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