backup_slave(slave); |
01844098 | 2153 | } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
f0c76d61 | 2154 | /* make it immediately active */ |
e30bc066 | 2155 | bond_set_active_slave(slave); |
059b47e8 | 2156 | } else if (slave != primary) { |
f0c76d61 | 2157 | /* prevent it from being the active one */ |
e30bc066 | 2158 | bond_set_backup_slave(slave); |
1da177e4 | 2159 | } |
1da177e4 | 2160 | |
76444f50 VF |
2161 | netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n", |
2162 | slave->dev->name, | |
2163 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, | |
2164 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2165 | |
f0c76d61 | 2166 | /* notify ad that the link status has changed */ |
01844098 | 2167 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 | 2168 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); |
059fe7a5 | 2169 | |
58402054 | 2170 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2171 | bond_alb_handle_link_change(bond, slave, |
2172 | BOND_LINK_UP); | |
1da177e4 | 2173 | |
ee637714 MB |
2174 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2175 | bond_update_slave_arr(bond, NULL); | |
2176 | ||
059b47e8 | 2177 | if (!bond->curr_active_slave || slave == primary) |
f0c76d61 | 2178 | goto do_failover; |
1da177e4 | 2179 | |
f0c76d61 | 2180 | continue; |
059fe7a5 | 2181 | |
f0c76d61 | 2182 | case BOND_LINK_DOWN: |
fba4acda JV |
2183 | if (slave->link_failure_count < UINT_MAX) |
2184 | slave->link_failure_count++; | |
2185 | ||
5d397061 JP |
2186 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2187 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 2188 | |
01844098 VF |
2189 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || |
2190 | BOND_MODE(bond) == BOND_MODE_8023AD) | |
5e5b0665 | 2191 | bond_set_slave_inactive_flags(slave, |
2192 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2193 | |
76444f50 VF |
2194 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2195 | slave->dev->name); | |
f0c76d61 | 2196 | |
01844098 | 2197 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 JV |
2198 | bond_3ad_handle_link_change(slave, |
2199 | BOND_LINK_DOWN); | |
2200 | ||
ae63e808 | 2201 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2202 | bond_alb_handle_link_change(bond, slave, |
2203 | BOND_LINK_DOWN); | |
2204 | ||
ee637714 MB |
2205 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2206 | bond_update_slave_arr(bond, NULL); | |
2207 | ||
4740d638 | 2208 | if (slave == rcu_access_pointer(bond->curr_active_slave)) |
f0c76d61 JV |
2209 | goto do_failover; |
2210 | ||
2211 | continue; | |
2212 | ||
2213 | default: | |
76444f50 VF |
2214 | netdev_err(bond->dev, "invalid new link %d on slave %s\n", |
2215 | slave->new_link, slave->dev->name); | |
f0c76d61 JV |
2216 | slave->new_link = BOND_LINK_NOCHANGE; |
2217 | ||
2218 | continue; | |
2219 | } | |
2220 | ||
2221 | do_failover: | |
e843fa50 | 2222 | block_netpoll_tx(); |
f0c76d61 | 2223 | bond_select_active_slave(bond); |
e843fa50 | 2224 | unblock_netpoll_tx(); |
f0c76d61 JV |
2225 | } |
2226 | ||
2227 | bond_set_carrier(bond); | |
1da177e4 LT |
2228 | } |
2229 | ||
547942ca | 2230 | /* bond_mii_monitor |
0b0eef66 JV |
2231 | * |
2232 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2233 | * inspection, then (if inspection indicates something needs to be done) |
2234 | * an acquisition of appropriate locks followed by a commit phase to | |
2235 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2236 | */ |
6da67d26 | 2237 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2238 | { |
2239 | struct bonding *bond = container_of(work, struct bonding, | |
2240 | mii_work.work); | |
ad246c99 | 2241 | bool should_notify_peers = false; |
1f2cd845 | 2242 | unsigned long delay; |
de77ecd4 MB |
2243 | struct slave *slave; |
2244 | struct list_head *iter; | |
0b0eef66 | 2245 | |
1f2cd845 DM |
2246 | delay = msecs_to_jiffies(bond->params.miimon); |
2247 | ||
2248 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2249 | goto re_arm; |
b59f9f74 | 2250 | |
4cb4f97b | 2251 | rcu_read_lock(); |
2252 | ||
ad246c99 BH |
2253 | should_notify_peers = bond_should_notify_peers(bond); |
2254 | ||
1f2cd845 | 2255 | if (bond_miimon_inspect(bond)) { |
4cb4f97b | 2256 | rcu_read_unlock(); |
f0c76d61 | 2257 | |
1f2cd845 DM |
2258 | /* Race avoidance with bond_close cancel of workqueue */ |
2259 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2260 | delay = 1; |
2261 | should_notify_peers = false; | |
2262 | goto re_arm; | |
2263 | } | |
6b6c5261 | 2264 | |
de77ecd4 MB |
2265 | bond_for_each_slave(bond, slave, iter) { |
2266 | bond_commit_link_state(slave, BOND_SLAVE_NOTIFY_LATER); | |
2267 | } | |
1f2cd845 DM |
2268 | bond_miimon_commit(bond); |
2269 | ||
1f2cd845 | 2270 | rtnl_unlock(); /* might sleep, hold no other locks */ |
4cb4f97b | 2271 | } else |
2272 | rcu_read_unlock(); | |
0b0eef66 | 2273 | |
f0c76d61 | 2274 | re_arm: |
e6d265e8 | 2275 | if (bond->params.miimon) |
1f2cd845 DM |
2276 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2277 | ||
1f2cd845 DM |
2278 | if (should_notify_peers) { |
2279 | if (!rtnl_trylock()) | |
2280 | return; | |
2281 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2282 | rtnl_unlock(); | |
2283 | } | |
0b0eef66 | 2284 | } |
c3ade5ca | 2285 | |
b3208b20 DA |
2286 | static int bond_upper_dev_walk(struct net_device *upper, void *data) |
2287 | { | |
2288 | __be32 ip = *((__be32 *)data); | |
2289 | ||
2290 | return ip == bond_confirm_addr(upper, 0, ip); | |
2291 | } | |
2292 | ||
50223ce4 | 2293 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2294 | { |
50223ce4 | 2295 | bool ret = false; |
f5b2b966 | 2296 | |
eaddcd76 | 2297 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2298 | return true; |
f5b2b966 | 2299 | |
50223ce4 | 2300 | rcu_read_lock(); |
b3208b20 DA |
2301 | if (netdev_walk_all_upper_dev_rcu(bond->dev, bond_upper_dev_walk, &ip)) |
2302 | ret = true; | |
50223ce4 | 2303 | rcu_read_unlock(); |
f5b2b966 | 2304 | |
50223ce4 | 2305 | return ret; |
f5b2b966 JV |
2306 | } |
2307 | ||
547942ca | 2308 | /* We go to the (large) trouble of VLAN tagging ARP frames because |
c3ade5ca JV |
2309 | * switches in VLAN mode (especially if ports are configured as |
2310 | * "native" to a VLAN) might not pass non-tagged frames. | |
2311 | */ | |
fbd929f2 | 2312 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, |
2313 | __be32 dest_ip, __be32 src_ip, | |
44a40855 | 2314 | struct bond_vlan_tag *tags) |
c3ade5ca JV |
2315 | { |
2316 | struct sk_buff *skb; | |
3e403a77 | 2317 | struct bond_vlan_tag *outer_tag = tags; |
c3ade5ca | 2318 | |
76444f50 VF |
2319 | netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n", |
2320 | arp_op, slave_dev->name, &dest_ip, &src_ip); | |
3d632c3f | 2321 | |
c3ade5ca JV |
2322 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2323 | NULL, slave_dev->dev_addr, NULL); | |
2324 | ||
2325 | if (!skb) { | |
4873ac3c | 2326 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2327 | return; |
2328 | } | |
fbd929f2 | 2329 | |
3e403a77 VF |
2330 | if (!tags || tags->vlan_proto == VLAN_N_VID) |
2331 | goto xmit; | |
2332 | ||
2333 | tags++; | |
2334 | ||
44a40855 | 2335 | /* Go through all the tags backwards and add them to the packet */ |
3e403a77 VF |
2336 | while (tags->vlan_proto != VLAN_N_VID) { |
2337 | if (!tags->vlan_id) { | |
2338 | tags++; | |
44a40855 | 2339 | continue; |
3e403a77 | 2340 | } |
44a40855 | 2341 | |
76444f50 | 2342 | netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n", |
3e403a77 | 2343 | ntohs(outer_tag->vlan_proto), tags->vlan_id); |
62749e2c JP |
2344 | skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto, |
2345 | tags->vlan_id); | |
44a40855 VY |
2346 | if (!skb) { |
2347 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2348 | return; | |
2349 | } | |
3e403a77 VF |
2350 | |
2351 | tags++; | |
44a40855 VY |
2352 | } |
2353 | /* Set the outer tag */ | |
3e403a77 | 2354 | if (outer_tag->vlan_id) { |
76444f50 | 2355 | netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n", |
3e403a77 | 2356 | ntohs(outer_tag->vlan_proto), outer_tag->vlan_id); |
b4bef1b5 JP |
2357 | __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto, |
2358 | outer_tag->vlan_id); | |
c3ade5ca | 2359 | } |
3e403a77 VF |
2360 | |
2361 | xmit: | |
c3ade5ca JV |
2362 | arp_xmit(skb); |
2363 | } | |
2364 | ||
44a40855 VY |
2365 | /* Validate the device path between the @start_dev and the @end_dev. |
2366 | * The path is valid if the @end_dev is reachable through device | |
2367 | * stacking. | |
2368 | * When the path is validated, collect any vlan information in the | |
2369 | * path. | |
2370 | */ | |
3e403a77 VF |
2371 | struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, |
2372 | struct net_device *end_dev, | |
2373 | int level) | |
44a40855 | 2374 | { |
3e403a77 | 2375 | struct bond_vlan_tag *tags; |
44a40855 VY |
2376 | struct net_device *upper; |
2377 | struct list_head *iter; | |
44a40855 | 2378 | |
3e403a77 VF |
2379 | if (start_dev == end_dev) { |
2380 | tags = kzalloc(sizeof(*tags) * (level + 1), GFP_ATOMIC); | |
2381 | if (!tags) | |
2382 | return ERR_PTR(-ENOMEM); | |
2383 | tags[level].vlan_proto = VLAN_N_VID; | |
2384 | return tags; | |
2385 | } | |
44a40855 VY |
2386 | |
2387 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
3e403a77 VF |
2388 | tags = bond_verify_device_path(upper, end_dev, level + 1); |
2389 | if (IS_ERR_OR_NULL(tags)) { | |
2390 | if (IS_ERR(tags)) | |
2391 | return tags; | |
2392 | continue; | |
44a40855 | 2393 | } |
3e403a77 VF |
2394 | if (is_vlan_dev(upper)) { |
2395 | tags[level].vlan_proto = vlan_dev_vlan_proto(upper); | |
2396 | tags[level].vlan_id = vlan_dev_vlan_id(upper); | |
2397 | } | |
2398 | ||
2399 | return tags; | |
44a40855 VY |
2400 | } |
2401 | ||
3e403a77 | 2402 | return NULL; |
44a40855 | 2403 | } |
c3ade5ca | 2404 | |
1da177e4 LT |
2405 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2406 | { | |
c3ade5ca | 2407 | struct rtable *rt; |
3e403a77 | 2408 | struct bond_vlan_tag *tags; |
27bc11e6 | 2409 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2410 | int i; |
1da177e4 | 2411 | |
27bc11e6 | 2412 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
76444f50 | 2413 | netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]); |
3e403a77 | 2414 | tags = NULL; |
c3ade5ca | 2415 | |
27bc11e6 | 2416 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2417 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2418 | RTO_ONLINK, 0); | |
b23dd4fe | 2419 | if (IS_ERR(rt)) { |
28572760 VF |
2420 | /* there's no route to target - try to send arp |
2421 | * probe to generate any traffic (arp_validate=0) | |
2422 | */ | |
4873ac3c | 2423 | if (bond->params.arp_validate) |
2424 | net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", | |
2425 | bond->dev->name, | |
2426 | &targets[i]); | |
44a40855 VY |
2427 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2428 | 0, tags); | |
c3ade5ca JV |
2429 | continue; |
2430 | } | |
2431 | ||
27bc11e6 VF |
2432 | /* bond device itself */ |
2433 | if (rt->dst.dev == bond->dev) | |
2434 | goto found; | |
2435 | ||
2436 | rcu_read_lock(); | |
3e403a77 | 2437 | tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0); |
27bc11e6 | 2438 | rcu_read_unlock(); |
c3ade5ca | 2439 | |
3e403a77 | 2440 | if (!IS_ERR_OR_NULL(tags)) |
44a40855 VY |
2441 | goto found; |
2442 | ||
27bc11e6 | 2443 | /* Not our device - skip */ |
76444f50 VF |
2444 | netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n", |
2445 | &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
3e32582f | 2446 | |
ed4b9f80 | 2447 | ip_rt_put(rt); |
27bc11e6 VF |
2448 | continue; |
2449 | ||
2450 | found: | |
2451 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2452 | ip_rt_put(rt); | |
2453 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
44a40855 | 2454 | addr, tags); |
a67eed57 | 2455 | kfree(tags); |
c3ade5ca JV |
2456 | } |
2457 | } | |
2458 | ||
d3bb52b0 | 2459 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2460 | { |
8599b52e VF |
2461 | int i; |
2462 | ||
87a7b84b | 2463 | if (!sip || !bond_has_this_ip(bond, tip)) { |
76444f50 VF |
2464 | netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n", |
2465 | &sip, &tip); | |
87a7b84b VF |
2466 | return; |
2467 | } | |
f5b2b966 | 2468 | |
8599b52e VF |
2469 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2470 | if (i == -1) { | |
76444f50 VF |
2471 | netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n", |
2472 | &sip); | |
87a7b84b | 2473 | return; |
f5b2b966 | 2474 | } |
49f17de7 | 2475 | slave->last_rx = jiffies; |
8599b52e | 2476 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2477 | } |
2478 | ||
5bb9e0b5 | 2479 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2480 | struct slave *slave) | |
f5b2b966 | 2481 | { |
de063b70 | 2482 | struct arphdr *arp = (struct arphdr *)skb->data; |
21a75f09 | 2483 | struct slave *curr_active_slave, *curr_arp_slave; |
f5b2b966 | 2484 | unsigned char *arp_ptr; |
d3bb52b0 | 2485 | __be32 sip, tip; |
f2cb691a | 2486 | int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
f5b2b966 | 2487 | |
f2cb691a | 2488 | if (!slave_do_arp_validate(bond, slave)) { |
bedabf90 | 2489 | if ((slave_do_arp_validate_only(bond) && is_arp) || |
2490 | !slave_do_arp_validate_only(bond)) | |
49f17de7 | 2491 | slave->last_rx = jiffies; |
13a8e0c8 | 2492 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2493 | } else if (!is_arp) { |
2494 | return RX_HANDLER_ANOTHER; | |
2495 | } | |
2c146102 | 2496 | |
de063b70 | 2497 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2498 | |
76444f50 VF |
2499 | netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n", |
2500 | skb->dev->name); | |
f5b2b966 | 2501 | |
de063b70 ED |
2502 | if (alen > skb_headlen(skb)) { |
2503 | arp = kmalloc(alen, GFP_ATOMIC); | |
2504 | if (!arp) | |
2505 | goto out_unlock; | |
2506 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2507 | goto out_unlock; | |
2508 | } | |
f5b2b966 | 2509 | |
3aba891d | 2510 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2511 | skb->pkt_type == PACKET_OTHERHOST || |
2512 | skb->pkt_type == PACKET_LOOPBACK || | |
2513 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2514 | arp->ar_pro != htons(ETH_P_IP) || | |
2515 | arp->ar_pln != 4) | |
2516 | goto out_unlock; | |
2517 | ||
2518 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2519 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2520 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2521 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2522 | memcpy(&tip, arp_ptr, 4); |
2523 | ||
76444f50 VF |
2524 | netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
2525 | slave->dev->name, bond_slave_state(slave), | |
2526 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2527 | &sip, &tip); | |
f5b2b966 | 2528 | |
010d3c39 | 2529 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
21a75f09 | 2530 | curr_arp_slave = rcu_dereference(bond->current_arp_slave); |
010d3c39 | 2531 | |
21a75f09 | 2532 | /* We 'trust' the received ARP enough to validate it if: |
aeea64ac | 2533 | * |
21a75f09 JV |
2534 | * (a) the slave receiving the ARP is active (which includes the |
2535 | * current ARP slave, if any), or | |
2536 | * | |
2537 | * (b) the receiving slave isn't active, but there is a currently | |
2538 | * active slave and it received valid arp reply(s) after it became | |
2539 | * the currently active slave, or | |
2540 | * | |
2541 | * (c) there is an ARP slave that sent an ARP during the prior ARP | |
2542 | * interval, and we receive an ARP reply on any slave. We accept | |
2543 | * these because switch FDB update delays may deliver the ARP | |
2544 | * reply to a slave other than the sender of the ARP request. | |
2545 | * | |
2546 | * Note: for (b), backup slaves are receiving the broadcast ARP | |
2547 | * request, not a reply. This request passes from the sending | |
2548 | * slave through the L2 switch(es) to the receiving slave. Since | |
2549 | * this is checking the request, sip/tip are swapped for | |
2550 | * validation. | |
2551 | * | |
2552 | * This is done to avoid endless looping when we can't reach the | |
aeea64ac | 2553 | * arp_ip_target and fool ourselves with our own arp requests. |
f5b2b966 | 2554 | */ |
e30bc066 | 2555 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2556 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
2557 | else if (curr_active_slave && |
2558 | time_after(slave_last_rx(bond, curr_active_slave), | |
2559 | curr_active_slave->last_link_up)) | |
f5b2b966 | 2560 | bond_validate_arp(bond, slave, tip, sip); |
21a75f09 JV |
2561 | else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) && |
2562 | bond_time_in_interval(bond, | |
2563 | dev_trans_start(curr_arp_slave->dev), 1)) | |
2564 | bond_validate_arp(bond, slave, sip, tip); | |
f5b2b966 JV |
2565 | |
2566 | out_unlock: | |
de063b70 ED |
2567 | if (arp != (struct arphdr *)skb->data) |
2568 | kfree(arp); | |
13a8e0c8 | 2569 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2570 | } |
2571 | ||
e7f63f1d VF |
2572 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2573 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2574 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2575 | */ | |
2576 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2577 | int mod) | |
2578 | { | |
2579 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2580 | ||
2581 | return time_in_range(jiffies, | |
2582 | last_act - delta_in_ticks, | |
2583 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2584 | } | |
2585 | ||
547942ca | 2586 | /* This function is called regularly to monitor each slave's link |
1da177e4 LT |
2587 | * ensuring that traffic is being sent and received when arp monitoring |
2588 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2589 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2590 | * arp monitoring in active backup mode. | |
2591 | */ | |
d5e73f7b | 2592 | static void bond_loadbalance_arp_mon(struct bonding *bond) |
1da177e4 | 2593 | { |
1da177e4 | 2594 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2595 | struct list_head *iter; |
6fde8f03 | 2596 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 2597 | |
1f2cd845 | 2598 | if (!bond_has_slaves(bond)) |
1da177e4 | 2599 | goto re_arm; |
1da177e4 | 2600 | |
2e52f4fe | 2601 | rcu_read_lock(); |
2602 | ||
4740d638 | 2603 | oldcurrent = rcu_dereference(bond->curr_active_slave); |
1da177e4 LT |
2604 | /* see if any of the previous devices are up now (i.e. they have |
2605 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
2606 | * the picture unless it is null. also, slave->last_link_up is not |
2607 | * needed here because we send an arp on each slave and give a slave | |
2608 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
2609 | * TODO: what about up/down delay in arp mode? it wasn't here before |
2610 | * so it can wait | |
2611 | */ | |
2e52f4fe | 2612 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2613 | unsigned long trans_start = dev_trans_start(slave->dev); |
2614 | ||
797a9364 NS |
2615 | slave->new_link = BOND_LINK_NOCHANGE; |
2616 | ||
1da177e4 | 2617 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2618 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 2619 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 | 2620 | |
797a9364 | 2621 | slave->new_link = BOND_LINK_UP; |
6fde8f03 | 2622 | slave_state_changed = 1; |
1da177e4 LT |
2623 | |
2624 | /* primary_slave has no meaning in round-robin | |
2625 | * mode. the window of a slave being up and | |
2626 | * curr_active_slave being null after enslaving | |
2627 | * is closed. | |
2628 | */ | |
2629 | if (!oldcurrent) { | |
76444f50 VF |
2630 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2631 | slave->dev->name); | |
1da177e4 LT |
2632 | do_failover = 1; |
2633 | } else { | |
76444f50 VF |
2634 | netdev_info(bond->dev, "interface %s is now up\n", |
2635 | slave->dev->name); | |
1da177e4 LT |
2636 | } |
2637 | } | |
2638 | } else { | |
2639 | /* slave->link == BOND_LINK_UP */ | |
2640 | ||
2641 | /* not all switches will respond to an arp request | |
2642 | * when the source ip is 0, so don't take the link down | |
2643 | * if we don't know our ip yet | |
2644 | */ | |
e7f63f1d | 2645 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 2646 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 | 2647 | |
797a9364 | 2648 | slave->new_link = BOND_LINK_DOWN; |
6fde8f03 | 2649 | slave_state_changed = 1; |
1da177e4 | 2650 | |
3d632c3f | 2651 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2652 | slave->link_failure_count++; |
1da177e4 | 2653 | |
76444f50 VF |
2654 | netdev_info(bond->dev, "interface %s is now down\n", |
2655 | slave->dev->name); | |
1da177e4 | 2656 | |
3d632c3f | 2657 | if (slave == oldcurrent) |
1da177e4 | 2658 | do_failover = 1; |
1da177e4 LT |
2659 | } |
2660 | } | |
2661 | ||
2662 | /* note: if switch is in round-robin mode, all links | |
2663 | * must tx arp to ensure all links rx an arp - otherwise | |
2664 | * links may oscillate or not come up at all; if switch is | |
2665 | * in something like xor mode, there is nothing we can | |
2666 | * do - all replies will be rx'ed on same link causing slaves | |
2667 | * to be unstable during low/no traffic periods | |
2668 | */ | |
b6adc610 | 2669 | if (bond_slave_is_up(slave)) |
1da177e4 | 2670 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2671 | } |
2672 | ||
2e52f4fe | 2673 | rcu_read_unlock(); |
2674 | ||
6fde8f03 | 2675 | if (do_failover || slave_state_changed) { |
2e52f4fe | 2676 | if (!rtnl_trylock()) |
2677 | goto re_arm; | |
1da177e4 | 2678 | |
797a9364 NS |
2679 | bond_for_each_slave(bond, slave, iter) { |
2680 | if (slave->new_link != BOND_LINK_NOCHANGE) | |
2681 | slave->link = slave->new_link; | |
2682 | } | |
2683 | ||
6fde8f03 DT |
2684 | if (slave_state_changed) { |
2685 | bond_slave_state_change(bond); | |
ee637714 MB |
2686 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2687 | bond_update_slave_arr(bond, NULL); | |
b8e4500f NA |
2688 | } |
2689 | if (do_failover) { | |
6fde8f03 | 2690 | block_netpoll_tx(); |
6fde8f03 | 2691 | bond_select_active_slave(bond); |
6fde8f03 DT |
2692 | unblock_netpoll_tx(); |
2693 | } | |
2e52f4fe | 2694 | rtnl_unlock(); |
1da177e4 LT |
2695 | } |
2696 | ||
2697 | re_arm: | |
e6d265e8 | 2698 | if (bond->params.arp_interval) |
e7f63f1d VF |
2699 | queue_delayed_work(bond->wq, &bond->arp_work, |
2700 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2701 | } |
2702 | ||
547942ca | 2703 | /* Called to inspect slaves for active-backup mode ARP monitor link state |
b2220cad JV |
2704 | * changes. Sets new_link in slaves to specify what action should take |
2705 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2706 | * to link states must be committed. | |
2707 | * | |
8c0bc550 | 2708 | * Called with rcu_read_lock held. |
1da177e4 | 2709 | */ |
e7f63f1d | 2710 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2711 | { |
def4460c | 2712 | unsigned long trans_start, last_rx; |
9caff1e7 | 2713 | struct list_head *iter; |
dec1e90e | 2714 | struct slave *slave; |
dec1e90e | 2715 | int commit = 0; |
da210f55 | 2716 | |
eb9fa4b0 | 2717 | bond_for_each_slave_rcu(bond, slave, iter) { |
b2220cad | 2718 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2719 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2720 | |
b2220cad | 2721 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2722 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2723 | slave->new_link = BOND_LINK_UP; |
2724 | commit++; | |
2725 | } | |
b2220cad JV |
2726 | continue; |
2727 | } | |
1da177e4 | 2728 | |
547942ca | 2729 | /* Give slaves 2*delta after being enslaved or made |
b2220cad JV |
2730 | * active. This avoids bouncing, as the last receive |
2731 | * times need a full ARP monitor cycle to be updated. | |
2732 | */ | |
8e603460 | 2733 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
2734 | continue; |
2735 | ||
547942ca | 2736 | /* Backup slave is down if: |
b2220cad JV |
2737 | * - No current_arp_slave AND |
2738 | * - more than 3*delta since last receive AND | |
2739 | * - the bond has an IP address | |
2740 | * | |
2741 | * Note: a non-null current_arp_slave indicates | |
2742 | * the curr_active_slave went down and we are | |
2743 | * searching for a new one; under this condition | |
2744 | * we only take the curr_active_slave down - this | |
2745 | * gives each slave a chance to tx/rx traffic | |
2746 | * before being taken out | |
2747 | */ | |
e30bc066 | 2748 | if (!bond_is_active_slave(slave) && |
85741718 | 2749 | !rcu_access_pointer(bond->current_arp_slave) && |
e7f63f1d | 2750 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2751 | slave->new_link = BOND_LINK_DOWN; |
2752 | commit++; | |
2753 | } | |
2754 | ||
547942ca | 2755 | /* Active slave is down if: |
b2220cad JV |
2756 | * - more than 2*delta since transmitting OR |
2757 | * - (more than 2*delta since receive AND | |
2758 | * the bond has an IP address) | |
2759 | */ | |
cb32f2a0 | 2760 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2761 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2762 | (!bond_time_in_interval(bond, trans_start, 2) || |
2763 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2764 | slave->new_link = BOND_LINK_DOWN; |
2765 | commit++; | |
2766 | } | |
1da177e4 LT |
2767 | } |
2768 | ||
b2220cad JV |
2769 | return commit; |
2770 | } | |
1da177e4 | 2771 | |
547942ca | 2772 | /* Called to commit link state changes noted by inspection step of |
b2220cad JV |
2773 | * active-backup mode ARP monitor. |
2774 | * | |
eb9fa4b0 | 2775 | * Called with RTNL hold. |
b2220cad | 2776 | */ |
e7f63f1d | 2777 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2778 | { |
cb32f2a0 | 2779 | unsigned long trans_start; |
9caff1e7 | 2780 | struct list_head *iter; |
dec1e90e | 2781 | struct slave *slave; |
1da177e4 | 2782 | |
9caff1e7 | 2783 | bond_for_each_slave(bond, slave, iter) { |
b2220cad JV |
2784 | switch (slave->new_link) { |
2785 | case BOND_LINK_NOCHANGE: | |
2786 | continue; | |
ff59c456 | 2787 | |
b2220cad | 2788 | case BOND_LINK_UP: |
cb32f2a0 | 2789 | trans_start = dev_trans_start(slave->dev); |
4740d638 ED |
2790 | if (rtnl_dereference(bond->curr_active_slave) != slave || |
2791 | (!rtnl_dereference(bond->curr_active_slave) && | |
e7f63f1d | 2792 | bond_time_in_interval(bond, trans_start, 1))) { |
85741718 ED |
2793 | struct slave *current_arp_slave; |
2794 | ||
2795 | current_arp_slave = rtnl_dereference(bond->current_arp_slave); | |
5d397061 JP |
2796 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2797 | BOND_SLAVE_NOTIFY_NOW); | |
85741718 | 2798 | if (current_arp_slave) { |
5a430974 | 2799 | bond_set_slave_inactive_flags( |
85741718 | 2800 | current_arp_slave, |
5e5b0665 | 2801 | BOND_SLAVE_NOTIFY_NOW); |
85741718 | 2802 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
5a430974 | 2803 | } |
b2220cad | 2804 | |
76444f50 VF |
2805 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2806 | slave->dev->name); | |
b2220cad | 2807 | |
4740d638 | 2808 | if (!rtnl_dereference(bond->curr_active_slave) || |
059b47e8 | 2809 | slave == rtnl_dereference(bond->primary_slave)) |
b9f60253 | 2810 | goto do_failover; |
1da177e4 | 2811 | |
b9f60253 | 2812 | } |
1da177e4 | 2813 | |
b9f60253 | 2814 | continue; |
1da177e4 | 2815 | |
b2220cad JV |
2816 | case BOND_LINK_DOWN: |
2817 | if (slave->link_failure_count < UINT_MAX) | |
2818 | slave->link_failure_count++; | |
2819 | ||
5d397061 JP |
2820 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2821 | BOND_SLAVE_NOTIFY_NOW); | |
5e5b0665 | 2822 | bond_set_slave_inactive_flags(slave, |
2823 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 2824 | |
76444f50 VF |
2825 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2826 | slave->dev->name); | |
b2220cad | 2827 | |
4740d638 | 2828 | if (slave == rtnl_dereference(bond->curr_active_slave)) { |
85741718 | 2829 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
b9f60253 | 2830 | goto do_failover; |
1da177e4 | 2831 | } |
b9f60253 JP |
2832 | |
2833 | continue; | |
b2220cad JV |
2834 | |
2835 | default: | |
76444f50 VF |
2836 | netdev_err(bond->dev, "impossible: new_link %d on slave %s\n", |
2837 | slave->new_link, slave->dev->name); | |
b9f60253 | 2838 | continue; |
1da177e4 | 2839 | } |
1da177e4 | 2840 | |
b9f60253 | 2841 | do_failover: |
e843fa50 | 2842 | block_netpoll_tx(); |
b9f60253 | 2843 | bond_select_active_slave(bond); |
e843fa50 | 2844 | unblock_netpoll_tx(); |
b2220cad | 2845 | } |
1da177e4 | 2846 | |
b2220cad JV |
2847 | bond_set_carrier(bond); |
2848 | } | |
1da177e4 | 2849 | |
547942ca | 2850 | /* Send ARP probes for active-backup mode ARP monitor. |
b0929915 | 2851 | * |
8c0bc550 | 2852 | * Called with rcu_read_lock held. |
b2220cad | 2853 | */ |
f2ebd477 | 2854 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 2855 | { |
eb9fa4b0 | 2856 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 2857 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
2858 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
2859 | struct list_head *iter; |
2860 | bool found = false; | |
b0929915 | 2861 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 2862 | |
98b90f26 | 2863 | if (curr_arp_slave && curr_active_slave) |
76444f50 VF |
2864 | netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n", |
2865 | curr_arp_slave->dev->name, | |
2866 | curr_active_slave->dev->name); | |
1da177e4 | 2867 | |
98b90f26 VF |
2868 | if (curr_active_slave) { |
2869 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 2870 | return should_notify_rtnl; |
b2220cad | 2871 | } |
1da177e4 | 2872 | |
b2220cad JV |
2873 | /* if we don't have a curr_active_slave, search for the next available |
2874 | * backup slave from the current_arp_slave and make it the candidate | |
2875 | * for becoming the curr_active_slave | |
2876 | */ | |
1da177e4 | 2877 | |
eb9fa4b0 | 2878 | if (!curr_arp_slave) { |
b0929915 | 2879 | curr_arp_slave = bond_first_slave_rcu(bond); |
2880 | if (!curr_arp_slave) | |
2881 | return should_notify_rtnl; | |
b2220cad | 2882 | } |
1da177e4 | 2883 | |
b0929915 | 2884 | bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER); |
059fe7a5 | 2885 | |
b0929915 | 2886 | bond_for_each_slave_rcu(bond, slave, iter) { |
b6adc610 | 2887 | if (!found && !before && bond_slave_is_up(slave)) |
4087df87 | 2888 | before = slave; |
1da177e4 | 2889 | |
b6adc610 | 2890 | if (found && !new_slave && bond_slave_is_up(slave)) |
4087df87 | 2891 | new_slave = slave; |
b2220cad JV |
2892 | /* if the link state is up at this point, we |
2893 | * mark it down - this can happen if we have | |
2894 | * simultaneous link failures and | |
2895 | * reselect_active_interface doesn't make this | |
2896 | * one the current slave so it is still marked | |
2897 | * up when it is actually down | |
1da177e4 | 2898 | */ |
b6adc610 | 2899 | if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
5d397061 JP |
2900 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2901 | BOND_SLAVE_NOTIFY_LATER); | |
b2220cad JV |
2902 | if (slave->link_failure_count < UINT_MAX) |
2903 | slave->link_failure_count++; | |
1da177e4 | 2904 | |
5e5b0665 | 2905 | bond_set_slave_inactive_flags(slave, |
b0929915 | 2906 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 2907 | |
76444f50 VF |
2908 | netdev_info(bond->dev, "backup interface %s is now down\n", |
2909 | slave->dev->name); | |
1da177e4 | 2910 | } |
eb9fa4b0 | 2911 | if (slave == curr_arp_slave) |
4087df87 | 2912 | found = true; |
b2220cad | 2913 | } |
4087df87 VF |
2914 | |
2915 | if (!new_slave && before) | |
2916 | new_slave = before; | |
2917 | ||
b0929915 | 2918 | if (!new_slave) |
2919 | goto check_state; | |
4087df87 | 2920 | |
5d397061 JP |
2921 | bond_set_slave_link_state(new_slave, BOND_LINK_BACK, |
2922 | BOND_SLAVE_NOTIFY_LATER); | |
b0929915 | 2923 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 2924 | bond_arp_send_all(bond, new_slave); |
8e603460 | 2925 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 2926 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 2927 | |
b0929915 | 2928 | check_state: |
2929 | bond_for_each_slave_rcu(bond, slave, iter) { | |
5d397061 | 2930 | if (slave->should_notify || slave->should_notify_link) { |
b0929915 | 2931 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; |
2932 | break; | |
2933 | } | |
2934 | } | |
2935 | return should_notify_rtnl; | |
b2220cad | 2936 | } |
1da177e4 | 2937 | |
d5e73f7b | 2938 | static void bond_activebackup_arp_mon(struct bonding *bond) |
b2220cad | 2939 | { |
b0929915 | 2940 | bool should_notify_peers = false; |
2941 | bool should_notify_rtnl = false; | |
1f2cd845 | 2942 | int delta_in_ticks; |
1da177e4 | 2943 | |
1f2cd845 DM |
2944 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
2945 | ||
2946 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
2947 | goto re_arm; |
2948 | ||
eb9fa4b0 | 2949 | rcu_read_lock(); |
b0929915 | 2950 | |
ad246c99 BH |
2951 | should_notify_peers = bond_should_notify_peers(bond); |
2952 | ||
b0929915 | 2953 | if (bond_ab_arp_inspect(bond)) { |
2954 | rcu_read_unlock(); | |
2955 | ||
1f2cd845 DM |
2956 | /* Race avoidance with bond_close flush of workqueue */ |
2957 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2958 | delta_in_ticks = 1; |
2959 | should_notify_peers = false; | |
2960 | goto re_arm; | |
2961 | } | |
b2220cad | 2962 | |
1f2cd845 | 2963 | bond_ab_arp_commit(bond); |
b0929915 | 2964 | |
1f2cd845 | 2965 | rtnl_unlock(); |
b0929915 | 2966 | rcu_read_lock(); |
1f2cd845 DM |
2967 | } |
2968 | ||
b0929915 | 2969 | should_notify_rtnl = bond_ab_arp_probe(bond); |
2970 | rcu_read_unlock(); | |
ad246c99 | 2971 | |
80b9d236 | 2972 | re_arm: |
2973 | if (bond->params.arp_interval) | |
1f2cd845 DM |
2974 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
2975 | ||
b0929915 | 2976 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
2977 | if (!rtnl_trylock()) |
2978 | return; | |
b0929915 | 2979 | |
2980 | if (should_notify_peers) | |
2981 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
2982 | bond->dev); | |
5d397061 | 2983 | if (should_notify_rtnl) { |
b0929915 | 2984 | bond_slave_state_notify(bond); |
5d397061 JP |
2985 | bond_slave_link_notify(bond); |
2986 | } | |
b0929915 | 2987 | |
1f2cd845 DM |
2988 | rtnl_unlock(); |
2989 | } | |
1da177e4 LT |
2990 | } |
2991 | ||
d5e73f7b MB |
2992 | static void bond_arp_monitor(struct work_struct *work) |
2993 | { | |
2994 | struct bonding *bond = container_of(work, struct bonding, | |
2995 | arp_work.work); | |
2996 | ||
2997 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) | |
2998 | bond_activebackup_arp_mon(bond); | |
2999 | else | |
3000 | bond_loadbalance_arp_mon(bond); | |
3001 | } | |
3002 | ||
1da177e4 LT |
3003 | /*-------------------------- netdev event handling --------------------------*/ |
3004 | ||
547942ca | 3005 | /* Change device name */ |
1da177e4 LT |
3006 | static int bond_event_changename(struct bonding *bond) |
3007 | { | |
1da177e4 LT |
3008 | bond_remove_proc_entry(bond); |
3009 | bond_create_proc_entry(bond); | |
7e083840 | 3010 | |
f073c7ca TI |
3011 | bond_debug_reregister(bond); |
3012 | ||
1da177e4 LT |
3013 | return NOTIFY_DONE; |
3014 | } | |
3015 | ||
3d632c3f SH |
3016 | static int bond_master_netdev_event(unsigned long event, |
3017 | struct net_device *bond_dev) | |
1da177e4 | 3018 | { |
454d7c9b | 3019 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3020 | |
3021 | switch (event) { | |
3022 | case NETDEV_CHANGENAME: | |
3023 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3024 | case NETDEV_UNREGISTER: |
3025 | bond_remove_proc_entry(event_bond); | |
3026 | break; | |
3027 | case NETDEV_REGISTER: | |
3028 | bond_create_proc_entry(event_bond); | |
3029 | break; | |
6c8d23f7 | 3030 | case NETDEV_NOTIFY_PEERS: |
3031 | if (event_bond->send_peer_notif) | |
3032 | event_bond->send_peer_notif--; | |
3033 | break; | |
1da177e4 LT |
3034 | default: |
3035 | break; | |
3036 | } | |
3037 | ||
3038 | return NOTIFY_DONE; | |
3039 | } | |
3040 | ||
3d632c3f SH |
3041 | static int bond_slave_netdev_event(unsigned long event, |
3042 | struct net_device *slave_dev) | |
1da177e4 | 3043 | { |
059b47e8 | 3044 | struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary; |
6101391d | 3045 | struct bonding *bond; |
3046 | struct net_device *bond_dev; | |
1da177e4 | 3047 | |
6101391d | 3048 | /* A netdev event can be generated while enslaving a device |
3049 | * before netdev_rx_handler_register is called in which case | |
3050 | * slave will be NULL | |
3051 | */ | |
3052 | if (!slave) | |
3053 | return NOTIFY_DONE; | |
3054 | bond_dev = slave->bond->dev; | |
3055 | bond = slave->bond; | |
059b47e8 | 3056 | primary = rtnl_dereference(bond->primary_slave); |
6101391d | 3057 | |
1da177e4 LT |
3058 | switch (event) { |
3059 | case NETDEV_UNREGISTER: | |
8d2ada77 | 3060 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
3061 | bond_release_and_destroy(bond_dev, slave_dev); |
3062 | else | |
3063 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 3064 | break; |
98f41f69 | 3065 | case NETDEV_UP: |
1da177e4 | 3066 | case NETDEV_CHANGE: |
471cb5a3 | 3067 | bond_update_speed_duplex(slave); |
52bc6716 MB |
3068 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3069 | bond_3ad_adapter_speed_duplex_changed(slave); | |
950ddcb1 MB |
3070 | /* Fallthrough */ |
3071 | case NETDEV_DOWN: | |
ee637714 MB |
3072 | /* Refresh slave-array if applicable! |
3073 | * If the setup does not use miimon or arpmon (mode-specific!), | |
3074 | * then these events will not cause the slave-array to be | |
3075 | * refreshed. This will cause xmit to use a slave that is not | |
3076 | * usable. Avoid such situation by refeshing the array at these | |
3077 | * events. If these (miimon/arpmon) parameters are configured | |
3078 | * then array gets refreshed twice and that should be fine! | |
3079 | */ | |
3080 | if (bond_mode_uses_xmit_hash(bond)) | |
3081 | bond_update_slave_arr(bond, NULL); | |
1da177e4 | 3082 | break; |
1da177e4 | 3083 | case NETDEV_CHANGEMTU: |
547942ca | 3084 | /* TODO: Should slaves be allowed to |
1da177e4 LT |
3085 | * independently alter their MTU? For |
3086 | * an active-backup bond, slaves need | |
3087 | * not be the same type of device, so | |
3088 | * MTUs may vary. For other modes, | |
3089 | * slaves arguably should have the | |
3090 | * same MTUs. To do this, we'd need to | |
3091 | * take over the slave's change_mtu | |
3092 | * function for the duration of their | |
3093 | * servitude. | |
3094 | */ | |
3095 | break; | |
3096 | case NETDEV_CHANGENAME: | |
3ec775b9 | 3097 | /* we don't care if we don't have primary set */ |
ec0865a9 | 3098 | if (!bond_uses_primary(bond) || |
3ec775b9 VF |
3099 | !bond->params.primary[0]) |
3100 | break; | |
3101 | ||
059b47e8 | 3102 | if (slave == primary) { |
3ec775b9 | 3103 | /* slave's name changed - he's no longer primary */ |
059b47e8 | 3104 | RCU_INIT_POINTER(bond->primary_slave, NULL); |
3ec775b9 VF |
3105 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { |
3106 | /* we have a new primary slave */ | |
059b47e8 | 3107 | rcu_assign_pointer(bond->primary_slave, slave); |
3ec775b9 VF |
3108 | } else { /* we didn't change primary - exit */ |
3109 | break; | |
3110 | } | |
3111 | ||
76444f50 | 3112 | netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n", |
059b47e8 | 3113 | primary ? slave_dev->name : "none"); |
f80889a5 | 3114 | |
3115 | block_netpoll_tx(); | |
3ec775b9 | 3116 | bond_select_active_slave(bond); |
f80889a5 | 3117 | unblock_netpoll_tx(); |
1da177e4 | 3118 | break; |
8531c5ff AK |
3119 | case NETDEV_FEAT_CHANGE: |
3120 | bond_compute_features(bond); | |
3121 | break; | |
4aa5dee4 JP |
3122 | case NETDEV_RESEND_IGMP: |
3123 | /* Propagate to master device */ | |
3124 | call_netdevice_notifiers(event, slave->bond->dev); | |
3125 | break; | |
1da177e4 LT |
3126 | default: |
3127 | break; | |
3128 | } | |
3129 | ||
3130 | return NOTIFY_DONE; | |
3131 | } | |
3132 | ||
547942ca | 3133 | /* bond_netdev_event: handle netdev notifier chain events. |
1da177e4 LT |
3134 | * |
3135 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3136 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3137 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3138 | * dev_probe_lock). | |
3139 | */ | |
3d632c3f SH |
3140 | static int bond_netdev_event(struct notifier_block *this, |
3141 | unsigned long event, void *ptr) | |
1da177e4 | 3142 | { |
351638e7 | 3143 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3144 | |
76444f50 | 3145 | netdev_dbg(event_dev, "event: %lx\n", event); |
1da177e4 | 3146 | |
0b680e75 JV |
3147 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3148 | return NOTIFY_DONE; | |
3149 | ||
1da177e4 | 3150 | if (event_dev->flags & IFF_MASTER) { |
76444f50 | 3151 | netdev_dbg(event_dev, "IFF_MASTER\n"); |
1da177e4 LT |
3152 | return bond_master_netdev_event(event, event_dev); |
3153 | } | |
3154 | ||
3155 | if (event_dev->flags & IFF_SLAVE) { | |
76444f50 | 3156 | netdev_dbg(event_dev, "IFF_SLAVE\n"); |
1da177e4 LT |
3157 | return bond_slave_netdev_event(event, event_dev); |
3158 | } | |
3159 | ||
3160 | return NOTIFY_DONE; | |
3161 | } | |
3162 | ||
3163 | static struct notifier_block bond_netdev_notifier = { | |
3164 | .notifier_call = bond_netdev_event, | |
3165 | }; | |
3166 | ||
169a3e66 JV |
3167 | /*---------------------------- Hashing Policies -----------------------------*/ |
3168 | ||
32819dc1 NA |
3169 | /* L2 hash helper */ |
3170 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 | 3171 | { |
ce04d635 | 3172 | struct ethhdr *ep, hdr_tmp; |
6b923cb7 | 3173 | |
ce04d635 JX |
3174 | ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp); |
3175 | if (ep) | |
3176 | return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto; | |
6b923cb7 JE |
3177 | return 0; |
3178 | } | |
3179 | ||
32819dc1 NA |
3180 | /* Extract the appropriate headers based on bond's xmit policy */ |
3181 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
3182 | struct flow_keys *fk) | |
6f6652be | 3183 | { |
32819dc1 | 3184 | const struct ipv6hdr *iph6; |
4394542c | 3185 | const struct iphdr *iph; |
32819dc1 | 3186 | int noff, proto = -1; |
6b923cb7 | 3187 | |
32819dc1 | 3188 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) |
cd79a238 | 3189 | return skb_flow_dissect_flow_keys(skb, fk, 0); |
32819dc1 | 3190 | |
06635a35 | 3191 | fk->ports.ports = 0; |
32819dc1 NA |
3192 | noff = skb_network_offset(skb); |
3193 | if (skb->protocol == htons(ETH_P_IP)) { | |
054bb880 | 3194 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
32819dc1 | 3195 | return false; |
6b923cb7 | 3196 | iph = ip_hdr(skb); |
c3f83241 | 3197 | iph_to_flow_copy_v4addrs(fk, iph); |
32819dc1 NA |
3198 | noff += iph->ihl << 2; |
3199 | if (!ip_is_fragment(iph)) | |
3200 | proto = iph->protocol; | |
3201 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
054bb880 | 3202 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6)))) |
32819dc1 NA |
3203 | return false; |
3204 | iph6 = ipv6_hdr(skb); | |
c3f83241 | 3205 | iph_to_flow_copy_v6addrs(fk, iph6); |
32819dc1 NA |
3206 | noff += sizeof(*iph6); |
3207 | proto = iph6->nexthdr; | |
3208 | } else { | |
3209 | return false; | |
6f6652be | 3210 | } |
32819dc1 | 3211 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) |
06635a35 | 3212 | fk->ports.ports = skb_flow_get_ports(skb, noff, proto); |
6f6652be | 3213 | |
32819dc1 | 3214 | return true; |
6f6652be JV |
3215 | } |
3216 | ||
32819dc1 NA |
3217 | /** |
3218 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3219 | * @bond: bonding device | |
3220 | * @skb: buffer to use for headers | |
32819dc1 NA |
3221 | * |
3222 | * This function will extract the necessary headers from the skb buffer and use | |
3223 | * them to generate a hash based on the xmit_policy set in the bonding device | |
169a3e66 | 3224 | */ |
ee62e868 | 3225 | u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) |
169a3e66 | 3226 | { |
32819dc1 NA |
3227 | struct flow_keys flow; |
3228 | u32 hash; | |
4394542c | 3229 | |
4b1b865e ED |
3230 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 && |
3231 | skb->l4_hash) | |
3232 | return skb->hash; | |
3233 | ||
32819dc1 NA |
3234 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3235 | !bond_flow_dissect(bond, skb, &flow)) | |
ee62e868 | 3236 | return bond_eth_hash(skb); |
169a3e66 | 3237 | |
32819dc1 NA |
3238 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
3239 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) | |
3240 | hash = bond_eth_hash(skb); | |
3241 | else | |
06635a35 | 3242 | hash = (__force u32)flow.ports.ports; |
c3f83241 TH |
3243 | hash ^= (__force u32)flow_get_u32_dst(&flow) ^ |
3244 | (__force u32)flow_get_u32_src(&flow); | |
32819dc1 NA |
3245 | hash ^= (hash >> 16); |
3246 | hash ^= (hash >> 8); | |
3247 | ||
ee62e868 | 3248 | return hash; |
169a3e66 JV |
3249 | } |
3250 | ||
1da177e4 LT |
3251 | /*-------------------------- Device entry points ----------------------------*/ |
3252 | ||
ea8ffc08 | 3253 | void bond_work_init_all(struct bonding *bond) |
fbb0c41b | 3254 | { |
3255 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3256 | bond_resend_igmp_join_requests_delayed); | |
3257 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3258 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
d5e73f7b | 3259 | INIT_DELAYED_WORK(&bond->arp_work, bond_arp_monitor); |
fbb0c41b | 3260 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
ee637714 | 3261 | INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler); |
fbb0c41b | 3262 | } |
3263 | ||
3264 | static void bond_work_cancel_all(struct bonding *bond) | |
3265 | { | |
3266 | cancel_delayed_work_sync(&bond->mii_work); | |
3267 | cancel_delayed_work_sync(&bond->arp_work); | |
3268 | cancel_delayed_work_sync(&bond->alb_work); | |
3269 | cancel_delayed_work_sync(&bond->ad_work); | |
3270 | cancel_delayed_work_sync(&bond->mcast_work); | |
ee637714 | 3271 | cancel_delayed_work_sync(&bond->slave_arr_work); |
fbb0c41b | 3272 | } |
3273 | ||
1da177e4 LT |
3274 | static int bond_open(struct net_device *bond_dev) |
3275 | { | |
454d7c9b | 3276 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3277 | struct list_head *iter; |
ba3211cc | 3278 | struct slave *slave; |
1da177e4 | 3279 | |
ba3211cc | 3280 | /* reset slave->backup and slave->inactive */ |
0965a1f3 | 3281 | if (bond_has_slaves(bond)) { |
9caff1e7 | 3282 | bond_for_each_slave(bond, slave, iter) { |
4740d638 ED |
3283 | if (bond_uses_primary(bond) && |
3284 | slave != rcu_access_pointer(bond->curr_active_slave)) { | |
5e5b0665 | 3285 | bond_set_slave_inactive_flags(slave, |
3286 | BOND_SLAVE_NOTIFY_NOW); | |
8bbe71a5 | 3287 | } else if (BOND_MODE(bond) != BOND_MODE_8023AD) { |
5e5b0665 | 3288 | bond_set_slave_active_flags(slave, |
3289 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3290 | } |
3291 | } | |
ba3211cc | 3292 | } |
ba3211cc | 3293 | |
58402054 | 3294 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3295 | /* bond_alb_initialize must be called before the timer |
3296 | * is started. | |
3297 | */ | |
01844098 | 3298 | if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) |
b473946a | 3299 | return -ENOMEM; |
e9f0fb88 MB |
3300 | if (bond->params.tlb_dynamic_lb) |
3301 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3302 | } |
3303 | ||
fbb0c41b | 3304 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3305 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3306 | |
3307 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3308 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3309 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3310 | } |
3311 | ||
01844098 | 3312 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1b76b316 | 3313 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3314 | /* register to receive LACPDUs */ |
3aba891d | 3315 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3316 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3317 | } |
3318 | ||
ee637714 MB |
3319 | if (bond_mode_uses_xmit_hash(bond)) |
3320 | bond_update_slave_arr(bond, NULL); | |
3321 | ||
1da177e4 LT |
3322 | return 0; |
3323 | } | |
3324 | ||
3325 | static int bond_close(struct net_device *bond_dev) | |
3326 | { | |
454d7c9b | 3327 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3328 | |
fbb0c41b | 3329 | bond_work_cancel_all(bond); |
6c8d23f7 | 3330 | bond->send_peer_notif = 0; |
ee8487c0 | 3331 | if (bond_is_lb(bond)) |
1da177e4 | 3332 | bond_alb_deinitialize(bond); |
3aba891d | 3333 | bond->recv_probe = NULL; |
1da177e4 LT |
3334 | |
3335 | return 0; | |
3336 | } | |
3337 | ||
fe30937b ED |
3338 | /* fold stats, assuming all rtnl_link_stats64 fields are u64, but |
3339 | * that some drivers can provide 32bit values only. | |
3340 | */ | |
3341 | static void bond_fold_stats(struct rtnl_link_stats64 *_res, | |
3342 | const struct rtnl_link_stats64 *_new, | |
3343 | const struct rtnl_link_stats64 *_old) | |
3344 | { | |
3345 | const u64 *new = (const u64 *)_new; | |
3346 | const u64 *old = (const u64 *)_old; | |
3347 | u64 *res = (u64 *)_res; | |
3348 | int i; | |
3349 | ||
3350 | for (i = 0; i < sizeof(*_res) / sizeof(u64); i++) { | |
3351 | u64 nv = new[i]; | |
3352 | u64 ov = old[i]; | |
142c6594 | 3353 | s64 delta = nv - ov; |
fe30937b ED |
3354 | |
3355 | /* detects if this particular field is 32bit only */ | |
3356 | if (((nv | ov) >> 32) == 0) | |
142c6594 ED |
3357 | delta = (s64)(s32)((u32)nv - (u32)ov); |
3358 | ||
3359 | /* filter anomalies, some drivers reset their stats | |
3360 | * at down/up events. | |
3361 | */ | |
3362 | if (delta > 0) | |
3363 | res[i] += delta; | |
fe30937b ED |
3364 | } |
3365 | } | |
3366 | ||
bc1f4470 | 3367 | static void bond_get_stats(struct net_device *bond_dev, |
3368 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3369 | { |
454d7c9b | 3370 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3371 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3372 | struct list_head *iter; |
1da177e4 | 3373 | struct slave *slave; |
1da177e4 | 3374 | |
fe30937b | 3375 | spin_lock(&bond->stats_lock); |
5f0c5f73 | 3376 | memcpy(stats, &bond->bond_stats, sizeof(*stats)); |
1da177e4 | 3377 | |
fe30937b ED |
3378 | rcu_read_lock(); |
3379 | bond_for_each_slave_rcu(bond, slave, iter) { | |
3380 | const struct rtnl_link_stats64 *new = | |
28172739 | 3381 | dev_get_stats(slave->dev, &temp); |
fe30937b ED |
3382 | |
3383 | bond_fold_stats(stats, new, &slave->slave_stats); | |
5f0c5f73 AG |
3384 | |
3385 | /* save off the slave stats for the next run */ | |
fe30937b | 3386 | memcpy(&slave->slave_stats, new, sizeof(*new)); |
5f0c5f73 | 3387 | } |
fe30937b ED |
3388 | rcu_read_unlock(); |
3389 | ||
5f0c5f73 | 3390 | memcpy(&bond->bond_stats, stats, sizeof(*stats)); |
fe30937b | 3391 | spin_unlock(&bond->stats_lock); |
1da177e4 LT |
3392 | } |
3393 | ||
3394 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3395 | { | |
080a06e1 | 3396 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3397 | struct net_device *slave_dev = NULL; |
3398 | struct ifbond k_binfo; | |
3399 | struct ifbond __user *u_binfo = NULL; | |
3400 | struct ifslave k_sinfo; | |
3401 | struct ifslave __user *u_sinfo = NULL; | |
3402 | struct mii_ioctl_data *mii = NULL; | |
d1fbd3ed | 3403 | struct bond_opt_value newval; |
387ff911 | 3404 | struct net *net; |
1da177e4 LT |
3405 | int res = 0; |
3406 | ||
76444f50 | 3407 | netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd); |
1da177e4 LT |
3408 | |
3409 | switch (cmd) { | |
1da177e4 LT |
3410 | case SIOCGMIIPHY: |
3411 | mii = if_mii(ifr); | |
3d632c3f | 3412 | if (!mii) |
1da177e4 | 3413 | return -EINVAL; |
3d632c3f | 3414 | |
1da177e4 LT |
3415 | mii->phy_id = 0; |
3416 | /* Fall Through */ | |
3417 | case SIOCGMIIREG: | |
547942ca | 3418 | /* We do this again just in case we were called by SIOCGMIIREG |
1da177e4 LT |
3419 | * instead of SIOCGMIIPHY. |
3420 | */ | |
3421 | mii = if_mii(ifr); | |
3d632c3f | 3422 | if (!mii) |
1da177e4 | 3423 | return -EINVAL; |
3d632c3f | 3424 | |
1da177e4 | 3425 | if (mii->reg_num == 1) { |
1da177e4 | 3426 | mii->val_out = 0; |
3d632c3f | 3427 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3428 | mii->val_out = BMSR_LSTATUS; |
1da177e4 LT |
3429 | } |
3430 | ||
3431 | return 0; | |
3432 | case BOND_INFO_QUERY_OLD: | |
3433 | case SIOCBONDINFOQUERY: | |
3434 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3435 | ||
3d632c3f | 3436 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3437 | return -EFAULT; |
1da177e4 | 3438 | |
3d67576d ZY |
3439 | bond_info_query(bond_dev, &k_binfo); |
3440 | if (copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3d632c3f | 3441 | return -EFAULT; |
1da177e4 | 3442 | |
3d67576d | 3443 | return 0; |
1da177e4 LT |
3444 | case BOND_SLAVE_INFO_QUERY_OLD: |
3445 | case SIOCBONDSLAVEINFOQUERY: | |
3446 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3447 | ||
3d632c3f | 3448 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3449 | return -EFAULT; |
1da177e4 LT |
3450 | |
3451 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3452 | if (res == 0 && |
3453 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3454 | return -EFAULT; | |
1da177e4 LT |
3455 | |
3456 | return res; | |
3457 | default: | |
1da177e4 LT |
3458 | break; |
3459 | } | |
3460 | ||
387ff911 G |
3461 | net = dev_net(bond_dev); |
3462 | ||
3463 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3464 | return -EPERM; |
1da177e4 | 3465 | |
0917b933 | 3466 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3467 | |
76444f50 | 3468 | netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev); |
1da177e4 | 3469 | |
3d632c3f | 3470 | if (!slave_dev) |
0917b933 | 3471 | return -ENODEV; |
1da177e4 | 3472 | |
76444f50 | 3473 | netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name); |
0917b933 YX |
3474 | switch (cmd) { |
3475 | case BOND_ENSLAVE_OLD: | |
3476 | case SIOCBONDENSLAVE: | |
3477 | res = bond_enslave(bond_dev, slave_dev); | |
3478 | break; | |
3479 | case BOND_RELEASE_OLD: | |
3480 | case SIOCBONDRELEASE: | |
3481 | res = bond_release(bond_dev, slave_dev); | |
3482 | break; | |
3483 | case BOND_SETHWADDR_OLD: | |
3484 | case SIOCBONDSETHWADDR: | |
3485 | bond_set_dev_addr(bond_dev, slave_dev); | |
3486 | res = 0; | |
3487 | break; | |
3488 | case BOND_CHANGE_ACTIVE_OLD: | |
3489 | case SIOCBONDCHANGEACTIVE: | |
d1fbd3ed NA |
3490 | bond_opt_initstr(&newval, slave_dev->name); |
3491 | res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval); | |
0917b933 YX |
3492 | break; |
3493 | default: | |
3494 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3495 | } |
3496 | ||
1da177e4 LT |
3497 | return res; |
3498 | } | |
3499 | ||
d03462b9 | 3500 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3501 | { |
454d7c9b | 3502 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3503 | |
d03462b9 JP |
3504 | if (change & IFF_PROMISC) |
3505 | bond_set_promiscuity(bond, | |
3506 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3507 | |
d03462b9 JP |
3508 | if (change & IFF_ALLMULTI) |
3509 | bond_set_allmulti(bond, | |
3510 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3511 | } | |
1da177e4 | 3512 | |
303d1cbf | 3513 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3514 | { |
3515 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3516 | struct list_head *iter; |
303d1cbf | 3517 | struct slave *slave; |
1da177e4 | 3518 | |
b3241870 | 3519 | rcu_read_lock(); |
ec0865a9 | 3520 | if (bond_uses_primary(bond)) { |
b3241870 | 3521 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3522 | if (slave) { |
3523 | dev_uc_sync(slave->dev, bond_dev); | |
3524 | dev_mc_sync(slave->dev, bond_dev); | |
3525 | } | |
303d1cbf | 3526 | } else { |
b3241870 | 3527 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3528 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3529 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3530 | } | |
1da177e4 | 3531 | } |
b3241870 | 3532 | rcu_read_unlock(); |
1da177e4 LT |
3533 | } |
3534 | ||
234bcf8a | 3535 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3536 | { |
234bcf8a | 3537 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3538 | const struct net_device_ops *slave_ops; |
3539 | struct neigh_parms parms; | |
dec1e90e | 3540 | struct slave *slave; |
234bcf8a SP |
3541 | int ret; |
3542 | ||
dec1e90e | 3543 | slave = bond_first_slave(bond); |
234bcf8a SP |
3544 | if (!slave) |
3545 | return 0; | |
234bcf8a | 3546 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3547 | if (!slave_ops->ndo_neigh_setup) |
3548 | return 0; | |
3549 | ||
3550 | parms.neigh_setup = NULL; | |
3551 | parms.neigh_cleanup = NULL; | |
3552 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3553 | if (ret) | |
3554 | return ret; | |
3555 | ||
547942ca | 3556 | /* Assign slave's neigh_cleanup to neighbour in case cleanup is called |
234bcf8a SP |
3557 | * after the last slave has been detached. Assumes that all slaves |
3558 | * utilize the same neigh_cleanup (true at this writing as only user | |
3559 | * is ipoib). | |
3560 | */ | |
3561 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3562 | ||
3563 | if (!parms.neigh_setup) | |
3564 | return 0; | |
3565 | ||
3566 | return parms.neigh_setup(n); | |
3567 | } | |
3568 | ||
547942ca | 3569 | /* The bonding ndo_neigh_setup is called at init time beofre any |
234bcf8a SP |
3570 | * slave exists. So we must declare proxy setup function which will |
3571 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3572 | * |
3573 | * It's also called by master devices (such as vlans) to setup their | |
3574 | * underlying devices. In that case - do nothing, we're already set up from | |
3575 | * our init. | |
234bcf8a SP |
3576 | */ |
3577 | static int bond_neigh_setup(struct net_device *dev, | |
3578 | struct neigh_parms *parms) | |
3579 | { | |
9918d5bf VF |
3580 | /* modify only our neigh_parms */ |
3581 | if (parms->dev == dev) | |
3582 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3583 | |
00829823 SH |
3584 | return 0; |
3585 | } | |
3586 | ||
547942ca | 3587 | /* Change the MTU of all of a master's slaves to match the master */ |
1da177e4 LT |
3588 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) |
3589 | { | |
454d7c9b | 3590 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3591 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3592 | struct list_head *iter; |
1da177e4 | 3593 | int res = 0; |
1da177e4 | 3594 | |
76444f50 | 3595 | netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu); |
1da177e4 | 3596 | |
9caff1e7 | 3597 | bond_for_each_slave(bond, slave, iter) { |
76444f50 VF |
3598 | netdev_dbg(bond_dev, "s %p c_m %p\n", |
3599 | slave, slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3600 | |
1da177e4 LT |
3601 | res = dev_set_mtu(slave->dev, new_mtu); |
3602 | ||
3603 | if (res) { | |
3604 | /* If we failed to set the slave's mtu to the new value | |
3605 | * we must abort the operation even in ACTIVE_BACKUP | |
3606 | * mode, because if we allow the backup slaves to have | |
3607 | * different mtu values than the active slave we'll | |
3608 | * need to change their mtu when doing a failover. That | |
3609 | * means changing their mtu from timer context, which | |
3610 | * is probably not a good idea. | |
3611 | */ | |
76444f50 VF |
3612 | netdev_dbg(bond_dev, "err %d %s\n", res, |
3613 | slave->dev->name); | |
1da177e4 LT |
3614 | goto unwind; |
3615 | } | |
3616 | } | |
3617 | ||
3618 | bond_dev->mtu = new_mtu; | |
3619 | ||
3620 | return 0; | |
3621 | ||
3622 | unwind: | |
3623 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3624 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3625 | int tmp_res; |
3626 | ||
81f23b13 VF |
3627 | if (rollback_slave == slave) |
3628 | break; | |
3629 | ||
3630 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3631 | if (tmp_res) { |
76444f50 VF |
3632 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3633 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3634 | } |
3635 | } | |
3636 | ||
3637 | return res; | |
3638 | } | |
3639 | ||
547942ca | 3640 | /* Change HW address |
1da177e4 LT |
3641 | * |
3642 | * Note that many devices must be down to change the HW address, and | |
3643 | * downing the master releases all slaves. We can make bonds full of | |
3644 | * bonding devices to test this, however. | |
3645 | */ | |
3646 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3647 | { | |
454d7c9b | 3648 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3649 | struct slave *slave, *rollback_slave; |
faeeb317 | 3650 | struct sockaddr_storage *ss = addr, tmp_ss; |
9caff1e7 | 3651 | struct list_head *iter; |
1da177e4 | 3652 | int res = 0; |
1da177e4 | 3653 | |
01844098 | 3654 | if (BOND_MODE(bond) == BOND_MODE_ALB) |
eb7cc59a SH |
3655 | return bond_alb_set_mac_address(bond_dev, addr); |
3656 | ||
3657 | ||
76444f50 | 3658 | netdev_dbg(bond_dev, "bond=%p\n", bond); |
1da177e4 | 3659 | |
1b5acd29 JV |
3660 | /* If fail_over_mac is enabled, do nothing and return success. |
3661 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3662 | */ |
cc689aaa | 3663 | if (bond->params.fail_over_mac && |
01844098 | 3664 | BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
dd957c57 | 3665 | return 0; |
2ab82852 | 3666 | |
faeeb317 | 3667 | if (!is_valid_ether_addr(ss->__data)) |
1da177e4 | 3668 | return -EADDRNOTAVAIL; |
1da177e4 | 3669 | |
9caff1e7 | 3670 | bond_for_each_slave(bond, slave, iter) { |
76444f50 | 3671 | netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name); |
1da177e4 LT |
3672 | res = dev_set_mac_address(slave->dev, addr); |
3673 | if (res) { | |
3674 | /* TODO: consider downing the slave | |
3675 | * and retry ? | |
3676 | * User should expect communications | |
3677 | * breakage anyway until ARP finish | |
3678 | * updating, so... | |
3679 | */ | |
76444f50 | 3680 | netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3681 | goto unwind; |
3682 | } | |
3683 | } | |
3684 | ||
3685 | /* success */ | |
faeeb317 | 3686 | memcpy(bond_dev->dev_addr, ss->__data, bond_dev->addr_len); |
1da177e4 LT |
3687 | return 0; |
3688 | ||
3689 | unwind: | |
faeeb317 JW |
3690 | memcpy(tmp_ss.__data, bond_dev->dev_addr, bond_dev->addr_len); |
3691 | tmp_ss.ss_family = bond_dev->type; | |
1da177e4 LT |
3692 | |
3693 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3694 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3695 | int tmp_res; |
3696 | ||
81f23b13 VF |
3697 | if (rollback_slave == slave) |
3698 | break; | |
3699 | ||
faeeb317 JW |
3700 | tmp_res = dev_set_mac_address(rollback_slave->dev, |
3701 | (struct sockaddr *)&tmp_ss); | |
1da177e4 | 3702 | if (tmp_res) { |
76444f50 VF |
3703 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3704 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3705 | } |
3706 | } | |
3707 | ||
3708 | return res; | |
3709 | } | |
3710 | ||
15077228 NA |
3711 | /** |
3712 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3713 | * @bond: bonding device that is transmitting | |
3714 | * @skb: buffer to transmit | |
3715 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3716 | * | |
3717 | * This function tries to transmit through slave with slave_id but in case | |
3718 | * it fails, it tries to find the first available slave for transmission. | |
3719 | * The skb is consumed in all cases, thus the function is void. | |
3720 | */ | |
da131ddb | 3721 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3722 | { |
9caff1e7 | 3723 | struct list_head *iter; |
15077228 NA |
3724 | struct slave *slave; |
3725 | int i = slave_id; | |
3726 | ||
3727 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3728 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 | 3729 | if (--i < 0) { |
891ab54d | 3730 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3731 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3732 | return; | |
3733 | } | |
3734 | } | |
3735 | } | |
3736 | ||
3737 | /* Here we start from the first slave up to slave_id */ | |
3738 | i = slave_id; | |
9caff1e7 | 3739 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3740 | if (--i < 0) |
3741 | break; | |
891ab54d | 3742 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3743 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3744 | return; | |
3745 | } | |
3746 | } | |
3747 | /* no slave that can tx has been found */ | |
31aa860e | 3748 | bond_tx_drop(bond->dev, skb); |
15077228 NA |
3749 | } |
3750 | ||
73958329 NA |
3751 | /** |
3752 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3753 | * @bond: bonding device to use | |
3754 | * | |
3755 | * Based on the value of the bonding device's packets_per_slave parameter | |
3756 | * this function generates a slave id, which is usually used as the next | |
3757 | * slave to transmit through. | |
3758 | */ | |
3759 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3760 | { | |
73958329 | 3761 | u32 slave_id; |
809fa972 HFS |
3762 | struct reciprocal_value reciprocal_packets_per_slave; |
3763 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
3764 | |
3765 | switch (packets_per_slave) { | |
3766 | case 0: | |
3767 | slave_id = prandom_u32(); | |
3768 | break; | |
3769 | case 1: | |
3770 | slave_id = bond->rr_tx_counter; | |
3771 | break; | |
3772 | default: | |
809fa972 HFS |
3773 | reciprocal_packets_per_slave = |
3774 | bond->params.reciprocal_packets_per_slave; | |
73958329 | 3775 | slave_id = reciprocal_divide(bond->rr_tx_counter, |
809fa972 | 3776 | reciprocal_packets_per_slave); |
73958329 NA |
3777 | break; |
3778 | } | |
3779 | bond->rr_tx_counter++; | |
3780 | ||
3781 | return slave_id; | |
3782 | } | |
3783 | ||
1da177e4 LT |
3784 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3785 | { | |
454d7c9b | 3786 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3787 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3788 | struct slave *slave; |
73958329 | 3789 | u32 slave_id; |
1da177e4 | 3790 | |
73958329 | 3791 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3792 | * default for sending IGMP traffic. For failover purposes one |
3793 | * needs to maintain some consistency for the interface that will | |
3794 | * send the join/membership reports. The curr_active_slave found | |
3795 | * will send all of this type of traffic. | |
cf5f9044 | 3796 | */ |
15077228 | 3797 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3798 | slave = rcu_dereference(bond->curr_active_slave); |
a64d044e | 3799 | if (slave) |
15077228 NA |
3800 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3801 | else | |
3802 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3803 | } else { |
9a72c2da NA |
3804 | int slave_cnt = ACCESS_ONCE(bond->slave_cnt); |
3805 | ||
3806 | if (likely(slave_cnt)) { | |
3807 | slave_id = bond_rr_gen_slave_id(bond); | |
3808 | bond_xmit_slave_id(bond, skb, slave_id % slave_cnt); | |
3809 | } else { | |
31aa860e | 3810 | bond_tx_drop(bond_dev, skb); |
9a72c2da | 3811 | } |
1da177e4 | 3812 | } |
0693e88e | 3813 | |
ec634fe3 | 3814 | return NETDEV_TX_OK; |
1da177e4 LT |
3815 | } |
3816 | ||
547942ca | 3817 | /* In active-backup mode, we know that bond->curr_active_slave is always valid if |
1da177e4 LT |
3818 | * the bond has a usable interface. |
3819 | */ | |
3820 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3821 | { | |
454d7c9b | 3822 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3823 | struct slave *slave; |
1da177e4 | 3824 | |
278b2083 | 3825 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3826 | if (slave) |
15077228 NA |
3827 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3828 | else | |
31aa860e | 3829 | bond_tx_drop(bond_dev, skb); |
0693e88e | 3830 | |
ec634fe3 | 3831 | return NETDEV_TX_OK; |
1da177e4 LT |
3832 | } |
3833 | ||
ee637714 MB |
3834 | /* Use this to update slave_array when (a) it's not appropriate to update |
3835 | * slave_array right away (note that update_slave_array() may sleep) | |
3836 | * and / or (b) RTNL is not held. | |
1da177e4 | 3837 | */ |
ee637714 | 3838 | void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay) |
1da177e4 | 3839 | { |
ee637714 MB |
3840 | queue_delayed_work(bond->wq, &bond->slave_arr_work, delay); |
3841 | } | |
1da177e4 | 3842 | |
ee637714 MB |
3843 | /* Slave array work handler. Holds only RTNL */ |
3844 | static void bond_slave_arr_handler(struct work_struct *work) | |
3845 | { | |
3846 | struct bonding *bond = container_of(work, struct bonding, | |
3847 | slave_arr_work.work); | |
3848 | int ret; | |
3849 | ||
3850 | if (!rtnl_trylock()) | |
3851 | goto err; | |
3852 | ||
3853 | ret = bond_update_slave_arr(bond, NULL); | |
3854 | rtnl_unlock(); | |
3855 | if (ret) { | |
3856 | pr_warn_ratelimited("Failed to update slave array from WT\n"); | |
3857 | goto err; | |
3858 | } | |
3859 | return; | |
3860 | ||
3861 | err: | |
3862 | bond_slave_arr_work_rearm(bond, 1); | |
3863 | } | |
3864 | ||
3865 | /* Build the usable slaves array in control path for modes that use xmit-hash | |
3866 | * to determine the slave interface - | |
3867 | * (a) BOND_MODE_8023AD | |
3868 | * (b) BOND_MODE_XOR | |
3869 | * (c) BOND_MODE_TLB && tlb_dynamic_lb == 0 | |
3870 | * | |
3871 | * The caller is expected to hold RTNL only and NO other lock! | |
3872 | */ | |
3873 | int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave) | |
3874 | { | |
3875 | struct slave *slave; | |
3876 | struct list_head *iter; | |
3877 | struct bond_up_slave *new_arr, *old_arr; | |
ee637714 MB |
3878 | int agg_id = 0; |
3879 | int ret = 0; | |
3880 | ||
3881 | #ifdef CONFIG_LOCKDEP | |
3882 | WARN_ON(lockdep_is_held(&bond->mode_lock)); | |
3883 | #endif | |
3884 | ||
3885 | new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]), | |
3886 | GFP_KERNEL); | |
3887 | if (!new_arr) { | |
3888 | ret = -ENOMEM; | |
3889 | pr_err("Failed to build slave-array.\n"); | |
3890 | goto out; | |
3891 | } | |
3892 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
3893 | struct ad_info ad_info; | |
3894 | ||
3895 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3896 | pr_debug("bond_3ad_get_active_agg_info failed\n"); | |
3897 | kfree_rcu(new_arr, rcu); | |
3898 | /* No active aggragator means it's not safe to use | |
3899 | * the previous array. | |
3900 | */ | |
3901 | old_arr = rtnl_dereference(bond->slave_arr); | |
3902 | if (old_arr) { | |
3903 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
3904 | kfree_rcu(old_arr, rcu); | |
3905 | } | |
3906 | goto out; | |
3907 | } | |
ee637714 MB |
3908 | agg_id = ad_info.aggregator_id; |
3909 | } | |
3910 | bond_for_each_slave(bond, slave, iter) { | |
3911 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
3912 | struct aggregator *agg; | |
3913 | ||
3914 | agg = SLAVE_AD_INFO(slave)->port.aggregator; | |
3915 | if (!agg || agg->aggregator_identifier != agg_id) | |
3916 | continue; | |
3917 | } | |
3918 | if (!bond_slave_can_tx(slave)) | |
3919 | continue; | |
3920 | if (skipslave == slave) | |
3921 | continue; | |
3922 | new_arr->arr[new_arr->count++] = slave; | |
3923 | } | |
3924 | ||
3925 | old_arr = rtnl_dereference(bond->slave_arr); | |
3926 | rcu_assign_pointer(bond->slave_arr, new_arr); | |
3927 | if (old_arr) | |
3928 | kfree_rcu(old_arr, rcu); | |
3929 | out: | |
3930 | if (ret != 0 && skipslave) { | |
3931 | int idx; | |
3932 | ||
3933 | /* Rare situation where caller has asked to skip a specific | |
3934 | * slave but allocation failed (most likely!). BTW this is | |
3935 | * only possible when the call is initiated from | |
3936 | * __bond_release_one(). In this situation; overwrite the | |
3937 | * skipslave entry in the array with the last entry from the | |
3938 | * array to avoid a situation where the xmit path may choose | |
3939 | * this to-be-skipped slave to send a packet out. | |
3940 | */ | |
3941 | old_arr = rtnl_dereference(bond->slave_arr); | |
3942 | for (idx = 0; idx < old_arr->count; idx++) { | |
3943 | if (skipslave == old_arr->arr[idx]) { | |
3944 | old_arr->arr[idx] = | |
3945 | old_arr->arr[old_arr->count-1]; | |
3946 | old_arr->count--; | |
3947 | break; | |
3948 | } | |
3949 | } | |
3950 | } | |
3951 | return ret; | |
3952 | } | |
3953 | ||
3954 | /* Use this Xmit function for 3AD as well as XOR modes. The current | |
3955 | * usable slave array is formed in the control path. The xmit function | |
3956 | * just calculates hash and sends the packet out. | |
3957 | */ | |
30369104 | 3958 | static int bond_3ad_xor_xmit(struct sk_buff *skb, struct net_device *dev) |
ee637714 MB |
3959 | { |
3960 | struct bonding *bond = netdev_priv(dev); | |
3961 | struct slave *slave; | |
3962 | struct bond_up_slave *slaves; | |
3963 | unsigned int count; | |
3964 | ||
3965 | slaves = rcu_dereference(bond->slave_arr); | |
3966 | count = slaves ? ACCESS_ONCE(slaves->count) : 0; | |
3967 | if (likely(count)) { | |
3968 | slave = slaves->arr[bond_xmit_hash(bond, skb) % count]; | |
3969 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3970 | } else { | |
31aa860e | 3971 | bond_tx_drop(dev, skb); |
ee637714 | 3972 | } |
0693e88e | 3973 | |
ec634fe3 | 3974 | return NETDEV_TX_OK; |
1da177e4 LT |
3975 | } |
3976 | ||
78a646ce | 3977 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3978 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3979 | { | |
454d7c9b | 3980 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3981 | struct slave *slave = NULL; |
9caff1e7 | 3982 | struct list_head *iter; |
1da177e4 | 3983 | |
9caff1e7 | 3984 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
3985 | if (bond_is_last_slave(bond, slave)) |
3986 | break; | |
b6adc610 | 3987 | if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
78a646ce | 3988 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 | 3989 | |
78a646ce | 3990 | if (!skb2) { |
9152e26d | 3991 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
3992 | bond_dev->name, __func__); | |
78a646ce | 3993 | continue; |
1da177e4 | 3994 | } |
78a646ce | 3995 | bond_dev_queue_xmit(bond, skb2, slave->dev); |
1da177e4 LT |
3996 | } |
3997 | } | |
b6adc610 | 3998 | if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) |
78a646ce NA |
3999 | bond_dev_queue_xmit(bond, skb, slave->dev); |
4000 | else | |
31aa860e | 4001 | bond_tx_drop(bond_dev, skb); |
3d632c3f | 4002 | |
ec634fe3 | 4003 | return NETDEV_TX_OK; |
1da177e4 LT |
4004 | } |
4005 | ||
4006 | /*------------------------- Device initialization ---------------------------*/ | |
4007 | ||
547942ca | 4008 | /* Lookup the slave that corresponds to a qid */ |
bb1d9123 AG |
4009 | static inline int bond_slave_override(struct bonding *bond, |
4010 | struct sk_buff *skb) | |
4011 | { | |
bb1d9123 | 4012 | struct slave *slave = NULL; |
9caff1e7 | 4013 | struct list_head *iter; |
bb1d9123 | 4014 | |
0693e88e MM |
4015 | if (!skb->queue_mapping) |
4016 | return 1; | |
bb1d9123 AG |
4017 | |
4018 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 4019 | bond_for_each_slave_rcu(bond, slave, iter) { |
4020 | if (slave->queue_id == skb->queue_mapping) { | |
f5e2dc5d AN |
4021 | if (bond_slave_is_up(slave) && |
4022 | slave->link == BOND_LINK_UP) { | |
3900f290 | 4023 | bond_dev_queue_xmit(bond, skb, slave->dev); |
4024 | return 0; | |
4025 | } | |
4026 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
4027 | break; |
4028 | } | |
4029 | } | |
4030 | ||
3900f290 | 4031 | return 1; |
bb1d9123 AG |
4032 | } |
4033 | ||
374eeb5a | 4034 | |
f663dd9a | 4035 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 4036 | void *accel_priv, select_queue_fallback_t fallback) |
bb1d9123 | 4037 | { |
547942ca | 4038 | /* This helper function exists to help dev_pick_tx get the correct |
fd0e435b | 4039 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4040 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4041 | * way down to the bonding driver. | |
4042 | */ | |
fd0e435b PO |
4043 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4044 | ||
547942ca | 4045 | /* Save the original txq to restore before passing to the driver */ |
df4ab5b3 | 4046 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 4047 | |
fd0e435b | 4048 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4049 | do { |
fd0e435b | 4050 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4051 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4052 | } |
4053 | return txq; | |
bb1d9123 AG |
4054 | } |
4055 | ||
0693e88e | 4056 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4057 | { |
bb1d9123 AG |
4058 | struct bonding *bond = netdev_priv(dev); |
4059 | ||
d1e2e5cd VF |
4060 | if (bond_should_override_tx_queue(bond) && |
4061 | !bond_slave_override(bond, skb)) | |
4062 | return NETDEV_TX_OK; | |
00829823 | 4063 | |
01844098 | 4064 | switch (BOND_MODE(bond)) { |
00829823 SH |
4065 | case BOND_MODE_ROUNDROBIN: |
4066 | return bond_xmit_roundrobin(skb, dev); | |
4067 | case BOND_MODE_ACTIVEBACKUP: | |
4068 | return bond_xmit_activebackup(skb, dev); | |
ee637714 | 4069 | case BOND_MODE_8023AD: |
00829823 | 4070 | case BOND_MODE_XOR: |
ee637714 | 4071 | return bond_3ad_xor_xmit(skb, dev); |
00829823 SH |
4072 | case BOND_MODE_BROADCAST: |
4073 | return bond_xmit_broadcast(skb, dev); | |
00829823 | 4074 | case BOND_MODE_ALB: |
00829823 | 4075 | return bond_alb_xmit(skb, dev); |
f05b42ea MB |
4076 | case BOND_MODE_TLB: |
4077 | return bond_tlb_xmit(skb, dev); | |
00829823 SH |
4078 | default: |
4079 | /* Should never happen, mode already checked */ | |
76444f50 | 4080 | netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); |
00829823 | 4081 | WARN_ON_ONCE(1); |
31aa860e | 4082 | bond_tx_drop(dev, skb); |
00829823 SH |
4083 | return NETDEV_TX_OK; |
4084 | } | |
4085 | } | |
4086 | ||
0693e88e MM |
4087 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4088 | { | |
4089 | struct bonding *bond = netdev_priv(dev); | |
4090 | netdev_tx_t ret = NETDEV_TX_OK; | |
4091 | ||
547942ca | 4092 | /* If we risk deadlock from transmitting this in the |
0693e88e MM |
4093 | * netpoll path, tell netpoll to queue the frame for later tx |
4094 | */ | |
054bb880 | 4095 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
4096 | return NETDEV_TX_BUSY; |
4097 | ||
278b2083 | 4098 | rcu_read_lock(); |
0965a1f3 | 4099 | if (bond_has_slaves(bond)) |
0693e88e MM |
4100 | ret = __bond_start_xmit(skb, dev); |
4101 | else | |
31aa860e | 4102 | bond_tx_drop(dev, skb); |
278b2083 | 4103 | rcu_read_unlock(); |
0693e88e MM |
4104 | |
4105 | return ret; | |
4106 | } | |
00829823 | 4107 | |
d46b6349 PR |
4108 | static int bond_ethtool_get_link_ksettings(struct net_device *bond_dev, |
4109 | struct ethtool_link_ksettings *cmd) | |
bb5b052f AG |
4110 | { |
4111 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 4112 | unsigned long speed = 0; |
9caff1e7 | 4113 | struct list_head *iter; |
dec1e90e | 4114 | struct slave *slave; |
bb5b052f | 4115 | |
d46b6349 PR |
4116 | cmd->base.duplex = DUPLEX_UNKNOWN; |
4117 | cmd->base.port = PORT_OTHER; | |
bb5b052f | 4118 | |
8557cd74 | 4119 | /* Since bond_slave_can_tx returns false for all inactive or down slaves, we |
bb5b052f AG |
4120 | * do not need to check mode. Though link speed might not represent |
4121 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
4122 | * this is an accurate maximum. | |
4123 | */ | |
9caff1e7 | 4124 | bond_for_each_slave(bond, slave, iter) { |
8557cd74 | 4125 | if (bond_slave_can_tx(slave)) { |
bb5b052f AG |
4126 | if (slave->speed != SPEED_UNKNOWN) |
4127 | speed += slave->speed; | |
d46b6349 | 4128 | if (cmd->base.duplex == DUPLEX_UNKNOWN && |
bb5b052f | 4129 | slave->duplex != DUPLEX_UNKNOWN) |
d46b6349 | 4130 | cmd->base.duplex = slave->duplex; |
bb5b052f AG |
4131 | } |
4132 | } | |
d46b6349 | 4133 | cmd->base.speed = speed ? : SPEED_UNKNOWN; |
dec1e90e | 4134 | |
bb5b052f AG |
4135 | return 0; |
4136 | } | |
4137 | ||
217df670 | 4138 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 4139 | struct ethtool_drvinfo *drvinfo) |
217df670 | 4140 | { |
7826d43f JP |
4141 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
4142 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
4143 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
4144 | BOND_ABI_VERSION); | |
217df670 JV |
4145 | } |
4146 | ||
7282d491 | 4147 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4148 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac | 4149 | .get_link = ethtool_op_get_link, |
d46b6349 | 4150 | .get_link_ksettings = bond_ethtool_get_link_ksettings, |
8531c5ff AK |
4151 | }; |
4152 | ||
eb7cc59a | 4153 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4154 | .ndo_init = bond_init, |
9e71626c | 4155 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4156 | .ndo_open = bond_open, |
4157 | .ndo_stop = bond_close, | |
00829823 | 4158 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4159 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4160 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4161 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4162 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 4163 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 4164 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4165 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4166 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4167 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4168 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4169 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4170 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4171 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4172 | .ndo_poll_controller = bond_poll_controller, | |
4173 | #endif | |
9232ecca JP |
4174 | .ndo_add_slave = bond_enslave, |
4175 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4176 | .ndo_fix_features = bond_fix_features, |
fc8f40d8 | 4177 | .ndo_bridge_setlink = switchdev_port_bridge_setlink, |
85fdb956 | 4178 | .ndo_bridge_getlink = switchdev_port_bridge_getlink, |
54ba5a0b | 4179 | .ndo_bridge_dellink = switchdev_port_bridge_dellink, |
45d4122c SS |
4180 | .ndo_fdb_add = switchdev_port_fdb_add, |
4181 | .ndo_fdb_del = switchdev_port_fdb_del, | |
4182 | .ndo_fdb_dump = switchdev_port_fdb_dump, | |
4847f049 | 4183 | .ndo_features_check = passthru_features_check, |
eb7cc59a SH |
4184 | }; |
4185 | ||
b3f92b63 DG |
4186 | static const struct device_type bond_type = { |
4187 | .name = "bond", | |
4188 | }; | |
4189 | ||
9e2e61fb AW |
4190 | static void bond_destructor(struct net_device *bond_dev) |
4191 | { | |
4192 | struct bonding *bond = netdev_priv(bond_dev); | |
4193 | if (bond->wq) | |
4194 | destroy_workqueue(bond->wq); | |
4195 | free_netdev(bond_dev); | |
4196 | } | |
4197 | ||
0a2a78c4 | 4198 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4199 | { |
454d7c9b | 4200 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4201 | |
b7435628 | 4202 | spin_lock_init(&bond->mode_lock); |
fe30937b | 4203 | spin_lock_init(&bond->stats_lock); |
d2991f75 | 4204 | bond->params = bonding_defaults; |
1da177e4 LT |
4205 | |
4206 | /* Initialize pointers */ | |
1da177e4 | 4207 | bond->dev = bond_dev; |
1da177e4 LT |
4208 | |
4209 | /* Initialize the device entry points */ | |
181470fc | 4210 | ether_setup(bond_dev); |
31c05415 | 4211 | bond_dev->max_mtu = ETH_MAX_MTU; |
eb7cc59a | 4212 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4213 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 4214 | |
9e2e61fb | 4215 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4216 | |
b3f92b63 DG |
4217 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4218 | ||
1da177e4 | 4219 | /* Initialize the device options */ |
1098cee6 | 4220 | bond_dev->flags |= IFF_MASTER; |
1e6f20ca | 4221 | bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE; |
550fd08c | 4222 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4223 | |
547942ca | 4224 | /* don't acquire bond device's netif_tx_lock when transmitting */ |
1da177e4 LT |
4225 | bond_dev->features |= NETIF_F_LLTX; |
4226 | ||
4227 | /* By default, we declare the bond to be fully | |
4228 | * VLAN hardware accelerated capable. Special | |
4229 | * care is taken in the various xmit functions | |
4230 | * when there are slaves that are not hw accel | |
4231 | * capable | |
4232 | */ | |
1da177e4 | 4233 | |
f9399814 WC |
4234 | /* Don't allow bond devices to change network namespaces. */ |
4235 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
4236 | ||
b2a103e6 | 4237 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4238 | NETIF_F_HW_VLAN_CTAG_TX | |
4239 | NETIF_F_HW_VLAN_CTAG_RX | | |
4240 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4241 | |
24f87d4c | 4242 | bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL; |
b2a103e6 | 4243 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4244 | } |
4245 | ||
547942ca NA |
4246 | /* Destroy a bonding device. |
4247 | * Must be under rtnl_lock when this function is called. | |
4248 | */ | |
c67dfb29 | 4249 | static void bond_uninit(struct net_device *bond_dev) |
a434e43f | 4250 | { |
454d7c9b | 4251 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
4252 | struct list_head *iter; |
4253 | struct slave *slave; | |
ee637714 | 4254 | struct bond_up_slave *arr; |
a434e43f | 4255 | |
f6dc31a8 WC |
4256 | bond_netpoll_cleanup(bond_dev); |
4257 | ||
c67dfb29 | 4258 | /* Release the bonded slaves */ |
544a028e | 4259 | bond_for_each_slave(bond, slave, iter) |
dec1e90e | 4260 | __bond_release_one(bond_dev, slave->dev, true); |
76444f50 | 4261 | netdev_info(bond_dev, "Released all slaves\n"); |
c67dfb29 | 4262 | |
ee637714 MB |
4263 | arr = rtnl_dereference(bond->slave_arr); |
4264 | if (arr) { | |
4265 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
4266 | kfree_rcu(arr, rcu); | |
4267 | } | |
4268 | ||
a434e43f JV |
4269 | list_del(&bond->bond_list); |
4270 | ||
f073c7ca | 4271 | bond_debug_unregister(bond); |
a434e43f JV |
4272 | } |
4273 | ||
1da177e4 LT |
4274 | /*------------------------- Module initialization ---------------------------*/ |
4275 | ||
1da177e4 LT |
4276 | static int bond_check_params(struct bond_params *params) |
4277 | { | |
5a5c5fd4 | 4278 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 4279 | struct bond_opt_value newval; |
4280 | const struct bond_opt_value *valptr; | |
72ccc471 | 4281 | int arp_all_targets_value = 0; |
6791e466 | 4282 | u16 ad_actor_sys_prio = 0; |
d22a5fc0 | 4283 | u16 ad_user_port_key = 0; |
72ccc471 | 4284 | __be32 arp_target[BOND_MAX_ARP_TARGETS] = { 0 }; |
dc9c4d0f MB |
4285 | int arp_ip_count; |
4286 | int bond_mode = BOND_MODE_ROUNDROBIN; | |
4287 | int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; | |
4288 | int lacp_fast = 0; | |
8b426dc5 | 4289 | int tlb_dynamic_lb = 0; |
f5b2b966 | 4290 | |
547942ca | 4291 | /* Convert string parameters. */ |
1da177e4 | 4292 | if (mode) { |
2b3798d5 NA |
4293 | bond_opt_initstr(&newval, mode); |
4294 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
4295 | if (!valptr) { | |
4296 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
4297 | return -EINVAL; |
4298 | } | |
2b3798d5 | 4299 | bond_mode = valptr->value; |
1da177e4 LT |
4300 | } |
4301 | ||
169a3e66 JV |
4302 | if (xmit_hash_policy) { |
4303 | if ((bond_mode != BOND_MODE_XOR) && | |
f05b42ea MB |
4304 | (bond_mode != BOND_MODE_8023AD) && |
4305 | (bond_mode != BOND_MODE_TLB)) { | |
a4aee5c8 | 4306 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 4307 | bond_mode_name(bond_mode)); |
169a3e66 | 4308 | } else { |
a4b32ce7 NA |
4309 | bond_opt_initstr(&newval, xmit_hash_policy); |
4310 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
4311 | &newval); | |
4312 | if (!valptr) { | |
a4aee5c8 | 4313 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
4314 | xmit_hash_policy); |
4315 | return -EINVAL; | |
4316 | } | |
a4b32ce7 | 4317 | xmit_hashtype = valptr->value; |
169a3e66 JV |
4318 | } |
4319 | } | |
4320 | ||
1da177e4 LT |
4321 | if (lacp_rate) { |
4322 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4323 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4324 | bond_mode_name(bond_mode)); | |
1da177e4 | 4325 | } else { |
d3131de7 NA |
4326 | bond_opt_initstr(&newval, lacp_rate); |
4327 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
4328 | &newval); | |
4329 | if (!valptr) { | |
a4aee5c8 | 4330 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 4331 | lacp_rate); |
1da177e4 LT |
4332 | return -EINVAL; |
4333 | } | |
d3131de7 | 4334 | lacp_fast = valptr->value; |
1da177e4 LT |
4335 | } |
4336 | } | |
4337 | ||
fd989c83 | 4338 | if (ad_select) { |
548d28bd | 4339 | bond_opt_initstr(&newval, ad_select); |
9e5f5eeb NA |
4340 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), |
4341 | &newval); | |
4342 | if (!valptr) { | |
4343 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
4344 | return -EINVAL; |
4345 | } | |
9e5f5eeb NA |
4346 | params->ad_select = valptr->value; |
4347 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 4348 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4349 | } else { |
4350 | params->ad_select = BOND_AD_STABLE; | |
4351 | } | |
4352 | ||
f5841306 | 4353 | if (max_bonds < 0) { |
91565ebb JP |
4354 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4355 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4356 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4357 | } | |
4358 | ||
4359 | if (miimon < 0) { | |
91565ebb JP |
4360 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4361 | miimon, INT_MAX); | |
b98d9c66 | 4362 | miimon = 0; |
1da177e4 LT |
4363 | } |
4364 | ||
4365 | if (updelay < 0) { | |
91565ebb JP |
4366 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4367 | updelay, INT_MAX); | |
1da177e4 LT |
4368 | updelay = 0; |
4369 | } | |
4370 | ||
4371 | if (downdelay < 0) { | |
91565ebb JP |
4372 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4373 | downdelay, INT_MAX); | |
1da177e4 LT |
4374 | downdelay = 0; |
4375 | } | |
4376 | ||
4377 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
91565ebb JP |
4378 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4379 | use_carrier); | |
1da177e4 LT |
4380 | use_carrier = 1; |
4381 | } | |
4382 | ||
ad246c99 | 4383 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
4384 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
4385 | num_peer_notif); | |
ad246c99 BH |
4386 | num_peer_notif = 1; |
4387 | } | |
4388 | ||
834db4bc | 4389 | /* reset values for 802.3ad/TLB/ALB */ |
267bed77 | 4390 | if (!bond_mode_uses_arp(bond_mode)) { |
1da177e4 | 4391 | if (!miimon) { |
91565ebb JP |
4392 | 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"); |
4393 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 4394 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4395 | } |
4396 | } | |
4397 | ||
bb1d9123 | 4398 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
4399 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
4400 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
4401 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
4402 | } | |
4403 | ||
ebd8e497 | 4404 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
4405 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
4406 | all_slaves_active); | |
ebd8e497 AG |
4407 | all_slaves_active = 0; |
4408 | } | |
4409 | ||
c2952c31 | 4410 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
4411 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
4412 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
4413 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
4414 | } | |
4415 | ||
aa59d851 NA |
4416 | bond_opt_initval(&newval, packets_per_slave); |
4417 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
4418 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
4419 | packets_per_slave, USHRT_MAX); | |
4420 | packets_per_slave = 1; | |
4421 | } | |
4422 | ||
1da177e4 | 4423 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4424 | 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", |
4425 | updelay); | |
1da177e4 LT |
4426 | } |
4427 | ||
4428 | if (!miimon) { | |
4429 | if (updelay || downdelay) { | |
4430 | /* just warn the user the up/down delay will have | |
4431 | * no effect since miimon is zero... | |
4432 | */ | |
91565ebb JP |
4433 | 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", |
4434 | updelay, downdelay); | |
1da177e4 LT |
4435 | } |
4436 | } else { | |
4437 | /* don't allow arp monitoring */ | |
4438 | if (arp_interval) { | |
91565ebb JP |
4439 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4440 | miimon, arp_interval); | |
1da177e4 LT |
4441 | arp_interval = 0; |
4442 | } | |
4443 | ||
4444 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
4445 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4446 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
4447 | } |
4448 | ||
4449 | updelay /= miimon; | |
4450 | ||
4451 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
4452 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4453 | downdelay, miimon, | |
4454 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4455 | } |
4456 | ||
4457 | downdelay /= miimon; | |
4458 | } | |
4459 | ||
4460 | if (arp_interval < 0) { | |
91565ebb JP |
4461 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4462 | arp_interval, INT_MAX); | |
7bdb04ed | 4463 | arp_interval = 0; |
1da177e4 LT |
4464 | } |
4465 | ||
5a5c5fd4 | 4466 | for (arp_ip_count = 0, i = 0; |
4467 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
89015c18 | 4468 | __be32 ip; |
547942ca NA |
4469 | |
4470 | /* not a complete check, but good enough to catch mistakes */ | |
89015c18 | 4471 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || |
2807a9fe | 4472 | !bond_is_ip_target_ok(ip)) { |
91565ebb JP |
4473 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4474 | arp_ip_target[i]); | |
1da177e4 LT |
4475 | arp_interval = 0; |
4476 | } else { | |
0afee4e8 VF |
4477 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4478 | arp_target[arp_ip_count++] = ip; | |
4479 | else | |
91565ebb JP |
4480 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
4481 | &ip); | |
1da177e4 LT |
4482 | } |
4483 | } | |
4484 | ||
4485 | if (arp_interval && !arp_ip_count) { | |
4486 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
4487 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4488 | arp_interval); | |
1da177e4 LT |
4489 | arp_interval = 0; |
4490 | } | |
4491 | ||
f5b2b966 | 4492 | if (arp_validate) { |
f5b2b966 | 4493 | if (!arp_interval) { |
a4aee5c8 | 4494 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4495 | return -EINVAL; |
4496 | } | |
4497 | ||
16228881 NA |
4498 | bond_opt_initstr(&newval, arp_validate); |
4499 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
4500 | &newval); | |
4501 | if (!valptr) { | |
a4aee5c8 | 4502 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 4503 | arp_validate); |
f5b2b966 JV |
4504 | return -EINVAL; |
4505 | } | |
16228881 NA |
4506 | arp_validate_value = valptr->value; |
4507 | } else { | |
f5b2b966 | 4508 | arp_validate_value = 0; |
16228881 | 4509 | } |
f5b2b966 | 4510 | |
8599b52e | 4511 | if (arp_all_targets) { |
edf36b24 NA |
4512 | bond_opt_initstr(&newval, arp_all_targets); |
4513 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
4514 | &newval); | |
4515 | if (!valptr) { | |
8599b52e VF |
4516 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
4517 | arp_all_targets); | |
4518 | arp_all_targets_value = 0; | |
edf36b24 NA |
4519 | } else { |
4520 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
4521 | } |
4522 | } | |
4523 | ||
1da177e4 | 4524 | if (miimon) { |
a4aee5c8 | 4525 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4526 | } else if (arp_interval) { |
16228881 NA |
4527 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
4528 | arp_validate_value); | |
a4aee5c8 | 4529 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 4530 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
4531 | |
4532 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 4533 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 4534 | |
90194264 | 4535 | pr_cont("\n"); |
1da177e4 | 4536 | |
b8a9787e | 4537 | } else if (max_bonds) { |
1da177e4 LT |
4538 | /* miimon and arp_interval not set, we need one so things |
4539 | * work as expected, see bonding.txt for details | |
4540 | */ | |
90194264 | 4541 | 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 |
4542 | } |
4543 | ||
ec0865a9 | 4544 | if (primary && !bond_mode_uses_primary(bond_mode)) { |
1da177e4 LT |
4545 | /* currently, using a primary only makes sense |
4546 | * in active backup, TLB or ALB modes | |
4547 | */ | |
91565ebb JP |
4548 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
4549 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4550 | primary = NULL; |
4551 | } | |
4552 | ||
a549952a | 4553 | if (primary && primary_reselect) { |
388d3a6d NA |
4554 | bond_opt_initstr(&newval, primary_reselect); |
4555 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
4556 | &newval); | |
4557 | if (!valptr) { | |
a4aee5c8 | 4558 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 4559 | primary_reselect); |
a549952a JP |
4560 | return -EINVAL; |
4561 | } | |
388d3a6d | 4562 | primary_reselect_value = valptr->value; |
a549952a JP |
4563 | } else { |
4564 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4565 | } | |
4566 | ||
3915c1e8 | 4567 | if (fail_over_mac) { |
1df6b6aa NA |
4568 | bond_opt_initstr(&newval, fail_over_mac); |
4569 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
4570 | &newval); | |
4571 | if (!valptr) { | |
a4aee5c8 | 4572 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 4573 | fail_over_mac); |
3915c1e8 JV |
4574 | return -EINVAL; |
4575 | } | |
1df6b6aa | 4576 | fail_over_mac_value = valptr->value; |
3915c1e8 | 4577 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 4578 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
4579 | } else { |
4580 | fail_over_mac_value = BOND_FOM_NONE; | |
4581 | } | |
dd957c57 | 4582 | |
6791e466 MB |
4583 | bond_opt_initstr(&newval, "default"); |
4584 | valptr = bond_opt_parse( | |
4585 | bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO), | |
4586 | &newval); | |
4587 | if (!valptr) { | |
4588 | pr_err("Error: No ad_actor_sys_prio default value"); | |
4589 | return -EINVAL; | |
4590 | } | |
4591 | ad_actor_sys_prio = valptr->value; | |
4592 | ||
d22a5fc0 MB |
4593 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY), |
4594 | &newval); | |
4595 | if (!valptr) { | |
4596 | pr_err("Error: No ad_user_port_key default value"); | |
4597 | return -EINVAL; | |
4598 | } | |
4599 | ad_user_port_key = valptr->value; | |
4600 | ||
8b426dc5 MB |
4601 | if (bond_mode == BOND_MODE_TLB) { |
4602 | bond_opt_initstr(&newval, "default"); | |
4603 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_TLB_DYNAMIC_LB), | |
4604 | &newval); | |
4605 | if (!valptr) { | |
4606 | pr_err("Error: No tlb_dynamic_lb default value"); | |
4607 | return -EINVAL; | |
4608 | } | |
4609 | tlb_dynamic_lb = valptr->value; | |
4610 | } | |
4611 | ||
3a7129e5 | 4612 | if (lp_interval == 0) { |
91565ebb JP |
4613 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
4614 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 4615 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
4616 | } | |
4617 | ||
1da177e4 LT |
4618 | /* fill params struct with the proper values */ |
4619 | params->mode = bond_mode; | |
169a3e66 | 4620 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4621 | params->miimon = miimon; |
ad246c99 | 4622 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4623 | params->arp_interval = arp_interval; |
f5b2b966 | 4624 | params->arp_validate = arp_validate_value; |
8599b52e | 4625 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4626 | params->updelay = updelay; |
4627 | params->downdelay = downdelay; | |
4628 | params->use_carrier = use_carrier; | |
4629 | params->lacp_fast = lacp_fast; | |
4630 | params->primary[0] = 0; | |
a549952a | 4631 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4632 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4633 | params->tx_queues = tx_queues; |
ebd8e497 | 4634 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4635 | params->resend_igmp = resend_igmp; |
655f8919 | 4636 | params->min_links = min_links; |
3a7129e5 | 4637 | params->lp_interval = lp_interval; |
809fa972 | 4638 | params->packets_per_slave = packets_per_slave; |
8b426dc5 | 4639 | params->tlb_dynamic_lb = tlb_dynamic_lb; |
6791e466 | 4640 | params->ad_actor_sys_prio = ad_actor_sys_prio; |
74514957 | 4641 | eth_zero_addr(params->ad_actor_system); |
d22a5fc0 | 4642 | params->ad_user_port_key = ad_user_port_key; |
809fa972 HFS |
4643 | if (packets_per_slave > 0) { |
4644 | params->reciprocal_packets_per_slave = | |
4645 | reciprocal_value(packets_per_slave); | |
4646 | } else { | |
4647 | /* reciprocal_packets_per_slave is unused if | |
4648 | * packets_per_slave is 0 or 1, just initialize it | |
4649 | */ | |
4650 | params->reciprocal_packets_per_slave = | |
4651 | (struct reciprocal_value) { 0 }; | |
4652 | } | |
4653 | ||
1da177e4 LT |
4654 | if (primary) { |
4655 | strncpy(params->primary, primary, IFNAMSIZ); | |
4656 | params->primary[IFNAMSIZ - 1] = 0; | |
4657 | } | |
4658 | ||
4659 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4660 | ||
4661 | return 0; | |
4662 | } | |
4663 | ||
547942ca | 4664 | /* Called from registration process */ |
181470fc SH |
4665 | static int bond_init(struct net_device *bond_dev) |
4666 | { | |
4667 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4668 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc | 4669 | |
76444f50 | 4670 | netdev_dbg(bond_dev, "Begin bond_init\n"); |
181470fc | 4671 | |
f9f225eb | 4672 | bond->wq = alloc_ordered_workqueue(bond_dev->name, WQ_MEM_RECLAIM); |
181470fc SH |
4673 | if (!bond->wq) |
4674 | return -ENOMEM; | |
4675 | ||
d3fff6c4 | 4676 | netdev_lockdep_set_classes(bond_dev); |
181470fc | 4677 | |
ec87fd3b | 4678 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4679 | |
6151b3d4 | 4680 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4681 | |
f073c7ca TI |
4682 | bond_debug_register(bond); |
4683 | ||
409cc1f8 JP |
4684 | /* Ensure valid dev_addr */ |
4685 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4686 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4687 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4688 | |
181470fc SH |
4689 | return 0; |
4690 | } | |
4691 | ||
0a2a78c4 | 4692 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4693 | { |
efacb309 | 4694 | return tx_queues; |
d5da4510 JP |
4695 | } |
4696 | ||
dfe60397 | 4697 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4698 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4699 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4700 | * set up our sysfs entries. | |
4701 | */ | |
ec87fd3b | 4702 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4703 | { |
4704 | struct net_device *bond_dev; | |
e913fb27 P |
4705 | struct bonding *bond; |
4706 | struct alb_bond_info *bond_info; | |
dfe60397 MW |
4707 | int res; |
4708 | ||
4709 | rtnl_lock(); | |
027ea041 | 4710 | |
1c5cae81 | 4711 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
c835a677 | 4712 | name ? name : "bond%d", NET_NAME_UNKNOWN, |
1c5cae81 | 4713 | bond_setup, tx_queues); |
dfe60397 | 4714 | if (!bond_dev) { |
a4aee5c8 | 4715 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4716 | rtnl_unlock(); |
4717 | return -ENOMEM; | |
dfe60397 MW |
4718 | } |
4719 | ||
e913fb27 P |
4720 | /* |
4721 | * Initialize rx_hashtbl_used_head to RLB_NULL_INDEX. | |
4722 | * It is set to 0 by default which is wrong. | |
4723 | */ | |
4724 | bond = netdev_priv(bond_dev); | |
4725 | bond_info = &(BOND_ALB_INFO(bond)); | |
4726 | bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX; | |
4727 | ||
ec87fd3b | 4728 | dev_net_set(bond_dev, net); |
88ead977 EB |
4729 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4730 | ||
dfe60397 | 4731 | res = register_netdevice(bond_dev); |
0daa2303 | 4732 | |
e826eafa PO |
4733 | netif_carrier_off(bond_dev); |
4734 | ||
4493b81b MB |
4735 | bond_work_init_all(bond); |
4736 | ||
7e083840 | 4737 | rtnl_unlock(); |
9e2e61fb AW |
4738 | if (res < 0) |
4739 | bond_destructor(bond_dev); | |
30c15ba9 | 4740 | return res; |
dfe60397 MW |
4741 | } |
4742 | ||
2c8c1e72 | 4743 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4744 | { |
15449745 | 4745 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4746 | |
4747 | bn->net = net; | |
4748 | INIT_LIST_HEAD(&bn->dev_list); | |
4749 | ||
ec87fd3b | 4750 | bond_create_proc_dir(bn); |
4c22400a | 4751 | bond_create_sysfs(bn); |
87a7b84b | 4752 | |
15449745 | 4753 | return 0; |
ec87fd3b EB |
4754 | } |
4755 | ||
2c8c1e72 | 4756 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4757 | { |
15449745 | 4758 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4759 | struct bonding *bond, *tmp_bond; |
4760 | LIST_HEAD(list); | |
ec87fd3b | 4761 | |
4c22400a | 4762 | bond_destroy_sysfs(bn); |
69b0216a | 4763 | |
4764 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4765 | rtnl_lock(); | |
4766 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4767 | unregister_netdevice_queue(bond->dev, &list); | |
4768 | unregister_netdevice_many(&list); | |
4769 | rtnl_unlock(); | |
23fa5c2c VF |
4770 | |
4771 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
4772 | } |
4773 | ||
4774 | static struct pernet_operations bond_net_ops = { | |
4775 | .init = bond_net_init, | |
4776 | .exit = bond_net_exit, | |
15449745 EB |
4777 | .id = &bond_net_id, |
4778 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4779 | }; |
4780 | ||
1da177e4 LT |
4781 | static int __init bonding_init(void) |
4782 | { | |
1da177e4 LT |
4783 | int i; |
4784 | int res; | |
4785 | ||
bd33acc3 | 4786 | pr_info("%s", bond_version); |
1da177e4 | 4787 | |
dfe60397 | 4788 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4789 | if (res) |
dfe60397 | 4790 | goto out; |
1da177e4 | 4791 | |
15449745 | 4792 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4793 | if (res) |
4794 | goto out; | |
027ea041 | 4795 | |
0a2a78c4 | 4796 | res = bond_netlink_init(); |
88ead977 | 4797 | if (res) |
6639104b | 4798 | goto err_link; |
88ead977 | 4799 | |
f073c7ca TI |
4800 | bond_create_debugfs(); |
4801 | ||
1da177e4 | 4802 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4803 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4804 | if (res) |
4805 | goto err; | |
1da177e4 LT |
4806 | } |
4807 | ||
1da177e4 | 4808 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4809 | out: |
1da177e4 | 4810 | return res; |
88ead977 | 4811 | err: |
db298686 | 4812 | bond_destroy_debugfs(); |
0a2a78c4 | 4813 | bond_netlink_fini(); |
6639104b | 4814 | err_link: |
15449745 | 4815 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4816 | goto out; |
dfe60397 | 4817 | |
1da177e4 LT |
4818 | } |
4819 | ||
4820 | static void __exit bonding_exit(void) | |
4821 | { | |
4822 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4823 | ||
f073c7ca | 4824 | bond_destroy_debugfs(); |
ae68c398 | 4825 | |
0a2a78c4 | 4826 | bond_netlink_fini(); |
ffcdedb6 | 4827 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4828 | |
4829 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
547942ca | 4830 | /* Make sure we don't have an imbalance on our netpoll blocking */ |
fb4fa76a | 4831 | WARN_ON(atomic_read(&netpoll_block_tx)); |
e843fa50 | 4832 | #endif |
1da177e4 LT |
4833 | } |
4834 | ||
4835 | module_init(bonding_init); | |
4836 | module_exit(bonding_exit); | |
4837 | MODULE_LICENSE("GPL"); | |
4838 | MODULE_VERSION(DRV_VERSION); | |
4839 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4840 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |