]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * originally based on the dummy device. | |
3 | * | |
4 | * Copyright 1999, Thomas Davis, tadavis@lbl.gov. | |
5 | * Licensed under the GPL. Based on dummy.c, and eql.c devices. | |
6 | * | |
7 | * bonding.c: an Ethernet Bonding driver | |
8 | * | |
9 | * This is useful to talk to a Cisco EtherChannel compatible equipment: | |
10 | * Cisco 5500 | |
11 | * Sun Trunking (Solaris) | |
12 | * Alteon AceDirector Trunks | |
13 | * Linux Bonding | |
14 | * and probably many L2 switches ... | |
15 | * | |
16 | * How it works: | |
17 | * ifconfig bond0 ipaddress netmask up | |
18 | * will setup a network device, with an ip address. No mac address | |
19 | * will be assigned at this time. The hw mac address will come from | |
20 | * the first slave bonded to the channel. All slaves will then use | |
21 | * this hw mac address. | |
22 | * | |
23 | * ifconfig bond0 down | |
24 | * will release all slaves, marking them as down. | |
25 | * | |
26 | * ifenslave bond0 eth0 | |
27 | * will attach eth0 to bond0 as a slave. eth0 hw mac address will either | |
28 | * a: be used as initial mac address | |
29 | * b: if a hw mac address already is there, eth0's hw mac address | |
30 | * will then be set from bond0. | |
31 | * | |
1da177e4 LT |
32 | */ |
33 | ||
a4aee5c8 JP |
34 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
35 | ||
1da177e4 LT |
36 | #include <linux/kernel.h> |
37 | #include <linux/module.h> | |
1da177e4 LT |
38 | #include <linux/types.h> |
39 | #include <linux/fcntl.h> | |
40 | #include <linux/interrupt.h> | |
41 | #include <linux/ptrace.h> | |
42 | #include <linux/ioport.h> | |
43 | #include <linux/in.h> | |
169a3e66 | 44 | #include <net/ip.h> |
1da177e4 | 45 | #include <linux/ip.h> |
169a3e66 JV |
46 | #include <linux/tcp.h> |
47 | #include <linux/udp.h> | |
1da177e4 LT |
48 | #include <linux/slab.h> |
49 | #include <linux/string.h> | |
50 | #include <linux/init.h> | |
51 | #include <linux/timer.h> | |
52 | #include <linux/socket.h> | |
53 | #include <linux/ctype.h> | |
54 | #include <linux/inet.h> | |
55 | #include <linux/bitops.h> | |
3d632c3f | 56 | #include <linux/io.h> |
1da177e4 | 57 | #include <asm/system.h> |
1da177e4 | 58 | #include <asm/dma.h> |
3d632c3f | 59 | #include <linux/uaccess.h> |
1da177e4 LT |
60 | #include <linux/errno.h> |
61 | #include <linux/netdevice.h> | |
f6dc31a8 | 62 | #include <linux/netpoll.h> |
1da177e4 | 63 | #include <linux/inetdevice.h> |
a816c7c7 | 64 | #include <linux/igmp.h> |
1da177e4 LT |
65 | #include <linux/etherdevice.h> |
66 | #include <linux/skbuff.h> | |
67 | #include <net/sock.h> | |
68 | #include <linux/rtnetlink.h> | |
69 | #include <linux/proc_fs.h> | |
70 | #include <linux/seq_file.h> | |
71 | #include <linux/smp.h> | |
72 | #include <linux/if_ether.h> | |
73 | #include <net/arp.h> | |
74 | #include <linux/mii.h> | |
75 | #include <linux/ethtool.h> | |
76 | #include <linux/if_vlan.h> | |
77 | #include <linux/if_bonding.h> | |
b63bb739 | 78 | #include <linux/jiffies.h> |
c3ade5ca | 79 | #include <net/route.h> |
457c4cbc | 80 | #include <net/net_namespace.h> |
ec87fd3b | 81 | #include <net/netns/generic.h> |
1da177e4 LT |
82 | #include "bonding.h" |
83 | #include "bond_3ad.h" | |
84 | #include "bond_alb.h" | |
85 | ||
86 | /*---------------------------- Module parameters ----------------------------*/ | |
87 | ||
88 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
89 | #define BOND_LINK_MON_INTERV 0 | |
90 | #define BOND_LINK_ARP_INTERV 0 | |
91 | ||
92 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 93 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
7893b249 | 94 | static int num_grat_arp = 1; |
305d552a | 95 | static int num_unsol_na = 1; |
1da177e4 | 96 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
97 | static int updelay; |
98 | static int downdelay; | |
1da177e4 | 99 | static int use_carrier = 1; |
3d632c3f SH |
100 | static char *mode; |
101 | static char *primary; | |
a549952a | 102 | static char *primary_reselect; |
3d632c3f SH |
103 | static char *lacp_rate; |
104 | static char *ad_select; | |
105 | static char *xmit_hash_policy; | |
1da177e4 | 106 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
107 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
108 | static char *arp_validate; | |
109 | static char *fail_over_mac; | |
ebd8e497 | 110 | static int all_slaves_active = 0; |
d2991f75 | 111 | static struct bond_params bonding_defaults; |
1da177e4 LT |
112 | |
113 | module_param(max_bonds, int, 0); | |
114 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
115 | module_param(tx_queues, int, 0); |
116 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
7893b249 MS |
117 | module_param(num_grat_arp, int, 0644); |
118 | MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packets to send on failover event"); | |
305d552a BH |
119 | module_param(num_unsol_na, int, 0644); |
120 | MODULE_PARM_DESC(num_unsol_na, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event"); | |
1da177e4 LT |
121 | module_param(miimon, int, 0); |
122 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
123 | module_param(updelay, int, 0); | |
124 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
125 | module_param(downdelay, int, 0); | |
2ac47660 MW |
126 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
127 | "in milliseconds"); | |
1da177e4 | 128 | module_param(use_carrier, int, 0); |
2ac47660 MW |
129 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
130 | "0 for off, 1 for on (default)"); | |
1da177e4 | 131 | module_param(mode, charp, 0); |
2ac47660 MW |
132 | MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, " |
133 | "1 for active-backup, 2 for balance-xor, " | |
134 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
135 | "6 for balance-alb"); | |
1da177e4 LT |
136 | module_param(primary, charp, 0); |
137 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
138 | module_param(primary_reselect, charp, 0); |
139 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
140 | "once it comes up; " | |
141 | "0 for always (default), " | |
142 | "1 for only if speed of primary is " | |
143 | "better, " | |
144 | "2 for only on active slave " | |
145 | "failure"); | |
1da177e4 | 146 | module_param(lacp_rate, charp, 0); |
2ac47660 MW |
147 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner " |
148 | "(slow/fast)"); | |
fd989c83 JV |
149 | module_param(ad_select, charp, 0); |
150 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)"); | |
169a3e66 | 151 | module_param(xmit_hash_policy, charp, 0); |
2ac47660 MW |
152 | MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)" |
153 | ", 1 for layer 3+4"); | |
1da177e4 LT |
154 | module_param(arp_interval, int, 0); |
155 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
156 | module_param_array(arp_ip_target, charp, NULL, 0); | |
157 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 JV |
158 | module_param(arp_validate, charp, 0); |
159 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all"); | |
3915c1e8 JV |
160 | module_param(fail_over_mac, charp, 0); |
161 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to the same MAC. none (default), active or follow"); | |
ebd8e497 AG |
162 | module_param(all_slaves_active, int, 0); |
163 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
164 | "by setting active flag for all slaves. " | |
165 | "0 for never (default), 1 for always."); | |
1da177e4 LT |
166 | |
167 | /*----------------------------- Global variables ----------------------------*/ | |
168 | ||
f71e1309 | 169 | static const char * const version = |
1da177e4 LT |
170 | DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n"; |
171 | ||
f99189b1 | 172 | int bond_net_id __read_mostly; |
1da177e4 | 173 | |
3d632c3f SH |
174 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
175 | static int arp_ip_count; | |
1da177e4 | 176 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
177 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
178 | static int lacp_fast; | |
217df670 | 179 | |
1da177e4 | 180 | |
e97fd7c6 | 181 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
182 | { "slow", AD_LACP_SLOW}, |
183 | { "fast", AD_LACP_FAST}, | |
184 | { NULL, -1}, | |
185 | }; | |
186 | ||
e97fd7c6 | 187 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
188 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
189 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
190 | { "balance-xor", BOND_MODE_XOR}, | |
191 | { "broadcast", BOND_MODE_BROADCAST}, | |
192 | { "802.3ad", BOND_MODE_8023AD}, | |
193 | { "balance-tlb", BOND_MODE_TLB}, | |
194 | { "balance-alb", BOND_MODE_ALB}, | |
195 | { NULL, -1}, | |
196 | }; | |
197 | ||
e97fd7c6 | 198 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
199 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
200 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 201 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
202 | { NULL, -1}, |
203 | }; | |
204 | ||
e97fd7c6 | 205 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
206 | { "none", BOND_ARP_VALIDATE_NONE}, |
207 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
208 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
209 | { "all", BOND_ARP_VALIDATE_ALL}, | |
210 | { NULL, -1}, | |
211 | }; | |
212 | ||
e97fd7c6 | 213 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
214 | { "none", BOND_FOM_NONE}, |
215 | { "active", BOND_FOM_ACTIVE}, | |
216 | { "follow", BOND_FOM_FOLLOW}, | |
217 | { NULL, -1}, | |
218 | }; | |
219 | ||
a549952a JP |
220 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
221 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
222 | { "better", BOND_PRI_RESELECT_BETTER}, | |
223 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
224 | { NULL, -1}, | |
225 | }; | |
226 | ||
fd989c83 JV |
227 | struct bond_parm_tbl ad_select_tbl[] = { |
228 | { "stable", BOND_AD_STABLE}, | |
229 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
230 | { "count", BOND_AD_COUNT}, | |
231 | { NULL, -1}, | |
232 | }; | |
233 | ||
1da177e4 LT |
234 | /*-------------------------- Forward declarations ---------------------------*/ |
235 | ||
c3ade5ca | 236 | static void bond_send_gratuitous_arp(struct bonding *bond); |
181470fc | 237 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 238 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
239 | |
240 | /*---------------------------- General routines -----------------------------*/ | |
241 | ||
4ad072c9 | 242 | static const char *bond_mode_name(int mode) |
1da177e4 | 243 | { |
77afc92b HE |
244 | static const char *names[] = { |
245 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
246 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
247 | [BOND_MODE_XOR] = "load balancing (xor)", | |
248 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 249 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
250 | [BOND_MODE_TLB] = "transmit load balancing", |
251 | [BOND_MODE_ALB] = "adaptive load balancing", | |
252 | }; | |
253 | ||
254 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 255 | return "unknown"; |
77afc92b HE |
256 | |
257 | return names[mode]; | |
1da177e4 LT |
258 | } |
259 | ||
260 | /*---------------------------------- VLAN -----------------------------------*/ | |
261 | ||
262 | /** | |
263 | * bond_add_vlan - add a new vlan id on bond | |
264 | * @bond: bond that got the notification | |
265 | * @vlan_id: the vlan id to add | |
266 | * | |
267 | * Returns -ENOMEM if allocation failed. | |
268 | */ | |
269 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
270 | { | |
271 | struct vlan_entry *vlan; | |
272 | ||
5a03cdb7 | 273 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 274 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 275 | |
305d552a | 276 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 277 | if (!vlan) |
1da177e4 | 278 | return -ENOMEM; |
1da177e4 LT |
279 | |
280 | INIT_LIST_HEAD(&vlan->vlan_list); | |
281 | vlan->vlan_id = vlan_id; | |
282 | ||
283 | write_lock_bh(&bond->lock); | |
284 | ||
285 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
286 | ||
287 | write_unlock_bh(&bond->lock); | |
288 | ||
5a03cdb7 | 289 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
290 | |
291 | return 0; | |
292 | } | |
293 | ||
294 | /** | |
295 | * bond_del_vlan - delete a vlan id from bond | |
296 | * @bond: bond that got the notification | |
297 | * @vlan_id: the vlan id to delete | |
298 | * | |
299 | * returns -ENODEV if @vlan_id was not found in @bond. | |
300 | */ | |
301 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
302 | { | |
0883beca | 303 | struct vlan_entry *vlan; |
1da177e4 LT |
304 | int res = -ENODEV; |
305 | ||
5a03cdb7 | 306 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 LT |
307 | |
308 | write_lock_bh(&bond->lock); | |
309 | ||
0883beca | 310 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
311 | if (vlan->vlan_id == vlan_id) { |
312 | list_del(&vlan->vlan_list); | |
313 | ||
58402054 | 314 | if (bond_is_lb(bond)) |
1da177e4 | 315 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 316 | |
a4aee5c8 JP |
317 | pr_debug("removed VLAN ID %d from bond %s\n", |
318 | vlan_id, bond->dev->name); | |
1da177e4 LT |
319 | |
320 | kfree(vlan); | |
321 | ||
322 | if (list_empty(&bond->vlan_list) && | |
323 | (bond->slave_cnt == 0)) { | |
324 | /* Last VLAN removed and no slaves, so | |
325 | * restore block on adding VLANs. This will | |
326 | * be removed once new slaves that are not | |
327 | * VLAN challenged will be added. | |
328 | */ | |
329 | bond->dev->features |= NETIF_F_VLAN_CHALLENGED; | |
330 | } | |
331 | ||
332 | res = 0; | |
333 | goto out; | |
334 | } | |
335 | } | |
336 | ||
a4aee5c8 JP |
337 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
338 | vlan_id, bond->dev->name); | |
1da177e4 LT |
339 | |
340 | out: | |
341 | write_unlock_bh(&bond->lock); | |
342 | return res; | |
343 | } | |
344 | ||
345 | /** | |
346 | * bond_has_challenged_slaves | |
347 | * @bond: the bond we're working on | |
348 | * | |
349 | * Searches the slave list. Returns 1 if a vlan challenged slave | |
350 | * was found, 0 otherwise. | |
351 | * | |
352 | * Assumes bond->lock is held. | |
353 | */ | |
354 | static int bond_has_challenged_slaves(struct bonding *bond) | |
355 | { | |
356 | struct slave *slave; | |
357 | int i; | |
358 | ||
359 | bond_for_each_slave(bond, slave, i) { | |
360 | if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 361 | pr_debug("found VLAN challenged slave - %s\n", |
a4aee5c8 | 362 | slave->dev->name); |
1da177e4 LT |
363 | return 1; |
364 | } | |
365 | } | |
366 | ||
5a03cdb7 | 367 | pr_debug("no VLAN challenged slaves found\n"); |
1da177e4 LT |
368 | return 0; |
369 | } | |
370 | ||
371 | /** | |
372 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
373 | * @bond: the bond we're working on | |
374 | * @curr: item we're advancing from | |
375 | * | |
376 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
377 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 378 | * |
1da177e4 LT |
379 | * Caller must hold bond->lock |
380 | */ | |
381 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
382 | { | |
383 | struct vlan_entry *next, *last; | |
384 | ||
3d632c3f | 385 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 386 | return NULL; |
1da177e4 LT |
387 | |
388 | if (!curr) { | |
389 | next = list_entry(bond->vlan_list.next, | |
390 | struct vlan_entry, vlan_list); | |
391 | } else { | |
392 | last = list_entry(bond->vlan_list.prev, | |
393 | struct vlan_entry, vlan_list); | |
394 | if (last == curr) { | |
395 | next = list_entry(bond->vlan_list.next, | |
396 | struct vlan_entry, vlan_list); | |
397 | } else { | |
398 | next = list_entry(curr->vlan_list.next, | |
399 | struct vlan_entry, vlan_list); | |
400 | } | |
401 | } | |
402 | ||
403 | return next; | |
404 | } | |
405 | ||
406 | /** | |
407 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 408 | * |
1da177e4 LT |
409 | * @bond: bond device that got this skb for tx. |
410 | * @skb: hw accel VLAN tagged skb to transmit | |
411 | * @slave_dev: slave that is supposed to xmit this skbuff | |
3d632c3f | 412 | * |
1da177e4 LT |
413 | * When the bond gets an skb to transmit that is |
414 | * already hardware accelerated VLAN tagged, and it | |
415 | * needs to relay this skb to a slave that is not | |
416 | * hw accel capable, the skb needs to be "unaccelerated", | |
417 | * i.e. strip the hwaccel tag and re-insert it as part | |
418 | * of the payload. | |
419 | */ | |
3d632c3f SH |
420 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
421 | struct net_device *slave_dev) | |
1da177e4 | 422 | { |
966bc6f4 | 423 | unsigned short uninitialized_var(vlan_id); |
1da177e4 LT |
424 | |
425 | if (!list_empty(&bond->vlan_list) && | |
426 | !(slave_dev->features & NETIF_F_HW_VLAN_TX) && | |
427 | vlan_get_tag(skb, &vlan_id) == 0) { | |
428 | skb->dev = slave_dev; | |
429 | skb = vlan_put_tag(skb, vlan_id); | |
430 | if (!skb) { | |
431 | /* vlan_put_tag() frees the skb in case of error, | |
432 | * so return success here so the calling functions | |
433 | * won't attempt to free is again. | |
434 | */ | |
435 | return 0; | |
436 | } | |
437 | } else { | |
438 | skb->dev = slave_dev; | |
439 | } | |
440 | ||
441 | skb->priority = 1; | |
f6dc31a8 WC |
442 | #ifdef CONFIG_NET_POLL_CONTROLLER |
443 | if (unlikely(bond->dev->priv_flags & IFF_IN_NETPOLL)) { | |
444 | struct netpoll *np = bond->dev->npinfo->netpoll; | |
445 | slave_dev->npinfo = bond->dev->npinfo; | |
446 | np->real_dev = np->dev = skb->dev; | |
447 | slave_dev->priv_flags |= IFF_IN_NETPOLL; | |
448 | netpoll_send_skb(np, skb); | |
449 | slave_dev->priv_flags &= ~IFF_IN_NETPOLL; | |
450 | np->dev = bond->dev; | |
451 | } else | |
452 | #endif | |
453 | dev_queue_xmit(skb); | |
1da177e4 LT |
454 | |
455 | return 0; | |
456 | } | |
457 | ||
458 | /* | |
459 | * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid | |
460 | * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a | |
461 | * lock because: | |
462 | * a. This operation is performed in IOCTL context, | |
463 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
464 | * c. Holding a lock with BH disabled while directly calling a base driver | |
465 | * entry point is generally a BAD idea. | |
3d632c3f | 466 | * |
1da177e4 LT |
467 | * The design of synchronization/protection for this operation in the 8021q |
468 | * module is good for one or more VLAN devices over a single physical device | |
469 | * and cannot be extended for a teaming solution like bonding, so there is a | |
470 | * potential race condition here where a net device from the vlan group might | |
471 | * be referenced (either by a base driver or the 8021q code) while it is being | |
472 | * removed from the system. However, it turns out we're not making matters | |
473 | * worse, and if it works for regular VLAN usage it will work here too. | |
474 | */ | |
475 | ||
476 | /** | |
477 | * bond_vlan_rx_register - Propagates registration to slaves | |
478 | * @bond_dev: bonding net device that got called | |
479 | * @grp: vlan group being registered | |
480 | */ | |
3d632c3f SH |
481 | static void bond_vlan_rx_register(struct net_device *bond_dev, |
482 | struct vlan_group *grp) | |
1da177e4 | 483 | { |
454d7c9b | 484 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
485 | struct slave *slave; |
486 | int i; | |
487 | ||
488 | bond->vlgrp = grp; | |
489 | ||
490 | bond_for_each_slave(bond, slave, i) { | |
491 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 492 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
493 | |
494 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
495 | slave_ops->ndo_vlan_rx_register) { |
496 | slave_ops->ndo_vlan_rx_register(slave_dev, grp); | |
1da177e4 LT |
497 | } |
498 | } | |
499 | } | |
500 | ||
501 | /** | |
502 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
503 | * @bond_dev: bonding net device that got called | |
504 | * @vid: vlan id being added | |
505 | */ | |
506 | static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) | |
507 | { | |
454d7c9b | 508 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
509 | struct slave *slave; |
510 | int i, res; | |
511 | ||
512 | bond_for_each_slave(bond, slave, i) { | |
513 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 514 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
515 | |
516 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a SH |
517 | slave_ops->ndo_vlan_rx_add_vid) { |
518 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid); | |
1da177e4 LT |
519 | } |
520 | } | |
521 | ||
522 | res = bond_add_vlan(bond, vid); | |
523 | if (res) { | |
a4aee5c8 | 524 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 LT |
525 | bond_dev->name, vid); |
526 | } | |
527 | } | |
528 | ||
529 | /** | |
530 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
531 | * @bond_dev: bonding net device that got called | |
532 | * @vid: vlan id being removed | |
533 | */ | |
534 | static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) | |
535 | { | |
454d7c9b | 536 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
537 | struct slave *slave; |
538 | struct net_device *vlan_dev; | |
539 | int i, res; | |
540 | ||
541 | bond_for_each_slave(bond, slave, i) { | |
542 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 543 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
544 | |
545 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a | 546 | slave_ops->ndo_vlan_rx_kill_vid) { |
1da177e4 LT |
547 | /* Save and then restore vlan_dev in the grp array, |
548 | * since the slave's driver might clear it. | |
549 | */ | |
5c15bdec | 550 | vlan_dev = vlan_group_get_device(bond->vlgrp, vid); |
eb7cc59a | 551 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid); |
5c15bdec | 552 | vlan_group_set_device(bond->vlgrp, vid, vlan_dev); |
1da177e4 LT |
553 | } |
554 | } | |
555 | ||
556 | res = bond_del_vlan(bond, vid); | |
557 | if (res) { | |
a4aee5c8 | 558 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 LT |
559 | bond_dev->name, vid); |
560 | } | |
561 | } | |
562 | ||
563 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
564 | { | |
565 | struct vlan_entry *vlan; | |
eb7cc59a | 566 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
567 | |
568 | write_lock_bh(&bond->lock); | |
569 | ||
eb7cc59a | 570 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 571 | goto out; |
1da177e4 LT |
572 | |
573 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
574 | slave_ops->ndo_vlan_rx_register) |
575 | slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp); | |
1da177e4 LT |
576 | |
577 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 578 | !(slave_ops->ndo_vlan_rx_add_vid)) |
1da177e4 | 579 | goto out; |
1da177e4 | 580 | |
eb7cc59a SH |
581 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) |
582 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id); | |
1da177e4 LT |
583 | |
584 | out: | |
585 | write_unlock_bh(&bond->lock); | |
586 | } | |
587 | ||
3d632c3f SH |
588 | static void bond_del_vlans_from_slave(struct bonding *bond, |
589 | struct net_device *slave_dev) | |
1da177e4 | 590 | { |
eb7cc59a | 591 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
592 | struct vlan_entry *vlan; |
593 | struct net_device *vlan_dev; | |
594 | ||
595 | write_lock_bh(&bond->lock); | |
596 | ||
eb7cc59a | 597 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 598 | goto out; |
1da177e4 LT |
599 | |
600 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 601 | !(slave_ops->ndo_vlan_rx_kill_vid)) |
1da177e4 | 602 | goto unreg; |
1da177e4 LT |
603 | |
604 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
605 | /* Save and then restore vlan_dev in the grp array, | |
606 | * since the slave's driver might clear it. | |
607 | */ | |
5c15bdec | 608 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
eb7cc59a | 609 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id); |
5c15bdec | 610 | vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev); |
1da177e4 LT |
611 | } |
612 | ||
613 | unreg: | |
614 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
615 | slave_ops->ndo_vlan_rx_register) |
616 | slave_ops->ndo_vlan_rx_register(slave_dev, NULL); | |
1da177e4 LT |
617 | |
618 | out: | |
619 | write_unlock_bh(&bond->lock); | |
620 | } | |
621 | ||
622 | /*------------------------------- Link status -------------------------------*/ | |
623 | ||
ff59c456 JV |
624 | /* |
625 | * Set the carrier state for the master according to the state of its | |
626 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
627 | * do special 802.3ad magic. | |
628 | * | |
629 | * Returns zero if carrier state does not change, nonzero if it does. | |
630 | */ | |
631 | static int bond_set_carrier(struct bonding *bond) | |
632 | { | |
633 | struct slave *slave; | |
634 | int i; | |
635 | ||
636 | if (bond->slave_cnt == 0) | |
637 | goto down; | |
638 | ||
639 | if (bond->params.mode == BOND_MODE_8023AD) | |
640 | return bond_3ad_set_carrier(bond); | |
641 | ||
642 | bond_for_each_slave(bond, slave, i) { | |
643 | if (slave->link == BOND_LINK_UP) { | |
644 | if (!netif_carrier_ok(bond->dev)) { | |
645 | netif_carrier_on(bond->dev); | |
646 | return 1; | |
647 | } | |
648 | return 0; | |
649 | } | |
650 | } | |
651 | ||
652 | down: | |
653 | if (netif_carrier_ok(bond->dev)) { | |
654 | netif_carrier_off(bond->dev); | |
655 | return 1; | |
656 | } | |
657 | return 0; | |
658 | } | |
659 | ||
1da177e4 LT |
660 | /* |
661 | * Get link speed and duplex from the slave's base driver | |
662 | * using ethtool. If for some reason the call fails or the | |
663 | * values are invalid, fake speed and duplex to 100/Full | |
664 | * and return error. | |
665 | */ | |
666 | static int bond_update_speed_duplex(struct slave *slave) | |
667 | { | |
668 | struct net_device *slave_dev = slave->dev; | |
1da177e4 | 669 | struct ethtool_cmd etool; |
61a44b9c | 670 | int res; |
1da177e4 LT |
671 | |
672 | /* Fake speed and duplex */ | |
673 | slave->speed = SPEED_100; | |
674 | slave->duplex = DUPLEX_FULL; | |
675 | ||
61a44b9c MW |
676 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
677 | return -1; | |
1da177e4 | 678 | |
61a44b9c MW |
679 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
680 | if (res < 0) | |
1da177e4 | 681 | return -1; |
1da177e4 | 682 | |
1da177e4 LT |
683 | switch (etool.speed) { |
684 | case SPEED_10: | |
685 | case SPEED_100: | |
686 | case SPEED_1000: | |
94dbffd5 | 687 | case SPEED_10000: |
1da177e4 LT |
688 | break; |
689 | default: | |
690 | return -1; | |
691 | } | |
692 | ||
693 | switch (etool.duplex) { | |
694 | case DUPLEX_FULL: | |
695 | case DUPLEX_HALF: | |
696 | break; | |
697 | default: | |
698 | return -1; | |
699 | } | |
700 | ||
701 | slave->speed = etool.speed; | |
702 | slave->duplex = etool.duplex; | |
703 | ||
704 | return 0; | |
705 | } | |
706 | ||
707 | /* | |
708 | * if <dev> supports MII link status reporting, check its link status. | |
709 | * | |
710 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 711 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
712 | * |
713 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
714 | * can't tell and just pretend it is), or 0, meaning that the link is | |
715 | * down. | |
716 | * | |
717 | * If reporting is non-zero, instead of faking link up, return -1 if | |
718 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
719 | * support them). If use_carrier is set, return whatever it says. | |
720 | * It'd be nice if there was a good way to tell if a driver supports | |
721 | * netif_carrier, but there really isn't. | |
722 | */ | |
3d632c3f SH |
723 | static int bond_check_dev_link(struct bonding *bond, |
724 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 725 | { |
eb7cc59a | 726 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 727 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
728 | struct ifreq ifr; |
729 | struct mii_ioctl_data *mii; | |
1da177e4 | 730 | |
6c988853 PG |
731 | if (!reporting && !netif_running(slave_dev)) |
732 | return 0; | |
733 | ||
eb7cc59a | 734 | if (bond->params.use_carrier) |
1da177e4 | 735 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 736 | |
29112f4e JP |
737 | /* Try to get link status using Ethtool first. */ |
738 | if (slave_dev->ethtool_ops) { | |
739 | if (slave_dev->ethtool_ops->get_link) { | |
740 | u32 link; | |
741 | ||
742 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
743 | ||
744 | return link ? BMSR_LSTATUS : 0; | |
745 | } | |
746 | } | |
747 | ||
3d632c3f | 748 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 749 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
750 | if (ioctl) { |
751 | /* TODO: set pointer to correct ioctl on a per team member */ | |
752 | /* bases to make this more efficient. that is, once */ | |
753 | /* we determine the correct ioctl, we will always */ | |
754 | /* call it and not the others for that team */ | |
755 | /* member. */ | |
756 | ||
757 | /* | |
758 | * We cannot assume that SIOCGMIIPHY will also read a | |
759 | * register; not all network drivers (e.g., e100) | |
760 | * support that. | |
761 | */ | |
762 | ||
763 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
764 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
765 | mii = if_mii(&ifr); | |
766 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
767 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
768 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
769 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
770 | } |
771 | } | |
772 | ||
1da177e4 LT |
773 | /* |
774 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 775 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
776 | * cannot report link status). If not reporting, pretend |
777 | * we're ok. | |
778 | */ | |
3d632c3f | 779 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
780 | } |
781 | ||
782 | /*----------------------------- Multicast list ------------------------------*/ | |
783 | ||
1da177e4 LT |
784 | /* |
785 | * Push the promiscuity flag down to appropriate slaves | |
786 | */ | |
7e1a1ac1 | 787 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 788 | { |
7e1a1ac1 | 789 | int err = 0; |
1da177e4 LT |
790 | if (USES_PRIMARY(bond->params.mode)) { |
791 | /* write lock already acquired */ | |
792 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
793 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
794 | inc); | |
1da177e4 LT |
795 | } |
796 | } else { | |
797 | struct slave *slave; | |
798 | int i; | |
799 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
800 | err = dev_set_promiscuity(slave->dev, inc); |
801 | if (err) | |
802 | return err; | |
1da177e4 LT |
803 | } |
804 | } | |
7e1a1ac1 | 805 | return err; |
1da177e4 LT |
806 | } |
807 | ||
808 | /* | |
809 | * Push the allmulti flag down to all slaves | |
810 | */ | |
7e1a1ac1 | 811 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 812 | { |
7e1a1ac1 | 813 | int err = 0; |
1da177e4 LT |
814 | if (USES_PRIMARY(bond->params.mode)) { |
815 | /* write lock already acquired */ | |
816 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
817 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
818 | inc); | |
1da177e4 LT |
819 | } |
820 | } else { | |
821 | struct slave *slave; | |
822 | int i; | |
823 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
824 | err = dev_set_allmulti(slave->dev, inc); |
825 | if (err) | |
826 | return err; | |
1da177e4 LT |
827 | } |
828 | } | |
7e1a1ac1 | 829 | return err; |
1da177e4 LT |
830 | } |
831 | ||
832 | /* | |
833 | * Add a Multicast address to slaves | |
834 | * according to mode | |
835 | */ | |
22bedad3 | 836 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
837 | { |
838 | if (USES_PRIMARY(bond->params.mode)) { | |
839 | /* write lock already acquired */ | |
3d632c3f | 840 | if (bond->curr_active_slave) |
22bedad3 | 841 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
842 | } else { |
843 | struct slave *slave; | |
844 | int i; | |
3d632c3f SH |
845 | |
846 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 847 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
848 | } |
849 | } | |
850 | ||
851 | /* | |
852 | * Remove a multicast address from slave | |
853 | * according to mode | |
854 | */ | |
22bedad3 | 855 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
856 | { |
857 | if (USES_PRIMARY(bond->params.mode)) { | |
858 | /* write lock already acquired */ | |
3d632c3f | 859 | if (bond->curr_active_slave) |
22bedad3 | 860 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
861 | } else { |
862 | struct slave *slave; | |
863 | int i; | |
864 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 865 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
866 | } |
867 | } | |
868 | } | |
869 | ||
a816c7c7 JV |
870 | |
871 | /* | |
872 | * Retrieve the list of registered multicast addresses for the bonding | |
873 | * device and retransmit an IGMP JOIN request to the current active | |
874 | * slave. | |
875 | */ | |
876 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
877 | { | |
878 | struct in_device *in_dev; | |
879 | struct ip_mc_list *im; | |
880 | ||
881 | rcu_read_lock(); | |
882 | in_dev = __in_dev_get_rcu(bond->dev); | |
883 | if (in_dev) { | |
3d632c3f | 884 | for (im = in_dev->mc_list; im; im = im->next) |
a816c7c7 | 885 | ip_mc_rejoin_group(im); |
a816c7c7 JV |
886 | } |
887 | ||
888 | rcu_read_unlock(); | |
889 | } | |
890 | ||
1da177e4 LT |
891 | /* |
892 | * flush all members of flush->mc_list from device dev->mc_list | |
893 | */ | |
3d632c3f SH |
894 | static void bond_mc_list_flush(struct net_device *bond_dev, |
895 | struct net_device *slave_dev) | |
1da177e4 | 896 | { |
454d7c9b | 897 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 898 | struct netdev_hw_addr *ha; |
1da177e4 | 899 | |
22bedad3 JP |
900 | netdev_for_each_mc_addr(ha, bond_dev) |
901 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
902 | |
903 | if (bond->params.mode == BOND_MODE_8023AD) { | |
904 | /* del lacpdu mc addr from mc list */ | |
905 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
906 | ||
22bedad3 | 907 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
908 | } |
909 | } | |
910 | ||
911 | /*--------------------------- Active slave change ---------------------------*/ | |
912 | ||
913 | /* | |
914 | * Update the mc list and multicast-related flags for the new and | |
915 | * old active slaves (if any) according to the multicast mode, and | |
916 | * promiscuous flags unconditionally. | |
917 | */ | |
3d632c3f SH |
918 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
919 | struct slave *old_active) | |
1da177e4 | 920 | { |
22bedad3 | 921 | struct netdev_hw_addr *ha; |
1da177e4 | 922 | |
3d632c3f | 923 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
924 | /* nothing to do - mc list is already up-to-date on |
925 | * all slaves | |
926 | */ | |
927 | return; | |
1da177e4 LT |
928 | |
929 | if (old_active) { | |
3d632c3f | 930 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 931 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 932 | |
3d632c3f | 933 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 934 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 935 | |
22bedad3 JP |
936 | netdev_for_each_mc_addr(ha, bond->dev) |
937 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
938 | } |
939 | ||
940 | if (new_active) { | |
7e1a1ac1 | 941 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 942 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 943 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 944 | |
3d632c3f | 945 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 946 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 947 | |
22bedad3 JP |
948 | netdev_for_each_mc_addr(ha, bond->dev) |
949 | dev_mc_add(new_active->dev, ha->addr); | |
a816c7c7 | 950 | bond_resend_igmp_join_requests(bond); |
1da177e4 LT |
951 | } |
952 | } | |
953 | ||
3915c1e8 JV |
954 | /* |
955 | * bond_do_fail_over_mac | |
956 | * | |
957 | * Perform special MAC address swapping for fail_over_mac settings | |
958 | * | |
959 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
960 | */ | |
961 | static void bond_do_fail_over_mac(struct bonding *bond, | |
962 | struct slave *new_active, | |
963 | struct slave *old_active) | |
1f78d9f9 HE |
964 | __releases(&bond->curr_slave_lock) |
965 | __releases(&bond->lock) | |
966 | __acquires(&bond->lock) | |
967 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
968 | { |
969 | u8 tmp_mac[ETH_ALEN]; | |
970 | struct sockaddr saddr; | |
971 | int rv; | |
972 | ||
973 | switch (bond->params.fail_over_mac) { | |
974 | case BOND_FOM_ACTIVE: | |
975 | if (new_active) | |
976 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
977 | new_active->dev->addr_len); | |
978 | break; | |
979 | case BOND_FOM_FOLLOW: | |
980 | /* | |
981 | * if new_active && old_active, swap them | |
982 | * if just old_active, do nothing (going to no active slave) | |
983 | * if just new_active, set new_active to bond's MAC | |
984 | */ | |
985 | if (!new_active) | |
986 | return; | |
987 | ||
988 | write_unlock_bh(&bond->curr_slave_lock); | |
989 | read_unlock(&bond->lock); | |
990 | ||
991 | if (old_active) { | |
992 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
993 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
994 | ETH_ALEN); | |
995 | saddr.sa_family = new_active->dev->type; | |
996 | } else { | |
997 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
998 | saddr.sa_family = bond->dev->type; | |
999 | } | |
1000 | ||
1001 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
1002 | if (rv) { | |
a4aee5c8 | 1003 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1004 | bond->dev->name, -rv, new_active->dev->name); |
1005 | goto out; | |
1006 | } | |
1007 | ||
1008 | if (!old_active) | |
1009 | goto out; | |
1010 | ||
1011 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1012 | saddr.sa_family = old_active->dev->type; | |
1013 | ||
1014 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1015 | if (rv) | |
a4aee5c8 | 1016 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1017 | bond->dev->name, -rv, new_active->dev->name); |
1018 | out: | |
1019 | read_lock(&bond->lock); | |
1020 | write_lock_bh(&bond->curr_slave_lock); | |
1021 | break; | |
1022 | default: | |
a4aee5c8 | 1023 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
1024 | bond->dev->name, bond->params.fail_over_mac); |
1025 | break; | |
1026 | } | |
1027 | ||
1028 | } | |
1029 | ||
a549952a JP |
1030 | static bool bond_should_change_active(struct bonding *bond) |
1031 | { | |
1032 | struct slave *prim = bond->primary_slave; | |
1033 | struct slave *curr = bond->curr_active_slave; | |
1034 | ||
1035 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1036 | return true; | |
1037 | if (bond->force_primary) { | |
1038 | bond->force_primary = false; | |
1039 | return true; | |
1040 | } | |
1041 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1042 | (prim->speed < curr->speed || | |
1043 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1044 | return false; | |
1045 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1046 | return false; | |
1047 | return true; | |
1048 | } | |
3915c1e8 | 1049 | |
1da177e4 LT |
1050 | /** |
1051 | * find_best_interface - select the best available slave to be the active one | |
1052 | * @bond: our bonding struct | |
1053 | * | |
1054 | * Warning: Caller must hold curr_slave_lock for writing. | |
1055 | */ | |
1056 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1057 | { | |
1058 | struct slave *new_active, *old_active; | |
1059 | struct slave *bestslave = NULL; | |
1060 | int mintime = bond->params.updelay; | |
1061 | int i; | |
1062 | ||
49b4ad92 | 1063 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1064 | |
1065 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1066 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1067 | new_active = bond->first_slave; |
3d632c3f | 1068 | else |
1da177e4 | 1069 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1070 | } |
1071 | ||
1da177e4 | 1072 | if ((bond->primary_slave) && |
a549952a JP |
1073 | bond->primary_slave->link == BOND_LINK_UP && |
1074 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1075 | new_active = bond->primary_slave; |
1076 | } | |
1077 | ||
1078 | /* remember where to stop iterating over the slaves */ | |
1079 | old_active = new_active; | |
1080 | ||
1081 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1082 | if (new_active->link == BOND_LINK_UP) { |
1083 | return new_active; | |
1084 | } else if (new_active->link == BOND_LINK_BACK && | |
1085 | IS_UP(new_active->dev)) { | |
1086 | /* link up, but waiting for stabilization */ | |
1087 | if (new_active->delay < mintime) { | |
1088 | mintime = new_active->delay; | |
1089 | bestslave = new_active; | |
1da177e4 LT |
1090 | } |
1091 | } | |
1092 | } | |
1093 | ||
1094 | return bestslave; | |
1095 | } | |
1096 | ||
1097 | /** | |
1098 | * change_active_interface - change the active slave into the specified one | |
1099 | * @bond: our bonding struct | |
1100 | * @new: the new slave to make the active one | |
1101 | * | |
1102 | * Set the new slave to the bond's settings and unset them on the old | |
1103 | * curr_active_slave. | |
1104 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1105 | * | |
1106 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1107 | * because it is apparently the best available slave we have, even though its | |
1108 | * updelay hasn't timed out yet. | |
1109 | * | |
3915c1e8 JV |
1110 | * If new_active is not NULL, caller must hold bond->lock for read and |
1111 | * curr_slave_lock for write_bh. | |
1da177e4 | 1112 | */ |
a77b5325 | 1113 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1114 | { |
1115 | struct slave *old_active = bond->curr_active_slave; | |
1116 | ||
3d632c3f | 1117 | if (old_active == new_active) |
1da177e4 | 1118 | return; |
1da177e4 LT |
1119 | |
1120 | if (new_active) { | |
b2220cad JV |
1121 | new_active->jiffies = jiffies; |
1122 | ||
1da177e4 LT |
1123 | if (new_active->link == BOND_LINK_BACK) { |
1124 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1125 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1126 | bond->dev->name, new_active->dev->name, | |
1127 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1128 | } |
1129 | ||
1130 | new_active->delay = 0; | |
1131 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1132 | |
3d632c3f | 1133 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1134 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1135 | |
58402054 | 1136 | if (bond_is_lb(bond)) |
1da177e4 | 1137 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1138 | } else { |
1139 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1140 | pr_info("%s: making interface %s the new active one.\n", |
1141 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1142 | } |
1143 | } | |
1144 | } | |
1145 | ||
3d632c3f | 1146 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1147 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1148 | |
58402054 | 1149 | if (bond_is_lb(bond)) { |
1da177e4 | 1150 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1151 | if (old_active) |
1152 | bond_set_slave_inactive_flags(old_active); | |
1153 | if (new_active) | |
1154 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1155 | } else { |
1156 | bond->curr_active_slave = new_active; | |
1157 | } | |
c3ade5ca JV |
1158 | |
1159 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1160 | if (old_active) |
c3ade5ca | 1161 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1162 | |
1163 | if (new_active) { | |
1164 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1165 | |
709f8a45 OG |
1166 | if (bond->params.fail_over_mac) |
1167 | bond_do_fail_over_mac(bond, new_active, | |
1168 | old_active); | |
3915c1e8 | 1169 | |
709f8a45 | 1170 | bond->send_grat_arp = bond->params.num_grat_arp; |
b59f9f74 | 1171 | bond_send_gratuitous_arp(bond); |
01f3109d | 1172 | |
305d552a BH |
1173 | bond->send_unsol_na = bond->params.num_unsol_na; |
1174 | bond_send_unsolicited_na(bond); | |
1175 | ||
01f3109d OG |
1176 | write_unlock_bh(&bond->curr_slave_lock); |
1177 | read_unlock(&bond->lock); | |
1178 | ||
75c78500 | 1179 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1180 | |
1181 | read_lock(&bond->lock); | |
1182 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1183 | } |
c3ade5ca | 1184 | } |
a2fd940f AG |
1185 | |
1186 | /* resend IGMP joins since all were sent on curr_active_slave */ | |
1187 | if (bond->params.mode == BOND_MODE_ROUNDROBIN) { | |
1188 | bond_resend_igmp_join_requests(bond); | |
1189 | } | |
1da177e4 LT |
1190 | } |
1191 | ||
1192 | /** | |
1193 | * bond_select_active_slave - select a new active slave, if needed | |
1194 | * @bond: our bonding struct | |
1195 | * | |
3d632c3f | 1196 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1197 | * - The old curr_active_slave has been released or lost its link. |
1198 | * - The primary_slave has got its link back. | |
1199 | * - A slave has got its link back and there's no old curr_active_slave. | |
1200 | * | |
3915c1e8 | 1201 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1202 | */ |
a77b5325 | 1203 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1204 | { |
1205 | struct slave *best_slave; | |
ff59c456 | 1206 | int rv; |
1da177e4 LT |
1207 | |
1208 | best_slave = bond_find_best_slave(bond); | |
1209 | if (best_slave != bond->curr_active_slave) { | |
1210 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1211 | rv = bond_set_carrier(bond); |
1212 | if (!rv) | |
1213 | return; | |
1214 | ||
1215 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1216 | pr_info("%s: first active interface up!\n", |
1217 | bond->dev->name); | |
ff59c456 | 1218 | } else { |
a4aee5c8 JP |
1219 | pr_info("%s: now running without any active interface !\n", |
1220 | bond->dev->name); | |
ff59c456 | 1221 | } |
1da177e4 LT |
1222 | } |
1223 | } | |
1224 | ||
1225 | /*--------------------------- slave list handling ---------------------------*/ | |
1226 | ||
1227 | /* | |
1228 | * This function attaches the slave to the end of list. | |
1229 | * | |
1230 | * bond->lock held for writing by caller. | |
1231 | */ | |
1232 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1233 | { | |
1234 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1235 | new_slave->next = new_slave; | |
1236 | new_slave->prev = new_slave; | |
1237 | bond->first_slave = new_slave; | |
1238 | } else { | |
1239 | new_slave->next = bond->first_slave; | |
1240 | new_slave->prev = bond->first_slave->prev; | |
1241 | new_slave->next->prev = new_slave; | |
1242 | new_slave->prev->next = new_slave; | |
1243 | } | |
1244 | ||
1245 | bond->slave_cnt++; | |
1246 | } | |
1247 | ||
1248 | /* | |
1249 | * This function detaches the slave from the list. | |
1250 | * WARNING: no check is made to verify if the slave effectively | |
1251 | * belongs to <bond>. | |
1252 | * Nothing is freed on return, structures are just unchained. | |
1253 | * If any slave pointer in bond was pointing to <slave>, | |
1254 | * it should be changed by the calling function. | |
1255 | * | |
1256 | * bond->lock held for writing by caller. | |
1257 | */ | |
1258 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1259 | { | |
3d632c3f | 1260 | if (slave->next) |
1da177e4 | 1261 | slave->next->prev = slave->prev; |
1da177e4 | 1262 | |
3d632c3f | 1263 | if (slave->prev) |
1da177e4 | 1264 | slave->prev->next = slave->next; |
1da177e4 LT |
1265 | |
1266 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1267 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1268 | bond->first_slave = slave->next; | |
1269 | } else { | |
1270 | bond->first_slave = NULL; /* slave was the last one */ | |
1271 | } | |
1272 | } | |
1273 | ||
1274 | slave->next = NULL; | |
1275 | slave->prev = NULL; | |
1276 | bond->slave_cnt--; | |
1277 | } | |
1278 | ||
f6dc31a8 WC |
1279 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1280 | /* | |
1281 | * You must hold read lock on bond->lock before calling this. | |
1282 | */ | |
1283 | static bool slaves_support_netpoll(struct net_device *bond_dev) | |
1284 | { | |
1285 | struct bonding *bond = netdev_priv(bond_dev); | |
1286 | struct slave *slave; | |
1287 | int i = 0; | |
1288 | bool ret = true; | |
1289 | ||
1290 | bond_for_each_slave(bond, slave, i) { | |
1291 | if ((slave->dev->priv_flags & IFF_DISABLE_NETPOLL) || | |
1292 | !slave->dev->netdev_ops->ndo_poll_controller) | |
1293 | ret = false; | |
1294 | } | |
1295 | return i != 0 && ret; | |
1296 | } | |
1297 | ||
1298 | static void bond_poll_controller(struct net_device *bond_dev) | |
1299 | { | |
1300 | struct net_device *dev = bond_dev->npinfo->netpoll->real_dev; | |
1301 | if (dev != bond_dev) | |
1302 | netpoll_poll_dev(dev); | |
1303 | } | |
1304 | ||
1305 | static void bond_netpoll_cleanup(struct net_device *bond_dev) | |
1306 | { | |
1307 | struct bonding *bond = netdev_priv(bond_dev); | |
1308 | struct slave *slave; | |
1309 | const struct net_device_ops *ops; | |
1310 | int i; | |
1311 | ||
1312 | read_lock(&bond->lock); | |
1313 | bond_dev->npinfo = NULL; | |
1314 | bond_for_each_slave(bond, slave, i) { | |
1315 | if (slave->dev) { | |
1316 | ops = slave->dev->netdev_ops; | |
1317 | if (ops->ndo_netpoll_cleanup) | |
1318 | ops->ndo_netpoll_cleanup(slave->dev); | |
1319 | else | |
1320 | slave->dev->npinfo = NULL; | |
1321 | } | |
1322 | } | |
1323 | read_unlock(&bond->lock); | |
1324 | } | |
1325 | ||
1326 | #else | |
1327 | ||
1328 | static void bond_netpoll_cleanup(struct net_device *bond_dev) | |
1329 | { | |
1330 | } | |
1331 | ||
1332 | #endif | |
1333 | ||
1da177e4 LT |
1334 | /*---------------------------------- IOCTL ----------------------------------*/ |
1335 | ||
4ad072c9 AB |
1336 | static int bond_sethwaddr(struct net_device *bond_dev, |
1337 | struct net_device *slave_dev) | |
1da177e4 | 1338 | { |
5a03cdb7 HE |
1339 | pr_debug("bond_dev=%p\n", bond_dev); |
1340 | pr_debug("slave_dev=%p\n", slave_dev); | |
1341 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1342 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1343 | return 0; | |
1344 | } | |
1345 | ||
7f353bf2 HX |
1346 | #define BOND_VLAN_FEATURES \ |
1347 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1348 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1349 | |
3d632c3f | 1350 | /* |
8e3babcd | 1351 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1352 | * feature bits are managed elsewhere, so preserve those feature bits |
1353 | * on the master device. | |
8531c5ff AK |
1354 | */ |
1355 | static int bond_compute_features(struct bonding *bond) | |
1356 | { | |
8531c5ff AK |
1357 | struct slave *slave; |
1358 | struct net_device *bond_dev = bond->dev; | |
7f353bf2 | 1359 | unsigned long features = bond_dev->features; |
278339a4 | 1360 | unsigned long vlan_features = 0; |
3158bf7d MS |
1361 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1362 | bond_dev->hard_header_len); | |
8e3babcd | 1363 | int i; |
8531c5ff | 1364 | |
7f353bf2 | 1365 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
b63365a2 HX |
1366 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM; |
1367 | ||
1368 | if (!bond->first_slave) | |
1369 | goto done; | |
1370 | ||
1371 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1372 | |
278339a4 | 1373 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1374 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1375 | features = netdev_increment_features(features, |
1376 | slave->dev->features, | |
1377 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1378 | vlan_features = netdev_increment_features(vlan_features, |
1379 | slave->dev->vlan_features, | |
1380 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1381 | if (slave->dev->hard_header_len > max_hard_header_len) |
1382 | max_hard_header_len = slave->dev->hard_header_len; | |
1383 | } | |
8531c5ff | 1384 | |
b63365a2 | 1385 | done: |
7f353bf2 | 1386 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
b63365a2 | 1387 | bond_dev->features = netdev_fix_features(features, NULL); |
278339a4 | 1388 | bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL); |
54ef3137 | 1389 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1390 | |
1391 | return 0; | |
1392 | } | |
1393 | ||
872254dd MS |
1394 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1395 | struct net_device *slave_dev) | |
1396 | { | |
454d7c9b | 1397 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1398 | |
00829823 | 1399 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1400 | |
1401 | bond_dev->type = slave_dev->type; | |
1402 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1403 | bond_dev->addr_len = slave_dev->addr_len; | |
1404 | ||
1405 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1406 | slave_dev->addr_len); | |
d90a162a | 1407 | bond->setup_by_slave = 1; |
872254dd MS |
1408 | } |
1409 | ||
1da177e4 | 1410 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1411 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1412 | { |
454d7c9b | 1413 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1414 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1415 | struct slave *new_slave = NULL; |
22bedad3 | 1416 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1417 | struct sockaddr addr; |
1418 | int link_reporting; | |
1419 | int old_features = bond_dev->features; | |
1420 | int res = 0; | |
1421 | ||
552709d5 | 1422 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1423 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1424 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1425 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1426 | } |
1427 | ||
1428 | /* bond must be initialized by bond_open() before enslaving */ | |
1429 | if (!(bond_dev->flags & IFF_UP)) { | |
a4aee5c8 JP |
1430 | pr_warning("%s: master_dev is not up in bond_enslave\n", |
1431 | bond_dev->name); | |
1da177e4 LT |
1432 | } |
1433 | ||
1434 | /* already enslaved */ | |
1435 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1436 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1437 | return -EBUSY; |
1438 | } | |
1439 | ||
1440 | /* vlan challenged mutual exclusion */ | |
1441 | /* no need to lock since we're protected by rtnl_lock */ | |
1442 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1443 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 | 1444 | if (!list_empty(&bond->vlan_list)) { |
a4aee5c8 JP |
1445 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1446 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1447 | return -EPERM; |
1448 | } else { | |
a4aee5c8 JP |
1449 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1450 | bond_dev->name, slave_dev->name, | |
1451 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1452 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1453 | } | |
1454 | } else { | |
5a03cdb7 | 1455 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1456 | if (bond->slave_cnt == 0) { |
1457 | /* First slave, and it is not VLAN challenged, | |
1458 | * so remove the block of adding VLANs over the bond. | |
1459 | */ | |
1460 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1461 | } | |
1462 | } | |
1463 | ||
217df670 JV |
1464 | /* |
1465 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1466 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1467 | * the current ifenslave will set the interface down prior to |
1468 | * enslaving it; the old ifenslave will not. | |
1469 | */ | |
1470 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1471 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1472 | slave_dev->name); |
1473 | res = -EPERM; | |
1474 | goto err_undo_flags; | |
1475 | } | |
1da177e4 | 1476 | |
872254dd MS |
1477 | /* set bonding device ether type by slave - bonding netdevices are |
1478 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1479 | * there is a need to override some of the type dependent attribs/funcs. | |
1480 | * | |
1481 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1482 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1483 | */ | |
1484 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1485 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1486 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1487 | bond_dev->name, |
1488 | bond_dev->type, slave_dev->type); | |
75c78500 | 1489 | |
3ca5b404 JP |
1490 | res = netdev_bonding_change(bond_dev, |
1491 | NETDEV_PRE_TYPE_CHANGE); | |
1492 | res = notifier_to_errno(res); | |
1493 | if (res) { | |
1494 | pr_err("%s: refused to change device type\n", | |
1495 | bond_dev->name); | |
1496 | res = -EBUSY; | |
1497 | goto err_undo_flags; | |
1498 | } | |
75c78500 | 1499 | |
32a806c1 | 1500 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1501 | dev_uc_flush(bond_dev); |
22bedad3 | 1502 | dev_mc_flush(bond_dev); |
32a806c1 | 1503 | |
e36b9d16 MS |
1504 | if (slave_dev->type != ARPHRD_ETHER) |
1505 | bond_setup_by_slave(bond_dev, slave_dev); | |
1506 | else | |
1507 | ether_setup(bond_dev); | |
75c78500 | 1508 | |
93d9b7d7 JP |
1509 | netdev_bonding_change(bond_dev, |
1510 | NETDEV_POST_TYPE_CHANGE); | |
e36b9d16 | 1511 | } |
872254dd | 1512 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1513 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1514 | slave_dev->name, | |
1515 | slave_dev->type, bond_dev->type); | |
1516 | res = -EINVAL; | |
1517 | goto err_undo_flags; | |
872254dd MS |
1518 | } |
1519 | ||
eb7cc59a | 1520 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1521 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1522 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1523 | bond_dev->name); | |
3915c1e8 JV |
1524 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1525 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1526 | pr_err("%s: Error: The slave device specified does not support setting the MAC address, but fail_over_mac is not set to active.\n", |
1527 | bond_dev->name); | |
2ab82852 MS |
1528 | res = -EOPNOTSUPP; |
1529 | goto err_undo_flags; | |
1530 | } | |
1da177e4 LT |
1531 | } |
1532 | ||
c20811a7 JP |
1533 | /* If this is the first slave, then we need to set the master's hardware |
1534 | * address to be the same as the slave's. */ | |
1535 | if (bond->slave_cnt == 0) | |
1536 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, | |
1537 | slave_dev->addr_len); | |
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 | ||
bb1d9123 AG |
1546 | /* |
1547 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1548 | * is set via sysfs or module option if desired. | |
1549 | */ | |
1550 | new_slave->queue_id = 0; | |
1551 | ||
b15ba0fb JP |
1552 | /* Save slave's original mtu and then set it to match the bond */ |
1553 | new_slave->original_mtu = slave_dev->mtu; | |
1554 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1555 | if (res) { | |
1556 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1557 | goto err_free; | |
1558 | } | |
1559 | ||
217df670 JV |
1560 | /* |
1561 | * Save slave's original ("permanent") mac address for modes | |
1562 | * that need it, and for restoring it upon release, and then | |
1563 | * set it to the master's address | |
1564 | */ | |
1565 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1566 | |
dd957c57 | 1567 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1568 | /* |
1569 | * Set slave to master's mac address. The application already | |
1570 | * set the master's mac address to that of the first slave | |
1571 | */ | |
1572 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1573 | addr.sa_family = slave_dev->type; | |
1574 | res = dev_set_mac_address(slave_dev, &addr); | |
1575 | if (res) { | |
5a03cdb7 | 1576 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1577 | goto err_restore_mtu; |
2ab82852 | 1578 | } |
217df670 | 1579 | } |
1da177e4 | 1580 | |
c2edacf8 JV |
1581 | res = netdev_set_master(slave_dev, bond_dev); |
1582 | if (res) { | |
5a03cdb7 | 1583 | pr_debug("Error %d calling netdev_set_master\n", res); |
569f0c4d | 1584 | goto err_restore_mac; |
c2edacf8 | 1585 | } |
217df670 JV |
1586 | /* open the slave since the application closed it */ |
1587 | res = dev_open(slave_dev); | |
1588 | if (res) { | |
3d632c3f | 1589 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
569f0c4d | 1590 | goto err_unset_master; |
1da177e4 LT |
1591 | } |
1592 | ||
1da177e4 | 1593 | new_slave->dev = slave_dev; |
0b680e75 | 1594 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1595 | |
58402054 | 1596 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1597 | /* bond_alb_init_slave() must be called before all other stages since |
1598 | * it might fail and we do not want to have to undo everything | |
1599 | */ | |
1600 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1601 | if (res) |
569f0c4d | 1602 | goto err_close; |
1da177e4 LT |
1603 | } |
1604 | ||
1605 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1606 | * master's promisc (and mc) settings only if it becomes the | |
1607 | * curr_active_slave, and that is taken care of later when calling | |
1608 | * bond_change_active() | |
1609 | */ | |
1610 | if (!USES_PRIMARY(bond->params.mode)) { | |
1611 | /* set promiscuity level to new slave */ | |
1612 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1613 | res = dev_set_promiscuity(slave_dev, 1); |
1614 | if (res) | |
1615 | goto err_close; | |
1da177e4 LT |
1616 | } |
1617 | ||
1618 | /* set allmulti level to new slave */ | |
1619 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1620 | res = dev_set_allmulti(slave_dev, 1); |
1621 | if (res) | |
1622 | goto err_close; | |
1da177e4 LT |
1623 | } |
1624 | ||
b9e40857 | 1625 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1626 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1627 | netdev_for_each_mc_addr(ha, bond_dev) |
1628 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1629 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1630 | } |
1631 | ||
1632 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1633 | /* add lacpdu mc addr to mc list */ | |
1634 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1635 | ||
22bedad3 | 1636 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1637 | } |
1638 | ||
1639 | bond_add_vlans_on_slave(bond, slave_dev); | |
1640 | ||
1641 | write_lock_bh(&bond->lock); | |
1642 | ||
1643 | bond_attach_slave(bond, new_slave); | |
1644 | ||
1645 | new_slave->delay = 0; | |
1646 | new_slave->link_failure_count = 0; | |
1647 | ||
8531c5ff AK |
1648 | bond_compute_features(bond); |
1649 | ||
3915c1e8 JV |
1650 | write_unlock_bh(&bond->lock); |
1651 | ||
1652 | read_lock(&bond->lock); | |
1653 | ||
f5b2b966 JV |
1654 | new_slave->last_arp_rx = jiffies; |
1655 | ||
1da177e4 LT |
1656 | if (bond->params.miimon && !bond->params.use_carrier) { |
1657 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1658 | ||
1659 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1660 | /* | |
1661 | * miimon is set but a bonded network driver | |
1662 | * does not support ETHTOOL/MII and | |
1663 | * arp_interval is not set. Note: if | |
1664 | * use_carrier is enabled, we will never go | |
1665 | * here (because netif_carrier is always | |
1666 | * supported); thus, we don't need to change | |
1667 | * the messages for netif_carrier. | |
1668 | */ | |
a4aee5c8 | 1669 | pr_warning("%s: Warning: MII and ETHTOOL support not available for interface %s, and arp_interval/arp_ip_target module parameters not specified, thus bonding will not detect link failures! see bonding.txt for details.\n", |
4e0952c7 | 1670 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1671 | } else if (link_reporting == -1) { |
1672 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1673 | pr_warning("%s: Warning: can't get link status from interface %s; the network driver associated with this interface does not support MII or ETHTOOL link status reporting, thus miimon has no effect on this interface.\n", |
1674 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1675 | } |
1676 | } | |
1677 | ||
1678 | /* check for initial state */ | |
1679 | if (!bond->params.miimon || | |
1680 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1681 | if (bond->params.updelay) { | |
a4aee5c8 | 1682 | pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n"); |
1da177e4 LT |
1683 | new_slave->link = BOND_LINK_BACK; |
1684 | new_slave->delay = bond->params.updelay; | |
1685 | } else { | |
a4aee5c8 | 1686 | pr_debug("Initial state of slave_dev is BOND_LINK_UP\n"); |
1da177e4 LT |
1687 | new_slave->link = BOND_LINK_UP; |
1688 | } | |
1689 | new_slave->jiffies = jiffies; | |
1690 | } else { | |
a4aee5c8 | 1691 | pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n"); |
1da177e4 LT |
1692 | new_slave->link = BOND_LINK_DOWN; |
1693 | } | |
1694 | ||
1695 | if (bond_update_speed_duplex(new_slave) && | |
1696 | (new_slave->link != BOND_LINK_DOWN)) { | |
a4aee5c8 JP |
1697 | pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n", |
1698 | bond_dev->name, new_slave->dev->name); | |
1da177e4 LT |
1699 | |
1700 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a4aee5c8 JP |
1701 | pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n", |
1702 | bond_dev->name); | |
1da177e4 LT |
1703 | } |
1704 | } | |
1705 | ||
1706 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1707 | /* if there is a primary slave, remember it */ | |
a549952a | 1708 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1709 | bond->primary_slave = new_slave; |
a549952a JP |
1710 | bond->force_primary = true; |
1711 | } | |
1da177e4 LT |
1712 | } |
1713 | ||
3915c1e8 JV |
1714 | write_lock_bh(&bond->curr_slave_lock); |
1715 | ||
1da177e4 LT |
1716 | switch (bond->params.mode) { |
1717 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1718 | bond_set_slave_inactive_flags(new_slave); |
1719 | bond_select_active_slave(bond); | |
1da177e4 LT |
1720 | break; |
1721 | case BOND_MODE_8023AD: | |
1722 | /* in 802.3ad mode, the internal mechanism | |
1723 | * will activate the slaves in the selected | |
1724 | * aggregator | |
1725 | */ | |
1726 | bond_set_slave_inactive_flags(new_slave); | |
1727 | /* if this is the first slave */ | |
1728 | if (bond->slave_cnt == 1) { | |
1729 | SLAVE_AD_INFO(new_slave).id = 1; | |
1730 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1731 | * can be called only after the mac address of the bond is set | |
1732 | */ | |
1733 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1734 | bond->params.lacp_fast); | |
1735 | } else { | |
1736 | SLAVE_AD_INFO(new_slave).id = | |
1737 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1738 | } | |
1739 | ||
1740 | bond_3ad_bind_slave(new_slave); | |
1741 | break; | |
1742 | case BOND_MODE_TLB: | |
1743 | case BOND_MODE_ALB: | |
1744 | new_slave->state = BOND_STATE_ACTIVE; | |
059fe7a5 | 1745 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1746 | bond_select_active_slave(bond); |
1da177e4 LT |
1747 | break; |
1748 | default: | |
5a03cdb7 | 1749 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1750 | |
1751 | /* always active in trunk mode */ | |
1752 | new_slave->state = BOND_STATE_ACTIVE; | |
1753 | ||
1754 | /* In trunking mode there is little meaning to curr_active_slave | |
1755 | * anyway (it holds no special properties of the bond device), | |
1756 | * so we can change it without calling change_active_interface() | |
1757 | */ | |
3d632c3f | 1758 | if (!bond->curr_active_slave) |
1da177e4 | 1759 | bond->curr_active_slave = new_slave; |
3d632c3f | 1760 | |
1da177e4 LT |
1761 | break; |
1762 | } /* switch(bond_mode) */ | |
1763 | ||
3915c1e8 JV |
1764 | write_unlock_bh(&bond->curr_slave_lock); |
1765 | ||
ff59c456 JV |
1766 | bond_set_carrier(bond); |
1767 | ||
f6dc31a8 WC |
1768 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1769 | if (slaves_support_netpoll(bond_dev)) { | |
1770 | bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL; | |
1771 | if (bond_dev->npinfo) | |
1772 | slave_dev->npinfo = bond_dev->npinfo; | |
1773 | } else if (!(bond_dev->priv_flags & IFF_DISABLE_NETPOLL)) { | |
1774 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; | |
1775 | pr_info("New slave device %s does not support netpoll\n", | |
1776 | slave_dev->name); | |
1777 | pr_info("Disabling netpoll support for %s\n", bond_dev->name); | |
1778 | } | |
1779 | #endif | |
3915c1e8 | 1780 | read_unlock(&bond->lock); |
1da177e4 | 1781 | |
b76cdba9 MW |
1782 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1783 | if (res) | |
569f0c4d | 1784 | goto err_close; |
b76cdba9 | 1785 | |
a4aee5c8 JP |
1786 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1787 | bond_dev->name, slave_dev->name, | |
1788 | new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup", | |
1789 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); | |
1da177e4 LT |
1790 | |
1791 | /* enslave is successful */ | |
1792 | return 0; | |
1793 | ||
1794 | /* Undo stages on error */ | |
1da177e4 LT |
1795 | err_close: |
1796 | dev_close(slave_dev); | |
1797 | ||
569f0c4d JV |
1798 | err_unset_master: |
1799 | netdev_set_master(slave_dev, NULL); | |
1800 | ||
1da177e4 | 1801 | err_restore_mac: |
dd957c57 | 1802 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1803 | /* XXX TODO - fom follow mode needs to change master's |
1804 | * MAC if this slave's MAC is in use by the bond, or at | |
1805 | * least print a warning. | |
1806 | */ | |
2ab82852 MS |
1807 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1808 | addr.sa_family = slave_dev->type; | |
1809 | dev_set_mac_address(slave_dev, &addr); | |
1810 | } | |
1da177e4 | 1811 | |
b15ba0fb JP |
1812 | err_restore_mtu: |
1813 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1814 | ||
1da177e4 LT |
1815 | err_free: |
1816 | kfree(new_slave); | |
1817 | ||
1818 | err_undo_flags: | |
1819 | bond_dev->features = old_features; | |
3d632c3f | 1820 | |
1da177e4 LT |
1821 | return res; |
1822 | } | |
1823 | ||
1824 | /* | |
1825 | * Try to release the slave device <slave> from the bond device <master> | |
1826 | * It is legal to access curr_active_slave without a lock because all the function | |
1827 | * is write-locked. | |
1828 | * | |
1829 | * The rules for slave state should be: | |
1830 | * for Active/Backup: | |
1831 | * Active stays on all backups go down | |
1832 | * for Bonded connections: | |
1833 | * The first up interface should be left on and all others downed. | |
1834 | */ | |
a77b5325 | 1835 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1836 | { |
454d7c9b | 1837 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1838 | struct slave *slave, *oldcurrent; |
1839 | struct sockaddr addr; | |
1da177e4 LT |
1840 | |
1841 | /* slave is not a slave or master is not master of this slave */ | |
1842 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1843 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1844 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1845 | bond_dev->name, slave_dev->name); |
1846 | return -EINVAL; | |
1847 | } | |
1848 | ||
f6dc31a8 | 1849 | netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE); |
1da177e4 LT |
1850 | write_lock_bh(&bond->lock); |
1851 | ||
1852 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1853 | if (!slave) { | |
1854 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1855 | pr_info("%s: %s not enslaved\n", |
1856 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1857 | write_unlock_bh(&bond->lock); |
1da177e4 LT |
1858 | return -EINVAL; |
1859 | } | |
1860 | ||
3915c1e8 | 1861 | if (!bond->params.fail_over_mac) { |
8e95a202 JP |
1862 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) && |
1863 | bond->slave_cnt > 1) | |
a4aee5c8 JP |
1864 | pr_warning("%s: Warning: the permanent HWaddr of %s - %pM - is still in use by %s. Set the HWaddr of %s to a different address to avoid conflicts.\n", |
1865 | bond_dev->name, slave_dev->name, | |
1866 | slave->perm_hwaddr, | |
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 | ||
a4aee5c8 JP |
1878 | pr_info("%s: releasing %s interface %s\n", |
1879 | bond_dev->name, | |
1880 | (slave->state == BOND_STATE_ACTIVE) ? "active" : "backup", | |
1881 | slave_dev->name); | |
1da177e4 LT |
1882 | |
1883 | oldcurrent = bond->curr_active_slave; | |
1884 | ||
1885 | bond->current_arp_slave = NULL; | |
1886 | ||
1887 | /* release the slave from its bond */ | |
1888 | bond_detach_slave(bond, slave); | |
1889 | ||
8531c5ff AK |
1890 | bond_compute_features(bond); |
1891 | ||
3d632c3f | 1892 | if (bond->primary_slave == slave) |
1da177e4 | 1893 | bond->primary_slave = NULL; |
1da177e4 | 1894 | |
3d632c3f | 1895 | if (oldcurrent == slave) |
1da177e4 | 1896 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1897 | |
58402054 | 1898 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1899 | /* Must be called only after the slave has been |
1900 | * detached from the list and the curr_active_slave | |
1901 | * has been cleared (if our_slave == old_current), | |
1902 | * but before a new active slave is selected. | |
1903 | */ | |
2543331d | 1904 | write_unlock_bh(&bond->lock); |
1da177e4 | 1905 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1906 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1907 | } |
1908 | ||
059fe7a5 JV |
1909 | if (oldcurrent == slave) { |
1910 | /* | |
1911 | * Note that we hold RTNL over this sequence, so there | |
1912 | * is no concern that another slave add/remove event | |
1913 | * will interfere. | |
1914 | */ | |
1915 | write_unlock_bh(&bond->lock); | |
1916 | read_lock(&bond->lock); | |
1917 | write_lock_bh(&bond->curr_slave_lock); | |
1918 | ||
1da177e4 LT |
1919 | bond_select_active_slave(bond); |
1920 | ||
059fe7a5 JV |
1921 | write_unlock_bh(&bond->curr_slave_lock); |
1922 | read_unlock(&bond->lock); | |
1923 | write_lock_bh(&bond->lock); | |
1924 | } | |
1925 | ||
1da177e4 | 1926 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
1927 | bond_set_carrier(bond); |
1928 | ||
1da177e4 LT |
1929 | /* if the last slave was removed, zero the mac address |
1930 | * of the master so it will be set by the application | |
1931 | * to the mac address of the first slave | |
1932 | */ | |
1933 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
1934 | ||
1935 | if (list_empty(&bond->vlan_list)) { | |
1936 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
1937 | } else { | |
a4aee5c8 JP |
1938 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
1939 | bond_dev->name, bond_dev->name); | |
1940 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
1941 | bond_dev->name); | |
1da177e4 LT |
1942 | } |
1943 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1944 | !bond_has_challenged_slaves(bond)) { | |
a4aee5c8 JP |
1945 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", |
1946 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1947 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
1948 | } | |
1949 | ||
1950 | write_unlock_bh(&bond->lock); | |
1951 | ||
b76cdba9 MW |
1952 | /* must do this from outside any spinlocks */ |
1953 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1954 | ||
1da177e4 LT |
1955 | bond_del_vlans_from_slave(bond, slave_dev); |
1956 | ||
1957 | /* If the mode USES_PRIMARY, then we should only remove its | |
1958 | * promisc and mc settings if it was the curr_active_slave, but that was | |
1959 | * already taken care of above when we detached the slave | |
1960 | */ | |
1961 | if (!USES_PRIMARY(bond->params.mode)) { | |
1962 | /* unset promiscuity level from slave */ | |
3d632c3f | 1963 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 1964 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1965 | |
1966 | /* unset allmulti level from slave */ | |
3d632c3f | 1967 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 1968 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
1969 | |
1970 | /* flush master's mc_list from slave */ | |
b9e40857 | 1971 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1972 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 1973 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1974 | } |
1975 | ||
1976 | netdev_set_master(slave_dev, NULL); | |
1977 | ||
f6dc31a8 WC |
1978 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1979 | read_lock_bh(&bond->lock); | |
1980 | if (slaves_support_netpoll(bond_dev)) | |
1981 | bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL; | |
1982 | read_unlock_bh(&bond->lock); | |
1983 | if (slave_dev->netdev_ops->ndo_netpoll_cleanup) | |
1984 | slave_dev->netdev_ops->ndo_netpoll_cleanup(slave_dev); | |
1985 | else | |
1986 | slave_dev->npinfo = NULL; | |
1987 | #endif | |
1988 | ||
1da177e4 LT |
1989 | /* close slave before restoring its mac address */ |
1990 | dev_close(slave_dev); | |
1991 | ||
3915c1e8 | 1992 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
1993 | /* restore original ("permanent") mac address */ |
1994 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
1995 | addr.sa_family = slave_dev->type; | |
1996 | dev_set_mac_address(slave_dev, &addr); | |
1997 | } | |
1da177e4 | 1998 | |
b15ba0fb JP |
1999 | dev_set_mtu(slave_dev, slave->original_mtu); |
2000 | ||
8f903c70 | 2001 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
f5b2b966 JV |
2002 | IFF_SLAVE_INACTIVE | IFF_BONDING | |
2003 | IFF_SLAVE_NEEDARP); | |
1da177e4 LT |
2004 | |
2005 | kfree(slave); | |
2006 | ||
2007 | return 0; /* deletion OK */ | |
2008 | } | |
2009 | ||
d90a162a | 2010 | /* |
3d632c3f | 2011 | * First release a slave and than destroy the bond if no more slaves are left. |
d90a162a MS |
2012 | * Must be under rtnl_lock when this function is called. |
2013 | */ | |
3d632c3f SH |
2014 | int bond_release_and_destroy(struct net_device *bond_dev, |
2015 | struct net_device *slave_dev) | |
d90a162a | 2016 | { |
454d7c9b | 2017 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2018 | int ret; |
2019 | ||
2020 | ret = bond_release(bond_dev, slave_dev); | |
2021 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
a4aee5c8 JP |
2022 | pr_info("%s: destroying bond %s.\n", |
2023 | bond_dev->name, bond_dev->name); | |
9e71626c | 2024 | unregister_netdevice(bond_dev); |
d90a162a MS |
2025 | } |
2026 | return ret; | |
2027 | } | |
2028 | ||
1da177e4 LT |
2029 | /* |
2030 | * This function releases all slaves. | |
2031 | */ | |
2032 | static int bond_release_all(struct net_device *bond_dev) | |
2033 | { | |
454d7c9b | 2034 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2035 | struct slave *slave; |
2036 | struct net_device *slave_dev; | |
2037 | struct sockaddr addr; | |
2038 | ||
2039 | write_lock_bh(&bond->lock); | |
2040 | ||
ff59c456 JV |
2041 | netif_carrier_off(bond_dev); |
2042 | ||
3d632c3f | 2043 | if (bond->slave_cnt == 0) |
1da177e4 | 2044 | goto out; |
1da177e4 LT |
2045 | |
2046 | bond->current_arp_slave = NULL; | |
2047 | bond->primary_slave = NULL; | |
2048 | bond_change_active_slave(bond, NULL); | |
2049 | ||
2050 | while ((slave = bond->first_slave) != NULL) { | |
2051 | /* Inform AD package of unbinding of slave | |
2052 | * before slave is detached from the list. | |
2053 | */ | |
3d632c3f | 2054 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2055 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2056 | |
2057 | slave_dev = slave->dev; | |
2058 | bond_detach_slave(bond, slave); | |
2059 | ||
2543331d JV |
2060 | /* now that the slave is detached, unlock and perform |
2061 | * all the undo steps that should not be called from | |
2062 | * within a lock. | |
2063 | */ | |
2064 | write_unlock_bh(&bond->lock); | |
2065 | ||
58402054 | 2066 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2067 | /* must be called only after the slave |
2068 | * has been detached from the list | |
2069 | */ | |
2070 | bond_alb_deinit_slave(bond, slave); | |
2071 | } | |
2072 | ||
8531c5ff AK |
2073 | bond_compute_features(bond); |
2074 | ||
b76cdba9 | 2075 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2076 | bond_del_vlans_from_slave(bond, slave_dev); |
2077 | ||
2078 | /* If the mode USES_PRIMARY, then we should only remove its | |
2079 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2080 | * already taken care of above when we detached the slave | |
2081 | */ | |
2082 | if (!USES_PRIMARY(bond->params.mode)) { | |
2083 | /* unset promiscuity level from slave */ | |
3d632c3f | 2084 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2085 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2086 | |
2087 | /* unset allmulti level from slave */ | |
3d632c3f | 2088 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2089 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2090 | |
2091 | /* flush master's mc_list from slave */ | |
b9e40857 | 2092 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2093 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2094 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2095 | } |
2096 | ||
2097 | netdev_set_master(slave_dev, NULL); | |
2098 | ||
2099 | /* close slave before restoring its mac address */ | |
2100 | dev_close(slave_dev); | |
2101 | ||
dd957c57 | 2102 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2103 | /* restore original ("permanent") mac address*/ |
2104 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2105 | addr.sa_family = slave_dev->type; | |
2106 | dev_set_mac_address(slave_dev, &addr); | |
2107 | } | |
1da177e4 | 2108 | |
8f903c70 JV |
2109 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
2110 | IFF_SLAVE_INACTIVE); | |
1da177e4 LT |
2111 | |
2112 | kfree(slave); | |
2113 | ||
2114 | /* re-acquire the lock before getting the next slave */ | |
2115 | write_lock_bh(&bond->lock); | |
2116 | } | |
2117 | ||
2118 | /* zero the mac address of the master so it will be | |
2119 | * set by the application to the mac address of the | |
2120 | * first slave | |
2121 | */ | |
2122 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2123 | ||
3d632c3f | 2124 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 2125 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
3d632c3f | 2126 | else { |
a4aee5c8 JP |
2127 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2128 | bond_dev->name, bond_dev->name); | |
2129 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2130 | bond_dev->name); | |
1da177e4 LT |
2131 | } |
2132 | ||
a4aee5c8 | 2133 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2134 | |
2135 | out: | |
2136 | write_unlock_bh(&bond->lock); | |
2137 | ||
2138 | return 0; | |
2139 | } | |
2140 | ||
2141 | /* | |
2142 | * This function changes the active slave to slave <slave_dev>. | |
2143 | * It returns -EINVAL in the following cases. | |
2144 | * - <slave_dev> is not found in the list. | |
2145 | * - There is not active slave now. | |
2146 | * - <slave_dev> is already active. | |
2147 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2148 | * - <slave_dev> is not running. | |
3d632c3f SH |
2149 | * In these cases, this function does nothing. |
2150 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2151 | */ |
2152 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2153 | { | |
454d7c9b | 2154 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2155 | struct slave *old_active = NULL; |
2156 | struct slave *new_active = NULL; | |
2157 | int res = 0; | |
2158 | ||
3d632c3f | 2159 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2160 | return -EINVAL; |
1da177e4 LT |
2161 | |
2162 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2163 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2164 | return -EINVAL; |
1da177e4 | 2165 | |
059fe7a5 | 2166 | read_lock(&bond->lock); |
1da177e4 | 2167 | |
059fe7a5 | 2168 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2169 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2170 | read_unlock(&bond->curr_slave_lock); |
2171 | ||
1da177e4 LT |
2172 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2173 | ||
2174 | /* | |
2175 | * Changing to the current active: do nothing; return success. | |
2176 | */ | |
2177 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2178 | read_unlock(&bond->lock); |
1da177e4 LT |
2179 | return 0; |
2180 | } | |
2181 | ||
2182 | if ((new_active) && | |
2183 | (old_active) && | |
2184 | (new_active->link == BOND_LINK_UP) && | |
2185 | IS_UP(new_active->dev)) { | |
059fe7a5 | 2186 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2187 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2188 | write_unlock_bh(&bond->curr_slave_lock); |
3d632c3f | 2189 | } else |
1da177e4 | 2190 | res = -EINVAL; |
1da177e4 | 2191 | |
059fe7a5 | 2192 | read_unlock(&bond->lock); |
1da177e4 LT |
2193 | |
2194 | return res; | |
2195 | } | |
2196 | ||
1da177e4 LT |
2197 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2198 | { | |
454d7c9b | 2199 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2200 | |
2201 | info->bond_mode = bond->params.mode; | |
2202 | info->miimon = bond->params.miimon; | |
2203 | ||
6603a6f2 | 2204 | read_lock(&bond->lock); |
1da177e4 | 2205 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2206 | read_unlock(&bond->lock); |
1da177e4 LT |
2207 | |
2208 | return 0; | |
2209 | } | |
2210 | ||
2211 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2212 | { | |
454d7c9b | 2213 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2214 | struct slave *slave; |
689c96cc | 2215 | int i, res = -ENODEV; |
1da177e4 | 2216 | |
6603a6f2 | 2217 | read_lock(&bond->lock); |
1da177e4 LT |
2218 | |
2219 | bond_for_each_slave(bond, slave, i) { | |
2220 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2221 | res = 0; |
2222 | strcpy(info->slave_name, slave->dev->name); | |
2223 | info->link = slave->link; | |
2224 | info->state = slave->state; | |
2225 | info->link_failure_count = slave->link_failure_count; | |
1da177e4 LT |
2226 | break; |
2227 | } | |
2228 | } | |
2229 | ||
6603a6f2 | 2230 | read_unlock(&bond->lock); |
1da177e4 | 2231 | |
689c96cc | 2232 | return res; |
1da177e4 LT |
2233 | } |
2234 | ||
2235 | /*-------------------------------- Monitoring -------------------------------*/ | |
2236 | ||
1da177e4 | 2237 | |
f0c76d61 JV |
2238 | static int bond_miimon_inspect(struct bonding *bond) |
2239 | { | |
2240 | struct slave *slave; | |
2241 | int i, link_state, commit = 0; | |
41f89100 JP |
2242 | bool ignore_updelay; |
2243 | ||
2244 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2245 | |
2246 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2247 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2248 | |
f0c76d61 | 2249 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2250 | |
2251 | switch (slave->link) { | |
f0c76d61 JV |
2252 | case BOND_LINK_UP: |
2253 | if (link_state) | |
2254 | continue; | |
1da177e4 | 2255 | |
f0c76d61 JV |
2256 | slave->link = BOND_LINK_FAIL; |
2257 | slave->delay = bond->params.downdelay; | |
2258 | if (slave->delay) { | |
a4aee5c8 JP |
2259 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2260 | bond->dev->name, | |
2261 | (bond->params.mode == | |
2262 | BOND_MODE_ACTIVEBACKUP) ? | |
2263 | ((slave->state == BOND_STATE_ACTIVE) ? | |
2264 | "active " : "backup ") : "", | |
2265 | slave->dev->name, | |
2266 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2267 | } |
f0c76d61 JV |
2268 | /*FALLTHRU*/ |
2269 | case BOND_LINK_FAIL: | |
2270 | if (link_state) { | |
2271 | /* | |
2272 | * recovered before downdelay expired | |
2273 | */ | |
2274 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2275 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2276 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2277 | bond->dev->name, | |
2278 | (bond->params.downdelay - slave->delay) * | |
2279 | bond->params.miimon, | |
2280 | slave->dev->name); | |
f0c76d61 | 2281 | continue; |
1da177e4 | 2282 | } |
f0c76d61 JV |
2283 | |
2284 | if (slave->delay <= 0) { | |
2285 | slave->new_link = BOND_LINK_DOWN; | |
2286 | commit++; | |
2287 | continue; | |
1da177e4 | 2288 | } |
1da177e4 | 2289 | |
f0c76d61 JV |
2290 | slave->delay--; |
2291 | break; | |
2292 | ||
2293 | case BOND_LINK_DOWN: | |
2294 | if (!link_state) | |
2295 | continue; | |
2296 | ||
2297 | slave->link = BOND_LINK_BACK; | |
2298 | slave->delay = bond->params.updelay; | |
2299 | ||
2300 | if (slave->delay) { | |
a4aee5c8 JP |
2301 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2302 | bond->dev->name, slave->dev->name, | |
2303 | ignore_updelay ? 0 : | |
2304 | bond->params.updelay * | |
2305 | bond->params.miimon); | |
f0c76d61 JV |
2306 | } |
2307 | /*FALLTHRU*/ | |
2308 | case BOND_LINK_BACK: | |
2309 | if (!link_state) { | |
2310 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2311 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2312 | bond->dev->name, | |
2313 | (bond->params.updelay - slave->delay) * | |
2314 | bond->params.miimon, | |
2315 | slave->dev->name); | |
f0c76d61 JV |
2316 | |
2317 | continue; | |
2318 | } | |
2319 | ||
41f89100 JP |
2320 | if (ignore_updelay) |
2321 | slave->delay = 0; | |
2322 | ||
f0c76d61 JV |
2323 | if (slave->delay <= 0) { |
2324 | slave->new_link = BOND_LINK_UP; | |
2325 | commit++; | |
41f89100 | 2326 | ignore_updelay = false; |
f0c76d61 | 2327 | continue; |
1da177e4 | 2328 | } |
f0c76d61 JV |
2329 | |
2330 | slave->delay--; | |
1da177e4 | 2331 | break; |
f0c76d61 JV |
2332 | } |
2333 | } | |
1da177e4 | 2334 | |
f0c76d61 JV |
2335 | return commit; |
2336 | } | |
1da177e4 | 2337 | |
f0c76d61 JV |
2338 | static void bond_miimon_commit(struct bonding *bond) |
2339 | { | |
2340 | struct slave *slave; | |
2341 | int i; | |
2342 | ||
2343 | bond_for_each_slave(bond, slave, i) { | |
2344 | switch (slave->new_link) { | |
2345 | case BOND_LINK_NOCHANGE: | |
2346 | continue; | |
1da177e4 | 2347 | |
f0c76d61 JV |
2348 | case BOND_LINK_UP: |
2349 | slave->link = BOND_LINK_UP; | |
2350 | slave->jiffies = jiffies; | |
2351 | ||
2352 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2353 | /* prevent it from being the active one */ | |
2354 | slave->state = BOND_STATE_BACKUP; | |
2355 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
2356 | /* make it immediately active */ | |
2357 | slave->state = BOND_STATE_ACTIVE; | |
2358 | } else if (slave != bond->primary_slave) { | |
2359 | /* prevent it from being the active one */ | |
2360 | slave->state = BOND_STATE_BACKUP; | |
1da177e4 | 2361 | } |
1da177e4 | 2362 | |
a4aee5c8 JP |
2363 | pr_info("%s: link status definitely up for interface %s.\n", |
2364 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2365 | |
f0c76d61 JV |
2366 | /* notify ad that the link status has changed */ |
2367 | if (bond->params.mode == BOND_MODE_8023AD) | |
2368 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2369 | |
58402054 | 2370 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2371 | bond_alb_handle_link_change(bond, slave, |
2372 | BOND_LINK_UP); | |
1da177e4 | 2373 | |
f0c76d61 JV |
2374 | if (!bond->curr_active_slave || |
2375 | (slave == bond->primary_slave)) | |
2376 | goto do_failover; | |
1da177e4 | 2377 | |
f0c76d61 | 2378 | continue; |
059fe7a5 | 2379 | |
f0c76d61 | 2380 | case BOND_LINK_DOWN: |
fba4acda JV |
2381 | if (slave->link_failure_count < UINT_MAX) |
2382 | slave->link_failure_count++; | |
2383 | ||
f0c76d61 | 2384 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2385 | |
f0c76d61 JV |
2386 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2387 | bond->params.mode == BOND_MODE_8023AD) | |
2388 | bond_set_slave_inactive_flags(slave); | |
2389 | ||
a4aee5c8 JP |
2390 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2391 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2392 | |
2393 | if (bond->params.mode == BOND_MODE_8023AD) | |
2394 | bond_3ad_handle_link_change(slave, | |
2395 | BOND_LINK_DOWN); | |
2396 | ||
ae63e808 | 2397 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2398 | bond_alb_handle_link_change(bond, slave, |
2399 | BOND_LINK_DOWN); | |
2400 | ||
2401 | if (slave == bond->curr_active_slave) | |
2402 | goto do_failover; | |
2403 | ||
2404 | continue; | |
2405 | ||
2406 | default: | |
a4aee5c8 | 2407 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2408 | bond->dev->name, slave->new_link, |
2409 | slave->dev->name); | |
2410 | slave->new_link = BOND_LINK_NOCHANGE; | |
2411 | ||
2412 | continue; | |
2413 | } | |
2414 | ||
2415 | do_failover: | |
2416 | ASSERT_RTNL(); | |
2417 | write_lock_bh(&bond->curr_slave_lock); | |
2418 | bond_select_active_slave(bond); | |
2419 | write_unlock_bh(&bond->curr_slave_lock); | |
2420 | } | |
2421 | ||
2422 | bond_set_carrier(bond); | |
1da177e4 LT |
2423 | } |
2424 | ||
0b0eef66 JV |
2425 | /* |
2426 | * bond_mii_monitor | |
2427 | * | |
2428 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2429 | * inspection, then (if inspection indicates something needs to be done) |
2430 | * an acquisition of appropriate locks followed by a commit phase to | |
2431 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2432 | */ |
2433 | void bond_mii_monitor(struct work_struct *work) | |
2434 | { | |
2435 | struct bonding *bond = container_of(work, struct bonding, | |
2436 | mii_work.work); | |
0b0eef66 JV |
2437 | |
2438 | read_lock(&bond->lock); | |
f0c76d61 JV |
2439 | if (bond->kill_timers) |
2440 | goto out; | |
2441 | ||
2442 | if (bond->slave_cnt == 0) | |
2443 | goto re_arm; | |
b59f9f74 JV |
2444 | |
2445 | if (bond->send_grat_arp) { | |
2446 | read_lock(&bond->curr_slave_lock); | |
2447 | bond_send_gratuitous_arp(bond); | |
2448 | read_unlock(&bond->curr_slave_lock); | |
2449 | } | |
2450 | ||
305d552a BH |
2451 | if (bond->send_unsol_na) { |
2452 | read_lock(&bond->curr_slave_lock); | |
2453 | bond_send_unsolicited_na(bond); | |
2454 | read_unlock(&bond->curr_slave_lock); | |
2455 | } | |
2456 | ||
f0c76d61 | 2457 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2458 | read_unlock(&bond->lock); |
2459 | rtnl_lock(); | |
2460 | read_lock(&bond->lock); | |
f0c76d61 JV |
2461 | |
2462 | bond_miimon_commit(bond); | |
2463 | ||
5655662d JV |
2464 | read_unlock(&bond->lock); |
2465 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2466 | read_lock(&bond->lock); | |
0b0eef66 JV |
2467 | } |
2468 | ||
f0c76d61 JV |
2469 | re_arm: |
2470 | if (bond->params.miimon) | |
2471 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2472 | msecs_to_jiffies(bond->params.miimon)); | |
2473 | out: | |
0b0eef66 | 2474 | read_unlock(&bond->lock); |
0b0eef66 | 2475 | } |
c3ade5ca | 2476 | |
d3bb52b0 | 2477 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2478 | { |
2479 | struct in_device *idev; | |
2480 | struct in_ifaddr *ifa; | |
a144ea4b | 2481 | __be32 addr = 0; |
c3ade5ca JV |
2482 | |
2483 | if (!dev) | |
2484 | return 0; | |
2485 | ||
2486 | rcu_read_lock(); | |
e5ed6399 | 2487 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2488 | if (!idev) |
2489 | goto out; | |
2490 | ||
2491 | ifa = idev->ifa_list; | |
2492 | if (!ifa) | |
2493 | goto out; | |
2494 | ||
2495 | addr = ifa->ifa_local; | |
2496 | out: | |
2497 | rcu_read_unlock(); | |
2498 | return addr; | |
2499 | } | |
2500 | ||
d3bb52b0 | 2501 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2502 | { |
0883beca | 2503 | struct vlan_entry *vlan; |
f5b2b966 JV |
2504 | |
2505 | if (ip == bond->master_ip) | |
2506 | return 1; | |
2507 | ||
0883beca | 2508 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2509 | if (ip == vlan->vlan_ip) |
2510 | return 1; | |
2511 | } | |
2512 | ||
2513 | return 0; | |
2514 | } | |
2515 | ||
c3ade5ca JV |
2516 | /* |
2517 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2518 | * switches in VLAN mode (especially if ports are configured as | |
2519 | * "native" to a VLAN) might not pass non-tagged frames. | |
2520 | */ | |
d3bb52b0 | 2521 | 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 |
2522 | { |
2523 | struct sk_buff *skb; | |
2524 | ||
5a03cdb7 | 2525 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2526 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2527 | |
c3ade5ca JV |
2528 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2529 | NULL, slave_dev->dev_addr, NULL); | |
2530 | ||
2531 | if (!skb) { | |
a4aee5c8 | 2532 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2533 | return; |
2534 | } | |
2535 | if (vlan_id) { | |
2536 | skb = vlan_put_tag(skb, vlan_id); | |
2537 | if (!skb) { | |
a4aee5c8 | 2538 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2539 | return; |
2540 | } | |
2541 | } | |
2542 | arp_xmit(skb); | |
2543 | } | |
2544 | ||
2545 | ||
1da177e4 LT |
2546 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2547 | { | |
c3ade5ca | 2548 | int i, vlan_id, rv; |
d3bb52b0 | 2549 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2550 | struct vlan_entry *vlan; |
c3ade5ca JV |
2551 | struct net_device *vlan_dev; |
2552 | struct flowi fl; | |
2553 | struct rtable *rt; | |
1da177e4 | 2554 | |
6b780567 MW |
2555 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2556 | if (!targets[i]) | |
5a31bec0 | 2557 | break; |
5a03cdb7 | 2558 | pr_debug("basa: target %x\n", targets[i]); |
c3ade5ca | 2559 | if (list_empty(&bond->vlan_list)) { |
5a03cdb7 | 2560 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2561 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2562 | bond->master_ip, 0); | |
2563 | continue; | |
2564 | } | |
2565 | ||
2566 | /* | |
2567 | * If VLANs are configured, we do a route lookup to | |
2568 | * determine which VLAN interface would be used, so we | |
2569 | * can tag the ARP with the proper VLAN tag. | |
2570 | */ | |
2571 | memset(&fl, 0, sizeof(fl)); | |
2572 | fl.fl4_dst = targets[i]; | |
2573 | fl.fl4_tos = RTO_ONLINK; | |
2574 | ||
ec87fd3b | 2575 | rv = ip_route_output_key(dev_net(bond->dev), &rt, &fl); |
c3ade5ca JV |
2576 | if (rv) { |
2577 | if (net_ratelimit()) { | |
a4aee5c8 JP |
2578 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
2579 | bond->dev->name, &fl.fl4_dst); | |
c3ade5ca JV |
2580 | } |
2581 | continue; | |
2582 | } | |
2583 | ||
2584 | /* | |
2585 | * This target is not on a VLAN | |
2586 | */ | |
d8d1f30b | 2587 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2588 | ip_rt_put(rt); |
5a03cdb7 | 2589 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2590 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2591 | bond->master_ip, 0); | |
2592 | continue; | |
2593 | } | |
2594 | ||
2595 | vlan_id = 0; | |
0883beca | 2596 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2597 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
d8d1f30b | 2598 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2599 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2600 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2601 | vlan_dev->name, vlan_id); |
2602 | break; | |
2603 | } | |
2604 | } | |
2605 | ||
2606 | if (vlan_id) { | |
ed4b9f80 | 2607 | ip_rt_put(rt); |
c3ade5ca JV |
2608 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2609 | vlan->vlan_ip, vlan_id); | |
2610 | continue; | |
2611 | } | |
2612 | ||
2613 | if (net_ratelimit()) { | |
a4aee5c8 JP |
2614 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2615 | bond->dev->name, &fl.fl4_dst, | |
d8d1f30b | 2616 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2617 | } |
ed4b9f80 | 2618 | ip_rt_put(rt); |
c3ade5ca JV |
2619 | } |
2620 | } | |
2621 | ||
2622 | /* | |
2623 | * Kick out a gratuitous ARP for an IP on the bonding master plus one | |
2624 | * for each VLAN above us. | |
b59f9f74 JV |
2625 | * |
2626 | * Caller must hold curr_slave_lock for read or better | |
c3ade5ca JV |
2627 | */ |
2628 | static void bond_send_gratuitous_arp(struct bonding *bond) | |
2629 | { | |
2630 | struct slave *slave = bond->curr_active_slave; | |
2631 | struct vlan_entry *vlan; | |
2632 | struct net_device *vlan_dev; | |
2633 | ||
a4aee5c8 JP |
2634 | pr_debug("bond_send_grat_arp: bond %s slave %s\n", |
2635 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
b59f9f74 JV |
2636 | |
2637 | if (!slave || !bond->send_grat_arp || | |
2638 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
c3ade5ca JV |
2639 | return; |
2640 | ||
b59f9f74 JV |
2641 | bond->send_grat_arp--; |
2642 | ||
c3ade5ca JV |
2643 | if (bond->master_ip) { |
2644 | bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip, | |
1053f62c | 2645 | bond->master_ip, 0); |
c3ade5ca JV |
2646 | } |
2647 | ||
2648 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
5c15bdec | 2649 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2650 | if (vlan->vlan_ip) { |
2651 | bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip, | |
2652 | vlan->vlan_ip, vlan->vlan_id); | |
2653 | } | |
1da177e4 LT |
2654 | } |
2655 | } | |
2656 | ||
d3bb52b0 | 2657 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2658 | { |
2659 | int i; | |
d3bb52b0 | 2660 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2661 | |
f5b2b966 | 2662 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2663 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2664 | &sip, &tip, i, &targets[i], |
2665 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2666 | if (sip == targets[i]) { |
2667 | if (bond_has_this_ip(bond, tip)) | |
2668 | slave->last_arp_rx = jiffies; | |
2669 | return; | |
2670 | } | |
2671 | } | |
2672 | } | |
2673 | ||
2674 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2675 | { | |
2676 | struct arphdr *arp; | |
2677 | struct slave *slave; | |
2678 | struct bonding *bond; | |
2679 | unsigned char *arp_ptr; | |
d3bb52b0 | 2680 | __be32 sip, tip; |
f5b2b966 | 2681 | |
1f3c8804 AG |
2682 | if (dev->priv_flags & IFF_802_1Q_VLAN) { |
2683 | /* | |
2684 | * When using VLANS and bonding, dev and oriv_dev may be | |
2685 | * incorrect if the physical interface supports VLAN | |
2686 | * acceleration. With this change ARP validation now | |
2687 | * works for hosts only reachable on the VLAN interface. | |
2688 | */ | |
2689 | dev = vlan_dev_real_dev(dev); | |
2690 | orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif); | |
2691 | } | |
2692 | ||
f5b2b966 JV |
2693 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2694 | goto out; | |
2695 | ||
454d7c9b | 2696 | bond = netdev_priv(dev); |
f5b2b966 JV |
2697 | read_lock(&bond->lock); |
2698 | ||
5a03cdb7 | 2699 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
a4aee5c8 JP |
2700 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2701 | orig_dev ? orig_dev->name : "NULL"); | |
f5b2b966 JV |
2702 | |
2703 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2704 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2705 | goto out_unlock; | |
2706 | ||
988b7050 | 2707 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2708 | goto out_unlock; |
2709 | ||
d0a92be0 | 2710 | arp = arp_hdr(skb); |
f5b2b966 JV |
2711 | if (arp->ar_hln != dev->addr_len || |
2712 | skb->pkt_type == PACKET_OTHERHOST || | |
2713 | skb->pkt_type == PACKET_LOOPBACK || | |
2714 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2715 | arp->ar_pro != htons(ETH_P_IP) || | |
2716 | arp->ar_pln != 4) | |
2717 | goto out_unlock; | |
2718 | ||
2719 | arp_ptr = (unsigned char *)(arp + 1); | |
2720 | arp_ptr += dev->addr_len; | |
2721 | memcpy(&sip, arp_ptr, 4); | |
2722 | arp_ptr += 4 + dev->addr_len; | |
2723 | memcpy(&tip, arp_ptr, 4); | |
2724 | ||
5a03cdb7 | 2725 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
a4aee5c8 JP |
2726 | bond->dev->name, slave->dev->name, slave->state, |
2727 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2728 | &sip, &tip); | |
f5b2b966 JV |
2729 | |
2730 | /* | |
2731 | * Backup slaves won't see the ARP reply, but do come through | |
2732 | * here for each ARP probe (so we swap the sip/tip to validate | |
2733 | * the probe). In a "redundant switch, common router" type of | |
2734 | * configuration, the ARP probe will (hopefully) travel from | |
2735 | * the active, through one switch, the router, then the other | |
2736 | * switch before reaching the backup. | |
2737 | */ | |
2738 | if (slave->state == BOND_STATE_ACTIVE) | |
2739 | bond_validate_arp(bond, slave, sip, tip); | |
2740 | else | |
2741 | bond_validate_arp(bond, slave, tip, sip); | |
2742 | ||
2743 | out_unlock: | |
2744 | read_unlock(&bond->lock); | |
2745 | out: | |
2746 | dev_kfree_skb(skb); | |
2747 | return NET_RX_SUCCESS; | |
2748 | } | |
2749 | ||
1da177e4 LT |
2750 | /* |
2751 | * this function is called regularly to monitor each slave's link | |
2752 | * ensuring that traffic is being sent and received when arp monitoring | |
2753 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2754 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2755 | * arp monitoring in active backup mode. | |
2756 | */ | |
1b76b316 | 2757 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2758 | { |
1b76b316 JV |
2759 | struct bonding *bond = container_of(work, struct bonding, |
2760 | arp_work.work); | |
1da177e4 LT |
2761 | struct slave *slave, *oldcurrent; |
2762 | int do_failover = 0; | |
2763 | int delta_in_ticks; | |
2764 | int i; | |
2765 | ||
2766 | read_lock(&bond->lock); | |
2767 | ||
5ce0da8f | 2768 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2769 | |
3d632c3f | 2770 | if (bond->kill_timers) |
1da177e4 | 2771 | goto out; |
1da177e4 | 2772 | |
3d632c3f | 2773 | if (bond->slave_cnt == 0) |
1da177e4 | 2774 | goto re_arm; |
1da177e4 LT |
2775 | |
2776 | read_lock(&bond->curr_slave_lock); | |
2777 | oldcurrent = bond->curr_active_slave; | |
2778 | read_unlock(&bond->curr_slave_lock); | |
2779 | ||
2780 | /* see if any of the previous devices are up now (i.e. they have | |
2781 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2782 | * the picture unless it is null. also, slave->jiffies is not needed | |
2783 | * here because we send an arp on each slave and give a slave as | |
2784 | * long as it needs to get the tx/rx within the delta. | |
2785 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2786 | * so it can wait | |
2787 | */ | |
2788 | bond_for_each_slave(bond, slave, i) { | |
2789 | if (slave->link != BOND_LINK_UP) { | |
9d21493b | 2790 | if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) && |
b63bb739 | 2791 | time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) { |
1da177e4 LT |
2792 | |
2793 | slave->link = BOND_LINK_UP; | |
2794 | slave->state = BOND_STATE_ACTIVE; | |
2795 | ||
2796 | /* primary_slave has no meaning in round-robin | |
2797 | * mode. the window of a slave being up and | |
2798 | * curr_active_slave being null after enslaving | |
2799 | * is closed. | |
2800 | */ | |
2801 | if (!oldcurrent) { | |
a4aee5c8 JP |
2802 | pr_info("%s: link status definitely up for interface %s, ", |
2803 | bond->dev->name, | |
2804 | slave->dev->name); | |
1da177e4 LT |
2805 | do_failover = 1; |
2806 | } else { | |
a4aee5c8 JP |
2807 | pr_info("%s: interface %s is now up\n", |
2808 | bond->dev->name, | |
2809 | slave->dev->name); | |
1da177e4 LT |
2810 | } |
2811 | } | |
2812 | } else { | |
2813 | /* slave->link == BOND_LINK_UP */ | |
2814 | ||
2815 | /* not all switches will respond to an arp request | |
2816 | * when the source ip is 0, so don't take the link down | |
2817 | * if we don't know our ip yet | |
2818 | */ | |
9d21493b | 2819 | if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) || |
4b8a9239 | 2820 | (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) { |
1da177e4 LT |
2821 | |
2822 | slave->link = BOND_LINK_DOWN; | |
2823 | slave->state = BOND_STATE_BACKUP; | |
2824 | ||
3d632c3f | 2825 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2826 | slave->link_failure_count++; |
1da177e4 | 2827 | |
a4aee5c8 JP |
2828 | pr_info("%s: interface %s is now down.\n", |
2829 | bond->dev->name, | |
2830 | slave->dev->name); | |
1da177e4 | 2831 | |
3d632c3f | 2832 | if (slave == oldcurrent) |
1da177e4 | 2833 | do_failover = 1; |
1da177e4 LT |
2834 | } |
2835 | } | |
2836 | ||
2837 | /* note: if switch is in round-robin mode, all links | |
2838 | * must tx arp to ensure all links rx an arp - otherwise | |
2839 | * links may oscillate or not come up at all; if switch is | |
2840 | * in something like xor mode, there is nothing we can | |
2841 | * do - all replies will be rx'ed on same link causing slaves | |
2842 | * to be unstable during low/no traffic periods | |
2843 | */ | |
3d632c3f | 2844 | if (IS_UP(slave->dev)) |
1da177e4 | 2845 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2846 | } |
2847 | ||
2848 | if (do_failover) { | |
059fe7a5 | 2849 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2850 | |
2851 | bond_select_active_slave(bond); | |
2852 | ||
059fe7a5 | 2853 | write_unlock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2854 | } |
2855 | ||
2856 | re_arm: | |
1b76b316 JV |
2857 | if (bond->params.arp_interval) |
2858 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
2859 | out: |
2860 | read_unlock(&bond->lock); | |
2861 | } | |
2862 | ||
2863 | /* | |
b2220cad JV |
2864 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2865 | * changes. Sets new_link in slaves to specify what action should take | |
2866 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2867 | * to link states must be committed. | |
2868 | * | |
2869 | * Called with bond->lock held for read. | |
1da177e4 | 2870 | */ |
b2220cad | 2871 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2872 | { |
1da177e4 | 2873 | struct slave *slave; |
b2220cad | 2874 | int i, commit = 0; |
1da177e4 | 2875 | |
b2220cad JV |
2876 | bond_for_each_slave(bond, slave, i) { |
2877 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2878 | |
b2220cad JV |
2879 | if (slave->link != BOND_LINK_UP) { |
2880 | if (time_before_eq(jiffies, slave_last_rx(bond, slave) + | |
2881 | delta_in_ticks)) { | |
2882 | slave->new_link = BOND_LINK_UP; | |
2883 | commit++; | |
2884 | } | |
1da177e4 | 2885 | |
b2220cad JV |
2886 | continue; |
2887 | } | |
1da177e4 | 2888 | |
b2220cad JV |
2889 | /* |
2890 | * Give slaves 2*delta after being enslaved or made | |
2891 | * active. This avoids bouncing, as the last receive | |
2892 | * times need a full ARP monitor cycle to be updated. | |
2893 | */ | |
2894 | if (!time_after_eq(jiffies, slave->jiffies + | |
2895 | 2 * delta_in_ticks)) | |
2896 | continue; | |
2897 | ||
2898 | /* | |
2899 | * Backup slave is down if: | |
2900 | * - No current_arp_slave AND | |
2901 | * - more than 3*delta since last receive AND | |
2902 | * - the bond has an IP address | |
2903 | * | |
2904 | * Note: a non-null current_arp_slave indicates | |
2905 | * the curr_active_slave went down and we are | |
2906 | * searching for a new one; under this condition | |
2907 | * we only take the curr_active_slave down - this | |
2908 | * gives each slave a chance to tx/rx traffic | |
2909 | * before being taken out | |
2910 | */ | |
2911 | if (slave->state == BOND_STATE_BACKUP && | |
2912 | !bond->current_arp_slave && | |
2913 | time_after(jiffies, slave_last_rx(bond, slave) + | |
2914 | 3 * delta_in_ticks)) { | |
2915 | slave->new_link = BOND_LINK_DOWN; | |
2916 | commit++; | |
2917 | } | |
2918 | ||
2919 | /* | |
2920 | * Active slave is down if: | |
2921 | * - more than 2*delta since transmitting OR | |
2922 | * - (more than 2*delta since receive AND | |
2923 | * the bond has an IP address) | |
2924 | */ | |
2925 | if ((slave->state == BOND_STATE_ACTIVE) && | |
9d21493b | 2926 | (time_after_eq(jiffies, dev_trans_start(slave->dev) + |
b2220cad JV |
2927 | 2 * delta_in_ticks) || |
2928 | (time_after_eq(jiffies, slave_last_rx(bond, slave) | |
2929 | + 2 * delta_in_ticks)))) { | |
2930 | slave->new_link = BOND_LINK_DOWN; | |
2931 | commit++; | |
2932 | } | |
1da177e4 LT |
2933 | } |
2934 | ||
b2220cad JV |
2935 | return commit; |
2936 | } | |
1da177e4 | 2937 | |
b2220cad JV |
2938 | /* |
2939 | * Called to commit link state changes noted by inspection step of | |
2940 | * active-backup mode ARP monitor. | |
2941 | * | |
2942 | * Called with RTNL and bond->lock for read. | |
2943 | */ | |
2944 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2945 | { | |
2946 | struct slave *slave; | |
2947 | int i; | |
1da177e4 | 2948 | |
b2220cad JV |
2949 | bond_for_each_slave(bond, slave, i) { |
2950 | switch (slave->new_link) { | |
2951 | case BOND_LINK_NOCHANGE: | |
2952 | continue; | |
ff59c456 | 2953 | |
b2220cad | 2954 | case BOND_LINK_UP: |
b9f60253 JP |
2955 | if ((!bond->curr_active_slave && |
2956 | time_before_eq(jiffies, | |
2957 | dev_trans_start(slave->dev) + | |
2958 | delta_in_ticks)) || | |
2959 | bond->curr_active_slave != slave) { | |
b2220cad | 2960 | slave->link = BOND_LINK_UP; |
b2220cad JV |
2961 | bond->current_arp_slave = NULL; |
2962 | ||
a4aee5c8 | 2963 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2964 | bond->dev->name, slave->dev->name); |
b2220cad | 2965 | |
b9f60253 JP |
2966 | if (!bond->curr_active_slave || |
2967 | (slave == bond->primary_slave)) | |
2968 | goto do_failover; | |
1da177e4 | 2969 | |
b9f60253 | 2970 | } |
1da177e4 | 2971 | |
b9f60253 | 2972 | continue; |
1da177e4 | 2973 | |
b2220cad JV |
2974 | case BOND_LINK_DOWN: |
2975 | if (slave->link_failure_count < UINT_MAX) | |
2976 | slave->link_failure_count++; | |
2977 | ||
2978 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2979 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2980 | |
a4aee5c8 | 2981 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2982 | bond->dev->name, slave->dev->name); |
b2220cad | 2983 | |
b9f60253 | 2984 | if (slave == bond->curr_active_slave) { |
b2220cad | 2985 | bond->current_arp_slave = NULL; |
b9f60253 | 2986 | goto do_failover; |
1da177e4 | 2987 | } |
b9f60253 JP |
2988 | |
2989 | continue; | |
b2220cad JV |
2990 | |
2991 | default: | |
a4aee5c8 | 2992 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2993 | bond->dev->name, slave->new_link, |
2994 | slave->dev->name); | |
b9f60253 | 2995 | continue; |
1da177e4 | 2996 | } |
1da177e4 | 2997 | |
b9f60253 JP |
2998 | do_failover: |
2999 | ASSERT_RTNL(); | |
b2220cad | 3000 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3001 | bond_select_active_slave(bond); |
b2220cad JV |
3002 | write_unlock_bh(&bond->curr_slave_lock); |
3003 | } | |
1da177e4 | 3004 | |
b2220cad JV |
3005 | bond_set_carrier(bond); |
3006 | } | |
1da177e4 | 3007 | |
b2220cad JV |
3008 | /* |
3009 | * Send ARP probes for active-backup mode ARP monitor. | |
3010 | * | |
3011 | * Called with bond->lock held for read. | |
3012 | */ | |
3013 | static void bond_ab_arp_probe(struct bonding *bond) | |
3014 | { | |
3015 | struct slave *slave; | |
3016 | int i; | |
1da177e4 | 3017 | |
b2220cad | 3018 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3019 | |
b2220cad | 3020 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3021 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3022 | bond->current_arp_slave->dev->name, | |
3023 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3024 | |
b2220cad JV |
3025 | if (bond->curr_active_slave) { |
3026 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3027 | read_unlock(&bond->curr_slave_lock); | |
3028 | return; | |
3029 | } | |
1da177e4 | 3030 | |
b2220cad | 3031 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3032 | |
b2220cad JV |
3033 | /* if we don't have a curr_active_slave, search for the next available |
3034 | * backup slave from the current_arp_slave and make it the candidate | |
3035 | * for becoming the curr_active_slave | |
3036 | */ | |
1da177e4 | 3037 | |
b2220cad JV |
3038 | if (!bond->current_arp_slave) { |
3039 | bond->current_arp_slave = bond->first_slave; | |
3040 | if (!bond->current_arp_slave) | |
3041 | return; | |
3042 | } | |
1da177e4 | 3043 | |
b2220cad | 3044 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3045 | |
b2220cad JV |
3046 | /* search for next candidate */ |
3047 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3048 | if (IS_UP(slave->dev)) { | |
3049 | slave->link = BOND_LINK_BACK; | |
3050 | bond_set_slave_active_flags(slave); | |
3051 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3052 | slave->jiffies = jiffies; |
b2220cad JV |
3053 | bond->current_arp_slave = slave; |
3054 | break; | |
1da177e4 LT |
3055 | } |
3056 | ||
b2220cad JV |
3057 | /* if the link state is up at this point, we |
3058 | * mark it down - this can happen if we have | |
3059 | * simultaneous link failures and | |
3060 | * reselect_active_interface doesn't make this | |
3061 | * one the current slave so it is still marked | |
3062 | * up when it is actually down | |
1da177e4 | 3063 | */ |
b2220cad JV |
3064 | if (slave->link == BOND_LINK_UP) { |
3065 | slave->link = BOND_LINK_DOWN; | |
3066 | if (slave->link_failure_count < UINT_MAX) | |
3067 | slave->link_failure_count++; | |
1da177e4 | 3068 | |
b2220cad JV |
3069 | bond_set_slave_inactive_flags(slave); |
3070 | ||
a4aee5c8 JP |
3071 | pr_info("%s: backup interface %s is now down.\n", |
3072 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3073 | } |
b2220cad JV |
3074 | } |
3075 | } | |
1da177e4 | 3076 | |
b2220cad JV |
3077 | void bond_activebackup_arp_mon(struct work_struct *work) |
3078 | { | |
3079 | struct bonding *bond = container_of(work, struct bonding, | |
3080 | arp_work.work); | |
3081 | int delta_in_ticks; | |
1da177e4 | 3082 | |
b2220cad | 3083 | read_lock(&bond->lock); |
1da177e4 | 3084 | |
b2220cad JV |
3085 | if (bond->kill_timers) |
3086 | goto out; | |
1da177e4 | 3087 | |
b2220cad | 3088 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3089 | |
b2220cad JV |
3090 | if (bond->slave_cnt == 0) |
3091 | goto re_arm; | |
3092 | ||
b59f9f74 JV |
3093 | if (bond->send_grat_arp) { |
3094 | read_lock(&bond->curr_slave_lock); | |
3095 | bond_send_gratuitous_arp(bond); | |
3096 | read_unlock(&bond->curr_slave_lock); | |
3097 | } | |
3098 | ||
305d552a BH |
3099 | if (bond->send_unsol_na) { |
3100 | read_lock(&bond->curr_slave_lock); | |
3101 | bond_send_unsolicited_na(bond); | |
3102 | read_unlock(&bond->curr_slave_lock); | |
3103 | } | |
3104 | ||
b2220cad JV |
3105 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3106 | read_unlock(&bond->lock); | |
3107 | rtnl_lock(); | |
3108 | read_lock(&bond->lock); | |
3109 | ||
3110 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3111 | ||
3112 | read_unlock(&bond->lock); | |
3113 | rtnl_unlock(); | |
3114 | read_lock(&bond->lock); | |
1da177e4 LT |
3115 | } |
3116 | ||
b2220cad JV |
3117 | bond_ab_arp_probe(bond); |
3118 | ||
1da177e4 | 3119 | re_arm: |
3d632c3f | 3120 | if (bond->params.arp_interval) |
1b76b316 | 3121 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3122 | out: |
3123 | read_unlock(&bond->lock); | |
3124 | } | |
3125 | ||
3126 | /*------------------------------ proc/seq_file-------------------------------*/ | |
3127 | ||
3128 | #ifdef CONFIG_PROC_FS | |
3129 | ||
1da177e4 | 3130 | static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos) |
1f78d9f9 HE |
3131 | __acquires(&dev_base_lock) |
3132 | __acquires(&bond->lock) | |
1da177e4 LT |
3133 | { |
3134 | struct bonding *bond = seq->private; | |
3135 | loff_t off = 0; | |
3136 | struct slave *slave; | |
3137 | int i; | |
3138 | ||
3139 | /* make sure the bond won't be taken away */ | |
3140 | read_lock(&dev_base_lock); | |
6603a6f2 | 3141 | read_lock(&bond->lock); |
1da177e4 | 3142 | |
3d632c3f | 3143 | if (*pos == 0) |
1da177e4 | 3144 | return SEQ_START_TOKEN; |
1da177e4 LT |
3145 | |
3146 | bond_for_each_slave(bond, slave, i) { | |
3d632c3f | 3147 | if (++off == *pos) |
1da177e4 | 3148 | return slave; |
1da177e4 LT |
3149 | } |
3150 | ||
3151 | return NULL; | |
3152 | } | |
3153 | ||
3154 | static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3155 | { | |
3156 | struct bonding *bond = seq->private; | |
3157 | struct slave *slave = v; | |
3158 | ||
3159 | ++*pos; | |
3d632c3f | 3160 | if (v == SEQ_START_TOKEN) |
1da177e4 | 3161 | return bond->first_slave; |
1da177e4 LT |
3162 | |
3163 | slave = slave->next; | |
3164 | ||
3165 | return (slave == bond->first_slave) ? NULL : slave; | |
3166 | } | |
3167 | ||
3168 | static void bond_info_seq_stop(struct seq_file *seq, void *v) | |
1f78d9f9 HE |
3169 | __releases(&bond->lock) |
3170 | __releases(&dev_base_lock) | |
1da177e4 LT |
3171 | { |
3172 | struct bonding *bond = seq->private; | |
3173 | ||
6603a6f2 | 3174 | read_unlock(&bond->lock); |
1da177e4 LT |
3175 | read_unlock(&dev_base_lock); |
3176 | } | |
3177 | ||
3178 | static void bond_info_show_master(struct seq_file *seq) | |
3179 | { | |
3180 | struct bonding *bond = seq->private; | |
3181 | struct slave *curr; | |
4756b02f | 3182 | int i; |
1da177e4 LT |
3183 | |
3184 | read_lock(&bond->curr_slave_lock); | |
3185 | curr = bond->curr_active_slave; | |
3186 | read_unlock(&bond->curr_slave_lock); | |
3187 | ||
dd957c57 | 3188 | seq_printf(seq, "Bonding Mode: %s", |
1da177e4 LT |
3189 | bond_mode_name(bond->params.mode)); |
3190 | ||
dd957c57 JV |
3191 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP && |
3192 | bond->params.fail_over_mac) | |
3915c1e8 JV |
3193 | seq_printf(seq, " (fail_over_mac %s)", |
3194 | fail_over_mac_tbl[bond->params.fail_over_mac].modename); | |
dd957c57 JV |
3195 | |
3196 | seq_printf(seq, "\n"); | |
3197 | ||
c61b75ad MW |
3198 | if (bond->params.mode == BOND_MODE_XOR || |
3199 | bond->params.mode == BOND_MODE_8023AD) { | |
3200 | seq_printf(seq, "Transmit Hash Policy: %s (%d)\n", | |
3201 | xmit_hashtype_tbl[bond->params.xmit_policy].modename, | |
3202 | bond->params.xmit_policy); | |
3203 | } | |
3204 | ||
1da177e4 | 3205 | if (USES_PRIMARY(bond->params.mode)) { |
a549952a | 3206 | seq_printf(seq, "Primary Slave: %s", |
0f418b2a MW |
3207 | (bond->primary_slave) ? |
3208 | bond->primary_slave->dev->name : "None"); | |
a549952a JP |
3209 | if (bond->primary_slave) |
3210 | seq_printf(seq, " (primary_reselect %s)", | |
3211 | pri_reselect_tbl[bond->params.primary_reselect].modename); | |
1da177e4 | 3212 | |
a549952a | 3213 | seq_printf(seq, "\nCurrently Active Slave: %s\n", |
1da177e4 LT |
3214 | (curr) ? curr->dev->name : "None"); |
3215 | } | |
3216 | ||
ff59c456 JV |
3217 | seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ? |
3218 | "up" : "down"); | |
1da177e4 LT |
3219 | seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon); |
3220 | seq_printf(seq, "Up Delay (ms): %d\n", | |
3221 | bond->params.updelay * bond->params.miimon); | |
3222 | seq_printf(seq, "Down Delay (ms): %d\n", | |
3223 | bond->params.downdelay * bond->params.miimon); | |
3224 | ||
4756b02f MW |
3225 | |
3226 | /* ARP information */ | |
3d632c3f SH |
3227 | if (bond->params.arp_interval > 0) { |
3228 | int printed = 0; | |
4756b02f MW |
3229 | seq_printf(seq, "ARP Polling Interval (ms): %d\n", |
3230 | bond->params.arp_interval); | |
3231 | ||
3232 | seq_printf(seq, "ARP IP target/s (n.n.n.n form):"); | |
3233 | ||
3d632c3f | 3234 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
4756b02f | 3235 | if (!bond->params.arp_targets[i]) |
5a31bec0 | 3236 | break; |
4756b02f MW |
3237 | if (printed) |
3238 | seq_printf(seq, ","); | |
8cf14e38 | 3239 | seq_printf(seq, " %pI4", &bond->params.arp_targets[i]); |
4756b02f MW |
3240 | printed = 1; |
3241 | } | |
3242 | seq_printf(seq, "\n"); | |
3243 | } | |
3244 | ||
1da177e4 LT |
3245 | if (bond->params.mode == BOND_MODE_8023AD) { |
3246 | struct ad_info ad_info; | |
3247 | ||
3248 | seq_puts(seq, "\n802.3ad info\n"); | |
3249 | seq_printf(seq, "LACP rate: %s\n", | |
3250 | (bond->params.lacp_fast) ? "fast" : "slow"); | |
fd989c83 JV |
3251 | seq_printf(seq, "Aggregator selection policy (ad_select): %s\n", |
3252 | ad_select_tbl[bond->params.ad_select].modename); | |
1da177e4 LT |
3253 | |
3254 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3255 | seq_printf(seq, "bond %s has no active aggregator\n", | |
3256 | bond->dev->name); | |
3257 | } else { | |
3258 | seq_printf(seq, "Active Aggregator Info:\n"); | |
3259 | ||
3260 | seq_printf(seq, "\tAggregator ID: %d\n", | |
3261 | ad_info.aggregator_id); | |
3262 | seq_printf(seq, "\tNumber of ports: %d\n", | |
3263 | ad_info.ports); | |
3264 | seq_printf(seq, "\tActor Key: %d\n", | |
3265 | ad_info.actor_key); | |
3266 | seq_printf(seq, "\tPartner Key: %d\n", | |
3267 | ad_info.partner_key); | |
e174961c JB |
3268 | seq_printf(seq, "\tPartner Mac Address: %pM\n", |
3269 | ad_info.partner_system); | |
1da177e4 LT |
3270 | } |
3271 | } | |
3272 | } | |
3273 | ||
3d632c3f SH |
3274 | static void bond_info_show_slave(struct seq_file *seq, |
3275 | const struct slave *slave) | |
1da177e4 LT |
3276 | { |
3277 | struct bonding *bond = seq->private; | |
3278 | ||
3279 | seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name); | |
3280 | seq_printf(seq, "MII Status: %s\n", | |
3281 | (slave->link == BOND_LINK_UP) ? "up" : "down"); | |
65509645 | 3282 | seq_printf(seq, "Link Failure Count: %u\n", |
1da177e4 LT |
3283 | slave->link_failure_count); |
3284 | ||
e174961c | 3285 | seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr); |
1da177e4 LT |
3286 | |
3287 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3288 | const struct aggregator *agg | |
3289 | = SLAVE_AD_INFO(slave).port.aggregator; | |
3290 | ||
3d632c3f | 3291 | if (agg) |
1da177e4 LT |
3292 | seq_printf(seq, "Aggregator ID: %d\n", |
3293 | agg->aggregator_identifier); | |
3d632c3f | 3294 | else |
1da177e4 | 3295 | seq_puts(seq, "Aggregator ID: N/A\n"); |
1da177e4 | 3296 | } |
bb1d9123 | 3297 | seq_printf(seq, "Slave queue ID: %d\n", slave->queue_id); |
1da177e4 LT |
3298 | } |
3299 | ||
3300 | static int bond_info_seq_show(struct seq_file *seq, void *v) | |
3301 | { | |
3302 | if (v == SEQ_START_TOKEN) { | |
3303 | seq_printf(seq, "%s\n", version); | |
3304 | bond_info_show_master(seq); | |
3d632c3f | 3305 | } else |
1da177e4 | 3306 | bond_info_show_slave(seq, v); |
1da177e4 LT |
3307 | |
3308 | return 0; | |
3309 | } | |
3310 | ||
4101dec9 | 3311 | static const struct seq_operations bond_info_seq_ops = { |
1da177e4 LT |
3312 | .start = bond_info_seq_start, |
3313 | .next = bond_info_seq_next, | |
3314 | .stop = bond_info_seq_stop, | |
3315 | .show = bond_info_seq_show, | |
3316 | }; | |
3317 | ||
3318 | static int bond_info_open(struct inode *inode, struct file *file) | |
3319 | { | |
3320 | struct seq_file *seq; | |
3321 | struct proc_dir_entry *proc; | |
3322 | int res; | |
3323 | ||
3324 | res = seq_open(file, &bond_info_seq_ops); | |
3325 | if (!res) { | |
3326 | /* recover the pointer buried in proc_dir_entry data */ | |
3327 | seq = file->private_data; | |
3328 | proc = PDE(inode); | |
3329 | seq->private = proc->data; | |
3330 | } | |
3331 | ||
3332 | return res; | |
3333 | } | |
3334 | ||
d54b1fdb | 3335 | static const struct file_operations bond_info_fops = { |
1da177e4 LT |
3336 | .owner = THIS_MODULE, |
3337 | .open = bond_info_open, | |
3338 | .read = seq_read, | |
3339 | .llseek = seq_lseek, | |
3340 | .release = seq_release, | |
3341 | }; | |
3342 | ||
38fc0026 | 3343 | static void bond_create_proc_entry(struct bonding *bond) |
1da177e4 LT |
3344 | { |
3345 | struct net_device *bond_dev = bond->dev; | |
ec87fd3b | 3346 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
1da177e4 | 3347 | |
ec87fd3b | 3348 | if (bn->proc_dir) { |
a95609cb | 3349 | bond->proc_entry = proc_create_data(bond_dev->name, |
ec87fd3b | 3350 | S_IRUGO, bn->proc_dir, |
a95609cb | 3351 | &bond_info_fops, bond); |
3d632c3f | 3352 | if (bond->proc_entry == NULL) |
a4aee5c8 JP |
3353 | pr_warning("Warning: Cannot create /proc/net/%s/%s\n", |
3354 | DRV_NAME, bond_dev->name); | |
3d632c3f | 3355 | else |
1da177e4 | 3356 | memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ); |
1da177e4 | 3357 | } |
1da177e4 LT |
3358 | } |
3359 | ||
3360 | static void bond_remove_proc_entry(struct bonding *bond) | |
3361 | { | |
ec87fd3b EB |
3362 | struct net_device *bond_dev = bond->dev; |
3363 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); | |
3364 | ||
3365 | if (bn->proc_dir && bond->proc_entry) { | |
3366 | remove_proc_entry(bond->proc_file_name, bn->proc_dir); | |
1da177e4 LT |
3367 | memset(bond->proc_file_name, 0, IFNAMSIZ); |
3368 | bond->proc_entry = NULL; | |
3369 | } | |
3370 | } | |
3371 | ||
3372 | /* Create the bonding directory under /proc/net, if doesn't exist yet. | |
3373 | * Caller must hold rtnl_lock. | |
3374 | */ | |
2c8c1e72 | 3375 | static void __net_init bond_create_proc_dir(struct bond_net *bn) |
1da177e4 | 3376 | { |
ec87fd3b EB |
3377 | if (!bn->proc_dir) { |
3378 | bn->proc_dir = proc_mkdir(DRV_NAME, bn->net->proc_net); | |
3379 | if (!bn->proc_dir) | |
a4aee5c8 JP |
3380 | pr_warning("Warning: cannot create /proc/net/%s\n", |
3381 | DRV_NAME); | |
1da177e4 LT |
3382 | } |
3383 | } | |
3384 | ||
3385 | /* Destroy the bonding directory under /proc/net, if empty. | |
3386 | * Caller must hold rtnl_lock. | |
3387 | */ | |
2c8c1e72 | 3388 | static void __net_exit bond_destroy_proc_dir(struct bond_net *bn) |
1da177e4 | 3389 | { |
ec87fd3b EB |
3390 | if (bn->proc_dir) { |
3391 | remove_proc_entry(DRV_NAME, bn->net->proc_net); | |
3392 | bn->proc_dir = NULL; | |
1da177e4 LT |
3393 | } |
3394 | } | |
aee64faf JP |
3395 | |
3396 | #else /* !CONFIG_PROC_FS */ | |
3397 | ||
38fc0026 | 3398 | static void bond_create_proc_entry(struct bonding *bond) |
aee64faf JP |
3399 | { |
3400 | } | |
3401 | ||
3402 | static void bond_remove_proc_entry(struct bonding *bond) | |
3403 | { | |
3404 | } | |
3405 | ||
2c8c1e72 | 3406 | static inline void bond_create_proc_dir(struct bond_net *bn) |
aee64faf JP |
3407 | { |
3408 | } | |
3409 | ||
2c8c1e72 | 3410 | static inline void bond_destroy_proc_dir(struct bond_net *bn) |
aee64faf JP |
3411 | { |
3412 | } | |
3413 | ||
1da177e4 LT |
3414 | #endif /* CONFIG_PROC_FS */ |
3415 | ||
aee64faf | 3416 | |
1da177e4 LT |
3417 | /*-------------------------- netdev event handling --------------------------*/ |
3418 | ||
3419 | /* | |
3420 | * Change device name | |
3421 | */ | |
3422 | static int bond_event_changename(struct bonding *bond) | |
3423 | { | |
1da177e4 LT |
3424 | bond_remove_proc_entry(bond); |
3425 | bond_create_proc_entry(bond); | |
7e083840 | 3426 | |
1da177e4 LT |
3427 | return NOTIFY_DONE; |
3428 | } | |
3429 | ||
3d632c3f SH |
3430 | static int bond_master_netdev_event(unsigned long event, |
3431 | struct net_device *bond_dev) | |
1da177e4 | 3432 | { |
454d7c9b | 3433 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3434 | |
3435 | switch (event) { | |
3436 | case NETDEV_CHANGENAME: | |
3437 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3438 | default: |
3439 | break; | |
3440 | } | |
3441 | ||
3442 | return NOTIFY_DONE; | |
3443 | } | |
3444 | ||
3d632c3f SH |
3445 | static int bond_slave_netdev_event(unsigned long event, |
3446 | struct net_device *slave_dev) | |
1da177e4 LT |
3447 | { |
3448 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3449 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3450 | |
3451 | switch (event) { | |
3452 | case NETDEV_UNREGISTER: | |
3453 | if (bond_dev) { | |
1284cd3a JV |
3454 | if (bond->setup_by_slave) |
3455 | bond_release_and_destroy(bond_dev, slave_dev); | |
3456 | else | |
3457 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3458 | } |
3459 | break; | |
3460 | case NETDEV_CHANGE: | |
17d04500 JV |
3461 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3462 | struct slave *slave; | |
3463 | ||
3464 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3465 | if (slave) { | |
3466 | u16 old_speed = slave->speed; | |
3467 | u16 old_duplex = slave->duplex; | |
3468 | ||
3469 | bond_update_speed_duplex(slave); | |
3470 | ||
3471 | if (bond_is_lb(bond)) | |
3472 | break; | |
3473 | ||
3474 | if (old_speed != slave->speed) | |
3475 | bond_3ad_adapter_speed_changed(slave); | |
3476 | if (old_duplex != slave->duplex) | |
3477 | bond_3ad_adapter_duplex_changed(slave); | |
3478 | } | |
3479 | } | |
3480 | ||
1da177e4 LT |
3481 | break; |
3482 | case NETDEV_DOWN: | |
3483 | /* | |
3484 | * ... Or is it this? | |
3485 | */ | |
3486 | break; | |
3487 | case NETDEV_CHANGEMTU: | |
3488 | /* | |
3489 | * TODO: Should slaves be allowed to | |
3490 | * independently alter their MTU? For | |
3491 | * an active-backup bond, slaves need | |
3492 | * not be the same type of device, so | |
3493 | * MTUs may vary. For other modes, | |
3494 | * slaves arguably should have the | |
3495 | * same MTUs. To do this, we'd need to | |
3496 | * take over the slave's change_mtu | |
3497 | * function for the duration of their | |
3498 | * servitude. | |
3499 | */ | |
3500 | break; | |
3501 | case NETDEV_CHANGENAME: | |
3502 | /* | |
3503 | * TODO: handle changing the primary's name | |
3504 | */ | |
3505 | break; | |
8531c5ff AK |
3506 | case NETDEV_FEAT_CHANGE: |
3507 | bond_compute_features(bond); | |
3508 | break; | |
1da177e4 LT |
3509 | default: |
3510 | break; | |
3511 | } | |
3512 | ||
3513 | return NOTIFY_DONE; | |
3514 | } | |
3515 | ||
3516 | /* | |
3517 | * bond_netdev_event: handle netdev notifier chain events. | |
3518 | * | |
3519 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3520 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3521 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3522 | * dev_probe_lock). | |
3523 | */ | |
3d632c3f SH |
3524 | static int bond_netdev_event(struct notifier_block *this, |
3525 | unsigned long event, void *ptr) | |
1da177e4 LT |
3526 | { |
3527 | struct net_device *event_dev = (struct net_device *)ptr; | |
3528 | ||
5a03cdb7 | 3529 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3530 | event_dev ? event_dev->name : "None", |
3531 | event); | |
1da177e4 | 3532 | |
0b680e75 JV |
3533 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3534 | return NOTIFY_DONE; | |
3535 | ||
1da177e4 | 3536 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3537 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3538 | return bond_master_netdev_event(event, event_dev); |
3539 | } | |
3540 | ||
3541 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3542 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3543 | return bond_slave_netdev_event(event, event_dev); |
3544 | } | |
3545 | ||
3546 | return NOTIFY_DONE; | |
3547 | } | |
3548 | ||
c3ade5ca JV |
3549 | /* |
3550 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3551 | * | |
3552 | * We keep track of device IPs primarily to use as source addresses in | |
3553 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3554 | * | |
3555 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3556 | */ | |
3557 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3558 | { | |
3559 | struct in_ifaddr *ifa = ptr; | |
3560 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
ec87fd3b | 3561 | struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id); |
0883beca PE |
3562 | struct bonding *bond; |
3563 | struct vlan_entry *vlan; | |
c3ade5ca | 3564 | |
ec87fd3b | 3565 | list_for_each_entry(bond, &bn->dev_list, bond_list) { |
c3ade5ca JV |
3566 | if (bond->dev == event_dev) { |
3567 | switch (event) { | |
3568 | case NETDEV_UP: | |
3569 | bond->master_ip = ifa->ifa_local; | |
3570 | return NOTIFY_OK; | |
3571 | case NETDEV_DOWN: | |
3572 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3573 | return NOTIFY_OK; | |
3574 | default: | |
3575 | return NOTIFY_DONE; | |
3576 | } | |
3577 | } | |
3578 | ||
0883beca | 3579 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 3580 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3581 | if (vlan_dev == event_dev) { |
3582 | switch (event) { | |
3583 | case NETDEV_UP: | |
3584 | vlan->vlan_ip = ifa->ifa_local; | |
3585 | return NOTIFY_OK; | |
3586 | case NETDEV_DOWN: | |
3587 | vlan->vlan_ip = | |
3588 | bond_glean_dev_ip(vlan_dev); | |
3589 | return NOTIFY_OK; | |
3590 | default: | |
3591 | return NOTIFY_DONE; | |
3592 | } | |
3593 | } | |
3594 | } | |
3595 | } | |
3596 | return NOTIFY_DONE; | |
3597 | } | |
3598 | ||
1da177e4 LT |
3599 | static struct notifier_block bond_netdev_notifier = { |
3600 | .notifier_call = bond_netdev_event, | |
3601 | }; | |
3602 | ||
c3ade5ca JV |
3603 | static struct notifier_block bond_inetaddr_notifier = { |
3604 | .notifier_call = bond_inetaddr_event, | |
3605 | }; | |
3606 | ||
1da177e4 LT |
3607 | /*-------------------------- Packet type handling ---------------------------*/ |
3608 | ||
3609 | /* register to receive lacpdus on a bond */ | |
3610 | static void bond_register_lacpdu(struct bonding *bond) | |
3611 | { | |
3612 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3613 | ||
3614 | /* initialize packet type */ | |
3615 | pk_type->type = PKT_TYPE_LACPDU; | |
3616 | pk_type->dev = bond->dev; | |
3617 | pk_type->func = bond_3ad_lacpdu_recv; | |
3618 | ||
3619 | dev_add_pack(pk_type); | |
3620 | } | |
3621 | ||
3622 | /* unregister to receive lacpdus on a bond */ | |
3623 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3624 | { | |
3625 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3626 | } | |
3627 | ||
f5b2b966 JV |
3628 | void bond_register_arp(struct bonding *bond) |
3629 | { | |
3630 | struct packet_type *pt = &bond->arp_mon_pt; | |
3631 | ||
c4f283b1 JV |
3632 | if (pt->type) |
3633 | return; | |
3634 | ||
f5b2b966 | 3635 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3636 | pt->dev = bond->dev; |
f5b2b966 JV |
3637 | pt->func = bond_arp_rcv; |
3638 | dev_add_pack(pt); | |
3639 | } | |
3640 | ||
3641 | void bond_unregister_arp(struct bonding *bond) | |
3642 | { | |
c4f283b1 JV |
3643 | struct packet_type *pt = &bond->arp_mon_pt; |
3644 | ||
3645 | dev_remove_pack(pt); | |
3646 | pt->type = 0; | |
f5b2b966 JV |
3647 | } |
3648 | ||
169a3e66 JV |
3649 | /*---------------------------- Hashing Policies -----------------------------*/ |
3650 | ||
6f6652be JV |
3651 | /* |
3652 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3653 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3654 | */ | |
a361c83c | 3655 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3656 | { |
3657 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3658 | struct iphdr *iph = ip_hdr(skb); | |
3659 | ||
f14c4e4e | 3660 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3661 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3662 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3663 | } |
3664 | ||
d3da6831 | 3665 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3666 | } |
3667 | ||
169a3e66 | 3668 | /* |
59c51591 | 3669 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3670 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3671 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3672 | */ | |
a361c83c | 3673 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3674 | { |
3675 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3676 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3677 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3678 | int layer4_xor = 0; |
3679 | ||
f14c4e4e BH |
3680 | if (skb->protocol == htons(ETH_P_IP)) { |
3681 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3682 | (iph->protocol == IPPROTO_TCP || |
3683 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3684 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3685 | } |
3686 | return (layer4_xor ^ | |
3687 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3688 | ||
3689 | } | |
3690 | ||
d3da6831 | 3691 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3692 | } |
3693 | ||
3694 | /* | |
3695 | * Hash for the output device based upon layer 2 data | |
3696 | */ | |
a361c83c | 3697 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3698 | { |
3699 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3700 | ||
d3da6831 | 3701 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3702 | } |
3703 | ||
1da177e4 LT |
3704 | /*-------------------------- Device entry points ----------------------------*/ |
3705 | ||
3706 | static int bond_open(struct net_device *bond_dev) | |
3707 | { | |
454d7c9b | 3708 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3709 | |
3710 | bond->kill_timers = 0; | |
3711 | ||
58402054 | 3712 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3713 | /* bond_alb_initialize must be called before the timer |
3714 | * is started. | |
3715 | */ | |
3716 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3717 | /* something went wrong - fail the open operation */ | |
b473946a | 3718 | return -ENOMEM; |
1da177e4 LT |
3719 | } |
3720 | ||
1b76b316 JV |
3721 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3722 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3723 | } |
3724 | ||
3725 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3726 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3727 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3728 | } |
3729 | ||
3730 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3731 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3732 | INIT_DELAYED_WORK(&bond->arp_work, | |
3733 | bond_activebackup_arp_mon); | |
3734 | else | |
3735 | INIT_DELAYED_WORK(&bond->arp_work, | |
3736 | bond_loadbalance_arp_mon); | |
3737 | ||
3738 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3739 | if (bond->params.arp_validate) |
3740 | bond_register_arp(bond); | |
1da177e4 LT |
3741 | } |
3742 | ||
3743 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3744 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3745 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3746 | /* register to receive LACPDUs */ |
3747 | bond_register_lacpdu(bond); | |
fd989c83 | 3748 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3749 | } |
3750 | ||
3751 | return 0; | |
3752 | } | |
3753 | ||
3754 | static int bond_close(struct net_device *bond_dev) | |
3755 | { | |
454d7c9b | 3756 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3757 | |
3758 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3759 | /* Unregister the receive of LACPDUs */ | |
3760 | bond_unregister_lacpdu(bond); | |
3761 | } | |
3762 | ||
f5b2b966 JV |
3763 | if (bond->params.arp_validate) |
3764 | bond_unregister_arp(bond); | |
3765 | ||
1da177e4 LT |
3766 | write_lock_bh(&bond->lock); |
3767 | ||
b59f9f74 | 3768 | bond->send_grat_arp = 0; |
305d552a | 3769 | bond->send_unsol_na = 0; |
1da177e4 LT |
3770 | |
3771 | /* signal timers not to re-arm */ | |
3772 | bond->kill_timers = 1; | |
3773 | ||
3774 | write_unlock_bh(&bond->lock); | |
3775 | ||
1da177e4 | 3776 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3777 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3778 | } |
3779 | ||
3780 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3781 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3782 | } |
3783 | ||
3784 | switch (bond->params.mode) { | |
3785 | case BOND_MODE_8023AD: | |
1b76b316 | 3786 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3787 | break; |
3788 | case BOND_MODE_TLB: | |
3789 | case BOND_MODE_ALB: | |
1b76b316 | 3790 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3791 | break; |
3792 | default: | |
3793 | break; | |
3794 | } | |
3795 | ||
1da177e4 | 3796 | |
58402054 | 3797 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3798 | /* Must be called only after all |
3799 | * slaves have been released | |
3800 | */ | |
3801 | bond_alb_deinitialize(bond); | |
3802 | } | |
3803 | ||
3804 | return 0; | |
3805 | } | |
3806 | ||
be1f3c2c | 3807 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev) |
1da177e4 | 3808 | { |
454d7c9b | 3809 | struct bonding *bond = netdev_priv(bond_dev); |
be1f3c2c BH |
3810 | struct rtnl_link_stats64 *stats = &bond_dev->stats64; |
3811 | struct rtnl_link_stats64 local_stats; | |
1da177e4 LT |
3812 | struct slave *slave; |
3813 | int i; | |
3814 | ||
be1f3c2c | 3815 | memset(&local_stats, 0, sizeof(local_stats)); |
1da177e4 LT |
3816 | |
3817 | read_lock_bh(&bond->lock); | |
3818 | ||
3819 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c BH |
3820 | const struct rtnl_link_stats64 *sstats = |
3821 | dev_get_stats(slave->dev); | |
eeda3fd6 | 3822 | |
2439f9eb AG |
3823 | local_stats.rx_packets += sstats->rx_packets; |
3824 | local_stats.rx_bytes += sstats->rx_bytes; | |
3825 | local_stats.rx_errors += sstats->rx_errors; | |
3826 | local_stats.rx_dropped += sstats->rx_dropped; | |
3827 | ||
3828 | local_stats.tx_packets += sstats->tx_packets; | |
3829 | local_stats.tx_bytes += sstats->tx_bytes; | |
3830 | local_stats.tx_errors += sstats->tx_errors; | |
3831 | local_stats.tx_dropped += sstats->tx_dropped; | |
3832 | ||
3833 | local_stats.multicast += sstats->multicast; | |
3834 | local_stats.collisions += sstats->collisions; | |
3835 | ||
3836 | local_stats.rx_length_errors += sstats->rx_length_errors; | |
3837 | local_stats.rx_over_errors += sstats->rx_over_errors; | |
3838 | local_stats.rx_crc_errors += sstats->rx_crc_errors; | |
3839 | local_stats.rx_frame_errors += sstats->rx_frame_errors; | |
3840 | local_stats.rx_fifo_errors += sstats->rx_fifo_errors; | |
3841 | local_stats.rx_missed_errors += sstats->rx_missed_errors; | |
3842 | ||
3843 | local_stats.tx_aborted_errors += sstats->tx_aborted_errors; | |
3844 | local_stats.tx_carrier_errors += sstats->tx_carrier_errors; | |
3845 | local_stats.tx_fifo_errors += sstats->tx_fifo_errors; | |
3846 | local_stats.tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3847 | local_stats.tx_window_errors += sstats->tx_window_errors; | |
3848 | } | |
3849 | ||
3850 | memcpy(stats, &local_stats, sizeof(struct net_device_stats)); | |
1da177e4 LT |
3851 | |
3852 | read_unlock_bh(&bond->lock); | |
3853 | ||
3854 | return stats; | |
3855 | } | |
3856 | ||
3857 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3858 | { | |
3859 | struct net_device *slave_dev = NULL; | |
3860 | struct ifbond k_binfo; | |
3861 | struct ifbond __user *u_binfo = NULL; | |
3862 | struct ifslave k_sinfo; | |
3863 | struct ifslave __user *u_sinfo = NULL; | |
3864 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3865 | int res = 0; |
3866 | ||
a4aee5c8 | 3867 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3868 | |
3869 | switch (cmd) { | |
1da177e4 LT |
3870 | case SIOCGMIIPHY: |
3871 | mii = if_mii(ifr); | |
3d632c3f | 3872 | if (!mii) |
1da177e4 | 3873 | return -EINVAL; |
3d632c3f | 3874 | |
1da177e4 LT |
3875 | mii->phy_id = 0; |
3876 | /* Fall Through */ | |
3877 | case SIOCGMIIREG: | |
3878 | /* | |
3879 | * We do this again just in case we were called by SIOCGMIIREG | |
3880 | * instead of SIOCGMIIPHY. | |
3881 | */ | |
3882 | mii = if_mii(ifr); | |
3d632c3f | 3883 | if (!mii) |
1da177e4 | 3884 | return -EINVAL; |
3d632c3f | 3885 | |
1da177e4 LT |
3886 | |
3887 | if (mii->reg_num == 1) { | |
454d7c9b | 3888 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3889 | mii->val_out = 0; |
6603a6f2 | 3890 | read_lock(&bond->lock); |
1da177e4 | 3891 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3892 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3893 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3894 | |
1da177e4 | 3895 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3896 | read_unlock(&bond->lock); |
1da177e4 LT |
3897 | } |
3898 | ||
3899 | return 0; | |
3900 | case BOND_INFO_QUERY_OLD: | |
3901 | case SIOCBONDINFOQUERY: | |
3902 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3903 | ||
3d632c3f | 3904 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3905 | return -EFAULT; |
1da177e4 LT |
3906 | |
3907 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3908 | if (res == 0 && |
3909 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3910 | return -EFAULT; | |
1da177e4 LT |
3911 | |
3912 | return res; | |
3913 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3914 | case SIOCBONDSLAVEINFOQUERY: | |
3915 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3916 | ||
3d632c3f | 3917 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3918 | return -EFAULT; |
1da177e4 LT |
3919 | |
3920 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3921 | if (res == 0 && |
3922 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3923 | return -EFAULT; | |
1da177e4 LT |
3924 | |
3925 | return res; | |
3926 | default: | |
3927 | /* Go on */ | |
3928 | break; | |
3929 | } | |
3930 | ||
3d632c3f | 3931 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3932 | return -EPERM; |
1da177e4 | 3933 | |
ec87fd3b | 3934 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3935 | |
a4aee5c8 | 3936 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3937 | |
3d632c3f | 3938 | if (!slave_dev) |
1da177e4 | 3939 | res = -ENODEV; |
3d632c3f | 3940 | else { |
a4aee5c8 | 3941 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3942 | switch (cmd) { |
3943 | case BOND_ENSLAVE_OLD: | |
3944 | case SIOCBONDENSLAVE: | |
3945 | res = bond_enslave(bond_dev, slave_dev); | |
3946 | break; | |
3947 | case BOND_RELEASE_OLD: | |
3948 | case SIOCBONDRELEASE: | |
3949 | res = bond_release(bond_dev, slave_dev); | |
3950 | break; | |
3951 | case BOND_SETHWADDR_OLD: | |
3952 | case SIOCBONDSETHWADDR: | |
3953 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3954 | break; | |
3955 | case BOND_CHANGE_ACTIVE_OLD: | |
3956 | case SIOCBONDCHANGEACTIVE: | |
3957 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3958 | break; | |
3959 | default: | |
3960 | res = -EOPNOTSUPP; | |
3961 | } | |
3962 | ||
3963 | dev_put(slave_dev); | |
3964 | } | |
3965 | ||
1da177e4 LT |
3966 | return res; |
3967 | } | |
3968 | ||
22bedad3 JP |
3969 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3970 | struct netdev_hw_addr_list *list, | |
3971 | int addrlen) | |
3972 | { | |
3973 | struct netdev_hw_addr *ha; | |
3974 | ||
3975 | netdev_hw_addr_list_for_each(ha, list) | |
3976 | if (!memcmp(ha->addr, addr, addrlen)) | |
3977 | return true; | |
3978 | ||
3979 | return false; | |
3980 | } | |
3981 | ||
1da177e4 LT |
3982 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3983 | { | |
454d7c9b | 3984 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 JP |
3985 | struct netdev_hw_addr *ha; |
3986 | bool found; | |
1da177e4 | 3987 | |
1da177e4 LT |
3988 | /* |
3989 | * Do promisc before checking multicast_mode | |
3990 | */ | |
3d632c3f | 3991 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
3992 | /* |
3993 | * FIXME: Need to handle the error when one of the multi-slaves | |
3994 | * encounters error. | |
3995 | */ | |
1da177e4 | 3996 | bond_set_promiscuity(bond, 1); |
1da177e4 | 3997 | |
3d632c3f SH |
3998 | |
3999 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 4000 | bond_set_promiscuity(bond, -1); |
3d632c3f | 4001 | |
1da177e4 LT |
4002 | |
4003 | /* set allmulti flag to slaves */ | |
3d632c3f | 4004 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
4005 | /* |
4006 | * FIXME: Need to handle the error when one of the multi-slaves | |
4007 | * encounters error. | |
4008 | */ | |
1da177e4 | 4009 | bond_set_allmulti(bond, 1); |
1da177e4 | 4010 | |
3d632c3f SH |
4011 | |
4012 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 4013 | bond_set_allmulti(bond, -1); |
3d632c3f | 4014 | |
1da177e4 | 4015 | |
80ee5ad2 JV |
4016 | read_lock(&bond->lock); |
4017 | ||
1da177e4 LT |
4018 | bond->flags = bond_dev->flags; |
4019 | ||
4020 | /* looking for addresses to add to slaves' mc list */ | |
22bedad3 JP |
4021 | netdev_for_each_mc_addr(ha, bond_dev) { |
4022 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
4023 | bond_dev->addr_len); | |
4024 | if (!found) | |
4025 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
4026 | } |
4027 | ||
4028 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
4029 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
4030 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
4031 | bond_dev->addr_len); | |
4032 | if (!found) | |
4033 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
4034 | } |
4035 | ||
4036 | /* save master's multicast list */ | |
22bedad3 JP |
4037 | __hw_addr_flush(&bond->mc_list); |
4038 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
4039 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 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, |
a4aee5c8 | 4069 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
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) { | |
a4aee5c8 JP |
4087 | pr_debug("s %p s->p %p c_m %p\n", |
4088 | slave, | |
4089 | slave->prev, | |
4090 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 4091 | |
1da177e4 LT |
4092 | res = dev_set_mtu(slave->dev, new_mtu); |
4093 | ||
4094 | if (res) { | |
4095 | /* If we failed to set the slave's mtu to the new value | |
4096 | * we must abort the operation even in ACTIVE_BACKUP | |
4097 | * mode, because if we allow the backup slaves to have | |
4098 | * different mtu values than the active slave we'll | |
4099 | * need to change their mtu when doing a failover. That | |
4100 | * means changing their mtu from timer context, which | |
4101 | * is probably not a good idea. | |
4102 | */ | |
5a03cdb7 | 4103 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4104 | goto unwind; |
4105 | } | |
4106 | } | |
4107 | ||
4108 | bond_dev->mtu = new_mtu; | |
4109 | ||
4110 | return 0; | |
4111 | ||
4112 | unwind: | |
4113 | /* unwind from head to the slave that failed */ | |
4114 | stop_at = slave; | |
4115 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4116 | int tmp_res; | |
4117 | ||
4118 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
4119 | if (tmp_res) { | |
a4aee5c8 JP |
4120 | pr_debug("unwind err %d dev %s\n", |
4121 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4122 | } |
4123 | } | |
4124 | ||
4125 | return res; | |
4126 | } | |
4127 | ||
4128 | /* | |
4129 | * Change HW address | |
4130 | * | |
4131 | * Note that many devices must be down to change the HW address, and | |
4132 | * downing the master releases all slaves. We can make bonds full of | |
4133 | * bonding devices to test this, however. | |
4134 | */ | |
4135 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4136 | { | |
454d7c9b | 4137 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4138 | struct sockaddr *sa = addr, tmp_sa; |
4139 | struct slave *slave, *stop_at; | |
4140 | int res = 0; | |
4141 | int i; | |
4142 | ||
eb7cc59a SH |
4143 | if (bond->params.mode == BOND_MODE_ALB) |
4144 | return bond_alb_set_mac_address(bond_dev, addr); | |
4145 | ||
4146 | ||
a4aee5c8 JP |
4147 | pr_debug("bond=%p, name=%s\n", |
4148 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 4149 | |
dd957c57 | 4150 | /* |
3915c1e8 JV |
4151 | * If fail_over_mac is set to active, do nothing and return |
4152 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 4153 | */ |
3915c1e8 | 4154 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 4155 | return 0; |
2ab82852 | 4156 | |
3d632c3f | 4157 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 4158 | return -EADDRNOTAVAIL; |
1da177e4 LT |
4159 | |
4160 | /* Can't hold bond->lock with bh disabled here since | |
4161 | * some base drivers panic. On the other hand we can't | |
4162 | * hold bond->lock without bh disabled because we'll | |
4163 | * deadlock. The only solution is to rely on the fact | |
4164 | * that we're under rtnl_lock here, and the slaves | |
4165 | * list won't change. This doesn't solve the problem | |
4166 | * of setting the slave's hw address while it is | |
4167 | * transmitting, but the assumption is that the base | |
4168 | * driver can handle that. | |
4169 | * | |
4170 | * TODO: figure out a way to safely iterate the slaves | |
4171 | * list, but without holding a lock around the actual | |
4172 | * call to the base driver. | |
4173 | */ | |
4174 | ||
4175 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 4176 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 4177 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 4178 | |
eb7cc59a | 4179 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 4180 | res = -EOPNOTSUPP; |
5a03cdb7 | 4181 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
4182 | goto unwind; |
4183 | } | |
4184 | ||
4185 | res = dev_set_mac_address(slave->dev, addr); | |
4186 | if (res) { | |
4187 | /* TODO: consider downing the slave | |
4188 | * and retry ? | |
4189 | * User should expect communications | |
4190 | * breakage anyway until ARP finish | |
4191 | * updating, so... | |
4192 | */ | |
5a03cdb7 | 4193 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4194 | goto unwind; |
4195 | } | |
4196 | } | |
4197 | ||
4198 | /* success */ | |
4199 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4200 | return 0; | |
4201 | ||
4202 | unwind: | |
4203 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4204 | tmp_sa.sa_family = bond_dev->type; | |
4205 | ||
4206 | /* unwind from head to the slave that failed */ | |
4207 | stop_at = slave; | |
4208 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4209 | int tmp_res; | |
4210 | ||
4211 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4212 | if (tmp_res) { | |
a4aee5c8 JP |
4213 | pr_debug("unwind err %d dev %s\n", |
4214 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4215 | } |
4216 | } | |
4217 | ||
4218 | return res; | |
4219 | } | |
4220 | ||
4221 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4222 | { | |
454d7c9b | 4223 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4224 | struct slave *slave, *start_at; |
cf5f9044 | 4225 | int i, slave_no, res = 1; |
a2fd940f | 4226 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
4227 | |
4228 | read_lock(&bond->lock); | |
4229 | ||
3d632c3f | 4230 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4231 | goto out; |
cf5f9044 | 4232 | /* |
a2fd940f AG |
4233 | * Start with the curr_active_slave that joined the bond as the |
4234 | * default for sending IGMP traffic. For failover purposes one | |
4235 | * needs to maintain some consistency for the interface that will | |
4236 | * send the join/membership reports. The curr_active_slave found | |
4237 | * will send all of this type of traffic. | |
cf5f9044 | 4238 | */ |
00ae7028 | 4239 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4240 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4241 | |
a2fd940f AG |
4242 | read_lock(&bond->curr_slave_lock); |
4243 | slave = bond->curr_active_slave; | |
4244 | read_unlock(&bond->curr_slave_lock); | |
4245 | ||
4246 | if (!slave) | |
4247 | goto out; | |
4248 | } else { | |
4249 | /* | |
4250 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4251 | * that as being rare enough not to justify using an | |
4252 | * atomic op here. | |
4253 | */ | |
4254 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4255 | ||
4256 | bond_for_each_slave(bond, slave, i) { | |
4257 | slave_no--; | |
4258 | if (slave_no < 0) | |
4259 | break; | |
4260 | } | |
1da177e4 LT |
4261 | } |
4262 | ||
cf5f9044 | 4263 | start_at = slave; |
1da177e4 LT |
4264 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4265 | if (IS_UP(slave->dev) && | |
4266 | (slave->link == BOND_LINK_UP) && | |
4267 | (slave->state == BOND_STATE_ACTIVE)) { | |
4268 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
1da177e4 LT |
4269 | break; |
4270 | } | |
4271 | } | |
4272 | ||
1da177e4 LT |
4273 | out: |
4274 | if (res) { | |
4275 | /* no suitable interface, frame not sent */ | |
4276 | dev_kfree_skb(skb); | |
4277 | } | |
4278 | read_unlock(&bond->lock); | |
ec634fe3 | 4279 | return NETDEV_TX_OK; |
1da177e4 LT |
4280 | } |
4281 | ||
075897ce | 4282 | |
1da177e4 LT |
4283 | /* |
4284 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4285 | * the bond has a usable interface. | |
4286 | */ | |
4287 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4288 | { | |
454d7c9b | 4289 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4290 | int res = 1; |
4291 | ||
1da177e4 LT |
4292 | read_lock(&bond->lock); |
4293 | read_lock(&bond->curr_slave_lock); | |
4294 | ||
3d632c3f | 4295 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4296 | goto out; |
1da177e4 | 4297 | |
075897ce JL |
4298 | if (!bond->curr_active_slave) |
4299 | goto out; | |
4300 | ||
075897ce JL |
4301 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4302 | ||
1da177e4 | 4303 | out: |
3d632c3f | 4304 | if (res) |
1da177e4 LT |
4305 | /* no suitable interface, frame not sent */ |
4306 | dev_kfree_skb(skb); | |
3d632c3f | 4307 | |
1da177e4 LT |
4308 | read_unlock(&bond->curr_slave_lock); |
4309 | read_unlock(&bond->lock); | |
ec634fe3 | 4310 | return NETDEV_TX_OK; |
1da177e4 LT |
4311 | } |
4312 | ||
4313 | /* | |
169a3e66 JV |
4314 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4315 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4316 | * find the next active slave. | |
1da177e4 LT |
4317 | */ |
4318 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4319 | { | |
454d7c9b | 4320 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4321 | struct slave *slave, *start_at; |
4322 | int slave_no; | |
4323 | int i; | |
4324 | int res = 1; | |
4325 | ||
4326 | read_lock(&bond->lock); | |
4327 | ||
3d632c3f | 4328 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4329 | goto out; |
1da177e4 | 4330 | |
a361c83c | 4331 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4332 | |
4333 | bond_for_each_slave(bond, slave, i) { | |
4334 | slave_no--; | |
3d632c3f | 4335 | if (slave_no < 0) |
1da177e4 | 4336 | break; |
1da177e4 LT |
4337 | } |
4338 | ||
4339 | start_at = slave; | |
4340 | ||
4341 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4342 | if (IS_UP(slave->dev) && | |
4343 | (slave->link == BOND_LINK_UP) && | |
4344 | (slave->state == BOND_STATE_ACTIVE)) { | |
4345 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4346 | break; | |
4347 | } | |
4348 | } | |
4349 | ||
4350 | out: | |
4351 | if (res) { | |
4352 | /* no suitable interface, frame not sent */ | |
4353 | dev_kfree_skb(skb); | |
4354 | } | |
4355 | read_unlock(&bond->lock); | |
ec634fe3 | 4356 | return NETDEV_TX_OK; |
1da177e4 LT |
4357 | } |
4358 | ||
4359 | /* | |
4360 | * in broadcast mode, we send everything to all usable interfaces. | |
4361 | */ | |
4362 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4363 | { | |
454d7c9b | 4364 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4365 | struct slave *slave, *start_at; |
4366 | struct net_device *tx_dev = NULL; | |
4367 | int i; | |
4368 | int res = 1; | |
4369 | ||
4370 | read_lock(&bond->lock); | |
4371 | ||
3d632c3f | 4372 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4373 | goto out; |
1da177e4 LT |
4374 | |
4375 | read_lock(&bond->curr_slave_lock); | |
4376 | start_at = bond->curr_active_slave; | |
4377 | read_unlock(&bond->curr_slave_lock); | |
4378 | ||
3d632c3f | 4379 | if (!start_at) |
1da177e4 | 4380 | goto out; |
1da177e4 LT |
4381 | |
4382 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4383 | if (IS_UP(slave->dev) && | |
4384 | (slave->link == BOND_LINK_UP) && | |
4385 | (slave->state == BOND_STATE_ACTIVE)) { | |
4386 | if (tx_dev) { | |
4387 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4388 | if (!skb2) { | |
a4aee5c8 | 4389 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4390 | bond_dev->name); |
1da177e4 LT |
4391 | continue; |
4392 | } | |
4393 | ||
4394 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4395 | if (res) { | |
4396 | dev_kfree_skb(skb2); | |
4397 | continue; | |
4398 | } | |
4399 | } | |
4400 | tx_dev = slave->dev; | |
4401 | } | |
4402 | } | |
4403 | ||
3d632c3f | 4404 | if (tx_dev) |
1da177e4 | 4405 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4406 | |
4407 | out: | |
3d632c3f | 4408 | if (res) |
1da177e4 LT |
4409 | /* no suitable interface, frame not sent */ |
4410 | dev_kfree_skb(skb); | |
3d632c3f | 4411 | |
1da177e4 LT |
4412 | /* frame sent to all suitable interfaces */ |
4413 | read_unlock(&bond->lock); | |
ec634fe3 | 4414 | return NETDEV_TX_OK; |
1da177e4 LT |
4415 | } |
4416 | ||
4417 | /*------------------------- Device initialization ---------------------------*/ | |
4418 | ||
6f6652be JV |
4419 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4420 | { | |
4421 | switch (bond->params.xmit_policy) { | |
4422 | case BOND_XMIT_POLICY_LAYER23: | |
4423 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4424 | break; | |
4425 | case BOND_XMIT_POLICY_LAYER34: | |
4426 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4427 | break; | |
4428 | case BOND_XMIT_POLICY_LAYER2: | |
4429 | default: | |
4430 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4431 | break; | |
4432 | } | |
4433 | } | |
4434 | ||
bb1d9123 AG |
4435 | /* |
4436 | * Lookup the slave that corresponds to a qid | |
4437 | */ | |
4438 | static inline int bond_slave_override(struct bonding *bond, | |
4439 | struct sk_buff *skb) | |
4440 | { | |
4441 | int i, res = 1; | |
4442 | struct slave *slave = NULL; | |
4443 | struct slave *check_slave; | |
4444 | ||
4445 | read_lock(&bond->lock); | |
4446 | ||
4447 | if (!BOND_IS_OK(bond) || !skb->queue_mapping) | |
4448 | goto out; | |
4449 | ||
4450 | /* Find out if any slaves have the same mapping as this skb. */ | |
4451 | bond_for_each_slave(bond, check_slave, i) { | |
4452 | if (check_slave->queue_id == skb->queue_mapping) { | |
4453 | slave = check_slave; | |
4454 | break; | |
4455 | } | |
4456 | } | |
4457 | ||
4458 | /* If the slave isn't UP, use default transmit policy. */ | |
4459 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4460 | (slave->link == BOND_LINK_UP)) { | |
4461 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4462 | } | |
4463 | ||
4464 | out: | |
4465 | read_unlock(&bond->lock); | |
4466 | return res; | |
4467 | } | |
4468 | ||
4469 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) | |
4470 | { | |
4471 | /* | |
4472 | * This helper function exists to help dev_pick_tx get the correct | |
4473 | * destination queue. Using a helper function skips the a call to | |
4474 | * skb_tx_hash and will put the skbs in the queue we expect on their | |
4475 | * way down to the bonding driver. | |
4476 | */ | |
4477 | return skb->queue_mapping; | |
4478 | } | |
4479 | ||
424efe9c | 4480 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4481 | { |
bb1d9123 AG |
4482 | struct bonding *bond = netdev_priv(dev); |
4483 | ||
4484 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
4485 | if (!bond_slave_override(bond, skb)) | |
4486 | return NETDEV_TX_OK; | |
4487 | } | |
00829823 SH |
4488 | |
4489 | switch (bond->params.mode) { | |
4490 | case BOND_MODE_ROUNDROBIN: | |
4491 | return bond_xmit_roundrobin(skb, dev); | |
4492 | case BOND_MODE_ACTIVEBACKUP: | |
4493 | return bond_xmit_activebackup(skb, dev); | |
4494 | case BOND_MODE_XOR: | |
4495 | return bond_xmit_xor(skb, dev); | |
4496 | case BOND_MODE_BROADCAST: | |
4497 | return bond_xmit_broadcast(skb, dev); | |
4498 | case BOND_MODE_8023AD: | |
4499 | return bond_3ad_xmit_xor(skb, dev); | |
4500 | case BOND_MODE_ALB: | |
4501 | case BOND_MODE_TLB: | |
4502 | return bond_alb_xmit(skb, dev); | |
4503 | default: | |
4504 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4505 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4506 | dev->name, bond->params.mode); | |
00829823 SH |
4507 | WARN_ON_ONCE(1); |
4508 | dev_kfree_skb(skb); | |
4509 | return NETDEV_TX_OK; | |
4510 | } | |
4511 | } | |
4512 | ||
4513 | ||
1da177e4 LT |
4514 | /* |
4515 | * set bond mode specific net device operations | |
4516 | */ | |
a77b5325 | 4517 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4518 | { |
169a3e66 JV |
4519 | struct net_device *bond_dev = bond->dev; |
4520 | ||
1da177e4 LT |
4521 | switch (mode) { |
4522 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4523 | break; |
4524 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4525 | break; |
4526 | case BOND_MODE_XOR: | |
6f6652be | 4527 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4528 | break; |
4529 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4530 | break; |
4531 | case BOND_MODE_8023AD: | |
8f903c70 | 4532 | bond_set_master_3ad_flags(bond); |
6f6652be | 4533 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4534 | break; |
1da177e4 | 4535 | case BOND_MODE_ALB: |
8f903c70 JV |
4536 | bond_set_master_alb_flags(bond); |
4537 | /* FALLTHRU */ | |
4538 | case BOND_MODE_TLB: | |
1da177e4 LT |
4539 | break; |
4540 | default: | |
4541 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4542 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4543 | bond_dev->name, mode); | |
1da177e4 LT |
4544 | break; |
4545 | } | |
4546 | } | |
4547 | ||
217df670 JV |
4548 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4549 | struct ethtool_drvinfo *drvinfo) | |
4550 | { | |
4551 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4552 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4553 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4554 | } | |
4555 | ||
7282d491 | 4556 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4557 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4558 | .get_link = ethtool_op_get_link, |
4559 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4560 | .get_sg = ethtool_op_get_sg, | |
4561 | .get_tso = ethtool_op_get_tso, | |
4562 | .get_ufo = ethtool_op_get_ufo, | |
4563 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4564 | }; |
4565 | ||
eb7cc59a | 4566 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4567 | .ndo_init = bond_init, |
9e71626c | 4568 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4569 | .ndo_open = bond_open, |
4570 | .ndo_stop = bond_close, | |
00829823 | 4571 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4572 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4573 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a SH |
4574 | .ndo_do_ioctl = bond_do_ioctl, |
4575 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4576 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4577 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4578 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4579 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4580 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4581 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
f6dc31a8 WC |
4582 | #ifdef CONFIG_NET_POLL_CONTROLLER |
4583 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, | |
4584 | .ndo_poll_controller = bond_poll_controller, | |
4585 | #endif | |
eb7cc59a SH |
4586 | }; |
4587 | ||
9e2e61fb AW |
4588 | static void bond_destructor(struct net_device *bond_dev) |
4589 | { | |
4590 | struct bonding *bond = netdev_priv(bond_dev); | |
4591 | if (bond->wq) | |
4592 | destroy_workqueue(bond->wq); | |
4593 | free_netdev(bond_dev); | |
4594 | } | |
4595 | ||
181470fc | 4596 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4597 | { |
454d7c9b | 4598 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4599 | |
1da177e4 LT |
4600 | /* initialize rwlocks */ |
4601 | rwlock_init(&bond->lock); | |
4602 | rwlock_init(&bond->curr_slave_lock); | |
4603 | ||
d2991f75 | 4604 | bond->params = bonding_defaults; |
1da177e4 LT |
4605 | |
4606 | /* Initialize pointers */ | |
1da177e4 LT |
4607 | bond->dev = bond_dev; |
4608 | INIT_LIST_HEAD(&bond->vlan_list); | |
4609 | ||
4610 | /* Initialize the device entry points */ | |
181470fc | 4611 | ether_setup(bond_dev); |
eb7cc59a | 4612 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4613 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4614 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4615 | |
9e2e61fb | 4616 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4617 | |
4618 | /* Initialize the device options */ | |
4619 | bond_dev->tx_queue_len = 0; | |
4620 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4621 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4622 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4623 | ||
6cf3f41e JV |
4624 | if (bond->params.arp_interval) |
4625 | bond_dev->priv_flags |= IFF_MASTER_ARPMON; | |
1da177e4 LT |
4626 | |
4627 | /* At first, we block adding VLANs. That's the only way to | |
4628 | * prevent problems that occur when adding VLANs over an | |
4629 | * empty bond. The block will be removed once non-challenged | |
4630 | * slaves are enslaved. | |
4631 | */ | |
4632 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4633 | ||
932ff279 | 4634 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4635 | * transmitting */ |
4636 | bond_dev->features |= NETIF_F_LLTX; | |
4637 | ||
4638 | /* By default, we declare the bond to be fully | |
4639 | * VLAN hardware accelerated capable. Special | |
4640 | * care is taken in the various xmit functions | |
4641 | * when there are slaves that are not hw accel | |
4642 | * capable | |
4643 | */ | |
1da177e4 LT |
4644 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4645 | NETIF_F_HW_VLAN_RX | | |
4646 | NETIF_F_HW_VLAN_FILTER); | |
4647 | ||
1da177e4 LT |
4648 | } |
4649 | ||
fdaea7a9 JV |
4650 | static void bond_work_cancel_all(struct bonding *bond) |
4651 | { | |
4652 | write_lock_bh(&bond->lock); | |
4653 | bond->kill_timers = 1; | |
4654 | write_unlock_bh(&bond->lock); | |
4655 | ||
4656 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4657 | cancel_delayed_work(&bond->mii_work); | |
4658 | ||
4659 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4660 | cancel_delayed_work(&bond->arp_work); | |
4661 | ||
4662 | if (bond->params.mode == BOND_MODE_ALB && | |
4663 | delayed_work_pending(&bond->alb_work)) | |
4664 | cancel_delayed_work(&bond->alb_work); | |
4665 | ||
4666 | if (bond->params.mode == BOND_MODE_8023AD && | |
4667 | delayed_work_pending(&bond->ad_work)) | |
4668 | cancel_delayed_work(&bond->ad_work); | |
4669 | } | |
4670 | ||
c67dfb29 EB |
4671 | /* |
4672 | * Destroy a bonding device. | |
4673 | * Must be under rtnl_lock when this function is called. | |
4674 | */ | |
4675 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4676 | { |
454d7c9b | 4677 | struct bonding *bond = netdev_priv(bond_dev); |
a434e43f | 4678 | |
f6dc31a8 WC |
4679 | bond_netpoll_cleanup(bond_dev); |
4680 | ||
c67dfb29 EB |
4681 | /* Release the bonded slaves */ |
4682 | bond_release_all(bond_dev); | |
4683 | ||
a434e43f JV |
4684 | list_del(&bond->bond_list); |
4685 | ||
4686 | bond_work_cancel_all(bond); | |
4687 | ||
a434e43f | 4688 | bond_remove_proc_entry(bond); |
c67dfb29 | 4689 | |
22bedad3 | 4690 | __hw_addr_flush(&bond->mc_list); |
a434e43f JV |
4691 | } |
4692 | ||
1da177e4 LT |
4693 | /*------------------------- Module initialization ---------------------------*/ |
4694 | ||
4695 | /* | |
4696 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4697 | * number of the mode or its string name. A bit complicated because |
4698 | * some mode names are substrings of other names, and calls from sysfs | |
4699 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4700 | */ |
325dcf7a | 4701 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4702 | { |
54b87323 | 4703 | int modeint = -1, i, rv; |
a42e534f | 4704 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4705 | |
a42e534f JV |
4706 | for (p = (char *)buf; *p; p++) |
4707 | if (!(isdigit(*p) || isspace(*p))) | |
4708 | break; | |
4709 | ||
4710 | if (*p) | |
ece95f7f | 4711 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4712 | else |
54b87323 | 4713 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4714 | |
4715 | if (!rv) | |
4716 | return -1; | |
1da177e4 LT |
4717 | |
4718 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4719 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4720 | return tbl[i].mode; |
4721 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4722 | return tbl[i].mode; |
1da177e4 LT |
4723 | } |
4724 | ||
4725 | return -1; | |
4726 | } | |
4727 | ||
4728 | static int bond_check_params(struct bond_params *params) | |
4729 | { | |
a549952a | 4730 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4731 | |
1da177e4 LT |
4732 | /* |
4733 | * Convert string parameters. | |
4734 | */ | |
4735 | if (mode) { | |
4736 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4737 | if (bond_mode == -1) { | |
a4aee5c8 | 4738 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4739 | mode == NULL ? "NULL" : mode); |
4740 | return -EINVAL; | |
4741 | } | |
4742 | } | |
4743 | ||
169a3e66 JV |
4744 | if (xmit_hash_policy) { |
4745 | if ((bond_mode != BOND_MODE_XOR) && | |
4746 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4747 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4748 | bond_mode_name(bond_mode)); |
4749 | } else { | |
4750 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4751 | xmit_hashtype_tbl); | |
4752 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4753 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4754 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4755 | xmit_hash_policy); |
4756 | return -EINVAL; | |
4757 | } | |
4758 | } | |
4759 | } | |
4760 | ||
1da177e4 LT |
4761 | if (lacp_rate) { |
4762 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4763 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4764 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4765 | } else { |
4766 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4767 | if (lacp_fast == -1) { | |
a4aee5c8 | 4768 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4769 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4770 | return -EINVAL; | |
4771 | } | |
4772 | } | |
4773 | } | |
4774 | ||
fd989c83 JV |
4775 | if (ad_select) { |
4776 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4777 | if (params->ad_select == -1) { | |
a4aee5c8 | 4778 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4779 | ad_select == NULL ? "NULL" : ad_select); |
4780 | return -EINVAL; | |
4781 | } | |
4782 | ||
4783 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4784 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4785 | } |
4786 | } else { | |
4787 | params->ad_select = BOND_AD_STABLE; | |
4788 | } | |
4789 | ||
f5841306 | 4790 | if (max_bonds < 0) { |
a4aee5c8 JP |
4791 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4792 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4793 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4794 | } | |
4795 | ||
4796 | if (miimon < 0) { | |
a4aee5c8 JP |
4797 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4798 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4799 | miimon = BOND_LINK_MON_INTERV; |
4800 | } | |
4801 | ||
4802 | if (updelay < 0) { | |
a4aee5c8 JP |
4803 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4804 | updelay, INT_MAX); | |
1da177e4 LT |
4805 | updelay = 0; |
4806 | } | |
4807 | ||
4808 | if (downdelay < 0) { | |
a4aee5c8 JP |
4809 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4810 | downdelay, INT_MAX); | |
1da177e4 LT |
4811 | downdelay = 0; |
4812 | } | |
4813 | ||
4814 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4815 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4816 | use_carrier); | |
1da177e4 LT |
4817 | use_carrier = 1; |
4818 | } | |
4819 | ||
7893b249 | 4820 | if (num_grat_arp < 0 || num_grat_arp > 255) { |
2381a55c | 4821 | pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4822 | num_grat_arp); |
7893b249 MS |
4823 | num_grat_arp = 1; |
4824 | } | |
4825 | ||
305d552a | 4826 | if (num_unsol_na < 0 || num_unsol_na > 255) { |
2381a55c | 4827 | pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4828 | num_unsol_na); |
305d552a BH |
4829 | num_unsol_na = 1; |
4830 | } | |
4831 | ||
1da177e4 LT |
4832 | /* reset values for 802.3ad */ |
4833 | if (bond_mode == BOND_MODE_8023AD) { | |
4834 | if (!miimon) { | |
a4aee5c8 | 4835 | pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure, speed and duplex which are essential for 802.3ad operation\n"); |
3d632c3f | 4836 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4837 | miimon = 100; |
4838 | } | |
4839 | } | |
4840 | ||
bb1d9123 AG |
4841 | if (tx_queues < 1 || tx_queues > 255) { |
4842 | pr_warning("Warning: tx_queues (%d) should be between " | |
4843 | "1 and 255, resetting to %d\n", | |
4844 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4845 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4846 | } | |
4847 | ||
ebd8e497 AG |
4848 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4849 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4850 | "not of valid value (0/1), so it was set to " | |
4851 | "0\n", all_slaves_active); | |
4852 | all_slaves_active = 0; | |
4853 | } | |
4854 | ||
1da177e4 LT |
4855 | /* reset values for TLB/ALB */ |
4856 | if ((bond_mode == BOND_MODE_TLB) || | |
4857 | (bond_mode == BOND_MODE_ALB)) { | |
4858 | if (!miimon) { | |
a4aee5c8 | 4859 | pr_warning("Warning: miimon must be specified, otherwise bonding will not detect link failure and link speed which are essential for TLB/ALB load balancing\n"); |
3d632c3f | 4860 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4861 | miimon = 100; |
4862 | } | |
4863 | } | |
4864 | ||
4865 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4866 | pr_notice("In ALB mode you might experience client disconnections upon reconnection of a link if the bonding module updelay parameter (%d msec) is incompatible with the forwarding delay time of the switch\n", |
4867 | updelay); | |
1da177e4 LT |
4868 | } |
4869 | ||
4870 | if (!miimon) { | |
4871 | if (updelay || downdelay) { | |
4872 | /* just warn the user the up/down delay will have | |
4873 | * no effect since miimon is zero... | |
4874 | */ | |
a4aee5c8 JP |
4875 | pr_warning("Warning: miimon module parameter not set and updelay (%d) or downdelay (%d) module parameter is set; updelay and downdelay have no effect unless miimon is set\n", |
4876 | updelay, downdelay); | |
1da177e4 LT |
4877 | } |
4878 | } else { | |
4879 | /* don't allow arp monitoring */ | |
4880 | if (arp_interval) { | |
a4aee5c8 JP |
4881 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4882 | miimon, arp_interval); | |
1da177e4 LT |
4883 | arp_interval = 0; |
4884 | } | |
4885 | ||
4886 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4887 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4888 | updelay, miimon, | |
4889 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4890 | } |
4891 | ||
4892 | updelay /= miimon; | |
4893 | ||
4894 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4895 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4896 | downdelay, miimon, | |
4897 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4898 | } |
4899 | ||
4900 | downdelay /= miimon; | |
4901 | } | |
4902 | ||
4903 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4904 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4905 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4906 | arp_interval = BOND_LINK_ARP_INTERV; |
4907 | } | |
4908 | ||
4909 | for (arp_ip_count = 0; | |
4910 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4911 | arp_ip_count++) { | |
4912 | /* not complete check, but should be good enough to | |
4913 | catch mistakes */ | |
4914 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4915 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4916 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4917 | arp_interval = 0; |
4918 | } else { | |
d3bb52b0 | 4919 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4920 | arp_target[arp_ip_count] = ip; |
4921 | } | |
4922 | } | |
4923 | ||
4924 | if (arp_interval && !arp_ip_count) { | |
4925 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4926 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4927 | arp_interval); | |
1da177e4 LT |
4928 | arp_interval = 0; |
4929 | } | |
4930 | ||
f5b2b966 JV |
4931 | if (arp_validate) { |
4932 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4933 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4934 | return -EINVAL; |
4935 | } | |
4936 | if (!arp_interval) { | |
a4aee5c8 | 4937 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4938 | return -EINVAL; |
4939 | } | |
4940 | ||
4941 | arp_validate_value = bond_parse_parm(arp_validate, | |
4942 | arp_validate_tbl); | |
4943 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4944 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4945 | arp_validate == NULL ? "NULL" : arp_validate); |
4946 | return -EINVAL; | |
4947 | } | |
4948 | } else | |
4949 | arp_validate_value = 0; | |
4950 | ||
1da177e4 | 4951 | if (miimon) { |
a4aee5c8 | 4952 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4953 | } else if (arp_interval) { |
4954 | int i; | |
4955 | ||
a4aee5c8 JP |
4956 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4957 | arp_interval, | |
4958 | arp_validate_tbl[arp_validate_value].modename, | |
4959 | arp_ip_count); | |
1da177e4 LT |
4960 | |
4961 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4962 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4963 | |
e5e2a8fd | 4964 | pr_info("\n"); |
1da177e4 | 4965 | |
b8a9787e | 4966 | } else if (max_bonds) { |
1da177e4 LT |
4967 | /* miimon and arp_interval not set, we need one so things |
4968 | * work as expected, see bonding.txt for details | |
4969 | */ | |
a4aee5c8 | 4970 | pr_warning("Warning: either miimon or arp_interval and arp_ip_target module parameters must be specified, otherwise bonding will not detect link failures! see bonding.txt for details.\n"); |
1da177e4 LT |
4971 | } |
4972 | ||
4973 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4974 | /* currently, using a primary only makes sense | |
4975 | * in active backup, TLB or ALB modes | |
4976 | */ | |
a4aee5c8 JP |
4977 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4978 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4979 | primary = NULL; |
4980 | } | |
4981 | ||
a549952a JP |
4982 | if (primary && primary_reselect) { |
4983 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4984 | pri_reselect_tbl); | |
4985 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4986 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4987 | primary_reselect == |
4988 | NULL ? "NULL" : primary_reselect); | |
4989 | return -EINVAL; | |
4990 | } | |
4991 | } else { | |
4992 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4993 | } | |
4994 | ||
3915c1e8 JV |
4995 | if (fail_over_mac) { |
4996 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4997 | fail_over_mac_tbl); | |
4998 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4999 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
5000 | arp_validate == NULL ? "NULL" : arp_validate); |
5001 | return -EINVAL; | |
5002 | } | |
5003 | ||
5004 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 5005 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
5006 | } else { |
5007 | fail_over_mac_value = BOND_FOM_NONE; | |
5008 | } | |
dd957c57 | 5009 | |
1da177e4 LT |
5010 | /* fill params struct with the proper values */ |
5011 | params->mode = bond_mode; | |
169a3e66 | 5012 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 5013 | params->miimon = miimon; |
7893b249 | 5014 | params->num_grat_arp = num_grat_arp; |
305d552a | 5015 | params->num_unsol_na = num_unsol_na; |
1da177e4 | 5016 | params->arp_interval = arp_interval; |
f5b2b966 | 5017 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
5018 | params->updelay = updelay; |
5019 | params->downdelay = downdelay; | |
5020 | params->use_carrier = use_carrier; | |
5021 | params->lacp_fast = lacp_fast; | |
5022 | params->primary[0] = 0; | |
a549952a | 5023 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 5024 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 5025 | params->tx_queues = tx_queues; |
ebd8e497 | 5026 | params->all_slaves_active = all_slaves_active; |
1da177e4 LT |
5027 | |
5028 | if (primary) { | |
5029 | strncpy(params->primary, primary, IFNAMSIZ); | |
5030 | params->primary[IFNAMSIZ - 1] = 0; | |
5031 | } | |
5032 | ||
5033 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
5034 | ||
5035 | return 0; | |
5036 | } | |
5037 | ||
0daa2303 | 5038 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 5039 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 5040 | |
e8a0464c DM |
5041 | static void bond_set_lockdep_class_one(struct net_device *dev, |
5042 | struct netdev_queue *txq, | |
5043 | void *_unused) | |
c773e847 DM |
5044 | { |
5045 | lockdep_set_class(&txq->_xmit_lock, | |
5046 | &bonding_netdev_xmit_lock_key); | |
5047 | } | |
5048 | ||
5049 | static void bond_set_lockdep_class(struct net_device *dev) | |
5050 | { | |
cf508b12 DM |
5051 | lockdep_set_class(&dev->addr_list_lock, |
5052 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 5053 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
5054 | } |
5055 | ||
181470fc SH |
5056 | /* |
5057 | * Called from registration process | |
5058 | */ | |
5059 | static int bond_init(struct net_device *bond_dev) | |
5060 | { | |
5061 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 5062 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc SH |
5063 | |
5064 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
5065 | ||
5066 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
5067 | if (!bond->wq) | |
5068 | return -ENOMEM; | |
5069 | ||
5070 | bond_set_lockdep_class(bond_dev); | |
5071 | ||
5072 | netif_carrier_off(bond_dev); | |
5073 | ||
5074 | bond_create_proc_entry(bond); | |
ec87fd3b | 5075 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 5076 | |
6151b3d4 | 5077 | bond_prepare_sysfs_group(bond); |
22bedad3 JP |
5078 | |
5079 | __hw_addr_init(&bond->mc_list); | |
181470fc SH |
5080 | return 0; |
5081 | } | |
5082 | ||
88ead977 EB |
5083 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
5084 | { | |
5085 | if (tb[IFLA_ADDRESS]) { | |
5086 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
5087 | return -EINVAL; | |
5088 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
5089 | return -EADDRNOTAVAIL; | |
5090 | } | |
5091 | return 0; | |
5092 | } | |
5093 | ||
5094 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
5095 | .kind = "bond", | |
6639104b | 5096 | .priv_size = sizeof(struct bonding), |
88ead977 EB |
5097 | .setup = bond_setup, |
5098 | .validate = bond_validate, | |
5099 | }; | |
5100 | ||
dfe60397 | 5101 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5102 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5103 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5104 | * set up our sysfs entries. | |
5105 | */ | |
ec87fd3b | 5106 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
5107 | { |
5108 | struct net_device *bond_dev; | |
5109 | int res; | |
5110 | ||
5111 | rtnl_lock(); | |
027ea041 | 5112 | |
bb1d9123 AG |
5113 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "", |
5114 | bond_setup, tx_queues); | |
dfe60397 | 5115 | if (!bond_dev) { |
a4aee5c8 | 5116 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
5117 | rtnl_unlock(); |
5118 | return -ENOMEM; | |
dfe60397 MW |
5119 | } |
5120 | ||
ec87fd3b | 5121 | dev_net_set(bond_dev, net); |
88ead977 EB |
5122 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5123 | ||
e4b91c48 JV |
5124 | if (!name) { |
5125 | res = dev_alloc_name(bond_dev, "bond%d"); | |
5126 | if (res < 0) | |
9e2e61fb | 5127 | goto out; |
e4b91c48 JV |
5128 | } |
5129 | ||
dfe60397 | 5130 | res = register_netdevice(bond_dev); |
0daa2303 | 5131 | |
30c15ba9 | 5132 | out: |
7e083840 | 5133 | rtnl_unlock(); |
9e2e61fb AW |
5134 | if (res < 0) |
5135 | bond_destructor(bond_dev); | |
30c15ba9 | 5136 | return res; |
dfe60397 MW |
5137 | } |
5138 | ||
2c8c1e72 | 5139 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 5140 | { |
15449745 | 5141 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5142 | |
5143 | bn->net = net; | |
5144 | INIT_LIST_HEAD(&bn->dev_list); | |
5145 | ||
ec87fd3b | 5146 | bond_create_proc_dir(bn); |
15449745 EB |
5147 | |
5148 | return 0; | |
ec87fd3b EB |
5149 | } |
5150 | ||
2c8c1e72 | 5151 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 5152 | { |
15449745 | 5153 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5154 | |
5155 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
5156 | } |
5157 | ||
5158 | static struct pernet_operations bond_net_ops = { | |
5159 | .init = bond_net_init, | |
5160 | .exit = bond_net_exit, | |
15449745 EB |
5161 | .id = &bond_net_id, |
5162 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
5163 | }; |
5164 | ||
1da177e4 LT |
5165 | static int __init bonding_init(void) |
5166 | { | |
1da177e4 LT |
5167 | int i; |
5168 | int res; | |
5169 | ||
3d632c3f | 5170 | pr_info("%s", version); |
1da177e4 | 5171 | |
dfe60397 | 5172 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5173 | if (res) |
dfe60397 | 5174 | goto out; |
1da177e4 | 5175 | |
15449745 | 5176 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
5177 | if (res) |
5178 | goto out; | |
027ea041 | 5179 | |
88ead977 EB |
5180 | res = rtnl_link_register(&bond_link_ops); |
5181 | if (res) | |
6639104b | 5182 | goto err_link; |
88ead977 | 5183 | |
1da177e4 | 5184 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5185 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5186 | if (res) |
5187 | goto err; | |
1da177e4 LT |
5188 | } |
5189 | ||
b76cdba9 MW |
5190 | res = bond_create_sysfs(); |
5191 | if (res) | |
5192 | goto err; | |
5193 | ||
1da177e4 | 5194 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5195 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5196 | bond_register_ipv6_notifier(); |
dfe60397 | 5197 | out: |
1da177e4 | 5198 | return res; |
88ead977 EB |
5199 | err: |
5200 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5201 | err_link: |
15449745 | 5202 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5203 | goto out; |
dfe60397 | 5204 | |
1da177e4 LT |
5205 | } |
5206 | ||
5207 | static void __exit bonding_exit(void) | |
5208 | { | |
5209 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5210 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5211 | bond_unregister_ipv6_notifier(); |
1da177e4 | 5212 | |
ae68c398 PE |
5213 | bond_destroy_sysfs(); |
5214 | ||
88ead977 | 5215 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5216 | unregister_pernet_subsys(&bond_net_ops); |
1da177e4 LT |
5217 | } |
5218 | ||
5219 | module_init(bonding_init); | |
5220 | module_exit(bonding_exit); | |
5221 | MODULE_LICENSE("GPL"); | |
5222 | MODULE_VERSION(DRV_VERSION); | |
5223 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5224 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5225 | MODULE_ALIAS_RTNL_LINK("bond"); |