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 | ||
a4aee5c8 JP |
34 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
35 | ||
1da177e4 LT |
36 | #include <linux/kernel.h> |
37 | #include <linux/module.h> | |
1da177e4 LT |
38 | #include <linux/types.h> |
39 | #include <linux/fcntl.h> | |
40 | #include <linux/interrupt.h> | |
41 | #include <linux/ptrace.h> | |
42 | #include <linux/ioport.h> | |
43 | #include <linux/in.h> | |
169a3e66 | 44 | #include <net/ip.h> |
1da177e4 | 45 | #include <linux/ip.h> |
169a3e66 JV |
46 | #include <linux/tcp.h> |
47 | #include <linux/udp.h> | |
1da177e4 LT |
48 | #include <linux/slab.h> |
49 | #include <linux/string.h> | |
50 | #include <linux/init.h> | |
51 | #include <linux/timer.h> | |
52 | #include <linux/socket.h> | |
53 | #include <linux/ctype.h> | |
54 | #include <linux/inet.h> | |
55 | #include <linux/bitops.h> | |
3d632c3f | 56 | #include <linux/io.h> |
1da177e4 | 57 | #include <asm/dma.h> |
3d632c3f | 58 | #include <linux/uaccess.h> |
1da177e4 LT |
59 | #include <linux/errno.h> |
60 | #include <linux/netdevice.h> | |
61 | #include <linux/inetdevice.h> | |
a816c7c7 | 62 | #include <linux/igmp.h> |
1da177e4 LT |
63 | #include <linux/etherdevice.h> |
64 | #include <linux/skbuff.h> | |
65 | #include <net/sock.h> | |
66 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
67 | #include <linux/smp.h> |
68 | #include <linux/if_ether.h> | |
69 | #include <net/arp.h> | |
70 | #include <linux/mii.h> | |
71 | #include <linux/ethtool.h> | |
72 | #include <linux/if_vlan.h> | |
73 | #include <linux/if_bonding.h> | |
b63bb739 | 74 | #include <linux/jiffies.h> |
e843fa50 | 75 | #include <linux/preempt.h> |
c3ade5ca | 76 | #include <net/route.h> |
457c4cbc | 77 | #include <net/net_namespace.h> |
ec87fd3b | 78 | #include <net/netns/generic.h> |
5ee31c68 | 79 | #include <net/pkt_sched.h> |
278b2083 | 80 | #include <linux/rculist.h> |
32819dc1 | 81 | #include <net/flow_keys.h> |
1da177e4 LT |
82 | #include "bonding.h" |
83 | #include "bond_3ad.h" | |
84 | #include "bond_alb.h" | |
85 | ||
86 | /*---------------------------- Module parameters ----------------------------*/ | |
87 | ||
88 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
1da177e4 LT |
89 | |
90 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 91 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 92 | static int num_peer_notif = 1; |
b98d9c66 | 93 | static int miimon; |
3d632c3f SH |
94 | static int updelay; |
95 | static int downdelay; | |
1da177e4 | 96 | static int use_carrier = 1; |
3d632c3f SH |
97 | static char *mode; |
98 | static char *primary; | |
a549952a | 99 | static char *primary_reselect; |
3d632c3f | 100 | static char *lacp_rate; |
655f8919 | 101 | static int min_links; |
3d632c3f SH |
102 | static char *ad_select; |
103 | static char *xmit_hash_policy; | |
7bdb04ed | 104 | static int arp_interval; |
3d632c3f SH |
105 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
106 | static char *arp_validate; | |
8599b52e | 107 | static char *arp_all_targets; |
3d632c3f | 108 | static char *fail_over_mac; |
b07ea07b | 109 | static int all_slaves_active; |
d2991f75 | 110 | static struct bond_params bonding_defaults; |
c2952c31 | 111 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
73958329 | 112 | static int packets_per_slave = 1; |
3a7129e5 | 113 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
114 | |
115 | module_param(max_bonds, int, 0); | |
116 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
117 | module_param(tx_queues, int, 0); |
118 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 119 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
120 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
121 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 122 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
123 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
124 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
125 | module_param(miimon, int, 0); |
126 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
127 | module_param(updelay, int, 0); | |
128 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
129 | module_param(downdelay, int, 0); | |
2ac47660 MW |
130 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
131 | "in milliseconds"); | |
1da177e4 | 132 | module_param(use_carrier, int, 0); |
2ac47660 MW |
133 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
134 | "0 for off, 1 for on (default)"); | |
1da177e4 | 135 | module_param(mode, charp, 0); |
90e62474 | 136 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
137 | "1 for active-backup, 2 for balance-xor, " |
138 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
139 | "6 for balance-alb"); | |
1da177e4 LT |
140 | module_param(primary, charp, 0); |
141 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
142 | module_param(primary_reselect, charp, 0); |
143 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
144 | "once it comes up; " | |
145 | "0 for always (default), " | |
146 | "1 for only if speed of primary is " | |
147 | "better, " | |
148 | "2 for only on active slave " | |
149 | "failure"); | |
1da177e4 | 150 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
151 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
152 | "0 for slow, 1 for fast"); | |
fd989c83 | 153 | module_param(ad_select, charp, 0); |
90e62474 AG |
154 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
155 | "0 for stable (default), 1 for bandwidth, " | |
156 | "2 for count"); | |
655f8919 | 157 | module_param(min_links, int, 0); |
158 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
159 | ||
169a3e66 | 160 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
161 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
162 | "0 for layer 2 (default), 1 for layer 3+4, " | |
32819dc1 NA |
163 | "2 for layer 2+3, 3 for encap layer 2+3, " |
164 | "4 for encap layer 3+4"); | |
1da177e4 LT |
165 | module_param(arp_interval, int, 0); |
166 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
167 | module_param_array(arp_ip_target, charp, NULL, 0); | |
168 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 169 | module_param(arp_validate, charp, 0); |
90e62474 AG |
170 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
171 | "0 for none (default), 1 for active, " | |
172 | "2 for backup, 3 for all"); | |
8599b52e VF |
173 | module_param(arp_all_targets, charp, 0); |
174 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 175 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
176 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
177 | "the same MAC; 0 for none (default), " | |
178 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
179 | module_param(all_slaves_active, int, 0); |
180 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 181 | "by setting active flag for all slaves; " |
ebd8e497 | 182 | "0 for never (default), 1 for always."); |
c2952c31 | 183 | module_param(resend_igmp, int, 0); |
90e62474 AG |
184 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
185 | "link failure"); | |
73958329 NA |
186 | module_param(packets_per_slave, int, 0); |
187 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
188 | "mode; 0 for a random slave, 1 packet per " | |
189 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 190 | module_param(lp_interval, uint, 0); |
191 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
192 | "the bonding driver sends learning packets to " | |
193 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
194 | |
195 | /*----------------------------- Global variables ----------------------------*/ | |
196 | ||
e843fa50 | 197 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 198 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
199 | #endif |
200 | ||
f99189b1 | 201 | int bond_net_id __read_mostly; |
1da177e4 | 202 | |
3d632c3f SH |
203 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
204 | static int arp_ip_count; | |
1da177e4 | 205 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
206 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
207 | static int lacp_fast; | |
1da177e4 | 208 | |
1da177e4 LT |
209 | /*-------------------------- Forward declarations ---------------------------*/ |
210 | ||
181470fc | 211 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 212 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
213 | |
214 | /*---------------------------- General routines -----------------------------*/ | |
215 | ||
bd33acc3 | 216 | const char *bond_mode_name(int mode) |
1da177e4 | 217 | { |
77afc92b HE |
218 | static const char *names[] = { |
219 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
220 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
221 | [BOND_MODE_XOR] = "load balancing (xor)", | |
222 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 223 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
224 | [BOND_MODE_TLB] = "transmit load balancing", |
225 | [BOND_MODE_ALB] = "adaptive load balancing", | |
226 | }; | |
227 | ||
f5280948 | 228 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 229 | return "unknown"; |
77afc92b HE |
230 | |
231 | return names[mode]; | |
1da177e4 LT |
232 | } |
233 | ||
234 | /*---------------------------------- VLAN -----------------------------------*/ | |
235 | ||
1da177e4 LT |
236 | /** |
237 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 238 | * |
1da177e4 LT |
239 | * @bond: bond device that got this skb for tx. |
240 | * @skb: hw accel VLAN tagged skb to transmit | |
241 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 242 | */ |
d316dedd | 243 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 244 | struct net_device *slave_dev) |
1da177e4 | 245 | { |
8387451e | 246 | skb->dev = slave_dev; |
374eeb5a | 247 | |
5ee31c68 | 248 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
249 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
250 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 251 | |
e15c3c22 | 252 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 253 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 254 | else |
f6dc31a8 | 255 | dev_queue_xmit(skb); |
1da177e4 LT |
256 | } |
257 | ||
258 | /* | |
cc0e4070 JP |
259 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
260 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
261 | * a. This operation is performed in IOCTL context, |
262 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
263 | * c. Holding a lock with BH disabled while directly calling a base driver | |
264 | * entry point is generally a BAD idea. | |
3d632c3f | 265 | * |
1da177e4 LT |
266 | * The design of synchronization/protection for this operation in the 8021q |
267 | * module is good for one or more VLAN devices over a single physical device | |
268 | * and cannot be extended for a teaming solution like bonding, so there is a | |
269 | * potential race condition here where a net device from the vlan group might | |
270 | * be referenced (either by a base driver or the 8021q code) while it is being | |
271 | * removed from the system. However, it turns out we're not making matters | |
272 | * worse, and if it works for regular VLAN usage it will work here too. | |
273 | */ | |
274 | ||
1da177e4 LT |
275 | /** |
276 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
277 | * @bond_dev: bonding net device that got called | |
278 | * @vid: vlan id being added | |
279 | */ | |
80d5c368 PM |
280 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
281 | __be16 proto, u16 vid) | |
1da177e4 | 282 | { |
454d7c9b | 283 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 284 | struct slave *slave, *rollback_slave; |
9caff1e7 | 285 | struct list_head *iter; |
dec1e90e | 286 | int res; |
1da177e4 | 287 | |
9caff1e7 | 288 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 289 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
290 | if (res) |
291 | goto unwind; | |
1da177e4 LT |
292 | } |
293 | ||
8e586137 | 294 | return 0; |
87002b03 JP |
295 | |
296 | unwind: | |
81f23b13 | 297 | /* unwind to the slave that failed */ |
9caff1e7 | 298 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
299 | if (rollback_slave == slave) |
300 | break; | |
301 | ||
302 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
303 | } | |
87002b03 JP |
304 | |
305 | return res; | |
1da177e4 LT |
306 | } |
307 | ||
308 | /** | |
309 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
310 | * @bond_dev: bonding net device that got called | |
311 | * @vid: vlan id being removed | |
312 | */ | |
80d5c368 PM |
313 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
314 | __be16 proto, u16 vid) | |
1da177e4 | 315 | { |
454d7c9b | 316 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 317 | struct list_head *iter; |
1da177e4 | 318 | struct slave *slave; |
1da177e4 | 319 | |
9caff1e7 | 320 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 321 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 322 | |
e868b0c9 VF |
323 | if (bond_is_lb(bond)) |
324 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
325 | |
326 | return 0; | |
1da177e4 LT |
327 | } |
328 | ||
1da177e4 LT |
329 | /*------------------------------- Link status -------------------------------*/ |
330 | ||
ff59c456 JV |
331 | /* |
332 | * Set the carrier state for the master according to the state of its | |
333 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
334 | * do special 802.3ad magic. | |
335 | * | |
336 | * Returns zero if carrier state does not change, nonzero if it does. | |
337 | */ | |
338 | static int bond_set_carrier(struct bonding *bond) | |
339 | { | |
9caff1e7 | 340 | struct list_head *iter; |
ff59c456 | 341 | struct slave *slave; |
ff59c456 | 342 | |
0965a1f3 | 343 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
344 | goto down; |
345 | ||
346 | if (bond->params.mode == BOND_MODE_8023AD) | |
347 | return bond_3ad_set_carrier(bond); | |
348 | ||
9caff1e7 | 349 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
350 | if (slave->link == BOND_LINK_UP) { |
351 | if (!netif_carrier_ok(bond->dev)) { | |
352 | netif_carrier_on(bond->dev); | |
353 | return 1; | |
354 | } | |
355 | return 0; | |
356 | } | |
357 | } | |
358 | ||
359 | down: | |
360 | if (netif_carrier_ok(bond->dev)) { | |
361 | netif_carrier_off(bond->dev); | |
362 | return 1; | |
363 | } | |
364 | return 0; | |
365 | } | |
366 | ||
1da177e4 LT |
367 | /* |
368 | * Get link speed and duplex from the slave's base driver | |
369 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 370 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 371 | * and return. |
1da177e4 | 372 | */ |
13b95fb7 | 373 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
374 | { |
375 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 376 | struct ethtool_cmd ecmd; |
5d30530e | 377 | u32 slave_speed; |
61a44b9c | 378 | int res; |
1da177e4 | 379 | |
589665f5 DC |
380 | slave->speed = SPEED_UNKNOWN; |
381 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 382 | |
4bc71cb9 | 383 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 384 | if (res < 0) |
13b95fb7 | 385 | return; |
1da177e4 | 386 | |
4bc71cb9 | 387 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 388 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 389 | return; |
1da177e4 | 390 | |
4bc71cb9 | 391 | switch (ecmd.duplex) { |
1da177e4 LT |
392 | case DUPLEX_FULL: |
393 | case DUPLEX_HALF: | |
394 | break; | |
395 | default: | |
13b95fb7 | 396 | return; |
1da177e4 LT |
397 | } |
398 | ||
5d30530e | 399 | slave->speed = slave_speed; |
4bc71cb9 | 400 | slave->duplex = ecmd.duplex; |
1da177e4 | 401 | |
13b95fb7 | 402 | return; |
1da177e4 LT |
403 | } |
404 | ||
07699f9a | 405 | const char *bond_slave_link_status(s8 link) |
406 | { | |
407 | switch (link) { | |
408 | case BOND_LINK_UP: | |
409 | return "up"; | |
410 | case BOND_LINK_FAIL: | |
411 | return "going down"; | |
412 | case BOND_LINK_DOWN: | |
413 | return "down"; | |
414 | case BOND_LINK_BACK: | |
415 | return "going back"; | |
416 | default: | |
417 | return "unknown"; | |
418 | } | |
419 | } | |
420 | ||
1da177e4 LT |
421 | /* |
422 | * if <dev> supports MII link status reporting, check its link status. | |
423 | * | |
424 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 425 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
426 | * |
427 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
428 | * can't tell and just pretend it is), or 0, meaning that the link is | |
429 | * down. | |
430 | * | |
431 | * If reporting is non-zero, instead of faking link up, return -1 if | |
432 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
433 | * support them). If use_carrier is set, return whatever it says. | |
434 | * It'd be nice if there was a good way to tell if a driver supports | |
435 | * netif_carrier, but there really isn't. | |
436 | */ | |
3d632c3f SH |
437 | static int bond_check_dev_link(struct bonding *bond, |
438 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 439 | { |
eb7cc59a | 440 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 441 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
442 | struct ifreq ifr; |
443 | struct mii_ioctl_data *mii; | |
1da177e4 | 444 | |
6c988853 PG |
445 | if (!reporting && !netif_running(slave_dev)) |
446 | return 0; | |
447 | ||
eb7cc59a | 448 | if (bond->params.use_carrier) |
1da177e4 | 449 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 450 | |
29112f4e | 451 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
452 | if (slave_dev->ethtool_ops->get_link) |
453 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
454 | BMSR_LSTATUS : 0; | |
29112f4e | 455 | |
3d632c3f | 456 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 457 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
458 | if (ioctl) { |
459 | /* TODO: set pointer to correct ioctl on a per team member */ | |
460 | /* bases to make this more efficient. that is, once */ | |
461 | /* we determine the correct ioctl, we will always */ | |
462 | /* call it and not the others for that team */ | |
463 | /* member. */ | |
464 | ||
465 | /* | |
466 | * We cannot assume that SIOCGMIIPHY will also read a | |
467 | * register; not all network drivers (e.g., e100) | |
468 | * support that. | |
469 | */ | |
470 | ||
471 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
472 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
473 | mii = if_mii(&ifr); | |
474 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
475 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
476 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
477 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
478 | } |
479 | } | |
480 | ||
1da177e4 LT |
481 | /* |
482 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 483 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
484 | * cannot report link status). If not reporting, pretend |
485 | * we're ok. | |
486 | */ | |
3d632c3f | 487 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
488 | } |
489 | ||
490 | /*----------------------------- Multicast list ------------------------------*/ | |
491 | ||
1da177e4 LT |
492 | /* |
493 | * Push the promiscuity flag down to appropriate slaves | |
494 | */ | |
7e1a1ac1 | 495 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 496 | { |
9caff1e7 | 497 | struct list_head *iter; |
7e1a1ac1 | 498 | int err = 0; |
9caff1e7 | 499 | |
1da177e4 LT |
500 | if (USES_PRIMARY(bond->params.mode)) { |
501 | /* write lock already acquired */ | |
502 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
503 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
504 | inc); | |
1da177e4 LT |
505 | } |
506 | } else { | |
507 | struct slave *slave; | |
dec1e90e | 508 | |
9caff1e7 | 509 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
510 | err = dev_set_promiscuity(slave->dev, inc); |
511 | if (err) | |
512 | return err; | |
1da177e4 LT |
513 | } |
514 | } | |
7e1a1ac1 | 515 | return err; |
1da177e4 LT |
516 | } |
517 | ||
518 | /* | |
519 | * Push the allmulti flag down to all slaves | |
520 | */ | |
7e1a1ac1 | 521 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 522 | { |
9caff1e7 | 523 | struct list_head *iter; |
7e1a1ac1 | 524 | int err = 0; |
9caff1e7 | 525 | |
1da177e4 LT |
526 | if (USES_PRIMARY(bond->params.mode)) { |
527 | /* write lock already acquired */ | |
528 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
529 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
530 | inc); | |
1da177e4 LT |
531 | } |
532 | } else { | |
533 | struct slave *slave; | |
dec1e90e | 534 | |
9caff1e7 | 535 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
536 | err = dev_set_allmulti(slave->dev, inc); |
537 | if (err) | |
538 | return err; | |
1da177e4 LT |
539 | } |
540 | } | |
7e1a1ac1 | 541 | return err; |
1da177e4 LT |
542 | } |
543 | ||
5a37e8ca FL |
544 | /* |
545 | * Retrieve the list of registered multicast addresses for the bonding | |
546 | * device and retransmit an IGMP JOIN request to the current active | |
547 | * slave. | |
548 | */ | |
f2369109 | 549 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 550 | { |
f2369109 | 551 | struct bonding *bond = container_of(work, struct bonding, |
552 | mcast_work.work); | |
553 | ||
4aa5dee4 | 554 | if (!rtnl_trylock()) { |
4beac029 | 555 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 556 | return; |
5a37e8ca | 557 | } |
4aa5dee4 | 558 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 559 | |
4f5474e7 NA |
560 | if (bond->igmp_retrans > 1) { |
561 | bond->igmp_retrans--; | |
c2952c31 | 562 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 563 | } |
f2369109 | 564 | rtnl_unlock(); |
5a37e8ca FL |
565 | } |
566 | ||
303d1cbf | 567 | /* Flush bond's hardware addresses from slave |
1da177e4 | 568 | */ |
303d1cbf | 569 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 570 | struct net_device *slave_dev) |
1da177e4 | 571 | { |
454d7c9b | 572 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 573 | |
303d1cbf JV |
574 | dev_uc_unsync(slave_dev, bond_dev); |
575 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 LT |
576 | |
577 | if (bond->params.mode == BOND_MODE_8023AD) { | |
578 | /* del lacpdu mc addr from mc list */ | |
579 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
580 | ||
22bedad3 | 581 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
582 | } |
583 | } | |
584 | ||
585 | /*--------------------------- Active slave change ---------------------------*/ | |
586 | ||
303d1cbf JV |
587 | /* Update the hardware address list and promisc/allmulti for the new and |
588 | * old active slaves (if any). Modes that are !USES_PRIMARY keep all | |
589 | * slaves up date at all times; only the USES_PRIMARY modes need to call | |
590 | * this function to swap these settings during a failover. | |
1da177e4 | 591 | */ |
303d1cbf JV |
592 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
593 | struct slave *old_active) | |
1da177e4 | 594 | { |
7864a1ad VF |
595 | ASSERT_RTNL(); |
596 | ||
1da177e4 | 597 | if (old_active) { |
3d632c3f | 598 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 599 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 600 | |
3d632c3f | 601 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 602 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 603 | |
303d1cbf | 604 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
605 | } |
606 | ||
607 | if (new_active) { | |
7e1a1ac1 | 608 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 609 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 610 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 611 | |
3d632c3f | 612 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 613 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 614 | |
d632ce98 | 615 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
616 | dev_uc_sync(new_active->dev, bond->dev); |
617 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 618 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
619 | } |
620 | } | |
621 | ||
ae0d6750 | 622 | /** |
623 | * bond_set_dev_addr - clone slave's address to bond | |
624 | * @bond_dev: bond net device | |
625 | * @slave_dev: slave net device | |
626 | * | |
627 | * Should be called with RTNL held. | |
628 | */ | |
629 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
630 | struct net_device *slave_dev) | |
631 | { | |
008aebde NA |
632 | pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
633 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 634 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
635 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
636 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
637 | } | |
638 | ||
3915c1e8 JV |
639 | /* |
640 | * bond_do_fail_over_mac | |
641 | * | |
642 | * Perform special MAC address swapping for fail_over_mac settings | |
643 | * | |
b2e7aceb | 644 | * Called with RTNL, curr_slave_lock for write_bh. |
3915c1e8 JV |
645 | */ |
646 | static void bond_do_fail_over_mac(struct bonding *bond, | |
647 | struct slave *new_active, | |
648 | struct slave *old_active) | |
1f78d9f9 | 649 | __releases(&bond->curr_slave_lock) |
1f78d9f9 | 650 | __acquires(&bond->curr_slave_lock) |
3915c1e8 JV |
651 | { |
652 | u8 tmp_mac[ETH_ALEN]; | |
653 | struct sockaddr saddr; | |
654 | int rv; | |
655 | ||
656 | switch (bond->params.fail_over_mac) { | |
657 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 658 | if (new_active) { |
7d26bb10 | 659 | write_unlock_bh(&bond->curr_slave_lock); |
ae0d6750 | 660 | bond_set_dev_addr(bond->dev, new_active->dev); |
7d26bb10 WP |
661 | write_lock_bh(&bond->curr_slave_lock); |
662 | } | |
3915c1e8 JV |
663 | break; |
664 | case BOND_FOM_FOLLOW: | |
665 | /* | |
666 | * if new_active && old_active, swap them | |
667 | * if just old_active, do nothing (going to no active slave) | |
668 | * if just new_active, set new_active to bond's MAC | |
669 | */ | |
670 | if (!new_active) | |
671 | return; | |
672 | ||
673 | write_unlock_bh(&bond->curr_slave_lock); | |
3915c1e8 JV |
674 | |
675 | if (old_active) { | |
ada0f863 | 676 | ether_addr_copy(tmp_mac, new_active->dev->dev_addr); |
2a7c183b JP |
677 | ether_addr_copy(saddr.sa_data, |
678 | old_active->dev->dev_addr); | |
3915c1e8 JV |
679 | saddr.sa_family = new_active->dev->type; |
680 | } else { | |
ada0f863 | 681 | ether_addr_copy(saddr.sa_data, bond->dev->dev_addr); |
3915c1e8 JV |
682 | saddr.sa_family = bond->dev->type; |
683 | } | |
684 | ||
685 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
686 | if (rv) { | |
a4aee5c8 | 687 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
688 | bond->dev->name, -rv, new_active->dev->name); |
689 | goto out; | |
690 | } | |
691 | ||
692 | if (!old_active) | |
693 | goto out; | |
694 | ||
ada0f863 | 695 | ether_addr_copy(saddr.sa_data, tmp_mac); |
3915c1e8 JV |
696 | saddr.sa_family = old_active->dev->type; |
697 | ||
698 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
699 | if (rv) | |
a4aee5c8 | 700 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
701 | bond->dev->name, -rv, new_active->dev->name); |
702 | out: | |
3915c1e8 JV |
703 | write_lock_bh(&bond->curr_slave_lock); |
704 | break; | |
705 | default: | |
a4aee5c8 | 706 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
707 | bond->dev->name, bond->params.fail_over_mac); |
708 | break; | |
709 | } | |
710 | ||
711 | } | |
712 | ||
a549952a JP |
713 | static bool bond_should_change_active(struct bonding *bond) |
714 | { | |
715 | struct slave *prim = bond->primary_slave; | |
716 | struct slave *curr = bond->curr_active_slave; | |
717 | ||
718 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
719 | return true; | |
720 | if (bond->force_primary) { | |
721 | bond->force_primary = false; | |
722 | return true; | |
723 | } | |
724 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
725 | (prim->speed < curr->speed || | |
726 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
727 | return false; | |
728 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
729 | return false; | |
730 | return true; | |
731 | } | |
3915c1e8 | 732 | |
1da177e4 LT |
733 | /** |
734 | * find_best_interface - select the best available slave to be the active one | |
735 | * @bond: our bonding struct | |
1da177e4 LT |
736 | */ |
737 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
738 | { | |
77140d29 VF |
739 | struct slave *slave, *bestslave = NULL; |
740 | struct list_head *iter; | |
1da177e4 | 741 | int mintime = bond->params.updelay; |
1da177e4 | 742 | |
77140d29 VF |
743 | if (bond->primary_slave && bond->primary_slave->link == BOND_LINK_UP && |
744 | bond_should_change_active(bond)) | |
745 | return bond->primary_slave; | |
1da177e4 | 746 | |
77140d29 VF |
747 | bond_for_each_slave(bond, slave, iter) { |
748 | if (slave->link == BOND_LINK_UP) | |
749 | return slave; | |
750 | if (slave->link == BOND_LINK_BACK && IS_UP(slave->dev) && | |
751 | slave->delay < mintime) { | |
752 | mintime = slave->delay; | |
753 | bestslave = slave; | |
1da177e4 LT |
754 | } |
755 | } | |
756 | ||
757 | return bestslave; | |
758 | } | |
759 | ||
ad246c99 BH |
760 | static bool bond_should_notify_peers(struct bonding *bond) |
761 | { | |
4cb4f97b | 762 | struct slave *slave; |
763 | ||
764 | rcu_read_lock(); | |
765 | slave = rcu_dereference(bond->curr_active_slave); | |
766 | rcu_read_unlock(); | |
ad246c99 BH |
767 | |
768 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
769 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
770 | ||
771 | if (!slave || !bond->send_peer_notif || | |
772 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
773 | return false; | |
774 | ||
ad246c99 BH |
775 | return true; |
776 | } | |
777 | ||
1da177e4 LT |
778 | /** |
779 | * change_active_interface - change the active slave into the specified one | |
780 | * @bond: our bonding struct | |
781 | * @new: the new slave to make the active one | |
782 | * | |
783 | * Set the new slave to the bond's settings and unset them on the old | |
784 | * curr_active_slave. | |
785 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
786 | * | |
787 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
788 | * because it is apparently the best available slave we have, even though its | |
789 | * updelay hasn't timed out yet. | |
790 | * | |
b2e7aceb | 791 | * If new_active is not NULL, caller must hold curr_slave_lock for write_bh. |
1da177e4 | 792 | */ |
a77b5325 | 793 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
794 | { |
795 | struct slave *old_active = bond->curr_active_slave; | |
796 | ||
3d632c3f | 797 | if (old_active == new_active) |
1da177e4 | 798 | return; |
1da177e4 LT |
799 | |
800 | if (new_active) { | |
8e603460 | 801 | new_active->last_link_up = jiffies; |
b2220cad | 802 | |
1da177e4 LT |
803 | if (new_active->link == BOND_LINK_BACK) { |
804 | if (USES_PRIMARY(bond->params.mode)) { | |
90194264 | 805 | pr_info("%s: making interface %s the new active one %d ms earlier\n", |
a4aee5c8 JP |
806 | bond->dev->name, new_active->dev->name, |
807 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
808 | } |
809 | ||
810 | new_active->delay = 0; | |
811 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 812 | |
3d632c3f | 813 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 814 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 815 | |
58402054 | 816 | if (bond_is_lb(bond)) |
1da177e4 | 817 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
818 | } else { |
819 | if (USES_PRIMARY(bond->params.mode)) { | |
90194264 | 820 | pr_info("%s: making interface %s the new active one\n", |
a4aee5c8 | 821 | bond->dev->name, new_active->dev->name); |
1da177e4 LT |
822 | } |
823 | } | |
824 | } | |
825 | ||
3d632c3f | 826 | if (USES_PRIMARY(bond->params.mode)) |
303d1cbf | 827 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 828 | |
58402054 | 829 | if (bond_is_lb(bond)) { |
1da177e4 | 830 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 | 831 | if (old_active) |
5e5b0665 | 832 | bond_set_slave_inactive_flags(old_active, |
833 | BOND_SLAVE_NOTIFY_NOW); | |
8f903c70 | 834 | if (new_active) |
5e5b0665 | 835 | bond_set_slave_active_flags(new_active, |
836 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 | 837 | } else { |
278b2083 | 838 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 839 | } |
c3ade5ca JV |
840 | |
841 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 842 | if (old_active) |
5e5b0665 | 843 | bond_set_slave_inactive_flags(old_active, |
844 | BOND_SLAVE_NOTIFY_NOW); | |
c3ade5ca JV |
845 | |
846 | if (new_active) { | |
ad246c99 BH |
847 | bool should_notify_peers = false; |
848 | ||
5e5b0665 | 849 | bond_set_slave_active_flags(new_active, |
850 | BOND_SLAVE_NOTIFY_NOW); | |
2ab82852 | 851 | |
709f8a45 OG |
852 | if (bond->params.fail_over_mac) |
853 | bond_do_fail_over_mac(bond, new_active, | |
854 | old_active); | |
3915c1e8 | 855 | |
ad246c99 BH |
856 | if (netif_running(bond->dev)) { |
857 | bond->send_peer_notif = | |
858 | bond->params.num_peer_notif; | |
859 | should_notify_peers = | |
860 | bond_should_notify_peers(bond); | |
861 | } | |
862 | ||
01f3109d | 863 | write_unlock_bh(&bond->curr_slave_lock); |
01f3109d | 864 | |
b7bc2a5b | 865 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 866 | if (should_notify_peers) |
b7bc2a5b AW |
867 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
868 | bond->dev); | |
01f3109d | 869 | |
01f3109d | 870 | write_lock_bh(&bond->curr_slave_lock); |
7893b249 | 871 | } |
c3ade5ca | 872 | } |
a2fd940f | 873 | |
5a37e8ca | 874 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
875 | * all were sent on curr_active_slave. |
876 | * resend only if bond is brought up with the affected | |
877 | * bonding modes and the retransmission is enabled */ | |
878 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
879 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
880 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 881 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 882 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 883 | } |
1da177e4 LT |
884 | } |
885 | ||
886 | /** | |
887 | * bond_select_active_slave - select a new active slave, if needed | |
888 | * @bond: our bonding struct | |
889 | * | |
3d632c3f | 890 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
891 | * - The old curr_active_slave has been released or lost its link. |
892 | * - The primary_slave has got its link back. | |
893 | * - A slave has got its link back and there's no old curr_active_slave. | |
894 | * | |
b2e7aceb | 895 | * Caller must hold curr_slave_lock for write_bh. |
1da177e4 | 896 | */ |
a77b5325 | 897 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
898 | { |
899 | struct slave *best_slave; | |
ff59c456 | 900 | int rv; |
1da177e4 LT |
901 | |
902 | best_slave = bond_find_best_slave(bond); | |
903 | if (best_slave != bond->curr_active_slave) { | |
904 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
905 | rv = bond_set_carrier(bond); |
906 | if (!rv) | |
907 | return; | |
908 | ||
909 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
910 | pr_info("%s: first active interface up!\n", |
911 | bond->dev->name); | |
ff59c456 | 912 | } else { |
90194264 | 913 | pr_info("%s: now running without any active interface!\n", |
a4aee5c8 | 914 | bond->dev->name); |
ff59c456 | 915 | } |
1da177e4 LT |
916 | } |
917 | } | |
918 | ||
f6dc31a8 | 919 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 920 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 921 | { |
8a8efa22 AW |
922 | struct netpoll *np; |
923 | int err = 0; | |
f6dc31a8 | 924 | |
a8779ec1 | 925 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
8a8efa22 AW |
926 | err = -ENOMEM; |
927 | if (!np) | |
928 | goto out; | |
929 | ||
a8779ec1 | 930 | err = __netpoll_setup(np, slave->dev); |
8a8efa22 AW |
931 | if (err) { |
932 | kfree(np); | |
933 | goto out; | |
f6dc31a8 | 934 | } |
8a8efa22 AW |
935 | slave->np = np; |
936 | out: | |
937 | return err; | |
938 | } | |
939 | static inline void slave_disable_netpoll(struct slave *slave) | |
940 | { | |
941 | struct netpoll *np = slave->np; | |
942 | ||
943 | if (!np) | |
944 | return; | |
945 | ||
946 | slave->np = NULL; | |
2cde6acd | 947 | __netpoll_free_async(np); |
8a8efa22 | 948 | } |
f6dc31a8 WC |
949 | |
950 | static void bond_poll_controller(struct net_device *bond_dev) | |
951 | { | |
8a8efa22 AW |
952 | } |
953 | ||
c4cdef9b | 954 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 955 | { |
c4cdef9b | 956 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 957 | struct list_head *iter; |
c2355e1a | 958 | struct slave *slave; |
c2355e1a | 959 | |
9caff1e7 | 960 | bond_for_each_slave(bond, slave, iter) |
8a8efa22 AW |
961 | if (IS_UP(slave->dev)) |
962 | slave_disable_netpoll(slave); | |
f6dc31a8 | 963 | } |
8a8efa22 | 964 | |
a8779ec1 | 965 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) |
8a8efa22 AW |
966 | { |
967 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 968 | struct list_head *iter; |
f6dc31a8 | 969 | struct slave *slave; |
dec1e90e | 970 | int err = 0; |
f6dc31a8 | 971 | |
9caff1e7 | 972 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
973 | err = slave_enable_netpoll(slave); |
974 | if (err) { | |
c4cdef9b | 975 | bond_netpoll_cleanup(dev); |
8a8efa22 | 976 | break; |
f6dc31a8 WC |
977 | } |
978 | } | |
8a8efa22 | 979 | return err; |
f6dc31a8 | 980 | } |
8a8efa22 AW |
981 | #else |
982 | static inline int slave_enable_netpoll(struct slave *slave) | |
983 | { | |
984 | return 0; | |
985 | } | |
986 | static inline void slave_disable_netpoll(struct slave *slave) | |
987 | { | |
988 | } | |
f6dc31a8 WC |
989 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
990 | { | |
991 | } | |
f6dc31a8 WC |
992 | #endif |
993 | ||
1da177e4 LT |
994 | /*---------------------------------- IOCTL ----------------------------------*/ |
995 | ||
c8f44aff | 996 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 997 | netdev_features_t features) |
8531c5ff | 998 | { |
b2a103e6 | 999 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 1000 | struct list_head *iter; |
c8f44aff | 1001 | netdev_features_t mask; |
9b7b165a | 1002 | struct slave *slave; |
b63365a2 | 1003 | |
0965a1f3 | 1004 | if (!bond_has_slaves(bond)) { |
b2a103e6 MM |
1005 | /* Disable adding VLANs to empty bond. But why? --mq */ |
1006 | features |= NETIF_F_VLAN_CHALLENGED; | |
9b7b165a | 1007 | return features; |
b2a103e6 | 1008 | } |
b63365a2 | 1009 | |
b2a103e6 | 1010 | mask = features; |
b63365a2 | 1011 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1012 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1013 | |
9caff1e7 | 1014 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
1015 | features = netdev_increment_features(features, |
1016 | slave->dev->features, | |
b2a103e6 MM |
1017 | mask); |
1018 | } | |
b0ce3508 | 1019 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1020 | |
b2a103e6 MM |
1021 | return features; |
1022 | } | |
1023 | ||
62f2a3a4 MM |
1024 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1025 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1026 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1027 | |
1028 | static void bond_compute_features(struct bonding *bond) | |
1029 | { | |
c4826861 | 1030 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
c8f44aff | 1031 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
9caff1e7 VF |
1032 | struct net_device *bond_dev = bond->dev; |
1033 | struct list_head *iter; | |
1034 | struct slave *slave; | |
b2a103e6 | 1035 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1036 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1037 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1038 | |
0965a1f3 | 1039 | if (!bond_has_slaves(bond)) |
b2a103e6 MM |
1040 | goto done; |
1041 | ||
9caff1e7 | 1042 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1043 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1044 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1045 | ||
b6fe83e9 | 1046 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1047 | if (slave->dev->hard_header_len > max_hard_header_len) |
1048 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1049 | |
1050 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1051 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1052 | } |
8531c5ff | 1053 | |
b63365a2 | 1054 | done: |
b2a103e6 | 1055 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1056 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1057 | bond_dev->gso_max_segs = gso_max_segs; |
1058 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1059 | |
b6fe83e9 ED |
1060 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1061 | bond_dev->priv_flags = flags | dst_release_flag; | |
1062 | ||
b2a103e6 | 1063 | netdev_change_features(bond_dev); |
8531c5ff AK |
1064 | } |
1065 | ||
872254dd MS |
1066 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1067 | struct net_device *slave_dev) | |
1068 | { | |
00829823 | 1069 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1070 | |
1071 | bond_dev->type = slave_dev->type; | |
1072 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1073 | bond_dev->addr_len = slave_dev->addr_len; | |
1074 | ||
1075 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1076 | slave_dev->addr_len); | |
1077 | } | |
1078 | ||
5b2c4dd2 | 1079 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1080 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1081 | */ |
1082 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1083 | struct slave *slave, |
1084 | struct bonding *bond) | |
5b2c4dd2 | 1085 | { |
2d7011ca | 1086 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1087 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1088 | skb->pkt_type != PACKET_BROADCAST && |
1089 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1090 | return false; |
5b2c4dd2 JP |
1091 | return true; |
1092 | } | |
1093 | return false; | |
1094 | } | |
1095 | ||
8a4eb573 | 1096 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1097 | { |
8a4eb573 | 1098 | struct sk_buff *skb = *pskb; |
f1c1775a | 1099 | struct slave *slave; |
0bd80dad | 1100 | struct bonding *bond; |
de063b70 ED |
1101 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1102 | struct slave *); | |
13a8e0c8 | 1103 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1104 | |
8a4eb573 JP |
1105 | skb = skb_share_check(skb, GFP_ATOMIC); |
1106 | if (unlikely(!skb)) | |
1107 | return RX_HANDLER_CONSUMED; | |
1108 | ||
1109 | *pskb = skb; | |
5b2c4dd2 | 1110 | |
35d48903 JP |
1111 | slave = bond_slave_get_rcu(skb->dev); |
1112 | bond = slave->bond; | |
0bd80dad | 1113 | |
4d97480b MH |
1114 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1115 | if (recv_probe) { | |
de063b70 ED |
1116 | ret = recv_probe(skb, bond, slave); |
1117 | if (ret == RX_HANDLER_CONSUMED) { | |
1118 | consume_skb(skb); | |
1119 | return ret; | |
3aba891d JP |
1120 | } |
1121 | } | |
1122 | ||
0bd80dad | 1123 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1124 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1125 | } |
1126 | ||
35d48903 | 1127 | skb->dev = bond->dev; |
5b2c4dd2 | 1128 | |
0bd80dad | 1129 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1130 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1131 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1132 | |
541ac7c9 CG |
1133 | if (unlikely(skb_cow_head(skb, |
1134 | skb->data - skb_mac_header(skb)))) { | |
1135 | kfree_skb(skb); | |
8a4eb573 | 1136 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1137 | } |
2a7c183b | 1138 | ether_addr_copy(eth_hdr(skb)->h_dest, bond->dev->dev_addr); |
5b2c4dd2 JP |
1139 | } |
1140 | ||
13a8e0c8 | 1141 | return ret; |
5b2c4dd2 JP |
1142 | } |
1143 | ||
471cb5a3 | 1144 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1f718f0f VF |
1145 | struct net_device *slave_dev, |
1146 | struct slave *slave) | |
471cb5a3 JP |
1147 | { |
1148 | int err; | |
1149 | ||
1f718f0f | 1150 | err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave); |
471cb5a3 JP |
1151 | if (err) |
1152 | return err; | |
1153 | slave_dev->flags |= IFF_SLAVE; | |
7f294054 | 1154 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1155 | return 0; |
1156 | } | |
1157 | ||
1158 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1159 | struct net_device *slave_dev) | |
1160 | { | |
1161 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1162 | slave_dev->flags &= ~IFF_SLAVE; | |
7f294054 | 1163 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1164 | } |
1165 | ||
1da177e4 | 1166 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1167 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1168 | { |
454d7c9b | 1169 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1170 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1171 | struct slave *new_slave = NULL, *prev_slave; |
1da177e4 LT |
1172 | struct sockaddr addr; |
1173 | int link_reporting; | |
8599b52e | 1174 | int res = 0, i; |
1da177e4 | 1175 | |
c772dde3 BH |
1176 | if (!bond->params.use_carrier && |
1177 | slave_dev->ethtool_ops->get_link == NULL && | |
1178 | slave_ops->ndo_do_ioctl == NULL) { | |
91565ebb JP |
1179 | pr_warn("%s: Warning: no link monitoring support for %s\n", |
1180 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1181 | } |
1182 | ||
1da177e4 LT |
1183 | /* already enslaved */ |
1184 | if (slave_dev->flags & IFF_SLAVE) { | |
90194264 | 1185 | pr_debug("Error: Device was already enslaved\n"); |
1da177e4 LT |
1186 | return -EBUSY; |
1187 | } | |
1188 | ||
09a89c21 JB |
1189 | if (bond_dev == slave_dev) { |
1190 | pr_err("%s: cannot enslave bond to itself.\n", bond_dev->name); | |
1191 | return -EPERM; | |
1192 | } | |
1193 | ||
1da177e4 LT |
1194 | /* vlan challenged mutual exclusion */ |
1195 | /* no need to lock since we're protected by rtnl_lock */ | |
1196 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1197 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1198 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1199 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1200 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1201 | return -EPERM; |
1202 | } else { | |
91565ebb JP |
1203 | pr_warn("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1204 | bond_dev->name, slave_dev->name, | |
1205 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1206 | } |
1207 | } else { | |
5a03cdb7 | 1208 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1209 | } |
1210 | ||
217df670 JV |
1211 | /* |
1212 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1213 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1214 | * the current ifenslave will set the interface down prior to |
1215 | * enslaving it; the old ifenslave will not. | |
1216 | */ | |
1217 | if ((slave_dev->flags & IFF_UP)) { | |
90194264 | 1218 | pr_err("%s is up - this may be due to an out of date ifenslave\n", |
217df670 JV |
1219 | slave_dev->name); |
1220 | res = -EPERM; | |
1221 | goto err_undo_flags; | |
1222 | } | |
1da177e4 | 1223 | |
872254dd MS |
1224 | /* set bonding device ether type by slave - bonding netdevices are |
1225 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1226 | * there is a need to override some of the type dependent attribs/funcs. | |
1227 | * | |
1228 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1229 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1230 | */ | |
0965a1f3 | 1231 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1232 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1233 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1234 | bond_dev->name, |
1235 | bond_dev->type, slave_dev->type); | |
75c78500 | 1236 | |
b7bc2a5b AW |
1237 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1238 | bond_dev); | |
3ca5b404 JP |
1239 | res = notifier_to_errno(res); |
1240 | if (res) { | |
1241 | pr_err("%s: refused to change device type\n", | |
1242 | bond_dev->name); | |
1243 | res = -EBUSY; | |
1244 | goto err_undo_flags; | |
1245 | } | |
75c78500 | 1246 | |
32a806c1 | 1247 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1248 | dev_uc_flush(bond_dev); |
22bedad3 | 1249 | dev_mc_flush(bond_dev); |
32a806c1 | 1250 | |
e36b9d16 MS |
1251 | if (slave_dev->type != ARPHRD_ETHER) |
1252 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1253 | else { |
e36b9d16 | 1254 | ether_setup(bond_dev); |
550fd08c NH |
1255 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1256 | } | |
75c78500 | 1257 | |
b7bc2a5b AW |
1258 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1259 | bond_dev); | |
e36b9d16 | 1260 | } |
872254dd | 1261 | } else if (bond_dev->type != slave_dev->type) { |
90194264 JP |
1262 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it\n", |
1263 | slave_dev->name, slave_dev->type, bond_dev->type); | |
a4aee5c8 JP |
1264 | res = -EINVAL; |
1265 | goto err_undo_flags; | |
872254dd MS |
1266 | } |
1267 | ||
eb7cc59a | 1268 | if (slave_ops->ndo_set_mac_address == NULL) { |
0965a1f3 | 1269 | if (!bond_has_slaves(bond)) { |
90194264 | 1270 | pr_warn("%s: Warning: The first slave device specified does not support setting the MAC address\n", |
00503b6f | 1271 | bond_dev->name); |
1272 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
1273 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; | |
90194264 | 1274 | pr_warn("%s: Setting fail_over_mac to active for active-backup mode\n", |
00503b6f | 1275 | bond_dev->name); |
1276 | } | |
3915c1e8 | 1277 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
90194264 | 1278 | pr_err("%s: Error: The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active\n", |
a4aee5c8 | 1279 | bond_dev->name); |
2ab82852 MS |
1280 | res = -EOPNOTSUPP; |
1281 | goto err_undo_flags; | |
1282 | } | |
1da177e4 LT |
1283 | } |
1284 | ||
8d8fc29d AW |
1285 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1286 | ||
c20811a7 JP |
1287 | /* If this is the first slave, then we need to set the master's hardware |
1288 | * address to be the same as the slave's. */ | |
0965a1f3 | 1289 | if (!bond_has_slaves(bond) && |
dec1e90e | 1290 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1291 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1292 | |
243cb4e5 | 1293 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1294 | if (!new_slave) { |
1295 | res = -ENOMEM; | |
1296 | goto err_undo_flags; | |
1297 | } | |
bb1d9123 AG |
1298 | /* |
1299 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1300 | * is set via sysfs or module option if desired. | |
1301 | */ | |
1302 | new_slave->queue_id = 0; | |
1303 | ||
b15ba0fb JP |
1304 | /* Save slave's original mtu and then set it to match the bond */ |
1305 | new_slave->original_mtu = slave_dev->mtu; | |
1306 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1307 | if (res) { | |
1308 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1309 | goto err_free; | |
1310 | } | |
1311 | ||
217df670 JV |
1312 | /* |
1313 | * Save slave's original ("permanent") mac address for modes | |
1314 | * that need it, and for restoring it upon release, and then | |
1315 | * set it to the master's address | |
1316 | */ | |
ada0f863 | 1317 | ether_addr_copy(new_slave->perm_hwaddr, slave_dev->dev_addr); |
1da177e4 | 1318 | |
00503b6f | 1319 | if (!bond->params.fail_over_mac || |
1320 | bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
2ab82852 MS |
1321 | /* |
1322 | * Set slave to master's mac address. The application already | |
1323 | * set the master's mac address to that of the first slave | |
1324 | */ | |
1325 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1326 | addr.sa_family = slave_dev->type; | |
1327 | res = dev_set_mac_address(slave_dev, &addr); | |
1328 | if (res) { | |
5a03cdb7 | 1329 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1330 | goto err_restore_mtu; |
2ab82852 | 1331 | } |
217df670 | 1332 | } |
1da177e4 | 1333 | |
217df670 JV |
1334 | /* open the slave since the application closed it */ |
1335 | res = dev_open(slave_dev); | |
1336 | if (res) { | |
3d632c3f | 1337 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1338 | goto err_restore_mac; |
1da177e4 LT |
1339 | } |
1340 | ||
35d48903 | 1341 | new_slave->bond = bond; |
1da177e4 | 1342 | new_slave->dev = slave_dev; |
0b680e75 | 1343 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1344 | |
58402054 | 1345 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1346 | /* bond_alb_init_slave() must be called before all other stages since |
1347 | * it might fail and we do not want to have to undo everything | |
1348 | */ | |
1349 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1350 | if (res) |
569f0c4d | 1351 | goto err_close; |
1da177e4 LT |
1352 | } |
1353 | ||
303d1cbf JV |
1354 | /* If the mode USES_PRIMARY, then the following is handled by |
1355 | * bond_change_active_slave(). | |
1da177e4 LT |
1356 | */ |
1357 | if (!USES_PRIMARY(bond->params.mode)) { | |
1358 | /* set promiscuity level to new slave */ | |
1359 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1360 | res = dev_set_promiscuity(slave_dev, 1); |
1361 | if (res) | |
1362 | goto err_close; | |
1da177e4 LT |
1363 | } |
1364 | ||
1365 | /* set allmulti level to new slave */ | |
1366 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1367 | res = dev_set_allmulti(slave_dev, 1); |
1368 | if (res) | |
1369 | goto err_close; | |
1da177e4 LT |
1370 | } |
1371 | ||
b9e40857 | 1372 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1373 | |
1374 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1375 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1376 | ||
b9e40857 | 1377 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1378 | } |
1379 | ||
1380 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1381 | /* add lacpdu mc addr to mc list */ | |
1382 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1383 | ||
22bedad3 | 1384 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1385 | } |
1386 | ||
b8e2fde4 WY |
1387 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1388 | if (res) { | |
1ff412ad | 1389 | pr_err("%s: Error: Couldn't add bond vlan ids to %s\n", |
1390 | bond_dev->name, slave_dev->name); | |
1391 | goto err_close; | |
1392 | } | |
1da177e4 | 1393 | |
c8c23903 | 1394 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1395 | |
1396 | new_slave->delay = 0; | |
1397 | new_slave->link_failure_count = 0; | |
1398 | ||
876254ae VF |
1399 | bond_update_speed_duplex(new_slave); |
1400 | ||
49f17de7 | 1401 | new_slave->last_rx = jiffies - |
f31c7937 | 1402 | (msecs_to_jiffies(bond->params.arp_interval) + 1); |
8599b52e | 1403 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
49f17de7 | 1404 | new_slave->target_last_arp_rx[i] = new_slave->last_rx; |
f5b2b966 | 1405 | |
1da177e4 LT |
1406 | if (bond->params.miimon && !bond->params.use_carrier) { |
1407 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1408 | ||
1409 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1410 | /* | |
1411 | * miimon is set but a bonded network driver | |
1412 | * does not support ETHTOOL/MII and | |
1413 | * arp_interval is not set. Note: if | |
1414 | * use_carrier is enabled, we will never go | |
1415 | * here (because netif_carrier is always | |
1416 | * supported); thus, we don't need to change | |
1417 | * the messages for netif_carrier. | |
1418 | */ | |
91565ebb JP |
1419 | pr_warn("%s: Warning: 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", |
1420 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1421 | } else if (link_reporting == -1) { |
1422 | /* unable get link status using mii/ethtool */ | |
91565ebb JP |
1423 | pr_warn("%s: Warning: 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", |
1424 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1425 | } |
1426 | } | |
1427 | ||
1428 | /* check for initial state */ | |
f31c7937 MK |
1429 | if (bond->params.miimon) { |
1430 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1431 | if (bond->params.updelay) { | |
1432 | new_slave->link = BOND_LINK_BACK; | |
1433 | new_slave->delay = bond->params.updelay; | |
1434 | } else { | |
1435 | new_slave->link = BOND_LINK_UP; | |
1436 | } | |
1da177e4 | 1437 | } else { |
f31c7937 | 1438 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1439 | } |
f31c7937 MK |
1440 | } else if (bond->params.arp_interval) { |
1441 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1442 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1443 | } else { |
f31c7937 | 1444 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1445 | } |
1446 | ||
f31c7937 | 1447 | if (new_slave->link != BOND_LINK_DOWN) |
8e603460 | 1448 | new_slave->last_link_up = jiffies; |
f31c7937 | 1449 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", |
90194264 JP |
1450 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : |
1451 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
f31c7937 | 1452 | |
1da177e4 LT |
1453 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1454 | /* if there is a primary slave, remember it */ | |
a549952a | 1455 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1456 | bond->primary_slave = new_slave; |
a549952a JP |
1457 | bond->force_primary = true; |
1458 | } | |
1da177e4 LT |
1459 | } |
1460 | ||
1461 | switch (bond->params.mode) { | |
1462 | case BOND_MODE_ACTIVEBACKUP: | |
5e5b0665 | 1463 | bond_set_slave_inactive_flags(new_slave, |
1464 | BOND_SLAVE_NOTIFY_NOW); | |
1da177e4 LT |
1465 | break; |
1466 | case BOND_MODE_8023AD: | |
1467 | /* in 802.3ad mode, the internal mechanism | |
1468 | * will activate the slaves in the selected | |
1469 | * aggregator | |
1470 | */ | |
5e5b0665 | 1471 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 | 1472 | /* if this is the first slave */ |
23c147e0 | 1473 | if (!prev_slave) { |
1da177e4 LT |
1474 | SLAVE_AD_INFO(new_slave).id = 1; |
1475 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1476 | * can be called only after the mac address of the bond is set | |
1477 | */ | |
56d00c67 | 1478 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 LT |
1479 | } else { |
1480 | SLAVE_AD_INFO(new_slave).id = | |
dec1e90e | 1481 | SLAVE_AD_INFO(prev_slave).id + 1; |
1da177e4 LT |
1482 | } |
1483 | ||
1484 | bond_3ad_bind_slave(new_slave); | |
1485 | break; | |
1486 | case BOND_MODE_TLB: | |
1487 | case BOND_MODE_ALB: | |
e30bc066 | 1488 | bond_set_active_slave(new_slave); |
5e5b0665 | 1489 | bond_set_slave_inactive_flags(new_slave, BOND_SLAVE_NOTIFY_NOW); |
1da177e4 LT |
1490 | break; |
1491 | default: | |
5a03cdb7 | 1492 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1493 | |
1494 | /* always active in trunk mode */ | |
e30bc066 | 1495 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1496 | |
1497 | /* In trunking mode there is little meaning to curr_active_slave | |
1498 | * anyway (it holds no special properties of the bond device), | |
1499 | * so we can change it without calling change_active_interface() | |
1500 | */ | |
4e591b93 | 1501 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
278b2083 | 1502 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1503 | |
1da177e4 LT |
1504 | break; |
1505 | } /* switch(bond_mode) */ | |
1506 | ||
f6dc31a8 | 1507 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1508 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1509 | if (slave_dev->npinfo) { |
1510 | if (slave_enable_netpoll(new_slave)) { | |
90194264 JP |
1511 | pr_info("Error, %s: master_dev is using netpoll, but new slave device does not support netpoll\n", |
1512 | bond_dev->name); | |
8a8efa22 | 1513 | res = -EBUSY; |
f7d9821a | 1514 | goto err_detach; |
8a8efa22 | 1515 | } |
f6dc31a8 WC |
1516 | } |
1517 | #endif | |
8a8efa22 | 1518 | |
35d48903 JP |
1519 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1520 | new_slave); | |
1521 | if (res) { | |
1522 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
5831d66e | 1523 | goto err_detach; |
35d48903 JP |
1524 | } |
1525 | ||
1f718f0f VF |
1526 | res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave); |
1527 | if (res) { | |
1528 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); | |
1529 | goto err_unregister; | |
1530 | } | |
1531 | ||
07699f9a | 1532 | res = bond_sysfs_slave_add(new_slave); |
1533 | if (res) { | |
1534 | pr_debug("Error %d calling bond_sysfs_slave_add\n", res); | |
1535 | goto err_upper_unlink; | |
1536 | } | |
1537 | ||
5378c2e6 VF |
1538 | bond->slave_cnt++; |
1539 | bond_compute_features(bond); | |
1540 | bond_set_carrier(bond); | |
1541 | ||
1542 | if (USES_PRIMARY(bond->params.mode)) { | |
f80889a5 | 1543 | block_netpoll_tx(); |
5378c2e6 VF |
1544 | write_lock_bh(&bond->curr_slave_lock); |
1545 | bond_select_active_slave(bond); | |
1546 | write_unlock_bh(&bond->curr_slave_lock); | |
f80889a5 | 1547 | unblock_netpoll_tx(); |
5378c2e6 | 1548 | } |
1f718f0f | 1549 | |
90194264 | 1550 | pr_info("%s: Enslaving %s as %s interface with %s link\n", |
a4aee5c8 | 1551 | bond_dev->name, slave_dev->name, |
90194264 JP |
1552 | bond_is_active_slave(new_slave) ? "an active" : "a backup", |
1553 | new_slave->link != BOND_LINK_DOWN ? "an up" : "a down"); | |
1da177e4 LT |
1554 | |
1555 | /* enslave is successful */ | |
1556 | return 0; | |
1557 | ||
1558 | /* Undo stages on error */ | |
07699f9a | 1559 | err_upper_unlink: |
1560 | bond_upper_dev_unlink(bond_dev, slave_dev); | |
1561 | ||
1f718f0f VF |
1562 | err_unregister: |
1563 | netdev_rx_handler_unregister(slave_dev); | |
1564 | ||
f7d9821a | 1565 | err_detach: |
303d1cbf JV |
1566 | if (!USES_PRIMARY(bond->params.mode)) |
1567 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1568 | ||
1ff412ad | 1569 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
3c5913b5 | 1570 | if (bond->primary_slave == new_slave) |
1571 | bond->primary_slave = NULL; | |
3c5913b5 | 1572 | if (bond->curr_active_slave == new_slave) { |
f80889a5 | 1573 | block_netpoll_tx(); |
3c5913b5 | 1574 | write_lock_bh(&bond->curr_slave_lock); |
c8517035 | 1575 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1576 | bond_select_active_slave(bond); |
1577 | write_unlock_bh(&bond->curr_slave_lock); | |
f80889a5 | 1578 | unblock_netpoll_tx(); |
3c5913b5 | 1579 | } |
fc7a72ac | 1580 | slave_disable_netpoll(new_slave); |
f7d9821a | 1581 | |
1da177e4 | 1582 | err_close: |
b6a5a7b9 | 1583 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1584 | dev_close(slave_dev); |
1585 | ||
1586 | err_restore_mac: | |
00503b6f | 1587 | if (!bond->params.fail_over_mac || |
1588 | bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
3915c1e8 JV |
1589 | /* XXX TODO - fom follow mode needs to change master's |
1590 | * MAC if this slave's MAC is in use by the bond, or at | |
1591 | * least print a warning. | |
1592 | */ | |
ada0f863 | 1593 | ether_addr_copy(addr.sa_data, new_slave->perm_hwaddr); |
2ab82852 MS |
1594 | addr.sa_family = slave_dev->type; |
1595 | dev_set_mac_address(slave_dev, &addr); | |
1596 | } | |
1da177e4 | 1597 | |
b15ba0fb JP |
1598 | err_restore_mtu: |
1599 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1600 | ||
1da177e4 LT |
1601 | err_free: |
1602 | kfree(new_slave); | |
1603 | ||
1604 | err_undo_flags: | |
b8fad459 | 1605 | /* Enslave of first slave has failed and we need to fix master's mac */ |
0965a1f3 | 1606 | if (!bond_has_slaves(bond) && |
844223ab | 1607 | ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr)) |
b8fad459 | 1608 | eth_hw_addr_random(bond_dev); |
3d632c3f | 1609 | |
1da177e4 LT |
1610 | return res; |
1611 | } | |
1612 | ||
1613 | /* | |
1614 | * Try to release the slave device <slave> from the bond device <master> | |
1615 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1616 | * is write-locked. If "all" is true it means that the function is being called |
1617 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1618 | * |
1619 | * The rules for slave state should be: | |
1620 | * for Active/Backup: | |
1621 | * Active stays on all backups go down | |
1622 | * for Bonded connections: | |
1623 | * The first up interface should be left on and all others downed. | |
1624 | */ | |
0896341a | 1625 | static int __bond_release_one(struct net_device *bond_dev, |
1626 | struct net_device *slave_dev, | |
1627 | bool all) | |
1da177e4 | 1628 | { |
454d7c9b | 1629 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1630 | struct slave *slave, *oldcurrent; |
1631 | struct sockaddr addr; | |
5a0068de | 1632 | int old_flags = bond_dev->flags; |
c8f44aff | 1633 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1634 | |
1635 | /* slave is not a slave or master is not master of this slave */ | |
1636 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1637 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
90194264 | 1638 | pr_err("%s: Error: cannot release %s\n", |
1da177e4 LT |
1639 | bond_dev->name, slave_dev->name); |
1640 | return -EINVAL; | |
1641 | } | |
1642 | ||
e843fa50 | 1643 | block_netpoll_tx(); |
1da177e4 LT |
1644 | |
1645 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1646 | if (!slave) { | |
1647 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1648 | pr_info("%s: %s not enslaved\n", |
1649 | bond_dev->name, slave_dev->name); | |
e843fa50 | 1650 | unblock_netpoll_tx(); |
1da177e4 LT |
1651 | return -EINVAL; |
1652 | } | |
1653 | ||
07699f9a | 1654 | bond_sysfs_slave_del(slave); |
1655 | ||
1f718f0f | 1656 | bond_upper_dev_unlink(bond_dev, slave_dev); |
35d48903 JP |
1657 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1658 | * for this slave anymore. | |
1659 | */ | |
1660 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1661 | write_lock_bh(&bond->lock); |
1662 | ||
1da177e4 | 1663 | /* Inform AD package of unbinding of slave. */ |
be79bd04 | 1664 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1665 | bond_3ad_unbind_slave(slave); |
be79bd04 | 1666 | |
c8517035 | 1667 | write_unlock_bh(&bond->lock); |
1da177e4 | 1668 | |
90194264 | 1669 | pr_info("%s: Releasing %s interface %s\n", |
a4aee5c8 | 1670 | bond_dev->name, |
e30bc066 | 1671 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1672 | slave_dev->name); |
1da177e4 LT |
1673 | |
1674 | oldcurrent = bond->curr_active_slave; | |
1675 | ||
1676 | bond->current_arp_slave = NULL; | |
1677 | ||
00503b6f | 1678 | if (!all && (!bond->params.fail_over_mac || |
1679 | bond->params.mode != BOND_MODE_ACTIVEBACKUP)) { | |
844223ab | 1680 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1681 | bond_has_slaves(bond)) |
90194264 JP |
1682 | pr_warn("%s: Warning: 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", |
1683 | bond_dev->name, slave_dev->name, | |
1684 | slave->perm_hwaddr, | |
1685 | bond_dev->name, slave_dev->name); | |
dec1e90e | 1686 | } |
1687 | ||
3d632c3f | 1688 | if (bond->primary_slave == slave) |
1da177e4 | 1689 | bond->primary_slave = NULL; |
1da177e4 | 1690 | |
c8517035 | 1691 | if (oldcurrent == slave) { |
1692 | write_lock_bh(&bond->curr_slave_lock); | |
1da177e4 | 1693 | bond_change_active_slave(bond, NULL); |
c8517035 | 1694 | write_unlock_bh(&bond->curr_slave_lock); |
1695 | } | |
1da177e4 | 1696 | |
58402054 | 1697 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1698 | /* Must be called only after the slave has been |
1699 | * detached from the list and the curr_active_slave | |
1700 | * has been cleared (if our_slave == old_current), | |
1701 | * but before a new active slave is selected. | |
1702 | */ | |
1703 | bond_alb_deinit_slave(bond, slave); | |
1704 | } | |
1705 | ||
0896341a | 1706 | if (all) { |
36708b89 | 1707 | RCU_INIT_POINTER(bond->curr_active_slave, NULL); |
0896341a | 1708 | } else if (oldcurrent == slave) { |
059fe7a5 JV |
1709 | /* |
1710 | * Note that we hold RTNL over this sequence, so there | |
1711 | * is no concern that another slave add/remove event | |
1712 | * will interfere. | |
1713 | */ | |
059fe7a5 JV |
1714 | write_lock_bh(&bond->curr_slave_lock); |
1715 | ||
1da177e4 LT |
1716 | bond_select_active_slave(bond); |
1717 | ||
059fe7a5 | 1718 | write_unlock_bh(&bond->curr_slave_lock); |
059fe7a5 JV |
1719 | } |
1720 | ||
0965a1f3 | 1721 | if (!bond_has_slaves(bond)) { |
ff59c456 | 1722 | bond_set_carrier(bond); |
409cc1f8 | 1723 | eth_hw_addr_random(bond_dev); |
1da177e4 | 1724 | |
a59d3d21 | 1725 | if (vlan_uses_dev(bond_dev)) { |
91565ebb JP |
1726 | pr_warn("%s: Warning: clearing HW address of %s while it still has VLANs\n", |
1727 | bond_dev->name, bond_dev->name); | |
1728 | pr_warn("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs\n", | |
1729 | bond_dev->name); | |
1da177e4 | 1730 | } |
1da177e4 LT |
1731 | } |
1732 | ||
e843fa50 | 1733 | unblock_netpoll_tx(); |
278b2083 | 1734 | synchronize_rcu(); |
ee6154e1 | 1735 | bond->slave_cnt--; |
1da177e4 | 1736 | |
0965a1f3 | 1737 | if (!bond_has_slaves(bond)) { |
2af73d4b | 1738 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1739 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1740 | } | |
2af73d4b | 1741 | |
b2a103e6 MM |
1742 | bond_compute_features(bond); |
1743 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1744 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
90194264 | 1745 | pr_info("%s: last VLAN challenged slave %s left bond %s - VLAN blocking is removed\n", |
b2a103e6 MM |
1746 | bond_dev->name, slave_dev->name, bond_dev->name); |
1747 | ||
b76cdba9 | 1748 | /* must do this from outside any spinlocks */ |
1ff412ad | 1749 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1750 | |
303d1cbf JV |
1751 | /* If the mode USES_PRIMARY, then this cases was handled above by |
1752 | * bond_change_active_slave(..., NULL) | |
1da177e4 LT |
1753 | */ |
1754 | if (!USES_PRIMARY(bond->params.mode)) { | |
5a0068de NH |
1755 | /* unset promiscuity level from slave |
1756 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
1757 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
1758 | * value of that flag before that change, as that was the value | |
1759 | * when this slave was attached, so we cache at the start of the | |
1760 | * function and use it here. Same goes for ALLMULTI below | |
1761 | */ | |
1762 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 1763 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1764 | |
1765 | /* unset allmulti level from slave */ | |
5a0068de | 1766 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 1767 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1768 | |
303d1cbf | 1769 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1770 | } |
1771 | ||
8a8efa22 | 1772 | slave_disable_netpoll(slave); |
f6dc31a8 | 1773 | |
1da177e4 LT |
1774 | /* close slave before restoring its mac address */ |
1775 | dev_close(slave_dev); | |
1776 | ||
00503b6f | 1777 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE || |
1778 | bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
2ab82852 | 1779 | /* restore original ("permanent") mac address */ |
ada0f863 | 1780 | ether_addr_copy(addr.sa_data, slave->perm_hwaddr); |
2ab82852 MS |
1781 | addr.sa_family = slave_dev->type; |
1782 | dev_set_mac_address(slave_dev, &addr); | |
1783 | } | |
1da177e4 | 1784 | |
b15ba0fb JP |
1785 | dev_set_mtu(slave_dev, slave->original_mtu); |
1786 | ||
2d7011ca | 1787 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1788 | |
1789 | kfree(slave); | |
1790 | ||
1791 | return 0; /* deletion OK */ | |
1792 | } | |
1793 | ||
0896341a | 1794 | /* A wrapper used because of ndo_del_link */ |
1795 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1796 | { | |
1797 | return __bond_release_one(bond_dev, slave_dev, false); | |
1798 | } | |
1799 | ||
d90a162a | 1800 | /* |
dadaa10b | 1801 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
1802 | * Must be under rtnl_lock when this function is called. |
1803 | */ | |
26d8ee75 | 1804 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1805 | struct net_device *slave_dev) | |
d90a162a | 1806 | { |
454d7c9b | 1807 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1808 | int ret; |
1809 | ||
1810 | ret = bond_release(bond_dev, slave_dev); | |
0965a1f3 | 1811 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 1812 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
90194264 | 1813 | pr_info("%s: Destroying bond %s\n", |
a4aee5c8 | 1814 | bond_dev->name, bond_dev->name); |
9e71626c | 1815 | unregister_netdevice(bond_dev); |
d90a162a MS |
1816 | } |
1817 | return ret; | |
1818 | } | |
1819 | ||
1da177e4 LT |
1820 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1821 | { | |
454d7c9b | 1822 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1823 | |
1824 | info->bond_mode = bond->params.mode; | |
1825 | info->miimon = bond->params.miimon; | |
1826 | ||
1da177e4 | 1827 | info->num_slaves = bond->slave_cnt; |
1da177e4 LT |
1828 | |
1829 | return 0; | |
1830 | } | |
1831 | ||
1832 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1833 | { | |
454d7c9b | 1834 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1835 | struct list_head *iter; |
dec1e90e | 1836 | int i = 0, res = -ENODEV; |
1da177e4 | 1837 | struct slave *slave; |
1da177e4 | 1838 | |
9caff1e7 | 1839 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 1840 | if (i++ == (int)info->slave_id) { |
689c96cc ED |
1841 | res = 0; |
1842 | strcpy(info->slave_name, slave->dev->name); | |
1843 | info->link = slave->link; | |
e30bc066 | 1844 | info->state = bond_slave_state(slave); |
689c96cc | 1845 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
1846 | break; |
1847 | } | |
1848 | } | |
1da177e4 | 1849 | |
689c96cc | 1850 | return res; |
1da177e4 LT |
1851 | } |
1852 | ||
1853 | /*-------------------------------- Monitoring -------------------------------*/ | |
1854 | ||
1da177e4 | 1855 | |
f0c76d61 JV |
1856 | static int bond_miimon_inspect(struct bonding *bond) |
1857 | { | |
dec1e90e | 1858 | int link_state, commit = 0; |
9caff1e7 | 1859 | struct list_head *iter; |
f0c76d61 | 1860 | struct slave *slave; |
41f89100 JP |
1861 | bool ignore_updelay; |
1862 | ||
1863 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 | 1864 | |
4cb4f97b | 1865 | bond_for_each_slave_rcu(bond, slave, iter) { |
f0c76d61 | 1866 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 1867 | |
f0c76d61 | 1868 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
1869 | |
1870 | switch (slave->link) { | |
f0c76d61 JV |
1871 | case BOND_LINK_UP: |
1872 | if (link_state) | |
1873 | continue; | |
1da177e4 | 1874 | |
f0c76d61 JV |
1875 | slave->link = BOND_LINK_FAIL; |
1876 | slave->delay = bond->params.downdelay; | |
1877 | if (slave->delay) { | |
90194264 | 1878 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms\n", |
a4aee5c8 JP |
1879 | bond->dev->name, |
1880 | (bond->params.mode == | |
1881 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 1882 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
1883 | "active " : "backup ") : "", |
1884 | slave->dev->name, | |
1885 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 1886 | } |
f0c76d61 JV |
1887 | /*FALLTHRU*/ |
1888 | case BOND_LINK_FAIL: | |
1889 | if (link_state) { | |
1890 | /* | |
1891 | * recovered before downdelay expired | |
1892 | */ | |
1893 | slave->link = BOND_LINK_UP; | |
8e603460 | 1894 | slave->last_link_up = jiffies; |
90194264 | 1895 | pr_info("%s: link status up again after %d ms for interface %s\n", |
a4aee5c8 JP |
1896 | bond->dev->name, |
1897 | (bond->params.downdelay - slave->delay) * | |
1898 | bond->params.miimon, | |
1899 | slave->dev->name); | |
f0c76d61 | 1900 | continue; |
1da177e4 | 1901 | } |
f0c76d61 JV |
1902 | |
1903 | if (slave->delay <= 0) { | |
1904 | slave->new_link = BOND_LINK_DOWN; | |
1905 | commit++; | |
1906 | continue; | |
1da177e4 | 1907 | } |
1da177e4 | 1908 | |
f0c76d61 JV |
1909 | slave->delay--; |
1910 | break; | |
1911 | ||
1912 | case BOND_LINK_DOWN: | |
1913 | if (!link_state) | |
1914 | continue; | |
1915 | ||
1916 | slave->link = BOND_LINK_BACK; | |
1917 | slave->delay = bond->params.updelay; | |
1918 | ||
1919 | if (slave->delay) { | |
90194264 | 1920 | pr_info("%s: link status up for interface %s, enabling it in %d ms\n", |
a4aee5c8 JP |
1921 | bond->dev->name, slave->dev->name, |
1922 | ignore_updelay ? 0 : | |
1923 | bond->params.updelay * | |
1924 | bond->params.miimon); | |
f0c76d61 JV |
1925 | } |
1926 | /*FALLTHRU*/ | |
1927 | case BOND_LINK_BACK: | |
1928 | if (!link_state) { | |
1929 | slave->link = BOND_LINK_DOWN; | |
90194264 | 1930 | pr_info("%s: link status down again after %d ms for interface %s\n", |
a4aee5c8 JP |
1931 | bond->dev->name, |
1932 | (bond->params.updelay - slave->delay) * | |
1933 | bond->params.miimon, | |
1934 | slave->dev->name); | |
f0c76d61 JV |
1935 | |
1936 | continue; | |
1937 | } | |
1938 | ||
41f89100 JP |
1939 | if (ignore_updelay) |
1940 | slave->delay = 0; | |
1941 | ||
f0c76d61 JV |
1942 | if (slave->delay <= 0) { |
1943 | slave->new_link = BOND_LINK_UP; | |
1944 | commit++; | |
41f89100 | 1945 | ignore_updelay = false; |
f0c76d61 | 1946 | continue; |
1da177e4 | 1947 | } |
f0c76d61 JV |
1948 | |
1949 | slave->delay--; | |
1da177e4 | 1950 | break; |
f0c76d61 JV |
1951 | } |
1952 | } | |
1da177e4 | 1953 | |
f0c76d61 JV |
1954 | return commit; |
1955 | } | |
1da177e4 | 1956 | |
f0c76d61 JV |
1957 | static void bond_miimon_commit(struct bonding *bond) |
1958 | { | |
9caff1e7 | 1959 | struct list_head *iter; |
f0c76d61 | 1960 | struct slave *slave; |
f0c76d61 | 1961 | |
9caff1e7 | 1962 | bond_for_each_slave(bond, slave, iter) { |
f0c76d61 JV |
1963 | switch (slave->new_link) { |
1964 | case BOND_LINK_NOCHANGE: | |
1965 | continue; | |
1da177e4 | 1966 | |
f0c76d61 JV |
1967 | case BOND_LINK_UP: |
1968 | slave->link = BOND_LINK_UP; | |
8e603460 | 1969 | slave->last_link_up = jiffies; |
f0c76d61 JV |
1970 | |
1971 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1972 | /* prevent it from being the active one */ | |
e30bc066 | 1973 | bond_set_backup_slave(slave); |
f0c76d61 JV |
1974 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
1975 | /* make it immediately active */ | |
e30bc066 | 1976 | bond_set_active_slave(slave); |
f0c76d61 JV |
1977 | } else if (slave != bond->primary_slave) { |
1978 | /* prevent it from being the active one */ | |
e30bc066 | 1979 | bond_set_backup_slave(slave); |
1da177e4 | 1980 | } |
1da177e4 | 1981 | |
90194264 | 1982 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex\n", |
700c2a77 | 1983 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
1984 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
1985 | slave->duplex ? "full" : "half"); | |
1da177e4 | 1986 | |
f0c76d61 JV |
1987 | /* notify ad that the link status has changed */ |
1988 | if (bond->params.mode == BOND_MODE_8023AD) | |
1989 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 1990 | |
58402054 | 1991 | if (bond_is_lb(bond)) |
f0c76d61 JV |
1992 | bond_alb_handle_link_change(bond, slave, |
1993 | BOND_LINK_UP); | |
1da177e4 | 1994 | |
f0c76d61 JV |
1995 | if (!bond->curr_active_slave || |
1996 | (slave == bond->primary_slave)) | |
1997 | goto do_failover; | |
1da177e4 | 1998 | |
f0c76d61 | 1999 | continue; |
059fe7a5 | 2000 | |
f0c76d61 | 2001 | case BOND_LINK_DOWN: |
fba4acda JV |
2002 | if (slave->link_failure_count < UINT_MAX) |
2003 | slave->link_failure_count++; | |
2004 | ||
f0c76d61 | 2005 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2006 | |
f0c76d61 JV |
2007 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2008 | bond->params.mode == BOND_MODE_8023AD) | |
5e5b0665 | 2009 | bond_set_slave_inactive_flags(slave, |
2010 | BOND_SLAVE_NOTIFY_NOW); | |
f0c76d61 | 2011 | |
a4aee5c8 JP |
2012 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2013 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2014 | |
2015 | if (bond->params.mode == BOND_MODE_8023AD) | |
2016 | bond_3ad_handle_link_change(slave, | |
2017 | BOND_LINK_DOWN); | |
2018 | ||
ae63e808 | 2019 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2020 | bond_alb_handle_link_change(bond, slave, |
2021 | BOND_LINK_DOWN); | |
2022 | ||
2023 | if (slave == bond->curr_active_slave) | |
2024 | goto do_failover; | |
2025 | ||
2026 | continue; | |
2027 | ||
2028 | default: | |
a4aee5c8 | 2029 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2030 | bond->dev->name, slave->new_link, |
2031 | slave->dev->name); | |
2032 | slave->new_link = BOND_LINK_NOCHANGE; | |
2033 | ||
2034 | continue; | |
2035 | } | |
2036 | ||
2037 | do_failover: | |
2038 | ASSERT_RTNL(); | |
e843fa50 | 2039 | block_netpoll_tx(); |
f0c76d61 JV |
2040 | write_lock_bh(&bond->curr_slave_lock); |
2041 | bond_select_active_slave(bond); | |
2042 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2043 | unblock_netpoll_tx(); |
f0c76d61 JV |
2044 | } |
2045 | ||
2046 | bond_set_carrier(bond); | |
1da177e4 LT |
2047 | } |
2048 | ||
0b0eef66 JV |
2049 | /* |
2050 | * bond_mii_monitor | |
2051 | * | |
2052 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2053 | * inspection, then (if inspection indicates something needs to be done) |
2054 | * an acquisition of appropriate locks followed by a commit phase to | |
2055 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2056 | */ |
6da67d26 | 2057 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2058 | { |
2059 | struct bonding *bond = container_of(work, struct bonding, | |
2060 | mii_work.work); | |
ad246c99 | 2061 | bool should_notify_peers = false; |
1f2cd845 | 2062 | unsigned long delay; |
0b0eef66 | 2063 | |
1f2cd845 DM |
2064 | delay = msecs_to_jiffies(bond->params.miimon); |
2065 | ||
2066 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2067 | goto re_arm; |
b59f9f74 | 2068 | |
4cb4f97b | 2069 | rcu_read_lock(); |
2070 | ||
ad246c99 BH |
2071 | should_notify_peers = bond_should_notify_peers(bond); |
2072 | ||
1f2cd845 | 2073 | if (bond_miimon_inspect(bond)) { |
4cb4f97b | 2074 | rcu_read_unlock(); |
f0c76d61 | 2075 | |
1f2cd845 DM |
2076 | /* Race avoidance with bond_close cancel of workqueue */ |
2077 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2078 | delay = 1; |
2079 | should_notify_peers = false; | |
2080 | goto re_arm; | |
2081 | } | |
6b6c5261 | 2082 | |
1f2cd845 DM |
2083 | bond_miimon_commit(bond); |
2084 | ||
1f2cd845 | 2085 | rtnl_unlock(); /* might sleep, hold no other locks */ |
4cb4f97b | 2086 | } else |
2087 | rcu_read_unlock(); | |
0b0eef66 | 2088 | |
f0c76d61 | 2089 | re_arm: |
e6d265e8 | 2090 | if (bond->params.miimon) |
1f2cd845 DM |
2091 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2092 | ||
1f2cd845 DM |
2093 | if (should_notify_peers) { |
2094 | if (!rtnl_trylock()) | |
2095 | return; | |
2096 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2097 | rtnl_unlock(); | |
2098 | } | |
0b0eef66 | 2099 | } |
c3ade5ca | 2100 | |
50223ce4 | 2101 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2102 | { |
50223ce4 VF |
2103 | struct net_device *upper; |
2104 | struct list_head *iter; | |
2105 | bool ret = false; | |
f5b2b966 | 2106 | |
eaddcd76 | 2107 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2108 | return true; |
f5b2b966 | 2109 | |
50223ce4 | 2110 | rcu_read_lock(); |
2f268f12 | 2111 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2112 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2113 | ret = true; | |
2114 | break; | |
2115 | } | |
f5b2b966 | 2116 | } |
50223ce4 | 2117 | rcu_read_unlock(); |
f5b2b966 | 2118 | |
50223ce4 | 2119 | return ret; |
f5b2b966 JV |
2120 | } |
2121 | ||
c3ade5ca JV |
2122 | /* |
2123 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2124 | * switches in VLAN mode (especially if ports are configured as | |
2125 | * "native" to a VLAN) might not pass non-tagged frames. | |
2126 | */ | |
fbd929f2 | 2127 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, |
2128 | __be32 dest_ip, __be32 src_ip, | |
44a40855 | 2129 | struct bond_vlan_tag *tags) |
c3ade5ca JV |
2130 | { |
2131 | struct sk_buff *skb; | |
44a40855 | 2132 | int i; |
c3ade5ca | 2133 | |
fbd929f2 | 2134 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4\n", |
2135 | arp_op, slave_dev->name, &dest_ip, &src_ip); | |
3d632c3f | 2136 | |
c3ade5ca JV |
2137 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2138 | NULL, slave_dev->dev_addr, NULL); | |
2139 | ||
2140 | if (!skb) { | |
4873ac3c | 2141 | net_err_ratelimited("ARP packet allocation failed\n"); |
c3ade5ca JV |
2142 | return; |
2143 | } | |
fbd929f2 | 2144 | |
44a40855 VY |
2145 | /* Go through all the tags backwards and add them to the packet */ |
2146 | for (i = BOND_MAX_VLAN_ENCAP - 1; i > 0; i--) { | |
2147 | if (!tags[i].vlan_id) | |
2148 | continue; | |
2149 | ||
2150 | pr_debug("inner tag: proto %X vid %X\n", | |
2151 | ntohs(tags[i].vlan_proto), tags[i].vlan_id); | |
2152 | skb = __vlan_put_tag(skb, tags[i].vlan_proto, | |
2153 | tags[i].vlan_id); | |
2154 | if (!skb) { | |
2155 | net_err_ratelimited("failed to insert inner VLAN tag\n"); | |
2156 | return; | |
2157 | } | |
2158 | } | |
2159 | /* Set the outer tag */ | |
2160 | if (tags[0].vlan_id) { | |
2161 | pr_debug("outer tag: proto %X vid %X\n", | |
2162 | ntohs(tags[0].vlan_proto), tags[0].vlan_id); | |
2163 | skb = vlan_put_tag(skb, tags[0].vlan_proto, tags[0].vlan_id); | |
c3ade5ca | 2164 | if (!skb) { |
4873ac3c | 2165 | net_err_ratelimited("failed to insert outer VLAN tag\n"); |
c3ade5ca JV |
2166 | return; |
2167 | } | |
2168 | } | |
2169 | arp_xmit(skb); | |
2170 | } | |
2171 | ||
44a40855 VY |
2172 | /* Validate the device path between the @start_dev and the @end_dev. |
2173 | * The path is valid if the @end_dev is reachable through device | |
2174 | * stacking. | |
2175 | * When the path is validated, collect any vlan information in the | |
2176 | * path. | |
2177 | */ | |
2178 | static bool bond_verify_device_path(struct net_device *start_dev, | |
2179 | struct net_device *end_dev, | |
2180 | struct bond_vlan_tag *tags) | |
2181 | { | |
2182 | struct net_device *upper; | |
2183 | struct list_head *iter; | |
2184 | int idx; | |
2185 | ||
2186 | if (start_dev == end_dev) | |
2187 | return true; | |
2188 | ||
2189 | netdev_for_each_upper_dev_rcu(start_dev, upper, iter) { | |
2190 | if (bond_verify_device_path(upper, end_dev, tags)) { | |
2191 | if (is_vlan_dev(upper)) { | |
2192 | idx = vlan_get_encap_level(upper); | |
2193 | if (idx >= BOND_MAX_VLAN_ENCAP) | |
2194 | return false; | |
2195 | ||
2196 | tags[idx].vlan_proto = | |
2197 | vlan_dev_vlan_proto(upper); | |
2198 | tags[idx].vlan_id = vlan_dev_vlan_id(upper); | |
2199 | } | |
2200 | return true; | |
2201 | } | |
2202 | } | |
2203 | ||
2204 | return false; | |
2205 | } | |
c3ade5ca | 2206 | |
1da177e4 LT |
2207 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2208 | { | |
c3ade5ca | 2209 | struct rtable *rt; |
44a40855 | 2210 | struct bond_vlan_tag tags[BOND_MAX_VLAN_ENCAP]; |
27bc11e6 | 2211 | __be32 *targets = bond->params.arp_targets, addr; |
fbd929f2 | 2212 | int i; |
44a40855 | 2213 | bool ret; |
1da177e4 | 2214 | |
27bc11e6 | 2215 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
acca2674 | 2216 | pr_debug("basa: target %pI4\n", &targets[i]); |
44a40855 | 2217 | memset(tags, 0, sizeof(tags)); |
c3ade5ca | 2218 | |
27bc11e6 | 2219 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2220 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2221 | RTO_ONLINK, 0); | |
b23dd4fe | 2222 | if (IS_ERR(rt)) { |
28572760 VF |
2223 | /* there's no route to target - try to send arp |
2224 | * probe to generate any traffic (arp_validate=0) | |
2225 | */ | |
4873ac3c | 2226 | if (bond->params.arp_validate) |
2227 | net_warn_ratelimited("%s: no route to arp_ip_target %pI4 and arp_validate is set\n", | |
2228 | bond->dev->name, | |
2229 | &targets[i]); | |
44a40855 VY |
2230 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2231 | 0, tags); | |
c3ade5ca JV |
2232 | continue; |
2233 | } | |
2234 | ||
27bc11e6 VF |
2235 | /* bond device itself */ |
2236 | if (rt->dst.dev == bond->dev) | |
2237 | goto found; | |
2238 | ||
2239 | rcu_read_lock(); | |
44a40855 | 2240 | ret = bond_verify_device_path(bond->dev, rt->dst.dev, tags); |
27bc11e6 | 2241 | rcu_read_unlock(); |
c3ade5ca | 2242 | |
44a40855 VY |
2243 | if (ret) |
2244 | goto found; | |
2245 | ||
27bc11e6 | 2246 | /* Not our device - skip */ |
3e32582f VF |
2247 | pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2248 | bond->dev->name, &targets[i], | |
2249 | rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
2250 | ||
ed4b9f80 | 2251 | ip_rt_put(rt); |
27bc11e6 VF |
2252 | continue; |
2253 | ||
2254 | found: | |
2255 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2256 | ip_rt_put(rt); | |
2257 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
44a40855 | 2258 | addr, tags); |
c3ade5ca JV |
2259 | } |
2260 | } | |
2261 | ||
d3bb52b0 | 2262 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2263 | { |
8599b52e VF |
2264 | int i; |
2265 | ||
87a7b84b VF |
2266 | if (!sip || !bond_has_this_ip(bond, tip)) { |
2267 | pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip); | |
2268 | return; | |
2269 | } | |
f5b2b966 | 2270 | |
8599b52e VF |
2271 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2272 | if (i == -1) { | |
87a7b84b VF |
2273 | pr_debug("bva: sip %pI4 not found in targets\n", &sip); |
2274 | return; | |
f5b2b966 | 2275 | } |
49f17de7 | 2276 | slave->last_rx = jiffies; |
8599b52e | 2277 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2278 | } |
2279 | ||
5bb9e0b5 | 2280 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2281 | struct slave *slave) | |
f5b2b966 | 2282 | { |
de063b70 | 2283 | struct arphdr *arp = (struct arphdr *)skb->data; |
010d3c39 | 2284 | struct slave *curr_active_slave; |
f5b2b966 | 2285 | unsigned char *arp_ptr; |
d3bb52b0 | 2286 | __be32 sip, tip; |
f2cb691a | 2287 | int alen, is_arp = skb->protocol == __cpu_to_be16(ETH_P_ARP); |
f5b2b966 | 2288 | |
f2cb691a VF |
2289 | if (!slave_do_arp_validate(bond, slave)) { |
2290 | if ((slave_do_arp_validate_only(bond, slave) && is_arp) || | |
2291 | !slave_do_arp_validate_only(bond, slave)) | |
49f17de7 | 2292 | slave->last_rx = jiffies; |
13a8e0c8 | 2293 | return RX_HANDLER_ANOTHER; |
f2cb691a VF |
2294 | } else if (!is_arp) { |
2295 | return RX_HANDLER_ANOTHER; | |
2296 | } | |
2c146102 | 2297 | |
de063b70 | 2298 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2299 | |
3aba891d JP |
2300 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2301 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2302 | |
de063b70 ED |
2303 | if (alen > skb_headlen(skb)) { |
2304 | arp = kmalloc(alen, GFP_ATOMIC); | |
2305 | if (!arp) | |
2306 | goto out_unlock; | |
2307 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2308 | goto out_unlock; | |
2309 | } | |
f5b2b966 | 2310 | |
3aba891d | 2311 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2312 | skb->pkt_type == PACKET_OTHERHOST || |
2313 | skb->pkt_type == PACKET_LOOPBACK || | |
2314 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2315 | arp->ar_pro != htons(ETH_P_IP) || | |
2316 | arp->ar_pln != 4) | |
2317 | goto out_unlock; | |
2318 | ||
2319 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2320 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2321 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2322 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2323 | memcpy(&tip, arp_ptr, 4); |
2324 | ||
5a03cdb7 | 2325 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2326 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2327 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2328 | &sip, &tip); | |
f5b2b966 | 2329 | |
010d3c39 VF |
2330 | curr_active_slave = rcu_dereference(bond->curr_active_slave); |
2331 | ||
f5b2b966 JV |
2332 | /* |
2333 | * Backup slaves won't see the ARP reply, but do come through | |
2334 | * here for each ARP probe (so we swap the sip/tip to validate | |
2335 | * the probe). In a "redundant switch, common router" type of | |
2336 | * configuration, the ARP probe will (hopefully) travel from | |
2337 | * the active, through one switch, the router, then the other | |
2338 | * switch before reaching the backup. | |
aeea64ac VF |
2339 | * |
2340 | * We 'trust' the arp requests if there is an active slave and | |
2341 | * it received valid arp reply(s) after it became active. This | |
2342 | * is done to avoid endless looping when we can't reach the | |
2343 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2344 | */ |
010d3c39 | 2345 | |
e30bc066 | 2346 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2347 | bond_validate_arp(bond, slave, sip, tip); |
010d3c39 VF |
2348 | else if (curr_active_slave && |
2349 | time_after(slave_last_rx(bond, curr_active_slave), | |
2350 | curr_active_slave->last_link_up)) | |
f5b2b966 JV |
2351 | bond_validate_arp(bond, slave, tip, sip); |
2352 | ||
2353 | out_unlock: | |
de063b70 ED |
2354 | if (arp != (struct arphdr *)skb->data) |
2355 | kfree(arp); | |
13a8e0c8 | 2356 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2357 | } |
2358 | ||
e7f63f1d VF |
2359 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2360 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2361 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2362 | */ | |
2363 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2364 | int mod) | |
2365 | { | |
2366 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2367 | ||
2368 | return time_in_range(jiffies, | |
2369 | last_act - delta_in_ticks, | |
2370 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2371 | } | |
2372 | ||
1da177e4 LT |
2373 | /* |
2374 | * this function is called regularly to monitor each slave's link | |
2375 | * ensuring that traffic is being sent and received when arp monitoring | |
2376 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2377 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2378 | * arp monitoring in active backup mode. | |
2379 | */ | |
6da67d26 | 2380 | static void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2381 | { |
1b76b316 JV |
2382 | struct bonding *bond = container_of(work, struct bonding, |
2383 | arp_work.work); | |
1da177e4 | 2384 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2385 | struct list_head *iter; |
6fde8f03 | 2386 | int do_failover = 0, slave_state_changed = 0; |
1da177e4 | 2387 | |
1f2cd845 | 2388 | if (!bond_has_slaves(bond)) |
1da177e4 | 2389 | goto re_arm; |
1da177e4 | 2390 | |
2e52f4fe | 2391 | rcu_read_lock(); |
2392 | ||
2393 | oldcurrent = ACCESS_ONCE(bond->curr_active_slave); | |
1da177e4 LT |
2394 | /* see if any of the previous devices are up now (i.e. they have |
2395 | * xmt and rcv traffic). the curr_active_slave does not come into | |
8e603460 VF |
2396 | * the picture unless it is null. also, slave->last_link_up is not |
2397 | * needed here because we send an arp on each slave and give a slave | |
2398 | * as long as it needs to get the tx/rx within the delta. | |
1da177e4 LT |
2399 | * TODO: what about up/down delay in arp mode? it wasn't here before |
2400 | * so it can wait | |
2401 | */ | |
2e52f4fe | 2402 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2403 | unsigned long trans_start = dev_trans_start(slave->dev); |
2404 | ||
1da177e4 | 2405 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2406 | if (bond_time_in_interval(bond, trans_start, 1) && |
49f17de7 | 2407 | bond_time_in_interval(bond, slave->last_rx, 1)) { |
1da177e4 LT |
2408 | |
2409 | slave->link = BOND_LINK_UP; | |
6fde8f03 | 2410 | slave_state_changed = 1; |
1da177e4 LT |
2411 | |
2412 | /* primary_slave has no meaning in round-robin | |
2413 | * mode. the window of a slave being up and | |
2414 | * curr_active_slave being null after enslaving | |
2415 | * is closed. | |
2416 | */ | |
2417 | if (!oldcurrent) { | |
90194264 | 2418 | pr_info("%s: link status definitely up for interface %s\n", |
a4aee5c8 JP |
2419 | bond->dev->name, |
2420 | slave->dev->name); | |
1da177e4 LT |
2421 | do_failover = 1; |
2422 | } else { | |
a4aee5c8 JP |
2423 | pr_info("%s: interface %s is now up\n", |
2424 | bond->dev->name, | |
2425 | slave->dev->name); | |
1da177e4 LT |
2426 | } |
2427 | } | |
2428 | } else { | |
2429 | /* slave->link == BOND_LINK_UP */ | |
2430 | ||
2431 | /* not all switches will respond to an arp request | |
2432 | * when the source ip is 0, so don't take the link down | |
2433 | * if we don't know our ip yet | |
2434 | */ | |
e7f63f1d | 2435 | if (!bond_time_in_interval(bond, trans_start, 2) || |
49f17de7 | 2436 | !bond_time_in_interval(bond, slave->last_rx, 2)) { |
1da177e4 LT |
2437 | |
2438 | slave->link = BOND_LINK_DOWN; | |
6fde8f03 | 2439 | slave_state_changed = 1; |
1da177e4 | 2440 | |
3d632c3f | 2441 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2442 | slave->link_failure_count++; |
1da177e4 | 2443 | |
90194264 JP |
2444 | pr_info("%s: interface %s is now down\n", |
2445 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2446 | |
3d632c3f | 2447 | if (slave == oldcurrent) |
1da177e4 | 2448 | do_failover = 1; |
1da177e4 LT |
2449 | } |
2450 | } | |
2451 | ||
2452 | /* note: if switch is in round-robin mode, all links | |
2453 | * must tx arp to ensure all links rx an arp - otherwise | |
2454 | * links may oscillate or not come up at all; if switch is | |
2455 | * in something like xor mode, there is nothing we can | |
2456 | * do - all replies will be rx'ed on same link causing slaves | |
2457 | * to be unstable during low/no traffic periods | |
2458 | */ | |
3d632c3f | 2459 | if (IS_UP(slave->dev)) |
1da177e4 | 2460 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2461 | } |
2462 | ||
2e52f4fe | 2463 | rcu_read_unlock(); |
2464 | ||
6fde8f03 | 2465 | if (do_failover || slave_state_changed) { |
2e52f4fe | 2466 | if (!rtnl_trylock()) |
2467 | goto re_arm; | |
1da177e4 | 2468 | |
6fde8f03 DT |
2469 | if (slave_state_changed) { |
2470 | bond_slave_state_change(bond); | |
2471 | } else if (do_failover) { | |
2472 | /* the bond_select_active_slave must hold RTNL | |
2473 | * and curr_slave_lock for write. | |
2474 | */ | |
2475 | block_netpoll_tx(); | |
2476 | write_lock_bh(&bond->curr_slave_lock); | |
1da177e4 | 2477 | |
6fde8f03 DT |
2478 | bond_select_active_slave(bond); |
2479 | ||
2480 | write_unlock_bh(&bond->curr_slave_lock); | |
2481 | unblock_netpoll_tx(); | |
2482 | } | |
2e52f4fe | 2483 | rtnl_unlock(); |
1da177e4 LT |
2484 | } |
2485 | ||
2486 | re_arm: | |
e6d265e8 | 2487 | if (bond->params.arp_interval) |
e7f63f1d VF |
2488 | queue_delayed_work(bond->wq, &bond->arp_work, |
2489 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2490 | } |
2491 | ||
2492 | /* | |
b2220cad JV |
2493 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2494 | * changes. Sets new_link in slaves to specify what action should take | |
2495 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2496 | * to link states must be committed. | |
2497 | * | |
eb9fa4b0 | 2498 | * Called with rcu_read_lock hold. |
1da177e4 | 2499 | */ |
e7f63f1d | 2500 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2501 | { |
def4460c | 2502 | unsigned long trans_start, last_rx; |
9caff1e7 | 2503 | struct list_head *iter; |
dec1e90e | 2504 | struct slave *slave; |
dec1e90e | 2505 | int commit = 0; |
da210f55 | 2506 | |
eb9fa4b0 | 2507 | bond_for_each_slave_rcu(bond, slave, iter) { |
b2220cad | 2508 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2509 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2510 | |
b2220cad | 2511 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2512 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2513 | slave->new_link = BOND_LINK_UP; |
2514 | commit++; | |
2515 | } | |
b2220cad JV |
2516 | continue; |
2517 | } | |
1da177e4 | 2518 | |
b2220cad JV |
2519 | /* |
2520 | * Give slaves 2*delta after being enslaved or made | |
2521 | * active. This avoids bouncing, as the last receive | |
2522 | * times need a full ARP monitor cycle to be updated. | |
2523 | */ | |
8e603460 | 2524 | if (bond_time_in_interval(bond, slave->last_link_up, 2)) |
b2220cad JV |
2525 | continue; |
2526 | ||
2527 | /* | |
2528 | * Backup slave is down if: | |
2529 | * - No current_arp_slave AND | |
2530 | * - more than 3*delta since last receive AND | |
2531 | * - the bond has an IP address | |
2532 | * | |
2533 | * Note: a non-null current_arp_slave indicates | |
2534 | * the curr_active_slave went down and we are | |
2535 | * searching for a new one; under this condition | |
2536 | * we only take the curr_active_slave down - this | |
2537 | * gives each slave a chance to tx/rx traffic | |
2538 | * before being taken out | |
2539 | */ | |
e30bc066 | 2540 | if (!bond_is_active_slave(slave) && |
b2220cad | 2541 | !bond->current_arp_slave && |
e7f63f1d | 2542 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2543 | slave->new_link = BOND_LINK_DOWN; |
2544 | commit++; | |
2545 | } | |
2546 | ||
2547 | /* | |
2548 | * Active slave is down if: | |
2549 | * - more than 2*delta since transmitting OR | |
2550 | * - (more than 2*delta since receive AND | |
2551 | * the bond has an IP address) | |
2552 | */ | |
cb32f2a0 | 2553 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2554 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2555 | (!bond_time_in_interval(bond, trans_start, 2) || |
2556 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2557 | slave->new_link = BOND_LINK_DOWN; |
2558 | commit++; | |
2559 | } | |
1da177e4 LT |
2560 | } |
2561 | ||
b2220cad JV |
2562 | return commit; |
2563 | } | |
1da177e4 | 2564 | |
b2220cad JV |
2565 | /* |
2566 | * Called to commit link state changes noted by inspection step of | |
2567 | * active-backup mode ARP monitor. | |
2568 | * | |
eb9fa4b0 | 2569 | * Called with RTNL hold. |
b2220cad | 2570 | */ |
e7f63f1d | 2571 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2572 | { |
cb32f2a0 | 2573 | unsigned long trans_start; |
9caff1e7 | 2574 | struct list_head *iter; |
dec1e90e | 2575 | struct slave *slave; |
1da177e4 | 2576 | |
9caff1e7 | 2577 | bond_for_each_slave(bond, slave, iter) { |
b2220cad JV |
2578 | switch (slave->new_link) { |
2579 | case BOND_LINK_NOCHANGE: | |
2580 | continue; | |
ff59c456 | 2581 | |
b2220cad | 2582 | case BOND_LINK_UP: |
cb32f2a0 | 2583 | trans_start = dev_trans_start(slave->dev); |
e7f63f1d VF |
2584 | if (bond->curr_active_slave != slave || |
2585 | (!bond->curr_active_slave && | |
2586 | bond_time_in_interval(bond, trans_start, 1))) { | |
b2220cad | 2587 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2588 | if (bond->current_arp_slave) { |
2589 | bond_set_slave_inactive_flags( | |
5e5b0665 | 2590 | bond->current_arp_slave, |
2591 | BOND_SLAVE_NOTIFY_NOW); | |
5a430974 VF |
2592 | bond->current_arp_slave = NULL; |
2593 | } | |
b2220cad | 2594 | |
90194264 | 2595 | pr_info("%s: link status definitely up for interface %s\n", |
b9f60253 | 2596 | bond->dev->name, slave->dev->name); |
b2220cad | 2597 | |
b9f60253 JP |
2598 | if (!bond->curr_active_slave || |
2599 | (slave == bond->primary_slave)) | |
2600 | goto do_failover; | |
1da177e4 | 2601 | |
b9f60253 | 2602 | } |
1da177e4 | 2603 | |
b9f60253 | 2604 | continue; |
1da177e4 | 2605 | |
b2220cad JV |
2606 | case BOND_LINK_DOWN: |
2607 | if (slave->link_failure_count < UINT_MAX) | |
2608 | slave->link_failure_count++; | |
2609 | ||
2610 | slave->link = BOND_LINK_DOWN; | |
5e5b0665 | 2611 | bond_set_slave_inactive_flags(slave, |
2612 | BOND_SLAVE_NOTIFY_NOW); | |
b2220cad | 2613 | |
a4aee5c8 | 2614 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2615 | bond->dev->name, slave->dev->name); |
b2220cad | 2616 | |
b9f60253 | 2617 | if (slave == bond->curr_active_slave) { |
b2220cad | 2618 | bond->current_arp_slave = NULL; |
b9f60253 | 2619 | goto do_failover; |
1da177e4 | 2620 | } |
b9f60253 JP |
2621 | |
2622 | continue; | |
b2220cad JV |
2623 | |
2624 | default: | |
a4aee5c8 | 2625 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2626 | bond->dev->name, slave->new_link, |
2627 | slave->dev->name); | |
b9f60253 | 2628 | continue; |
1da177e4 | 2629 | } |
1da177e4 | 2630 | |
b9f60253 JP |
2631 | do_failover: |
2632 | ASSERT_RTNL(); | |
e843fa50 | 2633 | block_netpoll_tx(); |
b2220cad | 2634 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2635 | bond_select_active_slave(bond); |
b2220cad | 2636 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2637 | unblock_netpoll_tx(); |
b2220cad | 2638 | } |
1da177e4 | 2639 | |
b2220cad JV |
2640 | bond_set_carrier(bond); |
2641 | } | |
1da177e4 | 2642 | |
b2220cad JV |
2643 | /* |
2644 | * Send ARP probes for active-backup mode ARP monitor. | |
b0929915 | 2645 | * |
2646 | * Called with rcu_read_lock hold. | |
b2220cad | 2647 | */ |
f2ebd477 | 2648 | static bool bond_ab_arp_probe(struct bonding *bond) |
b2220cad | 2649 | { |
eb9fa4b0 | 2650 | struct slave *slave, *before = NULL, *new_slave = NULL, |
b0929915 | 2651 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave), |
2652 | *curr_active_slave = rcu_dereference(bond->curr_active_slave); | |
4087df87 VF |
2653 | struct list_head *iter; |
2654 | bool found = false; | |
b0929915 | 2655 | bool should_notify_rtnl = BOND_SLAVE_NOTIFY_LATER; |
f2ebd477 | 2656 | |
98b90f26 | 2657 | if (curr_arp_slave && curr_active_slave) |
a4aee5c8 | 2658 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
eb9fa4b0 | 2659 | curr_arp_slave->dev->name, |
98b90f26 | 2660 | curr_active_slave->dev->name); |
1da177e4 | 2661 | |
98b90f26 VF |
2662 | if (curr_active_slave) { |
2663 | bond_arp_send_all(bond, curr_active_slave); | |
b0929915 | 2664 | return should_notify_rtnl; |
b2220cad | 2665 | } |
1da177e4 | 2666 | |
b2220cad JV |
2667 | /* if we don't have a curr_active_slave, search for the next available |
2668 | * backup slave from the current_arp_slave and make it the candidate | |
2669 | * for becoming the curr_active_slave | |
2670 | */ | |
1da177e4 | 2671 | |
eb9fa4b0 | 2672 | if (!curr_arp_slave) { |
b0929915 | 2673 | curr_arp_slave = bond_first_slave_rcu(bond); |
2674 | if (!curr_arp_slave) | |
2675 | return should_notify_rtnl; | |
b2220cad | 2676 | } |
1da177e4 | 2677 | |
b0929915 | 2678 | bond_set_slave_inactive_flags(curr_arp_slave, BOND_SLAVE_NOTIFY_LATER); |
059fe7a5 | 2679 | |
b0929915 | 2680 | bond_for_each_slave_rcu(bond, slave, iter) { |
4087df87 VF |
2681 | if (!found && !before && IS_UP(slave->dev)) |
2682 | before = slave; | |
1da177e4 | 2683 | |
4087df87 VF |
2684 | if (found && !new_slave && IS_UP(slave->dev)) |
2685 | new_slave = slave; | |
b2220cad JV |
2686 | /* if the link state is up at this point, we |
2687 | * mark it down - this can happen if we have | |
2688 | * simultaneous link failures and | |
2689 | * reselect_active_interface doesn't make this | |
2690 | * one the current slave so it is still marked | |
2691 | * up when it is actually down | |
1da177e4 | 2692 | */ |
4087df87 | 2693 | if (!IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { |
b2220cad JV |
2694 | slave->link = BOND_LINK_DOWN; |
2695 | if (slave->link_failure_count < UINT_MAX) | |
2696 | slave->link_failure_count++; | |
1da177e4 | 2697 | |
5e5b0665 | 2698 | bond_set_slave_inactive_flags(slave, |
b0929915 | 2699 | BOND_SLAVE_NOTIFY_LATER); |
b2220cad | 2700 | |
90194264 | 2701 | pr_info("%s: backup interface %s is now down\n", |
a4aee5c8 | 2702 | bond->dev->name, slave->dev->name); |
1da177e4 | 2703 | } |
eb9fa4b0 | 2704 | if (slave == curr_arp_slave) |
4087df87 | 2705 | found = true; |
b2220cad | 2706 | } |
4087df87 VF |
2707 | |
2708 | if (!new_slave && before) | |
2709 | new_slave = before; | |
2710 | ||
b0929915 | 2711 | if (!new_slave) |
2712 | goto check_state; | |
4087df87 VF |
2713 | |
2714 | new_slave->link = BOND_LINK_BACK; | |
b0929915 | 2715 | bond_set_slave_active_flags(new_slave, BOND_SLAVE_NOTIFY_LATER); |
4087df87 | 2716 | bond_arp_send_all(bond, new_slave); |
8e603460 | 2717 | new_slave->last_link_up = jiffies; |
eb9fa4b0 | 2718 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
f2ebd477 | 2719 | |
b0929915 | 2720 | check_state: |
2721 | bond_for_each_slave_rcu(bond, slave, iter) { | |
2722 | if (slave->should_notify) { | |
2723 | should_notify_rtnl = BOND_SLAVE_NOTIFY_NOW; | |
2724 | break; | |
2725 | } | |
2726 | } | |
2727 | return should_notify_rtnl; | |
b2220cad | 2728 | } |
1da177e4 | 2729 | |
6da67d26 | 2730 | static void bond_activebackup_arp_mon(struct work_struct *work) |
b2220cad JV |
2731 | { |
2732 | struct bonding *bond = container_of(work, struct bonding, | |
2733 | arp_work.work); | |
b0929915 | 2734 | bool should_notify_peers = false; |
2735 | bool should_notify_rtnl = false; | |
1f2cd845 | 2736 | int delta_in_ticks; |
1da177e4 | 2737 | |
1f2cd845 DM |
2738 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
2739 | ||
2740 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
2741 | goto re_arm; |
2742 | ||
eb9fa4b0 | 2743 | rcu_read_lock(); |
b0929915 | 2744 | |
ad246c99 BH |
2745 | should_notify_peers = bond_should_notify_peers(bond); |
2746 | ||
b0929915 | 2747 | if (bond_ab_arp_inspect(bond)) { |
2748 | rcu_read_unlock(); | |
2749 | ||
1f2cd845 DM |
2750 | /* Race avoidance with bond_close flush of workqueue */ |
2751 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2752 | delta_in_ticks = 1; |
2753 | should_notify_peers = false; | |
2754 | goto re_arm; | |
2755 | } | |
b2220cad | 2756 | |
1f2cd845 | 2757 | bond_ab_arp_commit(bond); |
b0929915 | 2758 | |
1f2cd845 | 2759 | rtnl_unlock(); |
b0929915 | 2760 | rcu_read_lock(); |
1f2cd845 DM |
2761 | } |
2762 | ||
b0929915 | 2763 | should_notify_rtnl = bond_ab_arp_probe(bond); |
2764 | rcu_read_unlock(); | |
ad246c99 | 2765 | |
80b9d236 | 2766 | re_arm: |
2767 | if (bond->params.arp_interval) | |
1f2cd845 DM |
2768 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
2769 | ||
b0929915 | 2770 | if (should_notify_peers || should_notify_rtnl) { |
1f2cd845 DM |
2771 | if (!rtnl_trylock()) |
2772 | return; | |
b0929915 | 2773 | |
2774 | if (should_notify_peers) | |
2775 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, | |
2776 | bond->dev); | |
2777 | if (should_notify_rtnl) | |
2778 | bond_slave_state_notify(bond); | |
2779 | ||
1f2cd845 DM |
2780 | rtnl_unlock(); |
2781 | } | |
1da177e4 LT |
2782 | } |
2783 | ||
1da177e4 LT |
2784 | /*-------------------------- netdev event handling --------------------------*/ |
2785 | ||
2786 | /* | |
2787 | * Change device name | |
2788 | */ | |
2789 | static int bond_event_changename(struct bonding *bond) | |
2790 | { | |
1da177e4 LT |
2791 | bond_remove_proc_entry(bond); |
2792 | bond_create_proc_entry(bond); | |
7e083840 | 2793 | |
f073c7ca TI |
2794 | bond_debug_reregister(bond); |
2795 | ||
1da177e4 LT |
2796 | return NOTIFY_DONE; |
2797 | } | |
2798 | ||
3d632c3f SH |
2799 | static int bond_master_netdev_event(unsigned long event, |
2800 | struct net_device *bond_dev) | |
1da177e4 | 2801 | { |
454d7c9b | 2802 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2803 | |
2804 | switch (event) { | |
2805 | case NETDEV_CHANGENAME: | |
2806 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2807 | case NETDEV_UNREGISTER: |
2808 | bond_remove_proc_entry(event_bond); | |
2809 | break; | |
2810 | case NETDEV_REGISTER: | |
2811 | bond_create_proc_entry(event_bond); | |
2812 | break; | |
6c8d23f7 | 2813 | case NETDEV_NOTIFY_PEERS: |
2814 | if (event_bond->send_peer_notif) | |
2815 | event_bond->send_peer_notif--; | |
2816 | break; | |
1da177e4 LT |
2817 | default: |
2818 | break; | |
2819 | } | |
2820 | ||
2821 | return NOTIFY_DONE; | |
2822 | } | |
2823 | ||
3d632c3f SH |
2824 | static int bond_slave_netdev_event(unsigned long event, |
2825 | struct net_device *slave_dev) | |
1da177e4 | 2826 | { |
471cb5a3 | 2827 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
6101391d | 2828 | struct bonding *bond; |
2829 | struct net_device *bond_dev; | |
471cb5a3 JP |
2830 | u32 old_speed; |
2831 | u8 old_duplex; | |
1da177e4 | 2832 | |
6101391d | 2833 | /* A netdev event can be generated while enslaving a device |
2834 | * before netdev_rx_handler_register is called in which case | |
2835 | * slave will be NULL | |
2836 | */ | |
2837 | if (!slave) | |
2838 | return NOTIFY_DONE; | |
2839 | bond_dev = slave->bond->dev; | |
2840 | bond = slave->bond; | |
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 | |
471cb5a3 JP |
2856 | if (bond->params.mode == BOND_MODE_8023AD) { |
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 VF |
2883 | /* we don't care if we don't have primary set */ |
2884 | if (!USES_PRIMARY(bond->params.mode) || | |
2885 | !bond->params.primary[0]) | |
2886 | break; | |
2887 | ||
2888 | if (slave == bond->primary_slave) { | |
2889 | /* slave's name changed - he's no longer primary */ | |
2890 | bond->primary_slave = NULL; | |
2891 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { | |
2892 | /* we have a new primary slave */ | |
2893 | bond->primary_slave = slave; | |
2894 | } else { /* we didn't change primary - exit */ | |
2895 | break; | |
2896 | } | |
2897 | ||
90194264 JP |
2898 | pr_info("%s: Primary slave changed to %s, reselecting active slave\n", |
2899 | bond->dev->name, | |
2900 | bond->primary_slave ? slave_dev->name : "none"); | |
f80889a5 | 2901 | |
2902 | block_netpoll_tx(); | |
3ec775b9 VF |
2903 | write_lock_bh(&bond->curr_slave_lock); |
2904 | bond_select_active_slave(bond); | |
2905 | write_unlock_bh(&bond->curr_slave_lock); | |
f80889a5 | 2906 | unblock_netpoll_tx(); |
1da177e4 | 2907 | break; |
8531c5ff AK |
2908 | case NETDEV_FEAT_CHANGE: |
2909 | bond_compute_features(bond); | |
2910 | break; | |
4aa5dee4 JP |
2911 | case NETDEV_RESEND_IGMP: |
2912 | /* Propagate to master device */ | |
2913 | call_netdevice_notifiers(event, slave->bond->dev); | |
2914 | break; | |
1da177e4 LT |
2915 | default: |
2916 | break; | |
2917 | } | |
2918 | ||
2919 | return NOTIFY_DONE; | |
2920 | } | |
2921 | ||
2922 | /* | |
2923 | * bond_netdev_event: handle netdev notifier chain events. | |
2924 | * | |
2925 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 2926 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
2927 | * locks for us to safely manipulate the slave devices (RTNL lock, |
2928 | * dev_probe_lock). | |
2929 | */ | |
3d632c3f SH |
2930 | static int bond_netdev_event(struct notifier_block *this, |
2931 | unsigned long event, void *ptr) | |
1da177e4 | 2932 | { |
351638e7 | 2933 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 2934 | |
5a03cdb7 | 2935 | pr_debug("event_dev: %s, event: %lx\n", |
90194264 | 2936 | event_dev ? event_dev->name : "None", event); |
1da177e4 | 2937 | |
0b680e75 JV |
2938 | if (!(event_dev->priv_flags & IFF_BONDING)) |
2939 | return NOTIFY_DONE; | |
2940 | ||
1da177e4 | 2941 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 2942 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
2943 | return bond_master_netdev_event(event, event_dev); |
2944 | } | |
2945 | ||
2946 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 2947 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
2948 | return bond_slave_netdev_event(event, event_dev); |
2949 | } | |
2950 | ||
2951 | return NOTIFY_DONE; | |
2952 | } | |
2953 | ||
2954 | static struct notifier_block bond_netdev_notifier = { | |
2955 | .notifier_call = bond_netdev_event, | |
2956 | }; | |
2957 | ||
169a3e66 JV |
2958 | /*---------------------------- Hashing Policies -----------------------------*/ |
2959 | ||
32819dc1 NA |
2960 | /* L2 hash helper */ |
2961 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 JE |
2962 | { |
2963 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
2964 | ||
2965 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
32819dc1 | 2966 | return data->h_dest[5] ^ data->h_source[5]; |
6b923cb7 JE |
2967 | |
2968 | return 0; | |
2969 | } | |
2970 | ||
32819dc1 NA |
2971 | /* Extract the appropriate headers based on bond's xmit policy */ |
2972 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
2973 | struct flow_keys *fk) | |
6f6652be | 2974 | { |
32819dc1 | 2975 | const struct ipv6hdr *iph6; |
4394542c | 2976 | const struct iphdr *iph; |
32819dc1 | 2977 | int noff, proto = -1; |
6b923cb7 | 2978 | |
32819dc1 NA |
2979 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) |
2980 | return skb_flow_dissect(skb, fk); | |
2981 | ||
2982 | fk->ports = 0; | |
2983 | noff = skb_network_offset(skb); | |
2984 | if (skb->protocol == htons(ETH_P_IP)) { | |
054bb880 | 2985 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph)))) |
32819dc1 | 2986 | return false; |
6b923cb7 | 2987 | iph = ip_hdr(skb); |
32819dc1 NA |
2988 | fk->src = iph->saddr; |
2989 | fk->dst = iph->daddr; | |
2990 | noff += iph->ihl << 2; | |
2991 | if (!ip_is_fragment(iph)) | |
2992 | proto = iph->protocol; | |
2993 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
054bb880 | 2994 | if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6)))) |
32819dc1 NA |
2995 | return false; |
2996 | iph6 = ipv6_hdr(skb); | |
2997 | fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr); | |
2998 | fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr); | |
2999 | noff += sizeof(*iph6); | |
3000 | proto = iph6->nexthdr; | |
3001 | } else { | |
3002 | return false; | |
6f6652be | 3003 | } |
32819dc1 NA |
3004 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) |
3005 | fk->ports = skb_flow_get_ports(skb, noff, proto); | |
6f6652be | 3006 | |
32819dc1 | 3007 | return true; |
6f6652be JV |
3008 | } |
3009 | ||
32819dc1 NA |
3010 | /** |
3011 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3012 | * @bond: bonding device | |
3013 | * @skb: buffer to use for headers | |
3014 | * @count: modulo value | |
3015 | * | |
3016 | * This function will extract the necessary headers from the skb buffer and use | |
3017 | * them to generate a hash based on the xmit_policy set in the bonding device | |
3018 | * which will be reduced modulo count before returning. | |
169a3e66 | 3019 | */ |
32819dc1 | 3020 | int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count) |
169a3e66 | 3021 | { |
32819dc1 NA |
3022 | struct flow_keys flow; |
3023 | u32 hash; | |
4394542c | 3024 | |
32819dc1 NA |
3025 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3026 | !bond_flow_dissect(bond, skb, &flow)) | |
3027 | return bond_eth_hash(skb) % count; | |
169a3e66 | 3028 | |
32819dc1 NA |
3029 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
3030 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) | |
3031 | hash = bond_eth_hash(skb); | |
3032 | else | |
3033 | hash = (__force u32)flow.ports; | |
3034 | hash ^= (__force u32)flow.dst ^ (__force u32)flow.src; | |
3035 | hash ^= (hash >> 16); | |
3036 | hash ^= (hash >> 8); | |
3037 | ||
3038 | return hash % count; | |
169a3e66 JV |
3039 | } |
3040 | ||
1da177e4 LT |
3041 | /*-------------------------- Device entry points ----------------------------*/ |
3042 | ||
fbb0c41b | 3043 | static void bond_work_init_all(struct bonding *bond) |
3044 | { | |
3045 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3046 | bond_resend_igmp_join_requests_delayed); | |
3047 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3048 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3049 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3050 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3051 | else | |
3052 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3053 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3054 | } | |
3055 | ||
3056 | static void bond_work_cancel_all(struct bonding *bond) | |
3057 | { | |
3058 | cancel_delayed_work_sync(&bond->mii_work); | |
3059 | cancel_delayed_work_sync(&bond->arp_work); | |
3060 | cancel_delayed_work_sync(&bond->alb_work); | |
3061 | cancel_delayed_work_sync(&bond->ad_work); | |
3062 | cancel_delayed_work_sync(&bond->mcast_work); | |
3063 | } | |
3064 | ||
1da177e4 LT |
3065 | static int bond_open(struct net_device *bond_dev) |
3066 | { | |
454d7c9b | 3067 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3068 | struct list_head *iter; |
ba3211cc | 3069 | struct slave *slave; |
1da177e4 | 3070 | |
ba3211cc PP |
3071 | /* reset slave->backup and slave->inactive */ |
3072 | read_lock(&bond->lock); | |
0965a1f3 | 3073 | if (bond_has_slaves(bond)) { |
ba3211cc | 3074 | read_lock(&bond->curr_slave_lock); |
9caff1e7 | 3075 | bond_for_each_slave(bond, slave, iter) { |
7db8df02 | 3076 | if (USES_PRIMARY(bond->params.mode) |
ba3211cc | 3077 | && (slave != bond->curr_active_slave)) { |
5e5b0665 | 3078 | bond_set_slave_inactive_flags(slave, |
3079 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc | 3080 | } else { |
5e5b0665 | 3081 | bond_set_slave_active_flags(slave, |
3082 | BOND_SLAVE_NOTIFY_NOW); | |
ba3211cc PP |
3083 | } |
3084 | } | |
3085 | read_unlock(&bond->curr_slave_lock); | |
3086 | } | |
3087 | read_unlock(&bond->lock); | |
3088 | ||
fbb0c41b | 3089 | bond_work_init_all(bond); |
5a37e8ca | 3090 | |
58402054 | 3091 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3092 | /* bond_alb_initialize must be called before the timer |
3093 | * is started. | |
3094 | */ | |
fbb0c41b | 3095 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3096 | return -ENOMEM; |
1b76b316 | 3097 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3098 | } |
3099 | ||
fbb0c41b | 3100 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3101 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3102 | |
3103 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3104 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
3fe68df9 | 3105 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3106 | } |
3107 | ||
3108 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3109 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3110 | /* register to receive LACPDUs */ |
3aba891d | 3111 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3112 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3113 | } |
3114 | ||
3115 | return 0; | |
3116 | } | |
3117 | ||
3118 | static int bond_close(struct net_device *bond_dev) | |
3119 | { | |
454d7c9b | 3120 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3121 | |
fbb0c41b | 3122 | bond_work_cancel_all(bond); |
6c8d23f7 | 3123 | bond->send_peer_notif = 0; |
ee8487c0 | 3124 | if (bond_is_lb(bond)) |
1da177e4 | 3125 | bond_alb_deinitialize(bond); |
3aba891d | 3126 | bond->recv_probe = NULL; |
1da177e4 LT |
3127 | |
3128 | return 0; | |
3129 | } | |
3130 | ||
28172739 ED |
3131 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3132 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3133 | { |
454d7c9b | 3134 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3135 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3136 | struct list_head *iter; |
1da177e4 | 3137 | struct slave *slave; |
1da177e4 | 3138 | |
28172739 | 3139 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3140 | |
3141 | read_lock_bh(&bond->lock); | |
9caff1e7 | 3142 | bond_for_each_slave(bond, slave, iter) { |
be1f3c2c | 3143 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3144 | dev_get_stats(slave->dev, &temp); |
3145 | ||
3146 | stats->rx_packets += sstats->rx_packets; | |
3147 | stats->rx_bytes += sstats->rx_bytes; | |
3148 | stats->rx_errors += sstats->rx_errors; | |
3149 | stats->rx_dropped += sstats->rx_dropped; | |
3150 | ||
3151 | stats->tx_packets += sstats->tx_packets; | |
3152 | stats->tx_bytes += sstats->tx_bytes; | |
3153 | stats->tx_errors += sstats->tx_errors; | |
3154 | stats->tx_dropped += sstats->tx_dropped; | |
3155 | ||
3156 | stats->multicast += sstats->multicast; | |
3157 | stats->collisions += sstats->collisions; | |
3158 | ||
3159 | stats->rx_length_errors += sstats->rx_length_errors; | |
3160 | stats->rx_over_errors += sstats->rx_over_errors; | |
3161 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3162 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3163 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3164 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3165 | ||
3166 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3167 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3168 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3169 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3170 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb | 3171 | } |
1da177e4 LT |
3172 | read_unlock_bh(&bond->lock); |
3173 | ||
3174 | return stats; | |
3175 | } | |
3176 | ||
3177 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3178 | { | |
080a06e1 | 3179 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3180 | struct net_device *slave_dev = NULL; |
3181 | struct ifbond k_binfo; | |
3182 | struct ifbond __user *u_binfo = NULL; | |
3183 | struct ifslave k_sinfo; | |
3184 | struct ifslave __user *u_sinfo = NULL; | |
3185 | struct mii_ioctl_data *mii = NULL; | |
d1fbd3ed | 3186 | struct bond_opt_value newval; |
387ff911 | 3187 | struct net *net; |
1da177e4 LT |
3188 | int res = 0; |
3189 | ||
a4aee5c8 | 3190 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3191 | |
3192 | switch (cmd) { | |
1da177e4 LT |
3193 | case SIOCGMIIPHY: |
3194 | mii = if_mii(ifr); | |
3d632c3f | 3195 | if (!mii) |
1da177e4 | 3196 | return -EINVAL; |
3d632c3f | 3197 | |
1da177e4 LT |
3198 | mii->phy_id = 0; |
3199 | /* Fall Through */ | |
3200 | case SIOCGMIIREG: | |
3201 | /* | |
3202 | * We do this again just in case we were called by SIOCGMIIREG | |
3203 | * instead of SIOCGMIIPHY. | |
3204 | */ | |
3205 | mii = if_mii(ifr); | |
3d632c3f | 3206 | if (!mii) |
1da177e4 | 3207 | return -EINVAL; |
3d632c3f | 3208 | |
1da177e4 LT |
3209 | |
3210 | if (mii->reg_num == 1) { | |
1da177e4 | 3211 | mii->val_out = 0; |
6603a6f2 | 3212 | read_lock(&bond->lock); |
1da177e4 | 3213 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3214 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3215 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3216 | |
1da177e4 | 3217 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3218 | read_unlock(&bond->lock); |
1da177e4 LT |
3219 | } |
3220 | ||
3221 | return 0; | |
3222 | case BOND_INFO_QUERY_OLD: | |
3223 | case SIOCBONDINFOQUERY: | |
3224 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3225 | ||
3d632c3f | 3226 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3227 | return -EFAULT; |
1da177e4 LT |
3228 | |
3229 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3230 | if (res == 0 && |
3231 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3232 | return -EFAULT; | |
1da177e4 LT |
3233 | |
3234 | return res; | |
3235 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3236 | case SIOCBONDSLAVEINFOQUERY: | |
3237 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3238 | ||
3d632c3f | 3239 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3240 | return -EFAULT; |
1da177e4 LT |
3241 | |
3242 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3243 | if (res == 0 && |
3244 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3245 | return -EFAULT; | |
1da177e4 LT |
3246 | |
3247 | return res; | |
3248 | default: | |
3249 | /* Go on */ | |
3250 | break; | |
3251 | } | |
3252 | ||
387ff911 G |
3253 | net = dev_net(bond_dev); |
3254 | ||
3255 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3256 | return -EPERM; |
1da177e4 | 3257 | |
0917b933 | 3258 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3259 | |
a4aee5c8 | 3260 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3261 | |
3d632c3f | 3262 | if (!slave_dev) |
0917b933 | 3263 | return -ENODEV; |
1da177e4 | 3264 | |
0917b933 YX |
3265 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
3266 | switch (cmd) { | |
3267 | case BOND_ENSLAVE_OLD: | |
3268 | case SIOCBONDENSLAVE: | |
3269 | res = bond_enslave(bond_dev, slave_dev); | |
3270 | break; | |
3271 | case BOND_RELEASE_OLD: | |
3272 | case SIOCBONDRELEASE: | |
3273 | res = bond_release(bond_dev, slave_dev); | |
3274 | break; | |
3275 | case BOND_SETHWADDR_OLD: | |
3276 | case SIOCBONDSETHWADDR: | |
3277 | bond_set_dev_addr(bond_dev, slave_dev); | |
3278 | res = 0; | |
3279 | break; | |
3280 | case BOND_CHANGE_ACTIVE_OLD: | |
3281 | case SIOCBONDCHANGEACTIVE: | |
d1fbd3ed NA |
3282 | bond_opt_initstr(&newval, slave_dev->name); |
3283 | res = __bond_opt_set(bond, BOND_OPT_ACTIVE_SLAVE, &newval); | |
0917b933 YX |
3284 | break; |
3285 | default: | |
3286 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3287 | } |
3288 | ||
1da177e4 LT |
3289 | return res; |
3290 | } | |
3291 | ||
d03462b9 | 3292 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3293 | { |
454d7c9b | 3294 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3295 | |
d03462b9 JP |
3296 | if (change & IFF_PROMISC) |
3297 | bond_set_promiscuity(bond, | |
3298 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3299 | |
d03462b9 JP |
3300 | if (change & IFF_ALLMULTI) |
3301 | bond_set_allmulti(bond, | |
3302 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3303 | } | |
1da177e4 | 3304 | |
303d1cbf | 3305 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3306 | { |
3307 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3308 | struct list_head *iter; |
303d1cbf | 3309 | struct slave *slave; |
1da177e4 | 3310 | |
80ee5ad2 | 3311 | |
b3241870 | 3312 | rcu_read_lock(); |
303d1cbf | 3313 | if (USES_PRIMARY(bond->params.mode)) { |
b3241870 | 3314 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3315 | if (slave) { |
3316 | dev_uc_sync(slave->dev, bond_dev); | |
3317 | dev_mc_sync(slave->dev, bond_dev); | |
3318 | } | |
303d1cbf | 3319 | } else { |
b3241870 | 3320 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3321 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3322 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3323 | } | |
1da177e4 | 3324 | } |
b3241870 | 3325 | rcu_read_unlock(); |
1da177e4 LT |
3326 | } |
3327 | ||
234bcf8a | 3328 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3329 | { |
234bcf8a | 3330 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3331 | const struct net_device_ops *slave_ops; |
3332 | struct neigh_parms parms; | |
dec1e90e | 3333 | struct slave *slave; |
234bcf8a SP |
3334 | int ret; |
3335 | ||
dec1e90e | 3336 | slave = bond_first_slave(bond); |
234bcf8a SP |
3337 | if (!slave) |
3338 | return 0; | |
234bcf8a | 3339 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3340 | if (!slave_ops->ndo_neigh_setup) |
3341 | return 0; | |
3342 | ||
3343 | parms.neigh_setup = NULL; | |
3344 | parms.neigh_cleanup = NULL; | |
3345 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3346 | if (ret) | |
3347 | return ret; | |
3348 | ||
3349 | /* | |
3350 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3351 | * after the last slave has been detached. Assumes that all slaves | |
3352 | * utilize the same neigh_cleanup (true at this writing as only user | |
3353 | * is ipoib). | |
3354 | */ | |
3355 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3356 | ||
3357 | if (!parms.neigh_setup) | |
3358 | return 0; | |
3359 | ||
3360 | return parms.neigh_setup(n); | |
3361 | } | |
3362 | ||
3363 | /* | |
3364 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3365 | * slave exists. So we must declare proxy setup function which will | |
3366 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3367 | * |
3368 | * It's also called by master devices (such as vlans) to setup their | |
3369 | * underlying devices. In that case - do nothing, we're already set up from | |
3370 | * our init. | |
234bcf8a SP |
3371 | */ |
3372 | static int bond_neigh_setup(struct net_device *dev, | |
3373 | struct neigh_parms *parms) | |
3374 | { | |
9918d5bf VF |
3375 | /* modify only our neigh_parms */ |
3376 | if (parms->dev == dev) | |
3377 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3378 | |
00829823 SH |
3379 | return 0; |
3380 | } | |
3381 | ||
1da177e4 LT |
3382 | /* |
3383 | * Change the MTU of all of a master's slaves to match the master | |
3384 | */ | |
3385 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3386 | { | |
454d7c9b | 3387 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3388 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3389 | struct list_head *iter; |
1da177e4 | 3390 | int res = 0; |
1da177e4 | 3391 | |
90194264 JP |
3392 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", |
3393 | bond, bond_dev ? bond_dev->name : "None", new_mtu); | |
1da177e4 LT |
3394 | |
3395 | /* Can't hold bond->lock with bh disabled here since | |
3396 | * some base drivers panic. On the other hand we can't | |
3397 | * hold bond->lock without bh disabled because we'll | |
3398 | * deadlock. The only solution is to rely on the fact | |
3399 | * that we're under rtnl_lock here, and the slaves | |
3400 | * list won't change. This doesn't solve the problem | |
3401 | * of setting the slave's MTU while it is | |
3402 | * transmitting, but the assumption is that the base | |
3403 | * driver can handle that. | |
3404 | * | |
3405 | * TODO: figure out a way to safely iterate the slaves | |
3406 | * list, but without holding a lock around the actual | |
3407 | * call to the base driver. | |
3408 | */ | |
3409 | ||
9caff1e7 | 3410 | bond_for_each_slave(bond, slave, iter) { |
c8c23903 | 3411 | pr_debug("s %p c_m %p\n", |
90194264 | 3412 | slave, slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 3413 | |
1da177e4 LT |
3414 | res = dev_set_mtu(slave->dev, new_mtu); |
3415 | ||
3416 | if (res) { | |
3417 | /* If we failed to set the slave's mtu to the new value | |
3418 | * we must abort the operation even in ACTIVE_BACKUP | |
3419 | * mode, because if we allow the backup slaves to have | |
3420 | * different mtu values than the active slave we'll | |
3421 | * need to change their mtu when doing a failover. That | |
3422 | * means changing their mtu from timer context, which | |
3423 | * is probably not a good idea. | |
3424 | */ | |
5a03cdb7 | 3425 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3426 | goto unwind; |
3427 | } | |
3428 | } | |
3429 | ||
3430 | bond_dev->mtu = new_mtu; | |
3431 | ||
3432 | return 0; | |
3433 | ||
3434 | unwind: | |
3435 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3436 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3437 | int tmp_res; |
3438 | ||
81f23b13 VF |
3439 | if (rollback_slave == slave) |
3440 | break; | |
3441 | ||
3442 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3443 | if (tmp_res) { |
a4aee5c8 | 3444 | pr_debug("unwind err %d dev %s\n", |
81f23b13 | 3445 | tmp_res, rollback_slave->dev->name); |
1da177e4 LT |
3446 | } |
3447 | } | |
3448 | ||
3449 | return res; | |
3450 | } | |
3451 | ||
3452 | /* | |
3453 | * Change HW address | |
3454 | * | |
3455 | * Note that many devices must be down to change the HW address, and | |
3456 | * downing the master releases all slaves. We can make bonds full of | |
3457 | * bonding devices to test this, however. | |
3458 | */ | |
3459 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3460 | { | |
454d7c9b | 3461 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3462 | struct slave *slave, *rollback_slave; |
1da177e4 | 3463 | struct sockaddr *sa = addr, tmp_sa; |
9caff1e7 | 3464 | struct list_head *iter; |
1da177e4 | 3465 | int res = 0; |
1da177e4 | 3466 | |
eb7cc59a SH |
3467 | if (bond->params.mode == BOND_MODE_ALB) |
3468 | return bond_alb_set_mac_address(bond_dev, addr); | |
3469 | ||
3470 | ||
a4aee5c8 JP |
3471 | pr_debug("bond=%p, name=%s\n", |
3472 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3473 | |
1b5acd29 JV |
3474 | /* If fail_over_mac is enabled, do nothing and return success. |
3475 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3476 | */ |
cc689aaa | 3477 | if (bond->params.fail_over_mac && |
3478 | bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
dd957c57 | 3479 | return 0; |
2ab82852 | 3480 | |
3d632c3f | 3481 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3482 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3483 | |
3484 | /* Can't hold bond->lock with bh disabled here since | |
3485 | * some base drivers panic. On the other hand we can't | |
3486 | * hold bond->lock without bh disabled because we'll | |
3487 | * deadlock. The only solution is to rely on the fact | |
3488 | * that we're under rtnl_lock here, and the slaves | |
3489 | * list won't change. This doesn't solve the problem | |
3490 | * of setting the slave's hw address while it is | |
3491 | * transmitting, but the assumption is that the base | |
3492 | * driver can handle that. | |
3493 | * | |
3494 | * TODO: figure out a way to safely iterate the slaves | |
3495 | * list, but without holding a lock around the actual | |
3496 | * call to the base driver. | |
3497 | */ | |
3498 | ||
9caff1e7 | 3499 | bond_for_each_slave(bond, slave, iter) { |
5a03cdb7 | 3500 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 LT |
3501 | res = dev_set_mac_address(slave->dev, addr); |
3502 | if (res) { | |
3503 | /* TODO: consider downing the slave | |
3504 | * and retry ? | |
3505 | * User should expect communications | |
3506 | * breakage anyway until ARP finish | |
3507 | * updating, so... | |
3508 | */ | |
5a03cdb7 | 3509 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3510 | goto unwind; |
3511 | } | |
3512 | } | |
3513 | ||
3514 | /* success */ | |
3515 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3516 | return 0; | |
3517 | ||
3518 | unwind: | |
3519 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3520 | tmp_sa.sa_family = bond_dev->type; | |
3521 | ||
3522 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3523 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3524 | int tmp_res; |
3525 | ||
81f23b13 VF |
3526 | if (rollback_slave == slave) |
3527 | break; | |
3528 | ||
3529 | tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa); | |
1da177e4 | 3530 | if (tmp_res) { |
a4aee5c8 | 3531 | pr_debug("unwind err %d dev %s\n", |
81f23b13 | 3532 | tmp_res, rollback_slave->dev->name); |
1da177e4 LT |
3533 | } |
3534 | } | |
3535 | ||
3536 | return res; | |
3537 | } | |
3538 | ||
15077228 NA |
3539 | /** |
3540 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3541 | * @bond: bonding device that is transmitting | |
3542 | * @skb: buffer to transmit | |
3543 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3544 | * | |
3545 | * This function tries to transmit through slave with slave_id but in case | |
3546 | * it fails, it tries to find the first available slave for transmission. | |
3547 | * The skb is consumed in all cases, thus the function is void. | |
3548 | */ | |
da131ddb | 3549 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3550 | { |
9caff1e7 | 3551 | struct list_head *iter; |
15077228 NA |
3552 | struct slave *slave; |
3553 | int i = slave_id; | |
3554 | ||
3555 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3556 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3557 | if (--i < 0) { |
3558 | if (slave_can_tx(slave)) { | |
3559 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3560 | return; | |
3561 | } | |
3562 | } | |
3563 | } | |
3564 | ||
3565 | /* Here we start from the first slave up to slave_id */ | |
3566 | i = slave_id; | |
9caff1e7 | 3567 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3568 | if (--i < 0) |
3569 | break; | |
3570 | if (slave_can_tx(slave)) { | |
3571 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3572 | return; | |
3573 | } | |
3574 | } | |
3575 | /* no slave that can tx has been found */ | |
2bb77ab4 | 3576 | dev_kfree_skb_any(skb); |
15077228 NA |
3577 | } |
3578 | ||
73958329 NA |
3579 | /** |
3580 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3581 | * @bond: bonding device to use | |
3582 | * | |
3583 | * Based on the value of the bonding device's packets_per_slave parameter | |
3584 | * this function generates a slave id, which is usually used as the next | |
3585 | * slave to transmit through. | |
3586 | */ | |
3587 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3588 | { | |
73958329 | 3589 | u32 slave_id; |
809fa972 HFS |
3590 | struct reciprocal_value reciprocal_packets_per_slave; |
3591 | int packets_per_slave = bond->params.packets_per_slave; | |
73958329 NA |
3592 | |
3593 | switch (packets_per_slave) { | |
3594 | case 0: | |
3595 | slave_id = prandom_u32(); | |
3596 | break; | |
3597 | case 1: | |
3598 | slave_id = bond->rr_tx_counter; | |
3599 | break; | |
3600 | default: | |
809fa972 HFS |
3601 | reciprocal_packets_per_slave = |
3602 | bond->params.reciprocal_packets_per_slave; | |
73958329 | 3603 | slave_id = reciprocal_divide(bond->rr_tx_counter, |
809fa972 | 3604 | reciprocal_packets_per_slave); |
73958329 NA |
3605 | break; |
3606 | } | |
3607 | bond->rr_tx_counter++; | |
3608 | ||
3609 | return slave_id; | |
3610 | } | |
3611 | ||
1da177e4 LT |
3612 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3613 | { | |
454d7c9b | 3614 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3615 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3616 | struct slave *slave; |
73958329 | 3617 | u32 slave_id; |
1da177e4 | 3618 | |
73958329 | 3619 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3620 | * default for sending IGMP traffic. For failover purposes one |
3621 | * needs to maintain some consistency for the interface that will | |
3622 | * send the join/membership reports. The curr_active_slave found | |
3623 | * will send all of this type of traffic. | |
cf5f9044 | 3624 | */ |
15077228 | 3625 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3626 | slave = rcu_dereference(bond->curr_active_slave); |
15077228 NA |
3627 | if (slave && slave_can_tx(slave)) |
3628 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3629 | else | |
3630 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3631 | } else { |
73958329 NA |
3632 | slave_id = bond_rr_gen_slave_id(bond); |
3633 | bond_xmit_slave_id(bond, skb, slave_id % bond->slave_cnt); | |
1da177e4 | 3634 | } |
0693e88e | 3635 | |
ec634fe3 | 3636 | return NETDEV_TX_OK; |
1da177e4 LT |
3637 | } |
3638 | ||
3639 | /* | |
3640 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3641 | * the bond has a usable interface. | |
3642 | */ | |
3643 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3644 | { | |
454d7c9b | 3645 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3646 | struct slave *slave; |
1da177e4 | 3647 | |
278b2083 | 3648 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3649 | if (slave) |
15077228 NA |
3650 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3651 | else | |
2bb77ab4 | 3652 | dev_kfree_skb_any(skb); |
0693e88e | 3653 | |
ec634fe3 | 3654 | return NETDEV_TX_OK; |
1da177e4 LT |
3655 | } |
3656 | ||
32819dc1 | 3657 | /* In bond_xmit_xor() , we determine the output device by using a pre- |
169a3e66 JV |
3658 | * determined xmit_hash_policy(), If the selected device is not enabled, |
3659 | * find the next active slave. | |
1da177e4 LT |
3660 | */ |
3661 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3662 | { | |
454d7c9b | 3663 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3664 | |
32819dc1 | 3665 | bond_xmit_slave_id(bond, skb, bond_xmit_hash(bond, skb, bond->slave_cnt)); |
0693e88e | 3666 | |
ec634fe3 | 3667 | return NETDEV_TX_OK; |
1da177e4 LT |
3668 | } |
3669 | ||
78a646ce | 3670 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3671 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3672 | { | |
454d7c9b | 3673 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3674 | struct slave *slave = NULL; |
9caff1e7 | 3675 | struct list_head *iter; |
1da177e4 | 3676 | |
9caff1e7 | 3677 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
3678 | if (bond_is_last_slave(bond, slave)) |
3679 | break; | |
3680 | if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { | |
3681 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1da177e4 | 3682 | |
78a646ce | 3683 | if (!skb2) { |
9152e26d | 3684 | net_err_ratelimited("%s: Error: %s: skb_clone() failed\n", |
3685 | bond_dev->name, __func__); | |
78a646ce | 3686 | continue; |
1da177e4 | 3687 | } |
78a646ce NA |
3688 | /* bond_dev_queue_xmit always returns 0 */ |
3689 | bond_dev_queue_xmit(bond, skb2, slave->dev); | |
1da177e4 LT |
3690 | } |
3691 | } | |
78a646ce NA |
3692 | if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP) |
3693 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3694 | else | |
2bb77ab4 | 3695 | dev_kfree_skb_any(skb); |
3d632c3f | 3696 | |
ec634fe3 | 3697 | return NETDEV_TX_OK; |
1da177e4 LT |
3698 | } |
3699 | ||
3700 | /*------------------------- Device initialization ---------------------------*/ | |
3701 | ||
bb1d9123 AG |
3702 | /* |
3703 | * Lookup the slave that corresponds to a qid | |
3704 | */ | |
3705 | static inline int bond_slave_override(struct bonding *bond, | |
3706 | struct sk_buff *skb) | |
3707 | { | |
bb1d9123 | 3708 | struct slave *slave = NULL; |
9caff1e7 | 3709 | struct list_head *iter; |
bb1d9123 | 3710 | |
0693e88e MM |
3711 | if (!skb->queue_mapping) |
3712 | return 1; | |
bb1d9123 AG |
3713 | |
3714 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 3715 | bond_for_each_slave_rcu(bond, slave, iter) { |
3716 | if (slave->queue_id == skb->queue_mapping) { | |
3717 | if (slave_can_tx(slave)) { | |
3718 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3719 | return 0; | |
3720 | } | |
3721 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
3722 | break; |
3723 | } | |
3724 | } | |
3725 | ||
3900f290 | 3726 | return 1; |
bb1d9123 AG |
3727 | } |
3728 | ||
374eeb5a | 3729 | |
f663dd9a | 3730 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
99932d4f | 3731 | void *accel_priv, select_queue_fallback_t fallback) |
bb1d9123 AG |
3732 | { |
3733 | /* | |
3734 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 3735 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3736 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3737 | * way down to the bonding driver. | |
3738 | */ | |
fd0e435b PO |
3739 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3740 | ||
374eeb5a NH |
3741 | /* |
3742 | * Save the original txq to restore before passing to the driver | |
3743 | */ | |
df4ab5b3 | 3744 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3745 | |
fd0e435b | 3746 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3747 | do { |
fd0e435b | 3748 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3749 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3750 | } |
3751 | return txq; | |
bb1d9123 AG |
3752 | } |
3753 | ||
0693e88e | 3754 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3755 | { |
bb1d9123 AG |
3756 | struct bonding *bond = netdev_priv(dev); |
3757 | ||
3758 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
3759 | if (!bond_slave_override(bond, skb)) | |
3760 | return NETDEV_TX_OK; | |
3761 | } | |
00829823 SH |
3762 | |
3763 | switch (bond->params.mode) { | |
3764 | case BOND_MODE_ROUNDROBIN: | |
3765 | return bond_xmit_roundrobin(skb, dev); | |
3766 | case BOND_MODE_ACTIVEBACKUP: | |
3767 | return bond_xmit_activebackup(skb, dev); | |
3768 | case BOND_MODE_XOR: | |
3769 | return bond_xmit_xor(skb, dev); | |
3770 | case BOND_MODE_BROADCAST: | |
3771 | return bond_xmit_broadcast(skb, dev); | |
3772 | case BOND_MODE_8023AD: | |
3773 | return bond_3ad_xmit_xor(skb, dev); | |
3774 | case BOND_MODE_ALB: | |
3775 | case BOND_MODE_TLB: | |
3776 | return bond_alb_xmit(skb, dev); | |
3777 | default: | |
3778 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3779 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3780 | dev->name, bond->params.mode); | |
00829823 | 3781 | WARN_ON_ONCE(1); |
2bb77ab4 | 3782 | dev_kfree_skb_any(skb); |
00829823 SH |
3783 | return NETDEV_TX_OK; |
3784 | } | |
3785 | } | |
3786 | ||
0693e88e MM |
3787 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3788 | { | |
3789 | struct bonding *bond = netdev_priv(dev); | |
3790 | netdev_tx_t ret = NETDEV_TX_OK; | |
3791 | ||
3792 | /* | |
3793 | * If we risk deadlock from transmitting this in the | |
3794 | * netpoll path, tell netpoll to queue the frame for later tx | |
3795 | */ | |
054bb880 | 3796 | if (unlikely(is_netpoll_tx_blocked(dev))) |
0693e88e MM |
3797 | return NETDEV_TX_BUSY; |
3798 | ||
278b2083 | 3799 | rcu_read_lock(); |
0965a1f3 | 3800 | if (bond_has_slaves(bond)) |
0693e88e MM |
3801 | ret = __bond_start_xmit(skb, dev); |
3802 | else | |
2bb77ab4 | 3803 | dev_kfree_skb_any(skb); |
278b2083 | 3804 | rcu_read_unlock(); |
0693e88e MM |
3805 | |
3806 | return ret; | |
3807 | } | |
00829823 | 3808 | |
bb5b052f AG |
3809 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3810 | struct ethtool_cmd *ecmd) | |
3811 | { | |
3812 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3813 | unsigned long speed = 0; |
9caff1e7 | 3814 | struct list_head *iter; |
dec1e90e | 3815 | struct slave *slave; |
bb5b052f AG |
3816 | |
3817 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3818 | ecmd->port = PORT_OTHER; | |
3819 | ||
3820 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
3821 | * do not need to check mode. Though link speed might not represent | |
3822 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3823 | * this is an accurate maximum. | |
3824 | */ | |
3825 | read_lock(&bond->lock); | |
9caff1e7 | 3826 | bond_for_each_slave(bond, slave, iter) { |
bb5b052f AG |
3827 | if (SLAVE_IS_OK(slave)) { |
3828 | if (slave->speed != SPEED_UNKNOWN) | |
3829 | speed += slave->speed; | |
3830 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3831 | slave->duplex != DUPLEX_UNKNOWN) | |
3832 | ecmd->duplex = slave->duplex; | |
3833 | } | |
3834 | } | |
3835 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
3836 | read_unlock(&bond->lock); | |
dec1e90e | 3837 | |
bb5b052f AG |
3838 | return 0; |
3839 | } | |
3840 | ||
217df670 | 3841 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3842 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3843 | { |
7826d43f JP |
3844 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3845 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3846 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3847 | BOND_ABI_VERSION); | |
217df670 JV |
3848 | } |
3849 | ||
7282d491 | 3850 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3851 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3852 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3853 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3854 | }; |
3855 | ||
eb7cc59a | 3856 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3857 | .ndo_init = bond_init, |
9e71626c | 3858 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3859 | .ndo_open = bond_open, |
3860 | .ndo_stop = bond_close, | |
00829823 | 3861 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3862 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3863 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3864 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3865 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 3866 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 3867 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 3868 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 3869 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 3870 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 3871 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 3872 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 3873 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
3874 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
3875 | .ndo_poll_controller = bond_poll_controller, | |
3876 | #endif | |
9232ecca JP |
3877 | .ndo_add_slave = bond_enslave, |
3878 | .ndo_del_slave = bond_release, | |
b2a103e6 | 3879 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
3880 | }; |
3881 | ||
b3f92b63 DG |
3882 | static const struct device_type bond_type = { |
3883 | .name = "bond", | |
3884 | }; | |
3885 | ||
9e2e61fb AW |
3886 | static void bond_destructor(struct net_device *bond_dev) |
3887 | { | |
3888 | struct bonding *bond = netdev_priv(bond_dev); | |
3889 | if (bond->wq) | |
3890 | destroy_workqueue(bond->wq); | |
3891 | free_netdev(bond_dev); | |
3892 | } | |
3893 | ||
0a2a78c4 | 3894 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 3895 | { |
454d7c9b | 3896 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3897 | |
1da177e4 LT |
3898 | /* initialize rwlocks */ |
3899 | rwlock_init(&bond->lock); | |
3900 | rwlock_init(&bond->curr_slave_lock); | |
d2991f75 | 3901 | bond->params = bonding_defaults; |
1da177e4 LT |
3902 | |
3903 | /* Initialize pointers */ | |
1da177e4 | 3904 | bond->dev = bond_dev; |
1da177e4 LT |
3905 | |
3906 | /* Initialize the device entry points */ | |
181470fc | 3907 | ether_setup(bond_dev); |
eb7cc59a | 3908 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 3909 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 3910 | |
9e2e61fb | 3911 | bond_dev->destructor = bond_destructor; |
1da177e4 | 3912 | |
b3f92b63 DG |
3913 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
3914 | ||
1da177e4 LT |
3915 | /* Initialize the device options */ |
3916 | bond_dev->tx_queue_len = 0; | |
3917 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 3918 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 3919 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 3920 | |
1da177e4 LT |
3921 | /* At first, we block adding VLANs. That's the only way to |
3922 | * prevent problems that occur when adding VLANs over an | |
3923 | * empty bond. The block will be removed once non-challenged | |
3924 | * slaves are enslaved. | |
3925 | */ | |
3926 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
3927 | ||
932ff279 | 3928 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
3929 | * transmitting */ |
3930 | bond_dev->features |= NETIF_F_LLTX; | |
3931 | ||
3932 | /* By default, we declare the bond to be fully | |
3933 | * VLAN hardware accelerated capable. Special | |
3934 | * care is taken in the various xmit functions | |
3935 | * when there are slaves that are not hw accel | |
3936 | * capable | |
3937 | */ | |
1da177e4 | 3938 | |
f9399814 WC |
3939 | /* Don't allow bond devices to change network namespaces. */ |
3940 | bond_dev->features |= NETIF_F_NETNS_LOCAL; | |
3941 | ||
b2a103e6 | 3942 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
3943 | NETIF_F_HW_VLAN_CTAG_TX | |
3944 | NETIF_F_HW_VLAN_CTAG_RX | | |
3945 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 3946 | |
34324dc2 | 3947 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 3948 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
3949 | } |
3950 | ||
c67dfb29 EB |
3951 | /* |
3952 | * Destroy a bonding device. | |
3953 | * Must be under rtnl_lock when this function is called. | |
3954 | */ | |
3955 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 3956 | { |
454d7c9b | 3957 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
3958 | struct list_head *iter; |
3959 | struct slave *slave; | |
a434e43f | 3960 | |
f6dc31a8 WC |
3961 | bond_netpoll_cleanup(bond_dev); |
3962 | ||
c67dfb29 | 3963 | /* Release the bonded slaves */ |
544a028e | 3964 | bond_for_each_slave(bond, slave, iter) |
dec1e90e | 3965 | __bond_release_one(bond_dev, slave->dev, true); |
90194264 | 3966 | pr_info("%s: Released all slaves\n", bond_dev->name); |
c67dfb29 | 3967 | |
a434e43f JV |
3968 | list_del(&bond->bond_list); |
3969 | ||
f073c7ca | 3970 | bond_debug_unregister(bond); |
a434e43f JV |
3971 | } |
3972 | ||
1da177e4 LT |
3973 | /*------------------------- Module initialization ---------------------------*/ |
3974 | ||
1da177e4 LT |
3975 | static int bond_check_params(struct bond_params *params) |
3976 | { | |
5a5c5fd4 | 3977 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
f3253339 | 3978 | struct bond_opt_value newval; |
3979 | const struct bond_opt_value *valptr; | |
8599b52e | 3980 | int arp_all_targets_value; |
f5b2b966 | 3981 | |
1da177e4 LT |
3982 | /* |
3983 | * Convert string parameters. | |
3984 | */ | |
3985 | if (mode) { | |
2b3798d5 NA |
3986 | bond_opt_initstr(&newval, mode); |
3987 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_MODE), &newval); | |
3988 | if (!valptr) { | |
3989 | pr_err("Error: Invalid bonding mode \"%s\"\n", mode); | |
1da177e4 LT |
3990 | return -EINVAL; |
3991 | } | |
2b3798d5 | 3992 | bond_mode = valptr->value; |
1da177e4 LT |
3993 | } |
3994 | ||
169a3e66 JV |
3995 | if (xmit_hash_policy) { |
3996 | if ((bond_mode != BOND_MODE_XOR) && | |
3997 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 3998 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
90194264 | 3999 | bond_mode_name(bond_mode)); |
169a3e66 | 4000 | } else { |
a4b32ce7 NA |
4001 | bond_opt_initstr(&newval, xmit_hash_policy); |
4002 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_XMIT_HASH), | |
4003 | &newval); | |
4004 | if (!valptr) { | |
a4aee5c8 | 4005 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
169a3e66 JV |
4006 | xmit_hash_policy); |
4007 | return -EINVAL; | |
4008 | } | |
a4b32ce7 | 4009 | xmit_hashtype = valptr->value; |
169a3e66 JV |
4010 | } |
4011 | } | |
4012 | ||
1da177e4 LT |
4013 | if (lacp_rate) { |
4014 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4015 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4016 | bond_mode_name(bond_mode)); | |
1da177e4 | 4017 | } else { |
d3131de7 NA |
4018 | bond_opt_initstr(&newval, lacp_rate); |
4019 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_LACP_RATE), | |
4020 | &newval); | |
4021 | if (!valptr) { | |
a4aee5c8 | 4022 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
d3131de7 | 4023 | lacp_rate); |
1da177e4 LT |
4024 | return -EINVAL; |
4025 | } | |
d3131de7 | 4026 | lacp_fast = valptr->value; |
1da177e4 LT |
4027 | } |
4028 | } | |
4029 | ||
fd989c83 | 4030 | if (ad_select) { |
9e5f5eeb NA |
4031 | bond_opt_initstr(&newval, lacp_rate); |
4032 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_AD_SELECT), | |
4033 | &newval); | |
4034 | if (!valptr) { | |
4035 | pr_err("Error: Invalid ad_select \"%s\"\n", ad_select); | |
fd989c83 JV |
4036 | return -EINVAL; |
4037 | } | |
9e5f5eeb NA |
4038 | params->ad_select = valptr->value; |
4039 | if (bond_mode != BOND_MODE_8023AD) | |
91565ebb | 4040 | pr_warn("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4041 | } else { |
4042 | params->ad_select = BOND_AD_STABLE; | |
4043 | } | |
4044 | ||
f5841306 | 4045 | if (max_bonds < 0) { |
91565ebb JP |
4046 | pr_warn("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4047 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4048 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4049 | } | |
4050 | ||
4051 | if (miimon < 0) { | |
91565ebb JP |
4052 | pr_warn("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4053 | miimon, INT_MAX); | |
b98d9c66 | 4054 | miimon = 0; |
1da177e4 LT |
4055 | } |
4056 | ||
4057 | if (updelay < 0) { | |
91565ebb JP |
4058 | pr_warn("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4059 | updelay, INT_MAX); | |
1da177e4 LT |
4060 | updelay = 0; |
4061 | } | |
4062 | ||
4063 | if (downdelay < 0) { | |
91565ebb JP |
4064 | pr_warn("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4065 | downdelay, INT_MAX); | |
1da177e4 LT |
4066 | downdelay = 0; |
4067 | } | |
4068 | ||
4069 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
91565ebb JP |
4070 | pr_warn("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4071 | use_carrier); | |
1da177e4 LT |
4072 | use_carrier = 1; |
4073 | } | |
4074 | ||
ad246c99 | 4075 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
91565ebb JP |
4076 | pr_warn("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
4077 | num_peer_notif); | |
ad246c99 BH |
4078 | num_peer_notif = 1; |
4079 | } | |
4080 | ||
834db4bc | 4081 | /* reset values for 802.3ad/TLB/ALB */ |
4082 | if (BOND_NO_USES_ARP(bond_mode)) { | |
1da177e4 | 4083 | if (!miimon) { |
91565ebb JP |
4084 | 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"); |
4085 | pr_warn("Forcing miimon to 100msec\n"); | |
fe9d04af | 4086 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4087 | } |
4088 | } | |
4089 | ||
bb1d9123 | 4090 | if (tx_queues < 1 || tx_queues > 255) { |
91565ebb JP |
4091 | pr_warn("Warning: tx_queues (%d) should be between 1 and 255, resetting to %d\n", |
4092 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
bb1d9123 AG |
4093 | tx_queues = BOND_DEFAULT_TX_QUEUES; |
4094 | } | |
4095 | ||
ebd8e497 | 4096 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
91565ebb JP |
4097 | pr_warn("Warning: all_slaves_active module parameter (%d), not of valid value (0/1), so it was set to 0\n", |
4098 | all_slaves_active); | |
ebd8e497 AG |
4099 | all_slaves_active = 0; |
4100 | } | |
4101 | ||
c2952c31 | 4102 | if (resend_igmp < 0 || resend_igmp > 255) { |
91565ebb JP |
4103 | pr_warn("Warning: resend_igmp (%d) should be between 0 and 255, resetting to %d\n", |
4104 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
c2952c31 FL |
4105 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
4106 | } | |
4107 | ||
aa59d851 NA |
4108 | bond_opt_initval(&newval, packets_per_slave); |
4109 | if (!bond_opt_parse(bond_opt_get(BOND_OPT_PACKETS_PER_SLAVE), &newval)) { | |
73958329 NA |
4110 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", |
4111 | packets_per_slave, USHRT_MAX); | |
4112 | packets_per_slave = 1; | |
4113 | } | |
4114 | ||
1da177e4 | 4115 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4116 | 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", |
4117 | updelay); | |
1da177e4 LT |
4118 | } |
4119 | ||
4120 | if (!miimon) { | |
4121 | if (updelay || downdelay) { | |
4122 | /* just warn the user the up/down delay will have | |
4123 | * no effect since miimon is zero... | |
4124 | */ | |
91565ebb JP |
4125 | 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", |
4126 | updelay, downdelay); | |
1da177e4 LT |
4127 | } |
4128 | } else { | |
4129 | /* don't allow arp monitoring */ | |
4130 | if (arp_interval) { | |
91565ebb JP |
4131 | pr_warn("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4132 | miimon, arp_interval); | |
1da177e4 LT |
4133 | arp_interval = 0; |
4134 | } | |
4135 | ||
4136 | if ((updelay % miimon) != 0) { | |
91565ebb JP |
4137 | pr_warn("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4138 | updelay, miimon, (updelay / miimon) * miimon); | |
1da177e4 LT |
4139 | } |
4140 | ||
4141 | updelay /= miimon; | |
4142 | ||
4143 | if ((downdelay % miimon) != 0) { | |
91565ebb JP |
4144 | pr_warn("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4145 | downdelay, miimon, | |
4146 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4147 | } |
4148 | ||
4149 | downdelay /= miimon; | |
4150 | } | |
4151 | ||
4152 | if (arp_interval < 0) { | |
91565ebb JP |
4153 | pr_warn("Warning: arp_interval module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4154 | arp_interval, INT_MAX); | |
7bdb04ed | 4155 | arp_interval = 0; |
1da177e4 LT |
4156 | } |
4157 | ||
5a5c5fd4 | 4158 | for (arp_ip_count = 0, i = 0; |
4159 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4160 | /* not complete check, but should be good enough to |
4161 | catch mistakes */ | |
89015c18 | 4162 | __be32 ip; |
4163 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || | |
4164 | IS_IP_TARGET_UNUSABLE_ADDRESS(ip)) { | |
91565ebb JP |
4165 | pr_warn("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4166 | arp_ip_target[i]); | |
1da177e4 LT |
4167 | arp_interval = 0; |
4168 | } else { | |
0afee4e8 VF |
4169 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4170 | arp_target[arp_ip_count++] = ip; | |
4171 | else | |
91565ebb JP |
4172 | pr_warn("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", |
4173 | &ip); | |
1da177e4 LT |
4174 | } |
4175 | } | |
4176 | ||
4177 | if (arp_interval && !arp_ip_count) { | |
4178 | /* don't allow arping if no arp_ip_target given... */ | |
91565ebb JP |
4179 | pr_warn("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4180 | arp_interval); | |
1da177e4 LT |
4181 | arp_interval = 0; |
4182 | } | |
4183 | ||
f5b2b966 | 4184 | if (arp_validate) { |
f5b2b966 | 4185 | if (!arp_interval) { |
a4aee5c8 | 4186 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4187 | return -EINVAL; |
4188 | } | |
4189 | ||
16228881 NA |
4190 | bond_opt_initstr(&newval, arp_validate); |
4191 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_VALIDATE), | |
4192 | &newval); | |
4193 | if (!valptr) { | |
a4aee5c8 | 4194 | pr_err("Error: invalid arp_validate \"%s\"\n", |
16228881 | 4195 | arp_validate); |
f5b2b966 JV |
4196 | return -EINVAL; |
4197 | } | |
16228881 NA |
4198 | arp_validate_value = valptr->value; |
4199 | } else { | |
f5b2b966 | 4200 | arp_validate_value = 0; |
16228881 | 4201 | } |
f5b2b966 | 4202 | |
8599b52e VF |
4203 | arp_all_targets_value = 0; |
4204 | if (arp_all_targets) { | |
edf36b24 NA |
4205 | bond_opt_initstr(&newval, arp_all_targets); |
4206 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_ARP_ALL_TARGETS), | |
4207 | &newval); | |
4208 | if (!valptr) { | |
8599b52e VF |
4209 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", |
4210 | arp_all_targets); | |
4211 | arp_all_targets_value = 0; | |
edf36b24 NA |
4212 | } else { |
4213 | arp_all_targets_value = valptr->value; | |
8599b52e VF |
4214 | } |
4215 | } | |
4216 | ||
1da177e4 | 4217 | if (miimon) { |
a4aee5c8 | 4218 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4219 | } else if (arp_interval) { |
16228881 NA |
4220 | valptr = bond_opt_get_val(BOND_OPT_ARP_VALIDATE, |
4221 | arp_validate_value); | |
a4aee5c8 | 4222 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
16228881 | 4223 | arp_interval, valptr->string, arp_ip_count); |
1da177e4 LT |
4224 | |
4225 | for (i = 0; i < arp_ip_count; i++) | |
90194264 | 4226 | pr_cont(" %s", arp_ip_target[i]); |
1da177e4 | 4227 | |
90194264 | 4228 | pr_cont("\n"); |
1da177e4 | 4229 | |
b8a9787e | 4230 | } else if (max_bonds) { |
1da177e4 LT |
4231 | /* miimon and arp_interval not set, we need one so things |
4232 | * work as expected, see bonding.txt for details | |
4233 | */ | |
90194264 | 4234 | 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 |
4235 | } |
4236 | ||
4237 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4238 | /* currently, using a primary only makes sense | |
4239 | * in active backup, TLB or ALB modes | |
4240 | */ | |
91565ebb JP |
4241 | pr_warn("Warning: %s primary device specified but has no effect in %s mode\n", |
4242 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4243 | primary = NULL; |
4244 | } | |
4245 | ||
a549952a | 4246 | if (primary && primary_reselect) { |
388d3a6d NA |
4247 | bond_opt_initstr(&newval, primary_reselect); |
4248 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_PRIMARY_RESELECT), | |
4249 | &newval); | |
4250 | if (!valptr) { | |
a4aee5c8 | 4251 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
388d3a6d | 4252 | primary_reselect); |
a549952a JP |
4253 | return -EINVAL; |
4254 | } | |
388d3a6d | 4255 | primary_reselect_value = valptr->value; |
a549952a JP |
4256 | } else { |
4257 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4258 | } | |
4259 | ||
3915c1e8 | 4260 | if (fail_over_mac) { |
1df6b6aa NA |
4261 | bond_opt_initstr(&newval, fail_over_mac); |
4262 | valptr = bond_opt_parse(bond_opt_get(BOND_OPT_FAIL_OVER_MAC), | |
4263 | &newval); | |
4264 | if (!valptr) { | |
a4aee5c8 | 4265 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
1df6b6aa | 4266 | fail_over_mac); |
3915c1e8 JV |
4267 | return -EINVAL; |
4268 | } | |
1df6b6aa | 4269 | fail_over_mac_value = valptr->value; |
3915c1e8 | 4270 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) |
91565ebb | 4271 | pr_warn("Warning: fail_over_mac only affects active-backup mode\n"); |
3915c1e8 JV |
4272 | } else { |
4273 | fail_over_mac_value = BOND_FOM_NONE; | |
4274 | } | |
dd957c57 | 4275 | |
3a7129e5 | 4276 | if (lp_interval == 0) { |
91565ebb JP |
4277 | pr_warn("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", |
4278 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
3a7129e5 | 4279 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
4280 | } | |
4281 | ||
1da177e4 LT |
4282 | /* fill params struct with the proper values */ |
4283 | params->mode = bond_mode; | |
169a3e66 | 4284 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4285 | params->miimon = miimon; |
ad246c99 | 4286 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4287 | params->arp_interval = arp_interval; |
f5b2b966 | 4288 | params->arp_validate = arp_validate_value; |
8599b52e | 4289 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4290 | params->updelay = updelay; |
4291 | params->downdelay = downdelay; | |
4292 | params->use_carrier = use_carrier; | |
4293 | params->lacp_fast = lacp_fast; | |
4294 | params->primary[0] = 0; | |
a549952a | 4295 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4296 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4297 | params->tx_queues = tx_queues; |
ebd8e497 | 4298 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4299 | params->resend_igmp = resend_igmp; |
655f8919 | 4300 | params->min_links = min_links; |
3a7129e5 | 4301 | params->lp_interval = lp_interval; |
809fa972 HFS |
4302 | params->packets_per_slave = packets_per_slave; |
4303 | if (packets_per_slave > 0) { | |
4304 | params->reciprocal_packets_per_slave = | |
4305 | reciprocal_value(packets_per_slave); | |
4306 | } else { | |
4307 | /* reciprocal_packets_per_slave is unused if | |
4308 | * packets_per_slave is 0 or 1, just initialize it | |
4309 | */ | |
4310 | params->reciprocal_packets_per_slave = | |
4311 | (struct reciprocal_value) { 0 }; | |
4312 | } | |
4313 | ||
1da177e4 LT |
4314 | if (primary) { |
4315 | strncpy(params->primary, primary, IFNAMSIZ); | |
4316 | params->primary[IFNAMSIZ - 1] = 0; | |
4317 | } | |
4318 | ||
4319 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4320 | ||
4321 | return 0; | |
4322 | } | |
4323 | ||
0daa2303 | 4324 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4325 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4326 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4327 | |
e8a0464c DM |
4328 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4329 | struct netdev_queue *txq, | |
4330 | void *_unused) | |
c773e847 DM |
4331 | { |
4332 | lockdep_set_class(&txq->_xmit_lock, | |
4333 | &bonding_netdev_xmit_lock_key); | |
4334 | } | |
4335 | ||
4336 | static void bond_set_lockdep_class(struct net_device *dev) | |
4337 | { | |
cf508b12 DM |
4338 | lockdep_set_class(&dev->addr_list_lock, |
4339 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4340 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4341 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4342 | } |
4343 | ||
181470fc SH |
4344 | /* |
4345 | * Called from registration process | |
4346 | */ | |
4347 | static int bond_init(struct net_device *bond_dev) | |
4348 | { | |
4349 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4350 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4351 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4352 | |
4353 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4354 | ||
9fe0617d NH |
4355 | /* |
4356 | * Initialize locks that may be required during | |
4357 | * en/deslave operations. All of the bond_open work | |
4358 | * (of which this is part) should really be moved to | |
4359 | * a phase prior to dev_open | |
4360 | */ | |
4361 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4362 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4363 | ||
181470fc SH |
4364 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4365 | if (!bond->wq) | |
4366 | return -ENOMEM; | |
4367 | ||
4368 | bond_set_lockdep_class(bond_dev); | |
4369 | ||
ec87fd3b | 4370 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4371 | |
6151b3d4 | 4372 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4373 | |
f073c7ca TI |
4374 | bond_debug_register(bond); |
4375 | ||
409cc1f8 JP |
4376 | /* Ensure valid dev_addr */ |
4377 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4378 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4379 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4380 | |
181470fc SH |
4381 | return 0; |
4382 | } | |
4383 | ||
0a2a78c4 | 4384 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4385 | { |
efacb309 | 4386 | return tx_queues; |
d5da4510 JP |
4387 | } |
4388 | ||
dfe60397 | 4389 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4390 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4391 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4392 | * set up our sysfs entries. | |
4393 | */ | |
ec87fd3b | 4394 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4395 | { |
4396 | struct net_device *bond_dev; | |
4397 | int res; | |
4398 | ||
4399 | rtnl_lock(); | |
027ea041 | 4400 | |
1c5cae81 JP |
4401 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4402 | name ? name : "bond%d", | |
4403 | bond_setup, tx_queues); | |
dfe60397 | 4404 | if (!bond_dev) { |
a4aee5c8 | 4405 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4406 | rtnl_unlock(); |
4407 | return -ENOMEM; | |
dfe60397 MW |
4408 | } |
4409 | ||
ec87fd3b | 4410 | dev_net_set(bond_dev, net); |
88ead977 EB |
4411 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4412 | ||
dfe60397 | 4413 | res = register_netdevice(bond_dev); |
0daa2303 | 4414 | |
e826eafa PO |
4415 | netif_carrier_off(bond_dev); |
4416 | ||
7e083840 | 4417 | rtnl_unlock(); |
9e2e61fb AW |
4418 | if (res < 0) |
4419 | bond_destructor(bond_dev); | |
30c15ba9 | 4420 | return res; |
dfe60397 MW |
4421 | } |
4422 | ||
2c8c1e72 | 4423 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4424 | { |
15449745 | 4425 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4426 | |
4427 | bn->net = net; | |
4428 | INIT_LIST_HEAD(&bn->dev_list); | |
4429 | ||
ec87fd3b | 4430 | bond_create_proc_dir(bn); |
4c22400a | 4431 | bond_create_sysfs(bn); |
87a7b84b | 4432 | |
15449745 | 4433 | return 0; |
ec87fd3b EB |
4434 | } |
4435 | ||
2c8c1e72 | 4436 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4437 | { |
15449745 | 4438 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4439 | struct bonding *bond, *tmp_bond; |
4440 | LIST_HEAD(list); | |
ec87fd3b | 4441 | |
4c22400a | 4442 | bond_destroy_sysfs(bn); |
ec87fd3b | 4443 | bond_destroy_proc_dir(bn); |
69b0216a | 4444 | |
4445 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4446 | rtnl_lock(); | |
4447 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4448 | unregister_netdevice_queue(bond->dev, &list); | |
4449 | unregister_netdevice_many(&list); | |
4450 | rtnl_unlock(); | |
ec87fd3b EB |
4451 | } |
4452 | ||
4453 | static struct pernet_operations bond_net_ops = { | |
4454 | .init = bond_net_init, | |
4455 | .exit = bond_net_exit, | |
15449745 EB |
4456 | .id = &bond_net_id, |
4457 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4458 | }; |
4459 | ||
1da177e4 LT |
4460 | static int __init bonding_init(void) |
4461 | { | |
1da177e4 LT |
4462 | int i; |
4463 | int res; | |
4464 | ||
bd33acc3 | 4465 | pr_info("%s", bond_version); |
1da177e4 | 4466 | |
dfe60397 | 4467 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4468 | if (res) |
dfe60397 | 4469 | goto out; |
1da177e4 | 4470 | |
15449745 | 4471 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4472 | if (res) |
4473 | goto out; | |
027ea041 | 4474 | |
0a2a78c4 | 4475 | res = bond_netlink_init(); |
88ead977 | 4476 | if (res) |
6639104b | 4477 | goto err_link; |
88ead977 | 4478 | |
f073c7ca TI |
4479 | bond_create_debugfs(); |
4480 | ||
1da177e4 | 4481 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4482 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4483 | if (res) |
4484 | goto err; | |
1da177e4 LT |
4485 | } |
4486 | ||
1da177e4 | 4487 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4488 | out: |
1da177e4 | 4489 | return res; |
88ead977 | 4490 | err: |
db298686 | 4491 | bond_destroy_debugfs(); |
0a2a78c4 | 4492 | bond_netlink_fini(); |
6639104b | 4493 | err_link: |
15449745 | 4494 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4495 | goto out; |
dfe60397 | 4496 | |
1da177e4 LT |
4497 | } |
4498 | ||
4499 | static void __exit bonding_exit(void) | |
4500 | { | |
4501 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4502 | ||
f073c7ca | 4503 | bond_destroy_debugfs(); |
ae68c398 | 4504 | |
0a2a78c4 | 4505 | bond_netlink_fini(); |
ffcdedb6 | 4506 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4507 | |
4508 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4509 | /* |
4510 | * Make sure we don't have an imbalance on our netpoll blocking | |
4511 | */ | |
4512 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4513 | #endif |
1da177e4 LT |
4514 | } |
4515 | ||
4516 | module_init(bonding_init); | |
4517 | module_exit(bonding_exit); | |
4518 | MODULE_LICENSE("GPL"); | |
4519 | MODULE_VERSION(DRV_VERSION); | |
4520 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4521 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |