]>
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> |
1da177e4 LT |
80 | #include "bonding.h" |
81 | #include "bond_3ad.h" | |
82 | #include "bond_alb.h" | |
83 | ||
84 | /*---------------------------- Module parameters ----------------------------*/ | |
85 | ||
86 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
87 | #define BOND_LINK_MON_INTERV 0 | |
88 | #define BOND_LINK_ARP_INTERV 0 | |
89 | ||
90 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 91 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
ad246c99 | 92 | static int num_peer_notif = 1; |
1da177e4 | 93 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
94 | static int updelay; |
95 | static int downdelay; | |
1da177e4 | 96 | static int use_carrier = 1; |
3d632c3f SH |
97 | static char *mode; |
98 | static char *primary; | |
a549952a | 99 | static char *primary_reselect; |
3d632c3f | 100 | static char *lacp_rate; |
655f8919 | 101 | static int min_links; |
3d632c3f SH |
102 | static char *ad_select; |
103 | static char *xmit_hash_policy; | |
1da177e4 | 104 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
105 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
106 | static char *arp_validate; | |
8599b52e | 107 | static char *arp_all_targets; |
3d632c3f | 108 | static char *fail_over_mac; |
b07ea07b | 109 | static int all_slaves_active; |
d2991f75 | 110 | static struct bond_params bonding_defaults; |
c2952c31 | 111 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
1da177e4 LT |
112 | |
113 | module_param(max_bonds, int, 0); | |
114 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
115 | module_param(tx_queues, int, 0); |
116 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 117 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
118 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
119 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 120 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
121 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
122 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
123 | module_param(miimon, int, 0); |
124 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
125 | module_param(updelay, int, 0); | |
126 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
127 | module_param(downdelay, int, 0); | |
2ac47660 MW |
128 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
129 | "in milliseconds"); | |
1da177e4 | 130 | module_param(use_carrier, int, 0); |
2ac47660 MW |
131 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
132 | "0 for off, 1 for on (default)"); | |
1da177e4 | 133 | module_param(mode, charp, 0); |
90e62474 | 134 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
135 | "1 for active-backup, 2 for balance-xor, " |
136 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
137 | "6 for balance-alb"); | |
1da177e4 LT |
138 | module_param(primary, charp, 0); |
139 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
140 | module_param(primary_reselect, charp, 0); |
141 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
142 | "once it comes up; " | |
143 | "0 for always (default), " | |
144 | "1 for only if speed of primary is " | |
145 | "better, " | |
146 | "2 for only on active slave " | |
147 | "failure"); | |
1da177e4 | 148 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
149 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
150 | "0 for slow, 1 for fast"); | |
fd989c83 | 151 | module_param(ad_select, charp, 0); |
90e62474 AG |
152 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
153 | "0 for stable (default), 1 for bandwidth, " | |
154 | "2 for count"); | |
655f8919 | 155 | module_param(min_links, int, 0); |
156 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
157 | ||
169a3e66 | 158 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
159 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
160 | "0 for layer 2 (default), 1 for layer 3+4, " | |
161 | "2 for layer 2+3"); | |
1da177e4 LT |
162 | module_param(arp_interval, int, 0); |
163 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
164 | module_param_array(arp_ip_target, charp, NULL, 0); | |
165 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 166 | module_param(arp_validate, charp, 0); |
90e62474 AG |
167 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
168 | "0 for none (default), 1 for active, " | |
169 | "2 for backup, 3 for all"); | |
8599b52e VF |
170 | module_param(arp_all_targets, charp, 0); |
171 | MODULE_PARM_DESC(arp_all_targets, "fail on any/all arp targets timeout; 0 for any (default), 1 for all"); | |
3915c1e8 | 172 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
173 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
174 | "the same MAC; 0 for none (default), " | |
175 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
176 | module_param(all_slaves_active, int, 0); |
177 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 178 | "by setting active flag for all slaves; " |
ebd8e497 | 179 | "0 for never (default), 1 for always."); |
c2952c31 | 180 | module_param(resend_igmp, int, 0); |
90e62474 AG |
181 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
182 | "link failure"); | |
1da177e4 LT |
183 | |
184 | /*----------------------------- Global variables ----------------------------*/ | |
185 | ||
e843fa50 | 186 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 187 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
188 | #endif |
189 | ||
f99189b1 | 190 | int bond_net_id __read_mostly; |
1da177e4 | 191 | |
3d632c3f SH |
192 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
193 | static int arp_ip_count; | |
1da177e4 | 194 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
195 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
196 | static int lacp_fast; | |
1da177e4 | 197 | |
e97fd7c6 | 198 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
199 | { "slow", AD_LACP_SLOW}, |
200 | { "fast", AD_LACP_FAST}, | |
201 | { NULL, -1}, | |
202 | }; | |
203 | ||
e97fd7c6 | 204 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
205 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
206 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
207 | { "balance-xor", BOND_MODE_XOR}, | |
208 | { "broadcast", BOND_MODE_BROADCAST}, | |
209 | { "802.3ad", BOND_MODE_8023AD}, | |
210 | { "balance-tlb", BOND_MODE_TLB}, | |
211 | { "balance-alb", BOND_MODE_ALB}, | |
212 | { NULL, -1}, | |
213 | }; | |
214 | ||
e97fd7c6 | 215 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
216 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
217 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 218 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
219 | { NULL, -1}, |
220 | }; | |
221 | ||
8599b52e VF |
222 | const struct bond_parm_tbl arp_all_targets_tbl[] = { |
223 | { "any", BOND_ARP_TARGETS_ANY}, | |
224 | { "all", BOND_ARP_TARGETS_ALL}, | |
225 | { NULL, -1}, | |
226 | }; | |
227 | ||
e97fd7c6 | 228 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
229 | { "none", BOND_ARP_VALIDATE_NONE}, |
230 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
231 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
232 | { "all", BOND_ARP_VALIDATE_ALL}, | |
233 | { NULL, -1}, | |
234 | }; | |
235 | ||
e97fd7c6 | 236 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
237 | { "none", BOND_FOM_NONE}, |
238 | { "active", BOND_FOM_ACTIVE}, | |
239 | { "follow", BOND_FOM_FOLLOW}, | |
240 | { NULL, -1}, | |
241 | }; | |
242 | ||
a549952a JP |
243 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
244 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
245 | { "better", BOND_PRI_RESELECT_BETTER}, | |
246 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
247 | { NULL, -1}, | |
248 | }; | |
249 | ||
fd989c83 JV |
250 | struct bond_parm_tbl ad_select_tbl[] = { |
251 | { "stable", BOND_AD_STABLE}, | |
252 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
253 | { "count", BOND_AD_COUNT}, | |
254 | { NULL, -1}, | |
255 | }; | |
256 | ||
1da177e4 LT |
257 | /*-------------------------- Forward declarations ---------------------------*/ |
258 | ||
181470fc | 259 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 260 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
261 | |
262 | /*---------------------------- General routines -----------------------------*/ | |
263 | ||
bd33acc3 | 264 | const char *bond_mode_name(int mode) |
1da177e4 | 265 | { |
77afc92b HE |
266 | static const char *names[] = { |
267 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
268 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
269 | [BOND_MODE_XOR] = "load balancing (xor)", | |
270 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 271 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
272 | [BOND_MODE_TLB] = "transmit load balancing", |
273 | [BOND_MODE_ALB] = "adaptive load balancing", | |
274 | }; | |
275 | ||
f5280948 | 276 | if (mode < BOND_MODE_ROUNDROBIN || mode > BOND_MODE_ALB) |
1da177e4 | 277 | return "unknown"; |
77afc92b HE |
278 | |
279 | return names[mode]; | |
1da177e4 LT |
280 | } |
281 | ||
282 | /*---------------------------------- VLAN -----------------------------------*/ | |
283 | ||
284 | /** | |
285 | * bond_add_vlan - add a new vlan id on bond | |
286 | * @bond: bond that got the notification | |
287 | * @vlan_id: the vlan id to add | |
288 | * | |
289 | * Returns -ENOMEM if allocation failed. | |
290 | */ | |
291 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
292 | { | |
293 | struct vlan_entry *vlan; | |
294 | ||
5a03cdb7 | 295 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 296 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 297 | |
305d552a | 298 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 299 | if (!vlan) |
1da177e4 | 300 | return -ENOMEM; |
1da177e4 LT |
301 | |
302 | INIT_LIST_HEAD(&vlan->vlan_list); | |
303 | vlan->vlan_id = vlan_id; | |
304 | ||
305 | write_lock_bh(&bond->lock); | |
306 | ||
307 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
308 | ||
309 | write_unlock_bh(&bond->lock); | |
310 | ||
5a03cdb7 | 311 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
312 | |
313 | return 0; | |
314 | } | |
315 | ||
316 | /** | |
317 | * bond_del_vlan - delete a vlan id from bond | |
318 | * @bond: bond that got the notification | |
319 | * @vlan_id: the vlan id to delete | |
320 | * | |
321 | * returns -ENODEV if @vlan_id was not found in @bond. | |
322 | */ | |
323 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
324 | { | |
0883beca | 325 | struct vlan_entry *vlan; |
1da177e4 LT |
326 | int res = -ENODEV; |
327 | ||
5a03cdb7 | 328 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 | 329 | |
e843fa50 | 330 | block_netpoll_tx(); |
1da177e4 LT |
331 | write_lock_bh(&bond->lock); |
332 | ||
0883beca | 333 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
334 | if (vlan->vlan_id == vlan_id) { |
335 | list_del(&vlan->vlan_list); | |
336 | ||
58402054 | 337 | if (bond_is_lb(bond)) |
1da177e4 | 338 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 339 | |
a4aee5c8 JP |
340 | pr_debug("removed VLAN ID %d from bond %s\n", |
341 | vlan_id, bond->dev->name); | |
1da177e4 LT |
342 | |
343 | kfree(vlan); | |
344 | ||
1da177e4 LT |
345 | res = 0; |
346 | goto out; | |
347 | } | |
348 | } | |
349 | ||
a4aee5c8 JP |
350 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
351 | vlan_id, bond->dev->name); | |
1da177e4 LT |
352 | |
353 | out: | |
354 | write_unlock_bh(&bond->lock); | |
e843fa50 | 355 | unblock_netpoll_tx(); |
1da177e4 LT |
356 | return res; |
357 | } | |
358 | ||
1da177e4 LT |
359 | /** |
360 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
361 | * @bond: the bond we're working on | |
362 | * @curr: item we're advancing from | |
363 | * | |
364 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
365 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 366 | * |
1da177e4 LT |
367 | * Caller must hold bond->lock |
368 | */ | |
369 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
370 | { | |
371 | struct vlan_entry *next, *last; | |
372 | ||
3d632c3f | 373 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 374 | return NULL; |
1da177e4 LT |
375 | |
376 | if (!curr) { | |
377 | next = list_entry(bond->vlan_list.next, | |
378 | struct vlan_entry, vlan_list); | |
379 | } else { | |
380 | last = list_entry(bond->vlan_list.prev, | |
381 | struct vlan_entry, vlan_list); | |
382 | if (last == curr) { | |
383 | next = list_entry(bond->vlan_list.next, | |
384 | struct vlan_entry, vlan_list); | |
385 | } else { | |
386 | next = list_entry(curr->vlan_list.next, | |
387 | struct vlan_entry, vlan_list); | |
388 | } | |
389 | } | |
390 | ||
391 | return next; | |
392 | } | |
393 | ||
394 | /** | |
395 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 396 | * |
1da177e4 LT |
397 | * @bond: bond device that got this skb for tx. |
398 | * @skb: hw accel VLAN tagged skb to transmit | |
399 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 400 | */ |
3d632c3f SH |
401 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
402 | struct net_device *slave_dev) | |
1da177e4 | 403 | { |
8387451e | 404 | skb->dev = slave_dev; |
374eeb5a | 405 | |
5ee31c68 | 406 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
407 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
408 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 409 | |
e15c3c22 | 410 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 411 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 412 | else |
f6dc31a8 | 413 | dev_queue_xmit(skb); |
1da177e4 LT |
414 | |
415 | return 0; | |
416 | } | |
417 | ||
418 | /* | |
cc0e4070 JP |
419 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
420 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
421 | * a. This operation is performed in IOCTL context, |
422 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
423 | * c. Holding a lock with BH disabled while directly calling a base driver | |
424 | * entry point is generally a BAD idea. | |
3d632c3f | 425 | * |
1da177e4 LT |
426 | * The design of synchronization/protection for this operation in the 8021q |
427 | * module is good for one or more VLAN devices over a single physical device | |
428 | * and cannot be extended for a teaming solution like bonding, so there is a | |
429 | * potential race condition here where a net device from the vlan group might | |
430 | * be referenced (either by a base driver or the 8021q code) while it is being | |
431 | * removed from the system. However, it turns out we're not making matters | |
432 | * worse, and if it works for regular VLAN usage it will work here too. | |
433 | */ | |
434 | ||
1da177e4 LT |
435 | /** |
436 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
437 | * @bond_dev: bonding net device that got called | |
438 | * @vid: vlan id being added | |
439 | */ | |
80d5c368 PM |
440 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, |
441 | __be16 proto, u16 vid) | |
1da177e4 | 442 | { |
454d7c9b | 443 | struct bonding *bond = netdev_priv(bond_dev); |
87002b03 | 444 | struct slave *slave, *stop_at; |
1da177e4 LT |
445 | int i, res; |
446 | ||
447 | bond_for_each_slave(bond, slave, i) { | |
80d5c368 | 448 | res = vlan_vid_add(slave->dev, proto, vid); |
87002b03 JP |
449 | if (res) |
450 | goto unwind; | |
1da177e4 LT |
451 | } |
452 | ||
453 | res = bond_add_vlan(bond, vid); | |
454 | if (res) { | |
a4aee5c8 | 455 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 | 456 | bond_dev->name, vid); |
8e586137 | 457 | return res; |
1da177e4 | 458 | } |
8e586137 JP |
459 | |
460 | return 0; | |
87002b03 JP |
461 | |
462 | unwind: | |
463 | /* unwind from head to the slave that failed */ | |
464 | stop_at = slave; | |
465 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) | |
80d5c368 | 466 | vlan_vid_del(slave->dev, proto, vid); |
87002b03 JP |
467 | |
468 | return res; | |
1da177e4 LT |
469 | } |
470 | ||
471 | /** | |
472 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
473 | * @bond_dev: bonding net device that got called | |
474 | * @vid: vlan id being removed | |
475 | */ | |
80d5c368 PM |
476 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, |
477 | __be16 proto, u16 vid) | |
1da177e4 | 478 | { |
454d7c9b | 479 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 480 | struct slave *slave; |
1da177e4 LT |
481 | int i, res; |
482 | ||
87002b03 | 483 | bond_for_each_slave(bond, slave, i) |
80d5c368 | 484 | vlan_vid_del(slave->dev, proto, vid); |
1da177e4 LT |
485 | |
486 | res = bond_del_vlan(bond, vid); | |
487 | if (res) { | |
a4aee5c8 | 488 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 | 489 | bond_dev->name, vid); |
8e586137 | 490 | return res; |
1da177e4 | 491 | } |
8e586137 JP |
492 | |
493 | return 0; | |
1da177e4 LT |
494 | } |
495 | ||
496 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
497 | { | |
498 | struct vlan_entry *vlan; | |
87002b03 | 499 | int res; |
1da177e4 | 500 | |
87002b03 | 501 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
80d5c368 PM |
502 | res = vlan_vid_add(slave_dev, htons(ETH_P_8021Q), |
503 | vlan->vlan_id); | |
87002b03 JP |
504 | if (res) |
505 | pr_warning("%s: Failed to add vlan id %d to device %s\n", | |
506 | bond->dev->name, vlan->vlan_id, | |
507 | slave_dev->name); | |
508 | } | |
1da177e4 LT |
509 | } |
510 | ||
3d632c3f SH |
511 | static void bond_del_vlans_from_slave(struct bonding *bond, |
512 | struct net_device *slave_dev) | |
1da177e4 LT |
513 | { |
514 | struct vlan_entry *vlan; | |
1da177e4 | 515 | |
1da177e4 | 516 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
517 | if (!vlan->vlan_id) |
518 | continue; | |
80d5c368 | 519 | vlan_vid_del(slave_dev, htons(ETH_P_8021Q), vlan->vlan_id); |
1da177e4 | 520 | } |
1da177e4 LT |
521 | } |
522 | ||
523 | /*------------------------------- Link status -------------------------------*/ | |
524 | ||
ff59c456 JV |
525 | /* |
526 | * Set the carrier state for the master according to the state of its | |
527 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
528 | * do special 802.3ad magic. | |
529 | * | |
530 | * Returns zero if carrier state does not change, nonzero if it does. | |
531 | */ | |
532 | static int bond_set_carrier(struct bonding *bond) | |
533 | { | |
534 | struct slave *slave; | |
535 | int i; | |
536 | ||
537 | if (bond->slave_cnt == 0) | |
538 | goto down; | |
539 | ||
540 | if (bond->params.mode == BOND_MODE_8023AD) | |
541 | return bond_3ad_set_carrier(bond); | |
542 | ||
543 | bond_for_each_slave(bond, slave, i) { | |
544 | if (slave->link == BOND_LINK_UP) { | |
545 | if (!netif_carrier_ok(bond->dev)) { | |
546 | netif_carrier_on(bond->dev); | |
547 | return 1; | |
548 | } | |
549 | return 0; | |
550 | } | |
551 | } | |
552 | ||
553 | down: | |
554 | if (netif_carrier_ok(bond->dev)) { | |
555 | netif_carrier_off(bond->dev); | |
556 | return 1; | |
557 | } | |
558 | return 0; | |
559 | } | |
560 | ||
1da177e4 LT |
561 | /* |
562 | * Get link speed and duplex from the slave's base driver | |
563 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 564 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 565 | * and return. |
1da177e4 | 566 | */ |
13b95fb7 | 567 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
568 | { |
569 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 570 | struct ethtool_cmd ecmd; |
5d30530e | 571 | u32 slave_speed; |
61a44b9c | 572 | int res; |
1da177e4 | 573 | |
589665f5 DC |
574 | slave->speed = SPEED_UNKNOWN; |
575 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 576 | |
4bc71cb9 | 577 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 578 | if (res < 0) |
13b95fb7 | 579 | return; |
1da177e4 | 580 | |
4bc71cb9 | 581 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 582 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 583 | return; |
1da177e4 | 584 | |
4bc71cb9 | 585 | switch (ecmd.duplex) { |
1da177e4 LT |
586 | case DUPLEX_FULL: |
587 | case DUPLEX_HALF: | |
588 | break; | |
589 | default: | |
13b95fb7 | 590 | return; |
1da177e4 LT |
591 | } |
592 | ||
5d30530e | 593 | slave->speed = slave_speed; |
4bc71cb9 | 594 | slave->duplex = ecmd.duplex; |
1da177e4 | 595 | |
13b95fb7 | 596 | return; |
1da177e4 LT |
597 | } |
598 | ||
599 | /* | |
600 | * if <dev> supports MII link status reporting, check its link status. | |
601 | * | |
602 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 603 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
604 | * |
605 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
606 | * can't tell and just pretend it is), or 0, meaning that the link is | |
607 | * down. | |
608 | * | |
609 | * If reporting is non-zero, instead of faking link up, return -1 if | |
610 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
611 | * support them). If use_carrier is set, return whatever it says. | |
612 | * It'd be nice if there was a good way to tell if a driver supports | |
613 | * netif_carrier, but there really isn't. | |
614 | */ | |
3d632c3f SH |
615 | static int bond_check_dev_link(struct bonding *bond, |
616 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 617 | { |
eb7cc59a | 618 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 619 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
620 | struct ifreq ifr; |
621 | struct mii_ioctl_data *mii; | |
1da177e4 | 622 | |
6c988853 PG |
623 | if (!reporting && !netif_running(slave_dev)) |
624 | return 0; | |
625 | ||
eb7cc59a | 626 | if (bond->params.use_carrier) |
1da177e4 | 627 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 628 | |
29112f4e | 629 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
630 | if (slave_dev->ethtool_ops->get_link) |
631 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
632 | BMSR_LSTATUS : 0; | |
29112f4e | 633 | |
3d632c3f | 634 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 635 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
636 | if (ioctl) { |
637 | /* TODO: set pointer to correct ioctl on a per team member */ | |
638 | /* bases to make this more efficient. that is, once */ | |
639 | /* we determine the correct ioctl, we will always */ | |
640 | /* call it and not the others for that team */ | |
641 | /* member. */ | |
642 | ||
643 | /* | |
644 | * We cannot assume that SIOCGMIIPHY will also read a | |
645 | * register; not all network drivers (e.g., e100) | |
646 | * support that. | |
647 | */ | |
648 | ||
649 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
650 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
651 | mii = if_mii(&ifr); | |
652 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
653 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
654 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
655 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
656 | } |
657 | } | |
658 | ||
1da177e4 LT |
659 | /* |
660 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 661 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
662 | * cannot report link status). If not reporting, pretend |
663 | * we're ok. | |
664 | */ | |
3d632c3f | 665 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
666 | } |
667 | ||
668 | /*----------------------------- Multicast list ------------------------------*/ | |
669 | ||
1da177e4 LT |
670 | /* |
671 | * Push the promiscuity flag down to appropriate slaves | |
672 | */ | |
7e1a1ac1 | 673 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 674 | { |
7e1a1ac1 | 675 | int err = 0; |
1da177e4 LT |
676 | if (USES_PRIMARY(bond->params.mode)) { |
677 | /* write lock already acquired */ | |
678 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
679 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
680 | inc); | |
1da177e4 LT |
681 | } |
682 | } else { | |
683 | struct slave *slave; | |
684 | int i; | |
685 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
686 | err = dev_set_promiscuity(slave->dev, inc); |
687 | if (err) | |
688 | return err; | |
1da177e4 LT |
689 | } |
690 | } | |
7e1a1ac1 | 691 | return err; |
1da177e4 LT |
692 | } |
693 | ||
694 | /* | |
695 | * Push the allmulti flag down to all slaves | |
696 | */ | |
7e1a1ac1 | 697 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 698 | { |
7e1a1ac1 | 699 | int err = 0; |
1da177e4 LT |
700 | if (USES_PRIMARY(bond->params.mode)) { |
701 | /* write lock already acquired */ | |
702 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
703 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
704 | inc); | |
1da177e4 LT |
705 | } |
706 | } else { | |
707 | struct slave *slave; | |
708 | int i; | |
709 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
710 | err = dev_set_allmulti(slave->dev, inc); |
711 | if (err) | |
712 | return err; | |
1da177e4 LT |
713 | } |
714 | } | |
7e1a1ac1 | 715 | return err; |
1da177e4 LT |
716 | } |
717 | ||
5a37e8ca FL |
718 | /* |
719 | * Retrieve the list of registered multicast addresses for the bonding | |
720 | * device and retransmit an IGMP JOIN request to the current active | |
721 | * slave. | |
722 | */ | |
723 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
724 | { | |
4aa5dee4 | 725 | if (!rtnl_trylock()) { |
4beac029 | 726 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
4aa5dee4 | 727 | return; |
5a37e8ca | 728 | } |
4aa5dee4 JP |
729 | call_netdevice_notifiers(NETDEV_RESEND_IGMP, bond->dev); |
730 | rtnl_unlock(); | |
5a37e8ca | 731 | |
4f5474e7 NA |
732 | /* We use curr_slave_lock to protect against concurrent access to |
733 | * igmp_retrans from multiple running instances of this function and | |
734 | * bond_change_active_slave | |
735 | */ | |
736 | write_lock_bh(&bond->curr_slave_lock); | |
737 | if (bond->igmp_retrans > 1) { | |
738 | bond->igmp_retrans--; | |
c2952c31 | 739 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
4f5474e7 NA |
740 | } |
741 | write_unlock_bh(&bond->curr_slave_lock); | |
5a37e8ca FL |
742 | } |
743 | ||
379b7383 | 744 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
745 | { |
746 | struct bonding *bond = container_of(work, struct bonding, | |
94265cf5 | 747 | mcast_work.work); |
ad999eee | 748 | |
5a37e8ca FL |
749 | bond_resend_igmp_join_requests(bond); |
750 | } | |
751 | ||
303d1cbf | 752 | /* Flush bond's hardware addresses from slave |
1da177e4 | 753 | */ |
303d1cbf | 754 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 755 | struct net_device *slave_dev) |
1da177e4 | 756 | { |
454d7c9b | 757 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 758 | |
303d1cbf JV |
759 | dev_uc_unsync(slave_dev, bond_dev); |
760 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 LT |
761 | |
762 | if (bond->params.mode == BOND_MODE_8023AD) { | |
763 | /* del lacpdu mc addr from mc list */ | |
764 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
765 | ||
22bedad3 | 766 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
767 | } |
768 | } | |
769 | ||
770 | /*--------------------------- Active slave change ---------------------------*/ | |
771 | ||
303d1cbf JV |
772 | /* Update the hardware address list and promisc/allmulti for the new and |
773 | * old active slaves (if any). Modes that are !USES_PRIMARY keep all | |
774 | * slaves up date at all times; only the USES_PRIMARY modes need to call | |
775 | * this function to swap these settings during a failover. | |
1da177e4 | 776 | */ |
303d1cbf JV |
777 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
778 | struct slave *old_active) | |
1da177e4 | 779 | { |
1da177e4 | 780 | if (old_active) { |
3d632c3f | 781 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 782 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 783 | |
3d632c3f | 784 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 785 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 786 | |
303d1cbf | 787 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
788 | } |
789 | ||
790 | if (new_active) { | |
7e1a1ac1 | 791 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 792 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 793 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 794 | |
3d632c3f | 795 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 796 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 797 | |
d632ce98 | 798 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
799 | dev_uc_sync(new_active->dev, bond->dev); |
800 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 801 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
802 | } |
803 | } | |
804 | ||
ae0d6750 | 805 | /** |
806 | * bond_set_dev_addr - clone slave's address to bond | |
807 | * @bond_dev: bond net device | |
808 | * @slave_dev: slave net device | |
809 | * | |
810 | * Should be called with RTNL held. | |
811 | */ | |
812 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
813 | struct net_device *slave_dev) | |
814 | { | |
008aebde NA |
815 | pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
816 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 817 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
818 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
819 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
820 | } | |
821 | ||
3915c1e8 JV |
822 | /* |
823 | * bond_do_fail_over_mac | |
824 | * | |
825 | * Perform special MAC address swapping for fail_over_mac settings | |
826 | * | |
827 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
828 | */ | |
829 | static void bond_do_fail_over_mac(struct bonding *bond, | |
830 | struct slave *new_active, | |
831 | struct slave *old_active) | |
1f78d9f9 HE |
832 | __releases(&bond->curr_slave_lock) |
833 | __releases(&bond->lock) | |
834 | __acquires(&bond->lock) | |
835 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
836 | { |
837 | u8 tmp_mac[ETH_ALEN]; | |
838 | struct sockaddr saddr; | |
839 | int rv; | |
840 | ||
841 | switch (bond->params.fail_over_mac) { | |
842 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 843 | if (new_active) { |
7d26bb10 WP |
844 | write_unlock_bh(&bond->curr_slave_lock); |
845 | read_unlock(&bond->lock); | |
ae0d6750 | 846 | bond_set_dev_addr(bond->dev, new_active->dev); |
7d26bb10 WP |
847 | read_lock(&bond->lock); |
848 | write_lock_bh(&bond->curr_slave_lock); | |
849 | } | |
3915c1e8 JV |
850 | break; |
851 | case BOND_FOM_FOLLOW: | |
852 | /* | |
853 | * if new_active && old_active, swap them | |
854 | * if just old_active, do nothing (going to no active slave) | |
855 | * if just new_active, set new_active to bond's MAC | |
856 | */ | |
857 | if (!new_active) | |
858 | return; | |
859 | ||
860 | write_unlock_bh(&bond->curr_slave_lock); | |
861 | read_unlock(&bond->lock); | |
862 | ||
863 | if (old_active) { | |
864 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
865 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
866 | ETH_ALEN); | |
867 | saddr.sa_family = new_active->dev->type; | |
868 | } else { | |
869 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
870 | saddr.sa_family = bond->dev->type; | |
871 | } | |
872 | ||
873 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
874 | if (rv) { | |
a4aee5c8 | 875 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
876 | bond->dev->name, -rv, new_active->dev->name); |
877 | goto out; | |
878 | } | |
879 | ||
880 | if (!old_active) | |
881 | goto out; | |
882 | ||
883 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
884 | saddr.sa_family = old_active->dev->type; | |
885 | ||
886 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
887 | if (rv) | |
a4aee5c8 | 888 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
889 | bond->dev->name, -rv, new_active->dev->name); |
890 | out: | |
891 | read_lock(&bond->lock); | |
892 | write_lock_bh(&bond->curr_slave_lock); | |
893 | break; | |
894 | default: | |
a4aee5c8 | 895 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
896 | bond->dev->name, bond->params.fail_over_mac); |
897 | break; | |
898 | } | |
899 | ||
900 | } | |
901 | ||
a549952a JP |
902 | static bool bond_should_change_active(struct bonding *bond) |
903 | { | |
904 | struct slave *prim = bond->primary_slave; | |
905 | struct slave *curr = bond->curr_active_slave; | |
906 | ||
907 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
908 | return true; | |
909 | if (bond->force_primary) { | |
910 | bond->force_primary = false; | |
911 | return true; | |
912 | } | |
913 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
914 | (prim->speed < curr->speed || | |
915 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
916 | return false; | |
917 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
918 | return false; | |
919 | return true; | |
920 | } | |
3915c1e8 | 921 | |
1da177e4 LT |
922 | /** |
923 | * find_best_interface - select the best available slave to be the active one | |
924 | * @bond: our bonding struct | |
925 | * | |
926 | * Warning: Caller must hold curr_slave_lock for writing. | |
927 | */ | |
928 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
929 | { | |
930 | struct slave *new_active, *old_active; | |
931 | struct slave *bestslave = NULL; | |
932 | int mintime = bond->params.updelay; | |
933 | int i; | |
934 | ||
49b4ad92 | 935 | new_active = bond->curr_active_slave; |
1da177e4 LT |
936 | |
937 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 938 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 939 | new_active = bond->first_slave; |
3d632c3f | 940 | else |
1da177e4 | 941 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
942 | } |
943 | ||
1da177e4 | 944 | if ((bond->primary_slave) && |
a549952a JP |
945 | bond->primary_slave->link == BOND_LINK_UP && |
946 | bond_should_change_active(bond)) { | |
1da177e4 LT |
947 | new_active = bond->primary_slave; |
948 | } | |
949 | ||
950 | /* remember where to stop iterating over the slaves */ | |
951 | old_active = new_active; | |
952 | ||
953 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
954 | if (new_active->link == BOND_LINK_UP) { |
955 | return new_active; | |
956 | } else if (new_active->link == BOND_LINK_BACK && | |
957 | IS_UP(new_active->dev)) { | |
958 | /* link up, but waiting for stabilization */ | |
959 | if (new_active->delay < mintime) { | |
960 | mintime = new_active->delay; | |
961 | bestslave = new_active; | |
1da177e4 LT |
962 | } |
963 | } | |
964 | } | |
965 | ||
966 | return bestslave; | |
967 | } | |
968 | ||
ad246c99 BH |
969 | static bool bond_should_notify_peers(struct bonding *bond) |
970 | { | |
971 | struct slave *slave = bond->curr_active_slave; | |
972 | ||
973 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
974 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
975 | ||
976 | if (!slave || !bond->send_peer_notif || | |
977 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
978 | return false; | |
979 | ||
980 | bond->send_peer_notif--; | |
981 | return true; | |
982 | } | |
983 | ||
1da177e4 LT |
984 | /** |
985 | * change_active_interface - change the active slave into the specified one | |
986 | * @bond: our bonding struct | |
987 | * @new: the new slave to make the active one | |
988 | * | |
989 | * Set the new slave to the bond's settings and unset them on the old | |
990 | * curr_active_slave. | |
991 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
992 | * | |
993 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
994 | * because it is apparently the best available slave we have, even though its | |
995 | * updelay hasn't timed out yet. | |
996 | * | |
3915c1e8 JV |
997 | * If new_active is not NULL, caller must hold bond->lock for read and |
998 | * curr_slave_lock for write_bh. | |
1da177e4 | 999 | */ |
a77b5325 | 1000 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1001 | { |
1002 | struct slave *old_active = bond->curr_active_slave; | |
1003 | ||
3d632c3f | 1004 | if (old_active == new_active) |
1da177e4 | 1005 | return; |
1da177e4 LT |
1006 | |
1007 | if (new_active) { | |
b2220cad JV |
1008 | new_active->jiffies = jiffies; |
1009 | ||
1da177e4 LT |
1010 | if (new_active->link == BOND_LINK_BACK) { |
1011 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1012 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1013 | bond->dev->name, new_active->dev->name, | |
1014 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1015 | } |
1016 | ||
1017 | new_active->delay = 0; | |
1018 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1019 | |
3d632c3f | 1020 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1021 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1022 | |
58402054 | 1023 | if (bond_is_lb(bond)) |
1da177e4 | 1024 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1025 | } else { |
1026 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1027 | pr_info("%s: making interface %s the new active one.\n", |
1028 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1029 | } |
1030 | } | |
1031 | } | |
1032 | ||
3d632c3f | 1033 | if (USES_PRIMARY(bond->params.mode)) |
303d1cbf | 1034 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 1035 | |
58402054 | 1036 | if (bond_is_lb(bond)) { |
1da177e4 | 1037 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1038 | if (old_active) |
1039 | bond_set_slave_inactive_flags(old_active); | |
1040 | if (new_active) | |
1041 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1042 | } else { |
1043 | bond->curr_active_slave = new_active; | |
1044 | } | |
c3ade5ca JV |
1045 | |
1046 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1047 | if (old_active) |
c3ade5ca | 1048 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1049 | |
1050 | if (new_active) { | |
ad246c99 BH |
1051 | bool should_notify_peers = false; |
1052 | ||
c3ade5ca | 1053 | bond_set_slave_active_flags(new_active); |
2ab82852 | 1054 | |
709f8a45 OG |
1055 | if (bond->params.fail_over_mac) |
1056 | bond_do_fail_over_mac(bond, new_active, | |
1057 | old_active); | |
3915c1e8 | 1058 | |
ad246c99 BH |
1059 | if (netif_running(bond->dev)) { |
1060 | bond->send_peer_notif = | |
1061 | bond->params.num_peer_notif; | |
1062 | should_notify_peers = | |
1063 | bond_should_notify_peers(bond); | |
1064 | } | |
1065 | ||
01f3109d OG |
1066 | write_unlock_bh(&bond->curr_slave_lock); |
1067 | read_unlock(&bond->lock); | |
1068 | ||
b7bc2a5b | 1069 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 1070 | if (should_notify_peers) |
b7bc2a5b AW |
1071 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
1072 | bond->dev); | |
01f3109d OG |
1073 | |
1074 | read_lock(&bond->lock); | |
1075 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1076 | } |
c3ade5ca | 1077 | } |
a2fd940f | 1078 | |
5a37e8ca | 1079 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
1080 | * all were sent on curr_active_slave. |
1081 | * resend only if bond is brought up with the affected | |
1082 | * bonding modes and the retransmission is enabled */ | |
1083 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
1084 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
1085 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 1086 | bond->igmp_retrans = bond->params.resend_igmp; |
4beac029 | 1087 | queue_delayed_work(bond->wq, &bond->mcast_work, 1); |
a2fd940f | 1088 | } |
1da177e4 LT |
1089 | } |
1090 | ||
1091 | /** | |
1092 | * bond_select_active_slave - select a new active slave, if needed | |
1093 | * @bond: our bonding struct | |
1094 | * | |
3d632c3f | 1095 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1096 | * - The old curr_active_slave has been released or lost its link. |
1097 | * - The primary_slave has got its link back. | |
1098 | * - A slave has got its link back and there's no old curr_active_slave. | |
1099 | * | |
3915c1e8 | 1100 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1101 | */ |
a77b5325 | 1102 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1103 | { |
1104 | struct slave *best_slave; | |
ff59c456 | 1105 | int rv; |
1da177e4 LT |
1106 | |
1107 | best_slave = bond_find_best_slave(bond); | |
1108 | if (best_slave != bond->curr_active_slave) { | |
1109 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1110 | rv = bond_set_carrier(bond); |
1111 | if (!rv) | |
1112 | return; | |
1113 | ||
1114 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1115 | pr_info("%s: first active interface up!\n", |
1116 | bond->dev->name); | |
ff59c456 | 1117 | } else { |
a4aee5c8 JP |
1118 | pr_info("%s: now running without any active interface !\n", |
1119 | bond->dev->name); | |
ff59c456 | 1120 | } |
1da177e4 LT |
1121 | } |
1122 | } | |
1123 | ||
1124 | /*--------------------------- slave list handling ---------------------------*/ | |
1125 | ||
1126 | /* | |
1127 | * This function attaches the slave to the end of list. | |
1128 | * | |
1129 | * bond->lock held for writing by caller. | |
1130 | */ | |
1131 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1132 | { | |
1133 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1134 | new_slave->next = new_slave; | |
1135 | new_slave->prev = new_slave; | |
1136 | bond->first_slave = new_slave; | |
1137 | } else { | |
1138 | new_slave->next = bond->first_slave; | |
1139 | new_slave->prev = bond->first_slave->prev; | |
1140 | new_slave->next->prev = new_slave; | |
1141 | new_slave->prev->next = new_slave; | |
1142 | } | |
1143 | ||
1144 | bond->slave_cnt++; | |
1145 | } | |
1146 | ||
1147 | /* | |
1148 | * This function detaches the slave from the list. | |
1149 | * WARNING: no check is made to verify if the slave effectively | |
1150 | * belongs to <bond>. | |
1151 | * Nothing is freed on return, structures are just unchained. | |
1152 | * If any slave pointer in bond was pointing to <slave>, | |
1153 | * it should be changed by the calling function. | |
1154 | * | |
1155 | * bond->lock held for writing by caller. | |
1156 | */ | |
1157 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1158 | { | |
3d632c3f | 1159 | if (slave->next) |
1da177e4 | 1160 | slave->next->prev = slave->prev; |
1da177e4 | 1161 | |
3d632c3f | 1162 | if (slave->prev) |
1da177e4 | 1163 | slave->prev->next = slave->next; |
1da177e4 LT |
1164 | |
1165 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1166 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1167 | bond->first_slave = slave->next; | |
1168 | } else { | |
1169 | bond->first_slave = NULL; /* slave was the last one */ | |
1170 | } | |
1171 | } | |
1172 | ||
1173 | slave->next = NULL; | |
1174 | slave->prev = NULL; | |
1175 | bond->slave_cnt--; | |
1176 | } | |
1177 | ||
f6dc31a8 | 1178 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1179 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1180 | { |
8a8efa22 AW |
1181 | struct netpoll *np; |
1182 | int err = 0; | |
f6dc31a8 | 1183 | |
47be03a2 | 1184 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
1185 | err = -ENOMEM; |
1186 | if (!np) | |
1187 | goto out; | |
1188 | ||
47be03a2 | 1189 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
1190 | if (err) { |
1191 | kfree(np); | |
1192 | goto out; | |
f6dc31a8 | 1193 | } |
8a8efa22 AW |
1194 | slave->np = np; |
1195 | out: | |
1196 | return err; | |
1197 | } | |
1198 | static inline void slave_disable_netpoll(struct slave *slave) | |
1199 | { | |
1200 | struct netpoll *np = slave->np; | |
1201 | ||
1202 | if (!np) | |
1203 | return; | |
1204 | ||
1205 | slave->np = NULL; | |
2cde6acd | 1206 | __netpoll_free_async(np); |
8a8efa22 AW |
1207 | } |
1208 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1209 | { | |
1210 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1211 | return false; | |
1212 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1213 | return false; | |
1214 | return true; | |
f6dc31a8 WC |
1215 | } |
1216 | ||
1217 | static void bond_poll_controller(struct net_device *bond_dev) | |
1218 | { | |
8a8efa22 AW |
1219 | } |
1220 | ||
c4cdef9b | 1221 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
8a8efa22 | 1222 | { |
c4cdef9b | 1223 | struct bonding *bond = netdev_priv(bond_dev); |
c2355e1a NH |
1224 | struct slave *slave; |
1225 | int i; | |
1226 | ||
8a8efa22 AW |
1227 | bond_for_each_slave(bond, slave, i) |
1228 | if (IS_UP(slave->dev)) | |
1229 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1230 | } |
8a8efa22 | 1231 | |
47be03a2 | 1232 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1233 | { |
1234 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1235 | struct slave *slave; |
8a8efa22 | 1236 | int i, err = 0; |
f6dc31a8 | 1237 | |
f6dc31a8 | 1238 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1239 | err = slave_enable_netpoll(slave); |
1240 | if (err) { | |
c4cdef9b | 1241 | bond_netpoll_cleanup(dev); |
8a8efa22 | 1242 | break; |
f6dc31a8 WC |
1243 | } |
1244 | } | |
8a8efa22 | 1245 | return err; |
f6dc31a8 | 1246 | } |
8a8efa22 AW |
1247 | #else |
1248 | static inline int slave_enable_netpoll(struct slave *slave) | |
1249 | { | |
1250 | return 0; | |
1251 | } | |
1252 | static inline void slave_disable_netpoll(struct slave *slave) | |
1253 | { | |
1254 | } | |
f6dc31a8 WC |
1255 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1256 | { | |
1257 | } | |
f6dc31a8 WC |
1258 | #endif |
1259 | ||
1da177e4 LT |
1260 | /*---------------------------------- IOCTL ----------------------------------*/ |
1261 | ||
c8f44aff MM |
1262 | static netdev_features_t bond_fix_features(struct net_device *dev, |
1263 | netdev_features_t features) | |
8531c5ff | 1264 | { |
8531c5ff | 1265 | struct slave *slave; |
b2a103e6 | 1266 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1267 | netdev_features_t mask; |
8e3babcd | 1268 | int i; |
8531c5ff | 1269 | |
b2a103e6 | 1270 | read_lock(&bond->lock); |
b63365a2 | 1271 | |
b2a103e6 MM |
1272 | if (!bond->first_slave) { |
1273 | /* Disable adding VLANs to empty bond. But why? --mq */ | |
1274 | features |= NETIF_F_VLAN_CHALLENGED; | |
1275 | goto out; | |
1276 | } | |
b63365a2 | 1277 | |
b2a103e6 | 1278 | mask = features; |
b63365a2 | 1279 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1280 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1281 | |
54ef3137 | 1282 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1283 | features = netdev_increment_features(features, |
1284 | slave->dev->features, | |
b2a103e6 MM |
1285 | mask); |
1286 | } | |
b0ce3508 | 1287 | features = netdev_add_tso_features(features, mask); |
b2a103e6 MM |
1288 | |
1289 | out: | |
1290 | read_unlock(&bond->lock); | |
1291 | return features; | |
1292 | } | |
1293 | ||
62f2a3a4 MM |
1294 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1295 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1296 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1297 | |
1298 | static void bond_compute_features(struct bonding *bond) | |
1299 | { | |
1300 | struct slave *slave; | |
1301 | struct net_device *bond_dev = bond->dev; | |
c8f44aff | 1302 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1303 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1304 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1305 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1306 | int i; |
b6fe83e9 | 1307 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
b2a103e6 MM |
1308 | |
1309 | read_lock(&bond->lock); | |
1310 | ||
1311 | if (!bond->first_slave) | |
1312 | goto done; | |
1313 | ||
1314 | bond_for_each_slave(bond, slave, i) { | |
278339a4 | 1315 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1316 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1317 | ||
b6fe83e9 | 1318 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1319 | if (slave->dev->hard_header_len > max_hard_header_len) |
1320 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1321 | |
1322 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1323 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1324 | } |
8531c5ff | 1325 | |
b63365a2 | 1326 | done: |
b2a103e6 | 1327 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1328 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1329 | bond_dev->gso_max_segs = gso_max_segs; |
1330 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1331 | |
b6fe83e9 ED |
1332 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1333 | bond_dev->priv_flags = flags | dst_release_flag; | |
1334 | ||
b2a103e6 MM |
1335 | read_unlock(&bond->lock); |
1336 | ||
1337 | netdev_change_features(bond_dev); | |
8531c5ff AK |
1338 | } |
1339 | ||
872254dd MS |
1340 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1341 | struct net_device *slave_dev) | |
1342 | { | |
00829823 | 1343 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1344 | |
1345 | bond_dev->type = slave_dev->type; | |
1346 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1347 | bond_dev->addr_len = slave_dev->addr_len; | |
1348 | ||
1349 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1350 | slave_dev->addr_len); | |
1351 | } | |
1352 | ||
5b2c4dd2 | 1353 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1354 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1355 | */ |
1356 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1357 | struct slave *slave, |
1358 | struct bonding *bond) | |
5b2c4dd2 | 1359 | { |
2d7011ca | 1360 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1361 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1362 | skb->pkt_type != PACKET_BROADCAST && |
1363 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1364 | return false; |
5b2c4dd2 JP |
1365 | return true; |
1366 | } | |
1367 | return false; | |
1368 | } | |
1369 | ||
8a4eb573 | 1370 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1371 | { |
8a4eb573 | 1372 | struct sk_buff *skb = *pskb; |
f1c1775a | 1373 | struct slave *slave; |
0bd80dad | 1374 | struct bonding *bond; |
de063b70 ED |
1375 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1376 | struct slave *); | |
13a8e0c8 | 1377 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1378 | |
8a4eb573 JP |
1379 | skb = skb_share_check(skb, GFP_ATOMIC); |
1380 | if (unlikely(!skb)) | |
1381 | return RX_HANDLER_CONSUMED; | |
1382 | ||
1383 | *pskb = skb; | |
5b2c4dd2 | 1384 | |
35d48903 JP |
1385 | slave = bond_slave_get_rcu(skb->dev); |
1386 | bond = slave->bond; | |
0bd80dad JP |
1387 | |
1388 | if (bond->params.arp_interval) | |
f1c1775a | 1389 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1390 | |
4d97480b MH |
1391 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1392 | if (recv_probe) { | |
de063b70 ED |
1393 | ret = recv_probe(skb, bond, slave); |
1394 | if (ret == RX_HANDLER_CONSUMED) { | |
1395 | consume_skb(skb); | |
1396 | return ret; | |
3aba891d JP |
1397 | } |
1398 | } | |
1399 | ||
0bd80dad | 1400 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1401 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1402 | } |
1403 | ||
35d48903 | 1404 | skb->dev = bond->dev; |
5b2c4dd2 | 1405 | |
0bd80dad | 1406 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1407 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1408 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1409 | |
541ac7c9 CG |
1410 | if (unlikely(skb_cow_head(skb, |
1411 | skb->data - skb_mac_header(skb)))) { | |
1412 | kfree_skb(skb); | |
8a4eb573 | 1413 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1414 | } |
35d48903 | 1415 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1416 | } |
1417 | ||
13a8e0c8 | 1418 | return ret; |
5b2c4dd2 JP |
1419 | } |
1420 | ||
471cb5a3 JP |
1421 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1422 | struct net_device *slave_dev) | |
1423 | { | |
1424 | int err; | |
1425 | ||
1426 | err = netdev_master_upper_dev_link(slave_dev, bond_dev); | |
1427 | if (err) | |
1428 | return err; | |
1429 | slave_dev->flags |= IFF_SLAVE; | |
1430 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1431 | return 0; | |
1432 | } | |
1433 | ||
1434 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1435 | struct net_device *slave_dev) | |
1436 | { | |
1437 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1438 | slave_dev->flags &= ~IFF_SLAVE; | |
1439 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1440 | } | |
1441 | ||
1da177e4 | 1442 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1443 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1444 | { |
454d7c9b | 1445 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1446 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1447 | struct slave *new_slave = NULL; |
1da177e4 LT |
1448 | struct sockaddr addr; |
1449 | int link_reporting; | |
8599b52e | 1450 | int res = 0, i; |
1da177e4 | 1451 | |
c772dde3 BH |
1452 | if (!bond->params.use_carrier && |
1453 | slave_dev->ethtool_ops->get_link == NULL && | |
1454 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1455 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1456 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1457 | } |
1458 | ||
1da177e4 LT |
1459 | /* already enslaved */ |
1460 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1461 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1462 | return -EBUSY; |
1463 | } | |
1464 | ||
1465 | /* vlan challenged mutual exclusion */ | |
1466 | /* no need to lock since we're protected by rtnl_lock */ | |
1467 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1468 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1469 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1470 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1471 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1472 | return -EPERM; |
1473 | } else { | |
a4aee5c8 JP |
1474 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1475 | bond_dev->name, slave_dev->name, | |
1476 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1477 | } |
1478 | } else { | |
5a03cdb7 | 1479 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1480 | } |
1481 | ||
217df670 JV |
1482 | /* |
1483 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1484 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1485 | * the current ifenslave will set the interface down prior to |
1486 | * enslaving it; the old ifenslave will not. | |
1487 | */ | |
1488 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1489 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1490 | slave_dev->name); |
1491 | res = -EPERM; | |
1492 | goto err_undo_flags; | |
1493 | } | |
1da177e4 | 1494 | |
872254dd MS |
1495 | /* set bonding device ether type by slave - bonding netdevices are |
1496 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1497 | * there is a need to override some of the type dependent attribs/funcs. | |
1498 | * | |
1499 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1500 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1501 | */ | |
1502 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1503 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1504 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1505 | bond_dev->name, |
1506 | bond_dev->type, slave_dev->type); | |
75c78500 | 1507 | |
b7bc2a5b AW |
1508 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1509 | bond_dev); | |
3ca5b404 JP |
1510 | res = notifier_to_errno(res); |
1511 | if (res) { | |
1512 | pr_err("%s: refused to change device type\n", | |
1513 | bond_dev->name); | |
1514 | res = -EBUSY; | |
1515 | goto err_undo_flags; | |
1516 | } | |
75c78500 | 1517 | |
32a806c1 | 1518 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1519 | dev_uc_flush(bond_dev); |
22bedad3 | 1520 | dev_mc_flush(bond_dev); |
32a806c1 | 1521 | |
e36b9d16 MS |
1522 | if (slave_dev->type != ARPHRD_ETHER) |
1523 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1524 | else { |
e36b9d16 | 1525 | ether_setup(bond_dev); |
550fd08c NH |
1526 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1527 | } | |
75c78500 | 1528 | |
b7bc2a5b AW |
1529 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1530 | bond_dev); | |
e36b9d16 | 1531 | } |
872254dd | 1532 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1533 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1534 | slave_dev->name, | |
1535 | slave_dev->type, bond_dev->type); | |
1536 | res = -EINVAL; | |
1537 | goto err_undo_flags; | |
872254dd MS |
1538 | } |
1539 | ||
eb7cc59a | 1540 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1541 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1542 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1543 | bond_dev->name); | |
3915c1e8 JV |
1544 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1545 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1546 | 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", |
1547 | bond_dev->name); | |
2ab82852 MS |
1548 | res = -EOPNOTSUPP; |
1549 | goto err_undo_flags; | |
1550 | } | |
1da177e4 LT |
1551 | } |
1552 | ||
8d8fc29d AW |
1553 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1554 | ||
c20811a7 JP |
1555 | /* If this is the first slave, then we need to set the master's hardware |
1556 | * address to be the same as the slave's. */ | |
97a1e639 | 1557 | if (!bond->slave_cnt && bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1558 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1559 | |
243cb4e5 | 1560 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1561 | if (!new_slave) { |
1562 | res = -ENOMEM; | |
1563 | goto err_undo_flags; | |
1564 | } | |
1565 | ||
bb1d9123 AG |
1566 | /* |
1567 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1568 | * is set via sysfs or module option if desired. | |
1569 | */ | |
1570 | new_slave->queue_id = 0; | |
1571 | ||
b15ba0fb JP |
1572 | /* Save slave's original mtu and then set it to match the bond */ |
1573 | new_slave->original_mtu = slave_dev->mtu; | |
1574 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1575 | if (res) { | |
1576 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1577 | goto err_free; | |
1578 | } | |
1579 | ||
217df670 JV |
1580 | /* |
1581 | * Save slave's original ("permanent") mac address for modes | |
1582 | * that need it, and for restoring it upon release, and then | |
1583 | * set it to the master's address | |
1584 | */ | |
1585 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1586 | |
dd957c57 | 1587 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1588 | /* |
1589 | * Set slave to master's mac address. The application already | |
1590 | * set the master's mac address to that of the first slave | |
1591 | */ | |
1592 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1593 | addr.sa_family = slave_dev->type; | |
1594 | res = dev_set_mac_address(slave_dev, &addr); | |
1595 | if (res) { | |
5a03cdb7 | 1596 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1597 | goto err_restore_mtu; |
2ab82852 | 1598 | } |
217df670 | 1599 | } |
1da177e4 | 1600 | |
471cb5a3 | 1601 | res = bond_master_upper_dev_link(bond_dev, slave_dev); |
c2edacf8 | 1602 | if (res) { |
471cb5a3 | 1603 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); |
569f0c4d | 1604 | goto err_restore_mac; |
c2edacf8 | 1605 | } |
5b2c4dd2 | 1606 | |
217df670 JV |
1607 | /* open the slave since the application closed it */ |
1608 | res = dev_open(slave_dev); | |
1609 | if (res) { | |
3d632c3f | 1610 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1611 | goto err_unset_master; |
1da177e4 LT |
1612 | } |
1613 | ||
35d48903 | 1614 | new_slave->bond = bond; |
1da177e4 | 1615 | new_slave->dev = slave_dev; |
0b680e75 | 1616 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1617 | |
58402054 | 1618 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1619 | /* bond_alb_init_slave() must be called before all other stages since |
1620 | * it might fail and we do not want to have to undo everything | |
1621 | */ | |
1622 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1623 | if (res) |
569f0c4d | 1624 | goto err_close; |
1da177e4 LT |
1625 | } |
1626 | ||
303d1cbf JV |
1627 | /* If the mode USES_PRIMARY, then the following is handled by |
1628 | * bond_change_active_slave(). | |
1da177e4 LT |
1629 | */ |
1630 | if (!USES_PRIMARY(bond->params.mode)) { | |
1631 | /* set promiscuity level to new slave */ | |
1632 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1633 | res = dev_set_promiscuity(slave_dev, 1); |
1634 | if (res) | |
1635 | goto err_close; | |
1da177e4 LT |
1636 | } |
1637 | ||
1638 | /* set allmulti level to new slave */ | |
1639 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1640 | res = dev_set_allmulti(slave_dev, 1); |
1641 | if (res) | |
1642 | goto err_close; | |
1da177e4 LT |
1643 | } |
1644 | ||
b9e40857 | 1645 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1646 | |
1647 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1648 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1649 | ||
b9e40857 | 1650 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1651 | } |
1652 | ||
1653 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1654 | /* add lacpdu mc addr to mc list */ | |
1655 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1656 | ||
22bedad3 | 1657 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1658 | } |
1659 | ||
1660 | bond_add_vlans_on_slave(bond, slave_dev); | |
1661 | ||
1662 | write_lock_bh(&bond->lock); | |
1663 | ||
1664 | bond_attach_slave(bond, new_slave); | |
1665 | ||
1666 | new_slave->delay = 0; | |
1667 | new_slave->link_failure_count = 0; | |
1668 | ||
3915c1e8 JV |
1669 | write_unlock_bh(&bond->lock); |
1670 | ||
b2a103e6 MM |
1671 | bond_compute_features(bond); |
1672 | ||
876254ae VF |
1673 | bond_update_speed_duplex(new_slave); |
1674 | ||
3915c1e8 JV |
1675 | read_lock(&bond->lock); |
1676 | ||
f31c7937 MK |
1677 | new_slave->last_arp_rx = jiffies - |
1678 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
8599b52e VF |
1679 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
1680 | new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx; | |
f5b2b966 | 1681 | |
1da177e4 LT |
1682 | if (bond->params.miimon && !bond->params.use_carrier) { |
1683 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1684 | ||
1685 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1686 | /* | |
1687 | * miimon is set but a bonded network driver | |
1688 | * does not support ETHTOOL/MII and | |
1689 | * arp_interval is not set. Note: if | |
1690 | * use_carrier is enabled, we will never go | |
1691 | * here (because netif_carrier is always | |
1692 | * supported); thus, we don't need to change | |
1693 | * the messages for netif_carrier. | |
1694 | */ | |
a4aee5c8 | 1695 | 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 | 1696 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1697 | } else if (link_reporting == -1) { |
1698 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1699 | 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", |
1700 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1701 | } |
1702 | } | |
1703 | ||
1704 | /* check for initial state */ | |
f31c7937 MK |
1705 | if (bond->params.miimon) { |
1706 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1707 | if (bond->params.updelay) { | |
1708 | new_slave->link = BOND_LINK_BACK; | |
1709 | new_slave->delay = bond->params.updelay; | |
1710 | } else { | |
1711 | new_slave->link = BOND_LINK_UP; | |
1712 | } | |
1da177e4 | 1713 | } else { |
f31c7937 | 1714 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1715 | } |
f31c7937 MK |
1716 | } else if (bond->params.arp_interval) { |
1717 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1718 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1719 | } else { |
f31c7937 | 1720 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1721 | } |
1722 | ||
f31c7937 MK |
1723 | if (new_slave->link != BOND_LINK_DOWN) |
1724 | new_slave->jiffies = jiffies; | |
1725 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1726 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1727 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1728 | ||
1da177e4 LT |
1729 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1730 | /* if there is a primary slave, remember it */ | |
a549952a | 1731 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1732 | bond->primary_slave = new_slave; |
a549952a JP |
1733 | bond->force_primary = true; |
1734 | } | |
1da177e4 LT |
1735 | } |
1736 | ||
3915c1e8 JV |
1737 | write_lock_bh(&bond->curr_slave_lock); |
1738 | ||
1da177e4 LT |
1739 | switch (bond->params.mode) { |
1740 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1741 | bond_set_slave_inactive_flags(new_slave); |
1742 | bond_select_active_slave(bond); | |
1da177e4 LT |
1743 | break; |
1744 | case BOND_MODE_8023AD: | |
1745 | /* in 802.3ad mode, the internal mechanism | |
1746 | * will activate the slaves in the selected | |
1747 | * aggregator | |
1748 | */ | |
1749 | bond_set_slave_inactive_flags(new_slave); | |
1750 | /* if this is the first slave */ | |
1751 | if (bond->slave_cnt == 1) { | |
1752 | SLAVE_AD_INFO(new_slave).id = 1; | |
1753 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1754 | * can be called only after the mac address of the bond is set | |
1755 | */ | |
56d00c67 | 1756 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 LT |
1757 | } else { |
1758 | SLAVE_AD_INFO(new_slave).id = | |
1759 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1760 | } | |
1761 | ||
1762 | bond_3ad_bind_slave(new_slave); | |
1763 | break; | |
1764 | case BOND_MODE_TLB: | |
1765 | case BOND_MODE_ALB: | |
e30bc066 | 1766 | bond_set_active_slave(new_slave); |
059fe7a5 | 1767 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1768 | bond_select_active_slave(bond); |
1da177e4 LT |
1769 | break; |
1770 | default: | |
5a03cdb7 | 1771 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1772 | |
1773 | /* always active in trunk mode */ | |
e30bc066 | 1774 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1775 | |
1776 | /* In trunking mode there is little meaning to curr_active_slave | |
1777 | * anyway (it holds no special properties of the bond device), | |
1778 | * so we can change it without calling change_active_interface() | |
1779 | */ | |
4e591b93 | 1780 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
1da177e4 | 1781 | bond->curr_active_slave = new_slave; |
3d632c3f | 1782 | |
1da177e4 LT |
1783 | break; |
1784 | } /* switch(bond_mode) */ | |
1785 | ||
3915c1e8 JV |
1786 | write_unlock_bh(&bond->curr_slave_lock); |
1787 | ||
ff59c456 JV |
1788 | bond_set_carrier(bond); |
1789 | ||
f6dc31a8 | 1790 | #ifdef CONFIG_NET_POLL_CONTROLLER |
10eccb46 | 1791 | slave_dev->npinfo = bond->dev->npinfo; |
8a8efa22 AW |
1792 | if (slave_dev->npinfo) { |
1793 | if (slave_enable_netpoll(new_slave)) { | |
1794 | read_unlock(&bond->lock); | |
1795 | pr_info("Error, %s: master_dev is using netpoll, " | |
1796 | "but new slave device does not support netpoll.\n", | |
1797 | bond_dev->name); | |
1798 | res = -EBUSY; | |
f7d9821a | 1799 | goto err_detach; |
8a8efa22 | 1800 | } |
f6dc31a8 WC |
1801 | } |
1802 | #endif | |
8a8efa22 | 1803 | |
3915c1e8 | 1804 | read_unlock(&bond->lock); |
1da177e4 | 1805 | |
b76cdba9 MW |
1806 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1807 | if (res) | |
f7d9821a | 1808 | goto err_detach; |
b76cdba9 | 1809 | |
35d48903 JP |
1810 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1811 | new_slave); | |
1812 | if (res) { | |
1813 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1814 | goto err_dest_symlinks; | |
1815 | } | |
1816 | ||
a4aee5c8 JP |
1817 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1818 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1819 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1820 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1821 | |
1822 | /* enslave is successful */ | |
1823 | return 0; | |
1824 | ||
1825 | /* Undo stages on error */ | |
35d48903 JP |
1826 | err_dest_symlinks: |
1827 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1828 | ||
f7d9821a | 1829 | err_detach: |
303d1cbf JV |
1830 | if (!USES_PRIMARY(bond->params.mode)) |
1831 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1832 | ||
a506e7b4 | 1833 | bond_del_vlans_from_slave(bond, slave_dev); |
f7d9821a | 1834 | write_lock_bh(&bond->lock); |
1835 | bond_detach_slave(bond, new_slave); | |
3c5913b5 | 1836 | if (bond->primary_slave == new_slave) |
1837 | bond->primary_slave = NULL; | |
3c5913b5 | 1838 | if (bond->curr_active_slave == new_slave) { |
c6cdcf6d | 1839 | bond_change_active_slave(bond, NULL); |
1840 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1841 | read_lock(&bond->lock); |
1842 | write_lock_bh(&bond->curr_slave_lock); | |
3c5913b5 | 1843 | bond_select_active_slave(bond); |
1844 | write_unlock_bh(&bond->curr_slave_lock); | |
1845 | read_unlock(&bond->lock); | |
c6cdcf6d | 1846 | } else { |
1847 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1848 | } |
fc7a72ac | 1849 | slave_disable_netpoll(new_slave); |
f7d9821a | 1850 | |
1da177e4 | 1851 | err_close: |
b6a5a7b9 | 1852 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1853 | dev_close(slave_dev); |
1854 | ||
569f0c4d | 1855 | err_unset_master: |
471cb5a3 | 1856 | bond_upper_dev_unlink(bond_dev, slave_dev); |
569f0c4d | 1857 | |
1da177e4 | 1858 | err_restore_mac: |
dd957c57 | 1859 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1860 | /* XXX TODO - fom follow mode needs to change master's |
1861 | * MAC if this slave's MAC is in use by the bond, or at | |
1862 | * least print a warning. | |
1863 | */ | |
2ab82852 MS |
1864 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1865 | addr.sa_family = slave_dev->type; | |
1866 | dev_set_mac_address(slave_dev, &addr); | |
1867 | } | |
1da177e4 | 1868 | |
b15ba0fb JP |
1869 | err_restore_mtu: |
1870 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1871 | ||
1da177e4 LT |
1872 | err_free: |
1873 | kfree(new_slave); | |
1874 | ||
1875 | err_undo_flags: | |
b2a103e6 | 1876 | bond_compute_features(bond); |
b8fad459 NA |
1877 | /* Enslave of first slave has failed and we need to fix master's mac */ |
1878 | if (bond->slave_cnt == 0 && | |
1879 | ether_addr_equal(bond_dev->dev_addr, slave_dev->dev_addr)) | |
1880 | eth_hw_addr_random(bond_dev); | |
3d632c3f | 1881 | |
1da177e4 LT |
1882 | return res; |
1883 | } | |
1884 | ||
1885 | /* | |
1886 | * Try to release the slave device <slave> from the bond device <master> | |
1887 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1888 | * is write-locked. If "all" is true it means that the function is being called |
1889 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1890 | * |
1891 | * The rules for slave state should be: | |
1892 | * for Active/Backup: | |
1893 | * Active stays on all backups go down | |
1894 | * for Bonded connections: | |
1895 | * The first up interface should be left on and all others downed. | |
1896 | */ | |
0896341a | 1897 | static int __bond_release_one(struct net_device *bond_dev, |
1898 | struct net_device *slave_dev, | |
1899 | bool all) | |
1da177e4 | 1900 | { |
454d7c9b | 1901 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1902 | struct slave *slave, *oldcurrent; |
1903 | struct sockaddr addr; | |
c8f44aff | 1904 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1905 | |
1906 | /* slave is not a slave or master is not master of this slave */ | |
1907 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1908 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1909 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1910 | bond_dev->name, slave_dev->name); |
1911 | return -EINVAL; | |
1912 | } | |
1913 | ||
e843fa50 | 1914 | block_netpoll_tx(); |
1da177e4 LT |
1915 | write_lock_bh(&bond->lock); |
1916 | ||
1917 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1918 | if (!slave) { | |
1919 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1920 | pr_info("%s: %s not enslaved\n", |
1921 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1922 | write_unlock_bh(&bond->lock); |
e843fa50 | 1923 | unblock_netpoll_tx(); |
1da177e4 LT |
1924 | return -EINVAL; |
1925 | } | |
1926 | ||
fcd99434 | 1927 | write_unlock_bh(&bond->lock); |
35d48903 JP |
1928 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1929 | * for this slave anymore. | |
1930 | */ | |
1931 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1932 | write_lock_bh(&bond->lock); |
1933 | ||
0896341a | 1934 | if (!all && !bond->params.fail_over_mac) { |
2e42e474 | 1935 | if (ether_addr_equal(bond_dev->dev_addr, slave->perm_hwaddr) && |
8e95a202 | 1936 | bond->slave_cnt > 1) |
a4aee5c8 JP |
1937 | pr_warning("%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", |
1938 | bond_dev->name, slave_dev->name, | |
1939 | slave->perm_hwaddr, | |
1940 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1941 | } |
1942 | ||
1943 | /* Inform AD package of unbinding of slave. */ | |
1944 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1945 | /* must be called before the slave is | |
1946 | * detached from the list | |
1947 | */ | |
1948 | bond_3ad_unbind_slave(slave); | |
1949 | } | |
1950 | ||
a4aee5c8 JP |
1951 | pr_info("%s: releasing %s interface %s\n", |
1952 | bond_dev->name, | |
e30bc066 | 1953 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1954 | slave_dev->name); |
1da177e4 LT |
1955 | |
1956 | oldcurrent = bond->curr_active_slave; | |
1957 | ||
1958 | bond->current_arp_slave = NULL; | |
1959 | ||
1960 | /* release the slave from its bond */ | |
1961 | bond_detach_slave(bond, slave); | |
1962 | ||
3d632c3f | 1963 | if (bond->primary_slave == slave) |
1da177e4 | 1964 | bond->primary_slave = NULL; |
1da177e4 | 1965 | |
3d632c3f | 1966 | if (oldcurrent == slave) |
1da177e4 | 1967 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1968 | |
58402054 | 1969 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1970 | /* Must be called only after the slave has been |
1971 | * detached from the list and the curr_active_slave | |
1972 | * has been cleared (if our_slave == old_current), | |
1973 | * but before a new active slave is selected. | |
1974 | */ | |
2543331d | 1975 | write_unlock_bh(&bond->lock); |
1da177e4 | 1976 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1977 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1978 | } |
1979 | ||
0896341a | 1980 | if (all) { |
1981 | bond->curr_active_slave = NULL; | |
1982 | } else if (oldcurrent == slave) { | |
059fe7a5 JV |
1983 | /* |
1984 | * Note that we hold RTNL over this sequence, so there | |
1985 | * is no concern that another slave add/remove event | |
1986 | * will interfere. | |
1987 | */ | |
1988 | write_unlock_bh(&bond->lock); | |
1989 | read_lock(&bond->lock); | |
1990 | write_lock_bh(&bond->curr_slave_lock); | |
1991 | ||
1da177e4 LT |
1992 | bond_select_active_slave(bond); |
1993 | ||
059fe7a5 JV |
1994 | write_unlock_bh(&bond->curr_slave_lock); |
1995 | read_unlock(&bond->lock); | |
1996 | write_lock_bh(&bond->lock); | |
1997 | } | |
1998 | ||
1da177e4 | 1999 | if (bond->slave_cnt == 0) { |
ff59c456 | 2000 | bond_set_carrier(bond); |
409cc1f8 | 2001 | eth_hw_addr_random(bond_dev); |
1da177e4 | 2002 | |
cc0e4070 | 2003 | if (bond_vlan_used(bond)) { |
a4aee5c8 JP |
2004 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2005 | bond_dev->name, bond_dev->name); | |
2006 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2007 | bond_dev->name); | |
1da177e4 | 2008 | } |
1da177e4 LT |
2009 | } |
2010 | ||
2011 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2012 | unblock_netpoll_tx(); |
1da177e4 | 2013 | |
80028ea1 | 2014 | if (bond->slave_cnt == 0) { |
2af73d4b | 2015 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
2016 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
2017 | } | |
2af73d4b | 2018 | |
b2a103e6 MM |
2019 | bond_compute_features(bond); |
2020 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2021 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
2022 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
2023 | bond_dev->name, slave_dev->name, bond_dev->name); | |
2024 | ||
b76cdba9 MW |
2025 | /* must do this from outside any spinlocks */ |
2026 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2027 | ||
1da177e4 LT |
2028 | bond_del_vlans_from_slave(bond, slave_dev); |
2029 | ||
303d1cbf JV |
2030 | /* If the mode USES_PRIMARY, then this cases was handled above by |
2031 | * bond_change_active_slave(..., NULL) | |
1da177e4 LT |
2032 | */ |
2033 | if (!USES_PRIMARY(bond->params.mode)) { | |
2034 | /* unset promiscuity level from slave */ | |
3d632c3f | 2035 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2036 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2037 | |
2038 | /* unset allmulti level from slave */ | |
3d632c3f | 2039 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2040 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 2041 | |
303d1cbf | 2042 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
2043 | } |
2044 | ||
471cb5a3 | 2045 | bond_upper_dev_unlink(bond_dev, slave_dev); |
1da177e4 | 2046 | |
8a8efa22 | 2047 | slave_disable_netpoll(slave); |
f6dc31a8 | 2048 | |
1da177e4 LT |
2049 | /* close slave before restoring its mac address */ |
2050 | dev_close(slave_dev); | |
2051 | ||
3915c1e8 | 2052 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2053 | /* restore original ("permanent") mac address */ |
2054 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2055 | addr.sa_family = slave_dev->type; | |
2056 | dev_set_mac_address(slave_dev, &addr); | |
2057 | } | |
1da177e4 | 2058 | |
b15ba0fb JP |
2059 | dev_set_mtu(slave_dev, slave->original_mtu); |
2060 | ||
2d7011ca | 2061 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2062 | |
2063 | kfree(slave); | |
2064 | ||
2065 | return 0; /* deletion OK */ | |
2066 | } | |
2067 | ||
0896341a | 2068 | /* A wrapper used because of ndo_del_link */ |
2069 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
2070 | { | |
2071 | return __bond_release_one(bond_dev, slave_dev, false); | |
2072 | } | |
2073 | ||
d90a162a | 2074 | /* |
dadaa10b | 2075 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2076 | * Must be under rtnl_lock when this function is called. |
2077 | */ | |
26d8ee75 | 2078 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2079 | struct net_device *slave_dev) | |
d90a162a | 2080 | { |
454d7c9b | 2081 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2082 | int ret; |
2083 | ||
2084 | ret = bond_release(bond_dev, slave_dev); | |
2085 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2086 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2087 | pr_info("%s: destroying bond %s.\n", |
2088 | bond_dev->name, bond_dev->name); | |
9e71626c | 2089 | unregister_netdevice(bond_dev); |
d90a162a MS |
2090 | } |
2091 | return ret; | |
2092 | } | |
2093 | ||
1da177e4 LT |
2094 | /* |
2095 | * This function changes the active slave to slave <slave_dev>. | |
2096 | * It returns -EINVAL in the following cases. | |
2097 | * - <slave_dev> is not found in the list. | |
2098 | * - There is not active slave now. | |
2099 | * - <slave_dev> is already active. | |
2100 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2101 | * - <slave_dev> is not running. | |
3d632c3f SH |
2102 | * In these cases, this function does nothing. |
2103 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2104 | */ |
2105 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2106 | { | |
454d7c9b | 2107 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2108 | struct slave *old_active = NULL; |
2109 | struct slave *new_active = NULL; | |
2110 | int res = 0; | |
2111 | ||
3d632c3f | 2112 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2113 | return -EINVAL; |
1da177e4 | 2114 | |
471cb5a3 JP |
2115 | /* Verify that bond_dev is indeed the master of slave_dev */ |
2116 | if (!(slave_dev->flags & IFF_SLAVE) || | |
2117 | !netdev_has_upper_dev(slave_dev, bond_dev)) | |
1da177e4 | 2118 | return -EINVAL; |
1da177e4 | 2119 | |
059fe7a5 | 2120 | read_lock(&bond->lock); |
1da177e4 | 2121 | |
059fe7a5 | 2122 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2123 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2124 | read_unlock(&bond->curr_slave_lock); |
2125 | ||
1da177e4 LT |
2126 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2127 | ||
2128 | /* | |
2129 | * Changing to the current active: do nothing; return success. | |
2130 | */ | |
2131 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2132 | read_unlock(&bond->lock); |
1da177e4 LT |
2133 | return 0; |
2134 | } | |
2135 | ||
2136 | if ((new_active) && | |
2137 | (old_active) && | |
2138 | (new_active->link == BOND_LINK_UP) && | |
2139 | IS_UP(new_active->dev)) { | |
e843fa50 | 2140 | block_netpoll_tx(); |
059fe7a5 | 2141 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2142 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2143 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2144 | unblock_netpoll_tx(); |
3d632c3f | 2145 | } else |
1da177e4 | 2146 | res = -EINVAL; |
1da177e4 | 2147 | |
059fe7a5 | 2148 | read_unlock(&bond->lock); |
1da177e4 LT |
2149 | |
2150 | return res; | |
2151 | } | |
2152 | ||
1da177e4 LT |
2153 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2154 | { | |
454d7c9b | 2155 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2156 | |
2157 | info->bond_mode = bond->params.mode; | |
2158 | info->miimon = bond->params.miimon; | |
2159 | ||
6603a6f2 | 2160 | read_lock(&bond->lock); |
1da177e4 | 2161 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2162 | read_unlock(&bond->lock); |
1da177e4 LT |
2163 | |
2164 | return 0; | |
2165 | } | |
2166 | ||
2167 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2168 | { | |
454d7c9b | 2169 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2170 | struct slave *slave; |
689c96cc | 2171 | int i, res = -ENODEV; |
1da177e4 | 2172 | |
6603a6f2 | 2173 | read_lock(&bond->lock); |
1da177e4 LT |
2174 | |
2175 | bond_for_each_slave(bond, slave, i) { | |
2176 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2177 | res = 0; |
2178 | strcpy(info->slave_name, slave->dev->name); | |
2179 | info->link = slave->link; | |
e30bc066 | 2180 | info->state = bond_slave_state(slave); |
689c96cc | 2181 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2182 | break; |
2183 | } | |
2184 | } | |
2185 | ||
6603a6f2 | 2186 | read_unlock(&bond->lock); |
1da177e4 | 2187 | |
689c96cc | 2188 | return res; |
1da177e4 LT |
2189 | } |
2190 | ||
2191 | /*-------------------------------- Monitoring -------------------------------*/ | |
2192 | ||
1da177e4 | 2193 | |
f0c76d61 JV |
2194 | static int bond_miimon_inspect(struct bonding *bond) |
2195 | { | |
2196 | struct slave *slave; | |
2197 | int i, link_state, commit = 0; | |
41f89100 JP |
2198 | bool ignore_updelay; |
2199 | ||
2200 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2201 | |
2202 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2203 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2204 | |
f0c76d61 | 2205 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2206 | |
2207 | switch (slave->link) { | |
f0c76d61 JV |
2208 | case BOND_LINK_UP: |
2209 | if (link_state) | |
2210 | continue; | |
1da177e4 | 2211 | |
f0c76d61 JV |
2212 | slave->link = BOND_LINK_FAIL; |
2213 | slave->delay = bond->params.downdelay; | |
2214 | if (slave->delay) { | |
a4aee5c8 JP |
2215 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2216 | bond->dev->name, | |
2217 | (bond->params.mode == | |
2218 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2219 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2220 | "active " : "backup ") : "", |
2221 | slave->dev->name, | |
2222 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2223 | } |
f0c76d61 JV |
2224 | /*FALLTHRU*/ |
2225 | case BOND_LINK_FAIL: | |
2226 | if (link_state) { | |
2227 | /* | |
2228 | * recovered before downdelay expired | |
2229 | */ | |
2230 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2231 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2232 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2233 | bond->dev->name, | |
2234 | (bond->params.downdelay - slave->delay) * | |
2235 | bond->params.miimon, | |
2236 | slave->dev->name); | |
f0c76d61 | 2237 | continue; |
1da177e4 | 2238 | } |
f0c76d61 JV |
2239 | |
2240 | if (slave->delay <= 0) { | |
2241 | slave->new_link = BOND_LINK_DOWN; | |
2242 | commit++; | |
2243 | continue; | |
1da177e4 | 2244 | } |
1da177e4 | 2245 | |
f0c76d61 JV |
2246 | slave->delay--; |
2247 | break; | |
2248 | ||
2249 | case BOND_LINK_DOWN: | |
2250 | if (!link_state) | |
2251 | continue; | |
2252 | ||
2253 | slave->link = BOND_LINK_BACK; | |
2254 | slave->delay = bond->params.updelay; | |
2255 | ||
2256 | if (slave->delay) { | |
a4aee5c8 JP |
2257 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2258 | bond->dev->name, slave->dev->name, | |
2259 | ignore_updelay ? 0 : | |
2260 | bond->params.updelay * | |
2261 | bond->params.miimon); | |
f0c76d61 JV |
2262 | } |
2263 | /*FALLTHRU*/ | |
2264 | case BOND_LINK_BACK: | |
2265 | if (!link_state) { | |
2266 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2267 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2268 | bond->dev->name, | |
2269 | (bond->params.updelay - slave->delay) * | |
2270 | bond->params.miimon, | |
2271 | slave->dev->name); | |
f0c76d61 JV |
2272 | |
2273 | continue; | |
2274 | } | |
2275 | ||
41f89100 JP |
2276 | if (ignore_updelay) |
2277 | slave->delay = 0; | |
2278 | ||
f0c76d61 JV |
2279 | if (slave->delay <= 0) { |
2280 | slave->new_link = BOND_LINK_UP; | |
2281 | commit++; | |
41f89100 | 2282 | ignore_updelay = false; |
f0c76d61 | 2283 | continue; |
1da177e4 | 2284 | } |
f0c76d61 JV |
2285 | |
2286 | slave->delay--; | |
1da177e4 | 2287 | break; |
f0c76d61 JV |
2288 | } |
2289 | } | |
1da177e4 | 2290 | |
f0c76d61 JV |
2291 | return commit; |
2292 | } | |
1da177e4 | 2293 | |
f0c76d61 JV |
2294 | static void bond_miimon_commit(struct bonding *bond) |
2295 | { | |
2296 | struct slave *slave; | |
2297 | int i; | |
2298 | ||
2299 | bond_for_each_slave(bond, slave, i) { | |
2300 | switch (slave->new_link) { | |
2301 | case BOND_LINK_NOCHANGE: | |
2302 | continue; | |
1da177e4 | 2303 | |
f0c76d61 JV |
2304 | case BOND_LINK_UP: |
2305 | slave->link = BOND_LINK_UP; | |
2306 | slave->jiffies = jiffies; | |
2307 | ||
2308 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2309 | /* prevent it from being the active one */ | |
e30bc066 | 2310 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2311 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2312 | /* make it immediately active */ | |
e30bc066 | 2313 | bond_set_active_slave(slave); |
f0c76d61 JV |
2314 | } else if (slave != bond->primary_slave) { |
2315 | /* prevent it from being the active one */ | |
e30bc066 | 2316 | bond_set_backup_slave(slave); |
1da177e4 | 2317 | } |
1da177e4 | 2318 | |
5d30530e | 2319 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 | 2320 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
2321 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
2322 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2323 | |
f0c76d61 JV |
2324 | /* notify ad that the link status has changed */ |
2325 | if (bond->params.mode == BOND_MODE_8023AD) | |
2326 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2327 | |
58402054 | 2328 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2329 | bond_alb_handle_link_change(bond, slave, |
2330 | BOND_LINK_UP); | |
1da177e4 | 2331 | |
f0c76d61 JV |
2332 | if (!bond->curr_active_slave || |
2333 | (slave == bond->primary_slave)) | |
2334 | goto do_failover; | |
1da177e4 | 2335 | |
f0c76d61 | 2336 | continue; |
059fe7a5 | 2337 | |
f0c76d61 | 2338 | case BOND_LINK_DOWN: |
fba4acda JV |
2339 | if (slave->link_failure_count < UINT_MAX) |
2340 | slave->link_failure_count++; | |
2341 | ||
f0c76d61 | 2342 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2343 | |
f0c76d61 JV |
2344 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2345 | bond->params.mode == BOND_MODE_8023AD) | |
2346 | bond_set_slave_inactive_flags(slave); | |
2347 | ||
a4aee5c8 JP |
2348 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2349 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2350 | |
2351 | if (bond->params.mode == BOND_MODE_8023AD) | |
2352 | bond_3ad_handle_link_change(slave, | |
2353 | BOND_LINK_DOWN); | |
2354 | ||
ae63e808 | 2355 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2356 | bond_alb_handle_link_change(bond, slave, |
2357 | BOND_LINK_DOWN); | |
2358 | ||
2359 | if (slave == bond->curr_active_slave) | |
2360 | goto do_failover; | |
2361 | ||
2362 | continue; | |
2363 | ||
2364 | default: | |
a4aee5c8 | 2365 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2366 | bond->dev->name, slave->new_link, |
2367 | slave->dev->name); | |
2368 | slave->new_link = BOND_LINK_NOCHANGE; | |
2369 | ||
2370 | continue; | |
2371 | } | |
2372 | ||
2373 | do_failover: | |
2374 | ASSERT_RTNL(); | |
e843fa50 | 2375 | block_netpoll_tx(); |
f0c76d61 JV |
2376 | write_lock_bh(&bond->curr_slave_lock); |
2377 | bond_select_active_slave(bond); | |
2378 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2379 | unblock_netpoll_tx(); |
f0c76d61 JV |
2380 | } |
2381 | ||
2382 | bond_set_carrier(bond); | |
1da177e4 LT |
2383 | } |
2384 | ||
0b0eef66 JV |
2385 | /* |
2386 | * bond_mii_monitor | |
2387 | * | |
2388 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2389 | * inspection, then (if inspection indicates something needs to be done) |
2390 | * an acquisition of appropriate locks followed by a commit phase to | |
2391 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2392 | */ |
2393 | void bond_mii_monitor(struct work_struct *work) | |
2394 | { | |
2395 | struct bonding *bond = container_of(work, struct bonding, | |
2396 | mii_work.work); | |
ad246c99 | 2397 | bool should_notify_peers = false; |
e6d265e8 | 2398 | unsigned long delay; |
0b0eef66 JV |
2399 | |
2400 | read_lock(&bond->lock); | |
e6d265e8 JV |
2401 | |
2402 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 JV |
2403 | |
2404 | if (bond->slave_cnt == 0) | |
2405 | goto re_arm; | |
b59f9f74 | 2406 | |
ad246c99 BH |
2407 | should_notify_peers = bond_should_notify_peers(bond); |
2408 | ||
f0c76d61 | 2409 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2410 | read_unlock(&bond->lock); |
e6d265e8 JV |
2411 | |
2412 | /* Race avoidance with bond_close cancel of workqueue */ | |
2413 | if (!rtnl_trylock()) { | |
2414 | read_lock(&bond->lock); | |
2415 | delay = 1; | |
2416 | should_notify_peers = false; | |
2417 | goto re_arm; | |
2418 | } | |
2419 | ||
0b0eef66 | 2420 | read_lock(&bond->lock); |
f0c76d61 JV |
2421 | |
2422 | bond_miimon_commit(bond); | |
2423 | ||
5655662d JV |
2424 | read_unlock(&bond->lock); |
2425 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2426 | read_lock(&bond->lock); | |
0b0eef66 JV |
2427 | } |
2428 | ||
f0c76d61 | 2429 | re_arm: |
e6d265e8 JV |
2430 | if (bond->params.miimon) |
2431 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2432 | ||
0b0eef66 | 2433 | read_unlock(&bond->lock); |
ad246c99 BH |
2434 | |
2435 | if (should_notify_peers) { | |
e6d265e8 JV |
2436 | if (!rtnl_trylock()) { |
2437 | read_lock(&bond->lock); | |
2438 | bond->send_peer_notif++; | |
2439 | read_unlock(&bond->lock); | |
2440 | return; | |
2441 | } | |
b7bc2a5b | 2442 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2443 | rtnl_unlock(); |
2444 | } | |
0b0eef66 | 2445 | } |
c3ade5ca | 2446 | |
d3bb52b0 | 2447 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2448 | { |
0883beca | 2449 | struct vlan_entry *vlan; |
eaddcd76 | 2450 | struct net_device *vlan_dev; |
f5b2b966 | 2451 | |
eaddcd76 | 2452 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
f5b2b966 JV |
2453 | return 1; |
2454 | ||
0883beca | 2455 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
eaddcd76 | 2456 | rcu_read_lock(); |
1fd9b1fc PM |
2457 | vlan_dev = __vlan_find_dev_deep(bond->dev, htons(ETH_P_8021Q), |
2458 | vlan->vlan_id); | |
eaddcd76 AG |
2459 | rcu_read_unlock(); |
2460 | if (vlan_dev && ip == bond_confirm_addr(vlan_dev, 0, ip)) | |
f5b2b966 JV |
2461 | return 1; |
2462 | } | |
2463 | ||
2464 | return 0; | |
2465 | } | |
2466 | ||
c3ade5ca JV |
2467 | /* |
2468 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2469 | * switches in VLAN mode (especially if ports are configured as | |
2470 | * "native" to a VLAN) might not pass non-tagged frames. | |
2471 | */ | |
d3bb52b0 | 2472 | 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 |
2473 | { |
2474 | struct sk_buff *skb; | |
2475 | ||
acca2674 | 2476 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op, |
2477 | slave_dev->name, &dest_ip, &src_ip, vlan_id); | |
3d632c3f | 2478 | |
c3ade5ca JV |
2479 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2480 | NULL, slave_dev->dev_addr, NULL); | |
2481 | ||
2482 | if (!skb) { | |
a4aee5c8 | 2483 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2484 | return; |
2485 | } | |
2486 | if (vlan_id) { | |
1fd9b1fc | 2487 | skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id); |
c3ade5ca | 2488 | if (!skb) { |
a4aee5c8 | 2489 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2490 | return; |
2491 | } | |
2492 | } | |
2493 | arp_xmit(skb); | |
2494 | } | |
2495 | ||
2496 | ||
1da177e4 LT |
2497 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2498 | { | |
b23dd4fe | 2499 | int i, vlan_id; |
d3bb52b0 | 2500 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2501 | struct vlan_entry *vlan; |
eaddcd76 | 2502 | struct net_device *vlan_dev = NULL; |
c3ade5ca | 2503 | struct rtable *rt; |
1da177e4 | 2504 | |
6b780567 | 2505 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
eaddcd76 | 2506 | __be32 addr; |
6b780567 | 2507 | if (!targets[i]) |
5a31bec0 | 2508 | break; |
acca2674 | 2509 | pr_debug("basa: target %pI4\n", &targets[i]); |
cc0e4070 | 2510 | if (!bond_vlan_used(bond)) { |
5a03cdb7 | 2511 | pr_debug("basa: empty vlan: arp_send\n"); |
eaddcd76 | 2512 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2513 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2514 | addr, 0); |
c3ade5ca JV |
2515 | continue; |
2516 | } | |
2517 | ||
2518 | /* | |
2519 | * If VLANs are configured, we do a route lookup to | |
2520 | * determine which VLAN interface would be used, so we | |
2521 | * can tag the ARP with the proper VLAN tag. | |
2522 | */ | |
78fbfd8a DM |
2523 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2524 | RTO_ONLINK, 0); | |
b23dd4fe | 2525 | if (IS_ERR(rt)) { |
c3ade5ca | 2526 | if (net_ratelimit()) { |
a4aee5c8 | 2527 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2528 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2529 | } |
2530 | continue; | |
2531 | } | |
2532 | ||
2533 | /* | |
2534 | * This target is not on a VLAN | |
2535 | */ | |
d8d1f30b | 2536 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2537 | ip_rt_put(rt); |
5a03cdb7 | 2538 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
eaddcd76 | 2539 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2540 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2541 | addr, 0); |
c3ade5ca JV |
2542 | continue; |
2543 | } | |
2544 | ||
2545 | vlan_id = 0; | |
0883beca | 2546 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
cc0e4070 JP |
2547 | rcu_read_lock(); |
2548 | vlan_dev = __vlan_find_dev_deep(bond->dev, | |
1fd9b1fc | 2549 | htons(ETH_P_8021Q), |
cc0e4070 JP |
2550 | vlan->vlan_id); |
2551 | rcu_read_unlock(); | |
d8d1f30b | 2552 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2553 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2554 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2555 | vlan_dev->name, vlan_id); |
2556 | break; | |
2557 | } | |
2558 | } | |
2559 | ||
eaddcd76 | 2560 | if (vlan_id && vlan_dev) { |
ed4b9f80 | 2561 | ip_rt_put(rt); |
eaddcd76 | 2562 | addr = bond_confirm_addr(vlan_dev, targets[i], 0); |
c3ade5ca | 2563 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2564 | addr, vlan_id); |
c3ade5ca JV |
2565 | continue; |
2566 | } | |
2567 | ||
2568 | if (net_ratelimit()) { | |
a4aee5c8 | 2569 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2570 | bond->dev->name, &targets[i], |
d8d1f30b | 2571 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2572 | } |
ed4b9f80 | 2573 | ip_rt_put(rt); |
c3ade5ca JV |
2574 | } |
2575 | } | |
2576 | ||
d3bb52b0 | 2577 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2578 | { |
8599b52e VF |
2579 | int i; |
2580 | ||
87a7b84b VF |
2581 | if (!sip || !bond_has_this_ip(bond, tip)) { |
2582 | pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip); | |
2583 | return; | |
2584 | } | |
f5b2b966 | 2585 | |
8599b52e VF |
2586 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2587 | if (i == -1) { | |
87a7b84b VF |
2588 | pr_debug("bva: sip %pI4 not found in targets\n", &sip); |
2589 | return; | |
f5b2b966 | 2590 | } |
87a7b84b | 2591 | slave->last_arp_rx = jiffies; |
8599b52e | 2592 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2593 | } |
2594 | ||
de063b70 ED |
2595 | static int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2596 | struct slave *slave) | |
f5b2b966 | 2597 | { |
de063b70 | 2598 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2599 | unsigned char *arp_ptr; |
d3bb52b0 | 2600 | __be32 sip, tip; |
de063b70 | 2601 | int alen; |
f5b2b966 | 2602 | |
3aba891d | 2603 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2604 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2605 | |
f5b2b966 | 2606 | read_lock(&bond->lock); |
2c146102 VF |
2607 | |
2608 | if (!slave_do_arp_validate(bond, slave)) | |
2609 | goto out_unlock; | |
2610 | ||
de063b70 | 2611 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2612 | |
3aba891d JP |
2613 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2614 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2615 | |
de063b70 ED |
2616 | if (alen > skb_headlen(skb)) { |
2617 | arp = kmalloc(alen, GFP_ATOMIC); | |
2618 | if (!arp) | |
2619 | goto out_unlock; | |
2620 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2621 | goto out_unlock; | |
2622 | } | |
f5b2b966 | 2623 | |
3aba891d | 2624 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2625 | skb->pkt_type == PACKET_OTHERHOST || |
2626 | skb->pkt_type == PACKET_LOOPBACK || | |
2627 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2628 | arp->ar_pro != htons(ETH_P_IP) || | |
2629 | arp->ar_pln != 4) | |
2630 | goto out_unlock; | |
2631 | ||
2632 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2633 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2634 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2635 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2636 | memcpy(&tip, arp_ptr, 4); |
2637 | ||
5a03cdb7 | 2638 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2639 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2640 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2641 | &sip, &tip); | |
f5b2b966 JV |
2642 | |
2643 | /* | |
2644 | * Backup slaves won't see the ARP reply, but do come through | |
2645 | * here for each ARP probe (so we swap the sip/tip to validate | |
2646 | * the probe). In a "redundant switch, common router" type of | |
2647 | * configuration, the ARP probe will (hopefully) travel from | |
2648 | * the active, through one switch, the router, then the other | |
2649 | * switch before reaching the backup. | |
aeea64ac VF |
2650 | * |
2651 | * We 'trust' the arp requests if there is an active slave and | |
2652 | * it received valid arp reply(s) after it became active. This | |
2653 | * is done to avoid endless looping when we can't reach the | |
2654 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2655 | */ |
e30bc066 | 2656 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2657 | bond_validate_arp(bond, slave, sip, tip); |
aeea64ac VF |
2658 | else if (bond->curr_active_slave && |
2659 | time_after(slave_last_rx(bond, bond->curr_active_slave), | |
2660 | bond->curr_active_slave->jiffies)) | |
f5b2b966 JV |
2661 | bond_validate_arp(bond, slave, tip, sip); |
2662 | ||
2663 | out_unlock: | |
2664 | read_unlock(&bond->lock); | |
de063b70 ED |
2665 | if (arp != (struct arphdr *)skb->data) |
2666 | kfree(arp); | |
13a8e0c8 | 2667 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2668 | } |
2669 | ||
1da177e4 LT |
2670 | /* |
2671 | * this function is called regularly to monitor each slave's link | |
2672 | * ensuring that traffic is being sent and received when arp monitoring | |
2673 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2674 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2675 | * arp monitoring in active backup mode. | |
2676 | */ | |
1b76b316 | 2677 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2678 | { |
1b76b316 JV |
2679 | struct bonding *bond = container_of(work, struct bonding, |
2680 | arp_work.work); | |
1da177e4 LT |
2681 | struct slave *slave, *oldcurrent; |
2682 | int do_failover = 0; | |
da210f55 | 2683 | int delta_in_ticks, extra_ticks; |
1da177e4 LT |
2684 | int i; |
2685 | ||
2686 | read_lock(&bond->lock); | |
2687 | ||
5ce0da8f | 2688 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
da210f55 | 2689 | extra_ticks = delta_in_ticks / 2; |
1da177e4 | 2690 | |
3d632c3f | 2691 | if (bond->slave_cnt == 0) |
1da177e4 | 2692 | goto re_arm; |
1da177e4 LT |
2693 | |
2694 | read_lock(&bond->curr_slave_lock); | |
2695 | oldcurrent = bond->curr_active_slave; | |
2696 | read_unlock(&bond->curr_slave_lock); | |
2697 | ||
2698 | /* see if any of the previous devices are up now (i.e. they have | |
2699 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2700 | * the picture unless it is null. also, slave->jiffies is not needed | |
2701 | * here because we send an arp on each slave and give a slave as | |
2702 | * long as it needs to get the tx/rx within the delta. | |
2703 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2704 | * so it can wait | |
2705 | */ | |
2706 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2707 | unsigned long trans_start = dev_trans_start(slave->dev); |
2708 | ||
1da177e4 | 2709 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2710 | if (time_in_range(jiffies, |
2711 | trans_start - delta_in_ticks, | |
da210f55 | 2712 | trans_start + delta_in_ticks + extra_ticks) && |
cb32f2a0 JB |
2713 | time_in_range(jiffies, |
2714 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2715 | slave->dev->last_rx + delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2716 | |
2717 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2718 | bond_set_active_slave(slave); |
1da177e4 LT |
2719 | |
2720 | /* primary_slave has no meaning in round-robin | |
2721 | * mode. the window of a slave being up and | |
2722 | * curr_active_slave being null after enslaving | |
2723 | * is closed. | |
2724 | */ | |
2725 | if (!oldcurrent) { | |
a4aee5c8 JP |
2726 | pr_info("%s: link status definitely up for interface %s, ", |
2727 | bond->dev->name, | |
2728 | slave->dev->name); | |
1da177e4 LT |
2729 | do_failover = 1; |
2730 | } else { | |
a4aee5c8 JP |
2731 | pr_info("%s: interface %s is now up\n", |
2732 | bond->dev->name, | |
2733 | slave->dev->name); | |
1da177e4 LT |
2734 | } |
2735 | } | |
2736 | } else { | |
2737 | /* slave->link == BOND_LINK_UP */ | |
2738 | ||
2739 | /* not all switches will respond to an arp request | |
2740 | * when the source ip is 0, so don't take the link down | |
2741 | * if we don't know our ip yet | |
2742 | */ | |
cb32f2a0 JB |
2743 | if (!time_in_range(jiffies, |
2744 | trans_start - delta_in_ticks, | |
da210f55 | 2745 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2746 | !time_in_range(jiffies, |
2747 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2748 | slave->dev->last_rx + 2 * delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2749 | |
2750 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2751 | bond_set_backup_slave(slave); |
1da177e4 | 2752 | |
3d632c3f | 2753 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2754 | slave->link_failure_count++; |
1da177e4 | 2755 | |
a4aee5c8 JP |
2756 | pr_info("%s: interface %s is now down.\n", |
2757 | bond->dev->name, | |
2758 | slave->dev->name); | |
1da177e4 | 2759 | |
3d632c3f | 2760 | if (slave == oldcurrent) |
1da177e4 | 2761 | do_failover = 1; |
1da177e4 LT |
2762 | } |
2763 | } | |
2764 | ||
2765 | /* note: if switch is in round-robin mode, all links | |
2766 | * must tx arp to ensure all links rx an arp - otherwise | |
2767 | * links may oscillate or not come up at all; if switch is | |
2768 | * in something like xor mode, there is nothing we can | |
2769 | * do - all replies will be rx'ed on same link causing slaves | |
2770 | * to be unstable during low/no traffic periods | |
2771 | */ | |
3d632c3f | 2772 | if (IS_UP(slave->dev)) |
1da177e4 | 2773 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2774 | } |
2775 | ||
2776 | if (do_failover) { | |
e843fa50 | 2777 | block_netpoll_tx(); |
059fe7a5 | 2778 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2779 | |
2780 | bond_select_active_slave(bond); | |
2781 | ||
059fe7a5 | 2782 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2783 | unblock_netpoll_tx(); |
1da177e4 LT |
2784 | } |
2785 | ||
2786 | re_arm: | |
e6d265e8 | 2787 | if (bond->params.arp_interval) |
1b76b316 | 2788 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 2789 | |
1da177e4 LT |
2790 | read_unlock(&bond->lock); |
2791 | } | |
2792 | ||
2793 | /* | |
b2220cad JV |
2794 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2795 | * changes. Sets new_link in slaves to specify what action should take | |
2796 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2797 | * to link states must be committed. | |
2798 | * | |
2799 | * Called with bond->lock held for read. | |
1da177e4 | 2800 | */ |
b2220cad | 2801 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2802 | { |
1da177e4 | 2803 | struct slave *slave; |
b2220cad | 2804 | int i, commit = 0; |
cb32f2a0 | 2805 | unsigned long trans_start; |
da210f55 JB |
2806 | int extra_ticks; |
2807 | ||
2808 | /* All the time comparisons below need some extra time. Otherwise, on | |
2809 | * fast networks the ARP probe/reply may arrive within the same jiffy | |
2810 | * as it was sent. Then, the next time the ARP monitor is run, one | |
2811 | * arp_interval will already have passed in the comparisons. | |
2812 | */ | |
2813 | extra_ticks = delta_in_ticks / 2; | |
1da177e4 | 2814 | |
b2220cad JV |
2815 | bond_for_each_slave(bond, slave, i) { |
2816 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2817 | |
b2220cad | 2818 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2819 | if (time_in_range(jiffies, |
2820 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2821 | slave_last_rx(bond, slave) + delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2822 | |
b2220cad JV |
2823 | slave->new_link = BOND_LINK_UP; |
2824 | commit++; | |
2825 | } | |
1da177e4 | 2826 | |
b2220cad JV |
2827 | continue; |
2828 | } | |
1da177e4 | 2829 | |
b2220cad JV |
2830 | /* |
2831 | * Give slaves 2*delta after being enslaved or made | |
2832 | * active. This avoids bouncing, as the last receive | |
2833 | * times need a full ARP monitor cycle to be updated. | |
2834 | */ | |
cb32f2a0 JB |
2835 | if (time_in_range(jiffies, |
2836 | slave->jiffies - delta_in_ticks, | |
da210f55 | 2837 | slave->jiffies + 2 * delta_in_ticks + extra_ticks)) |
b2220cad JV |
2838 | continue; |
2839 | ||
2840 | /* | |
2841 | * Backup slave is down if: | |
2842 | * - No current_arp_slave AND | |
2843 | * - more than 3*delta since last receive AND | |
2844 | * - the bond has an IP address | |
2845 | * | |
2846 | * Note: a non-null current_arp_slave indicates | |
2847 | * the curr_active_slave went down and we are | |
2848 | * searching for a new one; under this condition | |
2849 | * we only take the curr_active_slave down - this | |
2850 | * gives each slave a chance to tx/rx traffic | |
2851 | * before being taken out | |
2852 | */ | |
e30bc066 | 2853 | if (!bond_is_active_slave(slave) && |
b2220cad | 2854 | !bond->current_arp_slave && |
cb32f2a0 JB |
2855 | !time_in_range(jiffies, |
2856 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2857 | slave_last_rx(bond, slave) + 3 * delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2858 | |
b2220cad JV |
2859 | slave->new_link = BOND_LINK_DOWN; |
2860 | commit++; | |
2861 | } | |
2862 | ||
2863 | /* | |
2864 | * Active slave is down if: | |
2865 | * - more than 2*delta since transmitting OR | |
2866 | * - (more than 2*delta since receive AND | |
2867 | * the bond has an IP address) | |
2868 | */ | |
cb32f2a0 | 2869 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2870 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
2871 | (!time_in_range(jiffies, |
2872 | trans_start - delta_in_ticks, | |
da210f55 | 2873 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2874 | !time_in_range(jiffies, |
2875 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2876 | slave_last_rx(bond, slave) + 2 * delta_in_ticks + extra_ticks))) { |
cb32f2a0 | 2877 | |
b2220cad JV |
2878 | slave->new_link = BOND_LINK_DOWN; |
2879 | commit++; | |
2880 | } | |
1da177e4 LT |
2881 | } |
2882 | ||
b2220cad JV |
2883 | return commit; |
2884 | } | |
1da177e4 | 2885 | |
b2220cad JV |
2886 | /* |
2887 | * Called to commit link state changes noted by inspection step of | |
2888 | * active-backup mode ARP monitor. | |
2889 | * | |
2890 | * Called with RTNL and bond->lock for read. | |
2891 | */ | |
2892 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2893 | { | |
2894 | struct slave *slave; | |
2895 | int i; | |
cb32f2a0 | 2896 | unsigned long trans_start; |
1da177e4 | 2897 | |
b2220cad JV |
2898 | bond_for_each_slave(bond, slave, i) { |
2899 | switch (slave->new_link) { | |
2900 | case BOND_LINK_NOCHANGE: | |
2901 | continue; | |
ff59c456 | 2902 | |
b2220cad | 2903 | case BOND_LINK_UP: |
cb32f2a0 | 2904 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 2905 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
2906 | time_in_range(jiffies, |
2907 | trans_start - delta_in_ticks, | |
da210f55 | 2908 | trans_start + delta_in_ticks + delta_in_ticks / 2)) || |
b9f60253 | 2909 | bond->curr_active_slave != slave) { |
b2220cad | 2910 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2911 | if (bond->current_arp_slave) { |
2912 | bond_set_slave_inactive_flags( | |
2913 | bond->current_arp_slave); | |
2914 | bond->current_arp_slave = NULL; | |
2915 | } | |
b2220cad | 2916 | |
a4aee5c8 | 2917 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2918 | bond->dev->name, slave->dev->name); |
b2220cad | 2919 | |
b9f60253 JP |
2920 | if (!bond->curr_active_slave || |
2921 | (slave == bond->primary_slave)) | |
2922 | goto do_failover; | |
1da177e4 | 2923 | |
b9f60253 | 2924 | } |
1da177e4 | 2925 | |
b9f60253 | 2926 | continue; |
1da177e4 | 2927 | |
b2220cad JV |
2928 | case BOND_LINK_DOWN: |
2929 | if (slave->link_failure_count < UINT_MAX) | |
2930 | slave->link_failure_count++; | |
2931 | ||
2932 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2933 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2934 | |
a4aee5c8 | 2935 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2936 | bond->dev->name, slave->dev->name); |
b2220cad | 2937 | |
b9f60253 | 2938 | if (slave == bond->curr_active_slave) { |
b2220cad | 2939 | bond->current_arp_slave = NULL; |
b9f60253 | 2940 | goto do_failover; |
1da177e4 | 2941 | } |
b9f60253 JP |
2942 | |
2943 | continue; | |
b2220cad JV |
2944 | |
2945 | default: | |
a4aee5c8 | 2946 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2947 | bond->dev->name, slave->new_link, |
2948 | slave->dev->name); | |
b9f60253 | 2949 | continue; |
1da177e4 | 2950 | } |
1da177e4 | 2951 | |
b9f60253 JP |
2952 | do_failover: |
2953 | ASSERT_RTNL(); | |
e843fa50 | 2954 | block_netpoll_tx(); |
b2220cad | 2955 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2956 | bond_select_active_slave(bond); |
b2220cad | 2957 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2958 | unblock_netpoll_tx(); |
b2220cad | 2959 | } |
1da177e4 | 2960 | |
b2220cad JV |
2961 | bond_set_carrier(bond); |
2962 | } | |
1da177e4 | 2963 | |
b2220cad JV |
2964 | /* |
2965 | * Send ARP probes for active-backup mode ARP monitor. | |
2966 | * | |
2967 | * Called with bond->lock held for read. | |
2968 | */ | |
2969 | static void bond_ab_arp_probe(struct bonding *bond) | |
2970 | { | |
2971 | struct slave *slave; | |
2972 | int i; | |
1da177e4 | 2973 | |
b2220cad | 2974 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2975 | |
b2220cad | 2976 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
2977 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
2978 | bond->current_arp_slave->dev->name, | |
2979 | bond->curr_active_slave->dev->name); | |
1da177e4 | 2980 | |
b2220cad JV |
2981 | if (bond->curr_active_slave) { |
2982 | bond_arp_send_all(bond, bond->curr_active_slave); | |
2983 | read_unlock(&bond->curr_slave_lock); | |
2984 | return; | |
2985 | } | |
1da177e4 | 2986 | |
b2220cad | 2987 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 2988 | |
b2220cad JV |
2989 | /* if we don't have a curr_active_slave, search for the next available |
2990 | * backup slave from the current_arp_slave and make it the candidate | |
2991 | * for becoming the curr_active_slave | |
2992 | */ | |
1da177e4 | 2993 | |
b2220cad JV |
2994 | if (!bond->current_arp_slave) { |
2995 | bond->current_arp_slave = bond->first_slave; | |
2996 | if (!bond->current_arp_slave) | |
2997 | return; | |
2998 | } | |
1da177e4 | 2999 | |
b2220cad | 3000 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3001 | |
b2220cad JV |
3002 | /* search for next candidate */ |
3003 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3004 | if (IS_UP(slave->dev)) { | |
3005 | slave->link = BOND_LINK_BACK; | |
3006 | bond_set_slave_active_flags(slave); | |
3007 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3008 | slave->jiffies = jiffies; |
b2220cad JV |
3009 | bond->current_arp_slave = slave; |
3010 | break; | |
1da177e4 LT |
3011 | } |
3012 | ||
b2220cad JV |
3013 | /* if the link state is up at this point, we |
3014 | * mark it down - this can happen if we have | |
3015 | * simultaneous link failures and | |
3016 | * reselect_active_interface doesn't make this | |
3017 | * one the current slave so it is still marked | |
3018 | * up when it is actually down | |
1da177e4 | 3019 | */ |
b2220cad JV |
3020 | if (slave->link == BOND_LINK_UP) { |
3021 | slave->link = BOND_LINK_DOWN; | |
3022 | if (slave->link_failure_count < UINT_MAX) | |
3023 | slave->link_failure_count++; | |
1da177e4 | 3024 | |
b2220cad JV |
3025 | bond_set_slave_inactive_flags(slave); |
3026 | ||
a4aee5c8 JP |
3027 | pr_info("%s: backup interface %s is now down.\n", |
3028 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3029 | } |
b2220cad JV |
3030 | } |
3031 | } | |
1da177e4 | 3032 | |
b2220cad JV |
3033 | void bond_activebackup_arp_mon(struct work_struct *work) |
3034 | { | |
3035 | struct bonding *bond = container_of(work, struct bonding, | |
3036 | arp_work.work); | |
ad246c99 | 3037 | bool should_notify_peers = false; |
b2220cad | 3038 | int delta_in_ticks; |
1da177e4 | 3039 | |
b2220cad | 3040 | read_lock(&bond->lock); |
1da177e4 | 3041 | |
b2220cad | 3042 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3043 | |
b2220cad JV |
3044 | if (bond->slave_cnt == 0) |
3045 | goto re_arm; | |
3046 | ||
ad246c99 BH |
3047 | should_notify_peers = bond_should_notify_peers(bond); |
3048 | ||
b2220cad JV |
3049 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3050 | read_unlock(&bond->lock); | |
e6d265e8 JV |
3051 | |
3052 | /* Race avoidance with bond_close flush of workqueue */ | |
3053 | if (!rtnl_trylock()) { | |
3054 | read_lock(&bond->lock); | |
3055 | delta_in_ticks = 1; | |
3056 | should_notify_peers = false; | |
3057 | goto re_arm; | |
3058 | } | |
3059 | ||
b2220cad JV |
3060 | read_lock(&bond->lock); |
3061 | ||
3062 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3063 | ||
3064 | read_unlock(&bond->lock); | |
3065 | rtnl_unlock(); | |
3066 | read_lock(&bond->lock); | |
1da177e4 LT |
3067 | } |
3068 | ||
b2220cad JV |
3069 | bond_ab_arp_probe(bond); |
3070 | ||
1da177e4 | 3071 | re_arm: |
e6d265e8 | 3072 | if (bond->params.arp_interval) |
1b76b316 | 3073 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 3074 | |
1da177e4 | 3075 | read_unlock(&bond->lock); |
ad246c99 BH |
3076 | |
3077 | if (should_notify_peers) { | |
e6d265e8 JV |
3078 | if (!rtnl_trylock()) { |
3079 | read_lock(&bond->lock); | |
3080 | bond->send_peer_notif++; | |
3081 | read_unlock(&bond->lock); | |
3082 | return; | |
3083 | } | |
b7bc2a5b | 3084 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
3085 | rtnl_unlock(); |
3086 | } | |
1da177e4 LT |
3087 | } |
3088 | ||
1da177e4 LT |
3089 | /*-------------------------- netdev event handling --------------------------*/ |
3090 | ||
3091 | /* | |
3092 | * Change device name | |
3093 | */ | |
3094 | static int bond_event_changename(struct bonding *bond) | |
3095 | { | |
1da177e4 LT |
3096 | bond_remove_proc_entry(bond); |
3097 | bond_create_proc_entry(bond); | |
7e083840 | 3098 | |
f073c7ca TI |
3099 | bond_debug_reregister(bond); |
3100 | ||
1da177e4 LT |
3101 | return NOTIFY_DONE; |
3102 | } | |
3103 | ||
3d632c3f SH |
3104 | static int bond_master_netdev_event(unsigned long event, |
3105 | struct net_device *bond_dev) | |
1da177e4 | 3106 | { |
454d7c9b | 3107 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3108 | |
3109 | switch (event) { | |
3110 | case NETDEV_CHANGENAME: | |
3111 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3112 | case NETDEV_UNREGISTER: |
3113 | bond_remove_proc_entry(event_bond); | |
3114 | break; | |
3115 | case NETDEV_REGISTER: | |
3116 | bond_create_proc_entry(event_bond); | |
3117 | break; | |
1da177e4 LT |
3118 | default: |
3119 | break; | |
3120 | } | |
3121 | ||
3122 | return NOTIFY_DONE; | |
3123 | } | |
3124 | ||
3d632c3f SH |
3125 | static int bond_slave_netdev_event(unsigned long event, |
3126 | struct net_device *slave_dev) | |
1da177e4 | 3127 | { |
471cb5a3 | 3128 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
6101391d | 3129 | struct bonding *bond; |
3130 | struct net_device *bond_dev; | |
471cb5a3 JP |
3131 | u32 old_speed; |
3132 | u8 old_duplex; | |
1da177e4 | 3133 | |
6101391d | 3134 | /* A netdev event can be generated while enslaving a device |
3135 | * before netdev_rx_handler_register is called in which case | |
3136 | * slave will be NULL | |
3137 | */ | |
3138 | if (!slave) | |
3139 | return NOTIFY_DONE; | |
3140 | bond_dev = slave->bond->dev; | |
3141 | bond = slave->bond; | |
3142 | ||
1da177e4 LT |
3143 | switch (event) { |
3144 | case NETDEV_UNREGISTER: | |
8d2ada77 | 3145 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
3146 | bond_release_and_destroy(bond_dev, slave_dev); |
3147 | else | |
3148 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 3149 | break; |
98f41f69 | 3150 | case NETDEV_UP: |
1da177e4 | 3151 | case NETDEV_CHANGE: |
471cb5a3 JP |
3152 | old_speed = slave->speed; |
3153 | old_duplex = slave->duplex; | |
17d04500 | 3154 | |
471cb5a3 | 3155 | bond_update_speed_duplex(slave); |
17d04500 | 3156 | |
471cb5a3 JP |
3157 | if (bond->params.mode == BOND_MODE_8023AD) { |
3158 | if (old_speed != slave->speed) | |
3159 | bond_3ad_adapter_speed_changed(slave); | |
3160 | if (old_duplex != slave->duplex) | |
3161 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 3162 | } |
1da177e4 LT |
3163 | break; |
3164 | case NETDEV_DOWN: | |
3165 | /* | |
3166 | * ... Or is it this? | |
3167 | */ | |
3168 | break; | |
3169 | case NETDEV_CHANGEMTU: | |
3170 | /* | |
3171 | * TODO: Should slaves be allowed to | |
3172 | * independently alter their MTU? For | |
3173 | * an active-backup bond, slaves need | |
3174 | * not be the same type of device, so | |
3175 | * MTUs may vary. For other modes, | |
3176 | * slaves arguably should have the | |
3177 | * same MTUs. To do this, we'd need to | |
3178 | * take over the slave's change_mtu | |
3179 | * function for the duration of their | |
3180 | * servitude. | |
3181 | */ | |
3182 | break; | |
3183 | case NETDEV_CHANGENAME: | |
3184 | /* | |
3185 | * TODO: handle changing the primary's name | |
3186 | */ | |
3187 | break; | |
8531c5ff AK |
3188 | case NETDEV_FEAT_CHANGE: |
3189 | bond_compute_features(bond); | |
3190 | break; | |
4aa5dee4 JP |
3191 | case NETDEV_RESEND_IGMP: |
3192 | /* Propagate to master device */ | |
3193 | call_netdevice_notifiers(event, slave->bond->dev); | |
3194 | break; | |
1da177e4 LT |
3195 | default: |
3196 | break; | |
3197 | } | |
3198 | ||
3199 | return NOTIFY_DONE; | |
3200 | } | |
3201 | ||
3202 | /* | |
3203 | * bond_netdev_event: handle netdev notifier chain events. | |
3204 | * | |
3205 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3206 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3207 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3208 | * dev_probe_lock). | |
3209 | */ | |
3d632c3f SH |
3210 | static int bond_netdev_event(struct notifier_block *this, |
3211 | unsigned long event, void *ptr) | |
1da177e4 | 3212 | { |
351638e7 | 3213 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3214 | |
5a03cdb7 | 3215 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3216 | event_dev ? event_dev->name : "None", |
3217 | event); | |
1da177e4 | 3218 | |
0b680e75 JV |
3219 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3220 | return NOTIFY_DONE; | |
3221 | ||
1da177e4 | 3222 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3223 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3224 | return bond_master_netdev_event(event, event_dev); |
3225 | } | |
3226 | ||
3227 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3228 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3229 | return bond_slave_netdev_event(event, event_dev); |
3230 | } | |
3231 | ||
3232 | return NOTIFY_DONE; | |
3233 | } | |
3234 | ||
3235 | static struct notifier_block bond_netdev_notifier = { | |
3236 | .notifier_call = bond_netdev_event, | |
3237 | }; | |
3238 | ||
169a3e66 JV |
3239 | /*---------------------------- Hashing Policies -----------------------------*/ |
3240 | ||
6b923cb7 JE |
3241 | /* |
3242 | * Hash for the output device based upon layer 2 data | |
3243 | */ | |
3244 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) | |
3245 | { | |
3246 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3247 | ||
3248 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
3249 | return (data->h_dest[5] ^ data->h_source[5]) % count; | |
3250 | ||
3251 | return 0; | |
3252 | } | |
3253 | ||
6f6652be JV |
3254 | /* |
3255 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
6b923cb7 | 3256 | * the packet is not IP, fall back on bond_xmit_hash_policy_l2() |
6f6652be | 3257 | */ |
a361c83c | 3258 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be | 3259 | { |
4394542c ED |
3260 | const struct ethhdr *data; |
3261 | const struct iphdr *iph; | |
3262 | const struct ipv6hdr *ipv6h; | |
6b923cb7 | 3263 | u32 v6hash; |
4394542c | 3264 | const __be32 *s, *d; |
6b923cb7 JE |
3265 | |
3266 | if (skb->protocol == htons(ETH_P_IP) && | |
4394542c | 3267 | pskb_network_may_pull(skb, sizeof(*iph))) { |
6b923cb7 | 3268 | iph = ip_hdr(skb); |
4394542c | 3269 | data = (struct ethhdr *)skb->data; |
6f6652be | 3270 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3271 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6b923cb7 | 3272 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3273 | pskb_network_may_pull(skb, sizeof(*ipv6h))) { |
6b923cb7 | 3274 | ipv6h = ipv6_hdr(skb); |
4394542c | 3275 | data = (struct ethhdr *)skb->data; |
6b923cb7 JE |
3276 | s = &ipv6h->saddr.s6_addr32[0]; |
3277 | d = &ipv6h->daddr.s6_addr32[0]; | |
3278 | v6hash = (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3279 | v6hash ^= (v6hash >> 24) ^ (v6hash >> 16) ^ (v6hash >> 8); | |
3280 | return (v6hash ^ data->h_dest[5] ^ data->h_source[5]) % count; | |
6f6652be JV |
3281 | } |
3282 | ||
6b923cb7 | 3283 | return bond_xmit_hash_policy_l2(skb, count); |
6f6652be JV |
3284 | } |
3285 | ||
169a3e66 | 3286 | /* |
59c51591 | 3287 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 | 3288 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
6b923cb7 | 3289 | * altogether not IP, fall back on bond_xmit_hash_policy_l2() |
169a3e66 | 3290 | */ |
a361c83c | 3291 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 | 3292 | { |
6b923cb7 | 3293 | u32 layer4_xor = 0; |
4394542c ED |
3294 | const struct iphdr *iph; |
3295 | const struct ipv6hdr *ipv6h; | |
3296 | const __be32 *s, *d; | |
3297 | const __be16 *l4 = NULL; | |
3298 | __be16 _l4[2]; | |
3299 | int noff = skb_network_offset(skb); | |
3300 | int poff; | |
169a3e66 | 3301 | |
6b923cb7 | 3302 | if (skb->protocol == htons(ETH_P_IP) && |
4394542c | 3303 | pskb_may_pull(skb, noff + sizeof(*iph))) { |
6b923cb7 | 3304 | iph = ip_hdr(skb); |
4394542c ED |
3305 | poff = proto_ports_offset(iph->protocol); |
3306 | ||
3307 | if (!ip_is_fragment(iph) && poff >= 0) { | |
3308 | l4 = skb_header_pointer(skb, noff + (iph->ihl << 2) + poff, | |
3309 | sizeof(_l4), &_l4); | |
3310 | if (l4) | |
3311 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
169a3e66 JV |
3312 | } |
3313 | return (layer4_xor ^ | |
3314 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
6b923cb7 | 3315 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3316 | pskb_may_pull(skb, noff + sizeof(*ipv6h))) { |
6b923cb7 | 3317 | ipv6h = ipv6_hdr(skb); |
4394542c ED |
3318 | poff = proto_ports_offset(ipv6h->nexthdr); |
3319 | if (poff >= 0) { | |
3320 | l4 = skb_header_pointer(skb, noff + sizeof(*ipv6h) + poff, | |
3321 | sizeof(_l4), &_l4); | |
3322 | if (l4) | |
3323 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
6b923cb7 JE |
3324 | } |
3325 | s = &ipv6h->saddr.s6_addr32[0]; | |
3326 | d = &ipv6h->daddr.s6_addr32[0]; | |
3327 | layer4_xor ^= (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3328 | layer4_xor ^= (layer4_xor >> 24) ^ (layer4_xor >> 16) ^ | |
3329 | (layer4_xor >> 8); | |
3330 | return layer4_xor % count; | |
169a3e66 JV |
3331 | } |
3332 | ||
6b923cb7 | 3333 | return bond_xmit_hash_policy_l2(skb, count); |
169a3e66 JV |
3334 | } |
3335 | ||
1da177e4 LT |
3336 | /*-------------------------- Device entry points ----------------------------*/ |
3337 | ||
fbb0c41b | 3338 | static void bond_work_init_all(struct bonding *bond) |
3339 | { | |
3340 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3341 | bond_resend_igmp_join_requests_delayed); | |
3342 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3343 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3344 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3345 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3346 | else | |
3347 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3348 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3349 | } | |
3350 | ||
3351 | static void bond_work_cancel_all(struct bonding *bond) | |
3352 | { | |
3353 | cancel_delayed_work_sync(&bond->mii_work); | |
3354 | cancel_delayed_work_sync(&bond->arp_work); | |
3355 | cancel_delayed_work_sync(&bond->alb_work); | |
3356 | cancel_delayed_work_sync(&bond->ad_work); | |
3357 | cancel_delayed_work_sync(&bond->mcast_work); | |
3358 | } | |
3359 | ||
1da177e4 LT |
3360 | static int bond_open(struct net_device *bond_dev) |
3361 | { | |
454d7c9b | 3362 | struct bonding *bond = netdev_priv(bond_dev); |
ba3211cc PP |
3363 | struct slave *slave; |
3364 | int i; | |
1da177e4 | 3365 | |
ba3211cc PP |
3366 | /* reset slave->backup and slave->inactive */ |
3367 | read_lock(&bond->lock); | |
3368 | if (bond->slave_cnt > 0) { | |
3369 | read_lock(&bond->curr_slave_lock); | |
3370 | bond_for_each_slave(bond, slave, i) { | |
3371 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3372 | && (slave != bond->curr_active_slave)) { | |
3373 | bond_set_slave_inactive_flags(slave); | |
3374 | } else { | |
3375 | bond_set_slave_active_flags(slave); | |
3376 | } | |
3377 | } | |
3378 | read_unlock(&bond->curr_slave_lock); | |
3379 | } | |
3380 | read_unlock(&bond->lock); | |
3381 | ||
fbb0c41b | 3382 | bond_work_init_all(bond); |
5a37e8ca | 3383 | |
58402054 | 3384 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3385 | /* bond_alb_initialize must be called before the timer |
3386 | * is started. | |
3387 | */ | |
fbb0c41b | 3388 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3389 | return -ENOMEM; |
1b76b316 | 3390 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3391 | } |
3392 | ||
fbb0c41b | 3393 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3394 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3395 | |
3396 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3397 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
f5b2b966 | 3398 | if (bond->params.arp_validate) |
3aba891d | 3399 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3400 | } |
3401 | ||
3402 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3403 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3404 | /* register to receive LACPDUs */ |
3aba891d | 3405 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3406 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3407 | } |
3408 | ||
3409 | return 0; | |
3410 | } | |
3411 | ||
3412 | static int bond_close(struct net_device *bond_dev) | |
3413 | { | |
454d7c9b | 3414 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3415 | |
1da177e4 | 3416 | write_lock_bh(&bond->lock); |
ad246c99 | 3417 | bond->send_peer_notif = 0; |
1da177e4 LT |
3418 | write_unlock_bh(&bond->lock); |
3419 | ||
fbb0c41b | 3420 | bond_work_cancel_all(bond); |
58402054 | 3421 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3422 | /* Must be called only after all |
3423 | * slaves have been released | |
3424 | */ | |
3425 | bond_alb_deinitialize(bond); | |
3426 | } | |
3aba891d | 3427 | bond->recv_probe = NULL; |
1da177e4 LT |
3428 | |
3429 | return 0; | |
3430 | } | |
3431 | ||
28172739 ED |
3432 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3433 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3434 | { |
454d7c9b | 3435 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3436 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3437 | struct slave *slave; |
3438 | int i; | |
3439 | ||
28172739 | 3440 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3441 | |
3442 | read_lock_bh(&bond->lock); | |
3443 | ||
3444 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3445 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3446 | dev_get_stats(slave->dev, &temp); |
3447 | ||
3448 | stats->rx_packets += sstats->rx_packets; | |
3449 | stats->rx_bytes += sstats->rx_bytes; | |
3450 | stats->rx_errors += sstats->rx_errors; | |
3451 | stats->rx_dropped += sstats->rx_dropped; | |
3452 | ||
3453 | stats->tx_packets += sstats->tx_packets; | |
3454 | stats->tx_bytes += sstats->tx_bytes; | |
3455 | stats->tx_errors += sstats->tx_errors; | |
3456 | stats->tx_dropped += sstats->tx_dropped; | |
3457 | ||
3458 | stats->multicast += sstats->multicast; | |
3459 | stats->collisions += sstats->collisions; | |
3460 | ||
3461 | stats->rx_length_errors += sstats->rx_length_errors; | |
3462 | stats->rx_over_errors += sstats->rx_over_errors; | |
3463 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3464 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3465 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3466 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3467 | ||
3468 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3469 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3470 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3471 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3472 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3473 | } |
3474 | ||
1da177e4 LT |
3475 | read_unlock_bh(&bond->lock); |
3476 | ||
3477 | return stats; | |
3478 | } | |
3479 | ||
3480 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3481 | { | |
3482 | struct net_device *slave_dev = NULL; | |
3483 | struct ifbond k_binfo; | |
3484 | struct ifbond __user *u_binfo = NULL; | |
3485 | struct ifslave k_sinfo; | |
3486 | struct ifslave __user *u_sinfo = NULL; | |
3487 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3488 | struct net *net; |
1da177e4 LT |
3489 | int res = 0; |
3490 | ||
a4aee5c8 | 3491 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3492 | |
3493 | switch (cmd) { | |
1da177e4 LT |
3494 | case SIOCGMIIPHY: |
3495 | mii = if_mii(ifr); | |
3d632c3f | 3496 | if (!mii) |
1da177e4 | 3497 | return -EINVAL; |
3d632c3f | 3498 | |
1da177e4 LT |
3499 | mii->phy_id = 0; |
3500 | /* Fall Through */ | |
3501 | case SIOCGMIIREG: | |
3502 | /* | |
3503 | * We do this again just in case we were called by SIOCGMIIREG | |
3504 | * instead of SIOCGMIIPHY. | |
3505 | */ | |
3506 | mii = if_mii(ifr); | |
3d632c3f | 3507 | if (!mii) |
1da177e4 | 3508 | return -EINVAL; |
3d632c3f | 3509 | |
1da177e4 LT |
3510 | |
3511 | if (mii->reg_num == 1) { | |
454d7c9b | 3512 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3513 | mii->val_out = 0; |
6603a6f2 | 3514 | read_lock(&bond->lock); |
1da177e4 | 3515 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3516 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3517 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3518 | |
1da177e4 | 3519 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3520 | read_unlock(&bond->lock); |
1da177e4 LT |
3521 | } |
3522 | ||
3523 | return 0; | |
3524 | case BOND_INFO_QUERY_OLD: | |
3525 | case SIOCBONDINFOQUERY: | |
3526 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3527 | ||
3d632c3f | 3528 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3529 | return -EFAULT; |
1da177e4 LT |
3530 | |
3531 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3532 | if (res == 0 && |
3533 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3534 | return -EFAULT; | |
1da177e4 LT |
3535 | |
3536 | return res; | |
3537 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3538 | case SIOCBONDSLAVEINFOQUERY: | |
3539 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3540 | ||
3d632c3f | 3541 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3542 | return -EFAULT; |
1da177e4 LT |
3543 | |
3544 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3545 | if (res == 0 && |
3546 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3547 | return -EFAULT; | |
1da177e4 LT |
3548 | |
3549 | return res; | |
3550 | default: | |
3551 | /* Go on */ | |
3552 | break; | |
3553 | } | |
3554 | ||
387ff911 G |
3555 | net = dev_net(bond_dev); |
3556 | ||
3557 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3558 | return -EPERM; |
1da177e4 | 3559 | |
387ff911 | 3560 | slave_dev = dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3561 | |
a4aee5c8 | 3562 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3563 | |
3d632c3f | 3564 | if (!slave_dev) |
1da177e4 | 3565 | res = -ENODEV; |
3d632c3f | 3566 | else { |
a4aee5c8 | 3567 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3568 | switch (cmd) { |
3569 | case BOND_ENSLAVE_OLD: | |
3570 | case SIOCBONDENSLAVE: | |
3571 | res = bond_enslave(bond_dev, slave_dev); | |
3572 | break; | |
3573 | case BOND_RELEASE_OLD: | |
3574 | case SIOCBONDRELEASE: | |
3575 | res = bond_release(bond_dev, slave_dev); | |
3576 | break; | |
3577 | case BOND_SETHWADDR_OLD: | |
3578 | case SIOCBONDSETHWADDR: | |
409cc1f8 JP |
3579 | bond_set_dev_addr(bond_dev, slave_dev); |
3580 | res = 0; | |
1da177e4 LT |
3581 | break; |
3582 | case BOND_CHANGE_ACTIVE_OLD: | |
3583 | case SIOCBONDCHANGEACTIVE: | |
3584 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3585 | break; | |
3586 | default: | |
3587 | res = -EOPNOTSUPP; | |
3588 | } | |
3589 | ||
3590 | dev_put(slave_dev); | |
3591 | } | |
3592 | ||
1da177e4 LT |
3593 | return res; |
3594 | } | |
3595 | ||
d03462b9 | 3596 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3597 | { |
454d7c9b | 3598 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3599 | |
d03462b9 JP |
3600 | if (change & IFF_PROMISC) |
3601 | bond_set_promiscuity(bond, | |
3602 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3603 | |
d03462b9 JP |
3604 | if (change & IFF_ALLMULTI) |
3605 | bond_set_allmulti(bond, | |
3606 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3607 | } | |
1da177e4 | 3608 | |
303d1cbf | 3609 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3610 | { |
3611 | struct bonding *bond = netdev_priv(bond_dev); | |
303d1cbf JV |
3612 | struct slave *slave; |
3613 | int i; | |
1da177e4 | 3614 | |
80ee5ad2 JV |
3615 | read_lock(&bond->lock); |
3616 | ||
303d1cbf JV |
3617 | if (USES_PRIMARY(bond->params.mode)) { |
3618 | read_lock(&bond->curr_slave_lock); | |
3619 | slave = bond->curr_active_slave; | |
3620 | if (slave) { | |
3621 | dev_uc_sync(slave->dev, bond_dev); | |
3622 | dev_mc_sync(slave->dev, bond_dev); | |
3623 | } | |
3624 | read_unlock(&bond->curr_slave_lock); | |
3625 | } else { | |
3626 | bond_for_each_slave(bond, slave, i) { | |
3627 | dev_uc_sync_multiple(slave->dev, bond_dev); | |
3628 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3629 | } | |
1da177e4 LT |
3630 | } |
3631 | ||
80ee5ad2 | 3632 | read_unlock(&bond->lock); |
1da177e4 LT |
3633 | } |
3634 | ||
234bcf8a | 3635 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3636 | { |
234bcf8a | 3637 | struct bonding *bond = netdev_priv(n->dev); |
00829823 | 3638 | struct slave *slave = bond->first_slave; |
234bcf8a SP |
3639 | const struct net_device_ops *slave_ops; |
3640 | struct neigh_parms parms; | |
3641 | int ret; | |
3642 | ||
3643 | if (!slave) | |
3644 | return 0; | |
3645 | ||
3646 | slave_ops = slave->dev->netdev_ops; | |
3647 | ||
3648 | if (!slave_ops->ndo_neigh_setup) | |
3649 | return 0; | |
3650 | ||
3651 | parms.neigh_setup = NULL; | |
3652 | parms.neigh_cleanup = NULL; | |
3653 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3654 | if (ret) | |
3655 | return ret; | |
3656 | ||
3657 | /* | |
3658 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3659 | * after the last slave has been detached. Assumes that all slaves | |
3660 | * utilize the same neigh_cleanup (true at this writing as only user | |
3661 | * is ipoib). | |
3662 | */ | |
3663 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3664 | ||
3665 | if (!parms.neigh_setup) | |
3666 | return 0; | |
3667 | ||
3668 | return parms.neigh_setup(n); | |
3669 | } | |
3670 | ||
3671 | /* | |
3672 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3673 | * slave exists. So we must declare proxy setup function which will | |
3674 | * be used at run time to resolve the actual slave neigh param setup. | |
3675 | */ | |
3676 | static int bond_neigh_setup(struct net_device *dev, | |
3677 | struct neigh_parms *parms) | |
3678 | { | |
3679 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3680 | |
00829823 SH |
3681 | return 0; |
3682 | } | |
3683 | ||
1da177e4 LT |
3684 | /* |
3685 | * Change the MTU of all of a master's slaves to match the master | |
3686 | */ | |
3687 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3688 | { | |
454d7c9b | 3689 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3690 | struct slave *slave, *stop_at; |
3691 | int res = 0; | |
3692 | int i; | |
3693 | ||
5a03cdb7 | 3694 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3695 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3696 | |
3697 | /* Can't hold bond->lock with bh disabled here since | |
3698 | * some base drivers panic. On the other hand we can't | |
3699 | * hold bond->lock without bh disabled because we'll | |
3700 | * deadlock. The only solution is to rely on the fact | |
3701 | * that we're under rtnl_lock here, and the slaves | |
3702 | * list won't change. This doesn't solve the problem | |
3703 | * of setting the slave's MTU while it is | |
3704 | * transmitting, but the assumption is that the base | |
3705 | * driver can handle that. | |
3706 | * | |
3707 | * TODO: figure out a way to safely iterate the slaves | |
3708 | * list, but without holding a lock around the actual | |
3709 | * call to the base driver. | |
3710 | */ | |
3711 | ||
3712 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3713 | pr_debug("s %p s->p %p c_m %p\n", |
3714 | slave, | |
3715 | slave->prev, | |
3716 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3717 | |
1da177e4 LT |
3718 | res = dev_set_mtu(slave->dev, new_mtu); |
3719 | ||
3720 | if (res) { | |
3721 | /* If we failed to set the slave's mtu to the new value | |
3722 | * we must abort the operation even in ACTIVE_BACKUP | |
3723 | * mode, because if we allow the backup slaves to have | |
3724 | * different mtu values than the active slave we'll | |
3725 | * need to change their mtu when doing a failover. That | |
3726 | * means changing their mtu from timer context, which | |
3727 | * is probably not a good idea. | |
3728 | */ | |
5a03cdb7 | 3729 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3730 | goto unwind; |
3731 | } | |
3732 | } | |
3733 | ||
3734 | bond_dev->mtu = new_mtu; | |
3735 | ||
3736 | return 0; | |
3737 | ||
3738 | unwind: | |
3739 | /* unwind from head to the slave that failed */ | |
3740 | stop_at = slave; | |
3741 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3742 | int tmp_res; | |
3743 | ||
3744 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3745 | if (tmp_res) { | |
a4aee5c8 JP |
3746 | pr_debug("unwind err %d dev %s\n", |
3747 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3748 | } |
3749 | } | |
3750 | ||
3751 | return res; | |
3752 | } | |
3753 | ||
3754 | /* | |
3755 | * Change HW address | |
3756 | * | |
3757 | * Note that many devices must be down to change the HW address, and | |
3758 | * downing the master releases all slaves. We can make bonds full of | |
3759 | * bonding devices to test this, however. | |
3760 | */ | |
3761 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3762 | { | |
454d7c9b | 3763 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3764 | struct sockaddr *sa = addr, tmp_sa; |
3765 | struct slave *slave, *stop_at; | |
3766 | int res = 0; | |
3767 | int i; | |
3768 | ||
eb7cc59a SH |
3769 | if (bond->params.mode == BOND_MODE_ALB) |
3770 | return bond_alb_set_mac_address(bond_dev, addr); | |
3771 | ||
3772 | ||
a4aee5c8 JP |
3773 | pr_debug("bond=%p, name=%s\n", |
3774 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3775 | |
1b5acd29 JV |
3776 | /* If fail_over_mac is enabled, do nothing and return success. |
3777 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3778 | */ |
1b5acd29 | 3779 | if (bond->params.fail_over_mac) |
dd957c57 | 3780 | return 0; |
2ab82852 | 3781 | |
3d632c3f | 3782 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3783 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3784 | |
3785 | /* Can't hold bond->lock with bh disabled here since | |
3786 | * some base drivers panic. On the other hand we can't | |
3787 | * hold bond->lock without bh disabled because we'll | |
3788 | * deadlock. The only solution is to rely on the fact | |
3789 | * that we're under rtnl_lock here, and the slaves | |
3790 | * list won't change. This doesn't solve the problem | |
3791 | * of setting the slave's hw address while it is | |
3792 | * transmitting, but the assumption is that the base | |
3793 | * driver can handle that. | |
3794 | * | |
3795 | * TODO: figure out a way to safely iterate the slaves | |
3796 | * list, but without holding a lock around the actual | |
3797 | * call to the base driver. | |
3798 | */ | |
3799 | ||
3800 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 3801 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3802 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3803 | |
eb7cc59a | 3804 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3805 | res = -EOPNOTSUPP; |
5a03cdb7 | 3806 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3807 | goto unwind; |
3808 | } | |
3809 | ||
3810 | res = dev_set_mac_address(slave->dev, addr); | |
3811 | if (res) { | |
3812 | /* TODO: consider downing the slave | |
3813 | * and retry ? | |
3814 | * User should expect communications | |
3815 | * breakage anyway until ARP finish | |
3816 | * updating, so... | |
3817 | */ | |
5a03cdb7 | 3818 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3819 | goto unwind; |
3820 | } | |
3821 | } | |
3822 | ||
3823 | /* success */ | |
3824 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3825 | return 0; | |
3826 | ||
3827 | unwind: | |
3828 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3829 | tmp_sa.sa_family = bond_dev->type; | |
3830 | ||
3831 | /* unwind from head to the slave that failed */ | |
3832 | stop_at = slave; | |
3833 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3834 | int tmp_res; | |
3835 | ||
3836 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
3837 | if (tmp_res) { | |
a4aee5c8 JP |
3838 | pr_debug("unwind err %d dev %s\n", |
3839 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3840 | } |
3841 | } | |
3842 | ||
3843 | return res; | |
3844 | } | |
3845 | ||
3846 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
3847 | { | |
454d7c9b | 3848 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3849 | struct slave *slave, *start_at; |
cf5f9044 | 3850 | int i, slave_no, res = 1; |
a2fd940f | 3851 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 3852 | |
cf5f9044 | 3853 | /* |
a2fd940f AG |
3854 | * Start with the curr_active_slave that joined the bond as the |
3855 | * default for sending IGMP traffic. For failover purposes one | |
3856 | * needs to maintain some consistency for the interface that will | |
3857 | * send the join/membership reports. The curr_active_slave found | |
3858 | * will send all of this type of traffic. | |
cf5f9044 | 3859 | */ |
00ae7028 | 3860 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 3861 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 3862 | |
a2fd940f AG |
3863 | read_lock(&bond->curr_slave_lock); |
3864 | slave = bond->curr_active_slave; | |
3865 | read_unlock(&bond->curr_slave_lock); | |
3866 | ||
3867 | if (!slave) | |
3868 | goto out; | |
3869 | } else { | |
3870 | /* | |
3871 | * Concurrent TX may collide on rr_tx_counter; we accept | |
3872 | * that as being rare enough not to justify using an | |
3873 | * atomic op here. | |
3874 | */ | |
3875 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
3876 | ||
3877 | bond_for_each_slave(bond, slave, i) { | |
3878 | slave_no--; | |
3879 | if (slave_no < 0) | |
3880 | break; | |
3881 | } | |
1da177e4 LT |
3882 | } |
3883 | ||
cf5f9044 | 3884 | start_at = slave; |
1da177e4 LT |
3885 | bond_for_each_slave_from(bond, slave, i, start_at) { |
3886 | if (IS_UP(slave->dev) && | |
3887 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3888 | bond_is_active_slave(slave)) { |
1da177e4 | 3889 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
3890 | break; |
3891 | } | |
3892 | } | |
3893 | ||
1da177e4 LT |
3894 | out: |
3895 | if (res) { | |
3896 | /* no suitable interface, frame not sent */ | |
04502430 | 3897 | kfree_skb(skb); |
1da177e4 | 3898 | } |
0693e88e | 3899 | |
ec634fe3 | 3900 | return NETDEV_TX_OK; |
1da177e4 LT |
3901 | } |
3902 | ||
075897ce | 3903 | |
1da177e4 LT |
3904 | /* |
3905 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3906 | * the bond has a usable interface. | |
3907 | */ | |
3908 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3909 | { | |
454d7c9b | 3910 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3911 | int res = 1; |
3912 | ||
1da177e4 LT |
3913 | read_lock(&bond->curr_slave_lock); |
3914 | ||
0693e88e MM |
3915 | if (bond->curr_active_slave) |
3916 | res = bond_dev_queue_xmit(bond, skb, | |
3917 | bond->curr_active_slave->dev); | |
1da177e4 | 3918 | |
04502430 ED |
3919 | read_unlock(&bond->curr_slave_lock); |
3920 | ||
3d632c3f | 3921 | if (res) |
1da177e4 | 3922 | /* no suitable interface, frame not sent */ |
04502430 | 3923 | kfree_skb(skb); |
0693e88e | 3924 | |
ec634fe3 | 3925 | return NETDEV_TX_OK; |
1da177e4 LT |
3926 | } |
3927 | ||
3928 | /* | |
169a3e66 JV |
3929 | * In bond_xmit_xor() , we determine the output device by using a pre- |
3930 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
3931 | * find the next active slave. | |
1da177e4 LT |
3932 | */ |
3933 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3934 | { | |
454d7c9b | 3935 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3936 | struct slave *slave, *start_at; |
3937 | int slave_no; | |
3938 | int i; | |
3939 | int res = 1; | |
3940 | ||
a361c83c | 3941 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
3942 | |
3943 | bond_for_each_slave(bond, slave, i) { | |
3944 | slave_no--; | |
3d632c3f | 3945 | if (slave_no < 0) |
1da177e4 | 3946 | break; |
1da177e4 LT |
3947 | } |
3948 | ||
3949 | start_at = slave; | |
3950 | ||
3951 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
3952 | if (IS_UP(slave->dev) && | |
3953 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3954 | bond_is_active_slave(slave)) { |
1da177e4 LT |
3955 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
3956 | break; | |
3957 | } | |
3958 | } | |
3959 | ||
1da177e4 LT |
3960 | if (res) { |
3961 | /* no suitable interface, frame not sent */ | |
04502430 | 3962 | kfree_skb(skb); |
1da177e4 | 3963 | } |
0693e88e | 3964 | |
ec634fe3 | 3965 | return NETDEV_TX_OK; |
1da177e4 LT |
3966 | } |
3967 | ||
3968 | /* | |
3969 | * in broadcast mode, we send everything to all usable interfaces. | |
3970 | */ | |
3971 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
3972 | { | |
454d7c9b | 3973 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3974 | struct slave *slave, *start_at; |
3975 | struct net_device *tx_dev = NULL; | |
3976 | int i; | |
3977 | int res = 1; | |
3978 | ||
1da177e4 LT |
3979 | read_lock(&bond->curr_slave_lock); |
3980 | start_at = bond->curr_active_slave; | |
3981 | read_unlock(&bond->curr_slave_lock); | |
3982 | ||
3d632c3f | 3983 | if (!start_at) |
1da177e4 | 3984 | goto out; |
1da177e4 LT |
3985 | |
3986 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
3987 | if (IS_UP(slave->dev) && | |
3988 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3989 | bond_is_active_slave(slave)) { |
1da177e4 LT |
3990 | if (tx_dev) { |
3991 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
3992 | if (!skb2) { | |
a4aee5c8 | 3993 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 3994 | bond_dev->name); |
1da177e4 LT |
3995 | continue; |
3996 | } | |
3997 | ||
3998 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
3999 | if (res) { | |
04502430 | 4000 | kfree_skb(skb2); |
1da177e4 LT |
4001 | continue; |
4002 | } | |
4003 | } | |
4004 | tx_dev = slave->dev; | |
4005 | } | |
4006 | } | |
4007 | ||
3d632c3f | 4008 | if (tx_dev) |
1da177e4 | 4009 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4010 | |
4011 | out: | |
3d632c3f | 4012 | if (res) |
1da177e4 | 4013 | /* no suitable interface, frame not sent */ |
04502430 | 4014 | kfree_skb(skb); |
3d632c3f | 4015 | |
1da177e4 | 4016 | /* frame sent to all suitable interfaces */ |
ec634fe3 | 4017 | return NETDEV_TX_OK; |
1da177e4 LT |
4018 | } |
4019 | ||
4020 | /*------------------------- Device initialization ---------------------------*/ | |
4021 | ||
6f6652be JV |
4022 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4023 | { | |
4024 | switch (bond->params.xmit_policy) { | |
4025 | case BOND_XMIT_POLICY_LAYER23: | |
4026 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4027 | break; | |
4028 | case BOND_XMIT_POLICY_LAYER34: | |
4029 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4030 | break; | |
4031 | case BOND_XMIT_POLICY_LAYER2: | |
4032 | default: | |
4033 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4034 | break; | |
4035 | } | |
4036 | } | |
4037 | ||
bb1d9123 AG |
4038 | /* |
4039 | * Lookup the slave that corresponds to a qid | |
4040 | */ | |
4041 | static inline int bond_slave_override(struct bonding *bond, | |
4042 | struct sk_buff *skb) | |
4043 | { | |
4044 | int i, res = 1; | |
4045 | struct slave *slave = NULL; | |
4046 | struct slave *check_slave; | |
4047 | ||
0693e88e MM |
4048 | if (!skb->queue_mapping) |
4049 | return 1; | |
bb1d9123 AG |
4050 | |
4051 | /* Find out if any slaves have the same mapping as this skb. */ | |
4052 | bond_for_each_slave(bond, check_slave, i) { | |
4053 | if (check_slave->queue_id == skb->queue_mapping) { | |
4054 | slave = check_slave; | |
4055 | break; | |
4056 | } | |
4057 | } | |
4058 | ||
4059 | /* If the slave isn't UP, use default transmit policy. */ | |
4060 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4061 | (slave->link == BOND_LINK_UP)) { | |
4062 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4063 | } | |
4064 | ||
bb1d9123 AG |
4065 | return res; |
4066 | } | |
4067 | ||
374eeb5a | 4068 | |
bb1d9123 AG |
4069 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
4070 | { | |
4071 | /* | |
4072 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4073 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4074 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4075 | * way down to the bonding driver. | |
4076 | */ | |
fd0e435b PO |
4077 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4078 | ||
374eeb5a NH |
4079 | /* |
4080 | * Save the original txq to restore before passing to the driver | |
4081 | */ | |
df4ab5b3 | 4082 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 4083 | |
fd0e435b | 4084 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4085 | do { |
fd0e435b | 4086 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4087 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4088 | } |
4089 | return txq; | |
bb1d9123 AG |
4090 | } |
4091 | ||
0693e88e | 4092 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4093 | { |
bb1d9123 AG |
4094 | struct bonding *bond = netdev_priv(dev); |
4095 | ||
4096 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
4097 | if (!bond_slave_override(bond, skb)) | |
4098 | return NETDEV_TX_OK; | |
4099 | } | |
00829823 SH |
4100 | |
4101 | switch (bond->params.mode) { | |
4102 | case BOND_MODE_ROUNDROBIN: | |
4103 | return bond_xmit_roundrobin(skb, dev); | |
4104 | case BOND_MODE_ACTIVEBACKUP: | |
4105 | return bond_xmit_activebackup(skb, dev); | |
4106 | case BOND_MODE_XOR: | |
4107 | return bond_xmit_xor(skb, dev); | |
4108 | case BOND_MODE_BROADCAST: | |
4109 | return bond_xmit_broadcast(skb, dev); | |
4110 | case BOND_MODE_8023AD: | |
4111 | return bond_3ad_xmit_xor(skb, dev); | |
4112 | case BOND_MODE_ALB: | |
4113 | case BOND_MODE_TLB: | |
4114 | return bond_alb_xmit(skb, dev); | |
4115 | default: | |
4116 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4117 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4118 | dev->name, bond->params.mode); | |
00829823 | 4119 | WARN_ON_ONCE(1); |
04502430 | 4120 | kfree_skb(skb); |
00829823 SH |
4121 | return NETDEV_TX_OK; |
4122 | } | |
4123 | } | |
4124 | ||
0693e88e MM |
4125 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4126 | { | |
4127 | struct bonding *bond = netdev_priv(dev); | |
4128 | netdev_tx_t ret = NETDEV_TX_OK; | |
4129 | ||
4130 | /* | |
4131 | * If we risk deadlock from transmitting this in the | |
4132 | * netpoll path, tell netpoll to queue the frame for later tx | |
4133 | */ | |
4134 | if (is_netpoll_tx_blocked(dev)) | |
4135 | return NETDEV_TX_BUSY; | |
4136 | ||
4137 | read_lock(&bond->lock); | |
4138 | ||
4139 | if (bond->slave_cnt) | |
4140 | ret = __bond_start_xmit(skb, dev); | |
4141 | else | |
04502430 | 4142 | kfree_skb(skb); |
0693e88e MM |
4143 | |
4144 | read_unlock(&bond->lock); | |
4145 | ||
4146 | return ret; | |
4147 | } | |
00829823 | 4148 | |
1da177e4 LT |
4149 | /* |
4150 | * set bond mode specific net device operations | |
4151 | */ | |
a77b5325 | 4152 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4153 | { |
169a3e66 JV |
4154 | struct net_device *bond_dev = bond->dev; |
4155 | ||
1da177e4 LT |
4156 | switch (mode) { |
4157 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4158 | break; |
4159 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4160 | break; |
4161 | case BOND_MODE_XOR: | |
6f6652be | 4162 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4163 | break; |
4164 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4165 | break; |
4166 | case BOND_MODE_8023AD: | |
6f6652be | 4167 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4168 | break; |
1da177e4 | 4169 | case BOND_MODE_ALB: |
8f903c70 JV |
4170 | /* FALLTHRU */ |
4171 | case BOND_MODE_TLB: | |
1da177e4 LT |
4172 | break; |
4173 | default: | |
4174 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4175 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4176 | bond_dev->name, mode); | |
1da177e4 LT |
4177 | break; |
4178 | } | |
4179 | } | |
4180 | ||
bb5b052f AG |
4181 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
4182 | struct ethtool_cmd *ecmd) | |
4183 | { | |
4184 | struct bonding *bond = netdev_priv(bond_dev); | |
4185 | struct slave *slave; | |
4186 | int i; | |
4187 | unsigned long speed = 0; | |
4188 | ||
4189 | ecmd->duplex = DUPLEX_UNKNOWN; | |
4190 | ecmd->port = PORT_OTHER; | |
4191 | ||
4192 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
4193 | * do not need to check mode. Though link speed might not represent | |
4194 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
4195 | * this is an accurate maximum. | |
4196 | */ | |
4197 | read_lock(&bond->lock); | |
4198 | bond_for_each_slave(bond, slave, i) { | |
4199 | if (SLAVE_IS_OK(slave)) { | |
4200 | if (slave->speed != SPEED_UNKNOWN) | |
4201 | speed += slave->speed; | |
4202 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
4203 | slave->duplex != DUPLEX_UNKNOWN) | |
4204 | ecmd->duplex = slave->duplex; | |
4205 | } | |
4206 | } | |
4207 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
4208 | read_unlock(&bond->lock); | |
4209 | return 0; | |
4210 | } | |
4211 | ||
217df670 | 4212 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 4213 | struct ethtool_drvinfo *drvinfo) |
217df670 | 4214 | { |
7826d43f JP |
4215 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
4216 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
4217 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
4218 | BOND_ABI_VERSION); | |
217df670 JV |
4219 | } |
4220 | ||
7282d491 | 4221 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4222 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 4223 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 4224 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
4225 | }; |
4226 | ||
eb7cc59a | 4227 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4228 | .ndo_init = bond_init, |
9e71626c | 4229 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4230 | .ndo_open = bond_open, |
4231 | .ndo_stop = bond_close, | |
00829823 | 4232 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4233 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4234 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4235 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4236 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 4237 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 4238 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4239 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4240 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4241 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4242 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4243 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4244 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4245 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4246 | .ndo_poll_controller = bond_poll_controller, | |
4247 | #endif | |
9232ecca JP |
4248 | .ndo_add_slave = bond_enslave, |
4249 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4250 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
4251 | }; |
4252 | ||
b3f92b63 DG |
4253 | static const struct device_type bond_type = { |
4254 | .name = "bond", | |
4255 | }; | |
4256 | ||
9e2e61fb AW |
4257 | static void bond_destructor(struct net_device *bond_dev) |
4258 | { | |
4259 | struct bonding *bond = netdev_priv(bond_dev); | |
4260 | if (bond->wq) | |
4261 | destroy_workqueue(bond->wq); | |
4262 | free_netdev(bond_dev); | |
4263 | } | |
4264 | ||
181470fc | 4265 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4266 | { |
454d7c9b | 4267 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4268 | |
1da177e4 LT |
4269 | /* initialize rwlocks */ |
4270 | rwlock_init(&bond->lock); | |
4271 | rwlock_init(&bond->curr_slave_lock); | |
4272 | ||
d2991f75 | 4273 | bond->params = bonding_defaults; |
1da177e4 LT |
4274 | |
4275 | /* Initialize pointers */ | |
1da177e4 LT |
4276 | bond->dev = bond_dev; |
4277 | INIT_LIST_HEAD(&bond->vlan_list); | |
4278 | ||
4279 | /* Initialize the device entry points */ | |
181470fc | 4280 | ether_setup(bond_dev); |
eb7cc59a | 4281 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4282 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4283 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4284 | |
9e2e61fb | 4285 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4286 | |
b3f92b63 DG |
4287 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4288 | ||
1da177e4 LT |
4289 | /* Initialize the device options */ |
4290 | bond_dev->tx_queue_len = 0; | |
4291 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4292 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4293 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4294 | |
1da177e4 LT |
4295 | /* At first, we block adding VLANs. That's the only way to |
4296 | * prevent problems that occur when adding VLANs over an | |
4297 | * empty bond. The block will be removed once non-challenged | |
4298 | * slaves are enslaved. | |
4299 | */ | |
4300 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4301 | ||
932ff279 | 4302 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4303 | * transmitting */ |
4304 | bond_dev->features |= NETIF_F_LLTX; | |
4305 | ||
4306 | /* By default, we declare the bond to be fully | |
4307 | * VLAN hardware accelerated capable. Special | |
4308 | * care is taken in the various xmit functions | |
4309 | * when there are slaves that are not hw accel | |
4310 | * capable | |
4311 | */ | |
1da177e4 | 4312 | |
b2a103e6 | 4313 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4314 | NETIF_F_HW_VLAN_CTAG_TX | |
4315 | NETIF_F_HW_VLAN_CTAG_RX | | |
4316 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4317 | |
34324dc2 | 4318 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4319 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4320 | } |
4321 | ||
c67dfb29 EB |
4322 | /* |
4323 | * Destroy a bonding device. | |
4324 | * Must be under rtnl_lock when this function is called. | |
4325 | */ | |
4326 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4327 | { |
454d7c9b | 4328 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4329 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4330 | |
f6dc31a8 WC |
4331 | bond_netpoll_cleanup(bond_dev); |
4332 | ||
c67dfb29 | 4333 | /* Release the bonded slaves */ |
0896341a | 4334 | while (bond->first_slave != NULL) |
4335 | __bond_release_one(bond_dev, bond->first_slave->dev, true); | |
4336 | pr_info("%s: released all slaves\n", bond_dev->name); | |
c67dfb29 | 4337 | |
a434e43f JV |
4338 | list_del(&bond->bond_list); |
4339 | ||
f073c7ca TI |
4340 | bond_debug_unregister(bond); |
4341 | ||
f35188fa JV |
4342 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { |
4343 | list_del(&vlan->vlan_list); | |
4344 | kfree(vlan); | |
4345 | } | |
a434e43f JV |
4346 | } |
4347 | ||
1da177e4 LT |
4348 | /*------------------------- Module initialization ---------------------------*/ |
4349 | ||
4350 | /* | |
4351 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4352 | * number of the mode or its string name. A bit complicated because |
4353 | * some mode names are substrings of other names, and calls from sysfs | |
4354 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4355 | */ |
325dcf7a | 4356 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4357 | { |
54b87323 | 4358 | int modeint = -1, i, rv; |
a42e534f | 4359 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4360 | |
a42e534f JV |
4361 | for (p = (char *)buf; *p; p++) |
4362 | if (!(isdigit(*p) || isspace(*p))) | |
4363 | break; | |
4364 | ||
4365 | if (*p) | |
ece95f7f | 4366 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4367 | else |
54b87323 | 4368 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4369 | |
4370 | if (!rv) | |
4371 | return -1; | |
1da177e4 LT |
4372 | |
4373 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4374 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4375 | return tbl[i].mode; |
4376 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4377 | return tbl[i].mode; |
1da177e4 LT |
4378 | } |
4379 | ||
4380 | return -1; | |
4381 | } | |
4382 | ||
4383 | static int bond_check_params(struct bond_params *params) | |
4384 | { | |
5a5c5fd4 | 4385 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
8599b52e | 4386 | int arp_all_targets_value; |
f5b2b966 | 4387 | |
1da177e4 LT |
4388 | /* |
4389 | * Convert string parameters. | |
4390 | */ | |
4391 | if (mode) { | |
4392 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4393 | if (bond_mode == -1) { | |
a4aee5c8 | 4394 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4395 | mode == NULL ? "NULL" : mode); |
4396 | return -EINVAL; | |
4397 | } | |
4398 | } | |
4399 | ||
169a3e66 JV |
4400 | if (xmit_hash_policy) { |
4401 | if ((bond_mode != BOND_MODE_XOR) && | |
4402 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4403 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4404 | bond_mode_name(bond_mode)); |
4405 | } else { | |
4406 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4407 | xmit_hashtype_tbl); | |
4408 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4409 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4410 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4411 | xmit_hash_policy); |
4412 | return -EINVAL; | |
4413 | } | |
4414 | } | |
4415 | } | |
4416 | ||
1da177e4 LT |
4417 | if (lacp_rate) { |
4418 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4419 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4420 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4421 | } else { |
4422 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4423 | if (lacp_fast == -1) { | |
a4aee5c8 | 4424 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4425 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4426 | return -EINVAL; | |
4427 | } | |
4428 | } | |
4429 | } | |
4430 | ||
fd989c83 JV |
4431 | if (ad_select) { |
4432 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4433 | if (params->ad_select == -1) { | |
a4aee5c8 | 4434 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4435 | ad_select == NULL ? "NULL" : ad_select); |
4436 | return -EINVAL; | |
4437 | } | |
4438 | ||
4439 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4440 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4441 | } |
4442 | } else { | |
4443 | params->ad_select = BOND_AD_STABLE; | |
4444 | } | |
4445 | ||
f5841306 | 4446 | if (max_bonds < 0) { |
a4aee5c8 JP |
4447 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4448 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4449 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4450 | } | |
4451 | ||
4452 | if (miimon < 0) { | |
a4aee5c8 JP |
4453 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4454 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4455 | miimon = BOND_LINK_MON_INTERV; |
4456 | } | |
4457 | ||
4458 | if (updelay < 0) { | |
a4aee5c8 JP |
4459 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4460 | updelay, INT_MAX); | |
1da177e4 LT |
4461 | updelay = 0; |
4462 | } | |
4463 | ||
4464 | if (downdelay < 0) { | |
a4aee5c8 JP |
4465 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4466 | downdelay, INT_MAX); | |
1da177e4 LT |
4467 | downdelay = 0; |
4468 | } | |
4469 | ||
4470 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4471 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4472 | use_carrier); | |
1da177e4 LT |
4473 | use_carrier = 1; |
4474 | } | |
4475 | ||
ad246c99 BH |
4476 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4477 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4478 | num_peer_notif); | |
4479 | num_peer_notif = 1; | |
4480 | } | |
4481 | ||
1da177e4 LT |
4482 | /* reset values for 802.3ad */ |
4483 | if (bond_mode == BOND_MODE_8023AD) { | |
4484 | if (!miimon) { | |
a4aee5c8 | 4485 | 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 | 4486 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4487 | miimon = 100; |
4488 | } | |
4489 | } | |
4490 | ||
bb1d9123 AG |
4491 | if (tx_queues < 1 || tx_queues > 255) { |
4492 | pr_warning("Warning: tx_queues (%d) should be between " | |
4493 | "1 and 255, resetting to %d\n", | |
4494 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4495 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4496 | } | |
4497 | ||
ebd8e497 AG |
4498 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4499 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4500 | "not of valid value (0/1), so it was set to " | |
4501 | "0\n", all_slaves_active); | |
4502 | all_slaves_active = 0; | |
4503 | } | |
4504 | ||
c2952c31 FL |
4505 | if (resend_igmp < 0 || resend_igmp > 255) { |
4506 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4507 | "0 and 255, resetting to %d\n", | |
4508 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4509 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4510 | } | |
4511 | ||
1da177e4 LT |
4512 | /* reset values for TLB/ALB */ |
4513 | if ((bond_mode == BOND_MODE_TLB) || | |
4514 | (bond_mode == BOND_MODE_ALB)) { | |
4515 | if (!miimon) { | |
a4aee5c8 | 4516 | pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure and link speed which are essential for TLB/ALB load balancing\n"); |
3d632c3f | 4517 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4518 | miimon = 100; |
4519 | } | |
4520 | } | |
4521 | ||
4522 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4523 | 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", |
4524 | updelay); | |
1da177e4 LT |
4525 | } |
4526 | ||
4527 | if (!miimon) { | |
4528 | if (updelay || downdelay) { | |
4529 | /* just warn the user the up/down delay will have | |
4530 | * no effect since miimon is zero... | |
4531 | */ | |
a4aee5c8 JP |
4532 | 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", |
4533 | updelay, downdelay); | |
1da177e4 LT |
4534 | } |
4535 | } else { | |
4536 | /* don't allow arp monitoring */ | |
4537 | if (arp_interval) { | |
a4aee5c8 JP |
4538 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4539 | miimon, arp_interval); | |
1da177e4 LT |
4540 | arp_interval = 0; |
4541 | } | |
4542 | ||
4543 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4544 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4545 | updelay, miimon, | |
4546 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4547 | } |
4548 | ||
4549 | updelay /= miimon; | |
4550 | ||
4551 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4552 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4553 | downdelay, miimon, | |
4554 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4555 | } |
4556 | ||
4557 | downdelay /= miimon; | |
4558 | } | |
4559 | ||
4560 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4561 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4562 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4563 | arp_interval = BOND_LINK_ARP_INTERV; |
4564 | } | |
4565 | ||
5a5c5fd4 | 4566 | for (arp_ip_count = 0, i = 0; |
4567 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4568 | /* not complete check, but should be good enough to |
4569 | catch mistakes */ | |
5a5c5fd4 | 4570 | __be32 ip = in_aton(arp_ip_target[i]); |
4571 | if (!isdigit(arp_ip_target[i][0]) || ip == 0 || | |
4572 | ip == htonl(INADDR_BROADCAST)) { | |
a4aee5c8 | 4573 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5a5c5fd4 | 4574 | arp_ip_target[i]); |
1da177e4 LT |
4575 | arp_interval = 0; |
4576 | } else { | |
0afee4e8 VF |
4577 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4578 | arp_target[arp_ip_count++] = ip; | |
4579 | else | |
4580 | pr_warning("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", | |
4581 | &ip); | |
1da177e4 LT |
4582 | } |
4583 | } | |
4584 | ||
4585 | if (arp_interval && !arp_ip_count) { | |
4586 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4587 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4588 | arp_interval); | |
1da177e4 LT |
4589 | arp_interval = 0; |
4590 | } | |
4591 | ||
f5b2b966 JV |
4592 | if (arp_validate) { |
4593 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4594 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4595 | return -EINVAL; |
4596 | } | |
4597 | if (!arp_interval) { | |
a4aee5c8 | 4598 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4599 | return -EINVAL; |
4600 | } | |
4601 | ||
4602 | arp_validate_value = bond_parse_parm(arp_validate, | |
4603 | arp_validate_tbl); | |
4604 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4605 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4606 | arp_validate == NULL ? "NULL" : arp_validate); |
4607 | return -EINVAL; | |
4608 | } | |
4609 | } else | |
4610 | arp_validate_value = 0; | |
4611 | ||
8599b52e VF |
4612 | arp_all_targets_value = 0; |
4613 | if (arp_all_targets) { | |
4614 | arp_all_targets_value = bond_parse_parm(arp_all_targets, | |
4615 | arp_all_targets_tbl); | |
4616 | ||
4617 | if (arp_all_targets_value == -1) { | |
4618 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", | |
4619 | arp_all_targets); | |
4620 | arp_all_targets_value = 0; | |
4621 | } | |
4622 | } | |
4623 | ||
1da177e4 | 4624 | if (miimon) { |
a4aee5c8 | 4625 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4626 | } else if (arp_interval) { |
a4aee5c8 JP |
4627 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4628 | arp_interval, | |
4629 | arp_validate_tbl[arp_validate_value].modename, | |
4630 | arp_ip_count); | |
1da177e4 LT |
4631 | |
4632 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4633 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4634 | |
e5e2a8fd | 4635 | pr_info("\n"); |
1da177e4 | 4636 | |
b8a9787e | 4637 | } else if (max_bonds) { |
1da177e4 LT |
4638 | /* miimon and arp_interval not set, we need one so things |
4639 | * work as expected, see bonding.txt for details | |
4640 | */ | |
b2730f4f | 4641 | 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 |
4642 | } |
4643 | ||
4644 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4645 | /* currently, using a primary only makes sense | |
4646 | * in active backup, TLB or ALB modes | |
4647 | */ | |
a4aee5c8 JP |
4648 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4649 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4650 | primary = NULL; |
4651 | } | |
4652 | ||
a549952a JP |
4653 | if (primary && primary_reselect) { |
4654 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4655 | pri_reselect_tbl); | |
4656 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4657 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4658 | primary_reselect == |
4659 | NULL ? "NULL" : primary_reselect); | |
4660 | return -EINVAL; | |
4661 | } | |
4662 | } else { | |
4663 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4664 | } | |
4665 | ||
3915c1e8 JV |
4666 | if (fail_over_mac) { |
4667 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4668 | fail_over_mac_tbl); | |
4669 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4670 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4671 | arp_validate == NULL ? "NULL" : arp_validate); |
4672 | return -EINVAL; | |
4673 | } | |
4674 | ||
4675 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4676 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4677 | } else { |
4678 | fail_over_mac_value = BOND_FOM_NONE; | |
4679 | } | |
dd957c57 | 4680 | |
1da177e4 LT |
4681 | /* fill params struct with the proper values */ |
4682 | params->mode = bond_mode; | |
169a3e66 | 4683 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4684 | params->miimon = miimon; |
ad246c99 | 4685 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4686 | params->arp_interval = arp_interval; |
f5b2b966 | 4687 | params->arp_validate = arp_validate_value; |
8599b52e | 4688 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4689 | params->updelay = updelay; |
4690 | params->downdelay = downdelay; | |
4691 | params->use_carrier = use_carrier; | |
4692 | params->lacp_fast = lacp_fast; | |
4693 | params->primary[0] = 0; | |
a549952a | 4694 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4695 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4696 | params->tx_queues = tx_queues; |
ebd8e497 | 4697 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4698 | params->resend_igmp = resend_igmp; |
655f8919 | 4699 | params->min_links = min_links; |
1da177e4 LT |
4700 | |
4701 | if (primary) { | |
4702 | strncpy(params->primary, primary, IFNAMSIZ); | |
4703 | params->primary[IFNAMSIZ - 1] = 0; | |
4704 | } | |
4705 | ||
4706 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4707 | ||
4708 | return 0; | |
4709 | } | |
4710 | ||
0daa2303 | 4711 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4712 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4713 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4714 | |
e8a0464c DM |
4715 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4716 | struct netdev_queue *txq, | |
4717 | void *_unused) | |
c773e847 DM |
4718 | { |
4719 | lockdep_set_class(&txq->_xmit_lock, | |
4720 | &bonding_netdev_xmit_lock_key); | |
4721 | } | |
4722 | ||
4723 | static void bond_set_lockdep_class(struct net_device *dev) | |
4724 | { | |
cf508b12 DM |
4725 | lockdep_set_class(&dev->addr_list_lock, |
4726 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4727 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4728 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4729 | } |
4730 | ||
181470fc SH |
4731 | /* |
4732 | * Called from registration process | |
4733 | */ | |
4734 | static int bond_init(struct net_device *bond_dev) | |
4735 | { | |
4736 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4737 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4738 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4739 | |
4740 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4741 | ||
9fe0617d NH |
4742 | /* |
4743 | * Initialize locks that may be required during | |
4744 | * en/deslave operations. All of the bond_open work | |
4745 | * (of which this is part) should really be moved to | |
4746 | * a phase prior to dev_open | |
4747 | */ | |
4748 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4749 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4750 | ||
181470fc SH |
4751 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4752 | if (!bond->wq) | |
4753 | return -ENOMEM; | |
4754 | ||
4755 | bond_set_lockdep_class(bond_dev); | |
4756 | ||
ec87fd3b | 4757 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4758 | |
6151b3d4 | 4759 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4760 | |
f073c7ca TI |
4761 | bond_debug_register(bond); |
4762 | ||
409cc1f8 JP |
4763 | /* Ensure valid dev_addr */ |
4764 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4765 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4766 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4767 | |
181470fc SH |
4768 | return 0; |
4769 | } | |
4770 | ||
88ead977 EB |
4771 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4772 | { | |
4773 | if (tb[IFLA_ADDRESS]) { | |
4774 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4775 | return -EINVAL; | |
4776 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4777 | return -EADDRNOTAVAIL; | |
4778 | } | |
4779 | return 0; | |
4780 | } | |
4781 | ||
d40156aa | 4782 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4783 | { |
efacb309 | 4784 | return tx_queues; |
d5da4510 JP |
4785 | } |
4786 | ||
88ead977 | 4787 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4788 | .kind = "bond", |
4789 | .priv_size = sizeof(struct bonding), | |
4790 | .setup = bond_setup, | |
4791 | .validate = bond_validate, | |
4792 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4793 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4794 | as for TX queues */ | |
88ead977 EB |
4795 | }; |
4796 | ||
dfe60397 | 4797 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4798 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4799 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4800 | * set up our sysfs entries. | |
4801 | */ | |
ec87fd3b | 4802 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4803 | { |
4804 | struct net_device *bond_dev; | |
4805 | int res; | |
4806 | ||
4807 | rtnl_lock(); | |
027ea041 | 4808 | |
1c5cae81 JP |
4809 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4810 | name ? name : "bond%d", | |
4811 | bond_setup, tx_queues); | |
dfe60397 | 4812 | if (!bond_dev) { |
a4aee5c8 | 4813 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4814 | rtnl_unlock(); |
4815 | return -ENOMEM; | |
dfe60397 MW |
4816 | } |
4817 | ||
ec87fd3b | 4818 | dev_net_set(bond_dev, net); |
88ead977 EB |
4819 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4820 | ||
dfe60397 | 4821 | res = register_netdevice(bond_dev); |
0daa2303 | 4822 | |
e826eafa PO |
4823 | netif_carrier_off(bond_dev); |
4824 | ||
7e083840 | 4825 | rtnl_unlock(); |
9e2e61fb AW |
4826 | if (res < 0) |
4827 | bond_destructor(bond_dev); | |
30c15ba9 | 4828 | return res; |
dfe60397 MW |
4829 | } |
4830 | ||
2c8c1e72 | 4831 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4832 | { |
15449745 | 4833 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4834 | |
4835 | bn->net = net; | |
4836 | INIT_LIST_HEAD(&bn->dev_list); | |
4837 | ||
ec87fd3b | 4838 | bond_create_proc_dir(bn); |
4c22400a | 4839 | bond_create_sysfs(bn); |
87a7b84b | 4840 | |
15449745 | 4841 | return 0; |
ec87fd3b EB |
4842 | } |
4843 | ||
2c8c1e72 | 4844 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4845 | { |
15449745 | 4846 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4847 | struct bonding *bond, *tmp_bond; |
4848 | LIST_HEAD(list); | |
ec87fd3b | 4849 | |
4c22400a | 4850 | bond_destroy_sysfs(bn); |
ec87fd3b | 4851 | bond_destroy_proc_dir(bn); |
69b0216a | 4852 | |
4853 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4854 | rtnl_lock(); | |
4855 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4856 | unregister_netdevice_queue(bond->dev, &list); | |
4857 | unregister_netdevice_many(&list); | |
4858 | rtnl_unlock(); | |
ec87fd3b EB |
4859 | } |
4860 | ||
4861 | static struct pernet_operations bond_net_ops = { | |
4862 | .init = bond_net_init, | |
4863 | .exit = bond_net_exit, | |
15449745 EB |
4864 | .id = &bond_net_id, |
4865 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4866 | }; |
4867 | ||
1da177e4 LT |
4868 | static int __init bonding_init(void) |
4869 | { | |
1da177e4 LT |
4870 | int i; |
4871 | int res; | |
4872 | ||
bd33acc3 | 4873 | pr_info("%s", bond_version); |
1da177e4 | 4874 | |
dfe60397 | 4875 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4876 | if (res) |
dfe60397 | 4877 | goto out; |
1da177e4 | 4878 | |
15449745 | 4879 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4880 | if (res) |
4881 | goto out; | |
027ea041 | 4882 | |
88ead977 EB |
4883 | res = rtnl_link_register(&bond_link_ops); |
4884 | if (res) | |
6639104b | 4885 | goto err_link; |
88ead977 | 4886 | |
f073c7ca TI |
4887 | bond_create_debugfs(); |
4888 | ||
1da177e4 | 4889 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4890 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4891 | if (res) |
4892 | goto err; | |
1da177e4 LT |
4893 | } |
4894 | ||
1da177e4 | 4895 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4896 | out: |
1da177e4 | 4897 | return res; |
88ead977 EB |
4898 | err: |
4899 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 4900 | err_link: |
15449745 | 4901 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4902 | goto out; |
dfe60397 | 4903 | |
1da177e4 LT |
4904 | } |
4905 | ||
4906 | static void __exit bonding_exit(void) | |
4907 | { | |
4908 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4909 | ||
f073c7ca | 4910 | bond_destroy_debugfs(); |
ae68c398 | 4911 | |
4de79c73 | 4912 | rtnl_link_unregister(&bond_link_ops); |
ffcdedb6 | 4913 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4914 | |
4915 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4916 | /* |
4917 | * Make sure we don't have an imbalance on our netpoll blocking | |
4918 | */ | |
4919 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4920 | #endif |
1da177e4 LT |
4921 | } |
4922 | ||
4923 | module_init(bonding_init); | |
4924 | module_exit(bonding_exit); | |
4925 | MODULE_LICENSE("GPL"); | |
4926 | MODULE_VERSION(DRV_VERSION); | |
4927 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4928 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 4929 | MODULE_ALIAS_RTNL_LINK("bond"); |