]>
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/system.h> |
1da177e4 | 58 | #include <asm/dma.h> |
3d632c3f | 59 | #include <linux/uaccess.h> |
1da177e4 LT |
60 | #include <linux/errno.h> |
61 | #include <linux/netdevice.h> | |
62 | #include <linux/inetdevice.h> | |
a816c7c7 | 63 | #include <linux/igmp.h> |
1da177e4 LT |
64 | #include <linux/etherdevice.h> |
65 | #include <linux/skbuff.h> | |
66 | #include <net/sock.h> | |
67 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
68 | #include <linux/smp.h> |
69 | #include <linux/if_ether.h> | |
70 | #include <net/arp.h> | |
71 | #include <linux/mii.h> | |
72 | #include <linux/ethtool.h> | |
73 | #include <linux/if_vlan.h> | |
74 | #include <linux/if_bonding.h> | |
b63bb739 | 75 | #include <linux/jiffies.h> |
e843fa50 | 76 | #include <linux/preempt.h> |
c3ade5ca | 77 | #include <net/route.h> |
457c4cbc | 78 | #include <net/net_namespace.h> |
ec87fd3b | 79 | #include <net/netns/generic.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; |
1da177e4 | 92 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
93 | static int updelay; |
94 | static int downdelay; | |
1da177e4 | 95 | static int use_carrier = 1; |
3d632c3f SH |
96 | static char *mode; |
97 | static char *primary; | |
a549952a | 98 | static char *primary_reselect; |
3d632c3f SH |
99 | static char *lacp_rate; |
100 | static char *ad_select; | |
101 | static char *xmit_hash_policy; | |
1da177e4 | 102 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
103 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
104 | static char *arp_validate; | |
105 | static char *fail_over_mac; | |
ebd8e497 | 106 | static int all_slaves_active = 0; |
d2991f75 | 107 | static struct bond_params bonding_defaults; |
c2952c31 | 108 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
1da177e4 LT |
109 | |
110 | module_param(max_bonds, int, 0); | |
111 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
112 | module_param(tx_queues, int, 0); |
113 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
1da177e4 LT |
114 | module_param(miimon, int, 0); |
115 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
116 | module_param(updelay, int, 0); | |
117 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
118 | module_param(downdelay, int, 0); | |
2ac47660 MW |
119 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
120 | "in milliseconds"); | |
1da177e4 | 121 | module_param(use_carrier, int, 0); |
2ac47660 MW |
122 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
123 | "0 for off, 1 for on (default)"); | |
1da177e4 | 124 | module_param(mode, charp, 0); |
2ac47660 MW |
125 | MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, " |
126 | "1 for active-backup, 2 for balance-xor, " | |
127 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
128 | "6 for balance-alb"); | |
1da177e4 LT |
129 | module_param(primary, charp, 0); |
130 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
131 | module_param(primary_reselect, charp, 0); |
132 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
133 | "once it comes up; " | |
134 | "0 for always (default), " | |
135 | "1 for only if speed of primary is " | |
136 | "better, " | |
137 | "2 for only on active slave " | |
138 | "failure"); | |
1da177e4 | 139 | module_param(lacp_rate, charp, 0); |
2ac47660 MW |
140 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner " |
141 | "(slow/fast)"); | |
fd989c83 JV |
142 | module_param(ad_select, charp, 0); |
143 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)"); | |
169a3e66 | 144 | module_param(xmit_hash_policy, charp, 0); |
2ac47660 MW |
145 | MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)" |
146 | ", 1 for layer 3+4"); | |
1da177e4 LT |
147 | module_param(arp_interval, int, 0); |
148 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
149 | module_param_array(arp_ip_target, charp, NULL, 0); | |
150 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 JV |
151 | module_param(arp_validate, charp, 0); |
152 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all"); | |
3915c1e8 JV |
153 | module_param(fail_over_mac, charp, 0); |
154 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow"); | |
ebd8e497 AG |
155 | module_param(all_slaves_active, int, 0); |
156 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
157 | "by setting active flag for all slaves. " | |
158 | "0 for never (default), 1 for always."); | |
c2952c31 FL |
159 | module_param(resend_igmp, int, 0); |
160 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on link failure"); | |
1da177e4 LT |
161 | |
162 | /*----------------------------- Global variables ----------------------------*/ | |
163 | ||
e843fa50 | 164 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 165 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
166 | #endif |
167 | ||
f99189b1 | 168 | int bond_net_id __read_mostly; |
1da177e4 | 169 | |
3d632c3f SH |
170 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
171 | static int arp_ip_count; | |
1da177e4 | 172 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
173 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
174 | static int lacp_fast; | |
1da177e4 | 175 | |
e97fd7c6 | 176 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
177 | { "slow", AD_LACP_SLOW}, |
178 | { "fast", AD_LACP_FAST}, | |
179 | { NULL, -1}, | |
180 | }; | |
181 | ||
e97fd7c6 | 182 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
183 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
184 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
185 | { "balance-xor", BOND_MODE_XOR}, | |
186 | { "broadcast", BOND_MODE_BROADCAST}, | |
187 | { "802.3ad", BOND_MODE_8023AD}, | |
188 | { "balance-tlb", BOND_MODE_TLB}, | |
189 | { "balance-alb", BOND_MODE_ALB}, | |
190 | { NULL, -1}, | |
191 | }; | |
192 | ||
e97fd7c6 | 193 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
194 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
195 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 196 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
197 | { NULL, -1}, |
198 | }; | |
199 | ||
e97fd7c6 | 200 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
201 | { "none", BOND_ARP_VALIDATE_NONE}, |
202 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
203 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
204 | { "all", BOND_ARP_VALIDATE_ALL}, | |
205 | { NULL, -1}, | |
206 | }; | |
207 | ||
e97fd7c6 | 208 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
209 | { "none", BOND_FOM_NONE}, |
210 | { "active", BOND_FOM_ACTIVE}, | |
211 | { "follow", BOND_FOM_FOLLOW}, | |
212 | { NULL, -1}, | |
213 | }; | |
214 | ||
a549952a JP |
215 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
216 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
217 | { "better", BOND_PRI_RESELECT_BETTER}, | |
218 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
219 | { NULL, -1}, | |
220 | }; | |
221 | ||
fd989c83 JV |
222 | struct bond_parm_tbl ad_select_tbl[] = { |
223 | { "stable", BOND_AD_STABLE}, | |
224 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
225 | { "count", BOND_AD_COUNT}, | |
226 | { NULL, -1}, | |
227 | }; | |
228 | ||
1da177e4 LT |
229 | /*-------------------------- Forward declarations ---------------------------*/ |
230 | ||
181470fc | 231 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 232 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
233 | |
234 | /*---------------------------- General routines -----------------------------*/ | |
235 | ||
bd33acc3 | 236 | const char *bond_mode_name(int mode) |
1da177e4 | 237 | { |
77afc92b HE |
238 | static const char *names[] = { |
239 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
240 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
241 | [BOND_MODE_XOR] = "load balancing (xor)", | |
242 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 243 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
244 | [BOND_MODE_TLB] = "transmit load balancing", |
245 | [BOND_MODE_ALB] = "adaptive load balancing", | |
246 | }; | |
247 | ||
248 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 249 | return "unknown"; |
77afc92b HE |
250 | |
251 | return names[mode]; | |
1da177e4 LT |
252 | } |
253 | ||
254 | /*---------------------------------- VLAN -----------------------------------*/ | |
255 | ||
256 | /** | |
257 | * bond_add_vlan - add a new vlan id on bond | |
258 | * @bond: bond that got the notification | |
259 | * @vlan_id: the vlan id to add | |
260 | * | |
261 | * Returns -ENOMEM if allocation failed. | |
262 | */ | |
263 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
264 | { | |
265 | struct vlan_entry *vlan; | |
266 | ||
5a03cdb7 | 267 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 268 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 269 | |
305d552a | 270 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 271 | if (!vlan) |
1da177e4 | 272 | return -ENOMEM; |
1da177e4 LT |
273 | |
274 | INIT_LIST_HEAD(&vlan->vlan_list); | |
275 | vlan->vlan_id = vlan_id; | |
276 | ||
277 | write_lock_bh(&bond->lock); | |
278 | ||
279 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
280 | ||
281 | write_unlock_bh(&bond->lock); | |
282 | ||
5a03cdb7 | 283 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
284 | |
285 | return 0; | |
286 | } | |
287 | ||
288 | /** | |
289 | * bond_del_vlan - delete a vlan id from bond | |
290 | * @bond: bond that got the notification | |
291 | * @vlan_id: the vlan id to delete | |
292 | * | |
293 | * returns -ENODEV if @vlan_id was not found in @bond. | |
294 | */ | |
295 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
296 | { | |
0883beca | 297 | struct vlan_entry *vlan; |
1da177e4 LT |
298 | int res = -ENODEV; |
299 | ||
5a03cdb7 | 300 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 | 301 | |
e843fa50 | 302 | block_netpoll_tx(); |
1da177e4 LT |
303 | write_lock_bh(&bond->lock); |
304 | ||
0883beca | 305 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
306 | if (vlan->vlan_id == vlan_id) { |
307 | list_del(&vlan->vlan_list); | |
308 | ||
58402054 | 309 | if (bond_is_lb(bond)) |
1da177e4 | 310 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 311 | |
a4aee5c8 JP |
312 | pr_debug("removed VLAN ID %d from bond %s\n", |
313 | vlan_id, bond->dev->name); | |
1da177e4 LT |
314 | |
315 | kfree(vlan); | |
316 | ||
317 | if (list_empty(&bond->vlan_list) && | |
318 | (bond->slave_cnt == 0)) { | |
319 | /* Last VLAN removed and no slaves, so | |
320 | * restore block on adding VLANs. This will | |
321 | * be removed once new slaves that are not | |
322 | * VLAN challenged will be added. | |
323 | */ | |
324 | bond->dev->features |= NETIF_F_VLAN_CHALLENGED; | |
325 | } | |
326 | ||
327 | res = 0; | |
328 | goto out; | |
329 | } | |
330 | } | |
331 | ||
a4aee5c8 JP |
332 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
333 | vlan_id, bond->dev->name); | |
1da177e4 LT |
334 | |
335 | out: | |
336 | write_unlock_bh(&bond->lock); | |
e843fa50 | 337 | unblock_netpoll_tx(); |
1da177e4 LT |
338 | return res; |
339 | } | |
340 | ||
341 | /** | |
342 | * bond_has_challenged_slaves | |
343 | * @bond: the bond we're working on | |
344 | * | |
345 | * Searches the slave list. Returns 1 if a vlan challenged slave | |
346 | * was found, 0 otherwise. | |
347 | * | |
348 | * Assumes bond->lock is held. | |
349 | */ | |
350 | static int bond_has_challenged_slaves(struct bonding *bond) | |
351 | { | |
352 | struct slave *slave; | |
353 | int i; | |
354 | ||
355 | bond_for_each_slave(bond, slave, i) { | |
356 | if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 357 | pr_debug("found VLAN challenged slave - %s\n", |
a4aee5c8 | 358 | slave->dev->name); |
1da177e4 LT |
359 | return 1; |
360 | } | |
361 | } | |
362 | ||
5a03cdb7 | 363 | pr_debug("no VLAN challenged slaves found\n"); |
1da177e4 LT |
364 | return 0; |
365 | } | |
366 | ||
367 | /** | |
368 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
369 | * @bond: the bond we're working on | |
370 | * @curr: item we're advancing from | |
371 | * | |
372 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
373 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 374 | * |
1da177e4 LT |
375 | * Caller must hold bond->lock |
376 | */ | |
377 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
378 | { | |
379 | struct vlan_entry *next, *last; | |
380 | ||
3d632c3f | 381 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 382 | return NULL; |
1da177e4 LT |
383 | |
384 | if (!curr) { | |
385 | next = list_entry(bond->vlan_list.next, | |
386 | struct vlan_entry, vlan_list); | |
387 | } else { | |
388 | last = list_entry(bond->vlan_list.prev, | |
389 | struct vlan_entry, vlan_list); | |
390 | if (last == curr) { | |
391 | next = list_entry(bond->vlan_list.next, | |
392 | struct vlan_entry, vlan_list); | |
393 | } else { | |
394 | next = list_entry(curr->vlan_list.next, | |
395 | struct vlan_entry, vlan_list); | |
396 | } | |
397 | } | |
398 | ||
399 | return next; | |
400 | } | |
401 | ||
402 | /** | |
403 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 404 | * |
1da177e4 LT |
405 | * @bond: bond device that got this skb for tx. |
406 | * @skb: hw accel VLAN tagged skb to transmit | |
407 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 408 | */ |
3d632c3f SH |
409 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
410 | struct net_device *slave_dev) | |
1da177e4 | 411 | { |
8387451e | 412 | skb->dev = slave_dev; |
1da177e4 | 413 | skb->priority = 1; |
080e4130 | 414 | if (unlikely(netpoll_tx_running(slave_dev))) |
8a8efa22 | 415 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 416 | else |
f6dc31a8 | 417 | dev_queue_xmit(skb); |
1da177e4 LT |
418 | |
419 | return 0; | |
420 | } | |
421 | ||
422 | /* | |
423 | * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid | |
424 | * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a | |
425 | * lock because: | |
426 | * a. This operation is performed in IOCTL context, | |
427 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
428 | * c. Holding a lock with BH disabled while directly calling a base driver | |
429 | * entry point is generally a BAD idea. | |
3d632c3f | 430 | * |
1da177e4 LT |
431 | * The design of synchronization/protection for this operation in the 8021q |
432 | * module is good for one or more VLAN devices over a single physical device | |
433 | * and cannot be extended for a teaming solution like bonding, so there is a | |
434 | * potential race condition here where a net device from the vlan group might | |
435 | * be referenced (either by a base driver or the 8021q code) while it is being | |
436 | * removed from the system. However, it turns out we're not making matters | |
437 | * worse, and if it works for regular VLAN usage it will work here too. | |
438 | */ | |
439 | ||
440 | /** | |
441 | * bond_vlan_rx_register - Propagates registration to slaves | |
442 | * @bond_dev: bonding net device that got called | |
443 | * @grp: vlan group being registered | |
444 | */ | |
3d632c3f SH |
445 | static void bond_vlan_rx_register(struct net_device *bond_dev, |
446 | struct vlan_group *grp) | |
1da177e4 | 447 | { |
454d7c9b | 448 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
449 | struct slave *slave; |
450 | int i; | |
451 | ||
a71fb881 | 452 | write_lock_bh(&bond->lock); |
1da177e4 | 453 | bond->vlgrp = grp; |
a71fb881 | 454 | write_unlock_bh(&bond->lock); |
1da177e4 LT |
455 | |
456 | bond_for_each_slave(bond, slave, i) { | |
457 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 458 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
459 | |
460 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
461 | slave_ops->ndo_vlan_rx_register) { |
462 | slave_ops->ndo_vlan_rx_register(slave_dev, grp); | |
1da177e4 LT |
463 | } |
464 | } | |
465 | } | |
466 | ||
467 | /** | |
468 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
469 | * @bond_dev: bonding net device that got called | |
470 | * @vid: vlan id being added | |
471 | */ | |
472 | static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) | |
473 | { | |
454d7c9b | 474 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
475 | struct slave *slave; |
476 | int i, res; | |
477 | ||
478 | bond_for_each_slave(bond, slave, i) { | |
479 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 480 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
481 | |
482 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a SH |
483 | slave_ops->ndo_vlan_rx_add_vid) { |
484 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid); | |
1da177e4 LT |
485 | } |
486 | } | |
487 | ||
488 | res = bond_add_vlan(bond, vid); | |
489 | if (res) { | |
a4aee5c8 | 490 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 LT |
491 | bond_dev->name, vid); |
492 | } | |
493 | } | |
494 | ||
495 | /** | |
496 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
497 | * @bond_dev: bonding net device that got called | |
498 | * @vid: vlan id being removed | |
499 | */ | |
500 | static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) | |
501 | { | |
454d7c9b | 502 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
503 | struct slave *slave; |
504 | struct net_device *vlan_dev; | |
505 | int i, res; | |
506 | ||
507 | bond_for_each_slave(bond, slave, i) { | |
508 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 509 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
510 | |
511 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a | 512 | slave_ops->ndo_vlan_rx_kill_vid) { |
1da177e4 LT |
513 | /* Save and then restore vlan_dev in the grp array, |
514 | * since the slave's driver might clear it. | |
515 | */ | |
5c15bdec | 516 | vlan_dev = vlan_group_get_device(bond->vlgrp, vid); |
eb7cc59a | 517 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid); |
5c15bdec | 518 | vlan_group_set_device(bond->vlgrp, vid, vlan_dev); |
1da177e4 LT |
519 | } |
520 | } | |
521 | ||
522 | res = bond_del_vlan(bond, vid); | |
523 | if (res) { | |
a4aee5c8 | 524 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 LT |
525 | bond_dev->name, vid); |
526 | } | |
527 | } | |
528 | ||
529 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
530 | { | |
531 | struct vlan_entry *vlan; | |
eb7cc59a | 532 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 533 | |
f35188fa | 534 | if (!bond->vlgrp) |
03dc2f4c | 535 | return; |
1da177e4 LT |
536 | |
537 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
538 | slave_ops->ndo_vlan_rx_register) |
539 | slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp); | |
1da177e4 LT |
540 | |
541 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 542 | !(slave_ops->ndo_vlan_rx_add_vid)) |
03dc2f4c | 543 | return; |
1da177e4 | 544 | |
eb7cc59a SH |
545 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) |
546 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id); | |
1da177e4 LT |
547 | } |
548 | ||
3d632c3f SH |
549 | static void bond_del_vlans_from_slave(struct bonding *bond, |
550 | struct net_device *slave_dev) | |
1da177e4 | 551 | { |
eb7cc59a | 552 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
553 | struct vlan_entry *vlan; |
554 | struct net_device *vlan_dev; | |
555 | ||
f35188fa | 556 | if (!bond->vlgrp) |
03dc2f4c | 557 | return; |
1da177e4 LT |
558 | |
559 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 560 | !(slave_ops->ndo_vlan_rx_kill_vid)) |
1da177e4 | 561 | goto unreg; |
1da177e4 LT |
562 | |
563 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
f35188fa JV |
564 | if (!vlan->vlan_id) |
565 | continue; | |
1da177e4 LT |
566 | /* Save and then restore vlan_dev in the grp array, |
567 | * since the slave's driver might clear it. | |
568 | */ | |
5c15bdec | 569 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
eb7cc59a | 570 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id); |
5c15bdec | 571 | vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev); |
1da177e4 LT |
572 | } |
573 | ||
574 | unreg: | |
575 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
576 | slave_ops->ndo_vlan_rx_register) |
577 | slave_ops->ndo_vlan_rx_register(slave_dev, NULL); | |
1da177e4 LT |
578 | } |
579 | ||
580 | /*------------------------------- Link status -------------------------------*/ | |
581 | ||
ff59c456 JV |
582 | /* |
583 | * Set the carrier state for the master according to the state of its | |
584 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
585 | * do special 802.3ad magic. | |
586 | * | |
587 | * Returns zero if carrier state does not change, nonzero if it does. | |
588 | */ | |
589 | static int bond_set_carrier(struct bonding *bond) | |
590 | { | |
591 | struct slave *slave; | |
592 | int i; | |
593 | ||
594 | if (bond->slave_cnt == 0) | |
595 | goto down; | |
596 | ||
597 | if (bond->params.mode == BOND_MODE_8023AD) | |
598 | return bond_3ad_set_carrier(bond); | |
599 | ||
600 | bond_for_each_slave(bond, slave, i) { | |
601 | if (slave->link == BOND_LINK_UP) { | |
602 | if (!netif_carrier_ok(bond->dev)) { | |
603 | netif_carrier_on(bond->dev); | |
604 | return 1; | |
605 | } | |
606 | return 0; | |
607 | } | |
608 | } | |
609 | ||
610 | down: | |
611 | if (netif_carrier_ok(bond->dev)) { | |
612 | netif_carrier_off(bond->dev); | |
613 | return 1; | |
614 | } | |
615 | return 0; | |
616 | } | |
617 | ||
1da177e4 LT |
618 | /* |
619 | * Get link speed and duplex from the slave's base driver | |
620 | * using ethtool. If for some reason the call fails or the | |
621 | * values are invalid, fake speed and duplex to 100/Full | |
622 | * and return error. | |
623 | */ | |
624 | static int bond_update_speed_duplex(struct slave *slave) | |
625 | { | |
626 | struct net_device *slave_dev = slave->dev; | |
5d30530e DD |
627 | struct ethtool_cmd etool = { .cmd = ETHTOOL_GSET }; |
628 | u32 slave_speed; | |
61a44b9c | 629 | int res; |
1da177e4 LT |
630 | |
631 | /* Fake speed and duplex */ | |
632 | slave->speed = SPEED_100; | |
633 | slave->duplex = DUPLEX_FULL; | |
634 | ||
61a44b9c MW |
635 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
636 | return -1; | |
1da177e4 | 637 | |
61a44b9c MW |
638 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
639 | if (res < 0) | |
1da177e4 | 640 | return -1; |
1da177e4 | 641 | |
5d30530e DD |
642 | slave_speed = ethtool_cmd_speed(&etool); |
643 | switch (slave_speed) { | |
1da177e4 LT |
644 | case SPEED_10: |
645 | case SPEED_100: | |
646 | case SPEED_1000: | |
94dbffd5 | 647 | case SPEED_10000: |
1da177e4 LT |
648 | break; |
649 | default: | |
650 | return -1; | |
651 | } | |
652 | ||
653 | switch (etool.duplex) { | |
654 | case DUPLEX_FULL: | |
655 | case DUPLEX_HALF: | |
656 | break; | |
657 | default: | |
658 | return -1; | |
659 | } | |
660 | ||
5d30530e | 661 | slave->speed = slave_speed; |
1da177e4 LT |
662 | slave->duplex = etool.duplex; |
663 | ||
664 | return 0; | |
665 | } | |
666 | ||
667 | /* | |
668 | * if <dev> supports MII link status reporting, check its link status. | |
669 | * | |
670 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 671 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
672 | * |
673 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
674 | * can't tell and just pretend it is), or 0, meaning that the link is | |
675 | * down. | |
676 | * | |
677 | * If reporting is non-zero, instead of faking link up, return -1 if | |
678 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
679 | * support them). If use_carrier is set, return whatever it says. | |
680 | * It'd be nice if there was a good way to tell if a driver supports | |
681 | * netif_carrier, but there really isn't. | |
682 | */ | |
3d632c3f SH |
683 | static int bond_check_dev_link(struct bonding *bond, |
684 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 685 | { |
eb7cc59a | 686 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 687 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
688 | struct ifreq ifr; |
689 | struct mii_ioctl_data *mii; | |
1da177e4 | 690 | |
6c988853 PG |
691 | if (!reporting && !netif_running(slave_dev)) |
692 | return 0; | |
693 | ||
eb7cc59a | 694 | if (bond->params.use_carrier) |
1da177e4 | 695 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 696 | |
29112f4e JP |
697 | /* Try to get link status using Ethtool first. */ |
698 | if (slave_dev->ethtool_ops) { | |
699 | if (slave_dev->ethtool_ops->get_link) { | |
700 | u32 link; | |
701 | ||
702 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
703 | ||
704 | return link ? BMSR_LSTATUS : 0; | |
705 | } | |
706 | } | |
707 | ||
3d632c3f | 708 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 709 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
710 | if (ioctl) { |
711 | /* TODO: set pointer to correct ioctl on a per team member */ | |
712 | /* bases to make this more efficient. that is, once */ | |
713 | /* we determine the correct ioctl, we will always */ | |
714 | /* call it and not the others for that team */ | |
715 | /* member. */ | |
716 | ||
717 | /* | |
718 | * We cannot assume that SIOCGMIIPHY will also read a | |
719 | * register; not all network drivers (e.g., e100) | |
720 | * support that. | |
721 | */ | |
722 | ||
723 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
724 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
725 | mii = if_mii(&ifr); | |
726 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
727 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
728 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
729 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
730 | } |
731 | } | |
732 | ||
1da177e4 LT |
733 | /* |
734 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 735 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
736 | * cannot report link status). If not reporting, pretend |
737 | * we're ok. | |
738 | */ | |
3d632c3f | 739 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
740 | } |
741 | ||
742 | /*----------------------------- Multicast list ------------------------------*/ | |
743 | ||
1da177e4 LT |
744 | /* |
745 | * Push the promiscuity flag down to appropriate slaves | |
746 | */ | |
7e1a1ac1 | 747 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 748 | { |
7e1a1ac1 | 749 | int err = 0; |
1da177e4 LT |
750 | if (USES_PRIMARY(bond->params.mode)) { |
751 | /* write lock already acquired */ | |
752 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
753 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
754 | inc); | |
1da177e4 LT |
755 | } |
756 | } else { | |
757 | struct slave *slave; | |
758 | int i; | |
759 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
760 | err = dev_set_promiscuity(slave->dev, inc); |
761 | if (err) | |
762 | return err; | |
1da177e4 LT |
763 | } |
764 | } | |
7e1a1ac1 | 765 | return err; |
1da177e4 LT |
766 | } |
767 | ||
768 | /* | |
769 | * Push the allmulti flag down to all slaves | |
770 | */ | |
7e1a1ac1 | 771 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 772 | { |
7e1a1ac1 | 773 | int err = 0; |
1da177e4 LT |
774 | if (USES_PRIMARY(bond->params.mode)) { |
775 | /* write lock already acquired */ | |
776 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
777 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
778 | inc); | |
1da177e4 LT |
779 | } |
780 | } else { | |
781 | struct slave *slave; | |
782 | int i; | |
783 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
784 | err = dev_set_allmulti(slave->dev, inc); |
785 | if (err) | |
786 | return err; | |
1da177e4 LT |
787 | } |
788 | } | |
7e1a1ac1 | 789 | return err; |
1da177e4 LT |
790 | } |
791 | ||
792 | /* | |
793 | * Add a Multicast address to slaves | |
794 | * according to mode | |
795 | */ | |
22bedad3 | 796 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
797 | { |
798 | if (USES_PRIMARY(bond->params.mode)) { | |
799 | /* write lock already acquired */ | |
3d632c3f | 800 | if (bond->curr_active_slave) |
22bedad3 | 801 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
802 | } else { |
803 | struct slave *slave; | |
804 | int i; | |
3d632c3f SH |
805 | |
806 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 807 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
808 | } |
809 | } | |
810 | ||
811 | /* | |
812 | * Remove a multicast address from slave | |
813 | * according to mode | |
814 | */ | |
22bedad3 | 815 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
816 | { |
817 | if (USES_PRIMARY(bond->params.mode)) { | |
818 | /* write lock already acquired */ | |
3d632c3f | 819 | if (bond->curr_active_slave) |
22bedad3 | 820 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
821 | } else { |
822 | struct slave *slave; | |
823 | int i; | |
824 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 825 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
826 | } |
827 | } | |
828 | } | |
829 | ||
a816c7c7 | 830 | |
5a37e8ca | 831 | static void __bond_resend_igmp_join_requests(struct net_device *dev) |
a816c7c7 JV |
832 | { |
833 | struct in_device *in_dev; | |
a816c7c7 JV |
834 | |
835 | rcu_read_lock(); | |
5a37e8ca | 836 | in_dev = __in_dev_get_rcu(dev); |
866f3b25 | 837 | if (in_dev) |
24912420 | 838 | ip_mc_rejoin_groups(in_dev); |
a816c7c7 JV |
839 | rcu_read_unlock(); |
840 | } | |
841 | ||
5a37e8ca FL |
842 | /* |
843 | * Retrieve the list of registered multicast addresses for the bonding | |
844 | * device and retransmit an IGMP JOIN request to the current active | |
845 | * slave. | |
846 | */ | |
847 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
848 | { | |
849 | struct net_device *vlan_dev; | |
850 | struct vlan_entry *vlan; | |
851 | ||
852 | read_lock(&bond->lock); | |
853 | ||
854 | /* rejoin all groups on bond device */ | |
855 | __bond_resend_igmp_join_requests(bond->dev); | |
856 | ||
857 | /* rejoin all groups on vlan devices */ | |
858 | if (bond->vlgrp) { | |
859 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
860 | vlan_dev = vlan_group_get_device(bond->vlgrp, | |
861 | vlan->vlan_id); | |
862 | if (vlan_dev) | |
863 | __bond_resend_igmp_join_requests(vlan_dev); | |
864 | } | |
865 | } | |
866 | ||
c2952c31 FL |
867 | if (--bond->igmp_retrans > 0) |
868 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); | |
869 | ||
5a37e8ca FL |
870 | read_unlock(&bond->lock); |
871 | } | |
872 | ||
379b7383 | 873 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
874 | { |
875 | struct bonding *bond = container_of(work, struct bonding, | |
876 | mcast_work.work); | |
877 | bond_resend_igmp_join_requests(bond); | |
878 | } | |
879 | ||
1da177e4 LT |
880 | /* |
881 | * flush all members of flush->mc_list from device dev->mc_list | |
882 | */ | |
3d632c3f SH |
883 | static void bond_mc_list_flush(struct net_device *bond_dev, |
884 | struct net_device *slave_dev) | |
1da177e4 | 885 | { |
454d7c9b | 886 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 887 | struct netdev_hw_addr *ha; |
1da177e4 | 888 | |
22bedad3 JP |
889 | netdev_for_each_mc_addr(ha, bond_dev) |
890 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
891 | |
892 | if (bond->params.mode == BOND_MODE_8023AD) { | |
893 | /* del lacpdu mc addr from mc list */ | |
894 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
895 | ||
22bedad3 | 896 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
897 | } |
898 | } | |
899 | ||
900 | /*--------------------------- Active slave change ---------------------------*/ | |
901 | ||
902 | /* | |
903 | * Update the mc list and multicast-related flags for the new and | |
904 | * old active slaves (if any) according to the multicast mode, and | |
905 | * promiscuous flags unconditionally. | |
906 | */ | |
3d632c3f SH |
907 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
908 | struct slave *old_active) | |
1da177e4 | 909 | { |
22bedad3 | 910 | struct netdev_hw_addr *ha; |
1da177e4 | 911 | |
3d632c3f | 912 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
913 | /* nothing to do - mc list is already up-to-date on |
914 | * all slaves | |
915 | */ | |
916 | return; | |
1da177e4 LT |
917 | |
918 | if (old_active) { | |
3d632c3f | 919 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 920 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 921 | |
3d632c3f | 922 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 923 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 924 | |
22bedad3 JP |
925 | netdev_for_each_mc_addr(ha, bond->dev) |
926 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
927 | } |
928 | ||
929 | if (new_active) { | |
7e1a1ac1 | 930 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 931 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 932 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 933 | |
3d632c3f | 934 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 935 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 936 | |
22bedad3 JP |
937 | netdev_for_each_mc_addr(ha, bond->dev) |
938 | dev_mc_add(new_active->dev, ha->addr); | |
1da177e4 LT |
939 | } |
940 | } | |
941 | ||
3915c1e8 JV |
942 | /* |
943 | * bond_do_fail_over_mac | |
944 | * | |
945 | * Perform special MAC address swapping for fail_over_mac settings | |
946 | * | |
947 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
948 | */ | |
949 | static void bond_do_fail_over_mac(struct bonding *bond, | |
950 | struct slave *new_active, | |
951 | struct slave *old_active) | |
1f78d9f9 HE |
952 | __releases(&bond->curr_slave_lock) |
953 | __releases(&bond->lock) | |
954 | __acquires(&bond->lock) | |
955 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
956 | { |
957 | u8 tmp_mac[ETH_ALEN]; | |
958 | struct sockaddr saddr; | |
959 | int rv; | |
960 | ||
961 | switch (bond->params.fail_over_mac) { | |
962 | case BOND_FOM_ACTIVE: | |
963 | if (new_active) | |
964 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
965 | new_active->dev->addr_len); | |
966 | break; | |
967 | case BOND_FOM_FOLLOW: | |
968 | /* | |
969 | * if new_active && old_active, swap them | |
970 | * if just old_active, do nothing (going to no active slave) | |
971 | * if just new_active, set new_active to bond's MAC | |
972 | */ | |
973 | if (!new_active) | |
974 | return; | |
975 | ||
976 | write_unlock_bh(&bond->curr_slave_lock); | |
977 | read_unlock(&bond->lock); | |
978 | ||
979 | if (old_active) { | |
980 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
981 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
982 | ETH_ALEN); | |
983 | saddr.sa_family = new_active->dev->type; | |
984 | } else { | |
985 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
986 | saddr.sa_family = bond->dev->type; | |
987 | } | |
988 | ||
989 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
990 | if (rv) { | |
a4aee5c8 | 991 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
992 | bond->dev->name, -rv, new_active->dev->name); |
993 | goto out; | |
994 | } | |
995 | ||
996 | if (!old_active) | |
997 | goto out; | |
998 | ||
999 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1000 | saddr.sa_family = old_active->dev->type; | |
1001 | ||
1002 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1003 | if (rv) | |
a4aee5c8 | 1004 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1005 | bond->dev->name, -rv, new_active->dev->name); |
1006 | out: | |
1007 | read_lock(&bond->lock); | |
1008 | write_lock_bh(&bond->curr_slave_lock); | |
1009 | break; | |
1010 | default: | |
a4aee5c8 | 1011 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
1012 | bond->dev->name, bond->params.fail_over_mac); |
1013 | break; | |
1014 | } | |
1015 | ||
1016 | } | |
1017 | ||
a549952a JP |
1018 | static bool bond_should_change_active(struct bonding *bond) |
1019 | { | |
1020 | struct slave *prim = bond->primary_slave; | |
1021 | struct slave *curr = bond->curr_active_slave; | |
1022 | ||
1023 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1024 | return true; | |
1025 | if (bond->force_primary) { | |
1026 | bond->force_primary = false; | |
1027 | return true; | |
1028 | } | |
1029 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1030 | (prim->speed < curr->speed || | |
1031 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1032 | return false; | |
1033 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1034 | return false; | |
1035 | return true; | |
1036 | } | |
3915c1e8 | 1037 | |
1da177e4 LT |
1038 | /** |
1039 | * find_best_interface - select the best available slave to be the active one | |
1040 | * @bond: our bonding struct | |
1041 | * | |
1042 | * Warning: Caller must hold curr_slave_lock for writing. | |
1043 | */ | |
1044 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1045 | { | |
1046 | struct slave *new_active, *old_active; | |
1047 | struct slave *bestslave = NULL; | |
1048 | int mintime = bond->params.updelay; | |
1049 | int i; | |
1050 | ||
49b4ad92 | 1051 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1052 | |
1053 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1054 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1055 | new_active = bond->first_slave; |
3d632c3f | 1056 | else |
1da177e4 | 1057 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1058 | } |
1059 | ||
1da177e4 | 1060 | if ((bond->primary_slave) && |
a549952a JP |
1061 | bond->primary_slave->link == BOND_LINK_UP && |
1062 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1063 | new_active = bond->primary_slave; |
1064 | } | |
1065 | ||
1066 | /* remember where to stop iterating over the slaves */ | |
1067 | old_active = new_active; | |
1068 | ||
1069 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1070 | if (new_active->link == BOND_LINK_UP) { |
1071 | return new_active; | |
1072 | } else if (new_active->link == BOND_LINK_BACK && | |
1073 | IS_UP(new_active->dev)) { | |
1074 | /* link up, but waiting for stabilization */ | |
1075 | if (new_active->delay < mintime) { | |
1076 | mintime = new_active->delay; | |
1077 | bestslave = new_active; | |
1da177e4 LT |
1078 | } |
1079 | } | |
1080 | } | |
1081 | ||
1082 | return bestslave; | |
1083 | } | |
1084 | ||
1085 | /** | |
1086 | * change_active_interface - change the active slave into the specified one | |
1087 | * @bond: our bonding struct | |
1088 | * @new: the new slave to make the active one | |
1089 | * | |
1090 | * Set the new slave to the bond's settings and unset them on the old | |
1091 | * curr_active_slave. | |
1092 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1093 | * | |
1094 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1095 | * because it is apparently the best available slave we have, even though its | |
1096 | * updelay hasn't timed out yet. | |
1097 | * | |
3915c1e8 JV |
1098 | * If new_active is not NULL, caller must hold bond->lock for read and |
1099 | * curr_slave_lock for write_bh. | |
1da177e4 | 1100 | */ |
a77b5325 | 1101 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1102 | { |
1103 | struct slave *old_active = bond->curr_active_slave; | |
1104 | ||
3d632c3f | 1105 | if (old_active == new_active) |
1da177e4 | 1106 | return; |
1da177e4 LT |
1107 | |
1108 | if (new_active) { | |
b2220cad JV |
1109 | new_active->jiffies = jiffies; |
1110 | ||
1da177e4 LT |
1111 | if (new_active->link == BOND_LINK_BACK) { |
1112 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1113 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1114 | bond->dev->name, new_active->dev->name, | |
1115 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1116 | } |
1117 | ||
1118 | new_active->delay = 0; | |
1119 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1120 | |
3d632c3f | 1121 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1122 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1123 | |
58402054 | 1124 | if (bond_is_lb(bond)) |
1da177e4 | 1125 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1126 | } else { |
1127 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1128 | pr_info("%s: making interface %s the new active one.\n", |
1129 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1130 | } |
1131 | } | |
1132 | } | |
1133 | ||
3d632c3f | 1134 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1135 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1136 | |
58402054 | 1137 | if (bond_is_lb(bond)) { |
1da177e4 | 1138 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1139 | if (old_active) |
1140 | bond_set_slave_inactive_flags(old_active); | |
1141 | if (new_active) | |
1142 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1143 | } else { |
1144 | bond->curr_active_slave = new_active; | |
1145 | } | |
c3ade5ca JV |
1146 | |
1147 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1148 | if (old_active) |
c3ade5ca | 1149 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1150 | |
1151 | if (new_active) { | |
1152 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1153 | |
709f8a45 OG |
1154 | if (bond->params.fail_over_mac) |
1155 | bond_do_fail_over_mac(bond, new_active, | |
1156 | old_active); | |
3915c1e8 | 1157 | |
01f3109d OG |
1158 | write_unlock_bh(&bond->curr_slave_lock); |
1159 | read_unlock(&bond->lock); | |
1160 | ||
75c78500 | 1161 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1162 | |
1163 | read_lock(&bond->lock); | |
1164 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1165 | } |
c3ade5ca | 1166 | } |
a2fd940f | 1167 | |
5a37e8ca FL |
1168 | /* resend IGMP joins since active slave has changed or |
1169 | * all were sent on curr_active_slave */ | |
ffa95ed5 BH |
1170 | if (((USES_PRIMARY(bond->params.mode) && new_active) || |
1171 | bond->params.mode == BOND_MODE_ROUNDROBIN) && | |
1172 | netif_running(bond->dev)) { | |
c2952c31 | 1173 | bond->igmp_retrans = bond->params.resend_igmp; |
5a37e8ca | 1174 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); |
a2fd940f | 1175 | } |
1da177e4 LT |
1176 | } |
1177 | ||
1178 | /** | |
1179 | * bond_select_active_slave - select a new active slave, if needed | |
1180 | * @bond: our bonding struct | |
1181 | * | |
3d632c3f | 1182 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1183 | * - The old curr_active_slave has been released or lost its link. |
1184 | * - The primary_slave has got its link back. | |
1185 | * - A slave has got its link back and there's no old curr_active_slave. | |
1186 | * | |
3915c1e8 | 1187 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1188 | */ |
a77b5325 | 1189 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1190 | { |
1191 | struct slave *best_slave; | |
ff59c456 | 1192 | int rv; |
1da177e4 LT |
1193 | |
1194 | best_slave = bond_find_best_slave(bond); | |
1195 | if (best_slave != bond->curr_active_slave) { | |
1196 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1197 | rv = bond_set_carrier(bond); |
1198 | if (!rv) | |
1199 | return; | |
1200 | ||
1201 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1202 | pr_info("%s: first active interface up!\n", |
1203 | bond->dev->name); | |
ff59c456 | 1204 | } else { |
a4aee5c8 JP |
1205 | pr_info("%s: now running without any active interface !\n", |
1206 | bond->dev->name); | |
ff59c456 | 1207 | } |
1da177e4 LT |
1208 | } |
1209 | } | |
1210 | ||
1211 | /*--------------------------- slave list handling ---------------------------*/ | |
1212 | ||
1213 | /* | |
1214 | * This function attaches the slave to the end of list. | |
1215 | * | |
1216 | * bond->lock held for writing by caller. | |
1217 | */ | |
1218 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1219 | { | |
1220 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1221 | new_slave->next = new_slave; | |
1222 | new_slave->prev = new_slave; | |
1223 | bond->first_slave = new_slave; | |
1224 | } else { | |
1225 | new_slave->next = bond->first_slave; | |
1226 | new_slave->prev = bond->first_slave->prev; | |
1227 | new_slave->next->prev = new_slave; | |
1228 | new_slave->prev->next = new_slave; | |
1229 | } | |
1230 | ||
1231 | bond->slave_cnt++; | |
1232 | } | |
1233 | ||
1234 | /* | |
1235 | * This function detaches the slave from the list. | |
1236 | * WARNING: no check is made to verify if the slave effectively | |
1237 | * belongs to <bond>. | |
1238 | * Nothing is freed on return, structures are just unchained. | |
1239 | * If any slave pointer in bond was pointing to <slave>, | |
1240 | * it should be changed by the calling function. | |
1241 | * | |
1242 | * bond->lock held for writing by caller. | |
1243 | */ | |
1244 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1245 | { | |
3d632c3f | 1246 | if (slave->next) |
1da177e4 | 1247 | slave->next->prev = slave->prev; |
1da177e4 | 1248 | |
3d632c3f | 1249 | if (slave->prev) |
1da177e4 | 1250 | slave->prev->next = slave->next; |
1da177e4 LT |
1251 | |
1252 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1253 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1254 | bond->first_slave = slave->next; | |
1255 | } else { | |
1256 | bond->first_slave = NULL; /* slave was the last one */ | |
1257 | } | |
1258 | } | |
1259 | ||
1260 | slave->next = NULL; | |
1261 | slave->prev = NULL; | |
1262 | bond->slave_cnt--; | |
1263 | } | |
1264 | ||
f6dc31a8 | 1265 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1266 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1267 | { |
8a8efa22 AW |
1268 | struct netpoll *np; |
1269 | int err = 0; | |
f6dc31a8 | 1270 | |
8a8efa22 AW |
1271 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
1272 | err = -ENOMEM; | |
1273 | if (!np) | |
1274 | goto out; | |
1275 | ||
1276 | np->dev = slave->dev; | |
1277 | err = __netpoll_setup(np); | |
1278 | if (err) { | |
1279 | kfree(np); | |
1280 | goto out; | |
f6dc31a8 | 1281 | } |
8a8efa22 AW |
1282 | slave->np = np; |
1283 | out: | |
1284 | return err; | |
1285 | } | |
1286 | static inline void slave_disable_netpoll(struct slave *slave) | |
1287 | { | |
1288 | struct netpoll *np = slave->np; | |
1289 | ||
1290 | if (!np) | |
1291 | return; | |
1292 | ||
1293 | slave->np = NULL; | |
1294 | synchronize_rcu_bh(); | |
1295 | __netpoll_cleanup(np); | |
1296 | kfree(np); | |
1297 | } | |
1298 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1299 | { | |
1300 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1301 | return false; | |
1302 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1303 | return false; | |
1304 | return true; | |
f6dc31a8 WC |
1305 | } |
1306 | ||
1307 | static void bond_poll_controller(struct net_device *bond_dev) | |
1308 | { | |
8a8efa22 AW |
1309 | } |
1310 | ||
1311 | static void __bond_netpoll_cleanup(struct bonding *bond) | |
1312 | { | |
c2355e1a NH |
1313 | struct slave *slave; |
1314 | int i; | |
1315 | ||
8a8efa22 AW |
1316 | bond_for_each_slave(bond, slave, i) |
1317 | if (IS_UP(slave->dev)) | |
1318 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1319 | } |
f6dc31a8 WC |
1320 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1321 | { | |
1322 | struct bonding *bond = netdev_priv(bond_dev); | |
8a8efa22 AW |
1323 | |
1324 | read_lock(&bond->lock); | |
1325 | __bond_netpoll_cleanup(bond); | |
1326 | read_unlock(&bond->lock); | |
1327 | } | |
1328 | ||
1329 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) | |
1330 | { | |
1331 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1332 | struct slave *slave; |
8a8efa22 | 1333 | int i, err = 0; |
f6dc31a8 WC |
1334 | |
1335 | read_lock(&bond->lock); | |
f6dc31a8 | 1336 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1337 | err = slave_enable_netpoll(slave); |
1338 | if (err) { | |
1339 | __bond_netpoll_cleanup(bond); | |
1340 | break; | |
f6dc31a8 WC |
1341 | } |
1342 | } | |
1343 | read_unlock(&bond->lock); | |
8a8efa22 | 1344 | return err; |
f6dc31a8 WC |
1345 | } |
1346 | ||
8a8efa22 AW |
1347 | static struct netpoll_info *bond_netpoll_info(struct bonding *bond) |
1348 | { | |
1349 | return bond->dev->npinfo; | |
1350 | } | |
f6dc31a8 | 1351 | |
8a8efa22 AW |
1352 | #else |
1353 | static inline int slave_enable_netpoll(struct slave *slave) | |
1354 | { | |
1355 | return 0; | |
1356 | } | |
1357 | static inline void slave_disable_netpoll(struct slave *slave) | |
1358 | { | |
1359 | } | |
f6dc31a8 WC |
1360 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1361 | { | |
1362 | } | |
f6dc31a8 WC |
1363 | #endif |
1364 | ||
1da177e4 LT |
1365 | /*---------------------------------- IOCTL ----------------------------------*/ |
1366 | ||
4ad072c9 AB |
1367 | static int bond_sethwaddr(struct net_device *bond_dev, |
1368 | struct net_device *slave_dev) | |
1da177e4 | 1369 | { |
5a03cdb7 HE |
1370 | pr_debug("bond_dev=%p\n", bond_dev); |
1371 | pr_debug("slave_dev=%p\n", slave_dev); | |
1372 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1373 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1374 | return 0; | |
1375 | } | |
1376 | ||
7f353bf2 HX |
1377 | #define BOND_VLAN_FEATURES \ |
1378 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1379 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1380 | |
3d632c3f | 1381 | /* |
8e3babcd | 1382 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1383 | * feature bits are managed elsewhere, so preserve those feature bits |
1384 | * on the master device. | |
8531c5ff AK |
1385 | */ |
1386 | static int bond_compute_features(struct bonding *bond) | |
1387 | { | |
8531c5ff AK |
1388 | struct slave *slave; |
1389 | struct net_device *bond_dev = bond->dev; | |
04ed3e74 MM |
1390 | u32 features = bond_dev->features; |
1391 | u32 vlan_features = 0; | |
3158bf7d MS |
1392 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1393 | bond_dev->hard_header_len); | |
8e3babcd | 1394 | int i; |
8531c5ff | 1395 | |
7f353bf2 | 1396 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
c6e1a0d1 | 1397 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM | NETIF_F_NOCACHE_COPY; |
b63365a2 HX |
1398 | |
1399 | if (!bond->first_slave) | |
1400 | goto done; | |
1401 | ||
1402 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1403 | |
278339a4 | 1404 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1405 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1406 | features = netdev_increment_features(features, |
1407 | slave->dev->features, | |
1408 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1409 | vlan_features = netdev_increment_features(vlan_features, |
1410 | slave->dev->vlan_features, | |
1411 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1412 | if (slave->dev->hard_header_len > max_hard_header_len) |
1413 | max_hard_header_len = slave->dev->hard_header_len; | |
1414 | } | |
8531c5ff | 1415 | |
b63365a2 | 1416 | done: |
7f353bf2 | 1417 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
acd1130e MM |
1418 | bond_dev->features = netdev_fix_features(bond_dev, features); |
1419 | bond_dev->vlan_features = netdev_fix_features(bond_dev, vlan_features); | |
54ef3137 | 1420 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1421 | |
1422 | return 0; | |
1423 | } | |
1424 | ||
872254dd MS |
1425 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1426 | struct net_device *slave_dev) | |
1427 | { | |
454d7c9b | 1428 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1429 | |
00829823 | 1430 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1431 | |
1432 | bond_dev->type = slave_dev->type; | |
1433 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1434 | bond_dev->addr_len = slave_dev->addr_len; | |
1435 | ||
1436 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1437 | slave_dev->addr_len); | |
d90a162a | 1438 | bond->setup_by_slave = 1; |
872254dd MS |
1439 | } |
1440 | ||
5b2c4dd2 JP |
1441 | /* On bonding slaves other than the currently active slave, suppress |
1442 | * duplicates except for 802.3ad ETH_P_SLOW, alb non-mcast/bcast, and | |
1443 | * ARP on active-backup slaves with arp_validate enabled. | |
1444 | */ | |
1445 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1446 | struct slave *slave, |
1447 | struct bonding *bond) | |
5b2c4dd2 | 1448 | { |
2d7011ca | 1449 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1450 | if (slave_do_arp_validate(bond, slave) && |
5b2c4dd2 JP |
1451 | skb->protocol == __cpu_to_be16(ETH_P_ARP)) |
1452 | return false; | |
1453 | ||
0bd80dad | 1454 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1455 | skb->pkt_type != PACKET_BROADCAST && |
1456 | skb->pkt_type != PACKET_MULTICAST) | |
1457 | return false; | |
1458 | ||
0bd80dad | 1459 | if (bond->params.mode == BOND_MODE_8023AD && |
5b2c4dd2 JP |
1460 | skb->protocol == __cpu_to_be16(ETH_P_SLOW)) |
1461 | return false; | |
1462 | ||
1463 | return true; | |
1464 | } | |
1465 | return false; | |
1466 | } | |
1467 | ||
8a4eb573 | 1468 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1469 | { |
8a4eb573 | 1470 | struct sk_buff *skb = *pskb; |
f1c1775a | 1471 | struct slave *slave; |
0bd80dad | 1472 | struct bonding *bond; |
5b2c4dd2 | 1473 | |
8a4eb573 JP |
1474 | skb = skb_share_check(skb, GFP_ATOMIC); |
1475 | if (unlikely(!skb)) | |
1476 | return RX_HANDLER_CONSUMED; | |
1477 | ||
1478 | *pskb = skb; | |
5b2c4dd2 | 1479 | |
35d48903 JP |
1480 | slave = bond_slave_get_rcu(skb->dev); |
1481 | bond = slave->bond; | |
0bd80dad JP |
1482 | |
1483 | if (bond->params.arp_interval) | |
f1c1775a | 1484 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1485 | |
0bd80dad | 1486 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1487 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1488 | } |
1489 | ||
35d48903 | 1490 | skb->dev = bond->dev; |
5b2c4dd2 | 1491 | |
0bd80dad | 1492 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1493 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1494 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1495 | |
541ac7c9 CG |
1496 | if (unlikely(skb_cow_head(skb, |
1497 | skb->data - skb_mac_header(skb)))) { | |
1498 | kfree_skb(skb); | |
8a4eb573 | 1499 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1500 | } |
35d48903 | 1501 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1502 | } |
1503 | ||
8a4eb573 | 1504 | return RX_HANDLER_ANOTHER; |
5b2c4dd2 JP |
1505 | } |
1506 | ||
1da177e4 | 1507 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1508 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1509 | { |
454d7c9b | 1510 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1511 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1512 | struct slave *new_slave = NULL; |
22bedad3 | 1513 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1514 | struct sockaddr addr; |
1515 | int link_reporting; | |
1516 | int old_features = bond_dev->features; | |
1517 | int res = 0; | |
1518 | ||
552709d5 | 1519 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1520 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1521 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1522 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1523 | } |
1524 | ||
1525 | /* bond must be initialized by bond_open() before enslaving */ | |
1526 | if (!(bond_dev->flags & IFF_UP)) { | |
a4aee5c8 JP |
1527 | pr_warning("%s: master_dev is not up in bond_enslave\n", |
1528 | bond_dev->name); | |
1da177e4 LT |
1529 | } |
1530 | ||
1531 | /* already enslaved */ | |
1532 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1533 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1534 | return -EBUSY; |
1535 | } | |
1536 | ||
1537 | /* vlan challenged mutual exclusion */ | |
1538 | /* no need to lock since we're protected by rtnl_lock */ | |
1539 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1540 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
f35188fa | 1541 | if (bond->vlgrp) { |
a4aee5c8 JP |
1542 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1543 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1544 | return -EPERM; |
1545 | } else { | |
a4aee5c8 JP |
1546 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1547 | bond_dev->name, slave_dev->name, | |
1548 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1549 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1550 | } | |
1551 | } else { | |
5a03cdb7 | 1552 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1553 | if (bond->slave_cnt == 0) { |
1554 | /* First slave, and it is not VLAN challenged, | |
1555 | * so remove the block of adding VLANs over the bond. | |
1556 | */ | |
1557 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1558 | } | |
1559 | } | |
1560 | ||
217df670 JV |
1561 | /* |
1562 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1563 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1564 | * the current ifenslave will set the interface down prior to |
1565 | * enslaving it; the old ifenslave will not. | |
1566 | */ | |
1567 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1568 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1569 | slave_dev->name); |
1570 | res = -EPERM; | |
1571 | goto err_undo_flags; | |
1572 | } | |
1da177e4 | 1573 | |
872254dd MS |
1574 | /* set bonding device ether type by slave - bonding netdevices are |
1575 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1576 | * there is a need to override some of the type dependent attribs/funcs. | |
1577 | * | |
1578 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1579 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1580 | */ | |
1581 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1582 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1583 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1584 | bond_dev->name, |
1585 | bond_dev->type, slave_dev->type); | |
75c78500 | 1586 | |
3ca5b404 JP |
1587 | res = netdev_bonding_change(bond_dev, |
1588 | NETDEV_PRE_TYPE_CHANGE); | |
1589 | res = notifier_to_errno(res); | |
1590 | if (res) { | |
1591 | pr_err("%s: refused to change device type\n", | |
1592 | bond_dev->name); | |
1593 | res = -EBUSY; | |
1594 | goto err_undo_flags; | |
1595 | } | |
75c78500 | 1596 | |
32a806c1 | 1597 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1598 | dev_uc_flush(bond_dev); |
22bedad3 | 1599 | dev_mc_flush(bond_dev); |
32a806c1 | 1600 | |
e36b9d16 MS |
1601 | if (slave_dev->type != ARPHRD_ETHER) |
1602 | bond_setup_by_slave(bond_dev, slave_dev); | |
1603 | else | |
1604 | ether_setup(bond_dev); | |
75c78500 | 1605 | |
93d9b7d7 JP |
1606 | netdev_bonding_change(bond_dev, |
1607 | NETDEV_POST_TYPE_CHANGE); | |
e36b9d16 | 1608 | } |
872254dd | 1609 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1610 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1611 | slave_dev->name, | |
1612 | slave_dev->type, bond_dev->type); | |
1613 | res = -EINVAL; | |
1614 | goto err_undo_flags; | |
872254dd MS |
1615 | } |
1616 | ||
eb7cc59a | 1617 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1618 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1619 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1620 | bond_dev->name); | |
3915c1e8 JV |
1621 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1622 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1623 | 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", |
1624 | bond_dev->name); | |
2ab82852 MS |
1625 | res = -EOPNOTSUPP; |
1626 | goto err_undo_flags; | |
1627 | } | |
1da177e4 LT |
1628 | } |
1629 | ||
c20811a7 JP |
1630 | /* If this is the first slave, then we need to set the master's hardware |
1631 | * address to be the same as the slave's. */ | |
d13a2cb6 | 1632 | if (is_zero_ether_addr(bond->dev->dev_addr)) |
c20811a7 JP |
1633 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, |
1634 | slave_dev->addr_len); | |
1635 | ||
1636 | ||
243cb4e5 | 1637 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1638 | if (!new_slave) { |
1639 | res = -ENOMEM; | |
1640 | goto err_undo_flags; | |
1641 | } | |
1642 | ||
bb1d9123 AG |
1643 | /* |
1644 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1645 | * is set via sysfs or module option if desired. | |
1646 | */ | |
1647 | new_slave->queue_id = 0; | |
1648 | ||
b15ba0fb JP |
1649 | /* Save slave's original mtu and then set it to match the bond */ |
1650 | new_slave->original_mtu = slave_dev->mtu; | |
1651 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1652 | if (res) { | |
1653 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1654 | goto err_free; | |
1655 | } | |
1656 | ||
217df670 JV |
1657 | /* |
1658 | * Save slave's original ("permanent") mac address for modes | |
1659 | * that need it, and for restoring it upon release, and then | |
1660 | * set it to the master's address | |
1661 | */ | |
1662 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1663 | |
dd957c57 | 1664 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1665 | /* |
1666 | * Set slave to master's mac address. The application already | |
1667 | * set the master's mac address to that of the first slave | |
1668 | */ | |
1669 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1670 | addr.sa_family = slave_dev->type; | |
1671 | res = dev_set_mac_address(slave_dev, &addr); | |
1672 | if (res) { | |
5a03cdb7 | 1673 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1674 | goto err_restore_mtu; |
2ab82852 | 1675 | } |
217df670 | 1676 | } |
1da177e4 | 1677 | |
1765a575 | 1678 | res = netdev_set_bond_master(slave_dev, bond_dev); |
c2edacf8 | 1679 | if (res) { |
1765a575 | 1680 | pr_debug("Error %d calling netdev_set_bond_master\n", res); |
569f0c4d | 1681 | goto err_restore_mac; |
c2edacf8 | 1682 | } |
5b2c4dd2 | 1683 | |
217df670 JV |
1684 | /* open the slave since the application closed it */ |
1685 | res = dev_open(slave_dev); | |
1686 | if (res) { | |
3d632c3f | 1687 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1688 | goto err_unset_master; |
1da177e4 LT |
1689 | } |
1690 | ||
35d48903 | 1691 | new_slave->bond = bond; |
1da177e4 | 1692 | new_slave->dev = slave_dev; |
0b680e75 | 1693 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1694 | |
58402054 | 1695 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1696 | /* bond_alb_init_slave() must be called before all other stages since |
1697 | * it might fail and we do not want to have to undo everything | |
1698 | */ | |
1699 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1700 | if (res) |
569f0c4d | 1701 | goto err_close; |
1da177e4 LT |
1702 | } |
1703 | ||
1704 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1705 | * master's promisc (and mc) settings only if it becomes the | |
1706 | * curr_active_slave, and that is taken care of later when calling | |
1707 | * bond_change_active() | |
1708 | */ | |
1709 | if (!USES_PRIMARY(bond->params.mode)) { | |
1710 | /* set promiscuity level to new slave */ | |
1711 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1712 | res = dev_set_promiscuity(slave_dev, 1); |
1713 | if (res) | |
1714 | goto err_close; | |
1da177e4 LT |
1715 | } |
1716 | ||
1717 | /* set allmulti level to new slave */ | |
1718 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1719 | res = dev_set_allmulti(slave_dev, 1); |
1720 | if (res) | |
1721 | goto err_close; | |
1da177e4 LT |
1722 | } |
1723 | ||
b9e40857 | 1724 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1725 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1726 | netdev_for_each_mc_addr(ha, bond_dev) |
1727 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1728 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1729 | } |
1730 | ||
1731 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1732 | /* add lacpdu mc addr to mc list */ | |
1733 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1734 | ||
22bedad3 | 1735 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1736 | } |
1737 | ||
1738 | bond_add_vlans_on_slave(bond, slave_dev); | |
1739 | ||
1740 | write_lock_bh(&bond->lock); | |
1741 | ||
1742 | bond_attach_slave(bond, new_slave); | |
1743 | ||
1744 | new_slave->delay = 0; | |
1745 | new_slave->link_failure_count = 0; | |
1746 | ||
8531c5ff AK |
1747 | bond_compute_features(bond); |
1748 | ||
3915c1e8 JV |
1749 | write_unlock_bh(&bond->lock); |
1750 | ||
1751 | read_lock(&bond->lock); | |
1752 | ||
f5b2b966 JV |
1753 | new_slave->last_arp_rx = jiffies; |
1754 | ||
1da177e4 LT |
1755 | if (bond->params.miimon && !bond->params.use_carrier) { |
1756 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1757 | ||
1758 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1759 | /* | |
1760 | * miimon is set but a bonded network driver | |
1761 | * does not support ETHTOOL/MII and | |
1762 | * arp_interval is not set. Note: if | |
1763 | * use_carrier is enabled, we will never go | |
1764 | * here (because netif_carrier is always | |
1765 | * supported); thus, we don't need to change | |
1766 | * the messages for netif_carrier. | |
1767 | */ | |
a4aee5c8 | 1768 | 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 | 1769 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1770 | } else if (link_reporting == -1) { |
1771 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1772 | 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", |
1773 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1774 | } |
1775 | } | |
1776 | ||
1777 | /* check for initial state */ | |
1778 | if (!bond->params.miimon || | |
1779 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1780 | if (bond->params.updelay) { | |
a4aee5c8 | 1781 | pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n"); |
1da177e4 LT |
1782 | new_slave->link = BOND_LINK_BACK; |
1783 | new_slave->delay = bond->params.updelay; | |
1784 | } else { | |
a4aee5c8 | 1785 | pr_debug("Initial state of slave_dev is BOND_LINK_UP\n"); |
1da177e4 LT |
1786 | new_slave->link = BOND_LINK_UP; |
1787 | } | |
1788 | new_slave->jiffies = jiffies; | |
1789 | } else { | |
a4aee5c8 | 1790 | pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n"); |
1da177e4 LT |
1791 | new_slave->link = BOND_LINK_DOWN; |
1792 | } | |
1793 | ||
1794 | if (bond_update_speed_duplex(new_slave) && | |
1795 | (new_slave->link != BOND_LINK_DOWN)) { | |
a4aee5c8 JP |
1796 | pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n", |
1797 | bond_dev->name, new_slave->dev->name); | |
1da177e4 LT |
1798 | |
1799 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a4aee5c8 JP |
1800 | pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n", |
1801 | bond_dev->name); | |
1da177e4 LT |
1802 | } |
1803 | } | |
1804 | ||
1805 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1806 | /* if there is a primary slave, remember it */ | |
a549952a | 1807 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1808 | bond->primary_slave = new_slave; |
a549952a JP |
1809 | bond->force_primary = true; |
1810 | } | |
1da177e4 LT |
1811 | } |
1812 | ||
3915c1e8 JV |
1813 | write_lock_bh(&bond->curr_slave_lock); |
1814 | ||
1da177e4 LT |
1815 | switch (bond->params.mode) { |
1816 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1817 | bond_set_slave_inactive_flags(new_slave); |
1818 | bond_select_active_slave(bond); | |
1da177e4 LT |
1819 | break; |
1820 | case BOND_MODE_8023AD: | |
1821 | /* in 802.3ad mode, the internal mechanism | |
1822 | * will activate the slaves in the selected | |
1823 | * aggregator | |
1824 | */ | |
1825 | bond_set_slave_inactive_flags(new_slave); | |
1826 | /* if this is the first slave */ | |
1827 | if (bond->slave_cnt == 1) { | |
1828 | SLAVE_AD_INFO(new_slave).id = 1; | |
1829 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1830 | * can be called only after the mac address of the bond is set | |
1831 | */ | |
1832 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1833 | bond->params.lacp_fast); | |
1834 | } else { | |
1835 | SLAVE_AD_INFO(new_slave).id = | |
1836 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1837 | } | |
1838 | ||
1839 | bond_3ad_bind_slave(new_slave); | |
1840 | break; | |
1841 | case BOND_MODE_TLB: | |
1842 | case BOND_MODE_ALB: | |
e30bc066 | 1843 | bond_set_active_slave(new_slave); |
059fe7a5 | 1844 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1845 | bond_select_active_slave(bond); |
1da177e4 LT |
1846 | break; |
1847 | default: | |
5a03cdb7 | 1848 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1849 | |
1850 | /* always active in trunk mode */ | |
e30bc066 | 1851 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1852 | |
1853 | /* In trunking mode there is little meaning to curr_active_slave | |
1854 | * anyway (it holds no special properties of the bond device), | |
1855 | * so we can change it without calling change_active_interface() | |
1856 | */ | |
3d632c3f | 1857 | if (!bond->curr_active_slave) |
1da177e4 | 1858 | bond->curr_active_slave = new_slave; |
3d632c3f | 1859 | |
1da177e4 LT |
1860 | break; |
1861 | } /* switch(bond_mode) */ | |
1862 | ||
3915c1e8 JV |
1863 | write_unlock_bh(&bond->curr_slave_lock); |
1864 | ||
ff59c456 JV |
1865 | bond_set_carrier(bond); |
1866 | ||
f6dc31a8 | 1867 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 AW |
1868 | slave_dev->npinfo = bond_netpoll_info(bond); |
1869 | if (slave_dev->npinfo) { | |
1870 | if (slave_enable_netpoll(new_slave)) { | |
1871 | read_unlock(&bond->lock); | |
1872 | pr_info("Error, %s: master_dev is using netpoll, " | |
1873 | "but new slave device does not support netpoll.\n", | |
1874 | bond_dev->name); | |
1875 | res = -EBUSY; | |
1876 | goto err_close; | |
1877 | } | |
f6dc31a8 WC |
1878 | } |
1879 | #endif | |
8a8efa22 | 1880 | |
3915c1e8 | 1881 | read_unlock(&bond->lock); |
1da177e4 | 1882 | |
b76cdba9 MW |
1883 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1884 | if (res) | |
569f0c4d | 1885 | goto err_close; |
b76cdba9 | 1886 | |
35d48903 JP |
1887 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1888 | new_slave); | |
1889 | if (res) { | |
1890 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1891 | goto err_dest_symlinks; | |
1892 | } | |
1893 | ||
a4aee5c8 JP |
1894 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1895 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1896 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1897 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1898 | |
1899 | /* enslave is successful */ | |
1900 | return 0; | |
1901 | ||
1902 | /* Undo stages on error */ | |
35d48903 JP |
1903 | err_dest_symlinks: |
1904 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1905 | ||
1da177e4 LT |
1906 | err_close: |
1907 | dev_close(slave_dev); | |
1908 | ||
569f0c4d | 1909 | err_unset_master: |
1765a575 | 1910 | netdev_set_bond_master(slave_dev, NULL); |
569f0c4d | 1911 | |
1da177e4 | 1912 | err_restore_mac: |
dd957c57 | 1913 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1914 | /* XXX TODO - fom follow mode needs to change master's |
1915 | * MAC if this slave's MAC is in use by the bond, or at | |
1916 | * least print a warning. | |
1917 | */ | |
2ab82852 MS |
1918 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1919 | addr.sa_family = slave_dev->type; | |
1920 | dev_set_mac_address(slave_dev, &addr); | |
1921 | } | |
1da177e4 | 1922 | |
b15ba0fb JP |
1923 | err_restore_mtu: |
1924 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1925 | ||
1da177e4 LT |
1926 | err_free: |
1927 | kfree(new_slave); | |
1928 | ||
1929 | err_undo_flags: | |
1930 | bond_dev->features = old_features; | |
3d632c3f | 1931 | |
1da177e4 LT |
1932 | return res; |
1933 | } | |
1934 | ||
1935 | /* | |
1936 | * Try to release the slave device <slave> from the bond device <master> | |
1937 | * It is legal to access curr_active_slave without a lock because all the function | |
1938 | * is write-locked. | |
1939 | * | |
1940 | * The rules for slave state should be: | |
1941 | * for Active/Backup: | |
1942 | * Active stays on all backups go down | |
1943 | * for Bonded connections: | |
1944 | * The first up interface should be left on and all others downed. | |
1945 | */ | |
a77b5325 | 1946 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1947 | { |
454d7c9b | 1948 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1949 | struct slave *slave, *oldcurrent; |
1950 | struct sockaddr addr; | |
1da177e4 LT |
1951 | |
1952 | /* slave is not a slave or master is not master of this slave */ | |
1953 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1954 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1955 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1956 | bond_dev->name, slave_dev->name); |
1957 | return -EINVAL; | |
1958 | } | |
1959 | ||
e843fa50 | 1960 | block_netpoll_tx(); |
f6dc31a8 | 1961 | netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE); |
1da177e4 LT |
1962 | write_lock_bh(&bond->lock); |
1963 | ||
1964 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1965 | if (!slave) { | |
1966 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1967 | pr_info("%s: %s not enslaved\n", |
1968 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1969 | write_unlock_bh(&bond->lock); |
e843fa50 | 1970 | unblock_netpoll_tx(); |
1da177e4 LT |
1971 | return -EINVAL; |
1972 | } | |
1973 | ||
35d48903 JP |
1974 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1975 | * for this slave anymore. | |
1976 | */ | |
1977 | netdev_rx_handler_unregister(slave_dev); | |
1978 | write_unlock_bh(&bond->lock); | |
1979 | synchronize_net(); | |
1980 | write_lock_bh(&bond->lock); | |
1981 | ||
3915c1e8 | 1982 | if (!bond->params.fail_over_mac) { |
8e95a202 JP |
1983 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) && |
1984 | bond->slave_cnt > 1) | |
a4aee5c8 JP |
1985 | 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", |
1986 | bond_dev->name, slave_dev->name, | |
1987 | slave->perm_hwaddr, | |
1988 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1989 | } |
1990 | ||
1991 | /* Inform AD package of unbinding of slave. */ | |
1992 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1993 | /* must be called before the slave is | |
1994 | * detached from the list | |
1995 | */ | |
1996 | bond_3ad_unbind_slave(slave); | |
1997 | } | |
1998 | ||
a4aee5c8 JP |
1999 | pr_info("%s: releasing %s interface %s\n", |
2000 | bond_dev->name, | |
e30bc066 | 2001 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 2002 | slave_dev->name); |
1da177e4 LT |
2003 | |
2004 | oldcurrent = bond->curr_active_slave; | |
2005 | ||
2006 | bond->current_arp_slave = NULL; | |
2007 | ||
2008 | /* release the slave from its bond */ | |
2009 | bond_detach_slave(bond, slave); | |
2010 | ||
8531c5ff AK |
2011 | bond_compute_features(bond); |
2012 | ||
3d632c3f | 2013 | if (bond->primary_slave == slave) |
1da177e4 | 2014 | bond->primary_slave = NULL; |
1da177e4 | 2015 | |
3d632c3f | 2016 | if (oldcurrent == slave) |
1da177e4 | 2017 | bond_change_active_slave(bond, NULL); |
1da177e4 | 2018 | |
58402054 | 2019 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2020 | /* Must be called only after the slave has been |
2021 | * detached from the list and the curr_active_slave | |
2022 | * has been cleared (if our_slave == old_current), | |
2023 | * but before a new active slave is selected. | |
2024 | */ | |
2543331d | 2025 | write_unlock_bh(&bond->lock); |
1da177e4 | 2026 | bond_alb_deinit_slave(bond, slave); |
2543331d | 2027 | write_lock_bh(&bond->lock); |
1da177e4 LT |
2028 | } |
2029 | ||
059fe7a5 JV |
2030 | if (oldcurrent == slave) { |
2031 | /* | |
2032 | * Note that we hold RTNL over this sequence, so there | |
2033 | * is no concern that another slave add/remove event | |
2034 | * will interfere. | |
2035 | */ | |
2036 | write_unlock_bh(&bond->lock); | |
2037 | read_lock(&bond->lock); | |
2038 | write_lock_bh(&bond->curr_slave_lock); | |
2039 | ||
1da177e4 LT |
2040 | bond_select_active_slave(bond); |
2041 | ||
059fe7a5 JV |
2042 | write_unlock_bh(&bond->curr_slave_lock); |
2043 | read_unlock(&bond->lock); | |
2044 | write_lock_bh(&bond->lock); | |
2045 | } | |
2046 | ||
1da177e4 | 2047 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
2048 | bond_set_carrier(bond); |
2049 | ||
1da177e4 LT |
2050 | /* if the last slave was removed, zero the mac address |
2051 | * of the master so it will be set by the application | |
2052 | * to the mac address of the first slave | |
2053 | */ | |
2054 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2055 | ||
f35188fa | 2056 | if (!bond->vlgrp) { |
1da177e4 LT |
2057 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
2058 | } else { | |
a4aee5c8 JP |
2059 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2060 | bond_dev->name, bond_dev->name); | |
2061 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2062 | bond_dev->name); | |
1da177e4 LT |
2063 | } |
2064 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2065 | !bond_has_challenged_slaves(bond)) { | |
a4aee5c8 JP |
2066 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", |
2067 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
2068 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
2069 | } | |
2070 | ||
2071 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2072 | unblock_netpoll_tx(); |
1da177e4 | 2073 | |
b76cdba9 MW |
2074 | /* must do this from outside any spinlocks */ |
2075 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2076 | ||
1da177e4 LT |
2077 | bond_del_vlans_from_slave(bond, slave_dev); |
2078 | ||
2079 | /* If the mode USES_PRIMARY, then we should only remove its | |
2080 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2081 | * already taken care of above when we detached the slave | |
2082 | */ | |
2083 | if (!USES_PRIMARY(bond->params.mode)) { | |
2084 | /* unset promiscuity level from slave */ | |
3d632c3f | 2085 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2086 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2087 | |
2088 | /* unset allmulti level from slave */ | |
3d632c3f | 2089 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2090 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2091 | |
2092 | /* flush master's mc_list from slave */ | |
b9e40857 | 2093 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2094 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2095 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2096 | } |
2097 | ||
1765a575 | 2098 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2099 | |
8a8efa22 | 2100 | slave_disable_netpoll(slave); |
f6dc31a8 | 2101 | |
1da177e4 LT |
2102 | /* close slave before restoring its mac address */ |
2103 | dev_close(slave_dev); | |
2104 | ||
3915c1e8 | 2105 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2106 | /* restore original ("permanent") mac address */ |
2107 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2108 | addr.sa_family = slave_dev->type; | |
2109 | dev_set_mac_address(slave_dev, &addr); | |
2110 | } | |
1da177e4 | 2111 | |
b15ba0fb JP |
2112 | dev_set_mtu(slave_dev, slave->original_mtu); |
2113 | ||
2d7011ca | 2114 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2115 | |
2116 | kfree(slave); | |
2117 | ||
2118 | return 0; /* deletion OK */ | |
2119 | } | |
2120 | ||
d90a162a | 2121 | /* |
dadaa10b | 2122 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2123 | * Must be under rtnl_lock when this function is called. |
2124 | */ | |
26d8ee75 | 2125 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2126 | struct net_device *slave_dev) | |
d90a162a | 2127 | { |
454d7c9b | 2128 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2129 | int ret; |
2130 | ||
2131 | ret = bond_release(bond_dev, slave_dev); | |
2132 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2133 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2134 | pr_info("%s: destroying bond %s.\n", |
2135 | bond_dev->name, bond_dev->name); | |
9e71626c | 2136 | unregister_netdevice(bond_dev); |
d90a162a MS |
2137 | } |
2138 | return ret; | |
2139 | } | |
2140 | ||
1da177e4 LT |
2141 | /* |
2142 | * This function releases all slaves. | |
2143 | */ | |
2144 | static int bond_release_all(struct net_device *bond_dev) | |
2145 | { | |
454d7c9b | 2146 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2147 | struct slave *slave; |
2148 | struct net_device *slave_dev; | |
2149 | struct sockaddr addr; | |
2150 | ||
2151 | write_lock_bh(&bond->lock); | |
2152 | ||
ff59c456 JV |
2153 | netif_carrier_off(bond_dev); |
2154 | ||
3d632c3f | 2155 | if (bond->slave_cnt == 0) |
1da177e4 | 2156 | goto out; |
1da177e4 LT |
2157 | |
2158 | bond->current_arp_slave = NULL; | |
2159 | bond->primary_slave = NULL; | |
2160 | bond_change_active_slave(bond, NULL); | |
2161 | ||
2162 | while ((slave = bond->first_slave) != NULL) { | |
2163 | /* Inform AD package of unbinding of slave | |
2164 | * before slave is detached from the list. | |
2165 | */ | |
3d632c3f | 2166 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2167 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2168 | |
2169 | slave_dev = slave->dev; | |
2170 | bond_detach_slave(bond, slave); | |
2171 | ||
2543331d JV |
2172 | /* now that the slave is detached, unlock and perform |
2173 | * all the undo steps that should not be called from | |
2174 | * within a lock. | |
2175 | */ | |
2176 | write_unlock_bh(&bond->lock); | |
2177 | ||
35d48903 JP |
2178 | /* unregister rx_handler early so bond_handle_frame wouldn't |
2179 | * be called for this slave anymore. | |
2180 | */ | |
2181 | netdev_rx_handler_unregister(slave_dev); | |
2182 | synchronize_net(); | |
2183 | ||
58402054 | 2184 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2185 | /* must be called only after the slave |
2186 | * has been detached from the list | |
2187 | */ | |
2188 | bond_alb_deinit_slave(bond, slave); | |
2189 | } | |
2190 | ||
8531c5ff AK |
2191 | bond_compute_features(bond); |
2192 | ||
b76cdba9 | 2193 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2194 | bond_del_vlans_from_slave(bond, slave_dev); |
2195 | ||
2196 | /* If the mode USES_PRIMARY, then we should only remove its | |
2197 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2198 | * already taken care of above when we detached the slave | |
2199 | */ | |
2200 | if (!USES_PRIMARY(bond->params.mode)) { | |
2201 | /* unset promiscuity level from slave */ | |
3d632c3f | 2202 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2203 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2204 | |
2205 | /* unset allmulti level from slave */ | |
3d632c3f | 2206 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2207 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2208 | |
2209 | /* flush master's mc_list from slave */ | |
b9e40857 | 2210 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2211 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2212 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2213 | } |
2214 | ||
1765a575 | 2215 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2216 | |
8a8efa22 AW |
2217 | slave_disable_netpoll(slave); |
2218 | ||
1da177e4 LT |
2219 | /* close slave before restoring its mac address */ |
2220 | dev_close(slave_dev); | |
2221 | ||
dd957c57 | 2222 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2223 | /* restore original ("permanent") mac address*/ |
2224 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2225 | addr.sa_family = slave_dev->type; | |
2226 | dev_set_mac_address(slave_dev, &addr); | |
2227 | } | |
1da177e4 | 2228 | |
1da177e4 LT |
2229 | kfree(slave); |
2230 | ||
2231 | /* re-acquire the lock before getting the next slave */ | |
2232 | write_lock_bh(&bond->lock); | |
2233 | } | |
2234 | ||
2235 | /* zero the mac address of the master so it will be | |
2236 | * set by the application to the mac address of the | |
2237 | * first slave | |
2238 | */ | |
2239 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2240 | ||
f35188fa | 2241 | if (!bond->vlgrp) { |
1da177e4 | 2242 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
f35188fa | 2243 | } else { |
a4aee5c8 JP |
2244 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2245 | bond_dev->name, bond_dev->name); | |
2246 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2247 | bond_dev->name); | |
1da177e4 LT |
2248 | } |
2249 | ||
a4aee5c8 | 2250 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2251 | |
2252 | out: | |
2253 | write_unlock_bh(&bond->lock); | |
1da177e4 LT |
2254 | return 0; |
2255 | } | |
2256 | ||
2257 | /* | |
2258 | * This function changes the active slave to slave <slave_dev>. | |
2259 | * It returns -EINVAL in the following cases. | |
2260 | * - <slave_dev> is not found in the list. | |
2261 | * - There is not active slave now. | |
2262 | * - <slave_dev> is already active. | |
2263 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2264 | * - <slave_dev> is not running. | |
3d632c3f SH |
2265 | * In these cases, this function does nothing. |
2266 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2267 | */ |
2268 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2269 | { | |
454d7c9b | 2270 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2271 | struct slave *old_active = NULL; |
2272 | struct slave *new_active = NULL; | |
2273 | int res = 0; | |
2274 | ||
3d632c3f | 2275 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2276 | return -EINVAL; |
1da177e4 LT |
2277 | |
2278 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2279 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2280 | return -EINVAL; |
1da177e4 | 2281 | |
059fe7a5 | 2282 | read_lock(&bond->lock); |
1da177e4 | 2283 | |
059fe7a5 | 2284 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2285 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2286 | read_unlock(&bond->curr_slave_lock); |
2287 | ||
1da177e4 LT |
2288 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2289 | ||
2290 | /* | |
2291 | * Changing to the current active: do nothing; return success. | |
2292 | */ | |
2293 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2294 | read_unlock(&bond->lock); |
1da177e4 LT |
2295 | return 0; |
2296 | } | |
2297 | ||
2298 | if ((new_active) && | |
2299 | (old_active) && | |
2300 | (new_active->link == BOND_LINK_UP) && | |
2301 | IS_UP(new_active->dev)) { | |
e843fa50 | 2302 | block_netpoll_tx(); |
059fe7a5 | 2303 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2304 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2305 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2306 | unblock_netpoll_tx(); |
3d632c3f | 2307 | } else |
1da177e4 | 2308 | res = -EINVAL; |
1da177e4 | 2309 | |
059fe7a5 | 2310 | read_unlock(&bond->lock); |
1da177e4 LT |
2311 | |
2312 | return res; | |
2313 | } | |
2314 | ||
1da177e4 LT |
2315 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2316 | { | |
454d7c9b | 2317 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2318 | |
2319 | info->bond_mode = bond->params.mode; | |
2320 | info->miimon = bond->params.miimon; | |
2321 | ||
6603a6f2 | 2322 | read_lock(&bond->lock); |
1da177e4 | 2323 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2324 | read_unlock(&bond->lock); |
1da177e4 LT |
2325 | |
2326 | return 0; | |
2327 | } | |
2328 | ||
2329 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2330 | { | |
454d7c9b | 2331 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2332 | struct slave *slave; |
689c96cc | 2333 | int i, res = -ENODEV; |
1da177e4 | 2334 | |
6603a6f2 | 2335 | read_lock(&bond->lock); |
1da177e4 LT |
2336 | |
2337 | bond_for_each_slave(bond, slave, i) { | |
2338 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2339 | res = 0; |
2340 | strcpy(info->slave_name, slave->dev->name); | |
2341 | info->link = slave->link; | |
e30bc066 | 2342 | info->state = bond_slave_state(slave); |
689c96cc | 2343 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2344 | break; |
2345 | } | |
2346 | } | |
2347 | ||
6603a6f2 | 2348 | read_unlock(&bond->lock); |
1da177e4 | 2349 | |
689c96cc | 2350 | return res; |
1da177e4 LT |
2351 | } |
2352 | ||
2353 | /*-------------------------------- Monitoring -------------------------------*/ | |
2354 | ||
1da177e4 | 2355 | |
f0c76d61 JV |
2356 | static int bond_miimon_inspect(struct bonding *bond) |
2357 | { | |
2358 | struct slave *slave; | |
2359 | int i, link_state, commit = 0; | |
41f89100 JP |
2360 | bool ignore_updelay; |
2361 | ||
2362 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2363 | |
2364 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2365 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2366 | |
f0c76d61 | 2367 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2368 | |
2369 | switch (slave->link) { | |
f0c76d61 JV |
2370 | case BOND_LINK_UP: |
2371 | if (link_state) | |
2372 | continue; | |
1da177e4 | 2373 | |
f0c76d61 JV |
2374 | slave->link = BOND_LINK_FAIL; |
2375 | slave->delay = bond->params.downdelay; | |
2376 | if (slave->delay) { | |
a4aee5c8 JP |
2377 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2378 | bond->dev->name, | |
2379 | (bond->params.mode == | |
2380 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2381 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2382 | "active " : "backup ") : "", |
2383 | slave->dev->name, | |
2384 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2385 | } |
f0c76d61 JV |
2386 | /*FALLTHRU*/ |
2387 | case BOND_LINK_FAIL: | |
2388 | if (link_state) { | |
2389 | /* | |
2390 | * recovered before downdelay expired | |
2391 | */ | |
2392 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2393 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2394 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2395 | bond->dev->name, | |
2396 | (bond->params.downdelay - slave->delay) * | |
2397 | bond->params.miimon, | |
2398 | slave->dev->name); | |
f0c76d61 | 2399 | continue; |
1da177e4 | 2400 | } |
f0c76d61 JV |
2401 | |
2402 | if (slave->delay <= 0) { | |
2403 | slave->new_link = BOND_LINK_DOWN; | |
2404 | commit++; | |
2405 | continue; | |
1da177e4 | 2406 | } |
1da177e4 | 2407 | |
f0c76d61 JV |
2408 | slave->delay--; |
2409 | break; | |
2410 | ||
2411 | case BOND_LINK_DOWN: | |
2412 | if (!link_state) | |
2413 | continue; | |
2414 | ||
2415 | slave->link = BOND_LINK_BACK; | |
2416 | slave->delay = bond->params.updelay; | |
2417 | ||
2418 | if (slave->delay) { | |
a4aee5c8 JP |
2419 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2420 | bond->dev->name, slave->dev->name, | |
2421 | ignore_updelay ? 0 : | |
2422 | bond->params.updelay * | |
2423 | bond->params.miimon); | |
f0c76d61 JV |
2424 | } |
2425 | /*FALLTHRU*/ | |
2426 | case BOND_LINK_BACK: | |
2427 | if (!link_state) { | |
2428 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2429 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2430 | bond->dev->name, | |
2431 | (bond->params.updelay - slave->delay) * | |
2432 | bond->params.miimon, | |
2433 | slave->dev->name); | |
f0c76d61 JV |
2434 | |
2435 | continue; | |
2436 | } | |
2437 | ||
41f89100 JP |
2438 | if (ignore_updelay) |
2439 | slave->delay = 0; | |
2440 | ||
f0c76d61 JV |
2441 | if (slave->delay <= 0) { |
2442 | slave->new_link = BOND_LINK_UP; | |
2443 | commit++; | |
41f89100 | 2444 | ignore_updelay = false; |
f0c76d61 | 2445 | continue; |
1da177e4 | 2446 | } |
f0c76d61 JV |
2447 | |
2448 | slave->delay--; | |
1da177e4 | 2449 | break; |
f0c76d61 JV |
2450 | } |
2451 | } | |
1da177e4 | 2452 | |
f0c76d61 JV |
2453 | return commit; |
2454 | } | |
1da177e4 | 2455 | |
f0c76d61 JV |
2456 | static void bond_miimon_commit(struct bonding *bond) |
2457 | { | |
2458 | struct slave *slave; | |
2459 | int i; | |
2460 | ||
2461 | bond_for_each_slave(bond, slave, i) { | |
2462 | switch (slave->new_link) { | |
2463 | case BOND_LINK_NOCHANGE: | |
2464 | continue; | |
1da177e4 | 2465 | |
f0c76d61 JV |
2466 | case BOND_LINK_UP: |
2467 | slave->link = BOND_LINK_UP; | |
2468 | slave->jiffies = jiffies; | |
2469 | ||
2470 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2471 | /* prevent it from being the active one */ | |
e30bc066 | 2472 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2473 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2474 | /* make it immediately active */ | |
e30bc066 | 2475 | bond_set_active_slave(slave); |
f0c76d61 JV |
2476 | } else if (slave != bond->primary_slave) { |
2477 | /* prevent it from being the active one */ | |
e30bc066 | 2478 | bond_set_backup_slave(slave); |
1da177e4 | 2479 | } |
1da177e4 | 2480 | |
546add79 KPO |
2481 | bond_update_speed_duplex(slave); |
2482 | ||
5d30530e | 2483 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 KPO |
2484 | bond->dev->name, slave->dev->name, |
2485 | slave->speed, slave->duplex ? "full" : "half"); | |
1da177e4 | 2486 | |
f0c76d61 JV |
2487 | /* notify ad that the link status has changed */ |
2488 | if (bond->params.mode == BOND_MODE_8023AD) | |
2489 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2490 | |
58402054 | 2491 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2492 | bond_alb_handle_link_change(bond, slave, |
2493 | BOND_LINK_UP); | |
1da177e4 | 2494 | |
f0c76d61 JV |
2495 | if (!bond->curr_active_slave || |
2496 | (slave == bond->primary_slave)) | |
2497 | goto do_failover; | |
1da177e4 | 2498 | |
f0c76d61 | 2499 | continue; |
059fe7a5 | 2500 | |
f0c76d61 | 2501 | case BOND_LINK_DOWN: |
fba4acda JV |
2502 | if (slave->link_failure_count < UINT_MAX) |
2503 | slave->link_failure_count++; | |
2504 | ||
f0c76d61 | 2505 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2506 | |
f0c76d61 JV |
2507 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2508 | bond->params.mode == BOND_MODE_8023AD) | |
2509 | bond_set_slave_inactive_flags(slave); | |
2510 | ||
a4aee5c8 JP |
2511 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2512 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2513 | |
2514 | if (bond->params.mode == BOND_MODE_8023AD) | |
2515 | bond_3ad_handle_link_change(slave, | |
2516 | BOND_LINK_DOWN); | |
2517 | ||
ae63e808 | 2518 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2519 | bond_alb_handle_link_change(bond, slave, |
2520 | BOND_LINK_DOWN); | |
2521 | ||
2522 | if (slave == bond->curr_active_slave) | |
2523 | goto do_failover; | |
2524 | ||
2525 | continue; | |
2526 | ||
2527 | default: | |
a4aee5c8 | 2528 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2529 | bond->dev->name, slave->new_link, |
2530 | slave->dev->name); | |
2531 | slave->new_link = BOND_LINK_NOCHANGE; | |
2532 | ||
2533 | continue; | |
2534 | } | |
2535 | ||
2536 | do_failover: | |
2537 | ASSERT_RTNL(); | |
e843fa50 | 2538 | block_netpoll_tx(); |
f0c76d61 JV |
2539 | write_lock_bh(&bond->curr_slave_lock); |
2540 | bond_select_active_slave(bond); | |
2541 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2542 | unblock_netpoll_tx(); |
f0c76d61 JV |
2543 | } |
2544 | ||
2545 | bond_set_carrier(bond); | |
1da177e4 LT |
2546 | } |
2547 | ||
0b0eef66 JV |
2548 | /* |
2549 | * bond_mii_monitor | |
2550 | * | |
2551 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2552 | * inspection, then (if inspection indicates something needs to be done) |
2553 | * an acquisition of appropriate locks followed by a commit phase to | |
2554 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2555 | */ |
2556 | void bond_mii_monitor(struct work_struct *work) | |
2557 | { | |
2558 | struct bonding *bond = container_of(work, struct bonding, | |
2559 | mii_work.work); | |
0b0eef66 JV |
2560 | |
2561 | read_lock(&bond->lock); | |
f0c76d61 JV |
2562 | if (bond->kill_timers) |
2563 | goto out; | |
2564 | ||
2565 | if (bond->slave_cnt == 0) | |
2566 | goto re_arm; | |
b59f9f74 | 2567 | |
f0c76d61 | 2568 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2569 | read_unlock(&bond->lock); |
2570 | rtnl_lock(); | |
2571 | read_lock(&bond->lock); | |
f0c76d61 JV |
2572 | |
2573 | bond_miimon_commit(bond); | |
2574 | ||
5655662d JV |
2575 | read_unlock(&bond->lock); |
2576 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2577 | read_lock(&bond->lock); | |
0b0eef66 JV |
2578 | } |
2579 | ||
f0c76d61 JV |
2580 | re_arm: |
2581 | if (bond->params.miimon) | |
2582 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2583 | msecs_to_jiffies(bond->params.miimon)); | |
2584 | out: | |
0b0eef66 | 2585 | read_unlock(&bond->lock); |
0b0eef66 | 2586 | } |
c3ade5ca | 2587 | |
d3bb52b0 | 2588 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2589 | { |
2590 | struct in_device *idev; | |
2591 | struct in_ifaddr *ifa; | |
a144ea4b | 2592 | __be32 addr = 0; |
c3ade5ca JV |
2593 | |
2594 | if (!dev) | |
2595 | return 0; | |
2596 | ||
2597 | rcu_read_lock(); | |
e5ed6399 | 2598 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2599 | if (!idev) |
2600 | goto out; | |
2601 | ||
2602 | ifa = idev->ifa_list; | |
2603 | if (!ifa) | |
2604 | goto out; | |
2605 | ||
2606 | addr = ifa->ifa_local; | |
2607 | out: | |
2608 | rcu_read_unlock(); | |
2609 | return addr; | |
2610 | } | |
2611 | ||
d3bb52b0 | 2612 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2613 | { |
0883beca | 2614 | struct vlan_entry *vlan; |
f5b2b966 JV |
2615 | |
2616 | if (ip == bond->master_ip) | |
2617 | return 1; | |
2618 | ||
0883beca | 2619 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2620 | if (ip == vlan->vlan_ip) |
2621 | return 1; | |
2622 | } | |
2623 | ||
2624 | return 0; | |
2625 | } | |
2626 | ||
c3ade5ca JV |
2627 | /* |
2628 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2629 | * switches in VLAN mode (especially if ports are configured as | |
2630 | * "native" to a VLAN) might not pass non-tagged frames. | |
2631 | */ | |
d3bb52b0 | 2632 | 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 |
2633 | { |
2634 | struct sk_buff *skb; | |
2635 | ||
5a03cdb7 | 2636 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2637 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2638 | |
c3ade5ca JV |
2639 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2640 | NULL, slave_dev->dev_addr, NULL); | |
2641 | ||
2642 | if (!skb) { | |
a4aee5c8 | 2643 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2644 | return; |
2645 | } | |
2646 | if (vlan_id) { | |
2647 | skb = vlan_put_tag(skb, vlan_id); | |
2648 | if (!skb) { | |
a4aee5c8 | 2649 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2650 | return; |
2651 | } | |
2652 | } | |
2653 | arp_xmit(skb); | |
2654 | } | |
2655 | ||
2656 | ||
1da177e4 LT |
2657 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2658 | { | |
b23dd4fe | 2659 | int i, vlan_id; |
d3bb52b0 | 2660 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2661 | struct vlan_entry *vlan; |
c3ade5ca | 2662 | struct net_device *vlan_dev; |
c3ade5ca | 2663 | struct rtable *rt; |
1da177e4 | 2664 | |
6b780567 MW |
2665 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2666 | if (!targets[i]) | |
5a31bec0 | 2667 | break; |
5a03cdb7 | 2668 | pr_debug("basa: target %x\n", targets[i]); |
f35188fa | 2669 | if (!bond->vlgrp) { |
5a03cdb7 | 2670 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2671 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2672 | bond->master_ip, 0); | |
2673 | continue; | |
2674 | } | |
2675 | ||
2676 | /* | |
2677 | * If VLANs are configured, we do a route lookup to | |
2678 | * determine which VLAN interface would be used, so we | |
2679 | * can tag the ARP with the proper VLAN tag. | |
2680 | */ | |
78fbfd8a DM |
2681 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2682 | RTO_ONLINK, 0); | |
b23dd4fe | 2683 | if (IS_ERR(rt)) { |
c3ade5ca | 2684 | if (net_ratelimit()) { |
a4aee5c8 | 2685 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2686 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2687 | } |
2688 | continue; | |
2689 | } | |
2690 | ||
2691 | /* | |
2692 | * This target is not on a VLAN | |
2693 | */ | |
d8d1f30b | 2694 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2695 | ip_rt_put(rt); |
5a03cdb7 | 2696 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2697 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2698 | bond->master_ip, 0); | |
2699 | continue; | |
2700 | } | |
2701 | ||
2702 | vlan_id = 0; | |
0883beca | 2703 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2704 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
d8d1f30b | 2705 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2706 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2707 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2708 | vlan_dev->name, vlan_id); |
2709 | break; | |
2710 | } | |
2711 | } | |
2712 | ||
2713 | if (vlan_id) { | |
ed4b9f80 | 2714 | ip_rt_put(rt); |
c3ade5ca JV |
2715 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2716 | vlan->vlan_ip, vlan_id); | |
2717 | continue; | |
2718 | } | |
2719 | ||
2720 | if (net_ratelimit()) { | |
a4aee5c8 | 2721 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2722 | bond->dev->name, &targets[i], |
d8d1f30b | 2723 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2724 | } |
ed4b9f80 | 2725 | ip_rt_put(rt); |
c3ade5ca JV |
2726 | } |
2727 | } | |
2728 | ||
d3bb52b0 | 2729 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2730 | { |
2731 | int i; | |
d3bb52b0 | 2732 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2733 | |
f5b2b966 | 2734 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2735 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2736 | &sip, &tip, i, &targets[i], |
2737 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2738 | if (sip == targets[i]) { |
2739 | if (bond_has_this_ip(bond, tip)) | |
2740 | slave->last_arp_rx = jiffies; | |
2741 | return; | |
2742 | } | |
2743 | } | |
2744 | } | |
2745 | ||
2746 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2747 | { | |
2748 | struct arphdr *arp; | |
2749 | struct slave *slave; | |
2750 | struct bonding *bond; | |
2751 | unsigned char *arp_ptr; | |
d3bb52b0 | 2752 | __be32 sip, tip; |
f5b2b966 | 2753 | |
1f3c8804 AG |
2754 | if (dev->priv_flags & IFF_802_1Q_VLAN) { |
2755 | /* | |
2756 | * When using VLANS and bonding, dev and oriv_dev may be | |
2757 | * incorrect if the physical interface supports VLAN | |
2758 | * acceleration. With this change ARP validation now | |
2759 | * works for hosts only reachable on the VLAN interface. | |
2760 | */ | |
2761 | dev = vlan_dev_real_dev(dev); | |
2762 | orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif); | |
2763 | } | |
2764 | ||
f5b2b966 JV |
2765 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2766 | goto out; | |
2767 | ||
454d7c9b | 2768 | bond = netdev_priv(dev); |
f5b2b966 JV |
2769 | read_lock(&bond->lock); |
2770 | ||
5a03cdb7 | 2771 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
a4aee5c8 JP |
2772 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2773 | orig_dev ? orig_dev->name : "NULL"); | |
f5b2b966 JV |
2774 | |
2775 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2776 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2777 | goto out_unlock; | |
2778 | ||
b3053251 NH |
2779 | skb = skb_share_check(skb, GFP_ATOMIC); |
2780 | if (!skb) | |
2781 | goto out_unlock; | |
2782 | ||
988b7050 | 2783 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2784 | goto out_unlock; |
2785 | ||
d0a92be0 | 2786 | arp = arp_hdr(skb); |
f5b2b966 JV |
2787 | if (arp->ar_hln != dev->addr_len || |
2788 | skb->pkt_type == PACKET_OTHERHOST || | |
2789 | skb->pkt_type == PACKET_LOOPBACK || | |
2790 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2791 | arp->ar_pro != htons(ETH_P_IP) || | |
2792 | arp->ar_pln != 4) | |
2793 | goto out_unlock; | |
2794 | ||
2795 | arp_ptr = (unsigned char *)(arp + 1); | |
2796 | arp_ptr += dev->addr_len; | |
2797 | memcpy(&sip, arp_ptr, 4); | |
2798 | arp_ptr += 4 + dev->addr_len; | |
2799 | memcpy(&tip, arp_ptr, 4); | |
2800 | ||
5a03cdb7 | 2801 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2802 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2803 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2804 | &sip, &tip); | |
f5b2b966 JV |
2805 | |
2806 | /* | |
2807 | * Backup slaves won't see the ARP reply, but do come through | |
2808 | * here for each ARP probe (so we swap the sip/tip to validate | |
2809 | * the probe). In a "redundant switch, common router" type of | |
2810 | * configuration, the ARP probe will (hopefully) travel from | |
2811 | * the active, through one switch, the router, then the other | |
2812 | * switch before reaching the backup. | |
2813 | */ | |
e30bc066 | 2814 | if (bond_is_active_slave(slave)) |
f5b2b966 JV |
2815 | bond_validate_arp(bond, slave, sip, tip); |
2816 | else | |
2817 | bond_validate_arp(bond, slave, tip, sip); | |
2818 | ||
2819 | out_unlock: | |
2820 | read_unlock(&bond->lock); | |
2821 | out: | |
2822 | dev_kfree_skb(skb); | |
2823 | return NET_RX_SUCCESS; | |
2824 | } | |
2825 | ||
1da177e4 LT |
2826 | /* |
2827 | * this function is called regularly to monitor each slave's link | |
2828 | * ensuring that traffic is being sent and received when arp monitoring | |
2829 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2830 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2831 | * arp monitoring in active backup mode. | |
2832 | */ | |
1b76b316 | 2833 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2834 | { |
1b76b316 JV |
2835 | struct bonding *bond = container_of(work, struct bonding, |
2836 | arp_work.work); | |
1da177e4 LT |
2837 | struct slave *slave, *oldcurrent; |
2838 | int do_failover = 0; | |
2839 | int delta_in_ticks; | |
2840 | int i; | |
2841 | ||
2842 | read_lock(&bond->lock); | |
2843 | ||
5ce0da8f | 2844 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2845 | |
3d632c3f | 2846 | if (bond->kill_timers) |
1da177e4 | 2847 | goto out; |
1da177e4 | 2848 | |
3d632c3f | 2849 | if (bond->slave_cnt == 0) |
1da177e4 | 2850 | goto re_arm; |
1da177e4 LT |
2851 | |
2852 | read_lock(&bond->curr_slave_lock); | |
2853 | oldcurrent = bond->curr_active_slave; | |
2854 | read_unlock(&bond->curr_slave_lock); | |
2855 | ||
2856 | /* see if any of the previous devices are up now (i.e. they have | |
2857 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2858 | * the picture unless it is null. also, slave->jiffies is not needed | |
2859 | * here because we send an arp on each slave and give a slave as | |
2860 | * long as it needs to get the tx/rx within the delta. | |
2861 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2862 | * so it can wait | |
2863 | */ | |
2864 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2865 | unsigned long trans_start = dev_trans_start(slave->dev); |
2866 | ||
1da177e4 | 2867 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2868 | if (time_in_range(jiffies, |
2869 | trans_start - delta_in_ticks, | |
2870 | trans_start + delta_in_ticks) && | |
2871 | time_in_range(jiffies, | |
2872 | slave->dev->last_rx - delta_in_ticks, | |
2873 | slave->dev->last_rx + delta_in_ticks)) { | |
1da177e4 LT |
2874 | |
2875 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2876 | bond_set_active_slave(slave); |
1da177e4 LT |
2877 | |
2878 | /* primary_slave has no meaning in round-robin | |
2879 | * mode. the window of a slave being up and | |
2880 | * curr_active_slave being null after enslaving | |
2881 | * is closed. | |
2882 | */ | |
2883 | if (!oldcurrent) { | |
a4aee5c8 JP |
2884 | pr_info("%s: link status definitely up for interface %s, ", |
2885 | bond->dev->name, | |
2886 | slave->dev->name); | |
1da177e4 LT |
2887 | do_failover = 1; |
2888 | } else { | |
a4aee5c8 JP |
2889 | pr_info("%s: interface %s is now up\n", |
2890 | bond->dev->name, | |
2891 | slave->dev->name); | |
1da177e4 LT |
2892 | } |
2893 | } | |
2894 | } else { | |
2895 | /* slave->link == BOND_LINK_UP */ | |
2896 | ||
2897 | /* not all switches will respond to an arp request | |
2898 | * when the source ip is 0, so don't take the link down | |
2899 | * if we don't know our ip yet | |
2900 | */ | |
cb32f2a0 JB |
2901 | if (!time_in_range(jiffies, |
2902 | trans_start - delta_in_ticks, | |
2903 | trans_start + 2 * delta_in_ticks) || | |
2904 | !time_in_range(jiffies, | |
2905 | slave->dev->last_rx - delta_in_ticks, | |
2906 | slave->dev->last_rx + 2 * delta_in_ticks)) { | |
1da177e4 LT |
2907 | |
2908 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2909 | bond_set_backup_slave(slave); |
1da177e4 | 2910 | |
3d632c3f | 2911 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2912 | slave->link_failure_count++; |
1da177e4 | 2913 | |
a4aee5c8 JP |
2914 | pr_info("%s: interface %s is now down.\n", |
2915 | bond->dev->name, | |
2916 | slave->dev->name); | |
1da177e4 | 2917 | |
3d632c3f | 2918 | if (slave == oldcurrent) |
1da177e4 | 2919 | do_failover = 1; |
1da177e4 LT |
2920 | } |
2921 | } | |
2922 | ||
2923 | /* note: if switch is in round-robin mode, all links | |
2924 | * must tx arp to ensure all links rx an arp - otherwise | |
2925 | * links may oscillate or not come up at all; if switch is | |
2926 | * in something like xor mode, there is nothing we can | |
2927 | * do - all replies will be rx'ed on same link causing slaves | |
2928 | * to be unstable during low/no traffic periods | |
2929 | */ | |
3d632c3f | 2930 | if (IS_UP(slave->dev)) |
1da177e4 | 2931 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2932 | } |
2933 | ||
2934 | if (do_failover) { | |
e843fa50 | 2935 | block_netpoll_tx(); |
059fe7a5 | 2936 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2937 | |
2938 | bond_select_active_slave(bond); | |
2939 | ||
059fe7a5 | 2940 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2941 | unblock_netpoll_tx(); |
1da177e4 LT |
2942 | } |
2943 | ||
2944 | re_arm: | |
1b76b316 JV |
2945 | if (bond->params.arp_interval) |
2946 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
2947 | out: |
2948 | read_unlock(&bond->lock); | |
2949 | } | |
2950 | ||
2951 | /* | |
b2220cad JV |
2952 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2953 | * changes. Sets new_link in slaves to specify what action should take | |
2954 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2955 | * to link states must be committed. | |
2956 | * | |
2957 | * Called with bond->lock held for read. | |
1da177e4 | 2958 | */ |
b2220cad | 2959 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2960 | { |
1da177e4 | 2961 | struct slave *slave; |
b2220cad | 2962 | int i, commit = 0; |
cb32f2a0 | 2963 | unsigned long trans_start; |
1da177e4 | 2964 | |
b2220cad JV |
2965 | bond_for_each_slave(bond, slave, i) { |
2966 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2967 | |
b2220cad | 2968 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2969 | if (time_in_range(jiffies, |
2970 | slave_last_rx(bond, slave) - delta_in_ticks, | |
2971 | slave_last_rx(bond, slave) + delta_in_ticks)) { | |
2972 | ||
b2220cad JV |
2973 | slave->new_link = BOND_LINK_UP; |
2974 | commit++; | |
2975 | } | |
1da177e4 | 2976 | |
b2220cad JV |
2977 | continue; |
2978 | } | |
1da177e4 | 2979 | |
b2220cad JV |
2980 | /* |
2981 | * Give slaves 2*delta after being enslaved or made | |
2982 | * active. This avoids bouncing, as the last receive | |
2983 | * times need a full ARP monitor cycle to be updated. | |
2984 | */ | |
cb32f2a0 JB |
2985 | if (time_in_range(jiffies, |
2986 | slave->jiffies - delta_in_ticks, | |
2987 | slave->jiffies + 2 * delta_in_ticks)) | |
b2220cad JV |
2988 | continue; |
2989 | ||
2990 | /* | |
2991 | * Backup slave is down if: | |
2992 | * - No current_arp_slave AND | |
2993 | * - more than 3*delta since last receive AND | |
2994 | * - the bond has an IP address | |
2995 | * | |
2996 | * Note: a non-null current_arp_slave indicates | |
2997 | * the curr_active_slave went down and we are | |
2998 | * searching for a new one; under this condition | |
2999 | * we only take the curr_active_slave down - this | |
3000 | * gives each slave a chance to tx/rx traffic | |
3001 | * before being taken out | |
3002 | */ | |
e30bc066 | 3003 | if (!bond_is_active_slave(slave) && |
b2220cad | 3004 | !bond->current_arp_slave && |
cb32f2a0 JB |
3005 | !time_in_range(jiffies, |
3006 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3007 | slave_last_rx(bond, slave) + 3 * delta_in_ticks)) { | |
3008 | ||
b2220cad JV |
3009 | slave->new_link = BOND_LINK_DOWN; |
3010 | commit++; | |
3011 | } | |
3012 | ||
3013 | /* | |
3014 | * Active slave is down if: | |
3015 | * - more than 2*delta since transmitting OR | |
3016 | * - (more than 2*delta since receive AND | |
3017 | * the bond has an IP address) | |
3018 | */ | |
cb32f2a0 | 3019 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 3020 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
3021 | (!time_in_range(jiffies, |
3022 | trans_start - delta_in_ticks, | |
3023 | trans_start + 2 * delta_in_ticks) || | |
3024 | !time_in_range(jiffies, | |
3025 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3026 | slave_last_rx(bond, slave) + 2 * delta_in_ticks))) { | |
3027 | ||
b2220cad JV |
3028 | slave->new_link = BOND_LINK_DOWN; |
3029 | commit++; | |
3030 | } | |
1da177e4 LT |
3031 | } |
3032 | ||
b2220cad JV |
3033 | return commit; |
3034 | } | |
1da177e4 | 3035 | |
b2220cad JV |
3036 | /* |
3037 | * Called to commit link state changes noted by inspection step of | |
3038 | * active-backup mode ARP monitor. | |
3039 | * | |
3040 | * Called with RTNL and bond->lock for read. | |
3041 | */ | |
3042 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
3043 | { | |
3044 | struct slave *slave; | |
3045 | int i; | |
cb32f2a0 | 3046 | unsigned long trans_start; |
1da177e4 | 3047 | |
b2220cad JV |
3048 | bond_for_each_slave(bond, slave, i) { |
3049 | switch (slave->new_link) { | |
3050 | case BOND_LINK_NOCHANGE: | |
3051 | continue; | |
ff59c456 | 3052 | |
b2220cad | 3053 | case BOND_LINK_UP: |
cb32f2a0 | 3054 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 3055 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
3056 | time_in_range(jiffies, |
3057 | trans_start - delta_in_ticks, | |
3058 | trans_start + delta_in_ticks)) || | |
b9f60253 | 3059 | bond->curr_active_slave != slave) { |
b2220cad | 3060 | slave->link = BOND_LINK_UP; |
b2220cad JV |
3061 | bond->current_arp_slave = NULL; |
3062 | ||
a4aee5c8 | 3063 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 3064 | bond->dev->name, slave->dev->name); |
b2220cad | 3065 | |
b9f60253 JP |
3066 | if (!bond->curr_active_slave || |
3067 | (slave == bond->primary_slave)) | |
3068 | goto do_failover; | |
1da177e4 | 3069 | |
b9f60253 | 3070 | } |
1da177e4 | 3071 | |
b9f60253 | 3072 | continue; |
1da177e4 | 3073 | |
b2220cad JV |
3074 | case BOND_LINK_DOWN: |
3075 | if (slave->link_failure_count < UINT_MAX) | |
3076 | slave->link_failure_count++; | |
3077 | ||
3078 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 3079 | bond_set_slave_inactive_flags(slave); |
b2220cad | 3080 | |
a4aee5c8 | 3081 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 3082 | bond->dev->name, slave->dev->name); |
b2220cad | 3083 | |
b9f60253 | 3084 | if (slave == bond->curr_active_slave) { |
b2220cad | 3085 | bond->current_arp_slave = NULL; |
b9f60253 | 3086 | goto do_failover; |
1da177e4 | 3087 | } |
b9f60253 JP |
3088 | |
3089 | continue; | |
b2220cad JV |
3090 | |
3091 | default: | |
a4aee5c8 | 3092 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
3093 | bond->dev->name, slave->new_link, |
3094 | slave->dev->name); | |
b9f60253 | 3095 | continue; |
1da177e4 | 3096 | } |
1da177e4 | 3097 | |
b9f60253 JP |
3098 | do_failover: |
3099 | ASSERT_RTNL(); | |
e843fa50 | 3100 | block_netpoll_tx(); |
b2220cad | 3101 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3102 | bond_select_active_slave(bond); |
b2220cad | 3103 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3104 | unblock_netpoll_tx(); |
b2220cad | 3105 | } |
1da177e4 | 3106 | |
b2220cad JV |
3107 | bond_set_carrier(bond); |
3108 | } | |
1da177e4 | 3109 | |
b2220cad JV |
3110 | /* |
3111 | * Send ARP probes for active-backup mode ARP monitor. | |
3112 | * | |
3113 | * Called with bond->lock held for read. | |
3114 | */ | |
3115 | static void bond_ab_arp_probe(struct bonding *bond) | |
3116 | { | |
3117 | struct slave *slave; | |
3118 | int i; | |
1da177e4 | 3119 | |
b2220cad | 3120 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3121 | |
b2220cad | 3122 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3123 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3124 | bond->current_arp_slave->dev->name, | |
3125 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3126 | |
b2220cad JV |
3127 | if (bond->curr_active_slave) { |
3128 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3129 | read_unlock(&bond->curr_slave_lock); | |
3130 | return; | |
3131 | } | |
1da177e4 | 3132 | |
b2220cad | 3133 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3134 | |
b2220cad JV |
3135 | /* if we don't have a curr_active_slave, search for the next available |
3136 | * backup slave from the current_arp_slave and make it the candidate | |
3137 | * for becoming the curr_active_slave | |
3138 | */ | |
1da177e4 | 3139 | |
b2220cad JV |
3140 | if (!bond->current_arp_slave) { |
3141 | bond->current_arp_slave = bond->first_slave; | |
3142 | if (!bond->current_arp_slave) | |
3143 | return; | |
3144 | } | |
1da177e4 | 3145 | |
b2220cad | 3146 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3147 | |
b2220cad JV |
3148 | /* search for next candidate */ |
3149 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3150 | if (IS_UP(slave->dev)) { | |
3151 | slave->link = BOND_LINK_BACK; | |
3152 | bond_set_slave_active_flags(slave); | |
3153 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3154 | slave->jiffies = jiffies; |
b2220cad JV |
3155 | bond->current_arp_slave = slave; |
3156 | break; | |
1da177e4 LT |
3157 | } |
3158 | ||
b2220cad JV |
3159 | /* if the link state is up at this point, we |
3160 | * mark it down - this can happen if we have | |
3161 | * simultaneous link failures and | |
3162 | * reselect_active_interface doesn't make this | |
3163 | * one the current slave so it is still marked | |
3164 | * up when it is actually down | |
1da177e4 | 3165 | */ |
b2220cad JV |
3166 | if (slave->link == BOND_LINK_UP) { |
3167 | slave->link = BOND_LINK_DOWN; | |
3168 | if (slave->link_failure_count < UINT_MAX) | |
3169 | slave->link_failure_count++; | |
1da177e4 | 3170 | |
b2220cad JV |
3171 | bond_set_slave_inactive_flags(slave); |
3172 | ||
a4aee5c8 JP |
3173 | pr_info("%s: backup interface %s is now down.\n", |
3174 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3175 | } |
b2220cad JV |
3176 | } |
3177 | } | |
1da177e4 | 3178 | |
b2220cad JV |
3179 | void bond_activebackup_arp_mon(struct work_struct *work) |
3180 | { | |
3181 | struct bonding *bond = container_of(work, struct bonding, | |
3182 | arp_work.work); | |
3183 | int delta_in_ticks; | |
1da177e4 | 3184 | |
b2220cad | 3185 | read_lock(&bond->lock); |
1da177e4 | 3186 | |
b2220cad JV |
3187 | if (bond->kill_timers) |
3188 | goto out; | |
1da177e4 | 3189 | |
b2220cad | 3190 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3191 | |
b2220cad JV |
3192 | if (bond->slave_cnt == 0) |
3193 | goto re_arm; | |
3194 | ||
3195 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { | |
3196 | read_unlock(&bond->lock); | |
3197 | rtnl_lock(); | |
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: |
3d632c3f | 3210 | if (bond->params.arp_interval) |
1b76b316 | 3211 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3212 | out: |
3213 | read_unlock(&bond->lock); | |
3214 | } | |
3215 | ||
1da177e4 LT |
3216 | /*-------------------------- netdev event handling --------------------------*/ |
3217 | ||
3218 | /* | |
3219 | * Change device name | |
3220 | */ | |
3221 | static int bond_event_changename(struct bonding *bond) | |
3222 | { | |
1da177e4 LT |
3223 | bond_remove_proc_entry(bond); |
3224 | bond_create_proc_entry(bond); | |
7e083840 | 3225 | |
f073c7ca TI |
3226 | bond_debug_reregister(bond); |
3227 | ||
1da177e4 LT |
3228 | return NOTIFY_DONE; |
3229 | } | |
3230 | ||
3d632c3f SH |
3231 | static int bond_master_netdev_event(unsigned long event, |
3232 | struct net_device *bond_dev) | |
1da177e4 | 3233 | { |
454d7c9b | 3234 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3235 | |
3236 | switch (event) { | |
3237 | case NETDEV_CHANGENAME: | |
3238 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3239 | default: |
3240 | break; | |
3241 | } | |
3242 | ||
3243 | return NOTIFY_DONE; | |
3244 | } | |
3245 | ||
3d632c3f SH |
3246 | static int bond_slave_netdev_event(unsigned long event, |
3247 | struct net_device *slave_dev) | |
1da177e4 LT |
3248 | { |
3249 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3250 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3251 | |
3252 | switch (event) { | |
3253 | case NETDEV_UNREGISTER: | |
3254 | if (bond_dev) { | |
1284cd3a JV |
3255 | if (bond->setup_by_slave) |
3256 | bond_release_and_destroy(bond_dev, slave_dev); | |
3257 | else | |
3258 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3259 | } |
3260 | break; | |
3261 | case NETDEV_CHANGE: | |
17d04500 JV |
3262 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3263 | struct slave *slave; | |
3264 | ||
3265 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3266 | if (slave) { | |
5d30530e | 3267 | u32 old_speed = slave->speed; |
65cce19c | 3268 | u8 old_duplex = slave->duplex; |
17d04500 JV |
3269 | |
3270 | bond_update_speed_duplex(slave); | |
3271 | ||
3272 | if (bond_is_lb(bond)) | |
3273 | break; | |
3274 | ||
3275 | if (old_speed != slave->speed) | |
3276 | bond_3ad_adapter_speed_changed(slave); | |
3277 | if (old_duplex != slave->duplex) | |
3278 | bond_3ad_adapter_duplex_changed(slave); | |
3279 | } | |
3280 | } | |
3281 | ||
1da177e4 LT |
3282 | break; |
3283 | case NETDEV_DOWN: | |
3284 | /* | |
3285 | * ... Or is it this? | |
3286 | */ | |
3287 | break; | |
3288 | case NETDEV_CHANGEMTU: | |
3289 | /* | |
3290 | * TODO: Should slaves be allowed to | |
3291 | * independently alter their MTU? For | |
3292 | * an active-backup bond, slaves need | |
3293 | * not be the same type of device, so | |
3294 | * MTUs may vary. For other modes, | |
3295 | * slaves arguably should have the | |
3296 | * same MTUs. To do this, we'd need to | |
3297 | * take over the slave's change_mtu | |
3298 | * function for the duration of their | |
3299 | * servitude. | |
3300 | */ | |
3301 | break; | |
3302 | case NETDEV_CHANGENAME: | |
3303 | /* | |
3304 | * TODO: handle changing the primary's name | |
3305 | */ | |
3306 | break; | |
8531c5ff AK |
3307 | case NETDEV_FEAT_CHANGE: |
3308 | bond_compute_features(bond); | |
3309 | break; | |
1da177e4 LT |
3310 | default: |
3311 | break; | |
3312 | } | |
3313 | ||
3314 | return NOTIFY_DONE; | |
3315 | } | |
3316 | ||
3317 | /* | |
3318 | * bond_netdev_event: handle netdev notifier chain events. | |
3319 | * | |
3320 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3321 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3322 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3323 | * dev_probe_lock). | |
3324 | */ | |
3d632c3f SH |
3325 | static int bond_netdev_event(struct notifier_block *this, |
3326 | unsigned long event, void *ptr) | |
1da177e4 LT |
3327 | { |
3328 | struct net_device *event_dev = (struct net_device *)ptr; | |
3329 | ||
5a03cdb7 | 3330 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3331 | event_dev ? event_dev->name : "None", |
3332 | event); | |
1da177e4 | 3333 | |
0b680e75 JV |
3334 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3335 | return NOTIFY_DONE; | |
3336 | ||
1da177e4 | 3337 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3338 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3339 | return bond_master_netdev_event(event, event_dev); |
3340 | } | |
3341 | ||
3342 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3343 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3344 | return bond_slave_netdev_event(event, event_dev); |
3345 | } | |
3346 | ||
3347 | return NOTIFY_DONE; | |
3348 | } | |
3349 | ||
c3ade5ca JV |
3350 | /* |
3351 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3352 | * | |
3353 | * We keep track of device IPs primarily to use as source addresses in | |
3354 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3355 | * | |
3356 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3357 | */ | |
3358 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3359 | { | |
3360 | struct in_ifaddr *ifa = ptr; | |
3361 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
ec87fd3b | 3362 | struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id); |
0883beca PE |
3363 | struct bonding *bond; |
3364 | struct vlan_entry *vlan; | |
c3ade5ca | 3365 | |
ec87fd3b | 3366 | list_for_each_entry(bond, &bn->dev_list, bond_list) { |
c3ade5ca JV |
3367 | if (bond->dev == event_dev) { |
3368 | switch (event) { | |
3369 | case NETDEV_UP: | |
3370 | bond->master_ip = ifa->ifa_local; | |
3371 | return NOTIFY_OK; | |
3372 | case NETDEV_DOWN: | |
3373 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3374 | return NOTIFY_OK; | |
3375 | default: | |
3376 | return NOTIFY_DONE; | |
3377 | } | |
3378 | } | |
3379 | ||
0883beca | 3380 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
3381 | if (!bond->vlgrp) |
3382 | continue; | |
5c15bdec | 3383 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3384 | if (vlan_dev == event_dev) { |
3385 | switch (event) { | |
3386 | case NETDEV_UP: | |
3387 | vlan->vlan_ip = ifa->ifa_local; | |
3388 | return NOTIFY_OK; | |
3389 | case NETDEV_DOWN: | |
3390 | vlan->vlan_ip = | |
3391 | bond_glean_dev_ip(vlan_dev); | |
3392 | return NOTIFY_OK; | |
3393 | default: | |
3394 | return NOTIFY_DONE; | |
3395 | } | |
3396 | } | |
3397 | } | |
3398 | } | |
3399 | return NOTIFY_DONE; | |
3400 | } | |
3401 | ||
1da177e4 LT |
3402 | static struct notifier_block bond_netdev_notifier = { |
3403 | .notifier_call = bond_netdev_event, | |
3404 | }; | |
3405 | ||
c3ade5ca JV |
3406 | static struct notifier_block bond_inetaddr_notifier = { |
3407 | .notifier_call = bond_inetaddr_event, | |
3408 | }; | |
3409 | ||
1da177e4 LT |
3410 | /*-------------------------- Packet type handling ---------------------------*/ |
3411 | ||
3412 | /* register to receive lacpdus on a bond */ | |
3413 | static void bond_register_lacpdu(struct bonding *bond) | |
3414 | { | |
3415 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3416 | ||
3417 | /* initialize packet type */ | |
3418 | pk_type->type = PKT_TYPE_LACPDU; | |
3419 | pk_type->dev = bond->dev; | |
3420 | pk_type->func = bond_3ad_lacpdu_recv; | |
3421 | ||
3422 | dev_add_pack(pk_type); | |
3423 | } | |
3424 | ||
3425 | /* unregister to receive lacpdus on a bond */ | |
3426 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3427 | { | |
3428 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3429 | } | |
3430 | ||
f5b2b966 JV |
3431 | void bond_register_arp(struct bonding *bond) |
3432 | { | |
3433 | struct packet_type *pt = &bond->arp_mon_pt; | |
3434 | ||
c4f283b1 JV |
3435 | if (pt->type) |
3436 | return; | |
3437 | ||
f5b2b966 | 3438 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3439 | pt->dev = bond->dev; |
f5b2b966 JV |
3440 | pt->func = bond_arp_rcv; |
3441 | dev_add_pack(pt); | |
3442 | } | |
3443 | ||
3444 | void bond_unregister_arp(struct bonding *bond) | |
3445 | { | |
c4f283b1 JV |
3446 | struct packet_type *pt = &bond->arp_mon_pt; |
3447 | ||
3448 | dev_remove_pack(pt); | |
3449 | pt->type = 0; | |
f5b2b966 JV |
3450 | } |
3451 | ||
169a3e66 JV |
3452 | /*---------------------------- Hashing Policies -----------------------------*/ |
3453 | ||
6f6652be JV |
3454 | /* |
3455 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3456 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3457 | */ | |
a361c83c | 3458 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3459 | { |
3460 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3461 | struct iphdr *iph = ip_hdr(skb); | |
3462 | ||
f14c4e4e | 3463 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3464 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3465 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3466 | } |
3467 | ||
d3da6831 | 3468 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3469 | } |
3470 | ||
169a3e66 | 3471 | /* |
59c51591 | 3472 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3473 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3474 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3475 | */ | |
a361c83c | 3476 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3477 | { |
3478 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3479 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3480 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3481 | int layer4_xor = 0; |
3482 | ||
f14c4e4e BH |
3483 | if (skb->protocol == htons(ETH_P_IP)) { |
3484 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3485 | (iph->protocol == IPPROTO_TCP || |
3486 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3487 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3488 | } |
3489 | return (layer4_xor ^ | |
3490 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3491 | ||
3492 | } | |
3493 | ||
d3da6831 | 3494 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3495 | } |
3496 | ||
3497 | /* | |
3498 | * Hash for the output device based upon layer 2 data | |
3499 | */ | |
a361c83c | 3500 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3501 | { |
3502 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3503 | ||
d3da6831 | 3504 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3505 | } |
3506 | ||
1da177e4 LT |
3507 | /*-------------------------- Device entry points ----------------------------*/ |
3508 | ||
3509 | static int bond_open(struct net_device *bond_dev) | |
3510 | { | |
454d7c9b | 3511 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3512 | |
3513 | bond->kill_timers = 0; | |
3514 | ||
5a37e8ca FL |
3515 | INIT_DELAYED_WORK(&bond->mcast_work, bond_resend_igmp_join_requests_delayed); |
3516 | ||
58402054 | 3517 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3518 | /* bond_alb_initialize must be called before the timer |
3519 | * is started. | |
3520 | */ | |
3521 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3522 | /* something went wrong - fail the open operation */ | |
b473946a | 3523 | return -ENOMEM; |
1da177e4 LT |
3524 | } |
3525 | ||
1b76b316 JV |
3526 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3527 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3528 | } |
3529 | ||
3530 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3531 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3532 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3533 | } |
3534 | ||
3535 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3536 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3537 | INIT_DELAYED_WORK(&bond->arp_work, | |
3538 | bond_activebackup_arp_mon); | |
3539 | else | |
3540 | INIT_DELAYED_WORK(&bond->arp_work, | |
3541 | bond_loadbalance_arp_mon); | |
3542 | ||
3543 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3544 | if (bond->params.arp_validate) |
3545 | bond_register_arp(bond); | |
1da177e4 LT |
3546 | } |
3547 | ||
3548 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3549 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3550 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3551 | /* register to receive LACPDUs */ |
3552 | bond_register_lacpdu(bond); | |
fd989c83 | 3553 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3554 | } |
3555 | ||
3556 | return 0; | |
3557 | } | |
3558 | ||
3559 | static int bond_close(struct net_device *bond_dev) | |
3560 | { | |
454d7c9b | 3561 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3562 | |
3563 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3564 | /* Unregister the receive of LACPDUs */ | |
3565 | bond_unregister_lacpdu(bond); | |
3566 | } | |
3567 | ||
f5b2b966 JV |
3568 | if (bond->params.arp_validate) |
3569 | bond_unregister_arp(bond); | |
3570 | ||
1da177e4 LT |
3571 | write_lock_bh(&bond->lock); |
3572 | ||
1da177e4 LT |
3573 | /* signal timers not to re-arm */ |
3574 | bond->kill_timers = 1; | |
3575 | ||
3576 | write_unlock_bh(&bond->lock); | |
3577 | ||
1da177e4 | 3578 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3579 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3580 | } |
3581 | ||
3582 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3583 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3584 | } |
3585 | ||
3586 | switch (bond->params.mode) { | |
3587 | case BOND_MODE_8023AD: | |
1b76b316 | 3588 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3589 | break; |
3590 | case BOND_MODE_TLB: | |
3591 | case BOND_MODE_ALB: | |
1b76b316 | 3592 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3593 | break; |
3594 | default: | |
3595 | break; | |
3596 | } | |
3597 | ||
5a37e8ca FL |
3598 | if (delayed_work_pending(&bond->mcast_work)) |
3599 | cancel_delayed_work(&bond->mcast_work); | |
1da177e4 | 3600 | |
58402054 | 3601 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3602 | /* Must be called only after all |
3603 | * slaves have been released | |
3604 | */ | |
3605 | bond_alb_deinitialize(bond); | |
3606 | } | |
3607 | ||
3608 | return 0; | |
3609 | } | |
3610 | ||
28172739 ED |
3611 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3612 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3613 | { |
454d7c9b | 3614 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3615 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3616 | struct slave *slave; |
3617 | int i; | |
3618 | ||
28172739 | 3619 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3620 | |
3621 | read_lock_bh(&bond->lock); | |
3622 | ||
3623 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3624 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3625 | dev_get_stats(slave->dev, &temp); |
3626 | ||
3627 | stats->rx_packets += sstats->rx_packets; | |
3628 | stats->rx_bytes += sstats->rx_bytes; | |
3629 | stats->rx_errors += sstats->rx_errors; | |
3630 | stats->rx_dropped += sstats->rx_dropped; | |
3631 | ||
3632 | stats->tx_packets += sstats->tx_packets; | |
3633 | stats->tx_bytes += sstats->tx_bytes; | |
3634 | stats->tx_errors += sstats->tx_errors; | |
3635 | stats->tx_dropped += sstats->tx_dropped; | |
3636 | ||
3637 | stats->multicast += sstats->multicast; | |
3638 | stats->collisions += sstats->collisions; | |
3639 | ||
3640 | stats->rx_length_errors += sstats->rx_length_errors; | |
3641 | stats->rx_over_errors += sstats->rx_over_errors; | |
3642 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3643 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3644 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3645 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3646 | ||
3647 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3648 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3649 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3650 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3651 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3652 | } |
3653 | ||
1da177e4 LT |
3654 | read_unlock_bh(&bond->lock); |
3655 | ||
3656 | return stats; | |
3657 | } | |
3658 | ||
3659 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3660 | { | |
3661 | struct net_device *slave_dev = NULL; | |
3662 | struct ifbond k_binfo; | |
3663 | struct ifbond __user *u_binfo = NULL; | |
3664 | struct ifslave k_sinfo; | |
3665 | struct ifslave __user *u_sinfo = NULL; | |
3666 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3667 | int res = 0; |
3668 | ||
a4aee5c8 | 3669 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3670 | |
3671 | switch (cmd) { | |
1da177e4 LT |
3672 | case SIOCGMIIPHY: |
3673 | mii = if_mii(ifr); | |
3d632c3f | 3674 | if (!mii) |
1da177e4 | 3675 | return -EINVAL; |
3d632c3f | 3676 | |
1da177e4 LT |
3677 | mii->phy_id = 0; |
3678 | /* Fall Through */ | |
3679 | case SIOCGMIIREG: | |
3680 | /* | |
3681 | * We do this again just in case we were called by SIOCGMIIREG | |
3682 | * instead of SIOCGMIIPHY. | |
3683 | */ | |
3684 | mii = if_mii(ifr); | |
3d632c3f | 3685 | if (!mii) |
1da177e4 | 3686 | return -EINVAL; |
3d632c3f | 3687 | |
1da177e4 LT |
3688 | |
3689 | if (mii->reg_num == 1) { | |
454d7c9b | 3690 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3691 | mii->val_out = 0; |
6603a6f2 | 3692 | read_lock(&bond->lock); |
1da177e4 | 3693 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3694 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3695 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3696 | |
1da177e4 | 3697 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3698 | read_unlock(&bond->lock); |
1da177e4 LT |
3699 | } |
3700 | ||
3701 | return 0; | |
3702 | case BOND_INFO_QUERY_OLD: | |
3703 | case SIOCBONDINFOQUERY: | |
3704 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3705 | ||
3d632c3f | 3706 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3707 | return -EFAULT; |
1da177e4 LT |
3708 | |
3709 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3710 | if (res == 0 && |
3711 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3712 | return -EFAULT; | |
1da177e4 LT |
3713 | |
3714 | return res; | |
3715 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3716 | case SIOCBONDSLAVEINFOQUERY: | |
3717 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3718 | ||
3d632c3f | 3719 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3720 | return -EFAULT; |
1da177e4 LT |
3721 | |
3722 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3723 | if (res == 0 && |
3724 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3725 | return -EFAULT; | |
1da177e4 LT |
3726 | |
3727 | return res; | |
3728 | default: | |
3729 | /* Go on */ | |
3730 | break; | |
3731 | } | |
3732 | ||
3d632c3f | 3733 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3734 | return -EPERM; |
1da177e4 | 3735 | |
ec87fd3b | 3736 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3737 | |
a4aee5c8 | 3738 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3739 | |
3d632c3f | 3740 | if (!slave_dev) |
1da177e4 | 3741 | res = -ENODEV; |
3d632c3f | 3742 | else { |
a4aee5c8 | 3743 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3744 | switch (cmd) { |
3745 | case BOND_ENSLAVE_OLD: | |
3746 | case SIOCBONDENSLAVE: | |
3747 | res = bond_enslave(bond_dev, slave_dev); | |
3748 | break; | |
3749 | case BOND_RELEASE_OLD: | |
3750 | case SIOCBONDRELEASE: | |
3751 | res = bond_release(bond_dev, slave_dev); | |
3752 | break; | |
3753 | case BOND_SETHWADDR_OLD: | |
3754 | case SIOCBONDSETHWADDR: | |
3755 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3756 | break; | |
3757 | case BOND_CHANGE_ACTIVE_OLD: | |
3758 | case SIOCBONDCHANGEACTIVE: | |
3759 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3760 | break; | |
3761 | default: | |
3762 | res = -EOPNOTSUPP; | |
3763 | } | |
3764 | ||
3765 | dev_put(slave_dev); | |
3766 | } | |
3767 | ||
1da177e4 LT |
3768 | return res; |
3769 | } | |
3770 | ||
22bedad3 JP |
3771 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3772 | struct netdev_hw_addr_list *list, | |
3773 | int addrlen) | |
3774 | { | |
3775 | struct netdev_hw_addr *ha; | |
3776 | ||
3777 | netdev_hw_addr_list_for_each(ha, list) | |
3778 | if (!memcmp(ha->addr, addr, addrlen)) | |
3779 | return true; | |
3780 | ||
3781 | return false; | |
3782 | } | |
3783 | ||
1da177e4 LT |
3784 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3785 | { | |
454d7c9b | 3786 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 JP |
3787 | struct netdev_hw_addr *ha; |
3788 | bool found; | |
1da177e4 | 3789 | |
1da177e4 LT |
3790 | /* |
3791 | * Do promisc before checking multicast_mode | |
3792 | */ | |
3d632c3f | 3793 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
3794 | /* |
3795 | * FIXME: Need to handle the error when one of the multi-slaves | |
3796 | * encounters error. | |
3797 | */ | |
1da177e4 | 3798 | bond_set_promiscuity(bond, 1); |
1da177e4 | 3799 | |
3d632c3f SH |
3800 | |
3801 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 3802 | bond_set_promiscuity(bond, -1); |
3d632c3f | 3803 | |
1da177e4 LT |
3804 | |
3805 | /* set allmulti flag to slaves */ | |
3d632c3f | 3806 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
3807 | /* |
3808 | * FIXME: Need to handle the error when one of the multi-slaves | |
3809 | * encounters error. | |
3810 | */ | |
1da177e4 | 3811 | bond_set_allmulti(bond, 1); |
1da177e4 | 3812 | |
3d632c3f SH |
3813 | |
3814 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 3815 | bond_set_allmulti(bond, -1); |
3d632c3f | 3816 | |
1da177e4 | 3817 | |
80ee5ad2 JV |
3818 | read_lock(&bond->lock); |
3819 | ||
1da177e4 LT |
3820 | bond->flags = bond_dev->flags; |
3821 | ||
3822 | /* looking for addresses to add to slaves' mc list */ | |
22bedad3 JP |
3823 | netdev_for_each_mc_addr(ha, bond_dev) { |
3824 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
3825 | bond_dev->addr_len); | |
3826 | if (!found) | |
3827 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
3828 | } |
3829 | ||
3830 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
3831 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
3832 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
3833 | bond_dev->addr_len); | |
3834 | if (!found) | |
3835 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
3836 | } |
3837 | ||
3838 | /* save master's multicast list */ | |
22bedad3 JP |
3839 | __hw_addr_flush(&bond->mc_list); |
3840 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
3841 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 3842 | |
80ee5ad2 | 3843 | read_unlock(&bond->lock); |
1da177e4 LT |
3844 | } |
3845 | ||
00829823 SH |
3846 | static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms) |
3847 | { | |
3848 | struct bonding *bond = netdev_priv(dev); | |
3849 | struct slave *slave = bond->first_slave; | |
3850 | ||
3851 | if (slave) { | |
3852 | const struct net_device_ops *slave_ops | |
3853 | = slave->dev->netdev_ops; | |
3854 | if (slave_ops->ndo_neigh_setup) | |
72e2240f | 3855 | return slave_ops->ndo_neigh_setup(slave->dev, parms); |
00829823 SH |
3856 | } |
3857 | return 0; | |
3858 | } | |
3859 | ||
1da177e4 LT |
3860 | /* |
3861 | * Change the MTU of all of a master's slaves to match the master | |
3862 | */ | |
3863 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3864 | { | |
454d7c9b | 3865 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3866 | struct slave *slave, *stop_at; |
3867 | int res = 0; | |
3868 | int i; | |
3869 | ||
5a03cdb7 | 3870 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3871 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3872 | |
3873 | /* Can't hold bond->lock with bh disabled here since | |
3874 | * some base drivers panic. On the other hand we can't | |
3875 | * hold bond->lock without bh disabled because we'll | |
3876 | * deadlock. The only solution is to rely on the fact | |
3877 | * that we're under rtnl_lock here, and the slaves | |
3878 | * list won't change. This doesn't solve the problem | |
3879 | * of setting the slave's MTU while it is | |
3880 | * transmitting, but the assumption is that the base | |
3881 | * driver can handle that. | |
3882 | * | |
3883 | * TODO: figure out a way to safely iterate the slaves | |
3884 | * list, but without holding a lock around the actual | |
3885 | * call to the base driver. | |
3886 | */ | |
3887 | ||
3888 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3889 | pr_debug("s %p s->p %p c_m %p\n", |
3890 | slave, | |
3891 | slave->prev, | |
3892 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3893 | |
1da177e4 LT |
3894 | res = dev_set_mtu(slave->dev, new_mtu); |
3895 | ||
3896 | if (res) { | |
3897 | /* If we failed to set the slave's mtu to the new value | |
3898 | * we must abort the operation even in ACTIVE_BACKUP | |
3899 | * mode, because if we allow the backup slaves to have | |
3900 | * different mtu values than the active slave we'll | |
3901 | * need to change their mtu when doing a failover. That | |
3902 | * means changing their mtu from timer context, which | |
3903 | * is probably not a good idea. | |
3904 | */ | |
5a03cdb7 | 3905 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3906 | goto unwind; |
3907 | } | |
3908 | } | |
3909 | ||
3910 | bond_dev->mtu = new_mtu; | |
3911 | ||
3912 | return 0; | |
3913 | ||
3914 | unwind: | |
3915 | /* unwind from head to the slave that failed */ | |
3916 | stop_at = slave; | |
3917 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3918 | int tmp_res; | |
3919 | ||
3920 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3921 | if (tmp_res) { | |
a4aee5c8 JP |
3922 | pr_debug("unwind err %d dev %s\n", |
3923 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3924 | } |
3925 | } | |
3926 | ||
3927 | return res; | |
3928 | } | |
3929 | ||
3930 | /* | |
3931 | * Change HW address | |
3932 | * | |
3933 | * Note that many devices must be down to change the HW address, and | |
3934 | * downing the master releases all slaves. We can make bonds full of | |
3935 | * bonding devices to test this, however. | |
3936 | */ | |
3937 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3938 | { | |
454d7c9b | 3939 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3940 | struct sockaddr *sa = addr, tmp_sa; |
3941 | struct slave *slave, *stop_at; | |
3942 | int res = 0; | |
3943 | int i; | |
3944 | ||
eb7cc59a SH |
3945 | if (bond->params.mode == BOND_MODE_ALB) |
3946 | return bond_alb_set_mac_address(bond_dev, addr); | |
3947 | ||
3948 | ||
a4aee5c8 JP |
3949 | pr_debug("bond=%p, name=%s\n", |
3950 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3951 | |
dd957c57 | 3952 | /* |
3915c1e8 JV |
3953 | * If fail_over_mac is set to active, do nothing and return |
3954 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 3955 | */ |
3915c1e8 | 3956 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 3957 | return 0; |
2ab82852 | 3958 | |
3d632c3f | 3959 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3960 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3961 | |
3962 | /* Can't hold bond->lock with bh disabled here since | |
3963 | * some base drivers panic. On the other hand we can't | |
3964 | * hold bond->lock without bh disabled because we'll | |
3965 | * deadlock. The only solution is to rely on the fact | |
3966 | * that we're under rtnl_lock here, and the slaves | |
3967 | * list won't change. This doesn't solve the problem | |
3968 | * of setting the slave's hw address while it is | |
3969 | * transmitting, but the assumption is that the base | |
3970 | * driver can handle that. | |
3971 | * | |
3972 | * TODO: figure out a way to safely iterate the slaves | |
3973 | * list, but without holding a lock around the actual | |
3974 | * call to the base driver. | |
3975 | */ | |
3976 | ||
3977 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 3978 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3979 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3980 | |
eb7cc59a | 3981 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3982 | res = -EOPNOTSUPP; |
5a03cdb7 | 3983 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3984 | goto unwind; |
3985 | } | |
3986 | ||
3987 | res = dev_set_mac_address(slave->dev, addr); | |
3988 | if (res) { | |
3989 | /* TODO: consider downing the slave | |
3990 | * and retry ? | |
3991 | * User should expect communications | |
3992 | * breakage anyway until ARP finish | |
3993 | * updating, so... | |
3994 | */ | |
5a03cdb7 | 3995 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3996 | goto unwind; |
3997 | } | |
3998 | } | |
3999 | ||
4000 | /* success */ | |
4001 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4002 | return 0; | |
4003 | ||
4004 | unwind: | |
4005 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4006 | tmp_sa.sa_family = bond_dev->type; | |
4007 | ||
4008 | /* unwind from head to the slave that failed */ | |
4009 | stop_at = slave; | |
4010 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4011 | int tmp_res; | |
4012 | ||
4013 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4014 | if (tmp_res) { | |
a4aee5c8 JP |
4015 | pr_debug("unwind err %d dev %s\n", |
4016 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4017 | } |
4018 | } | |
4019 | ||
4020 | return res; | |
4021 | } | |
4022 | ||
4023 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4024 | { | |
454d7c9b | 4025 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4026 | struct slave *slave, *start_at; |
cf5f9044 | 4027 | int i, slave_no, res = 1; |
a2fd940f | 4028 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
4029 | |
4030 | read_lock(&bond->lock); | |
4031 | ||
3d632c3f | 4032 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4033 | goto out; |
cf5f9044 | 4034 | /* |
a2fd940f AG |
4035 | * Start with the curr_active_slave that joined the bond as the |
4036 | * default for sending IGMP traffic. For failover purposes one | |
4037 | * needs to maintain some consistency for the interface that will | |
4038 | * send the join/membership reports. The curr_active_slave found | |
4039 | * will send all of this type of traffic. | |
cf5f9044 | 4040 | */ |
00ae7028 | 4041 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4042 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4043 | |
a2fd940f AG |
4044 | read_lock(&bond->curr_slave_lock); |
4045 | slave = bond->curr_active_slave; | |
4046 | read_unlock(&bond->curr_slave_lock); | |
4047 | ||
4048 | if (!slave) | |
4049 | goto out; | |
4050 | } else { | |
4051 | /* | |
4052 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4053 | * that as being rare enough not to justify using an | |
4054 | * atomic op here. | |
4055 | */ | |
4056 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4057 | ||
4058 | bond_for_each_slave(bond, slave, i) { | |
4059 | slave_no--; | |
4060 | if (slave_no < 0) | |
4061 | break; | |
4062 | } | |
1da177e4 LT |
4063 | } |
4064 | ||
cf5f9044 | 4065 | start_at = slave; |
1da177e4 LT |
4066 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4067 | if (IS_UP(slave->dev) && | |
4068 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4069 | bond_is_active_slave(slave)) { |
1da177e4 | 4070 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
4071 | break; |
4072 | } | |
4073 | } | |
4074 | ||
1da177e4 LT |
4075 | out: |
4076 | if (res) { | |
4077 | /* no suitable interface, frame not sent */ | |
4078 | dev_kfree_skb(skb); | |
4079 | } | |
4080 | read_unlock(&bond->lock); | |
ec634fe3 | 4081 | return NETDEV_TX_OK; |
1da177e4 LT |
4082 | } |
4083 | ||
075897ce | 4084 | |
1da177e4 LT |
4085 | /* |
4086 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4087 | * the bond has a usable interface. | |
4088 | */ | |
4089 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4090 | { | |
454d7c9b | 4091 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4092 | int res = 1; |
4093 | ||
1da177e4 LT |
4094 | read_lock(&bond->lock); |
4095 | read_lock(&bond->curr_slave_lock); | |
4096 | ||
3d632c3f | 4097 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4098 | goto out; |
1da177e4 | 4099 | |
075897ce JL |
4100 | if (!bond->curr_active_slave) |
4101 | goto out; | |
4102 | ||
075897ce JL |
4103 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4104 | ||
1da177e4 | 4105 | out: |
3d632c3f | 4106 | if (res) |
1da177e4 LT |
4107 | /* no suitable interface, frame not sent */ |
4108 | dev_kfree_skb(skb); | |
3d632c3f | 4109 | |
1da177e4 LT |
4110 | read_unlock(&bond->curr_slave_lock); |
4111 | read_unlock(&bond->lock); | |
ec634fe3 | 4112 | return NETDEV_TX_OK; |
1da177e4 LT |
4113 | } |
4114 | ||
4115 | /* | |
169a3e66 JV |
4116 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4117 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4118 | * find the next active slave. | |
1da177e4 LT |
4119 | */ |
4120 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4121 | { | |
454d7c9b | 4122 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4123 | struct slave *slave, *start_at; |
4124 | int slave_no; | |
4125 | int i; | |
4126 | int res = 1; | |
4127 | ||
4128 | read_lock(&bond->lock); | |
4129 | ||
3d632c3f | 4130 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4131 | goto out; |
1da177e4 | 4132 | |
a361c83c | 4133 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4134 | |
4135 | bond_for_each_slave(bond, slave, i) { | |
4136 | slave_no--; | |
3d632c3f | 4137 | if (slave_no < 0) |
1da177e4 | 4138 | break; |
1da177e4 LT |
4139 | } |
4140 | ||
4141 | start_at = slave; | |
4142 | ||
4143 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4144 | if (IS_UP(slave->dev) && | |
4145 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4146 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4147 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
4148 | break; | |
4149 | } | |
4150 | } | |
4151 | ||
4152 | out: | |
4153 | if (res) { | |
4154 | /* no suitable interface, frame not sent */ | |
4155 | dev_kfree_skb(skb); | |
4156 | } | |
4157 | read_unlock(&bond->lock); | |
ec634fe3 | 4158 | return NETDEV_TX_OK; |
1da177e4 LT |
4159 | } |
4160 | ||
4161 | /* | |
4162 | * in broadcast mode, we send everything to all usable interfaces. | |
4163 | */ | |
4164 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4165 | { | |
454d7c9b | 4166 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4167 | struct slave *slave, *start_at; |
4168 | struct net_device *tx_dev = NULL; | |
4169 | int i; | |
4170 | int res = 1; | |
4171 | ||
4172 | read_lock(&bond->lock); | |
4173 | ||
3d632c3f | 4174 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4175 | goto out; |
1da177e4 LT |
4176 | |
4177 | read_lock(&bond->curr_slave_lock); | |
4178 | start_at = bond->curr_active_slave; | |
4179 | read_unlock(&bond->curr_slave_lock); | |
4180 | ||
3d632c3f | 4181 | if (!start_at) |
1da177e4 | 4182 | goto out; |
1da177e4 LT |
4183 | |
4184 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4185 | if (IS_UP(slave->dev) && | |
4186 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4187 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4188 | if (tx_dev) { |
4189 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4190 | if (!skb2) { | |
a4aee5c8 | 4191 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4192 | bond_dev->name); |
1da177e4 LT |
4193 | continue; |
4194 | } | |
4195 | ||
4196 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4197 | if (res) { | |
4198 | dev_kfree_skb(skb2); | |
4199 | continue; | |
4200 | } | |
4201 | } | |
4202 | tx_dev = slave->dev; | |
4203 | } | |
4204 | } | |
4205 | ||
3d632c3f | 4206 | if (tx_dev) |
1da177e4 | 4207 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4208 | |
4209 | out: | |
3d632c3f | 4210 | if (res) |
1da177e4 LT |
4211 | /* no suitable interface, frame not sent */ |
4212 | dev_kfree_skb(skb); | |
3d632c3f | 4213 | |
1da177e4 LT |
4214 | /* frame sent to all suitable interfaces */ |
4215 | read_unlock(&bond->lock); | |
ec634fe3 | 4216 | return NETDEV_TX_OK; |
1da177e4 LT |
4217 | } |
4218 | ||
4219 | /*------------------------- Device initialization ---------------------------*/ | |
4220 | ||
6f6652be JV |
4221 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4222 | { | |
4223 | switch (bond->params.xmit_policy) { | |
4224 | case BOND_XMIT_POLICY_LAYER23: | |
4225 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4226 | break; | |
4227 | case BOND_XMIT_POLICY_LAYER34: | |
4228 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4229 | break; | |
4230 | case BOND_XMIT_POLICY_LAYER2: | |
4231 | default: | |
4232 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4233 | break; | |
4234 | } | |
4235 | } | |
4236 | ||
bb1d9123 AG |
4237 | /* |
4238 | * Lookup the slave that corresponds to a qid | |
4239 | */ | |
4240 | static inline int bond_slave_override(struct bonding *bond, | |
4241 | struct sk_buff *skb) | |
4242 | { | |
4243 | int i, res = 1; | |
4244 | struct slave *slave = NULL; | |
4245 | struct slave *check_slave; | |
4246 | ||
4247 | read_lock(&bond->lock); | |
4248 | ||
4249 | if (!BOND_IS_OK(bond) || !skb->queue_mapping) | |
4250 | goto out; | |
4251 | ||
4252 | /* Find out if any slaves have the same mapping as this skb. */ | |
4253 | bond_for_each_slave(bond, check_slave, i) { | |
4254 | if (check_slave->queue_id == skb->queue_mapping) { | |
4255 | slave = check_slave; | |
4256 | break; | |
4257 | } | |
4258 | } | |
4259 | ||
4260 | /* If the slave isn't UP, use default transmit policy. */ | |
4261 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4262 | (slave->link == BOND_LINK_UP)) { | |
4263 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4264 | } | |
4265 | ||
4266 | out: | |
4267 | read_unlock(&bond->lock); | |
4268 | return res; | |
4269 | } | |
4270 | ||
4271 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) | |
4272 | { | |
4273 | /* | |
4274 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4275 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4276 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4277 | * way down to the bonding driver. | |
4278 | */ | |
fd0e435b PO |
4279 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4280 | ||
4281 | if (unlikely(txq >= dev->real_num_tx_queues)) { | |
d30ee670 | 4282 | do { |
fd0e435b | 4283 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4284 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4285 | } |
4286 | return txq; | |
bb1d9123 AG |
4287 | } |
4288 | ||
424efe9c | 4289 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4290 | { |
bb1d9123 AG |
4291 | struct bonding *bond = netdev_priv(dev); |
4292 | ||
e843fa50 NH |
4293 | /* |
4294 | * If we risk deadlock from transmitting this in the | |
4295 | * netpoll path, tell netpoll to queue the frame for later tx | |
4296 | */ | |
4297 | if (is_netpoll_tx_blocked(dev)) | |
4298 | return NETDEV_TX_BUSY; | |
4299 | ||
bb1d9123 AG |
4300 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { |
4301 | if (!bond_slave_override(bond, skb)) | |
4302 | return NETDEV_TX_OK; | |
4303 | } | |
00829823 SH |
4304 | |
4305 | switch (bond->params.mode) { | |
4306 | case BOND_MODE_ROUNDROBIN: | |
4307 | return bond_xmit_roundrobin(skb, dev); | |
4308 | case BOND_MODE_ACTIVEBACKUP: | |
4309 | return bond_xmit_activebackup(skb, dev); | |
4310 | case BOND_MODE_XOR: | |
4311 | return bond_xmit_xor(skb, dev); | |
4312 | case BOND_MODE_BROADCAST: | |
4313 | return bond_xmit_broadcast(skb, dev); | |
4314 | case BOND_MODE_8023AD: | |
4315 | return bond_3ad_xmit_xor(skb, dev); | |
4316 | case BOND_MODE_ALB: | |
4317 | case BOND_MODE_TLB: | |
4318 | return bond_alb_xmit(skb, dev); | |
4319 | default: | |
4320 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4321 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4322 | dev->name, bond->params.mode); | |
00829823 SH |
4323 | WARN_ON_ONCE(1); |
4324 | dev_kfree_skb(skb); | |
4325 | return NETDEV_TX_OK; | |
4326 | } | |
4327 | } | |
4328 | ||
4329 | ||
1da177e4 LT |
4330 | /* |
4331 | * set bond mode specific net device operations | |
4332 | */ | |
a77b5325 | 4333 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4334 | { |
169a3e66 JV |
4335 | struct net_device *bond_dev = bond->dev; |
4336 | ||
1da177e4 LT |
4337 | switch (mode) { |
4338 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4339 | break; |
4340 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4341 | break; |
4342 | case BOND_MODE_XOR: | |
6f6652be | 4343 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4344 | break; |
4345 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4346 | break; |
4347 | case BOND_MODE_8023AD: | |
6f6652be | 4348 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4349 | break; |
1da177e4 | 4350 | case BOND_MODE_ALB: |
8f903c70 JV |
4351 | /* FALLTHRU */ |
4352 | case BOND_MODE_TLB: | |
1da177e4 LT |
4353 | break; |
4354 | default: | |
4355 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4356 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4357 | bond_dev->name, mode); | |
1da177e4 LT |
4358 | break; |
4359 | } | |
4360 | } | |
4361 | ||
217df670 JV |
4362 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4363 | struct ethtool_drvinfo *drvinfo) | |
4364 | { | |
4365 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4366 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4367 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4368 | } | |
4369 | ||
7282d491 | 4370 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4371 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4372 | .get_link = ethtool_op_get_link, |
4373 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4374 | .get_sg = ethtool_op_get_sg, | |
4375 | .get_tso = ethtool_op_get_tso, | |
4376 | .get_ufo = ethtool_op_get_ufo, | |
4377 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4378 | }; |
4379 | ||
eb7cc59a | 4380 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4381 | .ndo_init = bond_init, |
9e71626c | 4382 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4383 | .ndo_open = bond_open, |
4384 | .ndo_stop = bond_close, | |
00829823 | 4385 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4386 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4387 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a SH |
4388 | .ndo_do_ioctl = bond_do_ioctl, |
4389 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4390 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4391 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4392 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4393 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4394 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4395 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
f6dc31a8 | 4396 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4397 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4398 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4399 | .ndo_poll_controller = bond_poll_controller, | |
4400 | #endif | |
9232ecca JP |
4401 | .ndo_add_slave = bond_enslave, |
4402 | .ndo_del_slave = bond_release, | |
eb7cc59a SH |
4403 | }; |
4404 | ||
9e2e61fb AW |
4405 | static void bond_destructor(struct net_device *bond_dev) |
4406 | { | |
4407 | struct bonding *bond = netdev_priv(bond_dev); | |
4408 | if (bond->wq) | |
4409 | destroy_workqueue(bond->wq); | |
4410 | free_netdev(bond_dev); | |
4411 | } | |
4412 | ||
181470fc | 4413 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4414 | { |
454d7c9b | 4415 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4416 | |
1da177e4 LT |
4417 | /* initialize rwlocks */ |
4418 | rwlock_init(&bond->lock); | |
4419 | rwlock_init(&bond->curr_slave_lock); | |
4420 | ||
d2991f75 | 4421 | bond->params = bonding_defaults; |
1da177e4 LT |
4422 | |
4423 | /* Initialize pointers */ | |
1da177e4 LT |
4424 | bond->dev = bond_dev; |
4425 | INIT_LIST_HEAD(&bond->vlan_list); | |
4426 | ||
4427 | /* Initialize the device entry points */ | |
181470fc | 4428 | ether_setup(bond_dev); |
eb7cc59a | 4429 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4430 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4431 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4432 | |
9e2e61fb | 4433 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4434 | |
4435 | /* Initialize the device options */ | |
4436 | bond_dev->tx_queue_len = 0; | |
4437 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4438 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4439 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4440 | ||
1da177e4 LT |
4441 | /* At first, we block adding VLANs. That's the only way to |
4442 | * prevent problems that occur when adding VLANs over an | |
4443 | * empty bond. The block will be removed once non-challenged | |
4444 | * slaves are enslaved. | |
4445 | */ | |
4446 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4447 | ||
932ff279 | 4448 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4449 | * transmitting */ |
4450 | bond_dev->features |= NETIF_F_LLTX; | |
4451 | ||
4452 | /* By default, we declare the bond to be fully | |
4453 | * VLAN hardware accelerated capable. Special | |
4454 | * care is taken in the various xmit functions | |
4455 | * when there are slaves that are not hw accel | |
4456 | * capable | |
4457 | */ | |
1da177e4 LT |
4458 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4459 | NETIF_F_HW_VLAN_RX | | |
4460 | NETIF_F_HW_VLAN_FILTER); | |
4461 | ||
e6599c2e ED |
4462 | /* By default, we enable GRO on bonding devices. |
4463 | * Actual support requires lowlevel drivers are GRO ready. | |
4464 | */ | |
4465 | bond_dev->features |= NETIF_F_GRO; | |
1da177e4 LT |
4466 | } |
4467 | ||
fdaea7a9 JV |
4468 | static void bond_work_cancel_all(struct bonding *bond) |
4469 | { | |
4470 | write_lock_bh(&bond->lock); | |
4471 | bond->kill_timers = 1; | |
4472 | write_unlock_bh(&bond->lock); | |
4473 | ||
4474 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4475 | cancel_delayed_work(&bond->mii_work); | |
4476 | ||
4477 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4478 | cancel_delayed_work(&bond->arp_work); | |
4479 | ||
4480 | if (bond->params.mode == BOND_MODE_ALB && | |
4481 | delayed_work_pending(&bond->alb_work)) | |
4482 | cancel_delayed_work(&bond->alb_work); | |
4483 | ||
4484 | if (bond->params.mode == BOND_MODE_8023AD && | |
4485 | delayed_work_pending(&bond->ad_work)) | |
4486 | cancel_delayed_work(&bond->ad_work); | |
5a37e8ca FL |
4487 | |
4488 | if (delayed_work_pending(&bond->mcast_work)) | |
4489 | cancel_delayed_work(&bond->mcast_work); | |
fdaea7a9 JV |
4490 | } |
4491 | ||
c67dfb29 EB |
4492 | /* |
4493 | * Destroy a bonding device. | |
4494 | * Must be under rtnl_lock when this function is called. | |
4495 | */ | |
4496 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4497 | { |
454d7c9b | 4498 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4499 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4500 | |
f6dc31a8 WC |
4501 | bond_netpoll_cleanup(bond_dev); |
4502 | ||
c67dfb29 EB |
4503 | /* Release the bonded slaves */ |
4504 | bond_release_all(bond_dev); | |
4505 | ||
a434e43f JV |
4506 | list_del(&bond->bond_list); |
4507 | ||
4508 | bond_work_cancel_all(bond); | |
4509 | ||
a434e43f | 4510 | bond_remove_proc_entry(bond); |
c67dfb29 | 4511 | |
f073c7ca TI |
4512 | bond_debug_unregister(bond); |
4513 | ||
22bedad3 | 4514 | __hw_addr_flush(&bond->mc_list); |
f35188fa JV |
4515 | |
4516 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { | |
4517 | list_del(&vlan->vlan_list); | |
4518 | kfree(vlan); | |
4519 | } | |
a434e43f JV |
4520 | } |
4521 | ||
1da177e4 LT |
4522 | /*------------------------- Module initialization ---------------------------*/ |
4523 | ||
4524 | /* | |
4525 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4526 | * number of the mode or its string name. A bit complicated because |
4527 | * some mode names are substrings of other names, and calls from sysfs | |
4528 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4529 | */ |
325dcf7a | 4530 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4531 | { |
54b87323 | 4532 | int modeint = -1, i, rv; |
a42e534f | 4533 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4534 | |
a42e534f JV |
4535 | for (p = (char *)buf; *p; p++) |
4536 | if (!(isdigit(*p) || isspace(*p))) | |
4537 | break; | |
4538 | ||
4539 | if (*p) | |
ece95f7f | 4540 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4541 | else |
54b87323 | 4542 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4543 | |
4544 | if (!rv) | |
4545 | return -1; | |
1da177e4 LT |
4546 | |
4547 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4548 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4549 | return tbl[i].mode; |
4550 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4551 | return tbl[i].mode; |
1da177e4 LT |
4552 | } |
4553 | ||
4554 | return -1; | |
4555 | } | |
4556 | ||
4557 | static int bond_check_params(struct bond_params *params) | |
4558 | { | |
a549952a | 4559 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4560 | |
1da177e4 LT |
4561 | /* |
4562 | * Convert string parameters. | |
4563 | */ | |
4564 | if (mode) { | |
4565 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4566 | if (bond_mode == -1) { | |
a4aee5c8 | 4567 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4568 | mode == NULL ? "NULL" : mode); |
4569 | return -EINVAL; | |
4570 | } | |
4571 | } | |
4572 | ||
169a3e66 JV |
4573 | if (xmit_hash_policy) { |
4574 | if ((bond_mode != BOND_MODE_XOR) && | |
4575 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4576 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4577 | bond_mode_name(bond_mode)); |
4578 | } else { | |
4579 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4580 | xmit_hashtype_tbl); | |
4581 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4582 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4583 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4584 | xmit_hash_policy); |
4585 | return -EINVAL; | |
4586 | } | |
4587 | } | |
4588 | } | |
4589 | ||
1da177e4 LT |
4590 | if (lacp_rate) { |
4591 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4592 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4593 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4594 | } else { |
4595 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4596 | if (lacp_fast == -1) { | |
a4aee5c8 | 4597 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4598 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4599 | return -EINVAL; | |
4600 | } | |
4601 | } | |
4602 | } | |
4603 | ||
fd989c83 JV |
4604 | if (ad_select) { |
4605 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4606 | if (params->ad_select == -1) { | |
a4aee5c8 | 4607 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4608 | ad_select == NULL ? "NULL" : ad_select); |
4609 | return -EINVAL; | |
4610 | } | |
4611 | ||
4612 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4613 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4614 | } |
4615 | } else { | |
4616 | params->ad_select = BOND_AD_STABLE; | |
4617 | } | |
4618 | ||
f5841306 | 4619 | if (max_bonds < 0) { |
a4aee5c8 JP |
4620 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4621 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4622 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4623 | } | |
4624 | ||
4625 | if (miimon < 0) { | |
a4aee5c8 JP |
4626 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4627 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4628 | miimon = BOND_LINK_MON_INTERV; |
4629 | } | |
4630 | ||
4631 | if (updelay < 0) { | |
a4aee5c8 JP |
4632 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4633 | updelay, INT_MAX); | |
1da177e4 LT |
4634 | updelay = 0; |
4635 | } | |
4636 | ||
4637 | if (downdelay < 0) { | |
a4aee5c8 JP |
4638 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4639 | downdelay, INT_MAX); | |
1da177e4 LT |
4640 | downdelay = 0; |
4641 | } | |
4642 | ||
4643 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4644 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4645 | use_carrier); | |
1da177e4 LT |
4646 | use_carrier = 1; |
4647 | } | |
4648 | ||
4649 | /* reset values for 802.3ad */ | |
4650 | if (bond_mode == BOND_MODE_8023AD) { | |
4651 | if (!miimon) { | |
a4aee5c8 | 4652 | 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 | 4653 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4654 | miimon = 100; |
4655 | } | |
4656 | } | |
4657 | ||
bb1d9123 AG |
4658 | if (tx_queues < 1 || tx_queues > 255) { |
4659 | pr_warning("Warning: tx_queues (%d) should be between " | |
4660 | "1 and 255, resetting to %d\n", | |
4661 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4662 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4663 | } | |
4664 | ||
ebd8e497 AG |
4665 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4666 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4667 | "not of valid value (0/1), so it was set to " | |
4668 | "0\n", all_slaves_active); | |
4669 | all_slaves_active = 0; | |
4670 | } | |
4671 | ||
c2952c31 FL |
4672 | if (resend_igmp < 0 || resend_igmp > 255) { |
4673 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4674 | "0 and 255, resetting to %d\n", | |
4675 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4676 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4677 | } | |
4678 | ||
1da177e4 LT |
4679 | /* reset values for TLB/ALB */ |
4680 | if ((bond_mode == BOND_MODE_TLB) || | |
4681 | (bond_mode == BOND_MODE_ALB)) { | |
4682 | if (!miimon) { | |
a4aee5c8 | 4683 | 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 | 4684 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4685 | miimon = 100; |
4686 | } | |
4687 | } | |
4688 | ||
4689 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4690 | 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", |
4691 | updelay); | |
1da177e4 LT |
4692 | } |
4693 | ||
4694 | if (!miimon) { | |
4695 | if (updelay || downdelay) { | |
4696 | /* just warn the user the up/down delay will have | |
4697 | * no effect since miimon is zero... | |
4698 | */ | |
a4aee5c8 JP |
4699 | 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", |
4700 | updelay, downdelay); | |
1da177e4 LT |
4701 | } |
4702 | } else { | |
4703 | /* don't allow arp monitoring */ | |
4704 | if (arp_interval) { | |
a4aee5c8 JP |
4705 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4706 | miimon, arp_interval); | |
1da177e4 LT |
4707 | arp_interval = 0; |
4708 | } | |
4709 | ||
4710 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4711 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4712 | updelay, miimon, | |
4713 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4714 | } |
4715 | ||
4716 | updelay /= miimon; | |
4717 | ||
4718 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4719 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4720 | downdelay, miimon, | |
4721 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4722 | } |
4723 | ||
4724 | downdelay /= miimon; | |
4725 | } | |
4726 | ||
4727 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4728 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4729 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4730 | arp_interval = BOND_LINK_ARP_INTERV; |
4731 | } | |
4732 | ||
4733 | for (arp_ip_count = 0; | |
4734 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4735 | arp_ip_count++) { | |
4736 | /* not complete check, but should be good enough to | |
4737 | catch mistakes */ | |
4738 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4739 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4740 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4741 | arp_interval = 0; |
4742 | } else { | |
d3bb52b0 | 4743 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4744 | arp_target[arp_ip_count] = ip; |
4745 | } | |
4746 | } | |
4747 | ||
4748 | if (arp_interval && !arp_ip_count) { | |
4749 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4750 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4751 | arp_interval); | |
1da177e4 LT |
4752 | arp_interval = 0; |
4753 | } | |
4754 | ||
f5b2b966 JV |
4755 | if (arp_validate) { |
4756 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4757 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4758 | return -EINVAL; |
4759 | } | |
4760 | if (!arp_interval) { | |
a4aee5c8 | 4761 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4762 | return -EINVAL; |
4763 | } | |
4764 | ||
4765 | arp_validate_value = bond_parse_parm(arp_validate, | |
4766 | arp_validate_tbl); | |
4767 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4768 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4769 | arp_validate == NULL ? "NULL" : arp_validate); |
4770 | return -EINVAL; | |
4771 | } | |
4772 | } else | |
4773 | arp_validate_value = 0; | |
4774 | ||
1da177e4 | 4775 | if (miimon) { |
a4aee5c8 | 4776 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4777 | } else if (arp_interval) { |
4778 | int i; | |
4779 | ||
a4aee5c8 JP |
4780 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4781 | arp_interval, | |
4782 | arp_validate_tbl[arp_validate_value].modename, | |
4783 | arp_ip_count); | |
1da177e4 LT |
4784 | |
4785 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4786 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4787 | |
e5e2a8fd | 4788 | pr_info("\n"); |
1da177e4 | 4789 | |
b8a9787e | 4790 | } else if (max_bonds) { |
1da177e4 LT |
4791 | /* miimon and arp_interval not set, we need one so things |
4792 | * work as expected, see bonding.txt for details | |
4793 | */ | |
a4aee5c8 | 4794 | pr_warning("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 |
4795 | } |
4796 | ||
4797 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4798 | /* currently, using a primary only makes sense | |
4799 | * in active backup, TLB or ALB modes | |
4800 | */ | |
a4aee5c8 JP |
4801 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4802 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4803 | primary = NULL; |
4804 | } | |
4805 | ||
a549952a JP |
4806 | if (primary && primary_reselect) { |
4807 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4808 | pri_reselect_tbl); | |
4809 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4810 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4811 | primary_reselect == |
4812 | NULL ? "NULL" : primary_reselect); | |
4813 | return -EINVAL; | |
4814 | } | |
4815 | } else { | |
4816 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4817 | } | |
4818 | ||
3915c1e8 JV |
4819 | if (fail_over_mac) { |
4820 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4821 | fail_over_mac_tbl); | |
4822 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4823 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4824 | arp_validate == NULL ? "NULL" : arp_validate); |
4825 | return -EINVAL; | |
4826 | } | |
4827 | ||
4828 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4829 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4830 | } else { |
4831 | fail_over_mac_value = BOND_FOM_NONE; | |
4832 | } | |
dd957c57 | 4833 | |
1da177e4 LT |
4834 | /* fill params struct with the proper values */ |
4835 | params->mode = bond_mode; | |
169a3e66 | 4836 | params->xmit_policy = xmit_hashtype; |
1da177e4 LT |
4837 | params->miimon = miimon; |
4838 | params->arp_interval = arp_interval; | |
f5b2b966 | 4839 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4840 | params->updelay = updelay; |
4841 | params->downdelay = downdelay; | |
4842 | params->use_carrier = use_carrier; | |
4843 | params->lacp_fast = lacp_fast; | |
4844 | params->primary[0] = 0; | |
a549952a | 4845 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4846 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4847 | params->tx_queues = tx_queues; |
ebd8e497 | 4848 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4849 | params->resend_igmp = resend_igmp; |
1da177e4 LT |
4850 | |
4851 | if (primary) { | |
4852 | strncpy(params->primary, primary, IFNAMSIZ); | |
4853 | params->primary[IFNAMSIZ - 1] = 0; | |
4854 | } | |
4855 | ||
4856 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4857 | ||
4858 | return 0; | |
4859 | } | |
4860 | ||
0daa2303 | 4861 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4862 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 4863 | |
e8a0464c DM |
4864 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4865 | struct netdev_queue *txq, | |
4866 | void *_unused) | |
c773e847 DM |
4867 | { |
4868 | lockdep_set_class(&txq->_xmit_lock, | |
4869 | &bonding_netdev_xmit_lock_key); | |
4870 | } | |
4871 | ||
4872 | static void bond_set_lockdep_class(struct net_device *dev) | |
4873 | { | |
cf508b12 DM |
4874 | lockdep_set_class(&dev->addr_list_lock, |
4875 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4876 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
4877 | } |
4878 | ||
181470fc SH |
4879 | /* |
4880 | * Called from registration process | |
4881 | */ | |
4882 | static int bond_init(struct net_device *bond_dev) | |
4883 | { | |
4884 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4885 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc SH |
4886 | |
4887 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4888 | ||
4889 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4890 | if (!bond->wq) | |
4891 | return -ENOMEM; | |
4892 | ||
4893 | bond_set_lockdep_class(bond_dev); | |
4894 | ||
181470fc | 4895 | bond_create_proc_entry(bond); |
ec87fd3b | 4896 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4897 | |
6151b3d4 | 4898 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4899 | |
f073c7ca TI |
4900 | bond_debug_register(bond); |
4901 | ||
22bedad3 | 4902 | __hw_addr_init(&bond->mc_list); |
181470fc SH |
4903 | return 0; |
4904 | } | |
4905 | ||
88ead977 EB |
4906 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4907 | { | |
4908 | if (tb[IFLA_ADDRESS]) { | |
4909 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4910 | return -EINVAL; | |
4911 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4912 | return -EADDRNOTAVAIL; | |
4913 | } | |
4914 | return 0; | |
4915 | } | |
4916 | ||
4917 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
4918 | .kind = "bond", | |
6639104b | 4919 | .priv_size = sizeof(struct bonding), |
88ead977 EB |
4920 | .setup = bond_setup, |
4921 | .validate = bond_validate, | |
4922 | }; | |
4923 | ||
dfe60397 | 4924 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4925 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4926 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4927 | * set up our sysfs entries. | |
4928 | */ | |
ec87fd3b | 4929 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4930 | { |
4931 | struct net_device *bond_dev; | |
4932 | int res; | |
4933 | ||
4934 | rtnl_lock(); | |
027ea041 | 4935 | |
bb1d9123 AG |
4936 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "", |
4937 | bond_setup, tx_queues); | |
dfe60397 | 4938 | if (!bond_dev) { |
a4aee5c8 | 4939 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4940 | rtnl_unlock(); |
4941 | return -ENOMEM; | |
dfe60397 MW |
4942 | } |
4943 | ||
ec87fd3b | 4944 | dev_net_set(bond_dev, net); |
88ead977 EB |
4945 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4946 | ||
e4b91c48 JV |
4947 | if (!name) { |
4948 | res = dev_alloc_name(bond_dev, "bond%d"); | |
4949 | if (res < 0) | |
9e2e61fb | 4950 | goto out; |
27e6f065 NH |
4951 | } else { |
4952 | /* | |
4953 | * If we're given a name to register | |
4954 | * we need to ensure that its not already | |
4955 | * registered | |
4956 | */ | |
4957 | res = -EEXIST; | |
4958 | if (__dev_get_by_name(net, name) != NULL) | |
4959 | goto out; | |
e4b91c48 JV |
4960 | } |
4961 | ||
dfe60397 | 4962 | res = register_netdevice(bond_dev); |
0daa2303 | 4963 | |
e826eafa PO |
4964 | netif_carrier_off(bond_dev); |
4965 | ||
30c15ba9 | 4966 | out: |
7e083840 | 4967 | rtnl_unlock(); |
9e2e61fb AW |
4968 | if (res < 0) |
4969 | bond_destructor(bond_dev); | |
30c15ba9 | 4970 | return res; |
dfe60397 MW |
4971 | } |
4972 | ||
2c8c1e72 | 4973 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4974 | { |
15449745 | 4975 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4976 | |
4977 | bn->net = net; | |
4978 | INIT_LIST_HEAD(&bn->dev_list); | |
4979 | ||
ec87fd3b | 4980 | bond_create_proc_dir(bn); |
15449745 EB |
4981 | |
4982 | return 0; | |
ec87fd3b EB |
4983 | } |
4984 | ||
2c8c1e72 | 4985 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4986 | { |
15449745 | 4987 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4988 | |
4989 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
4990 | } |
4991 | ||
4992 | static struct pernet_operations bond_net_ops = { | |
4993 | .init = bond_net_init, | |
4994 | .exit = bond_net_exit, | |
15449745 EB |
4995 | .id = &bond_net_id, |
4996 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4997 | }; |
4998 | ||
1da177e4 LT |
4999 | static int __init bonding_init(void) |
5000 | { | |
1da177e4 LT |
5001 | int i; |
5002 | int res; | |
5003 | ||
bd33acc3 | 5004 | pr_info("%s", bond_version); |
1da177e4 | 5005 | |
dfe60397 | 5006 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5007 | if (res) |
dfe60397 | 5008 | goto out; |
1da177e4 | 5009 | |
15449745 | 5010 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
5011 | if (res) |
5012 | goto out; | |
027ea041 | 5013 | |
88ead977 EB |
5014 | res = rtnl_link_register(&bond_link_ops); |
5015 | if (res) | |
6639104b | 5016 | goto err_link; |
88ead977 | 5017 | |
f073c7ca TI |
5018 | bond_create_debugfs(); |
5019 | ||
1da177e4 | 5020 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5021 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5022 | if (res) |
5023 | goto err; | |
1da177e4 LT |
5024 | } |
5025 | ||
b76cdba9 MW |
5026 | res = bond_create_sysfs(); |
5027 | if (res) | |
5028 | goto err; | |
5029 | ||
1da177e4 | 5030 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5031 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
dfe60397 | 5032 | out: |
1da177e4 | 5033 | return res; |
88ead977 EB |
5034 | err: |
5035 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5036 | err_link: |
15449745 | 5037 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5038 | goto out; |
dfe60397 | 5039 | |
1da177e4 LT |
5040 | } |
5041 | ||
5042 | static void __exit bonding_exit(void) | |
5043 | { | |
5044 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5045 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
1da177e4 | 5046 | |
ae68c398 | 5047 | bond_destroy_sysfs(); |
f073c7ca | 5048 | bond_destroy_debugfs(); |
ae68c398 | 5049 | |
88ead977 | 5050 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5051 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
5052 | |
5053 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
5054 | /* |
5055 | * Make sure we don't have an imbalance on our netpoll blocking | |
5056 | */ | |
5057 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 5058 | #endif |
1da177e4 LT |
5059 | } |
5060 | ||
5061 | module_init(bonding_init); | |
5062 | module_exit(bonding_exit); | |
5063 | MODULE_LICENSE("GPL"); | |
5064 | MODULE_VERSION(DRV_VERSION); | |
5065 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5066 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5067 | MODULE_ALIAS_RTNL_LINK("bond"); |