]>
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> |
1da177e4 LT |
80 | #include "bonding.h" |
81 | #include "bond_3ad.h" | |
82 | #include "bond_alb.h" | |
83 | ||
84 | /*---------------------------- Module parameters ----------------------------*/ | |
85 | ||
86 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
1da177e4 LT |
87 | |
88 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 89 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 90 | static int num_peer_notif = 1; |
b98d9c66 | 91 | static int miimon; |
3d632c3f SH |
92 | static int updelay; |
93 | static int downdelay; | |
1da177e4 | 94 | static int use_carrier = 1; |
3d632c3f SH |
95 | static char *mode; |
96 | static char *primary; | |
a549952a | 97 | static char *primary_reselect; |
3d632c3f | 98 | static char *lacp_rate; |
655f8919 | 99 | static int min_links; |
3d632c3f SH |
100 | static char *ad_select; |
101 | static char *xmit_hash_policy; | |
7bdb04ed | 102 | static int arp_interval; |
3d632c3f SH |
103 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
104 | static char *arp_validate; | |
8599b52e | 105 | static char *arp_all_targets; |
3d632c3f | 106 | static char *fail_over_mac; |
b07ea07b | 107 | static int all_slaves_active; |
d2991f75 | 108 | static struct bond_params bonding_defaults; |
c2952c31 | 109 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
73958329 | 110 | static int packets_per_slave = 1; |
3a7129e5 | 111 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
112 | |
113 | module_param(max_bonds, int, 0); | |
114 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
115 | module_param(tx_queues, int, 0); |
116 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 117 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
118 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
119 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 120 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
121 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
122 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
123 | module_param(miimon, int, 0); |
124 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
125 | module_param(updelay, int, 0); | |
126 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
127 | module_param(downdelay, int, 0); | |
2ac47660 MW |
128 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
129 | "in milliseconds"); | |
1da177e4 | 130 | module_param(use_carrier, int, 0); |
2ac47660 MW |
131 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
132 | "0 for off, 1 for on (default)"); | |
1da177e4 | 133 | module_param(mode, charp, 0); |
90e62474 | 134 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
135 | "1 for active-backup, 2 for balance-xor, " |
136 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
137 | "6 for balance-alb"); | |
1da177e4 LT |
138 | module_param(primary, charp, 0); |
139 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
140 | module_param(primary_reselect, charp, 0); |
141 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
142 | "once it comes up; " | |
143 | "0 for always (default), " | |
144 | "1 for only if speed of primary is " | |
145 | "better, " | |
146 | "2 for only on active slave " | |
147 | "failure"); | |
1da177e4 | 148 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
149 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
150 | "0 for slow, 1 for fast"); | |
fd989c83 | 151 | module_param(ad_select, charp, 0); |
90e62474 AG |
152 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
153 | "0 for stable (default), 1 for bandwidth, " | |
154 | "2 for count"); | |
655f8919 | 155 | module_param(min_links, int, 0); |
156 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
157 | ||
169a3e66 | 158 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
159 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
160 | "0 for layer 2 (default), 1 for layer 3+4, " | |
32819dc1 NA |
161 | "2 for layer 2+3, 3 for encap layer 2+3, " |
162 | "4 for encap layer 3+4"); | |
1da177e4 LT |
163 | module_param(arp_interval, int, 0); |
164 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
165 | module_param_array(arp_ip_target, charp, NULL, 0); | |
166 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 167 | module_param(arp_validate, charp, 0); |
90e62474 AG |
168 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
169 | "0 for none (default), 1 for active, " | |
170 | "2 for backup, 3 for all"); | |
8599b52e VF |
171 | module_param(arp_all_targets, charp, 0); |
172 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 173 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
174 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
175 | "the same MAC; 0 for none (default), " | |
176 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
177 | module_param(all_slaves_active, int, 0); |
178 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 179 | "by setting active flag for all slaves; " |
ebd8e497 | 180 | "0 for never (default), 1 for always."); |
c2952c31 | 181 | module_param(resend_igmp, int, 0); |
90e62474 AG |
182 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
183 | "link failure"); | |
73958329 NA |
184 | module_param(packets_per_slave, int, 0); |
185 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
186 | "mode; 0 for a random slave, 1 packet per " | |
187 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 188 | module_param(lp_interval, uint, 0); |
189 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
190 | "the bonding driver sends learning packets to " | |
191 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
192 | |
193 | /*----------------------------- Global variables ----------------------------*/ | |
194 | ||
e843fa50 | 195 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 196 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
197 | #endif |
198 | ||
f99189b1 | 199 | int bond_net_id __read_mostly; |
1da177e4 | 200 | |
3d632c3f SH |
201 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
202 | static int arp_ip_count; | |
1da177e4 | 203 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
204 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
205 | static int lacp_fast; | |
1da177e4 | 206 | |
1da177e4 LT |
207 | /*-------------------------- Forward declarations ---------------------------*/ |
208 | ||
181470fc | 209 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 210 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
211 | |
212 | /*---------------------------- General routines -----------------------------*/ | |
213 | ||
bd33acc3 | 214 | const char *bond_mode_name(int mode) |
1da177e4 | 215 | { |
77afc92b HE |
216 | static const char *names[] = { |
217 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
218 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
219 | [BOND_MODE_XOR] = "load balancing (xor)", | |
220 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 221 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
222 | [BOND_MODE_TLB] = "transmit load balancing", |
223 | [BOND_MODE_ALB] = "adaptive load balancing", | |
224 | }; | |
225 | ||
f5280948 | 226 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 227 | return "unknown"; |
77afc92b HE |
228 | |
229 | return names[mode]; | |
1da177e4 LT |
230 | } |
231 | ||
232 | /*---------------------------------- VLAN -----------------------------------*/ | |
233 | ||
1da177e4 LT |
234 | /** |
235 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 236 | * |
1da177e4 LT |
237 | * @bond: bond device that got this skb for tx. |
238 | * @skb: hw accel VLAN tagged skb to transmit | |
239 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 240 | */ |
d316dedd | 241 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 242 | struct net_device *slave_dev) |
1da177e4 | 243 | { |
8387451e | 244 | skb->dev = slave_dev; |
374eeb5a | 245 | |
5ee31c68 | 246 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
247 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
248 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 249 | |
e15c3c22 | 250 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 251 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 252 | else |
f6dc31a8 | 253 | dev_queue_xmit(skb); |
1da177e4 LT |
254 | } |
255 | ||
256 | /* | |
cc0e4070 JP |
257 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
258 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
259 | * a. This operation is performed in IOCTL context, |
260 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
261 | * c. Holding a lock with BH disabled while directly calling a base driver | |
262 | * entry point is generally a BAD idea. | |
3d632c3f | 263 | * |
1da177e4 LT |
264 | * The design of synchronization/protection for this operation in the 8021q |
265 | * module is good for one or more VLAN devices over a single physical device | |
266 | * and cannot be extended for a teaming solution like bonding, so there is a | |
267 | * potential race condition here where a net device from the vlan group might | |
268 | * be referenced (either by a base driver or the 8021q code) while it is being | |
269 | * removed from the system. However, it turns out we're not making matters | |
270 | * worse, and if it works for regular VLAN usage it will work here too. | |
271 | */ | |
272 | ||
1da177e4 LT |
273 | /** |
274 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
275 | * @bond_dev: bonding net device that got called | |
276 | * @vid: vlan id being added | |
277 | */ | |
80d5c368 PM |
278 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
279 | __be16 proto, u16 vid) | |
1da177e4 | 280 | { |
454d7c9b | 281 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 282 | struct slave *slave, *rollback_slave; |
9caff1e7 | 283 | struct list_head *iter; |
dec1e90e | 284 | int res; |
1da177e4 | 285 | |
9caff1e7 | 286 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 287 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
288 | if (res) |
289 | goto unwind; | |
1da177e4 LT |
290 | } |
291 | ||
8e586137 | 292 | return 0; |
87002b03 JP |
293 | |
294 | unwind: | |
81f23b13 | 295 | /* unwind to the slave that failed */ |
9caff1e7 | 296 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
297 | if (rollback_slave == slave) |
298 | break; | |
299 | ||
300 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
301 | } | |
87002b03 JP |
302 | |
303 | return res; | |
1da177e4 LT |
304 | } |
305 | ||
306 | /** | |
307 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
308 | * @bond_dev: bonding net device that got called | |
309 | * @vid: vlan id being removed | |
310 | */ | |
80d5c368 PM |
311 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
312 | __be16 proto, u16 vid) | |
1da177e4 | 313 | { |
454d7c9b | 314 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 315 | struct list_head *iter; |
1da177e4 | 316 | struct slave *slave; |
1da177e4 | 317 | |
9caff1e7 | 318 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 319 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 320 | |
e868b0c9 VF |
321 | if (bond_is_lb(bond)) |
322 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
323 | |
324 | return 0; | |
1da177e4 LT |
325 | } |
326 | ||
1da177e4 LT |
327 | /*------------------------------- Link status -------------------------------*/ |
328 | ||
ff59c456 JV |
329 | /* |
330 | * Set the carrier state for the master according to the state of its | |
331 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
332 | * do special 802.3ad magic. | |
333 | * | |
334 | * Returns zero if carrier state does not change, nonzero if it does. | |
335 | */ | |
336 | static int bond_set_carrier(struct bonding *bond) | |
337 | { | |
9caff1e7 | 338 | struct list_head *iter; |
ff59c456 | 339 | struct slave *slave; |
ff59c456 | 340 | |
0965a1f3 | 341 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
342 | goto down; |
343 | ||
01844098 | 344 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
ff59c456 JV |
345 | return bond_3ad_set_carrier(bond); |
346 | ||
9caff1e7 | 347 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
348 | if (slave->link == BOND_LINK_UP) { |
349 | if (!netif_carrier_ok(bond->dev)) { | |
350 | netif_carrier_on(bond->dev); | |
351 | return 1; | |
352 | } | |
353 | return 0; | |
354 | } | |
355 | } | |
356 | ||
357 | down: | |
358 | if (netif_carrier_ok(bond->dev)) { | |
359 | netif_carrier_off(bond->dev); | |
360 | return 1; | |
361 | } | |
362 | return 0; | |
363 | } | |
364 | ||
1da177e4 LT |
365 | /* |
366 | * Get link speed and duplex from the slave's base driver | |
367 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 368 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 369 | * and return. |
1da177e4 | 370 | */ |
13b95fb7 | 371 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
372 | { |
373 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 374 | struct ethtool_cmd ecmd; |
5d30530e | 375 | u32 slave_speed; |
61a44b9c | 376 | int res; |
1da177e4 | 377 | |
589665f5 DC |
378 | slave->speed = SPEED_UNKNOWN; |
379 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 380 | |
4bc71cb9 | 381 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 382 | if (res < 0) |
13b95fb7 | 383 | return; |
1da177e4 | 384 | |
4bc71cb9 | 385 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 386 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 387 | return; |
1da177e4 | 388 | |
4bc71cb9 | 389 | switch (ecmd.duplex) { |
1da177e4 LT |
390 | case DUPLEX_FULL: |
391 | case DUPLEX_HALF: | |
392 | break; | |
393 | default: | |
13b95fb7 | 394 | return; |
1da177e4 LT |
395 | } |
396 | ||
5d30530e | 397 | slave->speed = slave_speed; |
4bc71cb9 | 398 | slave->duplex = ecmd.duplex; |
1da177e4 | 399 | |
13b95fb7 | 400 | return; |
1da177e4 LT |
401 | } |
402 | ||
07699f9a | 403 | const char *bond_slave_link_status(s8 link) |
404 | { | |
405 | switch (link) { | |
406 | case BOND_LINK_UP: | |
407 | return "up"; | |
408 | case BOND_LINK_FAIL: | |
409 | return "going down"; | |
410 | case BOND_LINK_DOWN: | |
411 | return "down"; | |
412 | case BOND_LINK_BACK: | |
413 | return "going back"; | |
414 | default: | |
415 | return "unknown"; | |
416 | } | |
417 | } | |
418 | ||
1da177e4 LT |
419 | /* |
420 | * if <dev> supports MII link status reporting, check its link status. | |
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 LT |
456 | if (ioctl) { |
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 | |
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 | ||
1da177e4 LT |
479 | /* |
480 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 481 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
482 | * cannot report link status). If not reporting, pretend |
483 | * we're ok. | |
484 | */ | |
3d632c3f | 485 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
486 | } |
487 | ||
488 | /*----------------------------- Multicast list ------------------------------*/ | |
489 | ||
1da177e4 LT |
490 | /* |
491 | * Push the promiscuity flag down to appropriate slaves | |
492 | */ | |
7e1a1ac1 | 493 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 494 | { |
9caff1e7 | 495 | struct list_head *iter; |
7e1a1ac1 | 496 | int err = 0; |
9caff1e7 | 497 | |
ec0865a9 | 498 | if (bond_uses_primary(bond)) { |
14056e79 | 499 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 500 | |
4740d638 ED |
501 | if (curr_active) |
502 | err = dev_set_promiscuity(curr_active->dev, inc); | |
1da177e4 LT |
503 | } else { |
504 | struct slave *slave; | |
dec1e90e | 505 | |
9caff1e7 | 506 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
507 | err = dev_set_promiscuity(slave->dev, inc); |
508 | if (err) | |
509 | return err; | |
1da177e4 LT |
510 | } |
511 | } | |
7e1a1ac1 | 512 | return err; |
1da177e4 LT |
513 | } |
514 | ||
515 | /* | |
516 | * Push the allmulti flag down to all slaves | |
517 | */ | |
7e1a1ac1 | 518 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 519 | { |
9caff1e7 | 520 | struct list_head *iter; |
7e1a1ac1 | 521 | int err = 0; |
9caff1e7 | 522 | |
ec0865a9 | 523 | if (bond_uses_primary(bond)) { |
14056e79 | 524 | struct slave *curr_active = rtnl_dereference(bond->curr_active_slave); |
4740d638 | 525 | |
4740d638 ED |
526 | if (curr_active) |
527 | err = dev_set_allmulti(curr_active->dev, inc); | |
1da177e4 LT |
528 | } else { |
529 | struct slave *slave; | |
dec1e90e | 530 | |
9caff1e7 | 531 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
532 | err = dev_set_allmulti(slave->dev, inc); |
533 | if (err) | |
534 | return err; | |
1da177e4 LT |
535 | } |
536 | } | |
7e1a1ac1 | 537 | return err; |
1da177e4 LT |
538 | } |
539 | ||
5a37e8ca FL |
540 | /* |
541 | * Retrieve the list of registered multicast addresses for the bonding | |
542 | * device and retransmit an IGMP JOIN request to the current active | |
543 | * slave. | |
544 | */ | |
f2369109 | 545 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 546 | { |
f2369109 | 547 | struct bonding *bond = container_of(work, struct bonding, |
548 | mcast_work.work); | |
549 | ||
4aa5dee4 | 550 | if (!rtnl_trylock()) { |
4beac029 | 551 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 552 | return; |
5a37e8ca | 553 | } |
4aa5dee4 | 554 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 555 | |
4f5474e7 NA |
556 | if (bond->igmp_retrans > 1) { |
557 | bond->igmp_retrans--; | |
c2952c31 | 558 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 559 | } |
f2369109 | 560 | rtnl_unlock(); |
5a37e8ca FL |
561 | } |
562 | ||
303d1cbf | 563 | /* Flush bond's hardware addresses from slave |
1da177e4 | 564 | */ |
303d1cbf | 565 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 566 | struct net_device *slave_dev) |
1da177e4 | 567 | { |
454d7c9b | 568 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 569 | |
303d1cbf JV |
570 | dev_uc_unsync(slave_dev, bond_dev); |
571 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 | 572 | |
01844098 | 573 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
574 | /* del lacpdu mc addr from mc list */ |
575 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
576 | ||
22bedad3 | 577 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
578 | } |
579 | } | |
580 | ||
581 | /*--------------------------- Active slave change ---------------------------*/ | |
582 | ||
303d1cbf | 583 | /* Update the hardware address list and promisc/allmulti for the new and |
ec0865a9 VF |
584 | * old active slaves (if any). Modes that are not using primary keep all |
585 | * slaves up date at all times; only the modes that use primary need to call | |
303d1cbf | 586 | * this function to swap these settings during a failover. |
1da177e4 | 587 | */ |
303d1cbf JV |
588 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
589 | struct slave *old_active) | |
1da177e4 | 590 | { |
7864a1ad VF |
591 | ASSERT_RTNL(); |
592 | ||
1da177e4 | 593 | if (old_active) { |
3d632c3f | 594 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 595 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 596 | |
3d632c3f | 597 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 598 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 599 | |
303d1cbf | 600 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
601 | } |
602 | ||
603 | if (new_active) { | |
7e1a1ac1 | 604 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 605 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 606 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 607 | |
3d632c3f | 608 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 609 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 610 | |
d632ce98 | 611 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
612 | dev_uc_sync(new_active->dev, bond->dev); |
613 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 614 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
615 | } |
616 | } | |
617 | ||
ae0d6750 | 618 | /** |
619 | * bond_set_dev_addr - clone slave's address to bond | |
620 | * @bond_dev: bond net device | |
621 | * @slave_dev: slave net device | |
622 | * | |
623 | * Should be called with RTNL held. | |
624 | */ | |
625 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
626 | struct net_device *slave_dev) | |
627 | { | |
76444f50 VF |
628 | netdev_dbg(bond_dev, "bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
629 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 630 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
631 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
632 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
633 | } | |
634 | ||
3915c1e8 JV |
635 | /* |
636 | * bond_do_fail_over_mac | |
637 | * | |
638 | * Perform special MAC address swapping for fail_over_mac settings | |
639 | * | |
1c72cfdc | 640 | * Called with RTNL |
3915c1e8 JV |
641 | */ |
642 | static void bond_do_fail_over_mac(struct bonding *bond, | |
643 | struct slave *new_active, | |
644 | struct slave *old_active) | |
645 | { | |
646 | u8 tmp_mac[ETH_ALEN]; | |
647 | struct sockaddr saddr; | |
648 | int rv; | |
649 | ||
650 | switch (bond->params.fail_over_mac) { | |
651 | case BOND_FOM_ACTIVE: | |
1c72cfdc | 652 | if (new_active) |
ae0d6750 | 653 | bond_set_dev_addr(bond->dev, new_active->dev); |
3915c1e8 JV |
654 | break; |
655 | case BOND_FOM_FOLLOW: | |
656 | /* | |
657 | * if new_active && old_active, swap them | |
658 | * if just old_active, do nothing (going to no active slave) | |
659 | * if just new_active, set new_active to bond's MAC | |
660 | */ | |
661 | if (!new_active) | |
662 | return; | |
663 | ||
3915c1e8 | 664 | if (old_active) { |
ada0f863 | 665 | ether_addr_copy(tmp_mac, new_active->dev->dev_addr); |
2a7c183b JP |
666 | ether_addr_copy(saddr.sa_data, |
667 | old_active->dev->dev_addr); | |
3915c1e8 JV |
668 | saddr.sa_family = new_active->dev->type; |
669 | } else { | |
ada0f863 | 670 | ether_addr_copy(saddr.sa_data, bond->dev->dev_addr); |
3915c1e8 JV |
671 | saddr.sa_family = bond->dev->type; |
672 | } | |
673 | ||
674 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
675 | if (rv) { | |
76444f50 VF |
676 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
677 | -rv, new_active->dev->name); | |
3915c1e8 JV |
678 | goto out; |
679 | } | |
680 | ||
681 | if (!old_active) | |
682 | goto out; | |
683 | ||
ada0f863 | 684 | ether_addr_copy(saddr.sa_data, tmp_mac); |
3915c1e8 JV |
685 | saddr.sa_family = old_active->dev->type; |
686 | ||
687 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
688 | if (rv) | |
76444f50 VF |
689 | netdev_err(bond->dev, "Error %d setting MAC of slave %s\n", |
690 | -rv, new_active->dev->name); | |
3915c1e8 | 691 | out: |
3915c1e8 JV |
692 | break; |
693 | default: | |
76444f50 VF |
694 | netdev_err(bond->dev, "bond_do_fail_over_mac impossible: bad policy %d\n", |
695 | bond->params.fail_over_mac); | |
3915c1e8 JV |
696 | break; |
697 | } | |
698 | ||
699 | } | |
700 | ||
a549952a JP |
701 | static bool bond_should_change_active(struct bonding *bond) |
702 | { | |
059b47e8 | 703 | struct slave *prim = rtnl_dereference(bond->primary_slave); |
1c72cfdc | 704 | struct slave *curr = rtnl_dereference(bond->curr_active_slave); |
a549952a JP |
705 | |
706 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
707 | return true; | |
708 | if (bond->force_primary) { | |
709 | bond->force_primary = false; | |
710 | return true; | |
711 | } | |
712 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
713 | (prim->speed < curr->speed || | |
714 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
715 | return false; | |
716 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
717 | return false; | |
718 | return true; | |
719 | } | |
3915c1e8 | 720 | |
1da177e4 LT |
721 | /** |
722 | * find_best_interface - select the best available slave to be the active one | |
723 | * @bond: our bonding struct | |
1da177e4 LT |
724 | */ |
725 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
726 | { | |
059b47e8 | 727 | struct slave *slave, *bestslave = NULL, *primary; |
77140d29 | 728 | struct list_head *iter; |
1da177e4 | 729 | int mintime = bond->params.updelay; |
1da177e4 | 730 | |
059b47e8 NA |
731 | primary = rtnl_dereference(bond->primary_slave); |
732 | if (primary && primary->link == BOND_LINK_UP && | |
77140d29 | 733 | bond_should_change_active(bond)) |
059b47e8 | 734 | return primary; |
1da177e4 | 735 | |
77140d29 VF |
736 | bond_for_each_slave(bond, slave, iter) { |
737 | if (slave->link == BOND_LINK_UP) | |
738 | return slave; | |
b6adc610 | 739 | if (slave->link == BOND_LINK_BACK && bond_slave_is_up(slave) && |
77140d29 VF |
740 | slave->delay < mintime) { |
741 | mintime = slave->delay; | |
742 | bestslave = slave; | |
1da177e4 LT |
743 | } |
744 | } | |
745 | ||
746 | return bestslave; | |
747 | } | |
748 | ||
ad246c99 BH |
749 | static bool bond_should_notify_peers(struct bonding *bond) |
750 | { | |
4cb4f97b | 751 | struct slave *slave; |
752 | ||
753 | rcu_read_lock(); | |
754 | slave = rcu_dereference(bond->curr_active_slave); | |
755 | rcu_read_unlock(); | |
ad246c99 | 756 | |
76444f50 VF |
757 | netdev_dbg(bond->dev, "bond_should_notify_peers: slave %s\n", |
758 | slave ? slave->dev->name : "NULL"); | |
ad246c99 BH |
759 | |
760 | if (!slave || !bond->send_peer_notif || | |
761 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
762 | return false; | |
763 | ||
ad246c99 BH |
764 | return true; |
765 | } | |
766 | ||
1da177e4 LT |
767 | /** |
768 | * change_active_interface - change the active slave into the specified one | |
769 | * @bond: our bonding struct | |
770 | * @new: the new slave to make the active one | |
771 | * | |
772 | * Set the new slave to the bond's settings and unset them on the old | |
773 | * curr_active_slave. | |
774 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
775 | * | |
776 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
777 | * because it is apparently the best available slave we have, even though its | |
778 | * updelay hasn't timed out yet. | |
779 | * | |
1c72cfdc | 780 | * Caller must hold RTNL. |
1da177e4 | 781 | */ |
a77b5325 | 782 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 | 783 | { |
4740d638 ED |
784 | struct slave *old_active; |
785 | ||
1c72cfdc NA |
786 | ASSERT_RTNL(); |
787 | ||
788 | old_active = rtnl_dereference(bond->curr_active_slave); | |
1da177e4 | 789 | |
3d632c3f | 790 | if (old_active == new_active) |
1da177e4 | 791 | return; |
1da177e4 LT |
792 | |
793 | if (new_active) { | |
8e603460 | 794 | new_active->last_link_up = jiffies; |
b2220cad | 795 | |
1da177e4 | 796 | if (new_active->link == BOND_LINK_BACK) { |
ec0865a9 | 797 | if (bond_uses_primary(bond)) { |
76444f50 VF |
798 | netdev_info(bond->dev, "making interface %s the new active one %d ms earlier\n", |
799 | new_active->dev->name, | |
800 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
801 | } |
802 | ||
803 | new_active->delay = 0; | |
804 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 805 | |
01844098 | 806 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 807 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 808 | |
58402054 | 809 | if (bond_is_lb(bond)) |
1da177e4 | 810 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 | 811 | } else { |
ec0865a9 | 812 | if (bond_uses_primary(bond)) { |
76444f50 VF |
813 | netdev_info(bond->dev, "making interface %s the new active one\n", |
814 | new_active->dev->name); | |
1da177e4 LT |
815 | } |
816 | } | |
817 | } | |
818 | ||
ec0865a9 | 819 | if (bond_uses_primary(bond)) |
303d1cbf | 820 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 821 | |
58402054 | 822 | if (bond_is_lb(bond)) { |
1da177e4 | 823 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 824 | if (old_active) |
5e5b0665 | 825 | bond_set_slave_inactive_flags(old_active, |
826 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 827 | if (new_active) |
5e5b0665 | 828 | bond_set_slave_active_flags(new_active, |
829 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 830 | } else { |
278b2083 | 831 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 832 | } |
c3ade5ca | 833 | |
01844098 | 834 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) { |
3d632c3f | 835 | if (old_active) |
5e5b0665 | 836 | bond_set_slave_inactive_flags(old_active, |
837 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
838 | |
839 | if (new_active) { | |
ad246c99 BH |
840 | bool should_notify_peers = false; |
841 | ||
5e5b0665 | 842 | bond_set_slave_active_flags(new_active, |
843 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 844 | |
709f8a45 OG |
845 | if (bond->params.fail_over_mac) |
846 | bond_do_fail_over_mac(bond, new_active, | |
847 | old_active); | |
3915c1e8 | 848 | |
ad246c99 BH |
849 | if (netif_running(bond->dev)) { |
850 | bond->send_peer_notif = | |
851 | bond->params.num_peer_notif; | |
852 | should_notify_peers = | |
853 | bond_should_notify_peers(bond); | |
854 | } | |
855 | ||
b7bc2a5b | 856 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 857 | if (should_notify_peers) |
b7bc2a5b AW |
858 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
859 | bond->dev); | |
7893b249 | 860 | } |
c3ade5ca | 861 | } |
a2fd940f | 862 | |
5a37e8ca | 863 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
864 | * all were sent on curr_active_slave. |
865 | * resend only if bond is brought up with the affected | |
866 | * bonding modes and the retransmission is enabled */ | |
867 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
ec0865a9 | 868 | ((bond_uses_primary(bond) && new_active) || |
01844098 | 869 | BOND_MODE(bond) == BOND_MODE_ROUNDROBIN)) { |
c2952c31 | 870 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 871 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 872 | } |
1da177e4 LT |
873 | } |
874 | ||
875 | /** | |
876 | * bond_select_active_slave - select a new active slave, if needed | |
877 | * @bond: our bonding struct | |
878 | * | |
3d632c3f | 879 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
880 | * - The old curr_active_slave has been released or lost its link. |
881 | * - The primary_slave has got its link back. | |
882 | * - A slave has got its link back and there's no old curr_active_slave. | |
883 | * | |
1c72cfdc | 884 | * Caller must hold RTNL. |
1da177e4 | 885 | */ |
a77b5325 | 886 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
887 | { |
888 | struct slave *best_slave; | |
ff59c456 | 889 | int rv; |
1da177e4 LT |
890 | |
891 | best_slave = bond_find_best_slave(bond); | |
1c72cfdc | 892 | if (best_slave != rtnl_dereference(bond->curr_active_slave)) { |
1da177e4 | 893 | bond_change_active_slave(bond, best_slave); |
ff59c456 JV |
894 | rv = bond_set_carrier(bond); |
895 | if (!rv) | |
896 | return; | |
897 | ||
898 | if (netif_carrier_ok(bond->dev)) { | |
76444f50 | 899 | netdev_info(bond->dev, "first active interface up!\n"); |
ff59c456 | 900 | } else { |
76444f50 | 901 | netdev_info(bond->dev, "now running without any active interface!\n"); |
ff59c456 | 902 | } |
1da177e4 LT |
903 | } |
904 | } | |
905 | ||
f6dc31a8 | 906 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 907 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 908 | { |
8a8efa22 AW |
909 | struct netpoll *np; |
910 | int err = 0; | |
f6dc31a8 | 911 | |
a8779ec1 | 912 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
913 | err = -ENOMEM; |
914 | if (!np) | |
915 | goto out; | |
916 | ||
a8779ec1 | 917 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
918 | if (err) { |
919 | kfree(np); | |
920 | goto out; | |
f6dc31a8 | 921 | } |
8a8efa22 AW |
922 | slave->np = np; |
923 | out: | |
924 | return err; | |
925 | } | |
926 | static inline void slave_disable_netpoll(struct slave *slave) | |
927 | { | |
928 | struct netpoll *np = slave->np; | |
929 | ||
930 | if (!np) | |
931 | return; | |
932 | ||
933 | slave->np = NULL; | |
2cde6acd | 934 | __netpoll_free_async(np); |
8a8efa22 | 935 | } |
f6dc31a8 WC |
936 | |
937 | static void bond_poll_controller(struct net_device *bond_dev) | |
938 | { | |
8a8efa22 AW |
939 | } |
940 | ||
c4cdef9b | 941 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 942 | { |
c4cdef9b | 943 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 944 | struct list_head *iter; |
c2355e1a | 945 | struct slave *slave; |
c2355e1a | 946 | |
9caff1e7 | 947 | bond_for_each_slave(bond, slave, iter) |
b6adc610 | 948 | if (bond_slave_is_up(slave)) |
8a8efa22 | 949 | slave_disable_netpoll(slave); |
f6dc31a8 | 950 | } |
8a8efa22 | 951 | |
a8779ec1 | 952 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
953 | { |
954 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 955 | struct list_head *iter; |
f6dc31a8 | 956 | struct slave *slave; |
dec1e90e | 957 | int err = 0; |
f6dc31a8 | 958 | |
9caff1e7 | 959 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
960 | err = slave_enable_netpoll(slave); |
961 | if (err) { | |
c4cdef9b | 962 | bond_netpoll_cleanup(dev); |
8a8efa22 | 963 | break; |
f6dc31a8 WC |
964 | } |
965 | } | |
8a8efa22 | 966 | return err; |
f6dc31a8 | 967 | } |
8a8efa22 AW |
968 | #else |
969 | static inline int slave_enable_netpoll(struct slave *slave) | |
970 | { | |
971 | return 0; | |
972 | } | |
973 | static inline void slave_disable_netpoll(struct slave *slave) | |
974 | { | |
975 | } | |
f6dc31a8 WC |
976 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
977 | { | |
978 | } | |
f6dc31a8 WC |
979 | #endif |
980 | ||
1da177e4 LT |
981 | /*---------------------------------- IOCTL ----------------------------------*/ |
982 | ||
c8f44aff | 983 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 984 | netdev_features_t features) |
8531c5ff | 985 | { |
b2a103e6 | 986 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 987 | struct list_head *iter; |
c8f44aff | 988 | netdev_features_t mask; |
9b7b165a | 989 | struct slave *slave; |
b63365a2 | 990 | |
b2a103e6 | 991 | mask = features; |
b63365a2 | 992 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 993 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 994 | |
9caff1e7 | 995 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
996 | features = netdev_increment_features(features, |
997 | slave->dev->features, | |
b2a103e6 MM |
998 | mask); |
999 | } | |
b0ce3508 | 1000 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1001 | |
b2a103e6 MM |
1002 | return features; |
1003 | } | |
1004 | ||
62f2a3a4 MM |
1005 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1006 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1007 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 | 1008 | |
5a7baa78 OG |
1009 | #define BOND_ENC_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | NETIF_F_RXCSUM |\ |
1010 | NETIF_F_TSO | NETIF_F_GSO_UDP_TUNNEL) | |
1011 | ||
b2a103e6 MM |
1012 | static void bond_compute_features(struct bonding *bond) |
1013 | { | |
c4826861 | 1014 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
c8f44aff | 1015 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
5a7baa78 | 1016 | netdev_features_t enc_features = BOND_ENC_FEATURES; |
9caff1e7 VF |
1017 | struct net_device *bond_dev = bond->dev; |
1018 | struct list_head *iter; | |
1019 | struct slave *slave; | |
b2a103e6 | 1020 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1021 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1022 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1023 | |
0965a1f3 | 1024 | if (!bond_has_slaves(bond)) |
b2a103e6 | 1025 | goto done; |
a9b3ace4 | 1026 | vlan_features &= NETIF_F_ALL_FOR_ALL; |
b2a103e6 | 1027 | |
9caff1e7 | 1028 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1029 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1030 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1031 | ||
5a7baa78 OG |
1032 | enc_features = netdev_increment_features(enc_features, |
1033 | slave->dev->hw_enc_features, | |
1034 | BOND_ENC_FEATURES); | |
b6fe83e9 | 1035 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1036 | if (slave->dev->hard_header_len > max_hard_header_len) |
1037 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1038 | |
1039 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1040 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1041 | } |
8531c5ff | 1042 | |
b63365a2 | 1043 | done: |
b2a103e6 | 1044 | bond_dev->vlan_features = vlan_features; |
5a7baa78 | 1045 | bond_dev->hw_enc_features = enc_features; |
54ef3137 | 1046 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1047 | bond_dev->gso_max_segs = gso_max_segs; |
1048 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1049 | |
b6fe83e9 ED |
1050 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1051 | bond_dev->priv_flags = flags | dst_release_flag; | |
1052 | ||
b2a103e6 | 1053 | netdev_change_features(bond_dev); |
8531c5ff AK |
1054 | } |
1055 | ||
872254dd MS |
1056 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1057 | struct net_device *slave_dev) | |
1058 | { | |
00829823 | 1059 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1060 | |
1061 | bond_dev->type = slave_dev->type; | |
1062 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1063 | bond_dev->addr_len = slave_dev->addr_len; | |
1064 | ||
1065 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1066 | slave_dev->addr_len); | |
1067 | } | |
1068 | ||
5b2c4dd2 | 1069 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1070 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1071 | */ |
1072 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1073 | struct slave *slave, |
1074 | struct bonding *bond) | |
5b2c4dd2 | 1075 | { |
2d7011ca | 1076 | if (bond_is_slave_inactive(slave)) { |
01844098 | 1077 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
5b2c4dd2 JP |
1078 | skb->pkt_type != PACKET_BROADCAST && |
1079 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1080 | return false; |
5b2c4dd2 JP |
1081 | return true; |
1082 | } | |
1083 | return false; | |
1084 | } | |
1085 | ||
8a4eb573 | 1086 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1087 | { |
8a4eb573 | 1088 | struct sk_buff *skb = *pskb; |
f1c1775a | 1089 | struct slave *slave; |
0bd80dad | 1090 | struct bonding *bond; |
de063b70 ED |
1091 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1092 | struct slave *); | |
13a8e0c8 | 1093 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1094 | |
8a4eb573 JP |
1095 | skb = skb_share_check(skb, GFP_ATOMIC); |
1096 | if (unlikely(!skb)) | |
1097 | return RX_HANDLER_CONSUMED; | |
1098 | ||
1099 | *pskb = skb; | |
5b2c4dd2 | 1100 | |
35d48903 JP |
1101 | slave = bond_slave_get_rcu(skb->dev); |
1102 | bond = slave->bond; | |
0bd80dad | 1103 | |
4d97480b MH |
1104 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1105 | if (recv_probe) { | |
de063b70 ED |
1106 | ret = recv_probe(skb, bond, slave); |
1107 | if (ret == RX_HANDLER_CONSUMED) { | |
1108 | consume_skb(skb); | |
1109 | return ret; | |
3aba891d JP |
1110 | } |
1111 | } | |
1112 | ||
0bd80dad | 1113 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1114 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1115 | } |
1116 | ||
35d48903 | 1117 | skb->dev = bond->dev; |
5b2c4dd2 | 1118 | |
01844098 | 1119 | if (BOND_MODE(bond) == BOND_MODE_ALB && |
35d48903 | 1120 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1121 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1122 | |
541ac7c9 CG |
1123 | if (unlikely(skb_cow_head(skb, |
1124 | skb->data - skb_mac_header(skb)))) { | |
1125 | kfree_skb(skb); | |
8a4eb573 | 1126 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1127 | } |
2a7c183b | 1128 | ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr); |
5b2c4dd2 JP |
1129 | } |
1130 | ||
13a8e0c8 | 1131 | return ret; |
5b2c4dd2 JP |
1132 | } |
1133 | ||
471cb5a3 | 1134 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1f718f0f VF |
1135 | struct net_device *slave_dev, |
1136 | struct slave *slave) | |
471cb5a3 JP |
1137 | { |
1138 | int err; | |
1139 | ||
1f718f0f | 1140 | err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave); |
471cb5a3 JP |
1141 | if (err) |
1142 | return err; | |
1143 | slave_dev->flags |= IFF_SLAVE; | |
7f294054 | 1144 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1145 | return 0; |
1146 | } | |
1147 | ||
1148 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1149 | struct net_device *slave_dev) | |
1150 | { | |
1151 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1152 | slave_dev->flags &= ~IFF_SLAVE; | |
7f294054 | 1153 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1154 | } |
1155 | ||
3fdddd85 | 1156 | static struct slave *bond_alloc_slave(struct bonding *bond) |
1157 | { | |
1158 | struct slave *slave = NULL; | |
1159 | ||
1160 | slave = kzalloc(sizeof(struct slave), GFP_KERNEL); | |
1161 | if (!slave) | |
1162 | return NULL; | |
1163 | ||
01844098 | 1164 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
3fdddd85 | 1165 | SLAVE_AD_INFO(slave) = kzalloc(sizeof(struct ad_slave_info), |
1166 | GFP_KERNEL); | |
1167 | if (!SLAVE_AD_INFO(slave)) { | |
1168 | kfree(slave); | |
1169 | return NULL; | |
1170 | } | |
1171 | } | |
1172 | return slave; | |
1173 | } | |
1174 | ||
1175 | static void bond_free_slave(struct slave *slave) | |
1176 | { | |
1177 | struct bonding *bond = bond_get_bond_by_slave(slave); | |
1178 | ||
01844098 | 1179 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
3fdddd85 | 1180 | kfree(SLAVE_AD_INFO(slave)); |
1181 | ||
1182 | kfree(slave); | |
1183 | } | |
1184 | ||
1da177e4 | 1185 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1186 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1187 | { |
454d7c9b | 1188 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1189 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1190 | struct slave *new_slave = NULL, *prev_slave; |
1da177e4 LT |
1191 | struct sockaddr addr; |
1192 | int link_reporting; | |
8599b52e | 1193 | int res = 0, i; |
1da177e4 | 1194 | |
c772dde3 BH |
1195 | if (!bond->params.use_carrier && |
1196 | slave_dev->ethtool_ops->get_link == NULL && | |
1197 | slave_ops->ndo_do_ioctl == NULL) { | |
76444f50 VF |
1198 | netdev_warn(bond_dev, "no link monitoring support for %s\n", |
1199 | slave_dev->name); | |
1da177e4 LT |
1200 | } |
1201 | ||
1da177e4 LT |
1202 | /* already enslaved */ |
1203 | if (slave_dev->flags & IFF_SLAVE) { | |
76444f50 | 1204 | netdev_dbg(bond_dev, "Error: Device was already enslaved\n"); |
1da177e4 LT |
1205 | return -EBUSY; |
1206 | } | |
1207 | ||
09a89c21 | 1208 | if (bond_dev == slave_dev) { |
76444f50 | 1209 | netdev_err(bond_dev, "cannot enslave bond to itself.\n"); |
09a89c21 JB |
1210 | return -EPERM; |
1211 | } | |
1212 | ||
1da177e4 LT |
1213 | /* vlan challenged mutual exclusion */ |
1214 | /* no need to lock since we're protected by rtnl_lock */ | |
1215 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
76444f50 VF |
1216 | netdev_dbg(bond_dev, "%s is NETIF_F_VLAN_CHALLENGED\n", |
1217 | slave_dev->name); | |
55462cf3 | 1218 | if (vlan_uses_dev(bond_dev)) { |
76444f50 VF |
1219 | netdev_err(bond_dev, "Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1220 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1221 | return -EPERM; |
1222 | } else { | |
76444f50 VF |
1223 | netdev_warn(bond_dev, "enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1224 | slave_dev->name, slave_dev->name, | |
1225 | bond_dev->name); | |
1da177e4 LT |
1226 | } |
1227 | } else { | |
76444f50 VF |
1228 | netdev_dbg(bond_dev, "%s is !NETIF_F_VLAN_CHALLENGED\n", |
1229 | slave_dev->name); | |
1da177e4 LT |
1230 | } |
1231 | ||
217df670 JV |
1232 | /* |
1233 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1234 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1235 | * the current ifenslave will set the interface down prior to |
1236 | * enslaving it; the old ifenslave will not. | |
1237 | */ | |
1238 | if ((slave_dev->flags & IFF_UP)) { | |
76444f50 VF |
1239 | netdev_err(bond_dev, "%s is up - this may be due to an out of date ifenslave\n", |
1240 | slave_dev->name); | |
217df670 JV |
1241 | res = -EPERM; |
1242 | goto err_undo_flags; | |
1243 | } | |
1da177e4 | 1244 | |
872254dd MS |
1245 | /* set bonding device ether type by slave - bonding netdevices are |
1246 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1247 | * there is a need to override some of the type dependent attribs/funcs. | |
1248 | * | |
1249 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1250 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1251 | */ | |
0965a1f3 | 1252 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1253 | if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1254 | netdev_dbg(bond_dev, "change device type from %d to %d\n", |
1255 | bond_dev->type, slave_dev->type); | |
75c78500 | 1256 | |
b7bc2a5b AW |
1257 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1258 | bond_dev); | |
3ca5b404 JP |
1259 | res = notifier_to_errno(res); |
1260 | if (res) { | |
76444f50 | 1261 | netdev_err(bond_dev, "refused to change device type\n"); |
3ca5b404 JP |
1262 | res = -EBUSY; |
1263 | goto err_undo_flags; | |
1264 | } | |
75c78500 | 1265 | |
32a806c1 | 1266 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1267 | dev_uc_flush(bond_dev); |
22bedad3 | 1268 | dev_mc_flush(bond_dev); |
32a806c1 | 1269 | |
e36b9d16 MS |
1270 | if (slave_dev->type != ARPHRD_ETHER) |
1271 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1272 | else { |
e36b9d16 | 1273 | ether_setup(bond_dev); |
550fd08c NH |
1274 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1275 | } | |
75c78500 | 1276 | |
b7bc2a5b AW |
1277 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1278 | bond_dev); | |
e36b9d16 | 1279 | } |
872254dd | 1280 | } else if (bond_dev->type != slave_dev->type) { |
76444f50 VF |
1281 | netdev_err(bond_dev, "%s ether type (%d) is different from other slaves (%d), can not enslave it\n", |
1282 | slave_dev->name, slave_dev->type, bond_dev->type); | |
a4aee5c8 JP |
1283 | res = -EINVAL; |
1284 | goto err_undo_flags; | |
872254dd MS |
1285 | } |
1286 | ||
eb7cc59a | 1287 | if (slave_ops->ndo_set_mac_address == NULL) { |
76444f50 | 1288 | netdev_warn(bond_dev, "The slave device specified does not support setting the MAC address\n"); |
f5442441 VF |
1289 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP && |
1290 | bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
1291 | if (!bond_has_slaves(bond)) { | |
00503b6f | 1292 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
76444f50 | 1293 | netdev_warn(bond_dev, "Setting fail_over_mac to active for active-backup mode\n"); |
f5442441 | 1294 | } else { |
76444f50 | 1295 | 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 |
1296 | res = -EOPNOTSUPP; |
1297 | goto err_undo_flags; | |
00503b6f | 1298 | } |
2ab82852 | 1299 | } |
1da177e4 LT |
1300 | } |
1301 | ||
8d8fc29d AW |
1302 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1303 | ||
c20811a7 JP |
1304 | /* If this is the first slave, then we need to set the master's hardware |
1305 | * address to be the same as the slave's. */ | |
0965a1f3 | 1306 | if (!bond_has_slaves(bond) && |
dec1e90e | 1307 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1308 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1309 | |
3fdddd85 | 1310 | new_slave = bond_alloc_slave(bond); |
1da177e4 LT |
1311 | if (!new_slave) { |
1312 | res = -ENOMEM; | |
1313 | goto err_undo_flags; | |
1314 | } | |
3fdddd85 | 1315 | |
dc73c41f VF |
1316 | new_slave->bond = bond; |
1317 | new_slave->dev = slave_dev; | |
bb1d9123 AG |
1318 | /* |
1319 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1320 | * is set via sysfs or module option if desired. | |
1321 | */ | |
1322 | new_slave->queue_id = 0; | |
1323 | ||
b15ba0fb JP |
1324 | /* Save slave's original mtu and then set it to match the bond */ |
1325 | new_slave->original_mtu = slave_dev->mtu; | |
1326 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1327 | if (res) { | |
76444f50 | 1328 | netdev_dbg(bond_dev, "Error %d calling dev_set_mtu\n", res); |
b15ba0fb JP |
1329 | goto err_free; |
1330 | } | |
1331 | ||
217df670 JV |
1332 | /* |
1333 | * Save slave's original ("permanent") mac address for modes | |
1334 | * that need it, and for restoring it upon release, and then | |
1335 | * set it to the master's address | |
1336 | */ | |
ada0f863 | 1337 | ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr); |
1da177e4 | 1338 | |
00503b6f | 1339 | if (!bond->params.fail_over_mac || |
01844098 | 1340 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 MS |
1341 | /* |
1342 | * Set slave to master's mac address. The application already | |
1343 | * set the master's mac address to that of the first slave | |
1344 | */ | |
1345 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1346 | addr.sa_family = slave_dev->type; | |
1347 | res = dev_set_mac_address(slave_dev, &addr); | |
1348 | if (res) { | |
76444f50 | 1349 | netdev_dbg(bond_dev, "Error %d calling set_mac_address\n", res); |
b15ba0fb | 1350 | goto err_restore_mtu; |
2ab82852 | 1351 | } |
217df670 | 1352 | } |
1da177e4 | 1353 | |
217df670 JV |
1354 | /* open the slave since the application closed it */ |
1355 | res = dev_open(slave_dev); | |
1356 | if (res) { | |
76444f50 | 1357 | netdev_dbg(bond_dev, "Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1358 | goto err_restore_mac; |
1da177e4 LT |
1359 | } |
1360 | ||
0b680e75 | 1361 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1362 | |
58402054 | 1363 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1364 | /* bond_alb_init_slave() must be called before all other stages since |
1365 | * it might fail and we do not want to have to undo everything | |
1366 | */ | |
1367 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1368 | if (res) |
569f0c4d | 1369 | goto err_close; |
1da177e4 LT |
1370 | } |
1371 | ||
ec0865a9 | 1372 | /* If the mode uses primary, then the following is handled by |
303d1cbf | 1373 | * bond_change_active_slave(). |
1da177e4 | 1374 | */ |
ec0865a9 | 1375 | if (!bond_uses_primary(bond)) { |
1da177e4 LT |
1376 | /* set promiscuity level to new slave */ |
1377 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1378 | res = dev_set_promiscuity(slave_dev, 1); |
1379 | if (res) | |
1380 | goto err_close; | |
1da177e4 LT |
1381 | } |
1382 | ||
1383 | /* set allmulti level to new slave */ | |
1384 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1385 | res = dev_set_allmulti(slave_dev, 1); |
1386 | if (res) | |
1387 | goto err_close; | |
1da177e4 LT |
1388 | } |
1389 | ||
b9e40857 | 1390 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1391 | |
1392 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1393 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1394 | ||
b9e40857 | 1395 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1396 | } |
1397 | ||
01844098 | 1398 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1da177e4 LT |
1399 | /* add lacpdu mc addr to mc list */ |
1400 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1401 | ||
22bedad3 | 1402 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1403 | } |
1404 | ||
b8e2fde4 WY |
1405 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1406 | if (res) { | |
76444f50 VF |
1407 | netdev_err(bond_dev, "Couldn't add bond vlan ids to %s\n", |
1408 | slave_dev->name); | |
1ff412ad | 1409 | goto err_close; |
1410 | } | |
1da177e4 | 1411 | |
c8c23903 | 1412 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1413 | |
1414 | new_slave->delay = 0; | |
1415 | new_slave->link_failure_count = 0; | |
1416 | ||
876254ae VF |
1417 | bond_update_speed_duplex(new_slave); |
1418 | ||
49f17de7 | 1419 | new_slave->last_rx = jiffies - |
f31c7937 | 1420 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1421 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1422 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1423 | |
1da177e4 LT |
1424 | if (bond->params.miimon && !bond->params.use_carrier) { |
1425 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1426 | ||
1427 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1428 | /* | |
1429 | * miimon is set but a bonded network driver | |
1430 | * does not support ETHTOOL/MII and | |
1431 | * arp_interval is not set. Note: if | |
1432 | * use_carrier is enabled, we will never go | |
1433 | * here (because netif_carrier is always | |
1434 | * supported); thus, we don't need to change | |
1435 | * the messages for netif_carrier. | |
1436 | */ | |
76444f50 VF |
1437 | 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", |
1438 | slave_dev->name); | |
1da177e4 LT |
1439 | } else if (link_reporting == -1) { |
1440 | /* unable get link status using mii/ethtool */ | |
76444f50 VF |
1441 | 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", |
1442 | slave_dev->name); | |
1da177e4 LT |
1443 | } |
1444 | } | |
1445 | ||
1446 | /* check for initial state */ | |
f31c7937 MK |
1447 | if (bond->params.miimon) { |
1448 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1449 | if (bond->params.updelay) { | |
1450 | new_slave->link = BOND_LINK_BACK; | |
1451 | new_slave->delay = bond->params.updelay; | |
1452 | } else { | |
1453 | new_slave->link = BOND_LINK_UP; | |
1454 | } | |
1da177e4 | 1455 | } else { |
f31c7937 | 1456 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1457 | } |
f31c7937 MK |
1458 | } else if (bond->params.arp_interval) { |
1459 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1460 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1461 | } else { |
f31c7937 | 1462 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1463 | } |
1464 | ||
f31c7937 | 1465 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 1466 | new_slave->last_link_up = jiffies; |
76444f50 VF |
1467 | netdev_dbg(bond_dev, "Initial state of slave_dev is BOND_LINK_%s\n", |
1468 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1469 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 1470 | |
ec0865a9 | 1471 | if (bond_uses_primary(bond) && bond->params.primary[0]) { |
1da177e4 | 1472 | /* if there is a primary slave, remember it */ |
a549952a | 1473 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
059b47e8 | 1474 | rcu_assign_pointer(bond->primary_slave, new_slave); |
a549952a JP |
1475 | bond->force_primary = true; |
1476 | } | |
1da177e4 LT |
1477 | } |
1478 | ||
01844098 | 1479 | switch (BOND_MODE(bond)) { |
1da177e4 | 1480 | case BOND_MODE_ACTIVEBACKUP: |
5e5b0665 | 1481 | bond_set_slave_inactive_flags(new_slave, |
1482 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1483 | break; |
1484 | case BOND_MODE_8023AD: | |
1485 | /* in 802.3ad mode, the internal mechanism | |
1486 | * will activate the slaves in the selected | |
1487 | * aggregator | |
1488 | */ | |
5e5b0665 | 1489 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 1490 | /* if this is the first slave */ |
23c147e0 | 1491 | if (!prev_slave) { |
3fdddd85 | 1492 | SLAVE_AD_INFO(new_slave)->id = 1; |
1da177e4 LT |
1493 | /* Initialize AD with the number of times that the AD timer is called in 1 second |
1494 | * can be called only after the mac address of the bond is set | |
1495 | */ | |
56d00c67 | 1496 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 | 1497 | } else { |
3fdddd85 | 1498 | SLAVE_AD_INFO(new_slave)->id = |
1499 | SLAVE_AD_INFO(prev_slave)->id + 1; | |
1da177e4 LT |
1500 | } |
1501 | ||
1502 | bond_3ad_bind_slave(new_slave); | |
1503 | break; | |
1504 | case BOND_MODE_TLB: | |
1505 | case BOND_MODE_ALB: | |
e30bc066 | 1506 | bond_set_active_slave(new_slave); |
5e5b0665 | 1507 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
1508 | break; |
1509 | default: | |
76444f50 | 1510 | netdev_dbg(bond_dev, "This slave is always active in trunk mode\n"); |
1da177e4 LT |
1511 | |
1512 | /* always active in trunk mode */ | |
e30bc066 | 1513 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1514 | |
1515 | /* In trunking mode there is little meaning to curr_active_slave | |
1516 | * anyway (it holds no special properties of the bond device), | |
1517 | * so we can change it without calling change_active_interface() | |
1518 | */ | |
4740d638 ED |
1519 | if (!rcu_access_pointer(bond->curr_active_slave) && |
1520 | new_slave->link == BOND_LINK_UP) | |
278b2083 | 1521 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1522 | |
1da177e4 LT |
1523 | break; |
1524 | } /* switch(bond_mode) */ | |
1525 | ||
f6dc31a8 | 1526 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1527 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1528 | if (slave_dev->npinfo) { |
1529 | if (slave_enable_netpoll(new_slave)) { | |
76444f50 | 1530 | netdev_info(bond_dev, "master_dev is using netpoll, but new slave device does not support netpoll\n"); |
8a8efa22 | 1531 | res = -EBUSY; |
f7d9821a | 1532 | goto err_detach; |
8a8efa22 | 1533 | } |
f6dc31a8 WC |
1534 | } |
1535 | #endif | |
8a8efa22 | 1536 | |
35d48903 JP |
1537 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1538 | new_slave); | |
1539 | if (res) { | |
76444f50 | 1540 | netdev_dbg(bond_dev, "Error %d calling netdev_rx_handler_register\n", res); |
5831d66e | 1541 | goto err_detach; |
35d48903 JP |
1542 | } |
1543 | ||
1f718f0f VF |
1544 | res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave); |
1545 | if (res) { | |
76444f50 | 1546 | netdev_dbg(bond_dev, "Error %d calling bond_master_upper_dev_link\n", res); |
1f718f0f VF |
1547 | goto err_unregister; |
1548 | } | |
1549 | ||
07699f9a | 1550 | res = bond_sysfs_slave_add(new_slave); |
1551 | if (res) { | |
76444f50 | 1552 | netdev_dbg(bond_dev, "Error %d calling bond_sysfs_slave_add\n", res); |
07699f9a | 1553 | goto err_upper_unlink; |
1554 | } | |
1555 | ||
5378c2e6 VF |
1556 | bond->slave_cnt++; |
1557 | bond_compute_features(bond); | |
1558 | bond_set_carrier(bond); | |
1559 | ||
ec0865a9 | 1560 | if (bond_uses_primary(bond)) { |
f80889a5 | 1561 | block_netpoll_tx(); |
5378c2e6 | 1562 | bond_select_active_slave(bond); |
f80889a5 | 1563 | unblock_netpoll_tx(); |
5378c2e6 | 1564 | } |
1f718f0f | 1565 | |
76444f50 VF |
1566 | netdev_info(bond_dev, "Enslaving %s as %s interface with %s link\n", |
1567 | slave_dev->name, | |
1568 | bond_is_active_slave(new_slave) ? "an active" : "a backup", | |
1569 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
1570 | |
1571 | /* enslave is successful */ | |
1572 | return 0; | |
1573 | ||
1574 | /* Undo stages on error */ | |
07699f9a | 1575 | err_upper_unlink: |
1576 | bond_upper_dev_unlink(bond_dev, slave_dev); | |
1577 | ||
1f718f0f VF |
1578 | err_unregister: |
1579 | netdev_rx_handler_unregister(slave_dev); | |
1580 | ||
f7d9821a | 1581 | err_detach: |
ec0865a9 | 1582 | if (!bond_uses_primary(bond)) |
303d1cbf JV |
1583 | bond_hw_addr_flush(bond_dev, slave_dev); |
1584 | ||
1ff412ad | 1585 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
059b47e8 NA |
1586 | if (rcu_access_pointer(bond->primary_slave) == new_slave) |
1587 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
4740d638 | 1588 | if (rcu_access_pointer(bond->curr_active_slave) == new_slave) { |
f80889a5 | 1589 | block_netpoll_tx(); |
c8517035 | 1590 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1591 | bond_select_active_slave(bond); |
f80889a5 | 1592 | unblock_netpoll_tx(); |
3c5913b5 | 1593 | } |
059b47e8 NA |
1594 | /* either primary_slave or curr_active_slave might've changed */ |
1595 | synchronize_rcu(); | |
fc7a72ac | 1596 | slave_disable_netpoll(new_slave); |
f7d9821a | 1597 | |
1da177e4 | 1598 | err_close: |
b6a5a7b9 | 1599 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1600 | dev_close(slave_dev); |
1601 | ||
1602 | err_restore_mac: | |
00503b6f | 1603 | if (!bond->params.fail_over_mac || |
01844098 | 1604 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
3915c1e8 JV |
1605 | /* XXX TODO - fom follow mode needs to change master's |
1606 | * MAC if this slave's MAC is in use by the bond, or at | |
1607 | * least print a warning. | |
1608 | */ | |
ada0f863 | 1609 | ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr); |
2ab82852 MS |
1610 | addr.sa_family = slave_dev->type; |
1611 | dev_set_mac_address(slave_dev, &addr); | |
1612 | } | |
1da177e4 | 1613 | |
b15ba0fb JP |
1614 | err_restore_mtu: |
1615 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1616 | ||
1da177e4 | 1617 | err_free: |
3fdddd85 | 1618 | bond_free_slave(new_slave); |
1da177e4 LT |
1619 | |
1620 | err_undo_flags: | |
b8fad459 | 1621 | /* Enslave of first slave has failed and we need to fix master's mac */ |
0965a1f3 | 1622 | if (!bond_has_slaves(bond) && |
844223ab | 1623 | ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr)) |
b8fad459 | 1624 | eth_hw_addr_random(bond_dev); |
3d632c3f | 1625 | |
1da177e4 LT |
1626 | return res; |
1627 | } | |
1628 | ||
1629 | /* | |
1630 | * Try to release the slave device <slave> from the bond device <master> | |
1631 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1632 | * is write-locked. If "all" is true it means that the function is being called |
1633 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1634 | * |
1635 | * The rules for slave state should be: | |
1636 | * for Active/Backup: | |
1637 | * Active stays on all backups go down | |
1638 | * for Bonded connections: | |
1639 | * The first up interface should be left on and all others downed. | |
1640 | */ | |
0896341a | 1641 | static int __bond_release_one(struct net_device *bond_dev, |
1642 | struct net_device *slave_dev, | |
1643 | bool all) | |
1da177e4 | 1644 | { |
454d7c9b | 1645 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1646 | struct slave *slave, *oldcurrent; |
1647 | struct sockaddr addr; | |
5a0068de | 1648 | int old_flags = bond_dev->flags; |
c8f44aff | 1649 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1650 | |
1651 | /* slave is not a slave or master is not master of this slave */ | |
1652 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1653 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
76444f50 VF |
1654 | netdev_err(bond_dev, "cannot release %s\n", |
1655 | slave_dev->name); | |
1da177e4 LT |
1656 | return -EINVAL; |
1657 | } | |
1658 | ||
e843fa50 | 1659 | block_netpoll_tx(); |
1da177e4 LT |
1660 | |
1661 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1662 | if (!slave) { | |
1663 | /* not a slave of this bond */ | |
76444f50 VF |
1664 | netdev_info(bond_dev, "%s not enslaved\n", |
1665 | slave_dev->name); | |
e843fa50 | 1666 | unblock_netpoll_tx(); |
1da177e4 LT |
1667 | return -EINVAL; |
1668 | } | |
1669 | ||
07699f9a | 1670 | bond_sysfs_slave_del(slave); |
1671 | ||
1f718f0f | 1672 | bond_upper_dev_unlink(bond_dev, slave_dev); |
35d48903 JP |
1673 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1674 | * for this slave anymore. | |
1675 | */ | |
1676 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 | 1677 | |
e470259f | 1678 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
1da177e4 | 1679 | bond_3ad_unbind_slave(slave); |
1da177e4 | 1680 | |
76444f50 VF |
1681 | netdev_info(bond_dev, "Releasing %s interface %s\n", |
1682 | bond_is_active_slave(slave) ? "active" : "backup", | |
1683 | slave_dev->name); | |
1da177e4 | 1684 | |
4740d638 | 1685 | oldcurrent = rcu_access_pointer(bond->curr_active_slave); |
1da177e4 | 1686 | |
85741718 | 1687 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
1da177e4 | 1688 | |
00503b6f | 1689 | if (!all && (!bond->params.fail_over_mac || |
01844098 | 1690 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP)) { |
844223ab | 1691 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1692 | bond_has_slaves(bond)) |
76444f50 VF |
1693 | 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", |
1694 | slave_dev->name, slave->perm_hwaddr, | |
1695 | bond_dev->name, slave_dev->name); | |
dec1e90e | 1696 | } |
1697 | ||
059b47e8 NA |
1698 | if (rtnl_dereference(bond->primary_slave) == slave) |
1699 | RCU_INIT_POINTER(bond->primary_slave, NULL); | |
1da177e4 | 1700 | |
1c72cfdc | 1701 | if (oldcurrent == slave) |
1da177e4 | 1702 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1703 | |
58402054 | 1704 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1705 | /* Must be called only after the slave has been |
1706 | * detached from the list and the curr_active_slave | |
1707 | * has been cleared (if our_slave == old_current), | |
1708 | * but before a new active slave is selected. | |
1709 | */ | |
1710 | bond_alb_deinit_slave(bond, slave); | |
1711 | } | |
1712 | ||
0896341a | 1713 | if (all) { |
36708b89 | 1714 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 1715 | } else if (oldcurrent == slave) { |
059fe7a5 JV |
1716 | /* |
1717 | * Note that we hold RTNL over this sequence, so there | |
1718 | * is no concern that another slave add/remove event | |
1719 | * will interfere. | |
1720 | */ | |
1da177e4 | 1721 | bond_select_active_slave(bond); |
059fe7a5 JV |
1722 | } |
1723 | ||
0965a1f3 | 1724 | if (!bond_has_slaves(bond)) { |
ff59c456 | 1725 | bond_set_carrier(bond); |
409cc1f8 | 1726 | eth_hw_addr_random(bond_dev); |
1da177e4 LT |
1727 | } |
1728 | ||
e843fa50 | 1729 | unblock_netpoll_tx(); |
278b2083 | 1730 | synchronize_rcu(); |
ee6154e1 | 1731 | bond->slave_cnt--; |
1da177e4 | 1732 | |
0965a1f3 | 1733 | if (!bond_has_slaves(bond)) { |
2af73d4b | 1734 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1735 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1736 | } | |
2af73d4b | 1737 | |
b2a103e6 MM |
1738 | bond_compute_features(bond); |
1739 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1740 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
76444f50 VF |
1741 | netdev_info(bond_dev, "last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n", |
1742 | slave_dev->name, bond_dev->name); | |
b2a103e6 | 1743 | |
b76cdba9 | 1744 | /* must do this from outside any spinlocks */ |
1ff412ad | 1745 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1746 | |
ec0865a9 | 1747 | /* If the mode uses primary, then this cases was handled above by |
303d1cbf | 1748 | * bond_change_active_slave(..., NULL) |
1da177e4 | 1749 | */ |
ec0865a9 | 1750 | if (!bond_uses_primary(bond)) { |
5a0068de NH |
1751 | /* unset promiscuity level from slave |
1752 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
1753 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
1754 | * value of that flag before that change, as that was the value | |
1755 | * when this slave was attached, so we cache at the start of the | |
1756 | * function and use it here. Same goes for ALLMULTI below | |
1757 | */ | |
1758 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 1759 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1760 | |
1761 | /* unset allmulti level from slave */ | |
5a0068de | 1762 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 1763 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1764 | |
303d1cbf | 1765 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1766 | } |
1767 | ||
8a8efa22 | 1768 | slave_disable_netpoll(slave); |
f6dc31a8 | 1769 | |
1da177e4 LT |
1770 | /* close slave before restoring its mac address */ |
1771 | dev_close(slave_dev); | |
1772 | ||
00503b6f | 1773 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
01844098 | 1774 | BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
2ab82852 | 1775 | /* restore original ("permanent") mac address */ |
ada0f863 | 1776 | ether_addr_copy(addr.sa_data, slave->perm_hwaddr); |
2ab82852 MS |
1777 | addr.sa_family = slave_dev->type; |
1778 | dev_set_mac_address(slave_dev, &addr); | |
1779 | } | |
1da177e4 | 1780 | |
b15ba0fb JP |
1781 | dev_set_mtu(slave_dev, slave->original_mtu); |
1782 | ||
2d7011ca | 1783 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 | 1784 | |
3fdddd85 | 1785 | bond_free_slave(slave); |
1da177e4 LT |
1786 | |
1787 | return 0; /* deletion OK */ | |
1788 | } | |
1789 | ||
0896341a | 1790 | /* A wrapper used because of ndo_del_link */ |
1791 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1792 | { | |
1793 | return __bond_release_one(bond_dev, slave_dev, false); | |
1794 | } | |
1795 | ||
d90a162a | 1796 | /* |
dadaa10b | 1797 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
1798 | * Must be under rtnl_lock when this function is called. |
1799 | */ | |
26d8ee75 | 1800 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1801 | struct net_device *slave_dev) | |
d90a162a | 1802 | { |
454d7c9b | 1803 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1804 | int ret; |
1805 | ||
1806 | ret = bond_release(bond_dev, slave_dev); | |
0965a1f3 | 1807 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 1808 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
76444f50 VF |
1809 | netdev_info(bond_dev, "Destroying bond %s\n", |
1810 | bond_dev->name); | |
9e71626c | 1811 | unregister_netdevice(bond_dev); |
d90a162a MS |
1812 | } |
1813 | return ret; | |
1814 | } | |
1815 | ||
1da177e4 LT |
1816 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1817 | { | |
454d7c9b | 1818 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 1819 | |
01844098 | 1820 | info->bond_mode = BOND_MODE(bond); |
1da177e4 LT |
1821 | info->miimon = bond->params.miimon; |
1822 | ||
1da177e4 | 1823 | info->num_slaves = bond->slave_cnt; |
1da177e4 LT |
1824 | |
1825 | return 0; | |
1826 | } | |
1827 | ||
1828 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1829 | { | |
454d7c9b | 1830 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1831 | struct list_head *iter; |
dec1e90e | 1832 | int i = 0, res = -ENODEV; |
1da177e4 | 1833 | struct slave *slave; |
1da177e4 | 1834 | |
9caff1e7 | 1835 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 1836 | if (i++ == (int)info->slave_id) { |
689c96cc ED |
1837 | res = 0; |
1838 | strcpy(info->slave_name, slave->dev->name); | |
1839 | info->link = slave->link; | |
e30bc066 | 1840 | info->state = bond_slave_state(slave); |
689c96cc | 1841 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
1842 | break; |
1843 | } | |
1844 | } | |
1da177e4 | 1845 | |
689c96cc | 1846 | return res; |
1da177e4 LT |
1847 | } |
1848 | ||
1849 | /*-------------------------------- Monitoring -------------------------------*/ | |
1850 | ||
4740d638 | 1851 | /* called with rcu_read_lock() */ |
f0c76d61 JV |
1852 | static int bond_miimon_inspect(struct bonding *bond) |
1853 | { | |
dec1e90e | 1854 | int link_state, commit = 0; |
9caff1e7 | 1855 | struct list_head *iter; |
f0c76d61 | 1856 | struct slave *slave; |
41f89100 JP |
1857 | bool ignore_updelay; |
1858 | ||
4740d638 | 1859 | ignore_updelay = !rcu_dereference(bond->curr_active_slave); |
1da177e4 | 1860 | |
4cb4f97b | 1861 | bond_for_each_slave_rcu(bond, slave, iter) { |
f0c76d61 | 1862 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 1863 | |
f0c76d61 | 1864 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
1865 | |
1866 | switch (slave->link) { | |
f0c76d61 JV |
1867 | case BOND_LINK_UP: |
1868 | if (link_state) | |
1869 | continue; | |
1da177e4 | 1870 | |
f0c76d61 JV |
1871 | slave->link = BOND_LINK_FAIL; |
1872 | slave->delay = bond->params.downdelay; | |
1873 | if (slave->delay) { | |
76444f50 VF |
1874 | netdev_info(bond->dev, "link status down for %sinterface %s, disabling it in %d ms\n", |
1875 | (BOND_MODE(bond) == | |
1876 | BOND_MODE_ACTIVEBACKUP) ? | |
1877 | (bond_is_active_slave(slave) ? | |
1878 | "active " : "backup ") : "", | |
1879 | slave->dev->name, | |
1880 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 1881 | } |
f0c76d61 JV |
1882 | /*FALLTHRU*/ |
1883 | case BOND_LINK_FAIL: | |
1884 | if (link_state) { | |
1885 | /* | |
1886 | * recovered before downdelay expired | |
1887 | */ | |
1888 | slave->link = BOND_LINK_UP; | |
8e603460 | 1889 | slave->last_link_up = jiffies; |
76444f50 VF |
1890 | netdev_info(bond->dev, "link status up again after %d ms for interface %s\n", |
1891 | (bond->params.downdelay - slave->delay) * | |
1892 | bond->params.miimon, | |
1893 | slave->dev->name); | |
f0c76d61 | 1894 | continue; |
1da177e4 | 1895 | } |
f0c76d61 JV |
1896 | |
1897 | if (slave->delay <= 0) { | |
1898 | slave->new_link = BOND_LINK_DOWN; | |
1899 | commit++; | |
1900 | continue; | |
1da177e4 | 1901 | } |
1da177e4 | 1902 | |
f0c76d61 JV |
1903 | slave->delay--; |
1904 | break; | |
1905 | ||
1906 | case BOND_LINK_DOWN: | |
1907 | if (!link_state) | |
1908 | continue; | |
1909 | ||
1910 | slave->link = BOND_LINK_BACK; | |
1911 | slave->delay = bond->params.updelay; | |
1912 | ||
1913 | if (slave->delay) { | |
76444f50 VF |
1914 | netdev_info(bond->dev, "link status up for interface %s, enabling it in %d ms\n", |
1915 | slave->dev->name, | |
1916 | ignore_updelay ? 0 : | |
1917 | bond->params.updelay * | |
1918 | bond->params.miimon); | |
f0c76d61 JV |
1919 | } |
1920 | /*FALLTHRU*/ | |
1921 | case BOND_LINK_BACK: | |
1922 | if (!link_state) { | |
1923 | slave->link = BOND_LINK_DOWN; | |
76444f50 VF |
1924 | netdev_info(bond->dev, "link status down again after %d ms for interface %s\n", |
1925 | (bond->params.updelay - slave->delay) * | |
1926 | bond->params.miimon, | |
1927 | slave->dev->name); | |
f0c76d61 JV |
1928 | |
1929 | continue; | |
1930 | } | |
1931 | ||
41f89100 JP |
1932 | if (ignore_updelay) |
1933 | slave->delay = 0; | |
1934 | ||
f0c76d61 JV |
1935 | if (slave->delay <= 0) { |
1936 | slave->new_link = BOND_LINK_UP; | |
1937 | commit++; | |
41f89100 | 1938 | ignore_updelay = false; |
f0c76d61 | 1939 | continue; |
1da177e4 | 1940 | } |
f0c76d61 JV |
1941 | |
1942 | slave->delay--; | |
1da177e4 | 1943 | break; |
f0c76d61 JV |
1944 | } |
1945 | } | |
1da177e4 | 1946 | |
f0c76d61 JV |
1947 | return commit; |
1948 | } | |
1da177e4 | 1949 | |
f0c76d61 JV |
1950 | static void bond_miimon_commit(struct bonding *bond) |
1951 | { | |
9caff1e7 | 1952 | struct list_head *iter; |
059b47e8 | 1953 | struct slave *slave, *primary; |
f0c76d61 | 1954 | |
9caff1e7 | 1955 | bond_for_each_slave(bond, slave, iter) { |
f0c76d61 JV |
1956 | switch (slave->new_link) { |
1957 | case BOND_LINK_NOCHANGE: | |
1958 | continue; | |
1da177e4 | 1959 | |
f0c76d61 JV |
1960 | case BOND_LINK_UP: |
1961 | slave->link = BOND_LINK_UP; | |
8e603460 | 1962 | slave->last_link_up = jiffies; |
f0c76d61 | 1963 | |
059b47e8 | 1964 | primary = rtnl_dereference(bond->primary_slave); |
01844098 | 1965 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
f0c76d61 | 1966 | /* prevent it from being the active one */ |
e30bc066 | 1967 | bond_set_backup_slave(slave); |
01844098 | 1968 | } else if (BOND_MODE(bond) != BOND_MODE_ACTIVEBACKUP) { |
f0c76d61 | 1969 | /* make it immediately active */ |
e30bc066 | 1970 | bond_set_active_slave(slave); |
059b47e8 | 1971 | } else if (slave != primary) { |
f0c76d61 | 1972 | /* prevent it from being the active one */ |
e30bc066 | 1973 | bond_set_backup_slave(slave); |
1da177e4 | 1974 | } |
1da177e4 | 1975 | |
76444f50 VF |
1976 | netdev_info(bond->dev, "link status definitely up for interface %s, %u Mbps %s duplex\n", |
1977 | slave->dev->name, | |
1978 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, | |
1979 | slave->duplex ? "full" : "half"); | |
1da177e4 | 1980 | |
f0c76d61 | 1981 | /* notify ad that the link status has changed */ |
01844098 | 1982 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 | 1983 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); |
059fe7a5 | 1984 | |
58402054 | 1985 | if (bond_is_lb(bond)) |
f0c76d61 JV |
1986 | bond_alb_handle_link_change(bond, slave, |
1987 | BOND_LINK_UP); | |
1da177e4 | 1988 | |
059b47e8 | 1989 | if (!bond->curr_active_slave || slave == primary) |
f0c76d61 | 1990 | goto do_failover; |
1da177e4 | 1991 | |
f0c76d61 | 1992 | continue; |
059fe7a5 | 1993 | |
f0c76d61 | 1994 | case BOND_LINK_DOWN: |
fba4acda JV |
1995 | if (slave->link_failure_count < UINT_MAX) |
1996 | slave->link_failure_count++; | |
1997 | ||
f0c76d61 | 1998 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 1999 | |
01844098 VF |
2000 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP || |
2001 | BOND_MODE(bond) == BOND_MODE_8023AD) | |
5e5b0665 | 2002 | bond_set_slave_inactive_flags(slave, |
2003 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2004 | |
76444f50 VF |
2005 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2006 | slave->dev->name); | |
f0c76d61 | 2007 | |
01844098 | 2008 | if (BOND_MODE(bond) == BOND_MODE_8023AD) |
f0c76d61 JV |
2009 | bond_3ad_handle_link_change(slave, |
2010 | BOND_LINK_DOWN); | |
2011 | ||
ae63e808 | 2012 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2013 | bond_alb_handle_link_change(bond, slave, |
2014 | BOND_LINK_DOWN); | |
2015 | ||
4740d638 | 2016 | if (slave == rcu_access_pointer(bond->curr_active_slave)) |
f0c76d61 JV |
2017 | goto do_failover; |
2018 | ||
2019 | continue; | |
2020 | ||
2021 | default: | |
76444f50 VF |
2022 | netdev_err(bond->dev, "invalid new link %d on slave %s\n", |
2023 | slave->new_link, slave->dev->name); | |
f0c76d61 JV |
2024 | slave->new_link = BOND_LINK_NOCHANGE; |
2025 | ||
2026 | continue; | |
2027 | } | |
2028 | ||
2029 | do_failover: | |
2030 | ASSERT_RTNL(); | |
e843fa50 | 2031 | block_netpoll_tx(); |
f0c76d61 | 2032 | bond_select_active_slave(bond); |
e843fa50 | 2033 | unblock_netpoll_tx(); |
f0c76d61 JV |
2034 | } |
2035 | ||
2036 | bond_set_carrier(bond); | |
1da177e4 LT |
2037 | } |
2038 | ||
0b0eef66 JV |
2039 | /* |
2040 | * bond_mii_monitor | |
2041 | * | |
2042 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2043 | * inspection, then (if inspection indicates something needs to be done) |
2044 | * an acquisition of appropriate locks followed by a commit phase to | |
2045 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2046 | */ |
6da67d26 | 2047 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2048 | { |
2049 | struct bonding *bond = container_of(work, struct bonding, | |
2050 | mii_work.work); | |
ad246c99 | 2051 | bool should_notify_peers = false; |
1f2cd845 | 2052 | unsigned long delay; |
0b0eef66 | 2053 | |
1f2cd845 DM |
2054 | delay = msecs_to_jiffies(bond->params.miimon); |
2055 | ||
2056 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2057 | goto re_arm; |
b59f9f74 | 2058 | |
4cb4f97b | 2059 | rcu_read_lock(); |
2060 | ||
ad246c99 BH |
2061 | should_notify_peers = bond_should_notify_peers(bond); |
2062 | ||
1f2cd845 | 2063 | if (bond_miimon_inspect(bond)) { |
4cb4f97b | 2064 | rcu_read_unlock(); |
f0c76d61 | 2065 | |
1f2cd845 DM |
2066 | /* Race avoidance with bond_close cancel of workqueue */ |
2067 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2068 | delay = 1; |
2069 | should_notify_peers = false; | |
2070 | goto re_arm; | |
2071 | } | |
6b6c5261 | 2072 | |
1f2cd845 DM |
2073 | bond_miimon_commit(bond); |
2074 | ||
1f2cd845 | 2075 | rtnl_unlock(); /* might sleep, hold no other locks */ |
4cb4f97b | 2076 | } else |
2077 | rcu_read_unlock(); | |
0b0eef66 | 2078 | |
f0c76d61 | 2079 | re_arm: |
e6d265e8 | 2080 | if (bond->params.miimon) |
1f2cd845 DM |
2081 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2082 | ||
1f2cd845 DM |
2083 | if (should_notify_peers) { |
2084 | if (!rtnl_trylock()) | |
2085 | return; | |
2086 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2087 | rtnl_unlock(); | |
2088 | } | |
0b0eef66 | 2089 | } |
c3ade5ca | 2090 | |
50223ce4 | 2091 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2092 | { |
50223ce4 VF |
2093 | struct net_device *upper; |
2094 | struct list_head *iter; | |
2095 | bool ret = false; | |
f5b2b966 | 2096 | |
eaddcd76 | 2097 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2098 | return true; |
f5b2b966 | 2099 | |
50223ce4 | 2100 | rcu_read_lock(); |
2f268f12 | 2101 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2102 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2103 | ret = true; | |
2104 | break; | |
2105 | } | |
f5b2b966 | 2106 | } |
50223ce4 | 2107 | rcu_read_unlock(); |
f5b2b966 | 2108 | |
50223ce4 | 2109 | return ret; |
f5b2b966 JV |
2110 | } |
2111 | ||
c3ade5ca JV |
2112 | /* |
2113 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2114 | * switches in VLAN mode (especially if ports are configured as | |
2115 | * "native" to a VLAN) might not pass non-tagged frames. | |
2116 | */ | |
fbd929f2 | 2117 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, |
2118 | __be32 dest_ip, __be32 src_ip, | |
44a40855 | 2119 | struct bond_vlan_tag *tags) |
c3ade5ca JV |
2120 | { |
2121 | struct sk_buff *skb; | |
3e403a77 | 2122 | struct bond_vlan_tag *outer_tag = tags; |
c3ade5ca | 2123 | |
76444f50 VF |
2124 | netdev_dbg(slave_dev, "arp %d on slave %s: dst %pI4 src %pI4\n", |
2125 | arp_op, slave_dev->name, &dest_ip, &src_ip); | |
3d632c3f | 2126 | |
c3ade5ca JV |
2127 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2128 | NULL, slave_dev->dev_addr, NULL); | |
2129 | ||
2130 | if (!skb) { | |
4873ac3c | 2131 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2132 | return; |
2133 | } | |
fbd929f2 | 2134 | |
3e403a77 VF |
2135 | if (!tags || tags->vlan_proto == VLAN_N_VID) |
2136 | goto xmit; | |
2137 | ||
2138 | tags++; | |
2139 | ||
44a40855 | 2140 | /* Go through all the tags backwards and add them to the packet */ |
3e403a77 VF |
2141 | while (tags->vlan_proto != VLAN_N_VID) { |
2142 | if (!tags->vlan_id) { | |
2143 | tags++; | |
44a40855 | 2144 | continue; |
3e403a77 | 2145 | } |
44a40855 | 2146 | |
76444f50 | 2147 | netdev_dbg(slave_dev, "inner tag: proto %X vid %X\n", |
3e403a77 VF |
2148 | ntohs(outer_tag->vlan_proto), tags->vlan_id); |
2149 | skb = __vlan_put_tag(skb, tags->vlan_proto, | |
2150 | tags->vlan_id); | |
44a40855 VY |
2151 | if (!skb) { |
2152 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2153 | return; | |
2154 | } | |
3e403a77 VF |
2155 | |
2156 | tags++; | |
44a40855 VY |
2157 | } |
2158 | /* Set the outer tag */ | |
3e403a77 | 2159 | if (outer_tag->vlan_id) { |
76444f50 | 2160 | netdev_dbg(slave_dev, "outer tag: proto %X vid %X\n", |
3e403a77 VF |
2161 | ntohs(outer_tag->vlan_proto), outer_tag->vlan_id); |
2162 | skb = vlan_put_tag(skb, outer_tag->vlan_proto, | |
2163 | outer_tag->vlan_id); | |
c3ade5ca | 2164 | if (!skb) { |
4873ac3c | 2165 | net_err_ratelimited("failed to insert outer VLAN tag\n"); |
c3ade5ca JV |
2166 | return; |
2167 | } | |
2168 | } | |
3e403a77 VF |
2169 | |
2170 | xmit: | |
c3ade5ca JV |
2171 | arp_xmit(skb); |
2172 | } | |
2173 | ||
44a40855 VY |
2174 | /* Validate the device path between the @start_dev and the @end_dev. |
2175 | * The path is valid if the @end_dev is reachable through device | |
2176 | * stacking. | |
2177 | * When the path is validated, collect any vlan information in the | |
2178 | * path. | |
2179 | */ | |
3e403a77 VF |
2180 | struct bond_vlan_tag *bond_verify_device_path(struct net_device *start_dev, |
2181 | struct net_device *end_dev, | |
2182 | int level) | |
44a40855 | 2183 | { |
3e403a77 | 2184 | struct bond_vlan_tag *tags; |
44a40855 VY |
2185 | struct net_device *upper; |
2186 | struct list_head *iter; | |
44a40855 | 2187 | |
3e403a77 VF |
2188 | if (start_dev == end_dev) { |
2189 | tags = kzalloc(sizeof(*tags) * (level + 1), GFP_ATOMIC); | |
2190 | if (!tags) | |
2191 | return ERR_PTR(-ENOMEM); | |
2192 | tags[level].vlan_proto = VLAN_N_VID; | |
2193 | return tags; | |
2194 | } | |
44a40855 VY |
2195 | |
2196 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
3e403a77 VF |
2197 | tags = bond_verify_device_path(upper, end_dev, level + 1); |
2198 | if (IS_ERR_OR_NULL(tags)) { | |
2199 | if (IS_ERR(tags)) | |
2200 | return tags; | |
2201 | continue; | |
44a40855 | 2202 | } |
3e403a77 VF |
2203 | if (is_vlan_dev(upper)) { |
2204 | tags[level].vlan_proto = vlan_dev_vlan_proto(upper); | |
2205 | tags[level].vlan_id = vlan_dev_vlan_id(upper); | |
2206 | } | |
2207 | ||
2208 | return tags; | |
44a40855 VY |
2209 | } |
2210 | ||
3e403a77 | 2211 | return NULL; |
44a40855 | 2212 | } |
c3ade5ca | 2213 | |
1da177e4 LT |
2214 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2215 | { | |
c3ade5ca | 2216 | struct rtable *rt; |
3e403a77 | 2217 | struct bond_vlan_tag *tags; |
27bc11e6 | 2218 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2219 | int i; |
1da177e4 | 2220 | |
27bc11e6 | 2221 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
76444f50 | 2222 | netdev_dbg(bond->dev, "basa: target %pI4\n", &targets[i]); |
3e403a77 | 2223 | tags = NULL; |
c3ade5ca | 2224 | |
27bc11e6 | 2225 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2226 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2227 | RTO_ONLINK, 0); | |
b23dd4fe | 2228 | if (IS_ERR(rt)) { |
28572760 VF |
2229 | /* there's no route to target - try to send arp |
2230 | * probe to generate any traffic (arp_validate=0) | |
2231 | */ | |
4873ac3c | 2232 | if (bond->params.arp_validate) |
2233 | net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", | |
2234 | bond->dev->name, | |
2235 | &targets[i]); | |
44a40855 VY |
2236 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2237 | 0, tags); | |
c3ade5ca JV |
2238 | continue; |
2239 | } | |
2240 | ||
27bc11e6 VF |
2241 | /* bond device itself */ |
2242 | if (rt->dst.dev == bond->dev) | |
2243 | goto found; | |
2244 | ||
2245 | rcu_read_lock(); | |
3e403a77 | 2246 | tags = bond_verify_device_path(bond->dev, rt->dst.dev, 0); |
27bc11e6 | 2247 | rcu_read_unlock(); |
c3ade5ca | 2248 | |
3e403a77 | 2249 | if (!IS_ERR_OR_NULL(tags)) |
44a40855 VY |
2250 | goto found; |
2251 | ||
27bc11e6 | 2252 | /* Not our device - skip */ |
76444f50 VF |
2253 | netdev_dbg(bond->dev, "no path to arp_ip_target %pI4 via rt.dev %s\n", |
2254 | &targets[i], rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
3e32582f | 2255 | |
ed4b9f80 | 2256 | ip_rt_put(rt); |
27bc11e6 VF |
2257 | continue; |
2258 | ||
2259 | found: | |
2260 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2261 | ip_rt_put(rt); | |
2262 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
44a40855 | 2263 | addr, tags); |
a67eed57 | 2264 | kfree(tags); |
c3ade5ca JV |
2265 | } |
2266 | } | |
2267 | ||
d3bb52b0 | 2268 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2269 | { |
8599b52e VF |
2270 | int i; |
2271 | ||
87a7b84b | 2272 | if (!sip || !bond_has_this_ip(bond, tip)) { |
76444f50 VF |
2273 | netdev_dbg(bond->dev, "bva: sip %pI4 tip %pI4 not found\n", |
2274 | &sip, &tip); | |
87a7b84b VF |
2275 | return; |
2276 | } | |
f5b2b966 | 2277 | |
8599b52e VF |
2278 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2279 | if (i == -1) { | |
76444f50 VF |
2280 | netdev_dbg(bond->dev, "bva: sip %pI4 not found in targets\n", |
2281 | &sip); | |
87a7b84b | 2282 | return; |
f5b2b966 | 2283 | } |
49f17de7 | 2284 | slave->last_rx = jiffies; |
8599b52e | 2285 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2286 | } |
2287 | ||
5bb9e0b5 | 2288 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2289 | struct slave *slave) | |
f5b2b966 | 2290 | { |
de063b70 | 2291 | struct arphdr *arp = (struct arphdr *)skb->data; |
010d3c39 | 2292 | struct slave *curr_active_slave; |
f5b2b966 | 2293 | unsigned char *arp_ptr; |
d3bb52b0 | 2294 | __be32 sip, tip; |
f2cb691a | 2295 | int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
f5b2b966 | 2296 | |
f2cb691a | 2297 | if (!slave_do_arp_validate(bond, slave)) { |
bedabf90 | 2298 | if ((slave_do_arp_validate_only(bond) && is_arp) || |
2299 | !slave_do_arp_validate_only(bond)) | |
49f17de7 | 2300 | slave->last_rx = jiffies; |
13a8e0c8 | 2301 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2302 | } else if (!is_arp) { |
2303 | return RX_HANDLER_ANOTHER; | |
2304 | } | |
2c146102 | 2305 | |
de063b70 | 2306 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2307 | |
76444f50 VF |
2308 | netdev_dbg(bond->dev, "bond_arp_rcv: skb->dev %s\n", |
2309 | skb->dev->name); | |
f5b2b966 | 2310 | |
de063b70 ED |
2311 | if (alen > skb_headlen(skb)) { |
2312 | arp = kmalloc(alen, GFP_ATOMIC); | |
2313 | if (!arp) | |
2314 | goto out_unlock; | |
2315 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2316 | goto out_unlock; | |
2317 | } | |
f5b2b966 | 2318 | |
3aba891d | 2319 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2320 | skb->pkt_type == PACKET_OTHERHOST || |
2321 | skb->pkt_type == PACKET_LOOPBACK || | |
2322 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2323 | arp->ar_pro != htons(ETH_P_IP) || | |
2324 | arp->ar_pln != 4) | |
2325 | goto out_unlock; | |
2326 | ||
2327 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2328 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2329 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2330 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2331 | memcpy(&tip, arp_ptr, 4); |
2332 | ||
76444f50 VF |
2333 | netdev_dbg(bond->dev, "bond_arp_rcv: %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
2334 | slave->dev->name, bond_slave_state(slave), | |
2335 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2336 | &sip, &tip); | |
f5b2b966 | 2337 | |
010d3c39 VF |
2338 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
2339 | ||
f5b2b966 JV |
2340 | /* |
2341 | * Backup slaves won't see the ARP reply, but do come through | |
2342 | * here for each ARP probe (so we swap the sip/tip to validate | |
2343 | * the probe). In a "redundant switch, common router" type of | |
2344 | * configuration, the ARP probe will (hopefully) travel from | |
2345 | * the active, through one switch, the router, then the other | |
2346 | * switch before reaching the backup. | |
aeea64ac VF |
2347 | * |
2348 | * We 'trust' the arp requests if there is an active slave and | |
2349 | * it received valid arp reply(s) after it became active. This | |
2350 | * is done to avoid endless looping when we can't reach the | |
2351 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2352 | */ |
010d3c39 | 2353 | |
e30bc066 | 2354 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2355 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
2356 | else if (curr_active_slave && |
2357 | time_after(slave_last_rx(bond, curr_active_slave), | |
2358 | curr_active_slave->last_link_up)) | |
f5b2b966 JV |
2359 | bond_validate_arp(bond, slave, tip, sip); |
2360 | ||
2361 | out_unlock: | |
de063b70 ED |
2362 | if (arp != (struct arphdr *)skb->data) |
2363 | kfree(arp); | |
13a8e0c8 | 2364 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2365 | } |
2366 | ||
e7f63f1d VF |
2367 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2368 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2369 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2370 | */ | |
2371 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2372 | int mod) | |
2373 | { | |
2374 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2375 | ||
2376 | return time_in_range(jiffies, | |
2377 | last_act - delta_in_ticks, | |
2378 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2379 | } | |
2380 | ||
1da177e4 LT |
2381 | /* |
2382 | * this function is called regularly to monitor each slave's link | |
2383 | * ensuring that traffic is being sent and received when arp monitoring | |
2384 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2385 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2386 | * arp monitoring in active backup mode. | |
2387 | */ | |
6da67d26 | 2388 | static void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2389 | { |
1b76b316 JV |
2390 | struct bonding *bond = container_of(work, struct bonding, |
2391 | arp_work.work); | |
1da177e4 | 2392 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2393 | struct list_head *iter; |
6fde8f03 | 2394 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 2395 | |
1f2cd845 | 2396 | if (!bond_has_slaves(bond)) |
1da177e4 | 2397 | goto re_arm; |
1da177e4 | 2398 | |
2e52f4fe | 2399 | rcu_read_lock(); |
2400 | ||
4740d638 | 2401 | oldcurrent = rcu_dereference(bond->curr_active_slave); |
1da177e4 LT |
2402 | /* see if any of the previous devices are up now (i.e. they have |
2403 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
2404 | * the picture unless it is null. also, slave->last_link_up is not |
2405 | * needed here because we send an arp on each slave and give a slave | |
2406 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
2407 | * TODO: what about up/down delay in arp mode? it wasn't here before |
2408 | * so it can wait | |
2409 | */ | |
2e52f4fe | 2410 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2411 | unsigned long trans_start = dev_trans_start(slave->dev); |
2412 | ||
1da177e4 | 2413 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2414 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 2415 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 LT |
2416 | |
2417 | slave->link = BOND_LINK_UP; | |
6fde8f03 | 2418 | slave_state_changed = 1; |
1da177e4 LT |
2419 | |
2420 | /* primary_slave has no meaning in round-robin | |
2421 | * mode. the window of a slave being up and | |
2422 | * curr_active_slave being null after enslaving | |
2423 | * is closed. | |
2424 | */ | |
2425 | if (!oldcurrent) { | |
76444f50 VF |
2426 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2427 | slave->dev->name); | |
1da177e4 LT |
2428 | do_failover = 1; |
2429 | } else { | |
76444f50 VF |
2430 | netdev_info(bond->dev, "interface %s is now up\n", |
2431 | slave->dev->name); | |
1da177e4 LT |
2432 | } |
2433 | } | |
2434 | } else { | |
2435 | /* slave->link == BOND_LINK_UP */ | |
2436 | ||
2437 | /* not all switches will respond to an arp request | |
2438 | * when the source ip is 0, so don't take the link down | |
2439 | * if we don't know our ip yet | |
2440 | */ | |
e7f63f1d | 2441 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 2442 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 LT |
2443 | |
2444 | slave->link = BOND_LINK_DOWN; | |
6fde8f03 | 2445 | slave_state_changed = 1; |
1da177e4 | 2446 | |
3d632c3f | 2447 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2448 | slave->link_failure_count++; |
1da177e4 | 2449 | |
76444f50 VF |
2450 | netdev_info(bond->dev, "interface %s is now down\n", |
2451 | slave->dev->name); | |
1da177e4 | 2452 | |
3d632c3f | 2453 | if (slave == oldcurrent) |
1da177e4 | 2454 | do_failover = 1; |
1da177e4 LT |
2455 | } |
2456 | } | |
2457 | ||
2458 | /* note: if switch is in round-robin mode, all links | |
2459 | * must tx arp to ensure all links rx an arp - otherwise | |
2460 | * links may oscillate or not come up at all; if switch is | |
2461 | * in something like xor mode, there is nothing we can | |
2462 | * do - all replies will be rx'ed on same link causing slaves | |
2463 | * to be unstable during low/no traffic periods | |
2464 | */ | |
b6adc610 | 2465 | if (bond_slave_is_up(slave)) |
1da177e4 | 2466 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2467 | } |
2468 | ||
2e52f4fe | 2469 | rcu_read_unlock(); |
2470 | ||
6fde8f03 | 2471 | if (do_failover || slave_state_changed) { |
2e52f4fe | 2472 | if (!rtnl_trylock()) |
2473 | goto re_arm; | |
1da177e4 | 2474 | |
6fde8f03 DT |
2475 | if (slave_state_changed) { |
2476 | bond_slave_state_change(bond); | |
2477 | } else if (do_failover) { | |
6fde8f03 | 2478 | block_netpoll_tx(); |
6fde8f03 | 2479 | bond_select_active_slave(bond); |
6fde8f03 DT |
2480 | unblock_netpoll_tx(); |
2481 | } | |
2e52f4fe | 2482 | rtnl_unlock(); |
1da177e4 LT |
2483 | } |
2484 | ||
2485 | re_arm: | |
e6d265e8 | 2486 | if (bond->params.arp_interval) |
e7f63f1d VF |
2487 | queue_delayed_work(bond->wq, &bond->arp_work, |
2488 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2489 | } |
2490 | ||
2491 | /* | |
b2220cad JV |
2492 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2493 | * changes. Sets new_link in slaves to specify what action should take | |
2494 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2495 | * to link states must be committed. | |
2496 | * | |
eb9fa4b0 | 2497 | * Called with rcu_read_lock hold. |
1da177e4 | 2498 | */ |
e7f63f1d | 2499 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2500 | { |
def4460c | 2501 | unsigned long trans_start, last_rx; |
9caff1e7 | 2502 | struct list_head *iter; |
dec1e90e | 2503 | struct slave *slave; |
dec1e90e | 2504 | int commit = 0; |
da210f55 | 2505 | |
eb9fa4b0 | 2506 | bond_for_each_slave_rcu(bond, slave, iter) { |
b2220cad | 2507 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2508 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2509 | |
b2220cad | 2510 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2511 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2512 | slave->new_link = BOND_LINK_UP; |
2513 | commit++; | |
2514 | } | |
b2220cad JV |
2515 | continue; |
2516 | } | |
1da177e4 | 2517 | |
b2220cad JV |
2518 | /* |
2519 | * Give slaves 2*delta after being enslaved or made | |
2520 | * active. This avoids bouncing, as the last receive | |
2521 | * times need a full ARP monitor cycle to be updated. | |
2522 | */ | |
8e603460 | 2523 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
2524 | continue; |
2525 | ||
2526 | /* | |
2527 | * Backup slave is down if: | |
2528 | * - No current_arp_slave AND | |
2529 | * - more than 3*delta since last receive AND | |
2530 | * - the bond has an IP address | |
2531 | * | |
2532 | * Note: a non-null current_arp_slave indicates | |
2533 | * the curr_active_slave went down and we are | |
2534 | * searching for a new one; under this condition | |
2535 | * we only take the curr_active_slave down - this | |
2536 | * gives each slave a chance to tx/rx traffic | |
2537 | * before being taken out | |
2538 | */ | |
e30bc066 | 2539 | if (!bond_is_active_slave(slave) && |
85741718 | 2540 | !rcu_access_pointer(bond->current_arp_slave) && |
e7f63f1d | 2541 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2542 | slave->new_link = BOND_LINK_DOWN; |
2543 | commit++; | |
2544 | } | |
2545 | ||
2546 | /* | |
2547 | * Active slave is down if: | |
2548 | * - more than 2*delta since transmitting OR | |
2549 | * - (more than 2*delta since receive AND | |
2550 | * the bond has an IP address) | |
2551 | */ | |
cb32f2a0 | 2552 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2553 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2554 | (!bond_time_in_interval(bond, trans_start, 2) || |
2555 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2556 | slave->new_link = BOND_LINK_DOWN; |
2557 | commit++; | |
2558 | } | |
1da177e4 LT |
2559 | } |
2560 | ||
b2220cad JV |
2561 | return commit; |
2562 | } | |
1da177e4 | 2563 | |
b2220cad JV |
2564 | /* |
2565 | * Called to commit link state changes noted by inspection step of | |
2566 | * active-backup mode ARP monitor. | |
2567 | * | |
eb9fa4b0 | 2568 | * Called with RTNL hold. |
b2220cad | 2569 | */ |
e7f63f1d | 2570 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2571 | { |
cb32f2a0 | 2572 | unsigned long trans_start; |
9caff1e7 | 2573 | struct list_head *iter; |
dec1e90e | 2574 | struct slave *slave; |
1da177e4 | 2575 | |
9caff1e7 | 2576 | bond_for_each_slave(bond, slave, iter) { |
b2220cad JV |
2577 | switch (slave->new_link) { |
2578 | case BOND_LINK_NOCHANGE: | |
2579 | continue; | |
ff59c456 | 2580 | |
b2220cad | 2581 | case BOND_LINK_UP: |
cb32f2a0 | 2582 | trans_start = dev_trans_start(slave->dev); |
4740d638 ED |
2583 | if (rtnl_dereference(bond->curr_active_slave) != slave || |
2584 | (!rtnl_dereference(bond->curr_active_slave) && | |
e7f63f1d | 2585 | bond_time_in_interval(bond, trans_start, 1))) { |
85741718 ED |
2586 | struct slave *current_arp_slave; |
2587 | ||
2588 | current_arp_slave = rtnl_dereference(bond->current_arp_slave); | |
b2220cad | 2589 | slave->link = BOND_LINK_UP; |
85741718 | 2590 | if (current_arp_slave) { |
5a430974 | 2591 | bond_set_slave_inactive_flags( |
85741718 | 2592 | current_arp_slave, |
5e5b0665 | 2593 | BOND_SLAVE_NOTIFY_NOW); |
85741718 | 2594 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
5a430974 | 2595 | } |
b2220cad | 2596 | |
76444f50 VF |
2597 | netdev_info(bond->dev, "link status definitely up for interface %s\n", |
2598 | slave->dev->name); | |
b2220cad | 2599 | |
4740d638 | 2600 | if (!rtnl_dereference(bond->curr_active_slave) || |
059b47e8 | 2601 | slave == rtnl_dereference(bond->primary_slave)) |
b9f60253 | 2602 | goto do_failover; |
1da177e4 | 2603 | |
b9f60253 | 2604 | } |
1da177e4 | 2605 | |
b9f60253 | 2606 | continue; |
1da177e4 | 2607 | |
b2220cad JV |
2608 | case BOND_LINK_DOWN: |
2609 | if (slave->link_failure_count < UINT_MAX) | |
2610 | slave->link_failure_count++; | |
2611 | ||
2612 | slave->link = BOND_LINK_DOWN; | |
5e5b0665 | 2613 | bond_set_slave_inactive_flags(slave, |
2614 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 2615 | |
76444f50 VF |
2616 | netdev_info(bond->dev, "link status definitely down for interface %s, disabling it\n", |
2617 | slave->dev->name); | |
b2220cad | 2618 | |
4740d638 | 2619 | if (slave == rtnl_dereference(bond->curr_active_slave)) { |
85741718 | 2620 | RCU_INIT_POINTER(bond->current_arp_slave, NULL); |
b9f60253 | 2621 | goto do_failover; |
1da177e4 | 2622 | } |
b9f60253 JP |
2623 | |
2624 | continue; | |
b2220cad JV |
2625 | |
2626 | default: | |
76444f50 VF |
2627 | netdev_err(bond->dev, "impossible: new_link %d on slave %s\n", |
2628 | slave->new_link, slave->dev->name); | |
b9f60253 | 2629 | continue; |
1da177e4 | 2630 | } |
1da177e4 | 2631 | |
b9f60253 JP |
2632 | do_failover: |
2633 | ASSERT_RTNL(); | |
e843fa50 | 2634 | block_netpoll_tx(); |
b9f60253 | 2635 | bond_select_active_slave(bond); |
e843fa50 | 2636 | unblock_netpoll_tx(); |
b2220cad | 2637 | } |
1da177e4 | 2638 | |
b2220cad JV |
2639 | bond_set_carrier(bond); |
2640 | } | |
1da177e4 | 2641 | |
b2220cad JV |
2642 | /* |
2643 | * Send ARP probes for active-backup mode ARP monitor. | |
b0929915 | 2644 | * |
2645 | * Called with rcu_read_lock hold. | |
b2220cad | 2646 | */ |
f2ebd477 | 2647 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 2648 | { |
eb9fa4b0 | 2649 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 2650 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
2651 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
2652 | struct list_head *iter; |
2653 | bool found = false; | |
b0929915 | 2654 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 2655 | |
98b90f26 | 2656 | if (curr_arp_slave && curr_active_slave) |
76444f50 VF |
2657 | netdev_info(bond->dev, "PROBE: c_arp %s && cas %s BAD\n", |
2658 | curr_arp_slave->dev->name, | |
2659 | curr_active_slave->dev->name); | |
1da177e4 | 2660 | |
98b90f26 VF |
2661 | if (curr_active_slave) { |
2662 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 2663 | return should_notify_rtnl; |
b2220cad | 2664 | } |
1da177e4 | 2665 | |
b2220cad JV |
2666 | /* if we don't have a curr_active_slave, search for the next available |
2667 | * backup slave from the current_arp_slave and make it the candidate | |
2668 | * for becoming the curr_active_slave | |
2669 | */ | |
1da177e4 | 2670 | |
eb9fa4b0 | 2671 | if (!curr_arp_slave) { |
b0929915 | 2672 | curr_arp_slave = bond_first_slave_rcu(bond); |
2673 | if (!curr_arp_slave) | |
2674 | return should_notify_rtnl; | |
b2220cad | 2675 | } |
1da177e4 | 2676 | |
b0929915 | 2677 | bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER); |
059fe7a5 | 2678 | |
b0929915 | 2679 | bond_for_each_slave_rcu(bond, slave, iter) { |
b6adc610 | 2680 | if (!found && !before && bond_slave_is_up(slave)) |
4087df87 | 2681 | before = slave; |
1da177e4 | 2682 | |
b6adc610 | 2683 | if (found && !new_slave && bond_slave_is_up(slave)) |
4087df87 | 2684 | new_slave = slave; |
b2220cad JV |
2685 | /* if the link state is up at this point, we |
2686 | * mark it down - this can happen if we have | |
2687 | * simultaneous link failures and | |
2688 | * reselect_active_interface doesn't make this | |
2689 | * one the current slave so it is still marked | |
2690 | * up when it is actually down | |
1da177e4 | 2691 | */ |
b6adc610 | 2692 | if (!bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
b2220cad JV |
2693 | slave->link = BOND_LINK_DOWN; |
2694 | if (slave->link_failure_count < UINT_MAX) | |
2695 | slave->link_failure_count++; | |
1da177e4 | 2696 | |
5e5b0665 | 2697 | bond_set_slave_inactive_flags(slave, |
b0929915 | 2698 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 2699 | |
76444f50 VF |
2700 | netdev_info(bond->dev, "backup interface %s is now down\n", |
2701 | slave->dev->name); | |
1da177e4 | 2702 | } |
eb9fa4b0 | 2703 | if (slave == curr_arp_slave) |
4087df87 | 2704 | found = true; |
b2220cad | 2705 | } |
4087df87 VF |
2706 | |
2707 | if (!new_slave && before) | |
2708 | new_slave = before; | |
2709 | ||
b0929915 | 2710 | if (!new_slave) |
2711 | goto check_state; | |
4087df87 VF |
2712 | |
2713 | new_slave->link = BOND_LINK_BACK; | |
b0929915 | 2714 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 2715 | bond_arp_send_all(bond, new_slave); |
8e603460 | 2716 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 2717 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 2718 | |
b0929915 | 2719 | check_state: |
2720 | bond_for_each_slave_rcu(bond, slave, iter) { | |
2721 | if (slave->should_notify) { | |
2722 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; | |
2723 | break; | |
2724 | } | |
2725 | } | |
2726 | return should_notify_rtnl; | |
b2220cad | 2727 | } |
1da177e4 | 2728 | |
6da67d26 | 2729 | static void bond_activebackup_arp_mon(struct work_struct *work) |
b2220cad JV |
2730 | { |
2731 | struct bonding *bond = container_of(work, struct bonding, | |
2732 | arp_work.work); | |
b0929915 | 2733 | bool should_notify_peers = false; |
2734 | bool should_notify_rtnl = false; | |
1f2cd845 | 2735 | int delta_in_ticks; |
1da177e4 | 2736 | |
1f2cd845 DM |
2737 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
2738 | ||
2739 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
2740 | goto re_arm; |
2741 | ||
eb9fa4b0 | 2742 | rcu_read_lock(); |
b0929915 | 2743 | |
ad246c99 BH |
2744 | should_notify_peers = bond_should_notify_peers(bond); |
2745 | ||
b0929915 | 2746 | if (bond_ab_arp_inspect(bond)) { |
2747 | rcu_read_unlock(); | |
2748 | ||
1f2cd845 DM |
2749 | /* Race avoidance with bond_close flush of workqueue */ |
2750 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2751 | delta_in_ticks = 1; |
2752 | should_notify_peers = false; | |
2753 | goto re_arm; | |
2754 | } | |
b2220cad | 2755 | |
1f2cd845 | 2756 | bond_ab_arp_commit(bond); |
b0929915 | 2757 | |
1f2cd845 | 2758 | rtnl_unlock(); |
b0929915 | 2759 | rcu_read_lock(); |
1f2cd845 DM |
2760 | } |
2761 | ||
b0929915 | 2762 | should_notify_rtnl = bond_ab_arp_probe(bond); |
2763 | rcu_read_unlock(); | |
ad246c99 | 2764 | |
80b9d236 | 2765 | re_arm: |
2766 | if (bond->params.arp_interval) | |
1f2cd845 DM |
2767 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
2768 | ||
b0929915 | 2769 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
2770 | if (!rtnl_trylock()) |
2771 | return; | |
b0929915 | 2772 | |
2773 | if (should_notify_peers) | |
2774 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
2775 | bond->dev); | |
2776 | if (should_notify_rtnl) | |
2777 | bond_slave_state_notify(bond); | |
2778 | ||
1f2cd845 DM |
2779 | rtnl_unlock(); |
2780 | } | |
1da177e4 LT |
2781 | } |
2782 | ||
1da177e4 LT |
2783 | /*-------------------------- netdev event handling --------------------------*/ |
2784 | ||
2785 | /* | |
2786 | * Change device name | |
2787 | */ | |
2788 | static int bond_event_changename(struct bonding *bond) | |
2789 | { | |
1da177e4 LT |
2790 | bond_remove_proc_entry(bond); |
2791 | bond_create_proc_entry(bond); | |
7e083840 | 2792 | |
f073c7ca TI |
2793 | bond_debug_reregister(bond); |
2794 | ||
1da177e4 LT |
2795 | return NOTIFY_DONE; |
2796 | } | |
2797 | ||
3d632c3f SH |
2798 | static int bond_master_netdev_event(unsigned long event, |
2799 | struct net_device *bond_dev) | |
1da177e4 | 2800 | { |
454d7c9b | 2801 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2802 | |
2803 | switch (event) { | |
2804 | case NETDEV_CHANGENAME: | |
2805 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2806 | case NETDEV_UNREGISTER: |
2807 | bond_remove_proc_entry(event_bond); | |
2808 | break; | |
2809 | case NETDEV_REGISTER: | |
2810 | bond_create_proc_entry(event_bond); | |
2811 | break; | |
6c8d23f7 | 2812 | case NETDEV_NOTIFY_PEERS: |
2813 | if (event_bond->send_peer_notif) | |
2814 | event_bond->send_peer_notif--; | |
2815 | break; | |
1da177e4 LT |
2816 | default: |
2817 | break; | |
2818 | } | |
2819 | ||
2820 | return NOTIFY_DONE; | |
2821 | } | |
2822 | ||
3d632c3f SH |
2823 | static int bond_slave_netdev_event(unsigned long event, |
2824 | struct net_device *slave_dev) | |
1da177e4 | 2825 | { |
059b47e8 | 2826 | struct slave *slave = bond_slave_get_rtnl(slave_dev), *primary; |
6101391d | 2827 | struct bonding *bond; |
2828 | struct net_device *bond_dev; | |
471cb5a3 JP |
2829 | u32 old_speed; |
2830 | u8 old_duplex; | |
1da177e4 | 2831 | |
6101391d | 2832 | /* A netdev event can be generated while enslaving a device |
2833 | * before netdev_rx_handler_register is called in which case | |
2834 | * slave will be NULL | |
2835 | */ | |
2836 | if (!slave) | |
2837 | return NOTIFY_DONE; | |
2838 | bond_dev = slave->bond->dev; | |
2839 | bond = slave->bond; | |
059b47e8 | 2840 | primary = rtnl_dereference(bond->primary_slave); |
6101391d | 2841 | |
1da177e4 LT |
2842 | switch (event) { |
2843 | case NETDEV_UNREGISTER: | |
8d2ada77 | 2844 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
2845 | bond_release_and_destroy(bond_dev, slave_dev); |
2846 | else | |
2847 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 2848 | break; |
98f41f69 | 2849 | case NETDEV_UP: |
1da177e4 | 2850 | case NETDEV_CHANGE: |
471cb5a3 JP |
2851 | old_speed = slave->speed; |
2852 | old_duplex = slave->duplex; | |
17d04500 | 2853 | |
471cb5a3 | 2854 | bond_update_speed_duplex(slave); |
17d04500 | 2855 | |
01844098 | 2856 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
471cb5a3 JP |
2857 | if (old_speed != slave->speed) |
2858 | bond_3ad_adapter_speed_changed(slave); | |
2859 | if (old_duplex != slave->duplex) | |
2860 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 2861 | } |
1da177e4 LT |
2862 | break; |
2863 | case NETDEV_DOWN: | |
2864 | /* | |
2865 | * ... Or is it this? | |
2866 | */ | |
2867 | break; | |
2868 | case NETDEV_CHANGEMTU: | |
2869 | /* | |
2870 | * TODO: Should slaves be allowed to | |
2871 | * independently alter their MTU? For | |
2872 | * an active-backup bond, slaves need | |
2873 | * not be the same type of device, so | |
2874 | * MTUs may vary. For other modes, | |
2875 | * slaves arguably should have the | |
2876 | * same MTUs. To do this, we'd need to | |
2877 | * take over the slave's change_mtu | |
2878 | * function for the duration of their | |
2879 | * servitude. | |
2880 | */ | |
2881 | break; | |
2882 | case NETDEV_CHANGENAME: | |
3ec775b9 | 2883 | /* we don't care if we don't have primary set */ |
ec0865a9 | 2884 | if (!bond_uses_primary(bond) || |
3ec775b9 VF |
2885 | !bond->params.primary[0]) |
2886 | break; | |
2887 | ||
059b47e8 | 2888 | if (slave == primary) { |
3ec775b9 | 2889 | /* slave's name changed - he's no longer primary */ |
059b47e8 | 2890 | RCU_INIT_POINTER(bond->primary_slave, NULL); |
3ec775b9 VF |
2891 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { |
2892 | /* we have a new primary slave */ | |
059b47e8 | 2893 | rcu_assign_pointer(bond->primary_slave, slave); |
3ec775b9 VF |
2894 | } else { /* we didn't change primary - exit */ |
2895 | break; | |
2896 | } | |
2897 | ||
76444f50 | 2898 | netdev_info(bond->dev, "Primary slave changed to %s, reselecting active slave\n", |
059b47e8 | 2899 | primary ? slave_dev->name : "none"); |
f80889a5 | 2900 | |
2901 | block_netpoll_tx(); | |
3ec775b9 | 2902 | bond_select_active_slave(bond); |
f80889a5 | 2903 | unblock_netpoll_tx(); |
1da177e4 | 2904 | break; |
8531c5ff AK |
2905 | case NETDEV_FEAT_CHANGE: |
2906 | bond_compute_features(bond); | |
2907 | break; | |
4aa5dee4 JP |
2908 | case NETDEV_RESEND_IGMP: |
2909 | /* Propagate to master device */ | |
2910 | call_netdevice_notifiers(event, slave->bond->dev); | |
2911 | break; | |
1da177e4 LT |
2912 | default: |
2913 | break; | |
2914 | } | |
2915 | ||
2916 | return NOTIFY_DONE; | |
2917 | } | |
2918 | ||
2919 | /* | |
2920 | * bond_netdev_event: handle netdev notifier chain events. | |
2921 | * | |
2922 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 2923 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
2924 | * locks for us to safely manipulate the slave devices (RTNL lock, |
2925 | * dev_probe_lock). | |
2926 | */ | |
3d632c3f SH |
2927 | static int bond_netdev_event(struct notifier_block *this, |
2928 | unsigned long event, void *ptr) | |
1da177e4 | 2929 | { |
351638e7 | 2930 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 2931 | |
76444f50 | 2932 | netdev_dbg(event_dev, "event: %lx\n", event); |
1da177e4 | 2933 | |
0b680e75 JV |
2934 | if (!(event_dev->priv_flags & IFF_BONDING)) |
2935 | return NOTIFY_DONE; | |
2936 | ||
1da177e4 | 2937 | if (event_dev->flags & IFF_MASTER) { |
76444f50 | 2938 | netdev_dbg(event_dev, "IFF_MASTER\n"); |
1da177e4 LT |
2939 | return bond_master_netdev_event(event, event_dev); |
2940 | } | |
2941 | ||
2942 | if (event_dev->flags & IFF_SLAVE) { | |
76444f50 | 2943 | netdev_dbg(event_dev, "IFF_SLAVE\n"); |
1da177e4 LT |
2944 | return bond_slave_netdev_event(event, event_dev); |
2945 | } | |
2946 | ||
2947 | return NOTIFY_DONE; | |
2948 | } | |
2949 | ||
2950 | static struct notifier_block bond_netdev_notifier = { | |
2951 | .notifier_call = bond_netdev_event, | |
2952 | }; | |
2953 | ||
169a3e66 JV |
2954 | /*---------------------------- Hashing Policies -----------------------------*/ |
2955 | ||
32819dc1 NA |
2956 | /* L2 hash helper */ |
2957 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 | 2958 | { |
ce04d635 | 2959 | struct ethhdr *ep, hdr_tmp; |
6b923cb7 | 2960 | |
ce04d635 JX |
2961 | ep = skb_header_pointer(skb, 0, sizeof(hdr_tmp), &hdr_tmp); |
2962 | if (ep) | |
2963 | return ep->h_dest[5] ^ ep->h_source[5] ^ ep->h_proto; | |
6b923cb7 JE |
2964 | return 0; |
2965 | } | |
2966 | ||
32819dc1 NA |
2967 | /* Extract the appropriate headers based on bond's xmit policy */ |
2968 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
2969 | struct flow_keys *fk) | |
6f6652be | 2970 | { |
32819dc1 | 2971 | const struct ipv6hdr *iph6; |
4394542c | 2972 | const struct iphdr *iph; |
32819dc1 | 2973 | int noff, proto = -1; |
6b923cb7 | 2974 | |
32819dc1 NA |
2975 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) |
2976 | return skb_flow_dissect(skb, fk); | |
2977 | ||
2978 | fk->ports = 0; | |
2979 | noff = skb_network_offset(skb); | |
2980 | if (skb->protocol == htons(ETH_P_IP)) { | |
054bb880 | 2981 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
32819dc1 | 2982 | return false; |
6b923cb7 | 2983 | iph = ip_hdr(skb); |
32819dc1 NA |
2984 | fk->src = iph->saddr; |
2985 | fk->dst = iph->daddr; | |
2986 | noff += iph->ihl << 2; | |
2987 | if (!ip_is_fragment(iph)) | |
2988 | proto = iph->protocol; | |
2989 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
054bb880 | 2990 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6)))) |
32819dc1 NA |
2991 | return false; |
2992 | iph6 = ipv6_hdr(skb); | |
2993 | fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr); | |
2994 | fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr); | |
2995 | noff += sizeof(*iph6); | |
2996 | proto = iph6->nexthdr; | |
2997 | } else { | |
2998 | return false; | |
6f6652be | 2999 | } |
32819dc1 NA |
3000 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) |
3001 | fk->ports = skb_flow_get_ports(skb, noff, proto); | |
6f6652be | 3002 | |
32819dc1 | 3003 | return true; |
6f6652be JV |
3004 | } |
3005 | ||
32819dc1 NA |
3006 | /** |
3007 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3008 | * @bond: bonding device | |
3009 | * @skb: buffer to use for headers | |
32819dc1 NA |
3010 | * |
3011 | * This function will extract the necessary headers from the skb buffer and use | |
3012 | * them to generate a hash based on the xmit_policy set in the bonding device | |
169a3e66 | 3013 | */ |
ee62e868 | 3014 | u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb) |
169a3e66 | 3015 | { |
32819dc1 NA |
3016 | struct flow_keys flow; |
3017 | u32 hash; | |
4394542c | 3018 | |
32819dc1 NA |
3019 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3020 | !bond_flow_dissect(bond, skb, &flow)) | |
ee62e868 | 3021 | return bond_eth_hash(skb); |
169a3e66 | 3022 | |
32819dc1 NA |
3023 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
3024 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) | |
3025 | hash = bond_eth_hash(skb); | |
3026 | else | |
3027 | hash = (__force u32)flow.ports; | |
3028 | hash ^= (__force u32)flow.dst ^ (__force u32)flow.src; | |
3029 | hash ^= (hash >> 16); | |
3030 | hash ^= (hash >> 8); | |
3031 | ||
ee62e868 | 3032 | return hash; |
169a3e66 JV |
3033 | } |
3034 | ||
1da177e4 LT |
3035 | /*-------------------------- Device entry points ----------------------------*/ |
3036 | ||
fbb0c41b | 3037 | static void bond_work_init_all(struct bonding *bond) |
3038 | { | |
3039 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3040 | bond_resend_igmp_join_requests_delayed); | |
3041 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3042 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
01844098 | 3043 | if (BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
fbb0c41b | 3044 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); |
3045 | else | |
3046 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3047 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3048 | } | |
3049 | ||
3050 | static void bond_work_cancel_all(struct bonding *bond) | |
3051 | { | |
3052 | cancel_delayed_work_sync(&bond->mii_work); | |
3053 | cancel_delayed_work_sync(&bond->arp_work); | |
3054 | cancel_delayed_work_sync(&bond->alb_work); | |
3055 | cancel_delayed_work_sync(&bond->ad_work); | |
3056 | cancel_delayed_work_sync(&bond->mcast_work); | |
3057 | } | |
3058 | ||
1da177e4 LT |
3059 | static int bond_open(struct net_device *bond_dev) |
3060 | { | |
454d7c9b | 3061 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3062 | struct list_head *iter; |
ba3211cc | 3063 | struct slave *slave; |
1da177e4 | 3064 | |
ba3211cc | 3065 | /* reset slave->backup and slave->inactive */ |
0965a1f3 | 3066 | if (bond_has_slaves(bond)) { |
9caff1e7 | 3067 | bond_for_each_slave(bond, slave, iter) { |
4740d638 ED |
3068 | if (bond_uses_primary(bond) && |
3069 | slave != rcu_access_pointer(bond->curr_active_slave)) { | |
5e5b0665 | 3070 | bond_set_slave_inactive_flags(slave, |
3071 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc | 3072 | } else { |
5e5b0665 | 3073 | bond_set_slave_active_flags(slave, |
3074 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3075 | } |
3076 | } | |
ba3211cc | 3077 | } |
ba3211cc | 3078 | |
fbb0c41b | 3079 | bond_work_init_all(bond); |
5a37e8ca | 3080 | |
58402054 | 3081 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3082 | /* bond_alb_initialize must be called before the timer |
3083 | * is started. | |
3084 | */ | |
01844098 | 3085 | if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) |
b473946a | 3086 | return -ENOMEM; |
e9f0fb88 MB |
3087 | if (bond->params.tlb_dynamic_lb) |
3088 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3089 | } |
3090 | ||
fbb0c41b | 3091 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3092 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3093 | |
3094 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3095 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3096 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3097 | } |
3098 | ||
01844098 | 3099 | if (BOND_MODE(bond) == BOND_MODE_8023AD) { |
1b76b316 | 3100 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3101 | /* register to receive LACPDUs */ |
3aba891d | 3102 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3103 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3104 | } |
3105 | ||
3106 | return 0; | |
3107 | } | |
3108 | ||
3109 | static int bond_close(struct net_device *bond_dev) | |
3110 | { | |
454d7c9b | 3111 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3112 | |
fbb0c41b | 3113 | bond_work_cancel_all(bond); |
6c8d23f7 | 3114 | bond->send_peer_notif = 0; |
ee8487c0 | 3115 | if (bond_is_lb(bond)) |
1da177e4 | 3116 | bond_alb_deinitialize(bond); |
3aba891d | 3117 | bond->recv_probe = NULL; |
1da177e4 LT |
3118 | |
3119 | return 0; | |
3120 | } | |
3121 | ||
28172739 ED |
3122 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3123 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3124 | { |
454d7c9b | 3125 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3126 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3127 | struct list_head *iter; |
1da177e4 | 3128 | struct slave *slave; |
1da177e4 | 3129 | |
28172739 | 3130 | memset(stats, 0, sizeof(*stats)); |
1da177e4 | 3131 | |
9caff1e7 | 3132 | bond_for_each_slave(bond, slave, iter) { |
be1f3c2c | 3133 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3134 | dev_get_stats(slave->dev, &temp); |
3135 | ||
3136 | stats->rx_packets += sstats->rx_packets; | |
3137 | stats->rx_bytes += sstats->rx_bytes; | |
3138 | stats->rx_errors += sstats->rx_errors; | |
3139 | stats->rx_dropped += sstats->rx_dropped; | |
3140 | ||
3141 | stats->tx_packets += sstats->tx_packets; | |
3142 | stats->tx_bytes += sstats->tx_bytes; | |
3143 | stats->tx_errors += sstats->tx_errors; | |
3144 | stats->tx_dropped += sstats->tx_dropped; | |
3145 | ||
3146 | stats->multicast += sstats->multicast; | |
3147 | stats->collisions += sstats->collisions; | |
3148 | ||
3149 | stats->rx_length_errors += sstats->rx_length_errors; | |
3150 | stats->rx_over_errors += sstats->rx_over_errors; | |
3151 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3152 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3153 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3154 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3155 | ||
3156 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3157 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3158 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3159 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3160 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb | 3161 | } |
1da177e4 LT |
3162 | |
3163 | return stats; | |
3164 | } | |
3165 | ||
3166 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3167 | { | |
080a06e1 | 3168 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3169 | struct net_device *slave_dev = NULL; |
3170 | struct ifbond k_binfo; | |
3171 | struct ifbond __user *u_binfo = NULL; | |
3172 | struct ifslave k_sinfo; | |
3173 | struct ifslave __user *u_sinfo = NULL; | |
3174 | struct mii_ioctl_data *mii = NULL; | |
d1fbd3ed | 3175 | struct bond_opt_value newval; |
387ff911 | 3176 | struct net *net; |
1da177e4 LT |
3177 | int res = 0; |
3178 | ||
76444f50 | 3179 | netdev_dbg(bond_dev, "bond_ioctl: cmd=%d\n", cmd); |
1da177e4 LT |
3180 | |
3181 | switch (cmd) { | |
1da177e4 LT |
3182 | case SIOCGMIIPHY: |
3183 | mii = if_mii(ifr); | |
3d632c3f | 3184 | if (!mii) |
1da177e4 | 3185 | return -EINVAL; |
3d632c3f | 3186 | |
1da177e4 LT |
3187 | mii->phy_id = 0; |
3188 | /* Fall Through */ | |
3189 | case SIOCGMIIREG: | |
3190 | /* | |
3191 | * We do this again just in case we were called by SIOCGMIIREG | |
3192 | * instead of SIOCGMIIPHY. | |
3193 | */ | |
3194 | mii = if_mii(ifr); | |
3d632c3f | 3195 | if (!mii) |
1da177e4 | 3196 | return -EINVAL; |
3d632c3f | 3197 | |
1da177e4 | 3198 | if (mii->reg_num == 1) { |
1da177e4 | 3199 | mii->val_out = 0; |
3d632c3f | 3200 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3201 | mii->val_out = BMSR_LSTATUS; |
1da177e4 LT |
3202 | } |
3203 | ||
3204 | return 0; | |
3205 | case BOND_INFO_QUERY_OLD: | |
3206 | case SIOCBONDINFOQUERY: | |
3207 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3208 | ||
3d632c3f | 3209 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3210 | return -EFAULT; |
1da177e4 LT |
3211 | |
3212 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3213 | if (res == 0 && |
3214 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3215 | return -EFAULT; | |
1da177e4 LT |
3216 | |
3217 | return res; | |
3218 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3219 | case SIOCBONDSLAVEINFOQUERY: | |
3220 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3221 | ||
3d632c3f | 3222 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3223 | return -EFAULT; |
1da177e4 LT |
3224 | |
3225 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3226 | if (res == 0 && |
3227 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3228 | return -EFAULT; | |
1da177e4 LT |
3229 | |
3230 | return res; | |
3231 | default: | |
3232 | /* Go on */ | |
3233 | break; | |
3234 | } | |
3235 | ||
387ff911 G |
3236 | net = dev_net(bond_dev); |
3237 | ||
3238 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3239 | return -EPERM; |
1da177e4 | 3240 | |
0917b933 | 3241 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3242 | |
76444f50 | 3243 | netdev_dbg(bond_dev, "slave_dev=%p:\n", slave_dev); |
1da177e4 | 3244 | |
3d632c3f | 3245 | if (!slave_dev) |
0917b933 | 3246 | return -ENODEV; |
1da177e4 | 3247 | |
76444f50 | 3248 | netdev_dbg(bond_dev, "slave_dev->name=%s:\n", slave_dev->name); |
0917b933 YX |
3249 | switch (cmd) { |
3250 | case BOND_ENSLAVE_OLD: | |
3251 | case SIOCBONDENSLAVE: | |
3252 | res = bond_enslave(bond_dev, slave_dev); | |
3253 | break; | |
3254 | case BOND_RELEASE_OLD: | |
3255 | case SIOCBONDRELEASE: | |
3256 | res = bond_release(bond_dev, slave_dev); | |
3257 | break; | |
3258 | case BOND_SETHWADDR_OLD: | |
3259 | case SIOCBONDSETHWADDR: | |
3260 | bond_set_dev_addr(bond_dev, slave_dev); | |
3261 | res = 0; | |
3262 | break; | |
3263 | case BOND_CHANGE_ACTIVE_OLD: | |
3264 | case SIOCBONDCHANGEACTIVE: | |
d1fbd3ed NA |
3265 | bond_opt_initstr(&newval, slave_dev->name); |
3266 | res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval); | |
0917b933 YX |
3267 | break; |
3268 | default: | |
3269 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3270 | } |
3271 | ||
1da177e4 LT |
3272 | return res; |
3273 | } | |
3274 | ||
d03462b9 | 3275 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3276 | { |
454d7c9b | 3277 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3278 | |
d03462b9 JP |
3279 | if (change & IFF_PROMISC) |
3280 | bond_set_promiscuity(bond, | |
3281 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3282 | |
d03462b9 JP |
3283 | if (change & IFF_ALLMULTI) |
3284 | bond_set_allmulti(bond, | |
3285 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3286 | } | |
1da177e4 | 3287 | |
303d1cbf | 3288 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3289 | { |
3290 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3291 | struct list_head *iter; |
303d1cbf | 3292 | struct slave *slave; |
1da177e4 | 3293 | |
80ee5ad2 | 3294 | |
b3241870 | 3295 | rcu_read_lock(); |
ec0865a9 | 3296 | if (bond_uses_primary(bond)) { |
b3241870 | 3297 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3298 | if (slave) { |
3299 | dev_uc_sync(slave->dev, bond_dev); | |
3300 | dev_mc_sync(slave->dev, bond_dev); | |
3301 | } | |
303d1cbf | 3302 | } else { |
b3241870 | 3303 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3304 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3305 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3306 | } | |
1da177e4 | 3307 | } |
b3241870 | 3308 | rcu_read_unlock(); |
1da177e4 LT |
3309 | } |
3310 | ||
234bcf8a | 3311 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3312 | { |
234bcf8a | 3313 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3314 | const struct net_device_ops *slave_ops; |
3315 | struct neigh_parms parms; | |
dec1e90e | 3316 | struct slave *slave; |
234bcf8a SP |
3317 | int ret; |
3318 | ||
dec1e90e | 3319 | slave = bond_first_slave(bond); |
234bcf8a SP |
3320 | if (!slave) |
3321 | return 0; | |
234bcf8a | 3322 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3323 | if (!slave_ops->ndo_neigh_setup) |
3324 | return 0; | |
3325 | ||
3326 | parms.neigh_setup = NULL; | |
3327 | parms.neigh_cleanup = NULL; | |
3328 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3329 | if (ret) | |
3330 | return ret; | |
3331 | ||
3332 | /* | |
3333 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3334 | * after the last slave has been detached. Assumes that all slaves | |
3335 | * utilize the same neigh_cleanup (true at this writing as only user | |
3336 | * is ipoib). | |
3337 | */ | |
3338 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3339 | ||
3340 | if (!parms.neigh_setup) | |
3341 | return 0; | |
3342 | ||
3343 | return parms.neigh_setup(n); | |
3344 | } | |
3345 | ||
3346 | /* | |
3347 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3348 | * slave exists. So we must declare proxy setup function which will | |
3349 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3350 | * |
3351 | * It's also called by master devices (such as vlans) to setup their | |
3352 | * underlying devices. In that case - do nothing, we're already set up from | |
3353 | * our init. | |
234bcf8a SP |
3354 | */ |
3355 | static int bond_neigh_setup(struct net_device *dev, | |
3356 | struct neigh_parms *parms) | |
3357 | { | |
9918d5bf VF |
3358 | /* modify only our neigh_parms */ |
3359 | if (parms->dev == dev) | |
3360 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3361 | |
00829823 SH |
3362 | return 0; |
3363 | } | |
3364 | ||
1da177e4 LT |
3365 | /* |
3366 | * Change the MTU of all of a master's slaves to match the master | |
3367 | */ | |
3368 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3369 | { | |
454d7c9b | 3370 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3371 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3372 | struct list_head *iter; |
1da177e4 | 3373 | int res = 0; |
1da177e4 | 3374 | |
76444f50 | 3375 | netdev_dbg(bond_dev, "bond=%p, new_mtu=%d\n", bond, new_mtu); |
1da177e4 | 3376 | |
9caff1e7 | 3377 | bond_for_each_slave(bond, slave, iter) { |
76444f50 VF |
3378 | netdev_dbg(bond_dev, "s %p c_m %p\n", |
3379 | slave, slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3380 | |
1da177e4 LT |
3381 | res = dev_set_mtu(slave->dev, new_mtu); |
3382 | ||
3383 | if (res) { | |
3384 | /* If we failed to set the slave's mtu to the new value | |
3385 | * we must abort the operation even in ACTIVE_BACKUP | |
3386 | * mode, because if we allow the backup slaves to have | |
3387 | * different mtu values than the active slave we'll | |
3388 | * need to change their mtu when doing a failover. That | |
3389 | * means changing their mtu from timer context, which | |
3390 | * is probably not a good idea. | |
3391 | */ | |
76444f50 VF |
3392 | netdev_dbg(bond_dev, "err %d %s\n", res, |
3393 | slave->dev->name); | |
1da177e4 LT |
3394 | goto unwind; |
3395 | } | |
3396 | } | |
3397 | ||
3398 | bond_dev->mtu = new_mtu; | |
3399 | ||
3400 | return 0; | |
3401 | ||
3402 | unwind: | |
3403 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3404 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3405 | int tmp_res; |
3406 | ||
81f23b13 VF |
3407 | if (rollback_slave == slave) |
3408 | break; | |
3409 | ||
3410 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3411 | if (tmp_res) { |
76444f50 VF |
3412 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3413 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3414 | } |
3415 | } | |
3416 | ||
3417 | return res; | |
3418 | } | |
3419 | ||
3420 | /* | |
3421 | * Change HW address | |
3422 | * | |
3423 | * Note that many devices must be down to change the HW address, and | |
3424 | * downing the master releases all slaves. We can make bonds full of | |
3425 | * bonding devices to test this, however. | |
3426 | */ | |
3427 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3428 | { | |
454d7c9b | 3429 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3430 | struct slave *slave, *rollback_slave; |
1da177e4 | 3431 | struct sockaddr *sa = addr, tmp_sa; |
9caff1e7 | 3432 | struct list_head *iter; |
1da177e4 | 3433 | int res = 0; |
1da177e4 | 3434 | |
01844098 | 3435 | if (BOND_MODE(bond) == BOND_MODE_ALB) |
eb7cc59a SH |
3436 | return bond_alb_set_mac_address(bond_dev, addr); |
3437 | ||
3438 | ||
76444f50 | 3439 | netdev_dbg(bond_dev, "bond=%p\n", bond); |
1da177e4 | 3440 | |
1b5acd29 JV |
3441 | /* If fail_over_mac is enabled, do nothing and return success. |
3442 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3443 | */ |
cc689aaa | 3444 | if (bond->params.fail_over_mac && |
01844098 | 3445 | BOND_MODE(bond) == BOND_MODE_ACTIVEBACKUP) |
dd957c57 | 3446 | return 0; |
2ab82852 | 3447 | |
3d632c3f | 3448 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3449 | return -EADDRNOTAVAIL; |
1da177e4 | 3450 | |
9caff1e7 | 3451 | bond_for_each_slave(bond, slave, iter) { |
76444f50 | 3452 | netdev_dbg(bond_dev, "slave %p %s\n", slave, slave->dev->name); |
1da177e4 LT |
3453 | res = dev_set_mac_address(slave->dev, addr); |
3454 | if (res) { | |
3455 | /* TODO: consider downing the slave | |
3456 | * and retry ? | |
3457 | * User should expect communications | |
3458 | * breakage anyway until ARP finish | |
3459 | * updating, so... | |
3460 | */ | |
76444f50 | 3461 | netdev_dbg(bond_dev, "err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3462 | goto unwind; |
3463 | } | |
3464 | } | |
3465 | ||
3466 | /* success */ | |
3467 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3468 | return 0; | |
3469 | ||
3470 | unwind: | |
3471 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3472 | tmp_sa.sa_family = bond_dev->type; | |
3473 | ||
3474 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3475 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3476 | int tmp_res; |
3477 | ||
81f23b13 VF |
3478 | if (rollback_slave == slave) |
3479 | break; | |
3480 | ||
3481 | tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa); | |
1da177e4 | 3482 | if (tmp_res) { |
76444f50 VF |
3483 | netdev_dbg(bond_dev, "unwind err %d dev %s\n", |
3484 | tmp_res, rollback_slave->dev->name); | |
1da177e4 LT |
3485 | } |
3486 | } | |
3487 | ||
3488 | return res; | |
3489 | } | |
3490 | ||
15077228 NA |
3491 | /** |
3492 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3493 | * @bond: bonding device that is transmitting | |
3494 | * @skb: buffer to transmit | |
3495 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3496 | * | |
3497 | * This function tries to transmit through slave with slave_id but in case | |
3498 | * it fails, it tries to find the first available slave for transmission. | |
3499 | * The skb is consumed in all cases, thus the function is void. | |
3500 | */ | |
da131ddb | 3501 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3502 | { |
9caff1e7 | 3503 | struct list_head *iter; |
15077228 NA |
3504 | struct slave *slave; |
3505 | int i = slave_id; | |
3506 | ||
3507 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3508 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 | 3509 | if (--i < 0) { |
891ab54d | 3510 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3511 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3512 | return; | |
3513 | } | |
3514 | } | |
3515 | } | |
3516 | ||
3517 | /* Here we start from the first slave up to slave_id */ | |
3518 | i = slave_id; | |
9caff1e7 | 3519 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3520 | if (--i < 0) |
3521 | break; | |
891ab54d | 3522 | if (bond_slave_can_tx(slave)) { |
15077228 NA |
3523 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3524 | return; | |
3525 | } | |
3526 | } | |
3527 | /* no slave that can tx has been found */ | |
2bb77ab4 | 3528 | dev_kfree_skb_any(skb); |
15077228 NA |
3529 | } |
3530 | ||
73958329 NA |
3531 | /** |
3532 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3533 | * @bond: bonding device to use | |
3534 | * | |
3535 | * Based on the value of the bonding device's packets_per_slave parameter | |
3536 | * this function generates a slave id, which is usually used as the next | |
3537 | * slave to transmit through. | |
3538 | */ | |
3539 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3540 | { | |
73958329 | 3541 | u32 slave_id; |
809fa972 HFS |
3542 | struct reciprocal_value reciprocal_packets_per_slave; |
3543 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
3544 | |
3545 | switch (packets_per_slave) { | |
3546 | case 0: | |
3547 | slave_id = prandom_u32(); | |
3548 | break; | |
3549 | case 1: | |
3550 | slave_id = bond->rr_tx_counter; | |
3551 | break; | |
3552 | default: | |
809fa972 HFS |
3553 | reciprocal_packets_per_slave = |
3554 | bond->params.reciprocal_packets_per_slave; | |
73958329 | 3555 | slave_id = reciprocal_divide(bond->rr_tx_counter, |
809fa972 | 3556 | reciprocal_packets_per_slave); |
73958329 NA |
3557 | break; |
3558 | } | |
3559 | bond->rr_tx_counter++; | |
3560 | ||
3561 | return slave_id; | |
3562 | } | |
3563 | ||
1da177e4 LT |
3564 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3565 | { | |
454d7c9b | 3566 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3567 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3568 | struct slave *slave; |
73958329 | 3569 | u32 slave_id; |
1da177e4 | 3570 | |
73958329 | 3571 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3572 | * default for sending IGMP traffic. For failover purposes one |
3573 | * needs to maintain some consistency for the interface that will | |
3574 | * send the join/membership reports. The curr_active_slave found | |
3575 | * will send all of this type of traffic. | |
cf5f9044 | 3576 | */ |
15077228 | 3577 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3578 | slave = rcu_dereference(bond->curr_active_slave); |
891ab54d | 3579 | if (slave && bond_slave_can_tx(slave)) |
15077228 NA |
3580 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3581 | else | |
3582 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3583 | } else { |
73958329 NA |
3584 | slave_id = bond_rr_gen_slave_id(bond); |
3585 | bond_xmit_slave_id(bond, skb, slave_id % bond->slave_cnt); | |
1da177e4 | 3586 | } |
0693e88e | 3587 | |
ec634fe3 | 3588 | return NETDEV_TX_OK; |
1da177e4 LT |
3589 | } |
3590 | ||
3591 | /* | |
3592 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3593 | * the bond has a usable interface. | |
3594 | */ | |
3595 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3596 | { | |
454d7c9b | 3597 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3598 | struct slave *slave; |
1da177e4 | 3599 | |
278b2083 | 3600 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3601 | if (slave) |
15077228 NA |
3602 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3603 | else | |
2bb77ab4 | 3604 | dev_kfree_skb_any(skb); |
0693e88e | 3605 | |
ec634fe3 | 3606 | return NETDEV_TX_OK; |
1da177e4 LT |
3607 | } |
3608 | ||
32819dc1 | 3609 | /* In bond_xmit_xor() , we determine the output device by using a pre- |
169a3e66 JV |
3610 | * determined xmit_hash_policy(), If the selected device is not enabled, |
3611 | * find the next active slave. | |
1da177e4 LT |
3612 | */ |
3613 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3614 | { | |
454d7c9b | 3615 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3616 | |
ee62e868 | 3617 | bond_xmit_slave_id(bond, skb, bond_xmit_hash(bond, skb) % bond->slave_cnt); |
0693e88e | 3618 | |
ec634fe3 | 3619 | return NETDEV_TX_OK; |
1da177e4 LT |
3620 | } |
3621 | ||
78a646ce | 3622 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3623 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3624 | { | |
454d7c9b | 3625 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3626 | struct slave *slave = NULL; |
9caff1e7 | 3627 | struct list_head *iter; |
1da177e4 | 3628 | |
9caff1e7 | 3629 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
3630 | if (bond_is_last_slave(bond, slave)) |
3631 | break; | |
b6adc610 | 3632 | if (bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) { |
78a646ce | 3633 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); |
1da177e4 | 3634 | |
78a646ce | 3635 | if (!skb2) { |
9152e26d | 3636 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
3637 | bond_dev->name, __func__); | |
78a646ce | 3638 | continue; |
1da177e4 | 3639 | } |
78a646ce NA |
3640 | /* bond_dev_queue_xmit always returns 0 */ |
3641 | bond_dev_queue_xmit(bond, skb2, slave->dev); | |
1da177e4 LT |
3642 | } |
3643 | } | |
b6adc610 | 3644 | if (slave && bond_slave_is_up(slave) && slave->link == BOND_LINK_UP) |
78a646ce NA |
3645 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3646 | else | |
2bb77ab4 | 3647 | dev_kfree_skb_any(skb); |
3d632c3f | 3648 | |
ec634fe3 | 3649 | return NETDEV_TX_OK; |
1da177e4 LT |
3650 | } |
3651 | ||
3652 | /*------------------------- Device initialization ---------------------------*/ | |
3653 | ||
bb1d9123 AG |
3654 | /* |
3655 | * Lookup the slave that corresponds to a qid | |
3656 | */ | |
3657 | static inline int bond_slave_override(struct bonding *bond, | |
3658 | struct sk_buff *skb) | |
3659 | { | |
bb1d9123 | 3660 | struct slave *slave = NULL; |
9caff1e7 | 3661 | struct list_head *iter; |
bb1d9123 | 3662 | |
0693e88e MM |
3663 | if (!skb->queue_mapping) |
3664 | return 1; | |
bb1d9123 AG |
3665 | |
3666 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 3667 | bond_for_each_slave_rcu(bond, slave, iter) { |
3668 | if (slave->queue_id == skb->queue_mapping) { | |
891ab54d | 3669 | if (bond_slave_can_tx(slave)) { |
3900f290 | 3670 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3671 | return 0; | |
3672 | } | |
3673 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
3674 | break; |
3675 | } | |
3676 | } | |
3677 | ||
3900f290 | 3678 | return 1; |
bb1d9123 AG |
3679 | } |
3680 | ||
374eeb5a | 3681 | |
f663dd9a | 3682 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 3683 | void *accel_priv, select_queue_fallback_t fallback) |
bb1d9123 AG |
3684 | { |
3685 | /* | |
3686 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 3687 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3688 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3689 | * way down to the bonding driver. | |
3690 | */ | |
fd0e435b PO |
3691 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3692 | ||
374eeb5a NH |
3693 | /* |
3694 | * Save the original txq to restore before passing to the driver | |
3695 | */ | |
df4ab5b3 | 3696 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3697 | |
fd0e435b | 3698 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3699 | do { |
fd0e435b | 3700 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3701 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3702 | } |
3703 | return txq; | |
bb1d9123 AG |
3704 | } |
3705 | ||
0693e88e | 3706 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3707 | { |
bb1d9123 AG |
3708 | struct bonding *bond = netdev_priv(dev); |
3709 | ||
d1e2e5cd VF |
3710 | if (bond_should_override_tx_queue(bond) && |
3711 | !bond_slave_override(bond, skb)) | |
3712 | return NETDEV_TX_OK; | |
00829823 | 3713 | |
01844098 | 3714 | switch (BOND_MODE(bond)) { |
00829823 SH |
3715 | case BOND_MODE_ROUNDROBIN: |
3716 | return bond_xmit_roundrobin(skb, dev); | |
3717 | case BOND_MODE_ACTIVEBACKUP: | |
3718 | return bond_xmit_activebackup(skb, dev); | |
3719 | case BOND_MODE_XOR: | |
3720 | return bond_xmit_xor(skb, dev); | |
3721 | case BOND_MODE_BROADCAST: | |
3722 | return bond_xmit_broadcast(skb, dev); | |
3723 | case BOND_MODE_8023AD: | |
3724 | return bond_3ad_xmit_xor(skb, dev); | |
3725 | case BOND_MODE_ALB: | |
00829823 | 3726 | return bond_alb_xmit(skb, dev); |
f05b42ea MB |
3727 | case BOND_MODE_TLB: |
3728 | return bond_tlb_xmit(skb, dev); | |
00829823 SH |
3729 | default: |
3730 | /* Should never happen, mode already checked */ | |
76444f50 | 3731 | netdev_err(dev, "Unknown bonding mode %d\n", BOND_MODE(bond)); |
00829823 | 3732 | WARN_ON_ONCE(1); |
2bb77ab4 | 3733 | dev_kfree_skb_any(skb); |
00829823 SH |
3734 | return NETDEV_TX_OK; |
3735 | } | |
3736 | } | |
3737 | ||
0693e88e MM |
3738 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3739 | { | |
3740 | struct bonding *bond = netdev_priv(dev); | |
3741 | netdev_tx_t ret = NETDEV_TX_OK; | |
3742 | ||
3743 | /* | |
3744 | * If we risk deadlock from transmitting this in the | |
3745 | * netpoll path, tell netpoll to queue the frame for later tx | |
3746 | */ | |
054bb880 | 3747 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
3748 | return NETDEV_TX_BUSY; |
3749 | ||
278b2083 | 3750 | rcu_read_lock(); |
0965a1f3 | 3751 | if (bond_has_slaves(bond)) |
0693e88e MM |
3752 | ret = __bond_start_xmit(skb, dev); |
3753 | else | |
2bb77ab4 | 3754 | dev_kfree_skb_any(skb); |
278b2083 | 3755 | rcu_read_unlock(); |
0693e88e MM |
3756 | |
3757 | return ret; | |
3758 | } | |
00829823 | 3759 | |
bb5b052f AG |
3760 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3761 | struct ethtool_cmd *ecmd) | |
3762 | { | |
3763 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3764 | unsigned long speed = 0; |
9caff1e7 | 3765 | struct list_head *iter; |
dec1e90e | 3766 | struct slave *slave; |
bb5b052f AG |
3767 | |
3768 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3769 | ecmd->port = PORT_OTHER; | |
3770 | ||
8557cd74 | 3771 | /* Since bond_slave_can_tx returns false for all inactive or down slaves, we |
bb5b052f AG |
3772 | * do not need to check mode. Though link speed might not represent |
3773 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3774 | * this is an accurate maximum. | |
3775 | */ | |
9caff1e7 | 3776 | bond_for_each_slave(bond, slave, iter) { |
8557cd74 | 3777 | if (bond_slave_can_tx(slave)) { |
bb5b052f AG |
3778 | if (slave->speed != SPEED_UNKNOWN) |
3779 | speed += slave->speed; | |
3780 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3781 | slave->duplex != DUPLEX_UNKNOWN) | |
3782 | ecmd->duplex = slave->duplex; | |
3783 | } | |
3784 | } | |
3785 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
dec1e90e | 3786 | |
bb5b052f AG |
3787 | return 0; |
3788 | } | |
3789 | ||
217df670 | 3790 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3791 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3792 | { |
7826d43f JP |
3793 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3794 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3795 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3796 | BOND_ABI_VERSION); | |
217df670 JV |
3797 | } |
3798 | ||
7282d491 | 3799 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3800 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3801 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3802 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3803 | }; |
3804 | ||
eb7cc59a | 3805 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3806 | .ndo_init = bond_init, |
9e71626c | 3807 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3808 | .ndo_open = bond_open, |
3809 | .ndo_stop = bond_close, | |
00829823 | 3810 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3811 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3812 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3813 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3814 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 3815 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 3816 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 3817 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 3818 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 3819 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 3820 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 3821 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 3822 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
3823 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
3824 | .ndo_poll_controller = bond_poll_controller, | |
3825 | #endif | |
9232ecca JP |
3826 | .ndo_add_slave = bond_enslave, |
3827 | .ndo_del_slave = bond_release, | |
b2a103e6 | 3828 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
3829 | }; |
3830 | ||
b3f92b63 DG |
3831 | static const struct device_type bond_type = { |
3832 | .name = "bond", | |
3833 | }; | |
3834 | ||
9e2e61fb AW |
3835 | static void bond_destructor(struct net_device *bond_dev) |
3836 | { | |
3837 | struct bonding *bond = netdev_priv(bond_dev); | |
3838 | if (bond->wq) | |
3839 | destroy_workqueue(bond->wq); | |
3840 | free_netdev(bond_dev); | |
3841 | } | |
3842 | ||
0a2a78c4 | 3843 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 3844 | { |
454d7c9b | 3845 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3846 | |
b7435628 | 3847 | spin_lock_init(&bond->mode_lock); |
d2991f75 | 3848 | bond->params = bonding_defaults; |
1da177e4 LT |
3849 | |
3850 | /* Initialize pointers */ | |
1da177e4 | 3851 | bond->dev = bond_dev; |
1da177e4 LT |
3852 | |
3853 | /* Initialize the device entry points */ | |
181470fc | 3854 | ether_setup(bond_dev); |
eb7cc59a | 3855 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 3856 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 3857 | |
9e2e61fb | 3858 | bond_dev->destructor = bond_destructor; |
1da177e4 | 3859 | |
b3f92b63 DG |
3860 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
3861 | ||
1da177e4 LT |
3862 | /* Initialize the device options */ |
3863 | bond_dev->tx_queue_len = 0; | |
3864 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
c565b488 | 3865 | bond_dev->priv_flags |= IFF_BONDING | IFF_UNICAST_FLT; |
550fd08c | 3866 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 3867 | |
932ff279 | 3868 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
3869 | * transmitting */ |
3870 | bond_dev->features |= NETIF_F_LLTX; | |
3871 | ||
3872 | /* By default, we declare the bond to be fully | |
3873 | * VLAN hardware accelerated capable. Special | |
3874 | * care is taken in the various xmit functions | |
3875 | * when there are slaves that are not hw accel | |
3876 | * capable | |
3877 | */ | |
1da177e4 | 3878 | |
f9399814 WC |
3879 | /* Don't allow bond devices to change network namespaces. */ |
3880 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
3881 | ||
b2a103e6 | 3882 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
3883 | NETIF_F_HW_VLAN_CTAG_TX | |
3884 | NETIF_F_HW_VLAN_CTAG_RX | | |
3885 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 3886 | |
34324dc2 | 3887 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
5a7baa78 | 3888 | bond_dev->hw_features |= NETIF_F_GSO_UDP_TUNNEL; |
b2a103e6 | 3889 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
3890 | } |
3891 | ||
c67dfb29 EB |
3892 | /* |
3893 | * Destroy a bonding device. | |
3894 | * Must be under rtnl_lock when this function is called. | |
3895 | */ | |
3896 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 3897 | { |
454d7c9b | 3898 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
3899 | struct list_head *iter; |
3900 | struct slave *slave; | |
a434e43f | 3901 | |
f6dc31a8 WC |
3902 | bond_netpoll_cleanup(bond_dev); |
3903 | ||
c67dfb29 | 3904 | /* Release the bonded slaves */ |
544a028e | 3905 | bond_for_each_slave(bond, slave, iter) |
dec1e90e | 3906 | __bond_release_one(bond_dev, slave->dev, true); |
76444f50 | 3907 | netdev_info(bond_dev, "Released all slaves\n"); |
c67dfb29 | 3908 | |
a434e43f JV |
3909 | list_del(&bond->bond_list); |
3910 | ||
f073c7ca | 3911 | bond_debug_unregister(bond); |
a434e43f JV |
3912 | } |
3913 | ||
1da177e4 LT |
3914 | /*------------------------- Module initialization ---------------------------*/ |
3915 | ||
1da177e4 LT |
3916 | static int bond_check_params(struct bond_params *params) |
3917 | { | |
5a5c5fd4 | 3918 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 3919 | struct bond_opt_value newval; |
3920 | const struct bond_opt_value *valptr; | |
8599b52e | 3921 | int arp_all_targets_value; |
f5b2b966 | 3922 | |
1da177e4 LT |
3923 | /* |
3924 | * Convert string parameters. | |
3925 | */ | |
3926 | if (mode) { | |
2b3798d5 NA |
3927 | bond_opt_initstr(&newval, mode); |
3928 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
3929 | if (!valptr) { | |
3930 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
3931 | return -EINVAL; |
3932 | } | |
2b3798d5 | 3933 | bond_mode = valptr->value; |
1da177e4 LT |
3934 | } |
3935 | ||
169a3e66 JV |
3936 | if (xmit_hash_policy) { |
3937 | if ((bond_mode != BOND_MODE_XOR) && | |
f05b42ea MB |
3938 | (bond_mode != BOND_MODE_8023AD) && |
3939 | (bond_mode != BOND_MODE_TLB)) { | |
a4aee5c8 | 3940 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 3941 | bond_mode_name(bond_mode)); |
169a3e66 | 3942 | } else { |
a4b32ce7 NA |
3943 | bond_opt_initstr(&newval, xmit_hash_policy); |
3944 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
3945 | &newval); | |
3946 | if (!valptr) { | |
a4aee5c8 | 3947 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
3948 | xmit_hash_policy); |
3949 | return -EINVAL; | |
3950 | } | |
a4b32ce7 | 3951 | xmit_hashtype = valptr->value; |
169a3e66 JV |
3952 | } |
3953 | } | |
3954 | ||
1da177e4 LT |
3955 | if (lacp_rate) { |
3956 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
3957 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
3958 | bond_mode_name(bond_mode)); | |
1da177e4 | 3959 | } else { |
d3131de7 NA |
3960 | bond_opt_initstr(&newval, lacp_rate); |
3961 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
3962 | &newval); | |
3963 | if (!valptr) { | |
a4aee5c8 | 3964 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 3965 | lacp_rate); |
1da177e4 LT |
3966 | return -EINVAL; |
3967 | } | |
d3131de7 | 3968 | lacp_fast = valptr->value; |
1da177e4 LT |
3969 | } |
3970 | } | |
3971 | ||
fd989c83 | 3972 | if (ad_select) { |
548d28bd | 3973 | bond_opt_initstr(&newval, ad_select); |
9e5f5eeb NA |
3974 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), |
3975 | &newval); | |
3976 | if (!valptr) { | |
3977 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
3978 | return -EINVAL; |
3979 | } | |
9e5f5eeb NA |
3980 | params->ad_select = valptr->value; |
3981 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 3982 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
3983 | } else { |
3984 | params->ad_select = BOND_AD_STABLE; | |
3985 | } | |
3986 | ||
f5841306 | 3987 | if (max_bonds < 0) { |
91565ebb JP |
3988 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
3989 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
3990 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
3991 | } | |
3992 | ||
3993 | if (miimon < 0) { | |
91565ebb JP |
3994 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
3995 | miimon, INT_MAX); | |
b98d9c66 | 3996 | miimon = 0; |
1da177e4 LT |
3997 | } |
3998 | ||
3999 | if (updelay < 0) { | |
91565ebb JP |
4000 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4001 | updelay, INT_MAX); | |
1da177e4 LT |
4002 | updelay = 0; |
4003 | } | |
4004 | ||
4005 | if (downdelay < 0) { | |
91565ebb JP |
4006 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4007 | downdelay, INT_MAX); | |
1da177e4 LT |
4008 | downdelay = 0; |
4009 | } | |
4010 | ||
4011 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
91565ebb JP |
4012 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4013 | use_carrier); | |
1da177e4 LT |
4014 | use_carrier = 1; |
4015 | } | |
4016 | ||
ad246c99 | 4017 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
4018 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
4019 | num_peer_notif); | |
ad246c99 BH |
4020 | num_peer_notif = 1; |
4021 | } | |
4022 | ||
834db4bc | 4023 | /* reset values for 802.3ad/TLB/ALB */ |
267bed77 | 4024 | if (!bond_mode_uses_arp(bond_mode)) { |
1da177e4 | 4025 | if (!miimon) { |
91565ebb JP |
4026 | 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"); |
4027 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 4028 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4029 | } |
4030 | } | |
4031 | ||
bb1d9123 | 4032 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
4033 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
4034 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
4035 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
4036 | } | |
4037 | ||
ebd8e497 | 4038 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
4039 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
4040 | all_slaves_active); | |
ebd8e497 AG |
4041 | all_slaves_active = 0; |
4042 | } | |
4043 | ||
c2952c31 | 4044 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
4045 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
4046 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
4047 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
4048 | } | |
4049 | ||
aa59d851 NA |
4050 | bond_opt_initval(&newval, packets_per_slave); |
4051 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
4052 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
4053 | packets_per_slave, USHRT_MAX); | |
4054 | packets_per_slave = 1; | |
4055 | } | |
4056 | ||
1da177e4 | 4057 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4058 | 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", |
4059 | updelay); | |
1da177e4 LT |
4060 | } |
4061 | ||
4062 | if (!miimon) { | |
4063 | if (updelay || downdelay) { | |
4064 | /* just warn the user the up/down delay will have | |
4065 | * no effect since miimon is zero... | |
4066 | */ | |
91565ebb JP |
4067 | 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", |
4068 | updelay, downdelay); | |
1da177e4 LT |
4069 | } |
4070 | } else { | |
4071 | /* don't allow arp monitoring */ | |
4072 | if (arp_interval) { | |
91565ebb JP |
4073 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4074 | miimon, arp_interval); | |
1da177e4 LT |
4075 | arp_interval = 0; |
4076 | } | |
4077 | ||
4078 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
4079 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4080 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
4081 | } |
4082 | ||
4083 | updelay /= miimon; | |
4084 | ||
4085 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
4086 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4087 | downdelay, miimon, | |
4088 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4089 | } |
4090 | ||
4091 | downdelay /= miimon; | |
4092 | } | |
4093 | ||
4094 | if (arp_interval < 0) { | |
91565ebb JP |
4095 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4096 | arp_interval, INT_MAX); | |
7bdb04ed | 4097 | arp_interval = 0; |
1da177e4 LT |
4098 | } |
4099 | ||
5a5c5fd4 | 4100 | for (arp_ip_count = 0, i = 0; |
4101 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4102 | /* not complete check, but should be good enough to |
4103 | catch mistakes */ | |
89015c18 | 4104 | __be32 ip; |
4105 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || | |
2807a9fe | 4106 | !bond_is_ip_target_ok(ip)) { |
91565ebb JP |
4107 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4108 | arp_ip_target[i]); | |
1da177e4 LT |
4109 | arp_interval = 0; |
4110 | } else { | |
0afee4e8 VF |
4111 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4112 | arp_target[arp_ip_count++] = ip; | |
4113 | else | |
91565ebb JP |
4114 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
4115 | &ip); | |
1da177e4 LT |
4116 | } |
4117 | } | |
4118 | ||
4119 | if (arp_interval && !arp_ip_count) { | |
4120 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
4121 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4122 | arp_interval); | |
1da177e4 LT |
4123 | arp_interval = 0; |
4124 | } | |
4125 | ||
f5b2b966 | 4126 | if (arp_validate) { |
f5b2b966 | 4127 | if (!arp_interval) { |
a4aee5c8 | 4128 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4129 | return -EINVAL; |
4130 | } | |
4131 | ||
16228881 NA |
4132 | bond_opt_initstr(&newval, arp_validate); |
4133 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
4134 | &newval); | |
4135 | if (!valptr) { | |
a4aee5c8 | 4136 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 4137 | arp_validate); |
f5b2b966 JV |
4138 | return -EINVAL; |
4139 | } | |
16228881 NA |
4140 | arp_validate_value = valptr->value; |
4141 | } else { | |
f5b2b966 | 4142 | arp_validate_value = 0; |
16228881 | 4143 | } |
f5b2b966 | 4144 | |
8599b52e VF |
4145 | arp_all_targets_value = 0; |
4146 | if (arp_all_targets) { | |
edf36b24 NA |
4147 | bond_opt_initstr(&newval, arp_all_targets); |
4148 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
4149 | &newval); | |
4150 | if (!valptr) { | |
8599b52e VF |
4151 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
4152 | arp_all_targets); | |
4153 | arp_all_targets_value = 0; | |
edf36b24 NA |
4154 | } else { |
4155 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
4156 | } |
4157 | } | |
4158 | ||
1da177e4 | 4159 | if (miimon) { |
a4aee5c8 | 4160 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4161 | } else if (arp_interval) { |
16228881 NA |
4162 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
4163 | arp_validate_value); | |
a4aee5c8 | 4164 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 4165 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
4166 | |
4167 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 4168 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 4169 | |
90194264 | 4170 | pr_cont("\n"); |
1da177e4 | 4171 | |
b8a9787e | 4172 | } else if (max_bonds) { |
1da177e4 LT |
4173 | /* miimon and arp_interval not set, we need one so things |
4174 | * work as expected, see bonding.txt for details | |
4175 | */ | |
90194264 | 4176 | 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 |
4177 | } |
4178 | ||
ec0865a9 | 4179 | if (primary && !bond_mode_uses_primary(bond_mode)) { |
1da177e4 LT |
4180 | /* currently, using a primary only makes sense |
4181 | * in active backup, TLB or ALB modes | |
4182 | */ | |
91565ebb JP |
4183 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
4184 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4185 | primary = NULL; |
4186 | } | |
4187 | ||
a549952a | 4188 | if (primary && primary_reselect) { |
388d3a6d NA |
4189 | bond_opt_initstr(&newval, primary_reselect); |
4190 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
4191 | &newval); | |
4192 | if (!valptr) { | |
a4aee5c8 | 4193 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 4194 | primary_reselect); |
a549952a JP |
4195 | return -EINVAL; |
4196 | } | |
388d3a6d | 4197 | primary_reselect_value = valptr->value; |
a549952a JP |
4198 | } else { |
4199 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4200 | } | |
4201 | ||
3915c1e8 | 4202 | if (fail_over_mac) { |
1df6b6aa NA |
4203 | bond_opt_initstr(&newval, fail_over_mac); |
4204 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
4205 | &newval); | |
4206 | if (!valptr) { | |
a4aee5c8 | 4207 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 4208 | fail_over_mac); |
3915c1e8 JV |
4209 | return -EINVAL; |
4210 | } | |
1df6b6aa | 4211 | fail_over_mac_value = valptr->value; |
3915c1e8 | 4212 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 4213 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
4214 | } else { |
4215 | fail_over_mac_value = BOND_FOM_NONE; | |
4216 | } | |
dd957c57 | 4217 | |
3a7129e5 | 4218 | if (lp_interval == 0) { |
91565ebb JP |
4219 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
4220 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 4221 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
4222 | } | |
4223 | ||
1da177e4 LT |
4224 | /* fill params struct with the proper values */ |
4225 | params->mode = bond_mode; | |
169a3e66 | 4226 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4227 | params->miimon = miimon; |
ad246c99 | 4228 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4229 | params->arp_interval = arp_interval; |
f5b2b966 | 4230 | params->arp_validate = arp_validate_value; |
8599b52e | 4231 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4232 | params->updelay = updelay; |
4233 | params->downdelay = downdelay; | |
4234 | params->use_carrier = use_carrier; | |
4235 | params->lacp_fast = lacp_fast; | |
4236 | params->primary[0] = 0; | |
a549952a | 4237 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4238 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4239 | params->tx_queues = tx_queues; |
ebd8e497 | 4240 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4241 | params->resend_igmp = resend_igmp; |
655f8919 | 4242 | params->min_links = min_links; |
3a7129e5 | 4243 | params->lp_interval = lp_interval; |
809fa972 | 4244 | params->packets_per_slave = packets_per_slave; |
e9f0fb88 | 4245 | params->tlb_dynamic_lb = 1; /* Default value */ |
809fa972 HFS |
4246 | if (packets_per_slave > 0) { |
4247 | params->reciprocal_packets_per_slave = | |
4248 | reciprocal_value(packets_per_slave); | |
4249 | } else { | |
4250 | /* reciprocal_packets_per_slave is unused if | |
4251 | * packets_per_slave is 0 or 1, just initialize it | |
4252 | */ | |
4253 | params->reciprocal_packets_per_slave = | |
4254 | (struct reciprocal_value) { 0 }; | |
4255 | } | |
4256 | ||
1da177e4 LT |
4257 | if (primary) { |
4258 | strncpy(params->primary, primary, IFNAMSIZ); | |
4259 | params->primary[IFNAMSIZ - 1] = 0; | |
4260 | } | |
4261 | ||
4262 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4263 | ||
4264 | return 0; | |
4265 | } | |
4266 | ||
0daa2303 | 4267 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4268 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4269 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4270 | |
e8a0464c DM |
4271 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4272 | struct netdev_queue *txq, | |
4273 | void *_unused) | |
c773e847 DM |
4274 | { |
4275 | lockdep_set_class(&txq->_xmit_lock, | |
4276 | &bonding_netdev_xmit_lock_key); | |
4277 | } | |
4278 | ||
4279 | static void bond_set_lockdep_class(struct net_device *dev) | |
4280 | { | |
cf508b12 DM |
4281 | lockdep_set_class(&dev->addr_list_lock, |
4282 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4283 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4284 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4285 | } |
4286 | ||
181470fc SH |
4287 | /* |
4288 | * Called from registration process | |
4289 | */ | |
4290 | static int bond_init(struct net_device *bond_dev) | |
4291 | { | |
4292 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4293 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc | 4294 | |
76444f50 | 4295 | netdev_dbg(bond_dev, "Begin bond_init\n"); |
181470fc SH |
4296 | |
4297 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4298 | if (!bond->wq) | |
4299 | return -ENOMEM; | |
4300 | ||
4301 | bond_set_lockdep_class(bond_dev); | |
4302 | ||
ec87fd3b | 4303 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4304 | |
6151b3d4 | 4305 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4306 | |
f073c7ca TI |
4307 | bond_debug_register(bond); |
4308 | ||
409cc1f8 JP |
4309 | /* Ensure valid dev_addr */ |
4310 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4311 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4312 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4313 | |
181470fc SH |
4314 | return 0; |
4315 | } | |
4316 | ||
0a2a78c4 | 4317 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4318 | { |
efacb309 | 4319 | return tx_queues; |
d5da4510 JP |
4320 | } |
4321 | ||
dfe60397 | 4322 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4323 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4324 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4325 | * set up our sysfs entries. | |
4326 | */ | |
ec87fd3b | 4327 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4328 | { |
4329 | struct net_device *bond_dev; | |
4330 | int res; | |
4331 | ||
4332 | rtnl_lock(); | |
027ea041 | 4333 | |
1c5cae81 | 4334 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
c835a677 | 4335 | name ? name : "bond%d", NET_NAME_UNKNOWN, |
1c5cae81 | 4336 | bond_setup, tx_queues); |
dfe60397 | 4337 | if (!bond_dev) { |
a4aee5c8 | 4338 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4339 | rtnl_unlock(); |
4340 | return -ENOMEM; | |
dfe60397 MW |
4341 | } |
4342 | ||
ec87fd3b | 4343 | dev_net_set(bond_dev, net); |
88ead977 EB |
4344 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4345 | ||
dfe60397 | 4346 | res = register_netdevice(bond_dev); |
0daa2303 | 4347 | |
e826eafa PO |
4348 | netif_carrier_off(bond_dev); |
4349 | ||
7e083840 | 4350 | rtnl_unlock(); |
9e2e61fb AW |
4351 | if (res < 0) |
4352 | bond_destructor(bond_dev); | |
30c15ba9 | 4353 | return res; |
dfe60397 MW |
4354 | } |
4355 | ||
2c8c1e72 | 4356 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4357 | { |
15449745 | 4358 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4359 | |
4360 | bn->net = net; | |
4361 | INIT_LIST_HEAD(&bn->dev_list); | |
4362 | ||
ec87fd3b | 4363 | bond_create_proc_dir(bn); |
4c22400a | 4364 | bond_create_sysfs(bn); |
87a7b84b | 4365 | |
15449745 | 4366 | return 0; |
ec87fd3b EB |
4367 | } |
4368 | ||
2c8c1e72 | 4369 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4370 | { |
15449745 | 4371 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4372 | struct bonding *bond, *tmp_bond; |
4373 | LIST_HEAD(list); | |
ec87fd3b | 4374 | |
4c22400a | 4375 | bond_destroy_sysfs(bn); |
69b0216a | 4376 | |
4377 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4378 | rtnl_lock(); | |
4379 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4380 | unregister_netdevice_queue(bond->dev, &list); | |
4381 | unregister_netdevice_many(&list); | |
4382 | rtnl_unlock(); | |
23fa5c2c VF |
4383 | |
4384 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
4385 | } |
4386 | ||
4387 | static struct pernet_operations bond_net_ops = { | |
4388 | .init = bond_net_init, | |
4389 | .exit = bond_net_exit, | |
15449745 EB |
4390 | .id = &bond_net_id, |
4391 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4392 | }; |
4393 | ||
1da177e4 LT |
4394 | static int __init bonding_init(void) |
4395 | { | |
1da177e4 LT |
4396 | int i; |
4397 | int res; | |
4398 | ||
bd33acc3 | 4399 | pr_info("%s", bond_version); |
1da177e4 | 4400 | |
dfe60397 | 4401 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4402 | if (res) |
dfe60397 | 4403 | goto out; |
1da177e4 | 4404 | |
15449745 | 4405 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4406 | if (res) |
4407 | goto out; | |
027ea041 | 4408 | |
0a2a78c4 | 4409 | res = bond_netlink_init(); |
88ead977 | 4410 | if (res) |
6639104b | 4411 | goto err_link; |
88ead977 | 4412 | |
f073c7ca TI |
4413 | bond_create_debugfs(); |
4414 | ||
1da177e4 | 4415 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4416 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4417 | if (res) |
4418 | goto err; | |
1da177e4 LT |
4419 | } |
4420 | ||
1da177e4 | 4421 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4422 | out: |
1da177e4 | 4423 | return res; |
88ead977 | 4424 | err: |
db298686 | 4425 | bond_destroy_debugfs(); |
0a2a78c4 | 4426 | bond_netlink_fini(); |
6639104b | 4427 | err_link: |
15449745 | 4428 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4429 | goto out; |
dfe60397 | 4430 | |
1da177e4 LT |
4431 | } |
4432 | ||
4433 | static void __exit bonding_exit(void) | |
4434 | { | |
4435 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4436 | ||
f073c7ca | 4437 | bond_destroy_debugfs(); |
ae68c398 | 4438 | |
0a2a78c4 | 4439 | bond_netlink_fini(); |
ffcdedb6 | 4440 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4441 | |
4442 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4443 | /* |
4444 | * Make sure we don't have an imbalance on our netpoll blocking | |
4445 | */ | |
4446 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4447 | #endif |
1da177e4 LT |
4448 | } |
4449 | ||
4450 | module_init(bonding_init); | |
4451 | module_exit(bonding_exit); | |
4452 | MODULE_LICENSE("GPL"); | |
4453 | MODULE_VERSION(DRV_VERSION); | |
4454 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4455 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |