]>
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; |
ebd8e497 | 109 | static int all_slaves_active = 0; |
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 | ||
276 | if (mode < 0 || 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 JP |
725 | if (!rtnl_trylock()) { |
726 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); | |
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 | read_unlock(&bond->lock); |
743 | } | |
744 | ||
379b7383 | 745 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
746 | { |
747 | struct bonding *bond = container_of(work, struct bonding, | |
94265cf5 | 748 | mcast_work.work); |
ad999eee | 749 | |
5a37e8ca FL |
750 | bond_resend_igmp_join_requests(bond); |
751 | } | |
752 | ||
303d1cbf | 753 | /* Flush bond's hardware addresses from slave |
1da177e4 | 754 | */ |
303d1cbf | 755 | static void bond_hw_addr_flush(struct net_device *bond_dev, |
3d632c3f | 756 | struct net_device *slave_dev) |
1da177e4 | 757 | { |
454d7c9b | 758 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 759 | |
303d1cbf JV |
760 | dev_uc_unsync(slave_dev, bond_dev); |
761 | dev_mc_unsync(slave_dev, bond_dev); | |
1da177e4 LT |
762 | |
763 | if (bond->params.mode == BOND_MODE_8023AD) { | |
764 | /* del lacpdu mc addr from mc list */ | |
765 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
766 | ||
22bedad3 | 767 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
768 | } |
769 | } | |
770 | ||
771 | /*--------------------------- Active slave change ---------------------------*/ | |
772 | ||
303d1cbf JV |
773 | /* Update the hardware address list and promisc/allmulti for the new and |
774 | * old active slaves (if any). Modes that are !USES_PRIMARY keep all | |
775 | * slaves up date at all times; only the USES_PRIMARY modes need to call | |
776 | * this function to swap these settings during a failover. | |
1da177e4 | 777 | */ |
303d1cbf JV |
778 | static void bond_hw_addr_swap(struct bonding *bond, struct slave *new_active, |
779 | struct slave *old_active) | |
1da177e4 | 780 | { |
1da177e4 | 781 | if (old_active) { |
3d632c3f | 782 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 783 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 784 | |
3d632c3f | 785 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 786 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 787 | |
303d1cbf | 788 | bond_hw_addr_flush(bond->dev, old_active->dev); |
1da177e4 LT |
789 | } |
790 | ||
791 | if (new_active) { | |
7e1a1ac1 | 792 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 793 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 794 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 795 | |
3d632c3f | 796 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 797 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 798 | |
d632ce98 | 799 | netif_addr_lock_bh(bond->dev); |
303d1cbf JV |
800 | dev_uc_sync(new_active->dev, bond->dev); |
801 | dev_mc_sync(new_active->dev, bond->dev); | |
d632ce98 | 802 | netif_addr_unlock_bh(bond->dev); |
1da177e4 LT |
803 | } |
804 | } | |
805 | ||
ae0d6750 | 806 | /** |
807 | * bond_set_dev_addr - clone slave's address to bond | |
808 | * @bond_dev: bond net device | |
809 | * @slave_dev: slave net device | |
810 | * | |
811 | * Should be called with RTNL held. | |
812 | */ | |
813 | static void bond_set_dev_addr(struct net_device *bond_dev, | |
814 | struct net_device *slave_dev) | |
815 | { | |
008aebde NA |
816 | pr_debug("bond_dev=%p slave_dev=%p slave_dev->addr_len=%d\n", |
817 | bond_dev, slave_dev, slave_dev->addr_len); | |
ae0d6750 | 818 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
819 | bond_dev->addr_assign_type = NET_ADDR_STOLEN; | |
820 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
821 | } | |
822 | ||
3915c1e8 JV |
823 | /* |
824 | * bond_do_fail_over_mac | |
825 | * | |
826 | * Perform special MAC address swapping for fail_over_mac settings | |
827 | * | |
828 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
829 | */ | |
830 | static void bond_do_fail_over_mac(struct bonding *bond, | |
831 | struct slave *new_active, | |
832 | struct slave *old_active) | |
1f78d9f9 HE |
833 | __releases(&bond->curr_slave_lock) |
834 | __releases(&bond->lock) | |
835 | __acquires(&bond->lock) | |
836 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
837 | { |
838 | u8 tmp_mac[ETH_ALEN]; | |
839 | struct sockaddr saddr; | |
840 | int rv; | |
841 | ||
842 | switch (bond->params.fail_over_mac) { | |
843 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 844 | if (new_active) { |
7d26bb10 WP |
845 | write_unlock_bh(&bond->curr_slave_lock); |
846 | read_unlock(&bond->lock); | |
ae0d6750 | 847 | bond_set_dev_addr(bond->dev, new_active->dev); |
7d26bb10 WP |
848 | read_lock(&bond->lock); |
849 | write_lock_bh(&bond->curr_slave_lock); | |
850 | } | |
3915c1e8 JV |
851 | break; |
852 | case BOND_FOM_FOLLOW: | |
853 | /* | |
854 | * if new_active && old_active, swap them | |
855 | * if just old_active, do nothing (going to no active slave) | |
856 | * if just new_active, set new_active to bond's MAC | |
857 | */ | |
858 | if (!new_active) | |
859 | return; | |
860 | ||
861 | write_unlock_bh(&bond->curr_slave_lock); | |
862 | read_unlock(&bond->lock); | |
863 | ||
864 | if (old_active) { | |
865 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
866 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
867 | ETH_ALEN); | |
868 | saddr.sa_family = new_active->dev->type; | |
869 | } else { | |
870 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
871 | saddr.sa_family = bond->dev->type; | |
872 | } | |
873 | ||
874 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
875 | if (rv) { | |
a4aee5c8 | 876 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
877 | bond->dev->name, -rv, new_active->dev->name); |
878 | goto out; | |
879 | } | |
880 | ||
881 | if (!old_active) | |
882 | goto out; | |
883 | ||
884 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
885 | saddr.sa_family = old_active->dev->type; | |
886 | ||
887 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
888 | if (rv) | |
a4aee5c8 | 889 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
890 | bond->dev->name, -rv, new_active->dev->name); |
891 | out: | |
892 | read_lock(&bond->lock); | |
893 | write_lock_bh(&bond->curr_slave_lock); | |
894 | break; | |
895 | default: | |
a4aee5c8 | 896 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
897 | bond->dev->name, bond->params.fail_over_mac); |
898 | break; | |
899 | } | |
900 | ||
901 | } | |
902 | ||
a549952a JP |
903 | static bool bond_should_change_active(struct bonding *bond) |
904 | { | |
905 | struct slave *prim = bond->primary_slave; | |
906 | struct slave *curr = bond->curr_active_slave; | |
907 | ||
908 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
909 | return true; | |
910 | if (bond->force_primary) { | |
911 | bond->force_primary = false; | |
912 | return true; | |
913 | } | |
914 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
915 | (prim->speed < curr->speed || | |
916 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
917 | return false; | |
918 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
919 | return false; | |
920 | return true; | |
921 | } | |
3915c1e8 | 922 | |
1da177e4 LT |
923 | /** |
924 | * find_best_interface - select the best available slave to be the active one | |
925 | * @bond: our bonding struct | |
926 | * | |
927 | * Warning: Caller must hold curr_slave_lock for writing. | |
928 | */ | |
929 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
930 | { | |
931 | struct slave *new_active, *old_active; | |
932 | struct slave *bestslave = NULL; | |
933 | int mintime = bond->params.updelay; | |
934 | int i; | |
935 | ||
49b4ad92 | 936 | new_active = bond->curr_active_slave; |
1da177e4 LT |
937 | |
938 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 939 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 940 | new_active = bond->first_slave; |
3d632c3f | 941 | else |
1da177e4 | 942 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
943 | } |
944 | ||
1da177e4 | 945 | if ((bond->primary_slave) && |
a549952a JP |
946 | bond->primary_slave->link == BOND_LINK_UP && |
947 | bond_should_change_active(bond)) { | |
1da177e4 LT |
948 | new_active = bond->primary_slave; |
949 | } | |
950 | ||
951 | /* remember where to stop iterating over the slaves */ | |
952 | old_active = new_active; | |
953 | ||
954 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
955 | if (new_active->link == BOND_LINK_UP) { |
956 | return new_active; | |
957 | } else if (new_active->link == BOND_LINK_BACK && | |
958 | IS_UP(new_active->dev)) { | |
959 | /* link up, but waiting for stabilization */ | |
960 | if (new_active->delay < mintime) { | |
961 | mintime = new_active->delay; | |
962 | bestslave = new_active; | |
1da177e4 LT |
963 | } |
964 | } | |
965 | } | |
966 | ||
967 | return bestslave; | |
968 | } | |
969 | ||
ad246c99 BH |
970 | static bool bond_should_notify_peers(struct bonding *bond) |
971 | { | |
972 | struct slave *slave = bond->curr_active_slave; | |
973 | ||
974 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
975 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
976 | ||
977 | if (!slave || !bond->send_peer_notif || | |
978 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
979 | return false; | |
980 | ||
981 | bond->send_peer_notif--; | |
982 | return true; | |
983 | } | |
984 | ||
1da177e4 LT |
985 | /** |
986 | * change_active_interface - change the active slave into the specified one | |
987 | * @bond: our bonding struct | |
988 | * @new: the new slave to make the active one | |
989 | * | |
990 | * Set the new slave to the bond's settings and unset them on the old | |
991 | * curr_active_slave. | |
992 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
993 | * | |
994 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
995 | * because it is apparently the best available slave we have, even though its | |
996 | * updelay hasn't timed out yet. | |
997 | * | |
3915c1e8 JV |
998 | * If new_active is not NULL, caller must hold bond->lock for read and |
999 | * curr_slave_lock for write_bh. | |
1da177e4 | 1000 | */ |
a77b5325 | 1001 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1002 | { |
1003 | struct slave *old_active = bond->curr_active_slave; | |
1004 | ||
3d632c3f | 1005 | if (old_active == new_active) |
1da177e4 | 1006 | return; |
1da177e4 LT |
1007 | |
1008 | if (new_active) { | |
b2220cad JV |
1009 | new_active->jiffies = jiffies; |
1010 | ||
1da177e4 LT |
1011 | if (new_active->link == BOND_LINK_BACK) { |
1012 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1013 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1014 | bond->dev->name, new_active->dev->name, | |
1015 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1016 | } |
1017 | ||
1018 | new_active->delay = 0; | |
1019 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1020 | |
3d632c3f | 1021 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1022 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1023 | |
58402054 | 1024 | if (bond_is_lb(bond)) |
1da177e4 | 1025 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1026 | } else { |
1027 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1028 | pr_info("%s: making interface %s the new active one.\n", |
1029 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1030 | } |
1031 | } | |
1032 | } | |
1033 | ||
3d632c3f | 1034 | if (USES_PRIMARY(bond->params.mode)) |
303d1cbf | 1035 | bond_hw_addr_swap(bond, new_active, old_active); |
1da177e4 | 1036 | |
58402054 | 1037 | if (bond_is_lb(bond)) { |
1da177e4 | 1038 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1039 | if (old_active) |
1040 | bond_set_slave_inactive_flags(old_active); | |
1041 | if (new_active) | |
1042 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1043 | } else { |
1044 | bond->curr_active_slave = new_active; | |
1045 | } | |
c3ade5ca JV |
1046 | |
1047 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1048 | if (old_active) |
c3ade5ca | 1049 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1050 | |
1051 | if (new_active) { | |
ad246c99 BH |
1052 | bool should_notify_peers = false; |
1053 | ||
c3ade5ca | 1054 | bond_set_slave_active_flags(new_active); |
2ab82852 | 1055 | |
709f8a45 OG |
1056 | if (bond->params.fail_over_mac) |
1057 | bond_do_fail_over_mac(bond, new_active, | |
1058 | old_active); | |
3915c1e8 | 1059 | |
ad246c99 BH |
1060 | if (netif_running(bond->dev)) { |
1061 | bond->send_peer_notif = | |
1062 | bond->params.num_peer_notif; | |
1063 | should_notify_peers = | |
1064 | bond_should_notify_peers(bond); | |
1065 | } | |
1066 | ||
01f3109d OG |
1067 | write_unlock_bh(&bond->curr_slave_lock); |
1068 | read_unlock(&bond->lock); | |
1069 | ||
b7bc2a5b | 1070 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 1071 | if (should_notify_peers) |
b7bc2a5b AW |
1072 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
1073 | bond->dev); | |
01f3109d OG |
1074 | |
1075 | read_lock(&bond->lock); | |
1076 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1077 | } |
c3ade5ca | 1078 | } |
a2fd940f | 1079 | |
5a37e8ca | 1080 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
1081 | * all were sent on curr_active_slave. |
1082 | * resend only if bond is brought up with the affected | |
1083 | * bonding modes and the retransmission is enabled */ | |
1084 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
1085 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
1086 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 1087 | bond->igmp_retrans = bond->params.resend_igmp; |
5a37e8ca | 1088 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); |
a2fd940f | 1089 | } |
1da177e4 LT |
1090 | } |
1091 | ||
1092 | /** | |
1093 | * bond_select_active_slave - select a new active slave, if needed | |
1094 | * @bond: our bonding struct | |
1095 | * | |
3d632c3f | 1096 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1097 | * - The old curr_active_slave has been released or lost its link. |
1098 | * - The primary_slave has got its link back. | |
1099 | * - A slave has got its link back and there's no old curr_active_slave. | |
1100 | * | |
3915c1e8 | 1101 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1102 | */ |
a77b5325 | 1103 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1104 | { |
1105 | struct slave *best_slave; | |
ff59c456 | 1106 | int rv; |
1da177e4 LT |
1107 | |
1108 | best_slave = bond_find_best_slave(bond); | |
1109 | if (best_slave != bond->curr_active_slave) { | |
1110 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1111 | rv = bond_set_carrier(bond); |
1112 | if (!rv) | |
1113 | return; | |
1114 | ||
1115 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1116 | pr_info("%s: first active interface up!\n", |
1117 | bond->dev->name); | |
ff59c456 | 1118 | } else { |
a4aee5c8 JP |
1119 | pr_info("%s: now running without any active interface !\n", |
1120 | bond->dev->name); | |
ff59c456 | 1121 | } |
1da177e4 LT |
1122 | } |
1123 | } | |
1124 | ||
1125 | /*--------------------------- slave list handling ---------------------------*/ | |
1126 | ||
1127 | /* | |
1128 | * This function attaches the slave to the end of list. | |
1129 | * | |
1130 | * bond->lock held for writing by caller. | |
1131 | */ | |
1132 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1133 | { | |
1134 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1135 | new_slave->next = new_slave; | |
1136 | new_slave->prev = new_slave; | |
1137 | bond->first_slave = new_slave; | |
1138 | } else { | |
1139 | new_slave->next = bond->first_slave; | |
1140 | new_slave->prev = bond->first_slave->prev; | |
1141 | new_slave->next->prev = new_slave; | |
1142 | new_slave->prev->next = new_slave; | |
1143 | } | |
1144 | ||
1145 | bond->slave_cnt++; | |
1146 | } | |
1147 | ||
1148 | /* | |
1149 | * This function detaches the slave from the list. | |
1150 | * WARNING: no check is made to verify if the slave effectively | |
1151 | * belongs to <bond>. | |
1152 | * Nothing is freed on return, structures are just unchained. | |
1153 | * If any slave pointer in bond was pointing to <slave>, | |
1154 | * it should be changed by the calling function. | |
1155 | * | |
1156 | * bond->lock held for writing by caller. | |
1157 | */ | |
1158 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1159 | { | |
3d632c3f | 1160 | if (slave->next) |
1da177e4 | 1161 | slave->next->prev = slave->prev; |
1da177e4 | 1162 | |
3d632c3f | 1163 | if (slave->prev) |
1da177e4 | 1164 | slave->prev->next = slave->next; |
1da177e4 LT |
1165 | |
1166 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1167 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1168 | bond->first_slave = slave->next; | |
1169 | } else { | |
1170 | bond->first_slave = NULL; /* slave was the last one */ | |
1171 | } | |
1172 | } | |
1173 | ||
1174 | slave->next = NULL; | |
1175 | slave->prev = NULL; | |
1176 | bond->slave_cnt--; | |
1177 | } | |
1178 | ||
f6dc31a8 | 1179 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1180 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1181 | { |
8a8efa22 AW |
1182 | struct netpoll *np; |
1183 | int err = 0; | |
f6dc31a8 | 1184 | |
47be03a2 | 1185 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
1186 | err = -ENOMEM; |
1187 | if (!np) | |
1188 | goto out; | |
1189 | ||
47be03a2 | 1190 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
1191 | if (err) { |
1192 | kfree(np); | |
1193 | goto out; | |
f6dc31a8 | 1194 | } |
8a8efa22 AW |
1195 | slave->np = np; |
1196 | out: | |
1197 | return err; | |
1198 | } | |
1199 | static inline void slave_disable_netpoll(struct slave *slave) | |
1200 | { | |
1201 | struct netpoll *np = slave->np; | |
1202 | ||
1203 | if (!np) | |
1204 | return; | |
1205 | ||
1206 | slave->np = NULL; | |
2cde6acd | 1207 | __netpoll_free_async(np); |
8a8efa22 AW |
1208 | } |
1209 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1210 | { | |
1211 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1212 | return false; | |
1213 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1214 | return false; | |
1215 | return true; | |
f6dc31a8 WC |
1216 | } |
1217 | ||
1218 | static void bond_poll_controller(struct net_device *bond_dev) | |
1219 | { | |
8a8efa22 AW |
1220 | } |
1221 | ||
1222 | static void __bond_netpoll_cleanup(struct bonding *bond) | |
1223 | { | |
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 | } |
f6dc31a8 WC |
1231 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1232 | { | |
1233 | struct bonding *bond = netdev_priv(bond_dev); | |
8a8efa22 AW |
1234 | |
1235 | read_lock(&bond->lock); | |
1236 | __bond_netpoll_cleanup(bond); | |
1237 | read_unlock(&bond->lock); | |
1238 | } | |
1239 | ||
47be03a2 | 1240 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1241 | { |
1242 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1243 | struct slave *slave; |
8a8efa22 | 1244 | int i, err = 0; |
f6dc31a8 WC |
1245 | |
1246 | read_lock(&bond->lock); | |
f6dc31a8 | 1247 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1248 | err = slave_enable_netpoll(slave); |
1249 | if (err) { | |
1250 | __bond_netpoll_cleanup(bond); | |
1251 | break; | |
f6dc31a8 WC |
1252 | } |
1253 | } | |
1254 | read_unlock(&bond->lock); | |
8a8efa22 | 1255 | return err; |
f6dc31a8 WC |
1256 | } |
1257 | ||
8a8efa22 AW |
1258 | static struct netpoll_info *bond_netpoll_info(struct bonding *bond) |
1259 | { | |
1260 | return bond->dev->npinfo; | |
1261 | } | |
f6dc31a8 | 1262 | |
8a8efa22 AW |
1263 | #else |
1264 | static inline int slave_enable_netpoll(struct slave *slave) | |
1265 | { | |
1266 | return 0; | |
1267 | } | |
1268 | static inline void slave_disable_netpoll(struct slave *slave) | |
1269 | { | |
1270 | } | |
f6dc31a8 WC |
1271 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1272 | { | |
1273 | } | |
f6dc31a8 WC |
1274 | #endif |
1275 | ||
1da177e4 LT |
1276 | /*---------------------------------- IOCTL ----------------------------------*/ |
1277 | ||
c8f44aff MM |
1278 | static netdev_features_t bond_fix_features(struct net_device *dev, |
1279 | netdev_features_t features) | |
8531c5ff | 1280 | { |
8531c5ff | 1281 | struct slave *slave; |
b2a103e6 | 1282 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1283 | netdev_features_t mask; |
8e3babcd | 1284 | int i; |
8531c5ff | 1285 | |
b2a103e6 | 1286 | read_lock(&bond->lock); |
b63365a2 | 1287 | |
b2a103e6 MM |
1288 | if (!bond->first_slave) { |
1289 | /* Disable adding VLANs to empty bond. But why? --mq */ | |
1290 | features |= NETIF_F_VLAN_CHALLENGED; | |
1291 | goto out; | |
1292 | } | |
b63365a2 | 1293 | |
b2a103e6 | 1294 | mask = features; |
b63365a2 | 1295 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1296 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1297 | |
54ef3137 | 1298 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1299 | features = netdev_increment_features(features, |
1300 | slave->dev->features, | |
b2a103e6 MM |
1301 | mask); |
1302 | } | |
b0ce3508 | 1303 | features = netdev_add_tso_features(features, mask); |
b2a103e6 MM |
1304 | |
1305 | out: | |
1306 | read_unlock(&bond->lock); | |
1307 | return features; | |
1308 | } | |
1309 | ||
62f2a3a4 MM |
1310 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1311 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1312 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1313 | |
1314 | static void bond_compute_features(struct bonding *bond) | |
1315 | { | |
1316 | struct slave *slave; | |
1317 | struct net_device *bond_dev = bond->dev; | |
c8f44aff | 1318 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1319 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1320 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1321 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1322 | int i; |
b6fe83e9 | 1323 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
b2a103e6 MM |
1324 | |
1325 | read_lock(&bond->lock); | |
1326 | ||
1327 | if (!bond->first_slave) | |
1328 | goto done; | |
1329 | ||
1330 | bond_for_each_slave(bond, slave, i) { | |
278339a4 | 1331 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1332 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1333 | ||
b6fe83e9 | 1334 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1335 | if (slave->dev->hard_header_len > max_hard_header_len) |
1336 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1337 | |
1338 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1339 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1340 | } |
8531c5ff | 1341 | |
b63365a2 | 1342 | done: |
b2a103e6 | 1343 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1344 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1345 | bond_dev->gso_max_segs = gso_max_segs; |
1346 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1347 | |
b6fe83e9 ED |
1348 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1349 | bond_dev->priv_flags = flags | dst_release_flag; | |
1350 | ||
b2a103e6 MM |
1351 | read_unlock(&bond->lock); |
1352 | ||
1353 | netdev_change_features(bond_dev); | |
8531c5ff AK |
1354 | } |
1355 | ||
872254dd MS |
1356 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1357 | struct net_device *slave_dev) | |
1358 | { | |
00829823 | 1359 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1360 | |
1361 | bond_dev->type = slave_dev->type; | |
1362 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1363 | bond_dev->addr_len = slave_dev->addr_len; | |
1364 | ||
1365 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1366 | slave_dev->addr_len); | |
1367 | } | |
1368 | ||
5b2c4dd2 | 1369 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1370 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1371 | */ |
1372 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1373 | struct slave *slave, |
1374 | struct bonding *bond) | |
5b2c4dd2 | 1375 | { |
2d7011ca | 1376 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1377 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1378 | skb->pkt_type != PACKET_BROADCAST && |
1379 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1380 | return false; |
5b2c4dd2 JP |
1381 | return true; |
1382 | } | |
1383 | return false; | |
1384 | } | |
1385 | ||
8a4eb573 | 1386 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1387 | { |
8a4eb573 | 1388 | struct sk_buff *skb = *pskb; |
f1c1775a | 1389 | struct slave *slave; |
0bd80dad | 1390 | struct bonding *bond; |
de063b70 ED |
1391 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1392 | struct slave *); | |
13a8e0c8 | 1393 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1394 | |
8a4eb573 JP |
1395 | skb = skb_share_check(skb, GFP_ATOMIC); |
1396 | if (unlikely(!skb)) | |
1397 | return RX_HANDLER_CONSUMED; | |
1398 | ||
1399 | *pskb = skb; | |
5b2c4dd2 | 1400 | |
35d48903 JP |
1401 | slave = bond_slave_get_rcu(skb->dev); |
1402 | bond = slave->bond; | |
0bd80dad JP |
1403 | |
1404 | if (bond->params.arp_interval) | |
f1c1775a | 1405 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1406 | |
4d97480b MH |
1407 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1408 | if (recv_probe) { | |
de063b70 ED |
1409 | ret = recv_probe(skb, bond, slave); |
1410 | if (ret == RX_HANDLER_CONSUMED) { | |
1411 | consume_skb(skb); | |
1412 | return ret; | |
3aba891d JP |
1413 | } |
1414 | } | |
1415 | ||
0bd80dad | 1416 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1417 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1418 | } |
1419 | ||
35d48903 | 1420 | skb->dev = bond->dev; |
5b2c4dd2 | 1421 | |
0bd80dad | 1422 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1423 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1424 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1425 | |
541ac7c9 CG |
1426 | if (unlikely(skb_cow_head(skb, |
1427 | skb->data - skb_mac_header(skb)))) { | |
1428 | kfree_skb(skb); | |
8a4eb573 | 1429 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1430 | } |
35d48903 | 1431 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1432 | } |
1433 | ||
13a8e0c8 | 1434 | return ret; |
5b2c4dd2 JP |
1435 | } |
1436 | ||
471cb5a3 JP |
1437 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1438 | struct net_device *slave_dev) | |
1439 | { | |
1440 | int err; | |
1441 | ||
1442 | err = netdev_master_upper_dev_link(slave_dev, bond_dev); | |
1443 | if (err) | |
1444 | return err; | |
1445 | slave_dev->flags |= IFF_SLAVE; | |
1446 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1447 | return 0; | |
1448 | } | |
1449 | ||
1450 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1451 | struct net_device *slave_dev) | |
1452 | { | |
1453 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1454 | slave_dev->flags &= ~IFF_SLAVE; | |
1455 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1456 | } | |
1457 | ||
1da177e4 | 1458 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1459 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1460 | { |
454d7c9b | 1461 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1462 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1463 | struct slave *new_slave = NULL; |
1da177e4 LT |
1464 | struct sockaddr addr; |
1465 | int link_reporting; | |
8599b52e | 1466 | int res = 0, i; |
1da177e4 | 1467 | |
c772dde3 BH |
1468 | if (!bond->params.use_carrier && |
1469 | slave_dev->ethtool_ops->get_link == NULL && | |
1470 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1471 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1472 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1473 | } |
1474 | ||
1da177e4 LT |
1475 | /* already enslaved */ |
1476 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1477 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1478 | return -EBUSY; |
1479 | } | |
1480 | ||
1481 | /* vlan challenged mutual exclusion */ | |
1482 | /* no need to lock since we're protected by rtnl_lock */ | |
1483 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1484 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1485 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1486 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1487 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1488 | return -EPERM; |
1489 | } else { | |
a4aee5c8 JP |
1490 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1491 | bond_dev->name, slave_dev->name, | |
1492 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1493 | } |
1494 | } else { | |
5a03cdb7 | 1495 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1496 | } |
1497 | ||
217df670 JV |
1498 | /* |
1499 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1500 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1501 | * the current ifenslave will set the interface down prior to |
1502 | * enslaving it; the old ifenslave will not. | |
1503 | */ | |
1504 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1505 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1506 | slave_dev->name); |
1507 | res = -EPERM; | |
1508 | goto err_undo_flags; | |
1509 | } | |
1da177e4 | 1510 | |
872254dd MS |
1511 | /* set bonding device ether type by slave - bonding netdevices are |
1512 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1513 | * there is a need to override some of the type dependent attribs/funcs. | |
1514 | * | |
1515 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1516 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1517 | */ | |
1518 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1519 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1520 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1521 | bond_dev->name, |
1522 | bond_dev->type, slave_dev->type); | |
75c78500 | 1523 | |
b7bc2a5b AW |
1524 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1525 | bond_dev); | |
3ca5b404 JP |
1526 | res = notifier_to_errno(res); |
1527 | if (res) { | |
1528 | pr_err("%s: refused to change device type\n", | |
1529 | bond_dev->name); | |
1530 | res = -EBUSY; | |
1531 | goto err_undo_flags; | |
1532 | } | |
75c78500 | 1533 | |
32a806c1 | 1534 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1535 | dev_uc_flush(bond_dev); |
22bedad3 | 1536 | dev_mc_flush(bond_dev); |
32a806c1 | 1537 | |
e36b9d16 MS |
1538 | if (slave_dev->type != ARPHRD_ETHER) |
1539 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1540 | else { |
e36b9d16 | 1541 | ether_setup(bond_dev); |
550fd08c NH |
1542 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1543 | } | |
75c78500 | 1544 | |
b7bc2a5b AW |
1545 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1546 | bond_dev); | |
e36b9d16 | 1547 | } |
872254dd | 1548 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1549 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1550 | slave_dev->name, | |
1551 | slave_dev->type, bond_dev->type); | |
1552 | res = -EINVAL; | |
1553 | goto err_undo_flags; | |
872254dd MS |
1554 | } |
1555 | ||
eb7cc59a | 1556 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1557 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1558 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1559 | bond_dev->name); | |
3915c1e8 JV |
1560 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1561 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1562 | 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", |
1563 | bond_dev->name); | |
2ab82852 MS |
1564 | res = -EOPNOTSUPP; |
1565 | goto err_undo_flags; | |
1566 | } | |
1da177e4 LT |
1567 | } |
1568 | ||
8d8fc29d AW |
1569 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1570 | ||
c20811a7 JP |
1571 | /* If this is the first slave, then we need to set the master's hardware |
1572 | * address to be the same as the slave's. */ | |
97a1e639 | 1573 | if (!bond->slave_cnt && bond->dev->addr_assign_type == NET_ADDR_RANDOM) |
409cc1f8 | 1574 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1575 | |
243cb4e5 | 1576 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1577 | if (!new_slave) { |
1578 | res = -ENOMEM; | |
1579 | goto err_undo_flags; | |
1580 | } | |
1581 | ||
bb1d9123 AG |
1582 | /* |
1583 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1584 | * is set via sysfs or module option if desired. | |
1585 | */ | |
1586 | new_slave->queue_id = 0; | |
1587 | ||
b15ba0fb JP |
1588 | /* Save slave's original mtu and then set it to match the bond */ |
1589 | new_slave->original_mtu = slave_dev->mtu; | |
1590 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1591 | if (res) { | |
1592 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1593 | goto err_free; | |
1594 | } | |
1595 | ||
217df670 JV |
1596 | /* |
1597 | * Save slave's original ("permanent") mac address for modes | |
1598 | * that need it, and for restoring it upon release, and then | |
1599 | * set it to the master's address | |
1600 | */ | |
1601 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1602 | |
dd957c57 | 1603 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1604 | /* |
1605 | * Set slave to master's mac address. The application already | |
1606 | * set the master's mac address to that of the first slave | |
1607 | */ | |
1608 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1609 | addr.sa_family = slave_dev->type; | |
1610 | res = dev_set_mac_address(slave_dev, &addr); | |
1611 | if (res) { | |
5a03cdb7 | 1612 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1613 | goto err_restore_mtu; |
2ab82852 | 1614 | } |
217df670 | 1615 | } |
1da177e4 | 1616 | |
471cb5a3 | 1617 | res = bond_master_upper_dev_link(bond_dev, slave_dev); |
c2edacf8 | 1618 | if (res) { |
471cb5a3 | 1619 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); |
569f0c4d | 1620 | goto err_restore_mac; |
c2edacf8 | 1621 | } |
5b2c4dd2 | 1622 | |
217df670 JV |
1623 | /* open the slave since the application closed it */ |
1624 | res = dev_open(slave_dev); | |
1625 | if (res) { | |
3d632c3f | 1626 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1627 | goto err_unset_master; |
1da177e4 LT |
1628 | } |
1629 | ||
35d48903 | 1630 | new_slave->bond = bond; |
1da177e4 | 1631 | new_slave->dev = slave_dev; |
0b680e75 | 1632 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1633 | |
58402054 | 1634 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1635 | /* bond_alb_init_slave() must be called before all other stages since |
1636 | * it might fail and we do not want to have to undo everything | |
1637 | */ | |
1638 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1639 | if (res) |
569f0c4d | 1640 | goto err_close; |
1da177e4 LT |
1641 | } |
1642 | ||
303d1cbf JV |
1643 | /* If the mode USES_PRIMARY, then the following is handled by |
1644 | * bond_change_active_slave(). | |
1da177e4 LT |
1645 | */ |
1646 | if (!USES_PRIMARY(bond->params.mode)) { | |
1647 | /* set promiscuity level to new slave */ | |
1648 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1649 | res = dev_set_promiscuity(slave_dev, 1); |
1650 | if (res) | |
1651 | goto err_close; | |
1da177e4 LT |
1652 | } |
1653 | ||
1654 | /* set allmulti level to new slave */ | |
1655 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1656 | res = dev_set_allmulti(slave_dev, 1); |
1657 | if (res) | |
1658 | goto err_close; | |
1da177e4 LT |
1659 | } |
1660 | ||
b9e40857 | 1661 | netif_addr_lock_bh(bond_dev); |
303d1cbf JV |
1662 | |
1663 | dev_mc_sync_multiple(slave_dev, bond_dev); | |
1664 | dev_uc_sync_multiple(slave_dev, bond_dev); | |
1665 | ||
b9e40857 | 1666 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1667 | } |
1668 | ||
1669 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1670 | /* add lacpdu mc addr to mc list */ | |
1671 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1672 | ||
22bedad3 | 1673 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1674 | } |
1675 | ||
1676 | bond_add_vlans_on_slave(bond, slave_dev); | |
1677 | ||
1678 | write_lock_bh(&bond->lock); | |
1679 | ||
1680 | bond_attach_slave(bond, new_slave); | |
1681 | ||
1682 | new_slave->delay = 0; | |
1683 | new_slave->link_failure_count = 0; | |
1684 | ||
3915c1e8 JV |
1685 | write_unlock_bh(&bond->lock); |
1686 | ||
b2a103e6 MM |
1687 | bond_compute_features(bond); |
1688 | ||
876254ae VF |
1689 | bond_update_speed_duplex(new_slave); |
1690 | ||
3915c1e8 JV |
1691 | read_lock(&bond->lock); |
1692 | ||
f31c7937 MK |
1693 | new_slave->last_arp_rx = jiffies - |
1694 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
8599b52e VF |
1695 | for (i = 0; i < BOND_MAX_ARP_TARGETS; i++) |
1696 | new_slave->target_last_arp_rx[i] = new_slave->last_arp_rx; | |
f5b2b966 | 1697 | |
1da177e4 LT |
1698 | if (bond->params.miimon && !bond->params.use_carrier) { |
1699 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1700 | ||
1701 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1702 | /* | |
1703 | * miimon is set but a bonded network driver | |
1704 | * does not support ETHTOOL/MII and | |
1705 | * arp_interval is not set. Note: if | |
1706 | * use_carrier is enabled, we will never go | |
1707 | * here (because netif_carrier is always | |
1708 | * supported); thus, we don't need to change | |
1709 | * the messages for netif_carrier. | |
1710 | */ | |
a4aee5c8 | 1711 | 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 | 1712 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1713 | } else if (link_reporting == -1) { |
1714 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1715 | 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", |
1716 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1717 | } |
1718 | } | |
1719 | ||
1720 | /* check for initial state */ | |
f31c7937 MK |
1721 | if (bond->params.miimon) { |
1722 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1723 | if (bond->params.updelay) { | |
1724 | new_slave->link = BOND_LINK_BACK; | |
1725 | new_slave->delay = bond->params.updelay; | |
1726 | } else { | |
1727 | new_slave->link = BOND_LINK_UP; | |
1728 | } | |
1da177e4 | 1729 | } else { |
f31c7937 | 1730 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1731 | } |
f31c7937 MK |
1732 | } else if (bond->params.arp_interval) { |
1733 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1734 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1735 | } else { |
f31c7937 | 1736 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1737 | } |
1738 | ||
f31c7937 MK |
1739 | if (new_slave->link != BOND_LINK_DOWN) |
1740 | new_slave->jiffies = jiffies; | |
1741 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1742 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1743 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1744 | ||
1da177e4 LT |
1745 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1746 | /* if there is a primary slave, remember it */ | |
a549952a | 1747 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1748 | bond->primary_slave = new_slave; |
a549952a JP |
1749 | bond->force_primary = true; |
1750 | } | |
1da177e4 LT |
1751 | } |
1752 | ||
3915c1e8 JV |
1753 | write_lock_bh(&bond->curr_slave_lock); |
1754 | ||
1da177e4 LT |
1755 | switch (bond->params.mode) { |
1756 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1757 | bond_set_slave_inactive_flags(new_slave); |
1758 | bond_select_active_slave(bond); | |
1da177e4 LT |
1759 | break; |
1760 | case BOND_MODE_8023AD: | |
1761 | /* in 802.3ad mode, the internal mechanism | |
1762 | * will activate the slaves in the selected | |
1763 | * aggregator | |
1764 | */ | |
1765 | bond_set_slave_inactive_flags(new_slave); | |
1766 | /* if this is the first slave */ | |
1767 | if (bond->slave_cnt == 1) { | |
1768 | SLAVE_AD_INFO(new_slave).id = 1; | |
1769 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1770 | * can be called only after the mac address of the bond is set | |
1771 | */ | |
56d00c67 | 1772 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 LT |
1773 | } else { |
1774 | SLAVE_AD_INFO(new_slave).id = | |
1775 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1776 | } | |
1777 | ||
1778 | bond_3ad_bind_slave(new_slave); | |
1779 | break; | |
1780 | case BOND_MODE_TLB: | |
1781 | case BOND_MODE_ALB: | |
e30bc066 | 1782 | bond_set_active_slave(new_slave); |
059fe7a5 | 1783 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1784 | bond_select_active_slave(bond); |
1da177e4 LT |
1785 | break; |
1786 | default: | |
5a03cdb7 | 1787 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1788 | |
1789 | /* always active in trunk mode */ | |
e30bc066 | 1790 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1791 | |
1792 | /* In trunking mode there is little meaning to curr_active_slave | |
1793 | * anyway (it holds no special properties of the bond device), | |
1794 | * so we can change it without calling change_active_interface() | |
1795 | */ | |
4e591b93 | 1796 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
1da177e4 | 1797 | bond->curr_active_slave = new_slave; |
3d632c3f | 1798 | |
1da177e4 LT |
1799 | break; |
1800 | } /* switch(bond_mode) */ | |
1801 | ||
3915c1e8 JV |
1802 | write_unlock_bh(&bond->curr_slave_lock); |
1803 | ||
ff59c456 JV |
1804 | bond_set_carrier(bond); |
1805 | ||
f6dc31a8 | 1806 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 AW |
1807 | slave_dev->npinfo = bond_netpoll_info(bond); |
1808 | if (slave_dev->npinfo) { | |
1809 | if (slave_enable_netpoll(new_slave)) { | |
1810 | read_unlock(&bond->lock); | |
1811 | pr_info("Error, %s: master_dev is using netpoll, " | |
1812 | "but new slave device does not support netpoll.\n", | |
1813 | bond_dev->name); | |
1814 | res = -EBUSY; | |
f7d9821a | 1815 | goto err_detach; |
8a8efa22 | 1816 | } |
f6dc31a8 WC |
1817 | } |
1818 | #endif | |
8a8efa22 | 1819 | |
3915c1e8 | 1820 | read_unlock(&bond->lock); |
1da177e4 | 1821 | |
b76cdba9 MW |
1822 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1823 | if (res) | |
f7d9821a | 1824 | goto err_detach; |
b76cdba9 | 1825 | |
35d48903 JP |
1826 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1827 | new_slave); | |
1828 | if (res) { | |
1829 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1830 | goto err_dest_symlinks; | |
1831 | } | |
1832 | ||
a4aee5c8 JP |
1833 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1834 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1835 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1836 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1837 | |
1838 | /* enslave is successful */ | |
1839 | return 0; | |
1840 | ||
1841 | /* Undo stages on error */ | |
35d48903 JP |
1842 | err_dest_symlinks: |
1843 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1844 | ||
f7d9821a | 1845 | err_detach: |
303d1cbf JV |
1846 | if (!USES_PRIMARY(bond->params.mode)) |
1847 | bond_hw_addr_flush(bond_dev, slave_dev); | |
1848 | ||
a506e7b4 | 1849 | bond_del_vlans_from_slave(bond, slave_dev); |
f7d9821a | 1850 | write_lock_bh(&bond->lock); |
1851 | bond_detach_slave(bond, new_slave); | |
3c5913b5 | 1852 | if (bond->primary_slave == new_slave) |
1853 | bond->primary_slave = NULL; | |
3c5913b5 | 1854 | if (bond->curr_active_slave == new_slave) { |
c6cdcf6d | 1855 | bond_change_active_slave(bond, NULL); |
1856 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1857 | read_lock(&bond->lock); |
1858 | write_lock_bh(&bond->curr_slave_lock); | |
3c5913b5 | 1859 | bond_select_active_slave(bond); |
1860 | write_unlock_bh(&bond->curr_slave_lock); | |
1861 | read_unlock(&bond->lock); | |
c6cdcf6d | 1862 | } else { |
1863 | write_unlock_bh(&bond->lock); | |
3c5913b5 | 1864 | } |
fc7a72ac | 1865 | slave_disable_netpoll(new_slave); |
f7d9821a | 1866 | |
1da177e4 | 1867 | err_close: |
b6a5a7b9 | 1868 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
1869 | dev_close(slave_dev); |
1870 | ||
569f0c4d | 1871 | err_unset_master: |
471cb5a3 | 1872 | bond_upper_dev_unlink(bond_dev, slave_dev); |
569f0c4d | 1873 | |
1da177e4 | 1874 | err_restore_mac: |
dd957c57 | 1875 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1876 | /* XXX TODO - fom follow mode needs to change master's |
1877 | * MAC if this slave's MAC is in use by the bond, or at | |
1878 | * least print a warning. | |
1879 | */ | |
2ab82852 MS |
1880 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1881 | addr.sa_family = slave_dev->type; | |
1882 | dev_set_mac_address(slave_dev, &addr); | |
1883 | } | |
1da177e4 | 1884 | |
b15ba0fb JP |
1885 | err_restore_mtu: |
1886 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1887 | ||
1da177e4 LT |
1888 | err_free: |
1889 | kfree(new_slave); | |
1890 | ||
1891 | err_undo_flags: | |
b2a103e6 | 1892 | bond_compute_features(bond); |
b8fad459 NA |
1893 | /* Enslave of first slave has failed and we need to fix master's mac */ |
1894 | if (bond->slave_cnt == 0 && | |
1895 | ether_addr_equal(bond_dev->dev_addr, slave_dev->dev_addr)) | |
1896 | eth_hw_addr_random(bond_dev); | |
3d632c3f | 1897 | |
1da177e4 LT |
1898 | return res; |
1899 | } | |
1900 | ||
1901 | /* | |
1902 | * Try to release the slave device <slave> from the bond device <master> | |
1903 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1904 | * is write-locked. If "all" is true it means that the function is being called |
1905 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1906 | * |
1907 | * The rules for slave state should be: | |
1908 | * for Active/Backup: | |
1909 | * Active stays on all backups go down | |
1910 | * for Bonded connections: | |
1911 | * The first up interface should be left on and all others downed. | |
1912 | */ | |
0896341a | 1913 | static int __bond_release_one(struct net_device *bond_dev, |
1914 | struct net_device *slave_dev, | |
1915 | bool all) | |
1da177e4 | 1916 | { |
454d7c9b | 1917 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1918 | struct slave *slave, *oldcurrent; |
1919 | struct sockaddr addr; | |
c8f44aff | 1920 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1921 | |
1922 | /* slave is not a slave or master is not master of this slave */ | |
1923 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1924 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1925 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1926 | bond_dev->name, slave_dev->name); |
1927 | return -EINVAL; | |
1928 | } | |
1929 | ||
e843fa50 | 1930 | block_netpoll_tx(); |
1da177e4 LT |
1931 | write_lock_bh(&bond->lock); |
1932 | ||
1933 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1934 | if (!slave) { | |
1935 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1936 | pr_info("%s: %s not enslaved\n", |
1937 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1938 | write_unlock_bh(&bond->lock); |
e843fa50 | 1939 | unblock_netpoll_tx(); |
1da177e4 LT |
1940 | return -EINVAL; |
1941 | } | |
1942 | ||
fcd99434 | 1943 | write_unlock_bh(&bond->lock); |
35d48903 JP |
1944 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1945 | * for this slave anymore. | |
1946 | */ | |
1947 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1948 | write_lock_bh(&bond->lock); |
1949 | ||
0896341a | 1950 | if (!all && !bond->params.fail_over_mac) { |
2e42e474 | 1951 | if (ether_addr_equal(bond_dev->dev_addr, slave->perm_hwaddr) && |
8e95a202 | 1952 | bond->slave_cnt > 1) |
a4aee5c8 JP |
1953 | 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", |
1954 | bond_dev->name, slave_dev->name, | |
1955 | slave->perm_hwaddr, | |
1956 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1957 | } |
1958 | ||
1959 | /* Inform AD package of unbinding of slave. */ | |
1960 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1961 | /* must be called before the slave is | |
1962 | * detached from the list | |
1963 | */ | |
1964 | bond_3ad_unbind_slave(slave); | |
1965 | } | |
1966 | ||
a4aee5c8 JP |
1967 | pr_info("%s: releasing %s interface %s\n", |
1968 | bond_dev->name, | |
e30bc066 | 1969 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1970 | slave_dev->name); |
1da177e4 LT |
1971 | |
1972 | oldcurrent = bond->curr_active_slave; | |
1973 | ||
1974 | bond->current_arp_slave = NULL; | |
1975 | ||
1976 | /* release the slave from its bond */ | |
1977 | bond_detach_slave(bond, slave); | |
1978 | ||
3d632c3f | 1979 | if (bond->primary_slave == slave) |
1da177e4 | 1980 | bond->primary_slave = NULL; |
1da177e4 | 1981 | |
3d632c3f | 1982 | if (oldcurrent == slave) |
1da177e4 | 1983 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1984 | |
58402054 | 1985 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1986 | /* Must be called only after the slave has been |
1987 | * detached from the list and the curr_active_slave | |
1988 | * has been cleared (if our_slave == old_current), | |
1989 | * but before a new active slave is selected. | |
1990 | */ | |
2543331d | 1991 | write_unlock_bh(&bond->lock); |
1da177e4 | 1992 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1993 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1994 | } |
1995 | ||
0896341a | 1996 | if (all) { |
1997 | bond->curr_active_slave = NULL; | |
1998 | } else if (oldcurrent == slave) { | |
059fe7a5 JV |
1999 | /* |
2000 | * Note that we hold RTNL over this sequence, so there | |
2001 | * is no concern that another slave add/remove event | |
2002 | * will interfere. | |
2003 | */ | |
2004 | write_unlock_bh(&bond->lock); | |
2005 | read_lock(&bond->lock); | |
2006 | write_lock_bh(&bond->curr_slave_lock); | |
2007 | ||
1da177e4 LT |
2008 | bond_select_active_slave(bond); |
2009 | ||
059fe7a5 JV |
2010 | write_unlock_bh(&bond->curr_slave_lock); |
2011 | read_unlock(&bond->lock); | |
2012 | write_lock_bh(&bond->lock); | |
2013 | } | |
2014 | ||
1da177e4 | 2015 | if (bond->slave_cnt == 0) { |
ff59c456 | 2016 | bond_set_carrier(bond); |
409cc1f8 | 2017 | eth_hw_addr_random(bond_dev); |
1da177e4 | 2018 | |
cc0e4070 | 2019 | if (bond_vlan_used(bond)) { |
a4aee5c8 JP |
2020 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2021 | bond_dev->name, bond_dev->name); | |
2022 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2023 | bond_dev->name); | |
1da177e4 | 2024 | } |
1da177e4 LT |
2025 | } |
2026 | ||
2027 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2028 | unblock_netpoll_tx(); |
1da177e4 | 2029 | |
80028ea1 | 2030 | if (bond->slave_cnt == 0) { |
2af73d4b | 2031 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
2032 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
2033 | } | |
2af73d4b | 2034 | |
b2a103e6 MM |
2035 | bond_compute_features(bond); |
2036 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2037 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
2038 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
2039 | bond_dev->name, slave_dev->name, bond_dev->name); | |
2040 | ||
b76cdba9 MW |
2041 | /* must do this from outside any spinlocks */ |
2042 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2043 | ||
1da177e4 LT |
2044 | bond_del_vlans_from_slave(bond, slave_dev); |
2045 | ||
303d1cbf JV |
2046 | /* If the mode USES_PRIMARY, then this cases was handled above by |
2047 | * bond_change_active_slave(..., NULL) | |
1da177e4 LT |
2048 | */ |
2049 | if (!USES_PRIMARY(bond->params.mode)) { | |
2050 | /* unset promiscuity level from slave */ | |
3d632c3f | 2051 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2052 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2053 | |
2054 | /* unset allmulti level from slave */ | |
3d632c3f | 2055 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2056 | dev_set_allmulti(slave_dev, -1); |
1da177e4 | 2057 | |
303d1cbf | 2058 | bond_hw_addr_flush(bond_dev, slave_dev); |
1da177e4 LT |
2059 | } |
2060 | ||
471cb5a3 | 2061 | bond_upper_dev_unlink(bond_dev, slave_dev); |
1da177e4 | 2062 | |
8a8efa22 | 2063 | slave_disable_netpoll(slave); |
f6dc31a8 | 2064 | |
1da177e4 LT |
2065 | /* close slave before restoring its mac address */ |
2066 | dev_close(slave_dev); | |
2067 | ||
3915c1e8 | 2068 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2069 | /* restore original ("permanent") mac address */ |
2070 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2071 | addr.sa_family = slave_dev->type; | |
2072 | dev_set_mac_address(slave_dev, &addr); | |
2073 | } | |
1da177e4 | 2074 | |
b15ba0fb JP |
2075 | dev_set_mtu(slave_dev, slave->original_mtu); |
2076 | ||
2d7011ca | 2077 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2078 | |
2079 | kfree(slave); | |
2080 | ||
2081 | return 0; /* deletion OK */ | |
2082 | } | |
2083 | ||
0896341a | 2084 | /* A wrapper used because of ndo_del_link */ |
2085 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
2086 | { | |
2087 | return __bond_release_one(bond_dev, slave_dev, false); | |
2088 | } | |
2089 | ||
d90a162a | 2090 | /* |
dadaa10b | 2091 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2092 | * Must be under rtnl_lock when this function is called. |
2093 | */ | |
26d8ee75 | 2094 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2095 | struct net_device *slave_dev) | |
d90a162a | 2096 | { |
454d7c9b | 2097 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2098 | int ret; |
2099 | ||
2100 | ret = bond_release(bond_dev, slave_dev); | |
2101 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2102 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2103 | pr_info("%s: destroying bond %s.\n", |
2104 | bond_dev->name, bond_dev->name); | |
9e71626c | 2105 | unregister_netdevice(bond_dev); |
d90a162a MS |
2106 | } |
2107 | return ret; | |
2108 | } | |
2109 | ||
1da177e4 LT |
2110 | /* |
2111 | * This function changes the active slave to slave <slave_dev>. | |
2112 | * It returns -EINVAL in the following cases. | |
2113 | * - <slave_dev> is not found in the list. | |
2114 | * - There is not active slave now. | |
2115 | * - <slave_dev> is already active. | |
2116 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2117 | * - <slave_dev> is not running. | |
3d632c3f SH |
2118 | * In these cases, this function does nothing. |
2119 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2120 | */ |
2121 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2122 | { | |
454d7c9b | 2123 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2124 | struct slave *old_active = NULL; |
2125 | struct slave *new_active = NULL; | |
2126 | int res = 0; | |
2127 | ||
3d632c3f | 2128 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2129 | return -EINVAL; |
1da177e4 | 2130 | |
471cb5a3 JP |
2131 | /* Verify that bond_dev is indeed the master of slave_dev */ |
2132 | if (!(slave_dev->flags & IFF_SLAVE) || | |
2133 | !netdev_has_upper_dev(slave_dev, bond_dev)) | |
1da177e4 | 2134 | return -EINVAL; |
1da177e4 | 2135 | |
059fe7a5 | 2136 | read_lock(&bond->lock); |
1da177e4 | 2137 | |
059fe7a5 | 2138 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2139 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2140 | read_unlock(&bond->curr_slave_lock); |
2141 | ||
1da177e4 LT |
2142 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2143 | ||
2144 | /* | |
2145 | * Changing to the current active: do nothing; return success. | |
2146 | */ | |
2147 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2148 | read_unlock(&bond->lock); |
1da177e4 LT |
2149 | return 0; |
2150 | } | |
2151 | ||
2152 | if ((new_active) && | |
2153 | (old_active) && | |
2154 | (new_active->link == BOND_LINK_UP) && | |
2155 | IS_UP(new_active->dev)) { | |
e843fa50 | 2156 | block_netpoll_tx(); |
059fe7a5 | 2157 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2158 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2159 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2160 | unblock_netpoll_tx(); |
3d632c3f | 2161 | } else |
1da177e4 | 2162 | res = -EINVAL; |
1da177e4 | 2163 | |
059fe7a5 | 2164 | read_unlock(&bond->lock); |
1da177e4 LT |
2165 | |
2166 | return res; | |
2167 | } | |
2168 | ||
1da177e4 LT |
2169 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2170 | { | |
454d7c9b | 2171 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2172 | |
2173 | info->bond_mode = bond->params.mode; | |
2174 | info->miimon = bond->params.miimon; | |
2175 | ||
6603a6f2 | 2176 | read_lock(&bond->lock); |
1da177e4 | 2177 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2178 | read_unlock(&bond->lock); |
1da177e4 LT |
2179 | |
2180 | return 0; | |
2181 | } | |
2182 | ||
2183 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2184 | { | |
454d7c9b | 2185 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2186 | struct slave *slave; |
689c96cc | 2187 | int i, res = -ENODEV; |
1da177e4 | 2188 | |
6603a6f2 | 2189 | read_lock(&bond->lock); |
1da177e4 LT |
2190 | |
2191 | bond_for_each_slave(bond, slave, i) { | |
2192 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2193 | res = 0; |
2194 | strcpy(info->slave_name, slave->dev->name); | |
2195 | info->link = slave->link; | |
e30bc066 | 2196 | info->state = bond_slave_state(slave); |
689c96cc | 2197 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2198 | break; |
2199 | } | |
2200 | } | |
2201 | ||
6603a6f2 | 2202 | read_unlock(&bond->lock); |
1da177e4 | 2203 | |
689c96cc | 2204 | return res; |
1da177e4 LT |
2205 | } |
2206 | ||
2207 | /*-------------------------------- Monitoring -------------------------------*/ | |
2208 | ||
1da177e4 | 2209 | |
f0c76d61 JV |
2210 | static int bond_miimon_inspect(struct bonding *bond) |
2211 | { | |
2212 | struct slave *slave; | |
2213 | int i, link_state, commit = 0; | |
41f89100 JP |
2214 | bool ignore_updelay; |
2215 | ||
2216 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2217 | |
2218 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2219 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2220 | |
f0c76d61 | 2221 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2222 | |
2223 | switch (slave->link) { | |
f0c76d61 JV |
2224 | case BOND_LINK_UP: |
2225 | if (link_state) | |
2226 | continue; | |
1da177e4 | 2227 | |
f0c76d61 JV |
2228 | slave->link = BOND_LINK_FAIL; |
2229 | slave->delay = bond->params.downdelay; | |
2230 | if (slave->delay) { | |
a4aee5c8 JP |
2231 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2232 | bond->dev->name, | |
2233 | (bond->params.mode == | |
2234 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2235 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2236 | "active " : "backup ") : "", |
2237 | slave->dev->name, | |
2238 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2239 | } |
f0c76d61 JV |
2240 | /*FALLTHRU*/ |
2241 | case BOND_LINK_FAIL: | |
2242 | if (link_state) { | |
2243 | /* | |
2244 | * recovered before downdelay expired | |
2245 | */ | |
2246 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2247 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2248 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2249 | bond->dev->name, | |
2250 | (bond->params.downdelay - slave->delay) * | |
2251 | bond->params.miimon, | |
2252 | slave->dev->name); | |
f0c76d61 | 2253 | continue; |
1da177e4 | 2254 | } |
f0c76d61 JV |
2255 | |
2256 | if (slave->delay <= 0) { | |
2257 | slave->new_link = BOND_LINK_DOWN; | |
2258 | commit++; | |
2259 | continue; | |
1da177e4 | 2260 | } |
1da177e4 | 2261 | |
f0c76d61 JV |
2262 | slave->delay--; |
2263 | break; | |
2264 | ||
2265 | case BOND_LINK_DOWN: | |
2266 | if (!link_state) | |
2267 | continue; | |
2268 | ||
2269 | slave->link = BOND_LINK_BACK; | |
2270 | slave->delay = bond->params.updelay; | |
2271 | ||
2272 | if (slave->delay) { | |
a4aee5c8 JP |
2273 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2274 | bond->dev->name, slave->dev->name, | |
2275 | ignore_updelay ? 0 : | |
2276 | bond->params.updelay * | |
2277 | bond->params.miimon); | |
f0c76d61 JV |
2278 | } |
2279 | /*FALLTHRU*/ | |
2280 | case BOND_LINK_BACK: | |
2281 | if (!link_state) { | |
2282 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2283 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2284 | bond->dev->name, | |
2285 | (bond->params.updelay - slave->delay) * | |
2286 | bond->params.miimon, | |
2287 | slave->dev->name); | |
f0c76d61 JV |
2288 | |
2289 | continue; | |
2290 | } | |
2291 | ||
41f89100 JP |
2292 | if (ignore_updelay) |
2293 | slave->delay = 0; | |
2294 | ||
f0c76d61 JV |
2295 | if (slave->delay <= 0) { |
2296 | slave->new_link = BOND_LINK_UP; | |
2297 | commit++; | |
41f89100 | 2298 | ignore_updelay = false; |
f0c76d61 | 2299 | continue; |
1da177e4 | 2300 | } |
f0c76d61 JV |
2301 | |
2302 | slave->delay--; | |
1da177e4 | 2303 | break; |
f0c76d61 JV |
2304 | } |
2305 | } | |
1da177e4 | 2306 | |
f0c76d61 JV |
2307 | return commit; |
2308 | } | |
1da177e4 | 2309 | |
f0c76d61 JV |
2310 | static void bond_miimon_commit(struct bonding *bond) |
2311 | { | |
2312 | struct slave *slave; | |
2313 | int i; | |
2314 | ||
2315 | bond_for_each_slave(bond, slave, i) { | |
2316 | switch (slave->new_link) { | |
2317 | case BOND_LINK_NOCHANGE: | |
2318 | continue; | |
1da177e4 | 2319 | |
f0c76d61 JV |
2320 | case BOND_LINK_UP: |
2321 | slave->link = BOND_LINK_UP; | |
2322 | slave->jiffies = jiffies; | |
2323 | ||
2324 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2325 | /* prevent it from being the active one */ | |
e30bc066 | 2326 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2327 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2328 | /* make it immediately active */ | |
e30bc066 | 2329 | bond_set_active_slave(slave); |
f0c76d61 JV |
2330 | } else if (slave != bond->primary_slave) { |
2331 | /* prevent it from being the active one */ | |
e30bc066 | 2332 | bond_set_backup_slave(slave); |
1da177e4 | 2333 | } |
1da177e4 | 2334 | |
5d30530e | 2335 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 | 2336 | bond->dev->name, slave->dev->name, |
db4e9b2b NA |
2337 | slave->speed == SPEED_UNKNOWN ? 0 : slave->speed, |
2338 | slave->duplex ? "full" : "half"); | |
1da177e4 | 2339 | |
f0c76d61 JV |
2340 | /* notify ad that the link status has changed */ |
2341 | if (bond->params.mode == BOND_MODE_8023AD) | |
2342 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2343 | |
58402054 | 2344 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2345 | bond_alb_handle_link_change(bond, slave, |
2346 | BOND_LINK_UP); | |
1da177e4 | 2347 | |
f0c76d61 JV |
2348 | if (!bond->curr_active_slave || |
2349 | (slave == bond->primary_slave)) | |
2350 | goto do_failover; | |
1da177e4 | 2351 | |
f0c76d61 | 2352 | continue; |
059fe7a5 | 2353 | |
f0c76d61 | 2354 | case BOND_LINK_DOWN: |
fba4acda JV |
2355 | if (slave->link_failure_count < UINT_MAX) |
2356 | slave->link_failure_count++; | |
2357 | ||
f0c76d61 | 2358 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2359 | |
f0c76d61 JV |
2360 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2361 | bond->params.mode == BOND_MODE_8023AD) | |
2362 | bond_set_slave_inactive_flags(slave); | |
2363 | ||
a4aee5c8 JP |
2364 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2365 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2366 | |
2367 | if (bond->params.mode == BOND_MODE_8023AD) | |
2368 | bond_3ad_handle_link_change(slave, | |
2369 | BOND_LINK_DOWN); | |
2370 | ||
ae63e808 | 2371 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2372 | bond_alb_handle_link_change(bond, slave, |
2373 | BOND_LINK_DOWN); | |
2374 | ||
2375 | if (slave == bond->curr_active_slave) | |
2376 | goto do_failover; | |
2377 | ||
2378 | continue; | |
2379 | ||
2380 | default: | |
a4aee5c8 | 2381 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2382 | bond->dev->name, slave->new_link, |
2383 | slave->dev->name); | |
2384 | slave->new_link = BOND_LINK_NOCHANGE; | |
2385 | ||
2386 | continue; | |
2387 | } | |
2388 | ||
2389 | do_failover: | |
2390 | ASSERT_RTNL(); | |
e843fa50 | 2391 | block_netpoll_tx(); |
f0c76d61 JV |
2392 | write_lock_bh(&bond->curr_slave_lock); |
2393 | bond_select_active_slave(bond); | |
2394 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2395 | unblock_netpoll_tx(); |
f0c76d61 JV |
2396 | } |
2397 | ||
2398 | bond_set_carrier(bond); | |
1da177e4 LT |
2399 | } |
2400 | ||
0b0eef66 JV |
2401 | /* |
2402 | * bond_mii_monitor | |
2403 | * | |
2404 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2405 | * inspection, then (if inspection indicates something needs to be done) |
2406 | * an acquisition of appropriate locks followed by a commit phase to | |
2407 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2408 | */ |
2409 | void bond_mii_monitor(struct work_struct *work) | |
2410 | { | |
2411 | struct bonding *bond = container_of(work, struct bonding, | |
2412 | mii_work.work); | |
ad246c99 | 2413 | bool should_notify_peers = false; |
e6d265e8 | 2414 | unsigned long delay; |
0b0eef66 JV |
2415 | |
2416 | read_lock(&bond->lock); | |
e6d265e8 JV |
2417 | |
2418 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 JV |
2419 | |
2420 | if (bond->slave_cnt == 0) | |
2421 | goto re_arm; | |
b59f9f74 | 2422 | |
ad246c99 BH |
2423 | should_notify_peers = bond_should_notify_peers(bond); |
2424 | ||
f0c76d61 | 2425 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2426 | read_unlock(&bond->lock); |
e6d265e8 JV |
2427 | |
2428 | /* Race avoidance with bond_close cancel of workqueue */ | |
2429 | if (!rtnl_trylock()) { | |
2430 | read_lock(&bond->lock); | |
2431 | delay = 1; | |
2432 | should_notify_peers = false; | |
2433 | goto re_arm; | |
2434 | } | |
2435 | ||
0b0eef66 | 2436 | read_lock(&bond->lock); |
f0c76d61 JV |
2437 | |
2438 | bond_miimon_commit(bond); | |
2439 | ||
5655662d JV |
2440 | read_unlock(&bond->lock); |
2441 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2442 | read_lock(&bond->lock); | |
0b0eef66 JV |
2443 | } |
2444 | ||
f0c76d61 | 2445 | re_arm: |
e6d265e8 JV |
2446 | if (bond->params.miimon) |
2447 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2448 | ||
0b0eef66 | 2449 | read_unlock(&bond->lock); |
ad246c99 BH |
2450 | |
2451 | if (should_notify_peers) { | |
e6d265e8 JV |
2452 | if (!rtnl_trylock()) { |
2453 | read_lock(&bond->lock); | |
2454 | bond->send_peer_notif++; | |
2455 | read_unlock(&bond->lock); | |
2456 | return; | |
2457 | } | |
b7bc2a5b | 2458 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2459 | rtnl_unlock(); |
2460 | } | |
0b0eef66 | 2461 | } |
c3ade5ca | 2462 | |
d3bb52b0 | 2463 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2464 | { |
0883beca | 2465 | struct vlan_entry *vlan; |
eaddcd76 | 2466 | struct net_device *vlan_dev; |
f5b2b966 | 2467 | |
eaddcd76 | 2468 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
f5b2b966 JV |
2469 | return 1; |
2470 | ||
0883beca | 2471 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
eaddcd76 | 2472 | rcu_read_lock(); |
1fd9b1fc PM |
2473 | vlan_dev = __vlan_find_dev_deep(bond->dev, htons(ETH_P_8021Q), |
2474 | vlan->vlan_id); | |
eaddcd76 AG |
2475 | rcu_read_unlock(); |
2476 | if (vlan_dev && ip == bond_confirm_addr(vlan_dev, 0, ip)) | |
f5b2b966 JV |
2477 | return 1; |
2478 | } | |
2479 | ||
2480 | return 0; | |
2481 | } | |
2482 | ||
c3ade5ca JV |
2483 | /* |
2484 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2485 | * switches in VLAN mode (especially if ports are configured as | |
2486 | * "native" to a VLAN) might not pass non-tagged frames. | |
2487 | */ | |
d3bb52b0 | 2488 | 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 |
2489 | { |
2490 | struct sk_buff *skb; | |
2491 | ||
acca2674 | 2492 | pr_debug("arp %d on slave %s: dst %pI4 src %pI4 vid %d\n", arp_op, |
2493 | slave_dev->name, &dest_ip, &src_ip, vlan_id); | |
3d632c3f | 2494 | |
c3ade5ca JV |
2495 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2496 | NULL, slave_dev->dev_addr, NULL); | |
2497 | ||
2498 | if (!skb) { | |
a4aee5c8 | 2499 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2500 | return; |
2501 | } | |
2502 | if (vlan_id) { | |
1fd9b1fc | 2503 | skb = vlan_put_tag(skb, htons(ETH_P_8021Q), vlan_id); |
c3ade5ca | 2504 | if (!skb) { |
a4aee5c8 | 2505 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2506 | return; |
2507 | } | |
2508 | } | |
2509 | arp_xmit(skb); | |
2510 | } | |
2511 | ||
2512 | ||
1da177e4 LT |
2513 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2514 | { | |
b23dd4fe | 2515 | int i, vlan_id; |
d3bb52b0 | 2516 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2517 | struct vlan_entry *vlan; |
eaddcd76 | 2518 | struct net_device *vlan_dev = NULL; |
c3ade5ca | 2519 | struct rtable *rt; |
1da177e4 | 2520 | |
6b780567 | 2521 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
eaddcd76 | 2522 | __be32 addr; |
6b780567 | 2523 | if (!targets[i]) |
5a31bec0 | 2524 | break; |
acca2674 | 2525 | pr_debug("basa: target %pI4\n", &targets[i]); |
cc0e4070 | 2526 | if (!bond_vlan_used(bond)) { |
5a03cdb7 | 2527 | pr_debug("basa: empty vlan: arp_send\n"); |
eaddcd76 | 2528 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2529 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2530 | addr, 0); |
c3ade5ca JV |
2531 | continue; |
2532 | } | |
2533 | ||
2534 | /* | |
2535 | * If VLANs are configured, we do a route lookup to | |
2536 | * determine which VLAN interface would be used, so we | |
2537 | * can tag the ARP with the proper VLAN tag. | |
2538 | */ | |
78fbfd8a DM |
2539 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2540 | RTO_ONLINK, 0); | |
b23dd4fe | 2541 | if (IS_ERR(rt)) { |
c3ade5ca | 2542 | if (net_ratelimit()) { |
a4aee5c8 | 2543 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2544 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2545 | } |
2546 | continue; | |
2547 | } | |
2548 | ||
2549 | /* | |
2550 | * This target is not on a VLAN | |
2551 | */ | |
d8d1f30b | 2552 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2553 | ip_rt_put(rt); |
5a03cdb7 | 2554 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
eaddcd76 | 2555 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2556 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2557 | addr, 0); |
c3ade5ca JV |
2558 | continue; |
2559 | } | |
2560 | ||
2561 | vlan_id = 0; | |
0883beca | 2562 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
cc0e4070 JP |
2563 | rcu_read_lock(); |
2564 | vlan_dev = __vlan_find_dev_deep(bond->dev, | |
1fd9b1fc | 2565 | htons(ETH_P_8021Q), |
cc0e4070 JP |
2566 | vlan->vlan_id); |
2567 | rcu_read_unlock(); | |
d8d1f30b | 2568 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2569 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2570 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2571 | vlan_dev->name, vlan_id); |
2572 | break; | |
2573 | } | |
2574 | } | |
2575 | ||
eaddcd76 | 2576 | if (vlan_id && vlan_dev) { |
ed4b9f80 | 2577 | ip_rt_put(rt); |
eaddcd76 | 2578 | addr = bond_confirm_addr(vlan_dev, targets[i], 0); |
c3ade5ca | 2579 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2580 | addr, vlan_id); |
c3ade5ca JV |
2581 | continue; |
2582 | } | |
2583 | ||
2584 | if (net_ratelimit()) { | |
a4aee5c8 | 2585 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2586 | bond->dev->name, &targets[i], |
d8d1f30b | 2587 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2588 | } |
ed4b9f80 | 2589 | ip_rt_put(rt); |
c3ade5ca JV |
2590 | } |
2591 | } | |
2592 | ||
d3bb52b0 | 2593 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 | 2594 | { |
8599b52e VF |
2595 | int i; |
2596 | ||
87a7b84b VF |
2597 | if (!sip || !bond_has_this_ip(bond, tip)) { |
2598 | pr_debug("bva: sip %pI4 tip %pI4 not found\n", &sip, &tip); | |
2599 | return; | |
2600 | } | |
f5b2b966 | 2601 | |
8599b52e VF |
2602 | i = bond_get_targets_ip(bond->params.arp_targets, sip); |
2603 | if (i == -1) { | |
87a7b84b VF |
2604 | pr_debug("bva: sip %pI4 not found in targets\n", &sip); |
2605 | return; | |
f5b2b966 | 2606 | } |
87a7b84b | 2607 | slave->last_arp_rx = jiffies; |
8599b52e | 2608 | slave->target_last_arp_rx[i] = jiffies; |
f5b2b966 JV |
2609 | } |
2610 | ||
de063b70 ED |
2611 | static int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2612 | struct slave *slave) | |
f5b2b966 | 2613 | { |
de063b70 | 2614 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2615 | unsigned char *arp_ptr; |
d3bb52b0 | 2616 | __be32 sip, tip; |
de063b70 | 2617 | int alen; |
f5b2b966 | 2618 | |
3aba891d | 2619 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2620 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2621 | |
f5b2b966 | 2622 | read_lock(&bond->lock); |
2c146102 VF |
2623 | |
2624 | if (!slave_do_arp_validate(bond, slave)) | |
2625 | goto out_unlock; | |
2626 | ||
de063b70 | 2627 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2628 | |
3aba891d JP |
2629 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2630 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2631 | |
de063b70 ED |
2632 | if (alen > skb_headlen(skb)) { |
2633 | arp = kmalloc(alen, GFP_ATOMIC); | |
2634 | if (!arp) | |
2635 | goto out_unlock; | |
2636 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2637 | goto out_unlock; | |
2638 | } | |
f5b2b966 | 2639 | |
3aba891d | 2640 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2641 | skb->pkt_type == PACKET_OTHERHOST || |
2642 | skb->pkt_type == PACKET_LOOPBACK || | |
2643 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2644 | arp->ar_pro != htons(ETH_P_IP) || | |
2645 | arp->ar_pln != 4) | |
2646 | goto out_unlock; | |
2647 | ||
2648 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2649 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2650 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2651 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2652 | memcpy(&tip, arp_ptr, 4); |
2653 | ||
5a03cdb7 | 2654 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2655 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2656 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2657 | &sip, &tip); | |
f5b2b966 JV |
2658 | |
2659 | /* | |
2660 | * Backup slaves won't see the ARP reply, but do come through | |
2661 | * here for each ARP probe (so we swap the sip/tip to validate | |
2662 | * the probe). In a "redundant switch, common router" type of | |
2663 | * configuration, the ARP probe will (hopefully) travel from | |
2664 | * the active, through one switch, the router, then the other | |
2665 | * switch before reaching the backup. | |
aeea64ac VF |
2666 | * |
2667 | * We 'trust' the arp requests if there is an active slave and | |
2668 | * it received valid arp reply(s) after it became active. This | |
2669 | * is done to avoid endless looping when we can't reach the | |
2670 | * arp_ip_target and fool ourselves with our own arp requests. | |
f5b2b966 | 2671 | */ |
e30bc066 | 2672 | if (bond_is_active_slave(slave)) |
f5b2b966 | 2673 | bond_validate_arp(bond, slave, sip, tip); |
aeea64ac VF |
2674 | else if (bond->curr_active_slave && |
2675 | time_after(slave_last_rx(bond, bond->curr_active_slave), | |
2676 | bond->curr_active_slave->jiffies)) | |
f5b2b966 JV |
2677 | bond_validate_arp(bond, slave, tip, sip); |
2678 | ||
2679 | out_unlock: | |
2680 | read_unlock(&bond->lock); | |
de063b70 ED |
2681 | if (arp != (struct arphdr *)skb->data) |
2682 | kfree(arp); | |
13a8e0c8 | 2683 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2684 | } |
2685 | ||
1da177e4 LT |
2686 | /* |
2687 | * this function is called regularly to monitor each slave's link | |
2688 | * ensuring that traffic is being sent and received when arp monitoring | |
2689 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2690 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2691 | * arp monitoring in active backup mode. | |
2692 | */ | |
1b76b316 | 2693 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2694 | { |
1b76b316 JV |
2695 | struct bonding *bond = container_of(work, struct bonding, |
2696 | arp_work.work); | |
1da177e4 LT |
2697 | struct slave *slave, *oldcurrent; |
2698 | int do_failover = 0; | |
da210f55 | 2699 | int delta_in_ticks, extra_ticks; |
1da177e4 LT |
2700 | int i; |
2701 | ||
2702 | read_lock(&bond->lock); | |
2703 | ||
5ce0da8f | 2704 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
da210f55 | 2705 | extra_ticks = delta_in_ticks / 2; |
1da177e4 | 2706 | |
3d632c3f | 2707 | if (bond->slave_cnt == 0) |
1da177e4 | 2708 | goto re_arm; |
1da177e4 LT |
2709 | |
2710 | read_lock(&bond->curr_slave_lock); | |
2711 | oldcurrent = bond->curr_active_slave; | |
2712 | read_unlock(&bond->curr_slave_lock); | |
2713 | ||
2714 | /* see if any of the previous devices are up now (i.e. they have | |
2715 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2716 | * the picture unless it is null. also, slave->jiffies is not needed | |
2717 | * here because we send an arp on each slave and give a slave as | |
2718 | * long as it needs to get the tx/rx within the delta. | |
2719 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2720 | * so it can wait | |
2721 | */ | |
2722 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2723 | unsigned long trans_start = dev_trans_start(slave->dev); |
2724 | ||
1da177e4 | 2725 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2726 | if (time_in_range(jiffies, |
2727 | trans_start - delta_in_ticks, | |
da210f55 | 2728 | trans_start + delta_in_ticks + extra_ticks) && |
cb32f2a0 JB |
2729 | time_in_range(jiffies, |
2730 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2731 | slave->dev->last_rx + delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2732 | |
2733 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2734 | bond_set_active_slave(slave); |
1da177e4 LT |
2735 | |
2736 | /* primary_slave has no meaning in round-robin | |
2737 | * mode. the window of a slave being up and | |
2738 | * curr_active_slave being null after enslaving | |
2739 | * is closed. | |
2740 | */ | |
2741 | if (!oldcurrent) { | |
a4aee5c8 JP |
2742 | pr_info("%s: link status definitely up for interface %s, ", |
2743 | bond->dev->name, | |
2744 | slave->dev->name); | |
1da177e4 LT |
2745 | do_failover = 1; |
2746 | } else { | |
a4aee5c8 JP |
2747 | pr_info("%s: interface %s is now up\n", |
2748 | bond->dev->name, | |
2749 | slave->dev->name); | |
1da177e4 LT |
2750 | } |
2751 | } | |
2752 | } else { | |
2753 | /* slave->link == BOND_LINK_UP */ | |
2754 | ||
2755 | /* not all switches will respond to an arp request | |
2756 | * when the source ip is 0, so don't take the link down | |
2757 | * if we don't know our ip yet | |
2758 | */ | |
cb32f2a0 JB |
2759 | if (!time_in_range(jiffies, |
2760 | trans_start - delta_in_ticks, | |
da210f55 | 2761 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2762 | !time_in_range(jiffies, |
2763 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2764 | slave->dev->last_rx + 2 * delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2765 | |
2766 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2767 | bond_set_backup_slave(slave); |
1da177e4 | 2768 | |
3d632c3f | 2769 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2770 | slave->link_failure_count++; |
1da177e4 | 2771 | |
a4aee5c8 JP |
2772 | pr_info("%s: interface %s is now down.\n", |
2773 | bond->dev->name, | |
2774 | slave->dev->name); | |
1da177e4 | 2775 | |
3d632c3f | 2776 | if (slave == oldcurrent) |
1da177e4 | 2777 | do_failover = 1; |
1da177e4 LT |
2778 | } |
2779 | } | |
2780 | ||
2781 | /* note: if switch is in round-robin mode, all links | |
2782 | * must tx arp to ensure all links rx an arp - otherwise | |
2783 | * links may oscillate or not come up at all; if switch is | |
2784 | * in something like xor mode, there is nothing we can | |
2785 | * do - all replies will be rx'ed on same link causing slaves | |
2786 | * to be unstable during low/no traffic periods | |
2787 | */ | |
3d632c3f | 2788 | if (IS_UP(slave->dev)) |
1da177e4 | 2789 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2790 | } |
2791 | ||
2792 | if (do_failover) { | |
e843fa50 | 2793 | block_netpoll_tx(); |
059fe7a5 | 2794 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2795 | |
2796 | bond_select_active_slave(bond); | |
2797 | ||
059fe7a5 | 2798 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2799 | unblock_netpoll_tx(); |
1da177e4 LT |
2800 | } |
2801 | ||
2802 | re_arm: | |
e6d265e8 | 2803 | if (bond->params.arp_interval) |
1b76b316 | 2804 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 2805 | |
1da177e4 LT |
2806 | read_unlock(&bond->lock); |
2807 | } | |
2808 | ||
2809 | /* | |
b2220cad JV |
2810 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2811 | * changes. Sets new_link in slaves to specify what action should take | |
2812 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2813 | * to link states must be committed. | |
2814 | * | |
2815 | * Called with bond->lock held for read. | |
1da177e4 | 2816 | */ |
b2220cad | 2817 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2818 | { |
1da177e4 | 2819 | struct slave *slave; |
b2220cad | 2820 | int i, commit = 0; |
cb32f2a0 | 2821 | unsigned long trans_start; |
da210f55 JB |
2822 | int extra_ticks; |
2823 | ||
2824 | /* All the time comparisons below need some extra time. Otherwise, on | |
2825 | * fast networks the ARP probe/reply may arrive within the same jiffy | |
2826 | * as it was sent. Then, the next time the ARP monitor is run, one | |
2827 | * arp_interval will already have passed in the comparisons. | |
2828 | */ | |
2829 | extra_ticks = delta_in_ticks / 2; | |
1da177e4 | 2830 | |
b2220cad JV |
2831 | bond_for_each_slave(bond, slave, i) { |
2832 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2833 | |
b2220cad | 2834 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2835 | if (time_in_range(jiffies, |
2836 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2837 | slave_last_rx(bond, slave) + delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2838 | |
b2220cad JV |
2839 | slave->new_link = BOND_LINK_UP; |
2840 | commit++; | |
2841 | } | |
1da177e4 | 2842 | |
b2220cad JV |
2843 | continue; |
2844 | } | |
1da177e4 | 2845 | |
b2220cad JV |
2846 | /* |
2847 | * Give slaves 2*delta after being enslaved or made | |
2848 | * active. This avoids bouncing, as the last receive | |
2849 | * times need a full ARP monitor cycle to be updated. | |
2850 | */ | |
cb32f2a0 JB |
2851 | if (time_in_range(jiffies, |
2852 | slave->jiffies - delta_in_ticks, | |
da210f55 | 2853 | slave->jiffies + 2 * delta_in_ticks + extra_ticks)) |
b2220cad JV |
2854 | continue; |
2855 | ||
2856 | /* | |
2857 | * Backup slave is down if: | |
2858 | * - No current_arp_slave AND | |
2859 | * - more than 3*delta since last receive AND | |
2860 | * - the bond has an IP address | |
2861 | * | |
2862 | * Note: a non-null current_arp_slave indicates | |
2863 | * the curr_active_slave went down and we are | |
2864 | * searching for a new one; under this condition | |
2865 | * we only take the curr_active_slave down - this | |
2866 | * gives each slave a chance to tx/rx traffic | |
2867 | * before being taken out | |
2868 | */ | |
e30bc066 | 2869 | if (!bond_is_active_slave(slave) && |
b2220cad | 2870 | !bond->current_arp_slave && |
cb32f2a0 JB |
2871 | !time_in_range(jiffies, |
2872 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2873 | slave_last_rx(bond, slave) + 3 * delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2874 | |
b2220cad JV |
2875 | slave->new_link = BOND_LINK_DOWN; |
2876 | commit++; | |
2877 | } | |
2878 | ||
2879 | /* | |
2880 | * Active slave is down if: | |
2881 | * - more than 2*delta since transmitting OR | |
2882 | * - (more than 2*delta since receive AND | |
2883 | * the bond has an IP address) | |
2884 | */ | |
cb32f2a0 | 2885 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2886 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
2887 | (!time_in_range(jiffies, |
2888 | trans_start - delta_in_ticks, | |
da210f55 | 2889 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2890 | !time_in_range(jiffies, |
2891 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2892 | slave_last_rx(bond, slave) + 2 * delta_in_ticks + extra_ticks))) { |
cb32f2a0 | 2893 | |
b2220cad JV |
2894 | slave->new_link = BOND_LINK_DOWN; |
2895 | commit++; | |
2896 | } | |
1da177e4 LT |
2897 | } |
2898 | ||
b2220cad JV |
2899 | return commit; |
2900 | } | |
1da177e4 | 2901 | |
b2220cad JV |
2902 | /* |
2903 | * Called to commit link state changes noted by inspection step of | |
2904 | * active-backup mode ARP monitor. | |
2905 | * | |
2906 | * Called with RTNL and bond->lock for read. | |
2907 | */ | |
2908 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2909 | { | |
2910 | struct slave *slave; | |
2911 | int i; | |
cb32f2a0 | 2912 | unsigned long trans_start; |
1da177e4 | 2913 | |
b2220cad JV |
2914 | bond_for_each_slave(bond, slave, i) { |
2915 | switch (slave->new_link) { | |
2916 | case BOND_LINK_NOCHANGE: | |
2917 | continue; | |
ff59c456 | 2918 | |
b2220cad | 2919 | case BOND_LINK_UP: |
cb32f2a0 | 2920 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 2921 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
2922 | time_in_range(jiffies, |
2923 | trans_start - delta_in_ticks, | |
da210f55 | 2924 | trans_start + delta_in_ticks + delta_in_ticks / 2)) || |
b9f60253 | 2925 | bond->curr_active_slave != slave) { |
b2220cad | 2926 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2927 | if (bond->current_arp_slave) { |
2928 | bond_set_slave_inactive_flags( | |
2929 | bond->current_arp_slave); | |
2930 | bond->current_arp_slave = NULL; | |
2931 | } | |
b2220cad | 2932 | |
a4aee5c8 | 2933 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2934 | bond->dev->name, slave->dev->name); |
b2220cad | 2935 | |
b9f60253 JP |
2936 | if (!bond->curr_active_slave || |
2937 | (slave == bond->primary_slave)) | |
2938 | goto do_failover; | |
1da177e4 | 2939 | |
b9f60253 | 2940 | } |
1da177e4 | 2941 | |
b9f60253 | 2942 | continue; |
1da177e4 | 2943 | |
b2220cad JV |
2944 | case BOND_LINK_DOWN: |
2945 | if (slave->link_failure_count < UINT_MAX) | |
2946 | slave->link_failure_count++; | |
2947 | ||
2948 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2949 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2950 | |
a4aee5c8 | 2951 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2952 | bond->dev->name, slave->dev->name); |
b2220cad | 2953 | |
b9f60253 | 2954 | if (slave == bond->curr_active_slave) { |
b2220cad | 2955 | bond->current_arp_slave = NULL; |
b9f60253 | 2956 | goto do_failover; |
1da177e4 | 2957 | } |
b9f60253 JP |
2958 | |
2959 | continue; | |
b2220cad JV |
2960 | |
2961 | default: | |
a4aee5c8 | 2962 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2963 | bond->dev->name, slave->new_link, |
2964 | slave->dev->name); | |
b9f60253 | 2965 | continue; |
1da177e4 | 2966 | } |
1da177e4 | 2967 | |
b9f60253 JP |
2968 | do_failover: |
2969 | ASSERT_RTNL(); | |
e843fa50 | 2970 | block_netpoll_tx(); |
b2220cad | 2971 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2972 | bond_select_active_slave(bond); |
b2220cad | 2973 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2974 | unblock_netpoll_tx(); |
b2220cad | 2975 | } |
1da177e4 | 2976 | |
b2220cad JV |
2977 | bond_set_carrier(bond); |
2978 | } | |
1da177e4 | 2979 | |
b2220cad JV |
2980 | /* |
2981 | * Send ARP probes for active-backup mode ARP monitor. | |
2982 | * | |
2983 | * Called with bond->lock held for read. | |
2984 | */ | |
2985 | static void bond_ab_arp_probe(struct bonding *bond) | |
2986 | { | |
2987 | struct slave *slave; | |
2988 | int i; | |
1da177e4 | 2989 | |
b2220cad | 2990 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2991 | |
b2220cad | 2992 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
2993 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
2994 | bond->current_arp_slave->dev->name, | |
2995 | bond->curr_active_slave->dev->name); | |
1da177e4 | 2996 | |
b2220cad JV |
2997 | if (bond->curr_active_slave) { |
2998 | bond_arp_send_all(bond, bond->curr_active_slave); | |
2999 | read_unlock(&bond->curr_slave_lock); | |
3000 | return; | |
3001 | } | |
1da177e4 | 3002 | |
b2220cad | 3003 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3004 | |
b2220cad JV |
3005 | /* if we don't have a curr_active_slave, search for the next available |
3006 | * backup slave from the current_arp_slave and make it the candidate | |
3007 | * for becoming the curr_active_slave | |
3008 | */ | |
1da177e4 | 3009 | |
b2220cad JV |
3010 | if (!bond->current_arp_slave) { |
3011 | bond->current_arp_slave = bond->first_slave; | |
3012 | if (!bond->current_arp_slave) | |
3013 | return; | |
3014 | } | |
1da177e4 | 3015 | |
b2220cad | 3016 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3017 | |
b2220cad JV |
3018 | /* search for next candidate */ |
3019 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3020 | if (IS_UP(slave->dev)) { | |
3021 | slave->link = BOND_LINK_BACK; | |
3022 | bond_set_slave_active_flags(slave); | |
3023 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3024 | slave->jiffies = jiffies; |
b2220cad JV |
3025 | bond->current_arp_slave = slave; |
3026 | break; | |
1da177e4 LT |
3027 | } |
3028 | ||
b2220cad JV |
3029 | /* if the link state is up at this point, we |
3030 | * mark it down - this can happen if we have | |
3031 | * simultaneous link failures and | |
3032 | * reselect_active_interface doesn't make this | |
3033 | * one the current slave so it is still marked | |
3034 | * up when it is actually down | |
1da177e4 | 3035 | */ |
b2220cad JV |
3036 | if (slave->link == BOND_LINK_UP) { |
3037 | slave->link = BOND_LINK_DOWN; | |
3038 | if (slave->link_failure_count < UINT_MAX) | |
3039 | slave->link_failure_count++; | |
1da177e4 | 3040 | |
b2220cad JV |
3041 | bond_set_slave_inactive_flags(slave); |
3042 | ||
a4aee5c8 JP |
3043 | pr_info("%s: backup interface %s is now down.\n", |
3044 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3045 | } |
b2220cad JV |
3046 | } |
3047 | } | |
1da177e4 | 3048 | |
b2220cad JV |
3049 | void bond_activebackup_arp_mon(struct work_struct *work) |
3050 | { | |
3051 | struct bonding *bond = container_of(work, struct bonding, | |
3052 | arp_work.work); | |
ad246c99 | 3053 | bool should_notify_peers = false; |
b2220cad | 3054 | int delta_in_ticks; |
1da177e4 | 3055 | |
b2220cad | 3056 | read_lock(&bond->lock); |
1da177e4 | 3057 | |
b2220cad | 3058 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3059 | |
b2220cad JV |
3060 | if (bond->slave_cnt == 0) |
3061 | goto re_arm; | |
3062 | ||
ad246c99 BH |
3063 | should_notify_peers = bond_should_notify_peers(bond); |
3064 | ||
b2220cad JV |
3065 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3066 | read_unlock(&bond->lock); | |
e6d265e8 JV |
3067 | |
3068 | /* Race avoidance with bond_close flush of workqueue */ | |
3069 | if (!rtnl_trylock()) { | |
3070 | read_lock(&bond->lock); | |
3071 | delta_in_ticks = 1; | |
3072 | should_notify_peers = false; | |
3073 | goto re_arm; | |
3074 | } | |
3075 | ||
b2220cad JV |
3076 | read_lock(&bond->lock); |
3077 | ||
3078 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3079 | ||
3080 | read_unlock(&bond->lock); | |
3081 | rtnl_unlock(); | |
3082 | read_lock(&bond->lock); | |
1da177e4 LT |
3083 | } |
3084 | ||
b2220cad JV |
3085 | bond_ab_arp_probe(bond); |
3086 | ||
1da177e4 | 3087 | re_arm: |
e6d265e8 | 3088 | if (bond->params.arp_interval) |
1b76b316 | 3089 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 3090 | |
1da177e4 | 3091 | read_unlock(&bond->lock); |
ad246c99 BH |
3092 | |
3093 | if (should_notify_peers) { | |
e6d265e8 JV |
3094 | if (!rtnl_trylock()) { |
3095 | read_lock(&bond->lock); | |
3096 | bond->send_peer_notif++; | |
3097 | read_unlock(&bond->lock); | |
3098 | return; | |
3099 | } | |
b7bc2a5b | 3100 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
3101 | rtnl_unlock(); |
3102 | } | |
1da177e4 LT |
3103 | } |
3104 | ||
1da177e4 LT |
3105 | /*-------------------------- netdev event handling --------------------------*/ |
3106 | ||
3107 | /* | |
3108 | * Change device name | |
3109 | */ | |
3110 | static int bond_event_changename(struct bonding *bond) | |
3111 | { | |
1da177e4 LT |
3112 | bond_remove_proc_entry(bond); |
3113 | bond_create_proc_entry(bond); | |
7e083840 | 3114 | |
f073c7ca TI |
3115 | bond_debug_reregister(bond); |
3116 | ||
1da177e4 LT |
3117 | return NOTIFY_DONE; |
3118 | } | |
3119 | ||
3d632c3f SH |
3120 | static int bond_master_netdev_event(unsigned long event, |
3121 | struct net_device *bond_dev) | |
1da177e4 | 3122 | { |
454d7c9b | 3123 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3124 | |
3125 | switch (event) { | |
3126 | case NETDEV_CHANGENAME: | |
3127 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3128 | case NETDEV_UNREGISTER: |
3129 | bond_remove_proc_entry(event_bond); | |
3130 | break; | |
3131 | case NETDEV_REGISTER: | |
3132 | bond_create_proc_entry(event_bond); | |
3133 | break; | |
1da177e4 LT |
3134 | default: |
3135 | break; | |
3136 | } | |
3137 | ||
3138 | return NOTIFY_DONE; | |
3139 | } | |
3140 | ||
3d632c3f SH |
3141 | static int bond_slave_netdev_event(unsigned long event, |
3142 | struct net_device *slave_dev) | |
1da177e4 | 3143 | { |
471cb5a3 | 3144 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
6101391d | 3145 | struct bonding *bond; |
3146 | struct net_device *bond_dev; | |
471cb5a3 JP |
3147 | u32 old_speed; |
3148 | u8 old_duplex; | |
1da177e4 | 3149 | |
6101391d | 3150 | /* A netdev event can be generated while enslaving a device |
3151 | * before netdev_rx_handler_register is called in which case | |
3152 | * slave will be NULL | |
3153 | */ | |
3154 | if (!slave) | |
3155 | return NOTIFY_DONE; | |
3156 | bond_dev = slave->bond->dev; | |
3157 | bond = slave->bond; | |
3158 | ||
1da177e4 LT |
3159 | switch (event) { |
3160 | case NETDEV_UNREGISTER: | |
8d2ada77 | 3161 | if (bond_dev->type != ARPHRD_ETHER) |
471cb5a3 JP |
3162 | bond_release_and_destroy(bond_dev, slave_dev); |
3163 | else | |
3164 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 3165 | break; |
98f41f69 | 3166 | case NETDEV_UP: |
1da177e4 | 3167 | case NETDEV_CHANGE: |
471cb5a3 JP |
3168 | old_speed = slave->speed; |
3169 | old_duplex = slave->duplex; | |
17d04500 | 3170 | |
471cb5a3 | 3171 | bond_update_speed_duplex(slave); |
17d04500 | 3172 | |
471cb5a3 JP |
3173 | if (bond->params.mode == BOND_MODE_8023AD) { |
3174 | if (old_speed != slave->speed) | |
3175 | bond_3ad_adapter_speed_changed(slave); | |
3176 | if (old_duplex != slave->duplex) | |
3177 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 3178 | } |
1da177e4 LT |
3179 | break; |
3180 | case NETDEV_DOWN: | |
3181 | /* | |
3182 | * ... Or is it this? | |
3183 | */ | |
3184 | break; | |
3185 | case NETDEV_CHANGEMTU: | |
3186 | /* | |
3187 | * TODO: Should slaves be allowed to | |
3188 | * independently alter their MTU? For | |
3189 | * an active-backup bond, slaves need | |
3190 | * not be the same type of device, so | |
3191 | * MTUs may vary. For other modes, | |
3192 | * slaves arguably should have the | |
3193 | * same MTUs. To do this, we'd need to | |
3194 | * take over the slave's change_mtu | |
3195 | * function for the duration of their | |
3196 | * servitude. | |
3197 | */ | |
3198 | break; | |
3199 | case NETDEV_CHANGENAME: | |
3200 | /* | |
3201 | * TODO: handle changing the primary's name | |
3202 | */ | |
3203 | break; | |
8531c5ff AK |
3204 | case NETDEV_FEAT_CHANGE: |
3205 | bond_compute_features(bond); | |
3206 | break; | |
4aa5dee4 JP |
3207 | case NETDEV_RESEND_IGMP: |
3208 | /* Propagate to master device */ | |
3209 | call_netdevice_notifiers(event, slave->bond->dev); | |
3210 | break; | |
1da177e4 LT |
3211 | default: |
3212 | break; | |
3213 | } | |
3214 | ||
3215 | return NOTIFY_DONE; | |
3216 | } | |
3217 | ||
3218 | /* | |
3219 | * bond_netdev_event: handle netdev notifier chain events. | |
3220 | * | |
3221 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3222 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3223 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3224 | * dev_probe_lock). | |
3225 | */ | |
3d632c3f SH |
3226 | static int bond_netdev_event(struct notifier_block *this, |
3227 | unsigned long event, void *ptr) | |
1da177e4 | 3228 | { |
351638e7 | 3229 | struct net_device *event_dev = netdev_notifier_info_to_dev(ptr); |
1da177e4 | 3230 | |
5a03cdb7 | 3231 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3232 | event_dev ? event_dev->name : "None", |
3233 | event); | |
1da177e4 | 3234 | |
0b680e75 JV |
3235 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3236 | return NOTIFY_DONE; | |
3237 | ||
1da177e4 | 3238 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3239 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3240 | return bond_master_netdev_event(event, event_dev); |
3241 | } | |
3242 | ||
3243 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3244 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3245 | return bond_slave_netdev_event(event, event_dev); |
3246 | } | |
3247 | ||
3248 | return NOTIFY_DONE; | |
3249 | } | |
3250 | ||
3251 | static struct notifier_block bond_netdev_notifier = { | |
3252 | .notifier_call = bond_netdev_event, | |
3253 | }; | |
3254 | ||
169a3e66 JV |
3255 | /*---------------------------- Hashing Policies -----------------------------*/ |
3256 | ||
6b923cb7 JE |
3257 | /* |
3258 | * Hash for the output device based upon layer 2 data | |
3259 | */ | |
3260 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) | |
3261 | { | |
3262 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3263 | ||
3264 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
3265 | return (data->h_dest[5] ^ data->h_source[5]) % count; | |
3266 | ||
3267 | return 0; | |
3268 | } | |
3269 | ||
6f6652be JV |
3270 | /* |
3271 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
6b923cb7 | 3272 | * the packet is not IP, fall back on bond_xmit_hash_policy_l2() |
6f6652be | 3273 | */ |
a361c83c | 3274 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be | 3275 | { |
4394542c ED |
3276 | const struct ethhdr *data; |
3277 | const struct iphdr *iph; | |
3278 | const struct ipv6hdr *ipv6h; | |
6b923cb7 | 3279 | u32 v6hash; |
4394542c | 3280 | const __be32 *s, *d; |
6b923cb7 JE |
3281 | |
3282 | if (skb->protocol == htons(ETH_P_IP) && | |
4394542c | 3283 | pskb_network_may_pull(skb, sizeof(*iph))) { |
6b923cb7 | 3284 | iph = ip_hdr(skb); |
4394542c | 3285 | data = (struct ethhdr *)skb->data; |
6f6652be | 3286 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3287 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6b923cb7 | 3288 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3289 | pskb_network_may_pull(skb, sizeof(*ipv6h))) { |
6b923cb7 | 3290 | ipv6h = ipv6_hdr(skb); |
4394542c | 3291 | data = (struct ethhdr *)skb->data; |
6b923cb7 JE |
3292 | s = &ipv6h->saddr.s6_addr32[0]; |
3293 | d = &ipv6h->daddr.s6_addr32[0]; | |
3294 | v6hash = (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3295 | v6hash ^= (v6hash >> 24) ^ (v6hash >> 16) ^ (v6hash >> 8); | |
3296 | return (v6hash ^ data->h_dest[5] ^ data->h_source[5]) % count; | |
6f6652be JV |
3297 | } |
3298 | ||
6b923cb7 | 3299 | return bond_xmit_hash_policy_l2(skb, count); |
6f6652be JV |
3300 | } |
3301 | ||
169a3e66 | 3302 | /* |
59c51591 | 3303 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 | 3304 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
6b923cb7 | 3305 | * altogether not IP, fall back on bond_xmit_hash_policy_l2() |
169a3e66 | 3306 | */ |
a361c83c | 3307 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 | 3308 | { |
6b923cb7 | 3309 | u32 layer4_xor = 0; |
4394542c ED |
3310 | const struct iphdr *iph; |
3311 | const struct ipv6hdr *ipv6h; | |
3312 | const __be32 *s, *d; | |
3313 | const __be16 *l4 = NULL; | |
3314 | __be16 _l4[2]; | |
3315 | int noff = skb_network_offset(skb); | |
3316 | int poff; | |
169a3e66 | 3317 | |
6b923cb7 | 3318 | if (skb->protocol == htons(ETH_P_IP) && |
4394542c | 3319 | pskb_may_pull(skb, noff + sizeof(*iph))) { |
6b923cb7 | 3320 | iph = ip_hdr(skb); |
4394542c ED |
3321 | poff = proto_ports_offset(iph->protocol); |
3322 | ||
3323 | if (!ip_is_fragment(iph) && poff >= 0) { | |
3324 | l4 = skb_header_pointer(skb, noff + (iph->ihl << 2) + poff, | |
3325 | sizeof(_l4), &_l4); | |
3326 | if (l4) | |
3327 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
169a3e66 JV |
3328 | } |
3329 | return (layer4_xor ^ | |
3330 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
6b923cb7 | 3331 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
4394542c | 3332 | pskb_may_pull(skb, noff + sizeof(*ipv6h))) { |
6b923cb7 | 3333 | ipv6h = ipv6_hdr(skb); |
4394542c ED |
3334 | poff = proto_ports_offset(ipv6h->nexthdr); |
3335 | if (poff >= 0) { | |
3336 | l4 = skb_header_pointer(skb, noff + sizeof(*ipv6h) + poff, | |
3337 | sizeof(_l4), &_l4); | |
3338 | if (l4) | |
3339 | layer4_xor = ntohs(l4[0] ^ l4[1]); | |
6b923cb7 JE |
3340 | } |
3341 | s = &ipv6h->saddr.s6_addr32[0]; | |
3342 | d = &ipv6h->daddr.s6_addr32[0]; | |
3343 | layer4_xor ^= (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3344 | layer4_xor ^= (layer4_xor >> 24) ^ (layer4_xor >> 16) ^ | |
3345 | (layer4_xor >> 8); | |
3346 | return layer4_xor % count; | |
169a3e66 JV |
3347 | } |
3348 | ||
6b923cb7 | 3349 | return bond_xmit_hash_policy_l2(skb, count); |
169a3e66 JV |
3350 | } |
3351 | ||
1da177e4 LT |
3352 | /*-------------------------- Device entry points ----------------------------*/ |
3353 | ||
fbb0c41b | 3354 | static void bond_work_init_all(struct bonding *bond) |
3355 | { | |
3356 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3357 | bond_resend_igmp_join_requests_delayed); | |
3358 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3359 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3360 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3361 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3362 | else | |
3363 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3364 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3365 | } | |
3366 | ||
3367 | static void bond_work_cancel_all(struct bonding *bond) | |
3368 | { | |
3369 | cancel_delayed_work_sync(&bond->mii_work); | |
3370 | cancel_delayed_work_sync(&bond->arp_work); | |
3371 | cancel_delayed_work_sync(&bond->alb_work); | |
3372 | cancel_delayed_work_sync(&bond->ad_work); | |
3373 | cancel_delayed_work_sync(&bond->mcast_work); | |
3374 | } | |
3375 | ||
1da177e4 LT |
3376 | static int bond_open(struct net_device *bond_dev) |
3377 | { | |
454d7c9b | 3378 | struct bonding *bond = netdev_priv(bond_dev); |
ba3211cc PP |
3379 | struct slave *slave; |
3380 | int i; | |
1da177e4 | 3381 | |
ba3211cc PP |
3382 | /* reset slave->backup and slave->inactive */ |
3383 | read_lock(&bond->lock); | |
3384 | if (bond->slave_cnt > 0) { | |
3385 | read_lock(&bond->curr_slave_lock); | |
3386 | bond_for_each_slave(bond, slave, i) { | |
3387 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3388 | && (slave != bond->curr_active_slave)) { | |
3389 | bond_set_slave_inactive_flags(slave); | |
3390 | } else { | |
3391 | bond_set_slave_active_flags(slave); | |
3392 | } | |
3393 | } | |
3394 | read_unlock(&bond->curr_slave_lock); | |
3395 | } | |
3396 | read_unlock(&bond->lock); | |
3397 | ||
fbb0c41b | 3398 | bond_work_init_all(bond); |
5a37e8ca | 3399 | |
58402054 | 3400 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3401 | /* bond_alb_initialize must be called before the timer |
3402 | * is started. | |
3403 | */ | |
fbb0c41b | 3404 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3405 | return -ENOMEM; |
1b76b316 | 3406 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3407 | } |
3408 | ||
fbb0c41b | 3409 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3410 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3411 | |
3412 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3413 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
f5b2b966 | 3414 | if (bond->params.arp_validate) |
3aba891d | 3415 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3416 | } |
3417 | ||
3418 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3419 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3420 | /* register to receive LACPDUs */ |
3aba891d | 3421 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3422 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3423 | } |
3424 | ||
3425 | return 0; | |
3426 | } | |
3427 | ||
3428 | static int bond_close(struct net_device *bond_dev) | |
3429 | { | |
454d7c9b | 3430 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3431 | |
1da177e4 | 3432 | write_lock_bh(&bond->lock); |
ad246c99 | 3433 | bond->send_peer_notif = 0; |
1da177e4 LT |
3434 | write_unlock_bh(&bond->lock); |
3435 | ||
fbb0c41b | 3436 | bond_work_cancel_all(bond); |
58402054 | 3437 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3438 | /* Must be called only after all |
3439 | * slaves have been released | |
3440 | */ | |
3441 | bond_alb_deinitialize(bond); | |
3442 | } | |
3aba891d | 3443 | bond->recv_probe = NULL; |
1da177e4 LT |
3444 | |
3445 | return 0; | |
3446 | } | |
3447 | ||
28172739 ED |
3448 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3449 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3450 | { |
454d7c9b | 3451 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3452 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3453 | struct slave *slave; |
3454 | int i; | |
3455 | ||
28172739 | 3456 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3457 | |
3458 | read_lock_bh(&bond->lock); | |
3459 | ||
3460 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3461 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3462 | dev_get_stats(slave->dev, &temp); |
3463 | ||
3464 | stats->rx_packets += sstats->rx_packets; | |
3465 | stats->rx_bytes += sstats->rx_bytes; | |
3466 | stats->rx_errors += sstats->rx_errors; | |
3467 | stats->rx_dropped += sstats->rx_dropped; | |
3468 | ||
3469 | stats->tx_packets += sstats->tx_packets; | |
3470 | stats->tx_bytes += sstats->tx_bytes; | |
3471 | stats->tx_errors += sstats->tx_errors; | |
3472 | stats->tx_dropped += sstats->tx_dropped; | |
3473 | ||
3474 | stats->multicast += sstats->multicast; | |
3475 | stats->collisions += sstats->collisions; | |
3476 | ||
3477 | stats->rx_length_errors += sstats->rx_length_errors; | |
3478 | stats->rx_over_errors += sstats->rx_over_errors; | |
3479 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3480 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3481 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3482 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3483 | ||
3484 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3485 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3486 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3487 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3488 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3489 | } |
3490 | ||
1da177e4 LT |
3491 | read_unlock_bh(&bond->lock); |
3492 | ||
3493 | return stats; | |
3494 | } | |
3495 | ||
3496 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3497 | { | |
3498 | struct net_device *slave_dev = NULL; | |
3499 | struct ifbond k_binfo; | |
3500 | struct ifbond __user *u_binfo = NULL; | |
3501 | struct ifslave k_sinfo; | |
3502 | struct ifslave __user *u_sinfo = NULL; | |
3503 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3504 | struct net *net; |
1da177e4 LT |
3505 | int res = 0; |
3506 | ||
a4aee5c8 | 3507 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3508 | |
3509 | switch (cmd) { | |
1da177e4 LT |
3510 | case SIOCGMIIPHY: |
3511 | mii = if_mii(ifr); | |
3d632c3f | 3512 | if (!mii) |
1da177e4 | 3513 | return -EINVAL; |
3d632c3f | 3514 | |
1da177e4 LT |
3515 | mii->phy_id = 0; |
3516 | /* Fall Through */ | |
3517 | case SIOCGMIIREG: | |
3518 | /* | |
3519 | * We do this again just in case we were called by SIOCGMIIREG | |
3520 | * instead of SIOCGMIIPHY. | |
3521 | */ | |
3522 | mii = if_mii(ifr); | |
3d632c3f | 3523 | if (!mii) |
1da177e4 | 3524 | return -EINVAL; |
3d632c3f | 3525 | |
1da177e4 LT |
3526 | |
3527 | if (mii->reg_num == 1) { | |
454d7c9b | 3528 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3529 | mii->val_out = 0; |
6603a6f2 | 3530 | read_lock(&bond->lock); |
1da177e4 | 3531 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3532 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3533 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3534 | |
1da177e4 | 3535 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3536 | read_unlock(&bond->lock); |
1da177e4 LT |
3537 | } |
3538 | ||
3539 | return 0; | |
3540 | case BOND_INFO_QUERY_OLD: | |
3541 | case SIOCBONDINFOQUERY: | |
3542 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3543 | ||
3d632c3f | 3544 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3545 | return -EFAULT; |
1da177e4 LT |
3546 | |
3547 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3548 | if (res == 0 && |
3549 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3550 | return -EFAULT; | |
1da177e4 LT |
3551 | |
3552 | return res; | |
3553 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3554 | case SIOCBONDSLAVEINFOQUERY: | |
3555 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3556 | ||
3d632c3f | 3557 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3558 | return -EFAULT; |
1da177e4 LT |
3559 | |
3560 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3561 | if (res == 0 && |
3562 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3563 | return -EFAULT; | |
1da177e4 LT |
3564 | |
3565 | return res; | |
3566 | default: | |
3567 | /* Go on */ | |
3568 | break; | |
3569 | } | |
3570 | ||
387ff911 G |
3571 | net = dev_net(bond_dev); |
3572 | ||
3573 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3574 | return -EPERM; |
1da177e4 | 3575 | |
387ff911 | 3576 | slave_dev = dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3577 | |
a4aee5c8 | 3578 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3579 | |
3d632c3f | 3580 | if (!slave_dev) |
1da177e4 | 3581 | res = -ENODEV; |
3d632c3f | 3582 | else { |
a4aee5c8 | 3583 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3584 | switch (cmd) { |
3585 | case BOND_ENSLAVE_OLD: | |
3586 | case SIOCBONDENSLAVE: | |
3587 | res = bond_enslave(bond_dev, slave_dev); | |
3588 | break; | |
3589 | case BOND_RELEASE_OLD: | |
3590 | case SIOCBONDRELEASE: | |
3591 | res = bond_release(bond_dev, slave_dev); | |
3592 | break; | |
3593 | case BOND_SETHWADDR_OLD: | |
3594 | case SIOCBONDSETHWADDR: | |
409cc1f8 JP |
3595 | bond_set_dev_addr(bond_dev, slave_dev); |
3596 | res = 0; | |
1da177e4 LT |
3597 | break; |
3598 | case BOND_CHANGE_ACTIVE_OLD: | |
3599 | case SIOCBONDCHANGEACTIVE: | |
3600 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3601 | break; | |
3602 | default: | |
3603 | res = -EOPNOTSUPP; | |
3604 | } | |
3605 | ||
3606 | dev_put(slave_dev); | |
3607 | } | |
3608 | ||
1da177e4 LT |
3609 | return res; |
3610 | } | |
3611 | ||
d03462b9 | 3612 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3613 | { |
454d7c9b | 3614 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3615 | |
d03462b9 JP |
3616 | if (change & IFF_PROMISC) |
3617 | bond_set_promiscuity(bond, | |
3618 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3619 | |
d03462b9 JP |
3620 | if (change & IFF_ALLMULTI) |
3621 | bond_set_allmulti(bond, | |
3622 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3623 | } | |
1da177e4 | 3624 | |
303d1cbf | 3625 | static void bond_set_rx_mode(struct net_device *bond_dev) |
d03462b9 JP |
3626 | { |
3627 | struct bonding *bond = netdev_priv(bond_dev); | |
303d1cbf JV |
3628 | struct slave *slave; |
3629 | int i; | |
1da177e4 | 3630 | |
80ee5ad2 JV |
3631 | read_lock(&bond->lock); |
3632 | ||
303d1cbf JV |
3633 | if (USES_PRIMARY(bond->params.mode)) { |
3634 | read_lock(&bond->curr_slave_lock); | |
3635 | slave = bond->curr_active_slave; | |
3636 | if (slave) { | |
3637 | dev_uc_sync(slave->dev, bond_dev); | |
3638 | dev_mc_sync(slave->dev, bond_dev); | |
3639 | } | |
3640 | read_unlock(&bond->curr_slave_lock); | |
3641 | } else { | |
3642 | bond_for_each_slave(bond, slave, i) { | |
3643 | dev_uc_sync_multiple(slave->dev, bond_dev); | |
3644 | dev_mc_sync_multiple(slave->dev, bond_dev); | |
3645 | } | |
1da177e4 LT |
3646 | } |
3647 | ||
80ee5ad2 | 3648 | read_unlock(&bond->lock); |
1da177e4 LT |
3649 | } |
3650 | ||
234bcf8a | 3651 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3652 | { |
234bcf8a | 3653 | struct bonding *bond = netdev_priv(n->dev); |
00829823 | 3654 | struct slave *slave = bond->first_slave; |
234bcf8a SP |
3655 | const struct net_device_ops *slave_ops; |
3656 | struct neigh_parms parms; | |
3657 | int ret; | |
3658 | ||
3659 | if (!slave) | |
3660 | return 0; | |
3661 | ||
3662 | slave_ops = slave->dev->netdev_ops; | |
3663 | ||
3664 | if (!slave_ops->ndo_neigh_setup) | |
3665 | return 0; | |
3666 | ||
3667 | parms.neigh_setup = NULL; | |
3668 | parms.neigh_cleanup = NULL; | |
3669 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3670 | if (ret) | |
3671 | return ret; | |
3672 | ||
3673 | /* | |
3674 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3675 | * after the last slave has been detached. Assumes that all slaves | |
3676 | * utilize the same neigh_cleanup (true at this writing as only user | |
3677 | * is ipoib). | |
3678 | */ | |
3679 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3680 | ||
3681 | if (!parms.neigh_setup) | |
3682 | return 0; | |
3683 | ||
3684 | return parms.neigh_setup(n); | |
3685 | } | |
3686 | ||
3687 | /* | |
3688 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3689 | * slave exists. So we must declare proxy setup function which will | |
3690 | * be used at run time to resolve the actual slave neigh param setup. | |
3691 | */ | |
3692 | static int bond_neigh_setup(struct net_device *dev, | |
3693 | struct neigh_parms *parms) | |
3694 | { | |
3695 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3696 | |
00829823 SH |
3697 | return 0; |
3698 | } | |
3699 | ||
1da177e4 LT |
3700 | /* |
3701 | * Change the MTU of all of a master's slaves to match the master | |
3702 | */ | |
3703 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3704 | { | |
454d7c9b | 3705 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3706 | struct slave *slave, *stop_at; |
3707 | int res = 0; | |
3708 | int i; | |
3709 | ||
5a03cdb7 | 3710 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3711 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3712 | |
3713 | /* Can't hold bond->lock with bh disabled here since | |
3714 | * some base drivers panic. On the other hand we can't | |
3715 | * hold bond->lock without bh disabled because we'll | |
3716 | * deadlock. The only solution is to rely on the fact | |
3717 | * that we're under rtnl_lock here, and the slaves | |
3718 | * list won't change. This doesn't solve the problem | |
3719 | * of setting the slave's MTU while it is | |
3720 | * transmitting, but the assumption is that the base | |
3721 | * driver can handle that. | |
3722 | * | |
3723 | * TODO: figure out a way to safely iterate the slaves | |
3724 | * list, but without holding a lock around the actual | |
3725 | * call to the base driver. | |
3726 | */ | |
3727 | ||
3728 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3729 | pr_debug("s %p s->p %p c_m %p\n", |
3730 | slave, | |
3731 | slave->prev, | |
3732 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3733 | |
1da177e4 LT |
3734 | res = dev_set_mtu(slave->dev, new_mtu); |
3735 | ||
3736 | if (res) { | |
3737 | /* If we failed to set the slave's mtu to the new value | |
3738 | * we must abort the operation even in ACTIVE_BACKUP | |
3739 | * mode, because if we allow the backup slaves to have | |
3740 | * different mtu values than the active slave we'll | |
3741 | * need to change their mtu when doing a failover. That | |
3742 | * means changing their mtu from timer context, which | |
3743 | * is probably not a good idea. | |
3744 | */ | |
5a03cdb7 | 3745 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3746 | goto unwind; |
3747 | } | |
3748 | } | |
3749 | ||
3750 | bond_dev->mtu = new_mtu; | |
3751 | ||
3752 | return 0; | |
3753 | ||
3754 | unwind: | |
3755 | /* unwind from head to the slave that failed */ | |
3756 | stop_at = slave; | |
3757 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3758 | int tmp_res; | |
3759 | ||
3760 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3761 | if (tmp_res) { | |
a4aee5c8 JP |
3762 | pr_debug("unwind err %d dev %s\n", |
3763 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3764 | } |
3765 | } | |
3766 | ||
3767 | return res; | |
3768 | } | |
3769 | ||
3770 | /* | |
3771 | * Change HW address | |
3772 | * | |
3773 | * Note that many devices must be down to change the HW address, and | |
3774 | * downing the master releases all slaves. We can make bonds full of | |
3775 | * bonding devices to test this, however. | |
3776 | */ | |
3777 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3778 | { | |
454d7c9b | 3779 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3780 | struct sockaddr *sa = addr, tmp_sa; |
3781 | struct slave *slave, *stop_at; | |
3782 | int res = 0; | |
3783 | int i; | |
3784 | ||
eb7cc59a SH |
3785 | if (bond->params.mode == BOND_MODE_ALB) |
3786 | return bond_alb_set_mac_address(bond_dev, addr); | |
3787 | ||
3788 | ||
a4aee5c8 JP |
3789 | pr_debug("bond=%p, name=%s\n", |
3790 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3791 | |
1b5acd29 JV |
3792 | /* If fail_over_mac is enabled, do nothing and return success. |
3793 | * Returning an error causes ifenslave to fail. | |
dd957c57 | 3794 | */ |
1b5acd29 | 3795 | if (bond->params.fail_over_mac) |
dd957c57 | 3796 | return 0; |
2ab82852 | 3797 | |
3d632c3f | 3798 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3799 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3800 | |
3801 | /* Can't hold bond->lock with bh disabled here since | |
3802 | * some base drivers panic. On the other hand we can't | |
3803 | * hold bond->lock without bh disabled because we'll | |
3804 | * deadlock. The only solution is to rely on the fact | |
3805 | * that we're under rtnl_lock here, and the slaves | |
3806 | * list won't change. This doesn't solve the problem | |
3807 | * of setting the slave's hw address while it is | |
3808 | * transmitting, but the assumption is that the base | |
3809 | * driver can handle that. | |
3810 | * | |
3811 | * TODO: figure out a way to safely iterate the slaves | |
3812 | * list, but without holding a lock around the actual | |
3813 | * call to the base driver. | |
3814 | */ | |
3815 | ||
3816 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 3817 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3818 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3819 | |
eb7cc59a | 3820 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3821 | res = -EOPNOTSUPP; |
5a03cdb7 | 3822 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3823 | goto unwind; |
3824 | } | |
3825 | ||
3826 | res = dev_set_mac_address(slave->dev, addr); | |
3827 | if (res) { | |
3828 | /* TODO: consider downing the slave | |
3829 | * and retry ? | |
3830 | * User should expect communications | |
3831 | * breakage anyway until ARP finish | |
3832 | * updating, so... | |
3833 | */ | |
5a03cdb7 | 3834 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3835 | goto unwind; |
3836 | } | |
3837 | } | |
3838 | ||
3839 | /* success */ | |
3840 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3841 | return 0; | |
3842 | ||
3843 | unwind: | |
3844 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3845 | tmp_sa.sa_family = bond_dev->type; | |
3846 | ||
3847 | /* unwind from head to the slave that failed */ | |
3848 | stop_at = slave; | |
3849 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3850 | int tmp_res; | |
3851 | ||
3852 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
3853 | if (tmp_res) { | |
a4aee5c8 JP |
3854 | pr_debug("unwind err %d dev %s\n", |
3855 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3856 | } |
3857 | } | |
3858 | ||
3859 | return res; | |
3860 | } | |
3861 | ||
3862 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
3863 | { | |
454d7c9b | 3864 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3865 | struct slave *slave, *start_at; |
cf5f9044 | 3866 | int i, slave_no, res = 1; |
a2fd940f | 3867 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 3868 | |
cf5f9044 | 3869 | /* |
a2fd940f AG |
3870 | * Start with the curr_active_slave that joined the bond as the |
3871 | * default for sending IGMP traffic. For failover purposes one | |
3872 | * needs to maintain some consistency for the interface that will | |
3873 | * send the join/membership reports. The curr_active_slave found | |
3874 | * will send all of this type of traffic. | |
cf5f9044 | 3875 | */ |
00ae7028 | 3876 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 3877 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 3878 | |
a2fd940f AG |
3879 | read_lock(&bond->curr_slave_lock); |
3880 | slave = bond->curr_active_slave; | |
3881 | read_unlock(&bond->curr_slave_lock); | |
3882 | ||
3883 | if (!slave) | |
3884 | goto out; | |
3885 | } else { | |
3886 | /* | |
3887 | * Concurrent TX may collide on rr_tx_counter; we accept | |
3888 | * that as being rare enough not to justify using an | |
3889 | * atomic op here. | |
3890 | */ | |
3891 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
3892 | ||
3893 | bond_for_each_slave(bond, slave, i) { | |
3894 | slave_no--; | |
3895 | if (slave_no < 0) | |
3896 | break; | |
3897 | } | |
1da177e4 LT |
3898 | } |
3899 | ||
cf5f9044 | 3900 | start_at = slave; |
1da177e4 LT |
3901 | bond_for_each_slave_from(bond, slave, i, start_at) { |
3902 | if (IS_UP(slave->dev) && | |
3903 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3904 | bond_is_active_slave(slave)) { |
1da177e4 | 3905 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
3906 | break; |
3907 | } | |
3908 | } | |
3909 | ||
1da177e4 LT |
3910 | out: |
3911 | if (res) { | |
3912 | /* no suitable interface, frame not sent */ | |
04502430 | 3913 | kfree_skb(skb); |
1da177e4 | 3914 | } |
0693e88e | 3915 | |
ec634fe3 | 3916 | return NETDEV_TX_OK; |
1da177e4 LT |
3917 | } |
3918 | ||
075897ce | 3919 | |
1da177e4 LT |
3920 | /* |
3921 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3922 | * the bond has a usable interface. | |
3923 | */ | |
3924 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3925 | { | |
454d7c9b | 3926 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3927 | int res = 1; |
3928 | ||
1da177e4 LT |
3929 | read_lock(&bond->curr_slave_lock); |
3930 | ||
0693e88e MM |
3931 | if (bond->curr_active_slave) |
3932 | res = bond_dev_queue_xmit(bond, skb, | |
3933 | bond->curr_active_slave->dev); | |
1da177e4 | 3934 | |
04502430 ED |
3935 | read_unlock(&bond->curr_slave_lock); |
3936 | ||
3d632c3f | 3937 | if (res) |
1da177e4 | 3938 | /* no suitable interface, frame not sent */ |
04502430 | 3939 | kfree_skb(skb); |
0693e88e | 3940 | |
ec634fe3 | 3941 | return NETDEV_TX_OK; |
1da177e4 LT |
3942 | } |
3943 | ||
3944 | /* | |
169a3e66 JV |
3945 | * In bond_xmit_xor() , we determine the output device by using a pre- |
3946 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
3947 | * find the next active slave. | |
1da177e4 LT |
3948 | */ |
3949 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3950 | { | |
454d7c9b | 3951 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3952 | struct slave *slave, *start_at; |
3953 | int slave_no; | |
3954 | int i; | |
3955 | int res = 1; | |
3956 | ||
a361c83c | 3957 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
3958 | |
3959 | bond_for_each_slave(bond, slave, i) { | |
3960 | slave_no--; | |
3d632c3f | 3961 | if (slave_no < 0) |
1da177e4 | 3962 | break; |
1da177e4 LT |
3963 | } |
3964 | ||
3965 | start_at = slave; | |
3966 | ||
3967 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
3968 | if (IS_UP(slave->dev) && | |
3969 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3970 | bond_is_active_slave(slave)) { |
1da177e4 LT |
3971 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
3972 | break; | |
3973 | } | |
3974 | } | |
3975 | ||
1da177e4 LT |
3976 | if (res) { |
3977 | /* no suitable interface, frame not sent */ | |
04502430 | 3978 | kfree_skb(skb); |
1da177e4 | 3979 | } |
0693e88e | 3980 | |
ec634fe3 | 3981 | return NETDEV_TX_OK; |
1da177e4 LT |
3982 | } |
3983 | ||
3984 | /* | |
3985 | * in broadcast mode, we send everything to all usable interfaces. | |
3986 | */ | |
3987 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
3988 | { | |
454d7c9b | 3989 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3990 | struct slave *slave, *start_at; |
3991 | struct net_device *tx_dev = NULL; | |
3992 | int i; | |
3993 | int res = 1; | |
3994 | ||
1da177e4 LT |
3995 | read_lock(&bond->curr_slave_lock); |
3996 | start_at = bond->curr_active_slave; | |
3997 | read_unlock(&bond->curr_slave_lock); | |
3998 | ||
3d632c3f | 3999 | if (!start_at) |
1da177e4 | 4000 | goto out; |
1da177e4 LT |
4001 | |
4002 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4003 | if (IS_UP(slave->dev) && | |
4004 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4005 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4006 | if (tx_dev) { |
4007 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4008 | if (!skb2) { | |
a4aee5c8 | 4009 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4010 | bond_dev->name); |
1da177e4 LT |
4011 | continue; |
4012 | } | |
4013 | ||
4014 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4015 | if (res) { | |
04502430 | 4016 | kfree_skb(skb2); |
1da177e4 LT |
4017 | continue; |
4018 | } | |
4019 | } | |
4020 | tx_dev = slave->dev; | |
4021 | } | |
4022 | } | |
4023 | ||
3d632c3f | 4024 | if (tx_dev) |
1da177e4 | 4025 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4026 | |
4027 | out: | |
3d632c3f | 4028 | if (res) |
1da177e4 | 4029 | /* no suitable interface, frame not sent */ |
04502430 | 4030 | kfree_skb(skb); |
3d632c3f | 4031 | |
1da177e4 | 4032 | /* frame sent to all suitable interfaces */ |
ec634fe3 | 4033 | return NETDEV_TX_OK; |
1da177e4 LT |
4034 | } |
4035 | ||
4036 | /*------------------------- Device initialization ---------------------------*/ | |
4037 | ||
6f6652be JV |
4038 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4039 | { | |
4040 | switch (bond->params.xmit_policy) { | |
4041 | case BOND_XMIT_POLICY_LAYER23: | |
4042 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4043 | break; | |
4044 | case BOND_XMIT_POLICY_LAYER34: | |
4045 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4046 | break; | |
4047 | case BOND_XMIT_POLICY_LAYER2: | |
4048 | default: | |
4049 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4050 | break; | |
4051 | } | |
4052 | } | |
4053 | ||
bb1d9123 AG |
4054 | /* |
4055 | * Lookup the slave that corresponds to a qid | |
4056 | */ | |
4057 | static inline int bond_slave_override(struct bonding *bond, | |
4058 | struct sk_buff *skb) | |
4059 | { | |
4060 | int i, res = 1; | |
4061 | struct slave *slave = NULL; | |
4062 | struct slave *check_slave; | |
4063 | ||
0693e88e MM |
4064 | if (!skb->queue_mapping) |
4065 | return 1; | |
bb1d9123 AG |
4066 | |
4067 | /* Find out if any slaves have the same mapping as this skb. */ | |
4068 | bond_for_each_slave(bond, check_slave, i) { | |
4069 | if (check_slave->queue_id == skb->queue_mapping) { | |
4070 | slave = check_slave; | |
4071 | break; | |
4072 | } | |
4073 | } | |
4074 | ||
4075 | /* If the slave isn't UP, use default transmit policy. */ | |
4076 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4077 | (slave->link == BOND_LINK_UP)) { | |
4078 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4079 | } | |
4080 | ||
bb1d9123 AG |
4081 | return res; |
4082 | } | |
4083 | ||
374eeb5a | 4084 | |
bb1d9123 AG |
4085 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
4086 | { | |
4087 | /* | |
4088 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4089 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4090 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4091 | * way down to the bonding driver. | |
4092 | */ | |
fd0e435b PO |
4093 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4094 | ||
374eeb5a NH |
4095 | /* |
4096 | * Save the original txq to restore before passing to the driver | |
4097 | */ | |
df4ab5b3 | 4098 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 4099 | |
fd0e435b | 4100 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4101 | do { |
fd0e435b | 4102 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4103 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4104 | } |
4105 | return txq; | |
bb1d9123 AG |
4106 | } |
4107 | ||
0693e88e | 4108 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4109 | { |
bb1d9123 AG |
4110 | struct bonding *bond = netdev_priv(dev); |
4111 | ||
4112 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
4113 | if (!bond_slave_override(bond, skb)) | |
4114 | return NETDEV_TX_OK; | |
4115 | } | |
00829823 SH |
4116 | |
4117 | switch (bond->params.mode) { | |
4118 | case BOND_MODE_ROUNDROBIN: | |
4119 | return bond_xmit_roundrobin(skb, dev); | |
4120 | case BOND_MODE_ACTIVEBACKUP: | |
4121 | return bond_xmit_activebackup(skb, dev); | |
4122 | case BOND_MODE_XOR: | |
4123 | return bond_xmit_xor(skb, dev); | |
4124 | case BOND_MODE_BROADCAST: | |
4125 | return bond_xmit_broadcast(skb, dev); | |
4126 | case BOND_MODE_8023AD: | |
4127 | return bond_3ad_xmit_xor(skb, dev); | |
4128 | case BOND_MODE_ALB: | |
4129 | case BOND_MODE_TLB: | |
4130 | return bond_alb_xmit(skb, dev); | |
4131 | default: | |
4132 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4133 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4134 | dev->name, bond->params.mode); | |
00829823 | 4135 | WARN_ON_ONCE(1); |
04502430 | 4136 | kfree_skb(skb); |
00829823 SH |
4137 | return NETDEV_TX_OK; |
4138 | } | |
4139 | } | |
4140 | ||
0693e88e MM |
4141 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4142 | { | |
4143 | struct bonding *bond = netdev_priv(dev); | |
4144 | netdev_tx_t ret = NETDEV_TX_OK; | |
4145 | ||
4146 | /* | |
4147 | * If we risk deadlock from transmitting this in the | |
4148 | * netpoll path, tell netpoll to queue the frame for later tx | |
4149 | */ | |
4150 | if (is_netpoll_tx_blocked(dev)) | |
4151 | return NETDEV_TX_BUSY; | |
4152 | ||
4153 | read_lock(&bond->lock); | |
4154 | ||
4155 | if (bond->slave_cnt) | |
4156 | ret = __bond_start_xmit(skb, dev); | |
4157 | else | |
04502430 | 4158 | kfree_skb(skb); |
0693e88e MM |
4159 | |
4160 | read_unlock(&bond->lock); | |
4161 | ||
4162 | return ret; | |
4163 | } | |
00829823 | 4164 | |
1da177e4 LT |
4165 | /* |
4166 | * set bond mode specific net device operations | |
4167 | */ | |
a77b5325 | 4168 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4169 | { |
169a3e66 JV |
4170 | struct net_device *bond_dev = bond->dev; |
4171 | ||
1da177e4 LT |
4172 | switch (mode) { |
4173 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4174 | break; |
4175 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4176 | break; |
4177 | case BOND_MODE_XOR: | |
6f6652be | 4178 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4179 | break; |
4180 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4181 | break; |
4182 | case BOND_MODE_8023AD: | |
6f6652be | 4183 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4184 | break; |
1da177e4 | 4185 | case BOND_MODE_ALB: |
8f903c70 JV |
4186 | /* FALLTHRU */ |
4187 | case BOND_MODE_TLB: | |
1da177e4 LT |
4188 | break; |
4189 | default: | |
4190 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4191 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4192 | bond_dev->name, mode); | |
1da177e4 LT |
4193 | break; |
4194 | } | |
4195 | } | |
4196 | ||
bb5b052f AG |
4197 | static int bond_ethtool_get_settings(struct net_device *bond_dev, |
4198 | struct ethtool_cmd *ecmd) | |
4199 | { | |
4200 | struct bonding *bond = netdev_priv(bond_dev); | |
4201 | struct slave *slave; | |
4202 | int i; | |
4203 | unsigned long speed = 0; | |
4204 | ||
4205 | ecmd->duplex = DUPLEX_UNKNOWN; | |
4206 | ecmd->port = PORT_OTHER; | |
4207 | ||
4208 | /* Since SLAVE_IS_OK returns false for all inactive or down slaves, we | |
4209 | * do not need to check mode. Though link speed might not represent | |
4210 | * the true receive or transmit bandwidth (not all modes are symmetric) | |
4211 | * this is an accurate maximum. | |
4212 | */ | |
4213 | read_lock(&bond->lock); | |
4214 | bond_for_each_slave(bond, slave, i) { | |
4215 | if (SLAVE_IS_OK(slave)) { | |
4216 | if (slave->speed != SPEED_UNKNOWN) | |
4217 | speed += slave->speed; | |
4218 | if (ecmd->duplex == DUPLEX_UNKNOWN && | |
4219 | slave->duplex != DUPLEX_UNKNOWN) | |
4220 | ecmd->duplex = slave->duplex; | |
4221 | } | |
4222 | } | |
4223 | ethtool_cmd_speed_set(ecmd, speed ? : SPEED_UNKNOWN); | |
4224 | read_unlock(&bond->lock); | |
4225 | return 0; | |
4226 | } | |
4227 | ||
217df670 | 4228 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 4229 | struct ethtool_drvinfo *drvinfo) |
217df670 | 4230 | { |
7826d43f JP |
4231 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
4232 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
4233 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
4234 | BOND_ABI_VERSION); | |
217df670 JV |
4235 | } |
4236 | ||
7282d491 | 4237 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4238 | .get_drvinfo = bond_ethtool_get_drvinfo, |
bb5b052f | 4239 | .get_settings = bond_ethtool_get_settings, |
fa53ebac | 4240 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
4241 | }; |
4242 | ||
eb7cc59a | 4243 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4244 | .ndo_init = bond_init, |
9e71626c | 4245 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4246 | .ndo_open = bond_open, |
4247 | .ndo_stop = bond_close, | |
00829823 | 4248 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4249 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4250 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4251 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4252 | .ndo_change_rx_flags = bond_change_rx_flags, |
303d1cbf | 4253 | .ndo_set_rx_mode = bond_set_rx_mode, |
eb7cc59a | 4254 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4255 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4256 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4257 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4258 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4259 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4260 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4261 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4262 | .ndo_poll_controller = bond_poll_controller, | |
4263 | #endif | |
9232ecca JP |
4264 | .ndo_add_slave = bond_enslave, |
4265 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4266 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
4267 | }; |
4268 | ||
b3f92b63 DG |
4269 | static const struct device_type bond_type = { |
4270 | .name = "bond", | |
4271 | }; | |
4272 | ||
9e2e61fb AW |
4273 | static void bond_destructor(struct net_device *bond_dev) |
4274 | { | |
4275 | struct bonding *bond = netdev_priv(bond_dev); | |
4276 | if (bond->wq) | |
4277 | destroy_workqueue(bond->wq); | |
4278 | free_netdev(bond_dev); | |
4279 | } | |
4280 | ||
181470fc | 4281 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4282 | { |
454d7c9b | 4283 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4284 | |
1da177e4 LT |
4285 | /* initialize rwlocks */ |
4286 | rwlock_init(&bond->lock); | |
4287 | rwlock_init(&bond->curr_slave_lock); | |
4288 | ||
d2991f75 | 4289 | bond->params = bonding_defaults; |
1da177e4 LT |
4290 | |
4291 | /* Initialize pointers */ | |
1da177e4 LT |
4292 | bond->dev = bond_dev; |
4293 | INIT_LIST_HEAD(&bond->vlan_list); | |
4294 | ||
4295 | /* Initialize the device entry points */ | |
181470fc | 4296 | ether_setup(bond_dev); |
eb7cc59a | 4297 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4298 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4299 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4300 | |
9e2e61fb | 4301 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4302 | |
b3f92b63 DG |
4303 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4304 | ||
1da177e4 LT |
4305 | /* Initialize the device options */ |
4306 | bond_dev->tx_queue_len = 0; | |
4307 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4308 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4309 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4310 | |
1da177e4 LT |
4311 | /* At first, we block adding VLANs. That's the only way to |
4312 | * prevent problems that occur when adding VLANs over an | |
4313 | * empty bond. The block will be removed once non-challenged | |
4314 | * slaves are enslaved. | |
4315 | */ | |
4316 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4317 | ||
932ff279 | 4318 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4319 | * transmitting */ |
4320 | bond_dev->features |= NETIF_F_LLTX; | |
4321 | ||
4322 | /* By default, we declare the bond to be fully | |
4323 | * VLAN hardware accelerated capable. Special | |
4324 | * care is taken in the various xmit functions | |
4325 | * when there are slaves that are not hw accel | |
4326 | * capable | |
4327 | */ | |
1da177e4 | 4328 | |
b2a103e6 | 4329 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
f646968f PM |
4330 | NETIF_F_HW_VLAN_CTAG_TX | |
4331 | NETIF_F_HW_VLAN_CTAG_RX | | |
4332 | NETIF_F_HW_VLAN_CTAG_FILTER; | |
b2a103e6 | 4333 | |
34324dc2 | 4334 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4335 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4336 | } |
4337 | ||
c67dfb29 EB |
4338 | /* |
4339 | * Destroy a bonding device. | |
4340 | * Must be under rtnl_lock when this function is called. | |
4341 | */ | |
4342 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4343 | { |
454d7c9b | 4344 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4345 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4346 | |
f6dc31a8 WC |
4347 | bond_netpoll_cleanup(bond_dev); |
4348 | ||
c67dfb29 | 4349 | /* Release the bonded slaves */ |
0896341a | 4350 | while (bond->first_slave != NULL) |
4351 | __bond_release_one(bond_dev, bond->first_slave->dev, true); | |
4352 | pr_info("%s: released all slaves\n", bond_dev->name); | |
c67dfb29 | 4353 | |
a434e43f JV |
4354 | list_del(&bond->bond_list); |
4355 | ||
f073c7ca TI |
4356 | bond_debug_unregister(bond); |
4357 | ||
f35188fa JV |
4358 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { |
4359 | list_del(&vlan->vlan_list); | |
4360 | kfree(vlan); | |
4361 | } | |
a434e43f JV |
4362 | } |
4363 | ||
1da177e4 LT |
4364 | /*------------------------- Module initialization ---------------------------*/ |
4365 | ||
4366 | /* | |
4367 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4368 | * number of the mode or its string name. A bit complicated because |
4369 | * some mode names are substrings of other names, and calls from sysfs | |
4370 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4371 | */ |
325dcf7a | 4372 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4373 | { |
54b87323 | 4374 | int modeint = -1, i, rv; |
a42e534f | 4375 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4376 | |
a42e534f JV |
4377 | for (p = (char *)buf; *p; p++) |
4378 | if (!(isdigit(*p) || isspace(*p))) | |
4379 | break; | |
4380 | ||
4381 | if (*p) | |
ece95f7f | 4382 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4383 | else |
54b87323 | 4384 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4385 | |
4386 | if (!rv) | |
4387 | return -1; | |
1da177e4 LT |
4388 | |
4389 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4390 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4391 | return tbl[i].mode; |
4392 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4393 | return tbl[i].mode; |
1da177e4 LT |
4394 | } |
4395 | ||
4396 | return -1; | |
4397 | } | |
4398 | ||
4399 | static int bond_check_params(struct bond_params *params) | |
4400 | { | |
5a5c5fd4 | 4401 | int arp_validate_value, fail_over_mac_value, primary_reselect_value, i; |
8599b52e | 4402 | int arp_all_targets_value; |
f5b2b966 | 4403 | |
1da177e4 LT |
4404 | /* |
4405 | * Convert string parameters. | |
4406 | */ | |
4407 | if (mode) { | |
4408 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4409 | if (bond_mode == -1) { | |
a4aee5c8 | 4410 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4411 | mode == NULL ? "NULL" : mode); |
4412 | return -EINVAL; | |
4413 | } | |
4414 | } | |
4415 | ||
169a3e66 JV |
4416 | if (xmit_hash_policy) { |
4417 | if ((bond_mode != BOND_MODE_XOR) && | |
4418 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4419 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4420 | bond_mode_name(bond_mode)); |
4421 | } else { | |
4422 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4423 | xmit_hashtype_tbl); | |
4424 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4425 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4426 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4427 | xmit_hash_policy); |
4428 | return -EINVAL; | |
4429 | } | |
4430 | } | |
4431 | } | |
4432 | ||
1da177e4 LT |
4433 | if (lacp_rate) { |
4434 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4435 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4436 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4437 | } else { |
4438 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4439 | if (lacp_fast == -1) { | |
a4aee5c8 | 4440 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4441 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4442 | return -EINVAL; | |
4443 | } | |
4444 | } | |
4445 | } | |
4446 | ||
fd989c83 JV |
4447 | if (ad_select) { |
4448 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4449 | if (params->ad_select == -1) { | |
a4aee5c8 | 4450 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4451 | ad_select == NULL ? "NULL" : ad_select); |
4452 | return -EINVAL; | |
4453 | } | |
4454 | ||
4455 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4456 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4457 | } |
4458 | } else { | |
4459 | params->ad_select = BOND_AD_STABLE; | |
4460 | } | |
4461 | ||
f5841306 | 4462 | if (max_bonds < 0) { |
a4aee5c8 JP |
4463 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4464 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4465 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4466 | } | |
4467 | ||
4468 | if (miimon < 0) { | |
a4aee5c8 JP |
4469 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4470 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4471 | miimon = BOND_LINK_MON_INTERV; |
4472 | } | |
4473 | ||
4474 | if (updelay < 0) { | |
a4aee5c8 JP |
4475 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4476 | updelay, INT_MAX); | |
1da177e4 LT |
4477 | updelay = 0; |
4478 | } | |
4479 | ||
4480 | if (downdelay < 0) { | |
a4aee5c8 JP |
4481 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4482 | downdelay, INT_MAX); | |
1da177e4 LT |
4483 | downdelay = 0; |
4484 | } | |
4485 | ||
4486 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4487 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4488 | use_carrier); | |
1da177e4 LT |
4489 | use_carrier = 1; |
4490 | } | |
4491 | ||
ad246c99 BH |
4492 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4493 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4494 | num_peer_notif); | |
4495 | num_peer_notif = 1; | |
4496 | } | |
4497 | ||
1da177e4 LT |
4498 | /* reset values for 802.3ad */ |
4499 | if (bond_mode == BOND_MODE_8023AD) { | |
4500 | if (!miimon) { | |
a4aee5c8 | 4501 | 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 | 4502 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4503 | miimon = 100; |
4504 | } | |
4505 | } | |
4506 | ||
bb1d9123 AG |
4507 | if (tx_queues < 1 || tx_queues > 255) { |
4508 | pr_warning("Warning: tx_queues (%d) should be between " | |
4509 | "1 and 255, resetting to %d\n", | |
4510 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4511 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4512 | } | |
4513 | ||
ebd8e497 AG |
4514 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4515 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4516 | "not of valid value (0/1), so it was set to " | |
4517 | "0\n", all_slaves_active); | |
4518 | all_slaves_active = 0; | |
4519 | } | |
4520 | ||
c2952c31 FL |
4521 | if (resend_igmp < 0 || resend_igmp > 255) { |
4522 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4523 | "0 and 255, resetting to %d\n", | |
4524 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4525 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4526 | } | |
4527 | ||
1da177e4 LT |
4528 | /* reset values for TLB/ALB */ |
4529 | if ((bond_mode == BOND_MODE_TLB) || | |
4530 | (bond_mode == BOND_MODE_ALB)) { | |
4531 | if (!miimon) { | |
a4aee5c8 | 4532 | 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 | 4533 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4534 | miimon = 100; |
4535 | } | |
4536 | } | |
4537 | ||
4538 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4539 | 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", |
4540 | updelay); | |
1da177e4 LT |
4541 | } |
4542 | ||
4543 | if (!miimon) { | |
4544 | if (updelay || downdelay) { | |
4545 | /* just warn the user the up/down delay will have | |
4546 | * no effect since miimon is zero... | |
4547 | */ | |
a4aee5c8 JP |
4548 | 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", |
4549 | updelay, downdelay); | |
1da177e4 LT |
4550 | } |
4551 | } else { | |
4552 | /* don't allow arp monitoring */ | |
4553 | if (arp_interval) { | |
a4aee5c8 JP |
4554 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4555 | miimon, arp_interval); | |
1da177e4 LT |
4556 | arp_interval = 0; |
4557 | } | |
4558 | ||
4559 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4560 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4561 | updelay, miimon, | |
4562 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4563 | } |
4564 | ||
4565 | updelay /= miimon; | |
4566 | ||
4567 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4568 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4569 | downdelay, miimon, | |
4570 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4571 | } |
4572 | ||
4573 | downdelay /= miimon; | |
4574 | } | |
4575 | ||
4576 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4577 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4578 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4579 | arp_interval = BOND_LINK_ARP_INTERV; |
4580 | } | |
4581 | ||
5a5c5fd4 | 4582 | for (arp_ip_count = 0, i = 0; |
4583 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[i]; i++) { | |
1da177e4 LT |
4584 | /* not complete check, but should be good enough to |
4585 | catch mistakes */ | |
5a5c5fd4 | 4586 | __be32 ip = in_aton(arp_ip_target[i]); |
4587 | if (!isdigit(arp_ip_target[i][0]) || ip == 0 || | |
4588 | ip == htonl(INADDR_BROADCAST)) { | |
a4aee5c8 | 4589 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
5a5c5fd4 | 4590 | arp_ip_target[i]); |
1da177e4 LT |
4591 | arp_interval = 0; |
4592 | } else { | |
0afee4e8 VF |
4593 | if (bond_get_targets_ip(arp_target, ip) == -1) |
4594 | arp_target[arp_ip_count++] = ip; | |
4595 | else | |
4596 | pr_warning("Warning: duplicate address %pI4 in arp_ip_target, skipping\n", | |
4597 | &ip); | |
1da177e4 LT |
4598 | } |
4599 | } | |
4600 | ||
4601 | if (arp_interval && !arp_ip_count) { | |
4602 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4603 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4604 | arp_interval); | |
1da177e4 LT |
4605 | arp_interval = 0; |
4606 | } | |
4607 | ||
f5b2b966 JV |
4608 | if (arp_validate) { |
4609 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4610 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4611 | return -EINVAL; |
4612 | } | |
4613 | if (!arp_interval) { | |
a4aee5c8 | 4614 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4615 | return -EINVAL; |
4616 | } | |
4617 | ||
4618 | arp_validate_value = bond_parse_parm(arp_validate, | |
4619 | arp_validate_tbl); | |
4620 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4621 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4622 | arp_validate == NULL ? "NULL" : arp_validate); |
4623 | return -EINVAL; | |
4624 | } | |
4625 | } else | |
4626 | arp_validate_value = 0; | |
4627 | ||
8599b52e VF |
4628 | arp_all_targets_value = 0; |
4629 | if (arp_all_targets) { | |
4630 | arp_all_targets_value = bond_parse_parm(arp_all_targets, | |
4631 | arp_all_targets_tbl); | |
4632 | ||
4633 | if (arp_all_targets_value == -1) { | |
4634 | pr_err("Error: invalid arp_all_targets_value \"%s\"\n", | |
4635 | arp_all_targets); | |
4636 | arp_all_targets_value = 0; | |
4637 | } | |
4638 | } | |
4639 | ||
1da177e4 | 4640 | if (miimon) { |
a4aee5c8 | 4641 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 | 4642 | } else if (arp_interval) { |
a4aee5c8 JP |
4643 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4644 | arp_interval, | |
4645 | arp_validate_tbl[arp_validate_value].modename, | |
4646 | arp_ip_count); | |
1da177e4 LT |
4647 | |
4648 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4649 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4650 | |
e5e2a8fd | 4651 | pr_info("\n"); |
1da177e4 | 4652 | |
b8a9787e | 4653 | } else if (max_bonds) { |
1da177e4 LT |
4654 | /* miimon and arp_interval not set, we need one so things |
4655 | * work as expected, see bonding.txt for details | |
4656 | */ | |
b2730f4f | 4657 | 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 |
4658 | } |
4659 | ||
4660 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4661 | /* currently, using a primary only makes sense | |
4662 | * in active backup, TLB or ALB modes | |
4663 | */ | |
a4aee5c8 JP |
4664 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4665 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4666 | primary = NULL; |
4667 | } | |
4668 | ||
a549952a JP |
4669 | if (primary && primary_reselect) { |
4670 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4671 | pri_reselect_tbl); | |
4672 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4673 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4674 | primary_reselect == |
4675 | NULL ? "NULL" : primary_reselect); | |
4676 | return -EINVAL; | |
4677 | } | |
4678 | } else { | |
4679 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4680 | } | |
4681 | ||
3915c1e8 JV |
4682 | if (fail_over_mac) { |
4683 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4684 | fail_over_mac_tbl); | |
4685 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4686 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4687 | arp_validate == NULL ? "NULL" : arp_validate); |
4688 | return -EINVAL; | |
4689 | } | |
4690 | ||
4691 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4692 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4693 | } else { |
4694 | fail_over_mac_value = BOND_FOM_NONE; | |
4695 | } | |
dd957c57 | 4696 | |
1da177e4 LT |
4697 | /* fill params struct with the proper values */ |
4698 | params->mode = bond_mode; | |
169a3e66 | 4699 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4700 | params->miimon = miimon; |
ad246c99 | 4701 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4702 | params->arp_interval = arp_interval; |
f5b2b966 | 4703 | params->arp_validate = arp_validate_value; |
8599b52e | 4704 | params->arp_all_targets = arp_all_targets_value; |
1da177e4 LT |
4705 | params->updelay = updelay; |
4706 | params->downdelay = downdelay; | |
4707 | params->use_carrier = use_carrier; | |
4708 | params->lacp_fast = lacp_fast; | |
4709 | params->primary[0] = 0; | |
a549952a | 4710 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4711 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4712 | params->tx_queues = tx_queues; |
ebd8e497 | 4713 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4714 | params->resend_igmp = resend_igmp; |
655f8919 | 4715 | params->min_links = min_links; |
1da177e4 LT |
4716 | |
4717 | if (primary) { | |
4718 | strncpy(params->primary, primary, IFNAMSIZ); | |
4719 | params->primary[IFNAMSIZ - 1] = 0; | |
4720 | } | |
4721 | ||
4722 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4723 | ||
4724 | return 0; | |
4725 | } | |
4726 | ||
0daa2303 | 4727 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4728 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4729 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4730 | |
e8a0464c DM |
4731 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4732 | struct netdev_queue *txq, | |
4733 | void *_unused) | |
c773e847 DM |
4734 | { |
4735 | lockdep_set_class(&txq->_xmit_lock, | |
4736 | &bonding_netdev_xmit_lock_key); | |
4737 | } | |
4738 | ||
4739 | static void bond_set_lockdep_class(struct net_device *dev) | |
4740 | { | |
cf508b12 DM |
4741 | lockdep_set_class(&dev->addr_list_lock, |
4742 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4743 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4744 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4745 | } |
4746 | ||
181470fc SH |
4747 | /* |
4748 | * Called from registration process | |
4749 | */ | |
4750 | static int bond_init(struct net_device *bond_dev) | |
4751 | { | |
4752 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4753 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4754 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4755 | |
4756 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4757 | ||
9fe0617d NH |
4758 | /* |
4759 | * Initialize locks that may be required during | |
4760 | * en/deslave operations. All of the bond_open work | |
4761 | * (of which this is part) should really be moved to | |
4762 | * a phase prior to dev_open | |
4763 | */ | |
4764 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4765 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4766 | ||
181470fc SH |
4767 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4768 | if (!bond->wq) | |
4769 | return -ENOMEM; | |
4770 | ||
4771 | bond_set_lockdep_class(bond_dev); | |
4772 | ||
ec87fd3b | 4773 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4774 | |
6151b3d4 | 4775 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4776 | |
f073c7ca TI |
4777 | bond_debug_register(bond); |
4778 | ||
409cc1f8 JP |
4779 | /* Ensure valid dev_addr */ |
4780 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
97a1e639 | 4781 | bond_dev->addr_assign_type == NET_ADDR_PERM) |
409cc1f8 | 4782 | eth_hw_addr_random(bond_dev); |
409cc1f8 | 4783 | |
181470fc SH |
4784 | return 0; |
4785 | } | |
4786 | ||
88ead977 EB |
4787 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4788 | { | |
4789 | if (tb[IFLA_ADDRESS]) { | |
4790 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4791 | return -EINVAL; | |
4792 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4793 | return -EADDRNOTAVAIL; | |
4794 | } | |
4795 | return 0; | |
4796 | } | |
4797 | ||
d40156aa | 4798 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4799 | { |
efacb309 | 4800 | return tx_queues; |
d5da4510 JP |
4801 | } |
4802 | ||
88ead977 | 4803 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4804 | .kind = "bond", |
4805 | .priv_size = sizeof(struct bonding), | |
4806 | .setup = bond_setup, | |
4807 | .validate = bond_validate, | |
4808 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4809 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4810 | as for TX queues */ | |
88ead977 EB |
4811 | }; |
4812 | ||
dfe60397 | 4813 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4814 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4815 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4816 | * set up our sysfs entries. | |
4817 | */ | |
ec87fd3b | 4818 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4819 | { |
4820 | struct net_device *bond_dev; | |
4821 | int res; | |
4822 | ||
4823 | rtnl_lock(); | |
027ea041 | 4824 | |
1c5cae81 JP |
4825 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4826 | name ? name : "bond%d", | |
4827 | bond_setup, tx_queues); | |
dfe60397 | 4828 | if (!bond_dev) { |
a4aee5c8 | 4829 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4830 | rtnl_unlock(); |
4831 | return -ENOMEM; | |
dfe60397 MW |
4832 | } |
4833 | ||
ec87fd3b | 4834 | dev_net_set(bond_dev, net); |
88ead977 EB |
4835 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4836 | ||
dfe60397 | 4837 | res = register_netdevice(bond_dev); |
0daa2303 | 4838 | |
e826eafa PO |
4839 | netif_carrier_off(bond_dev); |
4840 | ||
7e083840 | 4841 | rtnl_unlock(); |
9e2e61fb AW |
4842 | if (res < 0) |
4843 | bond_destructor(bond_dev); | |
30c15ba9 | 4844 | return res; |
dfe60397 MW |
4845 | } |
4846 | ||
2c8c1e72 | 4847 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4848 | { |
15449745 | 4849 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4850 | |
4851 | bn->net = net; | |
4852 | INIT_LIST_HEAD(&bn->dev_list); | |
4853 | ||
ec87fd3b | 4854 | bond_create_proc_dir(bn); |
4c22400a | 4855 | bond_create_sysfs(bn); |
87a7b84b | 4856 | |
15449745 | 4857 | return 0; |
ec87fd3b EB |
4858 | } |
4859 | ||
2c8c1e72 | 4860 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4861 | { |
15449745 | 4862 | struct bond_net *bn = net_generic(net, bond_net_id); |
69b0216a | 4863 | struct bonding *bond, *tmp_bond; |
4864 | LIST_HEAD(list); | |
ec87fd3b | 4865 | |
4c22400a | 4866 | bond_destroy_sysfs(bn); |
ec87fd3b | 4867 | bond_destroy_proc_dir(bn); |
69b0216a | 4868 | |
4869 | /* Kill off any bonds created after unregistering bond rtnl ops */ | |
4870 | rtnl_lock(); | |
4871 | list_for_each_entry_safe(bond, tmp_bond, &bn->dev_list, bond_list) | |
4872 | unregister_netdevice_queue(bond->dev, &list); | |
4873 | unregister_netdevice_many(&list); | |
4874 | rtnl_unlock(); | |
ec87fd3b EB |
4875 | } |
4876 | ||
4877 | static struct pernet_operations bond_net_ops = { | |
4878 | .init = bond_net_init, | |
4879 | .exit = bond_net_exit, | |
15449745 EB |
4880 | .id = &bond_net_id, |
4881 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4882 | }; |
4883 | ||
1da177e4 LT |
4884 | static int __init bonding_init(void) |
4885 | { | |
1da177e4 LT |
4886 | int i; |
4887 | int res; | |
4888 | ||
bd33acc3 | 4889 | pr_info("%s", bond_version); |
1da177e4 | 4890 | |
dfe60397 | 4891 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4892 | if (res) |
dfe60397 | 4893 | goto out; |
1da177e4 | 4894 | |
15449745 | 4895 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4896 | if (res) |
4897 | goto out; | |
027ea041 | 4898 | |
88ead977 EB |
4899 | res = rtnl_link_register(&bond_link_ops); |
4900 | if (res) | |
6639104b | 4901 | goto err_link; |
88ead977 | 4902 | |
f073c7ca TI |
4903 | bond_create_debugfs(); |
4904 | ||
1da177e4 | 4905 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4906 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4907 | if (res) |
4908 | goto err; | |
1da177e4 LT |
4909 | } |
4910 | ||
1da177e4 | 4911 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4912 | out: |
1da177e4 | 4913 | return res; |
88ead977 EB |
4914 | err: |
4915 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 4916 | err_link: |
15449745 | 4917 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4918 | goto out; |
dfe60397 | 4919 | |
1da177e4 LT |
4920 | } |
4921 | ||
4922 | static void __exit bonding_exit(void) | |
4923 | { | |
4924 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4925 | ||
f073c7ca | 4926 | bond_destroy_debugfs(); |
ae68c398 | 4927 | |
4de79c73 | 4928 | rtnl_link_unregister(&bond_link_ops); |
ffcdedb6 | 4929 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
4930 | |
4931 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4932 | /* |
4933 | * Make sure we don't have an imbalance on our netpoll blocking | |
4934 | */ | |
4935 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4936 | #endif |
1da177e4 LT |
4937 | } |
4938 | ||
4939 | module_init(bonding_init); | |
4940 | module_exit(bonding_exit); | |
4941 | MODULE_LICENSE("GPL"); | |
4942 | MODULE_VERSION(DRV_VERSION); | |
4943 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4944 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 4945 | MODULE_ALIAS_RTNL_LINK("bond"); |