]>
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> |
73958329 | 82 | #include <linux/reciprocal_div.h> |
1da177e4 LT |
83 | #include "bonding.h" |
84 | #include "bond_3ad.h" | |
85 | #include "bond_alb.h" | |
86 | ||
87 | /*---------------------------- Module parameters ----------------------------*/ | |
88 | ||
89 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
90 | #define BOND_LINK_MON_INTERV 0 | |
91 | #define BOND_LINK_ARP_INTERV 0 | |
92 | ||
93 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 94 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 95 | static int num_peer_notif = 1; |
1da177e4 | 96 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
97 | static int updelay; |
98 | static int downdelay; | |
1da177e4 | 99 | static int use_carrier = 1; |
3d632c3f SH |
100 | static char *mode; |
101 | static char *primary; | |
a549952a | 102 | static char *primary_reselect; |
3d632c3f | 103 | static char *lacp_rate; |
655f8919 | 104 | static int min_links; |
3d632c3f SH |
105 | static char *ad_select; |
106 | static char *xmit_hash_policy; | |
1da177e4 | 107 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
108 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
109 | static char *arp_validate; | |
8599b52e | 110 | static char *arp_all_targets; |
3d632c3f | 111 | static char *fail_over_mac; |
b07ea07b | 112 | static int all_slaves_active; |
d2991f75 | 113 | static struct bond_params bonding_defaults; |
c2952c31 | 114 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
73958329 | 115 | static int packets_per_slave = 1; |
3a7129e5 | 116 | static int lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; |
1da177e4 LT |
117 | |
118 | module_param(max_bonds, int, 0); | |
119 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
120 | module_param(tx_queues, int, 0); |
121 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 122 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
123 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
124 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 125 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
126 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
127 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
128 | module_param(miimon, int, 0); |
129 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
130 | module_param(updelay, int, 0); | |
131 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
132 | module_param(downdelay, int, 0); | |
2ac47660 MW |
133 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
134 | "in milliseconds"); | |
1da177e4 | 135 | module_param(use_carrier, int, 0); |
2ac47660 MW |
136 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
137 | "0 for off, 1 for on (default)"); | |
1da177e4 | 138 | module_param(mode, charp, 0); |
90e62474 | 139 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
140 | "1 for active-backup, 2 for balance-xor, " |
141 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
142 | "6 for balance-alb"); | |
1da177e4 LT |
143 | module_param(primary, charp, 0); |
144 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
145 | module_param(primary_reselect, charp, 0); |
146 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
147 | "once it comes up; " | |
148 | "0 for always (default), " | |
149 | "1 for only if speed of primary is " | |
150 | "better, " | |
151 | "2 for only on active slave " | |
152 | "failure"); | |
1da177e4 | 153 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
154 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
155 | "0 for slow, 1 for fast"); | |
fd989c83 | 156 | module_param(ad_select, charp, 0); |
90e62474 AG |
157 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
158 | "0 for stable (default), 1 for bandwidth, " | |
159 | "2 for count"); | |
655f8919 | 160 | module_param(min_links, int, 0); |
161 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
162 | ||
169a3e66 | 163 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
164 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
165 | "0 for layer 2 (default), 1 for layer 3+4, " | |
32819dc1 NA |
166 | "2 for layer 2+3, 3 for encap layer 2+3, " |
167 | "4 for encap layer 3+4"); | |
1da177e4 LT |
168 | module_param(arp_interval, int, 0); |
169 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
170 | module_param_array(arp_ip_target, charp, NULL, 0); | |
171 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 172 | module_param(arp_validate, charp, 0); |
90e62474 AG |
173 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
174 | "0 for none (default), 1 for active, " | |
175 | "2 for backup, 3 for all"); | |
8599b52e VF |
176 | module_param(arp_all_targets, charp, 0); |
177 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 178 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
179 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
180 | "the same MAC; 0 for none (default), " | |
181 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
182 | module_param(all_slaves_active, int, 0); |
183 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 184 | "by setting active flag for all slaves; " |
ebd8e497 | 185 | "0 for never (default), 1 for always."); |
c2952c31 | 186 | module_param(resend_igmp, int, 0); |
90e62474 AG |
187 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
188 | "link failure"); | |
73958329 NA |
189 | module_param(packets_per_slave, int, 0); |
190 | MODULE_PARM_DESC(packets_per_slave, "Packets to send per slave in balance-rr " | |
191 | "mode; 0 for a random slave, 1 packet per " | |
192 | "slave (default), >1 packets per slave."); | |
3a7129e5 | 193 | module_param(lp_interval, uint, 0); |
194 | MODULE_PARM_DESC(lp_interval, "The number of seconds between instances where " | |
195 | "the bonding driver sends learning packets to " | |
196 | "each slaves peer switch. The default is 1."); | |
1da177e4 LT |
197 | |
198 | /*----------------------------- Global variables ----------------------------*/ | |
199 | ||
e843fa50 | 200 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 201 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
202 | #endif |
203 | ||
f99189b1 | 204 | int bond_net_id __read_mostly; |
1da177e4 | 205 | |
3d632c3f SH |
206 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
207 | static int arp_ip_count; | |
1da177e4 | 208 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
209 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
210 | static int lacp_fast; | |
1da177e4 | 211 | |
e97fd7c6 | 212 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
213 | { "slow", AD_LACP_SLOW}, |
214 | { "fast", AD_LACP_FAST}, | |
215 | { NULL, -1}, | |
216 | }; | |
217 | ||
e97fd7c6 | 218 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
219 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
220 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
221 | { "balance-xor", BOND_MODE_XOR}, | |
222 | { "broadcast", BOND_MODE_BROADCAST}, | |
223 | { "802.3ad", BOND_MODE_8023AD}, | |
224 | { "balance-tlb", BOND_MODE_TLB}, | |
225 | { "balance-alb", BOND_MODE_ALB}, | |
226 | { NULL, -1}, | |
227 | }; | |
228 | ||
e97fd7c6 | 229 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
230 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
231 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 232 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
32819dc1 NA |
233 | { "encap2+3", BOND_XMIT_POLICY_ENCAP23}, |
234 | { "encap3+4", BOND_XMIT_POLICY_ENCAP34}, | |
169a3e66 JV |
235 | { NULL, -1}, |
236 | }; | |
237 | ||
8599b52e VF |
238 | const struct bond_parm_tbl arp_all_targets_tbl[] = { |
239 | { "any", BOND_ARP_TARGETS_ANY}, | |
240 | { "all", BOND_ARP_TARGETS_ALL}, | |
241 | { NULL, -1}, | |
242 | }; | |
243 | ||
e97fd7c6 | 244 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
245 | { "none", BOND_ARP_VALIDATE_NONE}, |
246 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
247 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
248 | { "all", BOND_ARP_VALIDATE_ALL}, | |
249 | { NULL, -1}, | |
250 | }; | |
251 | ||
e97fd7c6 | 252 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
253 | { "none", BOND_FOM_NONE}, |
254 | { "active", BOND_FOM_ACTIVE}, | |
255 | { "follow", BOND_FOM_FOLLOW}, | |
256 | { NULL, -1}, | |
257 | }; | |
258 | ||
a549952a JP |
259 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
260 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
261 | { "better", BOND_PRI_RESELECT_BETTER}, | |
262 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
263 | { NULL, -1}, | |
264 | }; | |
265 | ||
fd989c83 JV |
266 | struct bond_parm_tbl ad_select_tbl[] = { |
267 | { "stable", BOND_AD_STABLE}, | |
268 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
269 | { "count", BOND_AD_COUNT}, | |
270 | { NULL, -1}, | |
271 | }; | |
272 | ||
1da177e4 LT |
273 | /*-------------------------- Forward declarations ---------------------------*/ |
274 | ||
181470fc | 275 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 276 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
277 | |
278 | /*---------------------------- General routines -----------------------------*/ | |
279 | ||
bd33acc3 | 280 | const char *bond_mode_name(int mode) |
1da177e4 | 281 | { |
77afc92b HE |
282 | static const char *names[] = { |
283 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
284 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
285 | [BOND_MODE_XOR] = "load balancing (xor)", | |
286 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 287 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
288 | [BOND_MODE_TLB] = "transmit load balancing", |
289 | [BOND_MODE_ALB] = "adaptive load balancing", | |
290 | }; | |
291 | ||
f5280948 | 292 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 293 | return "unknown"; |
77afc92b HE |
294 | |
295 | return names[mode]; | |
1da177e4 LT |
296 | } |
297 | ||
298 | /*---------------------------------- VLAN -----------------------------------*/ | |
299 | ||
1da177e4 LT |
300 | /** |
301 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 302 | * |
1da177e4 LT |
303 | * @bond: bond device that got this skb for tx. |
304 | * @skb: hw accel VLAN tagged skb to transmit | |
305 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 306 | */ |
d316dedd | 307 | void bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
3d632c3f | 308 | struct net_device *slave_dev) |
1da177e4 | 309 | { |
8387451e | 310 | skb->dev = slave_dev; |
374eeb5a | 311 | |
5ee31c68 | 312 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
313 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
314 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 315 | |
e15c3c22 | 316 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 317 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 318 | else |
f6dc31a8 | 319 | dev_queue_xmit(skb); |
1da177e4 LT |
320 | } |
321 | ||
322 | /* | |
cc0e4070 JP |
323 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
324 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
325 | * a. This operation is performed in IOCTL context, |
326 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
327 | * c. Holding a lock with BH disabled while directly calling a base driver | |
328 | * entry point is generally a BAD idea. | |
3d632c3f | 329 | * |
1da177e4 LT |
330 | * The design of synchronization/protection for this operation in the 8021q |
331 | * module is good for one or more VLAN devices over a single physical device | |
332 | * and cannot be extended for a teaming solution like bonding, so there is a | |
333 | * potential race condition here where a net device from the vlan group might | |
334 | * be referenced (either by a base driver or the 8021q code) while it is being | |
335 | * removed from the system. However, it turns out we're not making matters | |
336 | * worse, and if it works for regular VLAN usage it will work here too. | |
337 | */ | |
338 | ||
1da177e4 LT |
339 | /** |
340 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
341 | * @bond_dev: bonding net device that got called | |
342 | * @vid: vlan id being added | |
343 | */ | |
80d5c368 PM |
344 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
345 | __be16 proto, u16 vid) | |
1da177e4 | 346 | { |
454d7c9b | 347 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 348 | struct slave *slave, *rollback_slave; |
9caff1e7 | 349 | struct list_head *iter; |
dec1e90e | 350 | int res; |
1da177e4 | 351 | |
9caff1e7 | 352 | bond_for_each_slave(bond, slave, iter) { |
80d5c368 | 353 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
354 | if (res) |
355 | goto unwind; | |
1da177e4 LT |
356 | } |
357 | ||
8e586137 | 358 | return 0; |
87002b03 JP |
359 | |
360 | unwind: | |
81f23b13 | 361 | /* unwind to the slave that failed */ |
9caff1e7 | 362 | bond_for_each_slave(bond, rollback_slave, iter) { |
81f23b13 VF |
363 | if (rollback_slave == slave) |
364 | break; | |
365 | ||
366 | vlan_vid_del(rollback_slave->dev, proto, vid); | |
367 | } | |
87002b03 JP |
368 | |
369 | return res; | |
1da177e4 LT |
370 | } |
371 | ||
372 | /** | |
373 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
374 | * @bond_dev: bonding net device that got called | |
375 | * @vid: vlan id being removed | |
376 | */ | |
80d5c368 PM |
377 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
378 | __be16 proto, u16 vid) | |
1da177e4 | 379 | { |
454d7c9b | 380 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 381 | struct list_head *iter; |
1da177e4 | 382 | struct slave *slave; |
1da177e4 | 383 | |
9caff1e7 | 384 | bond_for_each_slave(bond, slave, iter) |
80d5c368 | 385 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 | 386 | |
e868b0c9 VF |
387 | if (bond_is_lb(bond)) |
388 | bond_alb_clear_vlan(bond, vid); | |
8e586137 JP |
389 | |
390 | return 0; | |
1da177e4 LT |
391 | } |
392 | ||
1da177e4 LT |
393 | /*------------------------------- Link status -------------------------------*/ |
394 | ||
ff59c456 JV |
395 | /* |
396 | * Set the carrier state for the master according to the state of its | |
397 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
398 | * do special 802.3ad magic. | |
399 | * | |
400 | * Returns zero if carrier state does not change, nonzero if it does. | |
401 | */ | |
402 | static int bond_set_carrier(struct bonding *bond) | |
403 | { | |
9caff1e7 | 404 | struct list_head *iter; |
ff59c456 | 405 | struct slave *slave; |
ff59c456 | 406 | |
0965a1f3 | 407 | if (!bond_has_slaves(bond)) |
ff59c456 JV |
408 | goto down; |
409 | ||
410 | if (bond->params.mode == BOND_MODE_8023AD) | |
411 | return bond_3ad_set_carrier(bond); | |
412 | ||
9caff1e7 | 413 | bond_for_each_slave(bond, slave, iter) { |
ff59c456 JV |
414 | if (slave->link == BOND_LINK_UP) { |
415 | if (!netif_carrier_ok(bond->dev)) { | |
416 | netif_carrier_on(bond->dev); | |
417 | return 1; | |
418 | } | |
419 | return 0; | |
420 | } | |
421 | } | |
422 | ||
423 | down: | |
424 | if (netif_carrier_ok(bond->dev)) { | |
425 | netif_carrier_off(bond->dev); | |
426 | return 1; | |
427 | } | |
428 | return 0; | |
429 | } | |
430 | ||
1da177e4 LT |
431 | /* |
432 | * Get link speed and duplex from the slave's base driver | |
433 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 434 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 435 | * and return. |
1da177e4 | 436 | */ |
13b95fb7 | 437 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
438 | { |
439 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 440 | struct ethtool_cmd ecmd; |
5d30530e | 441 | u32 slave_speed; |
61a44b9c | 442 | int res; |
1da177e4 | 443 | |
589665f5 DC |
444 | slave->speed = SPEED_UNKNOWN; |
445 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 446 | |
4bc71cb9 | 447 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 448 | if (res < 0) |
13b95fb7 | 449 | return; |
1da177e4 | 450 | |
4bc71cb9 | 451 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 452 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 453 | return; |
1da177e4 | 454 | |
4bc71cb9 | 455 | switch (ecmd.duplex) { |
1da177e4 LT |
456 | case DUPLEX_FULL: |
457 | case DUPLEX_HALF: | |
458 | break; | |
459 | default: | |
13b95fb7 | 460 | return; |
1da177e4 LT |
461 | } |
462 | ||
5d30530e | 463 | slave->speed = slave_speed; |
4bc71cb9 | 464 | slave->duplex = ecmd.duplex; |
1da177e4 | 465 | |
13b95fb7 | 466 | return; |
1da177e4 LT |
467 | } |
468 | ||
07699f9a | 469 | const char *bond_slave_link_status(s8 link) |
470 | { | |
471 | switch (link) { | |
472 | case BOND_LINK_UP: | |
473 | return "up"; | |
474 | case BOND_LINK_FAIL: | |
475 | return "going down"; | |
476 | case BOND_LINK_DOWN: | |
477 | return "down"; | |
478 | case BOND_LINK_BACK: | |
479 | return "going back"; | |
480 | default: | |
481 | return "unknown"; | |
482 | } | |
483 | } | |
484 | ||
1da177e4 LT |
485 | /* |
486 | * if <dev> supports MII link status reporting, check its link status. | |
487 | * | |
488 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 489 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
490 | * |
491 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
492 | * can't tell and just pretend it is), or 0, meaning that the link is | |
493 | * down. | |
494 | * | |
495 | * If reporting is non-zero, instead of faking link up, return -1 if | |
496 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
497 | * support them). If use_carrier is set, return whatever it says. | |
498 | * It'd be nice if there was a good way to tell if a driver supports | |
499 | * netif_carrier, but there really isn't. | |
500 | */ | |
3d632c3f SH |
501 | static int bond_check_dev_link(struct bonding *bond, |
502 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 503 | { |
eb7cc59a | 504 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 505 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
506 | struct ifreq ifr; |
507 | struct mii_ioctl_data *mii; | |
1da177e4 | 508 | |
6c988853 PG |
509 | if (!reporting && !netif_running(slave_dev)) |
510 | return 0; | |
511 | ||
eb7cc59a | 512 | if (bond->params.use_carrier) |
1da177e4 | 513 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 514 | |
29112f4e | 515 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
516 | if (slave_dev->ethtool_ops->get_link) |
517 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
518 | BMSR_LSTATUS : 0; | |
29112f4e | 519 | |
3d632c3f | 520 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 521 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
522 | if (ioctl) { |
523 | /* TODO: set pointer to correct ioctl on a per team member */ | |
524 | /* bases to make this more efficient. that is, once */ | |
525 | /* we determine the correct ioctl, we will always */ | |
526 | /* call it and not the others for that team */ | |
527 | /* member. */ | |
528 | ||
529 | /* | |
530 | * We cannot assume that SIOCGMIIPHY will also read a | |
531 | * register; not all network drivers (e.g., e100) | |
532 | * support that. | |
533 | */ | |
534 | ||
535 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
536 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
537 | mii = if_mii(&ifr); | |
538 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
539 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
540 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
541 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
542 | } |
543 | } | |
544 | ||
1da177e4 LT |
545 | /* |
546 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 547 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
548 | * cannot report link status). If not reporting, pretend |
549 | * we're ok. | |
550 | */ | |
3d632c3f | 551 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
552 | } |
553 | ||
554 | /*----------------------------- Multicast list ------------------------------*/ | |
555 | ||
1da177e4 LT |
556 | /* |
557 | * Push the promiscuity flag down to appropriate slaves | |
558 | */ | |
7e1a1ac1 | 559 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 560 | { |
9caff1e7 | 561 | struct list_head *iter; |
7e1a1ac1 | 562 | int err = 0; |
9caff1e7 | 563 | |
1da177e4 LT |
564 | if (USES_PRIMARY(bond->params.mode)) { |
565 | /* write lock already acquired */ | |
566 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
567 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
568 | inc); | |
1da177e4 LT |
569 | } |
570 | } else { | |
571 | struct slave *slave; | |
dec1e90e | 572 | |
9caff1e7 | 573 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
574 | err = dev_set_promiscuity(slave->dev, inc); |
575 | if (err) | |
576 | return err; | |
1da177e4 LT |
577 | } |
578 | } | |
7e1a1ac1 | 579 | return err; |
1da177e4 LT |
580 | } |
581 | ||
582 | /* | |
583 | * Push the allmulti flag down to all slaves | |
584 | */ | |
7e1a1ac1 | 585 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 586 | { |
9caff1e7 | 587 | struct list_head *iter; |
7e1a1ac1 | 588 | int err = 0; |
9caff1e7 | 589 | |
1da177e4 LT |
590 | if (USES_PRIMARY(bond->params.mode)) { |
591 | /* write lock already acquired */ | |
592 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
593 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
594 | inc); | |
1da177e4 LT |
595 | } |
596 | } else { | |
597 | struct slave *slave; | |
dec1e90e | 598 | |
9caff1e7 | 599 | bond_for_each_slave(bond, slave, iter) { |
7e1a1ac1 WC |
600 | err = dev_set_allmulti(slave->dev, inc); |
601 | if (err) | |
602 | return err; | |
1da177e4 LT |
603 | } |
604 | } | |
7e1a1ac1 | 605 | return err; |
1da177e4 LT |
606 | } |
607 | ||
5a37e8ca FL |
608 | /* |
609 | * Retrieve the list of registered multicast addresses for the bonding | |
610 | * device and retransmit an IGMP JOIN request to the current active | |
611 | * slave. | |
612 | */ | |
f2369109 | 613 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca | 614 | { |
f2369109 | 615 | struct bonding *bond = container_of(work, struct bonding, |
616 | mcast_work.work); | |
617 | ||
4aa5dee4 | 618 | if (!rtnl_trylock()) { |
4beac029 | 619 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 620 | return; |
5a37e8ca | 621 | } |
4aa5dee4 | 622 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
5a37e8ca | 623 | |
4f5474e7 NA |
624 | if (bond->igmp_retrans > 1) { |
625 | bond->igmp_retrans--; | |
c2952c31 | 626 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 | 627 | } |
f2369109 | 628 | rtnl_unlock(); |
5a37e8ca FL |
629 | } |
630 | ||
303d1cbf | 631 | /* Flush bond's hardware addresses from slave |
1da177e4 | 632 | */ |
303d1cbf | 633 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 634 | struct net_device *slave_dev) |
1da177e4 | 635 | { |
454d7c9b | 636 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 637 | |
303d1cbf JV |
638 | dev_uc_unsync(slave_dev, bond_dev); |
639 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 LT |
640 | |
641 | if (bond->params.mode == BOND_MODE_8023AD) { | |
642 | /* del lacpdu mc addr from mc list */ | |
643 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
644 | ||
22bedad3 | 645 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
646 | } |
647 | } | |
648 | ||
649 | /*--------------------------- Active slave change ---------------------------*/ | |
650 | ||
303d1cbf JV |
651 | /* Update the hardware address list and promisc/allmulti for the new and |
652 | * old active slaves (if any). Modes that are !USES_PRIMARY keep all | |
653 | * slaves up date at all times; only the USES_PRIMARY modes need to call | |
654 | * this function to swap these settings during a failover. | |
1da177e4 | 655 | */ |
303d1cbf JV |
656 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
657 | struct slave *old_active) | |
1da177e4 | 658 | { |
7864a1ad VF |
659 | ASSERT_RTNL(); |
660 | ||
1da177e4 | 661 | if (old_active) { |
3d632c3f | 662 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 663 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 664 | |
3d632c3f | 665 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 666 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 667 | |
303d1cbf | 668 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
669 | } |
670 | ||
671 | if (new_active) { | |
7e1a1ac1 | 672 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 673 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 674 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 675 | |
3d632c3f | 676 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 677 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 678 | |
d632ce98 | 679 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
680 | dev_uc_sync(new_active->dev, bond->dev); |
681 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 682 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
683 | } |
684 | } | |
685 | ||
ae0d6750 | 686 | /** |
687 | * bond_set_dev_addr - clone slave's address to bond | |
688 | * @bond_dev: bond net device | |
689 | * @slave_dev: slave net device | |
690 | * | |
691 | * Should be called with RTNL held. | |
692 | */ | |
693 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
694 | struct net_device *slave_dev) | |
695 | { | |
008aebde NA |
696 | pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
697 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 698 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
699 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
700 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
701 | } | |
702 | ||
3915c1e8 JV |
703 | /* |
704 | * bond_do_fail_over_mac | |
705 | * | |
706 | * Perform special MAC address swapping for fail_over_mac settings | |
707 | * | |
b2e7aceb | 708 | * Called with RTNL, curr_slave_lock for write_bh. |
3915c1e8 JV |
709 | */ |
710 | static void bond_do_fail_over_mac(struct bonding *bond, | |
711 | struct slave *new_active, | |
712 | struct slave *old_active) | |
1f78d9f9 | 713 | __releases(&bond->curr_slave_lock) |
1f78d9f9 | 714 | __acquires(&bond->curr_slave_lock) |
3915c1e8 JV |
715 | { |
716 | u8 tmp_mac[ETH_ALEN]; | |
717 | struct sockaddr saddr; | |
718 | int rv; | |
719 | ||
720 | switch (bond->params.fail_over_mac) { | |
721 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 722 | if (new_active) { |
7d26bb10 | 723 | write_unlock_bh(&bond->curr_slave_lock); |
ae0d6750 | 724 | bond_set_dev_addr(bond->dev, new_active->dev); |
7d26bb10 WP |
725 | write_lock_bh(&bond->curr_slave_lock); |
726 | } | |
3915c1e8 JV |
727 | break; |
728 | case BOND_FOM_FOLLOW: | |
729 | /* | |
730 | * if new_active && old_active, swap them | |
731 | * if just old_active, do nothing (going to no active slave) | |
732 | * if just new_active, set new_active to bond's MAC | |
733 | */ | |
734 | if (!new_active) | |
735 | return; | |
736 | ||
737 | write_unlock_bh(&bond->curr_slave_lock); | |
3915c1e8 JV |
738 | |
739 | if (old_active) { | |
740 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
741 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
742 | ETH_ALEN); | |
743 | saddr.sa_family = new_active->dev->type; | |
744 | } else { | |
745 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
746 | saddr.sa_family = bond->dev->type; | |
747 | } | |
748 | ||
749 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
750 | if (rv) { | |
a4aee5c8 | 751 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
752 | bond->dev->name, -rv, new_active->dev->name); |
753 | goto out; | |
754 | } | |
755 | ||
756 | if (!old_active) | |
757 | goto out; | |
758 | ||
759 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
760 | saddr.sa_family = old_active->dev->type; | |
761 | ||
762 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
763 | if (rv) | |
a4aee5c8 | 764 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
765 | bond->dev->name, -rv, new_active->dev->name); |
766 | out: | |
3915c1e8 JV |
767 | write_lock_bh(&bond->curr_slave_lock); |
768 | break; | |
769 | default: | |
a4aee5c8 | 770 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
771 | bond->dev->name, bond->params.fail_over_mac); |
772 | break; | |
773 | } | |
774 | ||
775 | } | |
776 | ||
a549952a JP |
777 | static bool bond_should_change_active(struct bonding *bond) |
778 | { | |
779 | struct slave *prim = bond->primary_slave; | |
780 | struct slave *curr = bond->curr_active_slave; | |
781 | ||
782 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
783 | return true; | |
784 | if (bond->force_primary) { | |
785 | bond->force_primary = false; | |
786 | return true; | |
787 | } | |
788 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
789 | (prim->speed < curr->speed || | |
790 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
791 | return false; | |
792 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
793 | return false; | |
794 | return true; | |
795 | } | |
3915c1e8 | 796 | |
1da177e4 LT |
797 | /** |
798 | * find_best_interface - select the best available slave to be the active one | |
799 | * @bond: our bonding struct | |
1da177e4 LT |
800 | */ |
801 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
802 | { | |
77140d29 VF |
803 | struct slave *slave, *bestslave = NULL; |
804 | struct list_head *iter; | |
1da177e4 | 805 | int mintime = bond->params.updelay; |
1da177e4 | 806 | |
77140d29 VF |
807 | if (bond->primary_slave && bond->primary_slave->link == BOND_LINK_UP && |
808 | bond_should_change_active(bond)) | |
809 | return bond->primary_slave; | |
1da177e4 | 810 | |
77140d29 VF |
811 | bond_for_each_slave(bond, slave, iter) { |
812 | if (slave->link == BOND_LINK_UP) | |
813 | return slave; | |
814 | if (slave->link == BOND_LINK_BACK && IS_UP(slave->dev) && | |
815 | slave->delay < mintime) { | |
816 | mintime = slave->delay; | |
817 | bestslave = slave; | |
1da177e4 LT |
818 | } |
819 | } | |
820 | ||
821 | return bestslave; | |
822 | } | |
823 | ||
ad246c99 BH |
824 | static bool bond_should_notify_peers(struct bonding *bond) |
825 | { | |
4cb4f97b | 826 | struct slave *slave; |
827 | ||
828 | rcu_read_lock(); | |
829 | slave = rcu_dereference(bond->curr_active_slave); | |
830 | rcu_read_unlock(); | |
ad246c99 BH |
831 | |
832 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
833 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
834 | ||
835 | if (!slave || !bond->send_peer_notif || | |
836 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
837 | return false; | |
838 | ||
ad246c99 BH |
839 | return true; |
840 | } | |
841 | ||
1da177e4 LT |
842 | /** |
843 | * change_active_interface - change the active slave into the specified one | |
844 | * @bond: our bonding struct | |
845 | * @new: the new slave to make the active one | |
846 | * | |
847 | * Set the new slave to the bond's settings and unset them on the old | |
848 | * curr_active_slave. | |
849 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
850 | * | |
851 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
852 | * because it is apparently the best available slave we have, even though its | |
853 | * updelay hasn't timed out yet. | |
854 | * | |
b2e7aceb | 855 | * If new_active is not NULL, caller must hold curr_slave_lock for write_bh. |
1da177e4 | 856 | */ |
a77b5325 | 857 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
858 | { |
859 | struct slave *old_active = bond->curr_active_slave; | |
860 | ||
3d632c3f | 861 | if (old_active == new_active) |
1da177e4 | 862 | return; |
1da177e4 LT |
863 | |
864 | if (new_active) { | |
b2220cad JV |
865 | new_active->jiffies = jiffies; |
866 | ||
1da177e4 LT |
867 | if (new_active->link == BOND_LINK_BACK) { |
868 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
869 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
870 | bond->dev->name, new_active->dev->name, | |
871 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
872 | } |
873 | ||
874 | new_active->delay = 0; | |
875 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 876 | |
3d632c3f | 877 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 878 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 879 | |
58402054 | 880 | if (bond_is_lb(bond)) |
1da177e4 | 881 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
882 | } else { |
883 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
884 | pr_info("%s: making interface %s the new active one.\n", |
885 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
886 | } |
887 | } | |
888 | } | |
889 | ||
3d632c3f | 890 | if (USES_PRIMARY(bond->params.mode)) |
303d1cbf | 891 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 892 | |
58402054 | 893 | if (bond_is_lb(bond)) { |
1da177e4 | 894 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
895 | if (old_active) |
896 | bond_set_slave_inactive_flags(old_active); | |
897 | if (new_active) | |
898 | bond_set_slave_active_flags(new_active); | |
1da177e4 | 899 | } else { |
278b2083 | 900 | rcu_assign_pointer(bond->curr_active_slave, new_active); |
1da177e4 | 901 | } |
c3ade5ca JV |
902 | |
903 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 904 | if (old_active) |
c3ade5ca | 905 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
906 | |
907 | if (new_active) { | |
ad246c99 BH |
908 | bool should_notify_peers = false; |
909 | ||
c3ade5ca | 910 | bond_set_slave_active_flags(new_active); |
2ab82852 | 911 | |
709f8a45 OG |
912 | if (bond->params.fail_over_mac) |
913 | bond_do_fail_over_mac(bond, new_active, | |
914 | old_active); | |
3915c1e8 | 915 | |
ad246c99 BH |
916 | if (netif_running(bond->dev)) { |
917 | bond->send_peer_notif = | |
918 | bond->params.num_peer_notif; | |
919 | should_notify_peers = | |
920 | bond_should_notify_peers(bond); | |
921 | } | |
922 | ||
01f3109d | 923 | write_unlock_bh(&bond->curr_slave_lock); |
01f3109d | 924 | |
b7bc2a5b | 925 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 926 | if (should_notify_peers) |
b7bc2a5b AW |
927 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
928 | bond->dev); | |
01f3109d | 929 | |
01f3109d | 930 | write_lock_bh(&bond->curr_slave_lock); |
7893b249 | 931 | } |
c3ade5ca | 932 | } |
a2fd940f | 933 | |
5a37e8ca | 934 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
935 | * all were sent on curr_active_slave. |
936 | * resend only if bond is brought up with the affected | |
937 | * bonding modes and the retransmission is enabled */ | |
938 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
939 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
940 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 941 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 942 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 943 | } |
1da177e4 LT |
944 | } |
945 | ||
946 | /** | |
947 | * bond_select_active_slave - select a new active slave, if needed | |
948 | * @bond: our bonding struct | |
949 | * | |
3d632c3f | 950 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
951 | * - The old curr_active_slave has been released or lost its link. |
952 | * - The primary_slave has got its link back. | |
953 | * - A slave has got its link back and there's no old curr_active_slave. | |
954 | * | |
b2e7aceb | 955 | * Caller must hold curr_slave_lock for write_bh. |
1da177e4 | 956 | */ |
a77b5325 | 957 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
958 | { |
959 | struct slave *best_slave; | |
ff59c456 | 960 | int rv; |
1da177e4 LT |
961 | |
962 | best_slave = bond_find_best_slave(bond); | |
963 | if (best_slave != bond->curr_active_slave) { | |
964 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
965 | rv = bond_set_carrier(bond); |
966 | if (!rv) | |
967 | return; | |
968 | ||
969 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
970 | pr_info("%s: first active interface up!\n", |
971 | bond->dev->name); | |
ff59c456 | 972 | } else { |
a4aee5c8 JP |
973 | pr_info("%s: now running without any active interface !\n", |
974 | bond->dev->name); | |
ff59c456 | 975 | } |
1da177e4 LT |
976 | } |
977 | } | |
978 | ||
f6dc31a8 | 979 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 980 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 981 | { |
8a8efa22 AW |
982 | struct netpoll *np; |
983 | int err = 0; | |
f6dc31a8 | 984 | |
47be03a2 | 985 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
986 | err = -ENOMEM; |
987 | if (!np) | |
988 | goto out; | |
989 | ||
47be03a2 | 990 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
991 | if (err) { |
992 | kfree(np); | |
993 | goto out; | |
f6dc31a8 | 994 | } |
8a8efa22 AW |
995 | slave->np = np; |
996 | out: | |
997 | return err; | |
998 | } | |
999 | static inline void slave_disable_netpoll(struct slave *slave) | |
1000 | { | |
1001 | struct netpoll *np = slave->np; | |
1002 | ||
1003 | if (!np) | |
1004 | return; | |
1005 | ||
1006 | slave->np = NULL; | |
2cde6acd | 1007 | __netpoll_free_async(np); |
8a8efa22 AW |
1008 | } |
1009 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1010 | { | |
1011 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1012 | return false; | |
1013 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1014 | return false; | |
1015 | return true; | |
f6dc31a8 WC |
1016 | } |
1017 | ||
1018 | static void bond_poll_controller(struct net_device *bond_dev) | |
1019 | { | |
8a8efa22 AW |
1020 | } |
1021 | ||
c4cdef9b | 1022 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1023 | { |
c4cdef9b | 1024 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1025 | struct list_head *iter; |
c2355e1a | 1026 | struct slave *slave; |
c2355e1a | 1027 | |
9caff1e7 | 1028 | bond_for_each_slave(bond, slave, iter) |
8a8efa22 AW |
1029 | if (IS_UP(slave->dev)) |
1030 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1031 | } |
8a8efa22 | 1032 | |
47be03a2 | 1033 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1034 | { |
1035 | struct bonding *bond = netdev_priv(dev); | |
9caff1e7 | 1036 | struct list_head *iter; |
f6dc31a8 | 1037 | struct slave *slave; |
dec1e90e | 1038 | int err = 0; |
f6dc31a8 | 1039 | |
9caff1e7 | 1040 | bond_for_each_slave(bond, slave, iter) { |
8a8efa22 AW |
1041 | err = slave_enable_netpoll(slave); |
1042 | if (err) { | |
c4cdef9b | 1043 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1044 | break; |
f6dc31a8 WC |
1045 | } |
1046 | } | |
8a8efa22 | 1047 | return err; |
f6dc31a8 | 1048 | } |
8a8efa22 AW |
1049 | #else |
1050 | static inline int slave_enable_netpoll(struct slave *slave) | |
1051 | { | |
1052 | return 0; | |
1053 | } | |
1054 | static inline void slave_disable_netpoll(struct slave *slave) | |
1055 | { | |
1056 | } | |
f6dc31a8 WC |
1057 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1058 | { | |
1059 | } | |
f6dc31a8 WC |
1060 | #endif |
1061 | ||
1da177e4 LT |
1062 | /*---------------------------------- IOCTL ----------------------------------*/ |
1063 | ||
c8f44aff | 1064 | static netdev_features_t bond_fix_features(struct net_device *dev, |
9b7b165a | 1065 | netdev_features_t features) |
8531c5ff | 1066 | { |
b2a103e6 | 1067 | struct bonding *bond = netdev_priv(dev); |
9caff1e7 | 1068 | struct list_head *iter; |
c8f44aff | 1069 | netdev_features_t mask; |
9b7b165a | 1070 | struct slave *slave; |
b63365a2 | 1071 | |
0965a1f3 | 1072 | if (!bond_has_slaves(bond)) { |
b2a103e6 MM |
1073 | /* Disable adding VLANs to empty bond. But why? --mq */ |
1074 | features |= NETIF_F_VLAN_CHALLENGED; | |
9b7b165a | 1075 | return features; |
b2a103e6 | 1076 | } |
b63365a2 | 1077 | |
b2a103e6 | 1078 | mask = features; |
b63365a2 | 1079 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1080 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1081 | |
9caff1e7 | 1082 | bond_for_each_slave(bond, slave, iter) { |
b63365a2 HX |
1083 | features = netdev_increment_features(features, |
1084 | slave->dev->features, | |
b2a103e6 MM |
1085 | mask); |
1086 | } | |
b0ce3508 | 1087 | features = netdev_add_tso_features(features, mask); |
b2a103e6 | 1088 | |
b2a103e6 MM |
1089 | return features; |
1090 | } | |
1091 | ||
62f2a3a4 MM |
1092 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1093 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1094 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1095 | |
1096 | static void bond_compute_features(struct bonding *bond) | |
1097 | { | |
c4826861 | 1098 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
c8f44aff | 1099 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
9caff1e7 VF |
1100 | struct net_device *bond_dev = bond->dev; |
1101 | struct list_head *iter; | |
1102 | struct slave *slave; | |
b2a103e6 | 1103 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1104 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1105 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1106 | |
0965a1f3 | 1107 | if (!bond_has_slaves(bond)) |
b2a103e6 MM |
1108 | goto done; |
1109 | ||
9caff1e7 | 1110 | bond_for_each_slave(bond, slave, iter) { |
278339a4 | 1111 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1112 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1113 | ||
b6fe83e9 | 1114 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1115 | if (slave->dev->hard_header_len > max_hard_header_len) |
1116 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1117 | |
1118 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1119 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1120 | } |
8531c5ff | 1121 | |
b63365a2 | 1122 | done: |
b2a103e6 | 1123 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1124 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1125 | bond_dev->gso_max_segs = gso_max_segs; |
1126 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1127 | |
b6fe83e9 ED |
1128 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1129 | bond_dev->priv_flags = flags | dst_release_flag; | |
1130 | ||
b2a103e6 | 1131 | netdev_change_features(bond_dev); |
8531c5ff AK |
1132 | } |
1133 | ||
872254dd MS |
1134 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1135 | struct net_device *slave_dev) | |
1136 | { | |
00829823 | 1137 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1138 | |
1139 | bond_dev->type = slave_dev->type; | |
1140 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1141 | bond_dev->addr_len = slave_dev->addr_len; | |
1142 | ||
1143 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1144 | slave_dev->addr_len); | |
1145 | } | |
1146 | ||
5b2c4dd2 | 1147 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1148 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1149 | */ |
1150 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1151 | struct slave *slave, |
1152 | struct bonding *bond) | |
5b2c4dd2 | 1153 | { |
2d7011ca | 1154 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1155 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1156 | skb->pkt_type != PACKET_BROADCAST && |
1157 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1158 | return false; |
5b2c4dd2 JP |
1159 | return true; |
1160 | } | |
1161 | return false; | |
1162 | } | |
1163 | ||
8a4eb573 | 1164 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1165 | { |
8a4eb573 | 1166 | struct sk_buff *skb = *pskb; |
f1c1775a | 1167 | struct slave *slave; |
0bd80dad | 1168 | struct bonding *bond; |
de063b70 ED |
1169 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1170 | struct slave *); | |
13a8e0c8 | 1171 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1172 | |
8a4eb573 JP |
1173 | skb = skb_share_check(skb, GFP_ATOMIC); |
1174 | if (unlikely(!skb)) | |
1175 | return RX_HANDLER_CONSUMED; | |
1176 | ||
1177 | *pskb = skb; | |
5b2c4dd2 | 1178 | |
35d48903 JP |
1179 | slave = bond_slave_get_rcu(skb->dev); |
1180 | bond = slave->bond; | |
0bd80dad JP |
1181 | |
1182 | if (bond->params.arp_interval) | |
f1c1775a | 1183 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1184 | |
4d97480b MH |
1185 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1186 | if (recv_probe) { | |
de063b70 ED |
1187 | ret = recv_probe(skb, bond, slave); |
1188 | if (ret == RX_HANDLER_CONSUMED) { | |
1189 | consume_skb(skb); | |
1190 | return ret; | |
3aba891d JP |
1191 | } |
1192 | } | |
1193 | ||
0bd80dad | 1194 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1195 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1196 | } |
1197 | ||
35d48903 | 1198 | skb->dev = bond->dev; |
5b2c4dd2 | 1199 | |
0bd80dad | 1200 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1201 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1202 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1203 | |
541ac7c9 CG |
1204 | if (unlikely(skb_cow_head(skb, |
1205 | skb->data - skb_mac_header(skb)))) { | |
1206 | kfree_skb(skb); | |
8a4eb573 | 1207 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1208 | } |
35d48903 | 1209 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1210 | } |
1211 | ||
13a8e0c8 | 1212 | return ret; |
5b2c4dd2 JP |
1213 | } |
1214 | ||
471cb5a3 | 1215 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1f718f0f VF |
1216 | struct net_device *slave_dev, |
1217 | struct slave *slave) | |
471cb5a3 JP |
1218 | { |
1219 | int err; | |
1220 | ||
1f718f0f | 1221 | err = netdev_master_upper_dev_link_private(slave_dev, bond_dev, slave); |
471cb5a3 JP |
1222 | if (err) |
1223 | return err; | |
1224 | slave_dev->flags |= IFF_SLAVE; | |
7f294054 | 1225 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1226 | return 0; |
1227 | } | |
1228 | ||
1229 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1230 | struct net_device *slave_dev) | |
1231 | { | |
1232 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1233 | slave_dev->flags &= ~IFF_SLAVE; | |
7f294054 | 1234 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE, GFP_KERNEL); |
471cb5a3 JP |
1235 | } |
1236 | ||
1da177e4 | 1237 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1238 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1239 | { |
454d7c9b | 1240 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1241 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
c8c23903 | 1242 | struct slave *new_slave = NULL, *prev_slave; |
1da177e4 LT |
1243 | struct sockaddr addr; |
1244 | int link_reporting; | |
8599b52e | 1245 | int res = 0, i; |
1da177e4 | 1246 | |
c772dde3 BH |
1247 | if (!bond->params.use_carrier && |
1248 | slave_dev->ethtool_ops->get_link == NULL && | |
1249 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1250 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1251 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1252 | } |
1253 | ||
1da177e4 LT |
1254 | /* already enslaved */ |
1255 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1256 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1257 | return -EBUSY; |
1258 | } | |
1259 | ||
1260 | /* vlan challenged mutual exclusion */ | |
1261 | /* no need to lock since we're protected by rtnl_lock */ | |
1262 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1263 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1264 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1265 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1266 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1267 | return -EPERM; |
1268 | } else { | |
a4aee5c8 JP |
1269 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1270 | bond_dev->name, slave_dev->name, | |
1271 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1272 | } |
1273 | } else { | |
5a03cdb7 | 1274 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1275 | } |
1276 | ||
217df670 JV |
1277 | /* |
1278 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1279 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1280 | * the current ifenslave will set the interface down prior to |
1281 | * enslaving it; the old ifenslave will not. | |
1282 | */ | |
1283 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1284 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1285 | slave_dev->name); |
1286 | res = -EPERM; | |
1287 | goto err_undo_flags; | |
1288 | } | |
1da177e4 | 1289 | |
872254dd MS |
1290 | /* set bonding device ether type by slave - bonding netdevices are |
1291 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1292 | * there is a need to override some of the type dependent attribs/funcs. | |
1293 | * | |
1294 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1295 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1296 | */ | |
0965a1f3 | 1297 | if (!bond_has_slaves(bond)) { |
e36b9d16 | 1298 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1299 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1300 | bond_dev->name, |
1301 | bond_dev->type, slave_dev->type); | |
75c78500 | 1302 | |
b7bc2a5b AW |
1303 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1304 | bond_dev); | |
3ca5b404 JP |
1305 | res = notifier_to_errno(res); |
1306 | if (res) { | |
1307 | pr_err("%s: refused to change device type\n", | |
1308 | bond_dev->name); | |
1309 | res = -EBUSY; | |
1310 | goto err_undo_flags; | |
1311 | } | |
75c78500 | 1312 | |
32a806c1 | 1313 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1314 | dev_uc_flush(bond_dev); |
22bedad3 | 1315 | dev_mc_flush(bond_dev); |
32a806c1 | 1316 | |
e36b9d16 MS |
1317 | if (slave_dev->type != ARPHRD_ETHER) |
1318 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1319 | else { |
e36b9d16 | 1320 | ether_setup(bond_dev); |
550fd08c NH |
1321 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1322 | } | |
75c78500 | 1323 | |
b7bc2a5b AW |
1324 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1325 | bond_dev); | |
e36b9d16 | 1326 | } |
872254dd | 1327 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1328 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1329 | slave_dev->name, | |
1330 | slave_dev->type, bond_dev->type); | |
1331 | res = -EINVAL; | |
1332 | goto err_undo_flags; | |
872254dd MS |
1333 | } |
1334 | ||
eb7cc59a | 1335 | if (slave_ops->ndo_set_mac_address == NULL) { |
0965a1f3 | 1336 | if (!bond_has_slaves(bond)) { |
a4aee5c8 JP |
1337 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1338 | bond_dev->name); | |
3915c1e8 JV |
1339 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1340 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1341 | 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", |
1342 | bond_dev->name); | |
2ab82852 MS |
1343 | res = -EOPNOTSUPP; |
1344 | goto err_undo_flags; | |
1345 | } | |
1da177e4 LT |
1346 | } |
1347 | ||
8d8fc29d AW |
1348 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1349 | ||
c20811a7 JP |
1350 | /* If this is the first slave, then we need to set the master's hardware |
1351 | * address to be the same as the slave's. */ | |
0965a1f3 | 1352 | if (!bond_has_slaves(bond) && |
dec1e90e | 1353 | bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1354 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1355 | |
243cb4e5 | 1356 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1357 | if (!new_slave) { |
1358 | res = -ENOMEM; | |
1359 | goto err_undo_flags; | |
1360 | } | |
bb1d9123 AG |
1361 | /* |
1362 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1363 | * is set via sysfs or module option if desired. | |
1364 | */ | |
1365 | new_slave->queue_id = 0; | |
1366 | ||
b15ba0fb JP |
1367 | /* Save slave's original mtu and then set it to match the bond */ |
1368 | new_slave->original_mtu = slave_dev->mtu; | |
1369 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1370 | if (res) { | |
1371 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1372 | goto err_free; | |
1373 | } | |
1374 | ||
217df670 JV |
1375 | /* |
1376 | * Save slave's original ("permanent") mac address for modes | |
1377 | * that need it, and for restoring it upon release, and then | |
1378 | * set it to the master's address | |
1379 | */ | |
1380 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1381 | |
dd957c57 | 1382 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1383 | /* |
1384 | * Set slave to master's mac address. The application already | |
1385 | * set the master's mac address to that of the first slave | |
1386 | */ | |
1387 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1388 | addr.sa_family = slave_dev->type; | |
1389 | res = dev_set_mac_address(slave_dev, &addr); | |
1390 | if (res) { | |
5a03cdb7 | 1391 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1392 | goto err_restore_mtu; |
2ab82852 | 1393 | } |
217df670 | 1394 | } |
1da177e4 | 1395 | |
217df670 JV |
1396 | /* open the slave since the application closed it */ |
1397 | res = dev_open(slave_dev); | |
1398 | if (res) { | |
3d632c3f | 1399 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
1f718f0f | 1400 | goto err_restore_mac; |
1da177e4 LT |
1401 | } |
1402 | ||
35d48903 | 1403 | new_slave->bond = bond; |
1da177e4 | 1404 | new_slave->dev = slave_dev; |
0b680e75 | 1405 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1406 | |
58402054 | 1407 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1408 | /* bond_alb_init_slave() must be called before all other stages since |
1409 | * it might fail and we do not want to have to undo everything | |
1410 | */ | |
1411 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1412 | if (res) |
569f0c4d | 1413 | goto err_close; |
1da177e4 LT |
1414 | } |
1415 | ||
303d1cbf JV |
1416 | /* If the mode USES_PRIMARY, then the following is handled by |
1417 | * bond_change_active_slave(). | |
1da177e4 LT |
1418 | */ |
1419 | if (!USES_PRIMARY(bond->params.mode)) { | |
1420 | /* set promiscuity level to new slave */ | |
1421 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1422 | res = dev_set_promiscuity(slave_dev, 1); |
1423 | if (res) | |
1424 | goto err_close; | |
1da177e4 LT |
1425 | } |
1426 | ||
1427 | /* set allmulti level to new slave */ | |
1428 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1429 | res = dev_set_allmulti(slave_dev, 1); |
1430 | if (res) | |
1431 | goto err_close; | |
1da177e4 LT |
1432 | } |
1433 | ||
b9e40857 | 1434 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1435 | |
1436 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1437 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1438 | ||
b9e40857 | 1439 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1440 | } |
1441 | ||
1442 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1443 | /* add lacpdu mc addr to mc list */ | |
1444 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1445 | ||
22bedad3 | 1446 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1447 | } |
1448 | ||
b8e2fde4 WY |
1449 | res = vlan_vids_add_by_dev(slave_dev, bond_dev); |
1450 | if (res) { | |
1ff412ad | 1451 | pr_err("%s: Error: Couldn't add bond vlan ids to %s\n", |
1452 | bond_dev->name, slave_dev->name); | |
1453 | goto err_close; | |
1454 | } | |
1da177e4 | 1455 | |
c8c23903 | 1456 | prev_slave = bond_last_slave(bond); |
1da177e4 LT |
1457 | |
1458 | new_slave->delay = 0; | |
1459 | new_slave->link_failure_count = 0; | |
1460 | ||
876254ae VF |
1461 | bond_update_speed_duplex(new_slave); |
1462 | ||
f31c7937 MK |
1463 | new_slave->last_arp_rx = jiffies - |
1464 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
8599b52e VF |
1465 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
1466 | new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx; | |
f5b2b966 | 1467 | |
1da177e4 LT |
1468 | if (bond->params.miimon && !bond->params.use_carrier) { |
1469 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1470 | ||
1471 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1472 | /* | |
1473 | * miimon is set but a bonded network driver | |
1474 | * does not support ETHTOOL/MII and | |
1475 | * arp_interval is not set. Note: if | |
1476 | * use_carrier is enabled, we will never go | |
1477 | * here (because netif_carrier is always | |
1478 | * supported); thus, we don't need to change | |
1479 | * the messages for netif_carrier. | |
1480 | */ | |
a4aee5c8 | 1481 | pr_warning("%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", |
4e0952c7 | 1482 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1483 | } else if (link_reporting == -1) { |
1484 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1485 | pr_warning("%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", |
1486 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1487 | } |
1488 | } | |
1489 | ||
1490 | /* check for initial state */ | |
f31c7937 MK |
1491 | if (bond->params.miimon) { |
1492 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1493 | if (bond->params.updelay) { | |
1494 | new_slave->link = BOND_LINK_BACK; | |
1495 | new_slave->delay = bond->params.updelay; | |
1496 | } else { | |
1497 | new_slave->link = BOND_LINK_UP; | |
1498 | } | |
1da177e4 | 1499 | } else { |
f31c7937 | 1500 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1501 | } |
f31c7937 MK |
1502 | } else if (bond->params.arp_interval) { |
1503 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1504 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1505 | } else { |
f31c7937 | 1506 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1507 | } |
1508 | ||
f31c7937 MK |
1509 | if (new_slave->link != BOND_LINK_DOWN) |
1510 | new_slave->jiffies = jiffies; | |
1511 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1512 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1513 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1514 | ||
1da177e4 LT |
1515 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1516 | /* if there is a primary slave, remember it */ | |
a549952a | 1517 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1518 | bond->primary_slave = new_slave; |
a549952a JP |
1519 | bond->force_primary = true; |
1520 | } | |
1da177e4 LT |
1521 | } |
1522 | ||
1523 | switch (bond->params.mode) { | |
1524 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b | 1525 | bond_set_slave_inactive_flags(new_slave); |
1da177e4 LT |
1526 | break; |
1527 | case BOND_MODE_8023AD: | |
1528 | /* in 802.3ad mode, the internal mechanism | |
1529 | * will activate the slaves in the selected | |
1530 | * aggregator | |
1531 | */ | |
1532 | bond_set_slave_inactive_flags(new_slave); | |
1533 | /* if this is the first slave */ | |
23c147e0 | 1534 | if (!prev_slave) { |
1da177e4 LT |
1535 | SLAVE_AD_INFO(new_slave).id = 1; |
1536 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1537 | * can be called only after the mac address of the bond is set | |
1538 | */ | |
56d00c67 | 1539 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 LT |
1540 | } else { |
1541 | SLAVE_AD_INFO(new_slave).id = | |
dec1e90e | 1542 | SLAVE_AD_INFO(prev_slave).id + 1; |
1da177e4 LT |
1543 | } |
1544 | ||
1545 | bond_3ad_bind_slave(new_slave); | |
1546 | break; | |
1547 | case BOND_MODE_TLB: | |
1548 | case BOND_MODE_ALB: | |
e30bc066 | 1549 | bond_set_active_slave(new_slave); |
059fe7a5 | 1550 | bond_set_slave_inactive_flags(new_slave); |
1da177e4 LT |
1551 | break; |
1552 | default: | |
5a03cdb7 | 1553 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1554 | |
1555 | /* always active in trunk mode */ | |
e30bc066 | 1556 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1557 | |
1558 | /* In trunking mode there is little meaning to curr_active_slave | |
1559 | * anyway (it holds no special properties of the bond device), | |
1560 | * so we can change it without calling change_active_interface() | |
1561 | */ | |
4e591b93 | 1562 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
278b2083 | 1563 | rcu_assign_pointer(bond->curr_active_slave, new_slave); |
3d632c3f | 1564 | |
1da177e4 LT |
1565 | break; |
1566 | } /* switch(bond_mode) */ | |
1567 | ||
f6dc31a8 | 1568 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1569 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1570 | if (slave_dev->npinfo) { |
1571 | if (slave_enable_netpoll(new_slave)) { | |
1572 | read_unlock(&bond->lock); | |
1573 | pr_info("Error, %s: master_dev is using netpoll, " | |
1574 | "but new slave device does not support netpoll.\n", | |
1575 | bond_dev->name); | |
1576 | res = -EBUSY; | |
f7d9821a | 1577 | goto err_detach; |
8a8efa22 | 1578 | } |
f6dc31a8 WC |
1579 | } |
1580 | #endif | |
8a8efa22 | 1581 | |
35d48903 JP |
1582 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1583 | new_slave); | |
1584 | if (res) { | |
1585 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
5831d66e | 1586 | goto err_detach; |
35d48903 JP |
1587 | } |
1588 | ||
1f718f0f VF |
1589 | res = bond_master_upper_dev_link(bond_dev, slave_dev, new_slave); |
1590 | if (res) { | |
1591 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); | |
1592 | goto err_unregister; | |
1593 | } | |
1594 | ||
07699f9a | 1595 | res = bond_sysfs_slave_add(new_slave); |
1596 | if (res) { | |
1597 | pr_debug("Error %d calling bond_sysfs_slave_add\n", res); | |
1598 | goto err_upper_unlink; | |
1599 | } | |
1600 | ||
5378c2e6 VF |
1601 | bond->slave_cnt++; |
1602 | bond_compute_features(bond); | |
1603 | bond_set_carrier(bond); | |
1604 | ||
1605 | if (USES_PRIMARY(bond->params.mode)) { | |
5378c2e6 VF |
1606 | write_lock_bh(&bond->curr_slave_lock); |
1607 | bond_select_active_slave(bond); | |
1608 | write_unlock_bh(&bond->curr_slave_lock); | |
5378c2e6 | 1609 | } |
1f718f0f | 1610 | |
a4aee5c8 JP |
1611 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1612 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1613 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1614 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1615 | |
1616 | /* enslave is successful */ | |
1617 | return 0; | |
1618 | ||
1619 | /* Undo stages on error */ | |
07699f9a | 1620 | err_upper_unlink: |
1621 | bond_upper_dev_unlink(bond_dev, slave_dev); | |
1622 | ||
1f718f0f VF |
1623 | err_unregister: |
1624 | netdev_rx_handler_unregister(slave_dev); | |
1625 | ||
f7d9821a | 1626 | err_detach: |
303d1cbf JV |
1627 | if (!USES_PRIMARY(bond->params.mode)) |
1628 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1629 | ||
1ff412ad | 1630 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
3c5913b5 | 1631 | if (bond->primary_slave == new_slave) |
1632 | bond->primary_slave = NULL; | |
3c5913b5 | 1633 | if (bond->curr_active_slave == new_slave) { |
3c5913b5 | 1634 | write_lock_bh(&bond->curr_slave_lock); |
c8517035 | 1635 | bond_change_active_slave(bond, NULL); |
3c5913b5 | 1636 | bond_select_active_slave(bond); |
1637 | write_unlock_bh(&bond->curr_slave_lock); | |
3c5913b5 | 1638 | } |
fc7a72ac | 1639 | slave_disable_netpoll(new_slave); |
f7d9821a | 1640 | |
1da177e4 | 1641 | err_close: |
b6a5a7b9 | 1642 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1643 | dev_close(slave_dev); |
1644 | ||
1645 | err_restore_mac: | |
dd957c57 | 1646 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1647 | /* XXX TODO - fom follow mode needs to change master's |
1648 | * MAC if this slave's MAC is in use by the bond, or at | |
1649 | * least print a warning. | |
1650 | */ | |
2ab82852 MS |
1651 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1652 | addr.sa_family = slave_dev->type; | |
1653 | dev_set_mac_address(slave_dev, &addr); | |
1654 | } | |
1da177e4 | 1655 | |
b15ba0fb JP |
1656 | err_restore_mtu: |
1657 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1658 | ||
1da177e4 LT |
1659 | err_free: |
1660 | kfree(new_slave); | |
1661 | ||
1662 | err_undo_flags: | |
b8fad459 | 1663 | /* Enslave of first slave has failed and we need to fix master's mac */ |
0965a1f3 | 1664 | if (!bond_has_slaves(bond) && |
844223ab | 1665 | ether_addr_equal_64bits(bond_dev->dev_addr, slave_dev->dev_addr)) |
b8fad459 | 1666 | eth_hw_addr_random(bond_dev); |
3d632c3f | 1667 | |
1da177e4 LT |
1668 | return res; |
1669 | } | |
1670 | ||
1671 | /* | |
1672 | * Try to release the slave device <slave> from the bond device <master> | |
1673 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1674 | * is write-locked. If "all" is true it means that the function is being called |
1675 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1676 | * |
1677 | * The rules for slave state should be: | |
1678 | * for Active/Backup: | |
1679 | * Active stays on all backups go down | |
1680 | * for Bonded connections: | |
1681 | * The first up interface should be left on and all others downed. | |
1682 | */ | |
0896341a | 1683 | static int __bond_release_one(struct net_device *bond_dev, |
1684 | struct net_device *slave_dev, | |
1685 | bool all) | |
1da177e4 | 1686 | { |
454d7c9b | 1687 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1688 | struct slave *slave, *oldcurrent; |
1689 | struct sockaddr addr; | |
5a0068de | 1690 | int old_flags = bond_dev->flags; |
c8f44aff | 1691 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1692 | |
1693 | /* slave is not a slave or master is not master of this slave */ | |
1694 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1695 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1696 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1697 | bond_dev->name, slave_dev->name); |
1698 | return -EINVAL; | |
1699 | } | |
1700 | ||
e843fa50 | 1701 | block_netpoll_tx(); |
1da177e4 LT |
1702 | |
1703 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1704 | if (!slave) { | |
1705 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1706 | pr_info("%s: %s not enslaved\n", |
1707 | bond_dev->name, slave_dev->name); | |
e843fa50 | 1708 | unblock_netpoll_tx(); |
1da177e4 LT |
1709 | return -EINVAL; |
1710 | } | |
1711 | ||
5378c2e6 VF |
1712 | /* release the slave from its bond */ |
1713 | bond->slave_cnt--; | |
1714 | ||
07699f9a | 1715 | bond_sysfs_slave_del(slave); |
1716 | ||
1f718f0f | 1717 | bond_upper_dev_unlink(bond_dev, slave_dev); |
35d48903 JP |
1718 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1719 | * for this slave anymore. | |
1720 | */ | |
1721 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1722 | write_lock_bh(&bond->lock); |
1723 | ||
1da177e4 | 1724 | /* Inform AD package of unbinding of slave. */ |
be79bd04 | 1725 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1726 | bond_3ad_unbind_slave(slave); |
be79bd04 | 1727 | |
c8517035 | 1728 | write_unlock_bh(&bond->lock); |
1da177e4 | 1729 | |
a4aee5c8 JP |
1730 | pr_info("%s: releasing %s interface %s\n", |
1731 | bond_dev->name, | |
e30bc066 | 1732 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1733 | slave_dev->name); |
1da177e4 LT |
1734 | |
1735 | oldcurrent = bond->curr_active_slave; | |
1736 | ||
1737 | bond->current_arp_slave = NULL; | |
1738 | ||
dec1e90e | 1739 | if (!all && !bond->params.fail_over_mac) { |
844223ab | 1740 | if (ether_addr_equal_64bits(bond_dev->dev_addr, slave->perm_hwaddr) && |
0965a1f3 | 1741 | bond_has_slaves(bond)) |
dec1e90e | 1742 | 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", |
1743 | bond_dev->name, slave_dev->name, | |
1744 | slave->perm_hwaddr, | |
1745 | bond_dev->name, slave_dev->name); | |
1746 | } | |
1747 | ||
3d632c3f | 1748 | if (bond->primary_slave == slave) |
1da177e4 | 1749 | bond->primary_slave = NULL; |
1da177e4 | 1750 | |
c8517035 | 1751 | if (oldcurrent == slave) { |
1752 | write_lock_bh(&bond->curr_slave_lock); | |
1da177e4 | 1753 | bond_change_active_slave(bond, NULL); |
c8517035 | 1754 | write_unlock_bh(&bond->curr_slave_lock); |
1755 | } | |
1da177e4 | 1756 | |
58402054 | 1757 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1758 | /* Must be called only after the slave has been |
1759 | * detached from the list and the curr_active_slave | |
1760 | * has been cleared (if our_slave == old_current), | |
1761 | * but before a new active slave is selected. | |
1762 | */ | |
1763 | bond_alb_deinit_slave(bond, slave); | |
1764 | } | |
1765 | ||
0896341a | 1766 | if (all) { |
278b2083 | 1767 | rcu_assign_pointer(bond->curr_active_slave, NULL); |
0896341a | 1768 | } else if (oldcurrent == slave) { |
059fe7a5 JV |
1769 | /* |
1770 | * Note that we hold RTNL over this sequence, so there | |
1771 | * is no concern that another slave add/remove event | |
1772 | * will interfere. | |
1773 | */ | |
059fe7a5 JV |
1774 | write_lock_bh(&bond->curr_slave_lock); |
1775 | ||
1da177e4 LT |
1776 | bond_select_active_slave(bond); |
1777 | ||
059fe7a5 | 1778 | write_unlock_bh(&bond->curr_slave_lock); |
059fe7a5 JV |
1779 | } |
1780 | ||
0965a1f3 | 1781 | if (!bond_has_slaves(bond)) { |
ff59c456 | 1782 | bond_set_carrier(bond); |
409cc1f8 | 1783 | eth_hw_addr_random(bond_dev); |
1da177e4 | 1784 | |
a59d3d21 | 1785 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1786 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
1787 | bond_dev->name, bond_dev->name); | |
1788 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
1789 | bond_dev->name); | |
1da177e4 | 1790 | } |
1da177e4 LT |
1791 | } |
1792 | ||
e843fa50 | 1793 | unblock_netpoll_tx(); |
278b2083 | 1794 | synchronize_rcu(); |
1da177e4 | 1795 | |
0965a1f3 | 1796 | if (!bond_has_slaves(bond)) { |
2af73d4b | 1797 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
1798 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
1799 | } | |
2af73d4b | 1800 | |
b2a103e6 MM |
1801 | bond_compute_features(bond); |
1802 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1803 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
1804 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
1805 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1806 | ||
b76cdba9 | 1807 | /* must do this from outside any spinlocks */ |
1ff412ad | 1808 | vlan_vids_del_by_dev(slave_dev, bond_dev); |
1da177e4 | 1809 | |
303d1cbf JV |
1810 | /* If the mode USES_PRIMARY, then this cases was handled above by |
1811 | * bond_change_active_slave(..., NULL) | |
1da177e4 LT |
1812 | */ |
1813 | if (!USES_PRIMARY(bond->params.mode)) { | |
5a0068de NH |
1814 | /* unset promiscuity level from slave |
1815 | * NOTE: The NETDEV_CHANGEADDR call above may change the value | |
1816 | * of the IFF_PROMISC flag in the bond_dev, but we need the | |
1817 | * value of that flag before that change, as that was the value | |
1818 | * when this slave was attached, so we cache at the start of the | |
1819 | * function and use it here. Same goes for ALLMULTI below | |
1820 | */ | |
1821 | if (old_flags & IFF_PROMISC) | |
1da177e4 | 1822 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1823 | |
1824 | /* unset allmulti level from slave */ | |
5a0068de | 1825 | if (old_flags & IFF_ALLMULTI) |
1da177e4 | 1826 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 1827 | |
303d1cbf | 1828 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
1829 | } |
1830 | ||
8a8efa22 | 1831 | slave_disable_netpoll(slave); |
f6dc31a8 | 1832 | |
1da177e4 LT |
1833 | /* close slave before restoring its mac address */ |
1834 | dev_close(slave_dev); | |
1835 | ||
3915c1e8 | 1836 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1837 | /* restore original ("permanent") mac address */ |
1838 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1839 | addr.sa_family = slave_dev->type; | |
1840 | dev_set_mac_address(slave_dev, &addr); | |
1841 | } | |
1da177e4 | 1842 | |
b15ba0fb JP |
1843 | dev_set_mtu(slave_dev, slave->original_mtu); |
1844 | ||
2d7011ca | 1845 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1846 | |
1847 | kfree(slave); | |
1848 | ||
1849 | return 0; /* deletion OK */ | |
1850 | } | |
1851 | ||
0896341a | 1852 | /* A wrapper used because of ndo_del_link */ |
1853 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
1854 | { | |
1855 | return __bond_release_one(bond_dev, slave_dev, false); | |
1856 | } | |
1857 | ||
d90a162a | 1858 | /* |
dadaa10b | 1859 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
1860 | * Must be under rtnl_lock when this function is called. |
1861 | */ | |
26d8ee75 | 1862 | static int bond_release_and_destroy(struct net_device *bond_dev, |
1863 | struct net_device *slave_dev) | |
d90a162a | 1864 | { |
454d7c9b | 1865 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
1866 | int ret; |
1867 | ||
1868 | ret = bond_release(bond_dev, slave_dev); | |
0965a1f3 | 1869 | if (ret == 0 && !bond_has_slaves(bond)) { |
8a8efa22 | 1870 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
1871 | pr_info("%s: destroying bond %s.\n", |
1872 | bond_dev->name, bond_dev->name); | |
9e71626c | 1873 | unregister_netdevice(bond_dev); |
d90a162a MS |
1874 | } |
1875 | return ret; | |
1876 | } | |
1877 | ||
1da177e4 LT |
1878 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
1879 | { | |
454d7c9b | 1880 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1881 | |
1882 | info->bond_mode = bond->params.mode; | |
1883 | info->miimon = bond->params.miimon; | |
1884 | ||
6603a6f2 | 1885 | read_lock(&bond->lock); |
1da177e4 | 1886 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 1887 | read_unlock(&bond->lock); |
1da177e4 LT |
1888 | |
1889 | return 0; | |
1890 | } | |
1891 | ||
1892 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
1893 | { | |
454d7c9b | 1894 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 1895 | struct list_head *iter; |
dec1e90e | 1896 | int i = 0, res = -ENODEV; |
1da177e4 | 1897 | struct slave *slave; |
1da177e4 | 1898 | |
6603a6f2 | 1899 | read_lock(&bond->lock); |
9caff1e7 | 1900 | bond_for_each_slave(bond, slave, iter) { |
dec1e90e | 1901 | if (i++ == (int)info->slave_id) { |
689c96cc ED |
1902 | res = 0; |
1903 | strcpy(info->slave_name, slave->dev->name); | |
1904 | info->link = slave->link; | |
e30bc066 | 1905 | info->state = bond_slave_state(slave); |
689c96cc | 1906 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
1907 | break; |
1908 | } | |
1909 | } | |
6603a6f2 | 1910 | read_unlock(&bond->lock); |
1da177e4 | 1911 | |
689c96cc | 1912 | return res; |
1da177e4 LT |
1913 | } |
1914 | ||
1915 | /*-------------------------------- Monitoring -------------------------------*/ | |
1916 | ||
1da177e4 | 1917 | |
f0c76d61 JV |
1918 | static int bond_miimon_inspect(struct bonding *bond) |
1919 | { | |
dec1e90e | 1920 | int link_state, commit = 0; |
9caff1e7 | 1921 | struct list_head *iter; |
f0c76d61 | 1922 | struct slave *slave; |
41f89100 JP |
1923 | bool ignore_updelay; |
1924 | ||
1925 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 | 1926 | |
4cb4f97b | 1927 | bond_for_each_slave_rcu(bond, slave, iter) { |
f0c76d61 | 1928 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 1929 | |
f0c76d61 | 1930 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
1931 | |
1932 | switch (slave->link) { | |
f0c76d61 JV |
1933 | case BOND_LINK_UP: |
1934 | if (link_state) | |
1935 | continue; | |
1da177e4 | 1936 | |
f0c76d61 JV |
1937 | slave->link = BOND_LINK_FAIL; |
1938 | slave->delay = bond->params.downdelay; | |
1939 | if (slave->delay) { | |
a4aee5c8 JP |
1940 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
1941 | bond->dev->name, | |
1942 | (bond->params.mode == | |
1943 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 1944 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
1945 | "active " : "backup ") : "", |
1946 | slave->dev->name, | |
1947 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 1948 | } |
f0c76d61 JV |
1949 | /*FALLTHRU*/ |
1950 | case BOND_LINK_FAIL: | |
1951 | if (link_state) { | |
1952 | /* | |
1953 | * recovered before downdelay expired | |
1954 | */ | |
1955 | slave->link = BOND_LINK_UP; | |
1da177e4 | 1956 | slave->jiffies = jiffies; |
a4aee5c8 JP |
1957 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
1958 | bond->dev->name, | |
1959 | (bond->params.downdelay - slave->delay) * | |
1960 | bond->params.miimon, | |
1961 | slave->dev->name); | |
f0c76d61 | 1962 | continue; |
1da177e4 | 1963 | } |
f0c76d61 JV |
1964 | |
1965 | if (slave->delay <= 0) { | |
1966 | slave->new_link = BOND_LINK_DOWN; | |
1967 | commit++; | |
1968 | continue; | |
1da177e4 | 1969 | } |
1da177e4 | 1970 | |
f0c76d61 JV |
1971 | slave->delay--; |
1972 | break; | |
1973 | ||
1974 | case BOND_LINK_DOWN: | |
1975 | if (!link_state) | |
1976 | continue; | |
1977 | ||
1978 | slave->link = BOND_LINK_BACK; | |
1979 | slave->delay = bond->params.updelay; | |
1980 | ||
1981 | if (slave->delay) { | |
a4aee5c8 JP |
1982 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
1983 | bond->dev->name, slave->dev->name, | |
1984 | ignore_updelay ? 0 : | |
1985 | bond->params.updelay * | |
1986 | bond->params.miimon); | |
f0c76d61 JV |
1987 | } |
1988 | /*FALLTHRU*/ | |
1989 | case BOND_LINK_BACK: | |
1990 | if (!link_state) { | |
1991 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
1992 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
1993 | bond->dev->name, | |
1994 | (bond->params.updelay - slave->delay) * | |
1995 | bond->params.miimon, | |
1996 | slave->dev->name); | |
f0c76d61 JV |
1997 | |
1998 | continue; | |
1999 | } | |
2000 | ||
41f89100 JP |
2001 | if (ignore_updelay) |
2002 | slave->delay = 0; | |
2003 | ||
f0c76d61 JV |
2004 | if (slave->delay <= 0) { |
2005 | slave->new_link = BOND_LINK_UP; | |
2006 | commit++; | |
41f89100 | 2007 | ignore_updelay = false; |
f0c76d61 | 2008 | continue; |
1da177e4 | 2009 | } |
f0c76d61 JV |
2010 | |
2011 | slave->delay--; | |
1da177e4 | 2012 | break; |
f0c76d61 JV |
2013 | } |
2014 | } | |
1da177e4 | 2015 | |
f0c76d61 JV |
2016 | return commit; |
2017 | } | |
1da177e4 | 2018 | |
f0c76d61 JV |
2019 | static void bond_miimon_commit(struct bonding *bond) |
2020 | { | |
9caff1e7 | 2021 | struct list_head *iter; |
f0c76d61 | 2022 | struct slave *slave; |
f0c76d61 | 2023 | |
9caff1e7 | 2024 | bond_for_each_slave(bond, slave, iter) { |
f0c76d61 JV |
2025 | switch (slave->new_link) { |
2026 | case BOND_LINK_NOCHANGE: | |
2027 | continue; | |
1da177e4 | 2028 | |
f0c76d61 JV |
2029 | case BOND_LINK_UP: |
2030 | slave->link = BOND_LINK_UP; | |
2031 | slave->jiffies = jiffies; | |
2032 | ||
2033 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2034 | /* prevent it from being the active one */ | |
e30bc066 | 2035 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2036 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2037 | /* make it immediately active */ | |
e30bc066 | 2038 | bond_set_active_slave(slave); |
f0c76d61 JV |
2039 | } else if (slave != bond->primary_slave) { |
2040 | /* prevent it from being the active one */ | |
e30bc066 | 2041 | bond_set_backup_slave(slave); |
1da177e4 | 2042 | } |
1da177e4 | 2043 | |
5d30530e | 2044 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 | 2045 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
2046 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
2047 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2048 | |
f0c76d61 JV |
2049 | /* notify ad that the link status has changed */ |
2050 | if (bond->params.mode == BOND_MODE_8023AD) | |
2051 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2052 | |
58402054 | 2053 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2054 | bond_alb_handle_link_change(bond, slave, |
2055 | BOND_LINK_UP); | |
1da177e4 | 2056 | |
f0c76d61 JV |
2057 | if (!bond->curr_active_slave || |
2058 | (slave == bond->primary_slave)) | |
2059 | goto do_failover; | |
1da177e4 | 2060 | |
f0c76d61 | 2061 | continue; |
059fe7a5 | 2062 | |
f0c76d61 | 2063 | case BOND_LINK_DOWN: |
fba4acda JV |
2064 | if (slave->link_failure_count < UINT_MAX) |
2065 | slave->link_failure_count++; | |
2066 | ||
f0c76d61 | 2067 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2068 | |
f0c76d61 JV |
2069 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2070 | bond->params.mode == BOND_MODE_8023AD) | |
2071 | bond_set_slave_inactive_flags(slave); | |
2072 | ||
a4aee5c8 JP |
2073 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2074 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2075 | |
2076 | if (bond->params.mode == BOND_MODE_8023AD) | |
2077 | bond_3ad_handle_link_change(slave, | |
2078 | BOND_LINK_DOWN); | |
2079 | ||
ae63e808 | 2080 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2081 | bond_alb_handle_link_change(bond, slave, |
2082 | BOND_LINK_DOWN); | |
2083 | ||
2084 | if (slave == bond->curr_active_slave) | |
2085 | goto do_failover; | |
2086 | ||
2087 | continue; | |
2088 | ||
2089 | default: | |
a4aee5c8 | 2090 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2091 | bond->dev->name, slave->new_link, |
2092 | slave->dev->name); | |
2093 | slave->new_link = BOND_LINK_NOCHANGE; | |
2094 | ||
2095 | continue; | |
2096 | } | |
2097 | ||
2098 | do_failover: | |
2099 | ASSERT_RTNL(); | |
e843fa50 | 2100 | block_netpoll_tx(); |
f0c76d61 JV |
2101 | write_lock_bh(&bond->curr_slave_lock); |
2102 | bond_select_active_slave(bond); | |
2103 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2104 | unblock_netpoll_tx(); |
f0c76d61 JV |
2105 | } |
2106 | ||
2107 | bond_set_carrier(bond); | |
1da177e4 LT |
2108 | } |
2109 | ||
0b0eef66 JV |
2110 | /* |
2111 | * bond_mii_monitor | |
2112 | * | |
2113 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2114 | * inspection, then (if inspection indicates something needs to be done) |
2115 | * an acquisition of appropriate locks followed by a commit phase to | |
2116 | * implement whatever link state changes are indicated. | |
0b0eef66 | 2117 | */ |
6da67d26 | 2118 | static void bond_mii_monitor(struct work_struct *work) |
0b0eef66 JV |
2119 | { |
2120 | struct bonding *bond = container_of(work, struct bonding, | |
2121 | mii_work.work); | |
ad246c99 | 2122 | bool should_notify_peers = false; |
1f2cd845 | 2123 | unsigned long delay; |
0b0eef66 | 2124 | |
1f2cd845 DM |
2125 | delay = msecs_to_jiffies(bond->params.miimon); |
2126 | ||
2127 | if (!bond_has_slaves(bond)) | |
f0c76d61 | 2128 | goto re_arm; |
b59f9f74 | 2129 | |
4cb4f97b | 2130 | rcu_read_lock(); |
2131 | ||
ad246c99 BH |
2132 | should_notify_peers = bond_should_notify_peers(bond); |
2133 | ||
1f2cd845 | 2134 | if (bond_miimon_inspect(bond)) { |
4cb4f97b | 2135 | rcu_read_unlock(); |
f0c76d61 | 2136 | |
1f2cd845 DM |
2137 | /* Race avoidance with bond_close cancel of workqueue */ |
2138 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2139 | delay = 1; |
2140 | should_notify_peers = false; | |
2141 | goto re_arm; | |
2142 | } | |
6b6c5261 | 2143 | |
1f2cd845 DM |
2144 | bond_miimon_commit(bond); |
2145 | ||
1f2cd845 | 2146 | rtnl_unlock(); /* might sleep, hold no other locks */ |
4cb4f97b | 2147 | } else |
2148 | rcu_read_unlock(); | |
0b0eef66 | 2149 | |
f0c76d61 | 2150 | re_arm: |
e6d265e8 | 2151 | if (bond->params.miimon) |
1f2cd845 DM |
2152 | queue_delayed_work(bond->wq, &bond->mii_work, delay); |
2153 | ||
1f2cd845 DM |
2154 | if (should_notify_peers) { |
2155 | if (!rtnl_trylock()) | |
2156 | return; | |
2157 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2158 | rtnl_unlock(); | |
2159 | } | |
0b0eef66 | 2160 | } |
c3ade5ca | 2161 | |
50223ce4 | 2162 | static bool bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2163 | { |
50223ce4 VF |
2164 | struct net_device *upper; |
2165 | struct list_head *iter; | |
2166 | bool ret = false; | |
f5b2b966 | 2167 | |
eaddcd76 | 2168 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
50223ce4 | 2169 | return true; |
f5b2b966 | 2170 | |
50223ce4 | 2171 | rcu_read_lock(); |
2f268f12 | 2172 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
50223ce4 VF |
2173 | if (ip == bond_confirm_addr(upper, 0, ip)) { |
2174 | ret = true; | |
2175 | break; | |
2176 | } | |
f5b2b966 | 2177 | } |
50223ce4 | 2178 | rcu_read_unlock(); |
f5b2b966 | 2179 | |
50223ce4 | 2180 | return ret; |
f5b2b966 JV |
2181 | } |
2182 | ||
c3ade5ca JV |
2183 | /* |
2184 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2185 | * switches in VLAN mode (especially if ports are configured as | |
2186 | * "native" to a VLAN) might not pass non-tagged frames. | |
2187 | */ | |
d3bb52b0 | 2188 | static void bond_arp_send(struct net_device *slave_dev, int arp_op, __be32 dest_ip, __be32 src_ip, unsigned short vlan_id) |
c3ade5ca JV |
2189 | { |
2190 | struct sk_buff *skb; | |
2191 | ||
acca2674 | 2192 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op, |
2193 | slave_dev->name, &dest_ip, &src_ip, vlan_id); | |
3d632c3f | 2194 | |
c3ade5ca JV |
2195 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2196 | NULL, slave_dev->dev_addr, NULL); | |
2197 | ||
2198 | if (!skb) { | |
a4aee5c8 | 2199 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2200 | return; |
2201 | } | |
2202 | if (vlan_id) { | |
1fd9b1fc | 2203 | skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id); |
c3ade5ca | 2204 | if (!skb) { |
a4aee5c8 | 2205 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2206 | return; |
2207 | } | |
2208 | } | |
2209 | arp_xmit(skb); | |
2210 | } | |
2211 | ||
2212 | ||
1da177e4 LT |
2213 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2214 | { | |
27bc11e6 VF |
2215 | struct net_device *upper, *vlan_upper; |
2216 | struct list_head *iter, *vlan_iter; | |
c3ade5ca | 2217 | struct rtable *rt; |
27bc11e6 VF |
2218 | __be32 *targets = bond->params.arp_targets, addr; |
2219 | int i, vlan_id; | |
1da177e4 | 2220 | |
27bc11e6 | 2221 | for (i = 0; i < BOND_MAX_ARP_TARGETS && targets[i]; i++) { |
acca2674 | 2222 | pr_debug("basa: target %pI4\n", &targets[i]); |
c3ade5ca | 2223 | |
27bc11e6 | 2224 | /* Find out through which dev should the packet go */ |
78fbfd8a DM |
2225 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2226 | RTO_ONLINK, 0); | |
b23dd4fe | 2227 | if (IS_ERR(rt)) { |
3e32582f VF |
2228 | pr_debug("%s: no route to arp_ip_target %pI4\n", |
2229 | bond->dev->name, &targets[i]); | |
c3ade5ca JV |
2230 | continue; |
2231 | } | |
2232 | ||
c3ade5ca | 2233 | vlan_id = 0; |
27bc11e6 VF |
2234 | |
2235 | /* bond device itself */ | |
2236 | if (rt->dst.dev == bond->dev) | |
2237 | goto found; | |
2238 | ||
2239 | rcu_read_lock(); | |
2240 | /* first we search only for vlan devices. for every vlan | |
2241 | * found we verify its upper dev list, searching for the | |
2242 | * rt->dst.dev. If found we save the tag of the vlan and | |
2243 | * proceed to send the packet. | |
2244 | * | |
2245 | * TODO: QinQ? | |
2246 | */ | |
2f268f12 VF |
2247 | netdev_for_each_all_upper_dev_rcu(bond->dev, vlan_upper, |
2248 | vlan_iter) { | |
27bc11e6 VF |
2249 | if (!is_vlan_dev(vlan_upper)) |
2250 | continue; | |
2f268f12 VF |
2251 | netdev_for_each_all_upper_dev_rcu(vlan_upper, upper, |
2252 | iter) { | |
27bc11e6 VF |
2253 | if (upper == rt->dst.dev) { |
2254 | vlan_id = vlan_dev_vlan_id(vlan_upper); | |
2255 | rcu_read_unlock(); | |
2256 | goto found; | |
2257 | } | |
c3ade5ca JV |
2258 | } |
2259 | } | |
2260 | ||
27bc11e6 VF |
2261 | /* if the device we're looking for is not on top of any of |
2262 | * our upper vlans, then just search for any dev that | |
2263 | * matches, and in case it's a vlan - save the id | |
2264 | */ | |
2f268f12 | 2265 | netdev_for_each_all_upper_dev_rcu(bond->dev, upper, iter) { |
27bc11e6 VF |
2266 | if (upper == rt->dst.dev) { |
2267 | /* if it's a vlan - get its VID */ | |
2268 | if (is_vlan_dev(upper)) | |
2269 | vlan_id = vlan_dev_vlan_id(upper); | |
2270 | ||
2271 | rcu_read_unlock(); | |
2272 | goto found; | |
2273 | } | |
c3ade5ca | 2274 | } |
27bc11e6 | 2275 | rcu_read_unlock(); |
c3ade5ca | 2276 | |
27bc11e6 | 2277 | /* Not our device - skip */ |
3e32582f VF |
2278 | pr_debug("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2279 | bond->dev->name, &targets[i], | |
2280 | rt->dst.dev ? rt->dst.dev->name : "NULL"); | |
2281 | ||
ed4b9f80 | 2282 | ip_rt_put(rt); |
27bc11e6 VF |
2283 | continue; |
2284 | ||
2285 | found: | |
2286 | addr = bond_confirm_addr(rt->dst.dev, targets[i], 0); | |
2287 | ip_rt_put(rt); | |
2288 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], | |
2289 | addr, vlan_id); | |
c3ade5ca JV |
2290 | } |
2291 | } | |
2292 | ||
d3bb52b0 | 2293 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2294 | { |
8599b52e VF |
2295 | int i; |
2296 | ||
87a7b84b VF |
2297 | if (!sip || !bond_has_this_ip(bond, tip)) { |
2298 | pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip); | |
2299 | return; | |
2300 | } | |
f5b2b966 | 2301 | |
8599b52e VF |
2302 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2303 | if (i == -1) { | |
87a7b84b VF |
2304 | pr_debug("bva: sip %pI4 not found in targets\n", &sip); |
2305 | return; | |
f5b2b966 | 2306 | } |
87a7b84b | 2307 | slave->last_arp_rx = jiffies; |
8599b52e | 2308 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2309 | } |
2310 | ||
5bb9e0b5 | 2311 | int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2312 | struct slave *slave) | |
f5b2b966 | 2313 | { |
de063b70 | 2314 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2315 | unsigned char *arp_ptr; |
d3bb52b0 | 2316 | __be32 sip, tip; |
de063b70 | 2317 | int alen; |
f5b2b966 | 2318 | |
3aba891d | 2319 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2320 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2321 | |
f5b2b966 | 2322 | read_lock(&bond->lock); |
2c146102 VF |
2323 | |
2324 | if (!slave_do_arp_validate(bond, slave)) | |
2325 | goto out_unlock; | |
2326 | ||
de063b70 | 2327 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2328 | |
3aba891d JP |
2329 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2330 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2331 | |
de063b70 ED |
2332 | if (alen > skb_headlen(skb)) { |
2333 | arp = kmalloc(alen, GFP_ATOMIC); | |
2334 | if (!arp) | |
2335 | goto out_unlock; | |
2336 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2337 | goto out_unlock; | |
2338 | } | |
f5b2b966 | 2339 | |
3aba891d | 2340 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2341 | skb->pkt_type == PACKET_OTHERHOST || |
2342 | skb->pkt_type == PACKET_LOOPBACK || | |
2343 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2344 | arp->ar_pro != htons(ETH_P_IP) || | |
2345 | arp->ar_pln != 4) | |
2346 | goto out_unlock; | |
2347 | ||
2348 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2349 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2350 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2351 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2352 | memcpy(&tip, arp_ptr, 4); |
2353 | ||
5a03cdb7 | 2354 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2355 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2356 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2357 | &sip, &tip); | |
f5b2b966 JV |
2358 | |
2359 | /* | |
2360 | * Backup slaves won't see the ARP reply, but do come through | |
2361 | * here for each ARP probe (so we swap the sip/tip to validate | |
2362 | * the probe). In a "redundant switch, common router" type of | |
2363 | * configuration, the ARP probe will (hopefully) travel from | |
2364 | * the active, through one switch, the router, then the other | |
2365 | * switch before reaching the backup. | |
aeea64ac VF |
2366 | * |
2367 | * We 'trust' the arp requests if there is an active slave and | |
2368 | * it received valid arp reply(s) after it became active. This | |
2369 | * is done to avoid endless looping when we can't reach the | |
2370 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2371 | */ |
e30bc066 | 2372 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2373 | bond_validate_arp(bond, slave, sip, tip); |
aeea64ac VF |
2374 | else if (bond->curr_active_slave && |
2375 | time_after(slave_last_rx(bond, bond->curr_active_slave), | |
2376 | bond->curr_active_slave->jiffies)) | |
f5b2b966 JV |
2377 | bond_validate_arp(bond, slave, tip, sip); |
2378 | ||
2379 | out_unlock: | |
2380 | read_unlock(&bond->lock); | |
de063b70 ED |
2381 | if (arp != (struct arphdr *)skb->data) |
2382 | kfree(arp); | |
13a8e0c8 | 2383 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2384 | } |
2385 | ||
e7f63f1d VF |
2386 | /* function to verify if we're in the arp_interval timeslice, returns true if |
2387 | * (last_act - arp_interval) <= jiffies <= (last_act + mod * arp_interval + | |
2388 | * arp_interval/2) . the arp_interval/2 is needed for really fast networks. | |
2389 | */ | |
2390 | static bool bond_time_in_interval(struct bonding *bond, unsigned long last_act, | |
2391 | int mod) | |
2392 | { | |
2393 | int delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); | |
2394 | ||
2395 | return time_in_range(jiffies, | |
2396 | last_act - delta_in_ticks, | |
2397 | last_act + mod * delta_in_ticks + delta_in_ticks/2); | |
2398 | } | |
2399 | ||
1da177e4 LT |
2400 | /* |
2401 | * this function is called regularly to monitor each slave's link | |
2402 | * ensuring that traffic is being sent and received when arp monitoring | |
2403 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2404 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2405 | * arp monitoring in active backup mode. | |
2406 | */ | |
6da67d26 | 2407 | static void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2408 | { |
1b76b316 JV |
2409 | struct bonding *bond = container_of(work, struct bonding, |
2410 | arp_work.work); | |
1da177e4 | 2411 | struct slave *slave, *oldcurrent; |
9caff1e7 | 2412 | struct list_head *iter; |
1da177e4 | 2413 | int do_failover = 0; |
1da177e4 | 2414 | |
1f2cd845 | 2415 | if (!bond_has_slaves(bond)) |
1da177e4 | 2416 | goto re_arm; |
1da177e4 | 2417 | |
2e52f4fe | 2418 | rcu_read_lock(); |
2419 | ||
2420 | oldcurrent = ACCESS_ONCE(bond->curr_active_slave); | |
1da177e4 LT |
2421 | /* see if any of the previous devices are up now (i.e. they have |
2422 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2423 | * the picture unless it is null. also, slave->jiffies is not needed | |
2424 | * here because we send an arp on each slave and give a slave as | |
2425 | * long as it needs to get the tx/rx within the delta. | |
2426 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2427 | * so it can wait | |
2428 | */ | |
2e52f4fe | 2429 | bond_for_each_slave_rcu(bond, slave, iter) { |
cb32f2a0 JB |
2430 | unsigned long trans_start = dev_trans_start(slave->dev); |
2431 | ||
1da177e4 | 2432 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d VF |
2433 | if (bond_time_in_interval(bond, trans_start, 1) && |
2434 | bond_time_in_interval(bond, slave->dev->last_rx, 1)) { | |
1da177e4 LT |
2435 | |
2436 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2437 | bond_set_active_slave(slave); |
1da177e4 LT |
2438 | |
2439 | /* primary_slave has no meaning in round-robin | |
2440 | * mode. the window of a slave being up and | |
2441 | * curr_active_slave being null after enslaving | |
2442 | * is closed. | |
2443 | */ | |
2444 | if (!oldcurrent) { | |
a4aee5c8 JP |
2445 | pr_info("%s: link status definitely up for interface %s, ", |
2446 | bond->dev->name, | |
2447 | slave->dev->name); | |
1da177e4 LT |
2448 | do_failover = 1; |
2449 | } else { | |
a4aee5c8 JP |
2450 | pr_info("%s: interface %s is now up\n", |
2451 | bond->dev->name, | |
2452 | slave->dev->name); | |
1da177e4 LT |
2453 | } |
2454 | } | |
2455 | } else { | |
2456 | /* slave->link == BOND_LINK_UP */ | |
2457 | ||
2458 | /* not all switches will respond to an arp request | |
2459 | * when the source ip is 0, so don't take the link down | |
2460 | * if we don't know our ip yet | |
2461 | */ | |
e7f63f1d VF |
2462 | if (!bond_time_in_interval(bond, trans_start, 2) || |
2463 | !bond_time_in_interval(bond, slave->dev->last_rx, 2)) { | |
1da177e4 LT |
2464 | |
2465 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2466 | bond_set_backup_slave(slave); |
1da177e4 | 2467 | |
3d632c3f | 2468 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2469 | slave->link_failure_count++; |
1da177e4 | 2470 | |
a4aee5c8 JP |
2471 | pr_info("%s: interface %s is now down.\n", |
2472 | bond->dev->name, | |
2473 | slave->dev->name); | |
1da177e4 | 2474 | |
3d632c3f | 2475 | if (slave == oldcurrent) |
1da177e4 | 2476 | do_failover = 1; |
1da177e4 LT |
2477 | } |
2478 | } | |
2479 | ||
2480 | /* note: if switch is in round-robin mode, all links | |
2481 | * must tx arp to ensure all links rx an arp - otherwise | |
2482 | * links may oscillate or not come up at all; if switch is | |
2483 | * in something like xor mode, there is nothing we can | |
2484 | * do - all replies will be rx'ed on same link causing slaves | |
2485 | * to be unstable during low/no traffic periods | |
2486 | */ | |
3d632c3f | 2487 | if (IS_UP(slave->dev)) |
1da177e4 | 2488 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2489 | } |
2490 | ||
2e52f4fe | 2491 | rcu_read_unlock(); |
2492 | ||
1da177e4 | 2493 | if (do_failover) { |
2e52f4fe | 2494 | /* the bond_select_active_slave must hold RTNL |
2495 | * and curr_slave_lock for write. | |
2496 | */ | |
2497 | if (!rtnl_trylock()) | |
2498 | goto re_arm; | |
e843fa50 | 2499 | block_netpoll_tx(); |
059fe7a5 | 2500 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2501 | |
2502 | bond_select_active_slave(bond); | |
2503 | ||
059fe7a5 | 2504 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2505 | unblock_netpoll_tx(); |
2e52f4fe | 2506 | rtnl_unlock(); |
1da177e4 LT |
2507 | } |
2508 | ||
2509 | re_arm: | |
e6d265e8 | 2510 | if (bond->params.arp_interval) |
e7f63f1d VF |
2511 | queue_delayed_work(bond->wq, &bond->arp_work, |
2512 | msecs_to_jiffies(bond->params.arp_interval)); | |
1da177e4 LT |
2513 | } |
2514 | ||
2515 | /* | |
b2220cad JV |
2516 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2517 | * changes. Sets new_link in slaves to specify what action should take | |
2518 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2519 | * to link states must be committed. | |
2520 | * | |
eb9fa4b0 | 2521 | * Called with rcu_read_lock hold. |
1da177e4 | 2522 | */ |
e7f63f1d | 2523 | static int bond_ab_arp_inspect(struct bonding *bond) |
1da177e4 | 2524 | { |
def4460c | 2525 | unsigned long trans_start, last_rx; |
9caff1e7 | 2526 | struct list_head *iter; |
dec1e90e | 2527 | struct slave *slave; |
dec1e90e | 2528 | int commit = 0; |
da210f55 | 2529 | |
eb9fa4b0 | 2530 | bond_for_each_slave_rcu(bond, slave, iter) { |
b2220cad | 2531 | slave->new_link = BOND_LINK_NOCHANGE; |
def4460c | 2532 | last_rx = slave_last_rx(bond, slave); |
1da177e4 | 2533 | |
b2220cad | 2534 | if (slave->link != BOND_LINK_UP) { |
e7f63f1d | 2535 | if (bond_time_in_interval(bond, last_rx, 1)) { |
b2220cad JV |
2536 | slave->new_link = BOND_LINK_UP; |
2537 | commit++; | |
2538 | } | |
b2220cad JV |
2539 | continue; |
2540 | } | |
1da177e4 | 2541 | |
b2220cad JV |
2542 | /* |
2543 | * Give slaves 2*delta after being enslaved or made | |
2544 | * active. This avoids bouncing, as the last receive | |
2545 | * times need a full ARP monitor cycle to be updated. | |
2546 | */ | |
e7f63f1d | 2547 | if (bond_time_in_interval(bond, slave->jiffies, 2)) |
b2220cad JV |
2548 | continue; |
2549 | ||
2550 | /* | |
2551 | * Backup slave is down if: | |
2552 | * - No current_arp_slave AND | |
2553 | * - more than 3*delta since last receive AND | |
2554 | * - the bond has an IP address | |
2555 | * | |
2556 | * Note: a non-null current_arp_slave indicates | |
2557 | * the curr_active_slave went down and we are | |
2558 | * searching for a new one; under this condition | |
2559 | * we only take the curr_active_slave down - this | |
2560 | * gives each slave a chance to tx/rx traffic | |
2561 | * before being taken out | |
2562 | */ | |
e30bc066 | 2563 | if (!bond_is_active_slave(slave) && |
b2220cad | 2564 | !bond->current_arp_slave && |
e7f63f1d | 2565 | !bond_time_in_interval(bond, last_rx, 3)) { |
b2220cad JV |
2566 | slave->new_link = BOND_LINK_DOWN; |
2567 | commit++; | |
2568 | } | |
2569 | ||
2570 | /* | |
2571 | * Active slave is down if: | |
2572 | * - more than 2*delta since transmitting OR | |
2573 | * - (more than 2*delta since receive AND | |
2574 | * the bond has an IP address) | |
2575 | */ | |
cb32f2a0 | 2576 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2577 | if (bond_is_active_slave(slave) && |
e7f63f1d VF |
2578 | (!bond_time_in_interval(bond, trans_start, 2) || |
2579 | !bond_time_in_interval(bond, last_rx, 2))) { | |
b2220cad JV |
2580 | slave->new_link = BOND_LINK_DOWN; |
2581 | commit++; | |
2582 | } | |
1da177e4 LT |
2583 | } |
2584 | ||
b2220cad JV |
2585 | return commit; |
2586 | } | |
1da177e4 | 2587 | |
b2220cad JV |
2588 | /* |
2589 | * Called to commit link state changes noted by inspection step of | |
2590 | * active-backup mode ARP monitor. | |
2591 | * | |
eb9fa4b0 | 2592 | * Called with RTNL hold. |
b2220cad | 2593 | */ |
e7f63f1d | 2594 | static void bond_ab_arp_commit(struct bonding *bond) |
b2220cad | 2595 | { |
cb32f2a0 | 2596 | unsigned long trans_start; |
9caff1e7 | 2597 | struct list_head *iter; |
dec1e90e | 2598 | struct slave *slave; |
1da177e4 | 2599 | |
9caff1e7 | 2600 | bond_for_each_slave(bond, slave, iter) { |
b2220cad JV |
2601 | switch (slave->new_link) { |
2602 | case BOND_LINK_NOCHANGE: | |
2603 | continue; | |
ff59c456 | 2604 | |
b2220cad | 2605 | case BOND_LINK_UP: |
cb32f2a0 | 2606 | trans_start = dev_trans_start(slave->dev); |
e7f63f1d VF |
2607 | if (bond->curr_active_slave != slave || |
2608 | (!bond->curr_active_slave && | |
2609 | bond_time_in_interval(bond, trans_start, 1))) { | |
b2220cad | 2610 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2611 | if (bond->current_arp_slave) { |
2612 | bond_set_slave_inactive_flags( | |
2613 | bond->current_arp_slave); | |
2614 | bond->current_arp_slave = NULL; | |
2615 | } | |
b2220cad | 2616 | |
a4aee5c8 | 2617 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2618 | bond->dev->name, slave->dev->name); |
b2220cad | 2619 | |
b9f60253 JP |
2620 | if (!bond->curr_active_slave || |
2621 | (slave == bond->primary_slave)) | |
2622 | goto do_failover; | |
1da177e4 | 2623 | |
b9f60253 | 2624 | } |
1da177e4 | 2625 | |
b9f60253 | 2626 | continue; |
1da177e4 | 2627 | |
b2220cad JV |
2628 | case BOND_LINK_DOWN: |
2629 | if (slave->link_failure_count < UINT_MAX) | |
2630 | slave->link_failure_count++; | |
2631 | ||
2632 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2633 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2634 | |
a4aee5c8 | 2635 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2636 | bond->dev->name, slave->dev->name); |
b2220cad | 2637 | |
b9f60253 | 2638 | if (slave == bond->curr_active_slave) { |
b2220cad | 2639 | bond->current_arp_slave = NULL; |
b9f60253 | 2640 | goto do_failover; |
1da177e4 | 2641 | } |
b9f60253 JP |
2642 | |
2643 | continue; | |
b2220cad JV |
2644 | |
2645 | default: | |
a4aee5c8 | 2646 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2647 | bond->dev->name, slave->new_link, |
2648 | slave->dev->name); | |
b9f60253 | 2649 | continue; |
1da177e4 | 2650 | } |
1da177e4 | 2651 | |
b9f60253 JP |
2652 | do_failover: |
2653 | ASSERT_RTNL(); | |
e843fa50 | 2654 | block_netpoll_tx(); |
b2220cad | 2655 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2656 | bond_select_active_slave(bond); |
b2220cad | 2657 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2658 | unblock_netpoll_tx(); |
b2220cad | 2659 | } |
1da177e4 | 2660 | |
b2220cad JV |
2661 | bond_set_carrier(bond); |
2662 | } | |
1da177e4 | 2663 | |
b2220cad JV |
2664 | /* |
2665 | * Send ARP probes for active-backup mode ARP monitor. | |
2666 | * | |
eb9fa4b0 | 2667 | * Called with rcu_read_lock hold. |
b2220cad JV |
2668 | */ |
2669 | static void bond_ab_arp_probe(struct bonding *bond) | |
2670 | { | |
eb9fa4b0 | 2671 | struct slave *slave, *before = NULL, *new_slave = NULL, |
2672 | *curr_arp_slave = rcu_dereference(bond->current_arp_slave); | |
4087df87 VF |
2673 | struct list_head *iter; |
2674 | bool found = false; | |
1da177e4 | 2675 | |
b2220cad | 2676 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2677 | |
eb9fa4b0 | 2678 | if (curr_arp_slave && bond->curr_active_slave) |
a4aee5c8 | 2679 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
eb9fa4b0 | 2680 | curr_arp_slave->dev->name, |
a4aee5c8 | 2681 | bond->curr_active_slave->dev->name); |
1da177e4 | 2682 | |
b2220cad JV |
2683 | if (bond->curr_active_slave) { |
2684 | bond_arp_send_all(bond, bond->curr_active_slave); | |
2685 | read_unlock(&bond->curr_slave_lock); | |
2686 | return; | |
2687 | } | |
1da177e4 | 2688 | |
b2220cad | 2689 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 2690 | |
b2220cad JV |
2691 | /* if we don't have a curr_active_slave, search for the next available |
2692 | * backup slave from the current_arp_slave and make it the candidate | |
2693 | * for becoming the curr_active_slave | |
2694 | */ | |
1da177e4 | 2695 | |
eb9fa4b0 | 2696 | if (!curr_arp_slave) { |
2697 | curr_arp_slave = bond_first_slave_rcu(bond); | |
2698 | if (!curr_arp_slave) | |
b2220cad JV |
2699 | return; |
2700 | } | |
1da177e4 | 2701 | |
eb9fa4b0 | 2702 | bond_set_slave_inactive_flags(curr_arp_slave); |
059fe7a5 | 2703 | |
eb9fa4b0 | 2704 | bond_for_each_slave_rcu(bond, slave, iter) { |
4087df87 VF |
2705 | if (!found && !before && IS_UP(slave->dev)) |
2706 | before = slave; | |
1da177e4 | 2707 | |
4087df87 VF |
2708 | if (found && !new_slave && IS_UP(slave->dev)) |
2709 | new_slave = slave; | |
b2220cad JV |
2710 | /* if the link state is up at this point, we |
2711 | * mark it down - this can happen if we have | |
2712 | * simultaneous link failures and | |
2713 | * reselect_active_interface doesn't make this | |
2714 | * one the current slave so it is still marked | |
2715 | * up when it is actually down | |
1da177e4 | 2716 | */ |
4087df87 | 2717 | if (!IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { |
b2220cad JV |
2718 | slave->link = BOND_LINK_DOWN; |
2719 | if (slave->link_failure_count < UINT_MAX) | |
2720 | slave->link_failure_count++; | |
1da177e4 | 2721 | |
b2220cad JV |
2722 | bond_set_slave_inactive_flags(slave); |
2723 | ||
a4aee5c8 JP |
2724 | pr_info("%s: backup interface %s is now down.\n", |
2725 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2726 | } |
eb9fa4b0 | 2727 | if (slave == curr_arp_slave) |
4087df87 | 2728 | found = true; |
b2220cad | 2729 | } |
4087df87 VF |
2730 | |
2731 | if (!new_slave && before) | |
2732 | new_slave = before; | |
2733 | ||
2734 | if (!new_slave) | |
2735 | return; | |
2736 | ||
2737 | new_slave->link = BOND_LINK_BACK; | |
2738 | bond_set_slave_active_flags(new_slave); | |
2739 | bond_arp_send_all(bond, new_slave); | |
2740 | new_slave->jiffies = jiffies; | |
eb9fa4b0 | 2741 | rcu_assign_pointer(bond->current_arp_slave, new_slave); |
b2220cad | 2742 | } |
1da177e4 | 2743 | |
6da67d26 | 2744 | static void bond_activebackup_arp_mon(struct work_struct *work) |
b2220cad JV |
2745 | { |
2746 | struct bonding *bond = container_of(work, struct bonding, | |
2747 | arp_work.work); | |
ad246c99 | 2748 | bool should_notify_peers = false; |
1f2cd845 | 2749 | int delta_in_ticks; |
1da177e4 | 2750 | |
1f2cd845 DM |
2751 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
2752 | ||
2753 | if (!bond_has_slaves(bond)) | |
b2220cad JV |
2754 | goto re_arm; |
2755 | ||
eb9fa4b0 | 2756 | rcu_read_lock(); |
2757 | ||
ad246c99 BH |
2758 | should_notify_peers = bond_should_notify_peers(bond); |
2759 | ||
1f2cd845 | 2760 | if (bond_ab_arp_inspect(bond)) { |
eb9fa4b0 | 2761 | rcu_read_unlock(); |
b2220cad | 2762 | |
1f2cd845 DM |
2763 | /* Race avoidance with bond_close flush of workqueue */ |
2764 | if (!rtnl_trylock()) { | |
1f2cd845 DM |
2765 | delta_in_ticks = 1; |
2766 | should_notify_peers = false; | |
2767 | goto re_arm; | |
2768 | } | |
b2220cad | 2769 | |
1f2cd845 DM |
2770 | bond_ab_arp_commit(bond); |
2771 | ||
1f2cd845 | 2772 | rtnl_unlock(); |
eb9fa4b0 | 2773 | rcu_read_lock(); |
1f2cd845 DM |
2774 | } |
2775 | ||
2776 | bond_ab_arp_probe(bond); | |
eb9fa4b0 | 2777 | rcu_read_unlock(); |
ad246c99 | 2778 | |
80b9d236 | 2779 | re_arm: |
2780 | if (bond->params.arp_interval) | |
1f2cd845 DM |
2781 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
2782 | ||
1f2cd845 DM |
2783 | if (should_notify_peers) { |
2784 | if (!rtnl_trylock()) | |
2785 | return; | |
2786 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); | |
2787 | rtnl_unlock(); | |
2788 | } | |
1da177e4 LT |
2789 | } |
2790 | ||
1da177e4 LT |
2791 | /*-------------------------- netdev event handling --------------------------*/ |
2792 | ||
2793 | /* | |
2794 | * Change device name | |
2795 | */ | |
2796 | static int bond_event_changename(struct bonding *bond) | |
2797 | { | |
1da177e4 LT |
2798 | bond_remove_proc_entry(bond); |
2799 | bond_create_proc_entry(bond); | |
7e083840 | 2800 | |
f073c7ca TI |
2801 | bond_debug_reregister(bond); |
2802 | ||
1da177e4 LT |
2803 | return NOTIFY_DONE; |
2804 | } | |
2805 | ||
3d632c3f SH |
2806 | static int bond_master_netdev_event(unsigned long event, |
2807 | struct net_device *bond_dev) | |
1da177e4 | 2808 | { |
454d7c9b | 2809 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
2810 | |
2811 | switch (event) { | |
2812 | case NETDEV_CHANGENAME: | |
2813 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
2814 | case NETDEV_UNREGISTER: |
2815 | bond_remove_proc_entry(event_bond); | |
2816 | break; | |
2817 | case NETDEV_REGISTER: | |
2818 | bond_create_proc_entry(event_bond); | |
2819 | break; | |
6c8d23f7 | 2820 | case NETDEV_NOTIFY_PEERS: |
2821 | if (event_bond->send_peer_notif) | |
2822 | event_bond->send_peer_notif--; | |
2823 | break; | |
1da177e4 LT |
2824 | default: |
2825 | break; | |
2826 | } | |
2827 | ||
2828 | return NOTIFY_DONE; | |
2829 | } | |
2830 | ||
3d632c3f SH |
2831 | static int bond_slave_netdev_event(unsigned long event, |
2832 | struct net_device *slave_dev) | |
1da177e4 | 2833 | { |
471cb5a3 | 2834 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
6101391d | 2835 | struct bonding *bond; |
2836 | struct net_device *bond_dev; | |
471cb5a3 JP |
2837 | u32 old_speed; |
2838 | u8 old_duplex; | |
1da177e4 | 2839 | |
6101391d | 2840 | /* A netdev event can be generated while enslaving a device |
2841 | * before netdev_rx_handler_register is called in which case | |
2842 | * slave will be NULL | |
2843 | */ | |
2844 | if (!slave) | |
2845 | return NOTIFY_DONE; | |
2846 | bond_dev = slave->bond->dev; | |
2847 | bond = slave->bond; | |
2848 | ||
1da177e4 LT |
2849 | switch (event) { |
2850 | case NETDEV_UNREGISTER: | |
8d2ada77 | 2851 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
2852 | bond_release_and_destroy(bond_dev, slave_dev); |
2853 | else | |
2854 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 2855 | break; |
98f41f69 | 2856 | case NETDEV_UP: |
1da177e4 | 2857 | case NETDEV_CHANGE: |
471cb5a3 JP |
2858 | old_speed = slave->speed; |
2859 | old_duplex = slave->duplex; | |
17d04500 | 2860 | |
471cb5a3 | 2861 | bond_update_speed_duplex(slave); |
17d04500 | 2862 | |
471cb5a3 JP |
2863 | if (bond->params.mode == BOND_MODE_8023AD) { |
2864 | if (old_speed != slave->speed) | |
2865 | bond_3ad_adapter_speed_changed(slave); | |
2866 | if (old_duplex != slave->duplex) | |
2867 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 2868 | } |
1da177e4 LT |
2869 | break; |
2870 | case NETDEV_DOWN: | |
2871 | /* | |
2872 | * ... Or is it this? | |
2873 | */ | |
2874 | break; | |
2875 | case NETDEV_CHANGEMTU: | |
2876 | /* | |
2877 | * TODO: Should slaves be allowed to | |
2878 | * independently alter their MTU? For | |
2879 | * an active-backup bond, slaves need | |
2880 | * not be the same type of device, so | |
2881 | * MTUs may vary. For other modes, | |
2882 | * slaves arguably should have the | |
2883 | * same MTUs. To do this, we'd need to | |
2884 | * take over the slave's change_mtu | |
2885 | * function for the duration of their | |
2886 | * servitude. | |
2887 | */ | |
2888 | break; | |
2889 | case NETDEV_CHANGENAME: | |
3ec775b9 VF |
2890 | /* we don't care if we don't have primary set */ |
2891 | if (!USES_PRIMARY(bond->params.mode) || | |
2892 | !bond->params.primary[0]) | |
2893 | break; | |
2894 | ||
2895 | if (slave == bond->primary_slave) { | |
2896 | /* slave's name changed - he's no longer primary */ | |
2897 | bond->primary_slave = NULL; | |
2898 | } else if (!strcmp(slave_dev->name, bond->params.primary)) { | |
2899 | /* we have a new primary slave */ | |
2900 | bond->primary_slave = slave; | |
2901 | } else { /* we didn't change primary - exit */ | |
2902 | break; | |
2903 | } | |
2904 | ||
2905 | pr_info("%s: Primary slave changed to %s, reselecting active slave.\n", | |
2906 | bond->dev->name, bond->primary_slave ? slave_dev->name : | |
2907 | "none"); | |
2908 | write_lock_bh(&bond->curr_slave_lock); | |
2909 | bond_select_active_slave(bond); | |
2910 | write_unlock_bh(&bond->curr_slave_lock); | |
1da177e4 | 2911 | break; |
8531c5ff AK |
2912 | case NETDEV_FEAT_CHANGE: |
2913 | bond_compute_features(bond); | |
2914 | break; | |
4aa5dee4 JP |
2915 | case NETDEV_RESEND_IGMP: |
2916 | /* Propagate to master device */ | |
2917 | call_netdevice_notifiers(event, slave->bond->dev); | |
2918 | break; | |
1da177e4 LT |
2919 | default: |
2920 | break; | |
2921 | } | |
2922 | ||
2923 | return NOTIFY_DONE; | |
2924 | } | |
2925 | ||
2926 | /* | |
2927 | * bond_netdev_event: handle netdev notifier chain events. | |
2928 | * | |
2929 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 2930 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
2931 | * locks for us to safely manipulate the slave devices (RTNL lock, |
2932 | * dev_probe_lock). | |
2933 | */ | |
3d632c3f SH |
2934 | static int bond_netdev_event(struct notifier_block *this, |
2935 | unsigned long event, void *ptr) | |
1da177e4 | 2936 | { |
351638e7 | 2937 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 2938 | |
5a03cdb7 | 2939 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
2940 | event_dev ? event_dev->name : "None", |
2941 | event); | |
1da177e4 | 2942 | |
0b680e75 JV |
2943 | if (!(event_dev->priv_flags & IFF_BONDING)) |
2944 | return NOTIFY_DONE; | |
2945 | ||
1da177e4 | 2946 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 2947 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
2948 | return bond_master_netdev_event(event, event_dev); |
2949 | } | |
2950 | ||
2951 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 2952 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
2953 | return bond_slave_netdev_event(event, event_dev); |
2954 | } | |
2955 | ||
2956 | return NOTIFY_DONE; | |
2957 | } | |
2958 | ||
2959 | static struct notifier_block bond_netdev_notifier = { | |
2960 | .notifier_call = bond_netdev_event, | |
2961 | }; | |
2962 | ||
169a3e66 JV |
2963 | /*---------------------------- Hashing Policies -----------------------------*/ |
2964 | ||
32819dc1 NA |
2965 | /* L2 hash helper */ |
2966 | static inline u32 bond_eth_hash(struct sk_buff *skb) | |
6b923cb7 JE |
2967 | { |
2968 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
2969 | ||
2970 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
32819dc1 | 2971 | return data->h_dest[5] ^ data->h_source[5]; |
6b923cb7 JE |
2972 | |
2973 | return 0; | |
2974 | } | |
2975 | ||
32819dc1 NA |
2976 | /* Extract the appropriate headers based on bond's xmit policy */ |
2977 | static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb, | |
2978 | struct flow_keys *fk) | |
6f6652be | 2979 | { |
32819dc1 | 2980 | const struct ipv6hdr *iph6; |
4394542c | 2981 | const struct iphdr *iph; |
32819dc1 | 2982 | int noff, proto = -1; |
6b923cb7 | 2983 | |
32819dc1 NA |
2984 | if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23) |
2985 | return skb_flow_dissect(skb, fk); | |
2986 | ||
2987 | fk->ports = 0; | |
2988 | noff = skb_network_offset(skb); | |
2989 | if (skb->protocol == htons(ETH_P_IP)) { | |
2990 | if (!pskb_may_pull(skb, noff + sizeof(*iph))) | |
2991 | return false; | |
6b923cb7 | 2992 | iph = ip_hdr(skb); |
32819dc1 NA |
2993 | fk->src = iph->saddr; |
2994 | fk->dst = iph->daddr; | |
2995 | noff += iph->ihl << 2; | |
2996 | if (!ip_is_fragment(iph)) | |
2997 | proto = iph->protocol; | |
2998 | } else if (skb->protocol == htons(ETH_P_IPV6)) { | |
2999 | if (!pskb_may_pull(skb, noff + sizeof(*iph6))) | |
3000 | return false; | |
3001 | iph6 = ipv6_hdr(skb); | |
3002 | fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr); | |
3003 | fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr); | |
3004 | noff += sizeof(*iph6); | |
3005 | proto = iph6->nexthdr; | |
3006 | } else { | |
3007 | return false; | |
6f6652be | 3008 | } |
32819dc1 NA |
3009 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0) |
3010 | fk->ports = skb_flow_get_ports(skb, noff, proto); | |
6f6652be | 3011 | |
32819dc1 | 3012 | return true; |
6f6652be JV |
3013 | } |
3014 | ||
32819dc1 NA |
3015 | /** |
3016 | * bond_xmit_hash - generate a hash value based on the xmit policy | |
3017 | * @bond: bonding device | |
3018 | * @skb: buffer to use for headers | |
3019 | * @count: modulo value | |
3020 | * | |
3021 | * This function will extract the necessary headers from the skb buffer and use | |
3022 | * them to generate a hash based on the xmit_policy set in the bonding device | |
3023 | * which will be reduced modulo count before returning. | |
169a3e66 | 3024 | */ |
32819dc1 | 3025 | int bond_xmit_hash(struct bonding *bond, struct sk_buff *skb, int count) |
169a3e66 | 3026 | { |
32819dc1 NA |
3027 | struct flow_keys flow; |
3028 | u32 hash; | |
4394542c | 3029 | |
32819dc1 NA |
3030 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER2 || |
3031 | !bond_flow_dissect(bond, skb, &flow)) | |
3032 | return bond_eth_hash(skb) % count; | |
169a3e66 | 3033 | |
32819dc1 NA |
3034 | if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER23 || |
3035 | bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23) | |
3036 | hash = bond_eth_hash(skb); | |
3037 | else | |
3038 | hash = (__force u32)flow.ports; | |
3039 | hash ^= (__force u32)flow.dst ^ (__force u32)flow.src; | |
3040 | hash ^= (hash >> 16); | |
3041 | hash ^= (hash >> 8); | |
3042 | ||
3043 | return hash % count; | |
169a3e66 JV |
3044 | } |
3045 | ||
1da177e4 LT |
3046 | /*-------------------------- Device entry points ----------------------------*/ |
3047 | ||
fbb0c41b | 3048 | static void bond_work_init_all(struct bonding *bond) |
3049 | { | |
3050 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3051 | bond_resend_igmp_join_requests_delayed); | |
3052 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3053 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3054 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3055 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3056 | else | |
3057 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3058 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3059 | } | |
3060 | ||
3061 | static void bond_work_cancel_all(struct bonding *bond) | |
3062 | { | |
3063 | cancel_delayed_work_sync(&bond->mii_work); | |
3064 | cancel_delayed_work_sync(&bond->arp_work); | |
3065 | cancel_delayed_work_sync(&bond->alb_work); | |
3066 | cancel_delayed_work_sync(&bond->ad_work); | |
3067 | cancel_delayed_work_sync(&bond->mcast_work); | |
3068 | } | |
3069 | ||
1da177e4 LT |
3070 | static int bond_open(struct net_device *bond_dev) |
3071 | { | |
454d7c9b | 3072 | struct bonding *bond = netdev_priv(bond_dev); |
9caff1e7 | 3073 | struct list_head *iter; |
ba3211cc | 3074 | struct slave *slave; |
1da177e4 | 3075 | |
ba3211cc PP |
3076 | /* reset slave->backup and slave->inactive */ |
3077 | read_lock(&bond->lock); | |
0965a1f3 | 3078 | if (bond_has_slaves(bond)) { |
ba3211cc | 3079 | read_lock(&bond->curr_slave_lock); |
9caff1e7 | 3080 | bond_for_each_slave(bond, slave, iter) { |
ba3211cc PP |
3081 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3082 | && (slave != bond->curr_active_slave)) { | |
3083 | bond_set_slave_inactive_flags(slave); | |
3084 | } else { | |
3085 | bond_set_slave_active_flags(slave); | |
3086 | } | |
3087 | } | |
3088 | read_unlock(&bond->curr_slave_lock); | |
3089 | } | |
3090 | read_unlock(&bond->lock); | |
3091 | ||
fbb0c41b | 3092 | bond_work_init_all(bond); |
5a37e8ca | 3093 | |
58402054 | 3094 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3095 | /* bond_alb_initialize must be called before the timer |
3096 | * is started. | |
3097 | */ | |
fbb0c41b | 3098 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3099 | return -ENOMEM; |
1b76b316 | 3100 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3101 | } |
3102 | ||
fbb0c41b | 3103 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3104 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3105 | |
3106 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3107 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
f5b2b966 | 3108 | if (bond->params.arp_validate) |
3aba891d | 3109 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3110 | } |
3111 | ||
3112 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3113 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3114 | /* register to receive LACPDUs */ |
3aba891d | 3115 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3116 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3117 | } |
3118 | ||
3119 | return 0; | |
3120 | } | |
3121 | ||
3122 | static int bond_close(struct net_device *bond_dev) | |
3123 | { | |
454d7c9b | 3124 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3125 | |
fbb0c41b | 3126 | bond_work_cancel_all(bond); |
6c8d23f7 | 3127 | bond->send_peer_notif = 0; |
ee8487c0 | 3128 | if (bond_is_lb(bond)) |
1da177e4 | 3129 | bond_alb_deinitialize(bond); |
3aba891d | 3130 | bond->recv_probe = NULL; |
1da177e4 LT |
3131 | |
3132 | return 0; | |
3133 | } | |
3134 | ||
28172739 ED |
3135 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3136 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3137 | { |
454d7c9b | 3138 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3139 | struct rtnl_link_stats64 temp; |
9caff1e7 | 3140 | struct list_head *iter; |
1da177e4 | 3141 | struct slave *slave; |
1da177e4 | 3142 | |
28172739 | 3143 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3144 | |
3145 | read_lock_bh(&bond->lock); | |
9caff1e7 | 3146 | bond_for_each_slave(bond, slave, iter) { |
be1f3c2c | 3147 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3148 | dev_get_stats(slave->dev, &temp); |
3149 | ||
3150 | stats->rx_packets += sstats->rx_packets; | |
3151 | stats->rx_bytes += sstats->rx_bytes; | |
3152 | stats->rx_errors += sstats->rx_errors; | |
3153 | stats->rx_dropped += sstats->rx_dropped; | |
3154 | ||
3155 | stats->tx_packets += sstats->tx_packets; | |
3156 | stats->tx_bytes += sstats->tx_bytes; | |
3157 | stats->tx_errors += sstats->tx_errors; | |
3158 | stats->tx_dropped += sstats->tx_dropped; | |
3159 | ||
3160 | stats->multicast += sstats->multicast; | |
3161 | stats->collisions += sstats->collisions; | |
3162 | ||
3163 | stats->rx_length_errors += sstats->rx_length_errors; | |
3164 | stats->rx_over_errors += sstats->rx_over_errors; | |
3165 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3166 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3167 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3168 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3169 | ||
3170 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3171 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3172 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3173 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3174 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb | 3175 | } |
1da177e4 LT |
3176 | read_unlock_bh(&bond->lock); |
3177 | ||
3178 | return stats; | |
3179 | } | |
3180 | ||
3181 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3182 | { | |
080a06e1 | 3183 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3184 | struct net_device *slave_dev = NULL; |
3185 | struct ifbond k_binfo; | |
3186 | struct ifbond __user *u_binfo = NULL; | |
3187 | struct ifslave k_sinfo; | |
3188 | struct ifslave __user *u_sinfo = NULL; | |
3189 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3190 | struct net *net; |
1da177e4 LT |
3191 | int res = 0; |
3192 | ||
a4aee5c8 | 3193 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3194 | |
3195 | switch (cmd) { | |
1da177e4 LT |
3196 | case SIOCGMIIPHY: |
3197 | mii = if_mii(ifr); | |
3d632c3f | 3198 | if (!mii) |
1da177e4 | 3199 | return -EINVAL; |
3d632c3f | 3200 | |
1da177e4 LT |
3201 | mii->phy_id = 0; |
3202 | /* Fall Through */ | |
3203 | case SIOCGMIIREG: | |
3204 | /* | |
3205 | * We do this again just in case we were called by SIOCGMIIREG | |
3206 | * instead of SIOCGMIIPHY. | |
3207 | */ | |
3208 | mii = if_mii(ifr); | |
3d632c3f | 3209 | if (!mii) |
1da177e4 | 3210 | return -EINVAL; |
3d632c3f | 3211 | |
1da177e4 LT |
3212 | |
3213 | if (mii->reg_num == 1) { | |
1da177e4 | 3214 | mii->val_out = 0; |
6603a6f2 | 3215 | read_lock(&bond->lock); |
1da177e4 | 3216 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3217 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3218 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3219 | |
1da177e4 | 3220 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3221 | read_unlock(&bond->lock); |
1da177e4 LT |
3222 | } |
3223 | ||
3224 | return 0; | |
3225 | case BOND_INFO_QUERY_OLD: | |
3226 | case SIOCBONDINFOQUERY: | |
3227 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3228 | ||
3d632c3f | 3229 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3230 | return -EFAULT; |
1da177e4 LT |
3231 | |
3232 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3233 | if (res == 0 && |
3234 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3235 | return -EFAULT; | |
1da177e4 LT |
3236 | |
3237 | return res; | |
3238 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3239 | case SIOCBONDSLAVEINFOQUERY: | |
3240 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3241 | ||
3d632c3f | 3242 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3243 | return -EFAULT; |
1da177e4 LT |
3244 | |
3245 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3246 | if (res == 0 && |
3247 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3248 | return -EFAULT; | |
1da177e4 LT |
3249 | |
3250 | return res; | |
3251 | default: | |
3252 | /* Go on */ | |
3253 | break; | |
3254 | } | |
3255 | ||
387ff911 G |
3256 | net = dev_net(bond_dev); |
3257 | ||
3258 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3259 | return -EPERM; |
1da177e4 | 3260 | |
0917b933 | 3261 | slave_dev = __dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3262 | |
a4aee5c8 | 3263 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3264 | |
3d632c3f | 3265 | if (!slave_dev) |
0917b933 | 3266 | return -ENODEV; |
1da177e4 | 3267 | |
0917b933 YX |
3268 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
3269 | switch (cmd) { | |
3270 | case BOND_ENSLAVE_OLD: | |
3271 | case SIOCBONDENSLAVE: | |
3272 | res = bond_enslave(bond_dev, slave_dev); | |
3273 | break; | |
3274 | case BOND_RELEASE_OLD: | |
3275 | case SIOCBONDRELEASE: | |
3276 | res = bond_release(bond_dev, slave_dev); | |
3277 | break; | |
3278 | case BOND_SETHWADDR_OLD: | |
3279 | case SIOCBONDSETHWADDR: | |
3280 | bond_set_dev_addr(bond_dev, slave_dev); | |
3281 | res = 0; | |
3282 | break; | |
3283 | case BOND_CHANGE_ACTIVE_OLD: | |
3284 | case SIOCBONDCHANGEACTIVE: | |
3285 | res = bond_option_active_slave_set(bond, slave_dev); | |
3286 | break; | |
3287 | default: | |
3288 | res = -EOPNOTSUPP; | |
1da177e4 LT |
3289 | } |
3290 | ||
1da177e4 LT |
3291 | return res; |
3292 | } | |
3293 | ||
d03462b9 | 3294 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3295 | { |
454d7c9b | 3296 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3297 | |
d03462b9 JP |
3298 | if (change & IFF_PROMISC) |
3299 | bond_set_promiscuity(bond, | |
3300 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3301 | |
d03462b9 JP |
3302 | if (change & IFF_ALLMULTI) |
3303 | bond_set_allmulti(bond, | |
3304 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3305 | } | |
1da177e4 | 3306 | |
303d1cbf | 3307 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3308 | { |
3309 | struct bonding *bond = netdev_priv(bond_dev); | |
9caff1e7 | 3310 | struct list_head *iter; |
303d1cbf | 3311 | struct slave *slave; |
1da177e4 | 3312 | |
80ee5ad2 | 3313 | |
b3241870 | 3314 | rcu_read_lock(); |
303d1cbf | 3315 | if (USES_PRIMARY(bond->params.mode)) { |
b3241870 | 3316 | slave = rcu_dereference(bond->curr_active_slave); |
303d1cbf JV |
3317 | if (slave) { |
3318 | dev_uc_sync(slave->dev, bond_dev); | |
3319 | dev_mc_sync(slave->dev, bond_dev); | |
3320 | } | |
303d1cbf | 3321 | } else { |
b3241870 | 3322 | bond_for_each_slave_rcu(bond, slave, iter) { |
303d1cbf JV |
3323 | dev_uc_sync_multiple(slave->dev, bond_dev); |
3324 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3325 | } | |
1da177e4 | 3326 | } |
b3241870 | 3327 | rcu_read_unlock(); |
1da177e4 LT |
3328 | } |
3329 | ||
234bcf8a | 3330 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3331 | { |
234bcf8a | 3332 | struct bonding *bond = netdev_priv(n->dev); |
234bcf8a SP |
3333 | const struct net_device_ops *slave_ops; |
3334 | struct neigh_parms parms; | |
dec1e90e | 3335 | struct slave *slave; |
234bcf8a SP |
3336 | int ret; |
3337 | ||
dec1e90e | 3338 | slave = bond_first_slave(bond); |
234bcf8a SP |
3339 | if (!slave) |
3340 | return 0; | |
234bcf8a | 3341 | slave_ops = slave->dev->netdev_ops; |
234bcf8a SP |
3342 | if (!slave_ops->ndo_neigh_setup) |
3343 | return 0; | |
3344 | ||
3345 | parms.neigh_setup = NULL; | |
3346 | parms.neigh_cleanup = NULL; | |
3347 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3348 | if (ret) | |
3349 | return ret; | |
3350 | ||
3351 | /* | |
3352 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3353 | * after the last slave has been detached. Assumes that all slaves | |
3354 | * utilize the same neigh_cleanup (true at this writing as only user | |
3355 | * is ipoib). | |
3356 | */ | |
3357 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3358 | ||
3359 | if (!parms.neigh_setup) | |
3360 | return 0; | |
3361 | ||
3362 | return parms.neigh_setup(n); | |
3363 | } | |
3364 | ||
3365 | /* | |
3366 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3367 | * slave exists. So we must declare proxy setup function which will | |
3368 | * be used at run time to resolve the actual slave neigh param setup. | |
9918d5bf VF |
3369 | * |
3370 | * It's also called by master devices (such as vlans) to setup their | |
3371 | * underlying devices. In that case - do nothing, we're already set up from | |
3372 | * our init. | |
234bcf8a SP |
3373 | */ |
3374 | static int bond_neigh_setup(struct net_device *dev, | |
3375 | struct neigh_parms *parms) | |
3376 | { | |
9918d5bf VF |
3377 | /* modify only our neigh_parms */ |
3378 | if (parms->dev == dev) | |
3379 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3380 | |
00829823 SH |
3381 | return 0; |
3382 | } | |
3383 | ||
1da177e4 LT |
3384 | /* |
3385 | * Change the MTU of all of a master's slaves to match the master | |
3386 | */ | |
3387 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3388 | { | |
454d7c9b | 3389 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3390 | struct slave *slave, *rollback_slave; |
9caff1e7 | 3391 | struct list_head *iter; |
1da177e4 | 3392 | int res = 0; |
1da177e4 | 3393 | |
5a03cdb7 | 3394 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3395 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3396 | |
3397 | /* Can't hold bond->lock with bh disabled here since | |
3398 | * some base drivers panic. On the other hand we can't | |
3399 | * hold bond->lock without bh disabled because we'll | |
3400 | * deadlock. The only solution is to rely on the fact | |
3401 | * that we're under rtnl_lock here, and the slaves | |
3402 | * list won't change. This doesn't solve the problem | |
3403 | * of setting the slave's MTU while it is | |
3404 | * transmitting, but the assumption is that the base | |
3405 | * driver can handle that. | |
3406 | * | |
3407 | * TODO: figure out a way to safely iterate the slaves | |
3408 | * list, but without holding a lock around the actual | |
3409 | * call to the base driver. | |
3410 | */ | |
3411 | ||
9caff1e7 | 3412 | bond_for_each_slave(bond, slave, iter) { |
c8c23903 | 3413 | pr_debug("s %p c_m %p\n", |
a4aee5c8 | 3414 | slave, |
a4aee5c8 | 3415 | slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 3416 | |
1da177e4 LT |
3417 | res = dev_set_mtu(slave->dev, new_mtu); |
3418 | ||
3419 | if (res) { | |
3420 | /* If we failed to set the slave's mtu to the new value | |
3421 | * we must abort the operation even in ACTIVE_BACKUP | |
3422 | * mode, because if we allow the backup slaves to have | |
3423 | * different mtu values than the active slave we'll | |
3424 | * need to change their mtu when doing a failover. That | |
3425 | * means changing their mtu from timer context, which | |
3426 | * is probably not a good idea. | |
3427 | */ | |
5a03cdb7 | 3428 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3429 | goto unwind; |
3430 | } | |
3431 | } | |
3432 | ||
3433 | bond_dev->mtu = new_mtu; | |
3434 | ||
3435 | return 0; | |
3436 | ||
3437 | unwind: | |
3438 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3439 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3440 | int tmp_res; |
3441 | ||
81f23b13 VF |
3442 | if (rollback_slave == slave) |
3443 | break; | |
3444 | ||
3445 | tmp_res = dev_set_mtu(rollback_slave->dev, bond_dev->mtu); | |
1da177e4 | 3446 | if (tmp_res) { |
a4aee5c8 | 3447 | pr_debug("unwind err %d dev %s\n", |
81f23b13 | 3448 | tmp_res, rollback_slave->dev->name); |
1da177e4 LT |
3449 | } |
3450 | } | |
3451 | ||
3452 | return res; | |
3453 | } | |
3454 | ||
3455 | /* | |
3456 | * Change HW address | |
3457 | * | |
3458 | * Note that many devices must be down to change the HW address, and | |
3459 | * downing the master releases all slaves. We can make bonds full of | |
3460 | * bonding devices to test this, however. | |
3461 | */ | |
3462 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3463 | { | |
454d7c9b | 3464 | struct bonding *bond = netdev_priv(bond_dev); |
81f23b13 | 3465 | struct slave *slave, *rollback_slave; |
1da177e4 | 3466 | struct sockaddr *sa = addr, tmp_sa; |
9caff1e7 | 3467 | struct list_head *iter; |
1da177e4 | 3468 | int res = 0; |
1da177e4 | 3469 | |
eb7cc59a SH |
3470 | if (bond->params.mode == BOND_MODE_ALB) |
3471 | return bond_alb_set_mac_address(bond_dev, addr); | |
3472 | ||
3473 | ||
a4aee5c8 JP |
3474 | pr_debug("bond=%p, name=%s\n", |
3475 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3476 | |
1b5acd29 JV |
3477 | /* If fail_over_mac is enabled, do nothing and return success. |
3478 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3479 | */ |
1b5acd29 | 3480 | if (bond->params.fail_over_mac) |
dd957c57 | 3481 | return 0; |
2ab82852 | 3482 | |
3d632c3f | 3483 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3484 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3485 | |
3486 | /* Can't hold bond->lock with bh disabled here since | |
3487 | * some base drivers panic. On the other hand we can't | |
3488 | * hold bond->lock without bh disabled because we'll | |
3489 | * deadlock. The only solution is to rely on the fact | |
3490 | * that we're under rtnl_lock here, and the slaves | |
3491 | * list won't change. This doesn't solve the problem | |
3492 | * of setting the slave's hw address while it is | |
3493 | * transmitting, but the assumption is that the base | |
3494 | * driver can handle that. | |
3495 | * | |
3496 | * TODO: figure out a way to safely iterate the slaves | |
3497 | * list, but without holding a lock around the actual | |
3498 | * call to the base driver. | |
3499 | */ | |
3500 | ||
9caff1e7 | 3501 | bond_for_each_slave(bond, slave, iter) { |
eb7cc59a | 3502 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3503 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3504 | |
eb7cc59a | 3505 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3506 | res = -EOPNOTSUPP; |
5a03cdb7 | 3507 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3508 | goto unwind; |
3509 | } | |
3510 | ||
3511 | res = dev_set_mac_address(slave->dev, addr); | |
3512 | if (res) { | |
3513 | /* TODO: consider downing the slave | |
3514 | * and retry ? | |
3515 | * User should expect communications | |
3516 | * breakage anyway until ARP finish | |
3517 | * updating, so... | |
3518 | */ | |
5a03cdb7 | 3519 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3520 | goto unwind; |
3521 | } | |
3522 | } | |
3523 | ||
3524 | /* success */ | |
3525 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3526 | return 0; | |
3527 | ||
3528 | unwind: | |
3529 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3530 | tmp_sa.sa_family = bond_dev->type; | |
3531 | ||
3532 | /* unwind from head to the slave that failed */ | |
9caff1e7 | 3533 | bond_for_each_slave(bond, rollback_slave, iter) { |
1da177e4 LT |
3534 | int tmp_res; |
3535 | ||
81f23b13 VF |
3536 | if (rollback_slave == slave) |
3537 | break; | |
3538 | ||
3539 | tmp_res = dev_set_mac_address(rollback_slave->dev, &tmp_sa); | |
1da177e4 | 3540 | if (tmp_res) { |
a4aee5c8 | 3541 | pr_debug("unwind err %d dev %s\n", |
81f23b13 | 3542 | tmp_res, rollback_slave->dev->name); |
1da177e4 LT |
3543 | } |
3544 | } | |
3545 | ||
3546 | return res; | |
3547 | } | |
3548 | ||
15077228 NA |
3549 | /** |
3550 | * bond_xmit_slave_id - transmit skb through slave with slave_id | |
3551 | * @bond: bonding device that is transmitting | |
3552 | * @skb: buffer to transmit | |
3553 | * @slave_id: slave id up to slave_cnt-1 through which to transmit | |
3554 | * | |
3555 | * This function tries to transmit through slave with slave_id but in case | |
3556 | * it fails, it tries to find the first available slave for transmission. | |
3557 | * The skb is consumed in all cases, thus the function is void. | |
3558 | */ | |
da131ddb | 3559 | static void bond_xmit_slave_id(struct bonding *bond, struct sk_buff *skb, int slave_id) |
15077228 | 3560 | { |
9caff1e7 | 3561 | struct list_head *iter; |
15077228 NA |
3562 | struct slave *slave; |
3563 | int i = slave_id; | |
3564 | ||
3565 | /* Here we start from the slave with slave_id */ | |
9caff1e7 | 3566 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3567 | if (--i < 0) { |
3568 | if (slave_can_tx(slave)) { | |
3569 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3570 | return; | |
3571 | } | |
3572 | } | |
3573 | } | |
3574 | ||
3575 | /* Here we start from the first slave up to slave_id */ | |
3576 | i = slave_id; | |
9caff1e7 | 3577 | bond_for_each_slave_rcu(bond, slave, iter) { |
15077228 NA |
3578 | if (--i < 0) |
3579 | break; | |
3580 | if (slave_can_tx(slave)) { | |
3581 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3582 | return; | |
3583 | } | |
3584 | } | |
3585 | /* no slave that can tx has been found */ | |
3586 | kfree_skb(skb); | |
3587 | } | |
3588 | ||
73958329 NA |
3589 | /** |
3590 | * bond_rr_gen_slave_id - generate slave id based on packets_per_slave | |
3591 | * @bond: bonding device to use | |
3592 | * | |
3593 | * Based on the value of the bonding device's packets_per_slave parameter | |
3594 | * this function generates a slave id, which is usually used as the next | |
3595 | * slave to transmit through. | |
3596 | */ | |
3597 | static u32 bond_rr_gen_slave_id(struct bonding *bond) | |
3598 | { | |
3599 | int packets_per_slave = bond->params.packets_per_slave; | |
3600 | u32 slave_id; | |
3601 | ||
3602 | switch (packets_per_slave) { | |
3603 | case 0: | |
3604 | slave_id = prandom_u32(); | |
3605 | break; | |
3606 | case 1: | |
3607 | slave_id = bond->rr_tx_counter; | |
3608 | break; | |
3609 | default: | |
3610 | slave_id = reciprocal_divide(bond->rr_tx_counter, | |
3611 | packets_per_slave); | |
3612 | break; | |
3613 | } | |
3614 | bond->rr_tx_counter++; | |
3615 | ||
3616 | return slave_id; | |
3617 | } | |
3618 | ||
1da177e4 LT |
3619 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) |
3620 | { | |
454d7c9b | 3621 | struct bonding *bond = netdev_priv(bond_dev); |
a2fd940f | 3622 | struct iphdr *iph = ip_hdr(skb); |
15077228 | 3623 | struct slave *slave; |
73958329 | 3624 | u32 slave_id; |
1da177e4 | 3625 | |
73958329 | 3626 | /* Start with the curr_active_slave that joined the bond as the |
a2fd940f AG |
3627 | * default for sending IGMP traffic. For failover purposes one |
3628 | * needs to maintain some consistency for the interface that will | |
3629 | * send the join/membership reports. The curr_active_slave found | |
3630 | * will send all of this type of traffic. | |
cf5f9044 | 3631 | */ |
15077228 | 3632 | if (iph->protocol == IPPROTO_IGMP && skb->protocol == htons(ETH_P_IP)) { |
278b2083 | 3633 | slave = rcu_dereference(bond->curr_active_slave); |
15077228 NA |
3634 | if (slave && slave_can_tx(slave)) |
3635 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3636 | else | |
3637 | bond_xmit_slave_id(bond, skb, 0); | |
a2fd940f | 3638 | } else { |
73958329 NA |
3639 | slave_id = bond_rr_gen_slave_id(bond); |
3640 | bond_xmit_slave_id(bond, skb, slave_id % bond->slave_cnt); | |
1da177e4 | 3641 | } |
0693e88e | 3642 | |
ec634fe3 | 3643 | return NETDEV_TX_OK; |
1da177e4 LT |
3644 | } |
3645 | ||
3646 | /* | |
3647 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3648 | * the bond has a usable interface. | |
3649 | */ | |
3650 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3651 | { | |
454d7c9b | 3652 | struct bonding *bond = netdev_priv(bond_dev); |
71bc3b2d | 3653 | struct slave *slave; |
1da177e4 | 3654 | |
278b2083 | 3655 | slave = rcu_dereference(bond->curr_active_slave); |
71bc3b2d | 3656 | if (slave) |
15077228 NA |
3657 | bond_dev_queue_xmit(bond, skb, slave->dev); |
3658 | else | |
04502430 | 3659 | kfree_skb(skb); |
0693e88e | 3660 | |
ec634fe3 | 3661 | return NETDEV_TX_OK; |
1da177e4 LT |
3662 | } |
3663 | ||
32819dc1 | 3664 | /* In bond_xmit_xor() , we determine the output device by using a pre- |
169a3e66 JV |
3665 | * determined xmit_hash_policy(), If the selected device is not enabled, |
3666 | * find the next active slave. | |
1da177e4 LT |
3667 | */ |
3668 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3669 | { | |
454d7c9b | 3670 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3671 | |
32819dc1 | 3672 | bond_xmit_slave_id(bond, skb, bond_xmit_hash(bond, skb, bond->slave_cnt)); |
0693e88e | 3673 | |
ec634fe3 | 3674 | return NETDEV_TX_OK; |
1da177e4 LT |
3675 | } |
3676 | ||
78a646ce | 3677 | /* in broadcast mode, we send everything to all usable interfaces. */ |
1da177e4 LT |
3678 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) |
3679 | { | |
454d7c9b | 3680 | struct bonding *bond = netdev_priv(bond_dev); |
78a646ce | 3681 | struct slave *slave = NULL; |
9caff1e7 | 3682 | struct list_head *iter; |
1da177e4 | 3683 | |
9caff1e7 | 3684 | bond_for_each_slave_rcu(bond, slave, iter) { |
78a646ce NA |
3685 | if (bond_is_last_slave(bond, slave)) |
3686 | break; | |
3687 | if (IS_UP(slave->dev) && slave->link == BOND_LINK_UP) { | |
3688 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
1da177e4 | 3689 | |
78a646ce NA |
3690 | if (!skb2) { |
3691 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", | |
3692 | bond_dev->name); | |
3693 | continue; | |
1da177e4 | 3694 | } |
78a646ce NA |
3695 | /* bond_dev_queue_xmit always returns 0 */ |
3696 | bond_dev_queue_xmit(bond, skb2, slave->dev); | |
1da177e4 LT |
3697 | } |
3698 | } | |
78a646ce NA |
3699 | if (slave && IS_UP(slave->dev) && slave->link == BOND_LINK_UP) |
3700 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3701 | else | |
04502430 | 3702 | kfree_skb(skb); |
3d632c3f | 3703 | |
ec634fe3 | 3704 | return NETDEV_TX_OK; |
1da177e4 LT |
3705 | } |
3706 | ||
3707 | /*------------------------- Device initialization ---------------------------*/ | |
3708 | ||
bb1d9123 AG |
3709 | /* |
3710 | * Lookup the slave that corresponds to a qid | |
3711 | */ | |
3712 | static inline int bond_slave_override(struct bonding *bond, | |
3713 | struct sk_buff *skb) | |
3714 | { | |
bb1d9123 | 3715 | struct slave *slave = NULL; |
9caff1e7 | 3716 | struct list_head *iter; |
bb1d9123 | 3717 | |
0693e88e MM |
3718 | if (!skb->queue_mapping) |
3719 | return 1; | |
bb1d9123 AG |
3720 | |
3721 | /* Find out if any slaves have the same mapping as this skb. */ | |
3900f290 | 3722 | bond_for_each_slave_rcu(bond, slave, iter) { |
3723 | if (slave->queue_id == skb->queue_mapping) { | |
3724 | if (slave_can_tx(slave)) { | |
3725 | bond_dev_queue_xmit(bond, skb, slave->dev); | |
3726 | return 0; | |
3727 | } | |
3728 | /* If the slave isn't UP, use default transmit policy. */ | |
bb1d9123 AG |
3729 | break; |
3730 | } | |
3731 | } | |
3732 | ||
3900f290 | 3733 | return 1; |
bb1d9123 AG |
3734 | } |
3735 | ||
374eeb5a | 3736 | |
f663dd9a JW |
3737 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb, |
3738 | void *accel_priv) | |
bb1d9123 AG |
3739 | { |
3740 | /* | |
3741 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 3742 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
3743 | * skb_tx_hash and will put the skbs in the queue we expect on their |
3744 | * way down to the bonding driver. | |
3745 | */ | |
fd0e435b PO |
3746 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
3747 | ||
374eeb5a NH |
3748 | /* |
3749 | * Save the original txq to restore before passing to the driver | |
3750 | */ | |
df4ab5b3 | 3751 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 3752 | |
fd0e435b | 3753 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 3754 | do { |
fd0e435b | 3755 | txq -= dev->real_num_tx_queues; |
d30ee670 | 3756 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
3757 | } |
3758 | return txq; | |
bb1d9123 AG |
3759 | } |
3760 | ||
0693e88e | 3761 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 3762 | { |
bb1d9123 AG |
3763 | struct bonding *bond = netdev_priv(dev); |
3764 | ||
3765 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
3766 | if (!bond_slave_override(bond, skb)) | |
3767 | return NETDEV_TX_OK; | |
3768 | } | |
00829823 SH |
3769 | |
3770 | switch (bond->params.mode) { | |
3771 | case BOND_MODE_ROUNDROBIN: | |
3772 | return bond_xmit_roundrobin(skb, dev); | |
3773 | case BOND_MODE_ACTIVEBACKUP: | |
3774 | return bond_xmit_activebackup(skb, dev); | |
3775 | case BOND_MODE_XOR: | |
3776 | return bond_xmit_xor(skb, dev); | |
3777 | case BOND_MODE_BROADCAST: | |
3778 | return bond_xmit_broadcast(skb, dev); | |
3779 | case BOND_MODE_8023AD: | |
3780 | return bond_3ad_xmit_xor(skb, dev); | |
3781 | case BOND_MODE_ALB: | |
3782 | case BOND_MODE_TLB: | |
3783 | return bond_alb_xmit(skb, dev); | |
3784 | default: | |
3785 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
3786 | pr_err("%s: Error: Unknown bonding mode %d\n", |
3787 | dev->name, bond->params.mode); | |
00829823 | 3788 | WARN_ON_ONCE(1); |
04502430 | 3789 | kfree_skb(skb); |
00829823 SH |
3790 | return NETDEV_TX_OK; |
3791 | } | |
3792 | } | |
3793 | ||
0693e88e MM |
3794 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
3795 | { | |
3796 | struct bonding *bond = netdev_priv(dev); | |
3797 | netdev_tx_t ret = NETDEV_TX_OK; | |
3798 | ||
3799 | /* | |
3800 | * If we risk deadlock from transmitting this in the | |
3801 | * netpoll path, tell netpoll to queue the frame for later tx | |
3802 | */ | |
3803 | if (is_netpoll_tx_blocked(dev)) | |
3804 | return NETDEV_TX_BUSY; | |
3805 | ||
278b2083 | 3806 | rcu_read_lock(); |
0965a1f3 | 3807 | if (bond_has_slaves(bond)) |
0693e88e MM |
3808 | ret = __bond_start_xmit(skb, dev); |
3809 | else | |
04502430 | 3810 | kfree_skb(skb); |
278b2083 | 3811 | rcu_read_unlock(); |
0693e88e MM |
3812 | |
3813 | return ret; | |
3814 | } | |
00829823 | 3815 | |
bb5b052f AG |
3816 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
3817 | struct ethtool_cmd *ecmd) | |
3818 | { | |
3819 | struct bonding *bond = netdev_priv(bond_dev); | |
bb5b052f | 3820 | unsigned long speed = 0; |
9caff1e7 | 3821 | struct list_head *iter; |
dec1e90e | 3822 | struct slave *slave; |
bb5b052f AG |
3823 | |
3824 | ecmd->duplex = DUPLEX_UNKNOWN; | |
3825 | ecmd->port = PORT_OTHER; | |
3826 | ||
3827 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
3828 | * do not need to check mode. Though link speed might not represent | |
3829 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
3830 | * this is an accurate maximum. | |
3831 | */ | |
3832 | read_lock(&bond->lock); | |
9caff1e7 | 3833 | bond_for_each_slave(bond, slave, iter) { |
bb5b052f AG |
3834 | if (SLAVE_IS_OK(slave)) { |
3835 | if (slave->speed != SPEED_UNKNOWN) | |
3836 | speed += slave->speed; | |
3837 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
3838 | slave->duplex != DUPLEX_UNKNOWN) | |
3839 | ecmd->duplex = slave->duplex; | |
3840 | } | |
3841 | } | |
3842 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
3843 | read_unlock(&bond->lock); | |
dec1e90e | 3844 | |
bb5b052f AG |
3845 | return 0; |
3846 | } | |
3847 | ||
217df670 | 3848 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 3849 | struct ethtool_drvinfo *drvinfo) |
217df670 | 3850 | { |
7826d43f JP |
3851 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
3852 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
3853 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
3854 | BOND_ABI_VERSION); | |
217df670 JV |
3855 | } |
3856 | ||
7282d491 | 3857 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 3858 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 3859 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 3860 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
3861 | }; |
3862 | ||
eb7cc59a | 3863 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 3864 | .ndo_init = bond_init, |
9e71626c | 3865 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
3866 | .ndo_open = bond_open, |
3867 | .ndo_stop = bond_close, | |
00829823 | 3868 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 3869 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 3870 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 3871 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 3872 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 3873 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 3874 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 3875 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 3876 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 3877 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 3878 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 3879 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 3880 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
3881 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
3882 | .ndo_poll_controller = bond_poll_controller, | |
3883 | #endif | |
9232ecca JP |
3884 | .ndo_add_slave = bond_enslave, |
3885 | .ndo_del_slave = bond_release, | |
1d3ee88a | 3886 | .ndo_get_slave = bond_get_slave, |
b2a103e6 | 3887 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
3888 | }; |
3889 | ||
b3f92b63 DG |
3890 | static const struct device_type bond_type = { |
3891 | .name = "bond", | |
3892 | }; | |
3893 | ||
9e2e61fb AW |
3894 | static void bond_destructor(struct net_device *bond_dev) |
3895 | { | |
3896 | struct bonding *bond = netdev_priv(bond_dev); | |
3897 | if (bond->wq) | |
3898 | destroy_workqueue(bond->wq); | |
3899 | free_netdev(bond_dev); | |
3900 | } | |
3901 | ||
0a2a78c4 | 3902 | void bond_setup(struct net_device *bond_dev) |
1da177e4 | 3903 | { |
454d7c9b | 3904 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3905 | |
1da177e4 LT |
3906 | /* initialize rwlocks */ |
3907 | rwlock_init(&bond->lock); | |
3908 | rwlock_init(&bond->curr_slave_lock); | |
d2991f75 | 3909 | bond->params = bonding_defaults; |
1da177e4 LT |
3910 | |
3911 | /* Initialize pointers */ | |
1da177e4 | 3912 | bond->dev = bond_dev; |
1da177e4 LT |
3913 | |
3914 | /* Initialize the device entry points */ | |
181470fc | 3915 | ether_setup(bond_dev); |
eb7cc59a | 3916 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 3917 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
1da177e4 | 3918 | |
9e2e61fb | 3919 | bond_dev->destructor = bond_destructor; |
1da177e4 | 3920 | |
b3f92b63 DG |
3921 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
3922 | ||
1da177e4 LT |
3923 | /* Initialize the device options */ |
3924 | bond_dev->tx_queue_len = 0; | |
3925 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 3926 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 3927 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 3928 | |
1da177e4 LT |
3929 | /* At first, we block adding VLANs. That's the only way to |
3930 | * prevent problems that occur when adding VLANs over an | |
3931 | * empty bond. The block will be removed once non-challenged | |
3932 | * slaves are enslaved. | |
3933 | */ | |
3934 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
3935 | ||
932ff279 | 3936 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
3937 | * transmitting */ |
3938 | bond_dev->features |= NETIF_F_LLTX; | |
3939 | ||
3940 | /* By default, we declare the bond to be fully | |
3941 | * VLAN hardware accelerated capable. Special | |
3942 | * care is taken in the various xmit functions | |
3943 | * when there are slaves that are not hw accel | |
3944 | * capable | |
3945 | */ | |
1da177e4 | 3946 | |
b2a103e6 | 3947 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
3948 | NETIF_F_HW_VLAN_CTAG_TX | |
3949 | NETIF_F_HW_VLAN_CTAG_RX | | |
3950 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 3951 | |
34324dc2 | 3952 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 3953 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
3954 | } |
3955 | ||
c67dfb29 EB |
3956 | /* |
3957 | * Destroy a bonding device. | |
3958 | * Must be under rtnl_lock when this function is called. | |
3959 | */ | |
3960 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 3961 | { |
454d7c9b | 3962 | struct bonding *bond = netdev_priv(bond_dev); |
544a028e VF |
3963 | struct list_head *iter; |
3964 | struct slave *slave; | |
a434e43f | 3965 | |
f6dc31a8 WC |
3966 | bond_netpoll_cleanup(bond_dev); |
3967 | ||
c67dfb29 | 3968 | /* Release the bonded slaves */ |
544a028e | 3969 | bond_for_each_slave(bond, slave, iter) |
dec1e90e | 3970 | __bond_release_one(bond_dev, slave->dev, true); |
0896341a | 3971 | pr_info("%s: released all slaves\n", bond_dev->name); |
c67dfb29 | 3972 | |
a434e43f JV |
3973 | list_del(&bond->bond_list); |
3974 | ||
f073c7ca | 3975 | bond_debug_unregister(bond); |
a434e43f JV |
3976 | } |
3977 | ||
1da177e4 LT |
3978 | /*------------------------- Module initialization ---------------------------*/ |
3979 | ||
ec029fac | 3980 | int bond_parm_tbl_lookup(int mode, const struct bond_parm_tbl *tbl) |
3981 | { | |
3982 | int i; | |
3983 | ||
3984 | for (i = 0; tbl[i].modename; i++) | |
3985 | if (mode == tbl[i].mode) | |
3986 | return tbl[i].mode; | |
3987 | ||
3988 | return -1; | |
3989 | } | |
3990 | ||
3991 | static int bond_parm_tbl_lookup_name(const char *modename, | |
3992 | const struct bond_parm_tbl *tbl) | |
3993 | { | |
3994 | int i; | |
3995 | ||
3996 | for (i = 0; tbl[i].modename; i++) | |
3997 | if (strcmp(modename, tbl[i].modename) == 0) | |
3998 | return tbl[i].mode; | |
3999 | ||
4000 | return -1; | |
4001 | } | |
4002 | ||
1da177e4 LT |
4003 | /* |
4004 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4005 | * number of the mode or its string name. A bit complicated because |
4006 | * some mode names are substrings of other names, and calls from sysfs | |
4007 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4008 | */ |
325dcf7a | 4009 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4010 | { |
ec029fac | 4011 | int modeint; |
4012 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1]; | |
ece95f7f | 4013 | |
a42e534f JV |
4014 | for (p = (char *)buf; *p; p++) |
4015 | if (!(isdigit(*p) || isspace(*p))) | |
4016 | break; | |
4017 | ||
ec029fac | 4018 | if (*p && sscanf(buf, "%20s", modestr) != 0) |
4019 | return bond_parm_tbl_lookup_name(modestr, tbl); | |
4020 | else if (sscanf(buf, "%d", &modeint) != 0) | |
4021 | return bond_parm_tbl_lookup(modeint, tbl); | |
1da177e4 LT |
4022 | |
4023 | return -1; | |
4024 | } | |
4025 | ||
4026 | static int bond_check_params(struct bond_params *params) | |
4027 | { | |
5a5c5fd4 | 4028 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
8599b52e | 4029 | int arp_all_targets_value; |
f5b2b966 | 4030 | |
1da177e4 LT |
4031 | /* |
4032 | * Convert string parameters. | |
4033 | */ | |
4034 | if (mode) { | |
4035 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4036 | if (bond_mode == -1) { | |
a4aee5c8 | 4037 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4038 | mode == NULL ? "NULL" : mode); |
4039 | return -EINVAL; | |
4040 | } | |
4041 | } | |
4042 | ||
169a3e66 JV |
4043 | if (xmit_hash_policy) { |
4044 | if ((bond_mode != BOND_MODE_XOR) && | |
4045 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4046 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4047 | bond_mode_name(bond_mode)); |
4048 | } else { | |
4049 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4050 | xmit_hashtype_tbl); | |
4051 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4052 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4053 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4054 | xmit_hash_policy); |
4055 | return -EINVAL; | |
4056 | } | |
4057 | } | |
4058 | } | |
4059 | ||
1da177e4 LT |
4060 | if (lacp_rate) { |
4061 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4062 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4063 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4064 | } else { |
4065 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4066 | if (lacp_fast == -1) { | |
a4aee5c8 | 4067 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4068 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4069 | return -EINVAL; | |
4070 | } | |
4071 | } | |
4072 | } | |
4073 | ||
fd989c83 JV |
4074 | if (ad_select) { |
4075 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4076 | if (params->ad_select == -1) { | |
a4aee5c8 | 4077 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4078 | ad_select == NULL ? "NULL" : ad_select); |
4079 | return -EINVAL; | |
4080 | } | |
4081 | ||
4082 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4083 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4084 | } |
4085 | } else { | |
4086 | params->ad_select = BOND_AD_STABLE; | |
4087 | } | |
4088 | ||
f5841306 | 4089 | if (max_bonds < 0) { |
a4aee5c8 JP |
4090 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4091 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4092 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4093 | } | |
4094 | ||
4095 | if (miimon < 0) { | |
a4aee5c8 JP |
4096 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4097 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4098 | miimon = BOND_LINK_MON_INTERV; |
4099 | } | |
4100 | ||
4101 | if (updelay < 0) { | |
a4aee5c8 JP |
4102 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4103 | updelay, INT_MAX); | |
1da177e4 LT |
4104 | updelay = 0; |
4105 | } | |
4106 | ||
4107 | if (downdelay < 0) { | |
a4aee5c8 JP |
4108 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4109 | downdelay, INT_MAX); | |
1da177e4 LT |
4110 | downdelay = 0; |
4111 | } | |
4112 | ||
4113 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4114 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4115 | use_carrier); | |
1da177e4 LT |
4116 | use_carrier = 1; |
4117 | } | |
4118 | ||
ad246c99 BH |
4119 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4120 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4121 | num_peer_notif); | |
4122 | num_peer_notif = 1; | |
4123 | } | |
4124 | ||
834db4bc | 4125 | /* reset values for 802.3ad/TLB/ALB */ |
4126 | if (BOND_NO_USES_ARP(bond_mode)) { | |
1da177e4 | 4127 | if (!miimon) { |
a4aee5c8 | 4128 | pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n"); |
3d632c3f | 4129 | pr_warning("Forcing miimon to 100msec\n"); |
fe9d04af | 4130 | miimon = BOND_DEFAULT_MIIMON; |
1da177e4 LT |
4131 | } |
4132 | } | |
4133 | ||
bb1d9123 AG |
4134 | if (tx_queues < 1 || tx_queues > 255) { |
4135 | pr_warning("Warning: tx_queues (%d) should be between " | |
4136 | "1 and 255, resetting to %d\n", | |
4137 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4138 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4139 | } | |
4140 | ||
ebd8e497 AG |
4141 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4142 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4143 | "not of valid value (0/1), so it was set to " | |
4144 | "0\n", all_slaves_active); | |
4145 | all_slaves_active = 0; | |
4146 | } | |
4147 | ||
c2952c31 FL |
4148 | if (resend_igmp < 0 || resend_igmp > 255) { |
4149 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4150 | "0 and 255, resetting to %d\n", | |
4151 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4152 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4153 | } | |
4154 | ||
73958329 NA |
4155 | if (packets_per_slave < 0 || packets_per_slave > USHRT_MAX) { |
4156 | pr_warn("Warning: packets_per_slave (%d) should be between 0 and %u resetting to 1\n", | |
4157 | packets_per_slave, USHRT_MAX); | |
4158 | packets_per_slave = 1; | |
4159 | } | |
4160 | ||
1da177e4 | 4161 | if (bond_mode == BOND_MODE_ALB) { |
a4aee5c8 JP |
4162 | 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", |
4163 | updelay); | |
1da177e4 LT |
4164 | } |
4165 | ||
4166 | if (!miimon) { | |
4167 | if (updelay || downdelay) { | |
4168 | /* just warn the user the up/down delay will have | |
4169 | * no effect since miimon is zero... | |
4170 | */ | |
a4aee5c8 JP |
4171 | pr_warning("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", |
4172 | updelay, downdelay); | |
1da177e4 LT |
4173 | } |
4174 | } else { | |
4175 | /* don't allow arp monitoring */ | |
4176 | if (arp_interval) { | |
a4aee5c8 JP |
4177 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4178 | miimon, arp_interval); | |
1da177e4 LT |
4179 | arp_interval = 0; |
4180 | } | |
4181 | ||
4182 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4183 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4184 | updelay, miimon, | |
4185 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4186 | } |
4187 | ||
4188 | updelay /= miimon; | |
4189 | ||
4190 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4191 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4192 | downdelay, miimon, | |
4193 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4194 | } |
4195 | ||
4196 | downdelay /= miimon; | |
4197 | } | |
4198 | ||
4199 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4200 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4201 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4202 | arp_interval = BOND_LINK_ARP_INTERV; |
4203 | } | |
4204 | ||
5a5c5fd4 | 4205 | for (arp_ip_count = 0, i = 0; |
4206 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4207 | /* not complete check, but should be good enough to |
4208 | catch mistakes */ | |
89015c18 | 4209 | __be32 ip; |
4210 | if (!in4_pton(arp_ip_target[i], -1, (u8 *)&ip, -1, NULL) || | |
4211 | IS_IP_TARGET_UNUSABLE_ADDRESS(ip)) { | |
a4aee5c8 | 4212 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5a5c5fd4 | 4213 | arp_ip_target[i]); |
1da177e4 LT |
4214 | arp_interval = 0; |
4215 | } else { | |
0afee4e8 VF |
4216 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4217 | arp_target[arp_ip_count++] = ip; | |
4218 | else | |
4219 | pr_warning("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", | |
4220 | &ip); | |
1da177e4 LT |
4221 | } |
4222 | } | |
4223 | ||
4224 | if (arp_interval && !arp_ip_count) { | |
4225 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4226 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4227 | arp_interval); | |
1da177e4 LT |
4228 | arp_interval = 0; |
4229 | } | |
4230 | ||
f5b2b966 JV |
4231 | if (arp_validate) { |
4232 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4233 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4234 | return -EINVAL; |
4235 | } | |
4236 | if (!arp_interval) { | |
a4aee5c8 | 4237 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4238 | return -EINVAL; |
4239 | } | |
4240 | ||
4241 | arp_validate_value = bond_parse_parm(arp_validate, | |
4242 | arp_validate_tbl); | |
4243 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4244 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4245 | arp_validate == NULL ? "NULL" : arp_validate); |
4246 | return -EINVAL; | |
4247 | } | |
4248 | } else | |
4249 | arp_validate_value = 0; | |
4250 | ||
8599b52e VF |
4251 | arp_all_targets_value = 0; |
4252 | if (arp_all_targets) { | |
4253 | arp_all_targets_value = bond_parse_parm(arp_all_targets, | |
4254 | arp_all_targets_tbl); | |
4255 | ||
4256 | if (arp_all_targets_value == -1) { | |
4257 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", | |
4258 | arp_all_targets); | |
4259 | arp_all_targets_value = 0; | |
4260 | } | |
4261 | } | |
4262 | ||
1da177e4 | 4263 | if (miimon) { |
a4aee5c8 | 4264 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4265 | } else if (arp_interval) { |
a4aee5c8 JP |
4266 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4267 | arp_interval, | |
4268 | arp_validate_tbl[arp_validate_value].modename, | |
4269 | arp_ip_count); | |
1da177e4 LT |
4270 | |
4271 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4272 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4273 | |
e5e2a8fd | 4274 | pr_info("\n"); |
1da177e4 | 4275 | |
b8a9787e | 4276 | } else if (max_bonds) { |
1da177e4 LT |
4277 | /* miimon and arp_interval not set, we need one so things |
4278 | * work as expected, see bonding.txt for details | |
4279 | */ | |
b2730f4f | 4280 | 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 |
4281 | } |
4282 | ||
4283 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4284 | /* currently, using a primary only makes sense | |
4285 | * in active backup, TLB or ALB modes | |
4286 | */ | |
a4aee5c8 JP |
4287 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4288 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4289 | primary = NULL; |
4290 | } | |
4291 | ||
a549952a JP |
4292 | if (primary && primary_reselect) { |
4293 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4294 | pri_reselect_tbl); | |
4295 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4296 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4297 | primary_reselect == |
4298 | NULL ? "NULL" : primary_reselect); | |
4299 | return -EINVAL; | |
4300 | } | |
4301 | } else { | |
4302 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4303 | } | |
4304 | ||
3915c1e8 JV |
4305 | if (fail_over_mac) { |
4306 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4307 | fail_over_mac_tbl); | |
4308 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4309 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4310 | arp_validate == NULL ? "NULL" : arp_validate); |
4311 | return -EINVAL; | |
4312 | } | |
4313 | ||
4314 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4315 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4316 | } else { |
4317 | fail_over_mac_value = BOND_FOM_NONE; | |
4318 | } | |
dd957c57 | 4319 | |
3a7129e5 | 4320 | if (lp_interval == 0) { |
4321 | pr_warning("Warning: ip_interval must be between 1 and %d, so it was reset to %d\n", | |
4322 | INT_MAX, BOND_ALB_DEFAULT_LP_INTERVAL); | |
4323 | lp_interval = BOND_ALB_DEFAULT_LP_INTERVAL; | |
4324 | } | |
4325 | ||
1da177e4 LT |
4326 | /* fill params struct with the proper values */ |
4327 | params->mode = bond_mode; | |
169a3e66 | 4328 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4329 | params->miimon = miimon; |
ad246c99 | 4330 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4331 | params->arp_interval = arp_interval; |
f5b2b966 | 4332 | params->arp_validate = arp_validate_value; |
8599b52e | 4333 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4334 | params->updelay = updelay; |
4335 | params->downdelay = downdelay; | |
4336 | params->use_carrier = use_carrier; | |
4337 | params->lacp_fast = lacp_fast; | |
4338 | params->primary[0] = 0; | |
a549952a | 4339 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4340 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4341 | params->tx_queues = tx_queues; |
ebd8e497 | 4342 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4343 | params->resend_igmp = resend_igmp; |
655f8919 | 4344 | params->min_links = min_links; |
3a7129e5 | 4345 | params->lp_interval = lp_interval; |
73958329 NA |
4346 | if (packets_per_slave > 1) |
4347 | params->packets_per_slave = reciprocal_value(packets_per_slave); | |
4348 | else | |
4349 | params->packets_per_slave = packets_per_slave; | |
1da177e4 LT |
4350 | if (primary) { |
4351 | strncpy(params->primary, primary, IFNAMSIZ); | |
4352 | params->primary[IFNAMSIZ - 1] = 0; | |
4353 | } | |
4354 | ||
4355 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4356 | ||
4357 | return 0; | |
4358 | } | |
4359 | ||
0daa2303 | 4360 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4361 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4362 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4363 | |
e8a0464c DM |
4364 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4365 | struct netdev_queue *txq, | |
4366 | void *_unused) | |
c773e847 DM |
4367 | { |
4368 | lockdep_set_class(&txq->_xmit_lock, | |
4369 | &bonding_netdev_xmit_lock_key); | |
4370 | } | |
4371 | ||
4372 | static void bond_set_lockdep_class(struct net_device *dev) | |
4373 | { | |
cf508b12 DM |
4374 | lockdep_set_class(&dev->addr_list_lock, |
4375 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4376 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4377 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4378 | } |
4379 | ||
181470fc SH |
4380 | /* |
4381 | * Called from registration process | |
4382 | */ | |
4383 | static int bond_init(struct net_device *bond_dev) | |
4384 | { | |
4385 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4386 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4387 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4388 | |
4389 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4390 | ||
9fe0617d NH |
4391 | /* |
4392 | * Initialize locks that may be required during | |
4393 | * en/deslave operations. All of the bond_open work | |
4394 | * (of which this is part) should really be moved to | |
4395 | * a phase prior to dev_open | |
4396 | */ | |
4397 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4398 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4399 | ||
181470fc SH |
4400 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4401 | if (!bond->wq) | |
4402 | return -ENOMEM; | |
4403 | ||
4404 | bond_set_lockdep_class(bond_dev); | |
4405 | ||
ec87fd3b | 4406 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4407 | |
6151b3d4 | 4408 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4409 | |
f073c7ca TI |
4410 | bond_debug_register(bond); |
4411 | ||
409cc1f8 JP |
4412 | /* Ensure valid dev_addr */ |
4413 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4414 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4415 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4416 | |
181470fc SH |
4417 | return 0; |
4418 | } | |
4419 | ||
0a2a78c4 | 4420 | unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4421 | { |
efacb309 | 4422 | return tx_queues; |
d5da4510 JP |
4423 | } |
4424 | ||
dfe60397 | 4425 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4426 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4427 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4428 | * set up our sysfs entries. | |
4429 | */ | |
ec87fd3b | 4430 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4431 | { |
4432 | struct net_device *bond_dev; | |
4433 | int res; | |
4434 | ||
4435 | rtnl_lock(); | |
027ea041 | 4436 | |
1c5cae81 JP |
4437 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4438 | name ? name : "bond%d", | |
4439 | bond_setup, tx_queues); | |
dfe60397 | 4440 | if (!bond_dev) { |
a4aee5c8 | 4441 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4442 | rtnl_unlock(); |
4443 | return -ENOMEM; | |
dfe60397 MW |
4444 | } |
4445 | ||
ec87fd3b | 4446 | dev_net_set(bond_dev, net); |
88ead977 EB |
4447 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4448 | ||
dfe60397 | 4449 | res = register_netdevice(bond_dev); |
0daa2303 | 4450 | |
e826eafa PO |
4451 | netif_carrier_off(bond_dev); |
4452 | ||
7e083840 | 4453 | rtnl_unlock(); |
9e2e61fb AW |
4454 | if (res < 0) |
4455 | bond_destructor(bond_dev); | |
30c15ba9 | 4456 | return res; |
dfe60397 MW |
4457 | } |
4458 | ||
2c8c1e72 | 4459 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4460 | { |
15449745 | 4461 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4462 | |
4463 | bn->net = net; | |
4464 | INIT_LIST_HEAD(&bn->dev_list); | |
4465 | ||
ec87fd3b | 4466 | bond_create_proc_dir(bn); |
4c22400a | 4467 | bond_create_sysfs(bn); |
87a7b84b | 4468 | |
15449745 | 4469 | return 0; |
ec87fd3b EB |
4470 | } |
4471 | ||
2c8c1e72 | 4472 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4473 | { |
15449745 | 4474 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4475 | struct bonding *bond, *tmp_bond; |
4476 | LIST_HEAD(list); | |
ec87fd3b | 4477 | |
4c22400a | 4478 | bond_destroy_sysfs(bn); |
ec87fd3b | 4479 | bond_destroy_proc_dir(bn); |
69b0216a | 4480 | |
4481 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4482 | rtnl_lock(); | |
4483 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4484 | unregister_netdevice_queue(bond->dev, &list); | |
4485 | unregister_netdevice_many(&list); | |
4486 | rtnl_unlock(); | |
ec87fd3b EB |
4487 | } |
4488 | ||
4489 | static struct pernet_operations bond_net_ops = { | |
4490 | .init = bond_net_init, | |
4491 | .exit = bond_net_exit, | |
15449745 EB |
4492 | .id = &bond_net_id, |
4493 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4494 | }; |
4495 | ||
1da177e4 LT |
4496 | static int __init bonding_init(void) |
4497 | { | |
1da177e4 LT |
4498 | int i; |
4499 | int res; | |
4500 | ||
bd33acc3 | 4501 | pr_info("%s", bond_version); |
1da177e4 | 4502 | |
dfe60397 | 4503 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4504 | if (res) |
dfe60397 | 4505 | goto out; |
1da177e4 | 4506 | |
15449745 | 4507 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4508 | if (res) |
4509 | goto out; | |
027ea041 | 4510 | |
0a2a78c4 | 4511 | res = bond_netlink_init(); |
88ead977 | 4512 | if (res) |
6639104b | 4513 | goto err_link; |
88ead977 | 4514 | |
f073c7ca TI |
4515 | bond_create_debugfs(); |
4516 | ||
1da177e4 | 4517 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4518 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4519 | if (res) |
4520 | goto err; | |
1da177e4 LT |
4521 | } |
4522 | ||
1da177e4 | 4523 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4524 | out: |
1da177e4 | 4525 | return res; |
88ead977 | 4526 | err: |
0a2a78c4 | 4527 | bond_netlink_fini(); |
6639104b | 4528 | err_link: |
15449745 | 4529 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4530 | goto out; |
dfe60397 | 4531 | |
1da177e4 LT |
4532 | } |
4533 | ||
4534 | static void __exit bonding_exit(void) | |
4535 | { | |
4536 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4537 | ||
f073c7ca | 4538 | bond_destroy_debugfs(); |
ae68c398 | 4539 | |
0a2a78c4 | 4540 | bond_netlink_fini(); |
ffcdedb6 | 4541 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4542 | |
4543 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4544 | /* |
4545 | * Make sure we don't have an imbalance on our netpoll blocking | |
4546 | */ | |
4547 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4548 | #endif |
1da177e4 LT |
4549 | } |
4550 | ||
4551 | module_init(bonding_init); | |
4552 | module_exit(bonding_exit); | |
4553 | MODULE_LICENSE("GPL"); | |
4554 | MODULE_VERSION(DRV_VERSION); | |
4555 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4556 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); |