Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * originally based on the dummy device. | |
3 | * | |
4 | * Copyright 1999, Thomas Davis, tadavis@lbl.gov. | |
5 | * Licensed under the GPL. Based on dummy.c, and eql.c devices. | |
6 | * | |
7 | * bonding.c: an Ethernet Bonding driver | |
8 | * | |
9 | * This is useful to talk to a Cisco EtherChannel compatible equipment: | |
10 | * Cisco 5500 | |
11 | * Sun Trunking (Solaris) | |
12 | * Alteon AceDirector Trunks | |
13 | * Linux Bonding | |
14 | * and probably many L2 switches ... | |
15 | * | |
16 | * How it works: | |
17 | * ifconfig bond0 ipaddress netmask up | |
18 | * will setup a network device, with an ip address. No mac address | |
19 | * will be assigned at this time. The hw mac address will come from | |
20 | * the first slave bonded to the channel. All slaves will then use | |
21 | * this hw mac address. | |
22 | * | |
23 | * ifconfig bond0 down | |
24 | * will release all slaves, marking them as down. | |
25 | * | |
26 | * ifenslave bond0 eth0 | |
27 | * will attach eth0 to bond0 as a slave. eth0 hw mac address will either | |
28 | * a: be used as initial mac address | |
29 | * b: if a hw mac address already is there, eth0's hw mac address | |
30 | * will then be set from bond0. | |
31 | * | |
1da177e4 LT |
32 | */ |
33 | ||
1da177e4 LT |
34 | #include <linux/kernel.h> |
35 | #include <linux/module.h> | |
1da177e4 LT |
36 | #include <linux/types.h> |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/interrupt.h> | |
39 | #include <linux/ptrace.h> | |
40 | #include <linux/ioport.h> | |
41 | #include <linux/in.h> | |
169a3e66 | 42 | #include <net/ip.h> |
1da177e4 | 43 | #include <linux/ip.h> |
169a3e66 JV |
44 | #include <linux/tcp.h> |
45 | #include <linux/udp.h> | |
1da177e4 LT |
46 | #include <linux/slab.h> |
47 | #include <linux/string.h> | |
48 | #include <linux/init.h> | |
49 | #include <linux/timer.h> | |
50 | #include <linux/socket.h> | |
51 | #include <linux/ctype.h> | |
52 | #include <linux/inet.h> | |
53 | #include <linux/bitops.h> | |
3d632c3f | 54 | #include <linux/io.h> |
1da177e4 | 55 | #include <asm/dma.h> |
3d632c3f | 56 | #include <linux/uaccess.h> |
1da177e4 LT |
57 | #include <linux/errno.h> |
58 | #include <linux/netdevice.h> | |
59 | #include <linux/inetdevice.h> | |
a816c7c7 | 60 | #include <linux/igmp.h> |
1da177e4 LT |
61 | #include <linux/etherdevice.h> |
62 | #include <linux/skbuff.h> | |
63 | #include <net/sock.h> | |
64 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
65 | #include <linux/smp.h> |
66 | #include <linux/if_ether.h> | |
67 | #include <net/arp.h> | |
68 | #include <linux/mii.h> | |
69 | #include <linux/ethtool.h> | |
70 | #include <linux/if_vlan.h> | |
71 | #include <linux/if_bonding.h> | |
b63bb739 | 72 | #include <linux/jiffies.h> |
e843fa50 | 73 | #include <linux/preempt.h> |
c3ade5ca | 74 | #include <net/route.h> |
457c4cbc | 75 | #include <net/net_namespace.h> |
ec87fd3b | 76 | #include <net/netns/generic.h> |
5ee31c68 | 77 | #include <net/pkt_sched.h> |
278b2083 | 78 | #include <linux/rculist.h> |
32819dc1 | 79 | #include <net/flow_keys.h> |
c158cba3 | 80 | #include <net/switchdev.h> |
1ef8019b DM |
81 | #include <net/bonding.h> |
82 | #include <net/bond_3ad.h> | |
83 | #include <net/bond_alb.h> | |
1da177e4 LT |
84 | |
85 | /*---------------------------- Module parameters ----------------------------*/ | |
86 | ||
87 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
1da177e4 LT |
88 | |
89 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 90 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 91 | static int num_peer_notif = 1; |
b98d9c66 | 92 | static int miimon; |
3d632c3f SH |
93 | static int updelay; |
94 | static int downdelay; | |
1da177e4 | 95 | static int use_carrier = 1; |
3d632c3f SH |
96 | static char *mode; |
97 | static char *primary; | |
a549952a | 98 | static char *primary_reselect; |
3d632c3f | 99 | static char *lacp_rate; |
655f8919 | 100 | static int min_links; |
3d632c3f SH |
101 | static char *ad_select; |
102 | static char *xmit_hash_policy; | |
7bdb04ed | 103 | static int arp_interval; |
3d632c3f SH |
104 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
105 | static char *arp_validate; | |
8599b52e | 106 | static char *arp_all_targets; |
3d632c3f | 107 | static char *fail_over_mac; |
b07ea07b | 108 | static int all_slaves_active; |
d2991f75 | 109 | static struct bond_params bonding_defaults; |
c2952c31 | 110 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
73958329 | 111 | static int packets_per_slave = 1; |
3a7129e5 | 112 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
113 | |
114 | module_param(max_bonds, int, 0); | |
115 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
116 | module_param(tx_queues, int, 0); |
117 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 118 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
119 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
120 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 121 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
122 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
123 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
124 | module_param(miimon, int, 0); |
125 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
126 | module_param(updelay, int, 0); | |
127 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
128 | module_param(downdelay, int, 0); | |
2ac47660 MW |
129 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
130 | "in milliseconds"); | |
1da177e4 | 131 | module_param(use_carrier, int, 0); |
2ac47660 MW |
132 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
133 | "0 for off, 1 for on (default)"); | |
1da177e4 | 134 | module_param(mode, charp, 0); |
90e62474 | 135 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
136 | "1 for active-backup, 2 for balance-xor, " |
137 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
138 | "6 for balance-alb"); | |
1da177e4 LT |
139 | module_param(primary, charp, 0); |
140 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
141 | module_param(primary_reselect, charp, 0); |
142 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
143 | "once it comes up; " | |
144 | "0 for always (default), " | |
145 | "1 for only if speed of primary is " | |
146 | "better, " | |
147 | "2 for only on active slave " | |
148 | "failure"); | |
1da177e4 | 149 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
150 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
151 | "0 for slow, 1 for fast"); | |
fd989c83 | 152 | module_param(ad_select, charp, 0); |
90e62474 AG |
153 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
154 | "0 for stable (default), 1 for bandwidth, " | |
155 | "2 for count"); | |
655f8919 | 156 | module_param(min_links, int, 0); |
157 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
158 | ||
169a3e66 | 159 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
160 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
161 | "0 for layer 2 (default), 1 for layer 3+4, " | |
32819dc1 NA |
162 | "2 for layer 2+3, 3 for encap layer 2+3, " |
163 | "4 for encap layer 3+4"); | |
1da177e4 LT |
164 | module_param(arp_interval, int, 0); |
165 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
166 | module_param_array(arp_ip_target, charp, NULL, 0); | |
167 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 168 | module_param(arp_validate, charp, 0); |
90e62474 AG |
169 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
170 | "0 for none (default), 1 for active, " | |
171 | "2 for backup, 3 for all"); | |
8599b52e VF |
172 | module_param(arp_all_targets, charp, 0); |
173 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 174 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
175 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
176 | "the same MAC; 0 for none (default), " | |
177 | "1 for active, 2 for follow"); | |
ebd8e497 | 178 | module_param(all_slaves_active, int, 0); |
37b7021d | 179 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface " |
90e62474 | 180 | "by setting active flag for all slaves; " |
ebd8e497 | 181 | "0 for never (default), 1 for always."); |
c2952c31 | 182 | module_param(resend_igmp, int, 0); |
90e62474 AG |
183 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
184 | "link failure"); | |
73958329 NA |
185 | module_param(packets_per_slave, int, 0); |
186 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
187 | "mode; 0 for a random slave, 1 packet per " | |
188 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 189 | module_param(lp_interval, uint, 0); |
190 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
191 | "the bonding driver sends learning packets to " | |
192 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
193 | |
194 | /*----------------------------- Global variables ----------------------------*/ | |
195 | ||
e843fa50 | 196 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 197 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
198 | #endif |
199 | ||
f99189b1 | 200 | int bond_net_id __read_mostly; |
1da177e4 | 201 | |
3d632c3f SH |
202 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
203 | static int arp_ip_count; | |
1da177e4 | 204 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
205 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
206 | static int lacp_fast; | |
1da177e4 | 207 | |
1da177e4 LT |
208 | /*-------------------------- Forward declarations ---------------------------*/ |
209 | ||
181470fc | 210 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 211 | static void bond_uninit(struct net_device *bond_dev); |
5f0c5f73 AG |
212 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
213 | struct rtnl_link_stats64 *stats); | |
ee637714 | 214 | static void bond_slave_arr_handler(struct work_struct *work); |
1da177e4 LT |
215 | |
216 | /*---------------------------- General routines -----------------------------*/ | |
217 | ||
bd33acc3 | 218 | const char *bond_mode_name(int mode) |
1da177e4 | 219 | { |
77afc92b HE |
220 | static const char *names[] = { |
221 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
222 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
223 | [BOND_MODE_XOR] = "load balancing (xor)", | |
224 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 225 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
226 | [BOND_MODE_TLB] = "transmit load balancing", |
227 | [BOND_MODE_ALB] = "adaptive load balancing", | |
228 | }; | |
229 | ||
f5280948 | 230 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 231 | return "unknown"; |
77afc92b HE |
232 | |
233 | return names[mode]; | |
1da177e4 LT |
234 | } |
235 | ||
236 | /*---------------------------------- VLAN -----------------------------------*/ | |
237 | ||
1da177e4 LT |
238 | /** |
239 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 240 | * |
1da177e4 LT |
241 | * @bond: bond device that got this skb for tx. |
242 | * @skb: hw accel VLAN tagged skb to transmit | |
243 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 244 | */ |
d316dedd | 245 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 246 | struct net_device *slave_dev) |
1da177e4 | 247 | { |
8387451e | 248 | skb->dev = slave_dev; |
374eeb5a | 249 | |
5ee31c68 | 250 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
251 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
252 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 253 | |
e15c3c22 | 254 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 255 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 256 | else |
f6dc31a8 | 257 | dev_queue_xmit(skb); |
1da177e4 LT |
258 | } |
259 | ||
547942ca | 260 | /* In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
cc0e4070 | 261 | * We don't protect the slave list iteration with a lock because: |
1da177e4 LT |
262 | * a. This operation is performed in IOCTL context, |
263 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
264 | * c. Holding a lock with BH disabled while directly calling a base driver | |
265 | * entry point is generally a BAD idea. | |
3d632c3f | 266 | * |
1da177e4 LT |
267 | * The design of synchronization/protection for this operation in the 8021q |
268 | * module is good for one or more VLAN devices over a single physical device | |
269 | * and cannot be extended for a teaming solution like bonding, so there is a | |
270 | * potential race condition here where a net device from the vlan group might | |
271 | * be referenced (either by a base driver or the 8021q code) while it is being | |
272 | * removed from the system. However, it turns out we're not making matters | |
273 | * worse, and if it works for regular VLAN usage it will work here too. | |
274 | */ | |
275 | ||
1da177e4 LT |
276 | /** |
277 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
278 | * @bond_dev: bonding net device that got called | |
279 | * @vid: vlan id being added | |
280 | */ | |
80d5c368 PM |
281 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
282 | __be16 proto, u16 vid) | |
1da177e4 | 283 | { |
454d7c9b | 284 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 285 | struct slave *slave, *rollback_slave; |
9caff1e7 | 286 | struct list_head *iter; |
dec1e90e | 287 | int res; |
1da177e4 | 288 | |
9caff1e7 | 289 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 290 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
291 | if (res) |
292 | goto unwind; | |
1da177e4 LT |
293 | } |
294 | ||
8e586137 | 295 | return 0; |
87002b03 JP |
296 | |
297 | unwind: | |
81f23b13 | 298 | /* unwind to the slave that failed */ |
9caff1e7 | 299 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
300 | if (rollback_slave == slave) |
301 | break; | |
302 | ||
303 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
304 | } | |
87002b03 JP |
305 | |
306 | return res; | |
1da177e4 LT |
307 | } |
308 | ||
309 | /** | |
310 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
311 | * @bond_dev: bonding net device that got called | |
312 | * @vid: vlan id being removed | |
313 | */ | |
80d5c368 PM |
314 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
315 | __be16 proto, u16 vid) | |
1da177e4 | 316 | { |
454d7c9b | 317 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 318 | struct list_head *iter; |
1da177e4 | 319 | struct slave *slave; |
1da177e4 | 320 | |
9caff1e7 | 321 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 322 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 323 | |
e868b0c9 VF |
324 | if (bond_is_lb(bond)) |
325 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
326 | |
327 | return 0; | |
1da177e4 LT |
328 | } |
329 | ||
1da177e4 LT |
330 | /*------------------------------- Link status -------------------------------*/ |
331 | ||
547942ca | 332 | /* Set the carrier state for the master according to the state of its |
ff59c456 JV |
333 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, |
334 | * do special 802.3ad magic. | |
335 | * | |
336 | * Returns zero if carrier state does not change, nonzero if it does. | |
337 | */ | |
2477bc9a | 338 | int bond_set_carrier(struct bonding *bond) |
ff59c456 | 339 | { |
9caff1e7 | 340 | struct list_head *iter; |
ff59c456 | 341 | struct slave *slave; |
ff59c456 | 342 | |
0965a1f3 | 343 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
344 | goto down; |
345 | ||
01844098 | 346 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
ff59c456 JV |
347 | return bond_3ad_set_carrier(bond); |
348 | ||
9caff1e7 | 349 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
350 | if (slave->link == BOND_LINK_UP) { |
351 | if (!netif_carrier_ok(bond->dev)) { | |
352 | netif_carrier_on(bond->dev); | |
353 | return 1; | |
354 | } | |
355 | return 0; | |
356 | } | |
357 | } | |
358 | ||
359 | down: | |
360 | if (netif_carrier_ok(bond->dev)) { | |
361 | netif_carrier_off(bond->dev); | |
362 | return 1; | |
363 | } | |
364 | return 0; | |
365 | } | |
366 | ||
547942ca | 367 | /* Get link speed and duplex from the slave's base driver |
1da177e4 | 368 | * using ethtool. If for some reason the call fails or the |
98f41f69 | 369 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 370 | * and return. |
1da177e4 | 371 | */ |
13b95fb7 | 372 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
373 | { |
374 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 375 | struct ethtool_cmd ecmd; |
5d30530e | 376 | u32 slave_speed; |
61a44b9c | 377 | int res; |
1da177e4 | 378 | |
589665f5 DC |
379 | slave->speed = SPEED_UNKNOWN; |
380 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 381 | |
4bc71cb9 | 382 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 383 | if (res < 0) |
13b95fb7 | 384 | return; |
1da177e4 | 385 | |
4bc71cb9 | 386 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 387 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 388 | return; |
1da177e4 | 389 | |
4bc71cb9 | 390 | switch (ecmd.duplex) { |
1da177e4 LT |
391 | case DUPLEX_FULL: |
392 | case DUPLEX_HALF: | |
393 | break; | |
394 | default: | |
13b95fb7 | 395 | return; |
1da177e4 LT |
396 | } |
397 | ||
5d30530e | 398 | slave->speed = slave_speed; |
4bc71cb9 | 399 | slave->duplex = ecmd.duplex; |
1da177e4 | 400 | |
13b95fb7 | 401 | return; |
1da177e4 LT |
402 | } |
403 | ||
07699f9a | 404 | const char *bond_slave_link_status(s8 link) |
405 | { | |
406 | switch (link) { | |
407 | case BOND_LINK_UP: | |
408 | return "up"; | |
409 | case BOND_LINK_FAIL: | |
410 | return "going down"; | |
411 | case BOND_LINK_DOWN: | |
412 | return "down"; | |
413 | case BOND_LINK_BACK: | |
414 | return "going back"; | |
415 | default: | |
416 | return "unknown"; | |
417 | } | |
418 | } | |
419 | ||
547942ca | 420 | /* if <dev> supports MII link status reporting, check its link status. |
1da177e4 LT |
421 | * |
422 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 423 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
424 | * |
425 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
426 | * can't tell and just pretend it is), or 0, meaning that the link is | |
427 | * down. | |
428 | * | |
429 | * If reporting is non-zero, instead of faking link up, return -1 if | |
430 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
431 | * support them). If use_carrier is set, return whatever it says. | |
432 | * It'd be nice if there was a good way to tell if a driver supports | |
433 | * netif_carrier, but there really isn't. | |
434 | */ | |
3d632c3f SH |
435 | static int bond_check_dev_link(struct bonding *bond, |
436 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 437 | { |
eb7cc59a | 438 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 439 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
440 | struct ifreq ifr; |
441 | struct mii_ioctl_data *mii; | |
1da177e4 | 442 | |
6c988853 PG |
443 | if (!reporting && !netif_running(slave_dev)) |
444 | return 0; | |
445 | ||
eb7cc59a | 446 | if (bond->params.use_carrier) |
1da177e4 | 447 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 448 | |
29112f4e | 449 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
450 | if (slave_dev->ethtool_ops->get_link) |
451 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
452 | BMSR_LSTATUS : 0; | |
29112f4e | 453 | |
3d632c3f | 454 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 455 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 | 456 | if (ioctl) { |
547942ca NA |
457 | /* TODO: set pointer to correct ioctl on a per team member |
458 | * bases to make this more efficient. that is, once | |
459 | * we determine the correct ioctl, we will always | |
460 | * call it and not the others for that team | |
461 | * member. | |
462 | */ | |
463 | ||
464 | /* We cannot assume that SIOCGMIIPHY will also read a | |
1da177e4 LT |
465 | * register; not all network drivers (e.g., e100) |
466 | * support that. | |
467 | */ | |
468 | ||
469 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
470 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
471 | mii = if_mii(&ifr); | |
472 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
473 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
474 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
475 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
476 | } |
477 | } | |
478 | ||
547942ca | 479 | /* If reporting, report that either there's no dev->do_ioctl, |
61a44b9c | 480 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
481 | * cannot report link status). If not reporting, pretend |
482 | * we're ok. | |
483 | */ | |
3d632c3f | 484 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
485 | } |
486 | ||
487 | /*----------------------------- Multicast list ------------------------------*/ | |
488 | ||
547942ca | 489 | /* Push the promiscuity flag down to appropriate slaves */ |
7e1a1ac1 | 490 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 491 | { |
9caff1e7 | 492 | struct list_head *iter; |
7e1a1ac1 | 493 | int err = 0; |
9caff1e7 | 494 | |
ec0865a9 | 495 | if (bond_uses_primary(bond)) { |
14056e79 | 496 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 497 | |
4740d638 ED |
498 | if (curr_active) |
499 | err = dev_set_promiscuity(curr_active->dev, inc); | |
1da177e4 LT |
500 | } else { |
501 | struct slave *slave; | |
dec1e90e | 502 | |
9caff1e7 | 503 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
504 | err = dev_set_promiscuity(slave->dev, inc); |
505 | if (err) | |
506 | return err; | |
1da177e4 LT |
507 | } |
508 | } | |
7e1a1ac1 | 509 | return err; |
1da177e4 LT |
510 | } |
511 | ||
547942ca | 512 | /* Push the allmulti flag down to all slaves */ |
7e1a1ac1 | 513 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 514 | { |
9caff1e7 | 515 | struct list_head *iter; |
7e1a1ac1 | 516 | int err = 0; |
9caff1e7 | 517 | |
ec0865a9 | 518 | if (bond_uses_primary(bond)) { |
14056e79 | 519 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 520 | |
4740d638 ED |
521 | if (curr_active) |
522 | err = dev_set_allmulti(curr_active->dev, inc); | |
1da177e4 LT |
523 | } else { |
524 | struct slave *slave; | |
dec1e90e | 525 | |
9caff1e7 | 526 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
527 | err = dev_set_allmulti(slave->dev, inc); |
528 | if (err) | |
529 | return err; | |
1da177e4 LT |
530 | } |
531 | } | |
7e1a1ac1 | 532 | return err; |
1da177e4 LT |
533 | } |
534 | ||
547942ca | 535 | /* Retrieve the list of registered multicast addresses for the bonding |
5a37e8ca FL |
536 | * device and retransmit an IGMP JOIN request to the current active |
537 | * slave. | |
538 | */ | |
f2369109 | 539 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 540 | { |
f2369109 | 541 | struct bonding *bond = container_of(work, struct bonding, |
542 | mcast_work.work); | |
543 | ||
4aa5dee4 | 544 | if (!rtnl_trylock()) { |
4beac029 | 545 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 546 | return; |
5a37e8ca | 547 | } |
4aa5dee4 | 548 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 549 | |
4f5474e7 NA |
550 | if (bond->igmp_retrans > 1) { |
551 | bond->igmp_retrans--; | |
c2952c31 | 552 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 553 | } |
f2369109 | 554 | rtnl_unlock(); |
5a37e8ca FL |
555 | } |
556 | ||
547942ca | 557 | /* Flush bond's hardware addresses from slave */ |
303d1cbf | 558 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 559 | struct net_device *slave_dev) |
1da177e4 | 560 | { |
454d7c9b | 561 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 562 | |
303d1cbf JV |
563 | dev_uc_unsync(slave_dev, bond_dev); |
564 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 | 565 | |
01844098 | 566 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
567 | /* del lacpdu mc addr from mc list */ |
568 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
569 | ||
22bedad3 | 570 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
571 | } |
572 | } | |
573 | ||
574 | /*--------------------------- Active slave change ---------------------------*/ | |
575 | ||
303d1cbf | 576 | /* Update the hardware address list and promisc/allmulti for the new and |
ec0865a9 VF |
577 | * old active slaves (if any). Modes that are not using primary keep all |
578 | * slaves up date at all times; only the modes that use primary need to call | |
303d1cbf | 579 | * this function to swap these settings during a failover. |
1da177e4 | 580 | */ |
303d1cbf JV |
581 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
582 | struct slave *old_active) | |
1da177e4 | 583 | { |
1da177e4 | 584 | if (old_active) { |
3d632c3f | 585 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 586 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 587 | |
3d632c3f | 588 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 589 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 590 | |
303d1cbf | 591 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
592 | } |
593 | ||
594 | if (new_active) { | |
7e1a1ac1 | 595 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 596 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 597 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 598 | |
3d632c3f | 599 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 600 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 601 | |
d632ce98 | 602 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
603 | dev_uc_sync(new_active->dev, bond->dev); |
604 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 605 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
606 | } |
607 | } | |
608 | ||
ae0d6750 | 609 | /** |
610 | * bond_set_dev_addr - clone slave's address to bond | |
611 | * @bond_dev: bond net device | |
612 | * @slave_dev: slave net device | |
613 | * | |
614 | * Should be called with RTNL held. | |
615 | */ | |
616 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
617 | struct net_device *slave_dev) | |
618 | { | |
76444f50 VF |
619 | netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
620 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 621 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
622 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
623 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
624 | } | |
625 | ||
547942ca | 626 | /* bond_do_fail_over_mac |
3915c1e8 JV |
627 | * |
628 | * Perform special MAC address swapping for fail_over_mac settings | |
629 | * | |
1c72cfdc | 630 | * Called with RTNL |
3915c1e8 JV |
631 | */ |
632 | static void bond_do_fail_over_mac(struct bonding *bond, | |
633 | struct slave *new_active, | |
634 | struct slave *old_active) | |
635 | { | |
636 | u8 tmp_mac[ETH_ALEN]; | |
637 | struct sockaddr saddr; | |
638 | int rv; | |
639 | ||
640 | switch (bond->params.fail_over_mac) { | |
641 | case BOND_FOM_ACTIVE: | |
1c72cfdc | 642 | if (new_active) |
ae0d6750 | 643 | bond_set_dev_addr(bond->dev, new_active->dev); |
3915c1e8 JV |
644 | break; |
645 | case BOND_FOM_FOLLOW: | |
547942ca | 646 | /* if new_active && old_active, swap them |
3915c1e8 JV |
647 | * if just old_active, do nothing (going to no active slave) |
648 | * if just new_active, set new_active to bond's MAC | |
649 | */ | |
650 | if (!new_active) | |
651 | return; | |
652 | ||
3915c1e8 | 653 | if (old_active) { |
ada0f863 | 654 | ether_addr_copy(tmp_mac, new_active->dev->dev_addr); |
2a7c183b JP |
655 | ether_addr_copy(saddr.sa_data, |
656 | old_active->dev->dev_addr); | |
3915c1e8 JV |
657 | saddr.sa_family = new_active->dev->type; |
658 | } else { | |
ada0f863 | 659 | ether_addr_copy(saddr.sa_data, bond->dev->dev_addr); |
3915c1e8 JV |
660 | saddr.sa_family = bond->dev->type; |
661 | } | |
662 | ||
663 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
664 | if (rv) { | |
76444f50 VF |
665 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
666 | -rv, new_active->dev->name); | |
3915c1e8 JV |
667 | goto out; |
668 | } | |
669 | ||
670 | if (!old_active) | |
671 | goto out; | |
672 | ||
ada0f863 | 673 | ether_addr_copy(saddr.sa_data, tmp_mac); |
3915c1e8 JV |
674 | saddr.sa_family = old_active->dev->type; |
675 | ||
676 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
677 | if (rv) | |
76444f50 VF |
678 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
679 | -rv, new_active->dev->name); | |
3915c1e8 | 680 | out: |
3915c1e8 JV |
681 | break; |
682 | default: | |
76444f50 VF |
683 | netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n", |
684 | bond->params.fail_over_mac); | |
3915c1e8 JV |
685 | break; |
686 | } | |
687 | ||
688 | } | |
689 | ||
a549952a JP |
690 | static bool bond_should_change_active(struct bonding *bond) |
691 | { | |
059b47e8 | 692 | struct slave *prim = rtnl_dereference(bond->primary_slave); |
1c72cfdc | 693 | struct slave *curr = rtnl_dereference(bond->curr_active_slave); |
a549952a JP |
694 | |
695 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
696 | return true; | |
697 | if (bond->force_primary) { | |
698 | bond->force_primary = false; | |
699 | return true; | |
700 | } | |
701 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
702 | (prim->speed < curr->speed || | |
703 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
704 | return false; | |
705 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
706 | return false; | |
707 | return true; | |
708 | } | |
3915c1e8 | 709 | |
1da177e4 LT |
710 | /** |
711 | * find_best_interface - select the best available slave to be the active one | |
712 | * @bond: our bonding struct | |
1da177e4 LT |
713 | */ |
714 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
715 | { | |
059b47e8 | 716 | struct slave *slave, *bestslave = NULL, *primary; |
77140d29 | 717 | struct list_head *iter; |
1da177e4 | 718 | int mintime = bond->params.updelay; |
1da177e4 | 719 | |
059b47e8 NA |
720 | primary = rtnl_dereference(bond->primary_slave); |
721 | if (primary && primary->link == BOND_LINK_UP && | |
77140d29 | 722 | bond_should_change_active(bond)) |
059b47e8 | 723 | return primary; |
1da177e4 | 724 | |
77140d29 VF |
725 | bond_for_each_slave(bond, slave, iter) { |
726 | if (slave->link == BOND_LINK_UP) | |
727 | return slave; | |
b6adc610 | 728 | if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) && |
77140d29 VF |
729 | slave->delay < mintime) { |
730 | mintime = slave->delay; | |
731 | bestslave = slave; | |
1da177e4 LT |
732 | } |
733 | } | |
734 | ||
735 | return bestslave; | |
736 | } | |
737 | ||
ad246c99 BH |
738 | static bool bond_should_notify_peers(struct bonding *bond) |
739 | { | |
4cb4f97b | 740 | struct slave *slave; |
741 | ||
742 | rcu_read_lock(); | |
743 | slave = rcu_dereference(bond->curr_active_slave); | |
744 | rcu_read_unlock(); | |
ad246c99 | 745 | |
76444f50 VF |
746 | netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n", |
747 | slave ? slave->dev->name : "NULL"); | |
ad246c99 BH |
748 | |
749 | if (!slave || !bond->send_peer_notif || | |
750 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
751 | return false; | |
752 | ||
ad246c99 BH |
753 | return true; |
754 | } | |
755 | ||
1da177e4 LT |
756 | /** |
757 | * change_active_interface - change the active slave into the specified one | |
758 | * @bond: our bonding struct | |
759 | * @new: the new slave to make the active one | |
760 | * | |
761 | * Set the new slave to the bond's settings and unset them on the old | |
762 | * curr_active_slave. | |
763 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
764 | * | |
765 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
766 | * because it is apparently the best available slave we have, even though its | |
767 | * updelay hasn't timed out yet. | |
768 | * | |
1c72cfdc | 769 | * Caller must hold RTNL. |
1da177e4 | 770 | */ |
a77b5325 | 771 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 | 772 | { |
4740d638 ED |
773 | struct slave *old_active; |
774 | ||
1c72cfdc NA |
775 | ASSERT_RTNL(); |
776 | ||
777 | old_active = rtnl_dereference(bond->curr_active_slave); | |
1da177e4 | 778 | |
3d632c3f | 779 | if (old_active == new_active) |
1da177e4 | 780 | return; |
1da177e4 LT |
781 | |
782 | if (new_active) { | |
8e603460 | 783 | new_active->last_link_up = jiffies; |
b2220cad | 784 | |
1da177e4 | 785 | if (new_active->link == BOND_LINK_BACK) { |
ec0865a9 | 786 | if (bond_uses_primary(bond)) { |
76444f50 VF |
787 | netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n", |
788 | new_active->dev->name, | |
789 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
790 | } |
791 | ||
792 | new_active->delay = 0; | |
69a2338e | 793 | bond_set_slave_link_state(new_active, BOND_LINK_UP); |
1da177e4 | 794 | |
01844098 | 795 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 796 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 797 | |
58402054 | 798 | if (bond_is_lb(bond)) |
1da177e4 | 799 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 | 800 | } else { |
ec0865a9 | 801 | if (bond_uses_primary(bond)) { |
76444f50 VF |
802 | netdev_info(bond->dev, "making interface %s the new active one\n", |
803 | new_active->dev->name); | |
1da177e4 LT |
804 | } |
805 | } | |
806 | } | |
807 | ||
ec0865a9 | 808 | if (bond_uses_primary(bond)) |
303d1cbf | 809 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 810 | |
58402054 | 811 | if (bond_is_lb(bond)) { |
1da177e4 | 812 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 813 | if (old_active) |
5e5b0665 | 814 | bond_set_slave_inactive_flags(old_active, |
815 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 816 | if (new_active) |
5e5b0665 | 817 | bond_set_slave_active_flags(new_active, |
818 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 819 | } else { |
278b2083 | 820 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 821 | } |
c3ade5ca | 822 | |
01844098 | 823 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) { |
3d632c3f | 824 | if (old_active) |
5e5b0665 | 825 | bond_set_slave_inactive_flags(old_active, |
826 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
827 | |
828 | if (new_active) { | |
ad246c99 BH |
829 | bool should_notify_peers = false; |
830 | ||
5e5b0665 | 831 | bond_set_slave_active_flags(new_active, |
832 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 833 | |
709f8a45 OG |
834 | if (bond->params.fail_over_mac) |
835 | bond_do_fail_over_mac(bond, new_active, | |
836 | old_active); | |
3915c1e8 | 837 | |
ad246c99 BH |
838 | if (netif_running(bond->dev)) { |
839 | bond->send_peer_notif = | |
840 | bond->params.num_peer_notif; | |
841 | should_notify_peers = | |
842 | bond_should_notify_peers(bond); | |
843 | } | |
844 | ||
b7bc2a5b | 845 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 846 | if (should_notify_peers) |
b7bc2a5b AW |
847 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
848 | bond->dev); | |
7893b249 | 849 | } |
c3ade5ca | 850 | } |
a2fd940f | 851 | |
5a37e8ca | 852 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
853 | * all were sent on curr_active_slave. |
854 | * resend only if bond is brought up with the affected | |
547942ca NA |
855 | * bonding modes and the retransmission is enabled |
856 | */ | |
94265cf5 | 857 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && |
ec0865a9 | 858 | ((bond_uses_primary(bond) && new_active) || |
01844098 | 859 | BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) { |
c2952c31 | 860 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 861 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 862 | } |
1da177e4 LT |
863 | } |
864 | ||
865 | /** | |
866 | * bond_select_active_slave - select a new active slave, if needed | |
867 | * @bond: our bonding struct | |
868 | * | |
3d632c3f | 869 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
870 | * - The old curr_active_slave has been released or lost its link. |
871 | * - The primary_slave has got its link back. | |
872 | * - A slave has got its link back and there's no old curr_active_slave. | |
873 | * | |
1c72cfdc | 874 | * Caller must hold RTNL. |
1da177e4 | 875 | */ |
a77b5325 | 876 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
877 | { |
878 | struct slave *best_slave; | |
ff59c456 | 879 | int rv; |
1da177e4 | 880 | |
e0974585 NA |
881 | ASSERT_RTNL(); |
882 | ||
1da177e4 | 883 | best_slave = bond_find_best_slave(bond); |
1c72cfdc | 884 | if (best_slave != rtnl_dereference(bond->curr_active_slave)) { |
1da177e4 | 885 | bond_change_active_slave(bond, best_slave); |
ff59c456 JV |
886 | rv = bond_set_carrier(bond); |
887 | if (!rv) | |
888 | return; | |
889 | ||
890 | if (netif_carrier_ok(bond->dev)) { | |
76444f50 | 891 | netdev_info(bond->dev, "first active interface up!\n"); |
ff59c456 | 892 | } else { |
76444f50 | 893 | netdev_info(bond->dev, "now running without any active interface!\n"); |
ff59c456 | 894 | } |
1da177e4 LT |
895 | } |
896 | } | |
897 | ||
f6dc31a8 | 898 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 899 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 900 | { |
8a8efa22 AW |
901 | struct netpoll *np; |
902 | int err = 0; | |
f6dc31a8 | 903 | |
a8779ec1 | 904 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
905 | err = -ENOMEM; |
906 | if (!np) | |
907 | goto out; | |
908 | ||
a8779ec1 | 909 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
910 | if (err) { |
911 | kfree(np); | |
912 | goto out; | |
f6dc31a8 | 913 | } |
8a8efa22 AW |
914 | slave->np = np; |
915 | out: | |
916 | return err; | |
917 | } | |
918 | static inline void slave_disable_netpoll(struct slave *slave) | |
919 | { | |
920 | struct netpoll *np = slave->np; | |
921 | ||
922 | if (!np) | |
923 | return; | |
924 | ||
925 | slave->np = NULL; | |
2cde6acd | 926 | __netpoll_free_async(np); |
8a8efa22 | 927 | } |
f6dc31a8 WC |
928 | |
929 | static void bond_poll_controller(struct net_device *bond_dev) | |
930 | { | |
8a8efa22 AW |
931 | } |
932 | ||
c4cdef9b | 933 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 934 | { |
c4cdef9b | 935 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 936 | struct list_head *iter; |
c2355e1a | 937 | struct slave *slave; |
c2355e1a | 938 | |
9caff1e7 | 939 | bond_for_each_slave(bond, slave, iter) |
b6adc610 | 940 | if (bond_slave_is_up(slave)) |
8a8efa22 | 941 | slave_disable_netpoll(slave); |
f6dc31a8 | 942 | } |
8a8efa22 | 943 | |
a8779ec1 | 944 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
945 | { |
946 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 947 | struct list_head *iter; |
f6dc31a8 | 948 | struct slave *slave; |
dec1e90e | 949 | int err = 0; |
f6dc31a8 | 950 | |
9caff1e7 | 951 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
952 | err = slave_enable_netpoll(slave); |
953 | if (err) { | |
c4cdef9b | 954 | bond_netpoll_cleanup(dev); |
8a8efa22 | 955 | break; |
f6dc31a8 WC |
956 | } |
957 | } | |
8a8efa22 | 958 | return err; |
f6dc31a8 | 959 | } |
8a8efa22 AW |
960 | #else |
961 | static inline int slave_enable_netpoll(struct slave *slave) | |
962 | { | |
963 | return 0; | |
964 | } | |
965 | static inline void slave_disable_netpoll(struct slave *slave) | |
966 | { | |
967 | } | |
f6dc31a8 WC |
968 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
969 | { | |
970 | } | |
f6dc31a8 WC |
971 | #endif |
972 | ||
1da177e4 LT |
973 | /*---------------------------------- IOCTL ----------------------------------*/ |
974 | ||
c8f44aff | 975 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 976 | netdev_features_t features) |
8531c5ff | 977 | { |
b2a103e6 | 978 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 979 | struct list_head *iter; |
c8f44aff | 980 | netdev_features_t mask; |
9b7b165a | 981 | struct slave *slave; |
b63365a2 | 982 | |
c158cba3 RP |
983 | /* If any slave has the offload feature flag set, |
984 | * set the offload flag on the bond. | |
985 | */ | |
986 | mask = features | NETIF_F_HW_SWITCH_OFFLOAD; | |
987 | ||
b63365a2 | 988 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 989 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 990 | |
9caff1e7 | 991 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
992 | features = netdev_increment_features(features, |
993 | slave->dev->features, | |
b2a103e6 MM |
994 | mask); |
995 | } | |
b0ce3508 | 996 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 997 | |
b2a103e6 MM |
998 | return features; |
999 | } | |
1000 | ||
62f2a3a4 MM |
1001 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1002 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1003 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 | 1004 | |
5a7baa78 | 1005 | #define BOND_ENC_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | NETIF_F_RXCSUM |\ |
24f87d4c | 1006 | NETIF_F_TSO) |
5a7baa78 | 1007 | |
b2a103e6 MM |
1008 | static void bond_compute_features(struct bonding *bond) |
1009 | { | |
02875878 ED |
1010 | unsigned int dst_release_flag = IFF_XMIT_DST_RELEASE | |
1011 | IFF_XMIT_DST_RELEASE_PERM; | |
c8f44aff | 1012 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
5a7baa78 | 1013 | netdev_features_t enc_features = BOND_ENC_FEATURES; |
9caff1e7 VF |
1014 | struct net_device *bond_dev = bond->dev; |
1015 | struct list_head *iter; | |
1016 | struct slave *slave; | |
b2a103e6 | 1017 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1018 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1019 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1020 | |
0965a1f3 | 1021 | if (!bond_has_slaves(bond)) |
b2a103e6 | 1022 | goto done; |
a9b3ace4 | 1023 | vlan_features &= NETIF_F_ALL_FOR_ALL; |
b2a103e6 | 1024 | |
9caff1e7 | 1025 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1026 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1027 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1028 | ||
5a7baa78 OG |
1029 | enc_features = netdev_increment_features(enc_features, |
1030 | slave->dev->hw_enc_features, | |
1031 | BOND_ENC_FEATURES); | |
b6fe83e9 | 1032 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1033 | if (slave->dev->hard_header_len > max_hard_header_len) |
1034 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1035 | |
1036 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1037 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1038 | } |
8531c5ff | 1039 | |
b63365a2 | 1040 | done: |
b2a103e6 | 1041 | bond_dev->vlan_features = vlan_features; |
24f87d4c | 1042 | bond_dev->hw_enc_features = enc_features | NETIF_F_GSO_ENCAP_ALL; |
54ef3137 | 1043 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1044 | bond_dev->gso_max_segs = gso_max_segs; |
1045 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1046 | |
02875878 ED |
1047 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
1048 | if ((bond_dev->priv_flags & IFF_XMIT_DST_RELEASE_PERM) && | |
1049 | dst_release_flag == (IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM)) | |
1050 | bond_dev->priv_flags |= IFF_XMIT_DST_RELEASE; | |
b6fe83e9 | 1051 | |
b2a103e6 | 1052 | netdev_change_features(bond_dev); |
8531c5ff AK |
1053 | } |
1054 | ||
872254dd MS |
1055 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1056 | struct net_device *slave_dev) | |
1057 | { | |
00829823 | 1058 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1059 | |
1060 | bond_dev->type = slave_dev->type; | |
1061 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1062 | bond_dev->addr_len = slave_dev->addr_len; | |
1063 | ||
1064 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1065 | slave_dev->addr_len); | |
1066 | } | |
1067 | ||
5b2c4dd2 | 1068 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1069 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1070 | */ |
1071 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1072 | struct slave *slave, |
1073 | struct bonding *bond) | |
5b2c4dd2 | 1074 | { |
2d7011ca | 1075 | if (bond_is_slave_inactive(slave)) { |
01844098 | 1076 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
5b2c4dd2 JP |
1077 | skb->pkt_type != PACKET_BROADCAST && |
1078 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1079 | return false; |
5b2c4dd2 JP |
1080 | return true; |
1081 | } | |
1082 | return false; | |
1083 | } | |
1084 | ||
8a4eb573 | 1085 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1086 | { |
8a4eb573 | 1087 | struct sk_buff *skb = *pskb; |
f1c1775a | 1088 | struct slave *slave; |
0bd80dad | 1089 | struct bonding *bond; |
de063b70 ED |
1090 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1091 | struct slave *); | |
13a8e0c8 | 1092 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1093 | |
8a4eb573 JP |
1094 | skb = skb_share_check(skb, GFP_ATOMIC); |
1095 | if (unlikely(!skb)) | |
1096 | return RX_HANDLER_CONSUMED; | |
1097 | ||
1098 | *pskb = skb; | |
5b2c4dd2 | 1099 | |
35d48903 JP |
1100 | slave = bond_slave_get_rcu(skb->dev); |
1101 | bond = slave->bond; | |
0bd80dad | 1102 | |
4d97480b MH |
1103 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1104 | if (recv_probe) { | |
de063b70 ED |
1105 | ret = recv_probe(skb, bond, slave); |
1106 | if (ret == RX_HANDLER_CONSUMED) { | |
1107 | consume_skb(skb); | |
1108 | return ret; | |
3aba891d JP |
1109 | } |
1110 | } | |
1111 | ||
0bd80dad | 1112 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1113 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1114 | } |
1115 | ||
35d48903 | 1116 | skb->dev = bond->dev; |
5b2c4dd2 | 1117 | |
01844098 | 1118 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
35d48903 | 1119 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1120 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1121 | |
541ac7c9 CG |
1122 | if (unlikely(skb_cow_head(skb, |
1123 | skb->data - skb_mac_header(skb)))) { | |
1124 | kfree_skb(skb); | |
8a4eb573 | 1125 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1126 | } |
2a7c183b | 1127 | ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr); |
5b2c4dd2 JP |
1128 | } |
1129 | ||
13a8e0c8 | 1130 | return ret; |
5b2c4dd2 JP |
1131 | } |
1132 | ||
471cb5a3 | 1133 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1f718f0f VF |
1134 | struct net_device *slave_dev, |
1135 | struct slave *slave) | |
471cb5a3 JP |
1136 | { |
1137 | int err; | |
1138 | ||
1f718f0f | 1139 | err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave); |
471cb5a3 JP |
1140 | if (err) |
1141 | return err; | |
1142 | slave_dev->flags |= IFF_SLAVE; | |
7f294054 | 1143 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1144 | return 0; |
1145 | } | |
1146 | ||
1147 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1148 | struct net_device *slave_dev) | |
1149 | { | |
1150 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1151 | slave_dev->flags &= ~IFF_SLAVE; | |
7f294054 | 1152 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1153 | } |
1154 | ||
3fdddd85 | 1155 | static struct slave *bond_alloc_slave(struct bonding *bond) |
1156 | { | |
1157 | struct slave *slave = NULL; | |
1158 | ||
1159 | slave = kzalloc(sizeof(struct slave), GFP_KERNEL); | |
1160 | if (!slave) | |
1161 | return NULL; | |
1162 | ||
01844098 | 1163 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
3fdddd85 | 1164 | SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info), |
1165 | GFP_KERNEL); | |
1166 | if (!SLAVE_AD_INFO(slave)) { | |
1167 | kfree(slave); | |
1168 | return NULL; | |
1169 | } | |
1170 | } | |
1171 | return slave; | |
1172 | } | |
1173 | ||
1174 | static void bond_free_slave(struct slave *slave) | |
1175 | { | |
1176 | struct bonding *bond = bond_get_bond_by_slave(slave); | |
1177 | ||
01844098 | 1178 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3fdddd85 | 1179 | kfree(SLAVE_AD_INFO(slave)); |
1180 | ||
1181 | kfree(slave); | |
1182 | } | |
1183 | ||
69a2338e MS |
1184 | static void bond_fill_ifbond(struct bonding *bond, struct ifbond *info) |
1185 | { | |
1186 | info->bond_mode = BOND_MODE(bond); | |
1187 | info->miimon = bond->params.miimon; | |
1188 | info->num_slaves = bond->slave_cnt; | |
1189 | } | |
1190 | ||
1191 | static void bond_fill_ifslave(struct slave *slave, struct ifslave *info) | |
1192 | { | |
1193 | strcpy(info->slave_name, slave->dev->name); | |
1194 | info->link = slave->link; | |
1195 | info->state = bond_slave_state(slave); | |
1196 | info->link_failure_count = slave->link_failure_count; | |
1197 | } | |
1198 | ||
69e61133 MS |
1199 | static void bond_netdev_notify(struct slave *slave, struct net_device *dev) |
1200 | { | |
1201 | struct bonding *bond = slave->bond; | |
1202 | struct netdev_bonding_info bonding_info; | |
1203 | ||
1204 | rtnl_lock(); | |
1205 | /* make sure that slave is still valid */ | |
1206 | if (dev->priv_flags & IFF_BONDING) { | |
1207 | bond_fill_ifslave(slave, &bonding_info.slave); | |
1208 | bond_fill_ifbond(bond, &bonding_info.master); | |
1209 | netdev_bonding_info_change(slave->dev, &bonding_info); | |
1210 | } | |
1211 | rtnl_unlock(); | |
1212 | } | |
1213 | ||
1214 | static void bond_netdev_notify_work(struct work_struct *_work) | |
1215 | { | |
1216 | struct netdev_notify_work *w = | |
1217 | container_of(_work, struct netdev_notify_work, work.work); | |
1218 | ||
1219 | bond_netdev_notify(w->slave, w->dev); | |
1220 | dev_put(w->dev); | |
1221 | } | |
1222 | ||
1223 | void bond_queue_slave_event(struct slave *slave) | |
1224 | { | |
1225 | struct netdev_notify_work *nnw = kzalloc(sizeof(*nnw), GFP_ATOMIC); | |
1226 | ||
1227 | if (!nnw) | |
1228 | return; | |
1229 | ||
1230 | INIT_DELAYED_WORK(&nnw->work, bond_netdev_notify_work); | |
1231 | nnw->slave = slave; | |
1232 | nnw->dev = slave->dev; | |
1233 | ||
1234 | if (queue_delayed_work(slave->bond->wq, &nnw->work, 0)) | |
1235 | dev_hold(slave->dev); | |
1236 | else | |
1237 | kfree(nnw); | |
1238 | } | |
1239 | ||
1da177e4 | 1240 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1241 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1242 | { |
454d7c9b | 1243 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1244 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1245 | struct slave *new_slave = NULL, *prev_slave; |
1da177e4 LT |
1246 | struct sockaddr addr; |
1247 | int link_reporting; | |
8599b52e | 1248 | int res = 0, i; |
1da177e4 | 1249 | |
c772dde3 BH |
1250 | if (!bond->params.use_carrier && |
1251 | slave_dev->ethtool_ops->get_link == NULL && | |
1252 | slave_ops->ndo_do_ioctl == NULL) { | |
76444f50 VF |
1253 | netdev_warn(bond_dev, "no link monitoring support for %s\n", |
1254 | slave_dev->name); | |
1da177e4 LT |
1255 | } |
1256 | ||
1da177e4 LT |
1257 | /* already enslaved */ |
1258 | if (slave_dev->flags & IFF_SLAVE) { | |
76444f50 | 1259 | netdev_dbg(bond_dev, "Error: Device was already enslaved\n"); |
1da177e4 LT |
1260 | return -EBUSY; |
1261 | } | |
1262 | ||
09a89c21 | 1263 | if (bond_dev == slave_dev) { |
76444f50 | 1264 | netdev_err(bond_dev, "cannot enslave bond to itself.\n"); |
09a89c21 JB |
1265 | return -EPERM; |
1266 | } | |
1267 | ||
1da177e4 LT |
1268 | /* vlan challenged mutual exclusion */ |
1269 | /* no need to lock since we're protected by rtnl_lock */ | |
1270 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
76444f50 VF |
1271 | netdev_dbg(bond_dev, "%s is NETIF_F_VLAN_CHALLENGED\n", |
1272 | slave_dev->name); | |
55462cf3 | 1273 | if (vlan_uses_dev(bond_dev)) { |
76444f50 VF |
1274 | netdev_err(bond_dev, "Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1275 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1276 | return -EPERM; |
1277 | } else { | |
76444f50 VF |
1278 | netdev_warn(bond_dev, "enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1279 | slave_dev->name, slave_dev->name, | |
1280 | bond_dev->name); | |
1da177e4 LT |
1281 | } |
1282 | } else { | |
76444f50 VF |
1283 | netdev_dbg(bond_dev, "%s is !NETIF_F_VLAN_CHALLENGED\n", |
1284 | slave_dev->name); | |
1da177e4 LT |
1285 | } |
1286 | ||
547942ca | 1287 | /* Old ifenslave binaries are no longer supported. These can |
3d632c3f | 1288 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1289 | * the current ifenslave will set the interface down prior to |
1290 | * enslaving it; the old ifenslave will not. | |
1291 | */ | |
1292 | if ((slave_dev->flags & IFF_UP)) { | |
76444f50 VF |
1293 | netdev_err(bond_dev, "%s is up - this may be due to an out of date ifenslave\n", |
1294 | slave_dev->name); | |
217df670 JV |
1295 | res = -EPERM; |
1296 | goto err_undo_flags; | |
1297 | } | |
1da177e4 | 1298 | |
872254dd MS |
1299 | /* set bonding device ether type by slave - bonding netdevices are |
1300 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1301 | * there is a need to override some of the type dependent attribs/funcs. | |
1302 | * | |
1303 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1304 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1305 | */ | |
0965a1f3 | 1306 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1307 | if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1308 | netdev_dbg(bond_dev, "change device type from %d to %d\n", |
1309 | bond_dev->type, slave_dev->type); | |
75c78500 | 1310 | |
b7bc2a5b AW |
1311 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1312 | bond_dev); | |
3ca5b404 JP |
1313 | res = notifier_to_errno(res); |
1314 | if (res) { | |
76444f50 | 1315 | netdev_err(bond_dev, "refused to change device type\n"); |
3ca5b404 JP |
1316 | res = -EBUSY; |
1317 | goto err_undo_flags; | |
1318 | } | |
75c78500 | 1319 | |
32a806c1 | 1320 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1321 | dev_uc_flush(bond_dev); |
22bedad3 | 1322 | dev_mc_flush(bond_dev); |
32a806c1 | 1323 | |
e36b9d16 MS |
1324 | if (slave_dev->type != ARPHRD_ETHER) |
1325 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1326 | else { |
e36b9d16 | 1327 | ether_setup(bond_dev); |
550fd08c NH |
1328 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1329 | } | |
75c78500 | 1330 | |
b7bc2a5b AW |
1331 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1332 | bond_dev); | |
e36b9d16 | 1333 | } |
872254dd | 1334 | } else if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1335 | netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n", |
1336 | slave_dev->name, slave_dev->type, bond_dev->type); | |
a4aee5c8 JP |
1337 | res = -EINVAL; |
1338 | goto err_undo_flags; | |
872254dd MS |
1339 | } |
1340 | ||
eb7cc59a | 1341 | if (slave_ops->ndo_set_mac_address == NULL) { |
76444f50 | 1342 | netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n"); |
f5442441 VF |
1343 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP && |
1344 | bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
1345 | if (!bond_has_slaves(bond)) { | |
00503b6f | 1346 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
76444f50 | 1347 | netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n"); |
f5442441 | 1348 | } else { |
76444f50 | 1349 | netdev_err(bond_dev, "The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n"); |
f5442441 VF |
1350 | res = -EOPNOTSUPP; |
1351 | goto err_undo_flags; | |
00503b6f | 1352 | } |
2ab82852 | 1353 | } |
1da177e4 LT |
1354 | } |
1355 | ||
8d8fc29d AW |
1356 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1357 | ||
c20811a7 | 1358 | /* If this is the first slave, then we need to set the master's hardware |
547942ca NA |
1359 | * address to be the same as the slave's. |
1360 | */ | |
0965a1f3 | 1361 | if (!bond_has_slaves(bond) && |
dec1e90e | 1362 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1363 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1364 | |
3fdddd85 | 1365 | new_slave = bond_alloc_slave(bond); |
1da177e4 LT |
1366 | if (!new_slave) { |
1367 | res = -ENOMEM; | |
1368 | goto err_undo_flags; | |
1369 | } | |
3fdddd85 | 1370 | |
dc73c41f VF |
1371 | new_slave->bond = bond; |
1372 | new_slave->dev = slave_dev; | |
547942ca | 1373 | /* Set the new_slave's queue_id to be zero. Queue ID mapping |
bb1d9123 AG |
1374 | * is set via sysfs or module option if desired. |
1375 | */ | |
1376 | new_slave->queue_id = 0; | |
1377 | ||
b15ba0fb JP |
1378 | /* Save slave's original mtu and then set it to match the bond */ |
1379 | new_slave->original_mtu = slave_dev->mtu; | |
1380 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1381 | if (res) { | |
76444f50 | 1382 | netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res); |
b15ba0fb JP |
1383 | goto err_free; |
1384 | } | |
1385 | ||
547942ca | 1386 | /* Save slave's original ("permanent") mac address for modes |
217df670 JV |
1387 | * that need it, and for restoring it upon release, and then |
1388 | * set it to the master's address | |
1389 | */ | |
ada0f863 | 1390 | ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr); |
1da177e4 | 1391 | |
00503b6f | 1392 | if (!bond->params.fail_over_mac || |
01844098 | 1393 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
547942ca | 1394 | /* Set slave to master's mac address. The application already |
2ab82852 MS |
1395 | * set the master's mac address to that of the first slave |
1396 | */ | |
1397 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1398 | addr.sa_family = slave_dev->type; | |
1399 | res = dev_set_mac_address(slave_dev, &addr); | |
1400 | if (res) { | |
76444f50 | 1401 | netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res); |
b15ba0fb | 1402 | goto err_restore_mtu; |
2ab82852 | 1403 | } |
217df670 | 1404 | } |
1da177e4 | 1405 | |
217df670 JV |
1406 | /* open the slave since the application closed it */ |
1407 | res = dev_open(slave_dev); | |
1408 | if (res) { | |
76444f50 | 1409 | netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1410 | goto err_restore_mac; |
1da177e4 LT |
1411 | } |
1412 | ||
0b680e75 | 1413 | slave_dev->priv_flags |= IFF_BONDING; |
5f0c5f73 AG |
1414 | /* initialize slave stats */ |
1415 | dev_get_stats(new_slave->dev, &new_slave->slave_stats); | |
1da177e4 | 1416 | |
58402054 | 1417 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1418 | /* bond_alb_init_slave() must be called before all other stages since |
1419 | * it might fail and we do not want to have to undo everything | |
1420 | */ | |
1421 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1422 | if (res) |
569f0c4d | 1423 | goto err_close; |
1da177e4 LT |
1424 | } |
1425 | ||
ec0865a9 | 1426 | /* If the mode uses primary, then the following is handled by |
303d1cbf | 1427 | * bond_change_active_slave(). |
1da177e4 | 1428 | */ |
ec0865a9 | 1429 | if (!bond_uses_primary(bond)) { |
1da177e4 LT |
1430 | /* set promiscuity level to new slave */ |
1431 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1432 | res = dev_set_promiscuity(slave_dev, 1); |
1433 | if (res) | |
1434 | goto err_close; | |
1da177e4 LT |
1435 | } |
1436 | ||
1437 | /* set allmulti level to new slave */ | |
1438 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1439 | res = dev_set_allmulti(slave_dev, 1); |
1440 | if (res) | |
1441 | goto err_close; | |
1da177e4 LT |
1442 | } |
1443 | ||
b9e40857 | 1444 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1445 | |
1446 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1447 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1448 | ||
b9e40857 | 1449 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1450 | } |
1451 | ||
01844098 | 1452 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
1453 | /* add lacpdu mc addr to mc list */ |
1454 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1455 | ||
22bedad3 | 1456 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1457 | } |
1458 | ||
b8e2fde4 WY |
1459 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1460 | if (res) { | |
76444f50 VF |
1461 | netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n", |
1462 | slave_dev->name); | |
1ff412ad | 1463 | goto err_close; |
1464 | } | |
1da177e4 | 1465 | |
c8c23903 | 1466 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1467 | |
1468 | new_slave->delay = 0; | |
1469 | new_slave->link_failure_count = 0; | |
1470 | ||
876254ae VF |
1471 | bond_update_speed_duplex(new_slave); |
1472 | ||
49f17de7 | 1473 | new_slave->last_rx = jiffies - |
f31c7937 | 1474 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1475 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1476 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1477 | |
1da177e4 LT |
1478 | if (bond->params.miimon && !bond->params.use_carrier) { |
1479 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1480 | ||
1481 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
547942ca | 1482 | /* miimon is set but a bonded network driver |
1da177e4 LT |
1483 | * does not support ETHTOOL/MII and |
1484 | * arp_interval is not set. Note: if | |
1485 | * use_carrier is enabled, we will never go | |
1486 | * here (because netif_carrier is always | |
1487 | * supported); thus, we don't need to change | |
1488 | * the messages for netif_carrier. | |
1489 | */ | |
76444f50 VF |
1490 | netdev_warn(bond_dev, "MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details\n", |
1491 | slave_dev->name); | |
1da177e4 LT |
1492 | } else if (link_reporting == -1) { |
1493 | /* unable get link status using mii/ethtool */ | |
76444f50 VF |
1494 | netdev_warn(bond_dev, "can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface\n", |
1495 | slave_dev->name); | |
1da177e4 LT |
1496 | } |
1497 | } | |
1498 | ||
1499 | /* check for initial state */ | |
f31c7937 MK |
1500 | if (bond->params.miimon) { |
1501 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1502 | if (bond->params.updelay) { | |
69a2338e MS |
1503 | bond_set_slave_link_state(new_slave, |
1504 | BOND_LINK_BACK); | |
f31c7937 MK |
1505 | new_slave->delay = bond->params.updelay; |
1506 | } else { | |
69a2338e MS |
1507 | bond_set_slave_link_state(new_slave, |
1508 | BOND_LINK_UP); | |
f31c7937 | 1509 | } |
1da177e4 | 1510 | } else { |
69a2338e | 1511 | bond_set_slave_link_state(new_slave, BOND_LINK_DOWN); |
1da177e4 | 1512 | } |
f31c7937 | 1513 | } else if (bond->params.arp_interval) { |
69a2338e MS |
1514 | bond_set_slave_link_state(new_slave, |
1515 | (netif_carrier_ok(slave_dev) ? | |
1516 | BOND_LINK_UP : BOND_LINK_DOWN)); | |
1da177e4 | 1517 | } else { |
69a2338e | 1518 | bond_set_slave_link_state(new_slave, BOND_LINK_UP); |
1da177e4 LT |
1519 | } |
1520 | ||
f31c7937 | 1521 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 1522 | new_slave->last_link_up = jiffies; |
76444f50 VF |
1523 | netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n", |
1524 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1525 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 1526 | |
ec0865a9 | 1527 | if (bond_uses_primary(bond) && bond->params.primary[0]) { |
1da177e4 | 1528 | /* if there is a primary slave, remember it */ |
a549952a | 1529 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
059b47e8 | 1530 | rcu_assign_pointer(bond->primary_slave, new_slave); |
a549952a JP |
1531 | bond->force_primary = true; |
1532 | } | |
1da177e4 LT |
1533 | } |
1534 | ||
01844098 | 1535 | switch (BOND_MODE(bond)) { |
1da177e4 | 1536 | case BOND_MODE_ACTIVEBACKUP: |
5e5b0665 | 1537 | bond_set_slave_inactive_flags(new_slave, |
1538 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1539 | break; |
1540 | case BOND_MODE_8023AD: | |
1541 | /* in 802.3ad mode, the internal mechanism | |
1542 | * will activate the slaves in the selected | |
1543 | * aggregator | |
1544 | */ | |
5e5b0665 | 1545 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 1546 | /* if this is the first slave */ |
23c147e0 | 1547 | if (!prev_slave) { |
3fdddd85 | 1548 | SLAVE_AD_INFO(new_slave)->id = 1; |
1da177e4 LT |
1549 | /* Initialize AD with the number of times that the AD timer is called in 1 second |
1550 | * can be called only after the mac address of the bond is set | |
1551 | */ | |
56d00c67 | 1552 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1553 | } else { |
3fdddd85 | 1554 | SLAVE_AD_INFO(new_slave)->id = |
1555 | SLAVE_AD_INFO(prev_slave)->id + 1; | |
1da177e4 LT |
1556 | } |
1557 | ||
1558 | bond_3ad_bind_slave(new_slave); | |
1559 | break; | |
1560 | case BOND_MODE_TLB: | |
1561 | case BOND_MODE_ALB: | |
e30bc066 | 1562 | bond_set_active_slave(new_slave); |
5e5b0665 | 1563 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
1564 | break; |
1565 | default: | |
76444f50 | 1566 | netdev_dbg(bond_dev, "This slave is always active in trunk mode\n"); |
1da177e4 LT |
1567 | |
1568 | /* always active in trunk mode */ | |
e30bc066 | 1569 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1570 | |
1571 | /* In trunking mode there is little meaning to curr_active_slave | |
1572 | * anyway (it holds no special properties of the bond device), | |
1573 | * so we can change it without calling change_active_interface() | |
1574 | */ | |
4740d638 ED |
1575 | if (!rcu_access_pointer(bond->curr_active_slave) && |
1576 | new_slave->link == BOND_LINK_UP) | |
278b2083 | 1577 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1578 | |
1da177e4 LT |
1579 | break; |
1580 | } /* switch(bond_mode) */ | |
1581 | ||
f6dc31a8 | 1582 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1583 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1584 | if (slave_dev->npinfo) { |
1585 | if (slave_enable_netpoll(new_slave)) { | |
76444f50 | 1586 | netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n"); |
8a8efa22 | 1587 | res = -EBUSY; |
f7d9821a | 1588 | goto err_detach; |
8a8efa22 | 1589 | } |
f6dc31a8 WC |
1590 | } |
1591 | #endif | |
8a8efa22 | 1592 | |
fbe168ba MK |
1593 | if (!(bond_dev->features & NETIF_F_LRO)) |
1594 | dev_disable_lro(slave_dev); | |
1595 | ||
35d48903 JP |
1596 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1597 | new_slave); | |
1598 | if (res) { | |
76444f50 | 1599 | netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res); |
5831d66e | 1600 | goto err_detach; |
35d48903 JP |
1601 | } |
1602 | ||
1f718f0f VF |
1603 | res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave); |
1604 | if (res) { | |
76444f50 | 1605 | netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res); |
1f718f0f VF |
1606 | goto err_unregister; |
1607 | } | |
1608 | ||
07699f9a | 1609 | res = bond_sysfs_slave_add(new_slave); |
1610 | if (res) { | |
76444f50 | 1611 | netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res); |
07699f9a | 1612 | goto err_upper_unlink; |
1613 | } | |
1614 | ||
5378c2e6 VF |
1615 | bond->slave_cnt++; |
1616 | bond_compute_features(bond); | |
1617 | bond_set_carrier(bond); | |
1618 | ||
ec0865a9 | 1619 | if (bond_uses_primary(bond)) { |
f80889a5 | 1620 | block_netpoll_tx(); |
5378c2e6 | 1621 | bond_select_active_slave(bond); |
f80889a5 | 1622 | unblock_netpoll_tx(); |
5378c2e6 | 1623 | } |
1f718f0f | 1624 | |
ee637714 MB |
1625 | if (bond_mode_uses_xmit_hash(bond)) |
1626 | bond_update_slave_arr(bond, NULL); | |
1627 | ||
76444f50 VF |
1628 | netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n", |
1629 | slave_dev->name, | |
1630 | bond_is_active_slave(new_slave) ? "an active" : "a backup", | |
1631 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
1632 | |
1633 | /* enslave is successful */ | |
69e61133 | 1634 | bond_queue_slave_event(new_slave); |
1da177e4 LT |
1635 | return 0; |
1636 | ||
1637 | /* Undo stages on error */ | |
07699f9a | 1638 | err_upper_unlink: |
1639 | bond_upper_dev_unlink(bond_dev, slave_dev); | |
1640 | ||
1f718f0f VF |
1641 | err_unregister: |
1642 | netdev_rx_handler_unregister(slave_dev); | |
1643 | ||
f7d9821a | 1644 | err_detach: |
ec0865a9 | 1645 | if (!bond_uses_primary(bond)) |
303d1cbf JV |
1646 | bond_hw_addr_flush(bond_dev, slave_dev); |
1647 | ||
1ff412ad | 1648 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
059b47e8 NA |
1649 | if (rcu_access_pointer(bond->primary_slave) == new_slave) |
1650 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
4740d638 | 1651 | if (rcu_access_pointer(bond->curr_active_slave) == new_slave) { |
f80889a5 | 1652 | block_netpoll_tx(); |
c8517035 | 1653 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1654 | bond_select_active_slave(bond); |
f80889a5 | 1655 | unblock_netpoll_tx(); |
3c5913b5 | 1656 | } |
059b47e8 NA |
1657 | /* either primary_slave or curr_active_slave might've changed */ |
1658 | synchronize_rcu(); | |
fc7a72ac | 1659 | slave_disable_netpoll(new_slave); |
f7d9821a | 1660 | |
1da177e4 | 1661 | err_close: |
b6a5a7b9 | 1662 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1663 | dev_close(slave_dev); |
1664 | ||
1665 | err_restore_mac: | |
00503b6f | 1666 | if (!bond->params.fail_over_mac || |
01844098 | 1667 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
3915c1e8 JV |
1668 | /* XXX TODO - fom follow mode needs to change master's |
1669 | * MAC if this slave's MAC is in use by the bond, or at | |
1670 | * least print a warning. | |
1671 | */ | |
ada0f863 | 1672 | ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr); |
2ab82852 MS |
1673 | addr.sa_family = slave_dev->type; |
1674 | dev_set_mac_address(slave_dev, &addr); | |
1675 | } | |
1da177e4 | 1676 | |
b15ba0fb JP |
1677 | err_restore_mtu: |
1678 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1679 | ||
1da177e4 | 1680 | err_free: |
3fdddd85 | 1681 | bond_free_slave(new_slave); |
1da177e4 LT |
1682 | |
1683 | err_undo_flags: | |
b8fad459 | 1684 | /* Enslave of first slave has failed and we need to fix master's mac */ |
0965a1f3 | 1685 | if (!bond_has_slaves(bond) && |
844223ab | 1686 | ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr)) |
b8fad459 | 1687 | eth_hw_addr_random(bond_dev); |
3d632c3f | 1688 | |
1da177e4 LT |
1689 | return res; |
1690 | } | |
1691 | ||
547942ca | 1692 | /* Try to release the slave device <slave> from the bond device <master> |
1da177e4 | 1693 | * It is legal to access curr_active_slave without a lock because all the function |
8c0bc550 | 1694 | * is RTNL-locked. If "all" is true it means that the function is being called |
0896341a | 1695 | * while destroying a bond interface and all slaves are being released. |
1da177e4 LT |
1696 | * |
1697 | * The rules for slave state should be: | |
1698 | * for Active/Backup: | |
1699 | * Active stays on all backups go down | |
1700 | * for Bonded connections: | |
1701 | * The first up interface should be left on and all others downed. | |
1702 | */ | |
0896341a | 1703 | static int __bond_release_one(struct net_device *bond_dev, |
1704 | struct net_device *slave_dev, | |
1705 | bool all) | |
1da177e4 | 1706 | { |
454d7c9b | 1707 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1708 | struct slave *slave, *oldcurrent; |
1709 | struct sockaddr addr; | |
5a0068de | 1710 | int old_flags = bond_dev->flags; |
c8f44aff | 1711 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1712 | |
1713 | /* slave is not a slave or master is not master of this slave */ | |
1714 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1715 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a22a9e41 | 1716 | netdev_dbg(bond_dev, "cannot release %s\n", |
76444f50 | 1717 | slave_dev->name); |
1da177e4 LT |
1718 | return -EINVAL; |
1719 | } | |
1720 | ||
e843fa50 | 1721 | block_netpoll_tx(); |
1da177e4 LT |
1722 | |
1723 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1724 | if (!slave) { | |
1725 | /* not a slave of this bond */ | |
76444f50 VF |
1726 | netdev_info(bond_dev, "%s not enslaved\n", |
1727 | slave_dev->name); | |
e843fa50 | 1728 | unblock_netpoll_tx(); |
1da177e4 LT |
1729 | return -EINVAL; |
1730 | } | |
1731 | ||
07699f9a | 1732 | bond_sysfs_slave_del(slave); |
1733 | ||
5f0c5f73 AG |
1734 | /* recompute stats just before removing the slave */ |
1735 | bond_get_stats(bond->dev, &bond->bond_stats); | |
1736 | ||
1f718f0f | 1737 | bond_upper_dev_unlink(bond_dev, slave_dev); |
35d48903 JP |
1738 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1739 | * for this slave anymore. | |
1740 | */ | |
1741 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 | 1742 | |
e470259f | 1743 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 1744 | bond_3ad_unbind_slave(slave); |
1da177e4 | 1745 | |
ee637714 MB |
1746 | if (bond_mode_uses_xmit_hash(bond)) |
1747 | bond_update_slave_arr(bond, slave); | |
1748 | ||
76444f50 VF |
1749 | netdev_info(bond_dev, "Releasing %s interface %s\n", |
1750 | bond_is_active_slave(slave) ? "active" : "backup", | |
1751 | slave_dev->name); | |
1da177e4 | 1752 | |
4740d638 | 1753 | oldcurrent = rcu_access_pointer(bond->curr_active_slave); |
1da177e4 | 1754 | |
85741718 | 1755 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
1da177e4 | 1756 | |
00503b6f | 1757 | if (!all && (!bond->params.fail_over_mac || |
01844098 | 1758 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) { |
844223ab | 1759 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1760 | bond_has_slaves(bond)) |
76444f50 VF |
1761 | netdev_warn(bond_dev, "the permanent HWaddr of %s - %pM - is still in use by %s - set the HWaddr of %s to a different address to avoid conflicts\n", |
1762 | slave_dev->name, slave->perm_hwaddr, | |
1763 | bond_dev->name, slave_dev->name); | |
dec1e90e | 1764 | } |
1765 | ||
059b47e8 NA |
1766 | if (rtnl_dereference(bond->primary_slave) == slave) |
1767 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
1da177e4 | 1768 | |
1c72cfdc | 1769 | if (oldcurrent == slave) |
1da177e4 | 1770 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1771 | |
58402054 | 1772 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1773 | /* Must be called only after the slave has been |
1774 | * detached from the list and the curr_active_slave | |
1775 | * has been cleared (if our_slave == old_current), | |
1776 | * but before a new active slave is selected. | |
1777 | */ | |
1778 | bond_alb_deinit_slave(bond, slave); | |
1779 | } | |
1780 | ||
0896341a | 1781 | if (all) { |
36708b89 | 1782 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 1783 | } else if (oldcurrent == slave) { |
547942ca | 1784 | /* Note that we hold RTNL over this sequence, so there |
059fe7a5 JV |
1785 | * is no concern that another slave add/remove event |
1786 | * will interfere. | |
1787 | */ | |
1da177e4 | 1788 | bond_select_active_slave(bond); |
059fe7a5 JV |
1789 | } |
1790 | ||
0965a1f3 | 1791 | if (!bond_has_slaves(bond)) { |
ff59c456 | 1792 | bond_set_carrier(bond); |
409cc1f8 | 1793 | eth_hw_addr_random(bond_dev); |
1da177e4 LT |
1794 | } |
1795 | ||
e843fa50 | 1796 | unblock_netpoll_tx(); |
278b2083 | 1797 | synchronize_rcu(); |
ee6154e1 | 1798 | bond->slave_cnt--; |
1da177e4 | 1799 | |
0965a1f3 | 1800 | if (!bond_has_slaves(bond)) { |
2af73d4b | 1801 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1802 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1803 | } | |
2af73d4b | 1804 | |
b2a103e6 MM |
1805 | bond_compute_features(bond); |
1806 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1807 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
76444f50 VF |
1808 | netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n", |
1809 | slave_dev->name, bond_dev->name); | |
b2a103e6 | 1810 | |
1ff412ad | 1811 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1812 | |
547942ca | 1813 | /* If the mode uses primary, then this case was handled above by |
303d1cbf | 1814 | * bond_change_active_slave(..., NULL) |
1da177e4 | 1815 | */ |
ec0865a9 | 1816 | if (!bond_uses_primary(bond)) { |
5a0068de NH |
1817 | /* unset promiscuity level from slave |
1818 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
1819 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
1820 | * value of that flag before that change, as that was the value | |
1821 | * when this slave was attached, so we cache at the start of the | |
1822 | * function and use it here. Same goes for ALLMULTI below | |
1823 | */ | |
1824 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 1825 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1826 | |
1827 | /* unset allmulti level from slave */ | |
5a0068de | 1828 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 1829 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1830 | |
303d1cbf | 1831 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1832 | } |
1833 | ||
8a8efa22 | 1834 | slave_disable_netpoll(slave); |
f6dc31a8 | 1835 | |
1da177e4 LT |
1836 | /* close slave before restoring its mac address */ |
1837 | dev_close(slave_dev); | |
1838 | ||
00503b6f | 1839 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
01844098 | 1840 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 | 1841 | /* restore original ("permanent") mac address */ |
ada0f863 | 1842 | ether_addr_copy(addr.sa_data, slave->perm_hwaddr); |
2ab82852 MS |
1843 | addr.sa_family = slave_dev->type; |
1844 | dev_set_mac_address(slave_dev, &addr); | |
1845 | } | |
1da177e4 | 1846 | |
b15ba0fb JP |
1847 | dev_set_mtu(slave_dev, slave->original_mtu); |
1848 | ||
2d7011ca | 1849 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 | 1850 | |
3fdddd85 | 1851 | bond_free_slave(slave); |
1da177e4 | 1852 | |
547942ca | 1853 | return 0; |
1da177e4 LT |
1854 | } |
1855 | ||
0896341a | 1856 | /* A wrapper used because of ndo_del_link */ |
1857 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1858 | { | |
1859 | return __bond_release_one(bond_dev, slave_dev, false); | |
1860 | } | |
1861 | ||
547942ca NA |
1862 | /* First release a slave and then destroy the bond if no more slaves are left. |
1863 | * Must be under rtnl_lock when this function is called. | |
1864 | */ | |
26d8ee75 | 1865 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1866 | struct net_device *slave_dev) | |
d90a162a | 1867 | { |
454d7c9b | 1868 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1869 | int ret; |
1870 | ||
1871 | ret = bond_release(bond_dev, slave_dev); | |
0965a1f3 | 1872 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 1873 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
76444f50 VF |
1874 | netdev_info(bond_dev, "Destroying bond %s\n", |
1875 | bond_dev->name); | |
9e71626c | 1876 | unregister_netdevice(bond_dev); |
d90a162a MS |
1877 | } |
1878 | return ret; | |
1879 | } | |
1880 | ||
1da177e4 LT |
1881 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1882 | { | |
454d7c9b | 1883 | struct bonding *bond = netdev_priv(bond_dev); |
69a2338e | 1884 | bond_fill_ifbond(bond, info); |
1da177e4 LT |
1885 | return 0; |
1886 | } | |
1887 | ||
1888 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1889 | { | |
454d7c9b | 1890 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1891 | struct list_head *iter; |
dec1e90e | 1892 | int i = 0, res = -ENODEV; |
1da177e4 | 1893 | struct slave *slave; |
1da177e4 | 1894 | |
9caff1e7 | 1895 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 1896 | if (i++ == (int)info->slave_id) { |
689c96cc | 1897 | res = 0; |
69a2338e | 1898 | bond_fill_ifslave(slave, info); |
1da177e4 LT |
1899 | break; |
1900 | } | |
1901 | } | |
1da177e4 | 1902 | |
689c96cc | 1903 | return res; |
1da177e4 LT |
1904 | } |
1905 | ||
1906 | /*-------------------------------- Monitoring -------------------------------*/ | |
1907 | ||
4740d638 | 1908 | /* called with rcu_read_lock() */ |
f0c76d61 JV |
1909 | static int bond_miimon_inspect(struct bonding *bond) |
1910 | { | |
dec1e90e | 1911 | int link_state, commit = 0; |
9caff1e7 | 1912 | struct list_head *iter; |
f0c76d61 | 1913 | struct slave *slave; |
41f89100 JP |
1914 | bool ignore_updelay; |
1915 | ||
4740d638 | 1916 | ignore_updelay = !rcu_dereference(bond->curr_active_slave); |
1da177e4 | 1917 | |
4cb4f97b | 1918 | bond_for_each_slave_rcu(bond, slave, iter) { |
f0c76d61 | 1919 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 1920 | |
f0c76d61 | 1921 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
1922 | |
1923 | switch (slave->link) { | |
f0c76d61 JV |
1924 | case BOND_LINK_UP: |
1925 | if (link_state) | |
1926 | continue; | |
1da177e4 | 1927 | |
69a2338e | 1928 | bond_set_slave_link_state(slave, BOND_LINK_FAIL); |
f0c76d61 JV |
1929 | slave->delay = bond->params.downdelay; |
1930 | if (slave->delay) { | |
76444f50 VF |
1931 | netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n", |
1932 | (BOND_MODE(bond) == | |
1933 | BOND_MODE_ACTIVEBACKUP) ? | |
1934 | (bond_is_active_slave(slave) ? | |
1935 | "active " : "backup ") : "", | |
1936 | slave->dev->name, | |
1937 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 1938 | } |
f0c76d61 JV |
1939 | /*FALLTHRU*/ |
1940 | case BOND_LINK_FAIL: | |
1941 | if (link_state) { | |
547942ca | 1942 | /* recovered before downdelay expired */ |
69a2338e | 1943 | bond_set_slave_link_state(slave, BOND_LINK_UP); |
8e603460 | 1944 | slave->last_link_up = jiffies; |
76444f50 VF |
1945 | netdev_info(bond->dev, "link status up again after %d ms for interface %s\n", |
1946 | (bond->params.downdelay - slave->delay) * | |
1947 | bond->params.miimon, | |
1948 | slave->dev->name); | |
f0c76d61 | 1949 | continue; |
1da177e4 | 1950 | } |
f0c76d61 JV |
1951 | |
1952 | if (slave->delay <= 0) { | |
1953 | slave->new_link = BOND_LINK_DOWN; | |
1954 | commit++; | |
1955 | continue; | |
1da177e4 | 1956 | } |
1da177e4 | 1957 | |
f0c76d61 JV |
1958 | slave->delay--; |
1959 | break; | |
1960 | ||
1961 | case BOND_LINK_DOWN: | |
1962 | if (!link_state) | |
1963 | continue; | |
1964 | ||
69a2338e | 1965 | bond_set_slave_link_state(slave, BOND_LINK_BACK); |
f0c76d61 JV |
1966 | slave->delay = bond->params.updelay; |
1967 | ||
1968 | if (slave->delay) { | |
76444f50 VF |
1969 | netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n", |
1970 | slave->dev->name, | |
1971 | ignore_updelay ? 0 : | |
1972 | bond->params.updelay * | |
1973 | bond->params.miimon); | |
f0c76d61 JV |
1974 | } |
1975 | /*FALLTHRU*/ | |
1976 | case BOND_LINK_BACK: | |
1977 | if (!link_state) { | |
69a2338e MS |
1978 | bond_set_slave_link_state(slave, |
1979 | BOND_LINK_DOWN); | |
76444f50 VF |
1980 | netdev_info(bond->dev, "link status down again after %d ms for interface %s\n", |
1981 | (bond->params.updelay - slave->delay) * | |
1982 | bond->params.miimon, | |
1983 | slave->dev->name); | |
f0c76d61 JV |
1984 | |
1985 | continue; | |
1986 | } | |
1987 | ||
41f89100 JP |
1988 | if (ignore_updelay) |
1989 | slave->delay = 0; | |
1990 | ||
f0c76d61 JV |
1991 | if (slave->delay <= 0) { |
1992 | slave->new_link = BOND_LINK_UP; | |
1993 | commit++; | |
41f89100 | 1994 | ignore_updelay = false; |
f0c76d61 | 1995 | continue; |
1da177e4 | 1996 | } |
f0c76d61 JV |
1997 | |
1998 | slave->delay--; | |
1da177e4 | 1999 | break; |
f0c76d61 JV |
2000 | } |
2001 | } | |
1da177e4 | 2002 | |
f0c76d61 JV |
2003 | return commit; |
2004 | } | |
1da177e4 | 2005 | |
f0c76d61 JV |
2006 | static void bond_miimon_commit(struct bonding *bond) |
2007 | { | |
9caff1e7 | 2008 | struct list_head *iter; |
059b47e8 | 2009 | struct slave *slave, *primary; |
f0c76d61 | 2010 | |
9caff1e7 | 2011 | bond_for_each_slave(bond, slave, iter) { |
f0c76d61 JV |
2012 | switch (slave->new_link) { |
2013 | case BOND_LINK_NOCHANGE: | |
2014 | continue; | |
1da177e4 | 2015 | |
f0c76d61 | 2016 | case BOND_LINK_UP: |
69a2338e | 2017 | bond_set_slave_link_state(slave, BOND_LINK_UP); |
8e603460 | 2018 | slave->last_link_up = jiffies; |
f0c76d61 | 2019 | |
059b47e8 | 2020 | primary = rtnl_dereference(bond->primary_slave); |
01844098 | 2021 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
f0c76d61 | 2022 | /* prevent it from being the active one */ |
e30bc066 | 2023 | bond_set_backup_slave(slave); |
01844098 | 2024 | } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
f0c76d61 | 2025 | /* make it immediately active */ |
e30bc066 | 2026 | bond_set_active_slave(slave); |
059b47e8 | 2027 | } else if (slave != primary) { |
f0c76d61 | 2028 | /* prevent it from being the active one */ |
e30bc066 | 2029 | bond_set_backup_slave(slave); |
1da177e4 | 2030 | } |
1da177e4 | 2031 | |
76444f50 VF |
2032 | netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n", |
2033 | slave->dev->name, | |
2034 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, | |
2035 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2036 | |
f0c76d61 | 2037 | /* notify ad that the link status has changed */ |
01844098 | 2038 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 | 2039 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); |
059fe7a5 | 2040 | |
58402054 | 2041 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2042 | bond_alb_handle_link_change(bond, slave, |
2043 | BOND_LINK_UP); | |
1da177e4 | 2044 | |
ee637714 MB |
2045 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2046 | bond_update_slave_arr(bond, NULL); | |
2047 | ||
059b47e8 | 2048 | if (!bond->curr_active_slave || slave == primary) |
f0c76d61 | 2049 | goto do_failover; |
1da177e4 | 2050 | |
f0c76d61 | 2051 | continue; |
059fe7a5 | 2052 | |
f0c76d61 | 2053 | case BOND_LINK_DOWN: |
fba4acda JV |
2054 | if (slave->link_failure_count < UINT_MAX) |
2055 | slave->link_failure_count++; | |
2056 | ||
69a2338e | 2057 | bond_set_slave_link_state(slave, BOND_LINK_DOWN); |
1da177e4 | 2058 | |
01844098 VF |
2059 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || |
2060 | BOND_MODE(bond) == BOND_MODE_8023AD) | |
5e5b0665 | 2061 | bond_set_slave_inactive_flags(slave, |
2062 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2063 | |
76444f50 VF |
2064 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2065 | slave->dev->name); | |
f0c76d61 | 2066 | |
01844098 | 2067 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 JV |
2068 | bond_3ad_handle_link_change(slave, |
2069 | BOND_LINK_DOWN); | |
2070 | ||
ae63e808 | 2071 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2072 | bond_alb_handle_link_change(bond, slave, |
2073 | BOND_LINK_DOWN); | |
2074 | ||
ee637714 MB |
2075 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2076 | bond_update_slave_arr(bond, NULL); | |
2077 | ||
4740d638 | 2078 | if (slave == rcu_access_pointer(bond->curr_active_slave)) |
f0c76d61 JV |
2079 | goto do_failover; |
2080 | ||
2081 | continue; | |
2082 | ||
2083 | default: | |
76444f50 VF |
2084 | netdev_err(bond->dev, "invalid new link %d on slave %s\n", |
2085 | slave->new_link, slave->dev->name); | |
f0c76d61 JV |
2086 | slave->new_link = BOND_LINK_NOCHANGE; |
2087 | ||
2088 | continue; | |
2089 | } | |
2090 | ||
2091 | do_failover: | |
e843fa50 | 2092 | block_netpoll_tx(); |
f0c76d61 | 2093 | bond_select_active_slave(bond); |
e843fa50 | 2094 | unblock_netpoll_tx(); |
f0c76d61 JV |
2095 | } |
2096 | ||
2097 | bond_set_carrier(bond); | |
1da177e4 LT |
2098 | } |
2099 | ||
547942ca | 2100 | /* bond_mii_monitor |
0b0eef66 JV |
2101 | * |
2102 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2103 | * inspection, then (if inspection indicates something needs to be done) |
2104 | * an acquisition of appropriate locks followed by a commit phase to | |
2105 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2106 | */ |
6da67d26 | 2107 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2108 | { |
2109 | struct bonding *bond = container_of(work, struct bonding, | |
2110 | mii_work.work); | |
ad246c99 | 2111 | bool should_notify_peers = false; |
1f2cd845 | 2112 | unsigned long delay; |
0b0eef66 | 2113 | |
1f2cd845 DM |
2114 | delay = msecs_to_jiffies(bond->params.miimon); |
2115 | ||
2116 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2117 | goto re_arm; |
b59f9f74 | 2118 | |
4cb4f97b | 2119 | rcu_read_lock(); |
2120 | ||
ad246c99 BH |
2121 | should_notify_peers = bond_should_notify_peers(bond); |
2122 | ||
1f2cd845 | 2123 | if (bond_miimon_inspect(bond)) { |
4cb4f97b | 2124 | rcu_read_unlock(); |
f0c76d61 | 2125 | |
1f2cd845 DM |
2126 | /* Race avoidance with bond_close cancel of workqueue */ |
2127 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2128 | delay = 1; |
2129 | should_notify_peers = false; | |
2130 | goto re_arm; | |
2131 | } | |
6b6c5261 | 2132 | |
1f2cd845 DM |
2133 | bond_miimon_commit(bond); |
2134 | ||
1f2cd845 | 2135 | rtnl_unlock(); /* might sleep, hold no other locks */ |
4cb4f97b | 2136 | } else |
2137 | rcu_read_unlock(); | |
0b0eef66 | 2138 | |
f0c76d61 | 2139 | re_arm: |
e6d265e8 | 2140 | if (bond->params.miimon) |
1f2cd845 DM |
2141 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2142 | ||
1f2cd845 DM |
2143 | if (should_notify_peers) { |
2144 | if (!rtnl_trylock()) | |
2145 | return; | |
2146 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2147 | rtnl_unlock(); | |
2148 | } | |
0b0eef66 | 2149 | } |
c3ade5ca | 2150 | |
50223ce4 | 2151 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2152 | { |
50223ce4 VF |
2153 | struct net_device *upper; |
2154 | struct list_head *iter; | |
2155 | bool ret = false; | |
f5b2b966 | 2156 | |
eaddcd76 | 2157 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2158 | return true; |
f5b2b966 | 2159 | |
50223ce4 | 2160 | rcu_read_lock(); |
2f268f12 | 2161 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2162 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2163 | ret = true; | |
2164 | break; | |
2165 | } | |
f5b2b966 | 2166 | } |
50223ce4 | 2167 | rcu_read_unlock(); |
f5b2b966 | 2168 | |
50223ce4 | 2169 | return ret; |
f5b2b966 JV |
2170 | } |
2171 | ||
547942ca | 2172 | /* We go to the (large) trouble of VLAN tagging ARP frames because |
c3ade5ca JV |
2173 | * switches in VLAN mode (especially if ports are configured as |
2174 | * "native" to a VLAN) might not pass non-tagged frames. | |
2175 | */ | |
fbd929f2 | 2176 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, |
2177 | __be32 dest_ip, __be32 src_ip, | |
44a40855 | 2178 | struct bond_vlan_tag *tags) |
c3ade5ca JV |
2179 | { |
2180 | struct sk_buff *skb; | |
3e403a77 | 2181 | struct bond_vlan_tag *outer_tag = tags; |
c3ade5ca | 2182 | |
76444f50 VF |
2183 | netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n", |
2184 | arp_op, slave_dev->name, &dest_ip, &src_ip); | |
3d632c3f | 2185 | |
c3ade5ca JV |
2186 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2187 | NULL, slave_dev->dev_addr, NULL); | |
2188 | ||
2189 | if (!skb) { | |
4873ac3c | 2190 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2191 | return; |
2192 | } | |
fbd929f2 | 2193 | |
3e403a77 VF |
2194 | if (!tags || tags->vlan_proto == VLAN_N_VID) |
2195 | goto xmit; | |
2196 | ||
2197 | tags++; | |
2198 | ||
44a40855 | 2199 | /* Go through all the tags backwards and add them to the packet */ |
3e403a77 VF |
2200 | while (tags->vlan_proto != VLAN_N_VID) { |
2201 | if (!tags->vlan_id) { | |
2202 | tags++; | |
44a40855 | 2203 | continue; |
3e403a77 | 2204 | } |
44a40855 | 2205 | |
76444f50 | 2206 | netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n", |
3e403a77 | 2207 | ntohs(outer_tag->vlan_proto), tags->vlan_id); |
62749e2c JP |
2208 | skb = vlan_insert_tag_set_proto(skb, tags->vlan_proto, |
2209 | tags->vlan_id); | |
44a40855 VY |
2210 | if (!skb) { |
2211 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2212 | return; | |
2213 | } | |
3e403a77 VF |
2214 | |
2215 | tags++; | |
44a40855 VY |
2216 | } |
2217 | /* Set the outer tag */ | |
3e403a77 | 2218 | if (outer_tag->vlan_id) { |
76444f50 | 2219 | netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n", |
3e403a77 | 2220 | ntohs(outer_tag->vlan_proto), outer_tag->vlan_id); |
b4bef1b5 JP |
2221 | __vlan_hwaccel_put_tag(skb, outer_tag->vlan_proto, |
2222 | outer_tag->vlan_id); | |
c3ade5ca | 2223 | } |
3e403a77 VF |
2224 | |
2225 | xmit: | |
c3ade5ca JV |
2226 | arp_xmit(skb); |
2227 | } | |
2228 | ||
44a40855 VY |
2229 | /* Validate the device path between the @start_dev and the @end_dev. |
2230 | * The path is valid if the @end_dev is reachable through device | |
2231 | * stacking. | |
2232 | * When the path is validated, collect any vlan information in the | |
2233 | * path. | |
2234 | */ | |
3e403a77 VF |
2235 | struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, |
2236 | struct net_device *end_dev, | |
2237 | int level) | |
44a40855 | 2238 | { |
3e403a77 | 2239 | struct bond_vlan_tag *tags; |
44a40855 VY |
2240 | struct net_device *upper; |
2241 | struct list_head *iter; | |
44a40855 | 2242 | |
3e403a77 VF |
2243 | if (start_dev == end_dev) { |
2244 | tags = kzalloc(sizeof(*tags) * (level + 1), GFP_ATOMIC); | |
2245 | if (!tags) | |
2246 | return ERR_PTR(-ENOMEM); | |
2247 | tags[level].vlan_proto = VLAN_N_VID; | |
2248 | return tags; | |
2249 | } | |
44a40855 VY |
2250 | |
2251 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
3e403a77 VF |
2252 | tags = bond_verify_device_path(upper, end_dev, level + 1); |
2253 | if (IS_ERR_OR_NULL(tags)) { | |
2254 | if (IS_ERR(tags)) | |
2255 | return tags; | |
2256 | continue; | |
44a40855 | 2257 | } |
3e403a77 VF |
2258 | if (is_vlan_dev(upper)) { |
2259 | tags[level].vlan_proto = vlan_dev_vlan_proto(upper); | |
2260 | tags[level].vlan_id = vlan_dev_vlan_id(upper); | |
2261 | } | |
2262 | ||
2263 | return tags; | |
44a40855 VY |
2264 | } |
2265 | ||
3e403a77 | 2266 | return NULL; |
44a40855 | 2267 | } |
c3ade5ca | 2268 | |
1da177e4 LT |
2269 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2270 | { | |
c3ade5ca | 2271 | struct rtable *rt; |
3e403a77 | 2272 | struct bond_vlan_tag *tags; |
27bc11e6 | 2273 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2274 | int i; |
1da177e4 | 2275 | |
27bc11e6 | 2276 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
76444f50 | 2277 | netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]); |
3e403a77 | 2278 | tags = NULL; |
c3ade5ca | 2279 | |
27bc11e6 | 2280 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2281 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2282 | RTO_ONLINK, 0); | |
b23dd4fe | 2283 | if (IS_ERR(rt)) { |
28572760 VF |
2284 | /* there's no route to target - try to send arp |
2285 | * probe to generate any traffic (arp_validate=0) | |
2286 | */ | |
4873ac3c | 2287 | if (bond->params.arp_validate) |
2288 | net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", | |
2289 | bond->dev->name, | |
2290 | &targets[i]); | |
44a40855 VY |
2291 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2292 | 0, tags); | |
c3ade5ca JV |
2293 | continue; |
2294 | } | |
2295 | ||
27bc11e6 VF |
2296 | /* bond device itself */ |
2297 | if (rt->dst.dev == bond->dev) | |
2298 | goto found; | |
2299 | ||
2300 | rcu_read_lock(); | |
3e403a77 | 2301 | tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0); |
27bc11e6 | 2302 | rcu_read_unlock(); |
c3ade5ca | 2303 | |
3e403a77 | 2304 | if (!IS_ERR_OR_NULL(tags)) |
44a40855 VY |
2305 | goto found; |
2306 | ||
27bc11e6 | 2307 | /* Not our device - skip */ |
76444f50 VF |
2308 | netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n", |
2309 | &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
3e32582f | 2310 | |
ed4b9f80 | 2311 | ip_rt_put(rt); |
27bc11e6 VF |
2312 | continue; |
2313 | ||
2314 | found: | |
2315 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2316 | ip_rt_put(rt); | |
2317 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
44a40855 | 2318 | addr, tags); |
a67eed57 | 2319 | kfree(tags); |
c3ade5ca JV |
2320 | } |
2321 | } | |
2322 | ||
d3bb52b0 | 2323 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2324 | { |
8599b52e VF |
2325 | int i; |
2326 | ||
87a7b84b | 2327 | if (!sip || !bond_has_this_ip(bond, tip)) { |
76444f50 VF |
2328 | netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n", |
2329 | &sip, &tip); | |
87a7b84b VF |
2330 | return; |
2331 | } | |
f5b2b966 | 2332 | |
8599b52e VF |
2333 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2334 | if (i == -1) { | |
76444f50 VF |
2335 | netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n", |
2336 | &sip); | |
87a7b84b | 2337 | return; |
f5b2b966 | 2338 | } |
49f17de7 | 2339 | slave->last_rx = jiffies; |
8599b52e | 2340 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2341 | } |
2342 | ||
5bb9e0b5 | 2343 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2344 | struct slave *slave) | |
f5b2b966 | 2345 | { |
de063b70 | 2346 | struct arphdr *arp = (struct arphdr *)skb->data; |
010d3c39 | 2347 | struct slave *curr_active_slave; |
f5b2b966 | 2348 | unsigned char *arp_ptr; |
d3bb52b0 | 2349 | __be32 sip, tip; |
f2cb691a | 2350 | int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
f5b2b966 | 2351 | |
f2cb691a | 2352 | if (!slave_do_arp_validate(bond, slave)) { |
bedabf90 | 2353 | if ((slave_do_arp_validate_only(bond) && is_arp) || |
2354 | !slave_do_arp_validate_only(bond)) | |
49f17de7 | 2355 | slave->last_rx = jiffies; |
13a8e0c8 | 2356 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2357 | } else if (!is_arp) { |
2358 | return RX_HANDLER_ANOTHER; | |
2359 | } | |
2c146102 | 2360 | |
de063b70 | 2361 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2362 | |
76444f50 VF |
2363 | netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n", |
2364 | skb->dev->name); | |
f5b2b966 | 2365 | |
de063b70 ED |
2366 | if (alen > skb_headlen(skb)) { |
2367 | arp = kmalloc(alen, GFP_ATOMIC); | |
2368 | if (!arp) | |
2369 | goto out_unlock; | |
2370 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2371 | goto out_unlock; | |
2372 | } | |
f5b2b966 | 2373 | |
3aba891d | 2374 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2375 | skb->pkt_type == PACKET_OTHERHOST || |
2376 | skb->pkt_type == PACKET_LOOPBACK || | |
2377 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2378 | arp->ar_pro != htons(ETH_P_IP) || | |
2379 | arp->ar_pln != 4) | |
2380 | goto out_unlock; | |
2381 | ||
2382 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2383 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2384 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2385 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2386 | memcpy(&tip, arp_ptr, 4); |
2387 | ||
76444f50 VF |
2388 | netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
2389 | slave->dev->name, bond_slave_state(slave), | |
2390 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2391 | &sip, &tip); | |
f5b2b966 | 2392 | |
010d3c39 VF |
2393 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
2394 | ||
547942ca | 2395 | /* Backup slaves won't see the ARP reply, but do come through |
f5b2b966 JV |
2396 | * here for each ARP probe (so we swap the sip/tip to validate |
2397 | * the probe). In a "redundant switch, common router" type of | |
2398 | * configuration, the ARP probe will (hopefully) travel from | |
2399 | * the active, through one switch, the router, then the other | |
2400 | * switch before reaching the backup. | |
aeea64ac VF |
2401 | * |
2402 | * We 'trust' the arp requests if there is an active slave and | |
2403 | * it received valid arp reply(s) after it became active. This | |
2404 | * is done to avoid endless looping when we can't reach the | |
2405 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2406 | */ |
010d3c39 | 2407 | |
e30bc066 | 2408 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2409 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
2410 | else if (curr_active_slave && |
2411 | time_after(slave_last_rx(bond, curr_active_slave), | |
2412 | curr_active_slave->last_link_up)) | |
f5b2b966 JV |
2413 | bond_validate_arp(bond, slave, tip, sip); |
2414 | ||
2415 | out_unlock: | |
de063b70 ED |
2416 | if (arp != (struct arphdr *)skb->data) |
2417 | kfree(arp); | |
13a8e0c8 | 2418 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2419 | } |
2420 | ||
e7f63f1d VF |
2421 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2422 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2423 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2424 | */ | |
2425 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2426 | int mod) | |
2427 | { | |
2428 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2429 | ||
2430 | return time_in_range(jiffies, | |
2431 | last_act - delta_in_ticks, | |
2432 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2433 | } | |
2434 | ||
547942ca | 2435 | /* This function is called regularly to monitor each slave's link |
1da177e4 LT |
2436 | * ensuring that traffic is being sent and received when arp monitoring |
2437 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2438 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2439 | * arp monitoring in active backup mode. | |
2440 | */ | |
6da67d26 | 2441 | static void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2442 | { |
1b76b316 JV |
2443 | struct bonding *bond = container_of(work, struct bonding, |
2444 | arp_work.work); | |
1da177e4 | 2445 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2446 | struct list_head *iter; |
6fde8f03 | 2447 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 2448 | |
1f2cd845 | 2449 | if (!bond_has_slaves(bond)) |
1da177e4 | 2450 | goto re_arm; |
1da177e4 | 2451 | |
2e52f4fe | 2452 | rcu_read_lock(); |
2453 | ||
4740d638 | 2454 | oldcurrent = rcu_dereference(bond->curr_active_slave); |
1da177e4 LT |
2455 | /* see if any of the previous devices are up now (i.e. they have |
2456 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
2457 | * the picture unless it is null. also, slave->last_link_up is not |
2458 | * needed here because we send an arp on each slave and give a slave | |
2459 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
2460 | * TODO: what about up/down delay in arp mode? it wasn't here before |
2461 | * so it can wait | |
2462 | */ | |
2e52f4fe | 2463 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2464 | unsigned long trans_start = dev_trans_start(slave->dev); |
2465 | ||
1da177e4 | 2466 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2467 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 2468 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 LT |
2469 | |
2470 | slave->link = BOND_LINK_UP; | |
6fde8f03 | 2471 | slave_state_changed = 1; |
1da177e4 LT |
2472 | |
2473 | /* primary_slave has no meaning in round-robin | |
2474 | * mode. the window of a slave being up and | |
2475 | * curr_active_slave being null after enslaving | |
2476 | * is closed. | |
2477 | */ | |
2478 | if (!oldcurrent) { | |
76444f50 VF |
2479 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2480 | slave->dev->name); | |
1da177e4 LT |
2481 | do_failover = 1; |
2482 | } else { | |
76444f50 VF |
2483 | netdev_info(bond->dev, "interface %s is now up\n", |
2484 | slave->dev->name); | |
1da177e4 LT |
2485 | } |
2486 | } | |
2487 | } else { | |
2488 | /* slave->link == BOND_LINK_UP */ | |
2489 | ||
2490 | /* not all switches will respond to an arp request | |
2491 | * when the source ip is 0, so don't take the link down | |
2492 | * if we don't know our ip yet | |
2493 | */ | |
e7f63f1d | 2494 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 2495 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 LT |
2496 | |
2497 | slave->link = BOND_LINK_DOWN; | |
6fde8f03 | 2498 | slave_state_changed = 1; |
1da177e4 | 2499 | |
3d632c3f | 2500 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2501 | slave->link_failure_count++; |
1da177e4 | 2502 | |
76444f50 VF |
2503 | netdev_info(bond->dev, "interface %s is now down\n", |
2504 | slave->dev->name); | |
1da177e4 | 2505 | |
3d632c3f | 2506 | if (slave == oldcurrent) |
1da177e4 | 2507 | do_failover = 1; |
1da177e4 LT |
2508 | } |
2509 | } | |
2510 | ||
2511 | /* note: if switch is in round-robin mode, all links | |
2512 | * must tx arp to ensure all links rx an arp - otherwise | |
2513 | * links may oscillate or not come up at all; if switch is | |
2514 | * in something like xor mode, there is nothing we can | |
2515 | * do - all replies will be rx'ed on same link causing slaves | |
2516 | * to be unstable during low/no traffic periods | |
2517 | */ | |
b6adc610 | 2518 | if (bond_slave_is_up(slave)) |
1da177e4 | 2519 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2520 | } |
2521 | ||
2e52f4fe | 2522 | rcu_read_unlock(); |
2523 | ||
6fde8f03 | 2524 | if (do_failover || slave_state_changed) { |
2e52f4fe | 2525 | if (!rtnl_trylock()) |
2526 | goto re_arm; | |
1da177e4 | 2527 | |
6fde8f03 DT |
2528 | if (slave_state_changed) { |
2529 | bond_slave_state_change(bond); | |
ee637714 MB |
2530 | if (BOND_MODE(bond) == BOND_MODE_XOR) |
2531 | bond_update_slave_arr(bond, NULL); | |
b8e4500f NA |
2532 | } |
2533 | if (do_failover) { | |
6fde8f03 | 2534 | block_netpoll_tx(); |
6fde8f03 | 2535 | bond_select_active_slave(bond); |
6fde8f03 DT |
2536 | unblock_netpoll_tx(); |
2537 | } | |
2e52f4fe | 2538 | rtnl_unlock(); |
1da177e4 LT |
2539 | } |
2540 | ||
2541 | re_arm: | |
e6d265e8 | 2542 | if (bond->params.arp_interval) |
e7f63f1d VF |
2543 | queue_delayed_work(bond->wq, &bond->arp_work, |
2544 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2545 | } |
2546 | ||
547942ca | 2547 | /* Called to inspect slaves for active-backup mode ARP monitor link state |
b2220cad JV |
2548 | * changes. Sets new_link in slaves to specify what action should take |
2549 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2550 | * to link states must be committed. | |
2551 | * | |
8c0bc550 | 2552 | * Called with rcu_read_lock held. |
1da177e4 | 2553 | */ |
e7f63f1d | 2554 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2555 | { |
def4460c | 2556 | unsigned long trans_start, last_rx; |
9caff1e7 | 2557 | struct list_head *iter; |
dec1e90e | 2558 | struct slave *slave; |
dec1e90e | 2559 | int commit = 0; |
da210f55 | 2560 | |
eb9fa4b0 | 2561 | bond_for_each_slave_rcu(bond, slave, iter) { |
b2220cad | 2562 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2563 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2564 | |
b2220cad | 2565 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2566 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2567 | slave->new_link = BOND_LINK_UP; |
2568 | commit++; | |
2569 | } | |
b2220cad JV |
2570 | continue; |
2571 | } | |
1da177e4 | 2572 | |
547942ca | 2573 | /* Give slaves 2*delta after being enslaved or made |
b2220cad JV |
2574 | * active. This avoids bouncing, as the last receive |
2575 | * times need a full ARP monitor cycle to be updated. | |
2576 | */ | |
8e603460 | 2577 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
2578 | continue; |
2579 | ||
547942ca | 2580 | /* Backup slave is down if: |
b2220cad JV |
2581 | * - No current_arp_slave AND |
2582 | * - more than 3*delta since last receive AND | |
2583 | * - the bond has an IP address | |
2584 | * | |
2585 | * Note: a non-null current_arp_slave indicates | |
2586 | * the curr_active_slave went down and we are | |
2587 | * searching for a new one; under this condition | |
2588 | * we only take the curr_active_slave down - this | |
2589 | * gives each slave a chance to tx/rx traffic | |
2590 | * before being taken out | |
2591 | */ | |
e30bc066 | 2592 | if (!bond_is_active_slave(slave) && |
85741718 | 2593 | !rcu_access_pointer(bond->current_arp_slave) && |
e7f63f1d | 2594 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2595 | slave->new_link = BOND_LINK_DOWN; |
2596 | commit++; | |
2597 | } | |
2598 | ||
547942ca | 2599 | /* Active slave is down if: |
b2220cad JV |
2600 | * - more than 2*delta since transmitting OR |
2601 | * - (more than 2*delta since receive AND | |
2602 | * the bond has an IP address) | |
2603 | */ | |
cb32f2a0 | 2604 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2605 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2606 | (!bond_time_in_interval(bond, trans_start, 2) || |
2607 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2608 | slave->new_link = BOND_LINK_DOWN; |
2609 | commit++; | |
2610 | } | |
1da177e4 LT |
2611 | } |
2612 | ||
b2220cad JV |
2613 | return commit; |
2614 | } | |
1da177e4 | 2615 | |
547942ca | 2616 | /* Called to commit link state changes noted by inspection step of |
b2220cad JV |
2617 | * active-backup mode ARP monitor. |
2618 | * | |
eb9fa4b0 | 2619 | * Called with RTNL hold. |
b2220cad | 2620 | */ |
e7f63f1d | 2621 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2622 | { |
cb32f2a0 | 2623 | unsigned long trans_start; |
9caff1e7 | 2624 | struct list_head *iter; |
dec1e90e | 2625 | struct slave *slave; |
1da177e4 | 2626 | |
9caff1e7 | 2627 | bond_for_each_slave(bond, slave, iter) { |
b2220cad JV |
2628 | switch (slave->new_link) { |
2629 | case BOND_LINK_NOCHANGE: | |
2630 | continue; | |
ff59c456 | 2631 | |
b2220cad | 2632 | case BOND_LINK_UP: |
cb32f2a0 | 2633 | trans_start = dev_trans_start(slave->dev); |
4740d638 ED |
2634 | if (rtnl_dereference(bond->curr_active_slave) != slave || |
2635 | (!rtnl_dereference(bond->curr_active_slave) && | |
e7f63f1d | 2636 | bond_time_in_interval(bond, trans_start, 1))) { |
85741718 ED |
2637 | struct slave *current_arp_slave; |
2638 | ||
2639 | current_arp_slave = rtnl_dereference(bond->current_arp_slave); | |
69a2338e | 2640 | bond_set_slave_link_state(slave, BOND_LINK_UP); |
85741718 | 2641 | if (current_arp_slave) { |
5a430974 | 2642 | bond_set_slave_inactive_flags( |
85741718 | 2643 | current_arp_slave, |
5e5b0665 | 2644 | BOND_SLAVE_NOTIFY_NOW); |
85741718 | 2645 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
5a430974 | 2646 | } |
b2220cad | 2647 | |
76444f50 VF |
2648 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2649 | slave->dev->name); | |
b2220cad | 2650 | |
4740d638 | 2651 | if (!rtnl_dereference(bond->curr_active_slave) || |
059b47e8 | 2652 | slave == rtnl_dereference(bond->primary_slave)) |
b9f60253 | 2653 | goto do_failover; |
1da177e4 | 2654 | |
b9f60253 | 2655 | } |
1da177e4 | 2656 | |
b9f60253 | 2657 | continue; |
1da177e4 | 2658 | |
b2220cad JV |
2659 | case BOND_LINK_DOWN: |
2660 | if (slave->link_failure_count < UINT_MAX) | |
2661 | slave->link_failure_count++; | |
2662 | ||
69a2338e | 2663 | bond_set_slave_link_state(slave, BOND_LINK_DOWN); |
5e5b0665 | 2664 | bond_set_slave_inactive_flags(slave, |
2665 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 2666 | |
76444f50 VF |
2667 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2668 | slave->dev->name); | |
b2220cad | 2669 | |
4740d638 | 2670 | if (slave == rtnl_dereference(bond->curr_active_slave)) { |
85741718 | 2671 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
b9f60253 | 2672 | goto do_failover; |
1da177e4 | 2673 | } |
b9f60253 JP |
2674 | |
2675 | continue; | |
b2220cad JV |
2676 | |
2677 | default: | |
76444f50 VF |
2678 | netdev_err(bond->dev, "impossible: new_link %d on slave %s\n", |
2679 | slave->new_link, slave->dev->name); | |
b9f60253 | 2680 | continue; |
1da177e4 | 2681 | } |
1da177e4 | 2682 | |
b9f60253 | 2683 | do_failover: |
e843fa50 | 2684 | block_netpoll_tx(); |
b9f60253 | 2685 | bond_select_active_slave(bond); |
e843fa50 | 2686 | unblock_netpoll_tx(); |
b2220cad | 2687 | } |
1da177e4 | 2688 | |
b2220cad JV |
2689 | bond_set_carrier(bond); |
2690 | } | |
1da177e4 | 2691 | |
547942ca | 2692 | /* Send ARP probes for active-backup mode ARP monitor. |
b0929915 | 2693 | * |
8c0bc550 | 2694 | * Called with rcu_read_lock held. |
b2220cad | 2695 | */ |
f2ebd477 | 2696 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 2697 | { |
eb9fa4b0 | 2698 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 2699 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
2700 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
2701 | struct list_head *iter; |
2702 | bool found = false; | |
b0929915 | 2703 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 2704 | |
98b90f26 | 2705 | if (curr_arp_slave && curr_active_slave) |
76444f50 VF |
2706 | netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n", |
2707 | curr_arp_slave->dev->name, | |
2708 | curr_active_slave->dev->name); | |
1da177e4 | 2709 | |
98b90f26 VF |
2710 | if (curr_active_slave) { |
2711 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 2712 | return should_notify_rtnl; |
b2220cad | 2713 | } |
1da177e4 | 2714 | |
b2220cad JV |
2715 | /* if we don't have a curr_active_slave, search for the next available |
2716 | * backup slave from the current_arp_slave and make it the candidate | |
2717 | * for becoming the curr_active_slave | |
2718 | */ | |
1da177e4 | 2719 | |
eb9fa4b0 | 2720 | if (!curr_arp_slave) { |
b0929915 | 2721 | curr_arp_slave = bond_first_slave_rcu(bond); |
2722 | if (!curr_arp_slave) | |
2723 | return should_notify_rtnl; | |
b2220cad | 2724 | } |
1da177e4 | 2725 | |
b0929915 | 2726 | bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER); |
059fe7a5 | 2727 | |
b0929915 | 2728 | bond_for_each_slave_rcu(bond, slave, iter) { |
b6adc610 | 2729 | if (!found && !before && bond_slave_is_up(slave)) |
4087df87 | 2730 | before = slave; |
1da177e4 | 2731 | |
b6adc610 | 2732 | if (found && !new_slave && bond_slave_is_up(slave)) |
4087df87 | 2733 | new_slave = slave; |
b2220cad JV |
2734 | /* if the link state is up at this point, we |
2735 | * mark it down - this can happen if we have | |
2736 | * simultaneous link failures and | |
2737 | * reselect_active_interface doesn't make this | |
2738 | * one the current slave so it is still marked | |
2739 | * up when it is actually down | |
1da177e4 | 2740 | */ |
b6adc610 | 2741 | if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
69a2338e | 2742 | bond_set_slave_link_state(slave, BOND_LINK_DOWN); |
b2220cad JV |
2743 | if (slave->link_failure_count < UINT_MAX) |
2744 | slave->link_failure_count++; | |
1da177e4 | 2745 | |
5e5b0665 | 2746 | bond_set_slave_inactive_flags(slave, |
b0929915 | 2747 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 2748 | |
76444f50 VF |
2749 | netdev_info(bond->dev, "backup interface %s is now down\n", |
2750 | slave->dev->name); | |
1da177e4 | 2751 | } |
eb9fa4b0 | 2752 | if (slave == curr_arp_slave) |
4087df87 | 2753 | found = true; |
b2220cad | 2754 | } |
4087df87 VF |
2755 | |
2756 | if (!new_slave && before) | |
2757 | new_slave = before; | |
2758 | ||
b0929915 | 2759 | if (!new_slave) |
2760 | goto check_state; | |
4087df87 | 2761 | |
69a2338e | 2762 | bond_set_slave_link_state(new_slave, BOND_LINK_BACK); |
b0929915 | 2763 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 2764 | bond_arp_send_all(bond, new_slave); |
8e603460 | 2765 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 2766 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 2767 | |
b0929915 | 2768 | check_state: |
2769 | bond_for_each_slave_rcu(bond, slave, iter) { | |
2770 | if (slave->should_notify) { | |
2771 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; | |
2772 | break; | |
2773 | } | |
2774 | } | |
2775 | return should_notify_rtnl; | |
b2220cad | 2776 | } |
1da177e4 | 2777 | |
6da67d26 | 2778 | static void bond_activebackup_arp_mon(struct work_struct *work) |
b2220cad JV |
2779 | { |
2780 | struct bonding *bond = container_of(work, struct bonding, | |
2781 | arp_work.work); | |
b0929915 | 2782 | bool should_notify_peers = false; |
2783 | bool should_notify_rtnl = false; | |
1f2cd845 | 2784 | int delta_in_ticks; |
1da177e4 | 2785 | |
1f2cd845 DM |
2786 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
2787 | ||
2788 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
2789 | goto re_arm; |
2790 | ||
eb9fa4b0 | 2791 | rcu_read_lock(); |
b0929915 | 2792 | |
ad246c99 BH |
2793 | should_notify_peers = bond_should_notify_peers(bond); |
2794 | ||
b0929915 | 2795 | if (bond_ab_arp_inspect(bond)) { |
2796 | rcu_read_unlock(); | |
2797 | ||
1f2cd845 DM |
2798 | /* Race avoidance with bond_close flush of workqueue */ |
2799 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2800 | delta_in_ticks = 1; |
2801 | should_notify_peers = false; | |
2802 | goto re_arm; | |
2803 | } | |
b2220cad | 2804 | |
1f2cd845 | 2805 | bond_ab_arp_commit(bond); |
b0929915 | 2806 | |
1f2cd845 | 2807 | rtnl_unlock(); |
b0929915 | 2808 | rcu_read_lock(); |
1f2cd845 DM |
2809 | } |
2810 | ||
b0929915 | 2811 | should_notify_rtnl = bond_ab_arp_probe(bond); |
2812 | rcu_read_unlock(); | |
ad246c99 | 2813 | |
80b9d236 | 2814 | re_arm: |
2815 | if (bond->params.arp_interval) | |
1f2cd845 DM |
2816 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
2817 | ||
b0929915 | 2818 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
2819 | if (!rtnl_trylock()) |
2820 | return; | |
b0929915 | 2821 | |
2822 | if (should_notify_peers) | |
2823 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
2824 | bond->dev); | |
2825 | if (should_notify_rtnl) | |
2826 | bond_slave_state_notify(bond); | |
2827 | ||
1f2cd845 DM |
2828 | rtnl_unlock(); |
2829 | } | |
1da177e4 LT |
2830 | } |
2831 | ||
1da177e4 LT |
2832 | /*-------------------------- netdev event handling --------------------------*/ |
2833 | ||
547942ca | 2834 | /* Change device name */ |
1da177e4 LT |
2835 | static int bond_event_changename(struct bonding *bond) |
2836 | { | |
1da177e4 LT |
2837 | bond_remove_proc_entry(bond); |
2838 | bond_create_proc_entry(bond); | |
7e083840 | 2839 | |
f073c7ca TI |
2840 | bond_debug_reregister(bond); |
2841 | ||
1da177e4 LT |
2842 | return NOTIFY_DONE; |
2843 | } | |
2844 | ||
3d632c3f SH |
2845 | static int bond_master_netdev_event(unsigned long event, |
2846 | struct net_device *bond_dev) | |
1da177e4 | 2847 | { |
454d7c9b | 2848 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2849 | |
2850 | switch (event) { | |
2851 | case NETDEV_CHANGENAME: | |
2852 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2853 | case NETDEV_UNREGISTER: |
2854 | bond_remove_proc_entry(event_bond); | |
2855 | break; | |
2856 | case NETDEV_REGISTER: | |
2857 | bond_create_proc_entry(event_bond); | |
2858 | break; | |
6c8d23f7 | 2859 | case NETDEV_NOTIFY_PEERS: |
2860 | if (event_bond->send_peer_notif) | |
2861 | event_bond->send_peer_notif--; | |
2862 | break; | |
1da177e4 LT |
2863 | default: |
2864 | break; | |
2865 | } | |
2866 | ||
2867 | return NOTIFY_DONE; | |
2868 | } | |
2869 | ||
3d632c3f SH |
2870 | static int bond_slave_netdev_event(unsigned long event, |
2871 | struct net_device *slave_dev) | |
1da177e4 | 2872 | { |
059b47e8 | 2873 | struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary; |
6101391d | 2874 | struct bonding *bond; |
2875 | struct net_device *bond_dev; | |
471cb5a3 JP |
2876 | u32 old_speed; |
2877 | u8 old_duplex; | |
1da177e4 | 2878 | |
6101391d | 2879 | /* A netdev event can be generated while enslaving a device |
2880 | * before netdev_rx_handler_register is called in which case | |
2881 | * slave will be NULL | |
2882 | */ | |
2883 | if (!slave) | |
2884 | return NOTIFY_DONE; | |
2885 | bond_dev = slave->bond->dev; | |
2886 | bond = slave->bond; | |
059b47e8 | 2887 | primary = rtnl_dereference(bond->primary_slave); |
6101391d | 2888 | |
1da177e4 LT |
2889 | switch (event) { |
2890 | case NETDEV_UNREGISTER: | |
8d2ada77 | 2891 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
2892 | bond_release_and_destroy(bond_dev, slave_dev); |
2893 | else | |
2894 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 2895 | break; |
98f41f69 | 2896 | case NETDEV_UP: |
1da177e4 | 2897 | case NETDEV_CHANGE: |
471cb5a3 JP |
2898 | old_speed = slave->speed; |
2899 | old_duplex = slave->duplex; | |
17d04500 | 2900 | |
471cb5a3 | 2901 | bond_update_speed_duplex(slave); |
17d04500 | 2902 | |
01844098 | 2903 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
471cb5a3 JP |
2904 | if (old_speed != slave->speed) |
2905 | bond_3ad_adapter_speed_changed(slave); | |
2906 | if (old_duplex != slave->duplex) | |
2907 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 2908 | } |
ee637714 MB |
2909 | /* Refresh slave-array if applicable! |
2910 | * If the setup does not use miimon or arpmon (mode-specific!), | |
2911 | * then these events will not cause the slave-array to be | |
2912 | * refreshed. This will cause xmit to use a slave that is not | |
2913 | * usable. Avoid such situation by refeshing the array at these | |
2914 | * events. If these (miimon/arpmon) parameters are configured | |
2915 | * then array gets refreshed twice and that should be fine! | |
2916 | */ | |
2917 | if (bond_mode_uses_xmit_hash(bond)) | |
2918 | bond_update_slave_arr(bond, NULL); | |
1da177e4 LT |
2919 | break; |
2920 | case NETDEV_DOWN: | |
ee637714 MB |
2921 | if (bond_mode_uses_xmit_hash(bond)) |
2922 | bond_update_slave_arr(bond, NULL); | |
1da177e4 LT |
2923 | break; |
2924 | case NETDEV_CHANGEMTU: | |
547942ca | 2925 | /* TODO: Should slaves be allowed to |
1da177e4 LT |
2926 | * independently alter their MTU? For |
2927 | * an active-backup bond, slaves need | |
2928 | * not be the same type of device, so | |
2929 | * MTUs may vary. For other modes, | |
2930 | * slaves arguably should have the | |
2931 | * same MTUs. To do this, we'd need to | |
2932 | * take over the slave's change_mtu | |
2933 | * function for the duration of their | |
2934 | * servitude. | |
2935 | */ | |
2936 | break; | |
2937 | case NETDEV_CHANGENAME: | |
3ec775b9 | 2938 | /* we don't care if we don't have primary set */ |
ec0865a9 | 2939 | if (!bond_uses_primary(bond) || |
3ec775b9 VF |
2940 | !bond->params.primary[0]) |
2941 | break; | |
2942 | ||
059b47e8 | 2943 | if (slave == primary) { |
3ec775b9 | 2944 | /* slave's name changed - he's no longer primary */ |
059b47e8 | 2945 | RCU_INIT_POINTER(bond->primary_slave, NULL); |
3ec775b9 VF |
2946 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { |
2947 | /* we have a new primary slave */ | |
059b47e8 | 2948 | rcu_assign_pointer(bond->primary_slave, slave); |
3ec775b9 VF |
2949 | } else { /* we didn't change primary - exit */ |
2950 | break; | |
2951 | } | |
2952 | ||
76444f50 | 2953 | netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n", |
059b47e8 | 2954 | primary ? slave_dev->name : "none"); |
f80889a5 | 2955 | |
2956 | block_netpoll_tx(); | |
3ec775b9 | 2957 | bond_select_active_slave(bond); |
f80889a5 | 2958 | unblock_netpoll_tx(); |
1da177e4 | 2959 | break; |
8531c5ff AK |
2960 | case NETDEV_FEAT_CHANGE: |
2961 | bond_compute_features(bond); | |
2962 | break; | |
4aa5dee4 JP |
2963 | case NETDEV_RESEND_IGMP: |
2964 | /* Propagate to master device */ | |
2965 | call_netdevice_notifiers(event, slave->bond->dev); | |
2966 | break; | |
1da177e4 LT |
2967 | default: |
2968 | break; | |
2969 | } | |
2970 | ||
2971 | return NOTIFY_DONE; | |
2972 | } | |
2973 | ||
547942ca | 2974 | /* bond_netdev_event: handle netdev notifier chain events. |
1da177e4 LT |
2975 | * |
2976 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 2977 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
2978 | * locks for us to safely manipulate the slave devices (RTNL lock, |
2979 | * dev_probe_lock). | |
2980 | */ | |
3d632c3f SH |
2981 | static int bond_netdev_event(struct notifier_block *this, |
2982 | unsigned long event, void *ptr) | |
1da177e4 | 2983 | { |
351638e7 | 2984 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 2985 | |
76444f50 | 2986 | netdev_dbg(event_dev, "event: %lx\n", event); |
1da177e4 | 2987 | |
0b680e75 JV |
2988 | if (!(event_dev->priv_flags & IFF_BONDING)) |
2989 | return NOTIFY_DONE; | |
2990 | ||
1da177e4 | 2991 | if (event_dev->flags & IFF_MASTER) { |
76444f50 | 2992 | netdev_dbg(event_dev, "IFF_MASTER\n"); |
1da177e4 LT |
2993 | return bond_master_netdev_event(event, event_dev); |
2994 | } | |
2995 | ||
2996 | if (event_dev->flags & IFF_SLAVE) { | |
76444f50 | 2997 | netdev_dbg(event_dev, "IFF_SLAVE\n"); |
1da177e4 LT |
2998 | return bond_slave_netdev_event(event, event_dev); |
2999 | } | |
3000 | ||
3001 | return NOTIFY_DONE; | |
3002 | } | |
3003 | ||
3004 | static struct notifier_block bond_netdev_notifier = { | |
3005 | .notifier_call = bond_netdev_event, | |
3006 | }; | |
3007 | ||
169a3e66 JV |
3008 | /*---------------------------- Hashing Policies -----------------------------*/ |
3009 | ||
32819dc1 NA |
3010 | /* L2 hash helper */ |
3011 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 | 3012 | { |
ce04d635 | 3013 | struct ethhdr *ep, hdr_tmp; |
6b923cb7 | 3014 | |
ce04d635 JX |
3015 | ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp); |
3016 | if (ep) | |
3017 | return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto; | |
6b923cb7 JE |
3018 | return 0; |
3019 | } | |
3020 | ||
32819dc1 NA |
3021 | /* Extract the appropriate headers based on bond's xmit policy */ |
3022 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
3023 | struct flow_keys *fk) | |
6f6652be | 3024 | { |
32819dc1 | 3025 | const struct ipv6hdr *iph6; |
4394542c | 3026 | const struct iphdr *iph; |
32819dc1 | 3027 | int noff, proto = -1; |
6b923cb7 | 3028 | |
32819dc1 NA |
3029 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) |
3030 | return skb_flow_dissect(skb, fk); | |
3031 | ||
3032 | fk->ports = 0; | |
3033 | noff = skb_network_offset(skb); | |
3034 | if (skb->protocol == htons(ETH_P_IP)) { | |
054bb880 | 3035 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
32819dc1 | 3036 | return false; |
6b923cb7 | 3037 | iph = ip_hdr(skb); |
32819dc1 NA |
3038 | fk->src = iph->saddr; |
3039 | fk->dst = iph->daddr; | |
3040 | noff += iph->ihl << 2; | |
3041 | if (!ip_is_fragment(iph)) | |
3042 | proto = iph->protocol; | |
3043 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
054bb880 | 3044 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6)))) |
32819dc1 NA |
3045 | return false; |
3046 | iph6 = ipv6_hdr(skb); | |
3047 | fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr); | |
3048 | fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr); | |
3049 | noff += sizeof(*iph6); | |
3050 | proto = iph6->nexthdr; | |
3051 | } else { | |
3052 | return false; | |
6f6652be | 3053 | } |
32819dc1 NA |
3054 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) |
3055 | fk->ports = skb_flow_get_ports(skb, noff, proto); | |
6f6652be | 3056 | |
32819dc1 | 3057 | return true; |
6f6652be JV |
3058 | } |
3059 | ||
32819dc1 NA |
3060 | /** |
3061 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3062 | * @bond: bonding device | |
3063 | * @skb: buffer to use for headers | |
32819dc1 NA |
3064 | * |
3065 | * This function will extract the necessary headers from the skb buffer and use | |
3066 | * them to generate a hash based on the xmit_policy set in the bonding device | |
169a3e66 | 3067 | */ |
ee62e868 | 3068 | u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) |
169a3e66 | 3069 | { |
32819dc1 NA |
3070 | struct flow_keys flow; |
3071 | u32 hash; | |
4394542c | 3072 | |
32819dc1 NA |
3073 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3074 | !bond_flow_dissect(bond, skb, &flow)) | |
ee62e868 | 3075 | return bond_eth_hash(skb); |
169a3e66 | 3076 | |
32819dc1 NA |
3077 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
3078 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) | |
3079 | hash = bond_eth_hash(skb); | |
3080 | else | |
3081 | hash = (__force u32)flow.ports; | |
3082 | hash ^= (__force u32)flow.dst ^ (__force u32)flow.src; | |
3083 | hash ^= (hash >> 16); | |
3084 | hash ^= (hash >> 8); | |
3085 | ||
ee62e868 | 3086 | return hash; |
169a3e66 JV |
3087 | } |
3088 | ||
1da177e4 LT |
3089 | /*-------------------------- Device entry points ----------------------------*/ |
3090 | ||
fbb0c41b | 3091 | static void bond_work_init_all(struct bonding *bond) |
3092 | { | |
3093 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3094 | bond_resend_igmp_join_requests_delayed); | |
3095 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3096 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
01844098 | 3097 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
fbb0c41b | 3098 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); |
3099 | else | |
3100 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3101 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
ee637714 | 3102 | INIT_DELAYED_WORK(&bond->slave_arr_work, bond_slave_arr_handler); |
fbb0c41b | 3103 | } |
3104 | ||
3105 | static void bond_work_cancel_all(struct bonding *bond) | |
3106 | { | |
3107 | cancel_delayed_work_sync(&bond->mii_work); | |
3108 | cancel_delayed_work_sync(&bond->arp_work); | |
3109 | cancel_delayed_work_sync(&bond->alb_work); | |
3110 | cancel_delayed_work_sync(&bond->ad_work); | |
3111 | cancel_delayed_work_sync(&bond->mcast_work); | |
ee637714 | 3112 | cancel_delayed_work_sync(&bond->slave_arr_work); |
fbb0c41b | 3113 | } |
3114 | ||
1da177e4 LT |
3115 | static int bond_open(struct net_device *bond_dev) |
3116 | { | |
454d7c9b | 3117 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3118 | struct list_head *iter; |
ba3211cc | 3119 | struct slave *slave; |
1da177e4 | 3120 | |
ba3211cc | 3121 | /* reset slave->backup and slave->inactive */ |
0965a1f3 | 3122 | if (bond_has_slaves(bond)) { |
9caff1e7 | 3123 | bond_for_each_slave(bond, slave, iter) { |
4740d638 ED |
3124 | if (bond_uses_primary(bond) && |
3125 | slave != rcu_access_pointer(bond->curr_active_slave)) { | |
5e5b0665 | 3126 | bond_set_slave_inactive_flags(slave, |
3127 | BOND_SLAVE_NOTIFY_NOW); | |
8bbe71a5 | 3128 | } else if (BOND_MODE(bond) != BOND_MODE_8023AD) { |
5e5b0665 | 3129 | bond_set_slave_active_flags(slave, |
3130 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3131 | } |
3132 | } | |
ba3211cc | 3133 | } |
ba3211cc | 3134 | |
fbb0c41b | 3135 | bond_work_init_all(bond); |
5a37e8ca | 3136 | |
58402054 | 3137 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3138 | /* bond_alb_initialize must be called before the timer |
3139 | * is started. | |
3140 | */ | |
01844098 | 3141 | if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) |
b473946a | 3142 | return -ENOMEM; |
e9f0fb88 MB |
3143 | if (bond->params.tlb_dynamic_lb) |
3144 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3145 | } |
3146 | ||
fbb0c41b | 3147 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3148 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3149 | |
3150 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3151 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3152 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3153 | } |
3154 | ||
01844098 | 3155 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1b76b316 | 3156 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3157 | /* register to receive LACPDUs */ |
3aba891d | 3158 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3159 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3160 | } |
3161 | ||
ee637714 MB |
3162 | if (bond_mode_uses_xmit_hash(bond)) |
3163 | bond_update_slave_arr(bond, NULL); | |
3164 | ||
1da177e4 LT |
3165 | return 0; |
3166 | } | |
3167 | ||
3168 | static int bond_close(struct net_device *bond_dev) | |
3169 | { | |
454d7c9b | 3170 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3171 | |
fbb0c41b | 3172 | bond_work_cancel_all(bond); |
6c8d23f7 | 3173 | bond->send_peer_notif = 0; |
ee8487c0 | 3174 | if (bond_is_lb(bond)) |
1da177e4 | 3175 | bond_alb_deinitialize(bond); |
3aba891d | 3176 | bond->recv_probe = NULL; |
1da177e4 LT |
3177 | |
3178 | return 0; | |
3179 | } | |
3180 | ||
28172739 ED |
3181 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3182 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3183 | { |
454d7c9b | 3184 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3185 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3186 | struct list_head *iter; |
1da177e4 | 3187 | struct slave *slave; |
1da177e4 | 3188 | |
5f0c5f73 | 3189 | memcpy(stats, &bond->bond_stats, sizeof(*stats)); |
1da177e4 | 3190 | |
9caff1e7 | 3191 | bond_for_each_slave(bond, slave, iter) { |
be1f3c2c | 3192 | const struct rtnl_link_stats64 *sstats = |
28172739 | 3193 | dev_get_stats(slave->dev, &temp); |
5f0c5f73 AG |
3194 | struct rtnl_link_stats64 *pstats = &slave->slave_stats; |
3195 | ||
3196 | stats->rx_packets += sstats->rx_packets - pstats->rx_packets; | |
3197 | stats->rx_bytes += sstats->rx_bytes - pstats->rx_bytes; | |
3198 | stats->rx_errors += sstats->rx_errors - pstats->rx_errors; | |
3199 | stats->rx_dropped += sstats->rx_dropped - pstats->rx_dropped; | |
3200 | ||
3201 | stats->tx_packets += sstats->tx_packets - pstats->tx_packets;; | |
3202 | stats->tx_bytes += sstats->tx_bytes - pstats->tx_bytes; | |
3203 | stats->tx_errors += sstats->tx_errors - pstats->tx_errors; | |
3204 | stats->tx_dropped += sstats->tx_dropped - pstats->tx_dropped; | |
3205 | ||
3206 | stats->multicast += sstats->multicast - pstats->multicast; | |
3207 | stats->collisions += sstats->collisions - pstats->collisions; | |
3208 | ||
3209 | stats->rx_length_errors += sstats->rx_length_errors - pstats->rx_length_errors; | |
3210 | stats->rx_over_errors += sstats->rx_over_errors - pstats->rx_over_errors; | |
3211 | stats->rx_crc_errors += sstats->rx_crc_errors - pstats->rx_crc_errors; | |
3212 | stats->rx_frame_errors += sstats->rx_frame_errors - pstats->rx_frame_errors; | |
3213 | stats->rx_fifo_errors += sstats->rx_fifo_errors - pstats->rx_fifo_errors; | |
3214 | stats->rx_missed_errors += sstats->rx_missed_errors - pstats->rx_missed_errors; | |
3215 | ||
3216 | stats->tx_aborted_errors += sstats->tx_aborted_errors - pstats->tx_aborted_errors; | |
3217 | stats->tx_carrier_errors += sstats->tx_carrier_errors - pstats->tx_carrier_errors; | |
3218 | stats->tx_fifo_errors += sstats->tx_fifo_errors - pstats->tx_fifo_errors; | |
3219 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors - pstats->tx_heartbeat_errors; | |
3220 | stats->tx_window_errors += sstats->tx_window_errors - pstats->tx_window_errors; | |
3221 | ||
3222 | /* save off the slave stats for the next run */ | |
3223 | memcpy(pstats, sstats, sizeof(*sstats)); | |
3224 | } | |
3225 | memcpy(&bond->bond_stats, stats, sizeof(*stats)); | |
1da177e4 LT |
3226 | |
3227 | return stats; | |
3228 | } | |
3229 | ||
3230 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3231 | { | |
080a06e1 | 3232 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3233 | struct net_device *slave_dev = NULL; |
3234 | struct ifbond k_binfo; | |
3235 | struct ifbond __user *u_binfo = NULL; | |
3236 | struct ifslave k_sinfo; | |
3237 | struct ifslave __user *u_sinfo = NULL; | |
3238 | struct mii_ioctl_data *mii = NULL; | |
d1fbd3ed | 3239 | struct bond_opt_value newval; |
387ff911 | 3240 | struct net *net; |
1da177e4 LT |
3241 | int res = 0; |
3242 | ||
76444f50 | 3243 | netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd); |
1da177e4 LT |
3244 | |
3245 | switch (cmd) { | |
1da177e4 LT |
3246 | case SIOCGMIIPHY: |
3247 | mii = if_mii(ifr); | |
3d632c3f | 3248 | if (!mii) |
1da177e4 | 3249 | return -EINVAL; |
3d632c3f | 3250 | |
1da177e4 LT |
3251 | mii->phy_id = 0; |
3252 | /* Fall Through */ | |
3253 | case SIOCGMIIREG: | |
547942ca | 3254 | /* We do this again just in case we were called by SIOCGMIIREG |
1da177e4 LT |
3255 | * instead of SIOCGMIIPHY. |
3256 | */ | |
3257 | mii = if_mii(ifr); | |
3d632c3f | 3258 | if (!mii) |
1da177e4 | 3259 | return -EINVAL; |
3d632c3f | 3260 | |
1da177e4 | 3261 | if (mii->reg_num == 1) { |
1da177e4 | 3262 | mii->val_out = 0; |
3d632c3f | 3263 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3264 | mii->val_out = BMSR_LSTATUS; |
1da177e4 LT |
3265 | } |
3266 | ||
3267 | return 0; | |
3268 | case BOND_INFO_QUERY_OLD: | |
3269 | case SIOCBONDINFOQUERY: | |
3270 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3271 | ||
3d632c3f | 3272 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3273 | return -EFAULT; |
1da177e4 LT |
3274 | |
3275 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3276 | if (res == 0 && |
3277 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3278 | return -EFAULT; | |
1da177e4 LT |
3279 | |
3280 | return res; | |
3281 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3282 | case SIOCBONDSLAVEINFOQUERY: | |
3283 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3284 | ||
3d632c3f | 3285 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3286 | return -EFAULT; |
1da177e4 LT |
3287 | |
3288 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3289 | if (res == 0 && |
3290 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3291 | return -EFAULT; | |
1da177e4 LT |
3292 | |
3293 | return res; | |
3294 | default: | |
1da177e4 LT |
3295 | break; |
3296 | } | |
3297 | ||
387ff911 G |
3298 | net = dev_net(bond_dev); |
3299 | ||
3300 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3301 | return -EPERM; |
1da177e4 | 3302 | |
0917b933 | 3303 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3304 | |
76444f50 | 3305 | netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev); |
1da177e4 | 3306 | |
3d632c3f | 3307 | if (!slave_dev) |
0917b933 | 3308 | return -ENODEV; |
1da177e4 | 3309 | |
76444f50 | 3310 | netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name); |
0917b933 YX |
3311 | switch (cmd) { |
3312 | case BOND_ENSLAVE_OLD: | |
3313 | case SIOCBONDENSLAVE: | |
3314 | res = bond_enslave(bond_dev, slave_dev); | |
3315 | break; | |
3316 | case BOND_RELEASE_OLD: | |
3317 | case SIOCBONDRELEASE: | |
3318 | res = bond_release(bond_dev, slave_dev); | |
3319 | break; | |
3320 | case BOND_SETHWADDR_OLD: | |
3321 | case SIOCBONDSETHWADDR: | |
3322 | bond_set_dev_addr(bond_dev, slave_dev); | |
3323 | res = 0; | |
3324 | break; | |
3325 | case BOND_CHANGE_ACTIVE_OLD: | |
3326 | case SIOCBONDCHANGEACTIVE: | |
d1fbd3ed NA |
3327 | bond_opt_initstr(&newval, slave_dev->name); |
3328 | res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval); | |
0917b933 YX |
3329 | break; |
3330 | default: | |
3331 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3332 | } |
3333 | ||
1da177e4 LT |
3334 | return res; |
3335 | } | |
3336 | ||
d03462b9 | 3337 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3338 | { |
454d7c9b | 3339 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3340 | |
d03462b9 JP |
3341 | if (change & IFF_PROMISC) |
3342 | bond_set_promiscuity(bond, | |
3343 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3344 | |
d03462b9 JP |
3345 | if (change & IFF_ALLMULTI) |
3346 | bond_set_allmulti(bond, | |
3347 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3348 | } | |
1da177e4 | 3349 | |
303d1cbf | 3350 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3351 | { |
3352 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3353 | struct list_head *iter; |
303d1cbf | 3354 | struct slave *slave; |
1da177e4 | 3355 | |
b3241870 | 3356 | rcu_read_lock(); |
ec0865a9 | 3357 | if (bond_uses_primary(bond)) { |
b3241870 | 3358 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3359 | if (slave) { |
3360 | dev_uc_sync(slave->dev, bond_dev); | |
3361 | dev_mc_sync(slave->dev, bond_dev); | |
3362 | } | |
303d1cbf | 3363 | } else { |
b3241870 | 3364 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3365 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3366 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3367 | } | |
1da177e4 | 3368 | } |
b3241870 | 3369 | rcu_read_unlock(); |
1da177e4 LT |
3370 | } |
3371 | ||
234bcf8a | 3372 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3373 | { |
234bcf8a | 3374 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3375 | const struct net_device_ops *slave_ops; |
3376 | struct neigh_parms parms; | |
dec1e90e | 3377 | struct slave *slave; |
234bcf8a SP |
3378 | int ret; |
3379 | ||
dec1e90e | 3380 | slave = bond_first_slave(bond); |
234bcf8a SP |
3381 | if (!slave) |
3382 | return 0; | |
234bcf8a | 3383 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3384 | if (!slave_ops->ndo_neigh_setup) |
3385 | return 0; | |
3386 | ||
3387 | parms.neigh_setup = NULL; | |
3388 | parms.neigh_cleanup = NULL; | |
3389 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3390 | if (ret) | |
3391 | return ret; | |
3392 | ||
547942ca | 3393 | /* Assign slave's neigh_cleanup to neighbour in case cleanup is called |
234bcf8a SP |
3394 | * after the last slave has been detached. Assumes that all slaves |
3395 | * utilize the same neigh_cleanup (true at this writing as only user | |
3396 | * is ipoib). | |
3397 | */ | |
3398 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3399 | ||
3400 | if (!parms.neigh_setup) | |
3401 | return 0; | |
3402 | ||
3403 | return parms.neigh_setup(n); | |
3404 | } | |
3405 | ||
547942ca | 3406 | /* The bonding ndo_neigh_setup is called at init time beofre any |
234bcf8a SP |
3407 | * slave exists. So we must declare proxy setup function which will |
3408 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3409 | * |
3410 | * It's also called by master devices (such as vlans) to setup their | |
3411 | * underlying devices. In that case - do nothing, we're already set up from | |
3412 | * our init. | |
234bcf8a SP |
3413 | */ |
3414 | static int bond_neigh_setup(struct net_device *dev, | |
3415 | struct neigh_parms *parms) | |
3416 | { | |
9918d5bf VF |
3417 | /* modify only our neigh_parms */ |
3418 | if (parms->dev == dev) | |
3419 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3420 | |
00829823 SH |
3421 | return 0; |
3422 | } | |
3423 | ||
547942ca | 3424 | /* Change the MTU of all of a master's slaves to match the master */ |
1da177e4 LT |
3425 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) |
3426 | { | |
454d7c9b | 3427 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3428 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3429 | struct list_head *iter; |
1da177e4 | 3430 | int res = 0; |
1da177e4 | 3431 | |
76444f50 | 3432 | netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu); |
1da177e4 | 3433 | |
9caff1e7 | 3434 | bond_for_each_slave(bond, slave, iter) { |
76444f50 VF |
3435 | netdev_dbg(bond_dev, "s %p c_m %p\n", |
3436 | slave, slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3437 | |
1da177e4 LT |
3438 | res = dev_set_mtu(slave->dev, new_mtu); |
3439 | ||
3440 | if (res) { | |
3441 | /* If we failed to set the slave's mtu to the new value | |
3442 | * we must abort the operation even in ACTIVE_BACKUP | |
3443 | * mode, because if we allow the backup slaves to have | |
3444 | * different mtu values than the active slave we'll | |
3445 | * need to change their mtu when doing a failover. That | |
3446 | * means changing their mtu from timer context, which | |
3447 | * is probably not a good idea. | |
3448 | */ | |
76444f50 VF |
3449 | netdev_dbg(bond_dev, "err %d %s\n", res, |
3450 | slave->dev->name); | |
1da177e4 LT |
3451 | goto unwind; |
3452 | } | |
3453 | } | |
3454 | ||
3455 | bond_dev->mtu = new_mtu; | |
3456 | ||
3457 | return 0; | |
3458 | ||
3459 | unwind: | |
3460 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3461 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3462 | int tmp_res; |
3463 | ||
81f23b13 VF |
3464 | if (rollback_slave == slave) |
3465 | break; | |
3466 | ||
3467 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3468 | if (tmp_res) { |
76444f50 VF |
3469 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3470 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3471 | } |
3472 | } | |
3473 | ||
3474 | return res; | |
3475 | } | |
3476 | ||
547942ca | 3477 | /* Change HW address |
1da177e4 LT |
3478 | * |
3479 | * Note that many devices must be down to change the HW address, and | |
3480 | * downing the master releases all slaves. We can make bonds full of | |
3481 | * bonding devices to test this, however. | |
3482 | */ | |
3483 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3484 | { | |
454d7c9b | 3485 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3486 | struct slave *slave, *rollback_slave; |
1da177e4 | 3487 | struct sockaddr *sa = addr, tmp_sa; |
9caff1e7 | 3488 | struct list_head *iter; |
1da177e4 | 3489 | int res = 0; |
1da177e4 | 3490 | |
01844098 | 3491 | if (BOND_MODE(bond) == BOND_MODE_ALB) |
eb7cc59a SH |
3492 | return bond_alb_set_mac_address(bond_dev, addr); |
3493 | ||
3494 | ||
76444f50 | 3495 | netdev_dbg(bond_dev, "bond=%p\n", bond); |
1da177e4 | 3496 | |
1b5acd29 JV |
3497 | /* If fail_over_mac is enabled, do nothing and return success. |
3498 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3499 | */ |
cc689aaa | 3500 | if (bond->params.fail_over_mac && |
01844098 | 3501 | BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
dd957c57 | 3502 | return 0; |
2ab82852 | 3503 | |
3d632c3f | 3504 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3505 | return -EADDRNOTAVAIL; |
1da177e4 | 3506 | |
9caff1e7 | 3507 | bond_for_each_slave(bond, slave, iter) { |
76444f50 | 3508 | netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name); |
1da177e4 LT |
3509 | res = dev_set_mac_address(slave->dev, addr); |
3510 | if (res) { | |
3511 | /* TODO: consider downing the slave | |
3512 | * and retry ? | |
3513 | * User should expect communications | |
3514 | * breakage anyway until ARP finish | |
3515 | * updating, so... | |
3516 | */ | |
76444f50 | 3517 | netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3518 | goto unwind; |
3519 | } | |
3520 | } | |
3521 | ||
3522 | /* success */ | |
3523 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3524 | return 0; | |
3525 | ||
3526 | unwind: | |
3527 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3528 | tmp_sa.sa_family = bond_dev->type; | |
3529 | ||
3530 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3531 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3532 | int tmp_res; |
3533 | ||
81f23b13 VF |
3534 | if (rollback_slave == slave) |
3535 | break; | |
3536 | ||
3537 | tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa); | |
1da177e4 | 3538 | if (tmp_res) { |
76444f50 VF |
3539 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3540 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3541 | } |
3542 | } | |
3543 | ||
3544 | return res; | |
3545 | } | |
3546 | ||
15077228 NA |
3547 | /** |
3548 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3549 | * @bond: bonding device that is transmitting | |
3550 | * @skb: buffer to transmit | |
3551 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3552 | * | |
3553 | * This function tries to transmit through slave with slave_id but in case | |
3554 | * it fails, it tries to find the first available slave for transmission. | |
3555 | * The skb is consumed in all cases, thus the function is void. | |
3556 | */ | |
da131ddb | 3557 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3558 | { |
9caff1e7 | 3559 | struct list_head *iter; |
15077228 NA |
3560 | struct slave *slave; |
3561 | int i = slave_id; | |
3562 | ||
3563 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3564 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 | 3565 | if (--i < 0) { |
891ab54d | 3566 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3567 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3568 | return; | |
3569 | } | |
3570 | } | |
3571 | } | |
3572 | ||
3573 | /* Here we start from the first slave up to slave_id */ | |
3574 | i = slave_id; | |
9caff1e7 | 3575 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3576 | if (--i < 0) |
3577 | break; | |
891ab54d | 3578 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3579 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3580 | return; | |
3581 | } | |
3582 | } | |
3583 | /* no slave that can tx has been found */ | |
31aa860e | 3584 | bond_tx_drop(bond->dev, skb); |
15077228 NA |
3585 | } |
3586 | ||
73958329 NA |
3587 | /** |
3588 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3589 | * @bond: bonding device to use | |
3590 | * | |
3591 | * Based on the value of the bonding device's packets_per_slave parameter | |
3592 | * this function generates a slave id, which is usually used as the next | |
3593 | * slave to transmit through. | |
3594 | */ | |
3595 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3596 | { | |
73958329 | 3597 | u32 slave_id; |
809fa972 HFS |
3598 | struct reciprocal_value reciprocal_packets_per_slave; |
3599 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
3600 | |
3601 | switch (packets_per_slave) { | |
3602 | case 0: | |
3603 | slave_id = prandom_u32(); | |
3604 | break; | |
3605 | case 1: | |
3606 | slave_id = bond->rr_tx_counter; | |
3607 | break; | |
3608 | default: | |
809fa972 HFS |
3609 | reciprocal_packets_per_slave = |
3610 | bond->params.reciprocal_packets_per_slave; | |
73958329 | 3611 | slave_id = reciprocal_divide(bond->rr_tx_counter, |
809fa972 | 3612 | reciprocal_packets_per_slave); |
73958329 NA |
3613 | break; |
3614 | } | |
3615 | bond->rr_tx_counter++; | |
3616 | ||
3617 | return slave_id; | |
3618 | } | |
3619 | ||
1da177e4 LT |
3620 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3621 | { | |
454d7c9b | 3622 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3623 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3624 | struct slave *slave; |
73958329 | 3625 | u32 slave_id; |
1da177e4 | 3626 | |
73958329 | 3627 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3628 | * default for sending IGMP traffic. For failover purposes one |
3629 | * needs to maintain some consistency for the interface that will | |
3630 | * send the join/membership reports. The curr_active_slave found | |
3631 | * will send all of this type of traffic. | |
cf5f9044 | 3632 | */ |
15077228 | 3633 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3634 | slave = rcu_dereference(bond->curr_active_slave); |
a64d044e | 3635 | if (slave) |
15077228 NA |
3636 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3637 | else | |
3638 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3639 | } else { |
9a72c2da NA |
3640 | int slave_cnt = ACCESS_ONCE(bond->slave_cnt); |
3641 | ||
3642 | if (likely(slave_cnt)) { | |
3643 | slave_id = bond_rr_gen_slave_id(bond); | |
3644 | bond_xmit_slave_id(bond, skb, slave_id % slave_cnt); | |
3645 | } else { | |
31aa860e | 3646 | bond_tx_drop(bond_dev, skb); |
9a72c2da | 3647 | } |
1da177e4 | 3648 | } |
0693e88e | 3649 | |
ec634fe3 | 3650 | return NETDEV_TX_OK; |
1da177e4 LT |
3651 | } |
3652 | ||
547942ca | 3653 | /* In active-backup mode, we know that bond->curr_active_slave is always valid if |
1da177e4 LT |
3654 | * the bond has a usable interface. |
3655 | */ | |
3656 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3657 | { | |
454d7c9b | 3658 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3659 | struct slave *slave; |
1da177e4 | 3660 | |
278b2083 | 3661 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3662 | if (slave) |
15077228 NA |
3663 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3664 | else | |
31aa860e | 3665 | bond_tx_drop(bond_dev, skb); |
0693e88e | 3666 | |
ec634fe3 | 3667 | return NETDEV_TX_OK; |
1da177e4 LT |
3668 | } |
3669 | ||
ee637714 MB |
3670 | /* Use this to update slave_array when (a) it's not appropriate to update |
3671 | * slave_array right away (note that update_slave_array() may sleep) | |
3672 | * and / or (b) RTNL is not held. | |
1da177e4 | 3673 | */ |
ee637714 | 3674 | void bond_slave_arr_work_rearm(struct bonding *bond, unsigned long delay) |
1da177e4 | 3675 | { |
ee637714 MB |
3676 | queue_delayed_work(bond->wq, &bond->slave_arr_work, delay); |
3677 | } | |
1da177e4 | 3678 | |
ee637714 MB |
3679 | /* Slave array work handler. Holds only RTNL */ |
3680 | static void bond_slave_arr_handler(struct work_struct *work) | |
3681 | { | |
3682 | struct bonding *bond = container_of(work, struct bonding, | |
3683 | slave_arr_work.work); | |
3684 | int ret; | |
3685 | ||
3686 | if (!rtnl_trylock()) | |
3687 | goto err; | |
3688 | ||
3689 | ret = bond_update_slave_arr(bond, NULL); | |
3690 | rtnl_unlock(); | |
3691 | if (ret) { | |
3692 | pr_warn_ratelimited("Failed to update slave array from WT\n"); | |
3693 | goto err; | |
3694 | } | |
3695 | return; | |
3696 | ||
3697 | err: | |
3698 | bond_slave_arr_work_rearm(bond, 1); | |
3699 | } | |
3700 | ||
3701 | /* Build the usable slaves array in control path for modes that use xmit-hash | |
3702 | * to determine the slave interface - | |
3703 | * (a) BOND_MODE_8023AD | |
3704 | * (b) BOND_MODE_XOR | |
3705 | * (c) BOND_MODE_TLB && tlb_dynamic_lb == 0 | |
3706 | * | |
3707 | * The caller is expected to hold RTNL only and NO other lock! | |
3708 | */ | |
3709 | int bond_update_slave_arr(struct bonding *bond, struct slave *skipslave) | |
3710 | { | |
3711 | struct slave *slave; | |
3712 | struct list_head *iter; | |
3713 | struct bond_up_slave *new_arr, *old_arr; | |
3714 | int slaves_in_agg; | |
3715 | int agg_id = 0; | |
3716 | int ret = 0; | |
3717 | ||
3718 | #ifdef CONFIG_LOCKDEP | |
3719 | WARN_ON(lockdep_is_held(&bond->mode_lock)); | |
3720 | #endif | |
3721 | ||
3722 | new_arr = kzalloc(offsetof(struct bond_up_slave, arr[bond->slave_cnt]), | |
3723 | GFP_KERNEL); | |
3724 | if (!new_arr) { | |
3725 | ret = -ENOMEM; | |
3726 | pr_err("Failed to build slave-array.\n"); | |
3727 | goto out; | |
3728 | } | |
3729 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
3730 | struct ad_info ad_info; | |
3731 | ||
3732 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3733 | pr_debug("bond_3ad_get_active_agg_info failed\n"); | |
3734 | kfree_rcu(new_arr, rcu); | |
3735 | /* No active aggragator means it's not safe to use | |
3736 | * the previous array. | |
3737 | */ | |
3738 | old_arr = rtnl_dereference(bond->slave_arr); | |
3739 | if (old_arr) { | |
3740 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
3741 | kfree_rcu(old_arr, rcu); | |
3742 | } | |
3743 | goto out; | |
3744 | } | |
3745 | slaves_in_agg = ad_info.ports; | |
3746 | agg_id = ad_info.aggregator_id; | |
3747 | } | |
3748 | bond_for_each_slave(bond, slave, iter) { | |
3749 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { | |
3750 | struct aggregator *agg; | |
3751 | ||
3752 | agg = SLAVE_AD_INFO(slave)->port.aggregator; | |
3753 | if (!agg || agg->aggregator_identifier != agg_id) | |
3754 | continue; | |
3755 | } | |
3756 | if (!bond_slave_can_tx(slave)) | |
3757 | continue; | |
3758 | if (skipslave == slave) | |
3759 | continue; | |
3760 | new_arr->arr[new_arr->count++] = slave; | |
3761 | } | |
3762 | ||
3763 | old_arr = rtnl_dereference(bond->slave_arr); | |
3764 | rcu_assign_pointer(bond->slave_arr, new_arr); | |
3765 | if (old_arr) | |
3766 | kfree_rcu(old_arr, rcu); | |
3767 | out: | |
3768 | if (ret != 0 && skipslave) { | |
3769 | int idx; | |
3770 | ||
3771 | /* Rare situation where caller has asked to skip a specific | |
3772 | * slave but allocation failed (most likely!). BTW this is | |
3773 | * only possible when the call is initiated from | |
3774 | * __bond_release_one(). In this situation; overwrite the | |
3775 | * skipslave entry in the array with the last entry from the | |
3776 | * array to avoid a situation where the xmit path may choose | |
3777 | * this to-be-skipped slave to send a packet out. | |
3778 | */ | |
3779 | old_arr = rtnl_dereference(bond->slave_arr); | |
3780 | for (idx = 0; idx < old_arr->count; idx++) { | |
3781 | if (skipslave == old_arr->arr[idx]) { | |
3782 | old_arr->arr[idx] = | |
3783 | old_arr->arr[old_arr->count-1]; | |
3784 | old_arr->count--; | |
3785 | break; | |
3786 | } | |
3787 | } | |
3788 | } | |
3789 | return ret; | |
3790 | } | |
3791 | ||
3792 | /* Use this Xmit function for 3AD as well as XOR modes. The current | |
3793 | * usable slave array is formed in the control path. The xmit function | |
3794 | * just calculates hash and sends the packet out. | |
3795 | */ | |
30369104 | 3796 | static int bond_3ad_xor_xmit(struct sk_buff *skb, struct net_device *dev) |
ee637714 MB |
3797 | { |
3798 | struct bonding *bond = netdev_priv(dev); | |
3799 | struct slave *slave; | |
3800 | struct bond_up_slave *slaves; | |
3801 | unsigned int count; | |
3802 | ||
3803 | slaves = rcu_dereference(bond->slave_arr); | |
3804 | count = slaves ? ACCESS_ONCE(slaves->count) : 0; | |
3805 | if (likely(count)) { | |
3806 | slave = slaves->arr[bond_xmit_hash(bond, skb) % count]; | |
3807 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3808 | } else { | |
31aa860e | 3809 | bond_tx_drop(dev, skb); |
ee637714 | 3810 | } |
0693e88e | 3811 | |
ec634fe3 | 3812 | return NETDEV_TX_OK; |
1da177e4 LT |
3813 | } |
3814 | ||
78a646ce | 3815 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3816 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3817 | { | |
454d7c9b | 3818 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3819 | struct slave *slave = NULL; |
9caff1e7 | 3820 | struct list_head *iter; |
1da177e4 | 3821 | |
9caff1e7 | 3822 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
3823 | if (bond_is_last_slave(bond, slave)) |
3824 | break; | |
b6adc610 | 3825 | if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
78a646ce | 3826 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 | 3827 | |
78a646ce | 3828 | if (!skb2) { |
9152e26d | 3829 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
3830 | bond_dev->name, __func__); | |
78a646ce | 3831 | continue; |
1da177e4 | 3832 | } |
78a646ce | 3833 | bond_dev_queue_xmit(bond, skb2, slave->dev); |
1da177e4 LT |
3834 | } |
3835 | } | |
b6adc610 | 3836 | if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) |
78a646ce NA |
3837 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3838 | else | |
31aa860e | 3839 | bond_tx_drop(bond_dev, skb); |
3d632c3f | 3840 | |
ec634fe3 | 3841 | return NETDEV_TX_OK; |
1da177e4 LT |
3842 | } |
3843 | ||
3844 | /*------------------------- Device initialization ---------------------------*/ | |
3845 | ||
547942ca | 3846 | /* Lookup the slave that corresponds to a qid */ |
bb1d9123 AG |
3847 | static inline int bond_slave_override(struct bonding *bond, |
3848 | struct sk_buff *skb) | |
3849 | { | |
bb1d9123 | 3850 | struct slave *slave = NULL; |
9caff1e7 | 3851 | struct list_head *iter; |
bb1d9123 | 3852 | |
0693e88e MM |
3853 | if (!skb->queue_mapping) |
3854 | return 1; | |
bb1d9123 AG |
3855 | |
3856 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 3857 | bond_for_each_slave_rcu(bond, slave, iter) { |
3858 | if (slave->queue_id == skb->queue_mapping) { | |
891ab54d | 3859 | if (bond_slave_can_tx(slave)) { |
3900f290 | 3860 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3861 | return 0; | |
3862 | } | |
3863 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
3864 | break; |
3865 | } | |
3866 | } | |
3867 | ||
3900f290 | 3868 | return 1; |
bb1d9123 AG |
3869 | } |
3870 | ||
374eeb5a | 3871 | |
f663dd9a | 3872 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 3873 | void *accel_priv, select_queue_fallback_t fallback) |
bb1d9123 | 3874 | { |
547942ca | 3875 | /* This helper function exists to help dev_pick_tx get the correct |
fd0e435b | 3876 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3877 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3878 | * way down to the bonding driver. | |
3879 | */ | |
fd0e435b PO |
3880 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3881 | ||
547942ca | 3882 | /* Save the original txq to restore before passing to the driver */ |
df4ab5b3 | 3883 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3884 | |
fd0e435b | 3885 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3886 | do { |
fd0e435b | 3887 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3888 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3889 | } |
3890 | return txq; | |
bb1d9123 AG |
3891 | } |
3892 | ||
0693e88e | 3893 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3894 | { |
bb1d9123 AG |
3895 | struct bonding *bond = netdev_priv(dev); |
3896 | ||
d1e2e5cd VF |
3897 | if (bond_should_override_tx_queue(bond) && |
3898 | !bond_slave_override(bond, skb)) | |
3899 | return NETDEV_TX_OK; | |
00829823 | 3900 | |
01844098 | 3901 | switch (BOND_MODE(bond)) { |
00829823 SH |
3902 | case BOND_MODE_ROUNDROBIN: |
3903 | return bond_xmit_roundrobin(skb, dev); | |
3904 | case BOND_MODE_ACTIVEBACKUP: | |
3905 | return bond_xmit_activebackup(skb, dev); | |
ee637714 | 3906 | case BOND_MODE_8023AD: |
00829823 | 3907 | case BOND_MODE_XOR: |
ee637714 | 3908 | return bond_3ad_xor_xmit(skb, dev); |
00829823 SH |
3909 | case BOND_MODE_BROADCAST: |
3910 | return bond_xmit_broadcast(skb, dev); | |
00829823 | 3911 | case BOND_MODE_ALB: |
00829823 | 3912 | return bond_alb_xmit(skb, dev); |
f05b42ea MB |
3913 | case BOND_MODE_TLB: |
3914 | return bond_tlb_xmit(skb, dev); | |
00829823 SH |
3915 | default: |
3916 | /* Should never happen, mode already checked */ | |
76444f50 | 3917 | netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); |
00829823 | 3918 | WARN_ON_ONCE(1); |
31aa860e | 3919 | bond_tx_drop(dev, skb); |
00829823 SH |
3920 | return NETDEV_TX_OK; |
3921 | } | |
3922 | } | |
3923 | ||
0693e88e MM |
3924 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3925 | { | |
3926 | struct bonding *bond = netdev_priv(dev); | |
3927 | netdev_tx_t ret = NETDEV_TX_OK; | |
3928 | ||
547942ca | 3929 | /* If we risk deadlock from transmitting this in the |
0693e88e MM |
3930 | * netpoll path, tell netpoll to queue the frame for later tx |
3931 | */ | |
054bb880 | 3932 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
3933 | return NETDEV_TX_BUSY; |
3934 | ||
278b2083 | 3935 | rcu_read_lock(); |
0965a1f3 | 3936 | if (bond_has_slaves(bond)) |
0693e88e MM |
3937 | ret = __bond_start_xmit(skb, dev); |
3938 | else | |
31aa860e | 3939 | bond_tx_drop(dev, skb); |
278b2083 | 3940 | rcu_read_unlock(); |
0693e88e MM |
3941 | |
3942 | return ret; | |
3943 | } | |
00829823 | 3944 | |
bb5b052f AG |
3945 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3946 | struct ethtool_cmd *ecmd) | |
3947 | { | |
3948 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3949 | unsigned long speed = 0; |
9caff1e7 | 3950 | struct list_head *iter; |
dec1e90e | 3951 | struct slave *slave; |
bb5b052f AG |
3952 | |
3953 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3954 | ecmd->port = PORT_OTHER; | |
3955 | ||
8557cd74 | 3956 | /* Since bond_slave_can_tx returns false for all inactive or down slaves, we |
bb5b052f AG |
3957 | * do not need to check mode. Though link speed might not represent |
3958 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3959 | * this is an accurate maximum. | |
3960 | */ | |
9caff1e7 | 3961 | bond_for_each_slave(bond, slave, iter) { |
8557cd74 | 3962 | if (bond_slave_can_tx(slave)) { |
bb5b052f AG |
3963 | if (slave->speed != SPEED_UNKNOWN) |
3964 | speed += slave->speed; | |
3965 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3966 | slave->duplex != DUPLEX_UNKNOWN) | |
3967 | ecmd->duplex = slave->duplex; | |
3968 | } | |
3969 | } | |
3970 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
dec1e90e | 3971 | |
bb5b052f AG |
3972 | return 0; |
3973 | } | |
3974 | ||
217df670 | 3975 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3976 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3977 | { |
7826d43f JP |
3978 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3979 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3980 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3981 | BOND_ABI_VERSION); | |
217df670 JV |
3982 | } |
3983 | ||
7282d491 | 3984 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3985 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3986 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3987 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3988 | }; |
3989 | ||
eb7cc59a | 3990 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3991 | .ndo_init = bond_init, |
9e71626c | 3992 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3993 | .ndo_open = bond_open, |
3994 | .ndo_stop = bond_close, | |
00829823 | 3995 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3996 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3997 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3998 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3999 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 4000 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 4001 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4002 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4003 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4004 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4005 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4006 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4007 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4008 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4009 | .ndo_poll_controller = bond_poll_controller, | |
4010 | #endif | |
9232ecca JP |
4011 | .ndo_add_slave = bond_enslave, |
4012 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4013 | .ndo_fix_features = bond_fix_features, |
c158cba3 RP |
4014 | .ndo_bridge_setlink = ndo_dflt_netdev_switch_port_bridge_setlink, |
4015 | .ndo_bridge_dellink = ndo_dflt_netdev_switch_port_bridge_dellink, | |
eb7cc59a SH |
4016 | }; |
4017 | ||
b3f92b63 DG |
4018 | static const struct device_type bond_type = { |
4019 | .name = "bond", | |
4020 | }; | |
4021 | ||
9e2e61fb AW |
4022 | static void bond_destructor(struct net_device *bond_dev) |
4023 | { | |
4024 | struct bonding *bond = netdev_priv(bond_dev); | |
4025 | if (bond->wq) | |
4026 | destroy_workqueue(bond->wq); | |
4027 | free_netdev(bond_dev); | |
4028 | } | |
4029 | ||
0a2a78c4 | 4030 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4031 | { |
454d7c9b | 4032 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4033 | |
b7435628 | 4034 | spin_lock_init(&bond->mode_lock); |
d2991f75 | 4035 | bond->params = bonding_defaults; |
1da177e4 LT |
4036 | |
4037 | /* Initialize pointers */ | |
1da177e4 | 4038 | bond->dev = bond_dev; |
1da177e4 LT |
4039 | |
4040 | /* Initialize the device entry points */ | |
181470fc | 4041 | ether_setup(bond_dev); |
eb7cc59a | 4042 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4043 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 4044 | |
9e2e61fb | 4045 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4046 | |
b3f92b63 DG |
4047 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4048 | ||
1da177e4 LT |
4049 | /* Initialize the device options */ |
4050 | bond_dev->tx_queue_len = 0; | |
4051 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
c565b488 | 4052 | bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT; |
550fd08c | 4053 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4054 | |
547942ca | 4055 | /* don't acquire bond device's netif_tx_lock when transmitting */ |
1da177e4 LT |
4056 | bond_dev->features |= NETIF_F_LLTX; |
4057 | ||
4058 | /* By default, we declare the bond to be fully | |
4059 | * VLAN hardware accelerated capable. Special | |
4060 | * care is taken in the various xmit functions | |
4061 | * when there are slaves that are not hw accel | |
4062 | * capable | |
4063 | */ | |
1da177e4 | 4064 | |
f9399814 WC |
4065 | /* Don't allow bond devices to change network namespaces. */ |
4066 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
4067 | ||
b2a103e6 | 4068 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4069 | NETIF_F_HW_VLAN_CTAG_TX | |
4070 | NETIF_F_HW_VLAN_CTAG_RX | | |
4071 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4072 | |
34324dc2 | 4073 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
24f87d4c | 4074 | bond_dev->hw_features |= NETIF_F_GSO_ENCAP_ALL; |
b2a103e6 | 4075 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4076 | } |
4077 | ||
547942ca NA |
4078 | /* Destroy a bonding device. |
4079 | * Must be under rtnl_lock when this function is called. | |
4080 | */ | |
c67dfb29 | 4081 | static void bond_uninit(struct net_device *bond_dev) |
a434e43f | 4082 | { |
454d7c9b | 4083 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
4084 | struct list_head *iter; |
4085 | struct slave *slave; | |
ee637714 | 4086 | struct bond_up_slave *arr; |
a434e43f | 4087 | |
f6dc31a8 WC |
4088 | bond_netpoll_cleanup(bond_dev); |
4089 | ||
c67dfb29 | 4090 | /* Release the bonded slaves */ |
544a028e | 4091 | bond_for_each_slave(bond, slave, iter) |
dec1e90e | 4092 | __bond_release_one(bond_dev, slave->dev, true); |
76444f50 | 4093 | netdev_info(bond_dev, "Released all slaves\n"); |
c67dfb29 | 4094 | |
ee637714 MB |
4095 | arr = rtnl_dereference(bond->slave_arr); |
4096 | if (arr) { | |
4097 | RCU_INIT_POINTER(bond->slave_arr, NULL); | |
4098 | kfree_rcu(arr, rcu); | |
4099 | } | |
4100 | ||
a434e43f JV |
4101 | list_del(&bond->bond_list); |
4102 | ||
f073c7ca | 4103 | bond_debug_unregister(bond); |
a434e43f JV |
4104 | } |
4105 | ||
1da177e4 LT |
4106 | /*------------------------- Module initialization ---------------------------*/ |
4107 | ||
1da177e4 LT |
4108 | static int bond_check_params(struct bond_params *params) |
4109 | { | |
5a5c5fd4 | 4110 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 4111 | struct bond_opt_value newval; |
4112 | const struct bond_opt_value *valptr; | |
8599b52e | 4113 | int arp_all_targets_value; |
f5b2b966 | 4114 | |
547942ca | 4115 | /* Convert string parameters. */ |
1da177e4 | 4116 | if (mode) { |
2b3798d5 NA |
4117 | bond_opt_initstr(&newval, mode); |
4118 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
4119 | if (!valptr) { | |
4120 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
4121 | return -EINVAL; |
4122 | } | |
2b3798d5 | 4123 | bond_mode = valptr->value; |
1da177e4 LT |
4124 | } |
4125 | ||
169a3e66 JV |
4126 | if (xmit_hash_policy) { |
4127 | if ((bond_mode != BOND_MODE_XOR) && | |
f05b42ea MB |
4128 | (bond_mode != BOND_MODE_8023AD) && |
4129 | (bond_mode != BOND_MODE_TLB)) { | |
a4aee5c8 | 4130 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 4131 | bond_mode_name(bond_mode)); |
169a3e66 | 4132 | } else { |
a4b32ce7 NA |
4133 | bond_opt_initstr(&newval, xmit_hash_policy); |
4134 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
4135 | &newval); | |
4136 | if (!valptr) { | |
a4aee5c8 | 4137 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
4138 | xmit_hash_policy); |
4139 | return -EINVAL; | |
4140 | } | |
a4b32ce7 | 4141 | xmit_hashtype = valptr->value; |
169a3e66 JV |
4142 | } |
4143 | } | |
4144 | ||
1da177e4 LT |
4145 | if (lacp_rate) { |
4146 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4147 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4148 | bond_mode_name(bond_mode)); | |
1da177e4 | 4149 | } else { |
d3131de7 NA |
4150 | bond_opt_initstr(&newval, lacp_rate); |
4151 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
4152 | &newval); | |
4153 | if (!valptr) { | |
a4aee5c8 | 4154 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 4155 | lacp_rate); |
1da177e4 LT |
4156 | return -EINVAL; |
4157 | } | |
d3131de7 | 4158 | lacp_fast = valptr->value; |
1da177e4 LT |
4159 | } |
4160 | } | |
4161 | ||
fd989c83 | 4162 | if (ad_select) { |
548d28bd | 4163 | bond_opt_initstr(&newval, ad_select); |
9e5f5eeb NA |
4164 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), |
4165 | &newval); | |
4166 | if (!valptr) { | |
4167 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
4168 | return -EINVAL; |
4169 | } | |
9e5f5eeb NA |
4170 | params->ad_select = valptr->value; |
4171 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 4172 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4173 | } else { |
4174 | params->ad_select = BOND_AD_STABLE; | |
4175 | } | |
4176 | ||
f5841306 | 4177 | if (max_bonds < 0) { |
91565ebb JP |
4178 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4179 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4180 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4181 | } | |
4182 | ||
4183 | if (miimon < 0) { | |
91565ebb JP |
4184 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4185 | miimon, INT_MAX); | |
b98d9c66 | 4186 | miimon = 0; |
1da177e4 LT |
4187 | } |
4188 | ||
4189 | if (updelay < 0) { | |
91565ebb JP |
4190 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4191 | updelay, INT_MAX); | |
1da177e4 LT |
4192 | updelay = 0; |
4193 | } | |
4194 | ||
4195 | if (downdelay < 0) { | |
91565ebb JP |
4196 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4197 | downdelay, INT_MAX); | |
1da177e4 LT |
4198 | downdelay = 0; |
4199 | } | |
4200 | ||
4201 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
91565ebb JP |
4202 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4203 | use_carrier); | |
1da177e4 LT |
4204 | use_carrier = 1; |
4205 | } | |
4206 | ||
ad246c99 | 4207 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
4208 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
4209 | num_peer_notif); | |
ad246c99 BH |
4210 | num_peer_notif = 1; |
4211 | } | |
4212 | ||
834db4bc | 4213 | /* reset values for 802.3ad/TLB/ALB */ |
267bed77 | 4214 | if (!bond_mode_uses_arp(bond_mode)) { |
1da177e4 | 4215 | if (!miimon) { |
91565ebb JP |
4216 | 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"); |
4217 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 4218 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4219 | } |
4220 | } | |
4221 | ||
bb1d9123 | 4222 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
4223 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
4224 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
4225 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
4226 | } | |
4227 | ||
ebd8e497 | 4228 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
4229 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
4230 | all_slaves_active); | |
ebd8e497 AG |
4231 | all_slaves_active = 0; |
4232 | } | |
4233 | ||
c2952c31 | 4234 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
4235 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
4236 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
4237 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
4238 | } | |
4239 | ||
aa59d851 NA |
4240 | bond_opt_initval(&newval, packets_per_slave); |
4241 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
4242 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
4243 | packets_per_slave, USHRT_MAX); | |
4244 | packets_per_slave = 1; | |
4245 | } | |
4246 | ||
1da177e4 | 4247 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4248 | 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", |
4249 | updelay); | |
1da177e4 LT |
4250 | } |
4251 | ||
4252 | if (!miimon) { | |
4253 | if (updelay || downdelay) { | |
4254 | /* just warn the user the up/down delay will have | |
4255 | * no effect since miimon is zero... | |
4256 | */ | |
91565ebb JP |
4257 | 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", |
4258 | updelay, downdelay); | |
1da177e4 LT |
4259 | } |
4260 | } else { | |
4261 | /* don't allow arp monitoring */ | |
4262 | if (arp_interval) { | |
91565ebb JP |
4263 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4264 | miimon, arp_interval); | |
1da177e4 LT |
4265 | arp_interval = 0; |
4266 | } | |
4267 | ||
4268 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
4269 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4270 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
4271 | } |
4272 | ||
4273 | updelay /= miimon; | |
4274 | ||
4275 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
4276 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4277 | downdelay, miimon, | |
4278 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4279 | } |
4280 | ||
4281 | downdelay /= miimon; | |
4282 | } | |
4283 | ||
4284 | if (arp_interval < 0) { | |
91565ebb JP |
4285 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4286 | arp_interval, INT_MAX); | |
7bdb04ed | 4287 | arp_interval = 0; |
1da177e4 LT |
4288 | } |
4289 | ||
5a5c5fd4 | 4290 | for (arp_ip_count = 0, i = 0; |
4291 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
89015c18 | 4292 | __be32 ip; |
547942ca NA |
4293 | |
4294 | /* not a complete check, but good enough to catch mistakes */ | |
89015c18 | 4295 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || |
2807a9fe | 4296 | !bond_is_ip_target_ok(ip)) { |
91565ebb JP |
4297 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4298 | arp_ip_target[i]); | |
1da177e4 LT |
4299 | arp_interval = 0; |
4300 | } else { | |
0afee4e8 VF |
4301 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4302 | arp_target[arp_ip_count++] = ip; | |
4303 | else | |
91565ebb JP |
4304 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
4305 | &ip); | |
1da177e4 LT |
4306 | } |
4307 | } | |
4308 | ||
4309 | if (arp_interval && !arp_ip_count) { | |
4310 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
4311 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4312 | arp_interval); | |
1da177e4 LT |
4313 | arp_interval = 0; |
4314 | } | |
4315 | ||
f5b2b966 | 4316 | if (arp_validate) { |
f5b2b966 | 4317 | if (!arp_interval) { |
a4aee5c8 | 4318 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4319 | return -EINVAL; |
4320 | } | |
4321 | ||
16228881 NA |
4322 | bond_opt_initstr(&newval, arp_validate); |
4323 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
4324 | &newval); | |
4325 | if (!valptr) { | |
a4aee5c8 | 4326 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 4327 | arp_validate); |
f5b2b966 JV |
4328 | return -EINVAL; |
4329 | } | |
16228881 NA |
4330 | arp_validate_value = valptr->value; |
4331 | } else { | |
f5b2b966 | 4332 | arp_validate_value = 0; |
16228881 | 4333 | } |
f5b2b966 | 4334 | |
8599b52e VF |
4335 | arp_all_targets_value = 0; |
4336 | if (arp_all_targets) { | |
edf36b24 NA |
4337 | bond_opt_initstr(&newval, arp_all_targets); |
4338 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
4339 | &newval); | |
4340 | if (!valptr) { | |
8599b52e VF |
4341 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
4342 | arp_all_targets); | |
4343 | arp_all_targets_value = 0; | |
edf36b24 NA |
4344 | } else { |
4345 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
4346 | } |
4347 | } | |
4348 | ||
1da177e4 | 4349 | if (miimon) { |
a4aee5c8 | 4350 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4351 | } else if (arp_interval) { |
16228881 NA |
4352 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
4353 | arp_validate_value); | |
a4aee5c8 | 4354 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 4355 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
4356 | |
4357 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 4358 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 4359 | |
90194264 | 4360 | pr_cont("\n"); |
1da177e4 | 4361 | |
b8a9787e | 4362 | } else if (max_bonds) { |
1da177e4 LT |
4363 | /* miimon and arp_interval not set, we need one so things |
4364 | * work as expected, see bonding.txt for details | |
4365 | */ | |
90194264 | 4366 | 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 |
4367 | } |
4368 | ||
ec0865a9 | 4369 | if (primary && !bond_mode_uses_primary(bond_mode)) { |
1da177e4 LT |
4370 | /* currently, using a primary only makes sense |
4371 | * in active backup, TLB or ALB modes | |
4372 | */ | |
91565ebb JP |
4373 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
4374 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4375 | primary = NULL; |
4376 | } | |
4377 | ||
a549952a | 4378 | if (primary && primary_reselect) { |
388d3a6d NA |
4379 | bond_opt_initstr(&newval, primary_reselect); |
4380 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
4381 | &newval); | |
4382 | if (!valptr) { | |
a4aee5c8 | 4383 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 4384 | primary_reselect); |
a549952a JP |
4385 | return -EINVAL; |
4386 | } | |
388d3a6d | 4387 | primary_reselect_value = valptr->value; |
a549952a JP |
4388 | } else { |
4389 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4390 | } | |
4391 | ||
3915c1e8 | 4392 | if (fail_over_mac) { |
1df6b6aa NA |
4393 | bond_opt_initstr(&newval, fail_over_mac); |
4394 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
4395 | &newval); | |
4396 | if (!valptr) { | |
a4aee5c8 | 4397 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 4398 | fail_over_mac); |
3915c1e8 JV |
4399 | return -EINVAL; |
4400 | } | |
1df6b6aa | 4401 | fail_over_mac_value = valptr->value; |
3915c1e8 | 4402 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 4403 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
4404 | } else { |
4405 | fail_over_mac_value = BOND_FOM_NONE; | |
4406 | } | |
dd957c57 | 4407 | |
3a7129e5 | 4408 | if (lp_interval == 0) { |
91565ebb JP |
4409 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
4410 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 4411 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
4412 | } | |
4413 | ||
1da177e4 LT |
4414 | /* fill params struct with the proper values */ |
4415 | params->mode = bond_mode; | |
169a3e66 | 4416 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4417 | params->miimon = miimon; |
ad246c99 | 4418 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4419 | params->arp_interval = arp_interval; |
f5b2b966 | 4420 | params->arp_validate = arp_validate_value; |
8599b52e | 4421 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4422 | params->updelay = updelay; |
4423 | params->downdelay = downdelay; | |
4424 | params->use_carrier = use_carrier; | |
4425 | params->lacp_fast = lacp_fast; | |
4426 | params->primary[0] = 0; | |
a549952a | 4427 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4428 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4429 | params->tx_queues = tx_queues; |
ebd8e497 | 4430 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4431 | params->resend_igmp = resend_igmp; |
655f8919 | 4432 | params->min_links = min_links; |
3a7129e5 | 4433 | params->lp_interval = lp_interval; |
809fa972 | 4434 | params->packets_per_slave = packets_per_slave; |
e9f0fb88 | 4435 | params->tlb_dynamic_lb = 1; /* Default value */ |
809fa972 HFS |
4436 | if (packets_per_slave > 0) { |
4437 | params->reciprocal_packets_per_slave = | |
4438 | reciprocal_value(packets_per_slave); | |
4439 | } else { | |
4440 | /* reciprocal_packets_per_slave is unused if | |
4441 | * packets_per_slave is 0 or 1, just initialize it | |
4442 | */ | |
4443 | params->reciprocal_packets_per_slave = | |
4444 | (struct reciprocal_value) { 0 }; | |
4445 | } | |
4446 | ||
1da177e4 LT |
4447 | if (primary) { |
4448 | strncpy(params->primary, primary, IFNAMSIZ); | |
4449 | params->primary[IFNAMSIZ - 1] = 0; | |
4450 | } | |
4451 | ||
4452 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4453 | ||
4454 | return 0; | |
4455 | } | |
4456 | ||
0daa2303 | 4457 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4458 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4459 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4460 | |
e8a0464c DM |
4461 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4462 | struct netdev_queue *txq, | |
4463 | void *_unused) | |
c773e847 DM |
4464 | { |
4465 | lockdep_set_class(&txq->_xmit_lock, | |
4466 | &bonding_netdev_xmit_lock_key); | |
4467 | } | |
4468 | ||
4469 | static void bond_set_lockdep_class(struct net_device *dev) | |
4470 | { | |
cf508b12 DM |
4471 | lockdep_set_class(&dev->addr_list_lock, |
4472 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4473 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4474 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4475 | } |
4476 | ||
547942ca | 4477 | /* Called from registration process */ |
181470fc SH |
4478 | static int bond_init(struct net_device *bond_dev) |
4479 | { | |
4480 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4481 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc | 4482 | |
76444f50 | 4483 | netdev_dbg(bond_dev, "Begin bond_init\n"); |
181470fc SH |
4484 | |
4485 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4486 | if (!bond->wq) | |
4487 | return -ENOMEM; | |
4488 | ||
4489 | bond_set_lockdep_class(bond_dev); | |
4490 | ||
ec87fd3b | 4491 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4492 | |
6151b3d4 | 4493 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4494 | |
f073c7ca TI |
4495 | bond_debug_register(bond); |
4496 | ||
409cc1f8 JP |
4497 | /* Ensure valid dev_addr */ |
4498 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4499 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4500 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4501 | |
181470fc SH |
4502 | return 0; |
4503 | } | |
4504 | ||
0a2a78c4 | 4505 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4506 | { |
efacb309 | 4507 | return tx_queues; |
d5da4510 JP |
4508 | } |
4509 | ||
dfe60397 | 4510 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4511 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4512 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4513 | * set up our sysfs entries. | |
4514 | */ | |
ec87fd3b | 4515 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4516 | { |
4517 | struct net_device *bond_dev; | |
4518 | int res; | |
4519 | ||
4520 | rtnl_lock(); | |
027ea041 | 4521 | |
1c5cae81 | 4522 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
c835a677 | 4523 | name ? name : "bond%d", NET_NAME_UNKNOWN, |
1c5cae81 | 4524 | bond_setup, tx_queues); |
dfe60397 | 4525 | if (!bond_dev) { |
a4aee5c8 | 4526 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4527 | rtnl_unlock(); |
4528 | return -ENOMEM; | |
dfe60397 MW |
4529 | } |
4530 | ||
ec87fd3b | 4531 | dev_net_set(bond_dev, net); |
88ead977 EB |
4532 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4533 | ||
dfe60397 | 4534 | res = register_netdevice(bond_dev); |
0daa2303 | 4535 | |
e826eafa PO |
4536 | netif_carrier_off(bond_dev); |
4537 | ||
7e083840 | 4538 | rtnl_unlock(); |
9e2e61fb AW |
4539 | if (res < 0) |
4540 | bond_destructor(bond_dev); | |
30c15ba9 | 4541 | return res; |
dfe60397 MW |
4542 | } |
4543 | ||
2c8c1e72 | 4544 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4545 | { |
15449745 | 4546 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4547 | |
4548 | bn->net = net; | |
4549 | INIT_LIST_HEAD(&bn->dev_list); | |
4550 | ||
ec87fd3b | 4551 | bond_create_proc_dir(bn); |
4c22400a | 4552 | bond_create_sysfs(bn); |
87a7b84b | 4553 | |
15449745 | 4554 | return 0; |
ec87fd3b EB |
4555 | } |
4556 | ||
2c8c1e72 | 4557 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4558 | { |
15449745 | 4559 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4560 | struct bonding *bond, *tmp_bond; |
4561 | LIST_HEAD(list); | |
ec87fd3b | 4562 | |
4c22400a | 4563 | bond_destroy_sysfs(bn); |
69b0216a | 4564 | |
4565 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4566 | rtnl_lock(); | |
4567 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4568 | unregister_netdevice_queue(bond->dev, &list); | |
4569 | unregister_netdevice_many(&list); | |
4570 | rtnl_unlock(); | |
23fa5c2c VF |
4571 | |
4572 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
4573 | } |
4574 | ||
4575 | static struct pernet_operations bond_net_ops = { | |
4576 | .init = bond_net_init, | |
4577 | .exit = bond_net_exit, | |
15449745 EB |
4578 | .id = &bond_net_id, |
4579 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4580 | }; |
4581 | ||
1da177e4 LT |
4582 | static int __init bonding_init(void) |
4583 | { | |
1da177e4 LT |
4584 | int i; |
4585 | int res; | |
4586 | ||
bd33acc3 | 4587 | pr_info("%s", bond_version); |
1da177e4 | 4588 | |
dfe60397 | 4589 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4590 | if (res) |
dfe60397 | 4591 | goto out; |
1da177e4 | 4592 | |
15449745 | 4593 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4594 | if (res) |
4595 | goto out; | |
027ea041 | 4596 | |
0a2a78c4 | 4597 | res = bond_netlink_init(); |
88ead977 | 4598 | if (res) |
6639104b | 4599 | goto err_link; |
88ead977 | 4600 | |
f073c7ca TI |
4601 | bond_create_debugfs(); |
4602 | ||
1da177e4 | 4603 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4604 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4605 | if (res) |
4606 | goto err; | |
1da177e4 LT |
4607 | } |
4608 | ||
1da177e4 | 4609 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4610 | out: |
1da177e4 | 4611 | return res; |
88ead977 | 4612 | err: |
db298686 | 4613 | bond_destroy_debugfs(); |
0a2a78c4 | 4614 | bond_netlink_fini(); |
6639104b | 4615 | err_link: |
15449745 | 4616 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4617 | goto out; |
dfe60397 | 4618 | |
1da177e4 LT |
4619 | } |
4620 | ||
4621 | static void __exit bonding_exit(void) | |
4622 | { | |
4623 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4624 | ||
f073c7ca | 4625 | bond_destroy_debugfs(); |
ae68c398 | 4626 | |
0a2a78c4 | 4627 | bond_netlink_fini(); |
ffcdedb6 | 4628 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4629 | |
4630 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
547942ca | 4631 | /* Make sure we don't have an imbalance on our netpoll blocking */ |
fb4fa76a | 4632 | WARN_ON(atomic_read(&netpoll_block_tx)); |
e843fa50 | 4633 | #endif |
1da177e4 LT |
4634 | } |
4635 | ||
4636 | module_init(bonding_init); | |
4637 | module_exit(bonding_exit); | |
4638 | MODULE_LICENSE("GPL"); | |
4639 | MODULE_VERSION(DRV_VERSION); | |
4640 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4641 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |