]>
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; | |
107 | static char *fail_over_mac; | |
ebd8e497 | 108 | static int all_slaves_active = 0; |
d2991f75 | 109 | static struct bond_params bonding_defaults; |
c2952c31 | 110 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
1da177e4 LT |
111 | |
112 | module_param(max_bonds, int, 0); | |
113 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
114 | module_param(tx_queues, int, 0); |
115 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
ad246c99 | 116 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
117 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
118 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 119 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
120 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
121 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
122 | module_param(miimon, int, 0); |
123 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
124 | module_param(updelay, int, 0); | |
125 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
126 | module_param(downdelay, int, 0); | |
2ac47660 MW |
127 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
128 | "in milliseconds"); | |
1da177e4 | 129 | module_param(use_carrier, int, 0); |
2ac47660 MW |
130 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
131 | "0 for off, 1 for on (default)"); | |
1da177e4 | 132 | module_param(mode, charp, 0); |
90e62474 | 133 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
134 | "1 for active-backup, 2 for balance-xor, " |
135 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
136 | "6 for balance-alb"); | |
1da177e4 LT |
137 | module_param(primary, charp, 0); |
138 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
139 | module_param(primary_reselect, charp, 0); |
140 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
141 | "once it comes up; " | |
142 | "0 for always (default), " | |
143 | "1 for only if speed of primary is " | |
144 | "better, " | |
145 | "2 for only on active slave " | |
146 | "failure"); | |
1da177e4 | 147 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
148 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
149 | "0 for slow, 1 for fast"); | |
fd989c83 | 150 | module_param(ad_select, charp, 0); |
90e62474 AG |
151 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
152 | "0 for stable (default), 1 for bandwidth, " | |
153 | "2 for count"); | |
655f8919 | 154 | module_param(min_links, int, 0); |
155 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
156 | ||
169a3e66 | 157 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
158 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
159 | "0 for layer 2 (default), 1 for layer 3+4, " | |
160 | "2 for layer 2+3"); | |
1da177e4 LT |
161 | module_param(arp_interval, int, 0); |
162 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
163 | module_param_array(arp_ip_target, charp, NULL, 0); | |
164 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 165 | module_param(arp_validate, charp, 0); |
90e62474 AG |
166 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
167 | "0 for none (default), 1 for active, " | |
168 | "2 for backup, 3 for all"); | |
3915c1e8 | 169 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
170 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
171 | "the same MAC; 0 for none (default), " | |
172 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
173 | module_param(all_slaves_active, int, 0); |
174 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 175 | "by setting active flag for all slaves; " |
ebd8e497 | 176 | "0 for never (default), 1 for always."); |
c2952c31 | 177 | module_param(resend_igmp, int, 0); |
90e62474 AG |
178 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
179 | "link failure"); | |
1da177e4 LT |
180 | |
181 | /*----------------------------- Global variables ----------------------------*/ | |
182 | ||
e843fa50 | 183 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 184 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
185 | #endif |
186 | ||
f99189b1 | 187 | int bond_net_id __read_mostly; |
1da177e4 | 188 | |
3d632c3f SH |
189 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
190 | static int arp_ip_count; | |
1da177e4 | 191 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
192 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
193 | static int lacp_fast; | |
1da177e4 | 194 | |
e97fd7c6 | 195 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
196 | { "slow", AD_LACP_SLOW}, |
197 | { "fast", AD_LACP_FAST}, | |
198 | { NULL, -1}, | |
199 | }; | |
200 | ||
e97fd7c6 | 201 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
202 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
203 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
204 | { "balance-xor", BOND_MODE_XOR}, | |
205 | { "broadcast", BOND_MODE_BROADCAST}, | |
206 | { "802.3ad", BOND_MODE_8023AD}, | |
207 | { "balance-tlb", BOND_MODE_TLB}, | |
208 | { "balance-alb", BOND_MODE_ALB}, | |
209 | { NULL, -1}, | |
210 | }; | |
211 | ||
e97fd7c6 | 212 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
213 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
214 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 215 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
216 | { NULL, -1}, |
217 | }; | |
218 | ||
e97fd7c6 | 219 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
220 | { "none", BOND_ARP_VALIDATE_NONE}, |
221 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
222 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
223 | { "all", BOND_ARP_VALIDATE_ALL}, | |
224 | { NULL, -1}, | |
225 | }; | |
226 | ||
e97fd7c6 | 227 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
228 | { "none", BOND_FOM_NONE}, |
229 | { "active", BOND_FOM_ACTIVE}, | |
230 | { "follow", BOND_FOM_FOLLOW}, | |
231 | { NULL, -1}, | |
232 | }; | |
233 | ||
a549952a JP |
234 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
235 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
236 | { "better", BOND_PRI_RESELECT_BETTER}, | |
237 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
238 | { NULL, -1}, | |
239 | }; | |
240 | ||
fd989c83 JV |
241 | struct bond_parm_tbl ad_select_tbl[] = { |
242 | { "stable", BOND_AD_STABLE}, | |
243 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
244 | { "count", BOND_AD_COUNT}, | |
245 | { NULL, -1}, | |
246 | }; | |
247 | ||
1da177e4 LT |
248 | /*-------------------------- Forward declarations ---------------------------*/ |
249 | ||
181470fc | 250 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 251 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
252 | |
253 | /*---------------------------- General routines -----------------------------*/ | |
254 | ||
bd33acc3 | 255 | const char *bond_mode_name(int mode) |
1da177e4 | 256 | { |
77afc92b HE |
257 | static const char *names[] = { |
258 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
259 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
260 | [BOND_MODE_XOR] = "load balancing (xor)", | |
261 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 262 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
263 | [BOND_MODE_TLB] = "transmit load balancing", |
264 | [BOND_MODE_ALB] = "adaptive load balancing", | |
265 | }; | |
266 | ||
267 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 268 | return "unknown"; |
77afc92b HE |
269 | |
270 | return names[mode]; | |
1da177e4 LT |
271 | } |
272 | ||
273 | /*---------------------------------- VLAN -----------------------------------*/ | |
274 | ||
275 | /** | |
276 | * bond_add_vlan - add a new vlan id on bond | |
277 | * @bond: bond that got the notification | |
278 | * @vlan_id: the vlan id to add | |
279 | * | |
280 | * Returns -ENOMEM if allocation failed. | |
281 | */ | |
282 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
283 | { | |
284 | struct vlan_entry *vlan; | |
285 | ||
5a03cdb7 | 286 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 287 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 288 | |
305d552a | 289 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 290 | if (!vlan) |
1da177e4 | 291 | return -ENOMEM; |
1da177e4 LT |
292 | |
293 | INIT_LIST_HEAD(&vlan->vlan_list); | |
294 | vlan->vlan_id = vlan_id; | |
295 | ||
296 | write_lock_bh(&bond->lock); | |
297 | ||
298 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
299 | ||
300 | write_unlock_bh(&bond->lock); | |
301 | ||
5a03cdb7 | 302 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
303 | |
304 | return 0; | |
305 | } | |
306 | ||
307 | /** | |
308 | * bond_del_vlan - delete a vlan id from bond | |
309 | * @bond: bond that got the notification | |
310 | * @vlan_id: the vlan id to delete | |
311 | * | |
312 | * returns -ENODEV if @vlan_id was not found in @bond. | |
313 | */ | |
314 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
315 | { | |
0883beca | 316 | struct vlan_entry *vlan; |
1da177e4 LT |
317 | int res = -ENODEV; |
318 | ||
5a03cdb7 | 319 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 | 320 | |
e843fa50 | 321 | block_netpoll_tx(); |
1da177e4 LT |
322 | write_lock_bh(&bond->lock); |
323 | ||
0883beca | 324 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
325 | if (vlan->vlan_id == vlan_id) { |
326 | list_del(&vlan->vlan_list); | |
327 | ||
58402054 | 328 | if (bond_is_lb(bond)) |
1da177e4 | 329 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 330 | |
a4aee5c8 JP |
331 | pr_debug("removed VLAN ID %d from bond %s\n", |
332 | vlan_id, bond->dev->name); | |
1da177e4 LT |
333 | |
334 | kfree(vlan); | |
335 | ||
1da177e4 LT |
336 | res = 0; |
337 | goto out; | |
338 | } | |
339 | } | |
340 | ||
a4aee5c8 JP |
341 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
342 | vlan_id, bond->dev->name); | |
1da177e4 LT |
343 | |
344 | out: | |
345 | write_unlock_bh(&bond->lock); | |
e843fa50 | 346 | unblock_netpoll_tx(); |
1da177e4 LT |
347 | return res; |
348 | } | |
349 | ||
1da177e4 LT |
350 | /** |
351 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
352 | * @bond: the bond we're working on | |
353 | * @curr: item we're advancing from | |
354 | * | |
355 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
356 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 357 | * |
1da177e4 LT |
358 | * Caller must hold bond->lock |
359 | */ | |
360 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
361 | { | |
362 | struct vlan_entry *next, *last; | |
363 | ||
3d632c3f | 364 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 365 | return NULL; |
1da177e4 LT |
366 | |
367 | if (!curr) { | |
368 | next = list_entry(bond->vlan_list.next, | |
369 | struct vlan_entry, vlan_list); | |
370 | } else { | |
371 | last = list_entry(bond->vlan_list.prev, | |
372 | struct vlan_entry, vlan_list); | |
373 | if (last == curr) { | |
374 | next = list_entry(bond->vlan_list.next, | |
375 | struct vlan_entry, vlan_list); | |
376 | } else { | |
377 | next = list_entry(curr->vlan_list.next, | |
378 | struct vlan_entry, vlan_list); | |
379 | } | |
380 | } | |
381 | ||
382 | return next; | |
383 | } | |
384 | ||
385 | /** | |
386 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 387 | * |
1da177e4 LT |
388 | * @bond: bond device that got this skb for tx. |
389 | * @skb: hw accel VLAN tagged skb to transmit | |
390 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 391 | */ |
3d632c3f SH |
392 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
393 | struct net_device *slave_dev) | |
1da177e4 | 394 | { |
8387451e | 395 | skb->dev = slave_dev; |
374eeb5a | 396 | |
5ee31c68 | 397 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
398 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
399 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 400 | |
e15c3c22 | 401 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 402 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 403 | else |
f6dc31a8 | 404 | dev_queue_xmit(skb); |
1da177e4 LT |
405 | |
406 | return 0; | |
407 | } | |
408 | ||
409 | /* | |
cc0e4070 JP |
410 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
411 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
412 | * a. This operation is performed in IOCTL context, |
413 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
414 | * c. Holding a lock with BH disabled while directly calling a base driver | |
415 | * entry point is generally a BAD idea. | |
3d632c3f | 416 | * |
1da177e4 LT |
417 | * The design of synchronization/protection for this operation in the 8021q |
418 | * module is good for one or more VLAN devices over a single physical device | |
419 | * and cannot be extended for a teaming solution like bonding, so there is a | |
420 | * potential race condition here where a net device from the vlan group might | |
421 | * be referenced (either by a base driver or the 8021q code) while it is being | |
422 | * removed from the system. However, it turns out we're not making matters | |
423 | * worse, and if it works for regular VLAN usage it will work here too. | |
424 | */ | |
425 | ||
1da177e4 LT |
426 | /** |
427 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
428 | * @bond_dev: bonding net device that got called | |
429 | * @vid: vlan id being added | |
430 | */ | |
8e586137 | 431 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) |
1da177e4 | 432 | { |
454d7c9b | 433 | struct bonding *bond = netdev_priv(bond_dev); |
87002b03 | 434 | struct slave *slave, *stop_at; |
1da177e4 LT |
435 | int i, res; |
436 | ||
437 | bond_for_each_slave(bond, slave, i) { | |
87002b03 JP |
438 | res = vlan_vid_add(slave->dev, vid); |
439 | if (res) | |
440 | goto unwind; | |
1da177e4 LT |
441 | } |
442 | ||
443 | res = bond_add_vlan(bond, vid); | |
444 | if (res) { | |
a4aee5c8 | 445 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 | 446 | bond_dev->name, vid); |
8e586137 | 447 | return res; |
1da177e4 | 448 | } |
8e586137 JP |
449 | |
450 | return 0; | |
87002b03 JP |
451 | |
452 | unwind: | |
453 | /* unwind from head to the slave that failed */ | |
454 | stop_at = slave; | |
455 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) | |
456 | vlan_vid_del(slave->dev, vid); | |
457 | ||
458 | return res; | |
1da177e4 LT |
459 | } |
460 | ||
461 | /** | |
462 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
463 | * @bond_dev: bonding net device that got called | |
464 | * @vid: vlan id being removed | |
465 | */ | |
8e586137 | 466 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) |
1da177e4 | 467 | { |
454d7c9b | 468 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 469 | struct slave *slave; |
1da177e4 LT |
470 | int i, res; |
471 | ||
87002b03 JP |
472 | bond_for_each_slave(bond, slave, i) |
473 | vlan_vid_del(slave->dev, vid); | |
1da177e4 LT |
474 | |
475 | res = bond_del_vlan(bond, vid); | |
476 | if (res) { | |
a4aee5c8 | 477 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 | 478 | bond_dev->name, vid); |
8e586137 | 479 | return res; |
1da177e4 | 480 | } |
8e586137 JP |
481 | |
482 | return 0; | |
1da177e4 LT |
483 | } |
484 | ||
485 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
486 | { | |
487 | struct vlan_entry *vlan; | |
87002b03 | 488 | int res; |
1da177e4 | 489 | |
87002b03 JP |
490 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
491 | res = vlan_vid_add(slave_dev, vlan->vlan_id); | |
492 | if (res) | |
493 | pr_warning("%s: Failed to add vlan id %d to device %s\n", | |
494 | bond->dev->name, vlan->vlan_id, | |
495 | slave_dev->name); | |
496 | } | |
1da177e4 LT |
497 | } |
498 | ||
3d632c3f SH |
499 | static void bond_del_vlans_from_slave(struct bonding *bond, |
500 | struct net_device *slave_dev) | |
1da177e4 LT |
501 | { |
502 | struct vlan_entry *vlan; | |
1da177e4 | 503 | |
1da177e4 | 504 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
505 | if (!vlan->vlan_id) |
506 | continue; | |
87002b03 | 507 | vlan_vid_del(slave_dev, vlan->vlan_id); |
1da177e4 | 508 | } |
1da177e4 LT |
509 | } |
510 | ||
511 | /*------------------------------- Link status -------------------------------*/ | |
512 | ||
ff59c456 JV |
513 | /* |
514 | * Set the carrier state for the master according to the state of its | |
515 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
516 | * do special 802.3ad magic. | |
517 | * | |
518 | * Returns zero if carrier state does not change, nonzero if it does. | |
519 | */ | |
520 | static int bond_set_carrier(struct bonding *bond) | |
521 | { | |
522 | struct slave *slave; | |
523 | int i; | |
524 | ||
525 | if (bond->slave_cnt == 0) | |
526 | goto down; | |
527 | ||
528 | if (bond->params.mode == BOND_MODE_8023AD) | |
529 | return bond_3ad_set_carrier(bond); | |
530 | ||
531 | bond_for_each_slave(bond, slave, i) { | |
532 | if (slave->link == BOND_LINK_UP) { | |
533 | if (!netif_carrier_ok(bond->dev)) { | |
534 | netif_carrier_on(bond->dev); | |
535 | return 1; | |
536 | } | |
537 | return 0; | |
538 | } | |
539 | } | |
540 | ||
541 | down: | |
542 | if (netif_carrier_ok(bond->dev)) { | |
543 | netif_carrier_off(bond->dev); | |
544 | return 1; | |
545 | } | |
546 | return 0; | |
547 | } | |
548 | ||
1da177e4 LT |
549 | /* |
550 | * Get link speed and duplex from the slave's base driver | |
551 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 552 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 553 | * and return. |
1da177e4 | 554 | */ |
13b95fb7 | 555 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
556 | { |
557 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 558 | struct ethtool_cmd ecmd; |
5d30530e | 559 | u32 slave_speed; |
61a44b9c | 560 | int res; |
1da177e4 | 561 | |
589665f5 DC |
562 | slave->speed = SPEED_UNKNOWN; |
563 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 564 | |
4bc71cb9 | 565 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 566 | if (res < 0) |
13b95fb7 | 567 | return; |
1da177e4 | 568 | |
4bc71cb9 | 569 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 570 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 571 | return; |
1da177e4 | 572 | |
4bc71cb9 | 573 | switch (ecmd.duplex) { |
1da177e4 LT |
574 | case DUPLEX_FULL: |
575 | case DUPLEX_HALF: | |
576 | break; | |
577 | default: | |
13b95fb7 | 578 | return; |
1da177e4 LT |
579 | } |
580 | ||
5d30530e | 581 | slave->speed = slave_speed; |
4bc71cb9 | 582 | slave->duplex = ecmd.duplex; |
1da177e4 | 583 | |
13b95fb7 | 584 | return; |
1da177e4 LT |
585 | } |
586 | ||
587 | /* | |
588 | * if <dev> supports MII link status reporting, check its link status. | |
589 | * | |
590 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 591 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
592 | * |
593 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
594 | * can't tell and just pretend it is), or 0, meaning that the link is | |
595 | * down. | |
596 | * | |
597 | * If reporting is non-zero, instead of faking link up, return -1 if | |
598 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
599 | * support them). If use_carrier is set, return whatever it says. | |
600 | * It'd be nice if there was a good way to tell if a driver supports | |
601 | * netif_carrier, but there really isn't. | |
602 | */ | |
3d632c3f SH |
603 | static int bond_check_dev_link(struct bonding *bond, |
604 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 605 | { |
eb7cc59a | 606 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 607 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
608 | struct ifreq ifr; |
609 | struct mii_ioctl_data *mii; | |
1da177e4 | 610 | |
6c988853 PG |
611 | if (!reporting && !netif_running(slave_dev)) |
612 | return 0; | |
613 | ||
eb7cc59a | 614 | if (bond->params.use_carrier) |
1da177e4 | 615 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 616 | |
29112f4e JP |
617 | /* Try to get link status using Ethtool first. */ |
618 | if (slave_dev->ethtool_ops) { | |
619 | if (slave_dev->ethtool_ops->get_link) { | |
620 | u32 link; | |
621 | ||
622 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
623 | ||
624 | return link ? BMSR_LSTATUS : 0; | |
625 | } | |
626 | } | |
627 | ||
3d632c3f | 628 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 629 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
630 | if (ioctl) { |
631 | /* TODO: set pointer to correct ioctl on a per team member */ | |
632 | /* bases to make this more efficient. that is, once */ | |
633 | /* we determine the correct ioctl, we will always */ | |
634 | /* call it and not the others for that team */ | |
635 | /* member. */ | |
636 | ||
637 | /* | |
638 | * We cannot assume that SIOCGMIIPHY will also read a | |
639 | * register; not all network drivers (e.g., e100) | |
640 | * support that. | |
641 | */ | |
642 | ||
643 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
644 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
645 | mii = if_mii(&ifr); | |
646 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
647 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
648 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
649 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
650 | } |
651 | } | |
652 | ||
1da177e4 LT |
653 | /* |
654 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 655 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
656 | * cannot report link status). If not reporting, pretend |
657 | * we're ok. | |
658 | */ | |
3d632c3f | 659 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
660 | } |
661 | ||
662 | /*----------------------------- Multicast list ------------------------------*/ | |
663 | ||
1da177e4 LT |
664 | /* |
665 | * Push the promiscuity flag down to appropriate slaves | |
666 | */ | |
7e1a1ac1 | 667 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 668 | { |
7e1a1ac1 | 669 | int err = 0; |
1da177e4 LT |
670 | if (USES_PRIMARY(bond->params.mode)) { |
671 | /* write lock already acquired */ | |
672 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
673 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
674 | inc); | |
1da177e4 LT |
675 | } |
676 | } else { | |
677 | struct slave *slave; | |
678 | int i; | |
679 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
680 | err = dev_set_promiscuity(slave->dev, inc); |
681 | if (err) | |
682 | return err; | |
1da177e4 LT |
683 | } |
684 | } | |
7e1a1ac1 | 685 | return err; |
1da177e4 LT |
686 | } |
687 | ||
688 | /* | |
689 | * Push the allmulti flag down to all slaves | |
690 | */ | |
7e1a1ac1 | 691 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 692 | { |
7e1a1ac1 | 693 | int err = 0; |
1da177e4 LT |
694 | if (USES_PRIMARY(bond->params.mode)) { |
695 | /* write lock already acquired */ | |
696 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
697 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
698 | inc); | |
1da177e4 LT |
699 | } |
700 | } else { | |
701 | struct slave *slave; | |
702 | int i; | |
703 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
704 | err = dev_set_allmulti(slave->dev, inc); |
705 | if (err) | |
706 | return err; | |
1da177e4 LT |
707 | } |
708 | } | |
7e1a1ac1 | 709 | return err; |
1da177e4 LT |
710 | } |
711 | ||
712 | /* | |
713 | * Add a Multicast address to slaves | |
714 | * according to mode | |
715 | */ | |
22bedad3 | 716 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
717 | { |
718 | if (USES_PRIMARY(bond->params.mode)) { | |
719 | /* write lock already acquired */ | |
3d632c3f | 720 | if (bond->curr_active_slave) |
22bedad3 | 721 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
722 | } else { |
723 | struct slave *slave; | |
724 | int i; | |
3d632c3f SH |
725 | |
726 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 727 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
728 | } |
729 | } | |
730 | ||
731 | /* | |
732 | * Remove a multicast address from slave | |
733 | * according to mode | |
734 | */ | |
22bedad3 | 735 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
736 | { |
737 | if (USES_PRIMARY(bond->params.mode)) { | |
738 | /* write lock already acquired */ | |
3d632c3f | 739 | if (bond->curr_active_slave) |
22bedad3 | 740 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
741 | } else { |
742 | struct slave *slave; | |
743 | int i; | |
744 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 745 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
746 | } |
747 | } | |
748 | } | |
749 | ||
a816c7c7 | 750 | |
5a37e8ca | 751 | static void __bond_resend_igmp_join_requests(struct net_device *dev) |
a816c7c7 JV |
752 | { |
753 | struct in_device *in_dev; | |
a816c7c7 JV |
754 | |
755 | rcu_read_lock(); | |
5a37e8ca | 756 | in_dev = __in_dev_get_rcu(dev); |
866f3b25 | 757 | if (in_dev) |
24912420 | 758 | ip_mc_rejoin_groups(in_dev); |
a816c7c7 JV |
759 | rcu_read_unlock(); |
760 | } | |
761 | ||
5a37e8ca FL |
762 | /* |
763 | * Retrieve the list of registered multicast addresses for the bonding | |
764 | * device and retransmit an IGMP JOIN request to the current active | |
765 | * slave. | |
766 | */ | |
767 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
768 | { | |
1c3ac428 | 769 | struct net_device *bond_dev, *vlan_dev, *master_dev; |
5a37e8ca FL |
770 | struct vlan_entry *vlan; |
771 | ||
772 | read_lock(&bond->lock); | |
773 | ||
1c3ac428 PP |
774 | bond_dev = bond->dev; |
775 | ||
5a37e8ca | 776 | /* rejoin all groups on bond device */ |
1c3ac428 PP |
777 | __bond_resend_igmp_join_requests(bond_dev); |
778 | ||
779 | /* | |
780 | * if bond is enslaved to a bridge, | |
781 | * then rejoin all groups on its master | |
782 | */ | |
783 | master_dev = bond_dev->master; | |
784 | if (master_dev) | |
785 | if ((master_dev->priv_flags & IFF_EBRIDGE) | |
786 | && (bond_dev->priv_flags & IFF_BRIDGE_PORT)) | |
787 | __bond_resend_igmp_join_requests(master_dev); | |
5a37e8ca FL |
788 | |
789 | /* rejoin all groups on vlan devices */ | |
cc0e4070 JP |
790 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
791 | rcu_read_lock(); | |
1c3ac428 | 792 | vlan_dev = __vlan_find_dev_deep(bond_dev, |
cc0e4070 JP |
793 | vlan->vlan_id); |
794 | rcu_read_unlock(); | |
795 | if (vlan_dev) | |
796 | __bond_resend_igmp_join_requests(vlan_dev); | |
5a37e8ca FL |
797 | } |
798 | ||
e6d265e8 | 799 | if (--bond->igmp_retrans > 0) |
c2952c31 | 800 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
e6d265e8 | 801 | |
5a37e8ca FL |
802 | read_unlock(&bond->lock); |
803 | } | |
804 | ||
379b7383 | 805 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
806 | { |
807 | struct bonding *bond = container_of(work, struct bonding, | |
94265cf5 | 808 | mcast_work.work); |
5a37e8ca FL |
809 | bond_resend_igmp_join_requests(bond); |
810 | } | |
811 | ||
1da177e4 LT |
812 | /* |
813 | * flush all members of flush->mc_list from device dev->mc_list | |
814 | */ | |
3d632c3f SH |
815 | static void bond_mc_list_flush(struct net_device *bond_dev, |
816 | struct net_device *slave_dev) | |
1da177e4 | 817 | { |
454d7c9b | 818 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 819 | struct netdev_hw_addr *ha; |
1da177e4 | 820 | |
22bedad3 JP |
821 | netdev_for_each_mc_addr(ha, bond_dev) |
822 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
823 | |
824 | if (bond->params.mode == BOND_MODE_8023AD) { | |
825 | /* del lacpdu mc addr from mc list */ | |
826 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
827 | ||
22bedad3 | 828 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
829 | } |
830 | } | |
831 | ||
832 | /*--------------------------- Active slave change ---------------------------*/ | |
833 | ||
834 | /* | |
835 | * Update the mc list and multicast-related flags for the new and | |
836 | * old active slaves (if any) according to the multicast mode, and | |
837 | * promiscuous flags unconditionally. | |
838 | */ | |
3d632c3f SH |
839 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
840 | struct slave *old_active) | |
1da177e4 | 841 | { |
22bedad3 | 842 | struct netdev_hw_addr *ha; |
1da177e4 | 843 | |
3d632c3f | 844 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
845 | /* nothing to do - mc list is already up-to-date on |
846 | * all slaves | |
847 | */ | |
848 | return; | |
1da177e4 LT |
849 | |
850 | if (old_active) { | |
3d632c3f | 851 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 852 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 853 | |
3d632c3f | 854 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 855 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 856 | |
22bedad3 JP |
857 | netdev_for_each_mc_addr(ha, bond->dev) |
858 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
859 | } |
860 | ||
861 | if (new_active) { | |
7e1a1ac1 | 862 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 863 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 864 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 865 | |
3d632c3f | 866 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 867 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 868 | |
22bedad3 JP |
869 | netdev_for_each_mc_addr(ha, bond->dev) |
870 | dev_mc_add(new_active->dev, ha->addr); | |
1da177e4 LT |
871 | } |
872 | } | |
873 | ||
3915c1e8 JV |
874 | /* |
875 | * bond_do_fail_over_mac | |
876 | * | |
877 | * Perform special MAC address swapping for fail_over_mac settings | |
878 | * | |
879 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
880 | */ | |
881 | static void bond_do_fail_over_mac(struct bonding *bond, | |
882 | struct slave *new_active, | |
883 | struct slave *old_active) | |
1f78d9f9 HE |
884 | __releases(&bond->curr_slave_lock) |
885 | __releases(&bond->lock) | |
886 | __acquires(&bond->lock) | |
887 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
888 | { |
889 | u8 tmp_mac[ETH_ALEN]; | |
890 | struct sockaddr saddr; | |
891 | int rv; | |
892 | ||
893 | switch (bond->params.fail_over_mac) { | |
894 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 895 | if (new_active) { |
3915c1e8 JV |
896 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, |
897 | new_active->dev->addr_len); | |
7d26bb10 WP |
898 | write_unlock_bh(&bond->curr_slave_lock); |
899 | read_unlock(&bond->lock); | |
900 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); | |
901 | read_lock(&bond->lock); | |
902 | write_lock_bh(&bond->curr_slave_lock); | |
903 | } | |
3915c1e8 JV |
904 | break; |
905 | case BOND_FOM_FOLLOW: | |
906 | /* | |
907 | * if new_active && old_active, swap them | |
908 | * if just old_active, do nothing (going to no active slave) | |
909 | * if just new_active, set new_active to bond's MAC | |
910 | */ | |
911 | if (!new_active) | |
912 | return; | |
913 | ||
914 | write_unlock_bh(&bond->curr_slave_lock); | |
915 | read_unlock(&bond->lock); | |
916 | ||
917 | if (old_active) { | |
918 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
919 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
920 | ETH_ALEN); | |
921 | saddr.sa_family = new_active->dev->type; | |
922 | } else { | |
923 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
924 | saddr.sa_family = bond->dev->type; | |
925 | } | |
926 | ||
927 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
928 | if (rv) { | |
a4aee5c8 | 929 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
930 | bond->dev->name, -rv, new_active->dev->name); |
931 | goto out; | |
932 | } | |
933 | ||
934 | if (!old_active) | |
935 | goto out; | |
936 | ||
937 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
938 | saddr.sa_family = old_active->dev->type; | |
939 | ||
940 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
941 | if (rv) | |
a4aee5c8 | 942 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
943 | bond->dev->name, -rv, new_active->dev->name); |
944 | out: | |
945 | read_lock(&bond->lock); | |
946 | write_lock_bh(&bond->curr_slave_lock); | |
947 | break; | |
948 | default: | |
a4aee5c8 | 949 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
950 | bond->dev->name, bond->params.fail_over_mac); |
951 | break; | |
952 | } | |
953 | ||
954 | } | |
955 | ||
a549952a JP |
956 | static bool bond_should_change_active(struct bonding *bond) |
957 | { | |
958 | struct slave *prim = bond->primary_slave; | |
959 | struct slave *curr = bond->curr_active_slave; | |
960 | ||
961 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
962 | return true; | |
963 | if (bond->force_primary) { | |
964 | bond->force_primary = false; | |
965 | return true; | |
966 | } | |
967 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
968 | (prim->speed < curr->speed || | |
969 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
970 | return false; | |
971 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
972 | return false; | |
973 | return true; | |
974 | } | |
3915c1e8 | 975 | |
1da177e4 LT |
976 | /** |
977 | * find_best_interface - select the best available slave to be the active one | |
978 | * @bond: our bonding struct | |
979 | * | |
980 | * Warning: Caller must hold curr_slave_lock for writing. | |
981 | */ | |
982 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
983 | { | |
984 | struct slave *new_active, *old_active; | |
985 | struct slave *bestslave = NULL; | |
986 | int mintime = bond->params.updelay; | |
987 | int i; | |
988 | ||
49b4ad92 | 989 | new_active = bond->curr_active_slave; |
1da177e4 LT |
990 | |
991 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 992 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 993 | new_active = bond->first_slave; |
3d632c3f | 994 | else |
1da177e4 | 995 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
996 | } |
997 | ||
1da177e4 | 998 | if ((bond->primary_slave) && |
a549952a JP |
999 | bond->primary_slave->link == BOND_LINK_UP && |
1000 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1001 | new_active = bond->primary_slave; |
1002 | } | |
1003 | ||
1004 | /* remember where to stop iterating over the slaves */ | |
1005 | old_active = new_active; | |
1006 | ||
1007 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1008 | if (new_active->link == BOND_LINK_UP) { |
1009 | return new_active; | |
1010 | } else if (new_active->link == BOND_LINK_BACK && | |
1011 | IS_UP(new_active->dev)) { | |
1012 | /* link up, but waiting for stabilization */ | |
1013 | if (new_active->delay < mintime) { | |
1014 | mintime = new_active->delay; | |
1015 | bestslave = new_active; | |
1da177e4 LT |
1016 | } |
1017 | } | |
1018 | } | |
1019 | ||
1020 | return bestslave; | |
1021 | } | |
1022 | ||
ad246c99 BH |
1023 | static bool bond_should_notify_peers(struct bonding *bond) |
1024 | { | |
1025 | struct slave *slave = bond->curr_active_slave; | |
1026 | ||
1027 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
1028 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
1029 | ||
1030 | if (!slave || !bond->send_peer_notif || | |
1031 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
1032 | return false; | |
1033 | ||
1034 | bond->send_peer_notif--; | |
1035 | return true; | |
1036 | } | |
1037 | ||
1da177e4 LT |
1038 | /** |
1039 | * change_active_interface - change the active slave into the specified one | |
1040 | * @bond: our bonding struct | |
1041 | * @new: the new slave to make the active one | |
1042 | * | |
1043 | * Set the new slave to the bond's settings and unset them on the old | |
1044 | * curr_active_slave. | |
1045 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1046 | * | |
1047 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1048 | * because it is apparently the best available slave we have, even though its | |
1049 | * updelay hasn't timed out yet. | |
1050 | * | |
3915c1e8 JV |
1051 | * If new_active is not NULL, caller must hold bond->lock for read and |
1052 | * curr_slave_lock for write_bh. | |
1da177e4 | 1053 | */ |
a77b5325 | 1054 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1055 | { |
1056 | struct slave *old_active = bond->curr_active_slave; | |
1057 | ||
3d632c3f | 1058 | if (old_active == new_active) |
1da177e4 | 1059 | return; |
1da177e4 LT |
1060 | |
1061 | if (new_active) { | |
b2220cad JV |
1062 | new_active->jiffies = jiffies; |
1063 | ||
1da177e4 LT |
1064 | if (new_active->link == BOND_LINK_BACK) { |
1065 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1066 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1067 | bond->dev->name, new_active->dev->name, | |
1068 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1069 | } |
1070 | ||
1071 | new_active->delay = 0; | |
1072 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1073 | |
3d632c3f | 1074 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1075 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1076 | |
58402054 | 1077 | if (bond_is_lb(bond)) |
1da177e4 | 1078 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1079 | } else { |
1080 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1081 | pr_info("%s: making interface %s the new active one.\n", |
1082 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1083 | } |
1084 | } | |
1085 | } | |
1086 | ||
3d632c3f | 1087 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1088 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1089 | |
58402054 | 1090 | if (bond_is_lb(bond)) { |
1da177e4 | 1091 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1092 | if (old_active) |
1093 | bond_set_slave_inactive_flags(old_active); | |
1094 | if (new_active) | |
1095 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1096 | } else { |
1097 | bond->curr_active_slave = new_active; | |
1098 | } | |
c3ade5ca JV |
1099 | |
1100 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1101 | if (old_active) |
c3ade5ca | 1102 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1103 | |
1104 | if (new_active) { | |
ad246c99 BH |
1105 | bool should_notify_peers = false; |
1106 | ||
c3ade5ca | 1107 | bond_set_slave_active_flags(new_active); |
2ab82852 | 1108 | |
709f8a45 OG |
1109 | if (bond->params.fail_over_mac) |
1110 | bond_do_fail_over_mac(bond, new_active, | |
1111 | old_active); | |
3915c1e8 | 1112 | |
ad246c99 BH |
1113 | if (netif_running(bond->dev)) { |
1114 | bond->send_peer_notif = | |
1115 | bond->params.num_peer_notif; | |
1116 | should_notify_peers = | |
1117 | bond_should_notify_peers(bond); | |
1118 | } | |
1119 | ||
01f3109d OG |
1120 | write_unlock_bh(&bond->curr_slave_lock); |
1121 | read_unlock(&bond->lock); | |
1122 | ||
b7bc2a5b | 1123 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 1124 | if (should_notify_peers) |
b7bc2a5b AW |
1125 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
1126 | bond->dev); | |
01f3109d OG |
1127 | |
1128 | read_lock(&bond->lock); | |
1129 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1130 | } |
c3ade5ca | 1131 | } |
a2fd940f | 1132 | |
5a37e8ca | 1133 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
1134 | * all were sent on curr_active_slave. |
1135 | * resend only if bond is brought up with the affected | |
1136 | * bonding modes and the retransmission is enabled */ | |
1137 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
1138 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
1139 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 1140 | bond->igmp_retrans = bond->params.resend_igmp; |
5a37e8ca | 1141 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); |
a2fd940f | 1142 | } |
1da177e4 LT |
1143 | } |
1144 | ||
1145 | /** | |
1146 | * bond_select_active_slave - select a new active slave, if needed | |
1147 | * @bond: our bonding struct | |
1148 | * | |
3d632c3f | 1149 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1150 | * - The old curr_active_slave has been released or lost its link. |
1151 | * - The primary_slave has got its link back. | |
1152 | * - A slave has got its link back and there's no old curr_active_slave. | |
1153 | * | |
3915c1e8 | 1154 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1155 | */ |
a77b5325 | 1156 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1157 | { |
1158 | struct slave *best_slave; | |
ff59c456 | 1159 | int rv; |
1da177e4 LT |
1160 | |
1161 | best_slave = bond_find_best_slave(bond); | |
1162 | if (best_slave != bond->curr_active_slave) { | |
1163 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1164 | rv = bond_set_carrier(bond); |
1165 | if (!rv) | |
1166 | return; | |
1167 | ||
1168 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1169 | pr_info("%s: first active interface up!\n", |
1170 | bond->dev->name); | |
ff59c456 | 1171 | } else { |
a4aee5c8 JP |
1172 | pr_info("%s: now running without any active interface !\n", |
1173 | bond->dev->name); | |
ff59c456 | 1174 | } |
1da177e4 LT |
1175 | } |
1176 | } | |
1177 | ||
1178 | /*--------------------------- slave list handling ---------------------------*/ | |
1179 | ||
1180 | /* | |
1181 | * This function attaches the slave to the end of list. | |
1182 | * | |
1183 | * bond->lock held for writing by caller. | |
1184 | */ | |
1185 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1186 | { | |
1187 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1188 | new_slave->next = new_slave; | |
1189 | new_slave->prev = new_slave; | |
1190 | bond->first_slave = new_slave; | |
1191 | } else { | |
1192 | new_slave->next = bond->first_slave; | |
1193 | new_slave->prev = bond->first_slave->prev; | |
1194 | new_slave->next->prev = new_slave; | |
1195 | new_slave->prev->next = new_slave; | |
1196 | } | |
1197 | ||
1198 | bond->slave_cnt++; | |
1199 | } | |
1200 | ||
1201 | /* | |
1202 | * This function detaches the slave from the list. | |
1203 | * WARNING: no check is made to verify if the slave effectively | |
1204 | * belongs to <bond>. | |
1205 | * Nothing is freed on return, structures are just unchained. | |
1206 | * If any slave pointer in bond was pointing to <slave>, | |
1207 | * it should be changed by the calling function. | |
1208 | * | |
1209 | * bond->lock held for writing by caller. | |
1210 | */ | |
1211 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1212 | { | |
3d632c3f | 1213 | if (slave->next) |
1da177e4 | 1214 | slave->next->prev = slave->prev; |
1da177e4 | 1215 | |
3d632c3f | 1216 | if (slave->prev) |
1da177e4 | 1217 | slave->prev->next = slave->next; |
1da177e4 LT |
1218 | |
1219 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1220 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1221 | bond->first_slave = slave->next; | |
1222 | } else { | |
1223 | bond->first_slave = NULL; /* slave was the last one */ | |
1224 | } | |
1225 | } | |
1226 | ||
1227 | slave->next = NULL; | |
1228 | slave->prev = NULL; | |
1229 | bond->slave_cnt--; | |
1230 | } | |
1231 | ||
f6dc31a8 | 1232 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1233 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1234 | { |
8a8efa22 AW |
1235 | struct netpoll *np; |
1236 | int err = 0; | |
f6dc31a8 | 1237 | |
47be03a2 | 1238 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
1239 | err = -ENOMEM; |
1240 | if (!np) | |
1241 | goto out; | |
1242 | ||
47be03a2 | 1243 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
1244 | if (err) { |
1245 | kfree(np); | |
1246 | goto out; | |
f6dc31a8 | 1247 | } |
8a8efa22 AW |
1248 | slave->np = np; |
1249 | out: | |
1250 | return err; | |
1251 | } | |
1252 | static inline void slave_disable_netpoll(struct slave *slave) | |
1253 | { | |
1254 | struct netpoll *np = slave->np; | |
1255 | ||
1256 | if (!np) | |
1257 | return; | |
1258 | ||
1259 | slave->np = NULL; | |
38e6bc18 | 1260 | __netpoll_free_rcu(np); |
8a8efa22 AW |
1261 | } |
1262 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1263 | { | |
1264 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1265 | return false; | |
1266 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1267 | return false; | |
1268 | return true; | |
f6dc31a8 WC |
1269 | } |
1270 | ||
1271 | static void bond_poll_controller(struct net_device *bond_dev) | |
1272 | { | |
8a8efa22 AW |
1273 | } |
1274 | ||
1275 | static void __bond_netpoll_cleanup(struct bonding *bond) | |
1276 | { | |
c2355e1a NH |
1277 | struct slave *slave; |
1278 | int i; | |
1279 | ||
8a8efa22 AW |
1280 | bond_for_each_slave(bond, slave, i) |
1281 | if (IS_UP(slave->dev)) | |
1282 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1283 | } |
f6dc31a8 WC |
1284 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1285 | { | |
1286 | struct bonding *bond = netdev_priv(bond_dev); | |
8a8efa22 AW |
1287 | |
1288 | read_lock(&bond->lock); | |
1289 | __bond_netpoll_cleanup(bond); | |
1290 | read_unlock(&bond->lock); | |
1291 | } | |
1292 | ||
47be03a2 | 1293 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1294 | { |
1295 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1296 | struct slave *slave; |
8a8efa22 | 1297 | int i, err = 0; |
f6dc31a8 WC |
1298 | |
1299 | read_lock(&bond->lock); | |
f6dc31a8 | 1300 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1301 | err = slave_enable_netpoll(slave); |
1302 | if (err) { | |
1303 | __bond_netpoll_cleanup(bond); | |
1304 | break; | |
f6dc31a8 WC |
1305 | } |
1306 | } | |
1307 | read_unlock(&bond->lock); | |
8a8efa22 | 1308 | return err; |
f6dc31a8 WC |
1309 | } |
1310 | ||
8a8efa22 AW |
1311 | static struct netpoll_info *bond_netpoll_info(struct bonding *bond) |
1312 | { | |
1313 | return bond->dev->npinfo; | |
1314 | } | |
f6dc31a8 | 1315 | |
8a8efa22 AW |
1316 | #else |
1317 | static inline int slave_enable_netpoll(struct slave *slave) | |
1318 | { | |
1319 | return 0; | |
1320 | } | |
1321 | static inline void slave_disable_netpoll(struct slave *slave) | |
1322 | { | |
1323 | } | |
f6dc31a8 WC |
1324 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1325 | { | |
1326 | } | |
f6dc31a8 WC |
1327 | #endif |
1328 | ||
1da177e4 LT |
1329 | /*---------------------------------- IOCTL ----------------------------------*/ |
1330 | ||
4ad072c9 AB |
1331 | static int bond_sethwaddr(struct net_device *bond_dev, |
1332 | struct net_device *slave_dev) | |
1da177e4 | 1333 | { |
5a03cdb7 HE |
1334 | pr_debug("bond_dev=%p\n", bond_dev); |
1335 | pr_debug("slave_dev=%p\n", slave_dev); | |
1336 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1337 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1338 | return 0; | |
1339 | } | |
1340 | ||
c8f44aff MM |
1341 | static netdev_features_t bond_fix_features(struct net_device *dev, |
1342 | netdev_features_t features) | |
8531c5ff | 1343 | { |
8531c5ff | 1344 | struct slave *slave; |
b2a103e6 | 1345 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1346 | netdev_features_t mask; |
8e3babcd | 1347 | int i; |
8531c5ff | 1348 | |
b2a103e6 | 1349 | read_lock(&bond->lock); |
b63365a2 | 1350 | |
b2a103e6 MM |
1351 | if (!bond->first_slave) { |
1352 | /* Disable adding VLANs to empty bond. But why? --mq */ | |
1353 | features |= NETIF_F_VLAN_CHALLENGED; | |
1354 | goto out; | |
1355 | } | |
b63365a2 | 1356 | |
b2a103e6 | 1357 | mask = features; |
b63365a2 | 1358 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1359 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1360 | |
54ef3137 | 1361 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1362 | features = netdev_increment_features(features, |
1363 | slave->dev->features, | |
b2a103e6 MM |
1364 | mask); |
1365 | } | |
1366 | ||
1367 | out: | |
1368 | read_unlock(&bond->lock); | |
1369 | return features; | |
1370 | } | |
1371 | ||
62f2a3a4 MM |
1372 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1373 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1374 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1375 | |
1376 | static void bond_compute_features(struct bonding *bond) | |
1377 | { | |
1378 | struct slave *slave; | |
1379 | struct net_device *bond_dev = bond->dev; | |
c8f44aff | 1380 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1381 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1382 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1383 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1384 | int i; |
b6fe83e9 | 1385 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
b2a103e6 MM |
1386 | |
1387 | read_lock(&bond->lock); | |
1388 | ||
1389 | if (!bond->first_slave) | |
1390 | goto done; | |
1391 | ||
1392 | bond_for_each_slave(bond, slave, i) { | |
278339a4 | 1393 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1394 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1395 | ||
b6fe83e9 | 1396 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1397 | if (slave->dev->hard_header_len > max_hard_header_len) |
1398 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1399 | |
1400 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1401 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1402 | } |
8531c5ff | 1403 | |
b63365a2 | 1404 | done: |
b2a103e6 | 1405 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1406 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1407 | bond_dev->gso_max_segs = gso_max_segs; |
1408 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1409 | |
b6fe83e9 ED |
1410 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1411 | bond_dev->priv_flags = flags | dst_release_flag; | |
1412 | ||
b2a103e6 MM |
1413 | read_unlock(&bond->lock); |
1414 | ||
1415 | netdev_change_features(bond_dev); | |
8531c5ff AK |
1416 | } |
1417 | ||
872254dd MS |
1418 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1419 | struct net_device *slave_dev) | |
1420 | { | |
454d7c9b | 1421 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1422 | |
00829823 | 1423 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1424 | |
1425 | bond_dev->type = slave_dev->type; | |
1426 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1427 | bond_dev->addr_len = slave_dev->addr_len; | |
1428 | ||
1429 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1430 | slave_dev->addr_len); | |
d90a162a | 1431 | bond->setup_by_slave = 1; |
872254dd MS |
1432 | } |
1433 | ||
5b2c4dd2 | 1434 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1435 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1436 | */ |
1437 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1438 | struct slave *slave, |
1439 | struct bonding *bond) | |
5b2c4dd2 | 1440 | { |
2d7011ca | 1441 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1442 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1443 | skb->pkt_type != PACKET_BROADCAST && |
1444 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1445 | return false; |
5b2c4dd2 JP |
1446 | return true; |
1447 | } | |
1448 | return false; | |
1449 | } | |
1450 | ||
8a4eb573 | 1451 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1452 | { |
8a4eb573 | 1453 | struct sk_buff *skb = *pskb; |
f1c1775a | 1454 | struct slave *slave; |
0bd80dad | 1455 | struct bonding *bond; |
de063b70 ED |
1456 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1457 | struct slave *); | |
13a8e0c8 | 1458 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1459 | |
8a4eb573 JP |
1460 | skb = skb_share_check(skb, GFP_ATOMIC); |
1461 | if (unlikely(!skb)) | |
1462 | return RX_HANDLER_CONSUMED; | |
1463 | ||
1464 | *pskb = skb; | |
5b2c4dd2 | 1465 | |
35d48903 JP |
1466 | slave = bond_slave_get_rcu(skb->dev); |
1467 | bond = slave->bond; | |
0bd80dad JP |
1468 | |
1469 | if (bond->params.arp_interval) | |
f1c1775a | 1470 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1471 | |
4d97480b MH |
1472 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1473 | if (recv_probe) { | |
de063b70 ED |
1474 | ret = recv_probe(skb, bond, slave); |
1475 | if (ret == RX_HANDLER_CONSUMED) { | |
1476 | consume_skb(skb); | |
1477 | return ret; | |
3aba891d JP |
1478 | } |
1479 | } | |
1480 | ||
0bd80dad | 1481 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1482 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1483 | } |
1484 | ||
35d48903 | 1485 | skb->dev = bond->dev; |
5b2c4dd2 | 1486 | |
0bd80dad | 1487 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1488 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1489 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1490 | |
541ac7c9 CG |
1491 | if (unlikely(skb_cow_head(skb, |
1492 | skb->data - skb_mac_header(skb)))) { | |
1493 | kfree_skb(skb); | |
8a4eb573 | 1494 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1495 | } |
35d48903 | 1496 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1497 | } |
1498 | ||
13a8e0c8 | 1499 | return ret; |
5b2c4dd2 JP |
1500 | } |
1501 | ||
1da177e4 | 1502 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1503 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1504 | { |
454d7c9b | 1505 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1506 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1507 | struct slave *new_slave = NULL; |
22bedad3 | 1508 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1509 | struct sockaddr addr; |
1510 | int link_reporting; | |
1da177e4 LT |
1511 | int res = 0; |
1512 | ||
552709d5 | 1513 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1514 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1515 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1516 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1517 | } |
1518 | ||
1da177e4 LT |
1519 | /* already enslaved */ |
1520 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1521 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1522 | return -EBUSY; |
1523 | } | |
1524 | ||
1525 | /* vlan challenged mutual exclusion */ | |
1526 | /* no need to lock since we're protected by rtnl_lock */ | |
1527 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1528 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1529 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1530 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1531 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1532 | return -EPERM; |
1533 | } else { | |
a4aee5c8 JP |
1534 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1535 | bond_dev->name, slave_dev->name, | |
1536 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1537 | } |
1538 | } else { | |
5a03cdb7 | 1539 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1540 | } |
1541 | ||
217df670 JV |
1542 | /* |
1543 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1544 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1545 | * the current ifenslave will set the interface down prior to |
1546 | * enslaving it; the old ifenslave will not. | |
1547 | */ | |
1548 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1549 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1550 | slave_dev->name); |
1551 | res = -EPERM; | |
1552 | goto err_undo_flags; | |
1553 | } | |
1da177e4 | 1554 | |
872254dd MS |
1555 | /* set bonding device ether type by slave - bonding netdevices are |
1556 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1557 | * there is a need to override some of the type dependent attribs/funcs. | |
1558 | * | |
1559 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1560 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1561 | */ | |
1562 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1563 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1564 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1565 | bond_dev->name, |
1566 | bond_dev->type, slave_dev->type); | |
75c78500 | 1567 | |
b7bc2a5b AW |
1568 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1569 | bond_dev); | |
3ca5b404 JP |
1570 | res = notifier_to_errno(res); |
1571 | if (res) { | |
1572 | pr_err("%s: refused to change device type\n", | |
1573 | bond_dev->name); | |
1574 | res = -EBUSY; | |
1575 | goto err_undo_flags; | |
1576 | } | |
75c78500 | 1577 | |
32a806c1 | 1578 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1579 | dev_uc_flush(bond_dev); |
22bedad3 | 1580 | dev_mc_flush(bond_dev); |
32a806c1 | 1581 | |
e36b9d16 MS |
1582 | if (slave_dev->type != ARPHRD_ETHER) |
1583 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1584 | else { |
e36b9d16 | 1585 | ether_setup(bond_dev); |
550fd08c NH |
1586 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1587 | } | |
75c78500 | 1588 | |
b7bc2a5b AW |
1589 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1590 | bond_dev); | |
e36b9d16 | 1591 | } |
872254dd | 1592 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1593 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1594 | slave_dev->name, | |
1595 | slave_dev->type, bond_dev->type); | |
1596 | res = -EINVAL; | |
1597 | goto err_undo_flags; | |
872254dd MS |
1598 | } |
1599 | ||
eb7cc59a | 1600 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1601 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1602 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1603 | bond_dev->name); | |
3915c1e8 JV |
1604 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1605 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1606 | 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", |
1607 | bond_dev->name); | |
2ab82852 MS |
1608 | res = -EOPNOTSUPP; |
1609 | goto err_undo_flags; | |
1610 | } | |
1da177e4 LT |
1611 | } |
1612 | ||
8d8fc29d AW |
1613 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1614 | ||
c20811a7 JP |
1615 | /* If this is the first slave, then we need to set the master's hardware |
1616 | * address to be the same as the slave's. */ | |
d13a2cb6 | 1617 | if (is_zero_ether_addr(bond->dev->dev_addr)) |
c20811a7 JP |
1618 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, |
1619 | slave_dev->addr_len); | |
1620 | ||
1621 | ||
243cb4e5 | 1622 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1623 | if (!new_slave) { |
1624 | res = -ENOMEM; | |
1625 | goto err_undo_flags; | |
1626 | } | |
1627 | ||
bb1d9123 AG |
1628 | /* |
1629 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1630 | * is set via sysfs or module option if desired. | |
1631 | */ | |
1632 | new_slave->queue_id = 0; | |
1633 | ||
b15ba0fb JP |
1634 | /* Save slave's original mtu and then set it to match the bond */ |
1635 | new_slave->original_mtu = slave_dev->mtu; | |
1636 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1637 | if (res) { | |
1638 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1639 | goto err_free; | |
1640 | } | |
1641 | ||
217df670 JV |
1642 | /* |
1643 | * Save slave's original ("permanent") mac address for modes | |
1644 | * that need it, and for restoring it upon release, and then | |
1645 | * set it to the master's address | |
1646 | */ | |
1647 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1648 | |
dd957c57 | 1649 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1650 | /* |
1651 | * Set slave to master's mac address. The application already | |
1652 | * set the master's mac address to that of the first slave | |
1653 | */ | |
1654 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1655 | addr.sa_family = slave_dev->type; | |
1656 | res = dev_set_mac_address(slave_dev, &addr); | |
1657 | if (res) { | |
5a03cdb7 | 1658 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1659 | goto err_restore_mtu; |
2ab82852 | 1660 | } |
217df670 | 1661 | } |
1da177e4 | 1662 | |
1765a575 | 1663 | res = netdev_set_bond_master(slave_dev, bond_dev); |
c2edacf8 | 1664 | if (res) { |
1765a575 | 1665 | pr_debug("Error %d calling netdev_set_bond_master\n", res); |
569f0c4d | 1666 | goto err_restore_mac; |
c2edacf8 | 1667 | } |
5b2c4dd2 | 1668 | |
217df670 JV |
1669 | /* open the slave since the application closed it */ |
1670 | res = dev_open(slave_dev); | |
1671 | if (res) { | |
3d632c3f | 1672 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1673 | goto err_unset_master; |
1da177e4 LT |
1674 | } |
1675 | ||
35d48903 | 1676 | new_slave->bond = bond; |
1da177e4 | 1677 | new_slave->dev = slave_dev; |
0b680e75 | 1678 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1679 | |
58402054 | 1680 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1681 | /* bond_alb_init_slave() must be called before all other stages since |
1682 | * it might fail and we do not want to have to undo everything | |
1683 | */ | |
1684 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1685 | if (res) |
569f0c4d | 1686 | goto err_close; |
1da177e4 LT |
1687 | } |
1688 | ||
1689 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1690 | * master's promisc (and mc) settings only if it becomes the | |
1691 | * curr_active_slave, and that is taken care of later when calling | |
1692 | * bond_change_active() | |
1693 | */ | |
1694 | if (!USES_PRIMARY(bond->params.mode)) { | |
1695 | /* set promiscuity level to new slave */ | |
1696 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1697 | res = dev_set_promiscuity(slave_dev, 1); |
1698 | if (res) | |
1699 | goto err_close; | |
1da177e4 LT |
1700 | } |
1701 | ||
1702 | /* set allmulti level to new slave */ | |
1703 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1704 | res = dev_set_allmulti(slave_dev, 1); |
1705 | if (res) | |
1706 | goto err_close; | |
1da177e4 LT |
1707 | } |
1708 | ||
b9e40857 | 1709 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1710 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1711 | netdev_for_each_mc_addr(ha, bond_dev) |
1712 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1713 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1714 | } |
1715 | ||
1716 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1717 | /* add lacpdu mc addr to mc list */ | |
1718 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1719 | ||
22bedad3 | 1720 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1721 | } |
1722 | ||
1723 | bond_add_vlans_on_slave(bond, slave_dev); | |
1724 | ||
1725 | write_lock_bh(&bond->lock); | |
1726 | ||
1727 | bond_attach_slave(bond, new_slave); | |
1728 | ||
1729 | new_slave->delay = 0; | |
1730 | new_slave->link_failure_count = 0; | |
1731 | ||
3915c1e8 JV |
1732 | write_unlock_bh(&bond->lock); |
1733 | ||
b2a103e6 MM |
1734 | bond_compute_features(bond); |
1735 | ||
3915c1e8 JV |
1736 | read_lock(&bond->lock); |
1737 | ||
f31c7937 MK |
1738 | new_slave->last_arp_rx = jiffies - |
1739 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
f5b2b966 | 1740 | |
1da177e4 LT |
1741 | if (bond->params.miimon && !bond->params.use_carrier) { |
1742 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1743 | ||
1744 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1745 | /* | |
1746 | * miimon is set but a bonded network driver | |
1747 | * does not support ETHTOOL/MII and | |
1748 | * arp_interval is not set. Note: if | |
1749 | * use_carrier is enabled, we will never go | |
1750 | * here (because netif_carrier is always | |
1751 | * supported); thus, we don't need to change | |
1752 | * the messages for netif_carrier. | |
1753 | */ | |
a4aee5c8 | 1754 | 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 | 1755 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1756 | } else if (link_reporting == -1) { |
1757 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1758 | 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", |
1759 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1760 | } |
1761 | } | |
1762 | ||
1763 | /* check for initial state */ | |
f31c7937 MK |
1764 | if (bond->params.miimon) { |
1765 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1766 | if (bond->params.updelay) { | |
1767 | new_slave->link = BOND_LINK_BACK; | |
1768 | new_slave->delay = bond->params.updelay; | |
1769 | } else { | |
1770 | new_slave->link = BOND_LINK_UP; | |
1771 | } | |
1da177e4 | 1772 | } else { |
f31c7937 | 1773 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1774 | } |
f31c7937 MK |
1775 | } else if (bond->params.arp_interval) { |
1776 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1777 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1778 | } else { |
f31c7937 | 1779 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1780 | } |
1781 | ||
f31c7937 MK |
1782 | if (new_slave->link != BOND_LINK_DOWN) |
1783 | new_slave->jiffies = jiffies; | |
1784 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1785 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1786 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1787 | ||
98f41f69 | 1788 | bond_update_speed_duplex(new_slave); |
1da177e4 LT |
1789 | |
1790 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1791 | /* if there is a primary slave, remember it */ | |
a549952a | 1792 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1793 | bond->primary_slave = new_slave; |
a549952a JP |
1794 | bond->force_primary = true; |
1795 | } | |
1da177e4 LT |
1796 | } |
1797 | ||
3915c1e8 JV |
1798 | write_lock_bh(&bond->curr_slave_lock); |
1799 | ||
1da177e4 LT |
1800 | switch (bond->params.mode) { |
1801 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1802 | bond_set_slave_inactive_flags(new_slave); |
1803 | bond_select_active_slave(bond); | |
1da177e4 LT |
1804 | break; |
1805 | case BOND_MODE_8023AD: | |
1806 | /* in 802.3ad mode, the internal mechanism | |
1807 | * will activate the slaves in the selected | |
1808 | * aggregator | |
1809 | */ | |
1810 | bond_set_slave_inactive_flags(new_slave); | |
1811 | /* if this is the first slave */ | |
1812 | if (bond->slave_cnt == 1) { | |
1813 | SLAVE_AD_INFO(new_slave).id = 1; | |
1814 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1815 | * can be called only after the mac address of the bond is set | |
1816 | */ | |
56d00c67 | 1817 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 LT |
1818 | } else { |
1819 | SLAVE_AD_INFO(new_slave).id = | |
1820 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1821 | } | |
1822 | ||
1823 | bond_3ad_bind_slave(new_slave); | |
1824 | break; | |
1825 | case BOND_MODE_TLB: | |
1826 | case BOND_MODE_ALB: | |
e30bc066 | 1827 | bond_set_active_slave(new_slave); |
059fe7a5 | 1828 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1829 | bond_select_active_slave(bond); |
1da177e4 LT |
1830 | break; |
1831 | default: | |
5a03cdb7 | 1832 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1833 | |
1834 | /* always active in trunk mode */ | |
e30bc066 | 1835 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1836 | |
1837 | /* In trunking mode there is little meaning to curr_active_slave | |
1838 | * anyway (it holds no special properties of the bond device), | |
1839 | * so we can change it without calling change_active_interface() | |
1840 | */ | |
4e591b93 | 1841 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
1da177e4 | 1842 | bond->curr_active_slave = new_slave; |
3d632c3f | 1843 | |
1da177e4 LT |
1844 | break; |
1845 | } /* switch(bond_mode) */ | |
1846 | ||
3915c1e8 JV |
1847 | write_unlock_bh(&bond->curr_slave_lock); |
1848 | ||
ff59c456 JV |
1849 | bond_set_carrier(bond); |
1850 | ||
f6dc31a8 | 1851 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 AW |
1852 | slave_dev->npinfo = bond_netpoll_info(bond); |
1853 | if (slave_dev->npinfo) { | |
1854 | if (slave_enable_netpoll(new_slave)) { | |
1855 | read_unlock(&bond->lock); | |
1856 | pr_info("Error, %s: master_dev is using netpoll, " | |
1857 | "but new slave device does not support netpoll.\n", | |
1858 | bond_dev->name); | |
1859 | res = -EBUSY; | |
f7d9821a | 1860 | goto err_detach; |
8a8efa22 | 1861 | } |
f6dc31a8 WC |
1862 | } |
1863 | #endif | |
8a8efa22 | 1864 | |
3915c1e8 | 1865 | read_unlock(&bond->lock); |
1da177e4 | 1866 | |
b76cdba9 MW |
1867 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1868 | if (res) | |
f7d9821a | 1869 | goto err_detach; |
b76cdba9 | 1870 | |
35d48903 JP |
1871 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1872 | new_slave); | |
1873 | if (res) { | |
1874 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1875 | goto err_dest_symlinks; | |
1876 | } | |
1877 | ||
a4aee5c8 JP |
1878 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1879 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1880 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1881 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1882 | |
1883 | /* enslave is successful */ | |
1884 | return 0; | |
1885 | ||
1886 | /* Undo stages on error */ | |
35d48903 JP |
1887 | err_dest_symlinks: |
1888 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1889 | ||
f7d9821a | 1890 | err_detach: |
1891 | write_lock_bh(&bond->lock); | |
1892 | bond_detach_slave(bond, new_slave); | |
1893 | write_unlock_bh(&bond->lock); | |
1894 | ||
1da177e4 LT |
1895 | err_close: |
1896 | dev_close(slave_dev); | |
1897 | ||
569f0c4d | 1898 | err_unset_master: |
1765a575 | 1899 | netdev_set_bond_master(slave_dev, NULL); |
569f0c4d | 1900 | |
1da177e4 | 1901 | err_restore_mac: |
dd957c57 | 1902 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1903 | /* XXX TODO - fom follow mode needs to change master's |
1904 | * MAC if this slave's MAC is in use by the bond, or at | |
1905 | * least print a warning. | |
1906 | */ | |
2ab82852 MS |
1907 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1908 | addr.sa_family = slave_dev->type; | |
1909 | dev_set_mac_address(slave_dev, &addr); | |
1910 | } | |
1da177e4 | 1911 | |
b15ba0fb JP |
1912 | err_restore_mtu: |
1913 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1914 | ||
1da177e4 LT |
1915 | err_free: |
1916 | kfree(new_slave); | |
1917 | ||
1918 | err_undo_flags: | |
b2a103e6 | 1919 | bond_compute_features(bond); |
3d632c3f | 1920 | |
1da177e4 LT |
1921 | return res; |
1922 | } | |
1923 | ||
1924 | /* | |
1925 | * Try to release the slave device <slave> from the bond device <master> | |
1926 | * It is legal to access curr_active_slave without a lock because all the function | |
1927 | * is write-locked. | |
1928 | * | |
1929 | * The rules for slave state should be: | |
1930 | * for Active/Backup: | |
1931 | * Active stays on all backups go down | |
1932 | * for Bonded connections: | |
1933 | * The first up interface should be left on and all others downed. | |
1934 | */ | |
a77b5325 | 1935 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1936 | { |
454d7c9b | 1937 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1938 | struct slave *slave, *oldcurrent; |
1939 | struct sockaddr addr; | |
c8f44aff | 1940 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1941 | |
1942 | /* slave is not a slave or master is not master of this slave */ | |
1943 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1944 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1945 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1946 | bond_dev->name, slave_dev->name); |
1947 | return -EINVAL; | |
1948 | } | |
1949 | ||
e843fa50 | 1950 | block_netpoll_tx(); |
b7bc2a5b | 1951 | call_netdevice_notifiers(NETDEV_RELEASE, bond_dev); |
1da177e4 LT |
1952 | write_lock_bh(&bond->lock); |
1953 | ||
1954 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1955 | if (!slave) { | |
1956 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1957 | pr_info("%s: %s not enslaved\n", |
1958 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1959 | write_unlock_bh(&bond->lock); |
e843fa50 | 1960 | unblock_netpoll_tx(); |
1da177e4 LT |
1961 | return -EINVAL; |
1962 | } | |
1963 | ||
35d48903 JP |
1964 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1965 | * for this slave anymore. | |
1966 | */ | |
1967 | netdev_rx_handler_unregister(slave_dev); | |
1968 | write_unlock_bh(&bond->lock); | |
1969 | synchronize_net(); | |
1970 | write_lock_bh(&bond->lock); | |
1971 | ||
3915c1e8 | 1972 | if (!bond->params.fail_over_mac) { |
2e42e474 | 1973 | if (ether_addr_equal(bond_dev->dev_addr, slave->perm_hwaddr) && |
8e95a202 | 1974 | bond->slave_cnt > 1) |
a4aee5c8 JP |
1975 | 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", |
1976 | bond_dev->name, slave_dev->name, | |
1977 | slave->perm_hwaddr, | |
1978 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1979 | } |
1980 | ||
1981 | /* Inform AD package of unbinding of slave. */ | |
1982 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1983 | /* must be called before the slave is | |
1984 | * detached from the list | |
1985 | */ | |
1986 | bond_3ad_unbind_slave(slave); | |
1987 | } | |
1988 | ||
a4aee5c8 JP |
1989 | pr_info("%s: releasing %s interface %s\n", |
1990 | bond_dev->name, | |
e30bc066 | 1991 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 1992 | slave_dev->name); |
1da177e4 LT |
1993 | |
1994 | oldcurrent = bond->curr_active_slave; | |
1995 | ||
1996 | bond->current_arp_slave = NULL; | |
1997 | ||
1998 | /* release the slave from its bond */ | |
1999 | bond_detach_slave(bond, slave); | |
2000 | ||
3d632c3f | 2001 | if (bond->primary_slave == slave) |
1da177e4 | 2002 | bond->primary_slave = NULL; |
1da177e4 | 2003 | |
3d632c3f | 2004 | if (oldcurrent == slave) |
1da177e4 | 2005 | bond_change_active_slave(bond, NULL); |
1da177e4 | 2006 | |
58402054 | 2007 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2008 | /* Must be called only after the slave has been |
2009 | * detached from the list and the curr_active_slave | |
2010 | * has been cleared (if our_slave == old_current), | |
2011 | * but before a new active slave is selected. | |
2012 | */ | |
2543331d | 2013 | write_unlock_bh(&bond->lock); |
1da177e4 | 2014 | bond_alb_deinit_slave(bond, slave); |
2543331d | 2015 | write_lock_bh(&bond->lock); |
1da177e4 LT |
2016 | } |
2017 | ||
059fe7a5 JV |
2018 | if (oldcurrent == slave) { |
2019 | /* | |
2020 | * Note that we hold RTNL over this sequence, so there | |
2021 | * is no concern that another slave add/remove event | |
2022 | * will interfere. | |
2023 | */ | |
2024 | write_unlock_bh(&bond->lock); | |
2025 | read_lock(&bond->lock); | |
2026 | write_lock_bh(&bond->curr_slave_lock); | |
2027 | ||
1da177e4 LT |
2028 | bond_select_active_slave(bond); |
2029 | ||
059fe7a5 JV |
2030 | write_unlock_bh(&bond->curr_slave_lock); |
2031 | read_unlock(&bond->lock); | |
2032 | write_lock_bh(&bond->lock); | |
2033 | } | |
2034 | ||
1da177e4 | 2035 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
2036 | bond_set_carrier(bond); |
2037 | ||
1da177e4 LT |
2038 | /* if the last slave was removed, zero the mac address |
2039 | * of the master so it will be set by the application | |
2040 | * to the mac address of the first slave | |
2041 | */ | |
2042 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2043 | ||
cc0e4070 | 2044 | if (bond_vlan_used(bond)) { |
a4aee5c8 JP |
2045 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2046 | bond_dev->name, bond_dev->name); | |
2047 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2048 | bond_dev->name); | |
1da177e4 | 2049 | } |
1da177e4 LT |
2050 | } |
2051 | ||
2052 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2053 | unblock_netpoll_tx(); |
1da177e4 | 2054 | |
2af73d4b SP |
2055 | if (bond->slave_cnt == 0) |
2056 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); | |
2057 | ||
b2a103e6 MM |
2058 | bond_compute_features(bond); |
2059 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2060 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
2061 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
2062 | bond_dev->name, slave_dev->name, bond_dev->name); | |
2063 | ||
b76cdba9 MW |
2064 | /* must do this from outside any spinlocks */ |
2065 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2066 | ||
1da177e4 LT |
2067 | bond_del_vlans_from_slave(bond, slave_dev); |
2068 | ||
2069 | /* If the mode USES_PRIMARY, then we should only remove its | |
2070 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2071 | * already taken care of above when we detached the slave | |
2072 | */ | |
2073 | if (!USES_PRIMARY(bond->params.mode)) { | |
2074 | /* unset promiscuity level from slave */ | |
3d632c3f | 2075 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2076 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2077 | |
2078 | /* unset allmulti level from slave */ | |
3d632c3f | 2079 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2080 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2081 | |
2082 | /* flush master's mc_list from slave */ | |
b9e40857 | 2083 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2084 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2085 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2086 | } |
2087 | ||
1765a575 | 2088 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2089 | |
8a8efa22 | 2090 | slave_disable_netpoll(slave); |
f6dc31a8 | 2091 | |
1da177e4 LT |
2092 | /* close slave before restoring its mac address */ |
2093 | dev_close(slave_dev); | |
2094 | ||
3915c1e8 | 2095 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2096 | /* restore original ("permanent") mac address */ |
2097 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2098 | addr.sa_family = slave_dev->type; | |
2099 | dev_set_mac_address(slave_dev, &addr); | |
2100 | } | |
1da177e4 | 2101 | |
b15ba0fb JP |
2102 | dev_set_mtu(slave_dev, slave->original_mtu); |
2103 | ||
2d7011ca | 2104 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2105 | |
2106 | kfree(slave); | |
2107 | ||
2108 | return 0; /* deletion OK */ | |
2109 | } | |
2110 | ||
d90a162a | 2111 | /* |
dadaa10b | 2112 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2113 | * Must be under rtnl_lock when this function is called. |
2114 | */ | |
26d8ee75 | 2115 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2116 | struct net_device *slave_dev) | |
d90a162a | 2117 | { |
454d7c9b | 2118 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2119 | int ret; |
2120 | ||
2121 | ret = bond_release(bond_dev, slave_dev); | |
2122 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2123 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2124 | pr_info("%s: destroying bond %s.\n", |
2125 | bond_dev->name, bond_dev->name); | |
9e71626c | 2126 | unregister_netdevice(bond_dev); |
d90a162a MS |
2127 | } |
2128 | return ret; | |
2129 | } | |
2130 | ||
1da177e4 LT |
2131 | /* |
2132 | * This function releases all slaves. | |
2133 | */ | |
2134 | static int bond_release_all(struct net_device *bond_dev) | |
2135 | { | |
454d7c9b | 2136 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2137 | struct slave *slave; |
2138 | struct net_device *slave_dev; | |
2139 | struct sockaddr addr; | |
2140 | ||
2141 | write_lock_bh(&bond->lock); | |
2142 | ||
ff59c456 JV |
2143 | netif_carrier_off(bond_dev); |
2144 | ||
3d632c3f | 2145 | if (bond->slave_cnt == 0) |
1da177e4 | 2146 | goto out; |
1da177e4 LT |
2147 | |
2148 | bond->current_arp_slave = NULL; | |
2149 | bond->primary_slave = NULL; | |
2150 | bond_change_active_slave(bond, NULL); | |
2151 | ||
2152 | while ((slave = bond->first_slave) != NULL) { | |
2153 | /* Inform AD package of unbinding of slave | |
2154 | * before slave is detached from the list. | |
2155 | */ | |
3d632c3f | 2156 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2157 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2158 | |
2159 | slave_dev = slave->dev; | |
2160 | bond_detach_slave(bond, slave); | |
2161 | ||
2543331d JV |
2162 | /* now that the slave is detached, unlock and perform |
2163 | * all the undo steps that should not be called from | |
2164 | * within a lock. | |
2165 | */ | |
2166 | write_unlock_bh(&bond->lock); | |
2167 | ||
35d48903 JP |
2168 | /* unregister rx_handler early so bond_handle_frame wouldn't |
2169 | * be called for this slave anymore. | |
2170 | */ | |
2171 | netdev_rx_handler_unregister(slave_dev); | |
2172 | synchronize_net(); | |
2173 | ||
58402054 | 2174 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2175 | /* must be called only after the slave |
2176 | * has been detached from the list | |
2177 | */ | |
2178 | bond_alb_deinit_slave(bond, slave); | |
2179 | } | |
2180 | ||
b76cdba9 | 2181 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2182 | bond_del_vlans_from_slave(bond, slave_dev); |
2183 | ||
2184 | /* If the mode USES_PRIMARY, then we should only remove its | |
2185 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2186 | * already taken care of above when we detached the slave | |
2187 | */ | |
2188 | if (!USES_PRIMARY(bond->params.mode)) { | |
2189 | /* unset promiscuity level from slave */ | |
3d632c3f | 2190 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2191 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2192 | |
2193 | /* unset allmulti level from slave */ | |
3d632c3f | 2194 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2195 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2196 | |
2197 | /* flush master's mc_list from slave */ | |
b9e40857 | 2198 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2199 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2200 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2201 | } |
2202 | ||
1765a575 | 2203 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2204 | |
8a8efa22 AW |
2205 | slave_disable_netpoll(slave); |
2206 | ||
1da177e4 LT |
2207 | /* close slave before restoring its mac address */ |
2208 | dev_close(slave_dev); | |
2209 | ||
dd957c57 | 2210 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2211 | /* restore original ("permanent") mac address*/ |
2212 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2213 | addr.sa_family = slave_dev->type; | |
2214 | dev_set_mac_address(slave_dev, &addr); | |
2215 | } | |
1da177e4 | 2216 | |
1da177e4 LT |
2217 | kfree(slave); |
2218 | ||
2219 | /* re-acquire the lock before getting the next slave */ | |
2220 | write_lock_bh(&bond->lock); | |
2221 | } | |
2222 | ||
2223 | /* zero the mac address of the master so it will be | |
2224 | * set by the application to the mac address of the | |
2225 | * first slave | |
2226 | */ | |
2227 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2228 | ||
cc0e4070 | 2229 | if (bond_vlan_used(bond)) { |
a4aee5c8 JP |
2230 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2231 | bond_dev->name, bond_dev->name); | |
2232 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2233 | bond_dev->name); | |
1da177e4 LT |
2234 | } |
2235 | ||
a4aee5c8 | 2236 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2237 | |
2238 | out: | |
2239 | write_unlock_bh(&bond->lock); | |
b2a103e6 MM |
2240 | |
2241 | bond_compute_features(bond); | |
2242 | ||
1da177e4 LT |
2243 | return 0; |
2244 | } | |
2245 | ||
2246 | /* | |
2247 | * This function changes the active slave to slave <slave_dev>. | |
2248 | * It returns -EINVAL in the following cases. | |
2249 | * - <slave_dev> is not found in the list. | |
2250 | * - There is not active slave now. | |
2251 | * - <slave_dev> is already active. | |
2252 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2253 | * - <slave_dev> is not running. | |
3d632c3f SH |
2254 | * In these cases, this function does nothing. |
2255 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2256 | */ |
2257 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2258 | { | |
454d7c9b | 2259 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2260 | struct slave *old_active = NULL; |
2261 | struct slave *new_active = NULL; | |
2262 | int res = 0; | |
2263 | ||
3d632c3f | 2264 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2265 | return -EINVAL; |
1da177e4 LT |
2266 | |
2267 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2268 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2269 | return -EINVAL; |
1da177e4 | 2270 | |
059fe7a5 | 2271 | read_lock(&bond->lock); |
1da177e4 | 2272 | |
059fe7a5 | 2273 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2274 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2275 | read_unlock(&bond->curr_slave_lock); |
2276 | ||
1da177e4 LT |
2277 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2278 | ||
2279 | /* | |
2280 | * Changing to the current active: do nothing; return success. | |
2281 | */ | |
2282 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2283 | read_unlock(&bond->lock); |
1da177e4 LT |
2284 | return 0; |
2285 | } | |
2286 | ||
2287 | if ((new_active) && | |
2288 | (old_active) && | |
2289 | (new_active->link == BOND_LINK_UP) && | |
2290 | IS_UP(new_active->dev)) { | |
e843fa50 | 2291 | block_netpoll_tx(); |
059fe7a5 | 2292 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2293 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2294 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2295 | unblock_netpoll_tx(); |
3d632c3f | 2296 | } else |
1da177e4 | 2297 | res = -EINVAL; |
1da177e4 | 2298 | |
059fe7a5 | 2299 | read_unlock(&bond->lock); |
1da177e4 LT |
2300 | |
2301 | return res; | |
2302 | } | |
2303 | ||
1da177e4 LT |
2304 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2305 | { | |
454d7c9b | 2306 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2307 | |
2308 | info->bond_mode = bond->params.mode; | |
2309 | info->miimon = bond->params.miimon; | |
2310 | ||
6603a6f2 | 2311 | read_lock(&bond->lock); |
1da177e4 | 2312 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2313 | read_unlock(&bond->lock); |
1da177e4 LT |
2314 | |
2315 | return 0; | |
2316 | } | |
2317 | ||
2318 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2319 | { | |
454d7c9b | 2320 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2321 | struct slave *slave; |
689c96cc | 2322 | int i, res = -ENODEV; |
1da177e4 | 2323 | |
6603a6f2 | 2324 | read_lock(&bond->lock); |
1da177e4 LT |
2325 | |
2326 | bond_for_each_slave(bond, slave, i) { | |
2327 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2328 | res = 0; |
2329 | strcpy(info->slave_name, slave->dev->name); | |
2330 | info->link = slave->link; | |
e30bc066 | 2331 | info->state = bond_slave_state(slave); |
689c96cc | 2332 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2333 | break; |
2334 | } | |
2335 | } | |
2336 | ||
6603a6f2 | 2337 | read_unlock(&bond->lock); |
1da177e4 | 2338 | |
689c96cc | 2339 | return res; |
1da177e4 LT |
2340 | } |
2341 | ||
2342 | /*-------------------------------- Monitoring -------------------------------*/ | |
2343 | ||
1da177e4 | 2344 | |
f0c76d61 JV |
2345 | static int bond_miimon_inspect(struct bonding *bond) |
2346 | { | |
2347 | struct slave *slave; | |
2348 | int i, link_state, commit = 0; | |
41f89100 JP |
2349 | bool ignore_updelay; |
2350 | ||
2351 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2352 | |
2353 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2354 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2355 | |
f0c76d61 | 2356 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2357 | |
2358 | switch (slave->link) { | |
f0c76d61 JV |
2359 | case BOND_LINK_UP: |
2360 | if (link_state) | |
2361 | continue; | |
1da177e4 | 2362 | |
f0c76d61 JV |
2363 | slave->link = BOND_LINK_FAIL; |
2364 | slave->delay = bond->params.downdelay; | |
2365 | if (slave->delay) { | |
a4aee5c8 JP |
2366 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2367 | bond->dev->name, | |
2368 | (bond->params.mode == | |
2369 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2370 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2371 | "active " : "backup ") : "", |
2372 | slave->dev->name, | |
2373 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2374 | } |
f0c76d61 JV |
2375 | /*FALLTHRU*/ |
2376 | case BOND_LINK_FAIL: | |
2377 | if (link_state) { | |
2378 | /* | |
2379 | * recovered before downdelay expired | |
2380 | */ | |
2381 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2382 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2383 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2384 | bond->dev->name, | |
2385 | (bond->params.downdelay - slave->delay) * | |
2386 | bond->params.miimon, | |
2387 | slave->dev->name); | |
f0c76d61 | 2388 | continue; |
1da177e4 | 2389 | } |
f0c76d61 JV |
2390 | |
2391 | if (slave->delay <= 0) { | |
2392 | slave->new_link = BOND_LINK_DOWN; | |
2393 | commit++; | |
2394 | continue; | |
1da177e4 | 2395 | } |
1da177e4 | 2396 | |
f0c76d61 JV |
2397 | slave->delay--; |
2398 | break; | |
2399 | ||
2400 | case BOND_LINK_DOWN: | |
2401 | if (!link_state) | |
2402 | continue; | |
2403 | ||
2404 | slave->link = BOND_LINK_BACK; | |
2405 | slave->delay = bond->params.updelay; | |
2406 | ||
2407 | if (slave->delay) { | |
a4aee5c8 JP |
2408 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2409 | bond->dev->name, slave->dev->name, | |
2410 | ignore_updelay ? 0 : | |
2411 | bond->params.updelay * | |
2412 | bond->params.miimon); | |
f0c76d61 JV |
2413 | } |
2414 | /*FALLTHRU*/ | |
2415 | case BOND_LINK_BACK: | |
2416 | if (!link_state) { | |
2417 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2418 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2419 | bond->dev->name, | |
2420 | (bond->params.updelay - slave->delay) * | |
2421 | bond->params.miimon, | |
2422 | slave->dev->name); | |
f0c76d61 JV |
2423 | |
2424 | continue; | |
2425 | } | |
2426 | ||
41f89100 JP |
2427 | if (ignore_updelay) |
2428 | slave->delay = 0; | |
2429 | ||
f0c76d61 JV |
2430 | if (slave->delay <= 0) { |
2431 | slave->new_link = BOND_LINK_UP; | |
2432 | commit++; | |
41f89100 | 2433 | ignore_updelay = false; |
f0c76d61 | 2434 | continue; |
1da177e4 | 2435 | } |
f0c76d61 JV |
2436 | |
2437 | slave->delay--; | |
1da177e4 | 2438 | break; |
f0c76d61 JV |
2439 | } |
2440 | } | |
1da177e4 | 2441 | |
f0c76d61 JV |
2442 | return commit; |
2443 | } | |
1da177e4 | 2444 | |
f0c76d61 JV |
2445 | static void bond_miimon_commit(struct bonding *bond) |
2446 | { | |
2447 | struct slave *slave; | |
2448 | int i; | |
2449 | ||
2450 | bond_for_each_slave(bond, slave, i) { | |
2451 | switch (slave->new_link) { | |
2452 | case BOND_LINK_NOCHANGE: | |
2453 | continue; | |
1da177e4 | 2454 | |
f0c76d61 JV |
2455 | case BOND_LINK_UP: |
2456 | slave->link = BOND_LINK_UP; | |
2457 | slave->jiffies = jiffies; | |
2458 | ||
2459 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2460 | /* prevent it from being the active one */ | |
e30bc066 | 2461 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2462 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2463 | /* make it immediately active */ | |
e30bc066 | 2464 | bond_set_active_slave(slave); |
f0c76d61 JV |
2465 | } else if (slave != bond->primary_slave) { |
2466 | /* prevent it from being the active one */ | |
e30bc066 | 2467 | bond_set_backup_slave(slave); |
1da177e4 | 2468 | } |
1da177e4 | 2469 | |
546add79 KPO |
2470 | bond_update_speed_duplex(slave); |
2471 | ||
5d30530e | 2472 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 KPO |
2473 | bond->dev->name, slave->dev->name, |
2474 | slave->speed, slave->duplex ? "full" : "half"); | |
1da177e4 | 2475 | |
f0c76d61 JV |
2476 | /* notify ad that the link status has changed */ |
2477 | if (bond->params.mode == BOND_MODE_8023AD) | |
2478 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2479 | |
58402054 | 2480 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2481 | bond_alb_handle_link_change(bond, slave, |
2482 | BOND_LINK_UP); | |
1da177e4 | 2483 | |
f0c76d61 JV |
2484 | if (!bond->curr_active_slave || |
2485 | (slave == bond->primary_slave)) | |
2486 | goto do_failover; | |
1da177e4 | 2487 | |
f0c76d61 | 2488 | continue; |
059fe7a5 | 2489 | |
f0c76d61 | 2490 | case BOND_LINK_DOWN: |
fba4acda JV |
2491 | if (slave->link_failure_count < UINT_MAX) |
2492 | slave->link_failure_count++; | |
2493 | ||
f0c76d61 | 2494 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2495 | |
f0c76d61 JV |
2496 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2497 | bond->params.mode == BOND_MODE_8023AD) | |
2498 | bond_set_slave_inactive_flags(slave); | |
2499 | ||
a4aee5c8 JP |
2500 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2501 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2502 | |
2503 | if (bond->params.mode == BOND_MODE_8023AD) | |
2504 | bond_3ad_handle_link_change(slave, | |
2505 | BOND_LINK_DOWN); | |
2506 | ||
ae63e808 | 2507 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2508 | bond_alb_handle_link_change(bond, slave, |
2509 | BOND_LINK_DOWN); | |
2510 | ||
2511 | if (slave == bond->curr_active_slave) | |
2512 | goto do_failover; | |
2513 | ||
2514 | continue; | |
2515 | ||
2516 | default: | |
a4aee5c8 | 2517 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2518 | bond->dev->name, slave->new_link, |
2519 | slave->dev->name); | |
2520 | slave->new_link = BOND_LINK_NOCHANGE; | |
2521 | ||
2522 | continue; | |
2523 | } | |
2524 | ||
2525 | do_failover: | |
2526 | ASSERT_RTNL(); | |
e843fa50 | 2527 | block_netpoll_tx(); |
f0c76d61 JV |
2528 | write_lock_bh(&bond->curr_slave_lock); |
2529 | bond_select_active_slave(bond); | |
2530 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2531 | unblock_netpoll_tx(); |
f0c76d61 JV |
2532 | } |
2533 | ||
2534 | bond_set_carrier(bond); | |
1da177e4 LT |
2535 | } |
2536 | ||
0b0eef66 JV |
2537 | /* |
2538 | * bond_mii_monitor | |
2539 | * | |
2540 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2541 | * inspection, then (if inspection indicates something needs to be done) |
2542 | * an acquisition of appropriate locks followed by a commit phase to | |
2543 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2544 | */ |
2545 | void bond_mii_monitor(struct work_struct *work) | |
2546 | { | |
2547 | struct bonding *bond = container_of(work, struct bonding, | |
2548 | mii_work.work); | |
ad246c99 | 2549 | bool should_notify_peers = false; |
e6d265e8 | 2550 | unsigned long delay; |
0b0eef66 JV |
2551 | |
2552 | read_lock(&bond->lock); | |
e6d265e8 JV |
2553 | |
2554 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 JV |
2555 | |
2556 | if (bond->slave_cnt == 0) | |
2557 | goto re_arm; | |
b59f9f74 | 2558 | |
ad246c99 BH |
2559 | should_notify_peers = bond_should_notify_peers(bond); |
2560 | ||
f0c76d61 | 2561 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2562 | read_unlock(&bond->lock); |
e6d265e8 JV |
2563 | |
2564 | /* Race avoidance with bond_close cancel of workqueue */ | |
2565 | if (!rtnl_trylock()) { | |
2566 | read_lock(&bond->lock); | |
2567 | delay = 1; | |
2568 | should_notify_peers = false; | |
2569 | goto re_arm; | |
2570 | } | |
2571 | ||
0b0eef66 | 2572 | read_lock(&bond->lock); |
f0c76d61 JV |
2573 | |
2574 | bond_miimon_commit(bond); | |
2575 | ||
5655662d JV |
2576 | read_unlock(&bond->lock); |
2577 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2578 | read_lock(&bond->lock); | |
0b0eef66 JV |
2579 | } |
2580 | ||
f0c76d61 | 2581 | re_arm: |
e6d265e8 JV |
2582 | if (bond->params.miimon) |
2583 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2584 | ||
0b0eef66 | 2585 | read_unlock(&bond->lock); |
ad246c99 BH |
2586 | |
2587 | if (should_notify_peers) { | |
e6d265e8 JV |
2588 | if (!rtnl_trylock()) { |
2589 | read_lock(&bond->lock); | |
2590 | bond->send_peer_notif++; | |
2591 | read_unlock(&bond->lock); | |
2592 | return; | |
2593 | } | |
b7bc2a5b | 2594 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2595 | rtnl_unlock(); |
2596 | } | |
0b0eef66 | 2597 | } |
c3ade5ca | 2598 | |
d3bb52b0 | 2599 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2600 | { |
0883beca | 2601 | struct vlan_entry *vlan; |
eaddcd76 | 2602 | struct net_device *vlan_dev; |
f5b2b966 | 2603 | |
eaddcd76 | 2604 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
f5b2b966 JV |
2605 | return 1; |
2606 | ||
0883beca | 2607 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
eaddcd76 AG |
2608 | rcu_read_lock(); |
2609 | vlan_dev = __vlan_find_dev_deep(bond->dev, vlan->vlan_id); | |
2610 | rcu_read_unlock(); | |
2611 | if (vlan_dev && ip == bond_confirm_addr(vlan_dev, 0, ip)) | |
f5b2b966 JV |
2612 | return 1; |
2613 | } | |
2614 | ||
2615 | return 0; | |
2616 | } | |
2617 | ||
c3ade5ca JV |
2618 | /* |
2619 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2620 | * switches in VLAN mode (especially if ports are configured as | |
2621 | * "native" to a VLAN) might not pass non-tagged frames. | |
2622 | */ | |
d3bb52b0 | 2623 | 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 |
2624 | { |
2625 | struct sk_buff *skb; | |
2626 | ||
5a03cdb7 | 2627 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2628 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2629 | |
c3ade5ca JV |
2630 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2631 | NULL, slave_dev->dev_addr, NULL); | |
2632 | ||
2633 | if (!skb) { | |
a4aee5c8 | 2634 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2635 | return; |
2636 | } | |
2637 | if (vlan_id) { | |
2638 | skb = vlan_put_tag(skb, vlan_id); | |
2639 | if (!skb) { | |
a4aee5c8 | 2640 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2641 | return; |
2642 | } | |
2643 | } | |
2644 | arp_xmit(skb); | |
2645 | } | |
2646 | ||
2647 | ||
1da177e4 LT |
2648 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2649 | { | |
b23dd4fe | 2650 | int i, vlan_id; |
d3bb52b0 | 2651 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2652 | struct vlan_entry *vlan; |
eaddcd76 | 2653 | struct net_device *vlan_dev = NULL; |
c3ade5ca | 2654 | struct rtable *rt; |
1da177e4 | 2655 | |
6b780567 | 2656 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
eaddcd76 | 2657 | __be32 addr; |
6b780567 | 2658 | if (!targets[i]) |
5a31bec0 | 2659 | break; |
5a03cdb7 | 2660 | pr_debug("basa: target %x\n", targets[i]); |
cc0e4070 | 2661 | if (!bond_vlan_used(bond)) { |
5a03cdb7 | 2662 | pr_debug("basa: empty vlan: arp_send\n"); |
eaddcd76 | 2663 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2664 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2665 | addr, 0); |
c3ade5ca JV |
2666 | continue; |
2667 | } | |
2668 | ||
2669 | /* | |
2670 | * If VLANs are configured, we do a route lookup to | |
2671 | * determine which VLAN interface would be used, so we | |
2672 | * can tag the ARP with the proper VLAN tag. | |
2673 | */ | |
78fbfd8a DM |
2674 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2675 | RTO_ONLINK, 0); | |
b23dd4fe | 2676 | if (IS_ERR(rt)) { |
c3ade5ca | 2677 | if (net_ratelimit()) { |
a4aee5c8 | 2678 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2679 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2680 | } |
2681 | continue; | |
2682 | } | |
2683 | ||
2684 | /* | |
2685 | * This target is not on a VLAN | |
2686 | */ | |
d8d1f30b | 2687 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2688 | ip_rt_put(rt); |
5a03cdb7 | 2689 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
eaddcd76 | 2690 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2691 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2692 | addr, 0); |
c3ade5ca JV |
2693 | continue; |
2694 | } | |
2695 | ||
2696 | vlan_id = 0; | |
0883beca | 2697 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
cc0e4070 JP |
2698 | rcu_read_lock(); |
2699 | vlan_dev = __vlan_find_dev_deep(bond->dev, | |
2700 | vlan->vlan_id); | |
2701 | rcu_read_unlock(); | |
d8d1f30b | 2702 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2703 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2704 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2705 | vlan_dev->name, vlan_id); |
2706 | break; | |
2707 | } | |
2708 | } | |
2709 | ||
eaddcd76 | 2710 | if (vlan_id && vlan_dev) { |
ed4b9f80 | 2711 | ip_rt_put(rt); |
eaddcd76 | 2712 | addr = bond_confirm_addr(vlan_dev, targets[i], 0); |
c3ade5ca | 2713 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2714 | addr, vlan_id); |
c3ade5ca JV |
2715 | continue; |
2716 | } | |
2717 | ||
2718 | if (net_ratelimit()) { | |
a4aee5c8 | 2719 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2720 | bond->dev->name, &targets[i], |
d8d1f30b | 2721 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2722 | } |
ed4b9f80 | 2723 | ip_rt_put(rt); |
c3ade5ca JV |
2724 | } |
2725 | } | |
2726 | ||
d3bb52b0 | 2727 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2728 | { |
2729 | int i; | |
d3bb52b0 | 2730 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2731 | |
f5b2b966 | 2732 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2733 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2734 | &sip, &tip, i, &targets[i], |
2735 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2736 | if (sip == targets[i]) { |
2737 | if (bond_has_this_ip(bond, tip)) | |
2738 | slave->last_arp_rx = jiffies; | |
2739 | return; | |
2740 | } | |
2741 | } | |
2742 | } | |
2743 | ||
de063b70 ED |
2744 | static int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2745 | struct slave *slave) | |
f5b2b966 | 2746 | { |
de063b70 | 2747 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2748 | unsigned char *arp_ptr; |
d3bb52b0 | 2749 | __be32 sip, tip; |
de063b70 | 2750 | int alen; |
f5b2b966 | 2751 | |
3aba891d | 2752 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2753 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2754 | |
f5b2b966 | 2755 | read_lock(&bond->lock); |
de063b70 | 2756 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2757 | |
3aba891d JP |
2758 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2759 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2760 | |
de063b70 ED |
2761 | if (alen > skb_headlen(skb)) { |
2762 | arp = kmalloc(alen, GFP_ATOMIC); | |
2763 | if (!arp) | |
2764 | goto out_unlock; | |
2765 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2766 | goto out_unlock; | |
2767 | } | |
f5b2b966 | 2768 | |
3aba891d | 2769 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2770 | skb->pkt_type == PACKET_OTHERHOST || |
2771 | skb->pkt_type == PACKET_LOOPBACK || | |
2772 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2773 | arp->ar_pro != htons(ETH_P_IP) || | |
2774 | arp->ar_pln != 4) | |
2775 | goto out_unlock; | |
2776 | ||
2777 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2778 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2779 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2780 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2781 | memcpy(&tip, arp_ptr, 4); |
2782 | ||
5a03cdb7 | 2783 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2784 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2785 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2786 | &sip, &tip); | |
f5b2b966 JV |
2787 | |
2788 | /* | |
2789 | * Backup slaves won't see the ARP reply, but do come through | |
2790 | * here for each ARP probe (so we swap the sip/tip to validate | |
2791 | * the probe). In a "redundant switch, common router" type of | |
2792 | * configuration, the ARP probe will (hopefully) travel from | |
2793 | * the active, through one switch, the router, then the other | |
2794 | * switch before reaching the backup. | |
2795 | */ | |
e30bc066 | 2796 | if (bond_is_active_slave(slave)) |
f5b2b966 JV |
2797 | bond_validate_arp(bond, slave, sip, tip); |
2798 | else | |
2799 | bond_validate_arp(bond, slave, tip, sip); | |
2800 | ||
2801 | out_unlock: | |
2802 | read_unlock(&bond->lock); | |
de063b70 ED |
2803 | if (arp != (struct arphdr *)skb->data) |
2804 | kfree(arp); | |
13a8e0c8 | 2805 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2806 | } |
2807 | ||
1da177e4 LT |
2808 | /* |
2809 | * this function is called regularly to monitor each slave's link | |
2810 | * ensuring that traffic is being sent and received when arp monitoring | |
2811 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2812 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2813 | * arp monitoring in active backup mode. | |
2814 | */ | |
1b76b316 | 2815 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2816 | { |
1b76b316 JV |
2817 | struct bonding *bond = container_of(work, struct bonding, |
2818 | arp_work.work); | |
1da177e4 LT |
2819 | struct slave *slave, *oldcurrent; |
2820 | int do_failover = 0; | |
da210f55 | 2821 | int delta_in_ticks, extra_ticks; |
1da177e4 LT |
2822 | int i; |
2823 | ||
2824 | read_lock(&bond->lock); | |
2825 | ||
5ce0da8f | 2826 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
da210f55 | 2827 | extra_ticks = delta_in_ticks / 2; |
1da177e4 | 2828 | |
3d632c3f | 2829 | if (bond->slave_cnt == 0) |
1da177e4 | 2830 | goto re_arm; |
1da177e4 LT |
2831 | |
2832 | read_lock(&bond->curr_slave_lock); | |
2833 | oldcurrent = bond->curr_active_slave; | |
2834 | read_unlock(&bond->curr_slave_lock); | |
2835 | ||
2836 | /* see if any of the previous devices are up now (i.e. they have | |
2837 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2838 | * the picture unless it is null. also, slave->jiffies is not needed | |
2839 | * here because we send an arp on each slave and give a slave as | |
2840 | * long as it needs to get the tx/rx within the delta. | |
2841 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2842 | * so it can wait | |
2843 | */ | |
2844 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2845 | unsigned long trans_start = dev_trans_start(slave->dev); |
2846 | ||
1da177e4 | 2847 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2848 | if (time_in_range(jiffies, |
2849 | trans_start - delta_in_ticks, | |
da210f55 | 2850 | trans_start + delta_in_ticks + extra_ticks) && |
cb32f2a0 JB |
2851 | time_in_range(jiffies, |
2852 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2853 | slave->dev->last_rx + delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2854 | |
2855 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2856 | bond_set_active_slave(slave); |
1da177e4 LT |
2857 | |
2858 | /* primary_slave has no meaning in round-robin | |
2859 | * mode. the window of a slave being up and | |
2860 | * curr_active_slave being null after enslaving | |
2861 | * is closed. | |
2862 | */ | |
2863 | if (!oldcurrent) { | |
a4aee5c8 JP |
2864 | pr_info("%s: link status definitely up for interface %s, ", |
2865 | bond->dev->name, | |
2866 | slave->dev->name); | |
1da177e4 LT |
2867 | do_failover = 1; |
2868 | } else { | |
a4aee5c8 JP |
2869 | pr_info("%s: interface %s is now up\n", |
2870 | bond->dev->name, | |
2871 | slave->dev->name); | |
1da177e4 LT |
2872 | } |
2873 | } | |
2874 | } else { | |
2875 | /* slave->link == BOND_LINK_UP */ | |
2876 | ||
2877 | /* not all switches will respond to an arp request | |
2878 | * when the source ip is 0, so don't take the link down | |
2879 | * if we don't know our ip yet | |
2880 | */ | |
cb32f2a0 JB |
2881 | if (!time_in_range(jiffies, |
2882 | trans_start - delta_in_ticks, | |
da210f55 | 2883 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2884 | !time_in_range(jiffies, |
2885 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2886 | slave->dev->last_rx + 2 * delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2887 | |
2888 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2889 | bond_set_backup_slave(slave); |
1da177e4 | 2890 | |
3d632c3f | 2891 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2892 | slave->link_failure_count++; |
1da177e4 | 2893 | |
a4aee5c8 JP |
2894 | pr_info("%s: interface %s is now down.\n", |
2895 | bond->dev->name, | |
2896 | slave->dev->name); | |
1da177e4 | 2897 | |
3d632c3f | 2898 | if (slave == oldcurrent) |
1da177e4 | 2899 | do_failover = 1; |
1da177e4 LT |
2900 | } |
2901 | } | |
2902 | ||
2903 | /* note: if switch is in round-robin mode, all links | |
2904 | * must tx arp to ensure all links rx an arp - otherwise | |
2905 | * links may oscillate or not come up at all; if switch is | |
2906 | * in something like xor mode, there is nothing we can | |
2907 | * do - all replies will be rx'ed on same link causing slaves | |
2908 | * to be unstable during low/no traffic periods | |
2909 | */ | |
3d632c3f | 2910 | if (IS_UP(slave->dev)) |
1da177e4 | 2911 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2912 | } |
2913 | ||
2914 | if (do_failover) { | |
e843fa50 | 2915 | block_netpoll_tx(); |
059fe7a5 | 2916 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2917 | |
2918 | bond_select_active_slave(bond); | |
2919 | ||
059fe7a5 | 2920 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2921 | unblock_netpoll_tx(); |
1da177e4 LT |
2922 | } |
2923 | ||
2924 | re_arm: | |
e6d265e8 | 2925 | if (bond->params.arp_interval) |
1b76b316 | 2926 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 2927 | |
1da177e4 LT |
2928 | read_unlock(&bond->lock); |
2929 | } | |
2930 | ||
2931 | /* | |
b2220cad JV |
2932 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2933 | * changes. Sets new_link in slaves to specify what action should take | |
2934 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2935 | * to link states must be committed. | |
2936 | * | |
2937 | * Called with bond->lock held for read. | |
1da177e4 | 2938 | */ |
b2220cad | 2939 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2940 | { |
1da177e4 | 2941 | struct slave *slave; |
b2220cad | 2942 | int i, commit = 0; |
cb32f2a0 | 2943 | unsigned long trans_start; |
da210f55 JB |
2944 | int extra_ticks; |
2945 | ||
2946 | /* All the time comparisons below need some extra time. Otherwise, on | |
2947 | * fast networks the ARP probe/reply may arrive within the same jiffy | |
2948 | * as it was sent. Then, the next time the ARP monitor is run, one | |
2949 | * arp_interval will already have passed in the comparisons. | |
2950 | */ | |
2951 | extra_ticks = delta_in_ticks / 2; | |
1da177e4 | 2952 | |
b2220cad JV |
2953 | bond_for_each_slave(bond, slave, i) { |
2954 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2955 | |
b2220cad | 2956 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2957 | if (time_in_range(jiffies, |
2958 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2959 | slave_last_rx(bond, slave) + delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2960 | |
b2220cad JV |
2961 | slave->new_link = BOND_LINK_UP; |
2962 | commit++; | |
2963 | } | |
1da177e4 | 2964 | |
b2220cad JV |
2965 | continue; |
2966 | } | |
1da177e4 | 2967 | |
b2220cad JV |
2968 | /* |
2969 | * Give slaves 2*delta after being enslaved or made | |
2970 | * active. This avoids bouncing, as the last receive | |
2971 | * times need a full ARP monitor cycle to be updated. | |
2972 | */ | |
cb32f2a0 JB |
2973 | if (time_in_range(jiffies, |
2974 | slave->jiffies - delta_in_ticks, | |
da210f55 | 2975 | slave->jiffies + 2 * delta_in_ticks + extra_ticks)) |
b2220cad JV |
2976 | continue; |
2977 | ||
2978 | /* | |
2979 | * Backup slave is down if: | |
2980 | * - No current_arp_slave AND | |
2981 | * - more than 3*delta since last receive AND | |
2982 | * - the bond has an IP address | |
2983 | * | |
2984 | * Note: a non-null current_arp_slave indicates | |
2985 | * the curr_active_slave went down and we are | |
2986 | * searching for a new one; under this condition | |
2987 | * we only take the curr_active_slave down - this | |
2988 | * gives each slave a chance to tx/rx traffic | |
2989 | * before being taken out | |
2990 | */ | |
e30bc066 | 2991 | if (!bond_is_active_slave(slave) && |
b2220cad | 2992 | !bond->current_arp_slave && |
cb32f2a0 JB |
2993 | !time_in_range(jiffies, |
2994 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2995 | slave_last_rx(bond, slave) + 3 * delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2996 | |
b2220cad JV |
2997 | slave->new_link = BOND_LINK_DOWN; |
2998 | commit++; | |
2999 | } | |
3000 | ||
3001 | /* | |
3002 | * Active slave is down if: | |
3003 | * - more than 2*delta since transmitting OR | |
3004 | * - (more than 2*delta since receive AND | |
3005 | * the bond has an IP address) | |
3006 | */ | |
cb32f2a0 | 3007 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 3008 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
3009 | (!time_in_range(jiffies, |
3010 | trans_start - delta_in_ticks, | |
da210f55 | 3011 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
3012 | !time_in_range(jiffies, |
3013 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 3014 | slave_last_rx(bond, slave) + 2 * delta_in_ticks + extra_ticks))) { |
cb32f2a0 | 3015 | |
b2220cad JV |
3016 | slave->new_link = BOND_LINK_DOWN; |
3017 | commit++; | |
3018 | } | |
1da177e4 LT |
3019 | } |
3020 | ||
b2220cad JV |
3021 | return commit; |
3022 | } | |
1da177e4 | 3023 | |
b2220cad JV |
3024 | /* |
3025 | * Called to commit link state changes noted by inspection step of | |
3026 | * active-backup mode ARP monitor. | |
3027 | * | |
3028 | * Called with RTNL and bond->lock for read. | |
3029 | */ | |
3030 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
3031 | { | |
3032 | struct slave *slave; | |
3033 | int i; | |
cb32f2a0 | 3034 | unsigned long trans_start; |
1da177e4 | 3035 | |
b2220cad JV |
3036 | bond_for_each_slave(bond, slave, i) { |
3037 | switch (slave->new_link) { | |
3038 | case BOND_LINK_NOCHANGE: | |
3039 | continue; | |
ff59c456 | 3040 | |
b2220cad | 3041 | case BOND_LINK_UP: |
cb32f2a0 | 3042 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 3043 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
3044 | time_in_range(jiffies, |
3045 | trans_start - delta_in_ticks, | |
da210f55 | 3046 | trans_start + delta_in_ticks + delta_in_ticks / 2)) || |
b9f60253 | 3047 | bond->curr_active_slave != slave) { |
b2220cad | 3048 | slave->link = BOND_LINK_UP; |
5a430974 VF |
3049 | if (bond->current_arp_slave) { |
3050 | bond_set_slave_inactive_flags( | |
3051 | bond->current_arp_slave); | |
3052 | bond->current_arp_slave = NULL; | |
3053 | } | |
b2220cad | 3054 | |
a4aee5c8 | 3055 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 3056 | bond->dev->name, slave->dev->name); |
b2220cad | 3057 | |
b9f60253 JP |
3058 | if (!bond->curr_active_slave || |
3059 | (slave == bond->primary_slave)) | |
3060 | goto do_failover; | |
1da177e4 | 3061 | |
b9f60253 | 3062 | } |
1da177e4 | 3063 | |
b9f60253 | 3064 | continue; |
1da177e4 | 3065 | |
b2220cad JV |
3066 | case BOND_LINK_DOWN: |
3067 | if (slave->link_failure_count < UINT_MAX) | |
3068 | slave->link_failure_count++; | |
3069 | ||
3070 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 3071 | bond_set_slave_inactive_flags(slave); |
b2220cad | 3072 | |
a4aee5c8 | 3073 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 3074 | bond->dev->name, slave->dev->name); |
b2220cad | 3075 | |
b9f60253 | 3076 | if (slave == bond->curr_active_slave) { |
b2220cad | 3077 | bond->current_arp_slave = NULL; |
b9f60253 | 3078 | goto do_failover; |
1da177e4 | 3079 | } |
b9f60253 JP |
3080 | |
3081 | continue; | |
b2220cad JV |
3082 | |
3083 | default: | |
a4aee5c8 | 3084 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
3085 | bond->dev->name, slave->new_link, |
3086 | slave->dev->name); | |
b9f60253 | 3087 | continue; |
1da177e4 | 3088 | } |
1da177e4 | 3089 | |
b9f60253 JP |
3090 | do_failover: |
3091 | ASSERT_RTNL(); | |
e843fa50 | 3092 | block_netpoll_tx(); |
b2220cad | 3093 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3094 | bond_select_active_slave(bond); |
b2220cad | 3095 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3096 | unblock_netpoll_tx(); |
b2220cad | 3097 | } |
1da177e4 | 3098 | |
b2220cad JV |
3099 | bond_set_carrier(bond); |
3100 | } | |
1da177e4 | 3101 | |
b2220cad JV |
3102 | /* |
3103 | * Send ARP probes for active-backup mode ARP monitor. | |
3104 | * | |
3105 | * Called with bond->lock held for read. | |
3106 | */ | |
3107 | static void bond_ab_arp_probe(struct bonding *bond) | |
3108 | { | |
3109 | struct slave *slave; | |
3110 | int i; | |
1da177e4 | 3111 | |
b2220cad | 3112 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3113 | |
b2220cad | 3114 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3115 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3116 | bond->current_arp_slave->dev->name, | |
3117 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3118 | |
b2220cad JV |
3119 | if (bond->curr_active_slave) { |
3120 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3121 | read_unlock(&bond->curr_slave_lock); | |
3122 | return; | |
3123 | } | |
1da177e4 | 3124 | |
b2220cad | 3125 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3126 | |
b2220cad JV |
3127 | /* if we don't have a curr_active_slave, search for the next available |
3128 | * backup slave from the current_arp_slave and make it the candidate | |
3129 | * for becoming the curr_active_slave | |
3130 | */ | |
1da177e4 | 3131 | |
b2220cad JV |
3132 | if (!bond->current_arp_slave) { |
3133 | bond->current_arp_slave = bond->first_slave; | |
3134 | if (!bond->current_arp_slave) | |
3135 | return; | |
3136 | } | |
1da177e4 | 3137 | |
b2220cad | 3138 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3139 | |
b2220cad JV |
3140 | /* search for next candidate */ |
3141 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3142 | if (IS_UP(slave->dev)) { | |
3143 | slave->link = BOND_LINK_BACK; | |
3144 | bond_set_slave_active_flags(slave); | |
3145 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3146 | slave->jiffies = jiffies; |
b2220cad JV |
3147 | bond->current_arp_slave = slave; |
3148 | break; | |
1da177e4 LT |
3149 | } |
3150 | ||
b2220cad JV |
3151 | /* if the link state is up at this point, we |
3152 | * mark it down - this can happen if we have | |
3153 | * simultaneous link failures and | |
3154 | * reselect_active_interface doesn't make this | |
3155 | * one the current slave so it is still marked | |
3156 | * up when it is actually down | |
1da177e4 | 3157 | */ |
b2220cad JV |
3158 | if (slave->link == BOND_LINK_UP) { |
3159 | slave->link = BOND_LINK_DOWN; | |
3160 | if (slave->link_failure_count < UINT_MAX) | |
3161 | slave->link_failure_count++; | |
1da177e4 | 3162 | |
b2220cad JV |
3163 | bond_set_slave_inactive_flags(slave); |
3164 | ||
a4aee5c8 JP |
3165 | pr_info("%s: backup interface %s is now down.\n", |
3166 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3167 | } |
b2220cad JV |
3168 | } |
3169 | } | |
1da177e4 | 3170 | |
b2220cad JV |
3171 | void bond_activebackup_arp_mon(struct work_struct *work) |
3172 | { | |
3173 | struct bonding *bond = container_of(work, struct bonding, | |
3174 | arp_work.work); | |
ad246c99 | 3175 | bool should_notify_peers = false; |
b2220cad | 3176 | int delta_in_ticks; |
1da177e4 | 3177 | |
b2220cad | 3178 | read_lock(&bond->lock); |
1da177e4 | 3179 | |
b2220cad | 3180 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3181 | |
b2220cad JV |
3182 | if (bond->slave_cnt == 0) |
3183 | goto re_arm; | |
3184 | ||
ad246c99 BH |
3185 | should_notify_peers = bond_should_notify_peers(bond); |
3186 | ||
b2220cad JV |
3187 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3188 | read_unlock(&bond->lock); | |
e6d265e8 JV |
3189 | |
3190 | /* Race avoidance with bond_close flush of workqueue */ | |
3191 | if (!rtnl_trylock()) { | |
3192 | read_lock(&bond->lock); | |
3193 | delta_in_ticks = 1; | |
3194 | should_notify_peers = false; | |
3195 | goto re_arm; | |
3196 | } | |
3197 | ||
b2220cad JV |
3198 | read_lock(&bond->lock); |
3199 | ||
3200 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3201 | ||
3202 | read_unlock(&bond->lock); | |
3203 | rtnl_unlock(); | |
3204 | read_lock(&bond->lock); | |
1da177e4 LT |
3205 | } |
3206 | ||
b2220cad JV |
3207 | bond_ab_arp_probe(bond); |
3208 | ||
1da177e4 | 3209 | re_arm: |
e6d265e8 | 3210 | if (bond->params.arp_interval) |
1b76b316 | 3211 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 3212 | |
1da177e4 | 3213 | read_unlock(&bond->lock); |
ad246c99 BH |
3214 | |
3215 | if (should_notify_peers) { | |
e6d265e8 JV |
3216 | if (!rtnl_trylock()) { |
3217 | read_lock(&bond->lock); | |
3218 | bond->send_peer_notif++; | |
3219 | read_unlock(&bond->lock); | |
3220 | return; | |
3221 | } | |
b7bc2a5b | 3222 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
3223 | rtnl_unlock(); |
3224 | } | |
1da177e4 LT |
3225 | } |
3226 | ||
1da177e4 LT |
3227 | /*-------------------------- netdev event handling --------------------------*/ |
3228 | ||
3229 | /* | |
3230 | * Change device name | |
3231 | */ | |
3232 | static int bond_event_changename(struct bonding *bond) | |
3233 | { | |
1da177e4 LT |
3234 | bond_remove_proc_entry(bond); |
3235 | bond_create_proc_entry(bond); | |
7e083840 | 3236 | |
f073c7ca TI |
3237 | bond_debug_reregister(bond); |
3238 | ||
1da177e4 LT |
3239 | return NOTIFY_DONE; |
3240 | } | |
3241 | ||
3d632c3f SH |
3242 | static int bond_master_netdev_event(unsigned long event, |
3243 | struct net_device *bond_dev) | |
1da177e4 | 3244 | { |
454d7c9b | 3245 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3246 | |
3247 | switch (event) { | |
3248 | case NETDEV_CHANGENAME: | |
3249 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3250 | case NETDEV_UNREGISTER: |
3251 | bond_remove_proc_entry(event_bond); | |
3252 | break; | |
3253 | case NETDEV_REGISTER: | |
3254 | bond_create_proc_entry(event_bond); | |
3255 | break; | |
1da177e4 LT |
3256 | default: |
3257 | break; | |
3258 | } | |
3259 | ||
3260 | return NOTIFY_DONE; | |
3261 | } | |
3262 | ||
3d632c3f SH |
3263 | static int bond_slave_netdev_event(unsigned long event, |
3264 | struct net_device *slave_dev) | |
1da177e4 LT |
3265 | { |
3266 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3267 | struct bonding *bond = netdev_priv(bond_dev); |
98f41f69 | 3268 | struct slave *slave = NULL; |
1da177e4 LT |
3269 | |
3270 | switch (event) { | |
3271 | case NETDEV_UNREGISTER: | |
3272 | if (bond_dev) { | |
1284cd3a JV |
3273 | if (bond->setup_by_slave) |
3274 | bond_release_and_destroy(bond_dev, slave_dev); | |
3275 | else | |
3276 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3277 | } |
3278 | break; | |
98f41f69 | 3279 | case NETDEV_UP: |
1da177e4 | 3280 | case NETDEV_CHANGE: |
98f41f69 WP |
3281 | slave = bond_get_slave_by_dev(bond, slave_dev); |
3282 | if (slave) { | |
3283 | u32 old_speed = slave->speed; | |
3284 | u8 old_duplex = slave->duplex; | |
17d04500 | 3285 | |
98f41f69 | 3286 | bond_update_speed_duplex(slave); |
17d04500 | 3287 | |
98f41f69 | 3288 | if (bond->params.mode == BOND_MODE_8023AD) { |
17d04500 JV |
3289 | if (old_speed != slave->speed) |
3290 | bond_3ad_adapter_speed_changed(slave); | |
3291 | if (old_duplex != slave->duplex) | |
3292 | bond_3ad_adapter_duplex_changed(slave); | |
3293 | } | |
3294 | } | |
3295 | ||
1da177e4 LT |
3296 | break; |
3297 | case NETDEV_DOWN: | |
3298 | /* | |
3299 | * ... Or is it this? | |
3300 | */ | |
3301 | break; | |
3302 | case NETDEV_CHANGEMTU: | |
3303 | /* | |
3304 | * TODO: Should slaves be allowed to | |
3305 | * independently alter their MTU? For | |
3306 | * an active-backup bond, slaves need | |
3307 | * not be the same type of device, so | |
3308 | * MTUs may vary. For other modes, | |
3309 | * slaves arguably should have the | |
3310 | * same MTUs. To do this, we'd need to | |
3311 | * take over the slave's change_mtu | |
3312 | * function for the duration of their | |
3313 | * servitude. | |
3314 | */ | |
3315 | break; | |
3316 | case NETDEV_CHANGENAME: | |
3317 | /* | |
3318 | * TODO: handle changing the primary's name | |
3319 | */ | |
3320 | break; | |
8531c5ff AK |
3321 | case NETDEV_FEAT_CHANGE: |
3322 | bond_compute_features(bond); | |
3323 | break; | |
1da177e4 LT |
3324 | default: |
3325 | break; | |
3326 | } | |
3327 | ||
3328 | return NOTIFY_DONE; | |
3329 | } | |
3330 | ||
3331 | /* | |
3332 | * bond_netdev_event: handle netdev notifier chain events. | |
3333 | * | |
3334 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3335 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3336 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3337 | * dev_probe_lock). | |
3338 | */ | |
3d632c3f SH |
3339 | static int bond_netdev_event(struct notifier_block *this, |
3340 | unsigned long event, void *ptr) | |
1da177e4 LT |
3341 | { |
3342 | struct net_device *event_dev = (struct net_device *)ptr; | |
3343 | ||
5a03cdb7 | 3344 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3345 | event_dev ? event_dev->name : "None", |
3346 | event); | |
1da177e4 | 3347 | |
0b680e75 JV |
3348 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3349 | return NOTIFY_DONE; | |
3350 | ||
1da177e4 | 3351 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3352 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3353 | return bond_master_netdev_event(event, event_dev); |
3354 | } | |
3355 | ||
3356 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3357 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3358 | return bond_slave_netdev_event(event, event_dev); |
3359 | } | |
3360 | ||
3361 | return NOTIFY_DONE; | |
3362 | } | |
3363 | ||
3364 | static struct notifier_block bond_netdev_notifier = { | |
3365 | .notifier_call = bond_netdev_event, | |
3366 | }; | |
3367 | ||
169a3e66 JV |
3368 | /*---------------------------- Hashing Policies -----------------------------*/ |
3369 | ||
6b923cb7 JE |
3370 | /* |
3371 | * Hash for the output device based upon layer 2 data | |
3372 | */ | |
3373 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) | |
3374 | { | |
3375 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3376 | ||
3377 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
3378 | return (data->h_dest[5] ^ data->h_source[5]) % count; | |
3379 | ||
3380 | return 0; | |
3381 | } | |
3382 | ||
6f6652be JV |
3383 | /* |
3384 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
6b923cb7 | 3385 | * the packet is not IP, fall back on bond_xmit_hash_policy_l2() |
6f6652be | 3386 | */ |
a361c83c | 3387 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3388 | { |
3389 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
6b923cb7 JE |
3390 | struct iphdr *iph; |
3391 | struct ipv6hdr *ipv6h; | |
3392 | u32 v6hash; | |
3393 | __be32 *s, *d; | |
3394 | ||
3395 | if (skb->protocol == htons(ETH_P_IP) && | |
3396 | skb_network_header_len(skb) >= sizeof(*iph)) { | |
3397 | iph = ip_hdr(skb); | |
6f6652be | 3398 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3399 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6b923cb7 JE |
3400 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
3401 | skb_network_header_len(skb) >= sizeof(*ipv6h)) { | |
3402 | ipv6h = ipv6_hdr(skb); | |
3403 | s = &ipv6h->saddr.s6_addr32[0]; | |
3404 | d = &ipv6h->daddr.s6_addr32[0]; | |
3405 | v6hash = (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3406 | v6hash ^= (v6hash >> 24) ^ (v6hash >> 16) ^ (v6hash >> 8); | |
3407 | return (v6hash ^ data->h_dest[5] ^ data->h_source[5]) % count; | |
6f6652be JV |
3408 | } |
3409 | ||
6b923cb7 | 3410 | return bond_xmit_hash_policy_l2(skb, count); |
6f6652be JV |
3411 | } |
3412 | ||
169a3e66 | 3413 | /* |
59c51591 | 3414 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 | 3415 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
6b923cb7 | 3416 | * altogether not IP, fall back on bond_xmit_hash_policy_l2() |
169a3e66 | 3417 | */ |
a361c83c | 3418 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 | 3419 | { |
6b923cb7 JE |
3420 | u32 layer4_xor = 0; |
3421 | struct iphdr *iph; | |
3422 | struct ipv6hdr *ipv6h; | |
3423 | __be32 *s, *d; | |
3424 | __be16 *layer4hdr; | |
169a3e66 | 3425 | |
6b923cb7 JE |
3426 | if (skb->protocol == htons(ETH_P_IP) && |
3427 | skb_network_header_len(skb) >= sizeof(*iph)) { | |
3428 | iph = ip_hdr(skb); | |
56f8a75c | 3429 | if (!ip_is_fragment(iph) && |
169a3e66 | 3430 | (iph->protocol == IPPROTO_TCP || |
6b923cb7 JE |
3431 | iph->protocol == IPPROTO_UDP) && |
3432 | (skb_headlen(skb) - skb_network_offset(skb) >= | |
3433 | iph->ihl * sizeof(u32) + sizeof(*layer4hdr) * 2)) { | |
3434 | layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); | |
3435 | layer4_xor = ntohs(*layer4hdr ^ *(layer4hdr + 1)); | |
169a3e66 JV |
3436 | } |
3437 | return (layer4_xor ^ | |
3438 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
6b923cb7 JE |
3439 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
3440 | skb_network_header_len(skb) >= sizeof(*ipv6h)) { | |
3441 | ipv6h = ipv6_hdr(skb); | |
3442 | if ((ipv6h->nexthdr == IPPROTO_TCP || | |
3443 | ipv6h->nexthdr == IPPROTO_UDP) && | |
3444 | (skb_headlen(skb) - skb_network_offset(skb) >= | |
3445 | sizeof(*ipv6h) + sizeof(*layer4hdr) * 2)) { | |
3446 | layer4hdr = (__be16 *)(ipv6h + 1); | |
3447 | layer4_xor = ntohs(*layer4hdr ^ *(layer4hdr + 1)); | |
3448 | } | |
3449 | s = &ipv6h->saddr.s6_addr32[0]; | |
3450 | d = &ipv6h->daddr.s6_addr32[0]; | |
3451 | layer4_xor ^= (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3452 | layer4_xor ^= (layer4_xor >> 24) ^ (layer4_xor >> 16) ^ | |
3453 | (layer4_xor >> 8); | |
3454 | return layer4_xor % count; | |
169a3e66 JV |
3455 | } |
3456 | ||
6b923cb7 | 3457 | return bond_xmit_hash_policy_l2(skb, count); |
169a3e66 JV |
3458 | } |
3459 | ||
1da177e4 LT |
3460 | /*-------------------------- Device entry points ----------------------------*/ |
3461 | ||
3462 | static int bond_open(struct net_device *bond_dev) | |
3463 | { | |
454d7c9b | 3464 | struct bonding *bond = netdev_priv(bond_dev); |
ba3211cc PP |
3465 | struct slave *slave; |
3466 | int i; | |
1da177e4 | 3467 | |
ba3211cc PP |
3468 | /* reset slave->backup and slave->inactive */ |
3469 | read_lock(&bond->lock); | |
3470 | if (bond->slave_cnt > 0) { | |
3471 | read_lock(&bond->curr_slave_lock); | |
3472 | bond_for_each_slave(bond, slave, i) { | |
3473 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3474 | && (slave != bond->curr_active_slave)) { | |
3475 | bond_set_slave_inactive_flags(slave); | |
3476 | } else { | |
3477 | bond_set_slave_active_flags(slave); | |
3478 | } | |
3479 | } | |
3480 | read_unlock(&bond->curr_slave_lock); | |
3481 | } | |
3482 | read_unlock(&bond->lock); | |
3483 | ||
5a37e8ca FL |
3484 | INIT_DELAYED_WORK(&bond->mcast_work, bond_resend_igmp_join_requests_delayed); |
3485 | ||
58402054 | 3486 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3487 | /* bond_alb_initialize must be called before the timer |
3488 | * is started. | |
3489 | */ | |
3490 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3491 | /* something went wrong - fail the open operation */ | |
b473946a | 3492 | return -ENOMEM; |
1da177e4 LT |
3493 | } |
3494 | ||
1b76b316 JV |
3495 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3496 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3497 | } |
3498 | ||
3499 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3500 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3501 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3502 | } |
3503 | ||
3504 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3505 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3506 | INIT_DELAYED_WORK(&bond->arp_work, | |
3507 | bond_activebackup_arp_mon); | |
3508 | else | |
3509 | INIT_DELAYED_WORK(&bond->arp_work, | |
3510 | bond_loadbalance_arp_mon); | |
3511 | ||
3512 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 | 3513 | if (bond->params.arp_validate) |
3aba891d | 3514 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3515 | } |
3516 | ||
3517 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3518 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3519 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3520 | /* register to receive LACPDUs */ |
3aba891d | 3521 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3522 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3523 | } |
3524 | ||
3525 | return 0; | |
3526 | } | |
3527 | ||
3528 | static int bond_close(struct net_device *bond_dev) | |
3529 | { | |
454d7c9b | 3530 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3531 | |
1da177e4 LT |
3532 | write_lock_bh(&bond->lock); |
3533 | ||
ad246c99 BH |
3534 | bond->send_peer_notif = 0; |
3535 | ||
1da177e4 LT |
3536 | write_unlock_bh(&bond->lock); |
3537 | ||
1da177e4 | 3538 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
e6d265e8 | 3539 | cancel_delayed_work_sync(&bond->mii_work); |
1da177e4 LT |
3540 | } |
3541 | ||
3542 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
e6d265e8 | 3543 | cancel_delayed_work_sync(&bond->arp_work); |
1da177e4 LT |
3544 | } |
3545 | ||
3546 | switch (bond->params.mode) { | |
3547 | case BOND_MODE_8023AD: | |
e6d265e8 | 3548 | cancel_delayed_work_sync(&bond->ad_work); |
1da177e4 LT |
3549 | break; |
3550 | case BOND_MODE_TLB: | |
3551 | case BOND_MODE_ALB: | |
e6d265e8 | 3552 | cancel_delayed_work_sync(&bond->alb_work); |
1da177e4 LT |
3553 | break; |
3554 | default: | |
3555 | break; | |
3556 | } | |
3557 | ||
5a37e8ca | 3558 | if (delayed_work_pending(&bond->mcast_work)) |
e6d265e8 | 3559 | cancel_delayed_work_sync(&bond->mcast_work); |
1da177e4 | 3560 | |
58402054 | 3561 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3562 | /* Must be called only after all |
3563 | * slaves have been released | |
3564 | */ | |
3565 | bond_alb_deinitialize(bond); | |
3566 | } | |
3aba891d | 3567 | bond->recv_probe = NULL; |
1da177e4 LT |
3568 | |
3569 | return 0; | |
3570 | } | |
3571 | ||
28172739 ED |
3572 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3573 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3574 | { |
454d7c9b | 3575 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3576 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3577 | struct slave *slave; |
3578 | int i; | |
3579 | ||
28172739 | 3580 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3581 | |
3582 | read_lock_bh(&bond->lock); | |
3583 | ||
3584 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3585 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3586 | dev_get_stats(slave->dev, &temp); |
3587 | ||
3588 | stats->rx_packets += sstats->rx_packets; | |
3589 | stats->rx_bytes += sstats->rx_bytes; | |
3590 | stats->rx_errors += sstats->rx_errors; | |
3591 | stats->rx_dropped += sstats->rx_dropped; | |
3592 | ||
3593 | stats->tx_packets += sstats->tx_packets; | |
3594 | stats->tx_bytes += sstats->tx_bytes; | |
3595 | stats->tx_errors += sstats->tx_errors; | |
3596 | stats->tx_dropped += sstats->tx_dropped; | |
3597 | ||
3598 | stats->multicast += sstats->multicast; | |
3599 | stats->collisions += sstats->collisions; | |
3600 | ||
3601 | stats->rx_length_errors += sstats->rx_length_errors; | |
3602 | stats->rx_over_errors += sstats->rx_over_errors; | |
3603 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3604 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3605 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3606 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3607 | ||
3608 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3609 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3610 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3611 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3612 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3613 | } |
3614 | ||
1da177e4 LT |
3615 | read_unlock_bh(&bond->lock); |
3616 | ||
3617 | return stats; | |
3618 | } | |
3619 | ||
3620 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3621 | { | |
3622 | struct net_device *slave_dev = NULL; | |
3623 | struct ifbond k_binfo; | |
3624 | struct ifbond __user *u_binfo = NULL; | |
3625 | struct ifslave k_sinfo; | |
3626 | struct ifslave __user *u_sinfo = NULL; | |
3627 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3628 | int res = 0; |
3629 | ||
a4aee5c8 | 3630 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3631 | |
3632 | switch (cmd) { | |
1da177e4 LT |
3633 | case SIOCGMIIPHY: |
3634 | mii = if_mii(ifr); | |
3d632c3f | 3635 | if (!mii) |
1da177e4 | 3636 | return -EINVAL; |
3d632c3f | 3637 | |
1da177e4 LT |
3638 | mii->phy_id = 0; |
3639 | /* Fall Through */ | |
3640 | case SIOCGMIIREG: | |
3641 | /* | |
3642 | * We do this again just in case we were called by SIOCGMIIREG | |
3643 | * instead of SIOCGMIIPHY. | |
3644 | */ | |
3645 | mii = if_mii(ifr); | |
3d632c3f | 3646 | if (!mii) |
1da177e4 | 3647 | return -EINVAL; |
3d632c3f | 3648 | |
1da177e4 LT |
3649 | |
3650 | if (mii->reg_num == 1) { | |
454d7c9b | 3651 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3652 | mii->val_out = 0; |
6603a6f2 | 3653 | read_lock(&bond->lock); |
1da177e4 | 3654 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3655 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3656 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3657 | |
1da177e4 | 3658 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3659 | read_unlock(&bond->lock); |
1da177e4 LT |
3660 | } |
3661 | ||
3662 | return 0; | |
3663 | case BOND_INFO_QUERY_OLD: | |
3664 | case SIOCBONDINFOQUERY: | |
3665 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3666 | ||
3d632c3f | 3667 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3668 | return -EFAULT; |
1da177e4 LT |
3669 | |
3670 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3671 | if (res == 0 && |
3672 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3673 | return -EFAULT; | |
1da177e4 LT |
3674 | |
3675 | return res; | |
3676 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3677 | case SIOCBONDSLAVEINFOQUERY: | |
3678 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3679 | ||
3d632c3f | 3680 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3681 | return -EFAULT; |
1da177e4 LT |
3682 | |
3683 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3684 | if (res == 0 && |
3685 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3686 | return -EFAULT; | |
1da177e4 LT |
3687 | |
3688 | return res; | |
3689 | default: | |
3690 | /* Go on */ | |
3691 | break; | |
3692 | } | |
3693 | ||
3d632c3f | 3694 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3695 | return -EPERM; |
1da177e4 | 3696 | |
ec87fd3b | 3697 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3698 | |
a4aee5c8 | 3699 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3700 | |
3d632c3f | 3701 | if (!slave_dev) |
1da177e4 | 3702 | res = -ENODEV; |
3d632c3f | 3703 | else { |
a4aee5c8 | 3704 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3705 | switch (cmd) { |
3706 | case BOND_ENSLAVE_OLD: | |
3707 | case SIOCBONDENSLAVE: | |
3708 | res = bond_enslave(bond_dev, slave_dev); | |
3709 | break; | |
3710 | case BOND_RELEASE_OLD: | |
3711 | case SIOCBONDRELEASE: | |
3712 | res = bond_release(bond_dev, slave_dev); | |
3713 | break; | |
3714 | case BOND_SETHWADDR_OLD: | |
3715 | case SIOCBONDSETHWADDR: | |
3716 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3717 | break; | |
3718 | case BOND_CHANGE_ACTIVE_OLD: | |
3719 | case SIOCBONDCHANGEACTIVE: | |
3720 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3721 | break; | |
3722 | default: | |
3723 | res = -EOPNOTSUPP; | |
3724 | } | |
3725 | ||
3726 | dev_put(slave_dev); | |
3727 | } | |
3728 | ||
1da177e4 LT |
3729 | return res; |
3730 | } | |
3731 | ||
22bedad3 JP |
3732 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3733 | struct netdev_hw_addr_list *list, | |
3734 | int addrlen) | |
3735 | { | |
3736 | struct netdev_hw_addr *ha; | |
3737 | ||
3738 | netdev_hw_addr_list_for_each(ha, list) | |
3739 | if (!memcmp(ha->addr, addr, addrlen)) | |
3740 | return true; | |
3741 | ||
3742 | return false; | |
3743 | } | |
3744 | ||
d03462b9 | 3745 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3746 | { |
454d7c9b | 3747 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3748 | |
d03462b9 JP |
3749 | if (change & IFF_PROMISC) |
3750 | bond_set_promiscuity(bond, | |
3751 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3752 | |
d03462b9 JP |
3753 | if (change & IFF_ALLMULTI) |
3754 | bond_set_allmulti(bond, | |
3755 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3756 | } | |
1da177e4 | 3757 | |
d03462b9 JP |
3758 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3759 | { | |
3760 | struct bonding *bond = netdev_priv(bond_dev); | |
3761 | struct netdev_hw_addr *ha; | |
3762 | bool found; | |
1da177e4 | 3763 | |
80ee5ad2 JV |
3764 | read_lock(&bond->lock); |
3765 | ||
1da177e4 | 3766 | /* looking for addresses to add to slaves' mc list */ |
22bedad3 JP |
3767 | netdev_for_each_mc_addr(ha, bond_dev) { |
3768 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
3769 | bond_dev->addr_len); | |
3770 | if (!found) | |
3771 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
3772 | } |
3773 | ||
3774 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
3775 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
3776 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
3777 | bond_dev->addr_len); | |
3778 | if (!found) | |
3779 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
3780 | } |
3781 | ||
3782 | /* save master's multicast list */ | |
22bedad3 JP |
3783 | __hw_addr_flush(&bond->mc_list); |
3784 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
3785 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 3786 | |
80ee5ad2 | 3787 | read_unlock(&bond->lock); |
1da177e4 LT |
3788 | } |
3789 | ||
234bcf8a | 3790 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3791 | { |
234bcf8a | 3792 | struct bonding *bond = netdev_priv(n->dev); |
00829823 | 3793 | struct slave *slave = bond->first_slave; |
234bcf8a SP |
3794 | const struct net_device_ops *slave_ops; |
3795 | struct neigh_parms parms; | |
3796 | int ret; | |
3797 | ||
3798 | if (!slave) | |
3799 | return 0; | |
3800 | ||
3801 | slave_ops = slave->dev->netdev_ops; | |
3802 | ||
3803 | if (!slave_ops->ndo_neigh_setup) | |
3804 | return 0; | |
3805 | ||
3806 | parms.neigh_setup = NULL; | |
3807 | parms.neigh_cleanup = NULL; | |
3808 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3809 | if (ret) | |
3810 | return ret; | |
3811 | ||
3812 | /* | |
3813 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3814 | * after the last slave has been detached. Assumes that all slaves | |
3815 | * utilize the same neigh_cleanup (true at this writing as only user | |
3816 | * is ipoib). | |
3817 | */ | |
3818 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3819 | ||
3820 | if (!parms.neigh_setup) | |
3821 | return 0; | |
3822 | ||
3823 | return parms.neigh_setup(n); | |
3824 | } | |
3825 | ||
3826 | /* | |
3827 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3828 | * slave exists. So we must declare proxy setup function which will | |
3829 | * be used at run time to resolve the actual slave neigh param setup. | |
3830 | */ | |
3831 | static int bond_neigh_setup(struct net_device *dev, | |
3832 | struct neigh_parms *parms) | |
3833 | { | |
3834 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3835 | |
00829823 SH |
3836 | return 0; |
3837 | } | |
3838 | ||
1da177e4 LT |
3839 | /* |
3840 | * Change the MTU of all of a master's slaves to match the master | |
3841 | */ | |
3842 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3843 | { | |
454d7c9b | 3844 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3845 | struct slave *slave, *stop_at; |
3846 | int res = 0; | |
3847 | int i; | |
3848 | ||
5a03cdb7 | 3849 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3850 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3851 | |
3852 | /* Can't hold bond->lock with bh disabled here since | |
3853 | * some base drivers panic. On the other hand we can't | |
3854 | * hold bond->lock without bh disabled because we'll | |
3855 | * deadlock. The only solution is to rely on the fact | |
3856 | * that we're under rtnl_lock here, and the slaves | |
3857 | * list won't change. This doesn't solve the problem | |
3858 | * of setting the slave's MTU while it is | |
3859 | * transmitting, but the assumption is that the base | |
3860 | * driver can handle that. | |
3861 | * | |
3862 | * TODO: figure out a way to safely iterate the slaves | |
3863 | * list, but without holding a lock around the actual | |
3864 | * call to the base driver. | |
3865 | */ | |
3866 | ||
3867 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3868 | pr_debug("s %p s->p %p c_m %p\n", |
3869 | slave, | |
3870 | slave->prev, | |
3871 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3872 | |
1da177e4 LT |
3873 | res = dev_set_mtu(slave->dev, new_mtu); |
3874 | ||
3875 | if (res) { | |
3876 | /* If we failed to set the slave's mtu to the new value | |
3877 | * we must abort the operation even in ACTIVE_BACKUP | |
3878 | * mode, because if we allow the backup slaves to have | |
3879 | * different mtu values than the active slave we'll | |
3880 | * need to change their mtu when doing a failover. That | |
3881 | * means changing their mtu from timer context, which | |
3882 | * is probably not a good idea. | |
3883 | */ | |
5a03cdb7 | 3884 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3885 | goto unwind; |
3886 | } | |
3887 | } | |
3888 | ||
3889 | bond_dev->mtu = new_mtu; | |
3890 | ||
3891 | return 0; | |
3892 | ||
3893 | unwind: | |
3894 | /* unwind from head to the slave that failed */ | |
3895 | stop_at = slave; | |
3896 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3897 | int tmp_res; | |
3898 | ||
3899 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3900 | if (tmp_res) { | |
a4aee5c8 JP |
3901 | pr_debug("unwind err %d dev %s\n", |
3902 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3903 | } |
3904 | } | |
3905 | ||
3906 | return res; | |
3907 | } | |
3908 | ||
3909 | /* | |
3910 | * Change HW address | |
3911 | * | |
3912 | * Note that many devices must be down to change the HW address, and | |
3913 | * downing the master releases all slaves. We can make bonds full of | |
3914 | * bonding devices to test this, however. | |
3915 | */ | |
3916 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3917 | { | |
454d7c9b | 3918 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3919 | struct sockaddr *sa = addr, tmp_sa; |
3920 | struct slave *slave, *stop_at; | |
3921 | int res = 0; | |
3922 | int i; | |
3923 | ||
eb7cc59a SH |
3924 | if (bond->params.mode == BOND_MODE_ALB) |
3925 | return bond_alb_set_mac_address(bond_dev, addr); | |
3926 | ||
3927 | ||
a4aee5c8 JP |
3928 | pr_debug("bond=%p, name=%s\n", |
3929 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3930 | |
dd957c57 | 3931 | /* |
3915c1e8 JV |
3932 | * If fail_over_mac is set to active, do nothing and return |
3933 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 3934 | */ |
3915c1e8 | 3935 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 3936 | return 0; |
2ab82852 | 3937 | |
3d632c3f | 3938 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3939 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3940 | |
3941 | /* Can't hold bond->lock with bh disabled here since | |
3942 | * some base drivers panic. On the other hand we can't | |
3943 | * hold bond->lock without bh disabled because we'll | |
3944 | * deadlock. The only solution is to rely on the fact | |
3945 | * that we're under rtnl_lock here, and the slaves | |
3946 | * list won't change. This doesn't solve the problem | |
3947 | * of setting the slave's hw address while it is | |
3948 | * transmitting, but the assumption is that the base | |
3949 | * driver can handle that. | |
3950 | * | |
3951 | * TODO: figure out a way to safely iterate the slaves | |
3952 | * list, but without holding a lock around the actual | |
3953 | * call to the base driver. | |
3954 | */ | |
3955 | ||
3956 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 3957 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3958 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3959 | |
eb7cc59a | 3960 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3961 | res = -EOPNOTSUPP; |
5a03cdb7 | 3962 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3963 | goto unwind; |
3964 | } | |
3965 | ||
3966 | res = dev_set_mac_address(slave->dev, addr); | |
3967 | if (res) { | |
3968 | /* TODO: consider downing the slave | |
3969 | * and retry ? | |
3970 | * User should expect communications | |
3971 | * breakage anyway until ARP finish | |
3972 | * updating, so... | |
3973 | */ | |
5a03cdb7 | 3974 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3975 | goto unwind; |
3976 | } | |
3977 | } | |
3978 | ||
3979 | /* success */ | |
3980 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3981 | return 0; | |
3982 | ||
3983 | unwind: | |
3984 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3985 | tmp_sa.sa_family = bond_dev->type; | |
3986 | ||
3987 | /* unwind from head to the slave that failed */ | |
3988 | stop_at = slave; | |
3989 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3990 | int tmp_res; | |
3991 | ||
3992 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
3993 | if (tmp_res) { | |
a4aee5c8 JP |
3994 | pr_debug("unwind err %d dev %s\n", |
3995 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3996 | } |
3997 | } | |
3998 | ||
3999 | return res; | |
4000 | } | |
4001 | ||
4002 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4003 | { | |
454d7c9b | 4004 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4005 | struct slave *slave, *start_at; |
cf5f9044 | 4006 | int i, slave_no, res = 1; |
a2fd940f | 4007 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 4008 | |
cf5f9044 | 4009 | /* |
a2fd940f AG |
4010 | * Start with the curr_active_slave that joined the bond as the |
4011 | * default for sending IGMP traffic. For failover purposes one | |
4012 | * needs to maintain some consistency for the interface that will | |
4013 | * send the join/membership reports. The curr_active_slave found | |
4014 | * will send all of this type of traffic. | |
cf5f9044 | 4015 | */ |
00ae7028 | 4016 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4017 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4018 | |
a2fd940f AG |
4019 | read_lock(&bond->curr_slave_lock); |
4020 | slave = bond->curr_active_slave; | |
4021 | read_unlock(&bond->curr_slave_lock); | |
4022 | ||
4023 | if (!slave) | |
4024 | goto out; | |
4025 | } else { | |
4026 | /* | |
4027 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4028 | * that as being rare enough not to justify using an | |
4029 | * atomic op here. | |
4030 | */ | |
4031 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4032 | ||
4033 | bond_for_each_slave(bond, slave, i) { | |
4034 | slave_no--; | |
4035 | if (slave_no < 0) | |
4036 | break; | |
4037 | } | |
1da177e4 LT |
4038 | } |
4039 | ||
cf5f9044 | 4040 | start_at = slave; |
1da177e4 LT |
4041 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4042 | if (IS_UP(slave->dev) && | |
4043 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4044 | bond_is_active_slave(slave)) { |
1da177e4 | 4045 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
4046 | break; |
4047 | } | |
4048 | } | |
4049 | ||
1da177e4 LT |
4050 | out: |
4051 | if (res) { | |
4052 | /* no suitable interface, frame not sent */ | |
04502430 | 4053 | kfree_skb(skb); |
1da177e4 | 4054 | } |
0693e88e | 4055 | |
ec634fe3 | 4056 | return NETDEV_TX_OK; |
1da177e4 LT |
4057 | } |
4058 | ||
075897ce | 4059 | |
1da177e4 LT |
4060 | /* |
4061 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4062 | * the bond has a usable interface. | |
4063 | */ | |
4064 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4065 | { | |
454d7c9b | 4066 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4067 | int res = 1; |
4068 | ||
1da177e4 LT |
4069 | read_lock(&bond->curr_slave_lock); |
4070 | ||
0693e88e MM |
4071 | if (bond->curr_active_slave) |
4072 | res = bond_dev_queue_xmit(bond, skb, | |
4073 | bond->curr_active_slave->dev); | |
1da177e4 | 4074 | |
04502430 ED |
4075 | read_unlock(&bond->curr_slave_lock); |
4076 | ||
3d632c3f | 4077 | if (res) |
1da177e4 | 4078 | /* no suitable interface, frame not sent */ |
04502430 | 4079 | kfree_skb(skb); |
0693e88e | 4080 | |
ec634fe3 | 4081 | return NETDEV_TX_OK; |
1da177e4 LT |
4082 | } |
4083 | ||
4084 | /* | |
169a3e66 JV |
4085 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4086 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4087 | * find the next active slave. | |
1da177e4 LT |
4088 | */ |
4089 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4090 | { | |
454d7c9b | 4091 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4092 | struct slave *slave, *start_at; |
4093 | int slave_no; | |
4094 | int i; | |
4095 | int res = 1; | |
4096 | ||
a361c83c | 4097 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4098 | |
4099 | bond_for_each_slave(bond, slave, i) { | |
4100 | slave_no--; | |
3d632c3f | 4101 | if (slave_no < 0) |
1da177e4 | 4102 | break; |
1da177e4 LT |
4103 | } |
4104 | ||
4105 | start_at = slave; | |
4106 | ||
4107 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4108 | if (IS_UP(slave->dev) && | |
4109 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4110 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4111 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
4112 | break; | |
4113 | } | |
4114 | } | |
4115 | ||
1da177e4 LT |
4116 | if (res) { |
4117 | /* no suitable interface, frame not sent */ | |
04502430 | 4118 | kfree_skb(skb); |
1da177e4 | 4119 | } |
0693e88e | 4120 | |
ec634fe3 | 4121 | return NETDEV_TX_OK; |
1da177e4 LT |
4122 | } |
4123 | ||
4124 | /* | |
4125 | * in broadcast mode, we send everything to all usable interfaces. | |
4126 | */ | |
4127 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4128 | { | |
454d7c9b | 4129 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4130 | struct slave *slave, *start_at; |
4131 | struct net_device *tx_dev = NULL; | |
4132 | int i; | |
4133 | int res = 1; | |
4134 | ||
1da177e4 LT |
4135 | read_lock(&bond->curr_slave_lock); |
4136 | start_at = bond->curr_active_slave; | |
4137 | read_unlock(&bond->curr_slave_lock); | |
4138 | ||
3d632c3f | 4139 | if (!start_at) |
1da177e4 | 4140 | goto out; |
1da177e4 LT |
4141 | |
4142 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4143 | if (IS_UP(slave->dev) && | |
4144 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4145 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4146 | if (tx_dev) { |
4147 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4148 | if (!skb2) { | |
a4aee5c8 | 4149 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4150 | bond_dev->name); |
1da177e4 LT |
4151 | continue; |
4152 | } | |
4153 | ||
4154 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4155 | if (res) { | |
04502430 | 4156 | kfree_skb(skb2); |
1da177e4 LT |
4157 | continue; |
4158 | } | |
4159 | } | |
4160 | tx_dev = slave->dev; | |
4161 | } | |
4162 | } | |
4163 | ||
3d632c3f | 4164 | if (tx_dev) |
1da177e4 | 4165 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4166 | |
4167 | out: | |
3d632c3f | 4168 | if (res) |
1da177e4 | 4169 | /* no suitable interface, frame not sent */ |
04502430 | 4170 | kfree_skb(skb); |
3d632c3f | 4171 | |
1da177e4 | 4172 | /* frame sent to all suitable interfaces */ |
ec634fe3 | 4173 | return NETDEV_TX_OK; |
1da177e4 LT |
4174 | } |
4175 | ||
4176 | /*------------------------- Device initialization ---------------------------*/ | |
4177 | ||
6f6652be JV |
4178 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4179 | { | |
4180 | switch (bond->params.xmit_policy) { | |
4181 | case BOND_XMIT_POLICY_LAYER23: | |
4182 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4183 | break; | |
4184 | case BOND_XMIT_POLICY_LAYER34: | |
4185 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4186 | break; | |
4187 | case BOND_XMIT_POLICY_LAYER2: | |
4188 | default: | |
4189 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4190 | break; | |
4191 | } | |
4192 | } | |
4193 | ||
bb1d9123 AG |
4194 | /* |
4195 | * Lookup the slave that corresponds to a qid | |
4196 | */ | |
4197 | static inline int bond_slave_override(struct bonding *bond, | |
4198 | struct sk_buff *skb) | |
4199 | { | |
4200 | int i, res = 1; | |
4201 | struct slave *slave = NULL; | |
4202 | struct slave *check_slave; | |
4203 | ||
0693e88e MM |
4204 | if (!skb->queue_mapping) |
4205 | return 1; | |
bb1d9123 AG |
4206 | |
4207 | /* Find out if any slaves have the same mapping as this skb. */ | |
4208 | bond_for_each_slave(bond, check_slave, i) { | |
4209 | if (check_slave->queue_id == skb->queue_mapping) { | |
4210 | slave = check_slave; | |
4211 | break; | |
4212 | } | |
4213 | } | |
4214 | ||
4215 | /* If the slave isn't UP, use default transmit policy. */ | |
4216 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4217 | (slave->link == BOND_LINK_UP)) { | |
4218 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4219 | } | |
4220 | ||
bb1d9123 AG |
4221 | return res; |
4222 | } | |
4223 | ||
374eeb5a | 4224 | |
bb1d9123 AG |
4225 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
4226 | { | |
4227 | /* | |
4228 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4229 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4230 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4231 | * way down to the bonding driver. | |
4232 | */ | |
fd0e435b PO |
4233 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4234 | ||
374eeb5a NH |
4235 | /* |
4236 | * Save the original txq to restore before passing to the driver | |
4237 | */ | |
df4ab5b3 | 4238 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 4239 | |
fd0e435b | 4240 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4241 | do { |
fd0e435b | 4242 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4243 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4244 | } |
4245 | return txq; | |
bb1d9123 AG |
4246 | } |
4247 | ||
0693e88e | 4248 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4249 | { |
bb1d9123 AG |
4250 | struct bonding *bond = netdev_priv(dev); |
4251 | ||
4252 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
4253 | if (!bond_slave_override(bond, skb)) | |
4254 | return NETDEV_TX_OK; | |
4255 | } | |
00829823 SH |
4256 | |
4257 | switch (bond->params.mode) { | |
4258 | case BOND_MODE_ROUNDROBIN: | |
4259 | return bond_xmit_roundrobin(skb, dev); | |
4260 | case BOND_MODE_ACTIVEBACKUP: | |
4261 | return bond_xmit_activebackup(skb, dev); | |
4262 | case BOND_MODE_XOR: | |
4263 | return bond_xmit_xor(skb, dev); | |
4264 | case BOND_MODE_BROADCAST: | |
4265 | return bond_xmit_broadcast(skb, dev); | |
4266 | case BOND_MODE_8023AD: | |
4267 | return bond_3ad_xmit_xor(skb, dev); | |
4268 | case BOND_MODE_ALB: | |
4269 | case BOND_MODE_TLB: | |
4270 | return bond_alb_xmit(skb, dev); | |
4271 | default: | |
4272 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4273 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4274 | dev->name, bond->params.mode); | |
00829823 | 4275 | WARN_ON_ONCE(1); |
04502430 | 4276 | kfree_skb(skb); |
00829823 SH |
4277 | return NETDEV_TX_OK; |
4278 | } | |
4279 | } | |
4280 | ||
0693e88e MM |
4281 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4282 | { | |
4283 | struct bonding *bond = netdev_priv(dev); | |
4284 | netdev_tx_t ret = NETDEV_TX_OK; | |
4285 | ||
4286 | /* | |
4287 | * If we risk deadlock from transmitting this in the | |
4288 | * netpoll path, tell netpoll to queue the frame for later tx | |
4289 | */ | |
4290 | if (is_netpoll_tx_blocked(dev)) | |
4291 | return NETDEV_TX_BUSY; | |
4292 | ||
4293 | read_lock(&bond->lock); | |
4294 | ||
4295 | if (bond->slave_cnt) | |
4296 | ret = __bond_start_xmit(skb, dev); | |
4297 | else | |
04502430 | 4298 | kfree_skb(skb); |
0693e88e MM |
4299 | |
4300 | read_unlock(&bond->lock); | |
4301 | ||
4302 | return ret; | |
4303 | } | |
00829823 | 4304 | |
1da177e4 LT |
4305 | /* |
4306 | * set bond mode specific net device operations | |
4307 | */ | |
a77b5325 | 4308 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4309 | { |
169a3e66 JV |
4310 | struct net_device *bond_dev = bond->dev; |
4311 | ||
1da177e4 LT |
4312 | switch (mode) { |
4313 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4314 | break; |
4315 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4316 | break; |
4317 | case BOND_MODE_XOR: | |
6f6652be | 4318 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4319 | break; |
4320 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4321 | break; |
4322 | case BOND_MODE_8023AD: | |
6f6652be | 4323 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4324 | break; |
1da177e4 | 4325 | case BOND_MODE_ALB: |
8f903c70 JV |
4326 | /* FALLTHRU */ |
4327 | case BOND_MODE_TLB: | |
1da177e4 LT |
4328 | break; |
4329 | default: | |
4330 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4331 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4332 | bond_dev->name, mode); | |
1da177e4 LT |
4333 | break; |
4334 | } | |
4335 | } | |
4336 | ||
217df670 JV |
4337 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4338 | struct ethtool_drvinfo *drvinfo) | |
4339 | { | |
4340 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4341 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4342 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4343 | } | |
4344 | ||
7282d491 | 4345 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4346 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac | 4347 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
4348 | }; |
4349 | ||
eb7cc59a | 4350 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4351 | .ndo_init = bond_init, |
9e71626c | 4352 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4353 | .ndo_open = bond_open, |
4354 | .ndo_stop = bond_close, | |
00829823 | 4355 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4356 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4357 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4358 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4359 | .ndo_change_rx_flags = bond_change_rx_flags, |
afc4b13d | 4360 | .ndo_set_rx_mode = bond_set_multicast_list, |
eb7cc59a | 4361 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4362 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4363 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4364 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4365 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4366 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4367 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4368 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4369 | .ndo_poll_controller = bond_poll_controller, | |
4370 | #endif | |
9232ecca JP |
4371 | .ndo_add_slave = bond_enslave, |
4372 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4373 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
4374 | }; |
4375 | ||
9e2e61fb AW |
4376 | static void bond_destructor(struct net_device *bond_dev) |
4377 | { | |
4378 | struct bonding *bond = netdev_priv(bond_dev); | |
4379 | if (bond->wq) | |
4380 | destroy_workqueue(bond->wq); | |
4381 | free_netdev(bond_dev); | |
4382 | } | |
4383 | ||
181470fc | 4384 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4385 | { |
454d7c9b | 4386 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4387 | |
1da177e4 LT |
4388 | /* initialize rwlocks */ |
4389 | rwlock_init(&bond->lock); | |
4390 | rwlock_init(&bond->curr_slave_lock); | |
4391 | ||
d2991f75 | 4392 | bond->params = bonding_defaults; |
1da177e4 LT |
4393 | |
4394 | /* Initialize pointers */ | |
1da177e4 LT |
4395 | bond->dev = bond_dev; |
4396 | INIT_LIST_HEAD(&bond->vlan_list); | |
4397 | ||
4398 | /* Initialize the device entry points */ | |
181470fc | 4399 | ether_setup(bond_dev); |
eb7cc59a | 4400 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4401 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4402 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4403 | |
9e2e61fb | 4404 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4405 | |
4406 | /* Initialize the device options */ | |
4407 | bond_dev->tx_queue_len = 0; | |
4408 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4409 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4410 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4411 | |
1da177e4 LT |
4412 | /* At first, we block adding VLANs. That's the only way to |
4413 | * prevent problems that occur when adding VLANs over an | |
4414 | * empty bond. The block will be removed once non-challenged | |
4415 | * slaves are enslaved. | |
4416 | */ | |
4417 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4418 | ||
932ff279 | 4419 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4420 | * transmitting */ |
4421 | bond_dev->features |= NETIF_F_LLTX; | |
4422 | ||
4423 | /* By default, we declare the bond to be fully | |
4424 | * VLAN hardware accelerated capable. Special | |
4425 | * care is taken in the various xmit functions | |
4426 | * when there are slaves that are not hw accel | |
4427 | * capable | |
4428 | */ | |
1da177e4 | 4429 | |
b2a103e6 MM |
4430 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
4431 | NETIF_F_HW_VLAN_TX | | |
4432 | NETIF_F_HW_VLAN_RX | | |
4433 | NETIF_F_HW_VLAN_FILTER; | |
4434 | ||
34324dc2 | 4435 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4436 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4437 | } |
4438 | ||
fdaea7a9 JV |
4439 | static void bond_work_cancel_all(struct bonding *bond) |
4440 | { | |
fdaea7a9 | 4441 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) |
e6d265e8 | 4442 | cancel_delayed_work_sync(&bond->mii_work); |
fdaea7a9 JV |
4443 | |
4444 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
e6d265e8 | 4445 | cancel_delayed_work_sync(&bond->arp_work); |
fdaea7a9 JV |
4446 | |
4447 | if (bond->params.mode == BOND_MODE_ALB && | |
4448 | delayed_work_pending(&bond->alb_work)) | |
e6d265e8 | 4449 | cancel_delayed_work_sync(&bond->alb_work); |
fdaea7a9 JV |
4450 | |
4451 | if (bond->params.mode == BOND_MODE_8023AD && | |
4452 | delayed_work_pending(&bond->ad_work)) | |
e6d265e8 | 4453 | cancel_delayed_work_sync(&bond->ad_work); |
5a37e8ca FL |
4454 | |
4455 | if (delayed_work_pending(&bond->mcast_work)) | |
e6d265e8 | 4456 | cancel_delayed_work_sync(&bond->mcast_work); |
fdaea7a9 JV |
4457 | } |
4458 | ||
c67dfb29 EB |
4459 | /* |
4460 | * Destroy a bonding device. | |
4461 | * Must be under rtnl_lock when this function is called. | |
4462 | */ | |
4463 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4464 | { |
454d7c9b | 4465 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4466 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4467 | |
f6dc31a8 WC |
4468 | bond_netpoll_cleanup(bond_dev); |
4469 | ||
c67dfb29 EB |
4470 | /* Release the bonded slaves */ |
4471 | bond_release_all(bond_dev); | |
4472 | ||
a434e43f JV |
4473 | list_del(&bond->bond_list); |
4474 | ||
4475 | bond_work_cancel_all(bond); | |
4476 | ||
f073c7ca TI |
4477 | bond_debug_unregister(bond); |
4478 | ||
22bedad3 | 4479 | __hw_addr_flush(&bond->mc_list); |
f35188fa JV |
4480 | |
4481 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { | |
4482 | list_del(&vlan->vlan_list); | |
4483 | kfree(vlan); | |
4484 | } | |
a434e43f JV |
4485 | } |
4486 | ||
1da177e4 LT |
4487 | /*------------------------- Module initialization ---------------------------*/ |
4488 | ||
4489 | /* | |
4490 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4491 | * number of the mode or its string name. A bit complicated because |
4492 | * some mode names are substrings of other names, and calls from sysfs | |
4493 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4494 | */ |
325dcf7a | 4495 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4496 | { |
54b87323 | 4497 | int modeint = -1, i, rv; |
a42e534f | 4498 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4499 | |
a42e534f JV |
4500 | for (p = (char *)buf; *p; p++) |
4501 | if (!(isdigit(*p) || isspace(*p))) | |
4502 | break; | |
4503 | ||
4504 | if (*p) | |
ece95f7f | 4505 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4506 | else |
54b87323 | 4507 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4508 | |
4509 | if (!rv) | |
4510 | return -1; | |
1da177e4 LT |
4511 | |
4512 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4513 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4514 | return tbl[i].mode; |
4515 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4516 | return tbl[i].mode; |
1da177e4 LT |
4517 | } |
4518 | ||
4519 | return -1; | |
4520 | } | |
4521 | ||
4522 | static int bond_check_params(struct bond_params *params) | |
4523 | { | |
a549952a | 4524 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4525 | |
1da177e4 LT |
4526 | /* |
4527 | * Convert string parameters. | |
4528 | */ | |
4529 | if (mode) { | |
4530 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4531 | if (bond_mode == -1) { | |
a4aee5c8 | 4532 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4533 | mode == NULL ? "NULL" : mode); |
4534 | return -EINVAL; | |
4535 | } | |
4536 | } | |
4537 | ||
169a3e66 JV |
4538 | if (xmit_hash_policy) { |
4539 | if ((bond_mode != BOND_MODE_XOR) && | |
4540 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4541 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4542 | bond_mode_name(bond_mode)); |
4543 | } else { | |
4544 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4545 | xmit_hashtype_tbl); | |
4546 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4547 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4548 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4549 | xmit_hash_policy); |
4550 | return -EINVAL; | |
4551 | } | |
4552 | } | |
4553 | } | |
4554 | ||
1da177e4 LT |
4555 | if (lacp_rate) { |
4556 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4557 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4558 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4559 | } else { |
4560 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4561 | if (lacp_fast == -1) { | |
a4aee5c8 | 4562 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4563 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4564 | return -EINVAL; | |
4565 | } | |
4566 | } | |
4567 | } | |
4568 | ||
fd989c83 JV |
4569 | if (ad_select) { |
4570 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4571 | if (params->ad_select == -1) { | |
a4aee5c8 | 4572 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4573 | ad_select == NULL ? "NULL" : ad_select); |
4574 | return -EINVAL; | |
4575 | } | |
4576 | ||
4577 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4578 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4579 | } |
4580 | } else { | |
4581 | params->ad_select = BOND_AD_STABLE; | |
4582 | } | |
4583 | ||
f5841306 | 4584 | if (max_bonds < 0) { |
a4aee5c8 JP |
4585 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4586 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4587 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4588 | } | |
4589 | ||
4590 | if (miimon < 0) { | |
a4aee5c8 JP |
4591 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4592 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4593 | miimon = BOND_LINK_MON_INTERV; |
4594 | } | |
4595 | ||
4596 | if (updelay < 0) { | |
a4aee5c8 JP |
4597 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4598 | updelay, INT_MAX); | |
1da177e4 LT |
4599 | updelay = 0; |
4600 | } | |
4601 | ||
4602 | if (downdelay < 0) { | |
a4aee5c8 JP |
4603 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4604 | downdelay, INT_MAX); | |
1da177e4 LT |
4605 | downdelay = 0; |
4606 | } | |
4607 | ||
4608 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4609 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4610 | use_carrier); | |
1da177e4 LT |
4611 | use_carrier = 1; |
4612 | } | |
4613 | ||
ad246c99 BH |
4614 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4615 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4616 | num_peer_notif); | |
4617 | num_peer_notif = 1; | |
4618 | } | |
4619 | ||
1da177e4 LT |
4620 | /* reset values for 802.3ad */ |
4621 | if (bond_mode == BOND_MODE_8023AD) { | |
4622 | if (!miimon) { | |
a4aee5c8 | 4623 | 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 | 4624 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4625 | miimon = 100; |
4626 | } | |
4627 | } | |
4628 | ||
bb1d9123 AG |
4629 | if (tx_queues < 1 || tx_queues > 255) { |
4630 | pr_warning("Warning: tx_queues (%d) should be between " | |
4631 | "1 and 255, resetting to %d\n", | |
4632 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4633 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4634 | } | |
4635 | ||
ebd8e497 AG |
4636 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4637 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4638 | "not of valid value (0/1), so it was set to " | |
4639 | "0\n", all_slaves_active); | |
4640 | all_slaves_active = 0; | |
4641 | } | |
4642 | ||
c2952c31 FL |
4643 | if (resend_igmp < 0 || resend_igmp > 255) { |
4644 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4645 | "0 and 255, resetting to %d\n", | |
4646 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4647 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4648 | } | |
4649 | ||
1da177e4 LT |
4650 | /* reset values for TLB/ALB */ |
4651 | if ((bond_mode == BOND_MODE_TLB) || | |
4652 | (bond_mode == BOND_MODE_ALB)) { | |
4653 | if (!miimon) { | |
a4aee5c8 | 4654 | 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 | 4655 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4656 | miimon = 100; |
4657 | } | |
4658 | } | |
4659 | ||
4660 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4661 | 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", |
4662 | updelay); | |
1da177e4 LT |
4663 | } |
4664 | ||
4665 | if (!miimon) { | |
4666 | if (updelay || downdelay) { | |
4667 | /* just warn the user the up/down delay will have | |
4668 | * no effect since miimon is zero... | |
4669 | */ | |
a4aee5c8 JP |
4670 | 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", |
4671 | updelay, downdelay); | |
1da177e4 LT |
4672 | } |
4673 | } else { | |
4674 | /* don't allow arp monitoring */ | |
4675 | if (arp_interval) { | |
a4aee5c8 JP |
4676 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4677 | miimon, arp_interval); | |
1da177e4 LT |
4678 | arp_interval = 0; |
4679 | } | |
4680 | ||
4681 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4682 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4683 | updelay, miimon, | |
4684 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4685 | } |
4686 | ||
4687 | updelay /= miimon; | |
4688 | ||
4689 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4690 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4691 | downdelay, miimon, | |
4692 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4693 | } |
4694 | ||
4695 | downdelay /= miimon; | |
4696 | } | |
4697 | ||
4698 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4699 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4700 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4701 | arp_interval = BOND_LINK_ARP_INTERV; |
4702 | } | |
4703 | ||
4704 | for (arp_ip_count = 0; | |
4705 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4706 | arp_ip_count++) { | |
4707 | /* not complete check, but should be good enough to | |
4708 | catch mistakes */ | |
4709 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4710 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4711 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4712 | arp_interval = 0; |
4713 | } else { | |
d3bb52b0 | 4714 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4715 | arp_target[arp_ip_count] = ip; |
4716 | } | |
4717 | } | |
4718 | ||
4719 | if (arp_interval && !arp_ip_count) { | |
4720 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4721 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4722 | arp_interval); | |
1da177e4 LT |
4723 | arp_interval = 0; |
4724 | } | |
4725 | ||
f5b2b966 JV |
4726 | if (arp_validate) { |
4727 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4728 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4729 | return -EINVAL; |
4730 | } | |
4731 | if (!arp_interval) { | |
a4aee5c8 | 4732 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4733 | return -EINVAL; |
4734 | } | |
4735 | ||
4736 | arp_validate_value = bond_parse_parm(arp_validate, | |
4737 | arp_validate_tbl); | |
4738 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4739 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4740 | arp_validate == NULL ? "NULL" : arp_validate); |
4741 | return -EINVAL; | |
4742 | } | |
4743 | } else | |
4744 | arp_validate_value = 0; | |
4745 | ||
1da177e4 | 4746 | if (miimon) { |
a4aee5c8 | 4747 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4748 | } else if (arp_interval) { |
4749 | int i; | |
4750 | ||
a4aee5c8 JP |
4751 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4752 | arp_interval, | |
4753 | arp_validate_tbl[arp_validate_value].modename, | |
4754 | arp_ip_count); | |
1da177e4 LT |
4755 | |
4756 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4757 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4758 | |
e5e2a8fd | 4759 | pr_info("\n"); |
1da177e4 | 4760 | |
b8a9787e | 4761 | } else if (max_bonds) { |
1da177e4 LT |
4762 | /* miimon and arp_interval not set, we need one so things |
4763 | * work as expected, see bonding.txt for details | |
4764 | */ | |
b2730f4f | 4765 | 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 |
4766 | } |
4767 | ||
4768 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4769 | /* currently, using a primary only makes sense | |
4770 | * in active backup, TLB or ALB modes | |
4771 | */ | |
a4aee5c8 JP |
4772 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4773 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4774 | primary = NULL; |
4775 | } | |
4776 | ||
a549952a JP |
4777 | if (primary && primary_reselect) { |
4778 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4779 | pri_reselect_tbl); | |
4780 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4781 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4782 | primary_reselect == |
4783 | NULL ? "NULL" : primary_reselect); | |
4784 | return -EINVAL; | |
4785 | } | |
4786 | } else { | |
4787 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4788 | } | |
4789 | ||
3915c1e8 JV |
4790 | if (fail_over_mac) { |
4791 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4792 | fail_over_mac_tbl); | |
4793 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4794 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4795 | arp_validate == NULL ? "NULL" : arp_validate); |
4796 | return -EINVAL; | |
4797 | } | |
4798 | ||
4799 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4800 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4801 | } else { |
4802 | fail_over_mac_value = BOND_FOM_NONE; | |
4803 | } | |
dd957c57 | 4804 | |
1da177e4 LT |
4805 | /* fill params struct with the proper values */ |
4806 | params->mode = bond_mode; | |
169a3e66 | 4807 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4808 | params->miimon = miimon; |
ad246c99 | 4809 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4810 | params->arp_interval = arp_interval; |
f5b2b966 | 4811 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4812 | params->updelay = updelay; |
4813 | params->downdelay = downdelay; | |
4814 | params->use_carrier = use_carrier; | |
4815 | params->lacp_fast = lacp_fast; | |
4816 | params->primary[0] = 0; | |
a549952a | 4817 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4818 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4819 | params->tx_queues = tx_queues; |
ebd8e497 | 4820 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4821 | params->resend_igmp = resend_igmp; |
655f8919 | 4822 | params->min_links = min_links; |
1da177e4 LT |
4823 | |
4824 | if (primary) { | |
4825 | strncpy(params->primary, primary, IFNAMSIZ); | |
4826 | params->primary[IFNAMSIZ - 1] = 0; | |
4827 | } | |
4828 | ||
4829 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4830 | ||
4831 | return 0; | |
4832 | } | |
4833 | ||
0daa2303 | 4834 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4835 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4836 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4837 | |
e8a0464c DM |
4838 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4839 | struct netdev_queue *txq, | |
4840 | void *_unused) | |
c773e847 DM |
4841 | { |
4842 | lockdep_set_class(&txq->_xmit_lock, | |
4843 | &bonding_netdev_xmit_lock_key); | |
4844 | } | |
4845 | ||
4846 | static void bond_set_lockdep_class(struct net_device *dev) | |
4847 | { | |
cf508b12 DM |
4848 | lockdep_set_class(&dev->addr_list_lock, |
4849 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4850 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4851 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4852 | } |
4853 | ||
181470fc SH |
4854 | /* |
4855 | * Called from registration process | |
4856 | */ | |
4857 | static int bond_init(struct net_device *bond_dev) | |
4858 | { | |
4859 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4860 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4861 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4862 | |
4863 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4864 | ||
9fe0617d NH |
4865 | /* |
4866 | * Initialize locks that may be required during | |
4867 | * en/deslave operations. All of the bond_open work | |
4868 | * (of which this is part) should really be moved to | |
4869 | * a phase prior to dev_open | |
4870 | */ | |
4871 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4872 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4873 | ||
181470fc SH |
4874 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4875 | if (!bond->wq) | |
4876 | return -ENOMEM; | |
4877 | ||
4878 | bond_set_lockdep_class(bond_dev); | |
4879 | ||
ec87fd3b | 4880 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4881 | |
6151b3d4 | 4882 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4883 | |
f073c7ca TI |
4884 | bond_debug_register(bond); |
4885 | ||
22bedad3 | 4886 | __hw_addr_init(&bond->mc_list); |
181470fc SH |
4887 | return 0; |
4888 | } | |
4889 | ||
88ead977 EB |
4890 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4891 | { | |
4892 | if (tb[IFLA_ADDRESS]) { | |
4893 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4894 | return -EINVAL; | |
4895 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4896 | return -EADDRNOTAVAIL; | |
4897 | } | |
4898 | return 0; | |
4899 | } | |
4900 | ||
d40156aa | 4901 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4902 | { |
efacb309 | 4903 | return tx_queues; |
d5da4510 JP |
4904 | } |
4905 | ||
88ead977 | 4906 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4907 | .kind = "bond", |
4908 | .priv_size = sizeof(struct bonding), | |
4909 | .setup = bond_setup, | |
4910 | .validate = bond_validate, | |
4911 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4912 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4913 | as for TX queues */ | |
88ead977 EB |
4914 | }; |
4915 | ||
dfe60397 | 4916 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4917 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4918 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4919 | * set up our sysfs entries. | |
4920 | */ | |
ec87fd3b | 4921 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4922 | { |
4923 | struct net_device *bond_dev; | |
4924 | int res; | |
4925 | ||
4926 | rtnl_lock(); | |
027ea041 | 4927 | |
1c5cae81 JP |
4928 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4929 | name ? name : "bond%d", | |
4930 | bond_setup, tx_queues); | |
dfe60397 | 4931 | if (!bond_dev) { |
a4aee5c8 | 4932 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4933 | rtnl_unlock(); |
4934 | return -ENOMEM; | |
dfe60397 MW |
4935 | } |
4936 | ||
ec87fd3b | 4937 | dev_net_set(bond_dev, net); |
88ead977 EB |
4938 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4939 | ||
dfe60397 | 4940 | res = register_netdevice(bond_dev); |
0daa2303 | 4941 | |
e826eafa PO |
4942 | netif_carrier_off(bond_dev); |
4943 | ||
7e083840 | 4944 | rtnl_unlock(); |
9e2e61fb AW |
4945 | if (res < 0) |
4946 | bond_destructor(bond_dev); | |
30c15ba9 | 4947 | return res; |
dfe60397 MW |
4948 | } |
4949 | ||
2c8c1e72 | 4950 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4951 | { |
15449745 | 4952 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4953 | |
4954 | bn->net = net; | |
4955 | INIT_LIST_HEAD(&bn->dev_list); | |
4956 | ||
ec87fd3b | 4957 | bond_create_proc_dir(bn); |
4c22400a | 4958 | bond_create_sysfs(bn); |
15449745 EB |
4959 | |
4960 | return 0; | |
ec87fd3b EB |
4961 | } |
4962 | ||
2c8c1e72 | 4963 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4964 | { |
15449745 | 4965 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b | 4966 | |
4c22400a | 4967 | bond_destroy_sysfs(bn); |
ec87fd3b | 4968 | bond_destroy_proc_dir(bn); |
ec87fd3b EB |
4969 | } |
4970 | ||
4971 | static struct pernet_operations bond_net_ops = { | |
4972 | .init = bond_net_init, | |
4973 | .exit = bond_net_exit, | |
15449745 EB |
4974 | .id = &bond_net_id, |
4975 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4976 | }; |
4977 | ||
1da177e4 LT |
4978 | static int __init bonding_init(void) |
4979 | { | |
1da177e4 LT |
4980 | int i; |
4981 | int res; | |
4982 | ||
bd33acc3 | 4983 | pr_info("%s", bond_version); |
1da177e4 | 4984 | |
dfe60397 | 4985 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4986 | if (res) |
dfe60397 | 4987 | goto out; |
1da177e4 | 4988 | |
15449745 | 4989 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4990 | if (res) |
4991 | goto out; | |
027ea041 | 4992 | |
88ead977 EB |
4993 | res = rtnl_link_register(&bond_link_ops); |
4994 | if (res) | |
6639104b | 4995 | goto err_link; |
88ead977 | 4996 | |
f073c7ca TI |
4997 | bond_create_debugfs(); |
4998 | ||
1da177e4 | 4999 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5000 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5001 | if (res) |
5002 | goto err; | |
1da177e4 LT |
5003 | } |
5004 | ||
1da177e4 | 5005 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 5006 | out: |
1da177e4 | 5007 | return res; |
88ead977 EB |
5008 | err: |
5009 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5010 | err_link: |
15449745 | 5011 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5012 | goto out; |
dfe60397 | 5013 | |
1da177e4 LT |
5014 | } |
5015 | ||
5016 | static void __exit bonding_exit(void) | |
5017 | { | |
5018 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
5019 | ||
f073c7ca | 5020 | bond_destroy_debugfs(); |
ae68c398 | 5021 | |
88ead977 | 5022 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5023 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
5024 | |
5025 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
5026 | /* |
5027 | * Make sure we don't have an imbalance on our netpoll blocking | |
5028 | */ | |
5029 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 5030 | #endif |
1da177e4 LT |
5031 | } |
5032 | ||
5033 | module_init(bonding_init); | |
5034 | module_exit(bonding_exit); | |
5035 | MODULE_LICENSE("GPL"); | |
5036 | MODULE_VERSION(DRV_VERSION); | |
5037 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5038 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5039 | MODULE_ALIAS_RTNL_LINK("bond"); |