]>
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 | ||
1da177e4 LT |
34 | #include <linux/kernel.h> |
35 | #include <linux/module.h> | |
1da177e4 LT |
36 | #include <linux/types.h> |
37 | #include <linux/fcntl.h> | |
38 | #include <linux/interrupt.h> | |
39 | #include <linux/ptrace.h> | |
40 | #include <linux/ioport.h> | |
41 | #include <linux/in.h> | |
169a3e66 | 42 | #include <net/ip.h> |
1da177e4 | 43 | #include <linux/ip.h> |
169a3e66 JV |
44 | #include <linux/tcp.h> |
45 | #include <linux/udp.h> | |
1da177e4 LT |
46 | #include <linux/slab.h> |
47 | #include <linux/string.h> | |
48 | #include <linux/init.h> | |
49 | #include <linux/timer.h> | |
50 | #include <linux/socket.h> | |
51 | #include <linux/ctype.h> | |
52 | #include <linux/inet.h> | |
53 | #include <linux/bitops.h> | |
3d632c3f | 54 | #include <linux/io.h> |
1da177e4 | 55 | #include <asm/system.h> |
1da177e4 | 56 | #include <asm/dma.h> |
3d632c3f | 57 | #include <linux/uaccess.h> |
1da177e4 LT |
58 | #include <linux/errno.h> |
59 | #include <linux/netdevice.h> | |
60 | #include <linux/inetdevice.h> | |
a816c7c7 | 61 | #include <linux/igmp.h> |
1da177e4 LT |
62 | #include <linux/etherdevice.h> |
63 | #include <linux/skbuff.h> | |
64 | #include <net/sock.h> | |
65 | #include <linux/rtnetlink.h> | |
66 | #include <linux/proc_fs.h> | |
67 | #include <linux/seq_file.h> | |
68 | #include <linux/smp.h> | |
69 | #include <linux/if_ether.h> | |
70 | #include <net/arp.h> | |
71 | #include <linux/mii.h> | |
72 | #include <linux/ethtool.h> | |
73 | #include <linux/if_vlan.h> | |
74 | #include <linux/if_bonding.h> | |
b63bb739 | 75 | #include <linux/jiffies.h> |
c3ade5ca | 76 | #include <net/route.h> |
457c4cbc | 77 | #include <net/net_namespace.h> |
1da177e4 LT |
78 | #include "bonding.h" |
79 | #include "bond_3ad.h" | |
80 | #include "bond_alb.h" | |
81 | ||
82 | /*---------------------------- Module parameters ----------------------------*/ | |
83 | ||
84 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
85 | #define BOND_LINK_MON_INTERV 0 | |
86 | #define BOND_LINK_ARP_INTERV 0 | |
87 | ||
88 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
7893b249 | 89 | static int num_grat_arp = 1; |
305d552a | 90 | static int num_unsol_na = 1; |
1da177e4 | 91 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
92 | static int updelay; |
93 | static int downdelay; | |
1da177e4 | 94 | static int use_carrier = 1; |
3d632c3f SH |
95 | static char *mode; |
96 | static char *primary; | |
a549952a | 97 | static char *primary_reselect; |
3d632c3f SH |
98 | static char *lacp_rate; |
99 | static char *ad_select; | |
100 | static char *xmit_hash_policy; | |
1da177e4 | 101 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
102 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
103 | static char *arp_validate; | |
104 | static char *fail_over_mac; | |
d2991f75 | 105 | static struct bond_params bonding_defaults; |
1da177e4 LT |
106 | |
107 | module_param(max_bonds, int, 0); | |
108 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
7893b249 MS |
109 | module_param(num_grat_arp, int, 0644); |
110 | MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packets to send on failover event"); | |
305d552a BH |
111 | module_param(num_unsol_na, int, 0644); |
112 | MODULE_PARM_DESC(num_unsol_na, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event"); | |
1da177e4 LT |
113 | module_param(miimon, int, 0); |
114 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
115 | module_param(updelay, int, 0); | |
116 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
117 | module_param(downdelay, int, 0); | |
2ac47660 MW |
118 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
119 | "in milliseconds"); | |
1da177e4 | 120 | module_param(use_carrier, int, 0); |
2ac47660 MW |
121 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
122 | "0 for off, 1 for on (default)"); | |
1da177e4 | 123 | module_param(mode, charp, 0); |
2ac47660 MW |
124 | MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, " |
125 | "1 for active-backup, 2 for balance-xor, " | |
126 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
127 | "6 for balance-alb"); | |
1da177e4 LT |
128 | module_param(primary, charp, 0); |
129 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
130 | module_param(primary_reselect, charp, 0); |
131 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
132 | "once it comes up; " | |
133 | "0 for always (default), " | |
134 | "1 for only if speed of primary is " | |
135 | "better, " | |
136 | "2 for only on active slave " | |
137 | "failure"); | |
1da177e4 | 138 | module_param(lacp_rate, charp, 0); |
2ac47660 MW |
139 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner " |
140 | "(slow/fast)"); | |
fd989c83 JV |
141 | module_param(ad_select, charp, 0); |
142 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)"); | |
169a3e66 | 143 | module_param(xmit_hash_policy, charp, 0); |
2ac47660 MW |
144 | MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)" |
145 | ", 1 for layer 3+4"); | |
1da177e4 LT |
146 | module_param(arp_interval, int, 0); |
147 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
148 | module_param_array(arp_ip_target, charp, NULL, 0); | |
149 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 JV |
150 | module_param(arp_validate, charp, 0); |
151 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all"); | |
3915c1e8 JV |
152 | module_param(fail_over_mac, charp, 0); |
153 | 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 |
154 | |
155 | /*----------------------------- Global variables ----------------------------*/ | |
156 | ||
f71e1309 | 157 | static const char * const version = |
1da177e4 LT |
158 | DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n"; |
159 | ||
12479f9a | 160 | LIST_HEAD(bond_dev_list); |
1da177e4 LT |
161 | |
162 | #ifdef CONFIG_PROC_FS | |
3d632c3f | 163 | static struct proc_dir_entry *bond_proc_dir; |
1da177e4 LT |
164 | #endif |
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", |
3d632c3f | 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 | |
5a03cdb7 | 309 | pr_debug("removed VLAN ID %d from bond %s\n", vlan_id, |
1da177e4 LT |
310 | bond->dev->name); |
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 | ||
5a03cdb7 | 329 | pr_debug("couldn't find VLAN ID %d in bond %s\n", vlan_id, |
1da177e4 LT |
330 | bond->dev->name); |
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", |
1da177e4 LT |
354 | slave->dev->name); |
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; | |
434 | dev_queue_xmit(skb); | |
435 | ||
436 | return 0; | |
437 | } | |
438 | ||
439 | /* | |
440 | * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid | |
441 | * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a | |
442 | * lock because: | |
443 | * a. This operation is performed in IOCTL context, | |
444 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
445 | * c. Holding a lock with BH disabled while directly calling a base driver | |
446 | * entry point is generally a BAD idea. | |
3d632c3f | 447 | * |
1da177e4 LT |
448 | * The design of synchronization/protection for this operation in the 8021q |
449 | * module is good for one or more VLAN devices over a single physical device | |
450 | * and cannot be extended for a teaming solution like bonding, so there is a | |
451 | * potential race condition here where a net device from the vlan group might | |
452 | * be referenced (either by a base driver or the 8021q code) while it is being | |
453 | * removed from the system. However, it turns out we're not making matters | |
454 | * worse, and if it works for regular VLAN usage it will work here too. | |
455 | */ | |
456 | ||
457 | /** | |
458 | * bond_vlan_rx_register - Propagates registration to slaves | |
459 | * @bond_dev: bonding net device that got called | |
460 | * @grp: vlan group being registered | |
461 | */ | |
3d632c3f SH |
462 | static void bond_vlan_rx_register(struct net_device *bond_dev, |
463 | struct vlan_group *grp) | |
1da177e4 | 464 | { |
454d7c9b | 465 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
466 | struct slave *slave; |
467 | int i; | |
468 | ||
469 | bond->vlgrp = grp; | |
470 | ||
471 | bond_for_each_slave(bond, slave, i) { | |
472 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 473 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
474 | |
475 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
476 | slave_ops->ndo_vlan_rx_register) { |
477 | slave_ops->ndo_vlan_rx_register(slave_dev, grp); | |
1da177e4 LT |
478 | } |
479 | } | |
480 | } | |
481 | ||
482 | /** | |
483 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
484 | * @bond_dev: bonding net device that got called | |
485 | * @vid: vlan id being added | |
486 | */ | |
487 | static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) | |
488 | { | |
454d7c9b | 489 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
490 | struct slave *slave; |
491 | int i, res; | |
492 | ||
493 | bond_for_each_slave(bond, slave, i) { | |
494 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 495 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
496 | |
497 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a SH |
498 | slave_ops->ndo_vlan_rx_add_vid) { |
499 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid); | |
1da177e4 LT |
500 | } |
501 | } | |
502 | ||
503 | res = bond_add_vlan(bond, vid); | |
504 | if (res) { | |
3d632c3f | 505 | pr_err(DRV_NAME |
4e0952c7 | 506 | ": %s: Error: Failed to add vlan id %d\n", |
1da177e4 LT |
507 | bond_dev->name, vid); |
508 | } | |
509 | } | |
510 | ||
511 | /** | |
512 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
513 | * @bond_dev: bonding net device that got called | |
514 | * @vid: vlan id being removed | |
515 | */ | |
516 | static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) | |
517 | { | |
454d7c9b | 518 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
519 | struct slave *slave; |
520 | struct net_device *vlan_dev; | |
521 | int i, res; | |
522 | ||
523 | bond_for_each_slave(bond, slave, i) { | |
524 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 525 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
526 | |
527 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a | 528 | slave_ops->ndo_vlan_rx_kill_vid) { |
1da177e4 LT |
529 | /* Save and then restore vlan_dev in the grp array, |
530 | * since the slave's driver might clear it. | |
531 | */ | |
5c15bdec | 532 | vlan_dev = vlan_group_get_device(bond->vlgrp, vid); |
eb7cc59a | 533 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid); |
5c15bdec | 534 | vlan_group_set_device(bond->vlgrp, vid, vlan_dev); |
1da177e4 LT |
535 | } |
536 | } | |
537 | ||
538 | res = bond_del_vlan(bond, vid); | |
539 | if (res) { | |
3d632c3f | 540 | pr_err(DRV_NAME |
4e0952c7 | 541 | ": %s: Error: Failed to remove vlan id %d\n", |
1da177e4 LT |
542 | bond_dev->name, vid); |
543 | } | |
544 | } | |
545 | ||
546 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
547 | { | |
548 | struct vlan_entry *vlan; | |
eb7cc59a | 549 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
550 | |
551 | write_lock_bh(&bond->lock); | |
552 | ||
eb7cc59a | 553 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 554 | goto out; |
1da177e4 LT |
555 | |
556 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
557 | slave_ops->ndo_vlan_rx_register) |
558 | slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp); | |
1da177e4 LT |
559 | |
560 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 561 | !(slave_ops->ndo_vlan_rx_add_vid)) |
1da177e4 | 562 | goto out; |
1da177e4 | 563 | |
eb7cc59a SH |
564 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) |
565 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id); | |
1da177e4 LT |
566 | |
567 | out: | |
568 | write_unlock_bh(&bond->lock); | |
569 | } | |
570 | ||
3d632c3f SH |
571 | static void bond_del_vlans_from_slave(struct bonding *bond, |
572 | struct net_device *slave_dev) | |
1da177e4 | 573 | { |
eb7cc59a | 574 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
575 | struct vlan_entry *vlan; |
576 | struct net_device *vlan_dev; | |
577 | ||
578 | write_lock_bh(&bond->lock); | |
579 | ||
eb7cc59a | 580 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 581 | goto out; |
1da177e4 LT |
582 | |
583 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 584 | !(slave_ops->ndo_vlan_rx_kill_vid)) |
1da177e4 | 585 | goto unreg; |
1da177e4 LT |
586 | |
587 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
588 | /* Save and then restore vlan_dev in the grp array, | |
589 | * since the slave's driver might clear it. | |
590 | */ | |
5c15bdec | 591 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
eb7cc59a | 592 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id); |
5c15bdec | 593 | vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev); |
1da177e4 LT |
594 | } |
595 | ||
596 | unreg: | |
597 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
598 | slave_ops->ndo_vlan_rx_register) |
599 | slave_ops->ndo_vlan_rx_register(slave_dev, NULL); | |
1da177e4 LT |
600 | |
601 | out: | |
602 | write_unlock_bh(&bond->lock); | |
603 | } | |
604 | ||
605 | /*------------------------------- Link status -------------------------------*/ | |
606 | ||
ff59c456 JV |
607 | /* |
608 | * Set the carrier state for the master according to the state of its | |
609 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
610 | * do special 802.3ad magic. | |
611 | * | |
612 | * Returns zero if carrier state does not change, nonzero if it does. | |
613 | */ | |
614 | static int bond_set_carrier(struct bonding *bond) | |
615 | { | |
616 | struct slave *slave; | |
617 | int i; | |
618 | ||
619 | if (bond->slave_cnt == 0) | |
620 | goto down; | |
621 | ||
622 | if (bond->params.mode == BOND_MODE_8023AD) | |
623 | return bond_3ad_set_carrier(bond); | |
624 | ||
625 | bond_for_each_slave(bond, slave, i) { | |
626 | if (slave->link == BOND_LINK_UP) { | |
627 | if (!netif_carrier_ok(bond->dev)) { | |
628 | netif_carrier_on(bond->dev); | |
629 | return 1; | |
630 | } | |
631 | return 0; | |
632 | } | |
633 | } | |
634 | ||
635 | down: | |
636 | if (netif_carrier_ok(bond->dev)) { | |
637 | netif_carrier_off(bond->dev); | |
638 | return 1; | |
639 | } | |
640 | return 0; | |
641 | } | |
642 | ||
1da177e4 LT |
643 | /* |
644 | * Get link speed and duplex from the slave's base driver | |
645 | * using ethtool. If for some reason the call fails or the | |
646 | * values are invalid, fake speed and duplex to 100/Full | |
647 | * and return error. | |
648 | */ | |
649 | static int bond_update_speed_duplex(struct slave *slave) | |
650 | { | |
651 | struct net_device *slave_dev = slave->dev; | |
1da177e4 | 652 | struct ethtool_cmd etool; |
61a44b9c | 653 | int res; |
1da177e4 LT |
654 | |
655 | /* Fake speed and duplex */ | |
656 | slave->speed = SPEED_100; | |
657 | slave->duplex = DUPLEX_FULL; | |
658 | ||
61a44b9c MW |
659 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
660 | return -1; | |
1da177e4 | 661 | |
61a44b9c MW |
662 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
663 | if (res < 0) | |
1da177e4 | 664 | return -1; |
1da177e4 | 665 | |
1da177e4 LT |
666 | switch (etool.speed) { |
667 | case SPEED_10: | |
668 | case SPEED_100: | |
669 | case SPEED_1000: | |
94dbffd5 | 670 | case SPEED_10000: |
1da177e4 LT |
671 | break; |
672 | default: | |
673 | return -1; | |
674 | } | |
675 | ||
676 | switch (etool.duplex) { | |
677 | case DUPLEX_FULL: | |
678 | case DUPLEX_HALF: | |
679 | break; | |
680 | default: | |
681 | return -1; | |
682 | } | |
683 | ||
684 | slave->speed = etool.speed; | |
685 | slave->duplex = etool.duplex; | |
686 | ||
687 | return 0; | |
688 | } | |
689 | ||
690 | /* | |
691 | * if <dev> supports MII link status reporting, check its link status. | |
692 | * | |
693 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 694 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
695 | * |
696 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
697 | * can't tell and just pretend it is), or 0, meaning that the link is | |
698 | * down. | |
699 | * | |
700 | * If reporting is non-zero, instead of faking link up, return -1 if | |
701 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
702 | * support them). If use_carrier is set, return whatever it says. | |
703 | * It'd be nice if there was a good way to tell if a driver supports | |
704 | * netif_carrier, but there really isn't. | |
705 | */ | |
3d632c3f SH |
706 | static int bond_check_dev_link(struct bonding *bond, |
707 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 708 | { |
eb7cc59a | 709 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 710 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
711 | struct ifreq ifr; |
712 | struct mii_ioctl_data *mii; | |
1da177e4 | 713 | |
6c988853 PG |
714 | if (!reporting && !netif_running(slave_dev)) |
715 | return 0; | |
716 | ||
eb7cc59a | 717 | if (bond->params.use_carrier) |
1da177e4 | 718 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 719 | |
29112f4e JP |
720 | /* Try to get link status using Ethtool first. */ |
721 | if (slave_dev->ethtool_ops) { | |
722 | if (slave_dev->ethtool_ops->get_link) { | |
723 | u32 link; | |
724 | ||
725 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
726 | ||
727 | return link ? BMSR_LSTATUS : 0; | |
728 | } | |
729 | } | |
730 | ||
3d632c3f | 731 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 732 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
733 | if (ioctl) { |
734 | /* TODO: set pointer to correct ioctl on a per team member */ | |
735 | /* bases to make this more efficient. that is, once */ | |
736 | /* we determine the correct ioctl, we will always */ | |
737 | /* call it and not the others for that team */ | |
738 | /* member. */ | |
739 | ||
740 | /* | |
741 | * We cannot assume that SIOCGMIIPHY will also read a | |
742 | * register; not all network drivers (e.g., e100) | |
743 | * support that. | |
744 | */ | |
745 | ||
746 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
747 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
748 | mii = if_mii(&ifr); | |
749 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
750 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
751 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
752 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
753 | } |
754 | } | |
755 | ||
1da177e4 LT |
756 | /* |
757 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 758 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
759 | * cannot report link status). If not reporting, pretend |
760 | * we're ok. | |
761 | */ | |
3d632c3f | 762 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
763 | } |
764 | ||
765 | /*----------------------------- Multicast list ------------------------------*/ | |
766 | ||
767 | /* | |
768 | * Returns 0 if dmi1 and dmi2 are the same, non-0 otherwise | |
769 | */ | |
3d632c3f SH |
770 | static inline int bond_is_dmi_same(const struct dev_mc_list *dmi1, |
771 | const struct dev_mc_list *dmi2) | |
1da177e4 LT |
772 | { |
773 | return memcmp(dmi1->dmi_addr, dmi2->dmi_addr, dmi1->dmi_addrlen) == 0 && | |
774 | dmi1->dmi_addrlen == dmi2->dmi_addrlen; | |
775 | } | |
776 | ||
777 | /* | |
778 | * returns dmi entry if found, NULL otherwise | |
779 | */ | |
3d632c3f SH |
780 | static struct dev_mc_list *bond_mc_list_find_dmi(struct dev_mc_list *dmi, |
781 | struct dev_mc_list *mc_list) | |
1da177e4 LT |
782 | { |
783 | struct dev_mc_list *idmi; | |
784 | ||
785 | for (idmi = mc_list; idmi; idmi = idmi->next) { | |
3d632c3f | 786 | if (bond_is_dmi_same(dmi, idmi)) |
1da177e4 | 787 | return idmi; |
1da177e4 LT |
788 | } |
789 | ||
790 | return NULL; | |
791 | } | |
792 | ||
793 | /* | |
794 | * Push the promiscuity flag down to appropriate slaves | |
795 | */ | |
7e1a1ac1 | 796 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 797 | { |
7e1a1ac1 | 798 | int err = 0; |
1da177e4 LT |
799 | if (USES_PRIMARY(bond->params.mode)) { |
800 | /* write lock already acquired */ | |
801 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
802 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
803 | inc); | |
1da177e4 LT |
804 | } |
805 | } else { | |
806 | struct slave *slave; | |
807 | int i; | |
808 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
809 | err = dev_set_promiscuity(slave->dev, inc); |
810 | if (err) | |
811 | return err; | |
1da177e4 LT |
812 | } |
813 | } | |
7e1a1ac1 | 814 | return err; |
1da177e4 LT |
815 | } |
816 | ||
817 | /* | |
818 | * Push the allmulti flag down to all slaves | |
819 | */ | |
7e1a1ac1 | 820 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 821 | { |
7e1a1ac1 | 822 | int err = 0; |
1da177e4 LT |
823 | if (USES_PRIMARY(bond->params.mode)) { |
824 | /* write lock already acquired */ | |
825 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
826 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
827 | inc); | |
1da177e4 LT |
828 | } |
829 | } else { | |
830 | struct slave *slave; | |
831 | int i; | |
832 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
833 | err = dev_set_allmulti(slave->dev, inc); |
834 | if (err) | |
835 | return err; | |
1da177e4 LT |
836 | } |
837 | } | |
7e1a1ac1 | 838 | return err; |
1da177e4 LT |
839 | } |
840 | ||
841 | /* | |
842 | * Add a Multicast address to slaves | |
843 | * according to mode | |
844 | */ | |
845 | static void bond_mc_add(struct bonding *bond, void *addr, int alen) | |
846 | { | |
847 | if (USES_PRIMARY(bond->params.mode)) { | |
848 | /* write lock already acquired */ | |
3d632c3f | 849 | if (bond->curr_active_slave) |
1da177e4 | 850 | dev_mc_add(bond->curr_active_slave->dev, addr, alen, 0); |
1da177e4 LT |
851 | } else { |
852 | struct slave *slave; | |
853 | int i; | |
3d632c3f SH |
854 | |
855 | bond_for_each_slave(bond, slave, i) | |
1da177e4 | 856 | dev_mc_add(slave->dev, addr, alen, 0); |
1da177e4 LT |
857 | } |
858 | } | |
859 | ||
860 | /* | |
861 | * Remove a multicast address from slave | |
862 | * according to mode | |
863 | */ | |
864 | static void bond_mc_delete(struct bonding *bond, void *addr, int alen) | |
865 | { | |
866 | if (USES_PRIMARY(bond->params.mode)) { | |
867 | /* write lock already acquired */ | |
3d632c3f SH |
868 | if (bond->curr_active_slave) |
869 | dev_mc_delete(bond->curr_active_slave->dev, addr, | |
870 | alen, 0); | |
1da177e4 LT |
871 | } else { |
872 | struct slave *slave; | |
873 | int i; | |
874 | bond_for_each_slave(bond, slave, i) { | |
875 | dev_mc_delete(slave->dev, addr, alen, 0); | |
876 | } | |
877 | } | |
878 | } | |
879 | ||
a816c7c7 JV |
880 | |
881 | /* | |
882 | * Retrieve the list of registered multicast addresses for the bonding | |
883 | * device and retransmit an IGMP JOIN request to the current active | |
884 | * slave. | |
885 | */ | |
886 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
887 | { | |
888 | struct in_device *in_dev; | |
889 | struct ip_mc_list *im; | |
890 | ||
891 | rcu_read_lock(); | |
892 | in_dev = __in_dev_get_rcu(bond->dev); | |
893 | if (in_dev) { | |
3d632c3f | 894 | for (im = in_dev->mc_list; im; im = im->next) |
a816c7c7 | 895 | ip_mc_rejoin_group(im); |
a816c7c7 JV |
896 | } |
897 | ||
898 | rcu_read_unlock(); | |
899 | } | |
900 | ||
1da177e4 LT |
901 | /* |
902 | * Totally destroys the mc_list in bond | |
903 | */ | |
904 | static void bond_mc_list_destroy(struct bonding *bond) | |
905 | { | |
906 | struct dev_mc_list *dmi; | |
907 | ||
908 | dmi = bond->mc_list; | |
909 | while (dmi) { | |
910 | bond->mc_list = dmi->next; | |
911 | kfree(dmi); | |
912 | dmi = bond->mc_list; | |
913 | } | |
3d632c3f SH |
914 | |
915 | bond->mc_list = NULL; | |
1da177e4 LT |
916 | } |
917 | ||
918 | /* | |
919 | * Copy all the Multicast addresses from src to the bonding device dst | |
920 | */ | |
de54f390 | 921 | static int bond_mc_list_copy(struct dev_mc_list *mc_list, struct bonding *bond, |
dd0fc66f | 922 | gfp_t gfp_flag) |
1da177e4 LT |
923 | { |
924 | struct dev_mc_list *dmi, *new_dmi; | |
925 | ||
926 | for (dmi = mc_list; dmi; dmi = dmi->next) { | |
de54f390 | 927 | new_dmi = kmalloc(sizeof(struct dev_mc_list), gfp_flag); |
1da177e4 LT |
928 | |
929 | if (!new_dmi) { | |
930 | /* FIXME: Potential memory leak !!! */ | |
931 | return -ENOMEM; | |
932 | } | |
933 | ||
934 | new_dmi->next = bond->mc_list; | |
935 | bond->mc_list = new_dmi; | |
936 | new_dmi->dmi_addrlen = dmi->dmi_addrlen; | |
937 | memcpy(new_dmi->dmi_addr, dmi->dmi_addr, dmi->dmi_addrlen); | |
938 | new_dmi->dmi_users = dmi->dmi_users; | |
939 | new_dmi->dmi_gusers = dmi->dmi_gusers; | |
940 | } | |
941 | ||
942 | return 0; | |
943 | } | |
944 | ||
945 | /* | |
946 | * flush all members of flush->mc_list from device dev->mc_list | |
947 | */ | |
3d632c3f SH |
948 | static void bond_mc_list_flush(struct net_device *bond_dev, |
949 | struct net_device *slave_dev) | |
1da177e4 | 950 | { |
454d7c9b | 951 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
952 | struct dev_mc_list *dmi; |
953 | ||
3d632c3f | 954 | for (dmi = bond_dev->mc_list; dmi; dmi = dmi->next) |
1da177e4 | 955 | dev_mc_delete(slave_dev, dmi->dmi_addr, dmi->dmi_addrlen, 0); |
1da177e4 LT |
956 | |
957 | if (bond->params.mode == BOND_MODE_8023AD) { | |
958 | /* del lacpdu mc addr from mc list */ | |
959 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
960 | ||
961 | dev_mc_delete(slave_dev, lacpdu_multicast, ETH_ALEN, 0); | |
962 | } | |
963 | } | |
964 | ||
965 | /*--------------------------- Active slave change ---------------------------*/ | |
966 | ||
967 | /* | |
968 | * Update the mc list and multicast-related flags for the new and | |
969 | * old active slaves (if any) according to the multicast mode, and | |
970 | * promiscuous flags unconditionally. | |
971 | */ | |
3d632c3f SH |
972 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
973 | struct slave *old_active) | |
1da177e4 LT |
974 | { |
975 | struct dev_mc_list *dmi; | |
976 | ||
3d632c3f | 977 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
978 | /* nothing to do - mc list is already up-to-date on |
979 | * all slaves | |
980 | */ | |
981 | return; | |
1da177e4 LT |
982 | |
983 | if (old_active) { | |
3d632c3f | 984 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 985 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 986 | |
3d632c3f | 987 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 988 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 989 | |
3d632c3f SH |
990 | for (dmi = bond->dev->mc_list; dmi; dmi = dmi->next) |
991 | dev_mc_delete(old_active->dev, dmi->dmi_addr, | |
992 | dmi->dmi_addrlen, 0); | |
1da177e4 LT |
993 | } |
994 | ||
995 | if (new_active) { | |
7e1a1ac1 | 996 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 997 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 998 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 999 | |
3d632c3f | 1000 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 1001 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 1002 | |
3d632c3f SH |
1003 | for (dmi = bond->dev->mc_list; dmi; dmi = dmi->next) |
1004 | dev_mc_add(new_active->dev, dmi->dmi_addr, | |
1005 | dmi->dmi_addrlen, 0); | |
a816c7c7 | 1006 | bond_resend_igmp_join_requests(bond); |
1da177e4 LT |
1007 | } |
1008 | } | |
1009 | ||
3915c1e8 JV |
1010 | /* |
1011 | * bond_do_fail_over_mac | |
1012 | * | |
1013 | * Perform special MAC address swapping for fail_over_mac settings | |
1014 | * | |
1015 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
1016 | */ | |
1017 | static void bond_do_fail_over_mac(struct bonding *bond, | |
1018 | struct slave *new_active, | |
1019 | struct slave *old_active) | |
1f78d9f9 HE |
1020 | __releases(&bond->curr_slave_lock) |
1021 | __releases(&bond->lock) | |
1022 | __acquires(&bond->lock) | |
1023 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
1024 | { |
1025 | u8 tmp_mac[ETH_ALEN]; | |
1026 | struct sockaddr saddr; | |
1027 | int rv; | |
1028 | ||
1029 | switch (bond->params.fail_over_mac) { | |
1030 | case BOND_FOM_ACTIVE: | |
1031 | if (new_active) | |
1032 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
1033 | new_active->dev->addr_len); | |
1034 | break; | |
1035 | case BOND_FOM_FOLLOW: | |
1036 | /* | |
1037 | * if new_active && old_active, swap them | |
1038 | * if just old_active, do nothing (going to no active slave) | |
1039 | * if just new_active, set new_active to bond's MAC | |
1040 | */ | |
1041 | if (!new_active) | |
1042 | return; | |
1043 | ||
1044 | write_unlock_bh(&bond->curr_slave_lock); | |
1045 | read_unlock(&bond->lock); | |
1046 | ||
1047 | if (old_active) { | |
1048 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
1049 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
1050 | ETH_ALEN); | |
1051 | saddr.sa_family = new_active->dev->type; | |
1052 | } else { | |
1053 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
1054 | saddr.sa_family = bond->dev->type; | |
1055 | } | |
1056 | ||
1057 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
1058 | if (rv) { | |
3d632c3f | 1059 | pr_err(DRV_NAME |
3915c1e8 JV |
1060 | ": %s: Error %d setting MAC of slave %s\n", |
1061 | bond->dev->name, -rv, new_active->dev->name); | |
1062 | goto out; | |
1063 | } | |
1064 | ||
1065 | if (!old_active) | |
1066 | goto out; | |
1067 | ||
1068 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1069 | saddr.sa_family = old_active->dev->type; | |
1070 | ||
1071 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1072 | if (rv) | |
3d632c3f | 1073 | pr_err(DRV_NAME |
3915c1e8 JV |
1074 | ": %s: Error %d setting MAC of slave %s\n", |
1075 | bond->dev->name, -rv, new_active->dev->name); | |
1076 | out: | |
1077 | read_lock(&bond->lock); | |
1078 | write_lock_bh(&bond->curr_slave_lock); | |
1079 | break; | |
1080 | default: | |
3d632c3f | 1081 | pr_err(DRV_NAME |
3915c1e8 JV |
1082 | ": %s: bond_do_fail_over_mac impossible: bad policy %d\n", |
1083 | bond->dev->name, bond->params.fail_over_mac); | |
1084 | break; | |
1085 | } | |
1086 | ||
1087 | } | |
1088 | ||
a549952a JP |
1089 | static bool bond_should_change_active(struct bonding *bond) |
1090 | { | |
1091 | struct slave *prim = bond->primary_slave; | |
1092 | struct slave *curr = bond->curr_active_slave; | |
1093 | ||
1094 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1095 | return true; | |
1096 | if (bond->force_primary) { | |
1097 | bond->force_primary = false; | |
1098 | return true; | |
1099 | } | |
1100 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1101 | (prim->speed < curr->speed || | |
1102 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1103 | return false; | |
1104 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1105 | return false; | |
1106 | return true; | |
1107 | } | |
3915c1e8 | 1108 | |
1da177e4 LT |
1109 | /** |
1110 | * find_best_interface - select the best available slave to be the active one | |
1111 | * @bond: our bonding struct | |
1112 | * | |
1113 | * Warning: Caller must hold curr_slave_lock for writing. | |
1114 | */ | |
1115 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1116 | { | |
1117 | struct slave *new_active, *old_active; | |
1118 | struct slave *bestslave = NULL; | |
1119 | int mintime = bond->params.updelay; | |
1120 | int i; | |
1121 | ||
49b4ad92 | 1122 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1123 | |
1124 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1125 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1126 | new_active = bond->first_slave; |
3d632c3f | 1127 | else |
1da177e4 | 1128 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1129 | } |
1130 | ||
1da177e4 | 1131 | if ((bond->primary_slave) && |
a549952a JP |
1132 | bond->primary_slave->link == BOND_LINK_UP && |
1133 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1134 | new_active = bond->primary_slave; |
1135 | } | |
1136 | ||
1137 | /* remember where to stop iterating over the slaves */ | |
1138 | old_active = new_active; | |
1139 | ||
1140 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1141 | if (new_active->link == BOND_LINK_UP) { |
1142 | return new_active; | |
1143 | } else if (new_active->link == BOND_LINK_BACK && | |
1144 | IS_UP(new_active->dev)) { | |
1145 | /* link up, but waiting for stabilization */ | |
1146 | if (new_active->delay < mintime) { | |
1147 | mintime = new_active->delay; | |
1148 | bestslave = new_active; | |
1da177e4 LT |
1149 | } |
1150 | } | |
1151 | } | |
1152 | ||
1153 | return bestslave; | |
1154 | } | |
1155 | ||
1156 | /** | |
1157 | * change_active_interface - change the active slave into the specified one | |
1158 | * @bond: our bonding struct | |
1159 | * @new: the new slave to make the active one | |
1160 | * | |
1161 | * Set the new slave to the bond's settings and unset them on the old | |
1162 | * curr_active_slave. | |
1163 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1164 | * | |
1165 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1166 | * because it is apparently the best available slave we have, even though its | |
1167 | * updelay hasn't timed out yet. | |
1168 | * | |
3915c1e8 JV |
1169 | * If new_active is not NULL, caller must hold bond->lock for read and |
1170 | * curr_slave_lock for write_bh. | |
1da177e4 | 1171 | */ |
a77b5325 | 1172 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1173 | { |
1174 | struct slave *old_active = bond->curr_active_slave; | |
1175 | ||
3d632c3f | 1176 | if (old_active == new_active) |
1da177e4 | 1177 | return; |
1da177e4 LT |
1178 | |
1179 | if (new_active) { | |
b2220cad JV |
1180 | new_active->jiffies = jiffies; |
1181 | ||
1da177e4 LT |
1182 | if (new_active->link == BOND_LINK_BACK) { |
1183 | if (USES_PRIMARY(bond->params.mode)) { | |
3d632c3f | 1184 | pr_info(DRV_NAME |
1da177e4 LT |
1185 | ": %s: making interface %s the new " |
1186 | "active one %d ms earlier.\n", | |
1187 | bond->dev->name, new_active->dev->name, | |
1188 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1189 | } | |
1190 | ||
1191 | new_active->delay = 0; | |
1192 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1193 | |
3d632c3f | 1194 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1195 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1196 | |
58402054 | 1197 | if (bond_is_lb(bond)) |
1da177e4 | 1198 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1199 | } else { |
1200 | if (USES_PRIMARY(bond->params.mode)) { | |
3d632c3f | 1201 | pr_info(DRV_NAME |
1da177e4 LT |
1202 | ": %s: making interface %s the new " |
1203 | "active one.\n", | |
1204 | bond->dev->name, new_active->dev->name); | |
1205 | } | |
1206 | } | |
1207 | } | |
1208 | ||
3d632c3f | 1209 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1210 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1211 | |
58402054 | 1212 | if (bond_is_lb(bond)) { |
1da177e4 | 1213 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1214 | if (old_active) |
1215 | bond_set_slave_inactive_flags(old_active); | |
1216 | if (new_active) | |
1217 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1218 | } else { |
1219 | bond->curr_active_slave = new_active; | |
1220 | } | |
c3ade5ca JV |
1221 | |
1222 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1223 | if (old_active) |
c3ade5ca | 1224 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1225 | |
1226 | if (new_active) { | |
1227 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1228 | |
709f8a45 OG |
1229 | if (bond->params.fail_over_mac) |
1230 | bond_do_fail_over_mac(bond, new_active, | |
1231 | old_active); | |
3915c1e8 | 1232 | |
709f8a45 | 1233 | bond->send_grat_arp = bond->params.num_grat_arp; |
b59f9f74 | 1234 | bond_send_gratuitous_arp(bond); |
01f3109d | 1235 | |
305d552a BH |
1236 | bond->send_unsol_na = bond->params.num_unsol_na; |
1237 | bond_send_unsolicited_na(bond); | |
1238 | ||
01f3109d OG |
1239 | write_unlock_bh(&bond->curr_slave_lock); |
1240 | read_unlock(&bond->lock); | |
1241 | ||
75c78500 | 1242 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1243 | |
1244 | read_lock(&bond->lock); | |
1245 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1246 | } |
c3ade5ca | 1247 | } |
1da177e4 LT |
1248 | } |
1249 | ||
1250 | /** | |
1251 | * bond_select_active_slave - select a new active slave, if needed | |
1252 | * @bond: our bonding struct | |
1253 | * | |
3d632c3f | 1254 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1255 | * - The old curr_active_slave has been released or lost its link. |
1256 | * - The primary_slave has got its link back. | |
1257 | * - A slave has got its link back and there's no old curr_active_slave. | |
1258 | * | |
3915c1e8 | 1259 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1260 | */ |
a77b5325 | 1261 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1262 | { |
1263 | struct slave *best_slave; | |
ff59c456 | 1264 | int rv; |
1da177e4 LT |
1265 | |
1266 | best_slave = bond_find_best_slave(bond); | |
1267 | if (best_slave != bond->curr_active_slave) { | |
1268 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1269 | rv = bond_set_carrier(bond); |
1270 | if (!rv) | |
1271 | return; | |
1272 | ||
1273 | if (netif_carrier_ok(bond->dev)) { | |
3d632c3f | 1274 | pr_info(DRV_NAME |
ff59c456 JV |
1275 | ": %s: first active interface up!\n", |
1276 | bond->dev->name); | |
1277 | } else { | |
3d632c3f | 1278 | pr_info(DRV_NAME ": %s: " |
ff59c456 JV |
1279 | "now running without any active interface !\n", |
1280 | bond->dev->name); | |
1281 | } | |
1da177e4 LT |
1282 | } |
1283 | } | |
1284 | ||
1285 | /*--------------------------- slave list handling ---------------------------*/ | |
1286 | ||
1287 | /* | |
1288 | * This function attaches the slave to the end of list. | |
1289 | * | |
1290 | * bond->lock held for writing by caller. | |
1291 | */ | |
1292 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1293 | { | |
1294 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1295 | new_slave->next = new_slave; | |
1296 | new_slave->prev = new_slave; | |
1297 | bond->first_slave = new_slave; | |
1298 | } else { | |
1299 | new_slave->next = bond->first_slave; | |
1300 | new_slave->prev = bond->first_slave->prev; | |
1301 | new_slave->next->prev = new_slave; | |
1302 | new_slave->prev->next = new_slave; | |
1303 | } | |
1304 | ||
1305 | bond->slave_cnt++; | |
1306 | } | |
1307 | ||
1308 | /* | |
1309 | * This function detaches the slave from the list. | |
1310 | * WARNING: no check is made to verify if the slave effectively | |
1311 | * belongs to <bond>. | |
1312 | * Nothing is freed on return, structures are just unchained. | |
1313 | * If any slave pointer in bond was pointing to <slave>, | |
1314 | * it should be changed by the calling function. | |
1315 | * | |
1316 | * bond->lock held for writing by caller. | |
1317 | */ | |
1318 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1319 | { | |
3d632c3f | 1320 | if (slave->next) |
1da177e4 | 1321 | slave->next->prev = slave->prev; |
1da177e4 | 1322 | |
3d632c3f | 1323 | if (slave->prev) |
1da177e4 | 1324 | slave->prev->next = slave->next; |
1da177e4 LT |
1325 | |
1326 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1327 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1328 | bond->first_slave = slave->next; | |
1329 | } else { | |
1330 | bond->first_slave = NULL; /* slave was the last one */ | |
1331 | } | |
1332 | } | |
1333 | ||
1334 | slave->next = NULL; | |
1335 | slave->prev = NULL; | |
1336 | bond->slave_cnt--; | |
1337 | } | |
1338 | ||
1339 | /*---------------------------------- IOCTL ----------------------------------*/ | |
1340 | ||
4ad072c9 AB |
1341 | static int bond_sethwaddr(struct net_device *bond_dev, |
1342 | struct net_device *slave_dev) | |
1da177e4 | 1343 | { |
5a03cdb7 HE |
1344 | pr_debug("bond_dev=%p\n", bond_dev); |
1345 | pr_debug("slave_dev=%p\n", slave_dev); | |
1346 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1347 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1348 | return 0; | |
1349 | } | |
1350 | ||
7f353bf2 HX |
1351 | #define BOND_VLAN_FEATURES \ |
1352 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1353 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1354 | |
3d632c3f | 1355 | /* |
8e3babcd | 1356 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1357 | * feature bits are managed elsewhere, so preserve those feature bits |
1358 | * on the master device. | |
8531c5ff AK |
1359 | */ |
1360 | static int bond_compute_features(struct bonding *bond) | |
1361 | { | |
8531c5ff AK |
1362 | struct slave *slave; |
1363 | struct net_device *bond_dev = bond->dev; | |
7f353bf2 | 1364 | unsigned long features = bond_dev->features; |
278339a4 | 1365 | unsigned long vlan_features = 0; |
3158bf7d MS |
1366 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1367 | bond_dev->hard_header_len); | |
8e3babcd | 1368 | int i; |
8531c5ff | 1369 | |
7f353bf2 | 1370 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
b63365a2 HX |
1371 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM; |
1372 | ||
1373 | if (!bond->first_slave) | |
1374 | goto done; | |
1375 | ||
1376 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1377 | |
278339a4 | 1378 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1379 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1380 | features = netdev_increment_features(features, |
1381 | slave->dev->features, | |
1382 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1383 | vlan_features = netdev_increment_features(vlan_features, |
1384 | slave->dev->vlan_features, | |
1385 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1386 | if (slave->dev->hard_header_len > max_hard_header_len) |
1387 | max_hard_header_len = slave->dev->hard_header_len; | |
1388 | } | |
8531c5ff | 1389 | |
b63365a2 | 1390 | done: |
7f353bf2 | 1391 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
b63365a2 | 1392 | bond_dev->features = netdev_fix_features(features, NULL); |
278339a4 | 1393 | bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL); |
54ef3137 | 1394 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1395 | |
1396 | return 0; | |
1397 | } | |
1398 | ||
872254dd MS |
1399 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1400 | struct net_device *slave_dev) | |
1401 | { | |
454d7c9b | 1402 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1403 | |
00829823 | 1404 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1405 | |
1406 | bond_dev->type = slave_dev->type; | |
1407 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1408 | bond_dev->addr_len = slave_dev->addr_len; | |
1409 | ||
1410 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1411 | slave_dev->addr_len); | |
d90a162a | 1412 | bond->setup_by_slave = 1; |
872254dd MS |
1413 | } |
1414 | ||
1da177e4 | 1415 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1416 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1417 | { |
454d7c9b | 1418 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1419 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
1420 | struct slave *new_slave = NULL; |
1421 | struct dev_mc_list *dmi; | |
1422 | struct sockaddr addr; | |
1423 | int link_reporting; | |
1424 | int old_features = bond_dev->features; | |
1425 | int res = 0; | |
1426 | ||
552709d5 | 1427 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1428 | slave_ops->ndo_do_ioctl == NULL) { |
3d632c3f | 1429 | pr_warning(DRV_NAME |
4e0952c7 MW |
1430 | ": %s: Warning: no link monitoring support for %s\n", |
1431 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1432 | } |
1433 | ||
1434 | /* bond must be initialized by bond_open() before enslaving */ | |
1435 | if (!(bond_dev->flags & IFF_UP)) { | |
3d632c3f | 1436 | pr_warning(DRV_NAME |
6b1bf096 MS |
1437 | " %s: master_dev is not up in bond_enslave\n", |
1438 | bond_dev->name); | |
1da177e4 LT |
1439 | } |
1440 | ||
1441 | /* already enslaved */ | |
1442 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1443 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1444 | return -EBUSY; |
1445 | } | |
1446 | ||
1447 | /* vlan challenged mutual exclusion */ | |
1448 | /* no need to lock since we're protected by rtnl_lock */ | |
1449 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1450 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 | 1451 | if (!list_empty(&bond->vlan_list)) { |
3d632c3f | 1452 | pr_err(DRV_NAME |
4e0952c7 | 1453 | ": %s: Error: cannot enslave VLAN " |
1da177e4 | 1454 | "challenged slave %s on VLAN enabled " |
4e0952c7 | 1455 | "bond %s\n", bond_dev->name, slave_dev->name, |
1da177e4 LT |
1456 | bond_dev->name); |
1457 | return -EPERM; | |
1458 | } else { | |
3d632c3f | 1459 | pr_warning(DRV_NAME |
4e0952c7 | 1460 | ": %s: Warning: enslaved VLAN challenged " |
1da177e4 LT |
1461 | "slave %s. Adding VLANs will be blocked as " |
1462 | "long as %s is part of bond %s\n", | |
4e0952c7 | 1463 | bond_dev->name, slave_dev->name, slave_dev->name, |
1da177e4 LT |
1464 | bond_dev->name); |
1465 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
1466 | } | |
1467 | } else { | |
5a03cdb7 | 1468 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1469 | if (bond->slave_cnt == 0) { |
1470 | /* First slave, and it is not VLAN challenged, | |
1471 | * so remove the block of adding VLANs over the bond. | |
1472 | */ | |
1473 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1474 | } | |
1475 | } | |
1476 | ||
217df670 JV |
1477 | /* |
1478 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1479 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1480 | * the current ifenslave will set the interface down prior to |
1481 | * enslaving it; the old ifenslave will not. | |
1482 | */ | |
1483 | if ((slave_dev->flags & IFF_UP)) { | |
3d632c3f | 1484 | pr_err(DRV_NAME ": %s is up. " |
217df670 JV |
1485 | "This may be due to an out of date ifenslave.\n", |
1486 | slave_dev->name); | |
1487 | res = -EPERM; | |
1488 | goto err_undo_flags; | |
1489 | } | |
1da177e4 | 1490 | |
872254dd MS |
1491 | /* set bonding device ether type by slave - bonding netdevices are |
1492 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1493 | * there is a need to override some of the type dependent attribs/funcs. | |
1494 | * | |
1495 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1496 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1497 | */ | |
1498 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1499 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 MS |
1500 | pr_debug("%s: change device type from %d to %d\n", |
1501 | bond_dev->name, bond_dev->type, slave_dev->type); | |
75c78500 MS |
1502 | |
1503 | netdev_bonding_change(bond_dev, NETDEV_BONDING_OLDTYPE); | |
1504 | ||
e36b9d16 MS |
1505 | if (slave_dev->type != ARPHRD_ETHER) |
1506 | bond_setup_by_slave(bond_dev, slave_dev); | |
1507 | else | |
1508 | ether_setup(bond_dev); | |
75c78500 MS |
1509 | |
1510 | netdev_bonding_change(bond_dev, NETDEV_BONDING_NEWTYPE); | |
e36b9d16 | 1511 | } |
872254dd | 1512 | } else if (bond_dev->type != slave_dev->type) { |
3d632c3f | 1513 | pr_err(DRV_NAME ": %s ether type (%d) is different " |
872254dd MS |
1514 | "from other slaves (%d), can not enslave it.\n", |
1515 | slave_dev->name, | |
1516 | slave_dev->type, bond_dev->type); | |
1517 | res = -EINVAL; | |
1518 | goto err_undo_flags; | |
1519 | } | |
1520 | ||
eb7cc59a | 1521 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1522 | if (bond->slave_cnt == 0) { |
3d632c3f | 1523 | pr_warning(DRV_NAME |
dd957c57 JV |
1524 | ": %s: Warning: The first slave device " |
1525 | "specified does not support setting the MAC " | |
3915c1e8 | 1526 | "address. Setting fail_over_mac to active.", |
dd957c57 | 1527 | bond_dev->name); |
3915c1e8 JV |
1528 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1529 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
3d632c3f | 1530 | pr_err(DRV_NAME |
dd957c57 JV |
1531 | ": %s: Error: The slave device specified " |
1532 | "does not support setting the MAC address, " | |
3915c1e8 | 1533 | "but fail_over_mac is not set to active.\n" |
2ab82852 MS |
1534 | , bond_dev->name); |
1535 | res = -EOPNOTSUPP; | |
1536 | goto err_undo_flags; | |
1537 | } | |
1da177e4 LT |
1538 | } |
1539 | ||
243cb4e5 | 1540 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1541 | if (!new_slave) { |
1542 | res = -ENOMEM; | |
1543 | goto err_undo_flags; | |
1544 | } | |
1545 | ||
1da177e4 LT |
1546 | /* save slave's original flags before calling |
1547 | * netdev_set_master and dev_open | |
1548 | */ | |
1549 | new_slave->original_flags = slave_dev->flags; | |
1550 | ||
217df670 JV |
1551 | /* |
1552 | * Save slave's original ("permanent") mac address for modes | |
1553 | * that need it, and for restoring it upon release, and then | |
1554 | * set it to the master's address | |
1555 | */ | |
1556 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1557 | |
dd957c57 | 1558 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1559 | /* |
1560 | * Set slave to master's mac address. The application already | |
1561 | * set the master's mac address to that of the first slave | |
1562 | */ | |
1563 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1564 | addr.sa_family = slave_dev->type; | |
1565 | res = dev_set_mac_address(slave_dev, &addr); | |
1566 | if (res) { | |
5a03cdb7 | 1567 | pr_debug("Error %d calling set_mac_address\n", res); |
2ab82852 MS |
1568 | goto err_free; |
1569 | } | |
217df670 | 1570 | } |
1da177e4 | 1571 | |
c2edacf8 JV |
1572 | res = netdev_set_master(slave_dev, bond_dev); |
1573 | if (res) { | |
5a03cdb7 | 1574 | pr_debug("Error %d calling netdev_set_master\n", res); |
569f0c4d | 1575 | goto err_restore_mac; |
c2edacf8 | 1576 | } |
217df670 JV |
1577 | /* open the slave since the application closed it */ |
1578 | res = dev_open(slave_dev); | |
1579 | if (res) { | |
3d632c3f | 1580 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
569f0c4d | 1581 | goto err_unset_master; |
1da177e4 LT |
1582 | } |
1583 | ||
1da177e4 | 1584 | new_slave->dev = slave_dev; |
0b680e75 | 1585 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1586 | |
58402054 | 1587 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1588 | /* bond_alb_init_slave() must be called before all other stages since |
1589 | * it might fail and we do not want to have to undo everything | |
1590 | */ | |
1591 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1592 | if (res) |
569f0c4d | 1593 | goto err_close; |
1da177e4 LT |
1594 | } |
1595 | ||
1596 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1597 | * master's promisc (and mc) settings only if it becomes the | |
1598 | * curr_active_slave, and that is taken care of later when calling | |
1599 | * bond_change_active() | |
1600 | */ | |
1601 | if (!USES_PRIMARY(bond->params.mode)) { | |
1602 | /* set promiscuity level to new slave */ | |
1603 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1604 | res = dev_set_promiscuity(slave_dev, 1); |
1605 | if (res) | |
1606 | goto err_close; | |
1da177e4 LT |
1607 | } |
1608 | ||
1609 | /* set allmulti level to new slave */ | |
1610 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1611 | res = dev_set_allmulti(slave_dev, 1); |
1612 | if (res) | |
1613 | goto err_close; | |
1da177e4 LT |
1614 | } |
1615 | ||
b9e40857 | 1616 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1617 | /* upload master's mc_list to new slave */ |
3d632c3f SH |
1618 | for (dmi = bond_dev->mc_list; dmi; dmi = dmi->next) |
1619 | dev_mc_add(slave_dev, dmi->dmi_addr, | |
1620 | dmi->dmi_addrlen, 0); | |
b9e40857 | 1621 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1622 | } |
1623 | ||
1624 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1625 | /* add lacpdu mc addr to mc list */ | |
1626 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1627 | ||
1628 | dev_mc_add(slave_dev, lacpdu_multicast, ETH_ALEN, 0); | |
1629 | } | |
1630 | ||
1631 | bond_add_vlans_on_slave(bond, slave_dev); | |
1632 | ||
1633 | write_lock_bh(&bond->lock); | |
1634 | ||
1635 | bond_attach_slave(bond, new_slave); | |
1636 | ||
1637 | new_slave->delay = 0; | |
1638 | new_slave->link_failure_count = 0; | |
1639 | ||
8531c5ff AK |
1640 | bond_compute_features(bond); |
1641 | ||
3915c1e8 JV |
1642 | write_unlock_bh(&bond->lock); |
1643 | ||
1644 | read_lock(&bond->lock); | |
1645 | ||
f5b2b966 JV |
1646 | new_slave->last_arp_rx = jiffies; |
1647 | ||
1da177e4 LT |
1648 | if (bond->params.miimon && !bond->params.use_carrier) { |
1649 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1650 | ||
1651 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1652 | /* | |
1653 | * miimon is set but a bonded network driver | |
1654 | * does not support ETHTOOL/MII and | |
1655 | * arp_interval is not set. Note: if | |
1656 | * use_carrier is enabled, we will never go | |
1657 | * here (because netif_carrier is always | |
1658 | * supported); thus, we don't need to change | |
1659 | * the messages for netif_carrier. | |
1660 | */ | |
3d632c3f | 1661 | pr_warning(DRV_NAME |
4e0952c7 | 1662 | ": %s: Warning: MII and ETHTOOL support not " |
1da177e4 LT |
1663 | "available for interface %s, and " |
1664 | "arp_interval/arp_ip_target module parameters " | |
1665 | "not specified, thus bonding will not detect " | |
1666 | "link failures! see bonding.txt for details.\n", | |
4e0952c7 | 1667 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1668 | } else if (link_reporting == -1) { |
1669 | /* unable get link status using mii/ethtool */ | |
3d632c3f | 1670 | pr_warning(DRV_NAME |
4e0952c7 | 1671 | ": %s: Warning: can't get link status from " |
1da177e4 LT |
1672 | "interface %s; the network driver associated " |
1673 | "with this interface does not support MII or " | |
1674 | "ETHTOOL link status reporting, thus miimon " | |
1675 | "has no effect on this interface.\n", | |
4e0952c7 | 1676 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1677 | } |
1678 | } | |
1679 | ||
1680 | /* check for initial state */ | |
1681 | if (!bond->params.miimon || | |
1682 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1683 | if (bond->params.updelay) { | |
5a03cdb7 | 1684 | pr_debug("Initial state of slave_dev is " |
1da177e4 LT |
1685 | "BOND_LINK_BACK\n"); |
1686 | new_slave->link = BOND_LINK_BACK; | |
1687 | new_slave->delay = bond->params.updelay; | |
1688 | } else { | |
5a03cdb7 | 1689 | pr_debug("Initial state of slave_dev is " |
1da177e4 LT |
1690 | "BOND_LINK_UP\n"); |
1691 | new_slave->link = BOND_LINK_UP; | |
1692 | } | |
1693 | new_slave->jiffies = jiffies; | |
1694 | } else { | |
5a03cdb7 | 1695 | pr_debug("Initial state of slave_dev is " |
1da177e4 LT |
1696 | "BOND_LINK_DOWN\n"); |
1697 | new_slave->link = BOND_LINK_DOWN; | |
1698 | } | |
1699 | ||
1700 | if (bond_update_speed_duplex(new_slave) && | |
1701 | (new_slave->link != BOND_LINK_DOWN)) { | |
3d632c3f | 1702 | pr_warning(DRV_NAME |
4e0952c7 | 1703 | ": %s: Warning: failed to get speed and duplex from %s, " |
1da177e4 | 1704 | "assumed to be 100Mb/sec and Full.\n", |
4e0952c7 | 1705 | bond_dev->name, new_slave->dev->name); |
1da177e4 LT |
1706 | |
1707 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3d632c3f | 1708 | pr_warning(DRV_NAME |
4e0952c7 | 1709 | ": %s: Warning: Operation of 802.3ad mode requires ETHTOOL " |
1da177e4 | 1710 | "support in base driver for proper aggregator " |
4e0952c7 | 1711 | "selection.\n", bond_dev->name); |
1da177e4 LT |
1712 | } |
1713 | } | |
1714 | ||
1715 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1716 | /* if there is a primary slave, remember it */ | |
a549952a | 1717 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1718 | bond->primary_slave = new_slave; |
a549952a JP |
1719 | bond->force_primary = true; |
1720 | } | |
1da177e4 LT |
1721 | } |
1722 | ||
3915c1e8 JV |
1723 | write_lock_bh(&bond->curr_slave_lock); |
1724 | ||
1da177e4 LT |
1725 | switch (bond->params.mode) { |
1726 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1727 | bond_set_slave_inactive_flags(new_slave); |
1728 | bond_select_active_slave(bond); | |
1da177e4 LT |
1729 | break; |
1730 | case BOND_MODE_8023AD: | |
1731 | /* in 802.3ad mode, the internal mechanism | |
1732 | * will activate the slaves in the selected | |
1733 | * aggregator | |
1734 | */ | |
1735 | bond_set_slave_inactive_flags(new_slave); | |
1736 | /* if this is the first slave */ | |
1737 | if (bond->slave_cnt == 1) { | |
1738 | SLAVE_AD_INFO(new_slave).id = 1; | |
1739 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1740 | * can be called only after the mac address of the bond is set | |
1741 | */ | |
1742 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1743 | bond->params.lacp_fast); | |
1744 | } else { | |
1745 | SLAVE_AD_INFO(new_slave).id = | |
1746 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1747 | } | |
1748 | ||
1749 | bond_3ad_bind_slave(new_slave); | |
1750 | break; | |
1751 | case BOND_MODE_TLB: | |
1752 | case BOND_MODE_ALB: | |
1753 | new_slave->state = BOND_STATE_ACTIVE; | |
059fe7a5 | 1754 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1755 | bond_select_active_slave(bond); |
1da177e4 LT |
1756 | break; |
1757 | default: | |
5a03cdb7 | 1758 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1759 | |
1760 | /* always active in trunk mode */ | |
1761 | new_slave->state = BOND_STATE_ACTIVE; | |
1762 | ||
1763 | /* In trunking mode there is little meaning to curr_active_slave | |
1764 | * anyway (it holds no special properties of the bond device), | |
1765 | * so we can change it without calling change_active_interface() | |
1766 | */ | |
3d632c3f | 1767 | if (!bond->curr_active_slave) |
1da177e4 | 1768 | bond->curr_active_slave = new_slave; |
3d632c3f | 1769 | |
1da177e4 LT |
1770 | break; |
1771 | } /* switch(bond_mode) */ | |
1772 | ||
3915c1e8 JV |
1773 | write_unlock_bh(&bond->curr_slave_lock); |
1774 | ||
ff59c456 JV |
1775 | bond_set_carrier(bond); |
1776 | ||
3915c1e8 | 1777 | read_unlock(&bond->lock); |
1da177e4 | 1778 | |
b76cdba9 MW |
1779 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1780 | if (res) | |
569f0c4d | 1781 | goto err_close; |
b76cdba9 | 1782 | |
3d632c3f | 1783 | pr_info(DRV_NAME |
1da177e4 LT |
1784 | ": %s: enslaving %s as a%s interface with a%s link.\n", |
1785 | bond_dev->name, slave_dev->name, | |
1786 | new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup", | |
1787 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); | |
1788 | ||
1789 | /* enslave is successful */ | |
1790 | return 0; | |
1791 | ||
1792 | /* Undo stages on error */ | |
1da177e4 LT |
1793 | err_close: |
1794 | dev_close(slave_dev); | |
1795 | ||
569f0c4d JV |
1796 | err_unset_master: |
1797 | netdev_set_master(slave_dev, NULL); | |
1798 | ||
1da177e4 | 1799 | err_restore_mac: |
dd957c57 | 1800 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1801 | /* XXX TODO - fom follow mode needs to change master's |
1802 | * MAC if this slave's MAC is in use by the bond, or at | |
1803 | * least print a warning. | |
1804 | */ | |
2ab82852 MS |
1805 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1806 | addr.sa_family = slave_dev->type; | |
1807 | dev_set_mac_address(slave_dev, &addr); | |
1808 | } | |
1da177e4 LT |
1809 | |
1810 | err_free: | |
1811 | kfree(new_slave); | |
1812 | ||
1813 | err_undo_flags: | |
1814 | bond_dev->features = old_features; | |
3d632c3f | 1815 | |
1da177e4 LT |
1816 | return res; |
1817 | } | |
1818 | ||
1819 | /* | |
1820 | * Try to release the slave device <slave> from the bond device <master> | |
1821 | * It is legal to access curr_active_slave without a lock because all the function | |
1822 | * is write-locked. | |
1823 | * | |
1824 | * The rules for slave state should be: | |
1825 | * for Active/Backup: | |
1826 | * Active stays on all backups go down | |
1827 | * for Bonded connections: | |
1828 | * The first up interface should be left on and all others downed. | |
1829 | */ | |
a77b5325 | 1830 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1831 | { |
454d7c9b | 1832 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1833 | struct slave *slave, *oldcurrent; |
1834 | struct sockaddr addr; | |
1da177e4 LT |
1835 | |
1836 | /* slave is not a slave or master is not master of this slave */ | |
1837 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1838 | (slave_dev->master != bond_dev)) { | |
3d632c3f | 1839 | pr_err(DRV_NAME |
4e0952c7 | 1840 | ": %s: Error: cannot release %s.\n", |
1da177e4 LT |
1841 | bond_dev->name, slave_dev->name); |
1842 | return -EINVAL; | |
1843 | } | |
1844 | ||
1845 | write_lock_bh(&bond->lock); | |
1846 | ||
1847 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1848 | if (!slave) { | |
1849 | /* not a slave of this bond */ | |
3d632c3f | 1850 | pr_info(DRV_NAME |
1da177e4 LT |
1851 | ": %s: %s not enslaved\n", |
1852 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1853 | write_unlock_bh(&bond->lock); |
1da177e4 LT |
1854 | return -EINVAL; |
1855 | } | |
1856 | ||
3915c1e8 | 1857 | if (!bond->params.fail_over_mac) { |
89c76c62 SH |
1858 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) |
1859 | && bond->slave_cnt > 1) | |
3d632c3f | 1860 | pr_warning(DRV_NAME |
3915c1e8 | 1861 | ": %s: Warning: the permanent HWaddr of %s - " |
e174961c | 1862 | "%pM - is still in use by %s. " |
3915c1e8 JV |
1863 | "Set the HWaddr of %s to a different address " |
1864 | "to avoid conflicts.\n", | |
1865 | bond_dev->name, slave_dev->name, | |
e174961c | 1866 | slave->perm_hwaddr, |
3915c1e8 | 1867 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1868 | } |
1869 | ||
1870 | /* Inform AD package of unbinding of slave. */ | |
1871 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1872 | /* must be called before the slave is | |
1873 | * detached from the list | |
1874 | */ | |
1875 | bond_3ad_unbind_slave(slave); | |
1876 | } | |
1877 | ||
3d632c3f | 1878 | pr_info(DRV_NAME |
1da177e4 LT |
1879 | ": %s: releasing %s interface %s\n", |
1880 | bond_dev->name, | |
1881 | (slave->state == BOND_STATE_ACTIVE) | |
1882 | ? "active" : "backup", | |
1883 | slave_dev->name); | |
1884 | ||
1885 | oldcurrent = bond->curr_active_slave; | |
1886 | ||
1887 | bond->current_arp_slave = NULL; | |
1888 | ||
1889 | /* release the slave from its bond */ | |
1890 | bond_detach_slave(bond, slave); | |
1891 | ||
8531c5ff AK |
1892 | bond_compute_features(bond); |
1893 | ||
3d632c3f | 1894 | if (bond->primary_slave == slave) |
1da177e4 | 1895 | bond->primary_slave = NULL; |
1da177e4 | 1896 | |
3d632c3f | 1897 | if (oldcurrent == slave) |
1da177e4 | 1898 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1899 | |
58402054 | 1900 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1901 | /* Must be called only after the slave has been |
1902 | * detached from the list and the curr_active_slave | |
1903 | * has been cleared (if our_slave == old_current), | |
1904 | * but before a new active slave is selected. | |
1905 | */ | |
2543331d | 1906 | write_unlock_bh(&bond->lock); |
1da177e4 | 1907 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1908 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1909 | } |
1910 | ||
059fe7a5 JV |
1911 | if (oldcurrent == slave) { |
1912 | /* | |
1913 | * Note that we hold RTNL over this sequence, so there | |
1914 | * is no concern that another slave add/remove event | |
1915 | * will interfere. | |
1916 | */ | |
1917 | write_unlock_bh(&bond->lock); | |
1918 | read_lock(&bond->lock); | |
1919 | write_lock_bh(&bond->curr_slave_lock); | |
1920 | ||
1da177e4 LT |
1921 | bond_select_active_slave(bond); |
1922 | ||
059fe7a5 JV |
1923 | write_unlock_bh(&bond->curr_slave_lock); |
1924 | read_unlock(&bond->lock); | |
1925 | write_lock_bh(&bond->lock); | |
1926 | } | |
1927 | ||
1da177e4 | 1928 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
1929 | bond_set_carrier(bond); |
1930 | ||
1da177e4 LT |
1931 | /* if the last slave was removed, zero the mac address |
1932 | * of the master so it will be set by the application | |
1933 | * to the mac address of the first slave | |
1934 | */ | |
1935 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
1936 | ||
1937 | if (list_empty(&bond->vlan_list)) { | |
1938 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
1939 | } else { | |
3d632c3f | 1940 | pr_warning(DRV_NAME |
4e0952c7 | 1941 | ": %s: Warning: clearing HW address of %s while it " |
1da177e4 | 1942 | "still has VLANs.\n", |
4e0952c7 | 1943 | bond_dev->name, bond_dev->name); |
3d632c3f | 1944 | pr_warning(DRV_NAME |
4e0952c7 MW |
1945 | ": %s: When re-adding slaves, make sure the bond's " |
1946 | "HW address matches its VLANs'.\n", | |
1947 | bond_dev->name); | |
1da177e4 LT |
1948 | } |
1949 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1950 | !bond_has_challenged_slaves(bond)) { | |
3d632c3f | 1951 | pr_info(DRV_NAME |
4e0952c7 | 1952 | ": %s: last VLAN challenged slave %s " |
1da177e4 | 1953 | "left bond %s. VLAN blocking is removed\n", |
4e0952c7 | 1954 | bond_dev->name, slave_dev->name, bond_dev->name); |
1da177e4 LT |
1955 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
1956 | } | |
1957 | ||
1958 | write_unlock_bh(&bond->lock); | |
1959 | ||
b76cdba9 MW |
1960 | /* must do this from outside any spinlocks */ |
1961 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1962 | ||
1da177e4 LT |
1963 | bond_del_vlans_from_slave(bond, slave_dev); |
1964 | ||
1965 | /* If the mode USES_PRIMARY, then we should only remove its | |
1966 | * promisc and mc settings if it was the curr_active_slave, but that was | |
1967 | * already taken care of above when we detached the slave | |
1968 | */ | |
1969 | if (!USES_PRIMARY(bond->params.mode)) { | |
1970 | /* unset promiscuity level from slave */ | |
3d632c3f | 1971 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 1972 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1973 | |
1974 | /* unset allmulti level from slave */ | |
3d632c3f | 1975 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 1976 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
1977 | |
1978 | /* flush master's mc_list from slave */ | |
b9e40857 | 1979 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1980 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 1981 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1982 | } |
1983 | ||
1984 | netdev_set_master(slave_dev, NULL); | |
1985 | ||
1986 | /* close slave before restoring its mac address */ | |
1987 | dev_close(slave_dev); | |
1988 | ||
3915c1e8 | 1989 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1990 | /* restore original ("permanent") mac address */ |
1991 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1992 | addr.sa_family = slave_dev->type; | |
1993 | dev_set_mac_address(slave_dev, &addr); | |
1994 | } | |
1da177e4 | 1995 | |
8f903c70 | 1996 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
f5b2b966 JV |
1997 | IFF_SLAVE_INACTIVE | IFF_BONDING | |
1998 | IFF_SLAVE_NEEDARP); | |
1da177e4 LT |
1999 | |
2000 | kfree(slave); | |
2001 | ||
2002 | return 0; /* deletion OK */ | |
2003 | } | |
2004 | ||
d90a162a | 2005 | /* |
3d632c3f | 2006 | * First release a slave and than destroy the bond if no more slaves are left. |
d90a162a MS |
2007 | * Must be under rtnl_lock when this function is called. |
2008 | */ | |
3d632c3f SH |
2009 | int bond_release_and_destroy(struct net_device *bond_dev, |
2010 | struct net_device *slave_dev) | |
d90a162a | 2011 | { |
454d7c9b | 2012 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2013 | int ret; |
2014 | ||
2015 | ret = bond_release(bond_dev, slave_dev); | |
2016 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
3d632c3f | 2017 | pr_info(DRV_NAME ": %s: destroying bond %s.\n", |
d90a162a | 2018 | bond_dev->name, bond_dev->name); |
9e71626c | 2019 | unregister_netdevice(bond_dev); |
d90a162a MS |
2020 | } |
2021 | return ret; | |
2022 | } | |
2023 | ||
1da177e4 LT |
2024 | /* |
2025 | * This function releases all slaves. | |
2026 | */ | |
2027 | static int bond_release_all(struct net_device *bond_dev) | |
2028 | { | |
454d7c9b | 2029 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2030 | struct slave *slave; |
2031 | struct net_device *slave_dev; | |
2032 | struct sockaddr addr; | |
2033 | ||
2034 | write_lock_bh(&bond->lock); | |
2035 | ||
ff59c456 JV |
2036 | netif_carrier_off(bond_dev); |
2037 | ||
3d632c3f | 2038 | if (bond->slave_cnt == 0) |
1da177e4 | 2039 | goto out; |
1da177e4 LT |
2040 | |
2041 | bond->current_arp_slave = NULL; | |
2042 | bond->primary_slave = NULL; | |
2043 | bond_change_active_slave(bond, NULL); | |
2044 | ||
2045 | while ((slave = bond->first_slave) != NULL) { | |
2046 | /* Inform AD package of unbinding of slave | |
2047 | * before slave is detached from the list. | |
2048 | */ | |
3d632c3f | 2049 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2050 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2051 | |
2052 | slave_dev = slave->dev; | |
2053 | bond_detach_slave(bond, slave); | |
2054 | ||
2543331d JV |
2055 | /* now that the slave is detached, unlock and perform |
2056 | * all the undo steps that should not be called from | |
2057 | * within a lock. | |
2058 | */ | |
2059 | write_unlock_bh(&bond->lock); | |
2060 | ||
58402054 | 2061 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2062 | /* must be called only after the slave |
2063 | * has been detached from the list | |
2064 | */ | |
2065 | bond_alb_deinit_slave(bond, slave); | |
2066 | } | |
2067 | ||
8531c5ff AK |
2068 | bond_compute_features(bond); |
2069 | ||
b76cdba9 | 2070 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2071 | bond_del_vlans_from_slave(bond, slave_dev); |
2072 | ||
2073 | /* If the mode USES_PRIMARY, then we should only remove its | |
2074 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2075 | * already taken care of above when we detached the slave | |
2076 | */ | |
2077 | if (!USES_PRIMARY(bond->params.mode)) { | |
2078 | /* unset promiscuity level from slave */ | |
3d632c3f | 2079 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2080 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2081 | |
2082 | /* unset allmulti level from slave */ | |
3d632c3f | 2083 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2084 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2085 | |
2086 | /* flush master's mc_list from slave */ | |
b9e40857 | 2087 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2088 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2089 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2090 | } |
2091 | ||
2092 | netdev_set_master(slave_dev, NULL); | |
2093 | ||
2094 | /* close slave before restoring its mac address */ | |
2095 | dev_close(slave_dev); | |
2096 | ||
dd957c57 | 2097 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2098 | /* restore original ("permanent") mac address*/ |
2099 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2100 | addr.sa_family = slave_dev->type; | |
2101 | dev_set_mac_address(slave_dev, &addr); | |
2102 | } | |
1da177e4 | 2103 | |
8f903c70 JV |
2104 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
2105 | IFF_SLAVE_INACTIVE); | |
1da177e4 LT |
2106 | |
2107 | kfree(slave); | |
2108 | ||
2109 | /* re-acquire the lock before getting the next slave */ | |
2110 | write_lock_bh(&bond->lock); | |
2111 | } | |
2112 | ||
2113 | /* zero the mac address of the master so it will be | |
2114 | * set by the application to the mac address of the | |
2115 | * first slave | |
2116 | */ | |
2117 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2118 | ||
3d632c3f | 2119 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 2120 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
3d632c3f SH |
2121 | else { |
2122 | pr_warning(DRV_NAME | |
4e0952c7 | 2123 | ": %s: Warning: clearing HW address of %s while it " |
1da177e4 | 2124 | "still has VLANs.\n", |
4e0952c7 | 2125 | bond_dev->name, bond_dev->name); |
3d632c3f | 2126 | pr_warning(DRV_NAME |
4e0952c7 MW |
2127 | ": %s: When re-adding slaves, make sure the bond's " |
2128 | "HW address matches its VLANs'.\n", | |
2129 | bond_dev->name); | |
1da177e4 LT |
2130 | } |
2131 | ||
3d632c3f | 2132 | pr_info(DRV_NAME |
1da177e4 LT |
2133 | ": %s: released all slaves\n", |
2134 | bond_dev->name); | |
2135 | ||
2136 | out: | |
2137 | write_unlock_bh(&bond->lock); | |
2138 | ||
2139 | return 0; | |
2140 | } | |
2141 | ||
2142 | /* | |
2143 | * This function changes the active slave to slave <slave_dev>. | |
2144 | * It returns -EINVAL in the following cases. | |
2145 | * - <slave_dev> is not found in the list. | |
2146 | * - There is not active slave now. | |
2147 | * - <slave_dev> is already active. | |
2148 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2149 | * - <slave_dev> is not running. | |
3d632c3f SH |
2150 | * In these cases, this function does nothing. |
2151 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2152 | */ |
2153 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2154 | { | |
454d7c9b | 2155 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2156 | struct slave *old_active = NULL; |
2157 | struct slave *new_active = NULL; | |
2158 | int res = 0; | |
2159 | ||
3d632c3f | 2160 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2161 | return -EINVAL; |
1da177e4 LT |
2162 | |
2163 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2164 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2165 | return -EINVAL; |
1da177e4 | 2166 | |
059fe7a5 | 2167 | read_lock(&bond->lock); |
1da177e4 | 2168 | |
059fe7a5 | 2169 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2170 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2171 | read_unlock(&bond->curr_slave_lock); |
2172 | ||
1da177e4 LT |
2173 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2174 | ||
2175 | /* | |
2176 | * Changing to the current active: do nothing; return success. | |
2177 | */ | |
2178 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2179 | read_unlock(&bond->lock); |
1da177e4 LT |
2180 | return 0; |
2181 | } | |
2182 | ||
2183 | if ((new_active) && | |
2184 | (old_active) && | |
2185 | (new_active->link == BOND_LINK_UP) && | |
2186 | IS_UP(new_active->dev)) { | |
059fe7a5 | 2187 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2188 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2189 | write_unlock_bh(&bond->curr_slave_lock); |
3d632c3f | 2190 | } else |
1da177e4 | 2191 | res = -EINVAL; |
1da177e4 | 2192 | |
059fe7a5 | 2193 | read_unlock(&bond->lock); |
1da177e4 LT |
2194 | |
2195 | return res; | |
2196 | } | |
2197 | ||
1da177e4 LT |
2198 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2199 | { | |
454d7c9b | 2200 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2201 | |
2202 | info->bond_mode = bond->params.mode; | |
2203 | info->miimon = bond->params.miimon; | |
2204 | ||
6603a6f2 | 2205 | read_lock(&bond->lock); |
1da177e4 | 2206 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2207 | read_unlock(&bond->lock); |
1da177e4 LT |
2208 | |
2209 | return 0; | |
2210 | } | |
2211 | ||
2212 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2213 | { | |
454d7c9b | 2214 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2215 | struct slave *slave; |
689c96cc | 2216 | int i, res = -ENODEV; |
1da177e4 | 2217 | |
6603a6f2 | 2218 | read_lock(&bond->lock); |
1da177e4 LT |
2219 | |
2220 | bond_for_each_slave(bond, slave, i) { | |
2221 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2222 | res = 0; |
2223 | strcpy(info->slave_name, slave->dev->name); | |
2224 | info->link = slave->link; | |
2225 | info->state = slave->state; | |
2226 | info->link_failure_count = slave->link_failure_count; | |
1da177e4 LT |
2227 | break; |
2228 | } | |
2229 | } | |
2230 | ||
6603a6f2 | 2231 | read_unlock(&bond->lock); |
1da177e4 | 2232 | |
689c96cc | 2233 | return res; |
1da177e4 LT |
2234 | } |
2235 | ||
2236 | /*-------------------------------- Monitoring -------------------------------*/ | |
2237 | ||
1da177e4 | 2238 | |
f0c76d61 JV |
2239 | static int bond_miimon_inspect(struct bonding *bond) |
2240 | { | |
2241 | struct slave *slave; | |
2242 | int i, link_state, commit = 0; | |
41f89100 JP |
2243 | bool ignore_updelay; |
2244 | ||
2245 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2246 | |
2247 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2248 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2249 | |
f0c76d61 | 2250 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2251 | |
2252 | switch (slave->link) { | |
f0c76d61 JV |
2253 | case BOND_LINK_UP: |
2254 | if (link_state) | |
2255 | continue; | |
1da177e4 | 2256 | |
f0c76d61 JV |
2257 | slave->link = BOND_LINK_FAIL; |
2258 | slave->delay = bond->params.downdelay; | |
2259 | if (slave->delay) { | |
3d632c3f | 2260 | pr_info(DRV_NAME |
f0c76d61 JV |
2261 | ": %s: link status down for %s" |
2262 | "interface %s, disabling it in %d ms.\n", | |
2263 | bond->dev->name, | |
2264 | (bond->params.mode == | |
2265 | BOND_MODE_ACTIVEBACKUP) ? | |
2266 | ((slave->state == BOND_STATE_ACTIVE) ? | |
2267 | "active " : "backup ") : "", | |
2268 | slave->dev->name, | |
2269 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2270 | } |
f0c76d61 JV |
2271 | /*FALLTHRU*/ |
2272 | case BOND_LINK_FAIL: | |
2273 | if (link_state) { | |
2274 | /* | |
2275 | * recovered before downdelay expired | |
2276 | */ | |
2277 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2278 | slave->jiffies = jiffies; |
3d632c3f | 2279 | pr_info(DRV_NAME |
1da177e4 LT |
2280 | ": %s: link status up again after %d " |
2281 | "ms for interface %s.\n", | |
1b76b316 | 2282 | bond->dev->name, |
f0c76d61 JV |
2283 | (bond->params.downdelay - slave->delay) * |
2284 | bond->params.miimon, | |
2285 | slave->dev->name); | |
2286 | continue; | |
1da177e4 | 2287 | } |
f0c76d61 JV |
2288 | |
2289 | if (slave->delay <= 0) { | |
2290 | slave->new_link = BOND_LINK_DOWN; | |
2291 | commit++; | |
2292 | continue; | |
1da177e4 | 2293 | } |
1da177e4 | 2294 | |
f0c76d61 JV |
2295 | slave->delay--; |
2296 | break; | |
2297 | ||
2298 | case BOND_LINK_DOWN: | |
2299 | if (!link_state) | |
2300 | continue; | |
2301 | ||
2302 | slave->link = BOND_LINK_BACK; | |
2303 | slave->delay = bond->params.updelay; | |
2304 | ||
2305 | if (slave->delay) { | |
3d632c3f | 2306 | pr_info(DRV_NAME |
f0c76d61 JV |
2307 | ": %s: link status up for " |
2308 | "interface %s, enabling it in %d ms.\n", | |
2309 | bond->dev->name, slave->dev->name, | |
41f89100 | 2310 | ignore_updelay ? 0 : |
f0c76d61 JV |
2311 | bond->params.updelay * |
2312 | bond->params.miimon); | |
2313 | } | |
2314 | /*FALLTHRU*/ | |
2315 | case BOND_LINK_BACK: | |
2316 | if (!link_state) { | |
2317 | slave->link = BOND_LINK_DOWN; | |
3d632c3f | 2318 | pr_info(DRV_NAME |
1da177e4 LT |
2319 | ": %s: link status down again after %d " |
2320 | "ms for interface %s.\n", | |
1b76b316 | 2321 | bond->dev->name, |
f0c76d61 JV |
2322 | (bond->params.updelay - slave->delay) * |
2323 | bond->params.miimon, | |
2324 | slave->dev->name); | |
2325 | ||
2326 | continue; | |
2327 | } | |
2328 | ||
41f89100 JP |
2329 | if (ignore_updelay) |
2330 | slave->delay = 0; | |
2331 | ||
f0c76d61 JV |
2332 | if (slave->delay <= 0) { |
2333 | slave->new_link = BOND_LINK_UP; | |
2334 | commit++; | |
41f89100 | 2335 | ignore_updelay = false; |
f0c76d61 | 2336 | continue; |
1da177e4 | 2337 | } |
f0c76d61 JV |
2338 | |
2339 | slave->delay--; | |
1da177e4 | 2340 | break; |
f0c76d61 JV |
2341 | } |
2342 | } | |
1da177e4 | 2343 | |
f0c76d61 JV |
2344 | return commit; |
2345 | } | |
1da177e4 | 2346 | |
f0c76d61 JV |
2347 | static void bond_miimon_commit(struct bonding *bond) |
2348 | { | |
2349 | struct slave *slave; | |
2350 | int i; | |
2351 | ||
2352 | bond_for_each_slave(bond, slave, i) { | |
2353 | switch (slave->new_link) { | |
2354 | case BOND_LINK_NOCHANGE: | |
2355 | continue; | |
1da177e4 | 2356 | |
f0c76d61 JV |
2357 | case BOND_LINK_UP: |
2358 | slave->link = BOND_LINK_UP; | |
2359 | slave->jiffies = jiffies; | |
2360 | ||
2361 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2362 | /* prevent it from being the active one */ | |
2363 | slave->state = BOND_STATE_BACKUP; | |
2364 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
2365 | /* make it immediately active */ | |
2366 | slave->state = BOND_STATE_ACTIVE; | |
2367 | } else if (slave != bond->primary_slave) { | |
2368 | /* prevent it from being the active one */ | |
2369 | slave->state = BOND_STATE_BACKUP; | |
1da177e4 | 2370 | } |
1da177e4 | 2371 | |
3d632c3f | 2372 | pr_info(DRV_NAME |
f0c76d61 JV |
2373 | ": %s: link status definitely " |
2374 | "up for interface %s.\n", | |
2375 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2376 | |
f0c76d61 JV |
2377 | /* notify ad that the link status has changed */ |
2378 | if (bond->params.mode == BOND_MODE_8023AD) | |
2379 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2380 | |
58402054 | 2381 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2382 | bond_alb_handle_link_change(bond, slave, |
2383 | BOND_LINK_UP); | |
1da177e4 | 2384 | |
f0c76d61 JV |
2385 | if (!bond->curr_active_slave || |
2386 | (slave == bond->primary_slave)) | |
2387 | goto do_failover; | |
1da177e4 | 2388 | |
f0c76d61 | 2389 | continue; |
059fe7a5 | 2390 | |
f0c76d61 | 2391 | case BOND_LINK_DOWN: |
fba4acda JV |
2392 | if (slave->link_failure_count < UINT_MAX) |
2393 | slave->link_failure_count++; | |
2394 | ||
f0c76d61 | 2395 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2396 | |
f0c76d61 JV |
2397 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2398 | bond->params.mode == BOND_MODE_8023AD) | |
2399 | bond_set_slave_inactive_flags(slave); | |
2400 | ||
3d632c3f | 2401 | pr_info(DRV_NAME |
f0c76d61 JV |
2402 | ": %s: link status definitely down for " |
2403 | "interface %s, disabling it\n", | |
2404 | bond->dev->name, slave->dev->name); | |
2405 | ||
2406 | if (bond->params.mode == BOND_MODE_8023AD) | |
2407 | bond_3ad_handle_link_change(slave, | |
2408 | BOND_LINK_DOWN); | |
2409 | ||
ae63e808 | 2410 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2411 | bond_alb_handle_link_change(bond, slave, |
2412 | BOND_LINK_DOWN); | |
2413 | ||
2414 | if (slave == bond->curr_active_slave) | |
2415 | goto do_failover; | |
2416 | ||
2417 | continue; | |
2418 | ||
2419 | default: | |
3d632c3f | 2420 | pr_err(DRV_NAME |
f0c76d61 JV |
2421 | ": %s: invalid new link %d on slave %s\n", |
2422 | bond->dev->name, slave->new_link, | |
2423 | slave->dev->name); | |
2424 | slave->new_link = BOND_LINK_NOCHANGE; | |
2425 | ||
2426 | continue; | |
2427 | } | |
2428 | ||
2429 | do_failover: | |
2430 | ASSERT_RTNL(); | |
2431 | write_lock_bh(&bond->curr_slave_lock); | |
2432 | bond_select_active_slave(bond); | |
2433 | write_unlock_bh(&bond->curr_slave_lock); | |
2434 | } | |
2435 | ||
2436 | bond_set_carrier(bond); | |
1da177e4 LT |
2437 | } |
2438 | ||
0b0eef66 JV |
2439 | /* |
2440 | * bond_mii_monitor | |
2441 | * | |
2442 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2443 | * inspection, then (if inspection indicates something needs to be done) |
2444 | * an acquisition of appropriate locks followed by a commit phase to | |
2445 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2446 | */ |
2447 | void bond_mii_monitor(struct work_struct *work) | |
2448 | { | |
2449 | struct bonding *bond = container_of(work, struct bonding, | |
2450 | mii_work.work); | |
0b0eef66 JV |
2451 | |
2452 | read_lock(&bond->lock); | |
f0c76d61 JV |
2453 | if (bond->kill_timers) |
2454 | goto out; | |
2455 | ||
2456 | if (bond->slave_cnt == 0) | |
2457 | goto re_arm; | |
b59f9f74 JV |
2458 | |
2459 | if (bond->send_grat_arp) { | |
2460 | read_lock(&bond->curr_slave_lock); | |
2461 | bond_send_gratuitous_arp(bond); | |
2462 | read_unlock(&bond->curr_slave_lock); | |
2463 | } | |
2464 | ||
305d552a BH |
2465 | if (bond->send_unsol_na) { |
2466 | read_lock(&bond->curr_slave_lock); | |
2467 | bond_send_unsolicited_na(bond); | |
2468 | read_unlock(&bond->curr_slave_lock); | |
2469 | } | |
2470 | ||
f0c76d61 | 2471 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2472 | read_unlock(&bond->lock); |
2473 | rtnl_lock(); | |
2474 | read_lock(&bond->lock); | |
f0c76d61 JV |
2475 | |
2476 | bond_miimon_commit(bond); | |
2477 | ||
5655662d JV |
2478 | read_unlock(&bond->lock); |
2479 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2480 | read_lock(&bond->lock); | |
0b0eef66 JV |
2481 | } |
2482 | ||
f0c76d61 JV |
2483 | re_arm: |
2484 | if (bond->params.miimon) | |
2485 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2486 | msecs_to_jiffies(bond->params.miimon)); | |
2487 | out: | |
0b0eef66 | 2488 | read_unlock(&bond->lock); |
0b0eef66 | 2489 | } |
c3ade5ca | 2490 | |
d3bb52b0 | 2491 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2492 | { |
2493 | struct in_device *idev; | |
2494 | struct in_ifaddr *ifa; | |
a144ea4b | 2495 | __be32 addr = 0; |
c3ade5ca JV |
2496 | |
2497 | if (!dev) | |
2498 | return 0; | |
2499 | ||
2500 | rcu_read_lock(); | |
e5ed6399 | 2501 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2502 | if (!idev) |
2503 | goto out; | |
2504 | ||
2505 | ifa = idev->ifa_list; | |
2506 | if (!ifa) | |
2507 | goto out; | |
2508 | ||
2509 | addr = ifa->ifa_local; | |
2510 | out: | |
2511 | rcu_read_unlock(); | |
2512 | return addr; | |
2513 | } | |
2514 | ||
d3bb52b0 | 2515 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2516 | { |
0883beca | 2517 | struct vlan_entry *vlan; |
f5b2b966 JV |
2518 | |
2519 | if (ip == bond->master_ip) | |
2520 | return 1; | |
2521 | ||
0883beca | 2522 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2523 | if (ip == vlan->vlan_ip) |
2524 | return 1; | |
2525 | } | |
2526 | ||
2527 | return 0; | |
2528 | } | |
2529 | ||
c3ade5ca JV |
2530 | /* |
2531 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2532 | * switches in VLAN mode (especially if ports are configured as | |
2533 | * "native" to a VLAN) might not pass non-tagged frames. | |
2534 | */ | |
d3bb52b0 | 2535 | 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 |
2536 | { |
2537 | struct sk_buff *skb; | |
2538 | ||
5a03cdb7 | 2539 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
c3ade5ca | 2540 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2541 | |
c3ade5ca JV |
2542 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2543 | NULL, slave_dev->dev_addr, NULL); | |
2544 | ||
2545 | if (!skb) { | |
3d632c3f | 2546 | pr_err(DRV_NAME ": ARP packet allocation failed\n"); |
c3ade5ca JV |
2547 | return; |
2548 | } | |
2549 | if (vlan_id) { | |
2550 | skb = vlan_put_tag(skb, vlan_id); | |
2551 | if (!skb) { | |
3d632c3f | 2552 | pr_err(DRV_NAME ": failed to insert VLAN tag\n"); |
c3ade5ca JV |
2553 | return; |
2554 | } | |
2555 | } | |
2556 | arp_xmit(skb); | |
2557 | } | |
2558 | ||
2559 | ||
1da177e4 LT |
2560 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2561 | { | |
c3ade5ca | 2562 | int i, vlan_id, rv; |
d3bb52b0 | 2563 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2564 | struct vlan_entry *vlan; |
c3ade5ca JV |
2565 | struct net_device *vlan_dev; |
2566 | struct flowi fl; | |
2567 | struct rtable *rt; | |
1da177e4 | 2568 | |
6b780567 MW |
2569 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2570 | if (!targets[i]) | |
5a31bec0 | 2571 | break; |
5a03cdb7 | 2572 | pr_debug("basa: target %x\n", targets[i]); |
c3ade5ca | 2573 | if (list_empty(&bond->vlan_list)) { |
5a03cdb7 | 2574 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2575 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2576 | bond->master_ip, 0); | |
2577 | continue; | |
2578 | } | |
2579 | ||
2580 | /* | |
2581 | * If VLANs are configured, we do a route lookup to | |
2582 | * determine which VLAN interface would be used, so we | |
2583 | * can tag the ARP with the proper VLAN tag. | |
2584 | */ | |
2585 | memset(&fl, 0, sizeof(fl)); | |
2586 | fl.fl4_dst = targets[i]; | |
2587 | fl.fl4_tos = RTO_ONLINK; | |
2588 | ||
f206351a | 2589 | rv = ip_route_output_key(&init_net, &rt, &fl); |
c3ade5ca JV |
2590 | if (rv) { |
2591 | if (net_ratelimit()) { | |
3d632c3f | 2592 | pr_warning(DRV_NAME |
63779436 HH |
2593 | ": %s: no route to arp_ip_target %pI4\n", |
2594 | bond->dev->name, &fl.fl4_dst); | |
c3ade5ca JV |
2595 | } |
2596 | continue; | |
2597 | } | |
2598 | ||
2599 | /* | |
2600 | * This target is not on a VLAN | |
2601 | */ | |
2602 | if (rt->u.dst.dev == bond->dev) { | |
ed4b9f80 | 2603 | ip_rt_put(rt); |
5a03cdb7 | 2604 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2605 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2606 | bond->master_ip, 0); | |
2607 | continue; | |
2608 | } | |
2609 | ||
2610 | vlan_id = 0; | |
0883beca | 2611 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2612 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2613 | if (vlan_dev == rt->u.dst.dev) { |
2614 | vlan_id = vlan->vlan_id; | |
5a03cdb7 | 2615 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2616 | vlan_dev->name, vlan_id); |
2617 | break; | |
2618 | } | |
2619 | } | |
2620 | ||
2621 | if (vlan_id) { | |
ed4b9f80 | 2622 | ip_rt_put(rt); |
c3ade5ca JV |
2623 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2624 | vlan->vlan_ip, vlan_id); | |
2625 | continue; | |
2626 | } | |
2627 | ||
2628 | if (net_ratelimit()) { | |
3d632c3f | 2629 | pr_warning(DRV_NAME |
63779436 HH |
2630 | ": %s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2631 | bond->dev->name, &fl.fl4_dst, | |
c3ade5ca JV |
2632 | rt->u.dst.dev ? rt->u.dst.dev->name : "NULL"); |
2633 | } | |
ed4b9f80 | 2634 | ip_rt_put(rt); |
c3ade5ca JV |
2635 | } |
2636 | } | |
2637 | ||
2638 | /* | |
2639 | * Kick out a gratuitous ARP for an IP on the bonding master plus one | |
2640 | * for each VLAN above us. | |
b59f9f74 JV |
2641 | * |
2642 | * Caller must hold curr_slave_lock for read or better | |
c3ade5ca JV |
2643 | */ |
2644 | static void bond_send_gratuitous_arp(struct bonding *bond) | |
2645 | { | |
2646 | struct slave *slave = bond->curr_active_slave; | |
2647 | struct vlan_entry *vlan; | |
2648 | struct net_device *vlan_dev; | |
2649 | ||
5a03cdb7 | 2650 | pr_debug("bond_send_grat_arp: bond %s slave %s\n", bond->dev->name, |
c3ade5ca | 2651 | slave ? slave->dev->name : "NULL"); |
b59f9f74 JV |
2652 | |
2653 | if (!slave || !bond->send_grat_arp || | |
2654 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
c3ade5ca JV |
2655 | return; |
2656 | ||
b59f9f74 JV |
2657 | bond->send_grat_arp--; |
2658 | ||
c3ade5ca JV |
2659 | if (bond->master_ip) { |
2660 | bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip, | |
1053f62c | 2661 | bond->master_ip, 0); |
c3ade5ca JV |
2662 | } |
2663 | ||
2664 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
5c15bdec | 2665 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2666 | if (vlan->vlan_ip) { |
2667 | bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip, | |
2668 | vlan->vlan_ip, vlan->vlan_id); | |
2669 | } | |
1da177e4 LT |
2670 | } |
2671 | } | |
2672 | ||
d3bb52b0 | 2673 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2674 | { |
2675 | int i; | |
d3bb52b0 | 2676 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2677 | |
f5b2b966 | 2678 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2679 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
63779436 | 2680 | &sip, &tip, i, &targets[i], bond_has_this_ip(bond, tip)); |
f5b2b966 JV |
2681 | if (sip == targets[i]) { |
2682 | if (bond_has_this_ip(bond, tip)) | |
2683 | slave->last_arp_rx = jiffies; | |
2684 | return; | |
2685 | } | |
2686 | } | |
2687 | } | |
2688 | ||
2689 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2690 | { | |
2691 | struct arphdr *arp; | |
2692 | struct slave *slave; | |
2693 | struct bonding *bond; | |
2694 | unsigned char *arp_ptr; | |
d3bb52b0 | 2695 | __be32 sip, tip; |
f5b2b966 | 2696 | |
c346dca1 | 2697 | if (dev_net(dev) != &init_net) |
e730c155 EB |
2698 | goto out; |
2699 | ||
f5b2b966 JV |
2700 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2701 | goto out; | |
2702 | ||
454d7c9b | 2703 | bond = netdev_priv(dev); |
f5b2b966 JV |
2704 | read_lock(&bond->lock); |
2705 | ||
5a03cdb7 | 2706 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
f5b2b966 JV |
2707 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2708 | orig_dev ? orig_dev->name : "NULL"); | |
2709 | ||
2710 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2711 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2712 | goto out_unlock; | |
2713 | ||
988b7050 | 2714 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2715 | goto out_unlock; |
2716 | ||
d0a92be0 | 2717 | arp = arp_hdr(skb); |
f5b2b966 JV |
2718 | if (arp->ar_hln != dev->addr_len || |
2719 | skb->pkt_type == PACKET_OTHERHOST || | |
2720 | skb->pkt_type == PACKET_LOOPBACK || | |
2721 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2722 | arp->ar_pro != htons(ETH_P_IP) || | |
2723 | arp->ar_pln != 4) | |
2724 | goto out_unlock; | |
2725 | ||
2726 | arp_ptr = (unsigned char *)(arp + 1); | |
2727 | arp_ptr += dev->addr_len; | |
2728 | memcpy(&sip, arp_ptr, 4); | |
2729 | arp_ptr += 4 + dev->addr_len; | |
2730 | memcpy(&tip, arp_ptr, 4); | |
2731 | ||
5a03cdb7 | 2732 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
63779436 HH |
2733 | bond->dev->name, slave->dev->name, slave->state, |
2734 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2735 | &sip, &tip); | |
f5b2b966 JV |
2736 | |
2737 | /* | |
2738 | * Backup slaves won't see the ARP reply, but do come through | |
2739 | * here for each ARP probe (so we swap the sip/tip to validate | |
2740 | * the probe). In a "redundant switch, common router" type of | |
2741 | * configuration, the ARP probe will (hopefully) travel from | |
2742 | * the active, through one switch, the router, then the other | |
2743 | * switch before reaching the backup. | |
2744 | */ | |
2745 | if (slave->state == BOND_STATE_ACTIVE) | |
2746 | bond_validate_arp(bond, slave, sip, tip); | |
2747 | else | |
2748 | bond_validate_arp(bond, slave, tip, sip); | |
2749 | ||
2750 | out_unlock: | |
2751 | read_unlock(&bond->lock); | |
2752 | out: | |
2753 | dev_kfree_skb(skb); | |
2754 | return NET_RX_SUCCESS; | |
2755 | } | |
2756 | ||
1da177e4 LT |
2757 | /* |
2758 | * this function is called regularly to monitor each slave's link | |
2759 | * ensuring that traffic is being sent and received when arp monitoring | |
2760 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2761 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2762 | * arp monitoring in active backup mode. | |
2763 | */ | |
1b76b316 | 2764 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2765 | { |
1b76b316 JV |
2766 | struct bonding *bond = container_of(work, struct bonding, |
2767 | arp_work.work); | |
1da177e4 LT |
2768 | struct slave *slave, *oldcurrent; |
2769 | int do_failover = 0; | |
2770 | int delta_in_ticks; | |
2771 | int i; | |
2772 | ||
2773 | read_lock(&bond->lock); | |
2774 | ||
5ce0da8f | 2775 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2776 | |
3d632c3f | 2777 | if (bond->kill_timers) |
1da177e4 | 2778 | goto out; |
1da177e4 | 2779 | |
3d632c3f | 2780 | if (bond->slave_cnt == 0) |
1da177e4 | 2781 | goto re_arm; |
1da177e4 LT |
2782 | |
2783 | read_lock(&bond->curr_slave_lock); | |
2784 | oldcurrent = bond->curr_active_slave; | |
2785 | read_unlock(&bond->curr_slave_lock); | |
2786 | ||
2787 | /* see if any of the previous devices are up now (i.e. they have | |
2788 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2789 | * the picture unless it is null. also, slave->jiffies is not needed | |
2790 | * here because we send an arp on each slave and give a slave as | |
2791 | * long as it needs to get the tx/rx within the delta. | |
2792 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2793 | * so it can wait | |
2794 | */ | |
2795 | bond_for_each_slave(bond, slave, i) { | |
2796 | if (slave->link != BOND_LINK_UP) { | |
9d21493b | 2797 | if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) && |
b63bb739 | 2798 | time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) { |
1da177e4 LT |
2799 | |
2800 | slave->link = BOND_LINK_UP; | |
2801 | slave->state = BOND_STATE_ACTIVE; | |
2802 | ||
2803 | /* primary_slave has no meaning in round-robin | |
2804 | * mode. the window of a slave being up and | |
2805 | * curr_active_slave being null after enslaving | |
2806 | * is closed. | |
2807 | */ | |
2808 | if (!oldcurrent) { | |
3d632c3f | 2809 | pr_info(DRV_NAME |
1da177e4 LT |
2810 | ": %s: link status definitely " |
2811 | "up for interface %s, ", | |
1b76b316 | 2812 | bond->dev->name, |
1da177e4 LT |
2813 | slave->dev->name); |
2814 | do_failover = 1; | |
2815 | } else { | |
3d632c3f | 2816 | pr_info(DRV_NAME |
1da177e4 | 2817 | ": %s: interface %s is now up\n", |
1b76b316 | 2818 | bond->dev->name, |
1da177e4 LT |
2819 | slave->dev->name); |
2820 | } | |
2821 | } | |
2822 | } else { | |
2823 | /* slave->link == BOND_LINK_UP */ | |
2824 | ||
2825 | /* not all switches will respond to an arp request | |
2826 | * when the source ip is 0, so don't take the link down | |
2827 | * if we don't know our ip yet | |
2828 | */ | |
9d21493b | 2829 | if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) || |
4b8a9239 | 2830 | (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) { |
1da177e4 LT |
2831 | |
2832 | slave->link = BOND_LINK_DOWN; | |
2833 | slave->state = BOND_STATE_BACKUP; | |
2834 | ||
3d632c3f | 2835 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2836 | slave->link_failure_count++; |
1da177e4 | 2837 | |
3d632c3f | 2838 | pr_info(DRV_NAME |
1da177e4 | 2839 | ": %s: interface %s is now down.\n", |
1b76b316 | 2840 | bond->dev->name, |
1da177e4 LT |
2841 | slave->dev->name); |
2842 | ||
3d632c3f | 2843 | if (slave == oldcurrent) |
1da177e4 | 2844 | do_failover = 1; |
1da177e4 LT |
2845 | } |
2846 | } | |
2847 | ||
2848 | /* note: if switch is in round-robin mode, all links | |
2849 | * must tx arp to ensure all links rx an arp - otherwise | |
2850 | * links may oscillate or not come up at all; if switch is | |
2851 | * in something like xor mode, there is nothing we can | |
2852 | * do - all replies will be rx'ed on same link causing slaves | |
2853 | * to be unstable during low/no traffic periods | |
2854 | */ | |
3d632c3f | 2855 | if (IS_UP(slave->dev)) |
1da177e4 | 2856 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2857 | } |
2858 | ||
2859 | if (do_failover) { | |
059fe7a5 | 2860 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2861 | |
2862 | bond_select_active_slave(bond); | |
2863 | ||
059fe7a5 | 2864 | write_unlock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2865 | } |
2866 | ||
2867 | re_arm: | |
1b76b316 JV |
2868 | if (bond->params.arp_interval) |
2869 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
2870 | out: |
2871 | read_unlock(&bond->lock); | |
2872 | } | |
2873 | ||
2874 | /* | |
b2220cad JV |
2875 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2876 | * changes. Sets new_link in slaves to specify what action should take | |
2877 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2878 | * to link states must be committed. | |
2879 | * | |
2880 | * Called with bond->lock held for read. | |
1da177e4 | 2881 | */ |
b2220cad | 2882 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2883 | { |
1da177e4 | 2884 | struct slave *slave; |
b2220cad | 2885 | int i, commit = 0; |
1da177e4 | 2886 | |
b2220cad JV |
2887 | bond_for_each_slave(bond, slave, i) { |
2888 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2889 | |
b2220cad JV |
2890 | if (slave->link != BOND_LINK_UP) { |
2891 | if (time_before_eq(jiffies, slave_last_rx(bond, slave) + | |
2892 | delta_in_ticks)) { | |
2893 | slave->new_link = BOND_LINK_UP; | |
2894 | commit++; | |
2895 | } | |
1da177e4 | 2896 | |
b2220cad JV |
2897 | continue; |
2898 | } | |
1da177e4 | 2899 | |
b2220cad JV |
2900 | /* |
2901 | * Give slaves 2*delta after being enslaved or made | |
2902 | * active. This avoids bouncing, as the last receive | |
2903 | * times need a full ARP monitor cycle to be updated. | |
2904 | */ | |
2905 | if (!time_after_eq(jiffies, slave->jiffies + | |
2906 | 2 * delta_in_ticks)) | |
2907 | continue; | |
2908 | ||
2909 | /* | |
2910 | * Backup slave is down if: | |
2911 | * - No current_arp_slave AND | |
2912 | * - more than 3*delta since last receive AND | |
2913 | * - the bond has an IP address | |
2914 | * | |
2915 | * Note: a non-null current_arp_slave indicates | |
2916 | * the curr_active_slave went down and we are | |
2917 | * searching for a new one; under this condition | |
2918 | * we only take the curr_active_slave down - this | |
2919 | * gives each slave a chance to tx/rx traffic | |
2920 | * before being taken out | |
2921 | */ | |
2922 | if (slave->state == BOND_STATE_BACKUP && | |
2923 | !bond->current_arp_slave && | |
2924 | time_after(jiffies, slave_last_rx(bond, slave) + | |
2925 | 3 * delta_in_ticks)) { | |
2926 | slave->new_link = BOND_LINK_DOWN; | |
2927 | commit++; | |
2928 | } | |
2929 | ||
2930 | /* | |
2931 | * Active slave is down if: | |
2932 | * - more than 2*delta since transmitting OR | |
2933 | * - (more than 2*delta since receive AND | |
2934 | * the bond has an IP address) | |
2935 | */ | |
2936 | if ((slave->state == BOND_STATE_ACTIVE) && | |
9d21493b | 2937 | (time_after_eq(jiffies, dev_trans_start(slave->dev) + |
b2220cad JV |
2938 | 2 * delta_in_ticks) || |
2939 | (time_after_eq(jiffies, slave_last_rx(bond, slave) | |
2940 | + 2 * delta_in_ticks)))) { | |
2941 | slave->new_link = BOND_LINK_DOWN; | |
2942 | commit++; | |
2943 | } | |
1da177e4 LT |
2944 | } |
2945 | ||
b2220cad JV |
2946 | return commit; |
2947 | } | |
1da177e4 | 2948 | |
b2220cad JV |
2949 | /* |
2950 | * Called to commit link state changes noted by inspection step of | |
2951 | * active-backup mode ARP monitor. | |
2952 | * | |
2953 | * Called with RTNL and bond->lock for read. | |
2954 | */ | |
2955 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2956 | { | |
2957 | struct slave *slave; | |
2958 | int i; | |
1da177e4 | 2959 | |
b2220cad JV |
2960 | bond_for_each_slave(bond, slave, i) { |
2961 | switch (slave->new_link) { | |
2962 | case BOND_LINK_NOCHANGE: | |
2963 | continue; | |
ff59c456 | 2964 | |
b2220cad | 2965 | case BOND_LINK_UP: |
b9f60253 JP |
2966 | if ((!bond->curr_active_slave && |
2967 | time_before_eq(jiffies, | |
2968 | dev_trans_start(slave->dev) + | |
2969 | delta_in_ticks)) || | |
2970 | bond->curr_active_slave != slave) { | |
b2220cad | 2971 | slave->link = BOND_LINK_UP; |
b2220cad JV |
2972 | bond->current_arp_slave = NULL; |
2973 | ||
3d632c3f | 2974 | pr_info(DRV_NAME |
b9f60253 JP |
2975 | ": %s: link status definitely " |
2976 | "up for interface %s.\n", | |
2977 | bond->dev->name, slave->dev->name); | |
b2220cad | 2978 | |
b9f60253 JP |
2979 | if (!bond->curr_active_slave || |
2980 | (slave == bond->primary_slave)) | |
2981 | goto do_failover; | |
1da177e4 | 2982 | |
b9f60253 | 2983 | } |
1da177e4 | 2984 | |
b9f60253 | 2985 | continue; |
1da177e4 | 2986 | |
b2220cad JV |
2987 | case BOND_LINK_DOWN: |
2988 | if (slave->link_failure_count < UINT_MAX) | |
2989 | slave->link_failure_count++; | |
2990 | ||
2991 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2992 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2993 | |
b9f60253 JP |
2994 | pr_info(DRV_NAME |
2995 | ": %s: link status definitely down for " | |
2996 | "interface %s, disabling it\n", | |
2997 | bond->dev->name, slave->dev->name); | |
b2220cad | 2998 | |
b9f60253 | 2999 | if (slave == bond->curr_active_slave) { |
b2220cad | 3000 | bond->current_arp_slave = NULL; |
b9f60253 | 3001 | goto do_failover; |
1da177e4 | 3002 | } |
b9f60253 JP |
3003 | |
3004 | continue; | |
b2220cad JV |
3005 | |
3006 | default: | |
3d632c3f | 3007 | pr_err(DRV_NAME |
b2220cad JV |
3008 | ": %s: impossible: new_link %d on slave %s\n", |
3009 | bond->dev->name, slave->new_link, | |
3010 | slave->dev->name); | |
b9f60253 | 3011 | continue; |
1da177e4 | 3012 | } |
1da177e4 | 3013 | |
b9f60253 JP |
3014 | do_failover: |
3015 | ASSERT_RTNL(); | |
b2220cad | 3016 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3017 | bond_select_active_slave(bond); |
b2220cad JV |
3018 | write_unlock_bh(&bond->curr_slave_lock); |
3019 | } | |
1da177e4 | 3020 | |
b2220cad JV |
3021 | bond_set_carrier(bond); |
3022 | } | |
1da177e4 | 3023 | |
b2220cad JV |
3024 | /* |
3025 | * Send ARP probes for active-backup mode ARP monitor. | |
3026 | * | |
3027 | * Called with bond->lock held for read. | |
3028 | */ | |
3029 | static void bond_ab_arp_probe(struct bonding *bond) | |
3030 | { | |
3031 | struct slave *slave; | |
3032 | int i; | |
1da177e4 | 3033 | |
b2220cad | 3034 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3035 | |
b2220cad | 3036 | if (bond->current_arp_slave && bond->curr_active_slave) |
3d632c3f | 3037 | pr_info(DRV_NAME "PROBE: c_arp %s && cas %s BAD\n", |
b2220cad JV |
3038 | bond->current_arp_slave->dev->name, |
3039 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3040 | |
b2220cad JV |
3041 | if (bond->curr_active_slave) { |
3042 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3043 | read_unlock(&bond->curr_slave_lock); | |
3044 | return; | |
3045 | } | |
1da177e4 | 3046 | |
b2220cad | 3047 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3048 | |
b2220cad JV |
3049 | /* if we don't have a curr_active_slave, search for the next available |
3050 | * backup slave from the current_arp_slave and make it the candidate | |
3051 | * for becoming the curr_active_slave | |
3052 | */ | |
1da177e4 | 3053 | |
b2220cad JV |
3054 | if (!bond->current_arp_slave) { |
3055 | bond->current_arp_slave = bond->first_slave; | |
3056 | if (!bond->current_arp_slave) | |
3057 | return; | |
3058 | } | |
1da177e4 | 3059 | |
b2220cad | 3060 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3061 | |
b2220cad JV |
3062 | /* search for next candidate */ |
3063 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3064 | if (IS_UP(slave->dev)) { | |
3065 | slave->link = BOND_LINK_BACK; | |
3066 | bond_set_slave_active_flags(slave); | |
3067 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3068 | slave->jiffies = jiffies; |
b2220cad JV |
3069 | bond->current_arp_slave = slave; |
3070 | break; | |
1da177e4 LT |
3071 | } |
3072 | ||
b2220cad JV |
3073 | /* if the link state is up at this point, we |
3074 | * mark it down - this can happen if we have | |
3075 | * simultaneous link failures and | |
3076 | * reselect_active_interface doesn't make this | |
3077 | * one the current slave so it is still marked | |
3078 | * up when it is actually down | |
1da177e4 | 3079 | */ |
b2220cad JV |
3080 | if (slave->link == BOND_LINK_UP) { |
3081 | slave->link = BOND_LINK_DOWN; | |
3082 | if (slave->link_failure_count < UINT_MAX) | |
3083 | slave->link_failure_count++; | |
1da177e4 | 3084 | |
b2220cad JV |
3085 | bond_set_slave_inactive_flags(slave); |
3086 | ||
3d632c3f | 3087 | pr_info(DRV_NAME |
b2220cad JV |
3088 | ": %s: backup interface %s is now down.\n", |
3089 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3090 | } |
b2220cad JV |
3091 | } |
3092 | } | |
1da177e4 | 3093 | |
b2220cad JV |
3094 | void bond_activebackup_arp_mon(struct work_struct *work) |
3095 | { | |
3096 | struct bonding *bond = container_of(work, struct bonding, | |
3097 | arp_work.work); | |
3098 | int delta_in_ticks; | |
1da177e4 | 3099 | |
b2220cad | 3100 | read_lock(&bond->lock); |
1da177e4 | 3101 | |
b2220cad JV |
3102 | if (bond->kill_timers) |
3103 | goto out; | |
1da177e4 | 3104 | |
b2220cad | 3105 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3106 | |
b2220cad JV |
3107 | if (bond->slave_cnt == 0) |
3108 | goto re_arm; | |
3109 | ||
b59f9f74 JV |
3110 | if (bond->send_grat_arp) { |
3111 | read_lock(&bond->curr_slave_lock); | |
3112 | bond_send_gratuitous_arp(bond); | |
3113 | read_unlock(&bond->curr_slave_lock); | |
3114 | } | |
3115 | ||
305d552a BH |
3116 | if (bond->send_unsol_na) { |
3117 | read_lock(&bond->curr_slave_lock); | |
3118 | bond_send_unsolicited_na(bond); | |
3119 | read_unlock(&bond->curr_slave_lock); | |
3120 | } | |
3121 | ||
b2220cad JV |
3122 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3123 | read_unlock(&bond->lock); | |
3124 | rtnl_lock(); | |
3125 | read_lock(&bond->lock); | |
3126 | ||
3127 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3128 | ||
3129 | read_unlock(&bond->lock); | |
3130 | rtnl_unlock(); | |
3131 | read_lock(&bond->lock); | |
1da177e4 LT |
3132 | } |
3133 | ||
b2220cad JV |
3134 | bond_ab_arp_probe(bond); |
3135 | ||
1da177e4 | 3136 | re_arm: |
3d632c3f | 3137 | if (bond->params.arp_interval) |
1b76b316 | 3138 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3139 | out: |
3140 | read_unlock(&bond->lock); | |
3141 | } | |
3142 | ||
3143 | /*------------------------------ proc/seq_file-------------------------------*/ | |
3144 | ||
3145 | #ifdef CONFIG_PROC_FS | |
3146 | ||
1da177e4 | 3147 | static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos) |
1f78d9f9 HE |
3148 | __acquires(&dev_base_lock) |
3149 | __acquires(&bond->lock) | |
1da177e4 LT |
3150 | { |
3151 | struct bonding *bond = seq->private; | |
3152 | loff_t off = 0; | |
3153 | struct slave *slave; | |
3154 | int i; | |
3155 | ||
3156 | /* make sure the bond won't be taken away */ | |
3157 | read_lock(&dev_base_lock); | |
6603a6f2 | 3158 | read_lock(&bond->lock); |
1da177e4 | 3159 | |
3d632c3f | 3160 | if (*pos == 0) |
1da177e4 | 3161 | return SEQ_START_TOKEN; |
1da177e4 LT |
3162 | |
3163 | bond_for_each_slave(bond, slave, i) { | |
3d632c3f | 3164 | if (++off == *pos) |
1da177e4 | 3165 | return slave; |
1da177e4 LT |
3166 | } |
3167 | ||
3168 | return NULL; | |
3169 | } | |
3170 | ||
3171 | static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3172 | { | |
3173 | struct bonding *bond = seq->private; | |
3174 | struct slave *slave = v; | |
3175 | ||
3176 | ++*pos; | |
3d632c3f | 3177 | if (v == SEQ_START_TOKEN) |
1da177e4 | 3178 | return bond->first_slave; |
1da177e4 LT |
3179 | |
3180 | slave = slave->next; | |
3181 | ||
3182 | return (slave == bond->first_slave) ? NULL : slave; | |
3183 | } | |
3184 | ||
3185 | static void bond_info_seq_stop(struct seq_file *seq, void *v) | |
1f78d9f9 HE |
3186 | __releases(&bond->lock) |
3187 | __releases(&dev_base_lock) | |
1da177e4 LT |
3188 | { |
3189 | struct bonding *bond = seq->private; | |
3190 | ||
6603a6f2 | 3191 | read_unlock(&bond->lock); |
1da177e4 LT |
3192 | read_unlock(&dev_base_lock); |
3193 | } | |
3194 | ||
3195 | static void bond_info_show_master(struct seq_file *seq) | |
3196 | { | |
3197 | struct bonding *bond = seq->private; | |
3198 | struct slave *curr; | |
4756b02f | 3199 | int i; |
1da177e4 LT |
3200 | |
3201 | read_lock(&bond->curr_slave_lock); | |
3202 | curr = bond->curr_active_slave; | |
3203 | read_unlock(&bond->curr_slave_lock); | |
3204 | ||
dd957c57 | 3205 | seq_printf(seq, "Bonding Mode: %s", |
1da177e4 LT |
3206 | bond_mode_name(bond->params.mode)); |
3207 | ||
dd957c57 JV |
3208 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP && |
3209 | bond->params.fail_over_mac) | |
3915c1e8 JV |
3210 | seq_printf(seq, " (fail_over_mac %s)", |
3211 | fail_over_mac_tbl[bond->params.fail_over_mac].modename); | |
dd957c57 JV |
3212 | |
3213 | seq_printf(seq, "\n"); | |
3214 | ||
c61b75ad MW |
3215 | if (bond->params.mode == BOND_MODE_XOR || |
3216 | bond->params.mode == BOND_MODE_8023AD) { | |
3217 | seq_printf(seq, "Transmit Hash Policy: %s (%d)\n", | |
3218 | xmit_hashtype_tbl[bond->params.xmit_policy].modename, | |
3219 | bond->params.xmit_policy); | |
3220 | } | |
3221 | ||
1da177e4 | 3222 | if (USES_PRIMARY(bond->params.mode)) { |
a549952a | 3223 | seq_printf(seq, "Primary Slave: %s", |
0f418b2a MW |
3224 | (bond->primary_slave) ? |
3225 | bond->primary_slave->dev->name : "None"); | |
a549952a JP |
3226 | if (bond->primary_slave) |
3227 | seq_printf(seq, " (primary_reselect %s)", | |
3228 | pri_reselect_tbl[bond->params.primary_reselect].modename); | |
1da177e4 | 3229 | |
a549952a | 3230 | seq_printf(seq, "\nCurrently Active Slave: %s\n", |
1da177e4 LT |
3231 | (curr) ? curr->dev->name : "None"); |
3232 | } | |
3233 | ||
ff59c456 JV |
3234 | seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ? |
3235 | "up" : "down"); | |
1da177e4 LT |
3236 | seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon); |
3237 | seq_printf(seq, "Up Delay (ms): %d\n", | |
3238 | bond->params.updelay * bond->params.miimon); | |
3239 | seq_printf(seq, "Down Delay (ms): %d\n", | |
3240 | bond->params.downdelay * bond->params.miimon); | |
3241 | ||
4756b02f MW |
3242 | |
3243 | /* ARP information */ | |
3d632c3f SH |
3244 | if (bond->params.arp_interval > 0) { |
3245 | int printed = 0; | |
4756b02f MW |
3246 | seq_printf(seq, "ARP Polling Interval (ms): %d\n", |
3247 | bond->params.arp_interval); | |
3248 | ||
3249 | seq_printf(seq, "ARP IP target/s (n.n.n.n form):"); | |
3250 | ||
3d632c3f | 3251 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
4756b02f | 3252 | if (!bond->params.arp_targets[i]) |
5a31bec0 | 3253 | break; |
4756b02f MW |
3254 | if (printed) |
3255 | seq_printf(seq, ","); | |
8cf14e38 | 3256 | seq_printf(seq, " %pI4", &bond->params.arp_targets[i]); |
4756b02f MW |
3257 | printed = 1; |
3258 | } | |
3259 | seq_printf(seq, "\n"); | |
3260 | } | |
3261 | ||
1da177e4 LT |
3262 | if (bond->params.mode == BOND_MODE_8023AD) { |
3263 | struct ad_info ad_info; | |
3264 | ||
3265 | seq_puts(seq, "\n802.3ad info\n"); | |
3266 | seq_printf(seq, "LACP rate: %s\n", | |
3267 | (bond->params.lacp_fast) ? "fast" : "slow"); | |
fd989c83 JV |
3268 | seq_printf(seq, "Aggregator selection policy (ad_select): %s\n", |
3269 | ad_select_tbl[bond->params.ad_select].modename); | |
1da177e4 LT |
3270 | |
3271 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3272 | seq_printf(seq, "bond %s has no active aggregator\n", | |
3273 | bond->dev->name); | |
3274 | } else { | |
3275 | seq_printf(seq, "Active Aggregator Info:\n"); | |
3276 | ||
3277 | seq_printf(seq, "\tAggregator ID: %d\n", | |
3278 | ad_info.aggregator_id); | |
3279 | seq_printf(seq, "\tNumber of ports: %d\n", | |
3280 | ad_info.ports); | |
3281 | seq_printf(seq, "\tActor Key: %d\n", | |
3282 | ad_info.actor_key); | |
3283 | seq_printf(seq, "\tPartner Key: %d\n", | |
3284 | ad_info.partner_key); | |
e174961c JB |
3285 | seq_printf(seq, "\tPartner Mac Address: %pM\n", |
3286 | ad_info.partner_system); | |
1da177e4 LT |
3287 | } |
3288 | } | |
3289 | } | |
3290 | ||
3d632c3f SH |
3291 | static void bond_info_show_slave(struct seq_file *seq, |
3292 | const struct slave *slave) | |
1da177e4 LT |
3293 | { |
3294 | struct bonding *bond = seq->private; | |
3295 | ||
3296 | seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name); | |
3297 | seq_printf(seq, "MII Status: %s\n", | |
3298 | (slave->link == BOND_LINK_UP) ? "up" : "down"); | |
65509645 | 3299 | seq_printf(seq, "Link Failure Count: %u\n", |
1da177e4 LT |
3300 | slave->link_failure_count); |
3301 | ||
e174961c | 3302 | seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr); |
1da177e4 LT |
3303 | |
3304 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3305 | const struct aggregator *agg | |
3306 | = SLAVE_AD_INFO(slave).port.aggregator; | |
3307 | ||
3d632c3f | 3308 | if (agg) |
1da177e4 LT |
3309 | seq_printf(seq, "Aggregator ID: %d\n", |
3310 | agg->aggregator_identifier); | |
3d632c3f | 3311 | else |
1da177e4 | 3312 | seq_puts(seq, "Aggregator ID: N/A\n"); |
1da177e4 LT |
3313 | } |
3314 | } | |
3315 | ||
3316 | static int bond_info_seq_show(struct seq_file *seq, void *v) | |
3317 | { | |
3318 | if (v == SEQ_START_TOKEN) { | |
3319 | seq_printf(seq, "%s\n", version); | |
3320 | bond_info_show_master(seq); | |
3d632c3f | 3321 | } else |
1da177e4 | 3322 | bond_info_show_slave(seq, v); |
1da177e4 LT |
3323 | |
3324 | return 0; | |
3325 | } | |
3326 | ||
4101dec9 | 3327 | static const struct seq_operations bond_info_seq_ops = { |
1da177e4 LT |
3328 | .start = bond_info_seq_start, |
3329 | .next = bond_info_seq_next, | |
3330 | .stop = bond_info_seq_stop, | |
3331 | .show = bond_info_seq_show, | |
3332 | }; | |
3333 | ||
3334 | static int bond_info_open(struct inode *inode, struct file *file) | |
3335 | { | |
3336 | struct seq_file *seq; | |
3337 | struct proc_dir_entry *proc; | |
3338 | int res; | |
3339 | ||
3340 | res = seq_open(file, &bond_info_seq_ops); | |
3341 | if (!res) { | |
3342 | /* recover the pointer buried in proc_dir_entry data */ | |
3343 | seq = file->private_data; | |
3344 | proc = PDE(inode); | |
3345 | seq->private = proc->data; | |
3346 | } | |
3347 | ||
3348 | return res; | |
3349 | } | |
3350 | ||
d54b1fdb | 3351 | static const struct file_operations bond_info_fops = { |
1da177e4 LT |
3352 | .owner = THIS_MODULE, |
3353 | .open = bond_info_open, | |
3354 | .read = seq_read, | |
3355 | .llseek = seq_lseek, | |
3356 | .release = seq_release, | |
3357 | }; | |
3358 | ||
38fc0026 | 3359 | static void bond_create_proc_entry(struct bonding *bond) |
1da177e4 LT |
3360 | { |
3361 | struct net_device *bond_dev = bond->dev; | |
3362 | ||
3363 | if (bond_proc_dir) { | |
a95609cb DL |
3364 | bond->proc_entry = proc_create_data(bond_dev->name, |
3365 | S_IRUGO, bond_proc_dir, | |
3366 | &bond_info_fops, bond); | |
3d632c3f SH |
3367 | if (bond->proc_entry == NULL) |
3368 | pr_warning(DRV_NAME | |
1da177e4 LT |
3369 | ": Warning: Cannot create /proc/net/%s/%s\n", |
3370 | DRV_NAME, bond_dev->name); | |
3d632c3f | 3371 | else |
1da177e4 | 3372 | memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ); |
1da177e4 | 3373 | } |
1da177e4 LT |
3374 | } |
3375 | ||
3376 | static void bond_remove_proc_entry(struct bonding *bond) | |
3377 | { | |
3378 | if (bond_proc_dir && bond->proc_entry) { | |
3379 | remove_proc_entry(bond->proc_file_name, bond_proc_dir); | |
3380 | memset(bond->proc_file_name, 0, IFNAMSIZ); | |
3381 | bond->proc_entry = NULL; | |
3382 | } | |
3383 | } | |
3384 | ||
3385 | /* Create the bonding directory under /proc/net, if doesn't exist yet. | |
3386 | * Caller must hold rtnl_lock. | |
3387 | */ | |
3388 | static void bond_create_proc_dir(void) | |
3389 | { | |
1da177e4 | 3390 | if (!bond_proc_dir) { |
457c4cbc | 3391 | bond_proc_dir = proc_mkdir(DRV_NAME, init_net.proc_net); |
99b76233 | 3392 | if (!bond_proc_dir) |
3d632c3f | 3393 | pr_warning(DRV_NAME |
1da177e4 LT |
3394 | ": Warning: cannot create /proc/net/%s\n", |
3395 | DRV_NAME); | |
1da177e4 LT |
3396 | } |
3397 | } | |
3398 | ||
3399 | /* Destroy the bonding directory under /proc/net, if empty. | |
3400 | * Caller must hold rtnl_lock. | |
3401 | */ | |
3402 | static void bond_destroy_proc_dir(void) | |
3403 | { | |
99b76233 | 3404 | if (bond_proc_dir) { |
457c4cbc | 3405 | remove_proc_entry(DRV_NAME, init_net.proc_net); |
1da177e4 LT |
3406 | bond_proc_dir = NULL; |
3407 | } | |
3408 | } | |
aee64faf JP |
3409 | |
3410 | #else /* !CONFIG_PROC_FS */ | |
3411 | ||
38fc0026 | 3412 | static void bond_create_proc_entry(struct bonding *bond) |
aee64faf JP |
3413 | { |
3414 | } | |
3415 | ||
3416 | static void bond_remove_proc_entry(struct bonding *bond) | |
3417 | { | |
3418 | } | |
3419 | ||
3420 | static void bond_create_proc_dir(void) | |
3421 | { | |
3422 | } | |
3423 | ||
3424 | static void bond_destroy_proc_dir(void) | |
3425 | { | |
3426 | } | |
3427 | ||
1da177e4 LT |
3428 | #endif /* CONFIG_PROC_FS */ |
3429 | ||
aee64faf | 3430 | |
1da177e4 LT |
3431 | /*-------------------------- netdev event handling --------------------------*/ |
3432 | ||
3433 | /* | |
3434 | * Change device name | |
3435 | */ | |
3436 | static int bond_event_changename(struct bonding *bond) | |
3437 | { | |
1da177e4 LT |
3438 | bond_remove_proc_entry(bond); |
3439 | bond_create_proc_entry(bond); | |
7e083840 | 3440 | |
1da177e4 LT |
3441 | return NOTIFY_DONE; |
3442 | } | |
3443 | ||
3d632c3f SH |
3444 | static int bond_master_netdev_event(unsigned long event, |
3445 | struct net_device *bond_dev) | |
1da177e4 | 3446 | { |
454d7c9b | 3447 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3448 | |
3449 | switch (event) { | |
3450 | case NETDEV_CHANGENAME: | |
3451 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3452 | default: |
3453 | break; | |
3454 | } | |
3455 | ||
3456 | return NOTIFY_DONE; | |
3457 | } | |
3458 | ||
3d632c3f SH |
3459 | static int bond_slave_netdev_event(unsigned long event, |
3460 | struct net_device *slave_dev) | |
1da177e4 LT |
3461 | { |
3462 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3463 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3464 | |
3465 | switch (event) { | |
3466 | case NETDEV_UNREGISTER: | |
3467 | if (bond_dev) { | |
1284cd3a JV |
3468 | if (bond->setup_by_slave) |
3469 | bond_release_and_destroy(bond_dev, slave_dev); | |
3470 | else | |
3471 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3472 | } |
3473 | break; | |
3474 | case NETDEV_CHANGE: | |
17d04500 JV |
3475 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3476 | struct slave *slave; | |
3477 | ||
3478 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3479 | if (slave) { | |
3480 | u16 old_speed = slave->speed; | |
3481 | u16 old_duplex = slave->duplex; | |
3482 | ||
3483 | bond_update_speed_duplex(slave); | |
3484 | ||
3485 | if (bond_is_lb(bond)) | |
3486 | break; | |
3487 | ||
3488 | if (old_speed != slave->speed) | |
3489 | bond_3ad_adapter_speed_changed(slave); | |
3490 | if (old_duplex != slave->duplex) | |
3491 | bond_3ad_adapter_duplex_changed(slave); | |
3492 | } | |
3493 | } | |
3494 | ||
1da177e4 LT |
3495 | break; |
3496 | case NETDEV_DOWN: | |
3497 | /* | |
3498 | * ... Or is it this? | |
3499 | */ | |
3500 | break; | |
3501 | case NETDEV_CHANGEMTU: | |
3502 | /* | |
3503 | * TODO: Should slaves be allowed to | |
3504 | * independently alter their MTU? For | |
3505 | * an active-backup bond, slaves need | |
3506 | * not be the same type of device, so | |
3507 | * MTUs may vary. For other modes, | |
3508 | * slaves arguably should have the | |
3509 | * same MTUs. To do this, we'd need to | |
3510 | * take over the slave's change_mtu | |
3511 | * function for the duration of their | |
3512 | * servitude. | |
3513 | */ | |
3514 | break; | |
3515 | case NETDEV_CHANGENAME: | |
3516 | /* | |
3517 | * TODO: handle changing the primary's name | |
3518 | */ | |
3519 | break; | |
8531c5ff AK |
3520 | case NETDEV_FEAT_CHANGE: |
3521 | bond_compute_features(bond); | |
3522 | break; | |
1da177e4 LT |
3523 | default: |
3524 | break; | |
3525 | } | |
3526 | ||
3527 | return NOTIFY_DONE; | |
3528 | } | |
3529 | ||
3530 | /* | |
3531 | * bond_netdev_event: handle netdev notifier chain events. | |
3532 | * | |
3533 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3534 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3535 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3536 | * dev_probe_lock). | |
3537 | */ | |
3d632c3f SH |
3538 | static int bond_netdev_event(struct notifier_block *this, |
3539 | unsigned long event, void *ptr) | |
1da177e4 LT |
3540 | { |
3541 | struct net_device *event_dev = (struct net_device *)ptr; | |
3542 | ||
c346dca1 | 3543 | if (dev_net(event_dev) != &init_net) |
e9dc8653 EB |
3544 | return NOTIFY_DONE; |
3545 | ||
5a03cdb7 | 3546 | pr_debug("event_dev: %s, event: %lx\n", |
1da177e4 LT |
3547 | (event_dev ? event_dev->name : "None"), |
3548 | event); | |
3549 | ||
0b680e75 JV |
3550 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3551 | return NOTIFY_DONE; | |
3552 | ||
1da177e4 | 3553 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3554 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3555 | return bond_master_netdev_event(event, event_dev); |
3556 | } | |
3557 | ||
3558 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3559 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3560 | return bond_slave_netdev_event(event, event_dev); |
3561 | } | |
3562 | ||
3563 | return NOTIFY_DONE; | |
3564 | } | |
3565 | ||
c3ade5ca JV |
3566 | /* |
3567 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3568 | * | |
3569 | * We keep track of device IPs primarily to use as source addresses in | |
3570 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3571 | * | |
3572 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3573 | */ | |
3574 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3575 | { | |
3576 | struct in_ifaddr *ifa = ptr; | |
3577 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
0883beca PE |
3578 | struct bonding *bond; |
3579 | struct vlan_entry *vlan; | |
c3ade5ca | 3580 | |
c346dca1 | 3581 | if (dev_net(ifa->ifa_dev->dev) != &init_net) |
6133fb1a DL |
3582 | return NOTIFY_DONE; |
3583 | ||
0883beca | 3584 | list_for_each_entry(bond, &bond_dev_list, bond_list) { |
c3ade5ca JV |
3585 | if (bond->dev == event_dev) { |
3586 | switch (event) { | |
3587 | case NETDEV_UP: | |
3588 | bond->master_ip = ifa->ifa_local; | |
3589 | return NOTIFY_OK; | |
3590 | case NETDEV_DOWN: | |
3591 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3592 | return NOTIFY_OK; | |
3593 | default: | |
3594 | return NOTIFY_DONE; | |
3595 | } | |
3596 | } | |
3597 | ||
0883beca | 3598 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 3599 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3600 | if (vlan_dev == event_dev) { |
3601 | switch (event) { | |
3602 | case NETDEV_UP: | |
3603 | vlan->vlan_ip = ifa->ifa_local; | |
3604 | return NOTIFY_OK; | |
3605 | case NETDEV_DOWN: | |
3606 | vlan->vlan_ip = | |
3607 | bond_glean_dev_ip(vlan_dev); | |
3608 | return NOTIFY_OK; | |
3609 | default: | |
3610 | return NOTIFY_DONE; | |
3611 | } | |
3612 | } | |
3613 | } | |
3614 | } | |
3615 | return NOTIFY_DONE; | |
3616 | } | |
3617 | ||
1da177e4 LT |
3618 | static struct notifier_block bond_netdev_notifier = { |
3619 | .notifier_call = bond_netdev_event, | |
3620 | }; | |
3621 | ||
c3ade5ca JV |
3622 | static struct notifier_block bond_inetaddr_notifier = { |
3623 | .notifier_call = bond_inetaddr_event, | |
3624 | }; | |
3625 | ||
1da177e4 LT |
3626 | /*-------------------------- Packet type handling ---------------------------*/ |
3627 | ||
3628 | /* register to receive lacpdus on a bond */ | |
3629 | static void bond_register_lacpdu(struct bonding *bond) | |
3630 | { | |
3631 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3632 | ||
3633 | /* initialize packet type */ | |
3634 | pk_type->type = PKT_TYPE_LACPDU; | |
3635 | pk_type->dev = bond->dev; | |
3636 | pk_type->func = bond_3ad_lacpdu_recv; | |
3637 | ||
3638 | dev_add_pack(pk_type); | |
3639 | } | |
3640 | ||
3641 | /* unregister to receive lacpdus on a bond */ | |
3642 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3643 | { | |
3644 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3645 | } | |
3646 | ||
f5b2b966 JV |
3647 | void bond_register_arp(struct bonding *bond) |
3648 | { | |
3649 | struct packet_type *pt = &bond->arp_mon_pt; | |
3650 | ||
c4f283b1 JV |
3651 | if (pt->type) |
3652 | return; | |
3653 | ||
f5b2b966 | 3654 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3655 | pt->dev = bond->dev; |
f5b2b966 JV |
3656 | pt->func = bond_arp_rcv; |
3657 | dev_add_pack(pt); | |
3658 | } | |
3659 | ||
3660 | void bond_unregister_arp(struct bonding *bond) | |
3661 | { | |
c4f283b1 JV |
3662 | struct packet_type *pt = &bond->arp_mon_pt; |
3663 | ||
3664 | dev_remove_pack(pt); | |
3665 | pt->type = 0; | |
f5b2b966 JV |
3666 | } |
3667 | ||
169a3e66 JV |
3668 | /*---------------------------- Hashing Policies -----------------------------*/ |
3669 | ||
6f6652be JV |
3670 | /* |
3671 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3672 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3673 | */ | |
a361c83c | 3674 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3675 | { |
3676 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3677 | struct iphdr *iph = ip_hdr(skb); | |
3678 | ||
f14c4e4e | 3679 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3680 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3681 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3682 | } |
3683 | ||
d3da6831 | 3684 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3685 | } |
3686 | ||
169a3e66 | 3687 | /* |
59c51591 | 3688 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3689 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3690 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3691 | */ | |
a361c83c | 3692 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3693 | { |
3694 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3695 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3696 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3697 | int layer4_xor = 0; |
3698 | ||
f14c4e4e BH |
3699 | if (skb->protocol == htons(ETH_P_IP)) { |
3700 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3701 | (iph->protocol == IPPROTO_TCP || |
3702 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3703 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3704 | } |
3705 | return (layer4_xor ^ | |
3706 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3707 | ||
3708 | } | |
3709 | ||
d3da6831 | 3710 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3711 | } |
3712 | ||
3713 | /* | |
3714 | * Hash for the output device based upon layer 2 data | |
3715 | */ | |
a361c83c | 3716 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3717 | { |
3718 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3719 | ||
d3da6831 | 3720 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3721 | } |
3722 | ||
1da177e4 LT |
3723 | /*-------------------------- Device entry points ----------------------------*/ |
3724 | ||
3725 | static int bond_open(struct net_device *bond_dev) | |
3726 | { | |
454d7c9b | 3727 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3728 | |
3729 | bond->kill_timers = 0; | |
3730 | ||
58402054 | 3731 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3732 | /* bond_alb_initialize must be called before the timer |
3733 | * is started. | |
3734 | */ | |
3735 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3736 | /* something went wrong - fail the open operation */ | |
3737 | return -1; | |
3738 | } | |
3739 | ||
1b76b316 JV |
3740 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3741 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3742 | } |
3743 | ||
3744 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3745 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3746 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3747 | } |
3748 | ||
3749 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3750 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3751 | INIT_DELAYED_WORK(&bond->arp_work, | |
3752 | bond_activebackup_arp_mon); | |
3753 | else | |
3754 | INIT_DELAYED_WORK(&bond->arp_work, | |
3755 | bond_loadbalance_arp_mon); | |
3756 | ||
3757 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3758 | if (bond->params.arp_validate) |
3759 | bond_register_arp(bond); | |
1da177e4 LT |
3760 | } |
3761 | ||
3762 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3763 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3764 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3765 | /* register to receive LACPDUs */ |
3766 | bond_register_lacpdu(bond); | |
fd989c83 | 3767 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3768 | } |
3769 | ||
3770 | return 0; | |
3771 | } | |
3772 | ||
3773 | static int bond_close(struct net_device *bond_dev) | |
3774 | { | |
454d7c9b | 3775 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3776 | |
3777 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3778 | /* Unregister the receive of LACPDUs */ | |
3779 | bond_unregister_lacpdu(bond); | |
3780 | } | |
3781 | ||
f5b2b966 JV |
3782 | if (bond->params.arp_validate) |
3783 | bond_unregister_arp(bond); | |
3784 | ||
1da177e4 LT |
3785 | write_lock_bh(&bond->lock); |
3786 | ||
b59f9f74 | 3787 | bond->send_grat_arp = 0; |
305d552a | 3788 | bond->send_unsol_na = 0; |
1da177e4 LT |
3789 | |
3790 | /* signal timers not to re-arm */ | |
3791 | bond->kill_timers = 1; | |
3792 | ||
3793 | write_unlock_bh(&bond->lock); | |
3794 | ||
1da177e4 | 3795 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3796 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3797 | } |
3798 | ||
3799 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3800 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3801 | } |
3802 | ||
3803 | switch (bond->params.mode) { | |
3804 | case BOND_MODE_8023AD: | |
1b76b316 | 3805 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3806 | break; |
3807 | case BOND_MODE_TLB: | |
3808 | case BOND_MODE_ALB: | |
1b76b316 | 3809 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3810 | break; |
3811 | default: | |
3812 | break; | |
3813 | } | |
3814 | ||
1da177e4 | 3815 | |
58402054 | 3816 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3817 | /* Must be called only after all |
3818 | * slaves have been released | |
3819 | */ | |
3820 | bond_alb_deinitialize(bond); | |
3821 | } | |
3822 | ||
3823 | return 0; | |
3824 | } | |
3825 | ||
3826 | static struct net_device_stats *bond_get_stats(struct net_device *bond_dev) | |
3827 | { | |
454d7c9b | 3828 | struct bonding *bond = netdev_priv(bond_dev); |
eeda3fd6 | 3829 | struct net_device_stats *stats = &bond->stats; |
2439f9eb | 3830 | struct net_device_stats local_stats; |
1da177e4 LT |
3831 | struct slave *slave; |
3832 | int i; | |
3833 | ||
2439f9eb | 3834 | memset(&local_stats, 0, sizeof(struct net_device_stats)); |
1da177e4 LT |
3835 | |
3836 | read_lock_bh(&bond->lock); | |
3837 | ||
3838 | bond_for_each_slave(bond, slave, i) { | |
eeda3fd6 SH |
3839 | const struct net_device_stats *sstats = dev_get_stats(slave->dev); |
3840 | ||
2439f9eb AG |
3841 | local_stats.rx_packets += sstats->rx_packets; |
3842 | local_stats.rx_bytes += sstats->rx_bytes; | |
3843 | local_stats.rx_errors += sstats->rx_errors; | |
3844 | local_stats.rx_dropped += sstats->rx_dropped; | |
3845 | ||
3846 | local_stats.tx_packets += sstats->tx_packets; | |
3847 | local_stats.tx_bytes += sstats->tx_bytes; | |
3848 | local_stats.tx_errors += sstats->tx_errors; | |
3849 | local_stats.tx_dropped += sstats->tx_dropped; | |
3850 | ||
3851 | local_stats.multicast += sstats->multicast; | |
3852 | local_stats.collisions += sstats->collisions; | |
3853 | ||
3854 | local_stats.rx_length_errors += sstats->rx_length_errors; | |
3855 | local_stats.rx_over_errors += sstats->rx_over_errors; | |
3856 | local_stats.rx_crc_errors += sstats->rx_crc_errors; | |
3857 | local_stats.rx_frame_errors += sstats->rx_frame_errors; | |
3858 | local_stats.rx_fifo_errors += sstats->rx_fifo_errors; | |
3859 | local_stats.rx_missed_errors += sstats->rx_missed_errors; | |
3860 | ||
3861 | local_stats.tx_aborted_errors += sstats->tx_aborted_errors; | |
3862 | local_stats.tx_carrier_errors += sstats->tx_carrier_errors; | |
3863 | local_stats.tx_fifo_errors += sstats->tx_fifo_errors; | |
3864 | local_stats.tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3865 | local_stats.tx_window_errors += sstats->tx_window_errors; | |
3866 | } | |
3867 | ||
3868 | memcpy(stats, &local_stats, sizeof(struct net_device_stats)); | |
1da177e4 LT |
3869 | |
3870 | read_unlock_bh(&bond->lock); | |
3871 | ||
3872 | return stats; | |
3873 | } | |
3874 | ||
3875 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3876 | { | |
3877 | struct net_device *slave_dev = NULL; | |
3878 | struct ifbond k_binfo; | |
3879 | struct ifbond __user *u_binfo = NULL; | |
3880 | struct ifslave k_sinfo; | |
3881 | struct ifslave __user *u_sinfo = NULL; | |
3882 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3883 | int res = 0; |
3884 | ||
5a03cdb7 | 3885 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", |
1da177e4 LT |
3886 | bond_dev->name, cmd); |
3887 | ||
3888 | switch (cmd) { | |
1da177e4 LT |
3889 | case SIOCGMIIPHY: |
3890 | mii = if_mii(ifr); | |
3d632c3f | 3891 | if (!mii) |
1da177e4 | 3892 | return -EINVAL; |
3d632c3f | 3893 | |
1da177e4 LT |
3894 | mii->phy_id = 0; |
3895 | /* Fall Through */ | |
3896 | case SIOCGMIIREG: | |
3897 | /* | |
3898 | * We do this again just in case we were called by SIOCGMIIREG | |
3899 | * instead of SIOCGMIIPHY. | |
3900 | */ | |
3901 | mii = if_mii(ifr); | |
3d632c3f | 3902 | if (!mii) |
1da177e4 | 3903 | return -EINVAL; |
3d632c3f | 3904 | |
1da177e4 LT |
3905 | |
3906 | if (mii->reg_num == 1) { | |
454d7c9b | 3907 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3908 | mii->val_out = 0; |
6603a6f2 | 3909 | read_lock(&bond->lock); |
1da177e4 | 3910 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3911 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3912 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3913 | |
1da177e4 | 3914 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3915 | read_unlock(&bond->lock); |
1da177e4 LT |
3916 | } |
3917 | ||
3918 | return 0; | |
3919 | case BOND_INFO_QUERY_OLD: | |
3920 | case SIOCBONDINFOQUERY: | |
3921 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3922 | ||
3d632c3f | 3923 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3924 | return -EFAULT; |
1da177e4 LT |
3925 | |
3926 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3927 | if (res == 0 && |
3928 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3929 | return -EFAULT; | |
1da177e4 LT |
3930 | |
3931 | return res; | |
3932 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3933 | case SIOCBONDSLAVEINFOQUERY: | |
3934 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3935 | ||
3d632c3f | 3936 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3937 | return -EFAULT; |
1da177e4 LT |
3938 | |
3939 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3940 | if (res == 0 && |
3941 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3942 | return -EFAULT; | |
1da177e4 LT |
3943 | |
3944 | return res; | |
3945 | default: | |
3946 | /* Go on */ | |
3947 | break; | |
3948 | } | |
3949 | ||
3d632c3f | 3950 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3951 | return -EPERM; |
1da177e4 | 3952 | |
881d966b | 3953 | slave_dev = dev_get_by_name(&init_net, ifr->ifr_slave); |
1da177e4 | 3954 | |
5a03cdb7 | 3955 | pr_debug("slave_dev=%p: \n", slave_dev); |
1da177e4 | 3956 | |
3d632c3f | 3957 | if (!slave_dev) |
1da177e4 | 3958 | res = -ENODEV; |
3d632c3f | 3959 | else { |
5a03cdb7 | 3960 | pr_debug("slave_dev->name=%s: \n", slave_dev->name); |
1da177e4 LT |
3961 | switch (cmd) { |
3962 | case BOND_ENSLAVE_OLD: | |
3963 | case SIOCBONDENSLAVE: | |
3964 | res = bond_enslave(bond_dev, slave_dev); | |
3965 | break; | |
3966 | case BOND_RELEASE_OLD: | |
3967 | case SIOCBONDRELEASE: | |
3968 | res = bond_release(bond_dev, slave_dev); | |
3969 | break; | |
3970 | case BOND_SETHWADDR_OLD: | |
3971 | case SIOCBONDSETHWADDR: | |
3972 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3973 | break; | |
3974 | case BOND_CHANGE_ACTIVE_OLD: | |
3975 | case SIOCBONDCHANGEACTIVE: | |
3976 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3977 | break; | |
3978 | default: | |
3979 | res = -EOPNOTSUPP; | |
3980 | } | |
3981 | ||
3982 | dev_put(slave_dev); | |
3983 | } | |
3984 | ||
1da177e4 LT |
3985 | return res; |
3986 | } | |
3987 | ||
3988 | static void bond_set_multicast_list(struct net_device *bond_dev) | |
3989 | { | |
454d7c9b | 3990 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3991 | struct dev_mc_list *dmi; |
3992 | ||
1da177e4 LT |
3993 | /* |
3994 | * Do promisc before checking multicast_mode | |
3995 | */ | |
3d632c3f | 3996 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
3997 | /* |
3998 | * FIXME: Need to handle the error when one of the multi-slaves | |
3999 | * encounters error. | |
4000 | */ | |
1da177e4 | 4001 | bond_set_promiscuity(bond, 1); |
1da177e4 | 4002 | |
3d632c3f SH |
4003 | |
4004 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 4005 | bond_set_promiscuity(bond, -1); |
3d632c3f | 4006 | |
1da177e4 LT |
4007 | |
4008 | /* set allmulti flag to slaves */ | |
3d632c3f | 4009 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
4010 | /* |
4011 | * FIXME: Need to handle the error when one of the multi-slaves | |
4012 | * encounters error. | |
4013 | */ | |
1da177e4 | 4014 | bond_set_allmulti(bond, 1); |
1da177e4 | 4015 | |
3d632c3f SH |
4016 | |
4017 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 4018 | bond_set_allmulti(bond, -1); |
3d632c3f | 4019 | |
1da177e4 | 4020 | |
80ee5ad2 JV |
4021 | read_lock(&bond->lock); |
4022 | ||
1da177e4 LT |
4023 | bond->flags = bond_dev->flags; |
4024 | ||
4025 | /* looking for addresses to add to slaves' mc list */ | |
4026 | for (dmi = bond_dev->mc_list; dmi; dmi = dmi->next) { | |
3d632c3f | 4027 | if (!bond_mc_list_find_dmi(dmi, bond->mc_list)) |
1da177e4 | 4028 | bond_mc_add(bond, dmi->dmi_addr, dmi->dmi_addrlen); |
1da177e4 LT |
4029 | } |
4030 | ||
4031 | /* looking for addresses to delete from slaves' list */ | |
4032 | for (dmi = bond->mc_list; dmi; dmi = dmi->next) { | |
3d632c3f | 4033 | if (!bond_mc_list_find_dmi(dmi, bond_dev->mc_list)) |
1da177e4 | 4034 | bond_mc_delete(bond, dmi->dmi_addr, dmi->dmi_addrlen); |
1da177e4 LT |
4035 | } |
4036 | ||
4037 | /* save master's multicast list */ | |
4038 | bond_mc_list_destroy(bond); | |
4039 | bond_mc_list_copy(bond_dev->mc_list, bond, GFP_ATOMIC); | |
4040 | ||
80ee5ad2 | 4041 | read_unlock(&bond->lock); |
1da177e4 LT |
4042 | } |
4043 | ||
00829823 SH |
4044 | static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms) |
4045 | { | |
4046 | struct bonding *bond = netdev_priv(dev); | |
4047 | struct slave *slave = bond->first_slave; | |
4048 | ||
4049 | if (slave) { | |
4050 | const struct net_device_ops *slave_ops | |
4051 | = slave->dev->netdev_ops; | |
4052 | if (slave_ops->ndo_neigh_setup) | |
72e2240f | 4053 | return slave_ops->ndo_neigh_setup(slave->dev, parms); |
00829823 SH |
4054 | } |
4055 | return 0; | |
4056 | } | |
4057 | ||
1da177e4 LT |
4058 | /* |
4059 | * Change the MTU of all of a master's slaves to match the master | |
4060 | */ | |
4061 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
4062 | { | |
454d7c9b | 4063 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4064 | struct slave *slave, *stop_at; |
4065 | int res = 0; | |
4066 | int i; | |
4067 | ||
5a03cdb7 | 4068 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
1da177e4 LT |
4069 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
4070 | ||
4071 | /* Can't hold bond->lock with bh disabled here since | |
4072 | * some base drivers panic. On the other hand we can't | |
4073 | * hold bond->lock without bh disabled because we'll | |
4074 | * deadlock. The only solution is to rely on the fact | |
4075 | * that we're under rtnl_lock here, and the slaves | |
4076 | * list won't change. This doesn't solve the problem | |
4077 | * of setting the slave's MTU while it is | |
4078 | * transmitting, but the assumption is that the base | |
4079 | * driver can handle that. | |
4080 | * | |
4081 | * TODO: figure out a way to safely iterate the slaves | |
4082 | * list, but without holding a lock around the actual | |
4083 | * call to the base driver. | |
4084 | */ | |
4085 | ||
4086 | bond_for_each_slave(bond, slave, i) { | |
5a03cdb7 | 4087 | pr_debug("s %p s->p %p c_m %p\n", slave, |
53a3294e | 4088 | slave->prev, slave->dev->netdev_ops->ndo_change_mtu); |
e944ef79 | 4089 | |
1da177e4 LT |
4090 | res = dev_set_mtu(slave->dev, new_mtu); |
4091 | ||
4092 | if (res) { | |
4093 | /* If we failed to set the slave's mtu to the new value | |
4094 | * we must abort the operation even in ACTIVE_BACKUP | |
4095 | * mode, because if we allow the backup slaves to have | |
4096 | * different mtu values than the active slave we'll | |
4097 | * need to change their mtu when doing a failover. That | |
4098 | * means changing their mtu from timer context, which | |
4099 | * is probably not a good idea. | |
4100 | */ | |
5a03cdb7 | 4101 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4102 | goto unwind; |
4103 | } | |
4104 | } | |
4105 | ||
4106 | bond_dev->mtu = new_mtu; | |
4107 | ||
4108 | return 0; | |
4109 | ||
4110 | unwind: | |
4111 | /* unwind from head to the slave that failed */ | |
4112 | stop_at = slave; | |
4113 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4114 | int tmp_res; | |
4115 | ||
4116 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
4117 | if (tmp_res) { | |
5a03cdb7 | 4118 | pr_debug("unwind err %d dev %s\n", tmp_res, |
1da177e4 LT |
4119 | slave->dev->name); |
4120 | } | |
4121 | } | |
4122 | ||
4123 | return res; | |
4124 | } | |
4125 | ||
4126 | /* | |
4127 | * Change HW address | |
4128 | * | |
4129 | * Note that many devices must be down to change the HW address, and | |
4130 | * downing the master releases all slaves. We can make bonds full of | |
4131 | * bonding devices to test this, however. | |
4132 | */ | |
4133 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4134 | { | |
454d7c9b | 4135 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4136 | struct sockaddr *sa = addr, tmp_sa; |
4137 | struct slave *slave, *stop_at; | |
4138 | int res = 0; | |
4139 | int i; | |
4140 | ||
eb7cc59a SH |
4141 | if (bond->params.mode == BOND_MODE_ALB) |
4142 | return bond_alb_set_mac_address(bond_dev, addr); | |
4143 | ||
4144 | ||
5a03cdb7 | 4145 | pr_debug("bond=%p, name=%s\n", bond, (bond_dev ? bond_dev->name : "None")); |
1da177e4 | 4146 | |
dd957c57 | 4147 | /* |
3915c1e8 JV |
4148 | * If fail_over_mac is set to active, do nothing and return |
4149 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 4150 | */ |
3915c1e8 | 4151 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 4152 | return 0; |
2ab82852 | 4153 | |
3d632c3f | 4154 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 4155 | return -EADDRNOTAVAIL; |
1da177e4 LT |
4156 | |
4157 | /* Can't hold bond->lock with bh disabled here since | |
4158 | * some base drivers panic. On the other hand we can't | |
4159 | * hold bond->lock without bh disabled because we'll | |
4160 | * deadlock. The only solution is to rely on the fact | |
4161 | * that we're under rtnl_lock here, and the slaves | |
4162 | * list won't change. This doesn't solve the problem | |
4163 | * of setting the slave's hw address while it is | |
4164 | * transmitting, but the assumption is that the base | |
4165 | * driver can handle that. | |
4166 | * | |
4167 | * TODO: figure out a way to safely iterate the slaves | |
4168 | * list, but without holding a lock around the actual | |
4169 | * call to the base driver. | |
4170 | */ | |
4171 | ||
4172 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 4173 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 4174 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 4175 | |
eb7cc59a | 4176 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 4177 | res = -EOPNOTSUPP; |
5a03cdb7 | 4178 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
4179 | goto unwind; |
4180 | } | |
4181 | ||
4182 | res = dev_set_mac_address(slave->dev, addr); | |
4183 | if (res) { | |
4184 | /* TODO: consider downing the slave | |
4185 | * and retry ? | |
4186 | * User should expect communications | |
4187 | * breakage anyway until ARP finish | |
4188 | * updating, so... | |
4189 | */ | |
5a03cdb7 | 4190 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4191 | goto unwind; |
4192 | } | |
4193 | } | |
4194 | ||
4195 | /* success */ | |
4196 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4197 | return 0; | |
4198 | ||
4199 | unwind: | |
4200 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4201 | tmp_sa.sa_family = bond_dev->type; | |
4202 | ||
4203 | /* unwind from head to the slave that failed */ | |
4204 | stop_at = slave; | |
4205 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4206 | int tmp_res; | |
4207 | ||
4208 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4209 | if (tmp_res) { | |
5a03cdb7 | 4210 | pr_debug("unwind err %d dev %s\n", tmp_res, |
1da177e4 LT |
4211 | slave->dev->name); |
4212 | } | |
4213 | } | |
4214 | ||
4215 | return res; | |
4216 | } | |
4217 | ||
4218 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4219 | { | |
454d7c9b | 4220 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4221 | struct slave *slave, *start_at; |
cf5f9044 | 4222 | int i, slave_no, res = 1; |
1da177e4 LT |
4223 | |
4224 | read_lock(&bond->lock); | |
4225 | ||
3d632c3f | 4226 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4227 | goto out; |
1da177e4 | 4228 | |
cf5f9044 JV |
4229 | /* |
4230 | * Concurrent TX may collide on rr_tx_counter; we accept that | |
4231 | * as being rare enough not to justify using an atomic op here | |
4232 | */ | |
4233 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
1da177e4 | 4234 | |
cf5f9044 JV |
4235 | bond_for_each_slave(bond, slave, i) { |
4236 | slave_no--; | |
3d632c3f | 4237 | if (slave_no < 0) |
cf5f9044 | 4238 | break; |
1da177e4 LT |
4239 | } |
4240 | ||
cf5f9044 | 4241 | start_at = slave; |
1da177e4 LT |
4242 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4243 | if (IS_UP(slave->dev) && | |
4244 | (slave->link == BOND_LINK_UP) && | |
4245 | (slave->state == BOND_STATE_ACTIVE)) { | |
4246 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
1da177e4 LT |
4247 | break; |
4248 | } | |
4249 | } | |
4250 | ||
1da177e4 LT |
4251 | out: |
4252 | if (res) { | |
4253 | /* no suitable interface, frame not sent */ | |
4254 | dev_kfree_skb(skb); | |
4255 | } | |
4256 | read_unlock(&bond->lock); | |
ec634fe3 | 4257 | return NETDEV_TX_OK; |
1da177e4 LT |
4258 | } |
4259 | ||
075897ce | 4260 | |
1da177e4 LT |
4261 | /* |
4262 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4263 | * the bond has a usable interface. | |
4264 | */ | |
4265 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4266 | { | |
454d7c9b | 4267 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4268 | int res = 1; |
4269 | ||
1da177e4 LT |
4270 | read_lock(&bond->lock); |
4271 | read_lock(&bond->curr_slave_lock); | |
4272 | ||
3d632c3f | 4273 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4274 | goto out; |
1da177e4 | 4275 | |
075897ce JL |
4276 | if (!bond->curr_active_slave) |
4277 | goto out; | |
4278 | ||
075897ce JL |
4279 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4280 | ||
1da177e4 | 4281 | out: |
3d632c3f | 4282 | if (res) |
1da177e4 LT |
4283 | /* no suitable interface, frame not sent */ |
4284 | dev_kfree_skb(skb); | |
3d632c3f | 4285 | |
1da177e4 LT |
4286 | read_unlock(&bond->curr_slave_lock); |
4287 | read_unlock(&bond->lock); | |
ec634fe3 | 4288 | return NETDEV_TX_OK; |
1da177e4 LT |
4289 | } |
4290 | ||
4291 | /* | |
169a3e66 JV |
4292 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4293 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4294 | * find the next active slave. | |
1da177e4 LT |
4295 | */ |
4296 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4297 | { | |
454d7c9b | 4298 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4299 | struct slave *slave, *start_at; |
4300 | int slave_no; | |
4301 | int i; | |
4302 | int res = 1; | |
4303 | ||
4304 | read_lock(&bond->lock); | |
4305 | ||
3d632c3f | 4306 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4307 | goto out; |
1da177e4 | 4308 | |
a361c83c | 4309 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4310 | |
4311 | bond_for_each_slave(bond, slave, i) { | |
4312 | slave_no--; | |
3d632c3f | 4313 | if (slave_no < 0) |
1da177e4 | 4314 | break; |
1da177e4 LT |
4315 | } |
4316 | ||
4317 | start_at = slave; | |
4318 | ||
4319 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4320 | if (IS_UP(slave->dev) && | |
4321 | (slave->link == BOND_LINK_UP) && | |
4322 | (slave->state == BOND_STATE_ACTIVE)) { | |
4323 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4324 | break; | |
4325 | } | |
4326 | } | |
4327 | ||
4328 | out: | |
4329 | if (res) { | |
4330 | /* no suitable interface, frame not sent */ | |
4331 | dev_kfree_skb(skb); | |
4332 | } | |
4333 | read_unlock(&bond->lock); | |
ec634fe3 | 4334 | return NETDEV_TX_OK; |
1da177e4 LT |
4335 | } |
4336 | ||
4337 | /* | |
4338 | * in broadcast mode, we send everything to all usable interfaces. | |
4339 | */ | |
4340 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4341 | { | |
454d7c9b | 4342 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4343 | struct slave *slave, *start_at; |
4344 | struct net_device *tx_dev = NULL; | |
4345 | int i; | |
4346 | int res = 1; | |
4347 | ||
4348 | read_lock(&bond->lock); | |
4349 | ||
3d632c3f | 4350 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4351 | goto out; |
1da177e4 LT |
4352 | |
4353 | read_lock(&bond->curr_slave_lock); | |
4354 | start_at = bond->curr_active_slave; | |
4355 | read_unlock(&bond->curr_slave_lock); | |
4356 | ||
3d632c3f | 4357 | if (!start_at) |
1da177e4 | 4358 | goto out; |
1da177e4 LT |
4359 | |
4360 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4361 | if (IS_UP(slave->dev) && | |
4362 | (slave->link == BOND_LINK_UP) && | |
4363 | (slave->state == BOND_STATE_ACTIVE)) { | |
4364 | if (tx_dev) { | |
4365 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4366 | if (!skb2) { | |
3d632c3f | 4367 | pr_err(DRV_NAME |
4e0952c7 MW |
4368 | ": %s: Error: bond_xmit_broadcast(): " |
4369 | "skb_clone() failed\n", | |
4370 | bond_dev->name); | |
1da177e4 LT |
4371 | continue; |
4372 | } | |
4373 | ||
4374 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4375 | if (res) { | |
4376 | dev_kfree_skb(skb2); | |
4377 | continue; | |
4378 | } | |
4379 | } | |
4380 | tx_dev = slave->dev; | |
4381 | } | |
4382 | } | |
4383 | ||
3d632c3f | 4384 | if (tx_dev) |
1da177e4 | 4385 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4386 | |
4387 | out: | |
3d632c3f | 4388 | if (res) |
1da177e4 LT |
4389 | /* no suitable interface, frame not sent */ |
4390 | dev_kfree_skb(skb); | |
3d632c3f | 4391 | |
1da177e4 LT |
4392 | /* frame sent to all suitable interfaces */ |
4393 | read_unlock(&bond->lock); | |
ec634fe3 | 4394 | return NETDEV_TX_OK; |
1da177e4 LT |
4395 | } |
4396 | ||
4397 | /*------------------------- Device initialization ---------------------------*/ | |
4398 | ||
6f6652be JV |
4399 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4400 | { | |
4401 | switch (bond->params.xmit_policy) { | |
4402 | case BOND_XMIT_POLICY_LAYER23: | |
4403 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4404 | break; | |
4405 | case BOND_XMIT_POLICY_LAYER34: | |
4406 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4407 | break; | |
4408 | case BOND_XMIT_POLICY_LAYER2: | |
4409 | default: | |
4410 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4411 | break; | |
4412 | } | |
4413 | } | |
4414 | ||
424efe9c | 4415 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 SH |
4416 | { |
4417 | const struct bonding *bond = netdev_priv(dev); | |
4418 | ||
4419 | switch (bond->params.mode) { | |
4420 | case BOND_MODE_ROUNDROBIN: | |
4421 | return bond_xmit_roundrobin(skb, dev); | |
4422 | case BOND_MODE_ACTIVEBACKUP: | |
4423 | return bond_xmit_activebackup(skb, dev); | |
4424 | case BOND_MODE_XOR: | |
4425 | return bond_xmit_xor(skb, dev); | |
4426 | case BOND_MODE_BROADCAST: | |
4427 | return bond_xmit_broadcast(skb, dev); | |
4428 | case BOND_MODE_8023AD: | |
4429 | return bond_3ad_xmit_xor(skb, dev); | |
4430 | case BOND_MODE_ALB: | |
4431 | case BOND_MODE_TLB: | |
4432 | return bond_alb_xmit(skb, dev); | |
4433 | default: | |
4434 | /* Should never happen, mode already checked */ | |
3d632c3f | 4435 | pr_err(DRV_NAME ": %s: Error: Unknown bonding mode %d\n", |
00829823 SH |
4436 | dev->name, bond->params.mode); |
4437 | WARN_ON_ONCE(1); | |
4438 | dev_kfree_skb(skb); | |
4439 | return NETDEV_TX_OK; | |
4440 | } | |
4441 | } | |
4442 | ||
4443 | ||
1da177e4 LT |
4444 | /* |
4445 | * set bond mode specific net device operations | |
4446 | */ | |
a77b5325 | 4447 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4448 | { |
169a3e66 JV |
4449 | struct net_device *bond_dev = bond->dev; |
4450 | ||
1da177e4 LT |
4451 | switch (mode) { |
4452 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4453 | break; |
4454 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4455 | break; |
4456 | case BOND_MODE_XOR: | |
6f6652be | 4457 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4458 | break; |
4459 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4460 | break; |
4461 | case BOND_MODE_8023AD: | |
8f903c70 | 4462 | bond_set_master_3ad_flags(bond); |
6f6652be | 4463 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4464 | break; |
1da177e4 | 4465 | case BOND_MODE_ALB: |
8f903c70 JV |
4466 | bond_set_master_alb_flags(bond); |
4467 | /* FALLTHRU */ | |
4468 | case BOND_MODE_TLB: | |
1da177e4 LT |
4469 | break; |
4470 | default: | |
4471 | /* Should never happen, mode already checked */ | |
3d632c3f | 4472 | pr_err(DRV_NAME |
4e0952c7 MW |
4473 | ": %s: Error: Unknown bonding mode %d\n", |
4474 | bond_dev->name, | |
1da177e4 LT |
4475 | mode); |
4476 | break; | |
4477 | } | |
4478 | } | |
4479 | ||
217df670 JV |
4480 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4481 | struct ethtool_drvinfo *drvinfo) | |
4482 | { | |
4483 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4484 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4485 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4486 | } | |
4487 | ||
7282d491 | 4488 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4489 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4490 | .get_link = ethtool_op_get_link, |
4491 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4492 | .get_sg = ethtool_op_get_sg, | |
4493 | .get_tso = ethtool_op_get_tso, | |
4494 | .get_ufo = ethtool_op_get_ufo, | |
4495 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4496 | }; |
4497 | ||
eb7cc59a | 4498 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4499 | .ndo_init = bond_init, |
9e71626c | 4500 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4501 | .ndo_open = bond_open, |
4502 | .ndo_stop = bond_close, | |
00829823 | 4503 | .ndo_start_xmit = bond_start_xmit, |
eb7cc59a SH |
4504 | .ndo_get_stats = bond_get_stats, |
4505 | .ndo_do_ioctl = bond_do_ioctl, | |
4506 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4507 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4508 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4509 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4510 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4511 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4512 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
4513 | }; | |
4514 | ||
181470fc | 4515 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4516 | { |
454d7c9b | 4517 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4518 | |
1da177e4 LT |
4519 | /* initialize rwlocks */ |
4520 | rwlock_init(&bond->lock); | |
4521 | rwlock_init(&bond->curr_slave_lock); | |
4522 | ||
d2991f75 | 4523 | bond->params = bonding_defaults; |
1da177e4 LT |
4524 | |
4525 | /* Initialize pointers */ | |
1da177e4 LT |
4526 | bond->dev = bond_dev; |
4527 | INIT_LIST_HEAD(&bond->vlan_list); | |
4528 | ||
4529 | /* Initialize the device entry points */ | |
181470fc | 4530 | ether_setup(bond_dev); |
eb7cc59a | 4531 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4532 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4533 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4534 | |
9e71626c | 4535 | bond_dev->destructor = free_netdev; |
1da177e4 LT |
4536 | |
4537 | /* Initialize the device options */ | |
4538 | bond_dev->tx_queue_len = 0; | |
4539 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4540 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4541 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4542 | ||
6cf3f41e JV |
4543 | if (bond->params.arp_interval) |
4544 | bond_dev->priv_flags |= IFF_MASTER_ARPMON; | |
1da177e4 LT |
4545 | |
4546 | /* At first, we block adding VLANs. That's the only way to | |
4547 | * prevent problems that occur when adding VLANs over an | |
4548 | * empty bond. The block will be removed once non-challenged | |
4549 | * slaves are enslaved. | |
4550 | */ | |
4551 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4552 | ||
932ff279 | 4553 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4554 | * transmitting */ |
4555 | bond_dev->features |= NETIF_F_LLTX; | |
4556 | ||
4557 | /* By default, we declare the bond to be fully | |
4558 | * VLAN hardware accelerated capable. Special | |
4559 | * care is taken in the various xmit functions | |
4560 | * when there are slaves that are not hw accel | |
4561 | * capable | |
4562 | */ | |
1da177e4 LT |
4563 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4564 | NETIF_F_HW_VLAN_RX | | |
4565 | NETIF_F_HW_VLAN_FILTER); | |
4566 | ||
1da177e4 LT |
4567 | } |
4568 | ||
fdaea7a9 JV |
4569 | static void bond_work_cancel_all(struct bonding *bond) |
4570 | { | |
4571 | write_lock_bh(&bond->lock); | |
4572 | bond->kill_timers = 1; | |
4573 | write_unlock_bh(&bond->lock); | |
4574 | ||
4575 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4576 | cancel_delayed_work(&bond->mii_work); | |
4577 | ||
4578 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4579 | cancel_delayed_work(&bond->arp_work); | |
4580 | ||
4581 | if (bond->params.mode == BOND_MODE_ALB && | |
4582 | delayed_work_pending(&bond->alb_work)) | |
4583 | cancel_delayed_work(&bond->alb_work); | |
4584 | ||
4585 | if (bond->params.mode == BOND_MODE_8023AD && | |
4586 | delayed_work_pending(&bond->ad_work)) | |
4587 | cancel_delayed_work(&bond->ad_work); | |
4588 | } | |
4589 | ||
c67dfb29 EB |
4590 | /* |
4591 | * Destroy a bonding device. | |
4592 | * Must be under rtnl_lock when this function is called. | |
4593 | */ | |
4594 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4595 | { |
454d7c9b | 4596 | struct bonding *bond = netdev_priv(bond_dev); |
a434e43f | 4597 | |
c67dfb29 EB |
4598 | /* Release the bonded slaves */ |
4599 | bond_release_all(bond_dev); | |
4600 | ||
a434e43f JV |
4601 | list_del(&bond->bond_list); |
4602 | ||
4603 | bond_work_cancel_all(bond); | |
4604 | ||
a434e43f | 4605 | bond_remove_proc_entry(bond); |
c67dfb29 EB |
4606 | |
4607 | if (bond->wq) | |
4608 | destroy_workqueue(bond->wq); | |
4609 | ||
4610 | netif_addr_lock_bh(bond_dev); | |
4611 | bond_mc_list_destroy(bond); | |
4612 | netif_addr_unlock_bh(bond_dev); | |
a434e43f JV |
4613 | } |
4614 | ||
1da177e4 LT |
4615 | /*------------------------- Module initialization ---------------------------*/ |
4616 | ||
4617 | /* | |
4618 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4619 | * number of the mode or its string name. A bit complicated because |
4620 | * some mode names are substrings of other names, and calls from sysfs | |
4621 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4622 | */ |
325dcf7a | 4623 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4624 | { |
54b87323 | 4625 | int modeint = -1, i, rv; |
a42e534f | 4626 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4627 | |
a42e534f JV |
4628 | for (p = (char *)buf; *p; p++) |
4629 | if (!(isdigit(*p) || isspace(*p))) | |
4630 | break; | |
4631 | ||
4632 | if (*p) | |
ece95f7f | 4633 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4634 | else |
54b87323 | 4635 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4636 | |
4637 | if (!rv) | |
4638 | return -1; | |
1da177e4 LT |
4639 | |
4640 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4641 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4642 | return tbl[i].mode; |
4643 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4644 | return tbl[i].mode; |
1da177e4 LT |
4645 | } |
4646 | ||
4647 | return -1; | |
4648 | } | |
4649 | ||
4650 | static int bond_check_params(struct bond_params *params) | |
4651 | { | |
a549952a | 4652 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4653 | |
1da177e4 LT |
4654 | /* |
4655 | * Convert string parameters. | |
4656 | */ | |
4657 | if (mode) { | |
4658 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4659 | if (bond_mode == -1) { | |
3d632c3f | 4660 | pr_err(DRV_NAME |
1da177e4 LT |
4661 | ": Error: Invalid bonding mode \"%s\"\n", |
4662 | mode == NULL ? "NULL" : mode); | |
4663 | return -EINVAL; | |
4664 | } | |
4665 | } | |
4666 | ||
169a3e66 JV |
4667 | if (xmit_hash_policy) { |
4668 | if ((bond_mode != BOND_MODE_XOR) && | |
4669 | (bond_mode != BOND_MODE_8023AD)) { | |
3d632c3f | 4670 | pr_info(DRV_NAME |
3c6aaa24 NP |
4671 | ": xmit_hash_policy param is irrelevant in" |
4672 | " mode %s\n", | |
169a3e66 JV |
4673 | bond_mode_name(bond_mode)); |
4674 | } else { | |
4675 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4676 | xmit_hashtype_tbl); | |
4677 | if (xmit_hashtype == -1) { | |
3d632c3f SH |
4678 | pr_err(DRV_NAME |
4679 | ": Error: Invalid xmit_hash_policy \"%s\"\n", | |
4680 | xmit_hash_policy == NULL ? "NULL" : | |
169a3e66 JV |
4681 | xmit_hash_policy); |
4682 | return -EINVAL; | |
4683 | } | |
4684 | } | |
4685 | } | |
4686 | ||
1da177e4 LT |
4687 | if (lacp_rate) { |
4688 | if (bond_mode != BOND_MODE_8023AD) { | |
3d632c3f | 4689 | pr_info(DRV_NAME |
1da177e4 LT |
4690 | ": lacp_rate param is irrelevant in mode %s\n", |
4691 | bond_mode_name(bond_mode)); | |
4692 | } else { | |
4693 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4694 | if (lacp_fast == -1) { | |
3d632c3f | 4695 | pr_err(DRV_NAME |
1da177e4 LT |
4696 | ": Error: Invalid lacp rate \"%s\"\n", |
4697 | lacp_rate == NULL ? "NULL" : lacp_rate); | |
4698 | return -EINVAL; | |
4699 | } | |
4700 | } | |
4701 | } | |
4702 | ||
fd989c83 JV |
4703 | if (ad_select) { |
4704 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4705 | if (params->ad_select == -1) { | |
3d632c3f | 4706 | pr_err(DRV_NAME |
fd989c83 JV |
4707 | ": Error: Invalid ad_select \"%s\"\n", |
4708 | ad_select == NULL ? "NULL" : ad_select); | |
4709 | return -EINVAL; | |
4710 | } | |
4711 | ||
4712 | if (bond_mode != BOND_MODE_8023AD) { | |
3d632c3f | 4713 | pr_warning(DRV_NAME |
fd989c83 JV |
4714 | ": ad_select param only affects 802.3ad mode\n"); |
4715 | } | |
4716 | } else { | |
4717 | params->ad_select = BOND_AD_STABLE; | |
4718 | } | |
4719 | ||
f5841306 | 4720 | if (max_bonds < 0) { |
3d632c3f | 4721 | pr_warning(DRV_NAME |
1da177e4 | 4722 | ": Warning: max_bonds (%d) not in range %d-%d, so it " |
4e0952c7 | 4723 | "was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
b8a9787e | 4724 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); |
1da177e4 LT |
4725 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4726 | } | |
4727 | ||
4728 | if (miimon < 0) { | |
3d632c3f | 4729 | pr_warning(DRV_NAME |
1da177e4 LT |
4730 | ": Warning: miimon module parameter (%d), " |
4731 | "not in range 0-%d, so it was reset to %d\n", | |
4732 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
4733 | miimon = BOND_LINK_MON_INTERV; | |
4734 | } | |
4735 | ||
4736 | if (updelay < 0) { | |
3d632c3f | 4737 | pr_warning(DRV_NAME |
1da177e4 LT |
4738 | ": Warning: updelay module parameter (%d), " |
4739 | "not in range 0-%d, so it was reset to 0\n", | |
4740 | updelay, INT_MAX); | |
4741 | updelay = 0; | |
4742 | } | |
4743 | ||
4744 | if (downdelay < 0) { | |
3d632c3f | 4745 | pr_warning(DRV_NAME |
1da177e4 LT |
4746 | ": Warning: downdelay module parameter (%d), " |
4747 | "not in range 0-%d, so it was reset to 0\n", | |
4748 | downdelay, INT_MAX); | |
4749 | downdelay = 0; | |
4750 | } | |
4751 | ||
4752 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
3d632c3f | 4753 | pr_warning(DRV_NAME |
1da177e4 LT |
4754 | ": Warning: use_carrier module parameter (%d), " |
4755 | "not of valid value (0/1), so it was set to 1\n", | |
4756 | use_carrier); | |
4757 | use_carrier = 1; | |
4758 | } | |
4759 | ||
7893b249 | 4760 | if (num_grat_arp < 0 || num_grat_arp > 255) { |
3d632c3f | 4761 | pr_warning(DRV_NAME |
7893b249 MS |
4762 | ": Warning: num_grat_arp (%d) not in range 0-255 so it " |
4763 | "was reset to 1 \n", num_grat_arp); | |
4764 | num_grat_arp = 1; | |
4765 | } | |
4766 | ||
305d552a | 4767 | if (num_unsol_na < 0 || num_unsol_na > 255) { |
3d632c3f | 4768 | pr_warning(DRV_NAME |
305d552a BH |
4769 | ": Warning: num_unsol_na (%d) not in range 0-255 so it " |
4770 | "was reset to 1 \n", num_unsol_na); | |
4771 | num_unsol_na = 1; | |
4772 | } | |
4773 | ||
1da177e4 LT |
4774 | /* reset values for 802.3ad */ |
4775 | if (bond_mode == BOND_MODE_8023AD) { | |
4776 | if (!miimon) { | |
3d632c3f | 4777 | pr_warning(DRV_NAME |
1da177e4 LT |
4778 | ": Warning: miimon must be specified, " |
4779 | "otherwise bonding will not detect link " | |
4780 | "failure, speed and duplex which are " | |
4781 | "essential for 802.3ad operation\n"); | |
3d632c3f | 4782 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4783 | miimon = 100; |
4784 | } | |
4785 | } | |
4786 | ||
4787 | /* reset values for TLB/ALB */ | |
4788 | if ((bond_mode == BOND_MODE_TLB) || | |
4789 | (bond_mode == BOND_MODE_ALB)) { | |
4790 | if (!miimon) { | |
3d632c3f | 4791 | pr_warning(DRV_NAME |
1da177e4 LT |
4792 | ": Warning: miimon must be specified, " |
4793 | "otherwise bonding will not detect link " | |
4794 | "failure and link speed which are essential " | |
4795 | "for TLB/ALB load balancing\n"); | |
3d632c3f | 4796 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4797 | miimon = 100; |
4798 | } | |
4799 | } | |
4800 | ||
4801 | if (bond_mode == BOND_MODE_ALB) { | |
e5e2a8fd | 4802 | pr_notice(DRV_NAME |
1da177e4 LT |
4803 | ": In ALB mode you might experience client " |
4804 | "disconnections upon reconnection of a link if the " | |
4805 | "bonding module updelay parameter (%d msec) is " | |
4806 | "incompatible with the forwarding delay time of the " | |
4807 | "switch\n", | |
4808 | updelay); | |
4809 | } | |
4810 | ||
4811 | if (!miimon) { | |
4812 | if (updelay || downdelay) { | |
4813 | /* just warn the user the up/down delay will have | |
4814 | * no effect since miimon is zero... | |
4815 | */ | |
3d632c3f | 4816 | pr_warning(DRV_NAME |
1da177e4 LT |
4817 | ": Warning: miimon module parameter not set " |
4818 | "and updelay (%d) or downdelay (%d) module " | |
4819 | "parameter is set; updelay and downdelay have " | |
4820 | "no effect unless miimon is set\n", | |
4821 | updelay, downdelay); | |
4822 | } | |
4823 | } else { | |
4824 | /* don't allow arp monitoring */ | |
4825 | if (arp_interval) { | |
3d632c3f | 4826 | pr_warning(DRV_NAME |
1da177e4 LT |
4827 | ": Warning: miimon (%d) and arp_interval (%d) " |
4828 | "can't be used simultaneously, disabling ARP " | |
4829 | "monitoring\n", | |
4830 | miimon, arp_interval); | |
4831 | arp_interval = 0; | |
4832 | } | |
4833 | ||
4834 | if ((updelay % miimon) != 0) { | |
3d632c3f | 4835 | pr_warning(DRV_NAME |
1da177e4 LT |
4836 | ": Warning: updelay (%d) is not a multiple " |
4837 | "of miimon (%d), updelay rounded to %d ms\n", | |
4838 | updelay, miimon, (updelay / miimon) * miimon); | |
4839 | } | |
4840 | ||
4841 | updelay /= miimon; | |
4842 | ||
4843 | if ((downdelay % miimon) != 0) { | |
3d632c3f | 4844 | pr_warning(DRV_NAME |
1da177e4 LT |
4845 | ": Warning: downdelay (%d) is not a multiple " |
4846 | "of miimon (%d), downdelay rounded to %d ms\n", | |
4847 | downdelay, miimon, | |
4848 | (downdelay / miimon) * miimon); | |
4849 | } | |
4850 | ||
4851 | downdelay /= miimon; | |
4852 | } | |
4853 | ||
4854 | if (arp_interval < 0) { | |
3d632c3f | 4855 | pr_warning(DRV_NAME |
1da177e4 LT |
4856 | ": Warning: arp_interval module parameter (%d) " |
4857 | ", not in range 0-%d, so it was reset to %d\n", | |
4858 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
4859 | arp_interval = BOND_LINK_ARP_INTERV; | |
4860 | } | |
4861 | ||
4862 | for (arp_ip_count = 0; | |
4863 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4864 | arp_ip_count++) { | |
4865 | /* not complete check, but should be good enough to | |
4866 | catch mistakes */ | |
4867 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
3d632c3f | 4868 | pr_warning(DRV_NAME |
1da177e4 LT |
4869 | ": Warning: bad arp_ip_target module parameter " |
4870 | "(%s), ARP monitoring will not be performed\n", | |
4871 | arp_ip_target[arp_ip_count]); | |
4872 | arp_interval = 0; | |
4873 | } else { | |
d3bb52b0 | 4874 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4875 | arp_target[arp_ip_count] = ip; |
4876 | } | |
4877 | } | |
4878 | ||
4879 | if (arp_interval && !arp_ip_count) { | |
4880 | /* don't allow arping if no arp_ip_target given... */ | |
3d632c3f | 4881 | pr_warning(DRV_NAME |
1da177e4 LT |
4882 | ": Warning: arp_interval module parameter (%d) " |
4883 | "specified without providing an arp_ip_target " | |
4884 | "parameter, arp_interval was reset to 0\n", | |
4885 | arp_interval); | |
4886 | arp_interval = 0; | |
4887 | } | |
4888 | ||
f5b2b966 JV |
4889 | if (arp_validate) { |
4890 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f SH |
4891 | pr_err(DRV_NAME |
4892 | ": arp_validate only supported in active-backup mode\n"); | |
f5b2b966 JV |
4893 | return -EINVAL; |
4894 | } | |
4895 | if (!arp_interval) { | |
3d632c3f | 4896 | pr_err(DRV_NAME |
f5b2b966 JV |
4897 | ": arp_validate requires arp_interval\n"); |
4898 | return -EINVAL; | |
4899 | } | |
4900 | ||
4901 | arp_validate_value = bond_parse_parm(arp_validate, | |
4902 | arp_validate_tbl); | |
4903 | if (arp_validate_value == -1) { | |
3d632c3f | 4904 | pr_err(DRV_NAME |
f5b2b966 JV |
4905 | ": Error: invalid arp_validate \"%s\"\n", |
4906 | arp_validate == NULL ? "NULL" : arp_validate); | |
4907 | return -EINVAL; | |
4908 | } | |
4909 | } else | |
4910 | arp_validate_value = 0; | |
4911 | ||
1da177e4 | 4912 | if (miimon) { |
3d632c3f | 4913 | pr_info(DRV_NAME |
1da177e4 LT |
4914 | ": MII link monitoring set to %d ms\n", |
4915 | miimon); | |
4916 | } else if (arp_interval) { | |
4917 | int i; | |
4918 | ||
3d632c3f SH |
4919 | pr_info(DRV_NAME ": ARP monitoring set to %d ms," |
4920 | " validate %s, with %d target(s):", | |
f5b2b966 JV |
4921 | arp_interval, |
4922 | arp_validate_tbl[arp_validate_value].modename, | |
4923 | arp_ip_count); | |
1da177e4 LT |
4924 | |
4925 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4926 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4927 | |
e5e2a8fd | 4928 | pr_info("\n"); |
1da177e4 | 4929 | |
b8a9787e | 4930 | } else if (max_bonds) { |
1da177e4 LT |
4931 | /* miimon and arp_interval not set, we need one so things |
4932 | * work as expected, see bonding.txt for details | |
4933 | */ | |
3d632c3f | 4934 | pr_warning(DRV_NAME |
1da177e4 LT |
4935 | ": Warning: either miimon or arp_interval and " |
4936 | "arp_ip_target module parameters must be specified, " | |
4937 | "otherwise bonding will not detect link failures! see " | |
4938 | "bonding.txt for details.\n"); | |
4939 | } | |
4940 | ||
4941 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4942 | /* currently, using a primary only makes sense | |
4943 | * in active backup, TLB or ALB modes | |
4944 | */ | |
3d632c3f | 4945 | pr_warning(DRV_NAME |
1da177e4 LT |
4946 | ": Warning: %s primary device specified but has no " |
4947 | "effect in %s mode\n", | |
4948 | primary, bond_mode_name(bond_mode)); | |
4949 | primary = NULL; | |
4950 | } | |
4951 | ||
a549952a JP |
4952 | if (primary && primary_reselect) { |
4953 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4954 | pri_reselect_tbl); | |
4955 | if (primary_reselect_value == -1) { | |
4956 | pr_err(DRV_NAME | |
4957 | ": Error: Invalid primary_reselect \"%s\"\n", | |
4958 | primary_reselect == | |
4959 | NULL ? "NULL" : primary_reselect); | |
4960 | return -EINVAL; | |
4961 | } | |
4962 | } else { | |
4963 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4964 | } | |
4965 | ||
3915c1e8 JV |
4966 | if (fail_over_mac) { |
4967 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4968 | fail_over_mac_tbl); | |
4969 | if (fail_over_mac_value == -1) { | |
3d632c3f | 4970 | pr_err(DRV_NAME |
3915c1e8 JV |
4971 | ": Error: invalid fail_over_mac \"%s\"\n", |
4972 | arp_validate == NULL ? "NULL" : arp_validate); | |
4973 | return -EINVAL; | |
4974 | } | |
4975 | ||
4976 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
3d632c3f | 4977 | pr_warning(DRV_NAME |
3915c1e8 JV |
4978 | ": Warning: fail_over_mac only affects " |
4979 | "active-backup mode.\n"); | |
4980 | } else { | |
4981 | fail_over_mac_value = BOND_FOM_NONE; | |
4982 | } | |
dd957c57 | 4983 | |
1da177e4 LT |
4984 | /* fill params struct with the proper values */ |
4985 | params->mode = bond_mode; | |
169a3e66 | 4986 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4987 | params->miimon = miimon; |
7893b249 | 4988 | params->num_grat_arp = num_grat_arp; |
305d552a | 4989 | params->num_unsol_na = num_unsol_na; |
1da177e4 | 4990 | params->arp_interval = arp_interval; |
f5b2b966 | 4991 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4992 | params->updelay = updelay; |
4993 | params->downdelay = downdelay; | |
4994 | params->use_carrier = use_carrier; | |
4995 | params->lacp_fast = lacp_fast; | |
4996 | params->primary[0] = 0; | |
a549952a | 4997 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4998 | params->fail_over_mac = fail_over_mac_value; |
1da177e4 LT |
4999 | |
5000 | if (primary) { | |
5001 | strncpy(params->primary, primary, IFNAMSIZ); | |
5002 | params->primary[IFNAMSIZ - 1] = 0; | |
5003 | } | |
5004 | ||
5005 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
5006 | ||
5007 | return 0; | |
5008 | } | |
5009 | ||
0daa2303 | 5010 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 5011 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 5012 | |
e8a0464c DM |
5013 | static void bond_set_lockdep_class_one(struct net_device *dev, |
5014 | struct netdev_queue *txq, | |
5015 | void *_unused) | |
c773e847 DM |
5016 | { |
5017 | lockdep_set_class(&txq->_xmit_lock, | |
5018 | &bonding_netdev_xmit_lock_key); | |
5019 | } | |
5020 | ||
5021 | static void bond_set_lockdep_class(struct net_device *dev) | |
5022 | { | |
cf508b12 DM |
5023 | lockdep_set_class(&dev->addr_list_lock, |
5024 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 5025 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
5026 | } |
5027 | ||
181470fc SH |
5028 | /* |
5029 | * Called from registration process | |
5030 | */ | |
5031 | static int bond_init(struct net_device *bond_dev) | |
5032 | { | |
5033 | struct bonding *bond = netdev_priv(bond_dev); | |
5034 | ||
5035 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
5036 | ||
5037 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
5038 | if (!bond->wq) | |
5039 | return -ENOMEM; | |
5040 | ||
5041 | bond_set_lockdep_class(bond_dev); | |
5042 | ||
5043 | netif_carrier_off(bond_dev); | |
5044 | ||
5045 | bond_create_proc_entry(bond); | |
5046 | list_add_tail(&bond->bond_list, &bond_dev_list); | |
5047 | ||
6151b3d4 | 5048 | bond_prepare_sysfs_group(bond); |
181470fc SH |
5049 | return 0; |
5050 | } | |
5051 | ||
88ead977 EB |
5052 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
5053 | { | |
5054 | if (tb[IFLA_ADDRESS]) { | |
5055 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
5056 | return -EINVAL; | |
5057 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
5058 | return -EADDRNOTAVAIL; | |
5059 | } | |
5060 | return 0; | |
5061 | } | |
5062 | ||
5063 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
5064 | .kind = "bond", | |
5065 | .setup = bond_setup, | |
5066 | .validate = bond_validate, | |
5067 | }; | |
5068 | ||
dfe60397 | 5069 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5070 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5071 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5072 | * set up our sysfs entries. | |
5073 | */ | |
d2991f75 | 5074 | int bond_create(const char *name) |
dfe60397 MW |
5075 | { |
5076 | struct net_device *bond_dev; | |
5077 | int res; | |
5078 | ||
5079 | rtnl_lock(); | |
027ea041 | 5080 | |
e4b91c48 | 5081 | bond_dev = alloc_netdev(sizeof(struct bonding), name ? name : "", |
181470fc | 5082 | bond_setup); |
dfe60397 | 5083 | if (!bond_dev) { |
3d632c3f | 5084 | pr_err(DRV_NAME ": %s: eek! can't alloc netdev!\n", |
dfe60397 MW |
5085 | name); |
5086 | res = -ENOMEM; | |
30c15ba9 | 5087 | goto out; |
dfe60397 MW |
5088 | } |
5089 | ||
88ead977 EB |
5090 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5091 | ||
e4b91c48 JV |
5092 | if (!name) { |
5093 | res = dev_alloc_name(bond_dev, "bond%d"); | |
5094 | if (res < 0) | |
5095 | goto out_netdev; | |
5096 | } | |
5097 | ||
dfe60397 | 5098 | res = register_netdevice(bond_dev); |
0daa2303 | 5099 | |
30c15ba9 | 5100 | out: |
7e083840 | 5101 | rtnl_unlock(); |
30c15ba9 | 5102 | return res; |
dfe60397 MW |
5103 | out_netdev: |
5104 | free_netdev(bond_dev); | |
30c15ba9 | 5105 | goto out; |
dfe60397 MW |
5106 | } |
5107 | ||
1da177e4 LT |
5108 | static int __init bonding_init(void) |
5109 | { | |
1da177e4 LT |
5110 | int i; |
5111 | int res; | |
5112 | ||
3d632c3f | 5113 | pr_info("%s", version); |
1da177e4 | 5114 | |
dfe60397 | 5115 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5116 | if (res) |
dfe60397 | 5117 | goto out; |
1da177e4 | 5118 | |
1da177e4 | 5119 | bond_create_proc_dir(); |
027ea041 | 5120 | |
88ead977 EB |
5121 | res = rtnl_link_register(&bond_link_ops); |
5122 | if (res) | |
5123 | goto err; | |
5124 | ||
1da177e4 | 5125 | for (i = 0; i < max_bonds; i++) { |
d2991f75 | 5126 | res = bond_create(NULL); |
dfe60397 MW |
5127 | if (res) |
5128 | goto err; | |
1da177e4 LT |
5129 | } |
5130 | ||
b76cdba9 MW |
5131 | res = bond_create_sysfs(); |
5132 | if (res) | |
5133 | goto err; | |
5134 | ||
1da177e4 | 5135 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5136 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5137 | bond_register_ipv6_notifier(); |
dfe60397 | 5138 | out: |
1da177e4 | 5139 | return res; |
88ead977 EB |
5140 | err: |
5141 | rtnl_link_unregister(&bond_link_ops); | |
5142 | bond_destroy_proc_dir(); | |
5143 | goto out; | |
dfe60397 | 5144 | |
1da177e4 LT |
5145 | } |
5146 | ||
5147 | static void __exit bonding_exit(void) | |
5148 | { | |
5149 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5150 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5151 | bond_unregister_ipv6_notifier(); |
1da177e4 | 5152 | |
ae68c398 PE |
5153 | bond_destroy_sysfs(); |
5154 | ||
88ead977 EB |
5155 | rtnl_link_unregister(&bond_link_ops); |
5156 | bond_destroy_proc_dir(); | |
1da177e4 LT |
5157 | } |
5158 | ||
5159 | module_init(bonding_init); | |
5160 | module_exit(bonding_exit); | |
5161 | MODULE_LICENSE("GPL"); | |
5162 | MODULE_VERSION(DRV_VERSION); | |
5163 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5164 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5165 | MODULE_ALIAS_RTNL_LINK("bond"); |