]>
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> |
df98be06 MC |
44 | #include <linux/icmp.h> |
45 | #include <linux/icmpv6.h> | |
169a3e66 JV |
46 | #include <linux/tcp.h> |
47 | #include <linux/udp.h> | |
1da177e4 LT |
48 | #include <linux/slab.h> |
49 | #include <linux/string.h> | |
50 | #include <linux/init.h> | |
51 | #include <linux/timer.h> | |
52 | #include <linux/socket.h> | |
53 | #include <linux/ctype.h> | |
54 | #include <linux/inet.h> | |
55 | #include <linux/bitops.h> | |
3d632c3f | 56 | #include <linux/io.h> |
1da177e4 | 57 | #include <asm/dma.h> |
3d632c3f | 58 | #include <linux/uaccess.h> |
1da177e4 LT |
59 | #include <linux/errno.h> |
60 | #include <linux/netdevice.h> | |
61 | #include <linux/inetdevice.h> | |
a816c7c7 | 62 | #include <linux/igmp.h> |
1da177e4 LT |
63 | #include <linux/etherdevice.h> |
64 | #include <linux/skbuff.h> | |
65 | #include <net/sock.h> | |
66 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
67 | #include <linux/smp.h> |
68 | #include <linux/if_ether.h> | |
69 | #include <net/arp.h> | |
70 | #include <linux/mii.h> | |
71 | #include <linux/ethtool.h> | |
72 | #include <linux/if_vlan.h> | |
73 | #include <linux/if_bonding.h> | |
b63bb739 | 74 | #include <linux/jiffies.h> |
e843fa50 | 75 | #include <linux/preempt.h> |
c3ade5ca | 76 | #include <net/route.h> |
457c4cbc | 77 | #include <net/net_namespace.h> |
ec87fd3b | 78 | #include <net/netns/generic.h> |
5ee31c68 | 79 | #include <net/pkt_sched.h> |
278b2083 | 80 | #include <linux/rculist.h> |
1bd758eb | 81 | #include <net/flow_dissector.h> |
1ef8019b DM |
82 | #include <net/bonding.h> |
83 | #include <net/bond_3ad.h> | |
84 | #include <net/bond_alb.h> | |
1da177e4 | 85 | |
73b5a6f2 MB |
86 | #include "bonding_priv.h" |
87 | ||
1da177e4 LT |
88 | /*---------------------------- Module parameters ----------------------------*/ |
89 | ||
90 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
1da177e4 LT |
91 | |
92 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 93 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 94 | static int num_peer_notif = 1; |
b98d9c66 | 95 | static int miimon; |
3d632c3f SH |
96 | static int updelay; |
97 | static int downdelay; | |
1da177e4 | 98 | static int use_carrier = 1; |
3d632c3f SH |
99 | static char *mode; |
100 | static char *primary; | |
a549952a | 101 | static char *primary_reselect; |
3d632c3f | 102 | static char *lacp_rate; |
655f8919 | 103 | static int min_links; |
3d632c3f SH |
104 | static char *ad_select; |
105 | static char *xmit_hash_policy; | |
7bdb04ed | 106 | static int arp_interval; |
3d632c3f SH |
107 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
108 | static char *arp_validate; | |
8599b52e | 109 | static char *arp_all_targets; |
3d632c3f | 110 | static char *fail_over_mac; |
b07ea07b | 111 | static int all_slaves_active; |
d2991f75 | 112 | static struct bond_params bonding_defaults; |
c2952c31 | 113 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
73958329 | 114 | static int packets_per_slave = 1; |
3a7129e5 | 115 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
116 | |
117 | module_param(max_bonds, int, 0); | |
118 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
119 | module_param(tx_queues, int, 0); |
120 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 121 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
122 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
123 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 124 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
125 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
126 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
127 | module_param(miimon, int, 0); |
128 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
129 | module_param(updelay, int, 0); | |
130 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
131 | module_param(downdelay, int, 0); | |
2ac47660 MW |
132 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
133 | "in milliseconds"); | |
1da177e4 | 134 | module_param(use_carrier, int, 0); |
2ac47660 | 135 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
b3c898e2 | 136 | "0 for off, 1 for on (default)"); |
1da177e4 | 137 | module_param(mode, charp, 0); |
90e62474 | 138 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
139 | "1 for active-backup, 2 for balance-xor, " |
140 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
141 | "6 for balance-alb"); | |
1da177e4 LT |
142 | module_param(primary, charp, 0); |
143 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
144 | module_param(primary_reselect, charp, 0); |
145 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
146 | "once it comes up; " | |
147 | "0 for always (default), " | |
148 | "1 for only if speed of primary is " | |
149 | "better, " | |
150 | "2 for only on active slave " | |
151 | "failure"); | |
1da177e4 | 152 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
153 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
154 | "0 for slow, 1 for fast"); | |
fd989c83 | 155 | module_param(ad_select, charp, 0); |
0d039f33 | 156 | MODULE_PARM_DESC(ad_select, "802.3ad aggregation selection logic; " |
90e62474 AG |
157 | "0 for stable (default), 1 for bandwidth, " |
158 | "2 for count"); | |
655f8919 | 159 | module_param(min_links, int, 0); |
160 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
161 | ||
169a3e66 | 162 | module_param(xmit_hash_policy, charp, 0); |
e79c1055 | 163 | MODULE_PARM_DESC(xmit_hash_policy, "balance-alb, balance-tlb, balance-xor, 802.3ad hashing method; " |
90e62474 | 164 | "0 for layer 2 (default), 1 for layer 3+4, " |
32819dc1 NA |
165 | "2 for layer 2+3, 3 for encap layer 2+3, " |
166 | "4 for encap layer 3+4"); | |
1da177e4 LT |
167 | module_param(arp_interval, int, 0); |
168 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
169 | module_param_array(arp_ip_target, charp, NULL, 0); | |
170 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 171 | module_param(arp_validate, charp, 0); |
90e62474 AG |
172 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
173 | "0 for none (default), 1 for active, " | |
174 | "2 for backup, 3 for all"); | |
8599b52e VF |
175 | module_param(arp_all_targets, charp, 0); |
176 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 177 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
178 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
179 | "the same MAC; 0 for none (default), " | |
180 | "1 for active, 2 for follow"); | |
ebd8e497 | 181 | module_param(all_slaves_active, int, 0); |
37b7021d | 182 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface " |
90e62474 | 183 | "by setting active flag for all slaves; " |
ebd8e497 | 184 | "0 for never (default), 1 for always."); |
c2952c31 | 185 | module_param(resend_igmp, int, 0); |
90e62474 AG |
186 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
187 | "link failure"); | |
73958329 NA |
188 | module_param(packets_per_slave, int, 0); |
189 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
190 | "mode; 0 for a random slave, 1 packet per " | |
191 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 192 | module_param(lp_interval, uint, 0); |
193 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
194 | "the bonding driver sends learning packets to " | |
195 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
196 | |
197 | /*----------------------------- Global variables ----------------------------*/ | |
198 | ||
e843fa50 | 199 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 200 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
201 | #endif |
202 | ||
c7d03a00 | 203 | unsigned int bond_net_id __read_mostly; |
1da177e4 | 204 | |
58deb77c MC |
205 | static const struct flow_dissector_key flow_keys_bonding_keys[] = { |
206 | { | |
207 | .key_id = FLOW_DISSECTOR_KEY_CONTROL, | |
208 | .offset = offsetof(struct flow_keys, control), | |
209 | }, | |
210 | { | |
211 | .key_id = FLOW_DISSECTOR_KEY_BASIC, | |
212 | .offset = offsetof(struct flow_keys, basic), | |
213 | }, | |
214 | { | |
215 | .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS, | |
216 | .offset = offsetof(struct flow_keys, addrs.v4addrs), | |
217 | }, | |
218 | { | |
219 | .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS, | |
220 | .offset = offsetof(struct flow_keys, addrs.v6addrs), | |
221 | }, | |
222 | { | |
223 | .key_id = FLOW_DISSECTOR_KEY_TIPC, | |
224 | .offset = offsetof(struct flow_keys, addrs.tipckey), | |
225 | }, | |
226 | { | |
227 | .key_id = FLOW_DISSECTOR_KEY_PORTS, | |
228 | .offset = offsetof(struct flow_keys, ports), | |
229 | }, | |
230 | { | |
231 | .key_id = FLOW_DISSECTOR_KEY_ICMP, | |
232 | .offset = offsetof(struct flow_keys, icmp), | |
233 | }, | |
234 | { | |
235 | .key_id = FLOW_DISSECTOR_KEY_VLAN, | |
236 | .offset = offsetof(struct flow_keys, vlan), | |
237 | }, | |
238 | { | |
239 | .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL, | |
240 | .offset = offsetof(struct flow_keys, tags), | |
241 | }, | |
242 | { | |
243 | .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID, | |
244 | .offset = offsetof(struct flow_keys, keyid), | |
245 | }, | |
246 | }; | |
247 | ||
248 | static struct flow_dissector flow_keys_bonding __read_mostly; | |
249 | ||
1da177e4 LT |
250 | /*-------------------------- Forward declarations ---------------------------*/ |
251 | ||
181470fc | 252 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 253 | static void bond_uninit(struct net_device *bond_dev); |
bc1f4470 | 254 | static void bond_get_stats(struct net_device *bond_dev, |
255 | struct rtnl_link_stats64 *stats); | |
ee637714 | 256 | static void bond_slave_arr_handler(struct work_struct *work); |
21a75f09 JV |
257 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, |
258 | int mod); | |
d4859d74 | 259 | static void bond_netdev_notify_work(struct work_struct *work); |
1da177e4 LT |
260 | |
261 | /*---------------------------- General routines -----------------------------*/ | |
262 | ||
bd33acc3 | 263 | const char *bond_mode_name(int mode) |
1da177e4 | 264 | { |
77afc92b HE |
265 | static const char *names[] = { |
266 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
267 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
268 | [BOND_MODE_XOR] = "load balancing (xor)", | |
269 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 270 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
271 | [BOND_MODE_TLB] = "transmit load balancing", |
272 | [BOND_MODE_ALB] = "adaptive load balancing", | |
273 | }; | |
274 | ||
f5280948 | 275 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 276 | return "unknown"; |
77afc92b HE |
277 | |
278 | return names[mode]; | |
1da177e4 LT |
279 | } |
280 | ||
281 | /*---------------------------------- VLAN -----------------------------------*/ | |
282 | ||
1da177e4 LT |
283 | /** |
284 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 285 | * |
1da177e4 LT |
286 | * @bond: bond device that got this skb for tx. |
287 | * @skb: hw accel VLAN tagged skb to transmit | |
288 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 289 | */ |
d316dedd | 290 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 291 | struct net_device *slave_dev) |
1da177e4 | 292 | { |
8387451e | 293 | skb->dev = slave_dev; |
374eeb5a | 294 | |
5ee31c68 | 295 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 | 296 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
ae35c6f7 | 297 | skb_set_queue_mapping(skb, qdisc_skb_cb(skb)->slave_dev_queue_mapping); |
374eeb5a | 298 | |
e15c3c22 | 299 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 300 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 301 | else |
f6dc31a8 | 302 | dev_queue_xmit(skb); |
1da177e4 LT |
303 | } |
304 | ||
547942ca | 305 | /* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
cc0e4070 | 306 | * We don't protect the slave list iteration with a lock because: |
1da177e4 LT |
307 | * a. This operation is performed in IOCTL context, |
308 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
309 | * c. Holding a lock with BH disabled while directly calling a base driver | |
310 | * entry point is generally a BAD idea. | |
3d632c3f | 311 | * |
1da177e4 LT |
312 | * The design of synchronization/protection for this operation in the 8021q |
313 | * module is good for one or more VLAN devices over a single physical device | |
314 | * and cannot be extended for a teaming solution like bonding, so there is a | |
315 | * potential race condition here where a net device from the vlan group might | |
316 | * be referenced (either by a base driver or the 8021q code) while it is being | |
317 | * removed from the system. However, it turns out we're not making matters | |
318 | * worse, and if it works for regular VLAN usage it will work here too. | |
319 | */ | |
320 | ||
1da177e4 LT |
321 | /** |
322 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
323 | * @bond_dev: bonding net device that got called | |
324 | * @vid: vlan id being added | |
325 | */ | |
80d5c368 PM |
326 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
327 | __be16 proto, u16 vid) | |
1da177e4 | 328 | { |
454d7c9b | 329 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 330 | struct slave *slave, *rollback_slave; |
9caff1e7 | 331 | struct list_head *iter; |
dec1e90e | 332 | int res; |
1da177e4 | 333 | |
9caff1e7 | 334 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 335 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
336 | if (res) |
337 | goto unwind; | |
1da177e4 LT |
338 | } |
339 | ||
8e586137 | 340 | return 0; |
87002b03 JP |
341 | |
342 | unwind: | |
81f23b13 | 343 | /* unwind to the slave that failed */ |
9caff1e7 | 344 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
345 | if (rollback_slave == slave) |
346 | break; | |
347 | ||
348 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
349 | } | |
87002b03 JP |
350 | |
351 | return res; | |
1da177e4 LT |
352 | } |
353 | ||
354 | /** | |
355 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
356 | * @bond_dev: bonding net device that got called | |
357 | * @vid: vlan id being removed | |
358 | */ | |
80d5c368 PM |
359 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
360 | __be16 proto, u16 vid) | |
1da177e4 | 361 | { |
454d7c9b | 362 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 363 | struct list_head *iter; |
1da177e4 | 364 | struct slave *slave; |
1da177e4 | 365 | |
9caff1e7 | 366 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 367 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 368 | |
e868b0c9 VF |
369 | if (bond_is_lb(bond)) |
370 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
371 | |
372 | return 0; | |
1da177e4 LT |
373 | } |
374 | ||
1da177e4 LT |
375 | /*------------------------------- Link status -------------------------------*/ |
376 | ||
547942ca | 377 | /* Set the carrier state for the master according to the state of its |
ff59c456 JV |
378 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, |
379 | * do special 802.3ad magic. | |
380 | * | |
381 | * Returns zero if carrier state does not change, nonzero if it does. | |
382 | */ | |
2477bc9a | 383 | int bond_set_carrier(struct bonding *bond) |
ff59c456 | 384 | { |
9caff1e7 | 385 | struct list_head *iter; |
ff59c456 | 386 | struct slave *slave; |
ff59c456 | 387 | |
0965a1f3 | 388 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
389 | goto down; |
390 | ||
01844098 | 391 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
ff59c456 JV |
392 | return bond_3ad_set_carrier(bond); |
393 | ||
9caff1e7 | 394 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
395 | if (slave->link == BOND_LINK_UP) { |
396 | if (!netif_carrier_ok(bond->dev)) { | |
397 | netif_carrier_on(bond->dev); | |
398 | return 1; | |
399 | } | |
400 | return 0; | |
401 | } | |
402 | } | |
403 | ||
404 | down: | |
405 | if (netif_carrier_ok(bond->dev)) { | |
406 | netif_carrier_off(bond->dev); | |
407 | return 1; | |
408 | } | |
409 | return 0; | |
410 | } | |
411 | ||
547942ca | 412 | /* Get link speed and duplex from the slave's base driver |
1da177e4 | 413 | * using ethtool. If for some reason the call fails or the |
98f41f69 | 414 | * values are invalid, set speed and duplex to -1, |
c4adfc82 MB |
415 | * and return. Return 1 if speed or duplex settings are |
416 | * UNKNOWN; 0 otherwise. | |
1da177e4 | 417 | */ |
c4adfc82 | 418 | static int bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
419 | { |
420 | struct net_device *slave_dev = slave->dev; | |
9856909c | 421 | struct ethtool_link_ksettings ecmd; |
61a44b9c | 422 | int res; |
1da177e4 | 423 | |
589665f5 DC |
424 | slave->speed = SPEED_UNKNOWN; |
425 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 426 | |
9856909c | 427 | res = __ethtool_get_link_ksettings(slave_dev, &ecmd); |
61a44b9c | 428 | if (res < 0) |
c4adfc82 | 429 | return 1; |
9856909c | 430 | if (ecmd.base.speed == 0 || ecmd.base.speed == ((__u32)-1)) |
c4adfc82 | 431 | return 1; |
9856909c | 432 | switch (ecmd.base.duplex) { |
1da177e4 LT |
433 | case DUPLEX_FULL: |
434 | case DUPLEX_HALF: | |
435 | break; | |
436 | default: | |
c4adfc82 | 437 | return 1; |
1da177e4 LT |
438 | } |
439 | ||
9856909c DD |
440 | slave->speed = ecmd.base.speed; |
441 | slave->duplex = ecmd.base.duplex; | |
1da177e4 | 442 | |
c4adfc82 | 443 | return 0; |
1da177e4 LT |
444 | } |
445 | ||
07699f9a | 446 | const char *bond_slave_link_status(s8 link) |
447 | { | |
448 | switch (link) { | |
449 | case BOND_LINK_UP: | |
450 | return "up"; | |
451 | case BOND_LINK_FAIL: | |
452 | return "going down"; | |
453 | case BOND_LINK_DOWN: | |
454 | return "down"; | |
455 | case BOND_LINK_BACK: | |
456 | return "going back"; | |
457 | default: | |
458 | return "unknown"; | |
459 | } | |
460 | } | |
461 | ||
547942ca | 462 | /* if <dev> supports MII link status reporting, check its link status. |
1da177e4 LT |
463 | * |
464 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 465 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
466 | * |
467 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
468 | * can't tell and just pretend it is), or 0, meaning that the link is | |
469 | * down. | |
470 | * | |
471 | * If reporting is non-zero, instead of faking link up, return -1 if | |
472 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
473 | * support them). If use_carrier is set, return whatever it says. | |
474 | * It'd be nice if there was a good way to tell if a driver supports | |
475 | * netif_carrier, but there really isn't. | |
476 | */ | |
3d632c3f SH |
477 | static int bond_check_dev_link(struct bonding *bond, |
478 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 479 | { |
eb7cc59a | 480 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 481 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
482 | struct ifreq ifr; |
483 | struct mii_ioctl_data *mii; | |
1da177e4 | 484 | |
6c988853 PG |
485 | if (!reporting && !netif_running(slave_dev)) |
486 | return 0; | |
487 | ||
eb7cc59a | 488 | if (bond->params.use_carrier) |
b3c898e2 | 489 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 490 | |
29112f4e | 491 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
492 | if (slave_dev->ethtool_ops->get_link) |
493 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
494 | BMSR_LSTATUS : 0; | |
29112f4e | 495 | |
3d632c3f | 496 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 497 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 | 498 | if (ioctl) { |
547942ca NA |
499 | /* TODO: set pointer to correct ioctl on a per team member |
500 | * bases to make this more efficient. that is, once | |
501 | * we determine the correct ioctl, we will always | |
502 | * call it and not the others for that team | |
503 | * member. | |
504 | */ | |
505 | ||
506 | /* We cannot assume that SIOCGMIIPHY will also read a | |
1da177e4 LT |
507 | * register; not all network drivers (e.g., e100) |
508 | * support that. | |
509 | */ | |
510 | ||
511 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
512 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
513 | mii = if_mii(&ifr); | |
4ad41c1e | 514 | if (ioctl(slave_dev, &ifr, SIOCGMIIPHY) == 0) { |
1da177e4 | 515 | mii->reg_num = MII_BMSR; |
4ad41c1e | 516 | if (ioctl(slave_dev, &ifr, SIOCGMIIREG) == 0) |
3d632c3f | 517 | return mii->val_out & BMSR_LSTATUS; |
1da177e4 LT |
518 | } |
519 | } | |
520 | ||
547942ca | 521 | /* If reporting, report that either there's no dev->do_ioctl, |
61a44b9c | 522 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
523 | * cannot report link status). If not reporting, pretend |
524 | * we're ok. | |
525 | */ | |
3d632c3f | 526 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
527 | } |
528 | ||
529 | /*----------------------------- Multicast list ------------------------------*/ | |
530 | ||
547942ca | 531 | /* Push the promiscuity flag down to appropriate slaves */ |
7e1a1ac1 | 532 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 533 | { |
9caff1e7 | 534 | struct list_head *iter; |
7e1a1ac1 | 535 | int err = 0; |
9caff1e7 | 536 | |
ec0865a9 | 537 | if (bond_uses_primary(bond)) { |
14056e79 | 538 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 539 | |
4740d638 ED |
540 | if (curr_active) |
541 | err = dev_set_promiscuity(curr_active->dev, inc); | |
1da177e4 LT |
542 | } else { |
543 | struct slave *slave; | |
dec1e90e | 544 | |
9caff1e7 | 545 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
546 | err = dev_set_promiscuity(slave->dev, inc); |
547 | if (err) | |
548 | return err; | |
1da177e4 LT |
549 | } |
550 | } | |
7e1a1ac1 | 551 | return err; |
1da177e4 LT |
552 | } |
553 | ||
547942ca | 554 | /* Push the allmulti flag down to all slaves */ |
7e1a1ac1 | 555 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 556 | { |
9caff1e7 | 557 | struct list_head *iter; |
7e1a1ac1 | 558 | int err = 0; |
9caff1e7 | 559 | |
ec0865a9 | 560 | if (bond_uses_primary(bond)) { |
14056e79 | 561 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 562 | |
4740d638 ED |
563 | if (curr_active) |
564 | err = dev_set_allmulti(curr_active->dev, inc); | |
1da177e4 LT |
565 | } else { |
566 | struct slave *slave; | |
dec1e90e | 567 | |
9caff1e7 | 568 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
569 | err = dev_set_allmulti(slave->dev, inc); |
570 | if (err) | |
571 | return err; | |
1da177e4 LT |
572 | } |
573 | } | |
7e1a1ac1 | 574 | return err; |
1da177e4 LT |
575 | } |
576 | ||
547942ca | 577 | /* Retrieve the list of registered multicast addresses for the bonding |
5a37e8ca FL |
578 | * device and retransmit an IGMP JOIN request to the current active |
579 | * slave. | |
580 | */ | |
f2369109 | 581 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 582 | { |
f2369109 | 583 | struct bonding *bond = container_of(work, struct bonding, |
584 | mcast_work.work); | |
585 | ||
4aa5dee4 | 586 | if (!rtnl_trylock()) { |
4beac029 | 587 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 588 | return; |
5a37e8ca | 589 | } |
4aa5dee4 | 590 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 591 | |
4f5474e7 NA |
592 | if (bond->igmp_retrans > 1) { |
593 | bond->igmp_retrans--; | |
c2952c31 | 594 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 595 | } |
f2369109 | 596 | rtnl_unlock(); |
5a37e8ca FL |
597 | } |
598 | ||
547942ca | 599 | /* Flush bond's hardware addresses from slave */ |
303d1cbf | 600 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 601 | struct net_device *slave_dev) |
1da177e4 | 602 | { |
454d7c9b | 603 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 604 | |
303d1cbf JV |
605 | dev_uc_unsync(slave_dev, bond_dev); |
606 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 | 607 | |
01844098 | 608 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
609 | /* del lacpdu mc addr from mc list */ |
610 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
611 | ||
22bedad3 | 612 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
613 | } |
614 | } | |
615 | ||
616 | /*--------------------------- Active slave change ---------------------------*/ | |
617 | ||
303d1cbf | 618 | /* Update the hardware address list and promisc/allmulti for the new and |
ec0865a9 VF |
619 | * old active slaves (if any). Modes that are not using primary keep all |
620 | * slaves up date at all times; only the modes that use primary need to call | |
303d1cbf | 621 | * this function to swap these settings during a failover. |
1da177e4 | 622 | */ |
303d1cbf JV |
623 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
624 | struct slave *old_active) | |
1da177e4 | 625 | { |
1da177e4 | 626 | if (old_active) { |
3d632c3f | 627 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 628 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 629 | |
3d632c3f | 630 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 631 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 632 | |
303d1cbf | 633 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
634 | } |
635 | ||
636 | if (new_active) { | |
7e1a1ac1 | 637 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 638 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 639 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 640 | |
3d632c3f | 641 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 642 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 643 | |
d632ce98 | 644 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
645 | dev_uc_sync(new_active->dev, bond->dev); |
646 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 647 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
648 | } |
649 | } | |
650 | ||
ae0d6750 | 651 | /** |
652 | * bond_set_dev_addr - clone slave's address to bond | |
653 | * @bond_dev: bond net device | |
654 | * @slave_dev: slave net device | |
655 | * | |
656 | * Should be called with RTNL held. | |
657 | */ | |
b9245914 PM |
658 | static int bond_set_dev_addr(struct net_device *bond_dev, |
659 | struct net_device *slave_dev) | |
ae0d6750 | 660 | { |
1caf40de PM |
661 | int err; |
662 | ||
e2a7420d JW |
663 | slave_dbg(bond_dev, slave_dev, "bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
664 | bond_dev, slave_dev, slave_dev->addr_len); | |
1caf40de PM |
665 | err = dev_pre_changeaddr_notify(bond_dev, slave_dev->dev_addr, NULL); |
666 | if (err) | |
667 | return err; | |
668 | ||
ae0d6750 | 669 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
670 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
671 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
b9245914 | 672 | return 0; |
ae0d6750 | 673 | } |
674 | ||
a951bc1e | 675 | static struct slave *bond_get_old_active(struct bonding *bond, |
676 | struct slave *new_active) | |
677 | { | |
678 | struct slave *slave; | |
679 | struct list_head *iter; | |
680 | ||
681 | bond_for_each_slave(bond, slave, iter) { | |
682 | if (slave == new_active) | |
683 | continue; | |
684 | ||
685 | if (ether_addr_equal(bond->dev->dev_addr, slave->dev->dev_addr)) | |
686 | return slave; | |
687 | } | |
688 | ||
689 | return NULL; | |
690 | } | |
691 | ||
547942ca | 692 | /* bond_do_fail_over_mac |
3915c1e8 JV |
693 | * |
694 | * Perform special MAC address swapping for fail_over_mac settings | |
695 | * | |
1c72cfdc | 696 | * Called with RTNL |
3915c1e8 JV |
697 | */ |
698 | static void bond_do_fail_over_mac(struct bonding *bond, | |
699 | struct slave *new_active, | |
700 | struct slave *old_active) | |
701 | { | |
faeeb317 JW |
702 | u8 tmp_mac[MAX_ADDR_LEN]; |
703 | struct sockaddr_storage ss; | |
3915c1e8 JV |
704 | int rv; |
705 | ||
706 | switch (bond->params.fail_over_mac) { | |
707 | case BOND_FOM_ACTIVE: | |
b9245914 PM |
708 | if (new_active) { |
709 | rv = bond_set_dev_addr(bond->dev, new_active->dev); | |
710 | if (rv) | |
e2a7420d JW |
711 | slave_err(bond->dev, new_active->dev, "Error %d setting bond MAC from slave\n", |
712 | -rv); | |
b9245914 | 713 | } |
3915c1e8 JV |
714 | break; |
715 | case BOND_FOM_FOLLOW: | |
547942ca | 716 | /* if new_active && old_active, swap them |
3915c1e8 JV |
717 | * if just old_active, do nothing (going to no active slave) |
718 | * if just new_active, set new_active to bond's MAC | |
719 | */ | |
720 | if (!new_active) | |
721 | return; | |
722 | ||
a951bc1e | 723 | if (!old_active) |
724 | old_active = bond_get_old_active(bond, new_active); | |
725 | ||
3915c1e8 | 726 | if (old_active) { |
faeeb317 JW |
727 | bond_hw_addr_copy(tmp_mac, new_active->dev->dev_addr, |
728 | new_active->dev->addr_len); | |
729 | bond_hw_addr_copy(ss.__data, | |
730 | old_active->dev->dev_addr, | |
731 | old_active->dev->addr_len); | |
732 | ss.ss_family = new_active->dev->type; | |
3915c1e8 | 733 | } else { |
faeeb317 JW |
734 | bond_hw_addr_copy(ss.__data, bond->dev->dev_addr, |
735 | bond->dev->addr_len); | |
736 | ss.ss_family = bond->dev->type; | |
3915c1e8 JV |
737 | } |
738 | ||
faeeb317 | 739 | rv = dev_set_mac_address(new_active->dev, |
3a37a963 | 740 | (struct sockaddr *)&ss, NULL); |
3915c1e8 | 741 | if (rv) { |
e2a7420d JW |
742 | slave_err(bond->dev, new_active->dev, "Error %d setting MAC of new active slave\n", |
743 | -rv); | |
3915c1e8 JV |
744 | goto out; |
745 | } | |
746 | ||
747 | if (!old_active) | |
748 | goto out; | |
749 | ||
faeeb317 JW |
750 | bond_hw_addr_copy(ss.__data, tmp_mac, |
751 | new_active->dev->addr_len); | |
752 | ss.ss_family = old_active->dev->type; | |
3915c1e8 | 753 | |
faeeb317 | 754 | rv = dev_set_mac_address(old_active->dev, |
3a37a963 | 755 | (struct sockaddr *)&ss, NULL); |
3915c1e8 | 756 | if (rv) |
e2a7420d JW |
757 | slave_err(bond->dev, old_active->dev, "Error %d setting MAC of old active slave\n", |
758 | -rv); | |
3915c1e8 | 759 | out: |
3915c1e8 JV |
760 | break; |
761 | default: | |
76444f50 VF |
762 | netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n", |
763 | bond->params.fail_over_mac); | |
3915c1e8 JV |
764 | break; |
765 | } | |
766 | ||
767 | } | |
768 | ||
b5a983f3 | 769 | static struct slave *bond_choose_primary_or_current(struct bonding *bond) |
a549952a | 770 | { |
059b47e8 | 771 | struct slave *prim = rtnl_dereference(bond->primary_slave); |
1c72cfdc | 772 | struct slave *curr = rtnl_dereference(bond->curr_active_slave); |
a549952a | 773 | |
b5a983f3 MR |
774 | if (!prim || prim->link != BOND_LINK_UP) { |
775 | if (!curr || curr->link != BOND_LINK_UP) | |
776 | return NULL; | |
777 | return curr; | |
778 | } | |
779 | ||
a549952a JP |
780 | if (bond->force_primary) { |
781 | bond->force_primary = false; | |
b5a983f3 MR |
782 | return prim; |
783 | } | |
784 | ||
785 | if (!curr || curr->link != BOND_LINK_UP) | |
786 | return prim; | |
787 | ||
788 | /* At this point, prim and curr are both up */ | |
789 | switch (bond->params.primary_reselect) { | |
790 | case BOND_PRI_RESELECT_ALWAYS: | |
791 | return prim; | |
792 | case BOND_PRI_RESELECT_BETTER: | |
793 | if (prim->speed < curr->speed) | |
794 | return curr; | |
795 | if (prim->speed == curr->speed && prim->duplex <= curr->duplex) | |
796 | return curr; | |
797 | return prim; | |
798 | case BOND_PRI_RESELECT_FAILURE: | |
799 | return curr; | |
800 | default: | |
801 | netdev_err(bond->dev, "impossible primary_reselect %d\n", | |
802 | bond->params.primary_reselect); | |
803 | return curr; | |
a549952a | 804 | } |
a549952a | 805 | } |
3915c1e8 | 806 | |
1da177e4 | 807 | /** |
b5a983f3 | 808 | * bond_find_best_slave - select the best available slave to be the active one |
1da177e4 | 809 | * @bond: our bonding struct |
1da177e4 LT |
810 | */ |
811 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
812 | { | |
b5a983f3 | 813 | struct slave *slave, *bestslave = NULL; |
77140d29 | 814 | struct list_head *iter; |
1da177e4 | 815 | int mintime = bond->params.updelay; |
1da177e4 | 816 | |
b5a983f3 MR |
817 | slave = bond_choose_primary_or_current(bond); |
818 | if (slave) | |
819 | return slave; | |
1da177e4 | 820 | |
77140d29 VF |
821 | bond_for_each_slave(bond, slave, iter) { |
822 | if (slave->link == BOND_LINK_UP) | |
823 | return slave; | |
b6adc610 | 824 | if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) && |
77140d29 VF |
825 | slave->delay < mintime) { |
826 | mintime = slave->delay; | |
827 | bestslave = slave; | |
1da177e4 LT |
828 | } |
829 | } | |
830 | ||
831 | return bestslave; | |
832 | } | |
833 | ||
ad246c99 BH |
834 | static bool bond_should_notify_peers(struct bonding *bond) |
835 | { | |
4cb4f97b | 836 | struct slave *slave; |
837 | ||
838 | rcu_read_lock(); | |
839 | slave = rcu_dereference(bond->curr_active_slave); | |
840 | rcu_read_unlock(); | |
ad246c99 | 841 | |
76444f50 VF |
842 | netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n", |
843 | slave ? slave->dev->name : "NULL"); | |
ad246c99 BH |
844 | |
845 | if (!slave || !bond->send_peer_notif || | |
07a4ddec VB |
846 | bond->send_peer_notif % |
847 | max(1, bond->params.peer_notif_delay) != 0 || | |
b02e3e94 | 848 | !netif_carrier_ok(bond->dev) || |
ad246c99 BH |
849 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) |
850 | return false; | |
851 | ||
ad246c99 BH |
852 | return true; |
853 | } | |
854 | ||
1da177e4 LT |
855 | /** |
856 | * change_active_interface - change the active slave into the specified one | |
857 | * @bond: our bonding struct | |
858 | * @new: the new slave to make the active one | |
859 | * | |
860 | * Set the new slave to the bond's settings and unset them on the old | |
861 | * curr_active_slave. | |
862 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
863 | * | |
864 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
865 | * because it is apparently the best available slave we have, even though its | |
866 | * updelay hasn't timed out yet. | |
867 | * | |
1c72cfdc | 868 | * Caller must hold RTNL. |
1da177e4 | 869 | */ |
a77b5325 | 870 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 | 871 | { |
4740d638 ED |
872 | struct slave *old_active; |
873 | ||
1c72cfdc NA |
874 | ASSERT_RTNL(); |
875 | ||
876 | old_active = rtnl_dereference(bond->curr_active_slave); | |
1da177e4 | 877 | |
3d632c3f | 878 | if (old_active == new_active) |
1da177e4 | 879 | return; |
1da177e4 LT |
880 | |
881 | if (new_active) { | |
8e603460 | 882 | new_active->last_link_up = jiffies; |
b2220cad | 883 | |
1da177e4 | 884 | if (new_active->link == BOND_LINK_BACK) { |
ec0865a9 | 885 | if (bond_uses_primary(bond)) { |
e2a7420d JW |
886 | slave_info(bond->dev, new_active->dev, "making interface the new active one %d ms earlier\n", |
887 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
888 | } |
889 | ||
890 | new_active->delay = 0; | |
5d397061 JP |
891 | bond_set_slave_link_state(new_active, BOND_LINK_UP, |
892 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 893 | |
01844098 | 894 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 895 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 896 | |
58402054 | 897 | if (bond_is_lb(bond)) |
1da177e4 | 898 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 | 899 | } else { |
ec0865a9 | 900 | if (bond_uses_primary(bond)) { |
e2a7420d | 901 | slave_info(bond->dev, new_active->dev, "making interface the new active one\n"); |
1da177e4 LT |
902 | } |
903 | } | |
904 | } | |
905 | ||
ec0865a9 | 906 | if (bond_uses_primary(bond)) |
303d1cbf | 907 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 908 | |
58402054 | 909 | if (bond_is_lb(bond)) { |
1da177e4 | 910 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 911 | if (old_active) |
5e5b0665 | 912 | bond_set_slave_inactive_flags(old_active, |
913 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 914 | if (new_active) |
5e5b0665 | 915 | bond_set_slave_active_flags(new_active, |
916 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 917 | } else { |
278b2083 | 918 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 919 | } |
c3ade5ca | 920 | |
01844098 | 921 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) { |
3d632c3f | 922 | if (old_active) |
5e5b0665 | 923 | bond_set_slave_inactive_flags(old_active, |
924 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
925 | |
926 | if (new_active) { | |
ad246c99 BH |
927 | bool should_notify_peers = false; |
928 | ||
5e5b0665 | 929 | bond_set_slave_active_flags(new_active, |
930 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 931 | |
709f8a45 OG |
932 | if (bond->params.fail_over_mac) |
933 | bond_do_fail_over_mac(bond, new_active, | |
934 | old_active); | |
3915c1e8 | 935 | |
ad246c99 BH |
936 | if (netif_running(bond->dev)) { |
937 | bond->send_peer_notif = | |
07a4ddec VB |
938 | bond->params.num_peer_notif * |
939 | max(1, bond->params.peer_notif_delay); | |
ad246c99 BH |
940 | should_notify_peers = |
941 | bond_should_notify_peers(bond); | |
942 | } | |
943 | ||
b7bc2a5b | 944 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
07a4ddec VB |
945 | if (should_notify_peers) { |
946 | bond->send_peer_notif--; | |
b7bc2a5b AW |
947 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
948 | bond->dev); | |
07a4ddec | 949 | } |
7893b249 | 950 | } |
c3ade5ca | 951 | } |
a2fd940f | 952 | |
5a37e8ca | 953 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
954 | * all were sent on curr_active_slave. |
955 | * resend only if bond is brought up with the affected | |
547942ca NA |
956 | * bonding modes and the retransmission is enabled |
957 | */ | |
94265cf5 | 958 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && |
ec0865a9 | 959 | ((bond_uses_primary(bond) && new_active) || |
01844098 | 960 | BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) { |
c2952c31 | 961 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 962 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 963 | } |
1da177e4 LT |
964 | } |
965 | ||
966 | /** | |
967 | * bond_select_active_slave - select a new active slave, if needed | |
968 | * @bond: our bonding struct | |
969 | * | |
3d632c3f | 970 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
971 | * - The old curr_active_slave has been released or lost its link. |
972 | * - The primary_slave has got its link back. | |
973 | * - A slave has got its link back and there's no old curr_active_slave. | |
974 | * | |
1c72cfdc | 975 | * Caller must hold RTNL. |
1da177e4 | 976 | */ |
a77b5325 | 977 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
978 | { |
979 | struct slave *best_slave; | |
ff59c456 | 980 | int rv; |
1da177e4 | 981 | |
e0974585 NA |
982 | ASSERT_RTNL(); |
983 | ||
1da177e4 | 984 | best_slave = bond_find_best_slave(bond); |
1c72cfdc | 985 | if (best_slave != rtnl_dereference(bond->curr_active_slave)) { |
1da177e4 | 986 | bond_change_active_slave(bond, best_slave); |
ff59c456 JV |
987 | rv = bond_set_carrier(bond); |
988 | if (!rv) | |
989 | return; | |
990 | ||
d66bd905 | 991 | if (netif_carrier_ok(bond->dev)) |
b8bd72d3 | 992 | netdev_info(bond->dev, "active interface up!\n"); |
d66bd905 | 993 | else |
76444f50 | 994 | netdev_info(bond->dev, "now running without any active interface!\n"); |
1da177e4 LT |
995 | } |
996 | } | |
997 | ||
f6dc31a8 | 998 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 999 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1000 | { |
8a8efa22 AW |
1001 | struct netpoll *np; |
1002 | int err = 0; | |
f6dc31a8 | 1003 | |
a8779ec1 | 1004 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
1005 | err = -ENOMEM; |
1006 | if (!np) | |
1007 | goto out; | |
1008 | ||
a8779ec1 | 1009 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
1010 | if (err) { |
1011 | kfree(np); | |
1012 | goto out; | |
f6dc31a8 | 1013 | } |
8a8efa22 AW |
1014 | slave->np = np; |
1015 | out: | |
1016 | return err; | |
1017 | } | |
1018 | static inline void slave_disable_netpoll(struct slave *slave) | |
1019 | { | |
1020 | struct netpoll *np = slave->np; | |
1021 | ||
1022 | if (!np) | |
1023 | return; | |
1024 | ||
1025 | slave->np = NULL; | |
c9fbd71f DB |
1026 | |
1027 | __netpoll_free(np); | |
8a8efa22 | 1028 | } |
f6dc31a8 WC |
1029 | |
1030 | static void bond_poll_controller(struct net_device *bond_dev) | |
1031 | { | |
616f4541 MB |
1032 | struct bonding *bond = netdev_priv(bond_dev); |
1033 | struct slave *slave = NULL; | |
1034 | struct list_head *iter; | |
1035 | struct ad_info ad_info; | |
616f4541 MB |
1036 | |
1037 | if (BOND_MODE(bond) == BOND_MODE_8023AD) | |
1038 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) | |
1039 | return; | |
1040 | ||
616f4541 | 1041 | bond_for_each_slave_rcu(bond, slave, iter) { |
93f62ad5 | 1042 | if (!bond_slave_is_up(slave)) |
616f4541 MB |
1043 | continue; |
1044 | ||
1045 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
1046 | struct aggregator *agg = | |
1047 | SLAVE_AD_INFO(slave)->port.aggregator; | |
1048 | ||
1049 | if (agg && | |
1050 | agg->aggregator_identifier != ad_info.aggregator_id) | |
1051 | continue; | |
1052 | } | |
1053 | ||
93f62ad5 | 1054 | netpoll_poll_dev(slave->dev); |
616f4541 | 1055 | } |
8a8efa22 AW |
1056 | } |
1057 | ||
c4cdef9b | 1058 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1059 | { |
c4cdef9b | 1060 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1061 | struct list_head *iter; |
c2355e1a | 1062 | struct slave *slave; |
c2355e1a | 1063 | |
9caff1e7 | 1064 | bond_for_each_slave(bond, slave, iter) |
b6adc610 | 1065 | if (bond_slave_is_up(slave)) |
8a8efa22 | 1066 | slave_disable_netpoll(slave); |
f6dc31a8 | 1067 | } |
8a8efa22 | 1068 | |
a8779ec1 | 1069 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
1070 | { |
1071 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 1072 | struct list_head *iter; |
f6dc31a8 | 1073 | struct slave *slave; |
dec1e90e | 1074 | int err = 0; |
f6dc31a8 | 1075 | |
9caff1e7 | 1076 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
1077 | err = slave_enable_netpoll(slave); |
1078 | if (err) { | |
c4cdef9b | 1079 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1080 | break; |
f6dc31a8 WC |
1081 | } |
1082 | } | |
8a8efa22 | 1083 | return err; |
f6dc31a8 | 1084 | } |
8a8efa22 AW |
1085 | #else |
1086 | static inline int slave_enable_netpoll(struct slave *slave) | |
1087 | { | |
1088 | return 0; | |
1089 | } | |
1090 | static inline void slave_disable_netpoll(struct slave *slave) | |
1091 | { | |
1092 | } | |
f6dc31a8 WC |
1093 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1094 | { | |
1095 | } | |
f6dc31a8 WC |
1096 | #endif |
1097 | ||
1da177e4 LT |
1098 | /*---------------------------------- IOCTL ----------------------------------*/ |
1099 | ||
c8f44aff | 1100 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1101 | netdev_features_t features) |
8531c5ff | 1102 | { |
b2a103e6 | 1103 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 1104 | struct list_head *iter; |
c8f44aff | 1105 | netdev_features_t mask; |
9b7b165a | 1106 | struct slave *slave; |
b63365a2 | 1107 | |
7889cbee | 1108 | mask = features; |
c158cba3 | 1109 | |
b63365a2 | 1110 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1111 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1112 | |
9caff1e7 | 1113 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
1114 | features = netdev_increment_features(features, |
1115 | slave->dev->features, | |
b2a103e6 MM |
1116 | mask); |
1117 | } | |
b0ce3508 | 1118 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1119 | |
b2a103e6 MM |
1120 | return features; |
1121 | } | |
1122 | ||
a188222b | 1123 | #define BOND_VLAN_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
62f2a3a4 MM |
1124 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ |
1125 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 | 1126 | |
a188222b TH |
1127 | #define BOND_ENC_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
1128 | NETIF_F_RXCSUM | NETIF_F_ALL_TSO) | |
5a7baa78 | 1129 | |
2e770b50 AL |
1130 | #define BOND_MPLS_FEATURES (NETIF_F_HW_CSUM | NETIF_F_SG | \ |
1131 | NETIF_F_ALL_TSO) | |
1132 | ||
b2a103e6 MM |
1133 | static void bond_compute_features(struct bonding *bond) |
1134 | { | |
02875878 ED |
1135 | unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | |
1136 | IFF_XMIT_DST_RELEASE_PERM; | |
c8f44aff | 1137 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
5a7baa78 | 1138 | netdev_features_t enc_features = BOND_ENC_FEATURES; |
2e770b50 | 1139 | netdev_features_t mpls_features = BOND_MPLS_FEATURES; |
9caff1e7 VF |
1140 | struct net_device *bond_dev = bond->dev; |
1141 | struct list_head *iter; | |
1142 | struct slave *slave; | |
b2a103e6 | 1143 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1144 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1145 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1146 | |
0965a1f3 | 1147 | if (!bond_has_slaves(bond)) |
b2a103e6 | 1148 | goto done; |
a9b3ace4 | 1149 | vlan_features &= NETIF_F_ALL_FOR_ALL; |
2e770b50 | 1150 | mpls_features &= NETIF_F_ALL_FOR_ALL; |
b2a103e6 | 1151 | |
9caff1e7 | 1152 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1153 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1154 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1155 | ||
5a7baa78 OG |
1156 | enc_features = netdev_increment_features(enc_features, |
1157 | slave->dev->hw_enc_features, | |
1158 | BOND_ENC_FEATURES); | |
2e770b50 AL |
1159 | |
1160 | mpls_features = netdev_increment_features(mpls_features, | |
1161 | slave->dev->mpls_features, | |
1162 | BOND_MPLS_FEATURES); | |
1163 | ||
b6fe83e9 | 1164 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1165 | if (slave->dev->hard_header_len > max_hard_header_len) |
1166 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1167 | |
1168 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1169 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1170 | } |
19cdead3 | 1171 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff | 1172 | |
b63365a2 | 1173 | done: |
b2a103e6 | 1174 | bond_dev->vlan_features = vlan_features; |
8eea1ca8 | 1175 | bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL | |
d595b03d Y |
1176 | NETIF_F_HW_VLAN_CTAG_TX | |
1177 | NETIF_F_HW_VLAN_STAG_TX | | |
8eea1ca8 | 1178 | NETIF_F_GSO_UDP_L4; |
2e770b50 | 1179 | bond_dev->mpls_features = mpls_features; |
0e376bd0 SB |
1180 | bond_dev->gso_max_segs = gso_max_segs; |
1181 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1182 | |
02875878 ED |
1183 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
1184 | if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) && | |
1185 | dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM)) | |
1186 | bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE; | |
b6fe83e9 | 1187 | |
b2a103e6 | 1188 | netdev_change_features(bond_dev); |
8531c5ff AK |
1189 | } |
1190 | ||
872254dd MS |
1191 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1192 | struct net_device *slave_dev) | |
1193 | { | |
00829823 | 1194 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1195 | |
1196 | bond_dev->type = slave_dev->type; | |
1197 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1198 | bond_dev->addr_len = slave_dev->addr_len; | |
1199 | ||
1200 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1201 | slave_dev->addr_len); | |
1202 | } | |
1203 | ||
5b2c4dd2 | 1204 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1205 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1206 | */ |
1207 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1208 | struct slave *slave, |
1209 | struct bonding *bond) | |
5b2c4dd2 | 1210 | { |
2d7011ca | 1211 | if (bond_is_slave_inactive(slave)) { |
01844098 | 1212 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
5b2c4dd2 JP |
1213 | skb->pkt_type != PACKET_BROADCAST && |
1214 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1215 | return false; |
5b2c4dd2 JP |
1216 | return true; |
1217 | } | |
1218 | return false; | |
1219 | } | |
1220 | ||
8a4eb573 | 1221 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1222 | { |
8a4eb573 | 1223 | struct sk_buff *skb = *pskb; |
f1c1775a | 1224 | struct slave *slave; |
0bd80dad | 1225 | struct bonding *bond; |
de063b70 ED |
1226 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1227 | struct slave *); | |
13a8e0c8 | 1228 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1229 | |
8a4eb573 JP |
1230 | skb = skb_share_check(skb, GFP_ATOMIC); |
1231 | if (unlikely(!skb)) | |
1232 | return RX_HANDLER_CONSUMED; | |
1233 | ||
1234 | *pskb = skb; | |
5b2c4dd2 | 1235 | |
35d48903 JP |
1236 | slave = bond_slave_get_rcu(skb->dev); |
1237 | bond = slave->bond; | |
0bd80dad | 1238 | |
6aa7de05 | 1239 | recv_probe = READ_ONCE(bond->recv_probe); |
4d97480b | 1240 | if (recv_probe) { |
de063b70 ED |
1241 | ret = recv_probe(skb, bond, slave); |
1242 | if (ret == RX_HANDLER_CONSUMED) { | |
1243 | consume_skb(skb); | |
1244 | return ret; | |
3aba891d JP |
1245 | } |
1246 | } | |
1247 | ||
3c963a33 MS |
1248 | /* |
1249 | * For packets determined by bond_should_deliver_exact_match() call to | |
1250 | * be suppressed we want to make an exception for link-local packets. | |
1251 | * This is necessary for e.g. LLDP daemons to be able to monitor | |
1252 | * inactive slave links without being forced to bind to them | |
1253 | * explicitly. | |
1254 | * | |
1255 | * At the same time, packets that are passed to the bonding master | |
1256 | * (including link-local ones) can have their originating interface | |
1257 | * determined via PACKET_ORIGDEV socket option. | |
6a9e461f | 1258 | */ |
3c963a33 MS |
1259 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
1260 | if (is_link_local_ether_addr(eth_hdr(skb)->h_dest)) | |
1261 | return RX_HANDLER_PASS; | |
8a4eb573 | 1262 | return RX_HANDLER_EXACT; |
3c963a33 | 1263 | } |
5b2c4dd2 | 1264 | |
35d48903 | 1265 | skb->dev = bond->dev; |
5b2c4dd2 | 1266 | |
01844098 | 1267 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
35d48903 | 1268 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1269 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1270 | |
541ac7c9 CG |
1271 | if (unlikely(skb_cow_head(skb, |
1272 | skb->data - skb_mac_header(skb)))) { | |
1273 | kfree_skb(skb); | |
8a4eb573 | 1274 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1275 | } |
faeeb317 JW |
1276 | bond_hw_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, |
1277 | bond->dev->addr_len); | |
5b2c4dd2 JP |
1278 | } |
1279 | ||
13a8e0c8 | 1280 | return ret; |
5b2c4dd2 JP |
1281 | } |
1282 | ||
41f0b049 | 1283 | static enum netdev_lag_tx_type bond_lag_tx_type(struct bonding *bond) |
471cb5a3 | 1284 | { |
41f0b049 JP |
1285 | switch (BOND_MODE(bond)) { |
1286 | case BOND_MODE_ROUNDROBIN: | |
1287 | return NETDEV_LAG_TX_TYPE_ROUNDROBIN; | |
1288 | case BOND_MODE_ACTIVEBACKUP: | |
1289 | return NETDEV_LAG_TX_TYPE_ACTIVEBACKUP; | |
1290 | case BOND_MODE_BROADCAST: | |
1291 | return NETDEV_LAG_TX_TYPE_BROADCAST; | |
1292 | case BOND_MODE_XOR: | |
1293 | case BOND_MODE_8023AD: | |
1294 | return NETDEV_LAG_TX_TYPE_HASH; | |
1295 | default: | |
1296 | return NETDEV_LAG_TX_TYPE_UNKNOWN; | |
1297 | } | |
1298 | } | |
1299 | ||
f44aa9ef JH |
1300 | static enum netdev_lag_hash bond_lag_hash_type(struct bonding *bond, |
1301 | enum netdev_lag_tx_type type) | |
1302 | { | |
1303 | if (type != NETDEV_LAG_TX_TYPE_HASH) | |
1304 | return NETDEV_LAG_HASH_NONE; | |
1305 | ||
1306 | switch (bond->params.xmit_policy) { | |
1307 | case BOND_XMIT_POLICY_LAYER2: | |
1308 | return NETDEV_LAG_HASH_L2; | |
1309 | case BOND_XMIT_POLICY_LAYER34: | |
1310 | return NETDEV_LAG_HASH_L34; | |
1311 | case BOND_XMIT_POLICY_LAYER23: | |
1312 | return NETDEV_LAG_HASH_L23; | |
1313 | case BOND_XMIT_POLICY_ENCAP23: | |
1314 | return NETDEV_LAG_HASH_E23; | |
1315 | case BOND_XMIT_POLICY_ENCAP34: | |
1316 | return NETDEV_LAG_HASH_E34; | |
1317 | default: | |
1318 | return NETDEV_LAG_HASH_UNKNOWN; | |
1319 | } | |
1320 | } | |
1321 | ||
42ab19ee DA |
1322 | static int bond_master_upper_dev_link(struct bonding *bond, struct slave *slave, |
1323 | struct netlink_ext_ack *extack) | |
41f0b049 JP |
1324 | { |
1325 | struct netdev_lag_upper_info lag_upper_info; | |
f44aa9ef | 1326 | enum netdev_lag_tx_type type; |
471cb5a3 | 1327 | |
f44aa9ef JH |
1328 | type = bond_lag_tx_type(bond); |
1329 | lag_upper_info.tx_type = type; | |
1330 | lag_upper_info.hash_type = bond_lag_hash_type(bond, type); | |
4597efe3 XL |
1331 | |
1332 | return netdev_master_upper_dev_link(slave->dev, bond->dev, slave, | |
1333 | &lag_upper_info, extack); | |
471cb5a3 JP |
1334 | } |
1335 | ||
41f0b049 | 1336 | static void bond_upper_dev_unlink(struct bonding *bond, struct slave *slave) |
471cb5a3 | 1337 | { |
41f0b049 JP |
1338 | netdev_upper_dev_unlink(slave->dev, bond->dev); |
1339 | slave->dev->flags &= ~IFF_SLAVE; | |
471cb5a3 JP |
1340 | } |
1341 | ||
3fdddd85 | 1342 | static struct slave *bond_alloc_slave(struct bonding *bond) |
1343 | { | |
1344 | struct slave *slave = NULL; | |
1345 | ||
d66bd905 | 1346 | slave = kzalloc(sizeof(*slave), GFP_KERNEL); |
3fdddd85 | 1347 | if (!slave) |
1348 | return NULL; | |
1349 | ||
01844098 | 1350 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
3fdddd85 | 1351 | SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info), |
1352 | GFP_KERNEL); | |
1353 | if (!SLAVE_AD_INFO(slave)) { | |
1354 | kfree(slave); | |
1355 | return NULL; | |
1356 | } | |
1357 | } | |
d4859d74 MB |
1358 | INIT_DELAYED_WORK(&slave->notify_work, bond_netdev_notify_work); |
1359 | ||
3fdddd85 | 1360 | return slave; |
1361 | } | |
1362 | ||
1363 | static void bond_free_slave(struct slave *slave) | |
1364 | { | |
1365 | struct bonding *bond = bond_get_bond_by_slave(slave); | |
1366 | ||
d4859d74 | 1367 | cancel_delayed_work_sync(&slave->notify_work); |
01844098 | 1368 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3fdddd85 | 1369 | kfree(SLAVE_AD_INFO(slave)); |
1370 | ||
1371 | kfree(slave); | |
1372 | } | |
1373 | ||
69a2338e MS |
1374 | static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info) |
1375 | { | |
1376 | info->bond_mode = BOND_MODE(bond); | |
1377 | info->miimon = bond->params.miimon; | |
1378 | info->num_slaves = bond->slave_cnt; | |
1379 | } | |
1380 | ||
1381 | static void bond_fill_ifslave(struct slave *slave, struct ifslave *info) | |
1382 | { | |
1383 | strcpy(info->slave_name, slave->dev->name); | |
1384 | info->link = slave->link; | |
1385 | info->state = bond_slave_state(slave); | |
1386 | info->link_failure_count = slave->link_failure_count; | |
1387 | } | |
1388 | ||
69e61133 MS |
1389 | static void bond_netdev_notify_work(struct work_struct *_work) |
1390 | { | |
d4859d74 MB |
1391 | struct slave *slave = container_of(_work, struct slave, |
1392 | notify_work.work); | |
1393 | ||
1394 | if (rtnl_trylock()) { | |
1395 | struct netdev_bonding_info binfo; | |
69e61133 | 1396 | |
d4859d74 MB |
1397 | bond_fill_ifslave(slave, &binfo.slave); |
1398 | bond_fill_ifbond(slave->bond, &binfo.master); | |
1399 | netdev_bonding_info_change(slave->dev, &binfo); | |
1400 | rtnl_unlock(); | |
1401 | } else { | |
1402 | queue_delayed_work(slave->bond->wq, &slave->notify_work, 1); | |
1403 | } | |
69e61133 MS |
1404 | } |
1405 | ||
1406 | void bond_queue_slave_event(struct slave *slave) | |
1407 | { | |
d4859d74 | 1408 | queue_delayed_work(slave->bond->wq, &slave->notify_work, 0); |
69e61133 MS |
1409 | } |
1410 | ||
f7c7eb7f JP |
1411 | void bond_lower_state_changed(struct slave *slave) |
1412 | { | |
1413 | struct netdev_lag_lower_state_info info; | |
1414 | ||
1415 | info.link_up = slave->link == BOND_LINK_UP || | |
1416 | slave->link == BOND_LINK_FAIL; | |
1417 | info.tx_enabled = bond_is_active_slave(slave); | |
1418 | netdev_lower_state_changed(slave->dev, &info); | |
1419 | } | |
1420 | ||
1da177e4 | 1421 | /* enslave device <slave> to bond device <master> */ |
33eaf2a6 DA |
1422 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev, |
1423 | struct netlink_ext_ack *extack) | |
1da177e4 | 1424 | { |
454d7c9b | 1425 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1426 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1427 | struct slave *new_slave = NULL, *prev_slave; |
faeeb317 | 1428 | struct sockaddr_storage ss; |
1da177e4 | 1429 | int link_reporting; |
8599b52e | 1430 | int res = 0, i; |
1da177e4 | 1431 | |
c772dde3 BH |
1432 | if (!bond->params.use_carrier && |
1433 | slave_dev->ethtool_ops->get_link == NULL && | |
1434 | slave_ops->ndo_do_ioctl == NULL) { | |
e2a7420d | 1435 | slave_warn(bond_dev, slave_dev, "no link monitoring support\n"); |
1da177e4 LT |
1436 | } |
1437 | ||
24b27fc4 MB |
1438 | /* already in-use? */ |
1439 | if (netdev_is_rx_handler_busy(slave_dev)) { | |
759088bd | 1440 | NL_SET_ERR_MSG(extack, "Device is in use and cannot be enslaved"); |
e2a7420d JW |
1441 | slave_err(bond_dev, slave_dev, |
1442 | "Error: Device is in use and cannot be enslaved\n"); | |
1da177e4 LT |
1443 | return -EBUSY; |
1444 | } | |
1445 | ||
09a89c21 | 1446 | if (bond_dev == slave_dev) { |
759088bd | 1447 | NL_SET_ERR_MSG(extack, "Cannot enslave bond to itself."); |
76444f50 | 1448 | netdev_err(bond_dev, "cannot enslave bond to itself.\n"); |
09a89c21 JB |
1449 | return -EPERM; |
1450 | } | |
1451 | ||
1da177e4 LT |
1452 | /* vlan challenged mutual exclusion */ |
1453 | /* no need to lock since we're protected by rtnl_lock */ | |
1454 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
e2a7420d | 1455 | slave_dbg(bond_dev, slave_dev, "is NETIF_F_VLAN_CHALLENGED\n"); |
55462cf3 | 1456 | if (vlan_uses_dev(bond_dev)) { |
759088bd | 1457 | NL_SET_ERR_MSG(extack, "Can not enslave VLAN challenged device to VLAN enabled bond"); |
e2a7420d | 1458 | slave_err(bond_dev, slave_dev, "Error: cannot enslave VLAN challenged slave on VLAN enabled bond\n"); |
1da177e4 LT |
1459 | return -EPERM; |
1460 | } else { | |
e2a7420d | 1461 | slave_warn(bond_dev, slave_dev, "enslaved VLAN challenged slave. Adding VLANs will be blocked as long as it is part of bond.\n"); |
1da177e4 LT |
1462 | } |
1463 | } else { | |
e2a7420d | 1464 | slave_dbg(bond_dev, slave_dev, "is !NETIF_F_VLAN_CHALLENGED\n"); |
1da177e4 LT |
1465 | } |
1466 | ||
547942ca | 1467 | /* Old ifenslave binaries are no longer supported. These can |
3d632c3f | 1468 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1469 | * the current ifenslave will set the interface down prior to |
1470 | * enslaving it; the old ifenslave will not. | |
1471 | */ | |
ce3ea1c7 | 1472 | if (slave_dev->flags & IFF_UP) { |
759088bd | 1473 | NL_SET_ERR_MSG(extack, "Device can not be enslaved while up"); |
e2a7420d | 1474 | slave_err(bond_dev, slave_dev, "slave is up - this may be due to an out of date ifenslave\n"); |
1e2a8868 | 1475 | return -EPERM; |
217df670 | 1476 | } |
1da177e4 | 1477 | |
872254dd MS |
1478 | /* set bonding device ether type by slave - bonding netdevices are |
1479 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1480 | * there is a need to override some of the type dependent attribs/funcs. | |
1481 | * | |
1482 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1483 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1484 | */ | |
0965a1f3 | 1485 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1486 | if (bond_dev->type != slave_dev->type) { |
e2a7420d JW |
1487 | slave_dbg(bond_dev, slave_dev, "change device type from %d to %d\n", |
1488 | bond_dev->type, slave_dev->type); | |
75c78500 | 1489 | |
b7bc2a5b AW |
1490 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1491 | bond_dev); | |
3ca5b404 JP |
1492 | res = notifier_to_errno(res); |
1493 | if (res) { | |
e2a7420d | 1494 | slave_err(bond_dev, slave_dev, "refused to change device type\n"); |
1e2a8868 | 1495 | return -EBUSY; |
3ca5b404 | 1496 | } |
75c78500 | 1497 | |
32a806c1 | 1498 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1499 | dev_uc_flush(bond_dev); |
22bedad3 | 1500 | dev_mc_flush(bond_dev); |
32a806c1 | 1501 | |
e36b9d16 MS |
1502 | if (slave_dev->type != ARPHRD_ETHER) |
1503 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1504 | else { |
e36b9d16 | 1505 | ether_setup(bond_dev); |
550fd08c NH |
1506 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1507 | } | |
75c78500 | 1508 | |
b7bc2a5b AW |
1509 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1510 | bond_dev); | |
e36b9d16 | 1511 | } |
872254dd | 1512 | } else if (bond_dev->type != slave_dev->type) { |
759088bd | 1513 | NL_SET_ERR_MSG(extack, "Device type is different from other slaves"); |
e2a7420d JW |
1514 | slave_err(bond_dev, slave_dev, "ether type (%d) is different from other slaves (%d), can not enslave it\n", |
1515 | slave_dev->type, bond_dev->type); | |
1e2a8868 | 1516 | return -EINVAL; |
872254dd MS |
1517 | } |
1518 | ||
1533e773 MB |
1519 | if (slave_dev->type == ARPHRD_INFINIBAND && |
1520 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { | |
759088bd | 1521 | NL_SET_ERR_MSG(extack, "Only active-backup mode is supported for infiniband slaves"); |
e2a7420d JW |
1522 | slave_warn(bond_dev, slave_dev, "Type (%d) supports only active-backup mode\n", |
1523 | slave_dev->type); | |
1533e773 MB |
1524 | res = -EOPNOTSUPP; |
1525 | goto err_undo_flags; | |
1526 | } | |
1527 | ||
1528 | if (!slave_ops->ndo_set_mac_address || | |
1529 | slave_dev->type == ARPHRD_INFINIBAND) { | |
e2a7420d | 1530 | slave_warn(bond_dev, slave_dev, "The slave device specified does not support setting the MAC address\n"); |
f5442441 VF |
1531 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP && |
1532 | bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
1533 | if (!bond_has_slaves(bond)) { | |
00503b6f | 1534 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
e2a7420d | 1535 | slave_warn(bond_dev, slave_dev, "Setting fail_over_mac to active for active-backup mode\n"); |
f5442441 | 1536 | } else { |
759088bd | 1537 | NL_SET_ERR_MSG(extack, "Slave device does not support setting the MAC address, but fail_over_mac is not set to active"); |
e2a7420d | 1538 | slave_err(bond_dev, slave_dev, "The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n"); |
f5442441 VF |
1539 | res = -EOPNOTSUPP; |
1540 | goto err_undo_flags; | |
00503b6f | 1541 | } |
2ab82852 | 1542 | } |
1da177e4 LT |
1543 | } |
1544 | ||
8d8fc29d AW |
1545 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1546 | ||
c20811a7 | 1547 | /* If this is the first slave, then we need to set the master's hardware |
547942ca NA |
1548 | * address to be the same as the slave's. |
1549 | */ | |
0965a1f3 | 1550 | if (!bond_has_slaves(bond) && |
b9245914 PM |
1551 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) { |
1552 | res = bond_set_dev_addr(bond->dev, slave_dev); | |
1553 | if (res) | |
1554 | goto err_undo_flags; | |
1555 | } | |
c20811a7 | 1556 | |
3fdddd85 | 1557 | new_slave = bond_alloc_slave(bond); |
1da177e4 LT |
1558 | if (!new_slave) { |
1559 | res = -ENOMEM; | |
1560 | goto err_undo_flags; | |
1561 | } | |
3fdddd85 | 1562 | |
dc73c41f VF |
1563 | new_slave->bond = bond; |
1564 | new_slave->dev = slave_dev; | |
547942ca | 1565 | /* Set the new_slave's queue_id to be zero. Queue ID mapping |
bb1d9123 AG |
1566 | * is set via sysfs or module option if desired. |
1567 | */ | |
1568 | new_slave->queue_id = 0; | |
1569 | ||
b15ba0fb JP |
1570 | /* Save slave's original mtu and then set it to match the bond */ |
1571 | new_slave->original_mtu = slave_dev->mtu; | |
1572 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1573 | if (res) { | |
e2a7420d | 1574 | slave_err(bond_dev, slave_dev, "Error %d calling dev_set_mtu\n", res); |
b15ba0fb JP |
1575 | goto err_free; |
1576 | } | |
1577 | ||
547942ca | 1578 | /* Save slave's original ("permanent") mac address for modes |
217df670 JV |
1579 | * that need it, and for restoring it upon release, and then |
1580 | * set it to the master's address | |
1581 | */ | |
faeeb317 JW |
1582 | bond_hw_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr, |
1583 | slave_dev->addr_len); | |
1da177e4 | 1584 | |
00503b6f | 1585 | if (!bond->params.fail_over_mac || |
01844098 | 1586 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
547942ca | 1587 | /* Set slave to master's mac address. The application already |
2ab82852 MS |
1588 | * set the master's mac address to that of the first slave |
1589 | */ | |
faeeb317 JW |
1590 | memcpy(ss.__data, bond_dev->dev_addr, bond_dev->addr_len); |
1591 | ss.ss_family = slave_dev->type; | |
3a37a963 PM |
1592 | res = dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, |
1593 | extack); | |
2ab82852 | 1594 | if (res) { |
e2a7420d | 1595 | slave_err(bond_dev, slave_dev, "Error %d calling set_mac_address\n", res); |
b15ba0fb | 1596 | goto err_restore_mtu; |
2ab82852 | 1597 | } |
217df670 | 1598 | } |
1da177e4 | 1599 | |
03d84a5f KH |
1600 | /* set slave flag before open to prevent IPv6 addrconf */ |
1601 | slave_dev->flags |= IFF_SLAVE; | |
1602 | ||
217df670 | 1603 | /* open the slave since the application closed it */ |
00f54e68 | 1604 | res = dev_open(slave_dev, extack); |
217df670 | 1605 | if (res) { |
e2a7420d | 1606 | slave_err(bond_dev, slave_dev, "Opening slave failed\n"); |
1f718f0f | 1607 | goto err_restore_mac; |
1da177e4 LT |
1608 | } |
1609 | ||
0b680e75 | 1610 | slave_dev->priv_flags |= IFF_BONDING; |
5f0c5f73 AG |
1611 | /* initialize slave stats */ |
1612 | dev_get_stats(new_slave->dev, &new_slave->slave_stats); | |
1da177e4 | 1613 | |
58402054 | 1614 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1615 | /* bond_alb_init_slave() must be called before all other stages since |
1616 | * it might fail and we do not want to have to undo everything | |
1617 | */ | |
1618 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1619 | if (res) |
569f0c4d | 1620 | goto err_close; |
1da177e4 LT |
1621 | } |
1622 | ||
b8e2fde4 WY |
1623 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1624 | if (res) { | |
e2a7420d | 1625 | slave_err(bond_dev, slave_dev, "Couldn't add bond vlan ids\n"); |
ae42cc62 | 1626 | goto err_close; |
1ff412ad | 1627 | } |
1da177e4 | 1628 | |
c8c23903 | 1629 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1630 | |
1631 | new_slave->delay = 0; | |
1632 | new_slave->link_failure_count = 0; | |
1633 | ||
ad729bc9 AB |
1634 | if (bond_update_speed_duplex(new_slave) && |
1635 | bond_needs_speed_duplex(bond)) | |
3f3c278c | 1636 | new_slave->link = BOND_LINK_DOWN; |
876254ae | 1637 | |
49f17de7 | 1638 | new_slave->last_rx = jiffies - |
f31c7937 | 1639 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1640 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1641 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1642 | |
1da177e4 LT |
1643 | if (bond->params.miimon && !bond->params.use_carrier) { |
1644 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1645 | ||
1646 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
547942ca | 1647 | /* miimon is set but a bonded network driver |
1da177e4 LT |
1648 | * does not support ETHTOOL/MII and |
1649 | * arp_interval is not set. Note: if | |
1650 | * use_carrier is enabled, we will never go | |
1651 | * here (because netif_carrier is always | |
1652 | * supported); thus, we don't need to change | |
1653 | * the messages for netif_carrier. | |
1654 | */ | |
e2a7420d | 1655 | slave_warn(bond_dev, slave_dev, "MII and ETHTOOL support not available for slave, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n"); |
1da177e4 LT |
1656 | } else if (link_reporting == -1) { |
1657 | /* unable get link status using mii/ethtool */ | |
e2a7420d | 1658 | slave_warn(bond_dev, slave_dev, "can't get link status from slave; 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"); |
1da177e4 LT |
1659 | } |
1660 | } | |
1661 | ||
1662 | /* check for initial state */ | |
a30b0168 | 1663 | new_slave->link = BOND_LINK_NOCHANGE; |
f31c7937 MK |
1664 | if (bond->params.miimon) { |
1665 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1666 | if (bond->params.updelay) { | |
69a2338e | 1667 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1668 | BOND_LINK_BACK, |
1669 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 MK |
1670 | new_slave->delay = bond->params.updelay; |
1671 | } else { | |
69a2338e | 1672 | bond_set_slave_link_state(new_slave, |
5d397061 JP |
1673 | BOND_LINK_UP, |
1674 | BOND_SLAVE_NOTIFY_NOW); | |
f31c7937 | 1675 | } |
1da177e4 | 1676 | } else { |
5d397061 JP |
1677 | bond_set_slave_link_state(new_slave, BOND_LINK_DOWN, |
1678 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1679 | } |
f31c7937 | 1680 | } else if (bond->params.arp_interval) { |
69a2338e MS |
1681 | bond_set_slave_link_state(new_slave, |
1682 | (netif_carrier_ok(slave_dev) ? | |
5d397061 JP |
1683 | BOND_LINK_UP : BOND_LINK_DOWN), |
1684 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 1685 | } else { |
5d397061 JP |
1686 | bond_set_slave_link_state(new_slave, BOND_LINK_UP, |
1687 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1688 | } |
1689 | ||
f31c7937 | 1690 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 1691 | new_slave->last_link_up = jiffies; |
e2a7420d JW |
1692 | slave_dbg(bond_dev, slave_dev, "Initial state of slave is BOND_LINK_%s\n", |
1693 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1694 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 1695 | |
ec0865a9 | 1696 | if (bond_uses_primary(bond) && bond->params.primary[0]) { |
1da177e4 | 1697 | /* if there is a primary slave, remember it */ |
a549952a | 1698 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
059b47e8 | 1699 | rcu_assign_pointer(bond->primary_slave, new_slave); |
a549952a JP |
1700 | bond->force_primary = true; |
1701 | } | |
1da177e4 LT |
1702 | } |
1703 | ||
01844098 | 1704 | switch (BOND_MODE(bond)) { |
1da177e4 | 1705 | case BOND_MODE_ACTIVEBACKUP: |
5e5b0665 | 1706 | bond_set_slave_inactive_flags(new_slave, |
1707 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1708 | break; |
1709 | case BOND_MODE_8023AD: | |
1710 | /* in 802.3ad mode, the internal mechanism | |
1711 | * will activate the slaves in the selected | |
1712 | * aggregator | |
1713 | */ | |
5e5b0665 | 1714 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 1715 | /* if this is the first slave */ |
23c147e0 | 1716 | if (!prev_slave) { |
3fdddd85 | 1717 | SLAVE_AD_INFO(new_slave)->id = 1; |
1da177e4 LT |
1718 | /* Initialize AD with the number of times that the AD timer is called in 1 second |
1719 | * can be called only after the mac address of the bond is set | |
1720 | */ | |
56d00c67 | 1721 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1722 | } else { |
3fdddd85 | 1723 | SLAVE_AD_INFO(new_slave)->id = |
1724 | SLAVE_AD_INFO(prev_slave)->id + 1; | |
1da177e4 LT |
1725 | } |
1726 | ||
1727 | bond_3ad_bind_slave(new_slave); | |
1728 | break; | |
1729 | case BOND_MODE_TLB: | |
1730 | case BOND_MODE_ALB: | |
e30bc066 | 1731 | bond_set_active_slave(new_slave); |
5e5b0665 | 1732 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
1733 | break; |
1734 | default: | |
e2a7420d | 1735 | slave_dbg(bond_dev, slave_dev, "This slave is always active in trunk mode\n"); |
1da177e4 LT |
1736 | |
1737 | /* always active in trunk mode */ | |
e30bc066 | 1738 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1739 | |
1740 | /* In trunking mode there is little meaning to curr_active_slave | |
1741 | * anyway (it holds no special properties of the bond device), | |
1742 | * so we can change it without calling change_active_interface() | |
1743 | */ | |
4740d638 ED |
1744 | if (!rcu_access_pointer(bond->curr_active_slave) && |
1745 | new_slave->link == BOND_LINK_UP) | |
278b2083 | 1746 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1747 | |
1da177e4 LT |
1748 | break; |
1749 | } /* switch(bond_mode) */ | |
1750 | ||
f6dc31a8 | 1751 | #ifdef CONFIG_NET_POLL_CONTROLLER |
ddea788c | 1752 | if (bond->dev->npinfo) { |
8a8efa22 | 1753 | if (slave_enable_netpoll(new_slave)) { |
e2a7420d | 1754 | slave_info(bond_dev, slave_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n"); |
8a8efa22 | 1755 | res = -EBUSY; |
f7d9821a | 1756 | goto err_detach; |
8a8efa22 | 1757 | } |
f6dc31a8 WC |
1758 | } |
1759 | #endif | |
8a8efa22 | 1760 | |
fbe168ba MK |
1761 | if (!(bond_dev->features & NETIF_F_LRO)) |
1762 | dev_disable_lro(slave_dev); | |
1763 | ||
35d48903 JP |
1764 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1765 | new_slave); | |
1766 | if (res) { | |
e2a7420d | 1767 | slave_dbg(bond_dev, slave_dev, "Error %d calling netdev_rx_handler_register\n", res); |
5831d66e | 1768 | goto err_detach; |
35d48903 JP |
1769 | } |
1770 | ||
42ab19ee | 1771 | res = bond_master_upper_dev_link(bond, new_slave, extack); |
1f718f0f | 1772 | if (res) { |
e2a7420d | 1773 | slave_dbg(bond_dev, slave_dev, "Error %d calling bond_master_upper_dev_link\n", res); |
1f718f0f VF |
1774 | goto err_unregister; |
1775 | } | |
1776 | ||
07699f9a | 1777 | res = bond_sysfs_slave_add(new_slave); |
1778 | if (res) { | |
e2a7420d | 1779 | slave_dbg(bond_dev, slave_dev, "Error %d calling bond_sysfs_slave_add\n", res); |
07699f9a | 1780 | goto err_upper_unlink; |
1781 | } | |
1782 | ||
ae42cc62 XL |
1783 | /* If the mode uses primary, then the following is handled by |
1784 | * bond_change_active_slave(). | |
1785 | */ | |
1786 | if (!bond_uses_primary(bond)) { | |
1787 | /* set promiscuity level to new slave */ | |
1788 | if (bond_dev->flags & IFF_PROMISC) { | |
1789 | res = dev_set_promiscuity(slave_dev, 1); | |
1790 | if (res) | |
1791 | goto err_sysfs_del; | |
1792 | } | |
1793 | ||
1794 | /* set allmulti level to new slave */ | |
1795 | if (bond_dev->flags & IFF_ALLMULTI) { | |
1796 | res = dev_set_allmulti(slave_dev, 1); | |
9f5a90c1 XL |
1797 | if (res) { |
1798 | if (bond_dev->flags & IFF_PROMISC) | |
1799 | dev_set_promiscuity(slave_dev, -1); | |
ae42cc62 | 1800 | goto err_sysfs_del; |
9f5a90c1 | 1801 | } |
ae42cc62 XL |
1802 | } |
1803 | ||
1804 | netif_addr_lock_bh(bond_dev); | |
1805 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1806 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1807 | netif_addr_unlock_bh(bond_dev); | |
1808 | ||
1809 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
1810 | /* add lacpdu mc addr to mc list */ | |
1811 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1812 | ||
1813 | dev_mc_add(slave_dev, lacpdu_multicast); | |
1814 | } | |
1815 | } | |
1816 | ||
5378c2e6 VF |
1817 | bond->slave_cnt++; |
1818 | bond_compute_features(bond); | |
1819 | bond_set_carrier(bond); | |
1820 | ||
ec0865a9 | 1821 | if (bond_uses_primary(bond)) { |
f80889a5 | 1822 | block_netpoll_tx(); |
5378c2e6 | 1823 | bond_select_active_slave(bond); |
f80889a5 | 1824 | unblock_netpoll_tx(); |
5378c2e6 | 1825 | } |
1f718f0f | 1826 | |
e79c1055 | 1827 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 MB |
1828 | bond_update_slave_arr(bond, NULL); |
1829 | ||
21706ee8 | 1830 | |
e2a7420d JW |
1831 | slave_info(bond_dev, slave_dev, "Enslaving as %s interface with %s link\n", |
1832 | bond_is_active_slave(new_slave) ? "an active" : "a backup", | |
1833 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
1834 | |
1835 | /* enslave is successful */ | |
69e61133 | 1836 | bond_queue_slave_event(new_slave); |
1da177e4 LT |
1837 | return 0; |
1838 | ||
1839 | /* Undo stages on error */ | |
ae42cc62 XL |
1840 | err_sysfs_del: |
1841 | bond_sysfs_slave_del(new_slave); | |
1842 | ||
07699f9a | 1843 | err_upper_unlink: |
41f0b049 | 1844 | bond_upper_dev_unlink(bond, new_slave); |
07699f9a | 1845 | |
1f718f0f VF |
1846 | err_unregister: |
1847 | netdev_rx_handler_unregister(slave_dev); | |
1848 | ||
f7d9821a | 1849 | err_detach: |
1ff412ad | 1850 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
059b47e8 NA |
1851 | if (rcu_access_pointer(bond->primary_slave) == new_slave) |
1852 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
4740d638 | 1853 | if (rcu_access_pointer(bond->curr_active_slave) == new_slave) { |
f80889a5 | 1854 | block_netpoll_tx(); |
c8517035 | 1855 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1856 | bond_select_active_slave(bond); |
f80889a5 | 1857 | unblock_netpoll_tx(); |
3c5913b5 | 1858 | } |
059b47e8 NA |
1859 | /* either primary_slave or curr_active_slave might've changed */ |
1860 | synchronize_rcu(); | |
fc7a72ac | 1861 | slave_disable_netpoll(new_slave); |
f7d9821a | 1862 | |
1da177e4 | 1863 | err_close: |
65de65d9 TY |
1864 | if (!netif_is_bond_master(slave_dev)) |
1865 | slave_dev->priv_flags &= ~IFF_BONDING; | |
1da177e4 LT |
1866 | dev_close(slave_dev); |
1867 | ||
1868 | err_restore_mac: | |
03d84a5f | 1869 | slave_dev->flags &= ~IFF_SLAVE; |
00503b6f | 1870 | if (!bond->params.fail_over_mac || |
01844098 | 1871 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
3915c1e8 JV |
1872 | /* XXX TODO - fom follow mode needs to change master's |
1873 | * MAC if this slave's MAC is in use by the bond, or at | |
1874 | * least print a warning. | |
1875 | */ | |
faeeb317 JW |
1876 | bond_hw_addr_copy(ss.__data, new_slave->perm_hwaddr, |
1877 | new_slave->dev->addr_len); | |
1878 | ss.ss_family = slave_dev->type; | |
3a37a963 | 1879 | dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, NULL); |
2ab82852 | 1880 | } |
1da177e4 | 1881 | |
b15ba0fb JP |
1882 | err_restore_mtu: |
1883 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1884 | ||
1da177e4 | 1885 | err_free: |
3fdddd85 | 1886 | bond_free_slave(new_slave); |
1da177e4 LT |
1887 | |
1888 | err_undo_flags: | |
b8fad459 | 1889 | /* Enslave of first slave has failed and we need to fix master's mac */ |
7d5cd2ce NA |
1890 | if (!bond_has_slaves(bond)) { |
1891 | if (ether_addr_equal_64bits(bond_dev->dev_addr, | |
1892 | slave_dev->dev_addr)) | |
1893 | eth_hw_addr_random(bond_dev); | |
1894 | if (bond_dev->type != ARPHRD_ETHER) { | |
40baec22 | 1895 | dev_close(bond_dev); |
7d5cd2ce NA |
1896 | ether_setup(bond_dev); |
1897 | bond_dev->flags |= IFF_MASTER; | |
1898 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; | |
1899 | } | |
1900 | } | |
3d632c3f | 1901 | |
1da177e4 LT |
1902 | return res; |
1903 | } | |
1904 | ||
547942ca | 1905 | /* Try to release the slave device <slave> from the bond device <master> |
1da177e4 | 1906 | * It is legal to access curr_active_slave without a lock because all the function |
8c0bc550 | 1907 | * is RTNL-locked. If "all" is true it means that the function is being called |
0896341a | 1908 | * while destroying a bond interface and all slaves are being released. |
1da177e4 LT |
1909 | * |
1910 | * The rules for slave state should be: | |
1911 | * for Active/Backup: | |
1912 | * Active stays on all backups go down | |
1913 | * for Bonded connections: | |
1914 | * The first up interface should be left on and all others downed. | |
1915 | */ | |
0896341a | 1916 | static int __bond_release_one(struct net_device *bond_dev, |
1917 | struct net_device *slave_dev, | |
f51048c3 | 1918 | bool all, bool unregister) |
1da177e4 | 1919 | { |
454d7c9b | 1920 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 1921 | struct slave *slave, *oldcurrent; |
faeeb317 | 1922 | struct sockaddr_storage ss; |
5a0068de | 1923 | int old_flags = bond_dev->flags; |
c8f44aff | 1924 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1925 | |
1926 | /* slave is not a slave or master is not master of this slave */ | |
1927 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1928 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
e2a7420d | 1929 | slave_dbg(bond_dev, slave_dev, "cannot release slave\n"); |
1da177e4 LT |
1930 | return -EINVAL; |
1931 | } | |
1932 | ||
e843fa50 | 1933 | block_netpoll_tx(); |
1da177e4 LT |
1934 | |
1935 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1936 | if (!slave) { | |
1937 | /* not a slave of this bond */ | |
e2a7420d | 1938 | slave_info(bond_dev, slave_dev, "interface not enslaved\n"); |
e843fa50 | 1939 | unblock_netpoll_tx(); |
1da177e4 LT |
1940 | return -EINVAL; |
1941 | } | |
1942 | ||
57beaca8 JP |
1943 | bond_set_slave_inactive_flags(slave, BOND_SLAVE_NOTIFY_NOW); |
1944 | ||
07699f9a | 1945 | bond_sysfs_slave_del(slave); |
1946 | ||
5f0c5f73 AG |
1947 | /* recompute stats just before removing the slave */ |
1948 | bond_get_stats(bond->dev, &bond->bond_stats); | |
1949 | ||
41f0b049 | 1950 | bond_upper_dev_unlink(bond, slave); |
35d48903 JP |
1951 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1952 | * for this slave anymore. | |
1953 | */ | |
1954 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 | 1955 | |
e470259f | 1956 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 1957 | bond_3ad_unbind_slave(slave); |
1da177e4 | 1958 | |
e79c1055 | 1959 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 MB |
1960 | bond_update_slave_arr(bond, slave); |
1961 | ||
e2a7420d JW |
1962 | slave_info(bond_dev, slave_dev, "Releasing %s interface\n", |
1963 | bond_is_active_slave(slave) ? "active" : "backup"); | |
1da177e4 | 1964 | |
4740d638 | 1965 | oldcurrent = rcu_access_pointer(bond->curr_active_slave); |
1da177e4 | 1966 | |
85741718 | 1967 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
1da177e4 | 1968 | |
00503b6f | 1969 | if (!all && (!bond->params.fail_over_mac || |
01844098 | 1970 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) { |
844223ab | 1971 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1972 | bond_has_slaves(bond)) |
e2a7420d JW |
1973 | slave_warn(bond_dev, slave_dev, "the permanent HWaddr of slave - %pM - is still in use by bond - set the HWaddr of slave to a different address to avoid conflicts\n", |
1974 | slave->perm_hwaddr); | |
dec1e90e | 1975 | } |
1976 | ||
059b47e8 NA |
1977 | if (rtnl_dereference(bond->primary_slave) == slave) |
1978 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
1da177e4 | 1979 | |
1c72cfdc | 1980 | if (oldcurrent == slave) |
1da177e4 | 1981 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1982 | |
58402054 | 1983 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1984 | /* Must be called only after the slave has been |
1985 | * detached from the list and the curr_active_slave | |
1986 | * has been cleared (if our_slave == old_current), | |
1987 | * but before a new active slave is selected. | |
1988 | */ | |
1989 | bond_alb_deinit_slave(bond, slave); | |
1990 | } | |
1991 | ||
0896341a | 1992 | if (all) { |
36708b89 | 1993 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 1994 | } else if (oldcurrent == slave) { |
547942ca | 1995 | /* Note that we hold RTNL over this sequence, so there |
059fe7a5 JV |
1996 | * is no concern that another slave add/remove event |
1997 | * will interfere. | |
1998 | */ | |
1da177e4 | 1999 | bond_select_active_slave(bond); |
059fe7a5 JV |
2000 | } |
2001 | ||
0965a1f3 | 2002 | if (!bond_has_slaves(bond)) { |
ff59c456 | 2003 | bond_set_carrier(bond); |
409cc1f8 | 2004 | eth_hw_addr_random(bond_dev); |
1da177e4 LT |
2005 | } |
2006 | ||
e843fa50 | 2007 | unblock_netpoll_tx(); |
278b2083 | 2008 | synchronize_rcu(); |
ee6154e1 | 2009 | bond->slave_cnt--; |
1da177e4 | 2010 | |
0965a1f3 | 2011 | if (!bond_has_slaves(bond)) { |
2af73d4b | 2012 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
2013 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
2014 | } | |
2af73d4b | 2015 | |
b2a103e6 MM |
2016 | bond_compute_features(bond); |
2017 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2018 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
e2a7420d | 2019 | slave_info(bond_dev, slave_dev, "last VLAN challenged slave left bond - VLAN blocking is removed\n"); |
b2a103e6 | 2020 | |
1ff412ad | 2021 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 2022 | |
547942ca | 2023 | /* If the mode uses primary, then this case was handled above by |
303d1cbf | 2024 | * bond_change_active_slave(..., NULL) |
1da177e4 | 2025 | */ |
ec0865a9 | 2026 | if (!bond_uses_primary(bond)) { |
5a0068de NH |
2027 | /* unset promiscuity level from slave |
2028 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
2029 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
2030 | * value of that flag before that change, as that was the value | |
2031 | * when this slave was attached, so we cache at the start of the | |
2032 | * function and use it here. Same goes for ALLMULTI below | |
2033 | */ | |
2034 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 2035 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2036 | |
2037 | /* unset allmulti level from slave */ | |
5a0068de | 2038 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 2039 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 2040 | |
303d1cbf | 2041 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
2042 | } |
2043 | ||
8a8efa22 | 2044 | slave_disable_netpoll(slave); |
f6dc31a8 | 2045 | |
1da177e4 LT |
2046 | /* close slave before restoring its mac address */ |
2047 | dev_close(slave_dev); | |
2048 | ||
00503b6f | 2049 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
01844098 | 2050 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 | 2051 | /* restore original ("permanent") mac address */ |
faeeb317 JW |
2052 | bond_hw_addr_copy(ss.__data, slave->perm_hwaddr, |
2053 | slave->dev->addr_len); | |
2054 | ss.ss_family = slave_dev->type; | |
3a37a963 | 2055 | dev_set_mac_address(slave_dev, (struct sockaddr *)&ss, NULL); |
2ab82852 | 2056 | } |
1da177e4 | 2057 | |
f51048c3 WC |
2058 | if (unregister) |
2059 | __dev_set_mtu(slave_dev, slave->original_mtu); | |
2060 | else | |
2061 | dev_set_mtu(slave_dev, slave->original_mtu); | |
b15ba0fb | 2062 | |
65de65d9 TY |
2063 | if (!netif_is_bond_master(slave_dev)) |
2064 | slave_dev->priv_flags &= ~IFF_BONDING; | |
1da177e4 | 2065 | |
3fdddd85 | 2066 | bond_free_slave(slave); |
1da177e4 | 2067 | |
547942ca | 2068 | return 0; |
1da177e4 LT |
2069 | } |
2070 | ||
0896341a | 2071 | /* A wrapper used because of ndo_del_link */ |
2072 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
2073 | { | |
f51048c3 | 2074 | return __bond_release_one(bond_dev, slave_dev, false, false); |
0896341a | 2075 | } |
2076 | ||
547942ca NA |
2077 | /* First release a slave and then destroy the bond if no more slaves are left. |
2078 | * Must be under rtnl_lock when this function is called. | |
2079 | */ | |
e2a7420d JW |
2080 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2081 | struct net_device *slave_dev) | |
d90a162a | 2082 | { |
454d7c9b | 2083 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2084 | int ret; |
2085 | ||
f51048c3 | 2086 | ret = __bond_release_one(bond_dev, slave_dev, false, true); |
0965a1f3 | 2087 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 2088 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
e2a7420d | 2089 | netdev_info(bond_dev, "Destroying bond\n"); |
06f6d109 | 2090 | bond_remove_proc_entry(bond); |
9e71626c | 2091 | unregister_netdevice(bond_dev); |
d90a162a MS |
2092 | } |
2093 | return ret; | |
2094 | } | |
2095 | ||
3d67576d | 2096 | static void bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1da177e4 | 2097 | { |
454d7c9b | 2098 | struct bonding *bond = netdev_priv(bond_dev); |
69a2338e | 2099 | bond_fill_ifbond(bond, info); |
1da177e4 LT |
2100 | } |
2101 | ||
2102 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2103 | { | |
454d7c9b | 2104 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 2105 | struct list_head *iter; |
dec1e90e | 2106 | int i = 0, res = -ENODEV; |
1da177e4 | 2107 | struct slave *slave; |
1da177e4 | 2108 | |
9caff1e7 | 2109 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 2110 | if (i++ == (int)info->slave_id) { |
689c96cc | 2111 | res = 0; |
69a2338e | 2112 | bond_fill_ifslave(slave, info); |
1da177e4 LT |
2113 | break; |
2114 | } | |
2115 | } | |
1da177e4 | 2116 | |
689c96cc | 2117 | return res; |
1da177e4 LT |
2118 | } |
2119 | ||
2120 | /*-------------------------------- Monitoring -------------------------------*/ | |
2121 | ||
4740d638 | 2122 | /* called with rcu_read_lock() */ |
f0c76d61 JV |
2123 | static int bond_miimon_inspect(struct bonding *bond) |
2124 | { | |
dec1e90e | 2125 | int link_state, commit = 0; |
9caff1e7 | 2126 | struct list_head *iter; |
f0c76d61 | 2127 | struct slave *slave; |
41f89100 JP |
2128 | bool ignore_updelay; |
2129 | ||
4740d638 | 2130 | ignore_updelay = !rcu_dereference(bond->curr_active_slave); |
1da177e4 | 2131 | |
4cb4f97b | 2132 | bond_for_each_slave_rcu(bond, slave, iter) { |
1899bb32 | 2133 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); |
1da177e4 | 2134 | |
f0c76d61 | 2135 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2136 | |
2137 | switch (slave->link) { | |
f0c76d61 JV |
2138 | case BOND_LINK_UP: |
2139 | if (link_state) | |
2140 | continue; | |
1da177e4 | 2141 | |
de77ecd4 | 2142 | bond_propose_link_state(slave, BOND_LINK_FAIL); |
d94708a5 | 2143 | commit++; |
f0c76d61 JV |
2144 | slave->delay = bond->params.downdelay; |
2145 | if (slave->delay) { | |
e2a7420d JW |
2146 | slave_info(bond->dev, slave->dev, "link status down for %sinterface, disabling it in %d ms\n", |
2147 | (BOND_MODE(bond) == | |
2148 | BOND_MODE_ACTIVEBACKUP) ? | |
2149 | (bond_is_active_slave(slave) ? | |
2150 | "active " : "backup ") : "", | |
2151 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2152 | } |
f0c76d61 JV |
2153 | /*FALLTHRU*/ |
2154 | case BOND_LINK_FAIL: | |
2155 | if (link_state) { | |
547942ca | 2156 | /* recovered before downdelay expired */ |
de77ecd4 | 2157 | bond_propose_link_state(slave, BOND_LINK_UP); |
8e603460 | 2158 | slave->last_link_up = jiffies; |
e2a7420d JW |
2159 | slave_info(bond->dev, slave->dev, "link status up again after %d ms\n", |
2160 | (bond->params.downdelay - slave->delay) * | |
2161 | bond->params.miimon); | |
fb9eb899 | 2162 | commit++; |
f0c76d61 | 2163 | continue; |
1da177e4 | 2164 | } |
f0c76d61 JV |
2165 | |
2166 | if (slave->delay <= 0) { | |
1899bb32 | 2167 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
f0c76d61 JV |
2168 | commit++; |
2169 | continue; | |
1da177e4 | 2170 | } |
1da177e4 | 2171 | |
f0c76d61 JV |
2172 | slave->delay--; |
2173 | break; | |
2174 | ||
2175 | case BOND_LINK_DOWN: | |
2176 | if (!link_state) | |
2177 | continue; | |
2178 | ||
de77ecd4 | 2179 | bond_propose_link_state(slave, BOND_LINK_BACK); |
d94708a5 | 2180 | commit++; |
f0c76d61 JV |
2181 | slave->delay = bond->params.updelay; |
2182 | ||
2183 | if (slave->delay) { | |
e2a7420d JW |
2184 | slave_info(bond->dev, slave->dev, "link status up, enabling it in %d ms\n", |
2185 | ignore_updelay ? 0 : | |
2186 | bond->params.updelay * | |
2187 | bond->params.miimon); | |
f0c76d61 JV |
2188 | } |
2189 | /*FALLTHRU*/ | |
2190 | case BOND_LINK_BACK: | |
2191 | if (!link_state) { | |
de77ecd4 | 2192 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
e2a7420d JW |
2193 | slave_info(bond->dev, slave->dev, "link status down again after %d ms\n", |
2194 | (bond->params.updelay - slave->delay) * | |
2195 | bond->params.miimon); | |
fb9eb899 | 2196 | commit++; |
f0c76d61 JV |
2197 | continue; |
2198 | } | |
2199 | ||
41f89100 JP |
2200 | if (ignore_updelay) |
2201 | slave->delay = 0; | |
2202 | ||
f0c76d61 | 2203 | if (slave->delay <= 0) { |
1899bb32 | 2204 | bond_propose_link_state(slave, BOND_LINK_UP); |
f0c76d61 | 2205 | commit++; |
41f89100 | 2206 | ignore_updelay = false; |
f0c76d61 | 2207 | continue; |
1da177e4 | 2208 | } |
f0c76d61 JV |
2209 | |
2210 | slave->delay--; | |
1da177e4 | 2211 | break; |
f0c76d61 JV |
2212 | } |
2213 | } | |
1da177e4 | 2214 | |
f0c76d61 JV |
2215 | return commit; |
2216 | } | |
1da177e4 | 2217 | |
7e878b60 TZ |
2218 | static void bond_miimon_link_change(struct bonding *bond, |
2219 | struct slave *slave, | |
2220 | char link) | |
2221 | { | |
2222 | switch (BOND_MODE(bond)) { | |
2223 | case BOND_MODE_8023AD: | |
2224 | bond_3ad_handle_link_change(slave, link); | |
2225 | break; | |
2226 | case BOND_MODE_TLB: | |
2227 | case BOND_MODE_ALB: | |
2228 | bond_alb_handle_link_change(bond, slave, link); | |
2229 | break; | |
2230 | case BOND_MODE_XOR: | |
2231 | bond_update_slave_arr(bond, NULL); | |
2232 | break; | |
2233 | } | |
2234 | } | |
2235 | ||
f0c76d61 JV |
2236 | static void bond_miimon_commit(struct bonding *bond) |
2237 | { | |
9caff1e7 | 2238 | struct list_head *iter; |
059b47e8 | 2239 | struct slave *slave, *primary; |
f0c76d61 | 2240 | |
9caff1e7 | 2241 | bond_for_each_slave(bond, slave, iter) { |
1899bb32 | 2242 | switch (slave->link_new_state) { |
f0c76d61 | 2243 | case BOND_LINK_NOCHANGE: |
12185dfe TF |
2244 | /* For 802.3ad mode, check current slave speed and |
2245 | * duplex again in case its port was disabled after | |
2246 | * invalid speed/duplex reporting but recovered before | |
2247 | * link monitoring could make a decision on the actual | |
2248 | * link status | |
2249 | */ | |
2250 | if (BOND_MODE(bond) == BOND_MODE_8023AD && | |
2251 | slave->link == BOND_LINK_UP) | |
2252 | bond_3ad_adapter_speed_duplex_changed(slave); | |
f0c76d61 | 2253 | continue; |
1da177e4 | 2254 | |
f0c76d61 | 2255 | case BOND_LINK_UP: |
ad729bc9 AB |
2256 | if (bond_update_speed_duplex(slave) && |
2257 | bond_needs_speed_duplex(bond)) { | |
3f3c278c | 2258 | slave->link = BOND_LINK_DOWN; |
11e9d782 | 2259 | if (net_ratelimit()) |
e2a7420d JW |
2260 | slave_warn(bond->dev, slave->dev, |
2261 | "failed to get link speed/duplex\n"); | |
b5bf0f5b MB |
2262 | continue; |
2263 | } | |
5d397061 JP |
2264 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2265 | BOND_SLAVE_NOTIFY_NOW); | |
8e603460 | 2266 | slave->last_link_up = jiffies; |
f0c76d61 | 2267 | |
059b47e8 | 2268 | primary = rtnl_dereference(bond->primary_slave); |
01844098 | 2269 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
f0c76d61 | 2270 | /* prevent it from being the active one */ |
e30bc066 | 2271 | bond_set_backup_slave(slave); |
01844098 | 2272 | } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
f0c76d61 | 2273 | /* make it immediately active */ |
e30bc066 | 2274 | bond_set_active_slave(slave); |
059b47e8 | 2275 | } else if (slave != primary) { |
f0c76d61 | 2276 | /* prevent it from being the active one */ |
e30bc066 | 2277 | bond_set_backup_slave(slave); |
1da177e4 | 2278 | } |
1da177e4 | 2279 | |
e2a7420d JW |
2280 | slave_info(bond->dev, slave->dev, "link status definitely up, %u Mbps %s duplex\n", |
2281 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, | |
2282 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2283 | |
7e878b60 | 2284 | bond_miimon_link_change(bond, slave, BOND_LINK_UP); |
ee637714 | 2285 | |
059b47e8 | 2286 | if (!bond->curr_active_slave || slave == primary) |
f0c76d61 | 2287 | goto do_failover; |
1da177e4 | 2288 | |
f0c76d61 | 2289 | continue; |
059fe7a5 | 2290 | |
f0c76d61 | 2291 | case BOND_LINK_DOWN: |
fba4acda JV |
2292 | if (slave->link_failure_count < UINT_MAX) |
2293 | slave->link_failure_count++; | |
2294 | ||
5d397061 JP |
2295 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2296 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 2297 | |
01844098 VF |
2298 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || |
2299 | BOND_MODE(bond) == BOND_MODE_8023AD) | |
5e5b0665 | 2300 | bond_set_slave_inactive_flags(slave, |
2301 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2302 | |
e2a7420d | 2303 | slave_info(bond->dev, slave->dev, "link status definitely down, disabling slave\n"); |
f0c76d61 | 2304 | |
7e878b60 | 2305 | bond_miimon_link_change(bond, slave, BOND_LINK_DOWN); |
ee637714 | 2306 | |
4740d638 | 2307 | if (slave == rcu_access_pointer(bond->curr_active_slave)) |
f0c76d61 JV |
2308 | goto do_failover; |
2309 | ||
2310 | continue; | |
2311 | ||
2312 | default: | |
e2a7420d | 2313 | slave_err(bond->dev, slave->dev, "invalid new link %d on slave\n", |
1899bb32 JV |
2314 | slave->link_new_state); |
2315 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); | |
f0c76d61 JV |
2316 | |
2317 | continue; | |
2318 | } | |
2319 | ||
2320 | do_failover: | |
e843fa50 | 2321 | block_netpoll_tx(); |
f0c76d61 | 2322 | bond_select_active_slave(bond); |
e843fa50 | 2323 | unblock_netpoll_tx(); |
f0c76d61 JV |
2324 | } |
2325 | ||
2326 | bond_set_carrier(bond); | |
1da177e4 LT |
2327 | } |
2328 | ||
547942ca | 2329 | /* bond_mii_monitor |
0b0eef66 JV |
2330 | * |
2331 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2332 | * inspection, then (if inspection indicates something needs to be done) |
2333 | * an acquisition of appropriate locks followed by a commit phase to | |
2334 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2335 | */ |
6da67d26 | 2336 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2337 | { |
2338 | struct bonding *bond = container_of(work, struct bonding, | |
2339 | mii_work.work); | |
ad246c99 | 2340 | bool should_notify_peers = false; |
07a4ddec | 2341 | bool commit; |
1f2cd845 | 2342 | unsigned long delay; |
de77ecd4 MB |
2343 | struct slave *slave; |
2344 | struct list_head *iter; | |
0b0eef66 | 2345 | |
1f2cd845 DM |
2346 | delay = msecs_to_jiffies(bond->params.miimon); |
2347 | ||
2348 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2349 | goto re_arm; |
b59f9f74 | 2350 | |
4cb4f97b | 2351 | rcu_read_lock(); |
ad246c99 | 2352 | should_notify_peers = bond_should_notify_peers(bond); |
07a4ddec VB |
2353 | commit = !!bond_miimon_inspect(bond); |
2354 | if (bond->send_peer_notif) { | |
2355 | rcu_read_unlock(); | |
2356 | if (rtnl_trylock()) { | |
2357 | bond->send_peer_notif--; | |
2358 | rtnl_unlock(); | |
2359 | } | |
2360 | } else { | |
4cb4f97b | 2361 | rcu_read_unlock(); |
07a4ddec | 2362 | } |
f0c76d61 | 2363 | |
07a4ddec | 2364 | if (commit) { |
1f2cd845 DM |
2365 | /* Race avoidance with bond_close cancel of workqueue */ |
2366 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2367 | delay = 1; |
2368 | should_notify_peers = false; | |
2369 | goto re_arm; | |
2370 | } | |
6b6c5261 | 2371 | |
de77ecd4 MB |
2372 | bond_for_each_slave(bond, slave, iter) { |
2373 | bond_commit_link_state(slave, BOND_SLAVE_NOTIFY_LATER); | |
2374 | } | |
1f2cd845 DM |
2375 | bond_miimon_commit(bond); |
2376 | ||
1f2cd845 | 2377 | rtnl_unlock(); /* might sleep, hold no other locks */ |
07a4ddec | 2378 | } |
0b0eef66 | 2379 | |
f0c76d61 | 2380 | re_arm: |
e6d265e8 | 2381 | if (bond->params.miimon) |
1f2cd845 DM |
2382 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2383 | ||
1f2cd845 DM |
2384 | if (should_notify_peers) { |
2385 | if (!rtnl_trylock()) | |
2386 | return; | |
2387 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2388 | rtnl_unlock(); | |
2389 | } | |
0b0eef66 | 2390 | } |
c3ade5ca | 2391 | |
b3208b20 DA |
2392 | static int bond_upper_dev_walk(struct net_device *upper, void *data) |
2393 | { | |
2394 | __be32 ip = *((__be32 *)data); | |
2395 | ||
2396 | return ip == bond_confirm_addr(upper, 0, ip); | |
2397 | } | |
2398 | ||
50223ce4 | 2399 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2400 | { |
50223ce4 | 2401 | bool ret = false; |
f5b2b966 | 2402 | |
eaddcd76 | 2403 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2404 | return true; |
f5b2b966 | 2405 | |
50223ce4 | 2406 | rcu_read_lock(); |
b3208b20 DA |
2407 | if (netdev_walk_all_upper_dev_rcu(bond->dev, bond_upper_dev_walk, &ip)) |
2408 | ret = true; | |
50223ce4 | 2409 | rcu_read_unlock(); |
f5b2b966 | 2410 | |
50223ce4 | 2411 | return ret; |
f5b2b966 JV |
2412 | } |
2413 | ||
547942ca | 2414 | /* We go to the (large) trouble of VLAN tagging ARP frames because |
c3ade5ca JV |
2415 | * switches in VLAN mode (especially if ports are configured as |
2416 | * "native" to a VLAN) might not pass non-tagged frames. | |
2417 | */ | |
e2a7420d JW |
2418 | static void bond_arp_send(struct slave *slave, int arp_op, __be32 dest_ip, |
2419 | __be32 src_ip, struct bond_vlan_tag *tags) | |
c3ade5ca JV |
2420 | { |
2421 | struct sk_buff *skb; | |
3e403a77 | 2422 | struct bond_vlan_tag *outer_tag = tags; |
e2a7420d JW |
2423 | struct net_device *slave_dev = slave->dev; |
2424 | struct net_device *bond_dev = slave->bond->dev; | |
c3ade5ca | 2425 | |
e2a7420d JW |
2426 | slave_dbg(bond_dev, slave_dev, "arp %d on slave: dst %pI4 src %pI4\n", |
2427 | arp_op, &dest_ip, &src_ip); | |
3d632c3f | 2428 | |
c3ade5ca JV |
2429 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2430 | NULL, slave_dev->dev_addr, NULL); | |
2431 | ||
2432 | if (!skb) { | |
4873ac3c | 2433 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2434 | return; |
2435 | } | |
fbd929f2 | 2436 | |
3e403a77 VF |
2437 | if (!tags || tags->vlan_proto == VLAN_N_VID) |
2438 | goto xmit; | |
2439 | ||
2440 | tags++; | |
2441 | ||
44a40855 | 2442 | /* Go through all the tags backwards and add them to the packet */ |
3e403a77 VF |
2443 | while (tags->vlan_proto != VLAN_N_VID) { |
2444 | if (!tags->vlan_id) { | |
2445 | tags++; | |
44a40855 | 2446 | continue; |
3e403a77 | 2447 | } |
44a40855 | 2448 | |
e2a7420d JW |
2449 | slave_dbg(bond_dev, slave_dev, "inner tag: proto %X vid %X\n", |
2450 | ntohs(outer_tag->vlan_proto), tags->vlan_id); | |
62749e2c JP |
2451 | skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto, |
2452 | tags->vlan_id); | |
44a40855 VY |
2453 | if (!skb) { |
2454 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2455 | return; | |
2456 | } | |
3e403a77 VF |
2457 | |
2458 | tags++; | |
44a40855 VY |
2459 | } |
2460 | /* Set the outer tag */ | |
3e403a77 | 2461 | if (outer_tag->vlan_id) { |
e2a7420d JW |
2462 | slave_dbg(bond_dev, slave_dev, "outer tag: proto %X vid %X\n", |
2463 | ntohs(outer_tag->vlan_proto), outer_tag->vlan_id); | |
b4bef1b5 JP |
2464 | __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto, |
2465 | outer_tag->vlan_id); | |
c3ade5ca | 2466 | } |
3e403a77 VF |
2467 | |
2468 | xmit: | |
c3ade5ca JV |
2469 | arp_xmit(skb); |
2470 | } | |
2471 | ||
44a40855 VY |
2472 | /* Validate the device path between the @start_dev and the @end_dev. |
2473 | * The path is valid if the @end_dev is reachable through device | |
2474 | * stacking. | |
2475 | * When the path is validated, collect any vlan information in the | |
2476 | * path. | |
2477 | */ | |
3e403a77 VF |
2478 | struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, |
2479 | struct net_device *end_dev, | |
2480 | int level) | |
44a40855 | 2481 | { |
3e403a77 | 2482 | struct bond_vlan_tag *tags; |
44a40855 VY |
2483 | struct net_device *upper; |
2484 | struct list_head *iter; | |
44a40855 | 2485 | |
3e403a77 | 2486 | if (start_dev == end_dev) { |
6396bb22 | 2487 | tags = kcalloc(level + 1, sizeof(*tags), GFP_ATOMIC); |
3e403a77 VF |
2488 | if (!tags) |
2489 | return ERR_PTR(-ENOMEM); | |
2490 | tags[level].vlan_proto = VLAN_N_VID; | |
2491 | return tags; | |
2492 | } | |
44a40855 VY |
2493 | |
2494 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
3e403a77 VF |
2495 | tags = bond_verify_device_path(upper, end_dev, level + 1); |
2496 | if (IS_ERR_OR_NULL(tags)) { | |
2497 | if (IS_ERR(tags)) | |
2498 | return tags; | |
2499 | continue; | |
44a40855 | 2500 | } |
3e403a77 VF |
2501 | if (is_vlan_dev(upper)) { |
2502 | tags[level].vlan_proto = vlan_dev_vlan_proto(upper); | |
2503 | tags[level].vlan_id = vlan_dev_vlan_id(upper); | |
2504 | } | |
2505 | ||
2506 | return tags; | |
44a40855 VY |
2507 | } |
2508 | ||
3e403a77 | 2509 | return NULL; |
44a40855 | 2510 | } |
c3ade5ca | 2511 | |
1da177e4 LT |
2512 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2513 | { | |
c3ade5ca | 2514 | struct rtable *rt; |
3e403a77 | 2515 | struct bond_vlan_tag *tags; |
27bc11e6 | 2516 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2517 | int i; |
1da177e4 | 2518 | |
27bc11e6 | 2519 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
e2a7420d JW |
2520 | slave_dbg(bond->dev, slave->dev, "%s: target %pI4\n", |
2521 | __func__, &targets[i]); | |
3e403a77 | 2522 | tags = NULL; |
c3ade5ca | 2523 | |
27bc11e6 | 2524 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2525 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2526 | RTO_ONLINK, 0); | |
b23dd4fe | 2527 | if (IS_ERR(rt)) { |
28572760 VF |
2528 | /* there's no route to target - try to send arp |
2529 | * probe to generate any traffic (arp_validate=0) | |
2530 | */ | |
4873ac3c | 2531 | if (bond->params.arp_validate) |
2532 | net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", | |
2533 | bond->dev->name, | |
2534 | &targets[i]); | |
e2a7420d | 2535 | bond_arp_send(slave, ARPOP_REQUEST, targets[i], |
44a40855 | 2536 | 0, tags); |
c3ade5ca JV |
2537 | continue; |
2538 | } | |
2539 | ||
27bc11e6 VF |
2540 | /* bond device itself */ |
2541 | if (rt->dst.dev == bond->dev) | |
2542 | goto found; | |
2543 | ||
2544 | rcu_read_lock(); | |
3e403a77 | 2545 | tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0); |
27bc11e6 | 2546 | rcu_read_unlock(); |
c3ade5ca | 2547 | |
3e403a77 | 2548 | if (!IS_ERR_OR_NULL(tags)) |
44a40855 VY |
2549 | goto found; |
2550 | ||
27bc11e6 | 2551 | /* Not our device - skip */ |
e2a7420d | 2552 | slave_dbg(bond->dev, slave->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n", |
76444f50 | 2553 | &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL"); |
3e32582f | 2554 | |
ed4b9f80 | 2555 | ip_rt_put(rt); |
27bc11e6 VF |
2556 | continue; |
2557 | ||
2558 | found: | |
2559 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2560 | ip_rt_put(rt); | |
e2a7420d | 2561 | bond_arp_send(slave, ARPOP_REQUEST, targets[i], addr, tags); |
a67eed57 | 2562 | kfree(tags); |
c3ade5ca JV |
2563 | } |
2564 | } | |
2565 | ||
d3bb52b0 | 2566 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2567 | { |
8599b52e VF |
2568 | int i; |
2569 | ||
87a7b84b | 2570 | if (!sip || !bond_has_this_ip(bond, tip)) { |
e2a7420d JW |
2571 | slave_dbg(bond->dev, slave->dev, "%s: sip %pI4 tip %pI4 not found\n", |
2572 | __func__, &sip, &tip); | |
87a7b84b VF |
2573 | return; |
2574 | } | |
f5b2b966 | 2575 | |
8599b52e VF |
2576 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2577 | if (i == -1) { | |
e2a7420d JW |
2578 | slave_dbg(bond->dev, slave->dev, "%s: sip %pI4 not found in targets\n", |
2579 | __func__, &sip); | |
87a7b84b | 2580 | return; |
f5b2b966 | 2581 | } |
49f17de7 | 2582 | slave->last_rx = jiffies; |
8599b52e | 2583 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2584 | } |
2585 | ||
5bb9e0b5 | 2586 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2587 | struct slave *slave) | |
f5b2b966 | 2588 | { |
de063b70 | 2589 | struct arphdr *arp = (struct arphdr *)skb->data; |
21a75f09 | 2590 | struct slave *curr_active_slave, *curr_arp_slave; |
f5b2b966 | 2591 | unsigned char *arp_ptr; |
d3bb52b0 | 2592 | __be32 sip, tip; |
6ade97da AD |
2593 | int is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
2594 | unsigned int alen; | |
f5b2b966 | 2595 | |
f2cb691a | 2596 | if (!slave_do_arp_validate(bond, slave)) { |
bedabf90 | 2597 | if ((slave_do_arp_validate_only(bond) && is_arp) || |
2598 | !slave_do_arp_validate_only(bond)) | |
49f17de7 | 2599 | slave->last_rx = jiffies; |
13a8e0c8 | 2600 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2601 | } else if (!is_arp) { |
2602 | return RX_HANDLER_ANOTHER; | |
2603 | } | |
2c146102 | 2604 | |
de063b70 | 2605 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2606 | |
e2a7420d JW |
2607 | slave_dbg(bond->dev, slave->dev, "%s: skb->dev %s\n", |
2608 | __func__, skb->dev->name); | |
f5b2b966 | 2609 | |
de063b70 ED |
2610 | if (alen > skb_headlen(skb)) { |
2611 | arp = kmalloc(alen, GFP_ATOMIC); | |
2612 | if (!arp) | |
2613 | goto out_unlock; | |
2614 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2615 | goto out_unlock; | |
2616 | } | |
f5b2b966 | 2617 | |
3aba891d | 2618 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2619 | skb->pkt_type == PACKET_OTHERHOST || |
2620 | skb->pkt_type == PACKET_LOOPBACK || | |
2621 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2622 | arp->ar_pro != htons(ETH_P_IP) || | |
2623 | arp->ar_pln != 4) | |
2624 | goto out_unlock; | |
2625 | ||
2626 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2627 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2628 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2629 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2630 | memcpy(&tip, arp_ptr, 4); |
2631 | ||
e2a7420d JW |
2632 | slave_dbg(bond->dev, slave->dev, "%s: %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
2633 | __func__, slave->dev->name, bond_slave_state(slave), | |
2634 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2635 | &sip, &tip); | |
f5b2b966 | 2636 | |
010d3c39 | 2637 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
21a75f09 | 2638 | curr_arp_slave = rcu_dereference(bond->current_arp_slave); |
010d3c39 | 2639 | |
21a75f09 | 2640 | /* We 'trust' the received ARP enough to validate it if: |
aeea64ac | 2641 | * |
21a75f09 JV |
2642 | * (a) the slave receiving the ARP is active (which includes the |
2643 | * current ARP slave, if any), or | |
2644 | * | |
2645 | * (b) the receiving slave isn't active, but there is a currently | |
2646 | * active slave and it received valid arp reply(s) after it became | |
2647 | * the currently active slave, or | |
2648 | * | |
2649 | * (c) there is an ARP slave that sent an ARP during the prior ARP | |
2650 | * interval, and we receive an ARP reply on any slave. We accept | |
2651 | * these because switch FDB update delays may deliver the ARP | |
2652 | * reply to a slave other than the sender of the ARP request. | |
2653 | * | |
2654 | * Note: for (b), backup slaves are receiving the broadcast ARP | |
2655 | * request, not a reply. This request passes from the sending | |
2656 | * slave through the L2 switch(es) to the receiving slave. Since | |
2657 | * this is checking the request, sip/tip are swapped for | |
2658 | * validation. | |
2659 | * | |
2660 | * This is done to avoid endless looping when we can't reach the | |
aeea64ac | 2661 | * arp_ip_target and fool ourselves with our own arp requests. |
f5b2b966 | 2662 | */ |
e30bc066 | 2663 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2664 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
2665 | else if (curr_active_slave && |
2666 | time_after(slave_last_rx(bond, curr_active_slave), | |
2667 | curr_active_slave->last_link_up)) | |
f5b2b966 | 2668 | bond_validate_arp(bond, slave, tip, sip); |
21a75f09 JV |
2669 | else if (curr_arp_slave && (arp->ar_op == htons(ARPOP_REPLY)) && |
2670 | bond_time_in_interval(bond, | |
2671 | dev_trans_start(curr_arp_slave->dev), 1)) | |
2672 | bond_validate_arp(bond, slave, sip, tip); | |
f5b2b966 JV |
2673 | |
2674 | out_unlock: | |
de063b70 ED |
2675 | if (arp != (struct arphdr *)skb->data) |
2676 | kfree(arp); | |
13a8e0c8 | 2677 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2678 | } |
2679 | ||
e7f63f1d VF |
2680 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2681 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2682 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2683 | */ | |
2684 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2685 | int mod) | |
2686 | { | |
2687 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2688 | ||
2689 | return time_in_range(jiffies, | |
2690 | last_act - delta_in_ticks, | |
2691 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2692 | } | |
2693 | ||
547942ca | 2694 | /* This function is called regularly to monitor each slave's link |
1da177e4 LT |
2695 | * ensuring that traffic is being sent and received when arp monitoring |
2696 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2697 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2698 | * arp monitoring in active backup mode. | |
2699 | */ | |
d5e73f7b | 2700 | static void bond_loadbalance_arp_mon(struct bonding *bond) |
1da177e4 | 2701 | { |
1da177e4 | 2702 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2703 | struct list_head *iter; |
6fde8f03 | 2704 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 2705 | |
1f2cd845 | 2706 | if (!bond_has_slaves(bond)) |
1da177e4 | 2707 | goto re_arm; |
1da177e4 | 2708 | |
2e52f4fe | 2709 | rcu_read_lock(); |
2710 | ||
4740d638 | 2711 | oldcurrent = rcu_dereference(bond->curr_active_slave); |
1da177e4 LT |
2712 | /* see if any of the previous devices are up now (i.e. they have |
2713 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
2714 | * the picture unless it is null. also, slave->last_link_up is not |
2715 | * needed here because we send an arp on each slave and give a slave | |
2716 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
2717 | * TODO: what about up/down delay in arp mode? it wasn't here before |
2718 | * so it can wait | |
2719 | */ | |
2e52f4fe | 2720 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2721 | unsigned long trans_start = dev_trans_start(slave->dev); |
2722 | ||
1899bb32 | 2723 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); |
797a9364 | 2724 | |
1da177e4 | 2725 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2726 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 2727 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 | 2728 | |
1899bb32 | 2729 | bond_propose_link_state(slave, BOND_LINK_UP); |
6fde8f03 | 2730 | slave_state_changed = 1; |
1da177e4 LT |
2731 | |
2732 | /* primary_slave has no meaning in round-robin | |
2733 | * mode. the window of a slave being up and | |
2734 | * curr_active_slave being null after enslaving | |
2735 | * is closed. | |
2736 | */ | |
2737 | if (!oldcurrent) { | |
e2a7420d | 2738 | slave_info(bond->dev, slave->dev, "link status definitely up\n"); |
1da177e4 LT |
2739 | do_failover = 1; |
2740 | } else { | |
e2a7420d | 2741 | slave_info(bond->dev, slave->dev, "interface is now up\n"); |
1da177e4 LT |
2742 | } |
2743 | } | |
2744 | } else { | |
2745 | /* slave->link == BOND_LINK_UP */ | |
2746 | ||
2747 | /* not all switches will respond to an arp request | |
2748 | * when the source ip is 0, so don't take the link down | |
2749 | * if we don't know our ip yet | |
2750 | */ | |
e7f63f1d | 2751 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 2752 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 | 2753 | |
1899bb32 | 2754 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
6fde8f03 | 2755 | slave_state_changed = 1; |
1da177e4 | 2756 | |
3d632c3f | 2757 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2758 | slave->link_failure_count++; |
1da177e4 | 2759 | |
e2a7420d | 2760 | slave_info(bond->dev, slave->dev, "interface is now down\n"); |
1da177e4 | 2761 | |
3d632c3f | 2762 | if (slave == oldcurrent) |
1da177e4 | 2763 | do_failover = 1; |
1da177e4 LT |
2764 | } |
2765 | } | |
2766 | ||
2767 | /* note: if switch is in round-robin mode, all links | |
2768 | * must tx arp to ensure all links rx an arp - otherwise | |
2769 | * links may oscillate or not come up at all; if switch is | |
2770 | * in something like xor mode, there is nothing we can | |
2771 | * do - all replies will be rx'ed on same link causing slaves | |
2772 | * to be unstable during low/no traffic periods | |
2773 | */ | |
b6adc610 | 2774 | if (bond_slave_is_up(slave)) |
1da177e4 | 2775 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2776 | } |
2777 | ||
2e52f4fe | 2778 | rcu_read_unlock(); |
2779 | ||
6fde8f03 | 2780 | if (do_failover || slave_state_changed) { |
2e52f4fe | 2781 | if (!rtnl_trylock()) |
2782 | goto re_arm; | |
1da177e4 | 2783 | |
797a9364 | 2784 | bond_for_each_slave(bond, slave, iter) { |
1899bb32 JV |
2785 | if (slave->link_new_state != BOND_LINK_NOCHANGE) |
2786 | slave->link = slave->link_new_state; | |
797a9364 NS |
2787 | } |
2788 | ||
6fde8f03 DT |
2789 | if (slave_state_changed) { |
2790 | bond_slave_state_change(bond); | |
ee637714 MB |
2791 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2792 | bond_update_slave_arr(bond, NULL); | |
b8e4500f NA |
2793 | } |
2794 | if (do_failover) { | |
6fde8f03 | 2795 | block_netpoll_tx(); |
6fde8f03 | 2796 | bond_select_active_slave(bond); |
6fde8f03 DT |
2797 | unblock_netpoll_tx(); |
2798 | } | |
2e52f4fe | 2799 | rtnl_unlock(); |
1da177e4 LT |
2800 | } |
2801 | ||
2802 | re_arm: | |
e6d265e8 | 2803 | if (bond->params.arp_interval) |
e7f63f1d VF |
2804 | queue_delayed_work(bond->wq, &bond->arp_work, |
2805 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2806 | } |
2807 | ||
547942ca | 2808 | /* Called to inspect slaves for active-backup mode ARP monitor link state |
1899bb32 JV |
2809 | * changes. Sets proposed link state in slaves to specify what action |
2810 | * should take place for the slave. Returns 0 if no changes are found, >0 | |
2811 | * if changes to link states must be committed. | |
b2220cad | 2812 | * |
8c0bc550 | 2813 | * Called with rcu_read_lock held. |
1da177e4 | 2814 | */ |
e7f63f1d | 2815 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2816 | { |
def4460c | 2817 | unsigned long trans_start, last_rx; |
9caff1e7 | 2818 | struct list_head *iter; |
dec1e90e | 2819 | struct slave *slave; |
dec1e90e | 2820 | int commit = 0; |
da210f55 | 2821 | |
eb9fa4b0 | 2822 | bond_for_each_slave_rcu(bond, slave, iter) { |
1899bb32 | 2823 | bond_propose_link_state(slave, BOND_LINK_NOCHANGE); |
def4460c | 2824 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2825 | |
b2220cad | 2826 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2827 | if (bond_time_in_interval(bond, last_rx, 1)) { |
1899bb32 | 2828 | bond_propose_link_state(slave, BOND_LINK_UP); |
b2220cad JV |
2829 | commit++; |
2830 | } | |
b2220cad JV |
2831 | continue; |
2832 | } | |
1da177e4 | 2833 | |
547942ca | 2834 | /* Give slaves 2*delta after being enslaved or made |
b2220cad JV |
2835 | * active. This avoids bouncing, as the last receive |
2836 | * times need a full ARP monitor cycle to be updated. | |
2837 | */ | |
8e603460 | 2838 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
2839 | continue; |
2840 | ||
547942ca | 2841 | /* Backup slave is down if: |
b2220cad JV |
2842 | * - No current_arp_slave AND |
2843 | * - more than 3*delta since last receive AND | |
2844 | * - the bond has an IP address | |
2845 | * | |
2846 | * Note: a non-null current_arp_slave indicates | |
2847 | * the curr_active_slave went down and we are | |
2848 | * searching for a new one; under this condition | |
2849 | * we only take the curr_active_slave down - this | |
2850 | * gives each slave a chance to tx/rx traffic | |
2851 | * before being taken out | |
2852 | */ | |
e30bc066 | 2853 | if (!bond_is_active_slave(slave) && |
85741718 | 2854 | !rcu_access_pointer(bond->current_arp_slave) && |
e7f63f1d | 2855 | !bond_time_in_interval(bond, last_rx, 3)) { |
1899bb32 | 2856 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
b2220cad JV |
2857 | commit++; |
2858 | } | |
2859 | ||
547942ca | 2860 | /* Active slave is down if: |
b2220cad JV |
2861 | * - more than 2*delta since transmitting OR |
2862 | * - (more than 2*delta since receive AND | |
2863 | * the bond has an IP address) | |
2864 | */ | |
cb32f2a0 | 2865 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2866 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2867 | (!bond_time_in_interval(bond, trans_start, 2) || |
2868 | !bond_time_in_interval(bond, last_rx, 2))) { | |
1899bb32 | 2869 | bond_propose_link_state(slave, BOND_LINK_DOWN); |
b2220cad JV |
2870 | commit++; |
2871 | } | |
1da177e4 LT |
2872 | } |
2873 | ||
b2220cad JV |
2874 | return commit; |
2875 | } | |
1da177e4 | 2876 | |
547942ca | 2877 | /* Called to commit link state changes noted by inspection step of |
b2220cad JV |
2878 | * active-backup mode ARP monitor. |
2879 | * | |
eb9fa4b0 | 2880 | * Called with RTNL hold. |
b2220cad | 2881 | */ |
e7f63f1d | 2882 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2883 | { |
cb32f2a0 | 2884 | unsigned long trans_start; |
9caff1e7 | 2885 | struct list_head *iter; |
dec1e90e | 2886 | struct slave *slave; |
1da177e4 | 2887 | |
9caff1e7 | 2888 | bond_for_each_slave(bond, slave, iter) { |
1899bb32 | 2889 | switch (slave->link_new_state) { |
b2220cad JV |
2890 | case BOND_LINK_NOCHANGE: |
2891 | continue; | |
ff59c456 | 2892 | |
b2220cad | 2893 | case BOND_LINK_UP: |
cb32f2a0 | 2894 | trans_start = dev_trans_start(slave->dev); |
4740d638 ED |
2895 | if (rtnl_dereference(bond->curr_active_slave) != slave || |
2896 | (!rtnl_dereference(bond->curr_active_slave) && | |
e7f63f1d | 2897 | bond_time_in_interval(bond, trans_start, 1))) { |
85741718 ED |
2898 | struct slave *current_arp_slave; |
2899 | ||
2900 | current_arp_slave = rtnl_dereference(bond->current_arp_slave); | |
5d397061 JP |
2901 | bond_set_slave_link_state(slave, BOND_LINK_UP, |
2902 | BOND_SLAVE_NOTIFY_NOW); | |
85741718 | 2903 | if (current_arp_slave) { |
5a430974 | 2904 | bond_set_slave_inactive_flags( |
85741718 | 2905 | current_arp_slave, |
5e5b0665 | 2906 | BOND_SLAVE_NOTIFY_NOW); |
85741718 | 2907 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
5a430974 | 2908 | } |
b2220cad | 2909 | |
e2a7420d | 2910 | slave_info(bond->dev, slave->dev, "link status definitely up\n"); |
b2220cad | 2911 | |
4740d638 | 2912 | if (!rtnl_dereference(bond->curr_active_slave) || |
059b47e8 | 2913 | slave == rtnl_dereference(bond->primary_slave)) |
b9f60253 | 2914 | goto do_failover; |
1da177e4 | 2915 | |
b9f60253 | 2916 | } |
1da177e4 | 2917 | |
b9f60253 | 2918 | continue; |
1da177e4 | 2919 | |
b2220cad JV |
2920 | case BOND_LINK_DOWN: |
2921 | if (slave->link_failure_count < UINT_MAX) | |
2922 | slave->link_failure_count++; | |
2923 | ||
5d397061 JP |
2924 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
2925 | BOND_SLAVE_NOTIFY_NOW); | |
5e5b0665 | 2926 | bond_set_slave_inactive_flags(slave, |
2927 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 2928 | |
e2a7420d | 2929 | slave_info(bond->dev, slave->dev, "link status definitely down, disabling slave\n"); |
b2220cad | 2930 | |
4740d638 | 2931 | if (slave == rtnl_dereference(bond->curr_active_slave)) { |
85741718 | 2932 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
b9f60253 | 2933 | goto do_failover; |
1da177e4 | 2934 | } |
b9f60253 JP |
2935 | |
2936 | continue; | |
b2220cad JV |
2937 | |
2938 | default: | |
1899bb32 JV |
2939 | slave_err(bond->dev, slave->dev, |
2940 | "impossible: link_new_state %d on slave\n", | |
2941 | slave->link_new_state); | |
b9f60253 | 2942 | continue; |
1da177e4 | 2943 | } |
1da177e4 | 2944 | |
b9f60253 | 2945 | do_failover: |
e843fa50 | 2946 | block_netpoll_tx(); |
b9f60253 | 2947 | bond_select_active_slave(bond); |
e843fa50 | 2948 | unblock_netpoll_tx(); |
b2220cad | 2949 | } |
1da177e4 | 2950 | |
b2220cad JV |
2951 | bond_set_carrier(bond); |
2952 | } | |
1da177e4 | 2953 | |
547942ca | 2954 | /* Send ARP probes for active-backup mode ARP monitor. |
b0929915 | 2955 | * |
8c0bc550 | 2956 | * Called with rcu_read_lock held. |
b2220cad | 2957 | */ |
f2ebd477 | 2958 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 2959 | { |
eb9fa4b0 | 2960 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 2961 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
2962 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
2963 | struct list_head *iter; |
2964 | bool found = false; | |
b0929915 | 2965 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 2966 | |
98b90f26 | 2967 | if (curr_arp_slave && curr_active_slave) |
76444f50 VF |
2968 | netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n", |
2969 | curr_arp_slave->dev->name, | |
2970 | curr_active_slave->dev->name); | |
1da177e4 | 2971 | |
98b90f26 VF |
2972 | if (curr_active_slave) { |
2973 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 2974 | return should_notify_rtnl; |
b2220cad | 2975 | } |
1da177e4 | 2976 | |
b2220cad JV |
2977 | /* if we don't have a curr_active_slave, search for the next available |
2978 | * backup slave from the current_arp_slave and make it the candidate | |
2979 | * for becoming the curr_active_slave | |
2980 | */ | |
1da177e4 | 2981 | |
eb9fa4b0 | 2982 | if (!curr_arp_slave) { |
b0929915 | 2983 | curr_arp_slave = bond_first_slave_rcu(bond); |
2984 | if (!curr_arp_slave) | |
2985 | return should_notify_rtnl; | |
b2220cad | 2986 | } |
1da177e4 | 2987 | |
b0929915 | 2988 | bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER); |
059fe7a5 | 2989 | |
b0929915 | 2990 | bond_for_each_slave_rcu(bond, slave, iter) { |
b6adc610 | 2991 | if (!found && !before && bond_slave_is_up(slave)) |
4087df87 | 2992 | before = slave; |
1da177e4 | 2993 | |
b6adc610 | 2994 | if (found && !new_slave && bond_slave_is_up(slave)) |
4087df87 | 2995 | new_slave = slave; |
b2220cad JV |
2996 | /* if the link state is up at this point, we |
2997 | * mark it down - this can happen if we have | |
2998 | * simultaneous link failures and | |
2999 | * reselect_active_interface doesn't make this | |
3000 | * one the current slave so it is still marked | |
3001 | * up when it is actually down | |
1da177e4 | 3002 | */ |
b6adc610 | 3003 | if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
5d397061 JP |
3004 | bond_set_slave_link_state(slave, BOND_LINK_DOWN, |
3005 | BOND_SLAVE_NOTIFY_LATER); | |
b2220cad JV |
3006 | if (slave->link_failure_count < UINT_MAX) |
3007 | slave->link_failure_count++; | |
1da177e4 | 3008 | |
5e5b0665 | 3009 | bond_set_slave_inactive_flags(slave, |
b0929915 | 3010 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 3011 | |
e2a7420d | 3012 | slave_info(bond->dev, slave->dev, "backup interface is now down\n"); |
1da177e4 | 3013 | } |
eb9fa4b0 | 3014 | if (slave == curr_arp_slave) |
4087df87 | 3015 | found = true; |
b2220cad | 3016 | } |
4087df87 VF |
3017 | |
3018 | if (!new_slave && before) | |
3019 | new_slave = before; | |
3020 | ||
b0929915 | 3021 | if (!new_slave) |
3022 | goto check_state; | |
4087df87 | 3023 | |
5d397061 JP |
3024 | bond_set_slave_link_state(new_slave, BOND_LINK_BACK, |
3025 | BOND_SLAVE_NOTIFY_LATER); | |
b0929915 | 3026 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 3027 | bond_arp_send_all(bond, new_slave); |
8e603460 | 3028 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 3029 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 3030 | |
b0929915 | 3031 | check_state: |
3032 | bond_for_each_slave_rcu(bond, slave, iter) { | |
5d397061 | 3033 | if (slave->should_notify || slave->should_notify_link) { |
b0929915 | 3034 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; |
3035 | break; | |
3036 | } | |
3037 | } | |
3038 | return should_notify_rtnl; | |
b2220cad | 3039 | } |
1da177e4 | 3040 | |
d5e73f7b | 3041 | static void bond_activebackup_arp_mon(struct bonding *bond) |
b2220cad | 3042 | { |
b0929915 | 3043 | bool should_notify_peers = false; |
3044 | bool should_notify_rtnl = false; | |
1f2cd845 | 3045 | int delta_in_ticks; |
1da177e4 | 3046 | |
1f2cd845 DM |
3047 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
3048 | ||
3049 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
3050 | goto re_arm; |
3051 | ||
eb9fa4b0 | 3052 | rcu_read_lock(); |
b0929915 | 3053 | |
ad246c99 BH |
3054 | should_notify_peers = bond_should_notify_peers(bond); |
3055 | ||
b0929915 | 3056 | if (bond_ab_arp_inspect(bond)) { |
3057 | rcu_read_unlock(); | |
3058 | ||
1f2cd845 DM |
3059 | /* Race avoidance with bond_close flush of workqueue */ |
3060 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
3061 | delta_in_ticks = 1; |
3062 | should_notify_peers = false; | |
3063 | goto re_arm; | |
3064 | } | |
b2220cad | 3065 | |
1f2cd845 | 3066 | bond_ab_arp_commit(bond); |
b0929915 | 3067 | |
1f2cd845 | 3068 | rtnl_unlock(); |
b0929915 | 3069 | rcu_read_lock(); |
1f2cd845 DM |
3070 | } |
3071 | ||
b0929915 | 3072 | should_notify_rtnl = bond_ab_arp_probe(bond); |
3073 | rcu_read_unlock(); | |
ad246c99 | 3074 | |
80b9d236 | 3075 | re_arm: |
3076 | if (bond->params.arp_interval) | |
1f2cd845 DM |
3077 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
3078 | ||
b0929915 | 3079 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
3080 | if (!rtnl_trylock()) |
3081 | return; | |
b0929915 | 3082 | |
3083 | if (should_notify_peers) | |
3084 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
3085 | bond->dev); | |
5d397061 | 3086 | if (should_notify_rtnl) { |
b0929915 | 3087 | bond_slave_state_notify(bond); |
5d397061 JP |
3088 | bond_slave_link_notify(bond); |
3089 | } | |
b0929915 | 3090 | |
1f2cd845 DM |
3091 | rtnl_unlock(); |
3092 | } | |
1da177e4 LT |
3093 | } |
3094 | ||
d5e73f7b MB |
3095 | static void bond_arp_monitor(struct work_struct *work) |
3096 | { | |
3097 | struct bonding *bond = container_of(work, struct bonding, | |
3098 | arp_work.work); | |
3099 | ||
3100 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) | |
3101 | bond_activebackup_arp_mon(bond); | |
3102 | else | |
3103 | bond_loadbalance_arp_mon(bond); | |
3104 | } | |
3105 | ||
1da177e4 LT |
3106 | /*-------------------------- netdev event handling --------------------------*/ |
3107 | ||
547942ca | 3108 | /* Change device name */ |
1da177e4 LT |
3109 | static int bond_event_changename(struct bonding *bond) |
3110 | { | |
1da177e4 LT |
3111 | bond_remove_proc_entry(bond); |
3112 | bond_create_proc_entry(bond); | |
7e083840 | 3113 | |
f073c7ca TI |
3114 | bond_debug_reregister(bond); |
3115 | ||
1da177e4 LT |
3116 | return NOTIFY_DONE; |
3117 | } | |
3118 | ||
3d632c3f SH |
3119 | static int bond_master_netdev_event(unsigned long event, |
3120 | struct net_device *bond_dev) | |
1da177e4 | 3121 | { |
454d7c9b | 3122 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 | 3123 | |
e2a7420d JW |
3124 | netdev_dbg(bond_dev, "%s called\n", __func__); |
3125 | ||
1da177e4 LT |
3126 | switch (event) { |
3127 | case NETDEV_CHANGENAME: | |
3128 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3129 | case NETDEV_UNREGISTER: |
3130 | bond_remove_proc_entry(event_bond); | |
3131 | break; | |
3132 | case NETDEV_REGISTER: | |
3133 | bond_create_proc_entry(event_bond); | |
3134 | break; | |
1da177e4 LT |
3135 | default: |
3136 | break; | |
3137 | } | |
3138 | ||
3139 | return NOTIFY_DONE; | |
3140 | } | |
3141 | ||
3d632c3f SH |
3142 | static int bond_slave_netdev_event(unsigned long event, |
3143 | struct net_device *slave_dev) | |
1da177e4 | 3144 | { |
059b47e8 | 3145 | struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary; |
6101391d | 3146 | struct bonding *bond; |
3147 | struct net_device *bond_dev; | |
1da177e4 | 3148 | |
6101391d | 3149 | /* A netdev event can be generated while enslaving a device |
3150 | * before netdev_rx_handler_register is called in which case | |
3151 | * slave will be NULL | |
3152 | */ | |
e2a7420d JW |
3153 | if (!slave) { |
3154 | netdev_dbg(slave_dev, "%s called on NULL slave\n", __func__); | |
6101391d | 3155 | return NOTIFY_DONE; |
e2a7420d JW |
3156 | } |
3157 | ||
6101391d | 3158 | bond_dev = slave->bond->dev; |
3159 | bond = slave->bond; | |
059b47e8 | 3160 | primary = rtnl_dereference(bond->primary_slave); |
6101391d | 3161 | |
e2a7420d JW |
3162 | slave_dbg(bond_dev, slave_dev, "%s called\n", __func__); |
3163 | ||
1da177e4 LT |
3164 | switch (event) { |
3165 | case NETDEV_UNREGISTER: | |
8d2ada77 | 3166 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
3167 | bond_release_and_destroy(bond_dev, slave_dev); |
3168 | else | |
f51048c3 | 3169 | __bond_release_one(bond_dev, slave_dev, false, true); |
1da177e4 | 3170 | break; |
98f41f69 | 3171 | case NETDEV_UP: |
1da177e4 | 3172 | case NETDEV_CHANGE: |
4d2c0cda MB |
3173 | /* For 802.3ad mode only: |
3174 | * Getting invalid Speed/Duplex values here will put slave | |
33403121 JW |
3175 | * in weird state. Mark it as link-fail if the link was |
3176 | * previously up or link-down if it hasn't yet come up, and | |
3177 | * let link-monitoring (miimon) set it right when correct | |
3178 | * speeds/duplex are available. | |
4d2c0cda MB |
3179 | */ |
3180 | if (bond_update_speed_duplex(slave) && | |
33403121 JW |
3181 | BOND_MODE(bond) == BOND_MODE_8023AD) { |
3182 | if (slave->last_link_up) | |
3183 | slave->link = BOND_LINK_FAIL; | |
3184 | else | |
3185 | slave->link = BOND_LINK_DOWN; | |
3186 | } | |
4d2c0cda | 3187 | |
52bc6716 MB |
3188 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3189 | bond_3ad_adapter_speed_duplex_changed(slave); | |
950ddcb1 MB |
3190 | /* Fallthrough */ |
3191 | case NETDEV_DOWN: | |
ee637714 MB |
3192 | /* Refresh slave-array if applicable! |
3193 | * If the setup does not use miimon or arpmon (mode-specific!), | |
3194 | * then these events will not cause the slave-array to be | |
3195 | * refreshed. This will cause xmit to use a slave that is not | |
3196 | * usable. Avoid such situation by refeshing the array at these | |
3197 | * events. If these (miimon/arpmon) parameters are configured | |
3198 | * then array gets refreshed twice and that should be fine! | |
3199 | */ | |
e79c1055 | 3200 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 | 3201 | bond_update_slave_arr(bond, NULL); |
1da177e4 | 3202 | break; |
1da177e4 | 3203 | case NETDEV_CHANGEMTU: |
547942ca | 3204 | /* TODO: Should slaves be allowed to |
1da177e4 LT |
3205 | * independently alter their MTU? For |
3206 | * an active-backup bond, slaves need | |
3207 | * not be the same type of device, so | |
3208 | * MTUs may vary. For other modes, | |
3209 | * slaves arguably should have the | |
3210 | * same MTUs. To do this, we'd need to | |
3211 | * take over the slave's change_mtu | |
3212 | * function for the duration of their | |
3213 | * servitude. | |
3214 | */ | |
3215 | break; | |
3216 | case NETDEV_CHANGENAME: | |
3ec775b9 | 3217 | /* we don't care if we don't have primary set */ |
ec0865a9 | 3218 | if (!bond_uses_primary(bond) || |
3ec775b9 VF |
3219 | !bond->params.primary[0]) |
3220 | break; | |
3221 | ||
059b47e8 | 3222 | if (slave == primary) { |
3ec775b9 | 3223 | /* slave's name changed - he's no longer primary */ |
059b47e8 | 3224 | RCU_INIT_POINTER(bond->primary_slave, NULL); |
3ec775b9 VF |
3225 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { |
3226 | /* we have a new primary slave */ | |
059b47e8 | 3227 | rcu_assign_pointer(bond->primary_slave, slave); |
3ec775b9 VF |
3228 | } else { /* we didn't change primary - exit */ |
3229 | break; | |
3230 | } | |
3231 | ||
76444f50 | 3232 | netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n", |
059b47e8 | 3233 | primary ? slave_dev->name : "none"); |
f80889a5 | 3234 | |
3235 | block_netpoll_tx(); | |
3ec775b9 | 3236 | bond_select_active_slave(bond); |
f80889a5 | 3237 | unblock_netpoll_tx(); |
1da177e4 | 3238 | break; |
8531c5ff AK |
3239 | case NETDEV_FEAT_CHANGE: |
3240 | bond_compute_features(bond); | |
3241 | break; | |
4aa5dee4 JP |
3242 | case NETDEV_RESEND_IGMP: |
3243 | /* Propagate to master device */ | |
3244 | call_netdevice_notifiers(event, slave->bond->dev); | |
3245 | break; | |
1da177e4 LT |
3246 | default: |
3247 | break; | |
3248 | } | |
3249 | ||
3250 | return NOTIFY_DONE; | |
3251 | } | |
3252 | ||
547942ca | 3253 | /* bond_netdev_event: handle netdev notifier chain events. |
1da177e4 LT |
3254 | * |
3255 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3256 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3257 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3258 | * dev_probe_lock). | |
3259 | */ | |
3d632c3f SH |
3260 | static int bond_netdev_event(struct notifier_block *this, |
3261 | unsigned long event, void *ptr) | |
1da177e4 | 3262 | { |
351638e7 | 3263 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3264 | |
75466dce JW |
3265 | netdev_dbg(event_dev, "%s received %s\n", |
3266 | __func__, netdev_cmd_to_name(event)); | |
1da177e4 | 3267 | |
0b680e75 JV |
3268 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3269 | return NOTIFY_DONE; | |
3270 | ||
1da177e4 | 3271 | if (event_dev->flags & IFF_MASTER) { |
92480b39 SD |
3272 | int ret; |
3273 | ||
92480b39 SD |
3274 | ret = bond_master_netdev_event(event, event_dev); |
3275 | if (ret != NOTIFY_DONE) | |
3276 | return ret; | |
1da177e4 LT |
3277 | } |
3278 | ||
e2a7420d | 3279 | if (event_dev->flags & IFF_SLAVE) |
1da177e4 | 3280 | return bond_slave_netdev_event(event, event_dev); |
1da177e4 LT |
3281 | |
3282 | return NOTIFY_DONE; | |
3283 | } | |
3284 | ||
3285 | static struct notifier_block bond_netdev_notifier = { | |
3286 | .notifier_call = bond_netdev_event, | |
3287 | }; | |
3288 | ||
169a3e66 JV |
3289 | /*---------------------------- Hashing Policies -----------------------------*/ |
3290 | ||
32819dc1 NA |
3291 | /* L2 hash helper */ |
3292 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 | 3293 | { |
ce04d635 | 3294 | struct ethhdr *ep, hdr_tmp; |
6b923cb7 | 3295 | |
ce04d635 JX |
3296 | ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp); |
3297 | if (ep) | |
3298 | return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto; | |
6b923cb7 JE |
3299 | return 0; |
3300 | } | |
3301 | ||
df98be06 MC |
3302 | static bool bond_flow_ip(struct sk_buff *skb, struct flow_keys *fk, |
3303 | int *noff, int *proto, bool l34) | |
3304 | { | |
3305 | const struct ipv6hdr *iph6; | |
3306 | const struct iphdr *iph; | |
3307 | ||
3308 | if (skb->protocol == htons(ETH_P_IP)) { | |
3309 | if (unlikely(!pskb_may_pull(skb, *noff + sizeof(*iph)))) | |
3310 | return false; | |
3311 | iph = (const struct iphdr *)(skb->data + *noff); | |
3312 | iph_to_flow_copy_v4addrs(fk, iph); | |
3313 | *noff += iph->ihl << 2; | |
3314 | if (!ip_is_fragment(iph)) | |
3315 | *proto = iph->protocol; | |
3316 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
3317 | if (unlikely(!pskb_may_pull(skb, *noff + sizeof(*iph6)))) | |
3318 | return false; | |
3319 | iph6 = (const struct ipv6hdr *)(skb->data + *noff); | |
3320 | iph_to_flow_copy_v6addrs(fk, iph6); | |
3321 | *noff += sizeof(*iph6); | |
3322 | *proto = iph6->nexthdr; | |
3323 | } else { | |
3324 | return false; | |
3325 | } | |
3326 | ||
3327 | if (l34 && *proto >= 0) | |
3328 | fk->ports.ports = skb_flow_get_ports(skb, *noff, *proto); | |
3329 | ||
3330 | return true; | |
3331 | } | |
3332 | ||
32819dc1 NA |
3333 | /* Extract the appropriate headers based on bond's xmit policy */ |
3334 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
3335 | struct flow_keys *fk) | |
6f6652be | 3336 | { |
df98be06 | 3337 | bool l34 = bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34; |
32819dc1 | 3338 | int noff, proto = -1; |
6b923cb7 | 3339 | |
58deb77c MC |
3340 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) { |
3341 | memset(fk, 0, sizeof(*fk)); | |
3342 | return __skb_flow_dissect(NULL, skb, &flow_keys_bonding, | |
3343 | fk, NULL, 0, 0, 0, 0); | |
3344 | } | |
32819dc1 | 3345 | |
06635a35 | 3346 | fk->ports.ports = 0; |
58deb77c | 3347 | memset(&fk->icmp, 0, sizeof(fk->icmp)); |
32819dc1 | 3348 | noff = skb_network_offset(skb); |
df98be06 | 3349 | if (!bond_flow_ip(skb, fk, &noff, &proto, l34)) |
32819dc1 | 3350 | return false; |
df98be06 MC |
3351 | |
3352 | /* ICMP error packets contains at least 8 bytes of the header | |
3353 | * of the packet which generated the error. Use this information | |
3354 | * to correlate ICMP error packets within the same flow which | |
3355 | * generated the error. | |
3356 | */ | |
3357 | if (proto == IPPROTO_ICMP || proto == IPPROTO_ICMPV6) { | |
3358 | skb_flow_get_icmp_tci(skb, &fk->icmp, skb->data, | |
3359 | skb_transport_offset(skb), | |
3360 | skb_headlen(skb)); | |
3361 | if (proto == IPPROTO_ICMP) { | |
3362 | if (!icmp_is_err(fk->icmp.type)) | |
3363 | return true; | |
3364 | ||
3365 | noff += sizeof(struct icmphdr); | |
3366 | } else if (proto == IPPROTO_ICMPV6) { | |
3367 | if (!icmpv6_is_err(fk->icmp.type)) | |
3368 | return true; | |
3369 | ||
3370 | noff += sizeof(struct icmp6hdr); | |
3371 | } | |
3372 | return bond_flow_ip(skb, fk, &noff, &proto, l34); | |
58deb77c | 3373 | } |
6f6652be | 3374 | |
32819dc1 | 3375 | return true; |
6f6652be JV |
3376 | } |
3377 | ||
32819dc1 NA |
3378 | /** |
3379 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3380 | * @bond: bonding device | |
3381 | * @skb: buffer to use for headers | |
32819dc1 NA |
3382 | * |
3383 | * This function will extract the necessary headers from the skb buffer and use | |
3384 | * them to generate a hash based on the xmit_policy set in the bonding device | |
169a3e66 | 3385 | */ |
ee62e868 | 3386 | u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) |
169a3e66 | 3387 | { |
32819dc1 NA |
3388 | struct flow_keys flow; |
3389 | u32 hash; | |
4394542c | 3390 | |
4b1b865e ED |
3391 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP34 && |
3392 | skb->l4_hash) | |
3393 | return skb->hash; | |
3394 | ||
32819dc1 NA |
3395 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3396 | !bond_flow_dissect(bond, skb, &flow)) | |
ee62e868 | 3397 | return bond_eth_hash(skb); |
169a3e66 | 3398 | |
32819dc1 | 3399 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
58deb77c | 3400 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) { |
32819dc1 | 3401 | hash = bond_eth_hash(skb); |
58deb77c MC |
3402 | } else { |
3403 | if (flow.icmp.id) | |
3404 | memcpy(&hash, &flow.icmp, sizeof(hash)); | |
3405 | else | |
3406 | memcpy(&hash, &flow.ports.ports, sizeof(hash)); | |
3407 | } | |
c3f83241 TH |
3408 | hash ^= (__force u32)flow_get_u32_dst(&flow) ^ |
3409 | (__force u32)flow_get_u32_src(&flow); | |
32819dc1 NA |
3410 | hash ^= (hash >> 16); |
3411 | hash ^= (hash >> 8); | |
3412 | ||
b5f86218 | 3413 | return hash >> 1; |
169a3e66 JV |
3414 | } |
3415 | ||
1da177e4 LT |
3416 | /*-------------------------- Device entry points ----------------------------*/ |
3417 | ||
ea8ffc08 | 3418 | void bond_work_init_all(struct bonding *bond) |
fbb0c41b | 3419 | { |
3420 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3421 | bond_resend_igmp_join_requests_delayed); | |
3422 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3423 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
d5e73f7b | 3424 | INIT_DELAYED_WORK(&bond->arp_work, bond_arp_monitor); |
fbb0c41b | 3425 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
ee637714 | 3426 | INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler); |
fbb0c41b | 3427 | } |
3428 | ||
3429 | static void bond_work_cancel_all(struct bonding *bond) | |
3430 | { | |
3431 | cancel_delayed_work_sync(&bond->mii_work); | |
3432 | cancel_delayed_work_sync(&bond->arp_work); | |
3433 | cancel_delayed_work_sync(&bond->alb_work); | |
3434 | cancel_delayed_work_sync(&bond->ad_work); | |
3435 | cancel_delayed_work_sync(&bond->mcast_work); | |
ee637714 | 3436 | cancel_delayed_work_sync(&bond->slave_arr_work); |
fbb0c41b | 3437 | } |
3438 | ||
1da177e4 LT |
3439 | static int bond_open(struct net_device *bond_dev) |
3440 | { | |
454d7c9b | 3441 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3442 | struct list_head *iter; |
ba3211cc | 3443 | struct slave *slave; |
1da177e4 | 3444 | |
ba3211cc | 3445 | /* reset slave->backup and slave->inactive */ |
0965a1f3 | 3446 | if (bond_has_slaves(bond)) { |
9caff1e7 | 3447 | bond_for_each_slave(bond, slave, iter) { |
4740d638 ED |
3448 | if (bond_uses_primary(bond) && |
3449 | slave != rcu_access_pointer(bond->curr_active_slave)) { | |
5e5b0665 | 3450 | bond_set_slave_inactive_flags(slave, |
3451 | BOND_SLAVE_NOTIFY_NOW); | |
8bbe71a5 | 3452 | } else if (BOND_MODE(bond) != BOND_MODE_8023AD) { |
5e5b0665 | 3453 | bond_set_slave_active_flags(slave, |
3454 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3455 | } |
3456 | } | |
ba3211cc | 3457 | } |
ba3211cc | 3458 | |
58402054 | 3459 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3460 | /* bond_alb_initialize must be called before the timer |
3461 | * is started. | |
3462 | */ | |
01844098 | 3463 | if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) |
b473946a | 3464 | return -ENOMEM; |
e79c1055 | 3465 | if (bond->params.tlb_dynamic_lb || BOND_MODE(bond) == BOND_MODE_ALB) |
e9f0fb88 | 3466 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3467 | } |
3468 | ||
fbb0c41b | 3469 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3470 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3471 | |
3472 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3473 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3474 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3475 | } |
3476 | ||
01844098 | 3477 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1b76b316 | 3478 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3479 | /* register to receive LACPDUs */ |
3aba891d | 3480 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3481 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3482 | } |
3483 | ||
e79c1055 | 3484 | if (bond_mode_can_use_xmit_hash(bond)) |
ee637714 MB |
3485 | bond_update_slave_arr(bond, NULL); |
3486 | ||
1da177e4 LT |
3487 | return 0; |
3488 | } | |
3489 | ||
3490 | static int bond_close(struct net_device *bond_dev) | |
3491 | { | |
454d7c9b | 3492 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3493 | |
fbb0c41b | 3494 | bond_work_cancel_all(bond); |
6c8d23f7 | 3495 | bond->send_peer_notif = 0; |
ee8487c0 | 3496 | if (bond_is_lb(bond)) |
1da177e4 | 3497 | bond_alb_deinitialize(bond); |
3aba891d | 3498 | bond->recv_probe = NULL; |
1da177e4 LT |
3499 | |
3500 | return 0; | |
3501 | } | |
3502 | ||
fe30937b ED |
3503 | /* fold stats, assuming all rtnl_link_stats64 fields are u64, but |
3504 | * that some drivers can provide 32bit values only. | |
3505 | */ | |
3506 | static void bond_fold_stats(struct rtnl_link_stats64 *_res, | |
3507 | const struct rtnl_link_stats64 *_new, | |
3508 | const struct rtnl_link_stats64 *_old) | |
3509 | { | |
3510 | const u64 *new = (const u64 *)_new; | |
3511 | const u64 *old = (const u64 *)_old; | |
3512 | u64 *res = (u64 *)_res; | |
3513 | int i; | |
3514 | ||
3515 | for (i = 0; i < sizeof(*_res) / sizeof(u64); i++) { | |
3516 | u64 nv = new[i]; | |
3517 | u64 ov = old[i]; | |
142c6594 | 3518 | s64 delta = nv - ov; |
fe30937b ED |
3519 | |
3520 | /* detects if this particular field is 32bit only */ | |
3521 | if (((nv | ov) >> 32) == 0) | |
142c6594 ED |
3522 | delta = (s64)(s32)((u32)nv - (u32)ov); |
3523 | ||
3524 | /* filter anomalies, some drivers reset their stats | |
3525 | * at down/up events. | |
3526 | */ | |
3527 | if (delta > 0) | |
3528 | res[i] += delta; | |
fe30937b ED |
3529 | } |
3530 | } | |
3531 | ||
bc1f4470 | 3532 | static void bond_get_stats(struct net_device *bond_dev, |
3533 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3534 | { |
454d7c9b | 3535 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3536 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3537 | struct list_head *iter; |
1da177e4 | 3538 | struct slave *slave; |
1da177e4 | 3539 | |
089bca2c | 3540 | spin_lock(&bond->stats_lock); |
5f0c5f73 | 3541 | memcpy(stats, &bond->bond_stats, sizeof(*stats)); |
1da177e4 | 3542 | |
fe30937b ED |
3543 | rcu_read_lock(); |
3544 | bond_for_each_slave_rcu(bond, slave, iter) { | |
3545 | const struct rtnl_link_stats64 *new = | |
28172739 | 3546 | dev_get_stats(slave->dev, &temp); |
fe30937b ED |
3547 | |
3548 | bond_fold_stats(stats, new, &slave->slave_stats); | |
5f0c5f73 AG |
3549 | |
3550 | /* save off the slave stats for the next run */ | |
fe30937b | 3551 | memcpy(&slave->slave_stats, new, sizeof(*new)); |
5f0c5f73 | 3552 | } |
fe30937b ED |
3553 | rcu_read_unlock(); |
3554 | ||
5f0c5f73 | 3555 | memcpy(&bond->bond_stats, stats, sizeof(*stats)); |
fe30937b | 3556 | spin_unlock(&bond->stats_lock); |
1da177e4 LT |
3557 | } |
3558 | ||
3559 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3560 | { | |
080a06e1 | 3561 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3562 | struct net_device *slave_dev = NULL; |
3563 | struct ifbond k_binfo; | |
3564 | struct ifbond __user *u_binfo = NULL; | |
3565 | struct ifslave k_sinfo; | |
3566 | struct ifslave __user *u_sinfo = NULL; | |
3567 | struct mii_ioctl_data *mii = NULL; | |
d1fbd3ed | 3568 | struct bond_opt_value newval; |
387ff911 | 3569 | struct net *net; |
1da177e4 LT |
3570 | int res = 0; |
3571 | ||
76444f50 | 3572 | netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd); |
1da177e4 LT |
3573 | |
3574 | switch (cmd) { | |
1da177e4 LT |
3575 | case SIOCGMIIPHY: |
3576 | mii = if_mii(ifr); | |
3d632c3f | 3577 | if (!mii) |
1da177e4 | 3578 | return -EINVAL; |
3d632c3f | 3579 | |
1da177e4 LT |
3580 | mii->phy_id = 0; |
3581 | /* Fall Through */ | |
3582 | case SIOCGMIIREG: | |
547942ca | 3583 | /* We do this again just in case we were called by SIOCGMIIREG |
1da177e4 LT |
3584 | * instead of SIOCGMIIPHY. |
3585 | */ | |
3586 | mii = if_mii(ifr); | |
3d632c3f | 3587 | if (!mii) |
1da177e4 | 3588 | return -EINVAL; |
3d632c3f | 3589 | |
1da177e4 | 3590 | if (mii->reg_num == 1) { |
1da177e4 | 3591 | mii->val_out = 0; |
3d632c3f | 3592 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3593 | mii->val_out = BMSR_LSTATUS; |
1da177e4 LT |
3594 | } |
3595 | ||
3596 | return 0; | |
3597 | case BOND_INFO_QUERY_OLD: | |
3598 | case SIOCBONDINFOQUERY: | |
3599 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3600 | ||
3d632c3f | 3601 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3602 | return -EFAULT; |
1da177e4 | 3603 | |
3d67576d ZY |
3604 | bond_info_query(bond_dev, &k_binfo); |
3605 | if (copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3d632c3f | 3606 | return -EFAULT; |
1da177e4 | 3607 | |
3d67576d | 3608 | return 0; |
1da177e4 LT |
3609 | case BOND_SLAVE_INFO_QUERY_OLD: |
3610 | case SIOCBONDSLAVEINFOQUERY: | |
3611 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3612 | ||
3d632c3f | 3613 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3614 | return -EFAULT; |
1da177e4 LT |
3615 | |
3616 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3617 | if (res == 0 && |
3618 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3619 | return -EFAULT; | |
1da177e4 LT |
3620 | |
3621 | return res; | |
3622 | default: | |
1da177e4 LT |
3623 | break; |
3624 | } | |
3625 | ||
387ff911 G |
3626 | net = dev_net(bond_dev); |
3627 | ||
3628 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3629 | return -EPERM; |
1da177e4 | 3630 | |
0917b933 | 3631 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3632 | |
e2a7420d | 3633 | slave_dbg(bond_dev, slave_dev, "slave_dev=%p:\n", slave_dev); |
1da177e4 | 3634 | |
3d632c3f | 3635 | if (!slave_dev) |
0917b933 | 3636 | return -ENODEV; |
1da177e4 | 3637 | |
0917b933 YX |
3638 | switch (cmd) { |
3639 | case BOND_ENSLAVE_OLD: | |
3640 | case SIOCBONDENSLAVE: | |
33eaf2a6 | 3641 | res = bond_enslave(bond_dev, slave_dev, NULL); |
0917b933 YX |
3642 | break; |
3643 | case BOND_RELEASE_OLD: | |
3644 | case SIOCBONDRELEASE: | |
3645 | res = bond_release(bond_dev, slave_dev); | |
3646 | break; | |
3647 | case BOND_SETHWADDR_OLD: | |
3648 | case SIOCBONDSETHWADDR: | |
b9245914 | 3649 | res = bond_set_dev_addr(bond_dev, slave_dev); |
0917b933 YX |
3650 | break; |
3651 | case BOND_CHANGE_ACTIVE_OLD: | |
3652 | case SIOCBONDCHANGEACTIVE: | |
d1fbd3ed | 3653 | bond_opt_initstr(&newval, slave_dev->name); |
7a7e96e0 VY |
3654 | res = __bond_opt_set_notify(bond, BOND_OPT_ACTIVE_SLAVE, |
3655 | &newval); | |
0917b933 YX |
3656 | break; |
3657 | default: | |
3658 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3659 | } |
3660 | ||
1da177e4 LT |
3661 | return res; |
3662 | } | |
3663 | ||
d03462b9 | 3664 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3665 | { |
454d7c9b | 3666 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3667 | |
d03462b9 JP |
3668 | if (change & IFF_PROMISC) |
3669 | bond_set_promiscuity(bond, | |
3670 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3671 | |
d03462b9 JP |
3672 | if (change & IFF_ALLMULTI) |
3673 | bond_set_allmulti(bond, | |
3674 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3675 | } | |
1da177e4 | 3676 | |
303d1cbf | 3677 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3678 | { |
3679 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3680 | struct list_head *iter; |
303d1cbf | 3681 | struct slave *slave; |
1da177e4 | 3682 | |
b3241870 | 3683 | rcu_read_lock(); |
ec0865a9 | 3684 | if (bond_uses_primary(bond)) { |
b3241870 | 3685 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3686 | if (slave) { |
3687 | dev_uc_sync(slave->dev, bond_dev); | |
3688 | dev_mc_sync(slave->dev, bond_dev); | |
3689 | } | |
303d1cbf | 3690 | } else { |
b3241870 | 3691 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3692 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3693 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3694 | } | |
1da177e4 | 3695 | } |
b3241870 | 3696 | rcu_read_unlock(); |
1da177e4 LT |
3697 | } |
3698 | ||
234bcf8a | 3699 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3700 | { |
234bcf8a | 3701 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3702 | const struct net_device_ops *slave_ops; |
3703 | struct neigh_parms parms; | |
dec1e90e | 3704 | struct slave *slave; |
234bcf8a SP |
3705 | int ret; |
3706 | ||
dec1e90e | 3707 | slave = bond_first_slave(bond); |
234bcf8a SP |
3708 | if (!slave) |
3709 | return 0; | |
234bcf8a | 3710 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3711 | if (!slave_ops->ndo_neigh_setup) |
3712 | return 0; | |
3713 | ||
3714 | parms.neigh_setup = NULL; | |
3715 | parms.neigh_cleanup = NULL; | |
3716 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3717 | if (ret) | |
3718 | return ret; | |
3719 | ||
547942ca | 3720 | /* Assign slave's neigh_cleanup to neighbour in case cleanup is called |
234bcf8a SP |
3721 | * after the last slave has been detached. Assumes that all slaves |
3722 | * utilize the same neigh_cleanup (true at this writing as only user | |
3723 | * is ipoib). | |
3724 | */ | |
3725 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3726 | ||
3727 | if (!parms.neigh_setup) | |
3728 | return 0; | |
3729 | ||
3730 | return parms.neigh_setup(n); | |
3731 | } | |
3732 | ||
547942ca | 3733 | /* The bonding ndo_neigh_setup is called at init time beofre any |
234bcf8a SP |
3734 | * slave exists. So we must declare proxy setup function which will |
3735 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3736 | * |
3737 | * It's also called by master devices (such as vlans) to setup their | |
3738 | * underlying devices. In that case - do nothing, we're already set up from | |
3739 | * our init. | |
234bcf8a SP |
3740 | */ |
3741 | static int bond_neigh_setup(struct net_device *dev, | |
3742 | struct neigh_parms *parms) | |
3743 | { | |
9918d5bf VF |
3744 | /* modify only our neigh_parms */ |
3745 | if (parms->dev == dev) | |
3746 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3747 | |
00829823 SH |
3748 | return 0; |
3749 | } | |
3750 | ||
547942ca | 3751 | /* Change the MTU of all of a master's slaves to match the master */ |
1da177e4 LT |
3752 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) |
3753 | { | |
454d7c9b | 3754 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3755 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3756 | struct list_head *iter; |
1da177e4 | 3757 | int res = 0; |
1da177e4 | 3758 | |
76444f50 | 3759 | netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu); |
1da177e4 | 3760 | |
9caff1e7 | 3761 | bond_for_each_slave(bond, slave, iter) { |
e2a7420d | 3762 | slave_dbg(bond_dev, slave->dev, "s %p c_m %p\n", |
76444f50 | 3763 | slave, slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 3764 | |
1da177e4 LT |
3765 | res = dev_set_mtu(slave->dev, new_mtu); |
3766 | ||
3767 | if (res) { | |
3768 | /* If we failed to set the slave's mtu to the new value | |
3769 | * we must abort the operation even in ACTIVE_BACKUP | |
3770 | * mode, because if we allow the backup slaves to have | |
3771 | * different mtu values than the active slave we'll | |
3772 | * need to change their mtu when doing a failover. That | |
3773 | * means changing their mtu from timer context, which | |
3774 | * is probably not a good idea. | |
3775 | */ | |
e2a7420d JW |
3776 | slave_dbg(bond_dev, slave->dev, "err %d setting mtu to %d\n", |
3777 | res, new_mtu); | |
1da177e4 LT |
3778 | goto unwind; |
3779 | } | |
3780 | } | |
3781 | ||
3782 | bond_dev->mtu = new_mtu; | |
3783 | ||
3784 | return 0; | |
3785 | ||
3786 | unwind: | |
3787 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3788 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3789 | int tmp_res; |
3790 | ||
81f23b13 VF |
3791 | if (rollback_slave == slave) |
3792 | break; | |
3793 | ||
3794 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
e2a7420d JW |
3795 | if (tmp_res) |
3796 | slave_dbg(bond_dev, rollback_slave->dev, "unwind err %d\n", | |
3797 | tmp_res); | |
1da177e4 LT |
3798 | } |
3799 | ||
3800 | return res; | |
3801 | } | |
3802 | ||
547942ca | 3803 | /* Change HW address |
1da177e4 LT |
3804 | * |
3805 | * Note that many devices must be down to change the HW address, and | |
3806 | * downing the master releases all slaves. We can make bonds full of | |
3807 | * bonding devices to test this, however. | |
3808 | */ | |
3809 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3810 | { | |
454d7c9b | 3811 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3812 | struct slave *slave, *rollback_slave; |
faeeb317 | 3813 | struct sockaddr_storage *ss = addr, tmp_ss; |
9caff1e7 | 3814 | struct list_head *iter; |
1da177e4 | 3815 | int res = 0; |
1da177e4 | 3816 | |
01844098 | 3817 | if (BOND_MODE(bond) == BOND_MODE_ALB) |
eb7cc59a SH |
3818 | return bond_alb_set_mac_address(bond_dev, addr); |
3819 | ||
3820 | ||
e2a7420d | 3821 | netdev_dbg(bond_dev, "%s: bond=%p\n", __func__, bond); |
1da177e4 | 3822 | |
1b5acd29 JV |
3823 | /* If fail_over_mac is enabled, do nothing and return success. |
3824 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3825 | */ |
cc689aaa | 3826 | if (bond->params.fail_over_mac && |
01844098 | 3827 | BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
dd957c57 | 3828 | return 0; |
2ab82852 | 3829 | |
faeeb317 | 3830 | if (!is_valid_ether_addr(ss->__data)) |
1da177e4 | 3831 | return -EADDRNOTAVAIL; |
1da177e4 | 3832 | |
9caff1e7 | 3833 | bond_for_each_slave(bond, slave, iter) { |
e2a7420d JW |
3834 | slave_dbg(bond_dev, slave->dev, "%s: slave=%p\n", |
3835 | __func__, slave); | |
3a37a963 | 3836 | res = dev_set_mac_address(slave->dev, addr, NULL); |
1da177e4 LT |
3837 | if (res) { |
3838 | /* TODO: consider downing the slave | |
3839 | * and retry ? | |
3840 | * User should expect communications | |
3841 | * breakage anyway until ARP finish | |
3842 | * updating, so... | |
3843 | */ | |
e2a7420d JW |
3844 | slave_dbg(bond_dev, slave->dev, "%s: err %d\n", |
3845 | __func__, res); | |
1da177e4 LT |
3846 | goto unwind; |
3847 | } | |
3848 | } | |
3849 | ||
3850 | /* success */ | |
faeeb317 | 3851 | memcpy(bond_dev->dev_addr, ss->__data, bond_dev->addr_len); |
1da177e4 LT |
3852 | return 0; |
3853 | ||
3854 | unwind: | |
faeeb317 JW |
3855 | memcpy(tmp_ss.__data, bond_dev->dev_addr, bond_dev->addr_len); |
3856 | tmp_ss.ss_family = bond_dev->type; | |
1da177e4 LT |
3857 | |
3858 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3859 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3860 | int tmp_res; |
3861 | ||
81f23b13 VF |
3862 | if (rollback_slave == slave) |
3863 | break; | |
3864 | ||
faeeb317 | 3865 | tmp_res = dev_set_mac_address(rollback_slave->dev, |
3a37a963 | 3866 | (struct sockaddr *)&tmp_ss, NULL); |
1da177e4 | 3867 | if (tmp_res) { |
e2a7420d JW |
3868 | slave_dbg(bond_dev, rollback_slave->dev, "%s: unwind err %d\n", |
3869 | __func__, tmp_res); | |
1da177e4 LT |
3870 | } |
3871 | } | |
3872 | ||
3873 | return res; | |
3874 | } | |
3875 | ||
15077228 NA |
3876 | /** |
3877 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3878 | * @bond: bonding device that is transmitting | |
3879 | * @skb: buffer to transmit | |
3880 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3881 | * | |
3882 | * This function tries to transmit through slave with slave_id but in case | |
3883 | * it fails, it tries to find the first available slave for transmission. | |
3884 | * The skb is consumed in all cases, thus the function is void. | |
3885 | */ | |
da131ddb | 3886 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3887 | { |
9caff1e7 | 3888 | struct list_head *iter; |
15077228 NA |
3889 | struct slave *slave; |
3890 | int i = slave_id; | |
3891 | ||
3892 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3893 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 | 3894 | if (--i < 0) { |
891ab54d | 3895 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3896 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3897 | return; | |
3898 | } | |
3899 | } | |
3900 | } | |
3901 | ||
3902 | /* Here we start from the first slave up to slave_id */ | |
3903 | i = slave_id; | |
9caff1e7 | 3904 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3905 | if (--i < 0) |
3906 | break; | |
891ab54d | 3907 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3908 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3909 | return; | |
3910 | } | |
3911 | } | |
3912 | /* no slave that can tx has been found */ | |
31aa860e | 3913 | bond_tx_drop(bond->dev, skb); |
15077228 NA |
3914 | } |
3915 | ||
73958329 NA |
3916 | /** |
3917 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3918 | * @bond: bonding device to use | |
3919 | * | |
3920 | * Based on the value of the bonding device's packets_per_slave parameter | |
3921 | * this function generates a slave id, which is usually used as the next | |
3922 | * slave to transmit through. | |
3923 | */ | |
3924 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3925 | { | |
73958329 | 3926 | u32 slave_id; |
809fa972 HFS |
3927 | struct reciprocal_value reciprocal_packets_per_slave; |
3928 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
3929 | |
3930 | switch (packets_per_slave) { | |
3931 | case 0: | |
3932 | slave_id = prandom_u32(); | |
3933 | break; | |
3934 | case 1: | |
3935 | slave_id = bond->rr_tx_counter; | |
3936 | break; | |
3937 | default: | |
809fa972 HFS |
3938 | reciprocal_packets_per_slave = |
3939 | bond->params.reciprocal_packets_per_slave; | |
73958329 | 3940 | slave_id = reciprocal_divide(bond->rr_tx_counter, |
809fa972 | 3941 | reciprocal_packets_per_slave); |
73958329 NA |
3942 | break; |
3943 | } | |
3944 | bond->rr_tx_counter++; | |
3945 | ||
3946 | return slave_id; | |
3947 | } | |
3948 | ||
dbdc8a21 TZ |
3949 | static netdev_tx_t bond_xmit_roundrobin(struct sk_buff *skb, |
3950 | struct net_device *bond_dev) | |
1da177e4 | 3951 | { |
454d7c9b | 3952 | struct bonding *bond = netdev_priv(bond_dev); |
15077228 | 3953 | struct slave *slave; |
9d1bc24b | 3954 | int slave_cnt; |
73958329 | 3955 | u32 slave_id; |
1da177e4 | 3956 | |
73958329 | 3957 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3958 | * default for sending IGMP traffic. For failover purposes one |
3959 | * needs to maintain some consistency for the interface that will | |
3960 | * send the join/membership reports. The curr_active_slave found | |
3961 | * will send all of this type of traffic. | |
cf5f9044 | 3962 | */ |
9d1bc24b CW |
3963 | if (skb->protocol == htons(ETH_P_IP)) { |
3964 | int noff = skb_network_offset(skb); | |
3965 | struct iphdr *iph; | |
9a72c2da | 3966 | |
9d1bc24b CW |
3967 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
3968 | goto non_igmp; | |
3969 | ||
3970 | iph = ip_hdr(skb); | |
3971 | if (iph->protocol == IPPROTO_IGMP) { | |
3972 | slave = rcu_dereference(bond->curr_active_slave); | |
3973 | if (slave) | |
3974 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3975 | else | |
3976 | bond_xmit_slave_id(bond, skb, 0); | |
3977 | return NETDEV_TX_OK; | |
9a72c2da | 3978 | } |
1da177e4 | 3979 | } |
0693e88e | 3980 | |
9d1bc24b CW |
3981 | non_igmp: |
3982 | slave_cnt = READ_ONCE(bond->slave_cnt); | |
3983 | if (likely(slave_cnt)) { | |
3984 | slave_id = bond_rr_gen_slave_id(bond); | |
3985 | bond_xmit_slave_id(bond, skb, slave_id % slave_cnt); | |
3986 | } else { | |
3987 | bond_tx_drop(bond_dev, skb); | |
3988 | } | |
ec634fe3 | 3989 | return NETDEV_TX_OK; |
1da177e4 LT |
3990 | } |
3991 | ||
547942ca | 3992 | /* In active-backup mode, we know that bond->curr_active_slave is always valid if |
1da177e4 LT |
3993 | * the bond has a usable interface. |
3994 | */ | |
dbdc8a21 TZ |
3995 | static netdev_tx_t bond_xmit_activebackup(struct sk_buff *skb, |
3996 | struct net_device *bond_dev) | |
1da177e4 | 3997 | { |
454d7c9b | 3998 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3999 | struct slave *slave; |
1da177e4 | 4000 | |
278b2083 | 4001 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 4002 | if (slave) |
15077228 NA |
4003 | bond_dev_queue_xmit(bond, skb, slave->dev); |
4004 | else | |
31aa860e | 4005 | bond_tx_drop(bond_dev, skb); |
0693e88e | 4006 | |
ec634fe3 | 4007 | return NETDEV_TX_OK; |
1da177e4 LT |
4008 | } |
4009 | ||
ee637714 MB |
4010 | /* Use this to update slave_array when (a) it's not appropriate to update |
4011 | * slave_array right away (note that update_slave_array() may sleep) | |
4012 | * and / or (b) RTNL is not held. | |
1da177e4 | 4013 | */ |
ee637714 | 4014 | void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay) |
1da177e4 | 4015 | { |
ee637714 MB |
4016 | queue_delayed_work(bond->wq, &bond->slave_arr_work, delay); |
4017 | } | |
1da177e4 | 4018 | |
ee637714 MB |
4019 | /* Slave array work handler. Holds only RTNL */ |
4020 | static void bond_slave_arr_handler(struct work_struct *work) | |
4021 | { | |
4022 | struct bonding *bond = container_of(work, struct bonding, | |
4023 | slave_arr_work.work); | |
4024 | int ret; | |
4025 | ||
4026 | if (!rtnl_trylock()) | |
4027 | goto err; | |
4028 | ||
4029 | ret = bond_update_slave_arr(bond, NULL); | |
4030 | rtnl_unlock(); | |
4031 | if (ret) { | |
4032 | pr_warn_ratelimited("Failed to update slave array from WT\n"); | |
4033 | goto err; | |
4034 | } | |
4035 | return; | |
4036 | ||
4037 | err: | |
4038 | bond_slave_arr_work_rearm(bond, 1); | |
4039 | } | |
4040 | ||
4041 | /* Build the usable slaves array in control path for modes that use xmit-hash | |
4042 | * to determine the slave interface - | |
4043 | * (a) BOND_MODE_8023AD | |
4044 | * (b) BOND_MODE_XOR | |
e79c1055 | 4045 | * (c) (BOND_MODE_TLB || BOND_MODE_ALB) && tlb_dynamic_lb == 0 |
ee637714 MB |
4046 | * |
4047 | * The caller is expected to hold RTNL only and NO other lock! | |
4048 | */ | |
4049 | int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave) | |
4050 | { | |
4051 | struct slave *slave; | |
4052 | struct list_head *iter; | |
4053 | struct bond_up_slave *new_arr, *old_arr; | |
ee637714 MB |
4054 | int agg_id = 0; |
4055 | int ret = 0; | |
4056 | ||
4057 | #ifdef CONFIG_LOCKDEP | |
4058 | WARN_ON(lockdep_is_held(&bond->mode_lock)); | |
4059 | #endif | |
4060 | ||
4061 | new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]), | |
4062 | GFP_KERNEL); | |
4063 | if (!new_arr) { | |
4064 | ret = -ENOMEM; | |
4065 | pr_err("Failed to build slave-array.\n"); | |
4066 | goto out; | |
4067 | } | |
4068 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
4069 | struct ad_info ad_info; | |
4070 | ||
4071 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
4072 | pr_debug("bond_3ad_get_active_agg_info failed\n"); | |
4073 | kfree_rcu(new_arr, rcu); | |
4074 | /* No active aggragator means it's not safe to use | |
4075 | * the previous array. | |
4076 | */ | |
4077 | old_arr = rtnl_dereference(bond->slave_arr); | |
4078 | if (old_arr) { | |
4079 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
4080 | kfree_rcu(old_arr, rcu); | |
4081 | } | |
4082 | goto out; | |
4083 | } | |
ee637714 MB |
4084 | agg_id = ad_info.aggregator_id; |
4085 | } | |
4086 | bond_for_each_slave(bond, slave, iter) { | |
4087 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
4088 | struct aggregator *agg; | |
4089 | ||
4090 | agg = SLAVE_AD_INFO(slave)->port.aggregator; | |
4091 | if (!agg || agg->aggregator_identifier != agg_id) | |
4092 | continue; | |
4093 | } | |
4094 | if (!bond_slave_can_tx(slave)) | |
4095 | continue; | |
4096 | if (skipslave == slave) | |
4097 | continue; | |
e79c1055 | 4098 | |
e2a7420d JW |
4099 | slave_dbg(bond->dev, slave->dev, "Adding slave to tx hash array[%d]\n", |
4100 | new_arr->count); | |
e79c1055 | 4101 | |
ee637714 MB |
4102 | new_arr->arr[new_arr->count++] = slave; |
4103 | } | |
4104 | ||
4105 | old_arr = rtnl_dereference(bond->slave_arr); | |
4106 | rcu_assign_pointer(bond->slave_arr, new_arr); | |
4107 | if (old_arr) | |
4108 | kfree_rcu(old_arr, rcu); | |
4109 | out: | |
4110 | if (ret != 0 && skipslave) { | |
4111 | int idx; | |
4112 | ||
4113 | /* Rare situation where caller has asked to skip a specific | |
4114 | * slave but allocation failed (most likely!). BTW this is | |
4115 | * only possible when the call is initiated from | |
4116 | * __bond_release_one(). In this situation; overwrite the | |
4117 | * skipslave entry in the array with the last entry from the | |
4118 | * array to avoid a situation where the xmit path may choose | |
4119 | * this to-be-skipped slave to send a packet out. | |
4120 | */ | |
4121 | old_arr = rtnl_dereference(bond->slave_arr); | |
a7137534 | 4122 | for (idx = 0; old_arr != NULL && idx < old_arr->count; idx++) { |
ee637714 MB |
4123 | if (skipslave == old_arr->arr[idx]) { |
4124 | old_arr->arr[idx] = | |
4125 | old_arr->arr[old_arr->count-1]; | |
4126 | old_arr->count--; | |
4127 | break; | |
4128 | } | |
4129 | } | |
4130 | } | |
4131 | return ret; | |
4132 | } | |
4133 | ||
4134 | /* Use this Xmit function for 3AD as well as XOR modes. The current | |
4135 | * usable slave array is formed in the control path. The xmit function | |
4136 | * just calculates hash and sends the packet out. | |
4137 | */ | |
dbdc8a21 TZ |
4138 | static netdev_tx_t bond_3ad_xor_xmit(struct sk_buff *skb, |
4139 | struct net_device *dev) | |
ee637714 MB |
4140 | { |
4141 | struct bonding *bond = netdev_priv(dev); | |
4142 | struct slave *slave; | |
4143 | struct bond_up_slave *slaves; | |
4144 | unsigned int count; | |
4145 | ||
4146 | slaves = rcu_dereference(bond->slave_arr); | |
6aa7de05 | 4147 | count = slaves ? READ_ONCE(slaves->count) : 0; |
ee637714 MB |
4148 | if (likely(count)) { |
4149 | slave = slaves->arr[bond_xmit_hash(bond, skb) % count]; | |
4150 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
4151 | } else { | |
31aa860e | 4152 | bond_tx_drop(dev, skb); |
ee637714 | 4153 | } |
0693e88e | 4154 | |
ec634fe3 | 4155 | return NETDEV_TX_OK; |
1da177e4 LT |
4156 | } |
4157 | ||
78a646ce | 4158 | /* in broadcast mode, we send everything to all usable interfaces. */ |
dbdc8a21 TZ |
4159 | static netdev_tx_t bond_xmit_broadcast(struct sk_buff *skb, |
4160 | struct net_device *bond_dev) | |
1da177e4 | 4161 | { |
454d7c9b | 4162 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 4163 | struct slave *slave = NULL; |
9caff1e7 | 4164 | struct list_head *iter; |
1da177e4 | 4165 | |
9caff1e7 | 4166 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
4167 | if (bond_is_last_slave(bond, slave)) |
4168 | break; | |
b6adc610 | 4169 | if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
78a646ce | 4170 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 | 4171 | |
78a646ce | 4172 | if (!skb2) { |
9152e26d | 4173 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
4174 | bond_dev->name, __func__); | |
78a646ce | 4175 | continue; |
1da177e4 | 4176 | } |
78a646ce | 4177 | bond_dev_queue_xmit(bond, skb2, slave->dev); |
1da177e4 LT |
4178 | } |
4179 | } | |
b6adc610 | 4180 | if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) |
78a646ce NA |
4181 | bond_dev_queue_xmit(bond, skb, slave->dev); |
4182 | else | |
31aa860e | 4183 | bond_tx_drop(bond_dev, skb); |
3d632c3f | 4184 | |
ec634fe3 | 4185 | return NETDEV_TX_OK; |
1da177e4 LT |
4186 | } |
4187 | ||
4188 | /*------------------------- Device initialization ---------------------------*/ | |
4189 | ||
547942ca | 4190 | /* Lookup the slave that corresponds to a qid */ |
bb1d9123 AG |
4191 | static inline int bond_slave_override(struct bonding *bond, |
4192 | struct sk_buff *skb) | |
4193 | { | |
bb1d9123 | 4194 | struct slave *slave = NULL; |
9caff1e7 | 4195 | struct list_head *iter; |
bb1d9123 | 4196 | |
ae35c6f7 | 4197 | if (!skb_rx_queue_recorded(skb)) |
0693e88e | 4198 | return 1; |
bb1d9123 AG |
4199 | |
4200 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 4201 | bond_for_each_slave_rcu(bond, slave, iter) { |
ae35c6f7 | 4202 | if (slave->queue_id == skb_get_queue_mapping(skb)) { |
f5e2dc5d AN |
4203 | if (bond_slave_is_up(slave) && |
4204 | slave->link == BOND_LINK_UP) { | |
3900f290 | 4205 | bond_dev_queue_xmit(bond, skb, slave->dev); |
4206 | return 0; | |
4207 | } | |
4208 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
4209 | break; |
4210 | } | |
4211 | } | |
4212 | ||
3900f290 | 4213 | return 1; |
bb1d9123 AG |
4214 | } |
4215 | ||
374eeb5a | 4216 | |
f663dd9a | 4217 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
a350ecce | 4218 | struct net_device *sb_dev) |
bb1d9123 | 4219 | { |
547942ca | 4220 | /* This helper function exists to help dev_pick_tx get the correct |
fd0e435b | 4221 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4222 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4223 | * way down to the bonding driver. | |
4224 | */ | |
fd0e435b PO |
4225 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4226 | ||
547942ca | 4227 | /* Save the original txq to restore before passing to the driver */ |
ae35c6f7 | 4228 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb_get_queue_mapping(skb); |
374eeb5a | 4229 | |
fd0e435b | 4230 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4231 | do { |
fd0e435b | 4232 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4233 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4234 | } |
4235 | return txq; | |
bb1d9123 AG |
4236 | } |
4237 | ||
0693e88e | 4238 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4239 | { |
bb1d9123 AG |
4240 | struct bonding *bond = netdev_priv(dev); |
4241 | ||
d1e2e5cd VF |
4242 | if (bond_should_override_tx_queue(bond) && |
4243 | !bond_slave_override(bond, skb)) | |
4244 | return NETDEV_TX_OK; | |
00829823 | 4245 | |
01844098 | 4246 | switch (BOND_MODE(bond)) { |
00829823 SH |
4247 | case BOND_MODE_ROUNDROBIN: |
4248 | return bond_xmit_roundrobin(skb, dev); | |
4249 | case BOND_MODE_ACTIVEBACKUP: | |
4250 | return bond_xmit_activebackup(skb, dev); | |
ee637714 | 4251 | case BOND_MODE_8023AD: |
00829823 | 4252 | case BOND_MODE_XOR: |
ee637714 | 4253 | return bond_3ad_xor_xmit(skb, dev); |
00829823 SH |
4254 | case BOND_MODE_BROADCAST: |
4255 | return bond_xmit_broadcast(skb, dev); | |
00829823 | 4256 | case BOND_MODE_ALB: |
00829823 | 4257 | return bond_alb_xmit(skb, dev); |
f05b42ea MB |
4258 | case BOND_MODE_TLB: |
4259 | return bond_tlb_xmit(skb, dev); | |
00829823 SH |
4260 | default: |
4261 | /* Should never happen, mode already checked */ | |
76444f50 | 4262 | netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); |
00829823 | 4263 | WARN_ON_ONCE(1); |
31aa860e | 4264 | bond_tx_drop(dev, skb); |
00829823 SH |
4265 | return NETDEV_TX_OK; |
4266 | } | |
4267 | } | |
4268 | ||
0693e88e MM |
4269 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4270 | { | |
4271 | struct bonding *bond = netdev_priv(dev); | |
4272 | netdev_tx_t ret = NETDEV_TX_OK; | |
4273 | ||
547942ca | 4274 | /* If we risk deadlock from transmitting this in the |
0693e88e MM |
4275 | * netpoll path, tell netpoll to queue the frame for later tx |
4276 | */ | |
054bb880 | 4277 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
4278 | return NETDEV_TX_BUSY; |
4279 | ||
278b2083 | 4280 | rcu_read_lock(); |
0965a1f3 | 4281 | if (bond_has_slaves(bond)) |
0693e88e MM |
4282 | ret = __bond_start_xmit(skb, dev); |
4283 | else | |
31aa860e | 4284 | bond_tx_drop(dev, skb); |
278b2083 | 4285 | rcu_read_unlock(); |
0693e88e MM |
4286 | |
4287 | return ret; | |
4288 | } | |
00829823 | 4289 | |
d46b6349 PR |
4290 | static int bond_ethtool_get_link_ksettings(struct net_device *bond_dev, |
4291 | struct ethtool_link_ksettings *cmd) | |
bb5b052f AG |
4292 | { |
4293 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 4294 | unsigned long speed = 0; |
9caff1e7 | 4295 | struct list_head *iter; |
dec1e90e | 4296 | struct slave *slave; |
bb5b052f | 4297 | |
d46b6349 PR |
4298 | cmd->base.duplex = DUPLEX_UNKNOWN; |
4299 | cmd->base.port = PORT_OTHER; | |
bb5b052f | 4300 | |
8557cd74 | 4301 | /* Since bond_slave_can_tx returns false for all inactive or down slaves, we |
bb5b052f AG |
4302 | * do not need to check mode. Though link speed might not represent |
4303 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
4304 | * this is an accurate maximum. | |
4305 | */ | |
9caff1e7 | 4306 | bond_for_each_slave(bond, slave, iter) { |
8557cd74 | 4307 | if (bond_slave_can_tx(slave)) { |
bb5b052f AG |
4308 | if (slave->speed != SPEED_UNKNOWN) |
4309 | speed += slave->speed; | |
d46b6349 | 4310 | if (cmd->base.duplex == DUPLEX_UNKNOWN && |
bb5b052f | 4311 | slave->duplex != DUPLEX_UNKNOWN) |
d46b6349 | 4312 | cmd->base.duplex = slave->duplex; |
bb5b052f AG |
4313 | } |
4314 | } | |
d46b6349 | 4315 | cmd->base.speed = speed ? : SPEED_UNKNOWN; |
dec1e90e | 4316 | |
bb5b052f AG |
4317 | return 0; |
4318 | } | |
4319 | ||
217df670 | 4320 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 4321 | struct ethtool_drvinfo *drvinfo) |
217df670 | 4322 | { |
7826d43f JP |
4323 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
4324 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
4325 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
4326 | BOND_ABI_VERSION); | |
217df670 JV |
4327 | } |
4328 | ||
7282d491 | 4329 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4330 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac | 4331 | .get_link = ethtool_op_get_link, |
d46b6349 | 4332 | .get_link_ksettings = bond_ethtool_get_link_ksettings, |
8531c5ff AK |
4333 | }; |
4334 | ||
eb7cc59a | 4335 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4336 | .ndo_init = bond_init, |
9e71626c | 4337 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4338 | .ndo_open = bond_open, |
4339 | .ndo_stop = bond_close, | |
00829823 | 4340 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4341 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4342 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4343 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4344 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 4345 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 4346 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4347 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4348 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4349 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4350 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4351 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4352 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4353 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4354 | .ndo_poll_controller = bond_poll_controller, | |
4355 | #endif | |
9232ecca JP |
4356 | .ndo_add_slave = bond_enslave, |
4357 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4358 | .ndo_fix_features = bond_fix_features, |
4847f049 | 4359 | .ndo_features_check = passthru_features_check, |
eb7cc59a SH |
4360 | }; |
4361 | ||
b3f92b63 DG |
4362 | static const struct device_type bond_type = { |
4363 | .name = "bond", | |
4364 | }; | |
4365 | ||
9e2e61fb AW |
4366 | static void bond_destructor(struct net_device *bond_dev) |
4367 | { | |
4368 | struct bonding *bond = netdev_priv(bond_dev); | |
4369 | if (bond->wq) | |
4370 | destroy_workqueue(bond->wq); | |
9e2e61fb AW |
4371 | } |
4372 | ||
0a2a78c4 | 4373 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4374 | { |
454d7c9b | 4375 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4376 | |
b7435628 | 4377 | spin_lock_init(&bond->mode_lock); |
d2991f75 | 4378 | bond->params = bonding_defaults; |
1da177e4 LT |
4379 | |
4380 | /* Initialize pointers */ | |
1da177e4 | 4381 | bond->dev = bond_dev; |
1da177e4 LT |
4382 | |
4383 | /* Initialize the device entry points */ | |
181470fc | 4384 | ether_setup(bond_dev); |
31c05415 | 4385 | bond_dev->max_mtu = ETH_MAX_MTU; |
eb7cc59a | 4386 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4387 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 4388 | |
cf124db5 DM |
4389 | bond_dev->needs_free_netdev = true; |
4390 | bond_dev->priv_destructor = bond_destructor; | |
1da177e4 | 4391 | |
b3f92b63 DG |
4392 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4393 | ||
1da177e4 | 4394 | /* Initialize the device options */ |
1098cee6 | 4395 | bond_dev->flags |= IFF_MASTER; |
1e6f20ca | 4396 | bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT | IFF_NO_QUEUE; |
550fd08c | 4397 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4398 | |
547942ca | 4399 | /* don't acquire bond device's netif_tx_lock when transmitting */ |
1da177e4 LT |
4400 | bond_dev->features |= NETIF_F_LLTX; |
4401 | ||
4402 | /* By default, we declare the bond to be fully | |
4403 | * VLAN hardware accelerated capable. Special | |
4404 | * care is taken in the various xmit functions | |
4405 | * when there are slaves that are not hw accel | |
4406 | * capable | |
4407 | */ | |
1da177e4 | 4408 | |
f9399814 WC |
4409 | /* Don't allow bond devices to change network namespaces. */ |
4410 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
4411 | ||
b2a103e6 | 4412 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4413 | NETIF_F_HW_VLAN_CTAG_RX | |
4414 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4415 | |
8eea1ca8 | 4416 | bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL | NETIF_F_GSO_UDP_L4; |
b2a103e6 | 4417 | bond_dev->features |= bond_dev->hw_features; |
30d8177e | 4418 | bond_dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; |
1da177e4 LT |
4419 | } |
4420 | ||
547942ca NA |
4421 | /* Destroy a bonding device. |
4422 | * Must be under rtnl_lock when this function is called. | |
4423 | */ | |
c67dfb29 | 4424 | static void bond_uninit(struct net_device *bond_dev) |
a434e43f | 4425 | { |
454d7c9b | 4426 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
4427 | struct list_head *iter; |
4428 | struct slave *slave; | |
ee637714 | 4429 | struct bond_up_slave *arr; |
a434e43f | 4430 | |
f6dc31a8 WC |
4431 | bond_netpoll_cleanup(bond_dev); |
4432 | ||
c67dfb29 | 4433 | /* Release the bonded slaves */ |
544a028e | 4434 | bond_for_each_slave(bond, slave, iter) |
f51048c3 | 4435 | __bond_release_one(bond_dev, slave->dev, true, true); |
76444f50 | 4436 | netdev_info(bond_dev, "Released all slaves\n"); |
c67dfb29 | 4437 | |
ee637714 MB |
4438 | arr = rtnl_dereference(bond->slave_arr); |
4439 | if (arr) { | |
4440 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
4441 | kfree_rcu(arr, rcu); | |
4442 | } | |
4443 | ||
a434e43f JV |
4444 | list_del(&bond->bond_list); |
4445 | ||
089bca2c | 4446 | lockdep_unregister_key(&bond->stats_lock_key); |
f073c7ca | 4447 | bond_debug_unregister(bond); |
a434e43f JV |
4448 | } |
4449 | ||
1da177e4 LT |
4450 | /*------------------------- Module initialization ---------------------------*/ |
4451 | ||
1da177e4 LT |
4452 | static int bond_check_params(struct bond_params *params) |
4453 | { | |
5a5c5fd4 | 4454 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 4455 | struct bond_opt_value newval; |
4456 | const struct bond_opt_value *valptr; | |
72ccc471 | 4457 | int arp_all_targets_value = 0; |
6791e466 | 4458 | u16 ad_actor_sys_prio = 0; |
d22a5fc0 | 4459 | u16 ad_user_port_key = 0; |
72ccc471 | 4460 | __be32 arp_target[BOND_MAX_ARP_TARGETS] = { 0 }; |
dc9c4d0f MB |
4461 | int arp_ip_count; |
4462 | int bond_mode = BOND_MODE_ROUNDROBIN; | |
4463 | int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; | |
4464 | int lacp_fast = 0; | |
f13ad104 | 4465 | int tlb_dynamic_lb; |
f5b2b966 | 4466 | |
547942ca | 4467 | /* Convert string parameters. */ |
1da177e4 | 4468 | if (mode) { |
2b3798d5 NA |
4469 | bond_opt_initstr(&newval, mode); |
4470 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
4471 | if (!valptr) { | |
4472 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
4473 | return -EINVAL; |
4474 | } | |
2b3798d5 | 4475 | bond_mode = valptr->value; |
1da177e4 LT |
4476 | } |
4477 | ||
169a3e66 | 4478 | if (xmit_hash_policy) { |
e79c1055 DB |
4479 | if (bond_mode == BOND_MODE_ROUNDROBIN || |
4480 | bond_mode == BOND_MODE_ACTIVEBACKUP || | |
4481 | bond_mode == BOND_MODE_BROADCAST) { | |
a4aee5c8 | 4482 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 4483 | bond_mode_name(bond_mode)); |
169a3e66 | 4484 | } else { |
a4b32ce7 NA |
4485 | bond_opt_initstr(&newval, xmit_hash_policy); |
4486 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
4487 | &newval); | |
4488 | if (!valptr) { | |
a4aee5c8 | 4489 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
4490 | xmit_hash_policy); |
4491 | return -EINVAL; | |
4492 | } | |
a4b32ce7 | 4493 | xmit_hashtype = valptr->value; |
169a3e66 JV |
4494 | } |
4495 | } | |
4496 | ||
1da177e4 LT |
4497 | if (lacp_rate) { |
4498 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4499 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4500 | bond_mode_name(bond_mode)); | |
1da177e4 | 4501 | } else { |
d3131de7 NA |
4502 | bond_opt_initstr(&newval, lacp_rate); |
4503 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
4504 | &newval); | |
4505 | if (!valptr) { | |
a4aee5c8 | 4506 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 4507 | lacp_rate); |
1da177e4 LT |
4508 | return -EINVAL; |
4509 | } | |
d3131de7 | 4510 | lacp_fast = valptr->value; |
1da177e4 LT |
4511 | } |
4512 | } | |
4513 | ||
fd989c83 | 4514 | if (ad_select) { |
548d28bd | 4515 | bond_opt_initstr(&newval, ad_select); |
9e5f5eeb NA |
4516 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), |
4517 | &newval); | |
4518 | if (!valptr) { | |
4519 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
4520 | return -EINVAL; |
4521 | } | |
9e5f5eeb NA |
4522 | params->ad_select = valptr->value; |
4523 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 4524 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4525 | } else { |
4526 | params->ad_select = BOND_AD_STABLE; | |
4527 | } | |
4528 | ||
f5841306 | 4529 | if (max_bonds < 0) { |
91565ebb JP |
4530 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4531 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4532 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4533 | } | |
4534 | ||
4535 | if (miimon < 0) { | |
91565ebb JP |
4536 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4537 | miimon, INT_MAX); | |
b98d9c66 | 4538 | miimon = 0; |
1da177e4 LT |
4539 | } |
4540 | ||
4541 | if (updelay < 0) { | |
91565ebb JP |
4542 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4543 | updelay, INT_MAX); | |
1da177e4 LT |
4544 | updelay = 0; |
4545 | } | |
4546 | ||
4547 | if (downdelay < 0) { | |
91565ebb JP |
4548 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4549 | downdelay, INT_MAX); | |
1da177e4 LT |
4550 | downdelay = 0; |
4551 | } | |
4552 | ||
b3c898e2 DB |
4553 | if ((use_carrier != 0) && (use_carrier != 1)) { |
4554 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", | |
91565ebb | 4555 | use_carrier); |
1da177e4 LT |
4556 | use_carrier = 1; |
4557 | } | |
4558 | ||
ad246c99 | 4559 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
4560 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
4561 | num_peer_notif); | |
ad246c99 BH |
4562 | num_peer_notif = 1; |
4563 | } | |
4564 | ||
834db4bc | 4565 | /* reset values for 802.3ad/TLB/ALB */ |
267bed77 | 4566 | if (!bond_mode_uses_arp(bond_mode)) { |
1da177e4 | 4567 | if (!miimon) { |
91565ebb JP |
4568 | 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"); |
4569 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 4570 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4571 | } |
4572 | } | |
4573 | ||
bb1d9123 | 4574 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
4575 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
4576 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
4577 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
4578 | } | |
4579 | ||
ebd8e497 | 4580 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
4581 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
4582 | all_slaves_active); | |
ebd8e497 AG |
4583 | all_slaves_active = 0; |
4584 | } | |
4585 | ||
c2952c31 | 4586 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
4587 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
4588 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
4589 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
4590 | } | |
4591 | ||
aa59d851 NA |
4592 | bond_opt_initval(&newval, packets_per_slave); |
4593 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
4594 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
4595 | packets_per_slave, USHRT_MAX); | |
4596 | packets_per_slave = 1; | |
4597 | } | |
4598 | ||
1da177e4 | 4599 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4600 | 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", |
4601 | updelay); | |
1da177e4 LT |
4602 | } |
4603 | ||
4604 | if (!miimon) { | |
4605 | if (updelay || downdelay) { | |
4606 | /* just warn the user the up/down delay will have | |
4607 | * no effect since miimon is zero... | |
4608 | */ | |
91565ebb JP |
4609 | 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", |
4610 | updelay, downdelay); | |
1da177e4 LT |
4611 | } |
4612 | } else { | |
4613 | /* don't allow arp monitoring */ | |
4614 | if (arp_interval) { | |
91565ebb JP |
4615 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4616 | miimon, arp_interval); | |
1da177e4 LT |
4617 | arp_interval = 0; |
4618 | } | |
4619 | ||
4620 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
4621 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4622 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
4623 | } |
4624 | ||
4625 | updelay /= miimon; | |
4626 | ||
4627 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
4628 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4629 | downdelay, miimon, | |
4630 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4631 | } |
4632 | ||
4633 | downdelay /= miimon; | |
4634 | } | |
4635 | ||
4636 | if (arp_interval < 0) { | |
91565ebb JP |
4637 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4638 | arp_interval, INT_MAX); | |
7bdb04ed | 4639 | arp_interval = 0; |
1da177e4 LT |
4640 | } |
4641 | ||
5a5c5fd4 | 4642 | for (arp_ip_count = 0, i = 0; |
4643 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
89015c18 | 4644 | __be32 ip; |
547942ca NA |
4645 | |
4646 | /* not a complete check, but good enough to catch mistakes */ | |
89015c18 | 4647 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || |
2807a9fe | 4648 | !bond_is_ip_target_ok(ip)) { |
91565ebb JP |
4649 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4650 | arp_ip_target[i]); | |
1da177e4 LT |
4651 | arp_interval = 0; |
4652 | } else { | |
0afee4e8 VF |
4653 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4654 | arp_target[arp_ip_count++] = ip; | |
4655 | else | |
91565ebb JP |
4656 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
4657 | &ip); | |
1da177e4 LT |
4658 | } |
4659 | } | |
4660 | ||
4661 | if (arp_interval && !arp_ip_count) { | |
4662 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
4663 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4664 | arp_interval); | |
1da177e4 LT |
4665 | arp_interval = 0; |
4666 | } | |
4667 | ||
f5b2b966 | 4668 | if (arp_validate) { |
f5b2b966 | 4669 | if (!arp_interval) { |
a4aee5c8 | 4670 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4671 | return -EINVAL; |
4672 | } | |
4673 | ||
16228881 NA |
4674 | bond_opt_initstr(&newval, arp_validate); |
4675 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
4676 | &newval); | |
4677 | if (!valptr) { | |
a4aee5c8 | 4678 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 4679 | arp_validate); |
f5b2b966 JV |
4680 | return -EINVAL; |
4681 | } | |
16228881 NA |
4682 | arp_validate_value = valptr->value; |
4683 | } else { | |
f5b2b966 | 4684 | arp_validate_value = 0; |
16228881 | 4685 | } |
f5b2b966 | 4686 | |
8599b52e | 4687 | if (arp_all_targets) { |
edf36b24 NA |
4688 | bond_opt_initstr(&newval, arp_all_targets); |
4689 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
4690 | &newval); | |
4691 | if (!valptr) { | |
8599b52e VF |
4692 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
4693 | arp_all_targets); | |
4694 | arp_all_targets_value = 0; | |
edf36b24 NA |
4695 | } else { |
4696 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
4697 | } |
4698 | } | |
4699 | ||
1da177e4 | 4700 | if (miimon) { |
a4aee5c8 | 4701 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4702 | } else if (arp_interval) { |
16228881 NA |
4703 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
4704 | arp_validate_value); | |
a4aee5c8 | 4705 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 4706 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
4707 | |
4708 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 4709 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 4710 | |
90194264 | 4711 | pr_cont("\n"); |
1da177e4 | 4712 | |
b8a9787e | 4713 | } else if (max_bonds) { |
1da177e4 LT |
4714 | /* miimon and arp_interval not set, we need one so things |
4715 | * work as expected, see bonding.txt for details | |
4716 | */ | |
90194264 | 4717 | 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 |
4718 | } |
4719 | ||
ec0865a9 | 4720 | if (primary && !bond_mode_uses_primary(bond_mode)) { |
1da177e4 LT |
4721 | /* currently, using a primary only makes sense |
4722 | * in active backup, TLB or ALB modes | |
4723 | */ | |
91565ebb JP |
4724 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
4725 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4726 | primary = NULL; |
4727 | } | |
4728 | ||
a549952a | 4729 | if (primary && primary_reselect) { |
388d3a6d NA |
4730 | bond_opt_initstr(&newval, primary_reselect); |
4731 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
4732 | &newval); | |
4733 | if (!valptr) { | |
a4aee5c8 | 4734 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 4735 | primary_reselect); |
a549952a JP |
4736 | return -EINVAL; |
4737 | } | |
388d3a6d | 4738 | primary_reselect_value = valptr->value; |
a549952a JP |
4739 | } else { |
4740 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4741 | } | |
4742 | ||
3915c1e8 | 4743 | if (fail_over_mac) { |
1df6b6aa NA |
4744 | bond_opt_initstr(&newval, fail_over_mac); |
4745 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
4746 | &newval); | |
4747 | if (!valptr) { | |
a4aee5c8 | 4748 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 4749 | fail_over_mac); |
3915c1e8 JV |
4750 | return -EINVAL; |
4751 | } | |
1df6b6aa | 4752 | fail_over_mac_value = valptr->value; |
3915c1e8 | 4753 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 4754 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
4755 | } else { |
4756 | fail_over_mac_value = BOND_FOM_NONE; | |
4757 | } | |
dd957c57 | 4758 | |
6791e466 MB |
4759 | bond_opt_initstr(&newval, "default"); |
4760 | valptr = bond_opt_parse( | |
4761 | bond_opt_get(BOND_OPT_AD_ACTOR_SYS_PRIO), | |
4762 | &newval); | |
4763 | if (!valptr) { | |
4764 | pr_err("Error: No ad_actor_sys_prio default value"); | |
4765 | return -EINVAL; | |
4766 | } | |
4767 | ad_actor_sys_prio = valptr->value; | |
4768 | ||
d22a5fc0 MB |
4769 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_USER_PORT_KEY), |
4770 | &newval); | |
4771 | if (!valptr) { | |
4772 | pr_err("Error: No ad_user_port_key default value"); | |
4773 | return -EINVAL; | |
4774 | } | |
4775 | ad_user_port_key = valptr->value; | |
4776 | ||
f13ad104 NA |
4777 | bond_opt_initstr(&newval, "default"); |
4778 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_TLB_DYNAMIC_LB), &newval); | |
4779 | if (!valptr) { | |
4780 | pr_err("Error: No tlb_dynamic_lb default value"); | |
4781 | return -EINVAL; | |
8b426dc5 | 4782 | } |
f13ad104 | 4783 | tlb_dynamic_lb = valptr->value; |
8b426dc5 | 4784 | |
3a7129e5 | 4785 | if (lp_interval == 0) { |
91565ebb JP |
4786 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
4787 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 4788 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
4789 | } | |
4790 | ||
1da177e4 LT |
4791 | /* fill params struct with the proper values */ |
4792 | params->mode = bond_mode; | |
169a3e66 | 4793 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4794 | params->miimon = miimon; |
ad246c99 | 4795 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4796 | params->arp_interval = arp_interval; |
f5b2b966 | 4797 | params->arp_validate = arp_validate_value; |
8599b52e | 4798 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4799 | params->updelay = updelay; |
4800 | params->downdelay = downdelay; | |
07a4ddec | 4801 | params->peer_notif_delay = 0; |
1da177e4 LT |
4802 | params->use_carrier = use_carrier; |
4803 | params->lacp_fast = lacp_fast; | |
4804 | params->primary[0] = 0; | |
a549952a | 4805 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4806 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4807 | params->tx_queues = tx_queues; |
ebd8e497 | 4808 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4809 | params->resend_igmp = resend_igmp; |
655f8919 | 4810 | params->min_links = min_links; |
3a7129e5 | 4811 | params->lp_interval = lp_interval; |
809fa972 | 4812 | params->packets_per_slave = packets_per_slave; |
8b426dc5 | 4813 | params->tlb_dynamic_lb = tlb_dynamic_lb; |
6791e466 | 4814 | params->ad_actor_sys_prio = ad_actor_sys_prio; |
74514957 | 4815 | eth_zero_addr(params->ad_actor_system); |
d22a5fc0 | 4816 | params->ad_user_port_key = ad_user_port_key; |
809fa972 HFS |
4817 | if (packets_per_slave > 0) { |
4818 | params->reciprocal_packets_per_slave = | |
4819 | reciprocal_value(packets_per_slave); | |
4820 | } else { | |
4821 | /* reciprocal_packets_per_slave is unused if | |
4822 | * packets_per_slave is 0 or 1, just initialize it | |
4823 | */ | |
4824 | params->reciprocal_packets_per_slave = | |
4825 | (struct reciprocal_value) { 0 }; | |
4826 | } | |
4827 | ||
1da177e4 LT |
4828 | if (primary) { |
4829 | strncpy(params->primary, primary, IFNAMSIZ); | |
4830 | params->primary[IFNAMSIZ - 1] = 0; | |
4831 | } | |
4832 | ||
4833 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4834 | ||
4835 | return 0; | |
4836 | } | |
4837 | ||
547942ca | 4838 | /* Called from registration process */ |
181470fc SH |
4839 | static int bond_init(struct net_device *bond_dev) |
4840 | { | |
4841 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4842 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc | 4843 | |
76444f50 | 4844 | netdev_dbg(bond_dev, "Begin bond_init\n"); |
181470fc | 4845 | |
f9f225eb | 4846 | bond->wq = alloc_ordered_workqueue(bond_dev->name, WQ_MEM_RECLAIM); |
181470fc SH |
4847 | if (!bond->wq) |
4848 | return -ENOMEM; | |
4849 | ||
089bca2c TY |
4850 | spin_lock_init(&bond->stats_lock); |
4851 | lockdep_register_key(&bond->stats_lock_key); | |
4852 | lockdep_set_class(&bond->stats_lock, &bond->stats_lock_key); | |
181470fc | 4853 | |
ec87fd3b | 4854 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4855 | |
6151b3d4 | 4856 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4857 | |
f073c7ca TI |
4858 | bond_debug_register(bond); |
4859 | ||
409cc1f8 JP |
4860 | /* Ensure valid dev_addr */ |
4861 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4862 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4863 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4864 | |
181470fc SH |
4865 | return 0; |
4866 | } | |
4867 | ||
0a2a78c4 | 4868 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4869 | { |
efacb309 | 4870 | return tx_queues; |
d5da4510 JP |
4871 | } |
4872 | ||
dfe60397 | 4873 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4874 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4875 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4876 | * set up our sysfs entries. | |
4877 | */ | |
ec87fd3b | 4878 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4879 | { |
4880 | struct net_device *bond_dev; | |
e913fb27 P |
4881 | struct bonding *bond; |
4882 | struct alb_bond_info *bond_info; | |
dfe60397 MW |
4883 | int res; |
4884 | ||
4885 | rtnl_lock(); | |
027ea041 | 4886 | |
1c5cae81 | 4887 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
c835a677 | 4888 | name ? name : "bond%d", NET_NAME_UNKNOWN, |
1c5cae81 | 4889 | bond_setup, tx_queues); |
dfe60397 | 4890 | if (!bond_dev) { |
a4aee5c8 | 4891 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4892 | rtnl_unlock(); |
4893 | return -ENOMEM; | |
dfe60397 MW |
4894 | } |
4895 | ||
e913fb27 P |
4896 | /* |
4897 | * Initialize rx_hashtbl_used_head to RLB_NULL_INDEX. | |
4898 | * It is set to 0 by default which is wrong. | |
4899 | */ | |
4900 | bond = netdev_priv(bond_dev); | |
4901 | bond_info = &(BOND_ALB_INFO(bond)); | |
4902 | bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX; | |
4903 | ||
ec87fd3b | 4904 | dev_net_set(bond_dev, net); |
88ead977 EB |
4905 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4906 | ||
dfe60397 | 4907 | res = register_netdevice(bond_dev); |
0daa2303 | 4908 | |
e826eafa PO |
4909 | netif_carrier_off(bond_dev); |
4910 | ||
4493b81b MB |
4911 | bond_work_init_all(bond); |
4912 | ||
7e083840 | 4913 | rtnl_unlock(); |
9e2e61fb | 4914 | if (res < 0) |
cf124db5 | 4915 | free_netdev(bond_dev); |
30c15ba9 | 4916 | return res; |
dfe60397 MW |
4917 | } |
4918 | ||
2c8c1e72 | 4919 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4920 | { |
15449745 | 4921 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4922 | |
4923 | bn->net = net; | |
4924 | INIT_LIST_HEAD(&bn->dev_list); | |
4925 | ||
ec87fd3b | 4926 | bond_create_proc_dir(bn); |
4c22400a | 4927 | bond_create_sysfs(bn); |
87a7b84b | 4928 | |
15449745 | 4929 | return 0; |
ec87fd3b EB |
4930 | } |
4931 | ||
2c8c1e72 | 4932 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4933 | { |
15449745 | 4934 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4935 | struct bonding *bond, *tmp_bond; |
4936 | LIST_HEAD(list); | |
ec87fd3b | 4937 | |
4c22400a | 4938 | bond_destroy_sysfs(bn); |
69b0216a | 4939 | |
4940 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4941 | rtnl_lock(); | |
4942 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4943 | unregister_netdevice_queue(bond->dev, &list); | |
4944 | unregister_netdevice_many(&list); | |
4945 | rtnl_unlock(); | |
23fa5c2c VF |
4946 | |
4947 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
4948 | } |
4949 | ||
4950 | static struct pernet_operations bond_net_ops = { | |
4951 | .init = bond_net_init, | |
4952 | .exit = bond_net_exit, | |
15449745 EB |
4953 | .id = &bond_net_id, |
4954 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4955 | }; |
4956 | ||
1da177e4 LT |
4957 | static int __init bonding_init(void) |
4958 | { | |
1da177e4 LT |
4959 | int i; |
4960 | int res; | |
4961 | ||
bd33acc3 | 4962 | pr_info("%s", bond_version); |
1da177e4 | 4963 | |
dfe60397 | 4964 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4965 | if (res) |
dfe60397 | 4966 | goto out; |
1da177e4 | 4967 | |
15449745 | 4968 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4969 | if (res) |
4970 | goto out; | |
027ea041 | 4971 | |
0a2a78c4 | 4972 | res = bond_netlink_init(); |
88ead977 | 4973 | if (res) |
6639104b | 4974 | goto err_link; |
88ead977 | 4975 | |
f073c7ca TI |
4976 | bond_create_debugfs(); |
4977 | ||
1da177e4 | 4978 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4979 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4980 | if (res) |
4981 | goto err; | |
1da177e4 LT |
4982 | } |
4983 | ||
58deb77c MC |
4984 | skb_flow_dissector_init(&flow_keys_bonding, |
4985 | flow_keys_bonding_keys, | |
4986 | ARRAY_SIZE(flow_keys_bonding_keys)); | |
4987 | ||
1da177e4 | 4988 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4989 | out: |
1da177e4 | 4990 | return res; |
88ead977 | 4991 | err: |
db298686 | 4992 | bond_destroy_debugfs(); |
0a2a78c4 | 4993 | bond_netlink_fini(); |
6639104b | 4994 | err_link: |
15449745 | 4995 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4996 | goto out; |
dfe60397 | 4997 | |
1da177e4 LT |
4998 | } |
4999 | ||
5000 | static void __exit bonding_exit(void) | |
5001 | { | |
5002 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
5003 | ||
f073c7ca | 5004 | bond_destroy_debugfs(); |
ae68c398 | 5005 | |
0a2a78c4 | 5006 | bond_netlink_fini(); |
ffcdedb6 | 5007 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
5008 | |
5009 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
547942ca | 5010 | /* Make sure we don't have an imbalance on our netpoll blocking */ |
fb4fa76a | 5011 | WARN_ON(atomic_read(&netpoll_block_tx)); |
e843fa50 | 5012 | #endif |
1da177e4 LT |
5013 | } |
5014 | ||
5015 | module_init(bonding_init); | |
5016 | module_exit(bonding_exit); | |
5017 | MODULE_LICENSE("GPL"); | |
5018 | MODULE_VERSION(DRV_VERSION); | |
5019 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5020 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |