]>
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> | |
f6dc31a8 | 62 | #include <linux/netpoll.h> |
1da177e4 | 63 | #include <linux/inetdevice.h> |
a816c7c7 | 64 | #include <linux/igmp.h> |
1da177e4 LT |
65 | #include <linux/etherdevice.h> |
66 | #include <linux/skbuff.h> | |
67 | #include <net/sock.h> | |
68 | #include <linux/rtnetlink.h> | |
69 | #include <linux/proc_fs.h> | |
70 | #include <linux/seq_file.h> | |
71 | #include <linux/smp.h> | |
72 | #include <linux/if_ether.h> | |
73 | #include <net/arp.h> | |
74 | #include <linux/mii.h> | |
75 | #include <linux/ethtool.h> | |
76 | #include <linux/if_vlan.h> | |
77 | #include <linux/if_bonding.h> | |
b63bb739 | 78 | #include <linux/jiffies.h> |
c3ade5ca | 79 | #include <net/route.h> |
457c4cbc | 80 | #include <net/net_namespace.h> |
ec87fd3b | 81 | #include <net/netns/generic.h> |
1da177e4 LT |
82 | #include "bonding.h" |
83 | #include "bond_3ad.h" | |
84 | #include "bond_alb.h" | |
85 | ||
86 | /*---------------------------- Module parameters ----------------------------*/ | |
87 | ||
88 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
89 | #define BOND_LINK_MON_INTERV 0 | |
90 | #define BOND_LINK_ARP_INTERV 0 | |
91 | ||
92 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
7893b249 | 93 | static int num_grat_arp = 1; |
305d552a | 94 | static int num_unsol_na = 1; |
1da177e4 | 95 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
96 | static int updelay; |
97 | static int downdelay; | |
1da177e4 | 98 | static int use_carrier = 1; |
3d632c3f SH |
99 | static char *mode; |
100 | static char *primary; | |
a549952a | 101 | static char *primary_reselect; |
3d632c3f SH |
102 | static char *lacp_rate; |
103 | static char *ad_select; | |
104 | static char *xmit_hash_policy; | |
1da177e4 | 105 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
106 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
107 | static char *arp_validate; | |
108 | static char *fail_over_mac; | |
d2991f75 | 109 | static struct bond_params bonding_defaults; |
1da177e4 LT |
110 | |
111 | module_param(max_bonds, int, 0); | |
112 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
7893b249 MS |
113 | module_param(num_grat_arp, int, 0644); |
114 | MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packets to send on failover event"); | |
305d552a BH |
115 | module_param(num_unsol_na, int, 0644); |
116 | MODULE_PARM_DESC(num_unsol_na, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event"); | |
1da177e4 LT |
117 | module_param(miimon, int, 0); |
118 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
119 | module_param(updelay, int, 0); | |
120 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
121 | module_param(downdelay, int, 0); | |
2ac47660 MW |
122 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
123 | "in milliseconds"); | |
1da177e4 | 124 | module_param(use_carrier, int, 0); |
2ac47660 MW |
125 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
126 | "0 for off, 1 for on (default)"); | |
1da177e4 | 127 | module_param(mode, charp, 0); |
2ac47660 MW |
128 | MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, " |
129 | "1 for active-backup, 2 for balance-xor, " | |
130 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
131 | "6 for balance-alb"); | |
1da177e4 LT |
132 | module_param(primary, charp, 0); |
133 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
134 | module_param(primary_reselect, charp, 0); |
135 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
136 | "once it comes up; " | |
137 | "0 for always (default), " | |
138 | "1 for only if speed of primary is " | |
139 | "better, " | |
140 | "2 for only on active slave " | |
141 | "failure"); | |
1da177e4 | 142 | module_param(lacp_rate, charp, 0); |
2ac47660 MW |
143 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner " |
144 | "(slow/fast)"); | |
fd989c83 JV |
145 | module_param(ad_select, charp, 0); |
146 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)"); | |
169a3e66 | 147 | module_param(xmit_hash_policy, charp, 0); |
2ac47660 MW |
148 | MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)" |
149 | ", 1 for layer 3+4"); | |
1da177e4 LT |
150 | module_param(arp_interval, int, 0); |
151 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
152 | module_param_array(arp_ip_target, charp, NULL, 0); | |
153 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 JV |
154 | module_param(arp_validate, charp, 0); |
155 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all"); | |
3915c1e8 JV |
156 | module_param(fail_over_mac, charp, 0); |
157 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow"); | |
1da177e4 LT |
158 | |
159 | /*----------------------------- Global variables ----------------------------*/ | |
160 | ||
f71e1309 | 161 | static const char * const version = |
1da177e4 LT |
162 | DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n"; |
163 | ||
f99189b1 | 164 | int bond_net_id __read_mostly; |
1da177e4 | 165 | |
3d632c3f SH |
166 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
167 | static int arp_ip_count; | |
1da177e4 | 168 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
169 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
170 | static int lacp_fast; | |
217df670 | 171 | |
1da177e4 | 172 | |
e97fd7c6 | 173 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
174 | { "slow", AD_LACP_SLOW}, |
175 | { "fast", AD_LACP_FAST}, | |
176 | { NULL, -1}, | |
177 | }; | |
178 | ||
e97fd7c6 | 179 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
180 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
181 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
182 | { "balance-xor", BOND_MODE_XOR}, | |
183 | { "broadcast", BOND_MODE_BROADCAST}, | |
184 | { "802.3ad", BOND_MODE_8023AD}, | |
185 | { "balance-tlb", BOND_MODE_TLB}, | |
186 | { "balance-alb", BOND_MODE_ALB}, | |
187 | { NULL, -1}, | |
188 | }; | |
189 | ||
e97fd7c6 | 190 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
191 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
192 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 193 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
194 | { NULL, -1}, |
195 | }; | |
196 | ||
e97fd7c6 | 197 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
198 | { "none", BOND_ARP_VALIDATE_NONE}, |
199 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
200 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
201 | { "all", BOND_ARP_VALIDATE_ALL}, | |
202 | { NULL, -1}, | |
203 | }; | |
204 | ||
e97fd7c6 | 205 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
206 | { "none", BOND_FOM_NONE}, |
207 | { "active", BOND_FOM_ACTIVE}, | |
208 | { "follow", BOND_FOM_FOLLOW}, | |
209 | { NULL, -1}, | |
210 | }; | |
211 | ||
a549952a JP |
212 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
213 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
214 | { "better", BOND_PRI_RESELECT_BETTER}, | |
215 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
216 | { NULL, -1}, | |
217 | }; | |
218 | ||
fd989c83 JV |
219 | struct bond_parm_tbl ad_select_tbl[] = { |
220 | { "stable", BOND_AD_STABLE}, | |
221 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
222 | { "count", BOND_AD_COUNT}, | |
223 | { NULL, -1}, | |
224 | }; | |
225 | ||
1da177e4 LT |
226 | /*-------------------------- Forward declarations ---------------------------*/ |
227 | ||
c3ade5ca | 228 | static void bond_send_gratuitous_arp(struct bonding *bond); |
181470fc | 229 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 230 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
231 | |
232 | /*---------------------------- General routines -----------------------------*/ | |
233 | ||
4ad072c9 | 234 | static const char *bond_mode_name(int mode) |
1da177e4 | 235 | { |
77afc92b HE |
236 | static const char *names[] = { |
237 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
238 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
239 | [BOND_MODE_XOR] = "load balancing (xor)", | |
240 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 241 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
242 | [BOND_MODE_TLB] = "transmit load balancing", |
243 | [BOND_MODE_ALB] = "adaptive load balancing", | |
244 | }; | |
245 | ||
246 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 247 | return "unknown"; |
77afc92b HE |
248 | |
249 | return names[mode]; | |
1da177e4 LT |
250 | } |
251 | ||
252 | /*---------------------------------- VLAN -----------------------------------*/ | |
253 | ||
254 | /** | |
255 | * bond_add_vlan - add a new vlan id on bond | |
256 | * @bond: bond that got the notification | |
257 | * @vlan_id: the vlan id to add | |
258 | * | |
259 | * Returns -ENOMEM if allocation failed. | |
260 | */ | |
261 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
262 | { | |
263 | struct vlan_entry *vlan; | |
264 | ||
5a03cdb7 | 265 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 266 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 267 | |
305d552a | 268 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 269 | if (!vlan) |
1da177e4 | 270 | return -ENOMEM; |
1da177e4 LT |
271 | |
272 | INIT_LIST_HEAD(&vlan->vlan_list); | |
273 | vlan->vlan_id = vlan_id; | |
274 | ||
275 | write_lock_bh(&bond->lock); | |
276 | ||
277 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
278 | ||
279 | write_unlock_bh(&bond->lock); | |
280 | ||
5a03cdb7 | 281 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
282 | |
283 | return 0; | |
284 | } | |
285 | ||
286 | /** | |
287 | * bond_del_vlan - delete a vlan id from bond | |
288 | * @bond: bond that got the notification | |
289 | * @vlan_id: the vlan id to delete | |
290 | * | |
291 | * returns -ENODEV if @vlan_id was not found in @bond. | |
292 | */ | |
293 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
294 | { | |
0883beca | 295 | struct vlan_entry *vlan; |
1da177e4 LT |
296 | int res = -ENODEV; |
297 | ||
5a03cdb7 | 298 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 LT |
299 | |
300 | write_lock_bh(&bond->lock); | |
301 | ||
0883beca | 302 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
303 | if (vlan->vlan_id == vlan_id) { |
304 | list_del(&vlan->vlan_list); | |
305 | ||
58402054 | 306 | if (bond_is_lb(bond)) |
1da177e4 | 307 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 308 | |
a4aee5c8 JP |
309 | pr_debug("removed VLAN ID %d from bond %s\n", |
310 | vlan_id, bond->dev->name); | |
1da177e4 LT |
311 | |
312 | kfree(vlan); | |
313 | ||
314 | if (list_empty(&bond->vlan_list) && | |
315 | (bond->slave_cnt == 0)) { | |
316 | /* Last VLAN removed and no slaves, so | |
317 | * restore block on adding VLANs. This will | |
318 | * be removed once new slaves that are not | |
319 | * VLAN challenged will be added. | |
320 | */ | |
321 | bond->dev->features |= NETIF_F_VLAN_CHALLENGED; | |
322 | } | |
323 | ||
324 | res = 0; | |
325 | goto out; | |
326 | } | |
327 | } | |
328 | ||
a4aee5c8 JP |
329 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
330 | vlan_id, bond->dev->name); | |
1da177e4 LT |
331 | |
332 | out: | |
333 | write_unlock_bh(&bond->lock); | |
334 | return res; | |
335 | } | |
336 | ||
337 | /** | |
338 | * bond_has_challenged_slaves | |
339 | * @bond: the bond we're working on | |
340 | * | |
341 | * Searches the slave list. Returns 1 if a vlan challenged slave | |
342 | * was found, 0 otherwise. | |
343 | * | |
344 | * Assumes bond->lock is held. | |
345 | */ | |
346 | static int bond_has_challenged_slaves(struct bonding *bond) | |
347 | { | |
348 | struct slave *slave; | |
349 | int i; | |
350 | ||
351 | bond_for_each_slave(bond, slave, i) { | |
352 | if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 353 | pr_debug("found VLAN challenged slave - %s\n", |
a4aee5c8 | 354 | slave->dev->name); |
1da177e4 LT |
355 | return 1; |
356 | } | |
357 | } | |
358 | ||
5a03cdb7 | 359 | pr_debug("no VLAN challenged slaves found\n"); |
1da177e4 LT |
360 | return 0; |
361 | } | |
362 | ||
363 | /** | |
364 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
365 | * @bond: the bond we're working on | |
366 | * @curr: item we're advancing from | |
367 | * | |
368 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
369 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 370 | * |
1da177e4 LT |
371 | * Caller must hold bond->lock |
372 | */ | |
373 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
374 | { | |
375 | struct vlan_entry *next, *last; | |
376 | ||
3d632c3f | 377 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 378 | return NULL; |
1da177e4 LT |
379 | |
380 | if (!curr) { | |
381 | next = list_entry(bond->vlan_list.next, | |
382 | struct vlan_entry, vlan_list); | |
383 | } else { | |
384 | last = list_entry(bond->vlan_list.prev, | |
385 | struct vlan_entry, vlan_list); | |
386 | if (last == curr) { | |
387 | next = list_entry(bond->vlan_list.next, | |
388 | struct vlan_entry, vlan_list); | |
389 | } else { | |
390 | next = list_entry(curr->vlan_list.next, | |
391 | struct vlan_entry, vlan_list); | |
392 | } | |
393 | } | |
394 | ||
395 | return next; | |
396 | } | |
397 | ||
398 | /** | |
399 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 400 | * |
1da177e4 LT |
401 | * @bond: bond device that got this skb for tx. |
402 | * @skb: hw accel VLAN tagged skb to transmit | |
403 | * @slave_dev: slave that is supposed to xmit this skbuff | |
3d632c3f | 404 | * |
1da177e4 LT |
405 | * When the bond gets an skb to transmit that is |
406 | * already hardware accelerated VLAN tagged, and it | |
407 | * needs to relay this skb to a slave that is not | |
408 | * hw accel capable, the skb needs to be "unaccelerated", | |
409 | * i.e. strip the hwaccel tag and re-insert it as part | |
410 | * of the payload. | |
411 | */ | |
3d632c3f SH |
412 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
413 | struct net_device *slave_dev) | |
1da177e4 | 414 | { |
966bc6f4 | 415 | unsigned short uninitialized_var(vlan_id); |
1da177e4 LT |
416 | |
417 | if (!list_empty(&bond->vlan_list) && | |
418 | !(slave_dev->features & NETIF_F_HW_VLAN_TX) && | |
419 | vlan_get_tag(skb, &vlan_id) == 0) { | |
420 | skb->dev = slave_dev; | |
421 | skb = vlan_put_tag(skb, vlan_id); | |
422 | if (!skb) { | |
423 | /* vlan_put_tag() frees the skb in case of error, | |
424 | * so return success here so the calling functions | |
425 | * won't attempt to free is again. | |
426 | */ | |
427 | return 0; | |
428 | } | |
429 | } else { | |
430 | skb->dev = slave_dev; | |
431 | } | |
432 | ||
433 | skb->priority = 1; | |
f6dc31a8 WC |
434 | #ifdef CONFIG_NET_POLL_CONTROLLER |
435 | if (unlikely(bond->dev->priv_flags & IFF_IN_NETPOLL)) { | |
436 | struct netpoll *np = bond->dev->npinfo->netpoll; | |
437 | slave_dev->npinfo = bond->dev->npinfo; | |
438 | np->real_dev = np->dev = skb->dev; | |
439 | slave_dev->priv_flags |= IFF_IN_NETPOLL; | |
440 | netpoll_send_skb(np, skb); | |
441 | slave_dev->priv_flags &= ~IFF_IN_NETPOLL; | |
442 | np->dev = bond->dev; | |
443 | } else | |
444 | #endif | |
445 | dev_queue_xmit(skb); | |
1da177e4 LT |
446 | |
447 | return 0; | |
448 | } | |
449 | ||
450 | /* | |
451 | * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid | |
452 | * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a | |
453 | * lock because: | |
454 | * a. This operation is performed in IOCTL context, | |
455 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
456 | * c. Holding a lock with BH disabled while directly calling a base driver | |
457 | * entry point is generally a BAD idea. | |
3d632c3f | 458 | * |
1da177e4 LT |
459 | * The design of synchronization/protection for this operation in the 8021q |
460 | * module is good for one or more VLAN devices over a single physical device | |
461 | * and cannot be extended for a teaming solution like bonding, so there is a | |
462 | * potential race condition here where a net device from the vlan group might | |
463 | * be referenced (either by a base driver or the 8021q code) while it is being | |
464 | * removed from the system. However, it turns out we're not making matters | |
465 | * worse, and if it works for regular VLAN usage it will work here too. | |
466 | */ | |
467 | ||
468 | /** | |
469 | * bond_vlan_rx_register - Propagates registration to slaves | |
470 | * @bond_dev: bonding net device that got called | |
471 | * @grp: vlan group being registered | |
472 | */ | |
3d632c3f SH |
473 | static void bond_vlan_rx_register(struct net_device *bond_dev, |
474 | struct vlan_group *grp) | |
1da177e4 | 475 | { |
454d7c9b | 476 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
477 | struct slave *slave; |
478 | int i; | |
479 | ||
480 | bond->vlgrp = grp; | |
481 | ||
482 | bond_for_each_slave(bond, slave, i) { | |
483 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 484 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
485 | |
486 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
487 | slave_ops->ndo_vlan_rx_register) { |
488 | slave_ops->ndo_vlan_rx_register(slave_dev, grp); | |
1da177e4 LT |
489 | } |
490 | } | |
491 | } | |
492 | ||
493 | /** | |
494 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
495 | * @bond_dev: bonding net device that got called | |
496 | * @vid: vlan id being added | |
497 | */ | |
498 | static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) | |
499 | { | |
454d7c9b | 500 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
501 | struct slave *slave; |
502 | int i, res; | |
503 | ||
504 | bond_for_each_slave(bond, slave, i) { | |
505 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 506 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
507 | |
508 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a SH |
509 | slave_ops->ndo_vlan_rx_add_vid) { |
510 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid); | |
1da177e4 LT |
511 | } |
512 | } | |
513 | ||
514 | res = bond_add_vlan(bond, vid); | |
515 | if (res) { | |
a4aee5c8 | 516 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 LT |
517 | bond_dev->name, vid); |
518 | } | |
519 | } | |
520 | ||
521 | /** | |
522 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
523 | * @bond_dev: bonding net device that got called | |
524 | * @vid: vlan id being removed | |
525 | */ | |
526 | static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) | |
527 | { | |
454d7c9b | 528 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
529 | struct slave *slave; |
530 | struct net_device *vlan_dev; | |
531 | int i, res; | |
532 | ||
533 | bond_for_each_slave(bond, slave, i) { | |
534 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 535 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
536 | |
537 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a | 538 | slave_ops->ndo_vlan_rx_kill_vid) { |
1da177e4 LT |
539 | /* Save and then restore vlan_dev in the grp array, |
540 | * since the slave's driver might clear it. | |
541 | */ | |
5c15bdec | 542 | vlan_dev = vlan_group_get_device(bond->vlgrp, vid); |
eb7cc59a | 543 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid); |
5c15bdec | 544 | vlan_group_set_device(bond->vlgrp, vid, vlan_dev); |
1da177e4 LT |
545 | } |
546 | } | |
547 | ||
548 | res = bond_del_vlan(bond, vid); | |
549 | if (res) { | |
a4aee5c8 | 550 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 LT |
551 | bond_dev->name, vid); |
552 | } | |
553 | } | |
554 | ||
555 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
556 | { | |
557 | struct vlan_entry *vlan; | |
eb7cc59a | 558 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
559 | |
560 | write_lock_bh(&bond->lock); | |
561 | ||
eb7cc59a | 562 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 563 | goto out; |
1da177e4 LT |
564 | |
565 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
566 | slave_ops->ndo_vlan_rx_register) |
567 | slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp); | |
1da177e4 LT |
568 | |
569 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 570 | !(slave_ops->ndo_vlan_rx_add_vid)) |
1da177e4 | 571 | goto out; |
1da177e4 | 572 | |
eb7cc59a SH |
573 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) |
574 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id); | |
1da177e4 LT |
575 | |
576 | out: | |
577 | write_unlock_bh(&bond->lock); | |
578 | } | |
579 | ||
3d632c3f SH |
580 | static void bond_del_vlans_from_slave(struct bonding *bond, |
581 | struct net_device *slave_dev) | |
1da177e4 | 582 | { |
eb7cc59a | 583 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
584 | struct vlan_entry *vlan; |
585 | struct net_device *vlan_dev; | |
586 | ||
587 | write_lock_bh(&bond->lock); | |
588 | ||
eb7cc59a | 589 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 590 | goto out; |
1da177e4 LT |
591 | |
592 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 593 | !(slave_ops->ndo_vlan_rx_kill_vid)) |
1da177e4 | 594 | goto unreg; |
1da177e4 LT |
595 | |
596 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
597 | /* Save and then restore vlan_dev in the grp array, | |
598 | * since the slave's driver might clear it. | |
599 | */ | |
5c15bdec | 600 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
eb7cc59a | 601 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id); |
5c15bdec | 602 | vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev); |
1da177e4 LT |
603 | } |
604 | ||
605 | unreg: | |
606 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
607 | slave_ops->ndo_vlan_rx_register) |
608 | slave_ops->ndo_vlan_rx_register(slave_dev, NULL); | |
1da177e4 LT |
609 | |
610 | out: | |
611 | write_unlock_bh(&bond->lock); | |
612 | } | |
613 | ||
614 | /*------------------------------- Link status -------------------------------*/ | |
615 | ||
ff59c456 JV |
616 | /* |
617 | * Set the carrier state for the master according to the state of its | |
618 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
619 | * do special 802.3ad magic. | |
620 | * | |
621 | * Returns zero if carrier state does not change, nonzero if it does. | |
622 | */ | |
623 | static int bond_set_carrier(struct bonding *bond) | |
624 | { | |
625 | struct slave *slave; | |
626 | int i; | |
627 | ||
628 | if (bond->slave_cnt == 0) | |
629 | goto down; | |
630 | ||
631 | if (bond->params.mode == BOND_MODE_8023AD) | |
632 | return bond_3ad_set_carrier(bond); | |
633 | ||
634 | bond_for_each_slave(bond, slave, i) { | |
635 | if (slave->link == BOND_LINK_UP) { | |
636 | if (!netif_carrier_ok(bond->dev)) { | |
637 | netif_carrier_on(bond->dev); | |
638 | return 1; | |
639 | } | |
640 | return 0; | |
641 | } | |
642 | } | |
643 | ||
644 | down: | |
645 | if (netif_carrier_ok(bond->dev)) { | |
646 | netif_carrier_off(bond->dev); | |
647 | return 1; | |
648 | } | |
649 | return 0; | |
650 | } | |
651 | ||
1da177e4 LT |
652 | /* |
653 | * Get link speed and duplex from the slave's base driver | |
654 | * using ethtool. If for some reason the call fails or the | |
655 | * values are invalid, fake speed and duplex to 100/Full | |
656 | * and return error. | |
657 | */ | |
658 | static int bond_update_speed_duplex(struct slave *slave) | |
659 | { | |
660 | struct net_device *slave_dev = slave->dev; | |
1da177e4 | 661 | struct ethtool_cmd etool; |
61a44b9c | 662 | int res; |
1da177e4 LT |
663 | |
664 | /* Fake speed and duplex */ | |
665 | slave->speed = SPEED_100; | |
666 | slave->duplex = DUPLEX_FULL; | |
667 | ||
61a44b9c MW |
668 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
669 | return -1; | |
1da177e4 | 670 | |
61a44b9c MW |
671 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
672 | if (res < 0) | |
1da177e4 | 673 | return -1; |
1da177e4 | 674 | |
1da177e4 LT |
675 | switch (etool.speed) { |
676 | case SPEED_10: | |
677 | case SPEED_100: | |
678 | case SPEED_1000: | |
94dbffd5 | 679 | case SPEED_10000: |
1da177e4 LT |
680 | break; |
681 | default: | |
682 | return -1; | |
683 | } | |
684 | ||
685 | switch (etool.duplex) { | |
686 | case DUPLEX_FULL: | |
687 | case DUPLEX_HALF: | |
688 | break; | |
689 | default: | |
690 | return -1; | |
691 | } | |
692 | ||
693 | slave->speed = etool.speed; | |
694 | slave->duplex = etool.duplex; | |
695 | ||
696 | return 0; | |
697 | } | |
698 | ||
699 | /* | |
700 | * if <dev> supports MII link status reporting, check its link status. | |
701 | * | |
702 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 703 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
704 | * |
705 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
706 | * can't tell and just pretend it is), or 0, meaning that the link is | |
707 | * down. | |
708 | * | |
709 | * If reporting is non-zero, instead of faking link up, return -1 if | |
710 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
711 | * support them). If use_carrier is set, return whatever it says. | |
712 | * It'd be nice if there was a good way to tell if a driver supports | |
713 | * netif_carrier, but there really isn't. | |
714 | */ | |
3d632c3f SH |
715 | static int bond_check_dev_link(struct bonding *bond, |
716 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 717 | { |
eb7cc59a | 718 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 719 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
720 | struct ifreq ifr; |
721 | struct mii_ioctl_data *mii; | |
1da177e4 | 722 | |
6c988853 PG |
723 | if (!reporting && !netif_running(slave_dev)) |
724 | return 0; | |
725 | ||
eb7cc59a | 726 | if (bond->params.use_carrier) |
1da177e4 | 727 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 728 | |
29112f4e JP |
729 | /* Try to get link status using Ethtool first. */ |
730 | if (slave_dev->ethtool_ops) { | |
731 | if (slave_dev->ethtool_ops->get_link) { | |
732 | u32 link; | |
733 | ||
734 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
735 | ||
736 | return link ? BMSR_LSTATUS : 0; | |
737 | } | |
738 | } | |
739 | ||
3d632c3f | 740 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 741 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
742 | if (ioctl) { |
743 | /* TODO: set pointer to correct ioctl on a per team member */ | |
744 | /* bases to make this more efficient. that is, once */ | |
745 | /* we determine the correct ioctl, we will always */ | |
746 | /* call it and not the others for that team */ | |
747 | /* member. */ | |
748 | ||
749 | /* | |
750 | * We cannot assume that SIOCGMIIPHY will also read a | |
751 | * register; not all network drivers (e.g., e100) | |
752 | * support that. | |
753 | */ | |
754 | ||
755 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
756 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
757 | mii = if_mii(&ifr); | |
758 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
759 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
760 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
761 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
762 | } |
763 | } | |
764 | ||
1da177e4 LT |
765 | /* |
766 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 767 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
768 | * cannot report link status). If not reporting, pretend |
769 | * we're ok. | |
770 | */ | |
3d632c3f | 771 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
772 | } |
773 | ||
774 | /*----------------------------- Multicast list ------------------------------*/ | |
775 | ||
1da177e4 LT |
776 | /* |
777 | * Push the promiscuity flag down to appropriate slaves | |
778 | */ | |
7e1a1ac1 | 779 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 780 | { |
7e1a1ac1 | 781 | int err = 0; |
1da177e4 LT |
782 | if (USES_PRIMARY(bond->params.mode)) { |
783 | /* write lock already acquired */ | |
784 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
785 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
786 | inc); | |
1da177e4 LT |
787 | } |
788 | } else { | |
789 | struct slave *slave; | |
790 | int i; | |
791 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
792 | err = dev_set_promiscuity(slave->dev, inc); |
793 | if (err) | |
794 | return err; | |
1da177e4 LT |
795 | } |
796 | } | |
7e1a1ac1 | 797 | return err; |
1da177e4 LT |
798 | } |
799 | ||
800 | /* | |
801 | * Push the allmulti flag down to all slaves | |
802 | */ | |
7e1a1ac1 | 803 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 804 | { |
7e1a1ac1 | 805 | int err = 0; |
1da177e4 LT |
806 | if (USES_PRIMARY(bond->params.mode)) { |
807 | /* write lock already acquired */ | |
808 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
809 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
810 | inc); | |
1da177e4 LT |
811 | } |
812 | } else { | |
813 | struct slave *slave; | |
814 | int i; | |
815 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
816 | err = dev_set_allmulti(slave->dev, inc); |
817 | if (err) | |
818 | return err; | |
1da177e4 LT |
819 | } |
820 | } | |
7e1a1ac1 | 821 | return err; |
1da177e4 LT |
822 | } |
823 | ||
824 | /* | |
825 | * Add a Multicast address to slaves | |
826 | * according to mode | |
827 | */ | |
22bedad3 | 828 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
829 | { |
830 | if (USES_PRIMARY(bond->params.mode)) { | |
831 | /* write lock already acquired */ | |
3d632c3f | 832 | if (bond->curr_active_slave) |
22bedad3 | 833 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
834 | } else { |
835 | struct slave *slave; | |
836 | int i; | |
3d632c3f SH |
837 | |
838 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 839 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
840 | } |
841 | } | |
842 | ||
843 | /* | |
844 | * Remove a multicast address from slave | |
845 | * according to mode | |
846 | */ | |
22bedad3 | 847 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
848 | { |
849 | if (USES_PRIMARY(bond->params.mode)) { | |
850 | /* write lock already acquired */ | |
3d632c3f | 851 | if (bond->curr_active_slave) |
22bedad3 | 852 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
853 | } else { |
854 | struct slave *slave; | |
855 | int i; | |
856 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 857 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
858 | } |
859 | } | |
860 | } | |
861 | ||
a816c7c7 JV |
862 | |
863 | /* | |
864 | * Retrieve the list of registered multicast addresses for the bonding | |
865 | * device and retransmit an IGMP JOIN request to the current active | |
866 | * slave. | |
867 | */ | |
868 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
869 | { | |
870 | struct in_device *in_dev; | |
871 | struct ip_mc_list *im; | |
872 | ||
873 | rcu_read_lock(); | |
874 | in_dev = __in_dev_get_rcu(bond->dev); | |
875 | if (in_dev) { | |
3d632c3f | 876 | for (im = in_dev->mc_list; im; im = im->next) |
a816c7c7 | 877 | ip_mc_rejoin_group(im); |
a816c7c7 JV |
878 | } |
879 | ||
880 | rcu_read_unlock(); | |
881 | } | |
882 | ||
1da177e4 LT |
883 | /* |
884 | * flush all members of flush->mc_list from device dev->mc_list | |
885 | */ | |
3d632c3f SH |
886 | static void bond_mc_list_flush(struct net_device *bond_dev, |
887 | struct net_device *slave_dev) | |
1da177e4 | 888 | { |
454d7c9b | 889 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 890 | struct netdev_hw_addr *ha; |
1da177e4 | 891 | |
22bedad3 JP |
892 | netdev_for_each_mc_addr(ha, bond_dev) |
893 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
894 | |
895 | if (bond->params.mode == BOND_MODE_8023AD) { | |
896 | /* del lacpdu mc addr from mc list */ | |
897 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
898 | ||
22bedad3 | 899 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
900 | } |
901 | } | |
902 | ||
903 | /*--------------------------- Active slave change ---------------------------*/ | |
904 | ||
905 | /* | |
906 | * Update the mc list and multicast-related flags for the new and | |
907 | * old active slaves (if any) according to the multicast mode, and | |
908 | * promiscuous flags unconditionally. | |
909 | */ | |
3d632c3f SH |
910 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
911 | struct slave *old_active) | |
1da177e4 | 912 | { |
22bedad3 | 913 | struct netdev_hw_addr *ha; |
1da177e4 | 914 | |
3d632c3f | 915 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
916 | /* nothing to do - mc list is already up-to-date on |
917 | * all slaves | |
918 | */ | |
919 | return; | |
1da177e4 LT |
920 | |
921 | if (old_active) { | |
3d632c3f | 922 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 923 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 924 | |
3d632c3f | 925 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 926 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 927 | |
22bedad3 JP |
928 | netdev_for_each_mc_addr(ha, bond->dev) |
929 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
930 | } |
931 | ||
932 | if (new_active) { | |
7e1a1ac1 | 933 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 934 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 935 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 936 | |
3d632c3f | 937 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 938 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 939 | |
22bedad3 JP |
940 | netdev_for_each_mc_addr(ha, bond->dev) |
941 | dev_mc_add(new_active->dev, ha->addr); | |
a816c7c7 | 942 | bond_resend_igmp_join_requests(bond); |
1da177e4 LT |
943 | } |
944 | } | |
945 | ||
3915c1e8 JV |
946 | /* |
947 | * bond_do_fail_over_mac | |
948 | * | |
949 | * Perform special MAC address swapping for fail_over_mac settings | |
950 | * | |
951 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
952 | */ | |
953 | static void bond_do_fail_over_mac(struct bonding *bond, | |
954 | struct slave *new_active, | |
955 | struct slave *old_active) | |
1f78d9f9 HE |
956 | __releases(&bond->curr_slave_lock) |
957 | __releases(&bond->lock) | |
958 | __acquires(&bond->lock) | |
959 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
960 | { |
961 | u8 tmp_mac[ETH_ALEN]; | |
962 | struct sockaddr saddr; | |
963 | int rv; | |
964 | ||
965 | switch (bond->params.fail_over_mac) { | |
966 | case BOND_FOM_ACTIVE: | |
967 | if (new_active) | |
968 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
969 | new_active->dev->addr_len); | |
970 | break; | |
971 | case BOND_FOM_FOLLOW: | |
972 | /* | |
973 | * if new_active && old_active, swap them | |
974 | * if just old_active, do nothing (going to no active slave) | |
975 | * if just new_active, set new_active to bond's MAC | |
976 | */ | |
977 | if (!new_active) | |
978 | return; | |
979 | ||
980 | write_unlock_bh(&bond->curr_slave_lock); | |
981 | read_unlock(&bond->lock); | |
982 | ||
983 | if (old_active) { | |
984 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
985 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
986 | ETH_ALEN); | |
987 | saddr.sa_family = new_active->dev->type; | |
988 | } else { | |
989 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
990 | saddr.sa_family = bond->dev->type; | |
991 | } | |
992 | ||
993 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
994 | if (rv) { | |
a4aee5c8 | 995 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
996 | bond->dev->name, -rv, new_active->dev->name); |
997 | goto out; | |
998 | } | |
999 | ||
1000 | if (!old_active) | |
1001 | goto out; | |
1002 | ||
1003 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1004 | saddr.sa_family = old_active->dev->type; | |
1005 | ||
1006 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1007 | if (rv) | |
a4aee5c8 | 1008 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1009 | bond->dev->name, -rv, new_active->dev->name); |
1010 | out: | |
1011 | read_lock(&bond->lock); | |
1012 | write_lock_bh(&bond->curr_slave_lock); | |
1013 | break; | |
1014 | default: | |
a4aee5c8 | 1015 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
1016 | bond->dev->name, bond->params.fail_over_mac); |
1017 | break; | |
1018 | } | |
1019 | ||
1020 | } | |
1021 | ||
a549952a JP |
1022 | static bool bond_should_change_active(struct bonding *bond) |
1023 | { | |
1024 | struct slave *prim = bond->primary_slave; | |
1025 | struct slave *curr = bond->curr_active_slave; | |
1026 | ||
1027 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1028 | return true; | |
1029 | if (bond->force_primary) { | |
1030 | bond->force_primary = false; | |
1031 | return true; | |
1032 | } | |
1033 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1034 | (prim->speed < curr->speed || | |
1035 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1036 | return false; | |
1037 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1038 | return false; | |
1039 | return true; | |
1040 | } | |
3915c1e8 | 1041 | |
1da177e4 LT |
1042 | /** |
1043 | * find_best_interface - select the best available slave to be the active one | |
1044 | * @bond: our bonding struct | |
1045 | * | |
1046 | * Warning: Caller must hold curr_slave_lock for writing. | |
1047 | */ | |
1048 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1049 | { | |
1050 | struct slave *new_active, *old_active; | |
1051 | struct slave *bestslave = NULL; | |
1052 | int mintime = bond->params.updelay; | |
1053 | int i; | |
1054 | ||
49b4ad92 | 1055 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1056 | |
1057 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1058 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1059 | new_active = bond->first_slave; |
3d632c3f | 1060 | else |
1da177e4 | 1061 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1062 | } |
1063 | ||
1da177e4 | 1064 | if ((bond->primary_slave) && |
a549952a JP |
1065 | bond->primary_slave->link == BOND_LINK_UP && |
1066 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1067 | new_active = bond->primary_slave; |
1068 | } | |
1069 | ||
1070 | /* remember where to stop iterating over the slaves */ | |
1071 | old_active = new_active; | |
1072 | ||
1073 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1074 | if (new_active->link == BOND_LINK_UP) { |
1075 | return new_active; | |
1076 | } else if (new_active->link == BOND_LINK_BACK && | |
1077 | IS_UP(new_active->dev)) { | |
1078 | /* link up, but waiting for stabilization */ | |
1079 | if (new_active->delay < mintime) { | |
1080 | mintime = new_active->delay; | |
1081 | bestslave = new_active; | |
1da177e4 LT |
1082 | } |
1083 | } | |
1084 | } | |
1085 | ||
1086 | return bestslave; | |
1087 | } | |
1088 | ||
1089 | /** | |
1090 | * change_active_interface - change the active slave into the specified one | |
1091 | * @bond: our bonding struct | |
1092 | * @new: the new slave to make the active one | |
1093 | * | |
1094 | * Set the new slave to the bond's settings and unset them on the old | |
1095 | * curr_active_slave. | |
1096 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1097 | * | |
1098 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1099 | * because it is apparently the best available slave we have, even though its | |
1100 | * updelay hasn't timed out yet. | |
1101 | * | |
3915c1e8 JV |
1102 | * If new_active is not NULL, caller must hold bond->lock for read and |
1103 | * curr_slave_lock for write_bh. | |
1da177e4 | 1104 | */ |
a77b5325 | 1105 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1106 | { |
1107 | struct slave *old_active = bond->curr_active_slave; | |
1108 | ||
3d632c3f | 1109 | if (old_active == new_active) |
1da177e4 | 1110 | return; |
1da177e4 LT |
1111 | |
1112 | if (new_active) { | |
b2220cad JV |
1113 | new_active->jiffies = jiffies; |
1114 | ||
1da177e4 LT |
1115 | if (new_active->link == BOND_LINK_BACK) { |
1116 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1117 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1118 | bond->dev->name, new_active->dev->name, | |
1119 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1120 | } |
1121 | ||
1122 | new_active->delay = 0; | |
1123 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1124 | |
3d632c3f | 1125 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1126 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1127 | |
58402054 | 1128 | if (bond_is_lb(bond)) |
1da177e4 | 1129 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1130 | } else { |
1131 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1132 | pr_info("%s: making interface %s the new active one.\n", |
1133 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1134 | } |
1135 | } | |
1136 | } | |
1137 | ||
3d632c3f | 1138 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1139 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1140 | |
58402054 | 1141 | if (bond_is_lb(bond)) { |
1da177e4 | 1142 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1143 | if (old_active) |
1144 | bond_set_slave_inactive_flags(old_active); | |
1145 | if (new_active) | |
1146 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1147 | } else { |
1148 | bond->curr_active_slave = new_active; | |
1149 | } | |
c3ade5ca JV |
1150 | |
1151 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1152 | if (old_active) |
c3ade5ca | 1153 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1154 | |
1155 | if (new_active) { | |
1156 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1157 | |
709f8a45 OG |
1158 | if (bond->params.fail_over_mac) |
1159 | bond_do_fail_over_mac(bond, new_active, | |
1160 | old_active); | |
3915c1e8 | 1161 | |
709f8a45 | 1162 | bond->send_grat_arp = bond->params.num_grat_arp; |
b59f9f74 | 1163 | bond_send_gratuitous_arp(bond); |
01f3109d | 1164 | |
305d552a BH |
1165 | bond->send_unsol_na = bond->params.num_unsol_na; |
1166 | bond_send_unsolicited_na(bond); | |
1167 | ||
01f3109d OG |
1168 | write_unlock_bh(&bond->curr_slave_lock); |
1169 | read_unlock(&bond->lock); | |
1170 | ||
75c78500 | 1171 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1172 | |
1173 | read_lock(&bond->lock); | |
1174 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1175 | } |
c3ade5ca | 1176 | } |
a2fd940f AG |
1177 | |
1178 | /* resend IGMP joins since all were sent on curr_active_slave */ | |
1179 | if (bond->params.mode == BOND_MODE_ROUNDROBIN) { | |
1180 | bond_resend_igmp_join_requests(bond); | |
1181 | } | |
1da177e4 LT |
1182 | } |
1183 | ||
1184 | /** | |
1185 | * bond_select_active_slave - select a new active slave, if needed | |
1186 | * @bond: our bonding struct | |
1187 | * | |
3d632c3f | 1188 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1189 | * - The old curr_active_slave has been released or lost its link. |
1190 | * - The primary_slave has got its link back. | |
1191 | * - A slave has got its link back and there's no old curr_active_slave. | |
1192 | * | |
3915c1e8 | 1193 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1194 | */ |
a77b5325 | 1195 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1196 | { |
1197 | struct slave *best_slave; | |
ff59c456 | 1198 | int rv; |
1da177e4 LT |
1199 | |
1200 | best_slave = bond_find_best_slave(bond); | |
1201 | if (best_slave != bond->curr_active_slave) { | |
1202 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1203 | rv = bond_set_carrier(bond); |
1204 | if (!rv) | |
1205 | return; | |
1206 | ||
1207 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1208 | pr_info("%s: first active interface up!\n", |
1209 | bond->dev->name); | |
ff59c456 | 1210 | } else { |
a4aee5c8 JP |
1211 | pr_info("%s: now running without any active interface !\n", |
1212 | bond->dev->name); | |
ff59c456 | 1213 | } |
1da177e4 LT |
1214 | } |
1215 | } | |
1216 | ||
1217 | /*--------------------------- slave list handling ---------------------------*/ | |
1218 | ||
1219 | /* | |
1220 | * This function attaches the slave to the end of list. | |
1221 | * | |
1222 | * bond->lock held for writing by caller. | |
1223 | */ | |
1224 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1225 | { | |
1226 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1227 | new_slave->next = new_slave; | |
1228 | new_slave->prev = new_slave; | |
1229 | bond->first_slave = new_slave; | |
1230 | } else { | |
1231 | new_slave->next = bond->first_slave; | |
1232 | new_slave->prev = bond->first_slave->prev; | |
1233 | new_slave->next->prev = new_slave; | |
1234 | new_slave->prev->next = new_slave; | |
1235 | } | |
1236 | ||
1237 | bond->slave_cnt++; | |
1238 | } | |
1239 | ||
1240 | /* | |
1241 | * This function detaches the slave from the list. | |
1242 | * WARNING: no check is made to verify if the slave effectively | |
1243 | * belongs to <bond>. | |
1244 | * Nothing is freed on return, structures are just unchained. | |
1245 | * If any slave pointer in bond was pointing to <slave>, | |
1246 | * it should be changed by the calling function. | |
1247 | * | |
1248 | * bond->lock held for writing by caller. | |
1249 | */ | |
1250 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1251 | { | |
3d632c3f | 1252 | if (slave->next) |
1da177e4 | 1253 | slave->next->prev = slave->prev; |
1da177e4 | 1254 | |
3d632c3f | 1255 | if (slave->prev) |
1da177e4 | 1256 | slave->prev->next = slave->next; |
1da177e4 LT |
1257 | |
1258 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1259 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1260 | bond->first_slave = slave->next; | |
1261 | } else { | |
1262 | bond->first_slave = NULL; /* slave was the last one */ | |
1263 | } | |
1264 | } | |
1265 | ||
1266 | slave->next = NULL; | |
1267 | slave->prev = NULL; | |
1268 | bond->slave_cnt--; | |
1269 | } | |
1270 | ||
f6dc31a8 WC |
1271 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1272 | /* | |
1273 | * You must hold read lock on bond->lock before calling this. | |
1274 | */ | |
1275 | static bool slaves_support_netpoll(struct net_device *bond_dev) | |
1276 | { | |
1277 | struct bonding *bond = netdev_priv(bond_dev); | |
1278 | struct slave *slave; | |
1279 | int i = 0; | |
1280 | bool ret = true; | |
1281 | ||
1282 | bond_for_each_slave(bond, slave, i) { | |
1283 | if ((slave->dev->priv_flags & IFF_DISABLE_NETPOLL) || | |
1284 | !slave->dev->netdev_ops->ndo_poll_controller) | |
1285 | ret = false; | |
1286 | } | |
1287 | return i != 0 && ret; | |
1288 | } | |
1289 | ||
1290 | static void bond_poll_controller(struct net_device *bond_dev) | |
1291 | { | |
1292 | struct net_device *dev = bond_dev->npinfo->netpoll->real_dev; | |
1293 | if (dev != bond_dev) | |
1294 | netpoll_poll_dev(dev); | |
1295 | } | |
1296 | ||
1297 | static void bond_netpoll_cleanup(struct net_device *bond_dev) | |
1298 | { | |
1299 | struct bonding *bond = netdev_priv(bond_dev); | |
1300 | struct slave *slave; | |
1301 | const struct net_device_ops *ops; | |
1302 | int i; | |
1303 | ||
1304 | read_lock(&bond->lock); | |
1305 | bond_dev->npinfo = NULL; | |
1306 | bond_for_each_slave(bond, slave, i) { | |
1307 | if (slave->dev) { | |
1308 | ops = slave->dev->netdev_ops; | |
1309 | if (ops->ndo_netpoll_cleanup) | |
1310 | ops->ndo_netpoll_cleanup(slave->dev); | |
1311 | else | |
1312 | slave->dev->npinfo = NULL; | |
1313 | } | |
1314 | } | |
1315 | read_unlock(&bond->lock); | |
1316 | } | |
1317 | ||
1318 | #else | |
1319 | ||
1320 | static void bond_netpoll_cleanup(struct net_device *bond_dev) | |
1321 | { | |
1322 | } | |
1323 | ||
1324 | #endif | |
1325 | ||
1da177e4 LT |
1326 | /*---------------------------------- IOCTL ----------------------------------*/ |
1327 | ||
4ad072c9 AB |
1328 | static int bond_sethwaddr(struct net_device *bond_dev, |
1329 | struct net_device *slave_dev) | |
1da177e4 | 1330 | { |
5a03cdb7 HE |
1331 | pr_debug("bond_dev=%p\n", bond_dev); |
1332 | pr_debug("slave_dev=%p\n", slave_dev); | |
1333 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1334 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1335 | return 0; | |
1336 | } | |
1337 | ||
7f353bf2 HX |
1338 | #define BOND_VLAN_FEATURES \ |
1339 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1340 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1341 | |
3d632c3f | 1342 | /* |
8e3babcd | 1343 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1344 | * feature bits are managed elsewhere, so preserve those feature bits |
1345 | * on the master device. | |
8531c5ff AK |
1346 | */ |
1347 | static int bond_compute_features(struct bonding *bond) | |
1348 | { | |
8531c5ff AK |
1349 | struct slave *slave; |
1350 | struct net_device *bond_dev = bond->dev; | |
7f353bf2 | 1351 | unsigned long features = bond_dev->features; |
278339a4 | 1352 | unsigned long vlan_features = 0; |
3158bf7d MS |
1353 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1354 | bond_dev->hard_header_len); | |
8e3babcd | 1355 | int i; |
8531c5ff | 1356 | |
7f353bf2 | 1357 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
b63365a2 HX |
1358 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM; |
1359 | ||
1360 | if (!bond->first_slave) | |
1361 | goto done; | |
1362 | ||
1363 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1364 | |
278339a4 | 1365 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1366 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1367 | features = netdev_increment_features(features, |
1368 | slave->dev->features, | |
1369 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1370 | vlan_features = netdev_increment_features(vlan_features, |
1371 | slave->dev->vlan_features, | |
1372 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1373 | if (slave->dev->hard_header_len > max_hard_header_len) |
1374 | max_hard_header_len = slave->dev->hard_header_len; | |
1375 | } | |
8531c5ff | 1376 | |
b63365a2 | 1377 | done: |
7f353bf2 | 1378 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
b63365a2 | 1379 | bond_dev->features = netdev_fix_features(features, NULL); |
278339a4 | 1380 | bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL); |
54ef3137 | 1381 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1382 | |
1383 | return 0; | |
1384 | } | |
1385 | ||
872254dd MS |
1386 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1387 | struct net_device *slave_dev) | |
1388 | { | |
454d7c9b | 1389 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1390 | |
00829823 | 1391 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1392 | |
1393 | bond_dev->type = slave_dev->type; | |
1394 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1395 | bond_dev->addr_len = slave_dev->addr_len; | |
1396 | ||
1397 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1398 | slave_dev->addr_len); | |
d90a162a | 1399 | bond->setup_by_slave = 1; |
872254dd MS |
1400 | } |
1401 | ||
1da177e4 | 1402 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1403 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1404 | { |
454d7c9b | 1405 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1406 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1407 | struct slave *new_slave = NULL; |
22bedad3 | 1408 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1409 | struct sockaddr addr; |
1410 | int link_reporting; | |
1411 | int old_features = bond_dev->features; | |
1412 | int res = 0; | |
1413 | ||
552709d5 | 1414 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1415 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1416 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1417 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1418 | } |
1419 | ||
1420 | /* bond must be initialized by bond_open() before enslaving */ | |
1421 | if (!(bond_dev->flags & IFF_UP)) { | |
a4aee5c8 JP |
1422 | pr_warning("%s: master_dev is not up in bond_enslave\n", |
1423 | bond_dev->name); | |
1da177e4 LT |
1424 | } |
1425 | ||
1426 | /* already enslaved */ | |
1427 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1428 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1429 | return -EBUSY; |
1430 | } | |
1431 | ||
1432 | /* vlan challenged mutual exclusion */ | |
1433 | /* no need to lock since we're protected by rtnl_lock */ | |
1434 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1435 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 | 1436 | if (!list_empty(&bond->vlan_list)) { |
a4aee5c8 JP |
1437 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1438 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1439 | return -EPERM; |
1440 | } else { | |
a4aee5c8 JP |
1441 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1442 | bond_dev->name, slave_dev->name, | |
1443 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1444 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1445 | } | |
1446 | } else { | |
5a03cdb7 | 1447 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1448 | if (bond->slave_cnt == 0) { |
1449 | /* First slave, and it is not VLAN challenged, | |
1450 | * so remove the block of adding VLANs over the bond. | |
1451 | */ | |
1452 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1453 | } | |
1454 | } | |
1455 | ||
217df670 JV |
1456 | /* |
1457 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1458 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1459 | * the current ifenslave will set the interface down prior to |
1460 | * enslaving it; the old ifenslave will not. | |
1461 | */ | |
1462 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1463 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1464 | slave_dev->name); |
1465 | res = -EPERM; | |
1466 | goto err_undo_flags; | |
1467 | } | |
1da177e4 | 1468 | |
872254dd MS |
1469 | /* set bonding device ether type by slave - bonding netdevices are |
1470 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1471 | * there is a need to override some of the type dependent attribs/funcs. | |
1472 | * | |
1473 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1474 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1475 | */ | |
1476 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1477 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1478 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1479 | bond_dev->name, |
1480 | bond_dev->type, slave_dev->type); | |
75c78500 | 1481 | |
3ca5b404 JP |
1482 | res = netdev_bonding_change(bond_dev, |
1483 | NETDEV_PRE_TYPE_CHANGE); | |
1484 | res = notifier_to_errno(res); | |
1485 | if (res) { | |
1486 | pr_err("%s: refused to change device type\n", | |
1487 | bond_dev->name); | |
1488 | res = -EBUSY; | |
1489 | goto err_undo_flags; | |
1490 | } | |
75c78500 | 1491 | |
32a806c1 | 1492 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1493 | dev_uc_flush(bond_dev); |
22bedad3 | 1494 | dev_mc_flush(bond_dev); |
32a806c1 | 1495 | |
e36b9d16 MS |
1496 | if (slave_dev->type != ARPHRD_ETHER) |
1497 | bond_setup_by_slave(bond_dev, slave_dev); | |
1498 | else | |
1499 | ether_setup(bond_dev); | |
75c78500 | 1500 | |
93d9b7d7 JP |
1501 | netdev_bonding_change(bond_dev, |
1502 | NETDEV_POST_TYPE_CHANGE); | |
e36b9d16 | 1503 | } |
872254dd | 1504 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1505 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1506 | slave_dev->name, | |
1507 | slave_dev->type, bond_dev->type); | |
1508 | res = -EINVAL; | |
1509 | goto err_undo_flags; | |
872254dd MS |
1510 | } |
1511 | ||
eb7cc59a | 1512 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1513 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1514 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1515 | bond_dev->name); | |
3915c1e8 JV |
1516 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1517 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1518 | 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", |
1519 | bond_dev->name); | |
2ab82852 MS |
1520 | res = -EOPNOTSUPP; |
1521 | goto err_undo_flags; | |
1522 | } | |
1da177e4 LT |
1523 | } |
1524 | ||
c20811a7 JP |
1525 | /* If this is the first slave, then we need to set the master's hardware |
1526 | * address to be the same as the slave's. */ | |
1527 | if (bond->slave_cnt == 0) | |
1528 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, | |
1529 | slave_dev->addr_len); | |
1530 | ||
1531 | ||
243cb4e5 | 1532 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1533 | if (!new_slave) { |
1534 | res = -ENOMEM; | |
1535 | goto err_undo_flags; | |
1536 | } | |
1537 | ||
b15ba0fb JP |
1538 | /* Save slave's original mtu and then set it to match the bond */ |
1539 | new_slave->original_mtu = slave_dev->mtu; | |
1540 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1541 | if (res) { | |
1542 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1543 | goto err_free; | |
1544 | } | |
1545 | ||
217df670 JV |
1546 | /* |
1547 | * Save slave's original ("permanent") mac address for modes | |
1548 | * that need it, and for restoring it upon release, and then | |
1549 | * set it to the master's address | |
1550 | */ | |
1551 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1552 | |
dd957c57 | 1553 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1554 | /* |
1555 | * Set slave to master's mac address. The application already | |
1556 | * set the master's mac address to that of the first slave | |
1557 | */ | |
1558 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1559 | addr.sa_family = slave_dev->type; | |
1560 | res = dev_set_mac_address(slave_dev, &addr); | |
1561 | if (res) { | |
5a03cdb7 | 1562 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1563 | goto err_restore_mtu; |
2ab82852 | 1564 | } |
217df670 | 1565 | } |
1da177e4 | 1566 | |
c2edacf8 JV |
1567 | res = netdev_set_master(slave_dev, bond_dev); |
1568 | if (res) { | |
5a03cdb7 | 1569 | pr_debug("Error %d calling netdev_set_master\n", res); |
569f0c4d | 1570 | goto err_restore_mac; |
c2edacf8 | 1571 | } |
217df670 JV |
1572 | /* open the slave since the application closed it */ |
1573 | res = dev_open(slave_dev); | |
1574 | if (res) { | |
3d632c3f | 1575 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
569f0c4d | 1576 | goto err_unset_master; |
1da177e4 LT |
1577 | } |
1578 | ||
1da177e4 | 1579 | new_slave->dev = slave_dev; |
0b680e75 | 1580 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1581 | |
58402054 | 1582 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1583 | /* bond_alb_init_slave() must be called before all other stages since |
1584 | * it might fail and we do not want to have to undo everything | |
1585 | */ | |
1586 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1587 | if (res) |
569f0c4d | 1588 | goto err_close; |
1da177e4 LT |
1589 | } |
1590 | ||
1591 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1592 | * master's promisc (and mc) settings only if it becomes the | |
1593 | * curr_active_slave, and that is taken care of later when calling | |
1594 | * bond_change_active() | |
1595 | */ | |
1596 | if (!USES_PRIMARY(bond->params.mode)) { | |
1597 | /* set promiscuity level to new slave */ | |
1598 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1599 | res = dev_set_promiscuity(slave_dev, 1); |
1600 | if (res) | |
1601 | goto err_close; | |
1da177e4 LT |
1602 | } |
1603 | ||
1604 | /* set allmulti level to new slave */ | |
1605 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1606 | res = dev_set_allmulti(slave_dev, 1); |
1607 | if (res) | |
1608 | goto err_close; | |
1da177e4 LT |
1609 | } |
1610 | ||
b9e40857 | 1611 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1612 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1613 | netdev_for_each_mc_addr(ha, bond_dev) |
1614 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1615 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1616 | } |
1617 | ||
1618 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1619 | /* add lacpdu mc addr to mc list */ | |
1620 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1621 | ||
22bedad3 | 1622 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1623 | } |
1624 | ||
1625 | bond_add_vlans_on_slave(bond, slave_dev); | |
1626 | ||
1627 | write_lock_bh(&bond->lock); | |
1628 | ||
1629 | bond_attach_slave(bond, new_slave); | |
1630 | ||
1631 | new_slave->delay = 0; | |
1632 | new_slave->link_failure_count = 0; | |
1633 | ||
8531c5ff AK |
1634 | bond_compute_features(bond); |
1635 | ||
3915c1e8 JV |
1636 | write_unlock_bh(&bond->lock); |
1637 | ||
1638 | read_lock(&bond->lock); | |
1639 | ||
f5b2b966 JV |
1640 | new_slave->last_arp_rx = jiffies; |
1641 | ||
1da177e4 LT |
1642 | if (bond->params.miimon && !bond->params.use_carrier) { |
1643 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1644 | ||
1645 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1646 | /* | |
1647 | * miimon is set but a bonded network driver | |
1648 | * does not support ETHTOOL/MII and | |
1649 | * arp_interval is not set. Note: if | |
1650 | * use_carrier is enabled, we will never go | |
1651 | * here (because netif_carrier is always | |
1652 | * supported); thus, we don't need to change | |
1653 | * the messages for netif_carrier. | |
1654 | */ | |
a4aee5c8 | 1655 | 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 | 1656 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1657 | } else if (link_reporting == -1) { |
1658 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1659 | 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", |
1660 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1661 | } |
1662 | } | |
1663 | ||
1664 | /* check for initial state */ | |
1665 | if (!bond->params.miimon || | |
1666 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1667 | if (bond->params.updelay) { | |
a4aee5c8 | 1668 | pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n"); |
1da177e4 LT |
1669 | new_slave->link = BOND_LINK_BACK; |
1670 | new_slave->delay = bond->params.updelay; | |
1671 | } else { | |
a4aee5c8 | 1672 | pr_debug("Initial state of slave_dev is BOND_LINK_UP\n"); |
1da177e4 LT |
1673 | new_slave->link = BOND_LINK_UP; |
1674 | } | |
1675 | new_slave->jiffies = jiffies; | |
1676 | } else { | |
a4aee5c8 | 1677 | pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n"); |
1da177e4 LT |
1678 | new_slave->link = BOND_LINK_DOWN; |
1679 | } | |
1680 | ||
1681 | if (bond_update_speed_duplex(new_slave) && | |
1682 | (new_slave->link != BOND_LINK_DOWN)) { | |
a4aee5c8 JP |
1683 | pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n", |
1684 | bond_dev->name, new_slave->dev->name); | |
1da177e4 LT |
1685 | |
1686 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a4aee5c8 JP |
1687 | pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n", |
1688 | bond_dev->name); | |
1da177e4 LT |
1689 | } |
1690 | } | |
1691 | ||
1692 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1693 | /* if there is a primary slave, remember it */ | |
a549952a | 1694 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1695 | bond->primary_slave = new_slave; |
a549952a JP |
1696 | bond->force_primary = true; |
1697 | } | |
1da177e4 LT |
1698 | } |
1699 | ||
3915c1e8 JV |
1700 | write_lock_bh(&bond->curr_slave_lock); |
1701 | ||
1da177e4 LT |
1702 | switch (bond->params.mode) { |
1703 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1704 | bond_set_slave_inactive_flags(new_slave); |
1705 | bond_select_active_slave(bond); | |
1da177e4 LT |
1706 | break; |
1707 | case BOND_MODE_8023AD: | |
1708 | /* in 802.3ad mode, the internal mechanism | |
1709 | * will activate the slaves in the selected | |
1710 | * aggregator | |
1711 | */ | |
1712 | bond_set_slave_inactive_flags(new_slave); | |
1713 | /* if this is the first slave */ | |
1714 | if (bond->slave_cnt == 1) { | |
1715 | SLAVE_AD_INFO(new_slave).id = 1; | |
1716 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1717 | * can be called only after the mac address of the bond is set | |
1718 | */ | |
1719 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1720 | bond->params.lacp_fast); | |
1721 | } else { | |
1722 | SLAVE_AD_INFO(new_slave).id = | |
1723 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1724 | } | |
1725 | ||
1726 | bond_3ad_bind_slave(new_slave); | |
1727 | break; | |
1728 | case BOND_MODE_TLB: | |
1729 | case BOND_MODE_ALB: | |
1730 | new_slave->state = BOND_STATE_ACTIVE; | |
059fe7a5 | 1731 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1732 | bond_select_active_slave(bond); |
1da177e4 LT |
1733 | break; |
1734 | default: | |
5a03cdb7 | 1735 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1736 | |
1737 | /* always active in trunk mode */ | |
1738 | new_slave->state = BOND_STATE_ACTIVE; | |
1739 | ||
1740 | /* In trunking mode there is little meaning to curr_active_slave | |
1741 | * anyway (it holds no special properties of the bond device), | |
1742 | * so we can change it without calling change_active_interface() | |
1743 | */ | |
3d632c3f | 1744 | if (!bond->curr_active_slave) |
1da177e4 | 1745 | bond->curr_active_slave = new_slave; |
3d632c3f | 1746 | |
1da177e4 LT |
1747 | break; |
1748 | } /* switch(bond_mode) */ | |
1749 | ||
3915c1e8 JV |
1750 | write_unlock_bh(&bond->curr_slave_lock); |
1751 | ||
ff59c456 JV |
1752 | bond_set_carrier(bond); |
1753 | ||
f6dc31a8 WC |
1754 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1755 | if (slaves_support_netpoll(bond_dev)) { | |
1756 | bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL; | |
1757 | if (bond_dev->npinfo) | |
1758 | slave_dev->npinfo = bond_dev->npinfo; | |
1759 | } else if (!(bond_dev->priv_flags & IFF_DISABLE_NETPOLL)) { | |
1760 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; | |
1761 | pr_info("New slave device %s does not support netpoll\n", | |
1762 | slave_dev->name); | |
1763 | pr_info("Disabling netpoll support for %s\n", bond_dev->name); | |
1764 | } | |
1765 | #endif | |
3915c1e8 | 1766 | read_unlock(&bond->lock); |
1da177e4 | 1767 | |
b76cdba9 MW |
1768 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1769 | if (res) | |
569f0c4d | 1770 | goto err_close; |
b76cdba9 | 1771 | |
a4aee5c8 JP |
1772 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1773 | bond_dev->name, slave_dev->name, | |
1774 | new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup", | |
1775 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); | |
1da177e4 LT |
1776 | |
1777 | /* enslave is successful */ | |
1778 | return 0; | |
1779 | ||
1780 | /* Undo stages on error */ | |
1da177e4 LT |
1781 | err_close: |
1782 | dev_close(slave_dev); | |
1783 | ||
569f0c4d JV |
1784 | err_unset_master: |
1785 | netdev_set_master(slave_dev, NULL); | |
1786 | ||
1da177e4 | 1787 | err_restore_mac: |
dd957c57 | 1788 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1789 | /* XXX TODO - fom follow mode needs to change master's |
1790 | * MAC if this slave's MAC is in use by the bond, or at | |
1791 | * least print a warning. | |
1792 | */ | |
2ab82852 MS |
1793 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1794 | addr.sa_family = slave_dev->type; | |
1795 | dev_set_mac_address(slave_dev, &addr); | |
1796 | } | |
1da177e4 | 1797 | |
b15ba0fb JP |
1798 | err_restore_mtu: |
1799 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1800 | ||
1da177e4 LT |
1801 | err_free: |
1802 | kfree(new_slave); | |
1803 | ||
1804 | err_undo_flags: | |
1805 | bond_dev->features = old_features; | |
3d632c3f | 1806 | |
1da177e4 LT |
1807 | return res; |
1808 | } | |
1809 | ||
1810 | /* | |
1811 | * Try to release the slave device <slave> from the bond device <master> | |
1812 | * It is legal to access curr_active_slave without a lock because all the function | |
1813 | * is write-locked. | |
1814 | * | |
1815 | * The rules for slave state should be: | |
1816 | * for Active/Backup: | |
1817 | * Active stays on all backups go down | |
1818 | * for Bonded connections: | |
1819 | * The first up interface should be left on and all others downed. | |
1820 | */ | |
a77b5325 | 1821 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1822 | { |
454d7c9b | 1823 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1824 | struct slave *slave, *oldcurrent; |
1825 | struct sockaddr addr; | |
1da177e4 LT |
1826 | |
1827 | /* slave is not a slave or master is not master of this slave */ | |
1828 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1829 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1830 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1831 | bond_dev->name, slave_dev->name); |
1832 | return -EINVAL; | |
1833 | } | |
1834 | ||
f6dc31a8 | 1835 | netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE); |
1da177e4 LT |
1836 | write_lock_bh(&bond->lock); |
1837 | ||
1838 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1839 | if (!slave) { | |
1840 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1841 | pr_info("%s: %s not enslaved\n", |
1842 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1843 | write_unlock_bh(&bond->lock); |
1da177e4 LT |
1844 | return -EINVAL; |
1845 | } | |
1846 | ||
3915c1e8 | 1847 | if (!bond->params.fail_over_mac) { |
8e95a202 JP |
1848 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) && |
1849 | bond->slave_cnt > 1) | |
a4aee5c8 JP |
1850 | 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", |
1851 | bond_dev->name, slave_dev->name, | |
1852 | slave->perm_hwaddr, | |
1853 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1854 | } |
1855 | ||
1856 | /* Inform AD package of unbinding of slave. */ | |
1857 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1858 | /* must be called before the slave is | |
1859 | * detached from the list | |
1860 | */ | |
1861 | bond_3ad_unbind_slave(slave); | |
1862 | } | |
1863 | ||
a4aee5c8 JP |
1864 | pr_info("%s: releasing %s interface %s\n", |
1865 | bond_dev->name, | |
1866 | (slave->state == BOND_STATE_ACTIVE) ? "active" : "backup", | |
1867 | slave_dev->name); | |
1da177e4 LT |
1868 | |
1869 | oldcurrent = bond->curr_active_slave; | |
1870 | ||
1871 | bond->current_arp_slave = NULL; | |
1872 | ||
1873 | /* release the slave from its bond */ | |
1874 | bond_detach_slave(bond, slave); | |
1875 | ||
8531c5ff AK |
1876 | bond_compute_features(bond); |
1877 | ||
3d632c3f | 1878 | if (bond->primary_slave == slave) |
1da177e4 | 1879 | bond->primary_slave = NULL; |
1da177e4 | 1880 | |
3d632c3f | 1881 | if (oldcurrent == slave) |
1da177e4 | 1882 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1883 | |
58402054 | 1884 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1885 | /* Must be called only after the slave has been |
1886 | * detached from the list and the curr_active_slave | |
1887 | * has been cleared (if our_slave == old_current), | |
1888 | * but before a new active slave is selected. | |
1889 | */ | |
2543331d | 1890 | write_unlock_bh(&bond->lock); |
1da177e4 | 1891 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1892 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1893 | } |
1894 | ||
059fe7a5 JV |
1895 | if (oldcurrent == slave) { |
1896 | /* | |
1897 | * Note that we hold RTNL over this sequence, so there | |
1898 | * is no concern that another slave add/remove event | |
1899 | * will interfere. | |
1900 | */ | |
1901 | write_unlock_bh(&bond->lock); | |
1902 | read_lock(&bond->lock); | |
1903 | write_lock_bh(&bond->curr_slave_lock); | |
1904 | ||
1da177e4 LT |
1905 | bond_select_active_slave(bond); |
1906 | ||
059fe7a5 JV |
1907 | write_unlock_bh(&bond->curr_slave_lock); |
1908 | read_unlock(&bond->lock); | |
1909 | write_lock_bh(&bond->lock); | |
1910 | } | |
1911 | ||
1da177e4 | 1912 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
1913 | bond_set_carrier(bond); |
1914 | ||
1da177e4 LT |
1915 | /* if the last slave was removed, zero the mac address |
1916 | * of the master so it will be set by the application | |
1917 | * to the mac address of the first slave | |
1918 | */ | |
1919 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
1920 | ||
1921 | if (list_empty(&bond->vlan_list)) { | |
1922 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
1923 | } else { | |
a4aee5c8 JP |
1924 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
1925 | bond_dev->name, bond_dev->name); | |
1926 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
1927 | bond_dev->name); | |
1da177e4 LT |
1928 | } |
1929 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1930 | !bond_has_challenged_slaves(bond)) { | |
a4aee5c8 JP |
1931 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", |
1932 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1933 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
1934 | } | |
1935 | ||
1936 | write_unlock_bh(&bond->lock); | |
1937 | ||
b76cdba9 MW |
1938 | /* must do this from outside any spinlocks */ |
1939 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1940 | ||
1da177e4 LT |
1941 | bond_del_vlans_from_slave(bond, slave_dev); |
1942 | ||
1943 | /* If the mode USES_PRIMARY, then we should only remove its | |
1944 | * promisc and mc settings if it was the curr_active_slave, but that was | |
1945 | * already taken care of above when we detached the slave | |
1946 | */ | |
1947 | if (!USES_PRIMARY(bond->params.mode)) { | |
1948 | /* unset promiscuity level from slave */ | |
3d632c3f | 1949 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 1950 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1951 | |
1952 | /* unset allmulti level from slave */ | |
3d632c3f | 1953 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 1954 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
1955 | |
1956 | /* flush master's mc_list from slave */ | |
b9e40857 | 1957 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1958 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 1959 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1960 | } |
1961 | ||
1962 | netdev_set_master(slave_dev, NULL); | |
1963 | ||
f6dc31a8 WC |
1964 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1965 | read_lock_bh(&bond->lock); | |
1966 | if (slaves_support_netpoll(bond_dev)) | |
1967 | bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL; | |
1968 | read_unlock_bh(&bond->lock); | |
1969 | if (slave_dev->netdev_ops->ndo_netpoll_cleanup) | |
1970 | slave_dev->netdev_ops->ndo_netpoll_cleanup(slave_dev); | |
1971 | else | |
1972 | slave_dev->npinfo = NULL; | |
1973 | #endif | |
1974 | ||
1da177e4 LT |
1975 | /* close slave before restoring its mac address */ |
1976 | dev_close(slave_dev); | |
1977 | ||
3915c1e8 | 1978 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1979 | /* restore original ("permanent") mac address */ |
1980 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1981 | addr.sa_family = slave_dev->type; | |
1982 | dev_set_mac_address(slave_dev, &addr); | |
1983 | } | |
1da177e4 | 1984 | |
b15ba0fb JP |
1985 | dev_set_mtu(slave_dev, slave->original_mtu); |
1986 | ||
8f903c70 | 1987 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
f5b2b966 JV |
1988 | IFF_SLAVE_INACTIVE | IFF_BONDING | |
1989 | IFF_SLAVE_NEEDARP); | |
1da177e4 LT |
1990 | |
1991 | kfree(slave); | |
1992 | ||
1993 | return 0; /* deletion OK */ | |
1994 | } | |
1995 | ||
d90a162a | 1996 | /* |
3d632c3f | 1997 | * First release a slave and than destroy the bond if no more slaves are left. |
d90a162a MS |
1998 | * Must be under rtnl_lock when this function is called. |
1999 | */ | |
3d632c3f SH |
2000 | int bond_release_and_destroy(struct net_device *bond_dev, |
2001 | struct net_device *slave_dev) | |
d90a162a | 2002 | { |
454d7c9b | 2003 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2004 | int ret; |
2005 | ||
2006 | ret = bond_release(bond_dev, slave_dev); | |
2007 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
a4aee5c8 JP |
2008 | pr_info("%s: destroying bond %s.\n", |
2009 | bond_dev->name, bond_dev->name); | |
9e71626c | 2010 | unregister_netdevice(bond_dev); |
d90a162a MS |
2011 | } |
2012 | return ret; | |
2013 | } | |
2014 | ||
1da177e4 LT |
2015 | /* |
2016 | * This function releases all slaves. | |
2017 | */ | |
2018 | static int bond_release_all(struct net_device *bond_dev) | |
2019 | { | |
454d7c9b | 2020 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2021 | struct slave *slave; |
2022 | struct net_device *slave_dev; | |
2023 | struct sockaddr addr; | |
2024 | ||
2025 | write_lock_bh(&bond->lock); | |
2026 | ||
ff59c456 JV |
2027 | netif_carrier_off(bond_dev); |
2028 | ||
3d632c3f | 2029 | if (bond->slave_cnt == 0) |
1da177e4 | 2030 | goto out; |
1da177e4 LT |
2031 | |
2032 | bond->current_arp_slave = NULL; | |
2033 | bond->primary_slave = NULL; | |
2034 | bond_change_active_slave(bond, NULL); | |
2035 | ||
2036 | while ((slave = bond->first_slave) != NULL) { | |
2037 | /* Inform AD package of unbinding of slave | |
2038 | * before slave is detached from the list. | |
2039 | */ | |
3d632c3f | 2040 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2041 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2042 | |
2043 | slave_dev = slave->dev; | |
2044 | bond_detach_slave(bond, slave); | |
2045 | ||
2543331d JV |
2046 | /* now that the slave is detached, unlock and perform |
2047 | * all the undo steps that should not be called from | |
2048 | * within a lock. | |
2049 | */ | |
2050 | write_unlock_bh(&bond->lock); | |
2051 | ||
58402054 | 2052 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2053 | /* must be called only after the slave |
2054 | * has been detached from the list | |
2055 | */ | |
2056 | bond_alb_deinit_slave(bond, slave); | |
2057 | } | |
2058 | ||
8531c5ff AK |
2059 | bond_compute_features(bond); |
2060 | ||
b76cdba9 | 2061 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2062 | bond_del_vlans_from_slave(bond, slave_dev); |
2063 | ||
2064 | /* If the mode USES_PRIMARY, then we should only remove its | |
2065 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2066 | * already taken care of above when we detached the slave | |
2067 | */ | |
2068 | if (!USES_PRIMARY(bond->params.mode)) { | |
2069 | /* unset promiscuity level from slave */ | |
3d632c3f | 2070 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2071 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2072 | |
2073 | /* unset allmulti level from slave */ | |
3d632c3f | 2074 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2075 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2076 | |
2077 | /* flush master's mc_list from slave */ | |
b9e40857 | 2078 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2079 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2080 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2081 | } |
2082 | ||
2083 | netdev_set_master(slave_dev, NULL); | |
2084 | ||
2085 | /* close slave before restoring its mac address */ | |
2086 | dev_close(slave_dev); | |
2087 | ||
dd957c57 | 2088 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2089 | /* restore original ("permanent") mac address*/ |
2090 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2091 | addr.sa_family = slave_dev->type; | |
2092 | dev_set_mac_address(slave_dev, &addr); | |
2093 | } | |
1da177e4 | 2094 | |
8f903c70 JV |
2095 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
2096 | IFF_SLAVE_INACTIVE); | |
1da177e4 LT |
2097 | |
2098 | kfree(slave); | |
2099 | ||
2100 | /* re-acquire the lock before getting the next slave */ | |
2101 | write_lock_bh(&bond->lock); | |
2102 | } | |
2103 | ||
2104 | /* zero the mac address of the master so it will be | |
2105 | * set by the application to the mac address of the | |
2106 | * first slave | |
2107 | */ | |
2108 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2109 | ||
3d632c3f | 2110 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 2111 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
3d632c3f | 2112 | else { |
a4aee5c8 JP |
2113 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2114 | bond_dev->name, bond_dev->name); | |
2115 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2116 | bond_dev->name); | |
1da177e4 LT |
2117 | } |
2118 | ||
a4aee5c8 | 2119 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2120 | |
2121 | out: | |
2122 | write_unlock_bh(&bond->lock); | |
2123 | ||
2124 | return 0; | |
2125 | } | |
2126 | ||
2127 | /* | |
2128 | * This function changes the active slave to slave <slave_dev>. | |
2129 | * It returns -EINVAL in the following cases. | |
2130 | * - <slave_dev> is not found in the list. | |
2131 | * - There is not active slave now. | |
2132 | * - <slave_dev> is already active. | |
2133 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2134 | * - <slave_dev> is not running. | |
3d632c3f SH |
2135 | * In these cases, this function does nothing. |
2136 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2137 | */ |
2138 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2139 | { | |
454d7c9b | 2140 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2141 | struct slave *old_active = NULL; |
2142 | struct slave *new_active = NULL; | |
2143 | int res = 0; | |
2144 | ||
3d632c3f | 2145 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2146 | return -EINVAL; |
1da177e4 LT |
2147 | |
2148 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2149 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2150 | return -EINVAL; |
1da177e4 | 2151 | |
059fe7a5 | 2152 | read_lock(&bond->lock); |
1da177e4 | 2153 | |
059fe7a5 | 2154 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2155 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2156 | read_unlock(&bond->curr_slave_lock); |
2157 | ||
1da177e4 LT |
2158 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2159 | ||
2160 | /* | |
2161 | * Changing to the current active: do nothing; return success. | |
2162 | */ | |
2163 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2164 | read_unlock(&bond->lock); |
1da177e4 LT |
2165 | return 0; |
2166 | } | |
2167 | ||
2168 | if ((new_active) && | |
2169 | (old_active) && | |
2170 | (new_active->link == BOND_LINK_UP) && | |
2171 | IS_UP(new_active->dev)) { | |
059fe7a5 | 2172 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2173 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2174 | write_unlock_bh(&bond->curr_slave_lock); |
3d632c3f | 2175 | } else |
1da177e4 | 2176 | res = -EINVAL; |
1da177e4 | 2177 | |
059fe7a5 | 2178 | read_unlock(&bond->lock); |
1da177e4 LT |
2179 | |
2180 | return res; | |
2181 | } | |
2182 | ||
1da177e4 LT |
2183 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2184 | { | |
454d7c9b | 2185 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2186 | |
2187 | info->bond_mode = bond->params.mode; | |
2188 | info->miimon = bond->params.miimon; | |
2189 | ||
6603a6f2 | 2190 | read_lock(&bond->lock); |
1da177e4 | 2191 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2192 | read_unlock(&bond->lock); |
1da177e4 LT |
2193 | |
2194 | return 0; | |
2195 | } | |
2196 | ||
2197 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2198 | { | |
454d7c9b | 2199 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2200 | struct slave *slave; |
689c96cc | 2201 | int i, res = -ENODEV; |
1da177e4 | 2202 | |
6603a6f2 | 2203 | read_lock(&bond->lock); |
1da177e4 LT |
2204 | |
2205 | bond_for_each_slave(bond, slave, i) { | |
2206 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2207 | res = 0; |
2208 | strcpy(info->slave_name, slave->dev->name); | |
2209 | info->link = slave->link; | |
2210 | info->state = slave->state; | |
2211 | info->link_failure_count = slave->link_failure_count; | |
1da177e4 LT |
2212 | break; |
2213 | } | |
2214 | } | |
2215 | ||
6603a6f2 | 2216 | read_unlock(&bond->lock); |
1da177e4 | 2217 | |
689c96cc | 2218 | return res; |
1da177e4 LT |
2219 | } |
2220 | ||
2221 | /*-------------------------------- Monitoring -------------------------------*/ | |
2222 | ||
1da177e4 | 2223 | |
f0c76d61 JV |
2224 | static int bond_miimon_inspect(struct bonding *bond) |
2225 | { | |
2226 | struct slave *slave; | |
2227 | int i, link_state, commit = 0; | |
41f89100 JP |
2228 | bool ignore_updelay; |
2229 | ||
2230 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2231 | |
2232 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2233 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2234 | |
f0c76d61 | 2235 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2236 | |
2237 | switch (slave->link) { | |
f0c76d61 JV |
2238 | case BOND_LINK_UP: |
2239 | if (link_state) | |
2240 | continue; | |
1da177e4 | 2241 | |
f0c76d61 JV |
2242 | slave->link = BOND_LINK_FAIL; |
2243 | slave->delay = bond->params.downdelay; | |
2244 | if (slave->delay) { | |
a4aee5c8 JP |
2245 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2246 | bond->dev->name, | |
2247 | (bond->params.mode == | |
2248 | BOND_MODE_ACTIVEBACKUP) ? | |
2249 | ((slave->state == BOND_STATE_ACTIVE) ? | |
2250 | "active " : "backup ") : "", | |
2251 | slave->dev->name, | |
2252 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2253 | } |
f0c76d61 JV |
2254 | /*FALLTHRU*/ |
2255 | case BOND_LINK_FAIL: | |
2256 | if (link_state) { | |
2257 | /* | |
2258 | * recovered before downdelay expired | |
2259 | */ | |
2260 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2261 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2262 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2263 | bond->dev->name, | |
2264 | (bond->params.downdelay - slave->delay) * | |
2265 | bond->params.miimon, | |
2266 | slave->dev->name); | |
f0c76d61 | 2267 | continue; |
1da177e4 | 2268 | } |
f0c76d61 JV |
2269 | |
2270 | if (slave->delay <= 0) { | |
2271 | slave->new_link = BOND_LINK_DOWN; | |
2272 | commit++; | |
2273 | continue; | |
1da177e4 | 2274 | } |
1da177e4 | 2275 | |
f0c76d61 JV |
2276 | slave->delay--; |
2277 | break; | |
2278 | ||
2279 | case BOND_LINK_DOWN: | |
2280 | if (!link_state) | |
2281 | continue; | |
2282 | ||
2283 | slave->link = BOND_LINK_BACK; | |
2284 | slave->delay = bond->params.updelay; | |
2285 | ||
2286 | if (slave->delay) { | |
a4aee5c8 JP |
2287 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2288 | bond->dev->name, slave->dev->name, | |
2289 | ignore_updelay ? 0 : | |
2290 | bond->params.updelay * | |
2291 | bond->params.miimon); | |
f0c76d61 JV |
2292 | } |
2293 | /*FALLTHRU*/ | |
2294 | case BOND_LINK_BACK: | |
2295 | if (!link_state) { | |
2296 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2297 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2298 | bond->dev->name, | |
2299 | (bond->params.updelay - slave->delay) * | |
2300 | bond->params.miimon, | |
2301 | slave->dev->name); | |
f0c76d61 JV |
2302 | |
2303 | continue; | |
2304 | } | |
2305 | ||
41f89100 JP |
2306 | if (ignore_updelay) |
2307 | slave->delay = 0; | |
2308 | ||
f0c76d61 JV |
2309 | if (slave->delay <= 0) { |
2310 | slave->new_link = BOND_LINK_UP; | |
2311 | commit++; | |
41f89100 | 2312 | ignore_updelay = false; |
f0c76d61 | 2313 | continue; |
1da177e4 | 2314 | } |
f0c76d61 JV |
2315 | |
2316 | slave->delay--; | |
1da177e4 | 2317 | break; |
f0c76d61 JV |
2318 | } |
2319 | } | |
1da177e4 | 2320 | |
f0c76d61 JV |
2321 | return commit; |
2322 | } | |
1da177e4 | 2323 | |
f0c76d61 JV |
2324 | static void bond_miimon_commit(struct bonding *bond) |
2325 | { | |
2326 | struct slave *slave; | |
2327 | int i; | |
2328 | ||
2329 | bond_for_each_slave(bond, slave, i) { | |
2330 | switch (slave->new_link) { | |
2331 | case BOND_LINK_NOCHANGE: | |
2332 | continue; | |
1da177e4 | 2333 | |
f0c76d61 JV |
2334 | case BOND_LINK_UP: |
2335 | slave->link = BOND_LINK_UP; | |
2336 | slave->jiffies = jiffies; | |
2337 | ||
2338 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2339 | /* prevent it from being the active one */ | |
2340 | slave->state = BOND_STATE_BACKUP; | |
2341 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
2342 | /* make it immediately active */ | |
2343 | slave->state = BOND_STATE_ACTIVE; | |
2344 | } else if (slave != bond->primary_slave) { | |
2345 | /* prevent it from being the active one */ | |
2346 | slave->state = BOND_STATE_BACKUP; | |
1da177e4 | 2347 | } |
1da177e4 | 2348 | |
a4aee5c8 JP |
2349 | pr_info("%s: link status definitely up for interface %s.\n", |
2350 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2351 | |
f0c76d61 JV |
2352 | /* notify ad that the link status has changed */ |
2353 | if (bond->params.mode == BOND_MODE_8023AD) | |
2354 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2355 | |
58402054 | 2356 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2357 | bond_alb_handle_link_change(bond, slave, |
2358 | BOND_LINK_UP); | |
1da177e4 | 2359 | |
f0c76d61 JV |
2360 | if (!bond->curr_active_slave || |
2361 | (slave == bond->primary_slave)) | |
2362 | goto do_failover; | |
1da177e4 | 2363 | |
f0c76d61 | 2364 | continue; |
059fe7a5 | 2365 | |
f0c76d61 | 2366 | case BOND_LINK_DOWN: |
fba4acda JV |
2367 | if (slave->link_failure_count < UINT_MAX) |
2368 | slave->link_failure_count++; | |
2369 | ||
f0c76d61 | 2370 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2371 | |
f0c76d61 JV |
2372 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2373 | bond->params.mode == BOND_MODE_8023AD) | |
2374 | bond_set_slave_inactive_flags(slave); | |
2375 | ||
a4aee5c8 JP |
2376 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2377 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2378 | |
2379 | if (bond->params.mode == BOND_MODE_8023AD) | |
2380 | bond_3ad_handle_link_change(slave, | |
2381 | BOND_LINK_DOWN); | |
2382 | ||
ae63e808 | 2383 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2384 | bond_alb_handle_link_change(bond, slave, |
2385 | BOND_LINK_DOWN); | |
2386 | ||
2387 | if (slave == bond->curr_active_slave) | |
2388 | goto do_failover; | |
2389 | ||
2390 | continue; | |
2391 | ||
2392 | default: | |
a4aee5c8 | 2393 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2394 | bond->dev->name, slave->new_link, |
2395 | slave->dev->name); | |
2396 | slave->new_link = BOND_LINK_NOCHANGE; | |
2397 | ||
2398 | continue; | |
2399 | } | |
2400 | ||
2401 | do_failover: | |
2402 | ASSERT_RTNL(); | |
2403 | write_lock_bh(&bond->curr_slave_lock); | |
2404 | bond_select_active_slave(bond); | |
2405 | write_unlock_bh(&bond->curr_slave_lock); | |
2406 | } | |
2407 | ||
2408 | bond_set_carrier(bond); | |
1da177e4 LT |
2409 | } |
2410 | ||
0b0eef66 JV |
2411 | /* |
2412 | * bond_mii_monitor | |
2413 | * | |
2414 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2415 | * inspection, then (if inspection indicates something needs to be done) |
2416 | * an acquisition of appropriate locks followed by a commit phase to | |
2417 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2418 | */ |
2419 | void bond_mii_monitor(struct work_struct *work) | |
2420 | { | |
2421 | struct bonding *bond = container_of(work, struct bonding, | |
2422 | mii_work.work); | |
0b0eef66 JV |
2423 | |
2424 | read_lock(&bond->lock); | |
f0c76d61 JV |
2425 | if (bond->kill_timers) |
2426 | goto out; | |
2427 | ||
2428 | if (bond->slave_cnt == 0) | |
2429 | goto re_arm; | |
b59f9f74 JV |
2430 | |
2431 | if (bond->send_grat_arp) { | |
2432 | read_lock(&bond->curr_slave_lock); | |
2433 | bond_send_gratuitous_arp(bond); | |
2434 | read_unlock(&bond->curr_slave_lock); | |
2435 | } | |
2436 | ||
305d552a BH |
2437 | if (bond->send_unsol_na) { |
2438 | read_lock(&bond->curr_slave_lock); | |
2439 | bond_send_unsolicited_na(bond); | |
2440 | read_unlock(&bond->curr_slave_lock); | |
2441 | } | |
2442 | ||
f0c76d61 | 2443 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2444 | read_unlock(&bond->lock); |
2445 | rtnl_lock(); | |
2446 | read_lock(&bond->lock); | |
f0c76d61 JV |
2447 | |
2448 | bond_miimon_commit(bond); | |
2449 | ||
5655662d JV |
2450 | read_unlock(&bond->lock); |
2451 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2452 | read_lock(&bond->lock); | |
0b0eef66 JV |
2453 | } |
2454 | ||
f0c76d61 JV |
2455 | re_arm: |
2456 | if (bond->params.miimon) | |
2457 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2458 | msecs_to_jiffies(bond->params.miimon)); | |
2459 | out: | |
0b0eef66 | 2460 | read_unlock(&bond->lock); |
0b0eef66 | 2461 | } |
c3ade5ca | 2462 | |
d3bb52b0 | 2463 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2464 | { |
2465 | struct in_device *idev; | |
2466 | struct in_ifaddr *ifa; | |
a144ea4b | 2467 | __be32 addr = 0; |
c3ade5ca JV |
2468 | |
2469 | if (!dev) | |
2470 | return 0; | |
2471 | ||
2472 | rcu_read_lock(); | |
e5ed6399 | 2473 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2474 | if (!idev) |
2475 | goto out; | |
2476 | ||
2477 | ifa = idev->ifa_list; | |
2478 | if (!ifa) | |
2479 | goto out; | |
2480 | ||
2481 | addr = ifa->ifa_local; | |
2482 | out: | |
2483 | rcu_read_unlock(); | |
2484 | return addr; | |
2485 | } | |
2486 | ||
d3bb52b0 | 2487 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2488 | { |
0883beca | 2489 | struct vlan_entry *vlan; |
f5b2b966 JV |
2490 | |
2491 | if (ip == bond->master_ip) | |
2492 | return 1; | |
2493 | ||
0883beca | 2494 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2495 | if (ip == vlan->vlan_ip) |
2496 | return 1; | |
2497 | } | |
2498 | ||
2499 | return 0; | |
2500 | } | |
2501 | ||
c3ade5ca JV |
2502 | /* |
2503 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2504 | * switches in VLAN mode (especially if ports are configured as | |
2505 | * "native" to a VLAN) might not pass non-tagged frames. | |
2506 | */ | |
d3bb52b0 | 2507 | 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 |
2508 | { |
2509 | struct sk_buff *skb; | |
2510 | ||
5a03cdb7 | 2511 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2512 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2513 | |
c3ade5ca JV |
2514 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2515 | NULL, slave_dev->dev_addr, NULL); | |
2516 | ||
2517 | if (!skb) { | |
a4aee5c8 | 2518 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2519 | return; |
2520 | } | |
2521 | if (vlan_id) { | |
2522 | skb = vlan_put_tag(skb, vlan_id); | |
2523 | if (!skb) { | |
a4aee5c8 | 2524 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2525 | return; |
2526 | } | |
2527 | } | |
2528 | arp_xmit(skb); | |
2529 | } | |
2530 | ||
2531 | ||
1da177e4 LT |
2532 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2533 | { | |
c3ade5ca | 2534 | int i, vlan_id, rv; |
d3bb52b0 | 2535 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2536 | struct vlan_entry *vlan; |
c3ade5ca JV |
2537 | struct net_device *vlan_dev; |
2538 | struct flowi fl; | |
2539 | struct rtable *rt; | |
1da177e4 | 2540 | |
6b780567 MW |
2541 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2542 | if (!targets[i]) | |
5a31bec0 | 2543 | break; |
5a03cdb7 | 2544 | pr_debug("basa: target %x\n", targets[i]); |
c3ade5ca | 2545 | if (list_empty(&bond->vlan_list)) { |
5a03cdb7 | 2546 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2547 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2548 | bond->master_ip, 0); | |
2549 | continue; | |
2550 | } | |
2551 | ||
2552 | /* | |
2553 | * If VLANs are configured, we do a route lookup to | |
2554 | * determine which VLAN interface would be used, so we | |
2555 | * can tag the ARP with the proper VLAN tag. | |
2556 | */ | |
2557 | memset(&fl, 0, sizeof(fl)); | |
2558 | fl.fl4_dst = targets[i]; | |
2559 | fl.fl4_tos = RTO_ONLINK; | |
2560 | ||
ec87fd3b | 2561 | rv = ip_route_output_key(dev_net(bond->dev), &rt, &fl); |
c3ade5ca JV |
2562 | if (rv) { |
2563 | if (net_ratelimit()) { | |
a4aee5c8 JP |
2564 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
2565 | bond->dev->name, &fl.fl4_dst); | |
c3ade5ca JV |
2566 | } |
2567 | continue; | |
2568 | } | |
2569 | ||
2570 | /* | |
2571 | * This target is not on a VLAN | |
2572 | */ | |
2573 | if (rt->u.dst.dev == bond->dev) { | |
ed4b9f80 | 2574 | ip_rt_put(rt); |
5a03cdb7 | 2575 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2576 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2577 | bond->master_ip, 0); | |
2578 | continue; | |
2579 | } | |
2580 | ||
2581 | vlan_id = 0; | |
0883beca | 2582 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2583 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2584 | if (vlan_dev == rt->u.dst.dev) { |
2585 | vlan_id = vlan->vlan_id; | |
5a03cdb7 | 2586 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2587 | vlan_dev->name, vlan_id); |
2588 | break; | |
2589 | } | |
2590 | } | |
2591 | ||
2592 | if (vlan_id) { | |
ed4b9f80 | 2593 | ip_rt_put(rt); |
c3ade5ca JV |
2594 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2595 | vlan->vlan_ip, vlan_id); | |
2596 | continue; | |
2597 | } | |
2598 | ||
2599 | if (net_ratelimit()) { | |
a4aee5c8 JP |
2600 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2601 | bond->dev->name, &fl.fl4_dst, | |
2602 | rt->u.dst.dev ? rt->u.dst.dev->name : "NULL"); | |
c3ade5ca | 2603 | } |
ed4b9f80 | 2604 | ip_rt_put(rt); |
c3ade5ca JV |
2605 | } |
2606 | } | |
2607 | ||
2608 | /* | |
2609 | * Kick out a gratuitous ARP for an IP on the bonding master plus one | |
2610 | * for each VLAN above us. | |
b59f9f74 JV |
2611 | * |
2612 | * Caller must hold curr_slave_lock for read or better | |
c3ade5ca JV |
2613 | */ |
2614 | static void bond_send_gratuitous_arp(struct bonding *bond) | |
2615 | { | |
2616 | struct slave *slave = bond->curr_active_slave; | |
2617 | struct vlan_entry *vlan; | |
2618 | struct net_device *vlan_dev; | |
2619 | ||
a4aee5c8 JP |
2620 | pr_debug("bond_send_grat_arp: bond %s slave %s\n", |
2621 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
b59f9f74 JV |
2622 | |
2623 | if (!slave || !bond->send_grat_arp || | |
2624 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
c3ade5ca JV |
2625 | return; |
2626 | ||
b59f9f74 JV |
2627 | bond->send_grat_arp--; |
2628 | ||
c3ade5ca JV |
2629 | if (bond->master_ip) { |
2630 | bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip, | |
1053f62c | 2631 | bond->master_ip, 0); |
c3ade5ca JV |
2632 | } |
2633 | ||
2634 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
5c15bdec | 2635 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2636 | if (vlan->vlan_ip) { |
2637 | bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip, | |
2638 | vlan->vlan_ip, vlan->vlan_id); | |
2639 | } | |
1da177e4 LT |
2640 | } |
2641 | } | |
2642 | ||
d3bb52b0 | 2643 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2644 | { |
2645 | int i; | |
d3bb52b0 | 2646 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2647 | |
f5b2b966 | 2648 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2649 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2650 | &sip, &tip, i, &targets[i], |
2651 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2652 | if (sip == targets[i]) { |
2653 | if (bond_has_this_ip(bond, tip)) | |
2654 | slave->last_arp_rx = jiffies; | |
2655 | return; | |
2656 | } | |
2657 | } | |
2658 | } | |
2659 | ||
2660 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2661 | { | |
2662 | struct arphdr *arp; | |
2663 | struct slave *slave; | |
2664 | struct bonding *bond; | |
2665 | unsigned char *arp_ptr; | |
d3bb52b0 | 2666 | __be32 sip, tip; |
f5b2b966 | 2667 | |
1f3c8804 AG |
2668 | if (dev->priv_flags & IFF_802_1Q_VLAN) { |
2669 | /* | |
2670 | * When using VLANS and bonding, dev and oriv_dev may be | |
2671 | * incorrect if the physical interface supports VLAN | |
2672 | * acceleration. With this change ARP validation now | |
2673 | * works for hosts only reachable on the VLAN interface. | |
2674 | */ | |
2675 | dev = vlan_dev_real_dev(dev); | |
2676 | orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif); | |
2677 | } | |
2678 | ||
f5b2b966 JV |
2679 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2680 | goto out; | |
2681 | ||
454d7c9b | 2682 | bond = netdev_priv(dev); |
f5b2b966 JV |
2683 | read_lock(&bond->lock); |
2684 | ||
5a03cdb7 | 2685 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
a4aee5c8 JP |
2686 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2687 | orig_dev ? orig_dev->name : "NULL"); | |
f5b2b966 JV |
2688 | |
2689 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2690 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2691 | goto out_unlock; | |
2692 | ||
988b7050 | 2693 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2694 | goto out_unlock; |
2695 | ||
d0a92be0 | 2696 | arp = arp_hdr(skb); |
f5b2b966 JV |
2697 | if (arp->ar_hln != dev->addr_len || |
2698 | skb->pkt_type == PACKET_OTHERHOST || | |
2699 | skb->pkt_type == PACKET_LOOPBACK || | |
2700 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2701 | arp->ar_pro != htons(ETH_P_IP) || | |
2702 | arp->ar_pln != 4) | |
2703 | goto out_unlock; | |
2704 | ||
2705 | arp_ptr = (unsigned char *)(arp + 1); | |
2706 | arp_ptr += dev->addr_len; | |
2707 | memcpy(&sip, arp_ptr, 4); | |
2708 | arp_ptr += 4 + dev->addr_len; | |
2709 | memcpy(&tip, arp_ptr, 4); | |
2710 | ||
5a03cdb7 | 2711 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
a4aee5c8 JP |
2712 | bond->dev->name, slave->dev->name, slave->state, |
2713 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2714 | &sip, &tip); | |
f5b2b966 JV |
2715 | |
2716 | /* | |
2717 | * Backup slaves won't see the ARP reply, but do come through | |
2718 | * here for each ARP probe (so we swap the sip/tip to validate | |
2719 | * the probe). In a "redundant switch, common router" type of | |
2720 | * configuration, the ARP probe will (hopefully) travel from | |
2721 | * the active, through one switch, the router, then the other | |
2722 | * switch before reaching the backup. | |
2723 | */ | |
2724 | if (slave->state == BOND_STATE_ACTIVE) | |
2725 | bond_validate_arp(bond, slave, sip, tip); | |
2726 | else | |
2727 | bond_validate_arp(bond, slave, tip, sip); | |
2728 | ||
2729 | out_unlock: | |
2730 | read_unlock(&bond->lock); | |
2731 | out: | |
2732 | dev_kfree_skb(skb); | |
2733 | return NET_RX_SUCCESS; | |
2734 | } | |
2735 | ||
1da177e4 LT |
2736 | /* |
2737 | * this function is called regularly to monitor each slave's link | |
2738 | * ensuring that traffic is being sent and received when arp monitoring | |
2739 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2740 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2741 | * arp monitoring in active backup mode. | |
2742 | */ | |
1b76b316 | 2743 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2744 | { |
1b76b316 JV |
2745 | struct bonding *bond = container_of(work, struct bonding, |
2746 | arp_work.work); | |
1da177e4 LT |
2747 | struct slave *slave, *oldcurrent; |
2748 | int do_failover = 0; | |
2749 | int delta_in_ticks; | |
2750 | int i; | |
2751 | ||
2752 | read_lock(&bond->lock); | |
2753 | ||
5ce0da8f | 2754 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2755 | |
3d632c3f | 2756 | if (bond->kill_timers) |
1da177e4 | 2757 | goto out; |
1da177e4 | 2758 | |
3d632c3f | 2759 | if (bond->slave_cnt == 0) |
1da177e4 | 2760 | goto re_arm; |
1da177e4 LT |
2761 | |
2762 | read_lock(&bond->curr_slave_lock); | |
2763 | oldcurrent = bond->curr_active_slave; | |
2764 | read_unlock(&bond->curr_slave_lock); | |
2765 | ||
2766 | /* see if any of the previous devices are up now (i.e. they have | |
2767 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2768 | * the picture unless it is null. also, slave->jiffies is not needed | |
2769 | * here because we send an arp on each slave and give a slave as | |
2770 | * long as it needs to get the tx/rx within the delta. | |
2771 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2772 | * so it can wait | |
2773 | */ | |
2774 | bond_for_each_slave(bond, slave, i) { | |
2775 | if (slave->link != BOND_LINK_UP) { | |
9d21493b | 2776 | if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) && |
b63bb739 | 2777 | time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) { |
1da177e4 LT |
2778 | |
2779 | slave->link = BOND_LINK_UP; | |
2780 | slave->state = BOND_STATE_ACTIVE; | |
2781 | ||
2782 | /* primary_slave has no meaning in round-robin | |
2783 | * mode. the window of a slave being up and | |
2784 | * curr_active_slave being null after enslaving | |
2785 | * is closed. | |
2786 | */ | |
2787 | if (!oldcurrent) { | |
a4aee5c8 JP |
2788 | pr_info("%s: link status definitely up for interface %s, ", |
2789 | bond->dev->name, | |
2790 | slave->dev->name); | |
1da177e4 LT |
2791 | do_failover = 1; |
2792 | } else { | |
a4aee5c8 JP |
2793 | pr_info("%s: interface %s is now up\n", |
2794 | bond->dev->name, | |
2795 | slave->dev->name); | |
1da177e4 LT |
2796 | } |
2797 | } | |
2798 | } else { | |
2799 | /* slave->link == BOND_LINK_UP */ | |
2800 | ||
2801 | /* not all switches will respond to an arp request | |
2802 | * when the source ip is 0, so don't take the link down | |
2803 | * if we don't know our ip yet | |
2804 | */ | |
9d21493b | 2805 | if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) || |
4b8a9239 | 2806 | (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) { |
1da177e4 LT |
2807 | |
2808 | slave->link = BOND_LINK_DOWN; | |
2809 | slave->state = BOND_STATE_BACKUP; | |
2810 | ||
3d632c3f | 2811 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2812 | slave->link_failure_count++; |
1da177e4 | 2813 | |
a4aee5c8 JP |
2814 | pr_info("%s: interface %s is now down.\n", |
2815 | bond->dev->name, | |
2816 | slave->dev->name); | |
1da177e4 | 2817 | |
3d632c3f | 2818 | if (slave == oldcurrent) |
1da177e4 | 2819 | do_failover = 1; |
1da177e4 LT |
2820 | } |
2821 | } | |
2822 | ||
2823 | /* note: if switch is in round-robin mode, all links | |
2824 | * must tx arp to ensure all links rx an arp - otherwise | |
2825 | * links may oscillate or not come up at all; if switch is | |
2826 | * in something like xor mode, there is nothing we can | |
2827 | * do - all replies will be rx'ed on same link causing slaves | |
2828 | * to be unstable during low/no traffic periods | |
2829 | */ | |
3d632c3f | 2830 | if (IS_UP(slave->dev)) |
1da177e4 | 2831 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2832 | } |
2833 | ||
2834 | if (do_failover) { | |
059fe7a5 | 2835 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2836 | |
2837 | bond_select_active_slave(bond); | |
2838 | ||
059fe7a5 | 2839 | write_unlock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2840 | } |
2841 | ||
2842 | re_arm: | |
1b76b316 JV |
2843 | if (bond->params.arp_interval) |
2844 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
2845 | out: |
2846 | read_unlock(&bond->lock); | |
2847 | } | |
2848 | ||
2849 | /* | |
b2220cad JV |
2850 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2851 | * changes. Sets new_link in slaves to specify what action should take | |
2852 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2853 | * to link states must be committed. | |
2854 | * | |
2855 | * Called with bond->lock held for read. | |
1da177e4 | 2856 | */ |
b2220cad | 2857 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2858 | { |
1da177e4 | 2859 | struct slave *slave; |
b2220cad | 2860 | int i, commit = 0; |
1da177e4 | 2861 | |
b2220cad JV |
2862 | bond_for_each_slave(bond, slave, i) { |
2863 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2864 | |
b2220cad JV |
2865 | if (slave->link != BOND_LINK_UP) { |
2866 | if (time_before_eq(jiffies, slave_last_rx(bond, slave) + | |
2867 | delta_in_ticks)) { | |
2868 | slave->new_link = BOND_LINK_UP; | |
2869 | commit++; | |
2870 | } | |
1da177e4 | 2871 | |
b2220cad JV |
2872 | continue; |
2873 | } | |
1da177e4 | 2874 | |
b2220cad JV |
2875 | /* |
2876 | * Give slaves 2*delta after being enslaved or made | |
2877 | * active. This avoids bouncing, as the last receive | |
2878 | * times need a full ARP monitor cycle to be updated. | |
2879 | */ | |
2880 | if (!time_after_eq(jiffies, slave->jiffies + | |
2881 | 2 * delta_in_ticks)) | |
2882 | continue; | |
2883 | ||
2884 | /* | |
2885 | * Backup slave is down if: | |
2886 | * - No current_arp_slave AND | |
2887 | * - more than 3*delta since last receive AND | |
2888 | * - the bond has an IP address | |
2889 | * | |
2890 | * Note: a non-null current_arp_slave indicates | |
2891 | * the curr_active_slave went down and we are | |
2892 | * searching for a new one; under this condition | |
2893 | * we only take the curr_active_slave down - this | |
2894 | * gives each slave a chance to tx/rx traffic | |
2895 | * before being taken out | |
2896 | */ | |
2897 | if (slave->state == BOND_STATE_BACKUP && | |
2898 | !bond->current_arp_slave && | |
2899 | time_after(jiffies, slave_last_rx(bond, slave) + | |
2900 | 3 * delta_in_ticks)) { | |
2901 | slave->new_link = BOND_LINK_DOWN; | |
2902 | commit++; | |
2903 | } | |
2904 | ||
2905 | /* | |
2906 | * Active slave is down if: | |
2907 | * - more than 2*delta since transmitting OR | |
2908 | * - (more than 2*delta since receive AND | |
2909 | * the bond has an IP address) | |
2910 | */ | |
2911 | if ((slave->state == BOND_STATE_ACTIVE) && | |
9d21493b | 2912 | (time_after_eq(jiffies, dev_trans_start(slave->dev) + |
b2220cad JV |
2913 | 2 * delta_in_ticks) || |
2914 | (time_after_eq(jiffies, slave_last_rx(bond, slave) | |
2915 | + 2 * delta_in_ticks)))) { | |
2916 | slave->new_link = BOND_LINK_DOWN; | |
2917 | commit++; | |
2918 | } | |
1da177e4 LT |
2919 | } |
2920 | ||
b2220cad JV |
2921 | return commit; |
2922 | } | |
1da177e4 | 2923 | |
b2220cad JV |
2924 | /* |
2925 | * Called to commit link state changes noted by inspection step of | |
2926 | * active-backup mode ARP monitor. | |
2927 | * | |
2928 | * Called with RTNL and bond->lock for read. | |
2929 | */ | |
2930 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2931 | { | |
2932 | struct slave *slave; | |
2933 | int i; | |
1da177e4 | 2934 | |
b2220cad JV |
2935 | bond_for_each_slave(bond, slave, i) { |
2936 | switch (slave->new_link) { | |
2937 | case BOND_LINK_NOCHANGE: | |
2938 | continue; | |
ff59c456 | 2939 | |
b2220cad | 2940 | case BOND_LINK_UP: |
b9f60253 JP |
2941 | if ((!bond->curr_active_slave && |
2942 | time_before_eq(jiffies, | |
2943 | dev_trans_start(slave->dev) + | |
2944 | delta_in_ticks)) || | |
2945 | bond->curr_active_slave != slave) { | |
b2220cad | 2946 | slave->link = BOND_LINK_UP; |
b2220cad JV |
2947 | bond->current_arp_slave = NULL; |
2948 | ||
a4aee5c8 | 2949 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2950 | bond->dev->name, slave->dev->name); |
b2220cad | 2951 | |
b9f60253 JP |
2952 | if (!bond->curr_active_slave || |
2953 | (slave == bond->primary_slave)) | |
2954 | goto do_failover; | |
1da177e4 | 2955 | |
b9f60253 | 2956 | } |
1da177e4 | 2957 | |
b9f60253 | 2958 | continue; |
1da177e4 | 2959 | |
b2220cad JV |
2960 | case BOND_LINK_DOWN: |
2961 | if (slave->link_failure_count < UINT_MAX) | |
2962 | slave->link_failure_count++; | |
2963 | ||
2964 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2965 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2966 | |
a4aee5c8 | 2967 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2968 | bond->dev->name, slave->dev->name); |
b2220cad | 2969 | |
b9f60253 | 2970 | if (slave == bond->curr_active_slave) { |
b2220cad | 2971 | bond->current_arp_slave = NULL; |
b9f60253 | 2972 | goto do_failover; |
1da177e4 | 2973 | } |
b9f60253 JP |
2974 | |
2975 | continue; | |
b2220cad JV |
2976 | |
2977 | default: | |
a4aee5c8 | 2978 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2979 | bond->dev->name, slave->new_link, |
2980 | slave->dev->name); | |
b9f60253 | 2981 | continue; |
1da177e4 | 2982 | } |
1da177e4 | 2983 | |
b9f60253 JP |
2984 | do_failover: |
2985 | ASSERT_RTNL(); | |
b2220cad | 2986 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2987 | bond_select_active_slave(bond); |
b2220cad JV |
2988 | write_unlock_bh(&bond->curr_slave_lock); |
2989 | } | |
1da177e4 | 2990 | |
b2220cad JV |
2991 | bond_set_carrier(bond); |
2992 | } | |
1da177e4 | 2993 | |
b2220cad JV |
2994 | /* |
2995 | * Send ARP probes for active-backup mode ARP monitor. | |
2996 | * | |
2997 | * Called with bond->lock held for read. | |
2998 | */ | |
2999 | static void bond_ab_arp_probe(struct bonding *bond) | |
3000 | { | |
3001 | struct slave *slave; | |
3002 | int i; | |
1da177e4 | 3003 | |
b2220cad | 3004 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3005 | |
b2220cad | 3006 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3007 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3008 | bond->current_arp_slave->dev->name, | |
3009 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3010 | |
b2220cad JV |
3011 | if (bond->curr_active_slave) { |
3012 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3013 | read_unlock(&bond->curr_slave_lock); | |
3014 | return; | |
3015 | } | |
1da177e4 | 3016 | |
b2220cad | 3017 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3018 | |
b2220cad JV |
3019 | /* if we don't have a curr_active_slave, search for the next available |
3020 | * backup slave from the current_arp_slave and make it the candidate | |
3021 | * for becoming the curr_active_slave | |
3022 | */ | |
1da177e4 | 3023 | |
b2220cad JV |
3024 | if (!bond->current_arp_slave) { |
3025 | bond->current_arp_slave = bond->first_slave; | |
3026 | if (!bond->current_arp_slave) | |
3027 | return; | |
3028 | } | |
1da177e4 | 3029 | |
b2220cad | 3030 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3031 | |
b2220cad JV |
3032 | /* search for next candidate */ |
3033 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3034 | if (IS_UP(slave->dev)) { | |
3035 | slave->link = BOND_LINK_BACK; | |
3036 | bond_set_slave_active_flags(slave); | |
3037 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3038 | slave->jiffies = jiffies; |
b2220cad JV |
3039 | bond->current_arp_slave = slave; |
3040 | break; | |
1da177e4 LT |
3041 | } |
3042 | ||
b2220cad JV |
3043 | /* if the link state is up at this point, we |
3044 | * mark it down - this can happen if we have | |
3045 | * simultaneous link failures and | |
3046 | * reselect_active_interface doesn't make this | |
3047 | * one the current slave so it is still marked | |
3048 | * up when it is actually down | |
1da177e4 | 3049 | */ |
b2220cad JV |
3050 | if (slave->link == BOND_LINK_UP) { |
3051 | slave->link = BOND_LINK_DOWN; | |
3052 | if (slave->link_failure_count < UINT_MAX) | |
3053 | slave->link_failure_count++; | |
1da177e4 | 3054 | |
b2220cad JV |
3055 | bond_set_slave_inactive_flags(slave); |
3056 | ||
a4aee5c8 JP |
3057 | pr_info("%s: backup interface %s is now down.\n", |
3058 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3059 | } |
b2220cad JV |
3060 | } |
3061 | } | |
1da177e4 | 3062 | |
b2220cad JV |
3063 | void bond_activebackup_arp_mon(struct work_struct *work) |
3064 | { | |
3065 | struct bonding *bond = container_of(work, struct bonding, | |
3066 | arp_work.work); | |
3067 | int delta_in_ticks; | |
1da177e4 | 3068 | |
b2220cad | 3069 | read_lock(&bond->lock); |
1da177e4 | 3070 | |
b2220cad JV |
3071 | if (bond->kill_timers) |
3072 | goto out; | |
1da177e4 | 3073 | |
b2220cad | 3074 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3075 | |
b2220cad JV |
3076 | if (bond->slave_cnt == 0) |
3077 | goto re_arm; | |
3078 | ||
b59f9f74 JV |
3079 | if (bond->send_grat_arp) { |
3080 | read_lock(&bond->curr_slave_lock); | |
3081 | bond_send_gratuitous_arp(bond); | |
3082 | read_unlock(&bond->curr_slave_lock); | |
3083 | } | |
3084 | ||
305d552a BH |
3085 | if (bond->send_unsol_na) { |
3086 | read_lock(&bond->curr_slave_lock); | |
3087 | bond_send_unsolicited_na(bond); | |
3088 | read_unlock(&bond->curr_slave_lock); | |
3089 | } | |
3090 | ||
b2220cad JV |
3091 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3092 | read_unlock(&bond->lock); | |
3093 | rtnl_lock(); | |
3094 | read_lock(&bond->lock); | |
3095 | ||
3096 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3097 | ||
3098 | read_unlock(&bond->lock); | |
3099 | rtnl_unlock(); | |
3100 | read_lock(&bond->lock); | |
1da177e4 LT |
3101 | } |
3102 | ||
b2220cad JV |
3103 | bond_ab_arp_probe(bond); |
3104 | ||
1da177e4 | 3105 | re_arm: |
3d632c3f | 3106 | if (bond->params.arp_interval) |
1b76b316 | 3107 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3108 | out: |
3109 | read_unlock(&bond->lock); | |
3110 | } | |
3111 | ||
3112 | /*------------------------------ proc/seq_file-------------------------------*/ | |
3113 | ||
3114 | #ifdef CONFIG_PROC_FS | |
3115 | ||
1da177e4 | 3116 | static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos) |
1f78d9f9 HE |
3117 | __acquires(&dev_base_lock) |
3118 | __acquires(&bond->lock) | |
1da177e4 LT |
3119 | { |
3120 | struct bonding *bond = seq->private; | |
3121 | loff_t off = 0; | |
3122 | struct slave *slave; | |
3123 | int i; | |
3124 | ||
3125 | /* make sure the bond won't be taken away */ | |
3126 | read_lock(&dev_base_lock); | |
6603a6f2 | 3127 | read_lock(&bond->lock); |
1da177e4 | 3128 | |
3d632c3f | 3129 | if (*pos == 0) |
1da177e4 | 3130 | return SEQ_START_TOKEN; |
1da177e4 LT |
3131 | |
3132 | bond_for_each_slave(bond, slave, i) { | |
3d632c3f | 3133 | if (++off == *pos) |
1da177e4 | 3134 | return slave; |
1da177e4 LT |
3135 | } |
3136 | ||
3137 | return NULL; | |
3138 | } | |
3139 | ||
3140 | static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3141 | { | |
3142 | struct bonding *bond = seq->private; | |
3143 | struct slave *slave = v; | |
3144 | ||
3145 | ++*pos; | |
3d632c3f | 3146 | if (v == SEQ_START_TOKEN) |
1da177e4 | 3147 | return bond->first_slave; |
1da177e4 LT |
3148 | |
3149 | slave = slave->next; | |
3150 | ||
3151 | return (slave == bond->first_slave) ? NULL : slave; | |
3152 | } | |
3153 | ||
3154 | static void bond_info_seq_stop(struct seq_file *seq, void *v) | |
1f78d9f9 HE |
3155 | __releases(&bond->lock) |
3156 | __releases(&dev_base_lock) | |
1da177e4 LT |
3157 | { |
3158 | struct bonding *bond = seq->private; | |
3159 | ||
6603a6f2 | 3160 | read_unlock(&bond->lock); |
1da177e4 LT |
3161 | read_unlock(&dev_base_lock); |
3162 | } | |
3163 | ||
3164 | static void bond_info_show_master(struct seq_file *seq) | |
3165 | { | |
3166 | struct bonding *bond = seq->private; | |
3167 | struct slave *curr; | |
4756b02f | 3168 | int i; |
1da177e4 LT |
3169 | |
3170 | read_lock(&bond->curr_slave_lock); | |
3171 | curr = bond->curr_active_slave; | |
3172 | read_unlock(&bond->curr_slave_lock); | |
3173 | ||
dd957c57 | 3174 | seq_printf(seq, "Bonding Mode: %s", |
1da177e4 LT |
3175 | bond_mode_name(bond->params.mode)); |
3176 | ||
dd957c57 JV |
3177 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP && |
3178 | bond->params.fail_over_mac) | |
3915c1e8 JV |
3179 | seq_printf(seq, " (fail_over_mac %s)", |
3180 | fail_over_mac_tbl[bond->params.fail_over_mac].modename); | |
dd957c57 JV |
3181 | |
3182 | seq_printf(seq, "\n"); | |
3183 | ||
c61b75ad MW |
3184 | if (bond->params.mode == BOND_MODE_XOR || |
3185 | bond->params.mode == BOND_MODE_8023AD) { | |
3186 | seq_printf(seq, "Transmit Hash Policy: %s (%d)\n", | |
3187 | xmit_hashtype_tbl[bond->params.xmit_policy].modename, | |
3188 | bond->params.xmit_policy); | |
3189 | } | |
3190 | ||
1da177e4 | 3191 | if (USES_PRIMARY(bond->params.mode)) { |
a549952a | 3192 | seq_printf(seq, "Primary Slave: %s", |
0f418b2a MW |
3193 | (bond->primary_slave) ? |
3194 | bond->primary_slave->dev->name : "None"); | |
a549952a JP |
3195 | if (bond->primary_slave) |
3196 | seq_printf(seq, " (primary_reselect %s)", | |
3197 | pri_reselect_tbl[bond->params.primary_reselect].modename); | |
1da177e4 | 3198 | |
a549952a | 3199 | seq_printf(seq, "\nCurrently Active Slave: %s\n", |
1da177e4 LT |
3200 | (curr) ? curr->dev->name : "None"); |
3201 | } | |
3202 | ||
ff59c456 JV |
3203 | seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ? |
3204 | "up" : "down"); | |
1da177e4 LT |
3205 | seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon); |
3206 | seq_printf(seq, "Up Delay (ms): %d\n", | |
3207 | bond->params.updelay * bond->params.miimon); | |
3208 | seq_printf(seq, "Down Delay (ms): %d\n", | |
3209 | bond->params.downdelay * bond->params.miimon); | |
3210 | ||
4756b02f MW |
3211 | |
3212 | /* ARP information */ | |
3d632c3f SH |
3213 | if (bond->params.arp_interval > 0) { |
3214 | int printed = 0; | |
4756b02f MW |
3215 | seq_printf(seq, "ARP Polling Interval (ms): %d\n", |
3216 | bond->params.arp_interval); | |
3217 | ||
3218 | seq_printf(seq, "ARP IP target/s (n.n.n.n form):"); | |
3219 | ||
3d632c3f | 3220 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
4756b02f | 3221 | if (!bond->params.arp_targets[i]) |
5a31bec0 | 3222 | break; |
4756b02f MW |
3223 | if (printed) |
3224 | seq_printf(seq, ","); | |
8cf14e38 | 3225 | seq_printf(seq, " %pI4", &bond->params.arp_targets[i]); |
4756b02f MW |
3226 | printed = 1; |
3227 | } | |
3228 | seq_printf(seq, "\n"); | |
3229 | } | |
3230 | ||
1da177e4 LT |
3231 | if (bond->params.mode == BOND_MODE_8023AD) { |
3232 | struct ad_info ad_info; | |
3233 | ||
3234 | seq_puts(seq, "\n802.3ad info\n"); | |
3235 | seq_printf(seq, "LACP rate: %s\n", | |
3236 | (bond->params.lacp_fast) ? "fast" : "slow"); | |
fd989c83 JV |
3237 | seq_printf(seq, "Aggregator selection policy (ad_select): %s\n", |
3238 | ad_select_tbl[bond->params.ad_select].modename); | |
1da177e4 LT |
3239 | |
3240 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3241 | seq_printf(seq, "bond %s has no active aggregator\n", | |
3242 | bond->dev->name); | |
3243 | } else { | |
3244 | seq_printf(seq, "Active Aggregator Info:\n"); | |
3245 | ||
3246 | seq_printf(seq, "\tAggregator ID: %d\n", | |
3247 | ad_info.aggregator_id); | |
3248 | seq_printf(seq, "\tNumber of ports: %d\n", | |
3249 | ad_info.ports); | |
3250 | seq_printf(seq, "\tActor Key: %d\n", | |
3251 | ad_info.actor_key); | |
3252 | seq_printf(seq, "\tPartner Key: %d\n", | |
3253 | ad_info.partner_key); | |
e174961c JB |
3254 | seq_printf(seq, "\tPartner Mac Address: %pM\n", |
3255 | ad_info.partner_system); | |
1da177e4 LT |
3256 | } |
3257 | } | |
3258 | } | |
3259 | ||
3d632c3f SH |
3260 | static void bond_info_show_slave(struct seq_file *seq, |
3261 | const struct slave *slave) | |
1da177e4 LT |
3262 | { |
3263 | struct bonding *bond = seq->private; | |
3264 | ||
3265 | seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name); | |
3266 | seq_printf(seq, "MII Status: %s\n", | |
3267 | (slave->link == BOND_LINK_UP) ? "up" : "down"); | |
65509645 | 3268 | seq_printf(seq, "Link Failure Count: %u\n", |
1da177e4 LT |
3269 | slave->link_failure_count); |
3270 | ||
e174961c | 3271 | seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr); |
1da177e4 LT |
3272 | |
3273 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3274 | const struct aggregator *agg | |
3275 | = SLAVE_AD_INFO(slave).port.aggregator; | |
3276 | ||
3d632c3f | 3277 | if (agg) |
1da177e4 LT |
3278 | seq_printf(seq, "Aggregator ID: %d\n", |
3279 | agg->aggregator_identifier); | |
3d632c3f | 3280 | else |
1da177e4 | 3281 | seq_puts(seq, "Aggregator ID: N/A\n"); |
1da177e4 LT |
3282 | } |
3283 | } | |
3284 | ||
3285 | static int bond_info_seq_show(struct seq_file *seq, void *v) | |
3286 | { | |
3287 | if (v == SEQ_START_TOKEN) { | |
3288 | seq_printf(seq, "%s\n", version); | |
3289 | bond_info_show_master(seq); | |
3d632c3f | 3290 | } else |
1da177e4 | 3291 | bond_info_show_slave(seq, v); |
1da177e4 LT |
3292 | |
3293 | return 0; | |
3294 | } | |
3295 | ||
4101dec9 | 3296 | static const struct seq_operations bond_info_seq_ops = { |
1da177e4 LT |
3297 | .start = bond_info_seq_start, |
3298 | .next = bond_info_seq_next, | |
3299 | .stop = bond_info_seq_stop, | |
3300 | .show = bond_info_seq_show, | |
3301 | }; | |
3302 | ||
3303 | static int bond_info_open(struct inode *inode, struct file *file) | |
3304 | { | |
3305 | struct seq_file *seq; | |
3306 | struct proc_dir_entry *proc; | |
3307 | int res; | |
3308 | ||
3309 | res = seq_open(file, &bond_info_seq_ops); | |
3310 | if (!res) { | |
3311 | /* recover the pointer buried in proc_dir_entry data */ | |
3312 | seq = file->private_data; | |
3313 | proc = PDE(inode); | |
3314 | seq->private = proc->data; | |
3315 | } | |
3316 | ||
3317 | return res; | |
3318 | } | |
3319 | ||
d54b1fdb | 3320 | static const struct file_operations bond_info_fops = { |
1da177e4 LT |
3321 | .owner = THIS_MODULE, |
3322 | .open = bond_info_open, | |
3323 | .read = seq_read, | |
3324 | .llseek = seq_lseek, | |
3325 | .release = seq_release, | |
3326 | }; | |
3327 | ||
38fc0026 | 3328 | static void bond_create_proc_entry(struct bonding *bond) |
1da177e4 LT |
3329 | { |
3330 | struct net_device *bond_dev = bond->dev; | |
ec87fd3b | 3331 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
1da177e4 | 3332 | |
ec87fd3b | 3333 | if (bn->proc_dir) { |
a95609cb | 3334 | bond->proc_entry = proc_create_data(bond_dev->name, |
ec87fd3b | 3335 | S_IRUGO, bn->proc_dir, |
a95609cb | 3336 | &bond_info_fops, bond); |
3d632c3f | 3337 | if (bond->proc_entry == NULL) |
a4aee5c8 JP |
3338 | pr_warning("Warning: Cannot create /proc/net/%s/%s\n", |
3339 | DRV_NAME, bond_dev->name); | |
3d632c3f | 3340 | else |
1da177e4 | 3341 | memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ); |
1da177e4 | 3342 | } |
1da177e4 LT |
3343 | } |
3344 | ||
3345 | static void bond_remove_proc_entry(struct bonding *bond) | |
3346 | { | |
ec87fd3b EB |
3347 | struct net_device *bond_dev = bond->dev; |
3348 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); | |
3349 | ||
3350 | if (bn->proc_dir && bond->proc_entry) { | |
3351 | remove_proc_entry(bond->proc_file_name, bn->proc_dir); | |
1da177e4 LT |
3352 | memset(bond->proc_file_name, 0, IFNAMSIZ); |
3353 | bond->proc_entry = NULL; | |
3354 | } | |
3355 | } | |
3356 | ||
3357 | /* Create the bonding directory under /proc/net, if doesn't exist yet. | |
3358 | * Caller must hold rtnl_lock. | |
3359 | */ | |
2c8c1e72 | 3360 | static void __net_init bond_create_proc_dir(struct bond_net *bn) |
1da177e4 | 3361 | { |
ec87fd3b EB |
3362 | if (!bn->proc_dir) { |
3363 | bn->proc_dir = proc_mkdir(DRV_NAME, bn->net->proc_net); | |
3364 | if (!bn->proc_dir) | |
a4aee5c8 JP |
3365 | pr_warning("Warning: cannot create /proc/net/%s\n", |
3366 | DRV_NAME); | |
1da177e4 LT |
3367 | } |
3368 | } | |
3369 | ||
3370 | /* Destroy the bonding directory under /proc/net, if empty. | |
3371 | * Caller must hold rtnl_lock. | |
3372 | */ | |
2c8c1e72 | 3373 | static void __net_exit bond_destroy_proc_dir(struct bond_net *bn) |
1da177e4 | 3374 | { |
ec87fd3b EB |
3375 | if (bn->proc_dir) { |
3376 | remove_proc_entry(DRV_NAME, bn->net->proc_net); | |
3377 | bn->proc_dir = NULL; | |
1da177e4 LT |
3378 | } |
3379 | } | |
aee64faf JP |
3380 | |
3381 | #else /* !CONFIG_PROC_FS */ | |
3382 | ||
38fc0026 | 3383 | static void bond_create_proc_entry(struct bonding *bond) |
aee64faf JP |
3384 | { |
3385 | } | |
3386 | ||
3387 | static void bond_remove_proc_entry(struct bonding *bond) | |
3388 | { | |
3389 | } | |
3390 | ||
2c8c1e72 | 3391 | static inline void bond_create_proc_dir(struct bond_net *bn) |
aee64faf JP |
3392 | { |
3393 | } | |
3394 | ||
2c8c1e72 | 3395 | static inline void bond_destroy_proc_dir(struct bond_net *bn) |
aee64faf JP |
3396 | { |
3397 | } | |
3398 | ||
1da177e4 LT |
3399 | #endif /* CONFIG_PROC_FS */ |
3400 | ||
aee64faf | 3401 | |
1da177e4 LT |
3402 | /*-------------------------- netdev event handling --------------------------*/ |
3403 | ||
3404 | /* | |
3405 | * Change device name | |
3406 | */ | |
3407 | static int bond_event_changename(struct bonding *bond) | |
3408 | { | |
1da177e4 LT |
3409 | bond_remove_proc_entry(bond); |
3410 | bond_create_proc_entry(bond); | |
7e083840 | 3411 | |
1da177e4 LT |
3412 | return NOTIFY_DONE; |
3413 | } | |
3414 | ||
3d632c3f SH |
3415 | static int bond_master_netdev_event(unsigned long event, |
3416 | struct net_device *bond_dev) | |
1da177e4 | 3417 | { |
454d7c9b | 3418 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3419 | |
3420 | switch (event) { | |
3421 | case NETDEV_CHANGENAME: | |
3422 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3423 | default: |
3424 | break; | |
3425 | } | |
3426 | ||
3427 | return NOTIFY_DONE; | |
3428 | } | |
3429 | ||
3d632c3f SH |
3430 | static int bond_slave_netdev_event(unsigned long event, |
3431 | struct net_device *slave_dev) | |
1da177e4 LT |
3432 | { |
3433 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3434 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3435 | |
3436 | switch (event) { | |
3437 | case NETDEV_UNREGISTER: | |
3438 | if (bond_dev) { | |
1284cd3a JV |
3439 | if (bond->setup_by_slave) |
3440 | bond_release_and_destroy(bond_dev, slave_dev); | |
3441 | else | |
3442 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3443 | } |
3444 | break; | |
3445 | case NETDEV_CHANGE: | |
17d04500 JV |
3446 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3447 | struct slave *slave; | |
3448 | ||
3449 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3450 | if (slave) { | |
3451 | u16 old_speed = slave->speed; | |
3452 | u16 old_duplex = slave->duplex; | |
3453 | ||
3454 | bond_update_speed_duplex(slave); | |
3455 | ||
3456 | if (bond_is_lb(bond)) | |
3457 | break; | |
3458 | ||
3459 | if (old_speed != slave->speed) | |
3460 | bond_3ad_adapter_speed_changed(slave); | |
3461 | if (old_duplex != slave->duplex) | |
3462 | bond_3ad_adapter_duplex_changed(slave); | |
3463 | } | |
3464 | } | |
3465 | ||
1da177e4 LT |
3466 | break; |
3467 | case NETDEV_DOWN: | |
3468 | /* | |
3469 | * ... Or is it this? | |
3470 | */ | |
3471 | break; | |
3472 | case NETDEV_CHANGEMTU: | |
3473 | /* | |
3474 | * TODO: Should slaves be allowed to | |
3475 | * independently alter their MTU? For | |
3476 | * an active-backup bond, slaves need | |
3477 | * not be the same type of device, so | |
3478 | * MTUs may vary. For other modes, | |
3479 | * slaves arguably should have the | |
3480 | * same MTUs. To do this, we'd need to | |
3481 | * take over the slave's change_mtu | |
3482 | * function for the duration of their | |
3483 | * servitude. | |
3484 | */ | |
3485 | break; | |
3486 | case NETDEV_CHANGENAME: | |
3487 | /* | |
3488 | * TODO: handle changing the primary's name | |
3489 | */ | |
3490 | break; | |
8531c5ff AK |
3491 | case NETDEV_FEAT_CHANGE: |
3492 | bond_compute_features(bond); | |
3493 | break; | |
1da177e4 LT |
3494 | default: |
3495 | break; | |
3496 | } | |
3497 | ||
3498 | return NOTIFY_DONE; | |
3499 | } | |
3500 | ||
3501 | /* | |
3502 | * bond_netdev_event: handle netdev notifier chain events. | |
3503 | * | |
3504 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3505 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3506 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3507 | * dev_probe_lock). | |
3508 | */ | |
3d632c3f SH |
3509 | static int bond_netdev_event(struct notifier_block *this, |
3510 | unsigned long event, void *ptr) | |
1da177e4 LT |
3511 | { |
3512 | struct net_device *event_dev = (struct net_device *)ptr; | |
3513 | ||
5a03cdb7 | 3514 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3515 | event_dev ? event_dev->name : "None", |
3516 | event); | |
1da177e4 | 3517 | |
0b680e75 JV |
3518 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3519 | return NOTIFY_DONE; | |
3520 | ||
1da177e4 | 3521 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3522 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3523 | return bond_master_netdev_event(event, event_dev); |
3524 | } | |
3525 | ||
3526 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3527 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3528 | return bond_slave_netdev_event(event, event_dev); |
3529 | } | |
3530 | ||
3531 | return NOTIFY_DONE; | |
3532 | } | |
3533 | ||
c3ade5ca JV |
3534 | /* |
3535 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3536 | * | |
3537 | * We keep track of device IPs primarily to use as source addresses in | |
3538 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3539 | * | |
3540 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3541 | */ | |
3542 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3543 | { | |
3544 | struct in_ifaddr *ifa = ptr; | |
3545 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
ec87fd3b | 3546 | struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id); |
0883beca PE |
3547 | struct bonding *bond; |
3548 | struct vlan_entry *vlan; | |
c3ade5ca | 3549 | |
ec87fd3b | 3550 | list_for_each_entry(bond, &bn->dev_list, bond_list) { |
c3ade5ca JV |
3551 | if (bond->dev == event_dev) { |
3552 | switch (event) { | |
3553 | case NETDEV_UP: | |
3554 | bond->master_ip = ifa->ifa_local; | |
3555 | return NOTIFY_OK; | |
3556 | case NETDEV_DOWN: | |
3557 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3558 | return NOTIFY_OK; | |
3559 | default: | |
3560 | return NOTIFY_DONE; | |
3561 | } | |
3562 | } | |
3563 | ||
0883beca | 3564 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 3565 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3566 | if (vlan_dev == event_dev) { |
3567 | switch (event) { | |
3568 | case NETDEV_UP: | |
3569 | vlan->vlan_ip = ifa->ifa_local; | |
3570 | return NOTIFY_OK; | |
3571 | case NETDEV_DOWN: | |
3572 | vlan->vlan_ip = | |
3573 | bond_glean_dev_ip(vlan_dev); | |
3574 | return NOTIFY_OK; | |
3575 | default: | |
3576 | return NOTIFY_DONE; | |
3577 | } | |
3578 | } | |
3579 | } | |
3580 | } | |
3581 | return NOTIFY_DONE; | |
3582 | } | |
3583 | ||
1da177e4 LT |
3584 | static struct notifier_block bond_netdev_notifier = { |
3585 | .notifier_call = bond_netdev_event, | |
3586 | }; | |
3587 | ||
c3ade5ca JV |
3588 | static struct notifier_block bond_inetaddr_notifier = { |
3589 | .notifier_call = bond_inetaddr_event, | |
3590 | }; | |
3591 | ||
1da177e4 LT |
3592 | /*-------------------------- Packet type handling ---------------------------*/ |
3593 | ||
3594 | /* register to receive lacpdus on a bond */ | |
3595 | static void bond_register_lacpdu(struct bonding *bond) | |
3596 | { | |
3597 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3598 | ||
3599 | /* initialize packet type */ | |
3600 | pk_type->type = PKT_TYPE_LACPDU; | |
3601 | pk_type->dev = bond->dev; | |
3602 | pk_type->func = bond_3ad_lacpdu_recv; | |
3603 | ||
3604 | dev_add_pack(pk_type); | |
3605 | } | |
3606 | ||
3607 | /* unregister to receive lacpdus on a bond */ | |
3608 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3609 | { | |
3610 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3611 | } | |
3612 | ||
f5b2b966 JV |
3613 | void bond_register_arp(struct bonding *bond) |
3614 | { | |
3615 | struct packet_type *pt = &bond->arp_mon_pt; | |
3616 | ||
c4f283b1 JV |
3617 | if (pt->type) |
3618 | return; | |
3619 | ||
f5b2b966 | 3620 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3621 | pt->dev = bond->dev; |
f5b2b966 JV |
3622 | pt->func = bond_arp_rcv; |
3623 | dev_add_pack(pt); | |
3624 | } | |
3625 | ||
3626 | void bond_unregister_arp(struct bonding *bond) | |
3627 | { | |
c4f283b1 JV |
3628 | struct packet_type *pt = &bond->arp_mon_pt; |
3629 | ||
3630 | dev_remove_pack(pt); | |
3631 | pt->type = 0; | |
f5b2b966 JV |
3632 | } |
3633 | ||
169a3e66 JV |
3634 | /*---------------------------- Hashing Policies -----------------------------*/ |
3635 | ||
6f6652be JV |
3636 | /* |
3637 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3638 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3639 | */ | |
a361c83c | 3640 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3641 | { |
3642 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3643 | struct iphdr *iph = ip_hdr(skb); | |
3644 | ||
f14c4e4e | 3645 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3646 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3647 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3648 | } |
3649 | ||
d3da6831 | 3650 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3651 | } |
3652 | ||
169a3e66 | 3653 | /* |
59c51591 | 3654 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3655 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3656 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3657 | */ | |
a361c83c | 3658 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3659 | { |
3660 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3661 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3662 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3663 | int layer4_xor = 0; |
3664 | ||
f14c4e4e BH |
3665 | if (skb->protocol == htons(ETH_P_IP)) { |
3666 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3667 | (iph->protocol == IPPROTO_TCP || |
3668 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3669 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3670 | } |
3671 | return (layer4_xor ^ | |
3672 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3673 | ||
3674 | } | |
3675 | ||
d3da6831 | 3676 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3677 | } |
3678 | ||
3679 | /* | |
3680 | * Hash for the output device based upon layer 2 data | |
3681 | */ | |
a361c83c | 3682 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3683 | { |
3684 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3685 | ||
d3da6831 | 3686 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3687 | } |
3688 | ||
1da177e4 LT |
3689 | /*-------------------------- Device entry points ----------------------------*/ |
3690 | ||
3691 | static int bond_open(struct net_device *bond_dev) | |
3692 | { | |
454d7c9b | 3693 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3694 | |
3695 | bond->kill_timers = 0; | |
3696 | ||
58402054 | 3697 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3698 | /* bond_alb_initialize must be called before the timer |
3699 | * is started. | |
3700 | */ | |
3701 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3702 | /* something went wrong - fail the open operation */ | |
b473946a | 3703 | return -ENOMEM; |
1da177e4 LT |
3704 | } |
3705 | ||
1b76b316 JV |
3706 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3707 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3708 | } |
3709 | ||
3710 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3711 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3712 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3713 | } |
3714 | ||
3715 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3716 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3717 | INIT_DELAYED_WORK(&bond->arp_work, | |
3718 | bond_activebackup_arp_mon); | |
3719 | else | |
3720 | INIT_DELAYED_WORK(&bond->arp_work, | |
3721 | bond_loadbalance_arp_mon); | |
3722 | ||
3723 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3724 | if (bond->params.arp_validate) |
3725 | bond_register_arp(bond); | |
1da177e4 LT |
3726 | } |
3727 | ||
3728 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3729 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3730 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3731 | /* register to receive LACPDUs */ |
3732 | bond_register_lacpdu(bond); | |
fd989c83 | 3733 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3734 | } |
3735 | ||
3736 | return 0; | |
3737 | } | |
3738 | ||
3739 | static int bond_close(struct net_device *bond_dev) | |
3740 | { | |
454d7c9b | 3741 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3742 | |
3743 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3744 | /* Unregister the receive of LACPDUs */ | |
3745 | bond_unregister_lacpdu(bond); | |
3746 | } | |
3747 | ||
f5b2b966 JV |
3748 | if (bond->params.arp_validate) |
3749 | bond_unregister_arp(bond); | |
3750 | ||
1da177e4 LT |
3751 | write_lock_bh(&bond->lock); |
3752 | ||
b59f9f74 | 3753 | bond->send_grat_arp = 0; |
305d552a | 3754 | bond->send_unsol_na = 0; |
1da177e4 LT |
3755 | |
3756 | /* signal timers not to re-arm */ | |
3757 | bond->kill_timers = 1; | |
3758 | ||
3759 | write_unlock_bh(&bond->lock); | |
3760 | ||
1da177e4 | 3761 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3762 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3763 | } |
3764 | ||
3765 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3766 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3767 | } |
3768 | ||
3769 | switch (bond->params.mode) { | |
3770 | case BOND_MODE_8023AD: | |
1b76b316 | 3771 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3772 | break; |
3773 | case BOND_MODE_TLB: | |
3774 | case BOND_MODE_ALB: | |
1b76b316 | 3775 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3776 | break; |
3777 | default: | |
3778 | break; | |
3779 | } | |
3780 | ||
1da177e4 | 3781 | |
58402054 | 3782 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3783 | /* Must be called only after all |
3784 | * slaves have been released | |
3785 | */ | |
3786 | bond_alb_deinitialize(bond); | |
3787 | } | |
3788 | ||
3789 | return 0; | |
3790 | } | |
3791 | ||
3792 | static struct net_device_stats *bond_get_stats(struct net_device *bond_dev) | |
3793 | { | |
454d7c9b | 3794 | struct bonding *bond = netdev_priv(bond_dev); |
35cfabdc | 3795 | struct net_device_stats *stats = &bond_dev->stats; |
2439f9eb | 3796 | struct net_device_stats local_stats; |
1da177e4 LT |
3797 | struct slave *slave; |
3798 | int i; | |
3799 | ||
2439f9eb | 3800 | memset(&local_stats, 0, sizeof(struct net_device_stats)); |
1da177e4 LT |
3801 | |
3802 | read_lock_bh(&bond->lock); | |
3803 | ||
3804 | bond_for_each_slave(bond, slave, i) { | |
eeda3fd6 SH |
3805 | const struct net_device_stats *sstats = dev_get_stats(slave->dev); |
3806 | ||
2439f9eb AG |
3807 | local_stats.rx_packets += sstats->rx_packets; |
3808 | local_stats.rx_bytes += sstats->rx_bytes; | |
3809 | local_stats.rx_errors += sstats->rx_errors; | |
3810 | local_stats.rx_dropped += sstats->rx_dropped; | |
3811 | ||
3812 | local_stats.tx_packets += sstats->tx_packets; | |
3813 | local_stats.tx_bytes += sstats->tx_bytes; | |
3814 | local_stats.tx_errors += sstats->tx_errors; | |
3815 | local_stats.tx_dropped += sstats->tx_dropped; | |
3816 | ||
3817 | local_stats.multicast += sstats->multicast; | |
3818 | local_stats.collisions += sstats->collisions; | |
3819 | ||
3820 | local_stats.rx_length_errors += sstats->rx_length_errors; | |
3821 | local_stats.rx_over_errors += sstats->rx_over_errors; | |
3822 | local_stats.rx_crc_errors += sstats->rx_crc_errors; | |
3823 | local_stats.rx_frame_errors += sstats->rx_frame_errors; | |
3824 | local_stats.rx_fifo_errors += sstats->rx_fifo_errors; | |
3825 | local_stats.rx_missed_errors += sstats->rx_missed_errors; | |
3826 | ||
3827 | local_stats.tx_aborted_errors += sstats->tx_aborted_errors; | |
3828 | local_stats.tx_carrier_errors += sstats->tx_carrier_errors; | |
3829 | local_stats.tx_fifo_errors += sstats->tx_fifo_errors; | |
3830 | local_stats.tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3831 | local_stats.tx_window_errors += sstats->tx_window_errors; | |
3832 | } | |
3833 | ||
3834 | memcpy(stats, &local_stats, sizeof(struct net_device_stats)); | |
1da177e4 LT |
3835 | |
3836 | read_unlock_bh(&bond->lock); | |
3837 | ||
3838 | return stats; | |
3839 | } | |
3840 | ||
3841 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3842 | { | |
3843 | struct net_device *slave_dev = NULL; | |
3844 | struct ifbond k_binfo; | |
3845 | struct ifbond __user *u_binfo = NULL; | |
3846 | struct ifslave k_sinfo; | |
3847 | struct ifslave __user *u_sinfo = NULL; | |
3848 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3849 | int res = 0; |
3850 | ||
a4aee5c8 | 3851 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3852 | |
3853 | switch (cmd) { | |
1da177e4 LT |
3854 | case SIOCGMIIPHY: |
3855 | mii = if_mii(ifr); | |
3d632c3f | 3856 | if (!mii) |
1da177e4 | 3857 | return -EINVAL; |
3d632c3f | 3858 | |
1da177e4 LT |
3859 | mii->phy_id = 0; |
3860 | /* Fall Through */ | |
3861 | case SIOCGMIIREG: | |
3862 | /* | |
3863 | * We do this again just in case we were called by SIOCGMIIREG | |
3864 | * instead of SIOCGMIIPHY. | |
3865 | */ | |
3866 | mii = if_mii(ifr); | |
3d632c3f | 3867 | if (!mii) |
1da177e4 | 3868 | return -EINVAL; |
3d632c3f | 3869 | |
1da177e4 LT |
3870 | |
3871 | if (mii->reg_num == 1) { | |
454d7c9b | 3872 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3873 | mii->val_out = 0; |
6603a6f2 | 3874 | read_lock(&bond->lock); |
1da177e4 | 3875 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3876 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3877 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3878 | |
1da177e4 | 3879 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3880 | read_unlock(&bond->lock); |
1da177e4 LT |
3881 | } |
3882 | ||
3883 | return 0; | |
3884 | case BOND_INFO_QUERY_OLD: | |
3885 | case SIOCBONDINFOQUERY: | |
3886 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3887 | ||
3d632c3f | 3888 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3889 | return -EFAULT; |
1da177e4 LT |
3890 | |
3891 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3892 | if (res == 0 && |
3893 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3894 | return -EFAULT; | |
1da177e4 LT |
3895 | |
3896 | return res; | |
3897 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3898 | case SIOCBONDSLAVEINFOQUERY: | |
3899 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3900 | ||
3d632c3f | 3901 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3902 | return -EFAULT; |
1da177e4 LT |
3903 | |
3904 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3905 | if (res == 0 && |
3906 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3907 | return -EFAULT; | |
1da177e4 LT |
3908 | |
3909 | return res; | |
3910 | default: | |
3911 | /* Go on */ | |
3912 | break; | |
3913 | } | |
3914 | ||
3d632c3f | 3915 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3916 | return -EPERM; |
1da177e4 | 3917 | |
ec87fd3b | 3918 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3919 | |
a4aee5c8 | 3920 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3921 | |
3d632c3f | 3922 | if (!slave_dev) |
1da177e4 | 3923 | res = -ENODEV; |
3d632c3f | 3924 | else { |
a4aee5c8 | 3925 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3926 | switch (cmd) { |
3927 | case BOND_ENSLAVE_OLD: | |
3928 | case SIOCBONDENSLAVE: | |
3929 | res = bond_enslave(bond_dev, slave_dev); | |
3930 | break; | |
3931 | case BOND_RELEASE_OLD: | |
3932 | case SIOCBONDRELEASE: | |
3933 | res = bond_release(bond_dev, slave_dev); | |
3934 | break; | |
3935 | case BOND_SETHWADDR_OLD: | |
3936 | case SIOCBONDSETHWADDR: | |
3937 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3938 | break; | |
3939 | case BOND_CHANGE_ACTIVE_OLD: | |
3940 | case SIOCBONDCHANGEACTIVE: | |
3941 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3942 | break; | |
3943 | default: | |
3944 | res = -EOPNOTSUPP; | |
3945 | } | |
3946 | ||
3947 | dev_put(slave_dev); | |
3948 | } | |
3949 | ||
1da177e4 LT |
3950 | return res; |
3951 | } | |
3952 | ||
22bedad3 JP |
3953 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3954 | struct netdev_hw_addr_list *list, | |
3955 | int addrlen) | |
3956 | { | |
3957 | struct netdev_hw_addr *ha; | |
3958 | ||
3959 | netdev_hw_addr_list_for_each(ha, list) | |
3960 | if (!memcmp(ha->addr, addr, addrlen)) | |
3961 | return true; | |
3962 | ||
3963 | return false; | |
3964 | } | |
3965 | ||
1da177e4 LT |
3966 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3967 | { | |
454d7c9b | 3968 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 JP |
3969 | struct netdev_hw_addr *ha; |
3970 | bool found; | |
1da177e4 | 3971 | |
1da177e4 LT |
3972 | /* |
3973 | * Do promisc before checking multicast_mode | |
3974 | */ | |
3d632c3f | 3975 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
3976 | /* |
3977 | * FIXME: Need to handle the error when one of the multi-slaves | |
3978 | * encounters error. | |
3979 | */ | |
1da177e4 | 3980 | bond_set_promiscuity(bond, 1); |
1da177e4 | 3981 | |
3d632c3f SH |
3982 | |
3983 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 3984 | bond_set_promiscuity(bond, -1); |
3d632c3f | 3985 | |
1da177e4 LT |
3986 | |
3987 | /* set allmulti flag to slaves */ | |
3d632c3f | 3988 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
3989 | /* |
3990 | * FIXME: Need to handle the error when one of the multi-slaves | |
3991 | * encounters error. | |
3992 | */ | |
1da177e4 | 3993 | bond_set_allmulti(bond, 1); |
1da177e4 | 3994 | |
3d632c3f SH |
3995 | |
3996 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 3997 | bond_set_allmulti(bond, -1); |
3d632c3f | 3998 | |
1da177e4 | 3999 | |
80ee5ad2 JV |
4000 | read_lock(&bond->lock); |
4001 | ||
1da177e4 LT |
4002 | bond->flags = bond_dev->flags; |
4003 | ||
4004 | /* looking for addresses to add to slaves' mc list */ | |
22bedad3 JP |
4005 | netdev_for_each_mc_addr(ha, bond_dev) { |
4006 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
4007 | bond_dev->addr_len); | |
4008 | if (!found) | |
4009 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
4010 | } |
4011 | ||
4012 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
4013 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
4014 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
4015 | bond_dev->addr_len); | |
4016 | if (!found) | |
4017 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
4018 | } |
4019 | ||
4020 | /* save master's multicast list */ | |
22bedad3 JP |
4021 | __hw_addr_flush(&bond->mc_list); |
4022 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
4023 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 4024 | |
80ee5ad2 | 4025 | read_unlock(&bond->lock); |
1da177e4 LT |
4026 | } |
4027 | ||
00829823 SH |
4028 | static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms) |
4029 | { | |
4030 | struct bonding *bond = netdev_priv(dev); | |
4031 | struct slave *slave = bond->first_slave; | |
4032 | ||
4033 | if (slave) { | |
4034 | const struct net_device_ops *slave_ops | |
4035 | = slave->dev->netdev_ops; | |
4036 | if (slave_ops->ndo_neigh_setup) | |
72e2240f | 4037 | return slave_ops->ndo_neigh_setup(slave->dev, parms); |
00829823 SH |
4038 | } |
4039 | return 0; | |
4040 | } | |
4041 | ||
1da177e4 LT |
4042 | /* |
4043 | * Change the MTU of all of a master's slaves to match the master | |
4044 | */ | |
4045 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
4046 | { | |
454d7c9b | 4047 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4048 | struct slave *slave, *stop_at; |
4049 | int res = 0; | |
4050 | int i; | |
4051 | ||
5a03cdb7 | 4052 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 4053 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
4054 | |
4055 | /* Can't hold bond->lock with bh disabled here since | |
4056 | * some base drivers panic. On the other hand we can't | |
4057 | * hold bond->lock without bh disabled because we'll | |
4058 | * deadlock. The only solution is to rely on the fact | |
4059 | * that we're under rtnl_lock here, and the slaves | |
4060 | * list won't change. This doesn't solve the problem | |
4061 | * of setting the slave's MTU while it is | |
4062 | * transmitting, but the assumption is that the base | |
4063 | * driver can handle that. | |
4064 | * | |
4065 | * TODO: figure out a way to safely iterate the slaves | |
4066 | * list, but without holding a lock around the actual | |
4067 | * call to the base driver. | |
4068 | */ | |
4069 | ||
4070 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
4071 | pr_debug("s %p s->p %p c_m %p\n", |
4072 | slave, | |
4073 | slave->prev, | |
4074 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 4075 | |
1da177e4 LT |
4076 | res = dev_set_mtu(slave->dev, new_mtu); |
4077 | ||
4078 | if (res) { | |
4079 | /* If we failed to set the slave's mtu to the new value | |
4080 | * we must abort the operation even in ACTIVE_BACKUP | |
4081 | * mode, because if we allow the backup slaves to have | |
4082 | * different mtu values than the active slave we'll | |
4083 | * need to change their mtu when doing a failover. That | |
4084 | * means changing their mtu from timer context, which | |
4085 | * is probably not a good idea. | |
4086 | */ | |
5a03cdb7 | 4087 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4088 | goto unwind; |
4089 | } | |
4090 | } | |
4091 | ||
4092 | bond_dev->mtu = new_mtu; | |
4093 | ||
4094 | return 0; | |
4095 | ||
4096 | unwind: | |
4097 | /* unwind from head to the slave that failed */ | |
4098 | stop_at = slave; | |
4099 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4100 | int tmp_res; | |
4101 | ||
4102 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
4103 | if (tmp_res) { | |
a4aee5c8 JP |
4104 | pr_debug("unwind err %d dev %s\n", |
4105 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4106 | } |
4107 | } | |
4108 | ||
4109 | return res; | |
4110 | } | |
4111 | ||
4112 | /* | |
4113 | * Change HW address | |
4114 | * | |
4115 | * Note that many devices must be down to change the HW address, and | |
4116 | * downing the master releases all slaves. We can make bonds full of | |
4117 | * bonding devices to test this, however. | |
4118 | */ | |
4119 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4120 | { | |
454d7c9b | 4121 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4122 | struct sockaddr *sa = addr, tmp_sa; |
4123 | struct slave *slave, *stop_at; | |
4124 | int res = 0; | |
4125 | int i; | |
4126 | ||
eb7cc59a SH |
4127 | if (bond->params.mode == BOND_MODE_ALB) |
4128 | return bond_alb_set_mac_address(bond_dev, addr); | |
4129 | ||
4130 | ||
a4aee5c8 JP |
4131 | pr_debug("bond=%p, name=%s\n", |
4132 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 4133 | |
dd957c57 | 4134 | /* |
3915c1e8 JV |
4135 | * If fail_over_mac is set to active, do nothing and return |
4136 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 4137 | */ |
3915c1e8 | 4138 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 4139 | return 0; |
2ab82852 | 4140 | |
3d632c3f | 4141 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 4142 | return -EADDRNOTAVAIL; |
1da177e4 LT |
4143 | |
4144 | /* Can't hold bond->lock with bh disabled here since | |
4145 | * some base drivers panic. On the other hand we can't | |
4146 | * hold bond->lock without bh disabled because we'll | |
4147 | * deadlock. The only solution is to rely on the fact | |
4148 | * that we're under rtnl_lock here, and the slaves | |
4149 | * list won't change. This doesn't solve the problem | |
4150 | * of setting the slave's hw address while it is | |
4151 | * transmitting, but the assumption is that the base | |
4152 | * driver can handle that. | |
4153 | * | |
4154 | * TODO: figure out a way to safely iterate the slaves | |
4155 | * list, but without holding a lock around the actual | |
4156 | * call to the base driver. | |
4157 | */ | |
4158 | ||
4159 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 4160 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 4161 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 4162 | |
eb7cc59a | 4163 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 4164 | res = -EOPNOTSUPP; |
5a03cdb7 | 4165 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
4166 | goto unwind; |
4167 | } | |
4168 | ||
4169 | res = dev_set_mac_address(slave->dev, addr); | |
4170 | if (res) { | |
4171 | /* TODO: consider downing the slave | |
4172 | * and retry ? | |
4173 | * User should expect communications | |
4174 | * breakage anyway until ARP finish | |
4175 | * updating, so... | |
4176 | */ | |
5a03cdb7 | 4177 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4178 | goto unwind; |
4179 | } | |
4180 | } | |
4181 | ||
4182 | /* success */ | |
4183 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4184 | return 0; | |
4185 | ||
4186 | unwind: | |
4187 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4188 | tmp_sa.sa_family = bond_dev->type; | |
4189 | ||
4190 | /* unwind from head to the slave that failed */ | |
4191 | stop_at = slave; | |
4192 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4193 | int tmp_res; | |
4194 | ||
4195 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4196 | if (tmp_res) { | |
a4aee5c8 JP |
4197 | pr_debug("unwind err %d dev %s\n", |
4198 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4199 | } |
4200 | } | |
4201 | ||
4202 | return res; | |
4203 | } | |
4204 | ||
4205 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4206 | { | |
454d7c9b | 4207 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4208 | struct slave *slave, *start_at; |
cf5f9044 | 4209 | int i, slave_no, res = 1; |
a2fd940f | 4210 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
4211 | |
4212 | read_lock(&bond->lock); | |
4213 | ||
3d632c3f | 4214 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4215 | goto out; |
cf5f9044 | 4216 | /* |
a2fd940f AG |
4217 | * Start with the curr_active_slave that joined the bond as the |
4218 | * default for sending IGMP traffic. For failover purposes one | |
4219 | * needs to maintain some consistency for the interface that will | |
4220 | * send the join/membership reports. The curr_active_slave found | |
4221 | * will send all of this type of traffic. | |
cf5f9044 | 4222 | */ |
00ae7028 | 4223 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4224 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4225 | |
a2fd940f AG |
4226 | read_lock(&bond->curr_slave_lock); |
4227 | slave = bond->curr_active_slave; | |
4228 | read_unlock(&bond->curr_slave_lock); | |
4229 | ||
4230 | if (!slave) | |
4231 | goto out; | |
4232 | } else { | |
4233 | /* | |
4234 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4235 | * that as being rare enough not to justify using an | |
4236 | * atomic op here. | |
4237 | */ | |
4238 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4239 | ||
4240 | bond_for_each_slave(bond, slave, i) { | |
4241 | slave_no--; | |
4242 | if (slave_no < 0) | |
4243 | break; | |
4244 | } | |
1da177e4 LT |
4245 | } |
4246 | ||
cf5f9044 | 4247 | start_at = slave; |
1da177e4 LT |
4248 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4249 | if (IS_UP(slave->dev) && | |
4250 | (slave->link == BOND_LINK_UP) && | |
4251 | (slave->state == BOND_STATE_ACTIVE)) { | |
4252 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
1da177e4 LT |
4253 | break; |
4254 | } | |
4255 | } | |
4256 | ||
1da177e4 LT |
4257 | out: |
4258 | if (res) { | |
4259 | /* no suitable interface, frame not sent */ | |
4260 | dev_kfree_skb(skb); | |
4261 | } | |
4262 | read_unlock(&bond->lock); | |
ec634fe3 | 4263 | return NETDEV_TX_OK; |
1da177e4 LT |
4264 | } |
4265 | ||
075897ce | 4266 | |
1da177e4 LT |
4267 | /* |
4268 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4269 | * the bond has a usable interface. | |
4270 | */ | |
4271 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4272 | { | |
454d7c9b | 4273 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4274 | int res = 1; |
4275 | ||
1da177e4 LT |
4276 | read_lock(&bond->lock); |
4277 | read_lock(&bond->curr_slave_lock); | |
4278 | ||
3d632c3f | 4279 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4280 | goto out; |
1da177e4 | 4281 | |
075897ce JL |
4282 | if (!bond->curr_active_slave) |
4283 | goto out; | |
4284 | ||
075897ce JL |
4285 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4286 | ||
1da177e4 | 4287 | out: |
3d632c3f | 4288 | if (res) |
1da177e4 LT |
4289 | /* no suitable interface, frame not sent */ |
4290 | dev_kfree_skb(skb); | |
3d632c3f | 4291 | |
1da177e4 LT |
4292 | read_unlock(&bond->curr_slave_lock); |
4293 | read_unlock(&bond->lock); | |
ec634fe3 | 4294 | return NETDEV_TX_OK; |
1da177e4 LT |
4295 | } |
4296 | ||
4297 | /* | |
169a3e66 JV |
4298 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4299 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4300 | * find the next active slave. | |
1da177e4 LT |
4301 | */ |
4302 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4303 | { | |
454d7c9b | 4304 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4305 | struct slave *slave, *start_at; |
4306 | int slave_no; | |
4307 | int i; | |
4308 | int res = 1; | |
4309 | ||
4310 | read_lock(&bond->lock); | |
4311 | ||
3d632c3f | 4312 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4313 | goto out; |
1da177e4 | 4314 | |
a361c83c | 4315 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4316 | |
4317 | bond_for_each_slave(bond, slave, i) { | |
4318 | slave_no--; | |
3d632c3f | 4319 | if (slave_no < 0) |
1da177e4 | 4320 | break; |
1da177e4 LT |
4321 | } |
4322 | ||
4323 | start_at = slave; | |
4324 | ||
4325 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4326 | if (IS_UP(slave->dev) && | |
4327 | (slave->link == BOND_LINK_UP) && | |
4328 | (slave->state == BOND_STATE_ACTIVE)) { | |
4329 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4330 | break; | |
4331 | } | |
4332 | } | |
4333 | ||
4334 | out: | |
4335 | if (res) { | |
4336 | /* no suitable interface, frame not sent */ | |
4337 | dev_kfree_skb(skb); | |
4338 | } | |
4339 | read_unlock(&bond->lock); | |
ec634fe3 | 4340 | return NETDEV_TX_OK; |
1da177e4 LT |
4341 | } |
4342 | ||
4343 | /* | |
4344 | * in broadcast mode, we send everything to all usable interfaces. | |
4345 | */ | |
4346 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4347 | { | |
454d7c9b | 4348 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4349 | struct slave *slave, *start_at; |
4350 | struct net_device *tx_dev = NULL; | |
4351 | int i; | |
4352 | int res = 1; | |
4353 | ||
4354 | read_lock(&bond->lock); | |
4355 | ||
3d632c3f | 4356 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4357 | goto out; |
1da177e4 LT |
4358 | |
4359 | read_lock(&bond->curr_slave_lock); | |
4360 | start_at = bond->curr_active_slave; | |
4361 | read_unlock(&bond->curr_slave_lock); | |
4362 | ||
3d632c3f | 4363 | if (!start_at) |
1da177e4 | 4364 | goto out; |
1da177e4 LT |
4365 | |
4366 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4367 | if (IS_UP(slave->dev) && | |
4368 | (slave->link == BOND_LINK_UP) && | |
4369 | (slave->state == BOND_STATE_ACTIVE)) { | |
4370 | if (tx_dev) { | |
4371 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4372 | if (!skb2) { | |
a4aee5c8 | 4373 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4374 | bond_dev->name); |
1da177e4 LT |
4375 | continue; |
4376 | } | |
4377 | ||
4378 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4379 | if (res) { | |
4380 | dev_kfree_skb(skb2); | |
4381 | continue; | |
4382 | } | |
4383 | } | |
4384 | tx_dev = slave->dev; | |
4385 | } | |
4386 | } | |
4387 | ||
3d632c3f | 4388 | if (tx_dev) |
1da177e4 | 4389 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4390 | |
4391 | out: | |
3d632c3f | 4392 | if (res) |
1da177e4 LT |
4393 | /* no suitable interface, frame not sent */ |
4394 | dev_kfree_skb(skb); | |
3d632c3f | 4395 | |
1da177e4 LT |
4396 | /* frame sent to all suitable interfaces */ |
4397 | read_unlock(&bond->lock); | |
ec634fe3 | 4398 | return NETDEV_TX_OK; |
1da177e4 LT |
4399 | } |
4400 | ||
4401 | /*------------------------- Device initialization ---------------------------*/ | |
4402 | ||
6f6652be JV |
4403 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4404 | { | |
4405 | switch (bond->params.xmit_policy) { | |
4406 | case BOND_XMIT_POLICY_LAYER23: | |
4407 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4408 | break; | |
4409 | case BOND_XMIT_POLICY_LAYER34: | |
4410 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4411 | break; | |
4412 | case BOND_XMIT_POLICY_LAYER2: | |
4413 | default: | |
4414 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4415 | break; | |
4416 | } | |
4417 | } | |
4418 | ||
424efe9c | 4419 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 SH |
4420 | { |
4421 | const struct bonding *bond = netdev_priv(dev); | |
4422 | ||
4423 | switch (bond->params.mode) { | |
4424 | case BOND_MODE_ROUNDROBIN: | |
4425 | return bond_xmit_roundrobin(skb, dev); | |
4426 | case BOND_MODE_ACTIVEBACKUP: | |
4427 | return bond_xmit_activebackup(skb, dev); | |
4428 | case BOND_MODE_XOR: | |
4429 | return bond_xmit_xor(skb, dev); | |
4430 | case BOND_MODE_BROADCAST: | |
4431 | return bond_xmit_broadcast(skb, dev); | |
4432 | case BOND_MODE_8023AD: | |
4433 | return bond_3ad_xmit_xor(skb, dev); | |
4434 | case BOND_MODE_ALB: | |
4435 | case BOND_MODE_TLB: | |
4436 | return bond_alb_xmit(skb, dev); | |
4437 | default: | |
4438 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4439 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4440 | dev->name, bond->params.mode); | |
00829823 SH |
4441 | WARN_ON_ONCE(1); |
4442 | dev_kfree_skb(skb); | |
4443 | return NETDEV_TX_OK; | |
4444 | } | |
4445 | } | |
4446 | ||
4447 | ||
1da177e4 LT |
4448 | /* |
4449 | * set bond mode specific net device operations | |
4450 | */ | |
a77b5325 | 4451 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4452 | { |
169a3e66 JV |
4453 | struct net_device *bond_dev = bond->dev; |
4454 | ||
1da177e4 LT |
4455 | switch (mode) { |
4456 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4457 | break; |
4458 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4459 | break; |
4460 | case BOND_MODE_XOR: | |
6f6652be | 4461 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4462 | break; |
4463 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4464 | break; |
4465 | case BOND_MODE_8023AD: | |
8f903c70 | 4466 | bond_set_master_3ad_flags(bond); |
6f6652be | 4467 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4468 | break; |
1da177e4 | 4469 | case BOND_MODE_ALB: |
8f903c70 JV |
4470 | bond_set_master_alb_flags(bond); |
4471 | /* FALLTHRU */ | |
4472 | case BOND_MODE_TLB: | |
1da177e4 LT |
4473 | break; |
4474 | default: | |
4475 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4476 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4477 | bond_dev->name, mode); | |
1da177e4 LT |
4478 | break; |
4479 | } | |
4480 | } | |
4481 | ||
217df670 JV |
4482 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4483 | struct ethtool_drvinfo *drvinfo) | |
4484 | { | |
4485 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4486 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4487 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4488 | } | |
4489 | ||
7282d491 | 4490 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4491 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4492 | .get_link = ethtool_op_get_link, |
4493 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4494 | .get_sg = ethtool_op_get_sg, | |
4495 | .get_tso = ethtool_op_get_tso, | |
4496 | .get_ufo = ethtool_op_get_ufo, | |
4497 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4498 | }; |
4499 | ||
eb7cc59a | 4500 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4501 | .ndo_init = bond_init, |
9e71626c | 4502 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4503 | .ndo_open = bond_open, |
4504 | .ndo_stop = bond_close, | |
00829823 | 4505 | .ndo_start_xmit = bond_start_xmit, |
eb7cc59a SH |
4506 | .ndo_get_stats = bond_get_stats, |
4507 | .ndo_do_ioctl = bond_do_ioctl, | |
4508 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4509 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4510 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4511 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4512 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4513 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4514 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
f6dc31a8 WC |
4515 | #ifdef CONFIG_NET_POLL_CONTROLLER |
4516 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, | |
4517 | .ndo_poll_controller = bond_poll_controller, | |
4518 | #endif | |
eb7cc59a SH |
4519 | }; |
4520 | ||
9e2e61fb AW |
4521 | static void bond_destructor(struct net_device *bond_dev) |
4522 | { | |
4523 | struct bonding *bond = netdev_priv(bond_dev); | |
4524 | if (bond->wq) | |
4525 | destroy_workqueue(bond->wq); | |
4526 | free_netdev(bond_dev); | |
4527 | } | |
4528 | ||
181470fc | 4529 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4530 | { |
454d7c9b | 4531 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4532 | |
1da177e4 LT |
4533 | /* initialize rwlocks */ |
4534 | rwlock_init(&bond->lock); | |
4535 | rwlock_init(&bond->curr_slave_lock); | |
4536 | ||
d2991f75 | 4537 | bond->params = bonding_defaults; |
1da177e4 LT |
4538 | |
4539 | /* Initialize pointers */ | |
1da177e4 LT |
4540 | bond->dev = bond_dev; |
4541 | INIT_LIST_HEAD(&bond->vlan_list); | |
4542 | ||
4543 | /* Initialize the device entry points */ | |
181470fc | 4544 | ether_setup(bond_dev); |
eb7cc59a | 4545 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4546 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4547 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4548 | |
9e2e61fb | 4549 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4550 | |
4551 | /* Initialize the device options */ | |
4552 | bond_dev->tx_queue_len = 0; | |
4553 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4554 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4555 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4556 | ||
6cf3f41e JV |
4557 | if (bond->params.arp_interval) |
4558 | bond_dev->priv_flags |= IFF_MASTER_ARPMON; | |
1da177e4 LT |
4559 | |
4560 | /* At first, we block adding VLANs. That's the only way to | |
4561 | * prevent problems that occur when adding VLANs over an | |
4562 | * empty bond. The block will be removed once non-challenged | |
4563 | * slaves are enslaved. | |
4564 | */ | |
4565 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4566 | ||
932ff279 | 4567 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4568 | * transmitting */ |
4569 | bond_dev->features |= NETIF_F_LLTX; | |
4570 | ||
4571 | /* By default, we declare the bond to be fully | |
4572 | * VLAN hardware accelerated capable. Special | |
4573 | * care is taken in the various xmit functions | |
4574 | * when there are slaves that are not hw accel | |
4575 | * capable | |
4576 | */ | |
1da177e4 LT |
4577 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4578 | NETIF_F_HW_VLAN_RX | | |
4579 | NETIF_F_HW_VLAN_FILTER); | |
4580 | ||
1da177e4 LT |
4581 | } |
4582 | ||
fdaea7a9 JV |
4583 | static void bond_work_cancel_all(struct bonding *bond) |
4584 | { | |
4585 | write_lock_bh(&bond->lock); | |
4586 | bond->kill_timers = 1; | |
4587 | write_unlock_bh(&bond->lock); | |
4588 | ||
4589 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4590 | cancel_delayed_work(&bond->mii_work); | |
4591 | ||
4592 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4593 | cancel_delayed_work(&bond->arp_work); | |
4594 | ||
4595 | if (bond->params.mode == BOND_MODE_ALB && | |
4596 | delayed_work_pending(&bond->alb_work)) | |
4597 | cancel_delayed_work(&bond->alb_work); | |
4598 | ||
4599 | if (bond->params.mode == BOND_MODE_8023AD && | |
4600 | delayed_work_pending(&bond->ad_work)) | |
4601 | cancel_delayed_work(&bond->ad_work); | |
4602 | } | |
4603 | ||
c67dfb29 EB |
4604 | /* |
4605 | * Destroy a bonding device. | |
4606 | * Must be under rtnl_lock when this function is called. | |
4607 | */ | |
4608 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4609 | { |
454d7c9b | 4610 | struct bonding *bond = netdev_priv(bond_dev); |
a434e43f | 4611 | |
f6dc31a8 WC |
4612 | bond_netpoll_cleanup(bond_dev); |
4613 | ||
c67dfb29 EB |
4614 | /* Release the bonded slaves */ |
4615 | bond_release_all(bond_dev); | |
4616 | ||
a434e43f JV |
4617 | list_del(&bond->bond_list); |
4618 | ||
4619 | bond_work_cancel_all(bond); | |
4620 | ||
a434e43f | 4621 | bond_remove_proc_entry(bond); |
c67dfb29 | 4622 | |
22bedad3 | 4623 | __hw_addr_flush(&bond->mc_list); |
a434e43f JV |
4624 | } |
4625 | ||
1da177e4 LT |
4626 | /*------------------------- Module initialization ---------------------------*/ |
4627 | ||
4628 | /* | |
4629 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4630 | * number of the mode or its string name. A bit complicated because |
4631 | * some mode names are substrings of other names, and calls from sysfs | |
4632 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4633 | */ |
325dcf7a | 4634 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4635 | { |
54b87323 | 4636 | int modeint = -1, i, rv; |
a42e534f | 4637 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4638 | |
a42e534f JV |
4639 | for (p = (char *)buf; *p; p++) |
4640 | if (!(isdigit(*p) || isspace(*p))) | |
4641 | break; | |
4642 | ||
4643 | if (*p) | |
ece95f7f | 4644 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4645 | else |
54b87323 | 4646 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4647 | |
4648 | if (!rv) | |
4649 | return -1; | |
1da177e4 LT |
4650 | |
4651 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4652 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4653 | return tbl[i].mode; |
4654 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4655 | return tbl[i].mode; |
1da177e4 LT |
4656 | } |
4657 | ||
4658 | return -1; | |
4659 | } | |
4660 | ||
4661 | static int bond_check_params(struct bond_params *params) | |
4662 | { | |
a549952a | 4663 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4664 | |
1da177e4 LT |
4665 | /* |
4666 | * Convert string parameters. | |
4667 | */ | |
4668 | if (mode) { | |
4669 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4670 | if (bond_mode == -1) { | |
a4aee5c8 | 4671 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4672 | mode == NULL ? "NULL" : mode); |
4673 | return -EINVAL; | |
4674 | } | |
4675 | } | |
4676 | ||
169a3e66 JV |
4677 | if (xmit_hash_policy) { |
4678 | if ((bond_mode != BOND_MODE_XOR) && | |
4679 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4680 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4681 | bond_mode_name(bond_mode)); |
4682 | } else { | |
4683 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4684 | xmit_hashtype_tbl); | |
4685 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4686 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4687 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4688 | xmit_hash_policy); |
4689 | return -EINVAL; | |
4690 | } | |
4691 | } | |
4692 | } | |
4693 | ||
1da177e4 LT |
4694 | if (lacp_rate) { |
4695 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4696 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4697 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4698 | } else { |
4699 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4700 | if (lacp_fast == -1) { | |
a4aee5c8 | 4701 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4702 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4703 | return -EINVAL; | |
4704 | } | |
4705 | } | |
4706 | } | |
4707 | ||
fd989c83 JV |
4708 | if (ad_select) { |
4709 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4710 | if (params->ad_select == -1) { | |
a4aee5c8 | 4711 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4712 | ad_select == NULL ? "NULL" : ad_select); |
4713 | return -EINVAL; | |
4714 | } | |
4715 | ||
4716 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4717 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4718 | } |
4719 | } else { | |
4720 | params->ad_select = BOND_AD_STABLE; | |
4721 | } | |
4722 | ||
f5841306 | 4723 | if (max_bonds < 0) { |
a4aee5c8 JP |
4724 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4725 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4726 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4727 | } | |
4728 | ||
4729 | if (miimon < 0) { | |
a4aee5c8 JP |
4730 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4731 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4732 | miimon = BOND_LINK_MON_INTERV; |
4733 | } | |
4734 | ||
4735 | if (updelay < 0) { | |
a4aee5c8 JP |
4736 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4737 | updelay, INT_MAX); | |
1da177e4 LT |
4738 | updelay = 0; |
4739 | } | |
4740 | ||
4741 | if (downdelay < 0) { | |
a4aee5c8 JP |
4742 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4743 | downdelay, INT_MAX); | |
1da177e4 LT |
4744 | downdelay = 0; |
4745 | } | |
4746 | ||
4747 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4748 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4749 | use_carrier); | |
1da177e4 LT |
4750 | use_carrier = 1; |
4751 | } | |
4752 | ||
7893b249 | 4753 | if (num_grat_arp < 0 || num_grat_arp > 255) { |
2381a55c | 4754 | pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4755 | num_grat_arp); |
7893b249 MS |
4756 | num_grat_arp = 1; |
4757 | } | |
4758 | ||
305d552a | 4759 | if (num_unsol_na < 0 || num_unsol_na > 255) { |
2381a55c | 4760 | pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4761 | num_unsol_na); |
305d552a BH |
4762 | num_unsol_na = 1; |
4763 | } | |
4764 | ||
1da177e4 LT |
4765 | /* reset values for 802.3ad */ |
4766 | if (bond_mode == BOND_MODE_8023AD) { | |
4767 | if (!miimon) { | |
a4aee5c8 | 4768 | 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 | 4769 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4770 | miimon = 100; |
4771 | } | |
4772 | } | |
4773 | ||
4774 | /* reset values for TLB/ALB */ | |
4775 | if ((bond_mode == BOND_MODE_TLB) || | |
4776 | (bond_mode == BOND_MODE_ALB)) { | |
4777 | if (!miimon) { | |
a4aee5c8 | 4778 | 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 | 4779 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4780 | miimon = 100; |
4781 | } | |
4782 | } | |
4783 | ||
4784 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4785 | 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", |
4786 | updelay); | |
1da177e4 LT |
4787 | } |
4788 | ||
4789 | if (!miimon) { | |
4790 | if (updelay || downdelay) { | |
4791 | /* just warn the user the up/down delay will have | |
4792 | * no effect since miimon is zero... | |
4793 | */ | |
a4aee5c8 JP |
4794 | 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", |
4795 | updelay, downdelay); | |
1da177e4 LT |
4796 | } |
4797 | } else { | |
4798 | /* don't allow arp monitoring */ | |
4799 | if (arp_interval) { | |
a4aee5c8 JP |
4800 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4801 | miimon, arp_interval); | |
1da177e4 LT |
4802 | arp_interval = 0; |
4803 | } | |
4804 | ||
4805 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4806 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4807 | updelay, miimon, | |
4808 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4809 | } |
4810 | ||
4811 | updelay /= miimon; | |
4812 | ||
4813 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4814 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4815 | downdelay, miimon, | |
4816 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4817 | } |
4818 | ||
4819 | downdelay /= miimon; | |
4820 | } | |
4821 | ||
4822 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4823 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4824 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4825 | arp_interval = BOND_LINK_ARP_INTERV; |
4826 | } | |
4827 | ||
4828 | for (arp_ip_count = 0; | |
4829 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4830 | arp_ip_count++) { | |
4831 | /* not complete check, but should be good enough to | |
4832 | catch mistakes */ | |
4833 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4834 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4835 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4836 | arp_interval = 0; |
4837 | } else { | |
d3bb52b0 | 4838 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4839 | arp_target[arp_ip_count] = ip; |
4840 | } | |
4841 | } | |
4842 | ||
4843 | if (arp_interval && !arp_ip_count) { | |
4844 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4845 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4846 | arp_interval); | |
1da177e4 LT |
4847 | arp_interval = 0; |
4848 | } | |
4849 | ||
f5b2b966 JV |
4850 | if (arp_validate) { |
4851 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4852 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4853 | return -EINVAL; |
4854 | } | |
4855 | if (!arp_interval) { | |
a4aee5c8 | 4856 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4857 | return -EINVAL; |
4858 | } | |
4859 | ||
4860 | arp_validate_value = bond_parse_parm(arp_validate, | |
4861 | arp_validate_tbl); | |
4862 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4863 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4864 | arp_validate == NULL ? "NULL" : arp_validate); |
4865 | return -EINVAL; | |
4866 | } | |
4867 | } else | |
4868 | arp_validate_value = 0; | |
4869 | ||
1da177e4 | 4870 | if (miimon) { |
a4aee5c8 | 4871 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4872 | } else if (arp_interval) { |
4873 | int i; | |
4874 | ||
a4aee5c8 JP |
4875 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4876 | arp_interval, | |
4877 | arp_validate_tbl[arp_validate_value].modename, | |
4878 | arp_ip_count); | |
1da177e4 LT |
4879 | |
4880 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4881 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4882 | |
e5e2a8fd | 4883 | pr_info("\n"); |
1da177e4 | 4884 | |
b8a9787e | 4885 | } else if (max_bonds) { |
1da177e4 LT |
4886 | /* miimon and arp_interval not set, we need one so things |
4887 | * work as expected, see bonding.txt for details | |
4888 | */ | |
a4aee5c8 | 4889 | 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 |
4890 | } |
4891 | ||
4892 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4893 | /* currently, using a primary only makes sense | |
4894 | * in active backup, TLB or ALB modes | |
4895 | */ | |
a4aee5c8 JP |
4896 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4897 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4898 | primary = NULL; |
4899 | } | |
4900 | ||
a549952a JP |
4901 | if (primary && primary_reselect) { |
4902 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4903 | pri_reselect_tbl); | |
4904 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4905 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4906 | primary_reselect == |
4907 | NULL ? "NULL" : primary_reselect); | |
4908 | return -EINVAL; | |
4909 | } | |
4910 | } else { | |
4911 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4912 | } | |
4913 | ||
3915c1e8 JV |
4914 | if (fail_over_mac) { |
4915 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4916 | fail_over_mac_tbl); | |
4917 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4918 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4919 | arp_validate == NULL ? "NULL" : arp_validate); |
4920 | return -EINVAL; | |
4921 | } | |
4922 | ||
4923 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4924 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4925 | } else { |
4926 | fail_over_mac_value = BOND_FOM_NONE; | |
4927 | } | |
dd957c57 | 4928 | |
1da177e4 LT |
4929 | /* fill params struct with the proper values */ |
4930 | params->mode = bond_mode; | |
169a3e66 | 4931 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4932 | params->miimon = miimon; |
7893b249 | 4933 | params->num_grat_arp = num_grat_arp; |
305d552a | 4934 | params->num_unsol_na = num_unsol_na; |
1da177e4 | 4935 | params->arp_interval = arp_interval; |
f5b2b966 | 4936 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4937 | params->updelay = updelay; |
4938 | params->downdelay = downdelay; | |
4939 | params->use_carrier = use_carrier; | |
4940 | params->lacp_fast = lacp_fast; | |
4941 | params->primary[0] = 0; | |
a549952a | 4942 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4943 | params->fail_over_mac = fail_over_mac_value; |
1da177e4 LT |
4944 | |
4945 | if (primary) { | |
4946 | strncpy(params->primary, primary, IFNAMSIZ); | |
4947 | params->primary[IFNAMSIZ - 1] = 0; | |
4948 | } | |
4949 | ||
4950 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4951 | ||
4952 | return 0; | |
4953 | } | |
4954 | ||
0daa2303 | 4955 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4956 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 4957 | |
e8a0464c DM |
4958 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4959 | struct netdev_queue *txq, | |
4960 | void *_unused) | |
c773e847 DM |
4961 | { |
4962 | lockdep_set_class(&txq->_xmit_lock, | |
4963 | &bonding_netdev_xmit_lock_key); | |
4964 | } | |
4965 | ||
4966 | static void bond_set_lockdep_class(struct net_device *dev) | |
4967 | { | |
cf508b12 DM |
4968 | lockdep_set_class(&dev->addr_list_lock, |
4969 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4970 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
4971 | } |
4972 | ||
181470fc SH |
4973 | /* |
4974 | * Called from registration process | |
4975 | */ | |
4976 | static int bond_init(struct net_device *bond_dev) | |
4977 | { | |
4978 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4979 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc SH |
4980 | |
4981 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4982 | ||
4983 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4984 | if (!bond->wq) | |
4985 | return -ENOMEM; | |
4986 | ||
4987 | bond_set_lockdep_class(bond_dev); | |
4988 | ||
4989 | netif_carrier_off(bond_dev); | |
4990 | ||
4991 | bond_create_proc_entry(bond); | |
ec87fd3b | 4992 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4993 | |
6151b3d4 | 4994 | bond_prepare_sysfs_group(bond); |
22bedad3 JP |
4995 | |
4996 | __hw_addr_init(&bond->mc_list); | |
181470fc SH |
4997 | return 0; |
4998 | } | |
4999 | ||
88ead977 EB |
5000 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
5001 | { | |
5002 | if (tb[IFLA_ADDRESS]) { | |
5003 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
5004 | return -EINVAL; | |
5005 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
5006 | return -EADDRNOTAVAIL; | |
5007 | } | |
5008 | return 0; | |
5009 | } | |
5010 | ||
5011 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
5012 | .kind = "bond", | |
6639104b | 5013 | .priv_size = sizeof(struct bonding), |
88ead977 EB |
5014 | .setup = bond_setup, |
5015 | .validate = bond_validate, | |
5016 | }; | |
5017 | ||
dfe60397 | 5018 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5019 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5020 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5021 | * set up our sysfs entries. | |
5022 | */ | |
ec87fd3b | 5023 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
5024 | { |
5025 | struct net_device *bond_dev; | |
5026 | int res; | |
5027 | ||
5028 | rtnl_lock(); | |
027ea041 | 5029 | |
e4b91c48 | 5030 | bond_dev = alloc_netdev(sizeof(struct bonding), name ? name : "", |
181470fc | 5031 | bond_setup); |
dfe60397 | 5032 | if (!bond_dev) { |
a4aee5c8 | 5033 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
5034 | rtnl_unlock(); |
5035 | return -ENOMEM; | |
dfe60397 MW |
5036 | } |
5037 | ||
ec87fd3b | 5038 | dev_net_set(bond_dev, net); |
88ead977 EB |
5039 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5040 | ||
e4b91c48 JV |
5041 | if (!name) { |
5042 | res = dev_alloc_name(bond_dev, "bond%d"); | |
5043 | if (res < 0) | |
9e2e61fb | 5044 | goto out; |
e4b91c48 JV |
5045 | } |
5046 | ||
dfe60397 | 5047 | res = register_netdevice(bond_dev); |
0daa2303 | 5048 | |
30c15ba9 | 5049 | out: |
7e083840 | 5050 | rtnl_unlock(); |
9e2e61fb AW |
5051 | if (res < 0) |
5052 | bond_destructor(bond_dev); | |
30c15ba9 | 5053 | return res; |
dfe60397 MW |
5054 | } |
5055 | ||
2c8c1e72 | 5056 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 5057 | { |
15449745 | 5058 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5059 | |
5060 | bn->net = net; | |
5061 | INIT_LIST_HEAD(&bn->dev_list); | |
5062 | ||
ec87fd3b | 5063 | bond_create_proc_dir(bn); |
15449745 EB |
5064 | |
5065 | return 0; | |
ec87fd3b EB |
5066 | } |
5067 | ||
2c8c1e72 | 5068 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 5069 | { |
15449745 | 5070 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5071 | |
5072 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
5073 | } |
5074 | ||
5075 | static struct pernet_operations bond_net_ops = { | |
5076 | .init = bond_net_init, | |
5077 | .exit = bond_net_exit, | |
15449745 EB |
5078 | .id = &bond_net_id, |
5079 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
5080 | }; |
5081 | ||
1da177e4 LT |
5082 | static int __init bonding_init(void) |
5083 | { | |
1da177e4 LT |
5084 | int i; |
5085 | int res; | |
5086 | ||
3d632c3f | 5087 | pr_info("%s", version); |
1da177e4 | 5088 | |
dfe60397 | 5089 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5090 | if (res) |
dfe60397 | 5091 | goto out; |
1da177e4 | 5092 | |
15449745 | 5093 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
5094 | if (res) |
5095 | goto out; | |
027ea041 | 5096 | |
88ead977 EB |
5097 | res = rtnl_link_register(&bond_link_ops); |
5098 | if (res) | |
6639104b | 5099 | goto err_link; |
88ead977 | 5100 | |
1da177e4 | 5101 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5102 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5103 | if (res) |
5104 | goto err; | |
1da177e4 LT |
5105 | } |
5106 | ||
b76cdba9 MW |
5107 | res = bond_create_sysfs(); |
5108 | if (res) | |
5109 | goto err; | |
5110 | ||
1da177e4 | 5111 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5112 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5113 | bond_register_ipv6_notifier(); |
dfe60397 | 5114 | out: |
1da177e4 | 5115 | return res; |
88ead977 EB |
5116 | err: |
5117 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5118 | err_link: |
15449745 | 5119 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5120 | goto out; |
dfe60397 | 5121 | |
1da177e4 LT |
5122 | } |
5123 | ||
5124 | static void __exit bonding_exit(void) | |
5125 | { | |
5126 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5127 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5128 | bond_unregister_ipv6_notifier(); |
1da177e4 | 5129 | |
ae68c398 PE |
5130 | bond_destroy_sysfs(); |
5131 | ||
88ead977 | 5132 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5133 | unregister_pernet_subsys(&bond_net_ops); |
1da177e4 LT |
5134 | } |
5135 | ||
5136 | module_init(bonding_init); | |
5137 | module_exit(bonding_exit); | |
5138 | MODULE_LICENSE("GPL"); | |
5139 | MODULE_VERSION(DRV_VERSION); | |
5140 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5141 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5142 | MODULE_ALIAS_RTNL_LINK("bond"); |