]>
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; | |
90e1795b | 179 | #ifdef CONFIG_NET_POLL_CONTROLLER |
c22d7ac8 | 180 | static int disable_netpoll = 1; |
90e1795b | 181 | #endif |
1da177e4 | 182 | |
e97fd7c6 | 183 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
184 | { "slow", AD_LACP_SLOW}, |
185 | { "fast", AD_LACP_FAST}, | |
186 | { NULL, -1}, | |
187 | }; | |
188 | ||
e97fd7c6 | 189 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
190 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
191 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
192 | { "balance-xor", BOND_MODE_XOR}, | |
193 | { "broadcast", BOND_MODE_BROADCAST}, | |
194 | { "802.3ad", BOND_MODE_8023AD}, | |
195 | { "balance-tlb", BOND_MODE_TLB}, | |
196 | { "balance-alb", BOND_MODE_ALB}, | |
197 | { NULL, -1}, | |
198 | }; | |
199 | ||
e97fd7c6 | 200 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
201 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
202 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 203 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
204 | { NULL, -1}, |
205 | }; | |
206 | ||
e97fd7c6 | 207 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
208 | { "none", BOND_ARP_VALIDATE_NONE}, |
209 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
210 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
211 | { "all", BOND_ARP_VALIDATE_ALL}, | |
212 | { NULL, -1}, | |
213 | }; | |
214 | ||
e97fd7c6 | 215 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
216 | { "none", BOND_FOM_NONE}, |
217 | { "active", BOND_FOM_ACTIVE}, | |
218 | { "follow", BOND_FOM_FOLLOW}, | |
219 | { NULL, -1}, | |
220 | }; | |
221 | ||
a549952a JP |
222 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
223 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
224 | { "better", BOND_PRI_RESELECT_BETTER}, | |
225 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
226 | { NULL, -1}, | |
227 | }; | |
228 | ||
fd989c83 JV |
229 | struct bond_parm_tbl ad_select_tbl[] = { |
230 | { "stable", BOND_AD_STABLE}, | |
231 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
232 | { "count", BOND_AD_COUNT}, | |
233 | { NULL, -1}, | |
234 | }; | |
235 | ||
1da177e4 LT |
236 | /*-------------------------- Forward declarations ---------------------------*/ |
237 | ||
c3ade5ca | 238 | static void bond_send_gratuitous_arp(struct bonding *bond); |
181470fc | 239 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 240 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
241 | |
242 | /*---------------------------- General routines -----------------------------*/ | |
243 | ||
4ad072c9 | 244 | static const char *bond_mode_name(int mode) |
1da177e4 | 245 | { |
77afc92b HE |
246 | static const char *names[] = { |
247 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
248 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
249 | [BOND_MODE_XOR] = "load balancing (xor)", | |
250 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 251 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
252 | [BOND_MODE_TLB] = "transmit load balancing", |
253 | [BOND_MODE_ALB] = "adaptive load balancing", | |
254 | }; | |
255 | ||
256 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 257 | return "unknown"; |
77afc92b HE |
258 | |
259 | return names[mode]; | |
1da177e4 LT |
260 | } |
261 | ||
262 | /*---------------------------------- VLAN -----------------------------------*/ | |
263 | ||
264 | /** | |
265 | * bond_add_vlan - add a new vlan id on bond | |
266 | * @bond: bond that got the notification | |
267 | * @vlan_id: the vlan id to add | |
268 | * | |
269 | * Returns -ENOMEM if allocation failed. | |
270 | */ | |
271 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
272 | { | |
273 | struct vlan_entry *vlan; | |
274 | ||
5a03cdb7 | 275 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 276 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 277 | |
305d552a | 278 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 279 | if (!vlan) |
1da177e4 | 280 | return -ENOMEM; |
1da177e4 LT |
281 | |
282 | INIT_LIST_HEAD(&vlan->vlan_list); | |
283 | vlan->vlan_id = vlan_id; | |
284 | ||
285 | write_lock_bh(&bond->lock); | |
286 | ||
287 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
288 | ||
289 | write_unlock_bh(&bond->lock); | |
290 | ||
5a03cdb7 | 291 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
292 | |
293 | return 0; | |
294 | } | |
295 | ||
296 | /** | |
297 | * bond_del_vlan - delete a vlan id from bond | |
298 | * @bond: bond that got the notification | |
299 | * @vlan_id: the vlan id to delete | |
300 | * | |
301 | * returns -ENODEV if @vlan_id was not found in @bond. | |
302 | */ | |
303 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
304 | { | |
0883beca | 305 | struct vlan_entry *vlan; |
1da177e4 LT |
306 | int res = -ENODEV; |
307 | ||
5a03cdb7 | 308 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 LT |
309 | |
310 | write_lock_bh(&bond->lock); | |
311 | ||
0883beca | 312 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
313 | if (vlan->vlan_id == vlan_id) { |
314 | list_del(&vlan->vlan_list); | |
315 | ||
58402054 | 316 | if (bond_is_lb(bond)) |
1da177e4 | 317 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 318 | |
a4aee5c8 JP |
319 | pr_debug("removed VLAN ID %d from bond %s\n", |
320 | vlan_id, bond->dev->name); | |
1da177e4 LT |
321 | |
322 | kfree(vlan); | |
323 | ||
324 | if (list_empty(&bond->vlan_list) && | |
325 | (bond->slave_cnt == 0)) { | |
326 | /* Last VLAN removed and no slaves, so | |
327 | * restore block on adding VLANs. This will | |
328 | * be removed once new slaves that are not | |
329 | * VLAN challenged will be added. | |
330 | */ | |
331 | bond->dev->features |= NETIF_F_VLAN_CHALLENGED; | |
332 | } | |
333 | ||
334 | res = 0; | |
335 | goto out; | |
336 | } | |
337 | } | |
338 | ||
a4aee5c8 JP |
339 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
340 | vlan_id, bond->dev->name); | |
1da177e4 LT |
341 | |
342 | out: | |
343 | write_unlock_bh(&bond->lock); | |
344 | return res; | |
345 | } | |
346 | ||
347 | /** | |
348 | * bond_has_challenged_slaves | |
349 | * @bond: the bond we're working on | |
350 | * | |
351 | * Searches the slave list. Returns 1 if a vlan challenged slave | |
352 | * was found, 0 otherwise. | |
353 | * | |
354 | * Assumes bond->lock is held. | |
355 | */ | |
356 | static int bond_has_challenged_slaves(struct bonding *bond) | |
357 | { | |
358 | struct slave *slave; | |
359 | int i; | |
360 | ||
361 | bond_for_each_slave(bond, slave, i) { | |
362 | if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 363 | pr_debug("found VLAN challenged slave - %s\n", |
a4aee5c8 | 364 | slave->dev->name); |
1da177e4 LT |
365 | return 1; |
366 | } | |
367 | } | |
368 | ||
5a03cdb7 | 369 | pr_debug("no VLAN challenged slaves found\n"); |
1da177e4 LT |
370 | return 0; |
371 | } | |
372 | ||
373 | /** | |
374 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
375 | * @bond: the bond we're working on | |
376 | * @curr: item we're advancing from | |
377 | * | |
378 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
379 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 380 | * |
1da177e4 LT |
381 | * Caller must hold bond->lock |
382 | */ | |
383 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
384 | { | |
385 | struct vlan_entry *next, *last; | |
386 | ||
3d632c3f | 387 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 388 | return NULL; |
1da177e4 LT |
389 | |
390 | if (!curr) { | |
391 | next = list_entry(bond->vlan_list.next, | |
392 | struct vlan_entry, vlan_list); | |
393 | } else { | |
394 | last = list_entry(bond->vlan_list.prev, | |
395 | struct vlan_entry, vlan_list); | |
396 | if (last == curr) { | |
397 | next = list_entry(bond->vlan_list.next, | |
398 | struct vlan_entry, vlan_list); | |
399 | } else { | |
400 | next = list_entry(curr->vlan_list.next, | |
401 | struct vlan_entry, vlan_list); | |
402 | } | |
403 | } | |
404 | ||
405 | return next; | |
406 | } | |
407 | ||
408 | /** | |
409 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 410 | * |
1da177e4 LT |
411 | * @bond: bond device that got this skb for tx. |
412 | * @skb: hw accel VLAN tagged skb to transmit | |
413 | * @slave_dev: slave that is supposed to xmit this skbuff | |
3d632c3f | 414 | * |
1da177e4 LT |
415 | * When the bond gets an skb to transmit that is |
416 | * already hardware accelerated VLAN tagged, and it | |
417 | * needs to relay this skb to a slave that is not | |
418 | * hw accel capable, the skb needs to be "unaccelerated", | |
419 | * i.e. strip the hwaccel tag and re-insert it as part | |
420 | * of the payload. | |
421 | */ | |
3d632c3f SH |
422 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
423 | struct net_device *slave_dev) | |
1da177e4 | 424 | { |
966bc6f4 | 425 | unsigned short uninitialized_var(vlan_id); |
1da177e4 | 426 | |
f35188fa | 427 | /* Test vlan_list not vlgrp to catch and handle 802.1p tags */ |
1da177e4 LT |
428 | if (!list_empty(&bond->vlan_list) && |
429 | !(slave_dev->features & NETIF_F_HW_VLAN_TX) && | |
430 | vlan_get_tag(skb, &vlan_id) == 0) { | |
431 | skb->dev = slave_dev; | |
432 | skb = vlan_put_tag(skb, vlan_id); | |
433 | if (!skb) { | |
434 | /* vlan_put_tag() frees the skb in case of error, | |
435 | * so return success here so the calling functions | |
436 | * won't attempt to free is again. | |
437 | */ | |
438 | return 0; | |
439 | } | |
440 | } else { | |
441 | skb->dev = slave_dev; | |
442 | } | |
443 | ||
444 | skb->priority = 1; | |
f6dc31a8 WC |
445 | #ifdef CONFIG_NET_POLL_CONTROLLER |
446 | if (unlikely(bond->dev->priv_flags & IFF_IN_NETPOLL)) { | |
447 | struct netpoll *np = bond->dev->npinfo->netpoll; | |
448 | slave_dev->npinfo = bond->dev->npinfo; | |
449 | np->real_dev = np->dev = skb->dev; | |
450 | slave_dev->priv_flags |= IFF_IN_NETPOLL; | |
451 | netpoll_send_skb(np, skb); | |
452 | slave_dev->priv_flags &= ~IFF_IN_NETPOLL; | |
453 | np->dev = bond->dev; | |
454 | } else | |
455 | #endif | |
456 | dev_queue_xmit(skb); | |
1da177e4 LT |
457 | |
458 | return 0; | |
459 | } | |
460 | ||
461 | /* | |
462 | * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid | |
463 | * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a | |
464 | * lock because: | |
465 | * a. This operation is performed in IOCTL context, | |
466 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
467 | * c. Holding a lock with BH disabled while directly calling a base driver | |
468 | * entry point is generally a BAD idea. | |
3d632c3f | 469 | * |
1da177e4 LT |
470 | * The design of synchronization/protection for this operation in the 8021q |
471 | * module is good for one or more VLAN devices over a single physical device | |
472 | * and cannot be extended for a teaming solution like bonding, so there is a | |
473 | * potential race condition here where a net device from the vlan group might | |
474 | * be referenced (either by a base driver or the 8021q code) while it is being | |
475 | * removed from the system. However, it turns out we're not making matters | |
476 | * worse, and if it works for regular VLAN usage it will work here too. | |
477 | */ | |
478 | ||
479 | /** | |
480 | * bond_vlan_rx_register - Propagates registration to slaves | |
481 | * @bond_dev: bonding net device that got called | |
482 | * @grp: vlan group being registered | |
483 | */ | |
3d632c3f SH |
484 | static void bond_vlan_rx_register(struct net_device *bond_dev, |
485 | struct vlan_group *grp) | |
1da177e4 | 486 | { |
454d7c9b | 487 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
488 | struct slave *slave; |
489 | int i; | |
490 | ||
f35188fa | 491 | write_lock(&bond->lock); |
1da177e4 | 492 | bond->vlgrp = grp; |
f35188fa | 493 | write_unlock(&bond->lock); |
1da177e4 LT |
494 | |
495 | bond_for_each_slave(bond, slave, i) { | |
496 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 497 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
498 | |
499 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
500 | slave_ops->ndo_vlan_rx_register) { |
501 | slave_ops->ndo_vlan_rx_register(slave_dev, grp); | |
1da177e4 LT |
502 | } |
503 | } | |
504 | } | |
505 | ||
506 | /** | |
507 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
508 | * @bond_dev: bonding net device that got called | |
509 | * @vid: vlan id being added | |
510 | */ | |
511 | static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) | |
512 | { | |
454d7c9b | 513 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
514 | struct slave *slave; |
515 | int i, res; | |
516 | ||
517 | bond_for_each_slave(bond, slave, i) { | |
518 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 519 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
520 | |
521 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a SH |
522 | slave_ops->ndo_vlan_rx_add_vid) { |
523 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid); | |
1da177e4 LT |
524 | } |
525 | } | |
526 | ||
527 | res = bond_add_vlan(bond, vid); | |
528 | if (res) { | |
a4aee5c8 | 529 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 LT |
530 | bond_dev->name, vid); |
531 | } | |
532 | } | |
533 | ||
534 | /** | |
535 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
536 | * @bond_dev: bonding net device that got called | |
537 | * @vid: vlan id being removed | |
538 | */ | |
539 | static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) | |
540 | { | |
454d7c9b | 541 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
542 | struct slave *slave; |
543 | struct net_device *vlan_dev; | |
544 | int i, res; | |
545 | ||
546 | bond_for_each_slave(bond, slave, i) { | |
547 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 548 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
549 | |
550 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a | 551 | slave_ops->ndo_vlan_rx_kill_vid) { |
1da177e4 LT |
552 | /* Save and then restore vlan_dev in the grp array, |
553 | * since the slave's driver might clear it. | |
554 | */ | |
5c15bdec | 555 | vlan_dev = vlan_group_get_device(bond->vlgrp, vid); |
eb7cc59a | 556 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid); |
5c15bdec | 557 | vlan_group_set_device(bond->vlgrp, vid, vlan_dev); |
1da177e4 LT |
558 | } |
559 | } | |
560 | ||
561 | res = bond_del_vlan(bond, vid); | |
562 | if (res) { | |
a4aee5c8 | 563 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 LT |
564 | bond_dev->name, vid); |
565 | } | |
566 | } | |
567 | ||
568 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
569 | { | |
570 | struct vlan_entry *vlan; | |
eb7cc59a | 571 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
572 | |
573 | write_lock_bh(&bond->lock); | |
574 | ||
f35188fa | 575 | if (!bond->vlgrp) |
1da177e4 | 576 | goto out; |
1da177e4 LT |
577 | |
578 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
579 | slave_ops->ndo_vlan_rx_register) |
580 | slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp); | |
1da177e4 LT |
581 | |
582 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 583 | !(slave_ops->ndo_vlan_rx_add_vid)) |
1da177e4 | 584 | goto out; |
1da177e4 | 585 | |
eb7cc59a SH |
586 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) |
587 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id); | |
1da177e4 LT |
588 | |
589 | out: | |
590 | write_unlock_bh(&bond->lock); | |
591 | } | |
592 | ||
3d632c3f SH |
593 | static void bond_del_vlans_from_slave(struct bonding *bond, |
594 | struct net_device *slave_dev) | |
1da177e4 | 595 | { |
eb7cc59a | 596 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
597 | struct vlan_entry *vlan; |
598 | struct net_device *vlan_dev; | |
599 | ||
600 | write_lock_bh(&bond->lock); | |
601 | ||
f35188fa | 602 | if (!bond->vlgrp) |
1da177e4 | 603 | goto out; |
1da177e4 LT |
604 | |
605 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 606 | !(slave_ops->ndo_vlan_rx_kill_vid)) |
1da177e4 | 607 | goto unreg; |
1da177e4 LT |
608 | |
609 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
f35188fa JV |
610 | if (!vlan->vlan_id) |
611 | continue; | |
1da177e4 LT |
612 | /* Save and then restore vlan_dev in the grp array, |
613 | * since the slave's driver might clear it. | |
614 | */ | |
5c15bdec | 615 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
eb7cc59a | 616 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id); |
5c15bdec | 617 | vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev); |
1da177e4 LT |
618 | } |
619 | ||
620 | unreg: | |
621 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
622 | slave_ops->ndo_vlan_rx_register) |
623 | slave_ops->ndo_vlan_rx_register(slave_dev, NULL); | |
1da177e4 LT |
624 | |
625 | out: | |
626 | write_unlock_bh(&bond->lock); | |
627 | } | |
628 | ||
629 | /*------------------------------- Link status -------------------------------*/ | |
630 | ||
ff59c456 JV |
631 | /* |
632 | * Set the carrier state for the master according to the state of its | |
633 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
634 | * do special 802.3ad magic. | |
635 | * | |
636 | * Returns zero if carrier state does not change, nonzero if it does. | |
637 | */ | |
638 | static int bond_set_carrier(struct bonding *bond) | |
639 | { | |
640 | struct slave *slave; | |
641 | int i; | |
642 | ||
643 | if (bond->slave_cnt == 0) | |
644 | goto down; | |
645 | ||
646 | if (bond->params.mode == BOND_MODE_8023AD) | |
647 | return bond_3ad_set_carrier(bond); | |
648 | ||
649 | bond_for_each_slave(bond, slave, i) { | |
650 | if (slave->link == BOND_LINK_UP) { | |
651 | if (!netif_carrier_ok(bond->dev)) { | |
652 | netif_carrier_on(bond->dev); | |
653 | return 1; | |
654 | } | |
655 | return 0; | |
656 | } | |
657 | } | |
658 | ||
659 | down: | |
660 | if (netif_carrier_ok(bond->dev)) { | |
661 | netif_carrier_off(bond->dev); | |
662 | return 1; | |
663 | } | |
664 | return 0; | |
665 | } | |
666 | ||
1da177e4 LT |
667 | /* |
668 | * Get link speed and duplex from the slave's base driver | |
669 | * using ethtool. If for some reason the call fails or the | |
670 | * values are invalid, fake speed and duplex to 100/Full | |
671 | * and return error. | |
672 | */ | |
673 | static int bond_update_speed_duplex(struct slave *slave) | |
674 | { | |
675 | struct net_device *slave_dev = slave->dev; | |
1da177e4 | 676 | struct ethtool_cmd etool; |
61a44b9c | 677 | int res; |
1da177e4 LT |
678 | |
679 | /* Fake speed and duplex */ | |
680 | slave->speed = SPEED_100; | |
681 | slave->duplex = DUPLEX_FULL; | |
682 | ||
61a44b9c MW |
683 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
684 | return -1; | |
1da177e4 | 685 | |
61a44b9c MW |
686 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
687 | if (res < 0) | |
1da177e4 | 688 | return -1; |
1da177e4 | 689 | |
1da177e4 LT |
690 | switch (etool.speed) { |
691 | case SPEED_10: | |
692 | case SPEED_100: | |
693 | case SPEED_1000: | |
94dbffd5 | 694 | case SPEED_10000: |
1da177e4 LT |
695 | break; |
696 | default: | |
697 | return -1; | |
698 | } | |
699 | ||
700 | switch (etool.duplex) { | |
701 | case DUPLEX_FULL: | |
702 | case DUPLEX_HALF: | |
703 | break; | |
704 | default: | |
705 | return -1; | |
706 | } | |
707 | ||
708 | slave->speed = etool.speed; | |
709 | slave->duplex = etool.duplex; | |
710 | ||
711 | return 0; | |
712 | } | |
713 | ||
714 | /* | |
715 | * if <dev> supports MII link status reporting, check its link status. | |
716 | * | |
717 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 718 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
719 | * |
720 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
721 | * can't tell and just pretend it is), or 0, meaning that the link is | |
722 | * down. | |
723 | * | |
724 | * If reporting is non-zero, instead of faking link up, return -1 if | |
725 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
726 | * support them). If use_carrier is set, return whatever it says. | |
727 | * It'd be nice if there was a good way to tell if a driver supports | |
728 | * netif_carrier, but there really isn't. | |
729 | */ | |
3d632c3f SH |
730 | static int bond_check_dev_link(struct bonding *bond, |
731 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 732 | { |
eb7cc59a | 733 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 734 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
735 | struct ifreq ifr; |
736 | struct mii_ioctl_data *mii; | |
1da177e4 | 737 | |
6c988853 PG |
738 | if (!reporting && !netif_running(slave_dev)) |
739 | return 0; | |
740 | ||
eb7cc59a | 741 | if (bond->params.use_carrier) |
1da177e4 | 742 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 743 | |
29112f4e JP |
744 | /* Try to get link status using Ethtool first. */ |
745 | if (slave_dev->ethtool_ops) { | |
746 | if (slave_dev->ethtool_ops->get_link) { | |
747 | u32 link; | |
748 | ||
749 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
750 | ||
751 | return link ? BMSR_LSTATUS : 0; | |
752 | } | |
753 | } | |
754 | ||
3d632c3f | 755 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 756 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
757 | if (ioctl) { |
758 | /* TODO: set pointer to correct ioctl on a per team member */ | |
759 | /* bases to make this more efficient. that is, once */ | |
760 | /* we determine the correct ioctl, we will always */ | |
761 | /* call it and not the others for that team */ | |
762 | /* member. */ | |
763 | ||
764 | /* | |
765 | * We cannot assume that SIOCGMIIPHY will also read a | |
766 | * register; not all network drivers (e.g., e100) | |
767 | * support that. | |
768 | */ | |
769 | ||
770 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
771 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
772 | mii = if_mii(&ifr); | |
773 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
774 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
775 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
776 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
777 | } |
778 | } | |
779 | ||
1da177e4 LT |
780 | /* |
781 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 782 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
783 | * cannot report link status). If not reporting, pretend |
784 | * we're ok. | |
785 | */ | |
3d632c3f | 786 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
787 | } |
788 | ||
789 | /*----------------------------- Multicast list ------------------------------*/ | |
790 | ||
1da177e4 LT |
791 | /* |
792 | * Push the promiscuity flag down to appropriate slaves | |
793 | */ | |
7e1a1ac1 | 794 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 795 | { |
7e1a1ac1 | 796 | int err = 0; |
1da177e4 LT |
797 | if (USES_PRIMARY(bond->params.mode)) { |
798 | /* write lock already acquired */ | |
799 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
800 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
801 | inc); | |
1da177e4 LT |
802 | } |
803 | } else { | |
804 | struct slave *slave; | |
805 | int i; | |
806 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
807 | err = dev_set_promiscuity(slave->dev, inc); |
808 | if (err) | |
809 | return err; | |
1da177e4 LT |
810 | } |
811 | } | |
7e1a1ac1 | 812 | return err; |
1da177e4 LT |
813 | } |
814 | ||
815 | /* | |
816 | * Push the allmulti flag down to all slaves | |
817 | */ | |
7e1a1ac1 | 818 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 819 | { |
7e1a1ac1 | 820 | int err = 0; |
1da177e4 LT |
821 | if (USES_PRIMARY(bond->params.mode)) { |
822 | /* write lock already acquired */ | |
823 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
824 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
825 | inc); | |
1da177e4 LT |
826 | } |
827 | } else { | |
828 | struct slave *slave; | |
829 | int i; | |
830 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
831 | err = dev_set_allmulti(slave->dev, inc); |
832 | if (err) | |
833 | return err; | |
1da177e4 LT |
834 | } |
835 | } | |
7e1a1ac1 | 836 | return err; |
1da177e4 LT |
837 | } |
838 | ||
839 | /* | |
840 | * Add a Multicast address to slaves | |
841 | * according to mode | |
842 | */ | |
22bedad3 | 843 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
844 | { |
845 | if (USES_PRIMARY(bond->params.mode)) { | |
846 | /* write lock already acquired */ | |
3d632c3f | 847 | if (bond->curr_active_slave) |
22bedad3 | 848 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
849 | } else { |
850 | struct slave *slave; | |
851 | int i; | |
3d632c3f SH |
852 | |
853 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 854 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
855 | } |
856 | } | |
857 | ||
858 | /* | |
859 | * Remove a multicast address from slave | |
860 | * according to mode | |
861 | */ | |
22bedad3 | 862 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
863 | { |
864 | if (USES_PRIMARY(bond->params.mode)) { | |
865 | /* write lock already acquired */ | |
3d632c3f | 866 | if (bond->curr_active_slave) |
22bedad3 | 867 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
868 | } else { |
869 | struct slave *slave; | |
870 | int i; | |
871 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 872 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
873 | } |
874 | } | |
875 | } | |
876 | ||
a816c7c7 JV |
877 | |
878 | /* | |
879 | * Retrieve the list of registered multicast addresses for the bonding | |
880 | * device and retransmit an IGMP JOIN request to the current active | |
881 | * slave. | |
882 | */ | |
883 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
884 | { | |
885 | struct in_device *in_dev; | |
886 | struct ip_mc_list *im; | |
887 | ||
888 | rcu_read_lock(); | |
889 | in_dev = __in_dev_get_rcu(bond->dev); | |
890 | if (in_dev) { | |
3d632c3f | 891 | for (im = in_dev->mc_list; im; im = im->next) |
a816c7c7 | 892 | ip_mc_rejoin_group(im); |
a816c7c7 JV |
893 | } |
894 | ||
895 | rcu_read_unlock(); | |
896 | } | |
897 | ||
1da177e4 LT |
898 | /* |
899 | * flush all members of flush->mc_list from device dev->mc_list | |
900 | */ | |
3d632c3f SH |
901 | static void bond_mc_list_flush(struct net_device *bond_dev, |
902 | struct net_device *slave_dev) | |
1da177e4 | 903 | { |
454d7c9b | 904 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 905 | struct netdev_hw_addr *ha; |
1da177e4 | 906 | |
22bedad3 JP |
907 | netdev_for_each_mc_addr(ha, bond_dev) |
908 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
909 | |
910 | if (bond->params.mode == BOND_MODE_8023AD) { | |
911 | /* del lacpdu mc addr from mc list */ | |
912 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
913 | ||
22bedad3 | 914 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
915 | } |
916 | } | |
917 | ||
918 | /*--------------------------- Active slave change ---------------------------*/ | |
919 | ||
920 | /* | |
921 | * Update the mc list and multicast-related flags for the new and | |
922 | * old active slaves (if any) according to the multicast mode, and | |
923 | * promiscuous flags unconditionally. | |
924 | */ | |
3d632c3f SH |
925 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
926 | struct slave *old_active) | |
1da177e4 | 927 | { |
22bedad3 | 928 | struct netdev_hw_addr *ha; |
1da177e4 | 929 | |
3d632c3f | 930 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
931 | /* nothing to do - mc list is already up-to-date on |
932 | * all slaves | |
933 | */ | |
934 | return; | |
1da177e4 LT |
935 | |
936 | if (old_active) { | |
3d632c3f | 937 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 938 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 939 | |
3d632c3f | 940 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 941 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 942 | |
22bedad3 JP |
943 | netdev_for_each_mc_addr(ha, bond->dev) |
944 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
945 | } |
946 | ||
947 | if (new_active) { | |
7e1a1ac1 | 948 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 949 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 950 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 951 | |
3d632c3f | 952 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 953 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 954 | |
22bedad3 JP |
955 | netdev_for_each_mc_addr(ha, bond->dev) |
956 | dev_mc_add(new_active->dev, ha->addr); | |
a816c7c7 | 957 | bond_resend_igmp_join_requests(bond); |
1da177e4 LT |
958 | } |
959 | } | |
960 | ||
3915c1e8 JV |
961 | /* |
962 | * bond_do_fail_over_mac | |
963 | * | |
964 | * Perform special MAC address swapping for fail_over_mac settings | |
965 | * | |
966 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
967 | */ | |
968 | static void bond_do_fail_over_mac(struct bonding *bond, | |
969 | struct slave *new_active, | |
970 | struct slave *old_active) | |
1f78d9f9 HE |
971 | __releases(&bond->curr_slave_lock) |
972 | __releases(&bond->lock) | |
973 | __acquires(&bond->lock) | |
974 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
975 | { |
976 | u8 tmp_mac[ETH_ALEN]; | |
977 | struct sockaddr saddr; | |
978 | int rv; | |
979 | ||
980 | switch (bond->params.fail_over_mac) { | |
981 | case BOND_FOM_ACTIVE: | |
982 | if (new_active) | |
983 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
984 | new_active->dev->addr_len); | |
985 | break; | |
986 | case BOND_FOM_FOLLOW: | |
987 | /* | |
988 | * if new_active && old_active, swap them | |
989 | * if just old_active, do nothing (going to no active slave) | |
990 | * if just new_active, set new_active to bond's MAC | |
991 | */ | |
992 | if (!new_active) | |
993 | return; | |
994 | ||
995 | write_unlock_bh(&bond->curr_slave_lock); | |
996 | read_unlock(&bond->lock); | |
997 | ||
998 | if (old_active) { | |
999 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
1000 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
1001 | ETH_ALEN); | |
1002 | saddr.sa_family = new_active->dev->type; | |
1003 | } else { | |
1004 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
1005 | saddr.sa_family = bond->dev->type; | |
1006 | } | |
1007 | ||
1008 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
1009 | if (rv) { | |
a4aee5c8 | 1010 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1011 | bond->dev->name, -rv, new_active->dev->name); |
1012 | goto out; | |
1013 | } | |
1014 | ||
1015 | if (!old_active) | |
1016 | goto out; | |
1017 | ||
1018 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1019 | saddr.sa_family = old_active->dev->type; | |
1020 | ||
1021 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1022 | if (rv) | |
a4aee5c8 | 1023 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1024 | bond->dev->name, -rv, new_active->dev->name); |
1025 | out: | |
1026 | read_lock(&bond->lock); | |
1027 | write_lock_bh(&bond->curr_slave_lock); | |
1028 | break; | |
1029 | default: | |
a4aee5c8 | 1030 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
1031 | bond->dev->name, bond->params.fail_over_mac); |
1032 | break; | |
1033 | } | |
1034 | ||
1035 | } | |
1036 | ||
a549952a JP |
1037 | static bool bond_should_change_active(struct bonding *bond) |
1038 | { | |
1039 | struct slave *prim = bond->primary_slave; | |
1040 | struct slave *curr = bond->curr_active_slave; | |
1041 | ||
1042 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1043 | return true; | |
1044 | if (bond->force_primary) { | |
1045 | bond->force_primary = false; | |
1046 | return true; | |
1047 | } | |
1048 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1049 | (prim->speed < curr->speed || | |
1050 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1051 | return false; | |
1052 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1053 | return false; | |
1054 | return true; | |
1055 | } | |
3915c1e8 | 1056 | |
1da177e4 LT |
1057 | /** |
1058 | * find_best_interface - select the best available slave to be the active one | |
1059 | * @bond: our bonding struct | |
1060 | * | |
1061 | * Warning: Caller must hold curr_slave_lock for writing. | |
1062 | */ | |
1063 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1064 | { | |
1065 | struct slave *new_active, *old_active; | |
1066 | struct slave *bestslave = NULL; | |
1067 | int mintime = bond->params.updelay; | |
1068 | int i; | |
1069 | ||
49b4ad92 | 1070 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1071 | |
1072 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1073 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1074 | new_active = bond->first_slave; |
3d632c3f | 1075 | else |
1da177e4 | 1076 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1077 | } |
1078 | ||
1da177e4 | 1079 | if ((bond->primary_slave) && |
a549952a JP |
1080 | bond->primary_slave->link == BOND_LINK_UP && |
1081 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1082 | new_active = bond->primary_slave; |
1083 | } | |
1084 | ||
1085 | /* remember where to stop iterating over the slaves */ | |
1086 | old_active = new_active; | |
1087 | ||
1088 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1089 | if (new_active->link == BOND_LINK_UP) { |
1090 | return new_active; | |
1091 | } else if (new_active->link == BOND_LINK_BACK && | |
1092 | IS_UP(new_active->dev)) { | |
1093 | /* link up, but waiting for stabilization */ | |
1094 | if (new_active->delay < mintime) { | |
1095 | mintime = new_active->delay; | |
1096 | bestslave = new_active; | |
1da177e4 LT |
1097 | } |
1098 | } | |
1099 | } | |
1100 | ||
1101 | return bestslave; | |
1102 | } | |
1103 | ||
1104 | /** | |
1105 | * change_active_interface - change the active slave into the specified one | |
1106 | * @bond: our bonding struct | |
1107 | * @new: the new slave to make the active one | |
1108 | * | |
1109 | * Set the new slave to the bond's settings and unset them on the old | |
1110 | * curr_active_slave. | |
1111 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1112 | * | |
1113 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1114 | * because it is apparently the best available slave we have, even though its | |
1115 | * updelay hasn't timed out yet. | |
1116 | * | |
3915c1e8 JV |
1117 | * If new_active is not NULL, caller must hold bond->lock for read and |
1118 | * curr_slave_lock for write_bh. | |
1da177e4 | 1119 | */ |
a77b5325 | 1120 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1121 | { |
1122 | struct slave *old_active = bond->curr_active_slave; | |
1123 | ||
3d632c3f | 1124 | if (old_active == new_active) |
1da177e4 | 1125 | return; |
1da177e4 LT |
1126 | |
1127 | if (new_active) { | |
b2220cad JV |
1128 | new_active->jiffies = jiffies; |
1129 | ||
1da177e4 LT |
1130 | if (new_active->link == BOND_LINK_BACK) { |
1131 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1132 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1133 | bond->dev->name, new_active->dev->name, | |
1134 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1135 | } |
1136 | ||
1137 | new_active->delay = 0; | |
1138 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1139 | |
3d632c3f | 1140 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1141 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1142 | |
58402054 | 1143 | if (bond_is_lb(bond)) |
1da177e4 | 1144 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1145 | } else { |
1146 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1147 | pr_info("%s: making interface %s the new active one.\n", |
1148 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1149 | } |
1150 | } | |
1151 | } | |
1152 | ||
3d632c3f | 1153 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1154 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1155 | |
58402054 | 1156 | if (bond_is_lb(bond)) { |
1da177e4 | 1157 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1158 | if (old_active) |
1159 | bond_set_slave_inactive_flags(old_active); | |
1160 | if (new_active) | |
1161 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1162 | } else { |
1163 | bond->curr_active_slave = new_active; | |
1164 | } | |
c3ade5ca JV |
1165 | |
1166 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1167 | if (old_active) |
c3ade5ca | 1168 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1169 | |
1170 | if (new_active) { | |
1171 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1172 | |
709f8a45 OG |
1173 | if (bond->params.fail_over_mac) |
1174 | bond_do_fail_over_mac(bond, new_active, | |
1175 | old_active); | |
3915c1e8 | 1176 | |
709f8a45 | 1177 | bond->send_grat_arp = bond->params.num_grat_arp; |
b59f9f74 | 1178 | bond_send_gratuitous_arp(bond); |
01f3109d | 1179 | |
305d552a BH |
1180 | bond->send_unsol_na = bond->params.num_unsol_na; |
1181 | bond_send_unsolicited_na(bond); | |
1182 | ||
01f3109d OG |
1183 | write_unlock_bh(&bond->curr_slave_lock); |
1184 | read_unlock(&bond->lock); | |
1185 | ||
75c78500 | 1186 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1187 | |
1188 | read_lock(&bond->lock); | |
1189 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1190 | } |
c3ade5ca | 1191 | } |
a2fd940f AG |
1192 | |
1193 | /* resend IGMP joins since all were sent on curr_active_slave */ | |
1194 | if (bond->params.mode == BOND_MODE_ROUNDROBIN) { | |
1195 | bond_resend_igmp_join_requests(bond); | |
1196 | } | |
1da177e4 LT |
1197 | } |
1198 | ||
1199 | /** | |
1200 | * bond_select_active_slave - select a new active slave, if needed | |
1201 | * @bond: our bonding struct | |
1202 | * | |
3d632c3f | 1203 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1204 | * - The old curr_active_slave has been released or lost its link. |
1205 | * - The primary_slave has got its link back. | |
1206 | * - A slave has got its link back and there's no old curr_active_slave. | |
1207 | * | |
3915c1e8 | 1208 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1209 | */ |
a77b5325 | 1210 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1211 | { |
1212 | struct slave *best_slave; | |
ff59c456 | 1213 | int rv; |
1da177e4 LT |
1214 | |
1215 | best_slave = bond_find_best_slave(bond); | |
1216 | if (best_slave != bond->curr_active_slave) { | |
1217 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1218 | rv = bond_set_carrier(bond); |
1219 | if (!rv) | |
1220 | return; | |
1221 | ||
1222 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1223 | pr_info("%s: first active interface up!\n", |
1224 | bond->dev->name); | |
ff59c456 | 1225 | } else { |
a4aee5c8 JP |
1226 | pr_info("%s: now running without any active interface !\n", |
1227 | bond->dev->name); | |
ff59c456 | 1228 | } |
1da177e4 LT |
1229 | } |
1230 | } | |
1231 | ||
1232 | /*--------------------------- slave list handling ---------------------------*/ | |
1233 | ||
1234 | /* | |
1235 | * This function attaches the slave to the end of list. | |
1236 | * | |
1237 | * bond->lock held for writing by caller. | |
1238 | */ | |
1239 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1240 | { | |
1241 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1242 | new_slave->next = new_slave; | |
1243 | new_slave->prev = new_slave; | |
1244 | bond->first_slave = new_slave; | |
1245 | } else { | |
1246 | new_slave->next = bond->first_slave; | |
1247 | new_slave->prev = bond->first_slave->prev; | |
1248 | new_slave->next->prev = new_slave; | |
1249 | new_slave->prev->next = new_slave; | |
1250 | } | |
1251 | ||
1252 | bond->slave_cnt++; | |
1253 | } | |
1254 | ||
1255 | /* | |
1256 | * This function detaches the slave from the list. | |
1257 | * WARNING: no check is made to verify if the slave effectively | |
1258 | * belongs to <bond>. | |
1259 | * Nothing is freed on return, structures are just unchained. | |
1260 | * If any slave pointer in bond was pointing to <slave>, | |
1261 | * it should be changed by the calling function. | |
1262 | * | |
1263 | * bond->lock held for writing by caller. | |
1264 | */ | |
1265 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1266 | { | |
3d632c3f | 1267 | if (slave->next) |
1da177e4 | 1268 | slave->next->prev = slave->prev; |
1da177e4 | 1269 | |
3d632c3f | 1270 | if (slave->prev) |
1da177e4 | 1271 | slave->prev->next = slave->next; |
1da177e4 LT |
1272 | |
1273 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1274 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1275 | bond->first_slave = slave->next; | |
1276 | } else { | |
1277 | bond->first_slave = NULL; /* slave was the last one */ | |
1278 | } | |
1279 | } | |
1280 | ||
1281 | slave->next = NULL; | |
1282 | slave->prev = NULL; | |
1283 | bond->slave_cnt--; | |
1284 | } | |
1285 | ||
f6dc31a8 WC |
1286 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1287 | /* | |
1288 | * You must hold read lock on bond->lock before calling this. | |
1289 | */ | |
1290 | static bool slaves_support_netpoll(struct net_device *bond_dev) | |
1291 | { | |
1292 | struct bonding *bond = netdev_priv(bond_dev); | |
1293 | struct slave *slave; | |
1294 | int i = 0; | |
1295 | bool ret = true; | |
1296 | ||
1297 | bond_for_each_slave(bond, slave, i) { | |
1298 | if ((slave->dev->priv_flags & IFF_DISABLE_NETPOLL) || | |
1299 | !slave->dev->netdev_ops->ndo_poll_controller) | |
1300 | ret = false; | |
1301 | } | |
1302 | return i != 0 && ret; | |
1303 | } | |
1304 | ||
1305 | static void bond_poll_controller(struct net_device *bond_dev) | |
1306 | { | |
1307 | struct net_device *dev = bond_dev->npinfo->netpoll->real_dev; | |
1308 | if (dev != bond_dev) | |
1309 | netpoll_poll_dev(dev); | |
1310 | } | |
1311 | ||
1312 | static void bond_netpoll_cleanup(struct net_device *bond_dev) | |
1313 | { | |
1314 | struct bonding *bond = netdev_priv(bond_dev); | |
1315 | struct slave *slave; | |
1316 | const struct net_device_ops *ops; | |
1317 | int i; | |
1318 | ||
1319 | read_lock(&bond->lock); | |
1320 | bond_dev->npinfo = NULL; | |
1321 | bond_for_each_slave(bond, slave, i) { | |
1322 | if (slave->dev) { | |
1323 | ops = slave->dev->netdev_ops; | |
1324 | if (ops->ndo_netpoll_cleanup) | |
1325 | ops->ndo_netpoll_cleanup(slave->dev); | |
1326 | else | |
1327 | slave->dev->npinfo = NULL; | |
1328 | } | |
1329 | } | |
1330 | read_unlock(&bond->lock); | |
1331 | } | |
1332 | ||
1333 | #else | |
1334 | ||
1335 | static void bond_netpoll_cleanup(struct net_device *bond_dev) | |
1336 | { | |
1337 | } | |
1338 | ||
1339 | #endif | |
1340 | ||
1da177e4 LT |
1341 | /*---------------------------------- IOCTL ----------------------------------*/ |
1342 | ||
4ad072c9 AB |
1343 | static int bond_sethwaddr(struct net_device *bond_dev, |
1344 | struct net_device *slave_dev) | |
1da177e4 | 1345 | { |
5a03cdb7 HE |
1346 | pr_debug("bond_dev=%p\n", bond_dev); |
1347 | pr_debug("slave_dev=%p\n", slave_dev); | |
1348 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1349 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1350 | return 0; | |
1351 | } | |
1352 | ||
7f353bf2 HX |
1353 | #define BOND_VLAN_FEATURES \ |
1354 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1355 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1356 | |
3d632c3f | 1357 | /* |
8e3babcd | 1358 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1359 | * feature bits are managed elsewhere, so preserve those feature bits |
1360 | * on the master device. | |
8531c5ff AK |
1361 | */ |
1362 | static int bond_compute_features(struct bonding *bond) | |
1363 | { | |
8531c5ff AK |
1364 | struct slave *slave; |
1365 | struct net_device *bond_dev = bond->dev; | |
7f353bf2 | 1366 | unsigned long features = bond_dev->features; |
278339a4 | 1367 | unsigned long vlan_features = 0; |
3158bf7d MS |
1368 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1369 | bond_dev->hard_header_len); | |
8e3babcd | 1370 | int i; |
8531c5ff | 1371 | |
7f353bf2 | 1372 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
b63365a2 HX |
1373 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM; |
1374 | ||
1375 | if (!bond->first_slave) | |
1376 | goto done; | |
1377 | ||
1378 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1379 | |
278339a4 | 1380 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1381 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1382 | features = netdev_increment_features(features, |
1383 | slave->dev->features, | |
1384 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1385 | vlan_features = netdev_increment_features(vlan_features, |
1386 | slave->dev->vlan_features, | |
1387 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1388 | if (slave->dev->hard_header_len > max_hard_header_len) |
1389 | max_hard_header_len = slave->dev->hard_header_len; | |
1390 | } | |
8531c5ff | 1391 | |
b63365a2 | 1392 | done: |
7f353bf2 | 1393 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
b63365a2 | 1394 | bond_dev->features = netdev_fix_features(features, NULL); |
278339a4 | 1395 | bond_dev->vlan_features = netdev_fix_features(vlan_features, NULL); |
54ef3137 | 1396 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1397 | |
1398 | return 0; | |
1399 | } | |
1400 | ||
872254dd MS |
1401 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1402 | struct net_device *slave_dev) | |
1403 | { | |
454d7c9b | 1404 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1405 | |
00829823 | 1406 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1407 | |
1408 | bond_dev->type = slave_dev->type; | |
1409 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1410 | bond_dev->addr_len = slave_dev->addr_len; | |
1411 | ||
1412 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1413 | slave_dev->addr_len); | |
d90a162a | 1414 | bond->setup_by_slave = 1; |
872254dd MS |
1415 | } |
1416 | ||
1da177e4 | 1417 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1418 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1419 | { |
454d7c9b | 1420 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1421 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1422 | struct slave *new_slave = NULL; |
22bedad3 | 1423 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1424 | struct sockaddr addr; |
1425 | int link_reporting; | |
1426 | int old_features = bond_dev->features; | |
1427 | int res = 0; | |
1428 | ||
552709d5 | 1429 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1430 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1431 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1432 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1433 | } |
1434 | ||
1435 | /* bond must be initialized by bond_open() before enslaving */ | |
1436 | if (!(bond_dev->flags & IFF_UP)) { | |
a4aee5c8 JP |
1437 | pr_warning("%s: master_dev is not up in bond_enslave\n", |
1438 | bond_dev->name); | |
1da177e4 LT |
1439 | } |
1440 | ||
1441 | /* already enslaved */ | |
1442 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1443 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1444 | return -EBUSY; |
1445 | } | |
1446 | ||
1447 | /* vlan challenged mutual exclusion */ | |
1448 | /* no need to lock since we're protected by rtnl_lock */ | |
1449 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1450 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
f35188fa | 1451 | if (bond->vlgrp) { |
a4aee5c8 JP |
1452 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1453 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1454 | return -EPERM; |
1455 | } else { | |
a4aee5c8 JP |
1456 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1457 | bond_dev->name, slave_dev->name, | |
1458 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1459 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1460 | } | |
1461 | } else { | |
5a03cdb7 | 1462 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1463 | if (bond->slave_cnt == 0) { |
1464 | /* First slave, and it is not VLAN challenged, | |
1465 | * so remove the block of adding VLANs over the bond. | |
1466 | */ | |
1467 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1468 | } | |
1469 | } | |
1470 | ||
217df670 JV |
1471 | /* |
1472 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1473 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1474 | * the current ifenslave will set the interface down prior to |
1475 | * enslaving it; the old ifenslave will not. | |
1476 | */ | |
1477 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1478 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1479 | slave_dev->name); |
1480 | res = -EPERM; | |
1481 | goto err_undo_flags; | |
1482 | } | |
1da177e4 | 1483 | |
872254dd MS |
1484 | /* set bonding device ether type by slave - bonding netdevices are |
1485 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1486 | * there is a need to override some of the type dependent attribs/funcs. | |
1487 | * | |
1488 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1489 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1490 | */ | |
1491 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1492 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1493 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1494 | bond_dev->name, |
1495 | bond_dev->type, slave_dev->type); | |
75c78500 | 1496 | |
3ca5b404 JP |
1497 | res = netdev_bonding_change(bond_dev, |
1498 | NETDEV_PRE_TYPE_CHANGE); | |
1499 | res = notifier_to_errno(res); | |
1500 | if (res) { | |
1501 | pr_err("%s: refused to change device type\n", | |
1502 | bond_dev->name); | |
1503 | res = -EBUSY; | |
1504 | goto err_undo_flags; | |
1505 | } | |
75c78500 | 1506 | |
32a806c1 | 1507 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1508 | dev_uc_flush(bond_dev); |
22bedad3 | 1509 | dev_mc_flush(bond_dev); |
32a806c1 | 1510 | |
e36b9d16 MS |
1511 | if (slave_dev->type != ARPHRD_ETHER) |
1512 | bond_setup_by_slave(bond_dev, slave_dev); | |
1513 | else | |
1514 | ether_setup(bond_dev); | |
75c78500 | 1515 | |
93d9b7d7 JP |
1516 | netdev_bonding_change(bond_dev, |
1517 | NETDEV_POST_TYPE_CHANGE); | |
e36b9d16 | 1518 | } |
872254dd | 1519 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1520 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1521 | slave_dev->name, | |
1522 | slave_dev->type, bond_dev->type); | |
1523 | res = -EINVAL; | |
1524 | goto err_undo_flags; | |
872254dd MS |
1525 | } |
1526 | ||
eb7cc59a | 1527 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1528 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1529 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1530 | bond_dev->name); | |
3915c1e8 JV |
1531 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1532 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1533 | 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", |
1534 | bond_dev->name); | |
2ab82852 MS |
1535 | res = -EOPNOTSUPP; |
1536 | goto err_undo_flags; | |
1537 | } | |
1da177e4 LT |
1538 | } |
1539 | ||
c20811a7 JP |
1540 | /* If this is the first slave, then we need to set the master's hardware |
1541 | * address to be the same as the slave's. */ | |
1542 | if (bond->slave_cnt == 0) | |
1543 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, | |
1544 | slave_dev->addr_len); | |
1545 | ||
1546 | ||
243cb4e5 | 1547 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1548 | if (!new_slave) { |
1549 | res = -ENOMEM; | |
1550 | goto err_undo_flags; | |
1551 | } | |
1552 | ||
bb1d9123 AG |
1553 | /* |
1554 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1555 | * is set via sysfs or module option if desired. | |
1556 | */ | |
1557 | new_slave->queue_id = 0; | |
1558 | ||
b15ba0fb JP |
1559 | /* Save slave's original mtu and then set it to match the bond */ |
1560 | new_slave->original_mtu = slave_dev->mtu; | |
1561 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1562 | if (res) { | |
1563 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1564 | goto err_free; | |
1565 | } | |
1566 | ||
217df670 JV |
1567 | /* |
1568 | * Save slave's original ("permanent") mac address for modes | |
1569 | * that need it, and for restoring it upon release, and then | |
1570 | * set it to the master's address | |
1571 | */ | |
1572 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1573 | |
dd957c57 | 1574 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1575 | /* |
1576 | * Set slave to master's mac address. The application already | |
1577 | * set the master's mac address to that of the first slave | |
1578 | */ | |
1579 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1580 | addr.sa_family = slave_dev->type; | |
1581 | res = dev_set_mac_address(slave_dev, &addr); | |
1582 | if (res) { | |
5a03cdb7 | 1583 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1584 | goto err_restore_mtu; |
2ab82852 | 1585 | } |
217df670 | 1586 | } |
1da177e4 | 1587 | |
c2edacf8 JV |
1588 | res = netdev_set_master(slave_dev, bond_dev); |
1589 | if (res) { | |
5a03cdb7 | 1590 | pr_debug("Error %d calling netdev_set_master\n", res); |
569f0c4d | 1591 | goto err_restore_mac; |
c2edacf8 | 1592 | } |
217df670 JV |
1593 | /* open the slave since the application closed it */ |
1594 | res = dev_open(slave_dev); | |
1595 | if (res) { | |
3d632c3f | 1596 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
569f0c4d | 1597 | goto err_unset_master; |
1da177e4 LT |
1598 | } |
1599 | ||
1da177e4 | 1600 | new_slave->dev = slave_dev; |
0b680e75 | 1601 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1602 | |
58402054 | 1603 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1604 | /* bond_alb_init_slave() must be called before all other stages since |
1605 | * it might fail and we do not want to have to undo everything | |
1606 | */ | |
1607 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1608 | if (res) |
569f0c4d | 1609 | goto err_close; |
1da177e4 LT |
1610 | } |
1611 | ||
1612 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1613 | * master's promisc (and mc) settings only if it becomes the | |
1614 | * curr_active_slave, and that is taken care of later when calling | |
1615 | * bond_change_active() | |
1616 | */ | |
1617 | if (!USES_PRIMARY(bond->params.mode)) { | |
1618 | /* set promiscuity level to new slave */ | |
1619 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1620 | res = dev_set_promiscuity(slave_dev, 1); |
1621 | if (res) | |
1622 | goto err_close; | |
1da177e4 LT |
1623 | } |
1624 | ||
1625 | /* set allmulti level to new slave */ | |
1626 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1627 | res = dev_set_allmulti(slave_dev, 1); |
1628 | if (res) | |
1629 | goto err_close; | |
1da177e4 LT |
1630 | } |
1631 | ||
b9e40857 | 1632 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1633 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1634 | netdev_for_each_mc_addr(ha, bond_dev) |
1635 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1636 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1637 | } |
1638 | ||
1639 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1640 | /* add lacpdu mc addr to mc list */ | |
1641 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1642 | ||
22bedad3 | 1643 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1644 | } |
1645 | ||
1646 | bond_add_vlans_on_slave(bond, slave_dev); | |
1647 | ||
1648 | write_lock_bh(&bond->lock); | |
1649 | ||
1650 | bond_attach_slave(bond, new_slave); | |
1651 | ||
1652 | new_slave->delay = 0; | |
1653 | new_slave->link_failure_count = 0; | |
1654 | ||
8531c5ff AK |
1655 | bond_compute_features(bond); |
1656 | ||
3915c1e8 JV |
1657 | write_unlock_bh(&bond->lock); |
1658 | ||
1659 | read_lock(&bond->lock); | |
1660 | ||
f5b2b966 JV |
1661 | new_slave->last_arp_rx = jiffies; |
1662 | ||
1da177e4 LT |
1663 | if (bond->params.miimon && !bond->params.use_carrier) { |
1664 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1665 | ||
1666 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1667 | /* | |
1668 | * miimon is set but a bonded network driver | |
1669 | * does not support ETHTOOL/MII and | |
1670 | * arp_interval is not set. Note: if | |
1671 | * use_carrier is enabled, we will never go | |
1672 | * here (because netif_carrier is always | |
1673 | * supported); thus, we don't need to change | |
1674 | * the messages for netif_carrier. | |
1675 | */ | |
a4aee5c8 | 1676 | 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 | 1677 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1678 | } else if (link_reporting == -1) { |
1679 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1680 | 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", |
1681 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1682 | } |
1683 | } | |
1684 | ||
1685 | /* check for initial state */ | |
1686 | if (!bond->params.miimon || | |
1687 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1688 | if (bond->params.updelay) { | |
a4aee5c8 | 1689 | pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n"); |
1da177e4 LT |
1690 | new_slave->link = BOND_LINK_BACK; |
1691 | new_slave->delay = bond->params.updelay; | |
1692 | } else { | |
a4aee5c8 | 1693 | pr_debug("Initial state of slave_dev is BOND_LINK_UP\n"); |
1da177e4 LT |
1694 | new_slave->link = BOND_LINK_UP; |
1695 | } | |
1696 | new_slave->jiffies = jiffies; | |
1697 | } else { | |
a4aee5c8 | 1698 | pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n"); |
1da177e4 LT |
1699 | new_slave->link = BOND_LINK_DOWN; |
1700 | } | |
1701 | ||
1702 | if (bond_update_speed_duplex(new_slave) && | |
1703 | (new_slave->link != BOND_LINK_DOWN)) { | |
a4aee5c8 JP |
1704 | pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n", |
1705 | bond_dev->name, new_slave->dev->name); | |
1da177e4 LT |
1706 | |
1707 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a4aee5c8 JP |
1708 | pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n", |
1709 | bond_dev->name); | |
1da177e4 LT |
1710 | } |
1711 | } | |
1712 | ||
1713 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1714 | /* if there is a primary slave, remember it */ | |
a549952a | 1715 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1716 | bond->primary_slave = new_slave; |
a549952a JP |
1717 | bond->force_primary = true; |
1718 | } | |
1da177e4 LT |
1719 | } |
1720 | ||
3915c1e8 JV |
1721 | write_lock_bh(&bond->curr_slave_lock); |
1722 | ||
1da177e4 LT |
1723 | switch (bond->params.mode) { |
1724 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1725 | bond_set_slave_inactive_flags(new_slave); |
1726 | bond_select_active_slave(bond); | |
1da177e4 LT |
1727 | break; |
1728 | case BOND_MODE_8023AD: | |
1729 | /* in 802.3ad mode, the internal mechanism | |
1730 | * will activate the slaves in the selected | |
1731 | * aggregator | |
1732 | */ | |
1733 | bond_set_slave_inactive_flags(new_slave); | |
1734 | /* if this is the first slave */ | |
1735 | if (bond->slave_cnt == 1) { | |
1736 | SLAVE_AD_INFO(new_slave).id = 1; | |
1737 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1738 | * can be called only after the mac address of the bond is set | |
1739 | */ | |
1740 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1741 | bond->params.lacp_fast); | |
1742 | } else { | |
1743 | SLAVE_AD_INFO(new_slave).id = | |
1744 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1745 | } | |
1746 | ||
1747 | bond_3ad_bind_slave(new_slave); | |
1748 | break; | |
1749 | case BOND_MODE_TLB: | |
1750 | case BOND_MODE_ALB: | |
1751 | new_slave->state = BOND_STATE_ACTIVE; | |
059fe7a5 | 1752 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1753 | bond_select_active_slave(bond); |
1da177e4 LT |
1754 | break; |
1755 | default: | |
5a03cdb7 | 1756 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1757 | |
1758 | /* always active in trunk mode */ | |
1759 | new_slave->state = BOND_STATE_ACTIVE; | |
1760 | ||
1761 | /* In trunking mode there is little meaning to curr_active_slave | |
1762 | * anyway (it holds no special properties of the bond device), | |
1763 | * so we can change it without calling change_active_interface() | |
1764 | */ | |
3d632c3f | 1765 | if (!bond->curr_active_slave) |
1da177e4 | 1766 | bond->curr_active_slave = new_slave; |
3d632c3f | 1767 | |
1da177e4 LT |
1768 | break; |
1769 | } /* switch(bond_mode) */ | |
1770 | ||
3915c1e8 JV |
1771 | write_unlock_bh(&bond->curr_slave_lock); |
1772 | ||
ff59c456 JV |
1773 | bond_set_carrier(bond); |
1774 | ||
f6dc31a8 | 1775 | #ifdef CONFIG_NET_POLL_CONTROLLER |
c22d7ac8 AG |
1776 | /* |
1777 | * Netpoll and bonding is broken, make sure it is not initialized | |
1778 | * until it is fixed. | |
1779 | */ | |
1780 | if (disable_netpoll) { | |
f6dc31a8 | 1781 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
c22d7ac8 AG |
1782 | } else { |
1783 | if (slaves_support_netpoll(bond_dev)) { | |
1784 | bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL; | |
1785 | if (bond_dev->npinfo) | |
1786 | slave_dev->npinfo = bond_dev->npinfo; | |
1787 | } else if (!(bond_dev->priv_flags & IFF_DISABLE_NETPOLL)) { | |
1788 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; | |
1789 | pr_info("New slave device %s does not support netpoll\n", | |
1790 | slave_dev->name); | |
1791 | pr_info("Disabling netpoll support for %s\n", bond_dev->name); | |
1792 | } | |
f6dc31a8 WC |
1793 | } |
1794 | #endif | |
3915c1e8 | 1795 | read_unlock(&bond->lock); |
1da177e4 | 1796 | |
b76cdba9 MW |
1797 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1798 | if (res) | |
569f0c4d | 1799 | goto err_close; |
b76cdba9 | 1800 | |
a4aee5c8 JP |
1801 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1802 | bond_dev->name, slave_dev->name, | |
1803 | new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup", | |
1804 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); | |
1da177e4 LT |
1805 | |
1806 | /* enslave is successful */ | |
1807 | return 0; | |
1808 | ||
1809 | /* Undo stages on error */ | |
1da177e4 LT |
1810 | err_close: |
1811 | dev_close(slave_dev); | |
1812 | ||
569f0c4d JV |
1813 | err_unset_master: |
1814 | netdev_set_master(slave_dev, NULL); | |
1815 | ||
1da177e4 | 1816 | err_restore_mac: |
dd957c57 | 1817 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1818 | /* XXX TODO - fom follow mode needs to change master's |
1819 | * MAC if this slave's MAC is in use by the bond, or at | |
1820 | * least print a warning. | |
1821 | */ | |
2ab82852 MS |
1822 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1823 | addr.sa_family = slave_dev->type; | |
1824 | dev_set_mac_address(slave_dev, &addr); | |
1825 | } | |
1da177e4 | 1826 | |
b15ba0fb JP |
1827 | err_restore_mtu: |
1828 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1829 | ||
1da177e4 LT |
1830 | err_free: |
1831 | kfree(new_slave); | |
1832 | ||
1833 | err_undo_flags: | |
1834 | bond_dev->features = old_features; | |
3d632c3f | 1835 | |
1da177e4 LT |
1836 | return res; |
1837 | } | |
1838 | ||
1839 | /* | |
1840 | * Try to release the slave device <slave> from the bond device <master> | |
1841 | * It is legal to access curr_active_slave without a lock because all the function | |
1842 | * is write-locked. | |
1843 | * | |
1844 | * The rules for slave state should be: | |
1845 | * for Active/Backup: | |
1846 | * Active stays on all backups go down | |
1847 | * for Bonded connections: | |
1848 | * The first up interface should be left on and all others downed. | |
1849 | */ | |
a77b5325 | 1850 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1851 | { |
454d7c9b | 1852 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1853 | struct slave *slave, *oldcurrent; |
1854 | struct sockaddr addr; | |
1da177e4 LT |
1855 | |
1856 | /* slave is not a slave or master is not master of this slave */ | |
1857 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1858 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1859 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1860 | bond_dev->name, slave_dev->name); |
1861 | return -EINVAL; | |
1862 | } | |
1863 | ||
f6dc31a8 | 1864 | netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE); |
1da177e4 LT |
1865 | write_lock_bh(&bond->lock); |
1866 | ||
1867 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1868 | if (!slave) { | |
1869 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1870 | pr_info("%s: %s not enslaved\n", |
1871 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1872 | write_unlock_bh(&bond->lock); |
1da177e4 LT |
1873 | return -EINVAL; |
1874 | } | |
1875 | ||
3915c1e8 | 1876 | if (!bond->params.fail_over_mac) { |
8e95a202 JP |
1877 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) && |
1878 | bond->slave_cnt > 1) | |
a4aee5c8 JP |
1879 | 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", |
1880 | bond_dev->name, slave_dev->name, | |
1881 | slave->perm_hwaddr, | |
1882 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1883 | } |
1884 | ||
1885 | /* Inform AD package of unbinding of slave. */ | |
1886 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1887 | /* must be called before the slave is | |
1888 | * detached from the list | |
1889 | */ | |
1890 | bond_3ad_unbind_slave(slave); | |
1891 | } | |
1892 | ||
a4aee5c8 JP |
1893 | pr_info("%s: releasing %s interface %s\n", |
1894 | bond_dev->name, | |
1895 | (slave->state == BOND_STATE_ACTIVE) ? "active" : "backup", | |
1896 | slave_dev->name); | |
1da177e4 LT |
1897 | |
1898 | oldcurrent = bond->curr_active_slave; | |
1899 | ||
1900 | bond->current_arp_slave = NULL; | |
1901 | ||
1902 | /* release the slave from its bond */ | |
1903 | bond_detach_slave(bond, slave); | |
1904 | ||
8531c5ff AK |
1905 | bond_compute_features(bond); |
1906 | ||
3d632c3f | 1907 | if (bond->primary_slave == slave) |
1da177e4 | 1908 | bond->primary_slave = NULL; |
1da177e4 | 1909 | |
3d632c3f | 1910 | if (oldcurrent == slave) |
1da177e4 | 1911 | bond_change_active_slave(bond, NULL); |
1da177e4 | 1912 | |
58402054 | 1913 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1914 | /* Must be called only after the slave has been |
1915 | * detached from the list and the curr_active_slave | |
1916 | * has been cleared (if our_slave == old_current), | |
1917 | * but before a new active slave is selected. | |
1918 | */ | |
2543331d | 1919 | write_unlock_bh(&bond->lock); |
1da177e4 | 1920 | bond_alb_deinit_slave(bond, slave); |
2543331d | 1921 | write_lock_bh(&bond->lock); |
1da177e4 LT |
1922 | } |
1923 | ||
059fe7a5 JV |
1924 | if (oldcurrent == slave) { |
1925 | /* | |
1926 | * Note that we hold RTNL over this sequence, so there | |
1927 | * is no concern that another slave add/remove event | |
1928 | * will interfere. | |
1929 | */ | |
1930 | write_unlock_bh(&bond->lock); | |
1931 | read_lock(&bond->lock); | |
1932 | write_lock_bh(&bond->curr_slave_lock); | |
1933 | ||
1da177e4 LT |
1934 | bond_select_active_slave(bond); |
1935 | ||
059fe7a5 JV |
1936 | write_unlock_bh(&bond->curr_slave_lock); |
1937 | read_unlock(&bond->lock); | |
1938 | write_lock_bh(&bond->lock); | |
1939 | } | |
1940 | ||
1da177e4 | 1941 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
1942 | bond_set_carrier(bond); |
1943 | ||
1da177e4 LT |
1944 | /* if the last slave was removed, zero the mac address |
1945 | * of the master so it will be set by the application | |
1946 | * to the mac address of the first slave | |
1947 | */ | |
1948 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
1949 | ||
f35188fa | 1950 | if (!bond->vlgrp) { |
1da177e4 LT |
1951 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1952 | } else { | |
a4aee5c8 JP |
1953 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
1954 | bond_dev->name, bond_dev->name); | |
1955 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
1956 | bond_dev->name); | |
1da177e4 LT |
1957 | } |
1958 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
1959 | !bond_has_challenged_slaves(bond)) { | |
a4aee5c8 JP |
1960 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", |
1961 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1962 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
1963 | } | |
1964 | ||
1965 | write_unlock_bh(&bond->lock); | |
1966 | ||
b76cdba9 MW |
1967 | /* must do this from outside any spinlocks */ |
1968 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1969 | ||
1da177e4 LT |
1970 | bond_del_vlans_from_slave(bond, slave_dev); |
1971 | ||
1972 | /* If the mode USES_PRIMARY, then we should only remove its | |
1973 | * promisc and mc settings if it was the curr_active_slave, but that was | |
1974 | * already taken care of above when we detached the slave | |
1975 | */ | |
1976 | if (!USES_PRIMARY(bond->params.mode)) { | |
1977 | /* unset promiscuity level from slave */ | |
3d632c3f | 1978 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 1979 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
1980 | |
1981 | /* unset allmulti level from slave */ | |
3d632c3f | 1982 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 1983 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
1984 | |
1985 | /* flush master's mc_list from slave */ | |
b9e40857 | 1986 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1987 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 1988 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1989 | } |
1990 | ||
1991 | netdev_set_master(slave_dev, NULL); | |
1992 | ||
f6dc31a8 WC |
1993 | #ifdef CONFIG_NET_POLL_CONTROLLER |
1994 | read_lock_bh(&bond->lock); | |
c22d7ac8 AG |
1995 | |
1996 | /* Make sure netpoll over stays disabled until fixed. */ | |
1997 | if (!disable_netpoll) | |
1998 | if (slaves_support_netpoll(bond_dev)) | |
1999 | bond_dev->priv_flags &= ~IFF_DISABLE_NETPOLL; | |
f6dc31a8 WC |
2000 | read_unlock_bh(&bond->lock); |
2001 | if (slave_dev->netdev_ops->ndo_netpoll_cleanup) | |
2002 | slave_dev->netdev_ops->ndo_netpoll_cleanup(slave_dev); | |
2003 | else | |
2004 | slave_dev->npinfo = NULL; | |
2005 | #endif | |
2006 | ||
1da177e4 LT |
2007 | /* close slave before restoring its mac address */ |
2008 | dev_close(slave_dev); | |
2009 | ||
3915c1e8 | 2010 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2011 | /* restore original ("permanent") mac address */ |
2012 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2013 | addr.sa_family = slave_dev->type; | |
2014 | dev_set_mac_address(slave_dev, &addr); | |
2015 | } | |
1da177e4 | 2016 | |
b15ba0fb JP |
2017 | dev_set_mtu(slave_dev, slave->original_mtu); |
2018 | ||
8f903c70 | 2019 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
f5b2b966 JV |
2020 | IFF_SLAVE_INACTIVE | IFF_BONDING | |
2021 | IFF_SLAVE_NEEDARP); | |
1da177e4 LT |
2022 | |
2023 | kfree(slave); | |
2024 | ||
2025 | return 0; /* deletion OK */ | |
2026 | } | |
2027 | ||
d90a162a | 2028 | /* |
3d632c3f | 2029 | * First release a slave and than destroy the bond if no more slaves are left. |
d90a162a MS |
2030 | * Must be under rtnl_lock when this function is called. |
2031 | */ | |
3d632c3f SH |
2032 | int bond_release_and_destroy(struct net_device *bond_dev, |
2033 | struct net_device *slave_dev) | |
d90a162a | 2034 | { |
454d7c9b | 2035 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2036 | int ret; |
2037 | ||
2038 | ret = bond_release(bond_dev, slave_dev); | |
2039 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
a4aee5c8 JP |
2040 | pr_info("%s: destroying bond %s.\n", |
2041 | bond_dev->name, bond_dev->name); | |
9e71626c | 2042 | unregister_netdevice(bond_dev); |
d90a162a MS |
2043 | } |
2044 | return ret; | |
2045 | } | |
2046 | ||
1da177e4 LT |
2047 | /* |
2048 | * This function releases all slaves. | |
2049 | */ | |
2050 | static int bond_release_all(struct net_device *bond_dev) | |
2051 | { | |
454d7c9b | 2052 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2053 | struct slave *slave; |
2054 | struct net_device *slave_dev; | |
2055 | struct sockaddr addr; | |
2056 | ||
2057 | write_lock_bh(&bond->lock); | |
2058 | ||
ff59c456 JV |
2059 | netif_carrier_off(bond_dev); |
2060 | ||
3d632c3f | 2061 | if (bond->slave_cnt == 0) |
1da177e4 | 2062 | goto out; |
1da177e4 LT |
2063 | |
2064 | bond->current_arp_slave = NULL; | |
2065 | bond->primary_slave = NULL; | |
2066 | bond_change_active_slave(bond, NULL); | |
2067 | ||
2068 | while ((slave = bond->first_slave) != NULL) { | |
2069 | /* Inform AD package of unbinding of slave | |
2070 | * before slave is detached from the list. | |
2071 | */ | |
3d632c3f | 2072 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2073 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2074 | |
2075 | slave_dev = slave->dev; | |
2076 | bond_detach_slave(bond, slave); | |
2077 | ||
2543331d JV |
2078 | /* now that the slave is detached, unlock and perform |
2079 | * all the undo steps that should not be called from | |
2080 | * within a lock. | |
2081 | */ | |
2082 | write_unlock_bh(&bond->lock); | |
2083 | ||
58402054 | 2084 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2085 | /* must be called only after the slave |
2086 | * has been detached from the list | |
2087 | */ | |
2088 | bond_alb_deinit_slave(bond, slave); | |
2089 | } | |
2090 | ||
8531c5ff AK |
2091 | bond_compute_features(bond); |
2092 | ||
b76cdba9 | 2093 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2094 | bond_del_vlans_from_slave(bond, slave_dev); |
2095 | ||
2096 | /* If the mode USES_PRIMARY, then we should only remove its | |
2097 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2098 | * already taken care of above when we detached the slave | |
2099 | */ | |
2100 | if (!USES_PRIMARY(bond->params.mode)) { | |
2101 | /* unset promiscuity level from slave */ | |
3d632c3f | 2102 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2103 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2104 | |
2105 | /* unset allmulti level from slave */ | |
3d632c3f | 2106 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2107 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2108 | |
2109 | /* flush master's mc_list from slave */ | |
b9e40857 | 2110 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2111 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2112 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2113 | } |
2114 | ||
2115 | netdev_set_master(slave_dev, NULL); | |
2116 | ||
2117 | /* close slave before restoring its mac address */ | |
2118 | dev_close(slave_dev); | |
2119 | ||
dd957c57 | 2120 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2121 | /* restore original ("permanent") mac address*/ |
2122 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2123 | addr.sa_family = slave_dev->type; | |
2124 | dev_set_mac_address(slave_dev, &addr); | |
2125 | } | |
1da177e4 | 2126 | |
8f903c70 JV |
2127 | slave_dev->priv_flags &= ~(IFF_MASTER_8023AD | IFF_MASTER_ALB | |
2128 | IFF_SLAVE_INACTIVE); | |
1da177e4 LT |
2129 | |
2130 | kfree(slave); | |
2131 | ||
2132 | /* re-acquire the lock before getting the next slave */ | |
2133 | write_lock_bh(&bond->lock); | |
2134 | } | |
2135 | ||
2136 | /* zero the mac address of the master so it will be | |
2137 | * set by the application to the mac address of the | |
2138 | * first slave | |
2139 | */ | |
2140 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2141 | ||
f35188fa | 2142 | if (!bond->vlgrp) { |
1da177e4 | 2143 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
f35188fa | 2144 | } else { |
a4aee5c8 JP |
2145 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2146 | bond_dev->name, bond_dev->name); | |
2147 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2148 | bond_dev->name); | |
1da177e4 LT |
2149 | } |
2150 | ||
a4aee5c8 | 2151 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2152 | |
2153 | out: | |
2154 | write_unlock_bh(&bond->lock); | |
2155 | ||
2156 | return 0; | |
2157 | } | |
2158 | ||
2159 | /* | |
2160 | * This function changes the active slave to slave <slave_dev>. | |
2161 | * It returns -EINVAL in the following cases. | |
2162 | * - <slave_dev> is not found in the list. | |
2163 | * - There is not active slave now. | |
2164 | * - <slave_dev> is already active. | |
2165 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2166 | * - <slave_dev> is not running. | |
3d632c3f SH |
2167 | * In these cases, this function does nothing. |
2168 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2169 | */ |
2170 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2171 | { | |
454d7c9b | 2172 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2173 | struct slave *old_active = NULL; |
2174 | struct slave *new_active = NULL; | |
2175 | int res = 0; | |
2176 | ||
3d632c3f | 2177 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2178 | return -EINVAL; |
1da177e4 LT |
2179 | |
2180 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2181 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2182 | return -EINVAL; |
1da177e4 | 2183 | |
059fe7a5 | 2184 | read_lock(&bond->lock); |
1da177e4 | 2185 | |
059fe7a5 | 2186 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2187 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2188 | read_unlock(&bond->curr_slave_lock); |
2189 | ||
1da177e4 LT |
2190 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2191 | ||
2192 | /* | |
2193 | * Changing to the current active: do nothing; return success. | |
2194 | */ | |
2195 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2196 | read_unlock(&bond->lock); |
1da177e4 LT |
2197 | return 0; |
2198 | } | |
2199 | ||
2200 | if ((new_active) && | |
2201 | (old_active) && | |
2202 | (new_active->link == BOND_LINK_UP) && | |
2203 | IS_UP(new_active->dev)) { | |
059fe7a5 | 2204 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2205 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2206 | write_unlock_bh(&bond->curr_slave_lock); |
3d632c3f | 2207 | } else |
1da177e4 | 2208 | res = -EINVAL; |
1da177e4 | 2209 | |
059fe7a5 | 2210 | read_unlock(&bond->lock); |
1da177e4 LT |
2211 | |
2212 | return res; | |
2213 | } | |
2214 | ||
1da177e4 LT |
2215 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2216 | { | |
454d7c9b | 2217 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2218 | |
2219 | info->bond_mode = bond->params.mode; | |
2220 | info->miimon = bond->params.miimon; | |
2221 | ||
6603a6f2 | 2222 | read_lock(&bond->lock); |
1da177e4 | 2223 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2224 | read_unlock(&bond->lock); |
1da177e4 LT |
2225 | |
2226 | return 0; | |
2227 | } | |
2228 | ||
2229 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2230 | { | |
454d7c9b | 2231 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2232 | struct slave *slave; |
689c96cc | 2233 | int i, res = -ENODEV; |
1da177e4 | 2234 | |
6603a6f2 | 2235 | read_lock(&bond->lock); |
1da177e4 LT |
2236 | |
2237 | bond_for_each_slave(bond, slave, i) { | |
2238 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2239 | res = 0; |
2240 | strcpy(info->slave_name, slave->dev->name); | |
2241 | info->link = slave->link; | |
2242 | info->state = slave->state; | |
2243 | info->link_failure_count = slave->link_failure_count; | |
1da177e4 LT |
2244 | break; |
2245 | } | |
2246 | } | |
2247 | ||
6603a6f2 | 2248 | read_unlock(&bond->lock); |
1da177e4 | 2249 | |
689c96cc | 2250 | return res; |
1da177e4 LT |
2251 | } |
2252 | ||
2253 | /*-------------------------------- Monitoring -------------------------------*/ | |
2254 | ||
1da177e4 | 2255 | |
f0c76d61 JV |
2256 | static int bond_miimon_inspect(struct bonding *bond) |
2257 | { | |
2258 | struct slave *slave; | |
2259 | int i, link_state, commit = 0; | |
41f89100 JP |
2260 | bool ignore_updelay; |
2261 | ||
2262 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2263 | |
2264 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2265 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2266 | |
f0c76d61 | 2267 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2268 | |
2269 | switch (slave->link) { | |
f0c76d61 JV |
2270 | case BOND_LINK_UP: |
2271 | if (link_state) | |
2272 | continue; | |
1da177e4 | 2273 | |
f0c76d61 JV |
2274 | slave->link = BOND_LINK_FAIL; |
2275 | slave->delay = bond->params.downdelay; | |
2276 | if (slave->delay) { | |
a4aee5c8 JP |
2277 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2278 | bond->dev->name, | |
2279 | (bond->params.mode == | |
2280 | BOND_MODE_ACTIVEBACKUP) ? | |
2281 | ((slave->state == BOND_STATE_ACTIVE) ? | |
2282 | "active " : "backup ") : "", | |
2283 | slave->dev->name, | |
2284 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2285 | } |
f0c76d61 JV |
2286 | /*FALLTHRU*/ |
2287 | case BOND_LINK_FAIL: | |
2288 | if (link_state) { | |
2289 | /* | |
2290 | * recovered before downdelay expired | |
2291 | */ | |
2292 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2293 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2294 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2295 | bond->dev->name, | |
2296 | (bond->params.downdelay - slave->delay) * | |
2297 | bond->params.miimon, | |
2298 | slave->dev->name); | |
f0c76d61 | 2299 | continue; |
1da177e4 | 2300 | } |
f0c76d61 JV |
2301 | |
2302 | if (slave->delay <= 0) { | |
2303 | slave->new_link = BOND_LINK_DOWN; | |
2304 | commit++; | |
2305 | continue; | |
1da177e4 | 2306 | } |
1da177e4 | 2307 | |
f0c76d61 JV |
2308 | slave->delay--; |
2309 | break; | |
2310 | ||
2311 | case BOND_LINK_DOWN: | |
2312 | if (!link_state) | |
2313 | continue; | |
2314 | ||
2315 | slave->link = BOND_LINK_BACK; | |
2316 | slave->delay = bond->params.updelay; | |
2317 | ||
2318 | if (slave->delay) { | |
a4aee5c8 JP |
2319 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2320 | bond->dev->name, slave->dev->name, | |
2321 | ignore_updelay ? 0 : | |
2322 | bond->params.updelay * | |
2323 | bond->params.miimon); | |
f0c76d61 JV |
2324 | } |
2325 | /*FALLTHRU*/ | |
2326 | case BOND_LINK_BACK: | |
2327 | if (!link_state) { | |
2328 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2329 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2330 | bond->dev->name, | |
2331 | (bond->params.updelay - slave->delay) * | |
2332 | bond->params.miimon, | |
2333 | slave->dev->name); | |
f0c76d61 JV |
2334 | |
2335 | continue; | |
2336 | } | |
2337 | ||
41f89100 JP |
2338 | if (ignore_updelay) |
2339 | slave->delay = 0; | |
2340 | ||
f0c76d61 JV |
2341 | if (slave->delay <= 0) { |
2342 | slave->new_link = BOND_LINK_UP; | |
2343 | commit++; | |
41f89100 | 2344 | ignore_updelay = false; |
f0c76d61 | 2345 | continue; |
1da177e4 | 2346 | } |
f0c76d61 JV |
2347 | |
2348 | slave->delay--; | |
1da177e4 | 2349 | break; |
f0c76d61 JV |
2350 | } |
2351 | } | |
1da177e4 | 2352 | |
f0c76d61 JV |
2353 | return commit; |
2354 | } | |
1da177e4 | 2355 | |
f0c76d61 JV |
2356 | static void bond_miimon_commit(struct bonding *bond) |
2357 | { | |
2358 | struct slave *slave; | |
2359 | int i; | |
2360 | ||
2361 | bond_for_each_slave(bond, slave, i) { | |
2362 | switch (slave->new_link) { | |
2363 | case BOND_LINK_NOCHANGE: | |
2364 | continue; | |
1da177e4 | 2365 | |
f0c76d61 JV |
2366 | case BOND_LINK_UP: |
2367 | slave->link = BOND_LINK_UP; | |
2368 | slave->jiffies = jiffies; | |
2369 | ||
2370 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2371 | /* prevent it from being the active one */ | |
2372 | slave->state = BOND_STATE_BACKUP; | |
2373 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { | |
2374 | /* make it immediately active */ | |
2375 | slave->state = BOND_STATE_ACTIVE; | |
2376 | } else if (slave != bond->primary_slave) { | |
2377 | /* prevent it from being the active one */ | |
2378 | slave->state = BOND_STATE_BACKUP; | |
1da177e4 | 2379 | } |
1da177e4 | 2380 | |
a4aee5c8 JP |
2381 | pr_info("%s: link status definitely up for interface %s.\n", |
2382 | bond->dev->name, slave->dev->name); | |
1da177e4 | 2383 | |
f0c76d61 JV |
2384 | /* notify ad that the link status has changed */ |
2385 | if (bond->params.mode == BOND_MODE_8023AD) | |
2386 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2387 | |
58402054 | 2388 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2389 | bond_alb_handle_link_change(bond, slave, |
2390 | BOND_LINK_UP); | |
1da177e4 | 2391 | |
f0c76d61 JV |
2392 | if (!bond->curr_active_slave || |
2393 | (slave == bond->primary_slave)) | |
2394 | goto do_failover; | |
1da177e4 | 2395 | |
f0c76d61 | 2396 | continue; |
059fe7a5 | 2397 | |
f0c76d61 | 2398 | case BOND_LINK_DOWN: |
fba4acda JV |
2399 | if (slave->link_failure_count < UINT_MAX) |
2400 | slave->link_failure_count++; | |
2401 | ||
f0c76d61 | 2402 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2403 | |
f0c76d61 JV |
2404 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2405 | bond->params.mode == BOND_MODE_8023AD) | |
2406 | bond_set_slave_inactive_flags(slave); | |
2407 | ||
a4aee5c8 JP |
2408 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2409 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2410 | |
2411 | if (bond->params.mode == BOND_MODE_8023AD) | |
2412 | bond_3ad_handle_link_change(slave, | |
2413 | BOND_LINK_DOWN); | |
2414 | ||
ae63e808 | 2415 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2416 | bond_alb_handle_link_change(bond, slave, |
2417 | BOND_LINK_DOWN); | |
2418 | ||
2419 | if (slave == bond->curr_active_slave) | |
2420 | goto do_failover; | |
2421 | ||
2422 | continue; | |
2423 | ||
2424 | default: | |
a4aee5c8 | 2425 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2426 | bond->dev->name, slave->new_link, |
2427 | slave->dev->name); | |
2428 | slave->new_link = BOND_LINK_NOCHANGE; | |
2429 | ||
2430 | continue; | |
2431 | } | |
2432 | ||
2433 | do_failover: | |
2434 | ASSERT_RTNL(); | |
2435 | write_lock_bh(&bond->curr_slave_lock); | |
2436 | bond_select_active_slave(bond); | |
2437 | write_unlock_bh(&bond->curr_slave_lock); | |
2438 | } | |
2439 | ||
2440 | bond_set_carrier(bond); | |
1da177e4 LT |
2441 | } |
2442 | ||
0b0eef66 JV |
2443 | /* |
2444 | * bond_mii_monitor | |
2445 | * | |
2446 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2447 | * inspection, then (if inspection indicates something needs to be done) |
2448 | * an acquisition of appropriate locks followed by a commit phase to | |
2449 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2450 | */ |
2451 | void bond_mii_monitor(struct work_struct *work) | |
2452 | { | |
2453 | struct bonding *bond = container_of(work, struct bonding, | |
2454 | mii_work.work); | |
0b0eef66 JV |
2455 | |
2456 | read_lock(&bond->lock); | |
f0c76d61 JV |
2457 | if (bond->kill_timers) |
2458 | goto out; | |
2459 | ||
2460 | if (bond->slave_cnt == 0) | |
2461 | goto re_arm; | |
b59f9f74 JV |
2462 | |
2463 | if (bond->send_grat_arp) { | |
2464 | read_lock(&bond->curr_slave_lock); | |
2465 | bond_send_gratuitous_arp(bond); | |
2466 | read_unlock(&bond->curr_slave_lock); | |
2467 | } | |
2468 | ||
305d552a BH |
2469 | if (bond->send_unsol_na) { |
2470 | read_lock(&bond->curr_slave_lock); | |
2471 | bond_send_unsolicited_na(bond); | |
2472 | read_unlock(&bond->curr_slave_lock); | |
2473 | } | |
2474 | ||
f0c76d61 | 2475 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2476 | read_unlock(&bond->lock); |
2477 | rtnl_lock(); | |
2478 | read_lock(&bond->lock); | |
f0c76d61 JV |
2479 | |
2480 | bond_miimon_commit(bond); | |
2481 | ||
5655662d JV |
2482 | read_unlock(&bond->lock); |
2483 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2484 | read_lock(&bond->lock); | |
0b0eef66 JV |
2485 | } |
2486 | ||
f0c76d61 JV |
2487 | re_arm: |
2488 | if (bond->params.miimon) | |
2489 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2490 | msecs_to_jiffies(bond->params.miimon)); | |
2491 | out: | |
0b0eef66 | 2492 | read_unlock(&bond->lock); |
0b0eef66 | 2493 | } |
c3ade5ca | 2494 | |
d3bb52b0 | 2495 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2496 | { |
2497 | struct in_device *idev; | |
2498 | struct in_ifaddr *ifa; | |
a144ea4b | 2499 | __be32 addr = 0; |
c3ade5ca JV |
2500 | |
2501 | if (!dev) | |
2502 | return 0; | |
2503 | ||
2504 | rcu_read_lock(); | |
e5ed6399 | 2505 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2506 | if (!idev) |
2507 | goto out; | |
2508 | ||
2509 | ifa = idev->ifa_list; | |
2510 | if (!ifa) | |
2511 | goto out; | |
2512 | ||
2513 | addr = ifa->ifa_local; | |
2514 | out: | |
2515 | rcu_read_unlock(); | |
2516 | return addr; | |
2517 | } | |
2518 | ||
d3bb52b0 | 2519 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2520 | { |
0883beca | 2521 | struct vlan_entry *vlan; |
f5b2b966 JV |
2522 | |
2523 | if (ip == bond->master_ip) | |
2524 | return 1; | |
2525 | ||
0883beca | 2526 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2527 | if (ip == vlan->vlan_ip) |
2528 | return 1; | |
2529 | } | |
2530 | ||
2531 | return 0; | |
2532 | } | |
2533 | ||
c3ade5ca JV |
2534 | /* |
2535 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2536 | * switches in VLAN mode (especially if ports are configured as | |
2537 | * "native" to a VLAN) might not pass non-tagged frames. | |
2538 | */ | |
d3bb52b0 | 2539 | 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 |
2540 | { |
2541 | struct sk_buff *skb; | |
2542 | ||
5a03cdb7 | 2543 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2544 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2545 | |
c3ade5ca JV |
2546 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2547 | NULL, slave_dev->dev_addr, NULL); | |
2548 | ||
2549 | if (!skb) { | |
a4aee5c8 | 2550 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2551 | return; |
2552 | } | |
2553 | if (vlan_id) { | |
2554 | skb = vlan_put_tag(skb, vlan_id); | |
2555 | if (!skb) { | |
a4aee5c8 | 2556 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2557 | return; |
2558 | } | |
2559 | } | |
2560 | arp_xmit(skb); | |
2561 | } | |
2562 | ||
2563 | ||
1da177e4 LT |
2564 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2565 | { | |
c3ade5ca | 2566 | int i, vlan_id, rv; |
d3bb52b0 | 2567 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2568 | struct vlan_entry *vlan; |
c3ade5ca JV |
2569 | struct net_device *vlan_dev; |
2570 | struct flowi fl; | |
2571 | struct rtable *rt; | |
1da177e4 | 2572 | |
6b780567 MW |
2573 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2574 | if (!targets[i]) | |
5a31bec0 | 2575 | break; |
5a03cdb7 | 2576 | pr_debug("basa: target %x\n", targets[i]); |
f35188fa | 2577 | if (!bond->vlgrp) { |
5a03cdb7 | 2578 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2579 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2580 | bond->master_ip, 0); | |
2581 | continue; | |
2582 | } | |
2583 | ||
2584 | /* | |
2585 | * If VLANs are configured, we do a route lookup to | |
2586 | * determine which VLAN interface would be used, so we | |
2587 | * can tag the ARP with the proper VLAN tag. | |
2588 | */ | |
2589 | memset(&fl, 0, sizeof(fl)); | |
2590 | fl.fl4_dst = targets[i]; | |
2591 | fl.fl4_tos = RTO_ONLINK; | |
2592 | ||
ec87fd3b | 2593 | rv = ip_route_output_key(dev_net(bond->dev), &rt, &fl); |
c3ade5ca JV |
2594 | if (rv) { |
2595 | if (net_ratelimit()) { | |
a4aee5c8 JP |
2596 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
2597 | bond->dev->name, &fl.fl4_dst); | |
c3ade5ca JV |
2598 | } |
2599 | continue; | |
2600 | } | |
2601 | ||
2602 | /* | |
2603 | * This target is not on a VLAN | |
2604 | */ | |
d8d1f30b | 2605 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2606 | ip_rt_put(rt); |
5a03cdb7 | 2607 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2608 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2609 | bond->master_ip, 0); | |
2610 | continue; | |
2611 | } | |
2612 | ||
2613 | vlan_id = 0; | |
0883beca | 2614 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2615 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
d8d1f30b | 2616 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2617 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2618 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2619 | vlan_dev->name, vlan_id); |
2620 | break; | |
2621 | } | |
2622 | } | |
2623 | ||
2624 | if (vlan_id) { | |
ed4b9f80 | 2625 | ip_rt_put(rt); |
c3ade5ca JV |
2626 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2627 | vlan->vlan_ip, vlan_id); | |
2628 | continue; | |
2629 | } | |
2630 | ||
2631 | if (net_ratelimit()) { | |
a4aee5c8 JP |
2632 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
2633 | bond->dev->name, &fl.fl4_dst, | |
d8d1f30b | 2634 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2635 | } |
ed4b9f80 | 2636 | ip_rt_put(rt); |
c3ade5ca JV |
2637 | } |
2638 | } | |
2639 | ||
2640 | /* | |
2641 | * Kick out a gratuitous ARP for an IP on the bonding master plus one | |
2642 | * for each VLAN above us. | |
b59f9f74 JV |
2643 | * |
2644 | * Caller must hold curr_slave_lock for read or better | |
c3ade5ca JV |
2645 | */ |
2646 | static void bond_send_gratuitous_arp(struct bonding *bond) | |
2647 | { | |
2648 | struct slave *slave = bond->curr_active_slave; | |
2649 | struct vlan_entry *vlan; | |
2650 | struct net_device *vlan_dev; | |
2651 | ||
a4aee5c8 JP |
2652 | pr_debug("bond_send_grat_arp: bond %s slave %s\n", |
2653 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
b59f9f74 JV |
2654 | |
2655 | if (!slave || !bond->send_grat_arp || | |
2656 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
c3ade5ca JV |
2657 | return; |
2658 | ||
b59f9f74 JV |
2659 | bond->send_grat_arp--; |
2660 | ||
c3ade5ca JV |
2661 | if (bond->master_ip) { |
2662 | bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip, | |
1053f62c | 2663 | bond->master_ip, 0); |
c3ade5ca JV |
2664 | } |
2665 | ||
f35188fa JV |
2666 | if (!bond->vlgrp) |
2667 | return; | |
2668 | ||
c3ade5ca | 2669 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2670 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2671 | if (vlan->vlan_ip) { |
2672 | bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip, | |
2673 | vlan->vlan_ip, vlan->vlan_id); | |
2674 | } | |
1da177e4 LT |
2675 | } |
2676 | } | |
2677 | ||
d3bb52b0 | 2678 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2679 | { |
2680 | int i; | |
d3bb52b0 | 2681 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2682 | |
f5b2b966 | 2683 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2684 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2685 | &sip, &tip, i, &targets[i], |
2686 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2687 | if (sip == targets[i]) { |
2688 | if (bond_has_this_ip(bond, tip)) | |
2689 | slave->last_arp_rx = jiffies; | |
2690 | return; | |
2691 | } | |
2692 | } | |
2693 | } | |
2694 | ||
2695 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2696 | { | |
2697 | struct arphdr *arp; | |
2698 | struct slave *slave; | |
2699 | struct bonding *bond; | |
2700 | unsigned char *arp_ptr; | |
d3bb52b0 | 2701 | __be32 sip, tip; |
f5b2b966 | 2702 | |
1f3c8804 AG |
2703 | if (dev->priv_flags & IFF_802_1Q_VLAN) { |
2704 | /* | |
2705 | * When using VLANS and bonding, dev and oriv_dev may be | |
2706 | * incorrect if the physical interface supports VLAN | |
2707 | * acceleration. With this change ARP validation now | |
2708 | * works for hosts only reachable on the VLAN interface. | |
2709 | */ | |
2710 | dev = vlan_dev_real_dev(dev); | |
2711 | orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif); | |
2712 | } | |
2713 | ||
f5b2b966 JV |
2714 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2715 | goto out; | |
2716 | ||
454d7c9b | 2717 | bond = netdev_priv(dev); |
f5b2b966 JV |
2718 | read_lock(&bond->lock); |
2719 | ||
5a03cdb7 | 2720 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
a4aee5c8 JP |
2721 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2722 | orig_dev ? orig_dev->name : "NULL"); | |
f5b2b966 JV |
2723 | |
2724 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2725 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2726 | goto out_unlock; | |
2727 | ||
988b7050 | 2728 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2729 | goto out_unlock; |
2730 | ||
d0a92be0 | 2731 | arp = arp_hdr(skb); |
f5b2b966 JV |
2732 | if (arp->ar_hln != dev->addr_len || |
2733 | skb->pkt_type == PACKET_OTHERHOST || | |
2734 | skb->pkt_type == PACKET_LOOPBACK || | |
2735 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2736 | arp->ar_pro != htons(ETH_P_IP) || | |
2737 | arp->ar_pln != 4) | |
2738 | goto out_unlock; | |
2739 | ||
2740 | arp_ptr = (unsigned char *)(arp + 1); | |
2741 | arp_ptr += dev->addr_len; | |
2742 | memcpy(&sip, arp_ptr, 4); | |
2743 | arp_ptr += 4 + dev->addr_len; | |
2744 | memcpy(&tip, arp_ptr, 4); | |
2745 | ||
5a03cdb7 | 2746 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
a4aee5c8 JP |
2747 | bond->dev->name, slave->dev->name, slave->state, |
2748 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), | |
2749 | &sip, &tip); | |
f5b2b966 JV |
2750 | |
2751 | /* | |
2752 | * Backup slaves won't see the ARP reply, but do come through | |
2753 | * here for each ARP probe (so we swap the sip/tip to validate | |
2754 | * the probe). In a "redundant switch, common router" type of | |
2755 | * configuration, the ARP probe will (hopefully) travel from | |
2756 | * the active, through one switch, the router, then the other | |
2757 | * switch before reaching the backup. | |
2758 | */ | |
2759 | if (slave->state == BOND_STATE_ACTIVE) | |
2760 | bond_validate_arp(bond, slave, sip, tip); | |
2761 | else | |
2762 | bond_validate_arp(bond, slave, tip, sip); | |
2763 | ||
2764 | out_unlock: | |
2765 | read_unlock(&bond->lock); | |
2766 | out: | |
2767 | dev_kfree_skb(skb); | |
2768 | return NET_RX_SUCCESS; | |
2769 | } | |
2770 | ||
1da177e4 LT |
2771 | /* |
2772 | * this function is called regularly to monitor each slave's link | |
2773 | * ensuring that traffic is being sent and received when arp monitoring | |
2774 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2775 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2776 | * arp monitoring in active backup mode. | |
2777 | */ | |
1b76b316 | 2778 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2779 | { |
1b76b316 JV |
2780 | struct bonding *bond = container_of(work, struct bonding, |
2781 | arp_work.work); | |
1da177e4 LT |
2782 | struct slave *slave, *oldcurrent; |
2783 | int do_failover = 0; | |
2784 | int delta_in_ticks; | |
2785 | int i; | |
2786 | ||
2787 | read_lock(&bond->lock); | |
2788 | ||
5ce0da8f | 2789 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2790 | |
3d632c3f | 2791 | if (bond->kill_timers) |
1da177e4 | 2792 | goto out; |
1da177e4 | 2793 | |
3d632c3f | 2794 | if (bond->slave_cnt == 0) |
1da177e4 | 2795 | goto re_arm; |
1da177e4 LT |
2796 | |
2797 | read_lock(&bond->curr_slave_lock); | |
2798 | oldcurrent = bond->curr_active_slave; | |
2799 | read_unlock(&bond->curr_slave_lock); | |
2800 | ||
2801 | /* see if any of the previous devices are up now (i.e. they have | |
2802 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2803 | * the picture unless it is null. also, slave->jiffies is not needed | |
2804 | * here because we send an arp on each slave and give a slave as | |
2805 | * long as it needs to get the tx/rx within the delta. | |
2806 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2807 | * so it can wait | |
2808 | */ | |
2809 | bond_for_each_slave(bond, slave, i) { | |
2810 | if (slave->link != BOND_LINK_UP) { | |
9d21493b | 2811 | if (time_before_eq(jiffies, dev_trans_start(slave->dev) + delta_in_ticks) && |
b63bb739 | 2812 | time_before_eq(jiffies, slave->dev->last_rx + delta_in_ticks)) { |
1da177e4 LT |
2813 | |
2814 | slave->link = BOND_LINK_UP; | |
2815 | slave->state = BOND_STATE_ACTIVE; | |
2816 | ||
2817 | /* primary_slave has no meaning in round-robin | |
2818 | * mode. the window of a slave being up and | |
2819 | * curr_active_slave being null after enslaving | |
2820 | * is closed. | |
2821 | */ | |
2822 | if (!oldcurrent) { | |
a4aee5c8 JP |
2823 | pr_info("%s: link status definitely up for interface %s, ", |
2824 | bond->dev->name, | |
2825 | slave->dev->name); | |
1da177e4 LT |
2826 | do_failover = 1; |
2827 | } else { | |
a4aee5c8 JP |
2828 | pr_info("%s: interface %s is now up\n", |
2829 | bond->dev->name, | |
2830 | slave->dev->name); | |
1da177e4 LT |
2831 | } |
2832 | } | |
2833 | } else { | |
2834 | /* slave->link == BOND_LINK_UP */ | |
2835 | ||
2836 | /* not all switches will respond to an arp request | |
2837 | * when the source ip is 0, so don't take the link down | |
2838 | * if we don't know our ip yet | |
2839 | */ | |
9d21493b | 2840 | if (time_after_eq(jiffies, dev_trans_start(slave->dev) + 2*delta_in_ticks) || |
4b8a9239 | 2841 | (time_after_eq(jiffies, slave->dev->last_rx + 2*delta_in_ticks))) { |
1da177e4 LT |
2842 | |
2843 | slave->link = BOND_LINK_DOWN; | |
2844 | slave->state = BOND_STATE_BACKUP; | |
2845 | ||
3d632c3f | 2846 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2847 | slave->link_failure_count++; |
1da177e4 | 2848 | |
a4aee5c8 JP |
2849 | pr_info("%s: interface %s is now down.\n", |
2850 | bond->dev->name, | |
2851 | slave->dev->name); | |
1da177e4 | 2852 | |
3d632c3f | 2853 | if (slave == oldcurrent) |
1da177e4 | 2854 | do_failover = 1; |
1da177e4 LT |
2855 | } |
2856 | } | |
2857 | ||
2858 | /* note: if switch is in round-robin mode, all links | |
2859 | * must tx arp to ensure all links rx an arp - otherwise | |
2860 | * links may oscillate or not come up at all; if switch is | |
2861 | * in something like xor mode, there is nothing we can | |
2862 | * do - all replies will be rx'ed on same link causing slaves | |
2863 | * to be unstable during low/no traffic periods | |
2864 | */ | |
3d632c3f | 2865 | if (IS_UP(slave->dev)) |
1da177e4 | 2866 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2867 | } |
2868 | ||
2869 | if (do_failover) { | |
059fe7a5 | 2870 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2871 | |
2872 | bond_select_active_slave(bond); | |
2873 | ||
059fe7a5 | 2874 | write_unlock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2875 | } |
2876 | ||
2877 | re_arm: | |
1b76b316 JV |
2878 | if (bond->params.arp_interval) |
2879 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
2880 | out: |
2881 | read_unlock(&bond->lock); | |
2882 | } | |
2883 | ||
2884 | /* | |
b2220cad JV |
2885 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2886 | * changes. Sets new_link in slaves to specify what action should take | |
2887 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2888 | * to link states must be committed. | |
2889 | * | |
2890 | * Called with bond->lock held for read. | |
1da177e4 | 2891 | */ |
b2220cad | 2892 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2893 | { |
1da177e4 | 2894 | struct slave *slave; |
b2220cad | 2895 | int i, commit = 0; |
1da177e4 | 2896 | |
b2220cad JV |
2897 | bond_for_each_slave(bond, slave, i) { |
2898 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2899 | |
b2220cad JV |
2900 | if (slave->link != BOND_LINK_UP) { |
2901 | if (time_before_eq(jiffies, slave_last_rx(bond, slave) + | |
2902 | delta_in_ticks)) { | |
2903 | slave->new_link = BOND_LINK_UP; | |
2904 | commit++; | |
2905 | } | |
1da177e4 | 2906 | |
b2220cad JV |
2907 | continue; |
2908 | } | |
1da177e4 | 2909 | |
b2220cad JV |
2910 | /* |
2911 | * Give slaves 2*delta after being enslaved or made | |
2912 | * active. This avoids bouncing, as the last receive | |
2913 | * times need a full ARP monitor cycle to be updated. | |
2914 | */ | |
2915 | if (!time_after_eq(jiffies, slave->jiffies + | |
2916 | 2 * delta_in_ticks)) | |
2917 | continue; | |
2918 | ||
2919 | /* | |
2920 | * Backup slave is down if: | |
2921 | * - No current_arp_slave AND | |
2922 | * - more than 3*delta since last receive AND | |
2923 | * - the bond has an IP address | |
2924 | * | |
2925 | * Note: a non-null current_arp_slave indicates | |
2926 | * the curr_active_slave went down and we are | |
2927 | * searching for a new one; under this condition | |
2928 | * we only take the curr_active_slave down - this | |
2929 | * gives each slave a chance to tx/rx traffic | |
2930 | * before being taken out | |
2931 | */ | |
2932 | if (slave->state == BOND_STATE_BACKUP && | |
2933 | !bond->current_arp_slave && | |
2934 | time_after(jiffies, slave_last_rx(bond, slave) + | |
2935 | 3 * delta_in_ticks)) { | |
2936 | slave->new_link = BOND_LINK_DOWN; | |
2937 | commit++; | |
2938 | } | |
2939 | ||
2940 | /* | |
2941 | * Active slave is down if: | |
2942 | * - more than 2*delta since transmitting OR | |
2943 | * - (more than 2*delta since receive AND | |
2944 | * the bond has an IP address) | |
2945 | */ | |
2946 | if ((slave->state == BOND_STATE_ACTIVE) && | |
9d21493b | 2947 | (time_after_eq(jiffies, dev_trans_start(slave->dev) + |
b2220cad JV |
2948 | 2 * delta_in_ticks) || |
2949 | (time_after_eq(jiffies, slave_last_rx(bond, slave) | |
2950 | + 2 * delta_in_ticks)))) { | |
2951 | slave->new_link = BOND_LINK_DOWN; | |
2952 | commit++; | |
2953 | } | |
1da177e4 LT |
2954 | } |
2955 | ||
b2220cad JV |
2956 | return commit; |
2957 | } | |
1da177e4 | 2958 | |
b2220cad JV |
2959 | /* |
2960 | * Called to commit link state changes noted by inspection step of | |
2961 | * active-backup mode ARP monitor. | |
2962 | * | |
2963 | * Called with RTNL and bond->lock for read. | |
2964 | */ | |
2965 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2966 | { | |
2967 | struct slave *slave; | |
2968 | int i; | |
1da177e4 | 2969 | |
b2220cad JV |
2970 | bond_for_each_slave(bond, slave, i) { |
2971 | switch (slave->new_link) { | |
2972 | case BOND_LINK_NOCHANGE: | |
2973 | continue; | |
ff59c456 | 2974 | |
b2220cad | 2975 | case BOND_LINK_UP: |
b9f60253 JP |
2976 | if ((!bond->curr_active_slave && |
2977 | time_before_eq(jiffies, | |
2978 | dev_trans_start(slave->dev) + | |
2979 | delta_in_ticks)) || | |
2980 | bond->curr_active_slave != slave) { | |
b2220cad | 2981 | slave->link = BOND_LINK_UP; |
b2220cad JV |
2982 | bond->current_arp_slave = NULL; |
2983 | ||
a4aee5c8 | 2984 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2985 | bond->dev->name, slave->dev->name); |
b2220cad | 2986 | |
b9f60253 JP |
2987 | if (!bond->curr_active_slave || |
2988 | (slave == bond->primary_slave)) | |
2989 | goto do_failover; | |
1da177e4 | 2990 | |
b9f60253 | 2991 | } |
1da177e4 | 2992 | |
b9f60253 | 2993 | continue; |
1da177e4 | 2994 | |
b2220cad JV |
2995 | case BOND_LINK_DOWN: |
2996 | if (slave->link_failure_count < UINT_MAX) | |
2997 | slave->link_failure_count++; | |
2998 | ||
2999 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 3000 | bond_set_slave_inactive_flags(slave); |
b2220cad | 3001 | |
a4aee5c8 | 3002 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 3003 | bond->dev->name, slave->dev->name); |
b2220cad | 3004 | |
b9f60253 | 3005 | if (slave == bond->curr_active_slave) { |
b2220cad | 3006 | bond->current_arp_slave = NULL; |
b9f60253 | 3007 | goto do_failover; |
1da177e4 | 3008 | } |
b9f60253 JP |
3009 | |
3010 | continue; | |
b2220cad JV |
3011 | |
3012 | default: | |
a4aee5c8 | 3013 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
3014 | bond->dev->name, slave->new_link, |
3015 | slave->dev->name); | |
b9f60253 | 3016 | continue; |
1da177e4 | 3017 | } |
1da177e4 | 3018 | |
b9f60253 JP |
3019 | do_failover: |
3020 | ASSERT_RTNL(); | |
b2220cad | 3021 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3022 | bond_select_active_slave(bond); |
b2220cad JV |
3023 | write_unlock_bh(&bond->curr_slave_lock); |
3024 | } | |
1da177e4 | 3025 | |
b2220cad JV |
3026 | bond_set_carrier(bond); |
3027 | } | |
1da177e4 | 3028 | |
b2220cad JV |
3029 | /* |
3030 | * Send ARP probes for active-backup mode ARP monitor. | |
3031 | * | |
3032 | * Called with bond->lock held for read. | |
3033 | */ | |
3034 | static void bond_ab_arp_probe(struct bonding *bond) | |
3035 | { | |
3036 | struct slave *slave; | |
3037 | int i; | |
1da177e4 | 3038 | |
b2220cad | 3039 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3040 | |
b2220cad | 3041 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3042 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3043 | bond->current_arp_slave->dev->name, | |
3044 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3045 | |
b2220cad JV |
3046 | if (bond->curr_active_slave) { |
3047 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3048 | read_unlock(&bond->curr_slave_lock); | |
3049 | return; | |
3050 | } | |
1da177e4 | 3051 | |
b2220cad | 3052 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3053 | |
b2220cad JV |
3054 | /* if we don't have a curr_active_slave, search for the next available |
3055 | * backup slave from the current_arp_slave and make it the candidate | |
3056 | * for becoming the curr_active_slave | |
3057 | */ | |
1da177e4 | 3058 | |
b2220cad JV |
3059 | if (!bond->current_arp_slave) { |
3060 | bond->current_arp_slave = bond->first_slave; | |
3061 | if (!bond->current_arp_slave) | |
3062 | return; | |
3063 | } | |
1da177e4 | 3064 | |
b2220cad | 3065 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3066 | |
b2220cad JV |
3067 | /* search for next candidate */ |
3068 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3069 | if (IS_UP(slave->dev)) { | |
3070 | slave->link = BOND_LINK_BACK; | |
3071 | bond_set_slave_active_flags(slave); | |
3072 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3073 | slave->jiffies = jiffies; |
b2220cad JV |
3074 | bond->current_arp_slave = slave; |
3075 | break; | |
1da177e4 LT |
3076 | } |
3077 | ||
b2220cad JV |
3078 | /* if the link state is up at this point, we |
3079 | * mark it down - this can happen if we have | |
3080 | * simultaneous link failures and | |
3081 | * reselect_active_interface doesn't make this | |
3082 | * one the current slave so it is still marked | |
3083 | * up when it is actually down | |
1da177e4 | 3084 | */ |
b2220cad JV |
3085 | if (slave->link == BOND_LINK_UP) { |
3086 | slave->link = BOND_LINK_DOWN; | |
3087 | if (slave->link_failure_count < UINT_MAX) | |
3088 | slave->link_failure_count++; | |
1da177e4 | 3089 | |
b2220cad JV |
3090 | bond_set_slave_inactive_flags(slave); |
3091 | ||
a4aee5c8 JP |
3092 | pr_info("%s: backup interface %s is now down.\n", |
3093 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3094 | } |
b2220cad JV |
3095 | } |
3096 | } | |
1da177e4 | 3097 | |
b2220cad JV |
3098 | void bond_activebackup_arp_mon(struct work_struct *work) |
3099 | { | |
3100 | struct bonding *bond = container_of(work, struct bonding, | |
3101 | arp_work.work); | |
3102 | int delta_in_ticks; | |
1da177e4 | 3103 | |
b2220cad | 3104 | read_lock(&bond->lock); |
1da177e4 | 3105 | |
b2220cad JV |
3106 | if (bond->kill_timers) |
3107 | goto out; | |
1da177e4 | 3108 | |
b2220cad | 3109 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3110 | |
b2220cad JV |
3111 | if (bond->slave_cnt == 0) |
3112 | goto re_arm; | |
3113 | ||
b59f9f74 JV |
3114 | if (bond->send_grat_arp) { |
3115 | read_lock(&bond->curr_slave_lock); | |
3116 | bond_send_gratuitous_arp(bond); | |
3117 | read_unlock(&bond->curr_slave_lock); | |
3118 | } | |
3119 | ||
305d552a BH |
3120 | if (bond->send_unsol_na) { |
3121 | read_lock(&bond->curr_slave_lock); | |
3122 | bond_send_unsolicited_na(bond); | |
3123 | read_unlock(&bond->curr_slave_lock); | |
3124 | } | |
3125 | ||
b2220cad JV |
3126 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3127 | read_unlock(&bond->lock); | |
3128 | rtnl_lock(); | |
3129 | read_lock(&bond->lock); | |
3130 | ||
3131 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3132 | ||
3133 | read_unlock(&bond->lock); | |
3134 | rtnl_unlock(); | |
3135 | read_lock(&bond->lock); | |
1da177e4 LT |
3136 | } |
3137 | ||
b2220cad JV |
3138 | bond_ab_arp_probe(bond); |
3139 | ||
1da177e4 | 3140 | re_arm: |
3d632c3f | 3141 | if (bond->params.arp_interval) |
1b76b316 | 3142 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3143 | out: |
3144 | read_unlock(&bond->lock); | |
3145 | } | |
3146 | ||
3147 | /*------------------------------ proc/seq_file-------------------------------*/ | |
3148 | ||
3149 | #ifdef CONFIG_PROC_FS | |
3150 | ||
1da177e4 | 3151 | static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos) |
1f78d9f9 HE |
3152 | __acquires(&dev_base_lock) |
3153 | __acquires(&bond->lock) | |
1da177e4 LT |
3154 | { |
3155 | struct bonding *bond = seq->private; | |
3156 | loff_t off = 0; | |
3157 | struct slave *slave; | |
3158 | int i; | |
3159 | ||
3160 | /* make sure the bond won't be taken away */ | |
3161 | read_lock(&dev_base_lock); | |
6603a6f2 | 3162 | read_lock(&bond->lock); |
1da177e4 | 3163 | |
3d632c3f | 3164 | if (*pos == 0) |
1da177e4 | 3165 | return SEQ_START_TOKEN; |
1da177e4 LT |
3166 | |
3167 | bond_for_each_slave(bond, slave, i) { | |
3d632c3f | 3168 | if (++off == *pos) |
1da177e4 | 3169 | return slave; |
1da177e4 LT |
3170 | } |
3171 | ||
3172 | return NULL; | |
3173 | } | |
3174 | ||
3175 | static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos) | |
3176 | { | |
3177 | struct bonding *bond = seq->private; | |
3178 | struct slave *slave = v; | |
3179 | ||
3180 | ++*pos; | |
3d632c3f | 3181 | if (v == SEQ_START_TOKEN) |
1da177e4 | 3182 | return bond->first_slave; |
1da177e4 LT |
3183 | |
3184 | slave = slave->next; | |
3185 | ||
3186 | return (slave == bond->first_slave) ? NULL : slave; | |
3187 | } | |
3188 | ||
3189 | static void bond_info_seq_stop(struct seq_file *seq, void *v) | |
1f78d9f9 HE |
3190 | __releases(&bond->lock) |
3191 | __releases(&dev_base_lock) | |
1da177e4 LT |
3192 | { |
3193 | struct bonding *bond = seq->private; | |
3194 | ||
6603a6f2 | 3195 | read_unlock(&bond->lock); |
1da177e4 LT |
3196 | read_unlock(&dev_base_lock); |
3197 | } | |
3198 | ||
3199 | static void bond_info_show_master(struct seq_file *seq) | |
3200 | { | |
3201 | struct bonding *bond = seq->private; | |
3202 | struct slave *curr; | |
4756b02f | 3203 | int i; |
1da177e4 LT |
3204 | |
3205 | read_lock(&bond->curr_slave_lock); | |
3206 | curr = bond->curr_active_slave; | |
3207 | read_unlock(&bond->curr_slave_lock); | |
3208 | ||
dd957c57 | 3209 | seq_printf(seq, "Bonding Mode: %s", |
1da177e4 LT |
3210 | bond_mode_name(bond->params.mode)); |
3211 | ||
dd957c57 JV |
3212 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP && |
3213 | bond->params.fail_over_mac) | |
3915c1e8 JV |
3214 | seq_printf(seq, " (fail_over_mac %s)", |
3215 | fail_over_mac_tbl[bond->params.fail_over_mac].modename); | |
dd957c57 JV |
3216 | |
3217 | seq_printf(seq, "\n"); | |
3218 | ||
c61b75ad MW |
3219 | if (bond->params.mode == BOND_MODE_XOR || |
3220 | bond->params.mode == BOND_MODE_8023AD) { | |
3221 | seq_printf(seq, "Transmit Hash Policy: %s (%d)\n", | |
3222 | xmit_hashtype_tbl[bond->params.xmit_policy].modename, | |
3223 | bond->params.xmit_policy); | |
3224 | } | |
3225 | ||
1da177e4 | 3226 | if (USES_PRIMARY(bond->params.mode)) { |
a549952a | 3227 | seq_printf(seq, "Primary Slave: %s", |
0f418b2a MW |
3228 | (bond->primary_slave) ? |
3229 | bond->primary_slave->dev->name : "None"); | |
a549952a JP |
3230 | if (bond->primary_slave) |
3231 | seq_printf(seq, " (primary_reselect %s)", | |
3232 | pri_reselect_tbl[bond->params.primary_reselect].modename); | |
1da177e4 | 3233 | |
a549952a | 3234 | seq_printf(seq, "\nCurrently Active Slave: %s\n", |
1da177e4 LT |
3235 | (curr) ? curr->dev->name : "None"); |
3236 | } | |
3237 | ||
ff59c456 JV |
3238 | seq_printf(seq, "MII Status: %s\n", netif_carrier_ok(bond->dev) ? |
3239 | "up" : "down"); | |
1da177e4 LT |
3240 | seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon); |
3241 | seq_printf(seq, "Up Delay (ms): %d\n", | |
3242 | bond->params.updelay * bond->params.miimon); | |
3243 | seq_printf(seq, "Down Delay (ms): %d\n", | |
3244 | bond->params.downdelay * bond->params.miimon); | |
3245 | ||
4756b02f MW |
3246 | |
3247 | /* ARP information */ | |
3d632c3f SH |
3248 | if (bond->params.arp_interval > 0) { |
3249 | int printed = 0; | |
4756b02f MW |
3250 | seq_printf(seq, "ARP Polling Interval (ms): %d\n", |
3251 | bond->params.arp_interval); | |
3252 | ||
3253 | seq_printf(seq, "ARP IP target/s (n.n.n.n form):"); | |
3254 | ||
3d632c3f | 3255 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
4756b02f | 3256 | if (!bond->params.arp_targets[i]) |
5a31bec0 | 3257 | break; |
4756b02f MW |
3258 | if (printed) |
3259 | seq_printf(seq, ","); | |
8cf14e38 | 3260 | seq_printf(seq, " %pI4", &bond->params.arp_targets[i]); |
4756b02f MW |
3261 | printed = 1; |
3262 | } | |
3263 | seq_printf(seq, "\n"); | |
3264 | } | |
3265 | ||
1da177e4 LT |
3266 | if (bond->params.mode == BOND_MODE_8023AD) { |
3267 | struct ad_info ad_info; | |
3268 | ||
3269 | seq_puts(seq, "\n802.3ad info\n"); | |
3270 | seq_printf(seq, "LACP rate: %s\n", | |
3271 | (bond->params.lacp_fast) ? "fast" : "slow"); | |
fd989c83 JV |
3272 | seq_printf(seq, "Aggregator selection policy (ad_select): %s\n", |
3273 | ad_select_tbl[bond->params.ad_select].modename); | |
1da177e4 LT |
3274 | |
3275 | if (bond_3ad_get_active_agg_info(bond, &ad_info)) { | |
3276 | seq_printf(seq, "bond %s has no active aggregator\n", | |
3277 | bond->dev->name); | |
3278 | } else { | |
3279 | seq_printf(seq, "Active Aggregator Info:\n"); | |
3280 | ||
3281 | seq_printf(seq, "\tAggregator ID: %d\n", | |
3282 | ad_info.aggregator_id); | |
3283 | seq_printf(seq, "\tNumber of ports: %d\n", | |
3284 | ad_info.ports); | |
3285 | seq_printf(seq, "\tActor Key: %d\n", | |
3286 | ad_info.actor_key); | |
3287 | seq_printf(seq, "\tPartner Key: %d\n", | |
3288 | ad_info.partner_key); | |
e174961c JB |
3289 | seq_printf(seq, "\tPartner Mac Address: %pM\n", |
3290 | ad_info.partner_system); | |
1da177e4 LT |
3291 | } |
3292 | } | |
3293 | } | |
3294 | ||
3d632c3f SH |
3295 | static void bond_info_show_slave(struct seq_file *seq, |
3296 | const struct slave *slave) | |
1da177e4 LT |
3297 | { |
3298 | struct bonding *bond = seq->private; | |
3299 | ||
3300 | seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name); | |
3301 | seq_printf(seq, "MII Status: %s\n", | |
3302 | (slave->link == BOND_LINK_UP) ? "up" : "down"); | |
65509645 | 3303 | seq_printf(seq, "Link Failure Count: %u\n", |
1da177e4 LT |
3304 | slave->link_failure_count); |
3305 | ||
e174961c | 3306 | seq_printf(seq, "Permanent HW addr: %pM\n", slave->perm_hwaddr); |
1da177e4 LT |
3307 | |
3308 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3309 | const struct aggregator *agg | |
3310 | = SLAVE_AD_INFO(slave).port.aggregator; | |
3311 | ||
3d632c3f | 3312 | if (agg) |
1da177e4 LT |
3313 | seq_printf(seq, "Aggregator ID: %d\n", |
3314 | agg->aggregator_identifier); | |
3d632c3f | 3315 | else |
1da177e4 | 3316 | seq_puts(seq, "Aggregator ID: N/A\n"); |
1da177e4 | 3317 | } |
bb1d9123 | 3318 | seq_printf(seq, "Slave queue ID: %d\n", slave->queue_id); |
1da177e4 LT |
3319 | } |
3320 | ||
3321 | static int bond_info_seq_show(struct seq_file *seq, void *v) | |
3322 | { | |
3323 | if (v == SEQ_START_TOKEN) { | |
3324 | seq_printf(seq, "%s\n", version); | |
3325 | bond_info_show_master(seq); | |
3d632c3f | 3326 | } else |
1da177e4 | 3327 | bond_info_show_slave(seq, v); |
1da177e4 LT |
3328 | |
3329 | return 0; | |
3330 | } | |
3331 | ||
4101dec9 | 3332 | static const struct seq_operations bond_info_seq_ops = { |
1da177e4 LT |
3333 | .start = bond_info_seq_start, |
3334 | .next = bond_info_seq_next, | |
3335 | .stop = bond_info_seq_stop, | |
3336 | .show = bond_info_seq_show, | |
3337 | }; | |
3338 | ||
3339 | static int bond_info_open(struct inode *inode, struct file *file) | |
3340 | { | |
3341 | struct seq_file *seq; | |
3342 | struct proc_dir_entry *proc; | |
3343 | int res; | |
3344 | ||
3345 | res = seq_open(file, &bond_info_seq_ops); | |
3346 | if (!res) { | |
3347 | /* recover the pointer buried in proc_dir_entry data */ | |
3348 | seq = file->private_data; | |
3349 | proc = PDE(inode); | |
3350 | seq->private = proc->data; | |
3351 | } | |
3352 | ||
3353 | return res; | |
3354 | } | |
3355 | ||
d54b1fdb | 3356 | static const struct file_operations bond_info_fops = { |
1da177e4 LT |
3357 | .owner = THIS_MODULE, |
3358 | .open = bond_info_open, | |
3359 | .read = seq_read, | |
3360 | .llseek = seq_lseek, | |
3361 | .release = seq_release, | |
3362 | }; | |
3363 | ||
38fc0026 | 3364 | static void bond_create_proc_entry(struct bonding *bond) |
1da177e4 LT |
3365 | { |
3366 | struct net_device *bond_dev = bond->dev; | |
ec87fd3b | 3367 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
1da177e4 | 3368 | |
ec87fd3b | 3369 | if (bn->proc_dir) { |
a95609cb | 3370 | bond->proc_entry = proc_create_data(bond_dev->name, |
ec87fd3b | 3371 | S_IRUGO, bn->proc_dir, |
a95609cb | 3372 | &bond_info_fops, bond); |
3d632c3f | 3373 | if (bond->proc_entry == NULL) |
a4aee5c8 JP |
3374 | pr_warning("Warning: Cannot create /proc/net/%s/%s\n", |
3375 | DRV_NAME, bond_dev->name); | |
3d632c3f | 3376 | else |
1da177e4 | 3377 | memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ); |
1da177e4 | 3378 | } |
1da177e4 LT |
3379 | } |
3380 | ||
3381 | static void bond_remove_proc_entry(struct bonding *bond) | |
3382 | { | |
ec87fd3b EB |
3383 | struct net_device *bond_dev = bond->dev; |
3384 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); | |
3385 | ||
3386 | if (bn->proc_dir && bond->proc_entry) { | |
3387 | remove_proc_entry(bond->proc_file_name, bn->proc_dir); | |
1da177e4 LT |
3388 | memset(bond->proc_file_name, 0, IFNAMSIZ); |
3389 | bond->proc_entry = NULL; | |
3390 | } | |
3391 | } | |
3392 | ||
3393 | /* Create the bonding directory under /proc/net, if doesn't exist yet. | |
3394 | * Caller must hold rtnl_lock. | |
3395 | */ | |
2c8c1e72 | 3396 | static void __net_init bond_create_proc_dir(struct bond_net *bn) |
1da177e4 | 3397 | { |
ec87fd3b EB |
3398 | if (!bn->proc_dir) { |
3399 | bn->proc_dir = proc_mkdir(DRV_NAME, bn->net->proc_net); | |
3400 | if (!bn->proc_dir) | |
a4aee5c8 JP |
3401 | pr_warning("Warning: cannot create /proc/net/%s\n", |
3402 | DRV_NAME); | |
1da177e4 LT |
3403 | } |
3404 | } | |
3405 | ||
3406 | /* Destroy the bonding directory under /proc/net, if empty. | |
3407 | * Caller must hold rtnl_lock. | |
3408 | */ | |
2c8c1e72 | 3409 | static void __net_exit bond_destroy_proc_dir(struct bond_net *bn) |
1da177e4 | 3410 | { |
ec87fd3b EB |
3411 | if (bn->proc_dir) { |
3412 | remove_proc_entry(DRV_NAME, bn->net->proc_net); | |
3413 | bn->proc_dir = NULL; | |
1da177e4 LT |
3414 | } |
3415 | } | |
aee64faf JP |
3416 | |
3417 | #else /* !CONFIG_PROC_FS */ | |
3418 | ||
38fc0026 | 3419 | static void bond_create_proc_entry(struct bonding *bond) |
aee64faf JP |
3420 | { |
3421 | } | |
3422 | ||
3423 | static void bond_remove_proc_entry(struct bonding *bond) | |
3424 | { | |
3425 | } | |
3426 | ||
2c8c1e72 | 3427 | static inline void bond_create_proc_dir(struct bond_net *bn) |
aee64faf JP |
3428 | { |
3429 | } | |
3430 | ||
2c8c1e72 | 3431 | static inline void bond_destroy_proc_dir(struct bond_net *bn) |
aee64faf JP |
3432 | { |
3433 | } | |
3434 | ||
1da177e4 LT |
3435 | #endif /* CONFIG_PROC_FS */ |
3436 | ||
aee64faf | 3437 | |
1da177e4 LT |
3438 | /*-------------------------- netdev event handling --------------------------*/ |
3439 | ||
3440 | /* | |
3441 | * Change device name | |
3442 | */ | |
3443 | static int bond_event_changename(struct bonding *bond) | |
3444 | { | |
1da177e4 LT |
3445 | bond_remove_proc_entry(bond); |
3446 | bond_create_proc_entry(bond); | |
7e083840 | 3447 | |
1da177e4 LT |
3448 | return NOTIFY_DONE; |
3449 | } | |
3450 | ||
3d632c3f SH |
3451 | static int bond_master_netdev_event(unsigned long event, |
3452 | struct net_device *bond_dev) | |
1da177e4 | 3453 | { |
454d7c9b | 3454 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3455 | |
3456 | switch (event) { | |
3457 | case NETDEV_CHANGENAME: | |
3458 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3459 | default: |
3460 | break; | |
3461 | } | |
3462 | ||
3463 | return NOTIFY_DONE; | |
3464 | } | |
3465 | ||
3d632c3f SH |
3466 | static int bond_slave_netdev_event(unsigned long event, |
3467 | struct net_device *slave_dev) | |
1da177e4 LT |
3468 | { |
3469 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3470 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3471 | |
3472 | switch (event) { | |
3473 | case NETDEV_UNREGISTER: | |
3474 | if (bond_dev) { | |
1284cd3a JV |
3475 | if (bond->setup_by_slave) |
3476 | bond_release_and_destroy(bond_dev, slave_dev); | |
3477 | else | |
3478 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3479 | } |
3480 | break; | |
3481 | case NETDEV_CHANGE: | |
17d04500 JV |
3482 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3483 | struct slave *slave; | |
3484 | ||
3485 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3486 | if (slave) { | |
3487 | u16 old_speed = slave->speed; | |
3488 | u16 old_duplex = slave->duplex; | |
3489 | ||
3490 | bond_update_speed_duplex(slave); | |
3491 | ||
3492 | if (bond_is_lb(bond)) | |
3493 | break; | |
3494 | ||
3495 | if (old_speed != slave->speed) | |
3496 | bond_3ad_adapter_speed_changed(slave); | |
3497 | if (old_duplex != slave->duplex) | |
3498 | bond_3ad_adapter_duplex_changed(slave); | |
3499 | } | |
3500 | } | |
3501 | ||
1da177e4 LT |
3502 | break; |
3503 | case NETDEV_DOWN: | |
3504 | /* | |
3505 | * ... Or is it this? | |
3506 | */ | |
3507 | break; | |
3508 | case NETDEV_CHANGEMTU: | |
3509 | /* | |
3510 | * TODO: Should slaves be allowed to | |
3511 | * independently alter their MTU? For | |
3512 | * an active-backup bond, slaves need | |
3513 | * not be the same type of device, so | |
3514 | * MTUs may vary. For other modes, | |
3515 | * slaves arguably should have the | |
3516 | * same MTUs. To do this, we'd need to | |
3517 | * take over the slave's change_mtu | |
3518 | * function for the duration of their | |
3519 | * servitude. | |
3520 | */ | |
3521 | break; | |
3522 | case NETDEV_CHANGENAME: | |
3523 | /* | |
3524 | * TODO: handle changing the primary's name | |
3525 | */ | |
3526 | break; | |
8531c5ff AK |
3527 | case NETDEV_FEAT_CHANGE: |
3528 | bond_compute_features(bond); | |
3529 | break; | |
1da177e4 LT |
3530 | default: |
3531 | break; | |
3532 | } | |
3533 | ||
3534 | return NOTIFY_DONE; | |
3535 | } | |
3536 | ||
3537 | /* | |
3538 | * bond_netdev_event: handle netdev notifier chain events. | |
3539 | * | |
3540 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3541 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3542 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3543 | * dev_probe_lock). | |
3544 | */ | |
3d632c3f SH |
3545 | static int bond_netdev_event(struct notifier_block *this, |
3546 | unsigned long event, void *ptr) | |
1da177e4 LT |
3547 | { |
3548 | struct net_device *event_dev = (struct net_device *)ptr; | |
3549 | ||
5a03cdb7 | 3550 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3551 | event_dev ? event_dev->name : "None", |
3552 | event); | |
1da177e4 | 3553 | |
0b680e75 JV |
3554 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3555 | return NOTIFY_DONE; | |
3556 | ||
1da177e4 | 3557 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3558 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3559 | return bond_master_netdev_event(event, event_dev); |
3560 | } | |
3561 | ||
3562 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3563 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3564 | return bond_slave_netdev_event(event, event_dev); |
3565 | } | |
3566 | ||
3567 | return NOTIFY_DONE; | |
3568 | } | |
3569 | ||
c3ade5ca JV |
3570 | /* |
3571 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3572 | * | |
3573 | * We keep track of device IPs primarily to use as source addresses in | |
3574 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3575 | * | |
3576 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3577 | */ | |
3578 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3579 | { | |
3580 | struct in_ifaddr *ifa = ptr; | |
3581 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
ec87fd3b | 3582 | struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id); |
0883beca PE |
3583 | struct bonding *bond; |
3584 | struct vlan_entry *vlan; | |
c3ade5ca | 3585 | |
ec87fd3b | 3586 | list_for_each_entry(bond, &bn->dev_list, bond_list) { |
c3ade5ca JV |
3587 | if (bond->dev == event_dev) { |
3588 | switch (event) { | |
3589 | case NETDEV_UP: | |
3590 | bond->master_ip = ifa->ifa_local; | |
3591 | return NOTIFY_OK; | |
3592 | case NETDEV_DOWN: | |
3593 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3594 | return NOTIFY_OK; | |
3595 | default: | |
3596 | return NOTIFY_DONE; | |
3597 | } | |
3598 | } | |
3599 | ||
0883beca | 3600 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
3601 | if (!bond->vlgrp) |
3602 | continue; | |
5c15bdec | 3603 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3604 | if (vlan_dev == event_dev) { |
3605 | switch (event) { | |
3606 | case NETDEV_UP: | |
3607 | vlan->vlan_ip = ifa->ifa_local; | |
3608 | return NOTIFY_OK; | |
3609 | case NETDEV_DOWN: | |
3610 | vlan->vlan_ip = | |
3611 | bond_glean_dev_ip(vlan_dev); | |
3612 | return NOTIFY_OK; | |
3613 | default: | |
3614 | return NOTIFY_DONE; | |
3615 | } | |
3616 | } | |
3617 | } | |
3618 | } | |
3619 | return NOTIFY_DONE; | |
3620 | } | |
3621 | ||
1da177e4 LT |
3622 | static struct notifier_block bond_netdev_notifier = { |
3623 | .notifier_call = bond_netdev_event, | |
3624 | }; | |
3625 | ||
c3ade5ca JV |
3626 | static struct notifier_block bond_inetaddr_notifier = { |
3627 | .notifier_call = bond_inetaddr_event, | |
3628 | }; | |
3629 | ||
1da177e4 LT |
3630 | /*-------------------------- Packet type handling ---------------------------*/ |
3631 | ||
3632 | /* register to receive lacpdus on a bond */ | |
3633 | static void bond_register_lacpdu(struct bonding *bond) | |
3634 | { | |
3635 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3636 | ||
3637 | /* initialize packet type */ | |
3638 | pk_type->type = PKT_TYPE_LACPDU; | |
3639 | pk_type->dev = bond->dev; | |
3640 | pk_type->func = bond_3ad_lacpdu_recv; | |
3641 | ||
3642 | dev_add_pack(pk_type); | |
3643 | } | |
3644 | ||
3645 | /* unregister to receive lacpdus on a bond */ | |
3646 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3647 | { | |
3648 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3649 | } | |
3650 | ||
f5b2b966 JV |
3651 | void bond_register_arp(struct bonding *bond) |
3652 | { | |
3653 | struct packet_type *pt = &bond->arp_mon_pt; | |
3654 | ||
c4f283b1 JV |
3655 | if (pt->type) |
3656 | return; | |
3657 | ||
f5b2b966 | 3658 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3659 | pt->dev = bond->dev; |
f5b2b966 JV |
3660 | pt->func = bond_arp_rcv; |
3661 | dev_add_pack(pt); | |
3662 | } | |
3663 | ||
3664 | void bond_unregister_arp(struct bonding *bond) | |
3665 | { | |
c4f283b1 JV |
3666 | struct packet_type *pt = &bond->arp_mon_pt; |
3667 | ||
3668 | dev_remove_pack(pt); | |
3669 | pt->type = 0; | |
f5b2b966 JV |
3670 | } |
3671 | ||
169a3e66 JV |
3672 | /*---------------------------- Hashing Policies -----------------------------*/ |
3673 | ||
6f6652be JV |
3674 | /* |
3675 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3676 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3677 | */ | |
a361c83c | 3678 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3679 | { |
3680 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3681 | struct iphdr *iph = ip_hdr(skb); | |
3682 | ||
f14c4e4e | 3683 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3684 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3685 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3686 | } |
3687 | ||
d3da6831 | 3688 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3689 | } |
3690 | ||
169a3e66 | 3691 | /* |
59c51591 | 3692 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3693 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3694 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3695 | */ | |
a361c83c | 3696 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3697 | { |
3698 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3699 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3700 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3701 | int layer4_xor = 0; |
3702 | ||
f14c4e4e BH |
3703 | if (skb->protocol == htons(ETH_P_IP)) { |
3704 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3705 | (iph->protocol == IPPROTO_TCP || |
3706 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3707 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3708 | } |
3709 | return (layer4_xor ^ | |
3710 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3711 | ||
3712 | } | |
3713 | ||
d3da6831 | 3714 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3715 | } |
3716 | ||
3717 | /* | |
3718 | * Hash for the output device based upon layer 2 data | |
3719 | */ | |
a361c83c | 3720 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3721 | { |
3722 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3723 | ||
d3da6831 | 3724 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3725 | } |
3726 | ||
1da177e4 LT |
3727 | /*-------------------------- Device entry points ----------------------------*/ |
3728 | ||
3729 | static int bond_open(struct net_device *bond_dev) | |
3730 | { | |
454d7c9b | 3731 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3732 | |
3733 | bond->kill_timers = 0; | |
3734 | ||
58402054 | 3735 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3736 | /* bond_alb_initialize must be called before the timer |
3737 | * is started. | |
3738 | */ | |
3739 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3740 | /* something went wrong - fail the open operation */ | |
b473946a | 3741 | return -ENOMEM; |
1da177e4 LT |
3742 | } |
3743 | ||
1b76b316 JV |
3744 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3745 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3746 | } |
3747 | ||
3748 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3749 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3750 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3751 | } |
3752 | ||
3753 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3754 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3755 | INIT_DELAYED_WORK(&bond->arp_work, | |
3756 | bond_activebackup_arp_mon); | |
3757 | else | |
3758 | INIT_DELAYED_WORK(&bond->arp_work, | |
3759 | bond_loadbalance_arp_mon); | |
3760 | ||
3761 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3762 | if (bond->params.arp_validate) |
3763 | bond_register_arp(bond); | |
1da177e4 LT |
3764 | } |
3765 | ||
3766 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3767 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3768 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3769 | /* register to receive LACPDUs */ |
3770 | bond_register_lacpdu(bond); | |
fd989c83 | 3771 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3772 | } |
3773 | ||
3774 | return 0; | |
3775 | } | |
3776 | ||
3777 | static int bond_close(struct net_device *bond_dev) | |
3778 | { | |
454d7c9b | 3779 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3780 | |
3781 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3782 | /* Unregister the receive of LACPDUs */ | |
3783 | bond_unregister_lacpdu(bond); | |
3784 | } | |
3785 | ||
f5b2b966 JV |
3786 | if (bond->params.arp_validate) |
3787 | bond_unregister_arp(bond); | |
3788 | ||
1da177e4 LT |
3789 | write_lock_bh(&bond->lock); |
3790 | ||
b59f9f74 | 3791 | bond->send_grat_arp = 0; |
305d552a | 3792 | bond->send_unsol_na = 0; |
1da177e4 LT |
3793 | |
3794 | /* signal timers not to re-arm */ | |
3795 | bond->kill_timers = 1; | |
3796 | ||
3797 | write_unlock_bh(&bond->lock); | |
3798 | ||
1da177e4 | 3799 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3800 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3801 | } |
3802 | ||
3803 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3804 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3805 | } |
3806 | ||
3807 | switch (bond->params.mode) { | |
3808 | case BOND_MODE_8023AD: | |
1b76b316 | 3809 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3810 | break; |
3811 | case BOND_MODE_TLB: | |
3812 | case BOND_MODE_ALB: | |
1b76b316 | 3813 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3814 | break; |
3815 | default: | |
3816 | break; | |
3817 | } | |
3818 | ||
1da177e4 | 3819 | |
58402054 | 3820 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3821 | /* Must be called only after all |
3822 | * slaves have been released | |
3823 | */ | |
3824 | bond_alb_deinitialize(bond); | |
3825 | } | |
3826 | ||
3827 | return 0; | |
3828 | } | |
3829 | ||
28172739 ED |
3830 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3831 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3832 | { |
454d7c9b | 3833 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3834 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3835 | struct slave *slave; |
3836 | int i; | |
3837 | ||
28172739 | 3838 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3839 | |
3840 | read_lock_bh(&bond->lock); | |
3841 | ||
3842 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3843 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3844 | dev_get_stats(slave->dev, &temp); |
3845 | ||
3846 | stats->rx_packets += sstats->rx_packets; | |
3847 | stats->rx_bytes += sstats->rx_bytes; | |
3848 | stats->rx_errors += sstats->rx_errors; | |
3849 | stats->rx_dropped += sstats->rx_dropped; | |
3850 | ||
3851 | stats->tx_packets += sstats->tx_packets; | |
3852 | stats->tx_bytes += sstats->tx_bytes; | |
3853 | stats->tx_errors += sstats->tx_errors; | |
3854 | stats->tx_dropped += sstats->tx_dropped; | |
3855 | ||
3856 | stats->multicast += sstats->multicast; | |
3857 | stats->collisions += sstats->collisions; | |
3858 | ||
3859 | stats->rx_length_errors += sstats->rx_length_errors; | |
3860 | stats->rx_over_errors += sstats->rx_over_errors; | |
3861 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3862 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3863 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3864 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3865 | ||
3866 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3867 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3868 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3869 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3870 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3871 | } |
3872 | ||
1da177e4 LT |
3873 | read_unlock_bh(&bond->lock); |
3874 | ||
3875 | return stats; | |
3876 | } | |
3877 | ||
3878 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3879 | { | |
3880 | struct net_device *slave_dev = NULL; | |
3881 | struct ifbond k_binfo; | |
3882 | struct ifbond __user *u_binfo = NULL; | |
3883 | struct ifslave k_sinfo; | |
3884 | struct ifslave __user *u_sinfo = NULL; | |
3885 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3886 | int res = 0; |
3887 | ||
a4aee5c8 | 3888 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3889 | |
3890 | switch (cmd) { | |
1da177e4 LT |
3891 | case SIOCGMIIPHY: |
3892 | mii = if_mii(ifr); | |
3d632c3f | 3893 | if (!mii) |
1da177e4 | 3894 | return -EINVAL; |
3d632c3f | 3895 | |
1da177e4 LT |
3896 | mii->phy_id = 0; |
3897 | /* Fall Through */ | |
3898 | case SIOCGMIIREG: | |
3899 | /* | |
3900 | * We do this again just in case we were called by SIOCGMIIREG | |
3901 | * instead of SIOCGMIIPHY. | |
3902 | */ | |
3903 | mii = if_mii(ifr); | |
3d632c3f | 3904 | if (!mii) |
1da177e4 | 3905 | return -EINVAL; |
3d632c3f | 3906 | |
1da177e4 LT |
3907 | |
3908 | if (mii->reg_num == 1) { | |
454d7c9b | 3909 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3910 | mii->val_out = 0; |
6603a6f2 | 3911 | read_lock(&bond->lock); |
1da177e4 | 3912 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3913 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3914 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3915 | |
1da177e4 | 3916 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3917 | read_unlock(&bond->lock); |
1da177e4 LT |
3918 | } |
3919 | ||
3920 | return 0; | |
3921 | case BOND_INFO_QUERY_OLD: | |
3922 | case SIOCBONDINFOQUERY: | |
3923 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3924 | ||
3d632c3f | 3925 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3926 | return -EFAULT; |
1da177e4 LT |
3927 | |
3928 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3929 | if (res == 0 && |
3930 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3931 | return -EFAULT; | |
1da177e4 LT |
3932 | |
3933 | return res; | |
3934 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3935 | case SIOCBONDSLAVEINFOQUERY: | |
3936 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3937 | ||
3d632c3f | 3938 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3939 | return -EFAULT; |
1da177e4 LT |
3940 | |
3941 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3942 | if (res == 0 && |
3943 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3944 | return -EFAULT; | |
1da177e4 LT |
3945 | |
3946 | return res; | |
3947 | default: | |
3948 | /* Go on */ | |
3949 | break; | |
3950 | } | |
3951 | ||
3d632c3f | 3952 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3953 | return -EPERM; |
1da177e4 | 3954 | |
ec87fd3b | 3955 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3956 | |
a4aee5c8 | 3957 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3958 | |
3d632c3f | 3959 | if (!slave_dev) |
1da177e4 | 3960 | res = -ENODEV; |
3d632c3f | 3961 | else { |
a4aee5c8 | 3962 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3963 | switch (cmd) { |
3964 | case BOND_ENSLAVE_OLD: | |
3965 | case SIOCBONDENSLAVE: | |
3966 | res = bond_enslave(bond_dev, slave_dev); | |
3967 | break; | |
3968 | case BOND_RELEASE_OLD: | |
3969 | case SIOCBONDRELEASE: | |
3970 | res = bond_release(bond_dev, slave_dev); | |
3971 | break; | |
3972 | case BOND_SETHWADDR_OLD: | |
3973 | case SIOCBONDSETHWADDR: | |
3974 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3975 | break; | |
3976 | case BOND_CHANGE_ACTIVE_OLD: | |
3977 | case SIOCBONDCHANGEACTIVE: | |
3978 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3979 | break; | |
3980 | default: | |
3981 | res = -EOPNOTSUPP; | |
3982 | } | |
3983 | ||
3984 | dev_put(slave_dev); | |
3985 | } | |
3986 | ||
1da177e4 LT |
3987 | return res; |
3988 | } | |
3989 | ||
22bedad3 JP |
3990 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3991 | struct netdev_hw_addr_list *list, | |
3992 | int addrlen) | |
3993 | { | |
3994 | struct netdev_hw_addr *ha; | |
3995 | ||
3996 | netdev_hw_addr_list_for_each(ha, list) | |
3997 | if (!memcmp(ha->addr, addr, addrlen)) | |
3998 | return true; | |
3999 | ||
4000 | return false; | |
4001 | } | |
4002 | ||
1da177e4 LT |
4003 | static void bond_set_multicast_list(struct net_device *bond_dev) |
4004 | { | |
454d7c9b | 4005 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 JP |
4006 | struct netdev_hw_addr *ha; |
4007 | bool found; | |
1da177e4 | 4008 | |
1da177e4 LT |
4009 | /* |
4010 | * Do promisc before checking multicast_mode | |
4011 | */ | |
3d632c3f | 4012 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
4013 | /* |
4014 | * FIXME: Need to handle the error when one of the multi-slaves | |
4015 | * encounters error. | |
4016 | */ | |
1da177e4 | 4017 | bond_set_promiscuity(bond, 1); |
1da177e4 | 4018 | |
3d632c3f SH |
4019 | |
4020 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 4021 | bond_set_promiscuity(bond, -1); |
3d632c3f | 4022 | |
1da177e4 LT |
4023 | |
4024 | /* set allmulti flag to slaves */ | |
3d632c3f | 4025 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
4026 | /* |
4027 | * FIXME: Need to handle the error when one of the multi-slaves | |
4028 | * encounters error. | |
4029 | */ | |
1da177e4 | 4030 | bond_set_allmulti(bond, 1); |
1da177e4 | 4031 | |
3d632c3f SH |
4032 | |
4033 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 4034 | bond_set_allmulti(bond, -1); |
3d632c3f | 4035 | |
1da177e4 | 4036 | |
80ee5ad2 JV |
4037 | read_lock(&bond->lock); |
4038 | ||
1da177e4 LT |
4039 | bond->flags = bond_dev->flags; |
4040 | ||
4041 | /* looking for addresses to add to slaves' mc list */ | |
22bedad3 JP |
4042 | netdev_for_each_mc_addr(ha, bond_dev) { |
4043 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
4044 | bond_dev->addr_len); | |
4045 | if (!found) | |
4046 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
4047 | } |
4048 | ||
4049 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
4050 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
4051 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
4052 | bond_dev->addr_len); | |
4053 | if (!found) | |
4054 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
4055 | } |
4056 | ||
4057 | /* save master's multicast list */ | |
22bedad3 JP |
4058 | __hw_addr_flush(&bond->mc_list); |
4059 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
4060 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 4061 | |
80ee5ad2 | 4062 | read_unlock(&bond->lock); |
1da177e4 LT |
4063 | } |
4064 | ||
00829823 SH |
4065 | static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms) |
4066 | { | |
4067 | struct bonding *bond = netdev_priv(dev); | |
4068 | struct slave *slave = bond->first_slave; | |
4069 | ||
4070 | if (slave) { | |
4071 | const struct net_device_ops *slave_ops | |
4072 | = slave->dev->netdev_ops; | |
4073 | if (slave_ops->ndo_neigh_setup) | |
72e2240f | 4074 | return slave_ops->ndo_neigh_setup(slave->dev, parms); |
00829823 SH |
4075 | } |
4076 | return 0; | |
4077 | } | |
4078 | ||
1da177e4 LT |
4079 | /* |
4080 | * Change the MTU of all of a master's slaves to match the master | |
4081 | */ | |
4082 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
4083 | { | |
454d7c9b | 4084 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4085 | struct slave *slave, *stop_at; |
4086 | int res = 0; | |
4087 | int i; | |
4088 | ||
5a03cdb7 | 4089 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 4090 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
4091 | |
4092 | /* Can't hold bond->lock with bh disabled here since | |
4093 | * some base drivers panic. On the other hand we can't | |
4094 | * hold bond->lock without bh disabled because we'll | |
4095 | * deadlock. The only solution is to rely on the fact | |
4096 | * that we're under rtnl_lock here, and the slaves | |
4097 | * list won't change. This doesn't solve the problem | |
4098 | * of setting the slave's MTU while it is | |
4099 | * transmitting, but the assumption is that the base | |
4100 | * driver can handle that. | |
4101 | * | |
4102 | * TODO: figure out a way to safely iterate the slaves | |
4103 | * list, but without holding a lock around the actual | |
4104 | * call to the base driver. | |
4105 | */ | |
4106 | ||
4107 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
4108 | pr_debug("s %p s->p %p c_m %p\n", |
4109 | slave, | |
4110 | slave->prev, | |
4111 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 4112 | |
1da177e4 LT |
4113 | res = dev_set_mtu(slave->dev, new_mtu); |
4114 | ||
4115 | if (res) { | |
4116 | /* If we failed to set the slave's mtu to the new value | |
4117 | * we must abort the operation even in ACTIVE_BACKUP | |
4118 | * mode, because if we allow the backup slaves to have | |
4119 | * different mtu values than the active slave we'll | |
4120 | * need to change their mtu when doing a failover. That | |
4121 | * means changing their mtu from timer context, which | |
4122 | * is probably not a good idea. | |
4123 | */ | |
5a03cdb7 | 4124 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4125 | goto unwind; |
4126 | } | |
4127 | } | |
4128 | ||
4129 | bond_dev->mtu = new_mtu; | |
4130 | ||
4131 | return 0; | |
4132 | ||
4133 | unwind: | |
4134 | /* unwind from head to the slave that failed */ | |
4135 | stop_at = slave; | |
4136 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4137 | int tmp_res; | |
4138 | ||
4139 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
4140 | if (tmp_res) { | |
a4aee5c8 JP |
4141 | pr_debug("unwind err %d dev %s\n", |
4142 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4143 | } |
4144 | } | |
4145 | ||
4146 | return res; | |
4147 | } | |
4148 | ||
4149 | /* | |
4150 | * Change HW address | |
4151 | * | |
4152 | * Note that many devices must be down to change the HW address, and | |
4153 | * downing the master releases all slaves. We can make bonds full of | |
4154 | * bonding devices to test this, however. | |
4155 | */ | |
4156 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4157 | { | |
454d7c9b | 4158 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4159 | struct sockaddr *sa = addr, tmp_sa; |
4160 | struct slave *slave, *stop_at; | |
4161 | int res = 0; | |
4162 | int i; | |
4163 | ||
eb7cc59a SH |
4164 | if (bond->params.mode == BOND_MODE_ALB) |
4165 | return bond_alb_set_mac_address(bond_dev, addr); | |
4166 | ||
4167 | ||
a4aee5c8 JP |
4168 | pr_debug("bond=%p, name=%s\n", |
4169 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 4170 | |
dd957c57 | 4171 | /* |
3915c1e8 JV |
4172 | * If fail_over_mac is set to active, do nothing and return |
4173 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 4174 | */ |
3915c1e8 | 4175 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 4176 | return 0; |
2ab82852 | 4177 | |
3d632c3f | 4178 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 4179 | return -EADDRNOTAVAIL; |
1da177e4 LT |
4180 | |
4181 | /* Can't hold bond->lock with bh disabled here since | |
4182 | * some base drivers panic. On the other hand we can't | |
4183 | * hold bond->lock without bh disabled because we'll | |
4184 | * deadlock. The only solution is to rely on the fact | |
4185 | * that we're under rtnl_lock here, and the slaves | |
4186 | * list won't change. This doesn't solve the problem | |
4187 | * of setting the slave's hw address while it is | |
4188 | * transmitting, but the assumption is that the base | |
4189 | * driver can handle that. | |
4190 | * | |
4191 | * TODO: figure out a way to safely iterate the slaves | |
4192 | * list, but without holding a lock around the actual | |
4193 | * call to the base driver. | |
4194 | */ | |
4195 | ||
4196 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 4197 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 4198 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 4199 | |
eb7cc59a | 4200 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 4201 | res = -EOPNOTSUPP; |
5a03cdb7 | 4202 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
4203 | goto unwind; |
4204 | } | |
4205 | ||
4206 | res = dev_set_mac_address(slave->dev, addr); | |
4207 | if (res) { | |
4208 | /* TODO: consider downing the slave | |
4209 | * and retry ? | |
4210 | * User should expect communications | |
4211 | * breakage anyway until ARP finish | |
4212 | * updating, so... | |
4213 | */ | |
5a03cdb7 | 4214 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4215 | goto unwind; |
4216 | } | |
4217 | } | |
4218 | ||
4219 | /* success */ | |
4220 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4221 | return 0; | |
4222 | ||
4223 | unwind: | |
4224 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4225 | tmp_sa.sa_family = bond_dev->type; | |
4226 | ||
4227 | /* unwind from head to the slave that failed */ | |
4228 | stop_at = slave; | |
4229 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4230 | int tmp_res; | |
4231 | ||
4232 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4233 | if (tmp_res) { | |
a4aee5c8 JP |
4234 | pr_debug("unwind err %d dev %s\n", |
4235 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4236 | } |
4237 | } | |
4238 | ||
4239 | return res; | |
4240 | } | |
4241 | ||
4242 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4243 | { | |
454d7c9b | 4244 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4245 | struct slave *slave, *start_at; |
cf5f9044 | 4246 | int i, slave_no, res = 1; |
a2fd940f | 4247 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
4248 | |
4249 | read_lock(&bond->lock); | |
4250 | ||
3d632c3f | 4251 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4252 | goto out; |
cf5f9044 | 4253 | /* |
a2fd940f AG |
4254 | * Start with the curr_active_slave that joined the bond as the |
4255 | * default for sending IGMP traffic. For failover purposes one | |
4256 | * needs to maintain some consistency for the interface that will | |
4257 | * send the join/membership reports. The curr_active_slave found | |
4258 | * will send all of this type of traffic. | |
cf5f9044 | 4259 | */ |
00ae7028 | 4260 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4261 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4262 | |
a2fd940f AG |
4263 | read_lock(&bond->curr_slave_lock); |
4264 | slave = bond->curr_active_slave; | |
4265 | read_unlock(&bond->curr_slave_lock); | |
4266 | ||
4267 | if (!slave) | |
4268 | goto out; | |
4269 | } else { | |
4270 | /* | |
4271 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4272 | * that as being rare enough not to justify using an | |
4273 | * atomic op here. | |
4274 | */ | |
4275 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4276 | ||
4277 | bond_for_each_slave(bond, slave, i) { | |
4278 | slave_no--; | |
4279 | if (slave_no < 0) | |
4280 | break; | |
4281 | } | |
1da177e4 LT |
4282 | } |
4283 | ||
cf5f9044 | 4284 | start_at = slave; |
1da177e4 LT |
4285 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4286 | if (IS_UP(slave->dev) && | |
4287 | (slave->link == BOND_LINK_UP) && | |
4288 | (slave->state == BOND_STATE_ACTIVE)) { | |
4289 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
1da177e4 LT |
4290 | break; |
4291 | } | |
4292 | } | |
4293 | ||
1da177e4 LT |
4294 | out: |
4295 | if (res) { | |
4296 | /* no suitable interface, frame not sent */ | |
4297 | dev_kfree_skb(skb); | |
4298 | } | |
4299 | read_unlock(&bond->lock); | |
ec634fe3 | 4300 | return NETDEV_TX_OK; |
1da177e4 LT |
4301 | } |
4302 | ||
075897ce | 4303 | |
1da177e4 LT |
4304 | /* |
4305 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4306 | * the bond has a usable interface. | |
4307 | */ | |
4308 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4309 | { | |
454d7c9b | 4310 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4311 | int res = 1; |
4312 | ||
1da177e4 LT |
4313 | read_lock(&bond->lock); |
4314 | read_lock(&bond->curr_slave_lock); | |
4315 | ||
3d632c3f | 4316 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4317 | goto out; |
1da177e4 | 4318 | |
075897ce JL |
4319 | if (!bond->curr_active_slave) |
4320 | goto out; | |
4321 | ||
075897ce JL |
4322 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4323 | ||
1da177e4 | 4324 | out: |
3d632c3f | 4325 | if (res) |
1da177e4 LT |
4326 | /* no suitable interface, frame not sent */ |
4327 | dev_kfree_skb(skb); | |
3d632c3f | 4328 | |
1da177e4 LT |
4329 | read_unlock(&bond->curr_slave_lock); |
4330 | read_unlock(&bond->lock); | |
ec634fe3 | 4331 | return NETDEV_TX_OK; |
1da177e4 LT |
4332 | } |
4333 | ||
4334 | /* | |
169a3e66 JV |
4335 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4336 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4337 | * find the next active slave. | |
1da177e4 LT |
4338 | */ |
4339 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4340 | { | |
454d7c9b | 4341 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4342 | struct slave *slave, *start_at; |
4343 | int slave_no; | |
4344 | int i; | |
4345 | int res = 1; | |
4346 | ||
4347 | read_lock(&bond->lock); | |
4348 | ||
3d632c3f | 4349 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4350 | goto out; |
1da177e4 | 4351 | |
a361c83c | 4352 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4353 | |
4354 | bond_for_each_slave(bond, slave, i) { | |
4355 | slave_no--; | |
3d632c3f | 4356 | if (slave_no < 0) |
1da177e4 | 4357 | break; |
1da177e4 LT |
4358 | } |
4359 | ||
4360 | start_at = slave; | |
4361 | ||
4362 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4363 | if (IS_UP(slave->dev) && | |
4364 | (slave->link == BOND_LINK_UP) && | |
4365 | (slave->state == BOND_STATE_ACTIVE)) { | |
4366 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4367 | break; | |
4368 | } | |
4369 | } | |
4370 | ||
4371 | out: | |
4372 | if (res) { | |
4373 | /* no suitable interface, frame not sent */ | |
4374 | dev_kfree_skb(skb); | |
4375 | } | |
4376 | read_unlock(&bond->lock); | |
ec634fe3 | 4377 | return NETDEV_TX_OK; |
1da177e4 LT |
4378 | } |
4379 | ||
4380 | /* | |
4381 | * in broadcast mode, we send everything to all usable interfaces. | |
4382 | */ | |
4383 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4384 | { | |
454d7c9b | 4385 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4386 | struct slave *slave, *start_at; |
4387 | struct net_device *tx_dev = NULL; | |
4388 | int i; | |
4389 | int res = 1; | |
4390 | ||
4391 | read_lock(&bond->lock); | |
4392 | ||
3d632c3f | 4393 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4394 | goto out; |
1da177e4 LT |
4395 | |
4396 | read_lock(&bond->curr_slave_lock); | |
4397 | start_at = bond->curr_active_slave; | |
4398 | read_unlock(&bond->curr_slave_lock); | |
4399 | ||
3d632c3f | 4400 | if (!start_at) |
1da177e4 | 4401 | goto out; |
1da177e4 LT |
4402 | |
4403 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4404 | if (IS_UP(slave->dev) && | |
4405 | (slave->link == BOND_LINK_UP) && | |
4406 | (slave->state == BOND_STATE_ACTIVE)) { | |
4407 | if (tx_dev) { | |
4408 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4409 | if (!skb2) { | |
a4aee5c8 | 4410 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4411 | bond_dev->name); |
1da177e4 LT |
4412 | continue; |
4413 | } | |
4414 | ||
4415 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4416 | if (res) { | |
4417 | dev_kfree_skb(skb2); | |
4418 | continue; | |
4419 | } | |
4420 | } | |
4421 | tx_dev = slave->dev; | |
4422 | } | |
4423 | } | |
4424 | ||
3d632c3f | 4425 | if (tx_dev) |
1da177e4 | 4426 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4427 | |
4428 | out: | |
3d632c3f | 4429 | if (res) |
1da177e4 LT |
4430 | /* no suitable interface, frame not sent */ |
4431 | dev_kfree_skb(skb); | |
3d632c3f | 4432 | |
1da177e4 LT |
4433 | /* frame sent to all suitable interfaces */ |
4434 | read_unlock(&bond->lock); | |
ec634fe3 | 4435 | return NETDEV_TX_OK; |
1da177e4 LT |
4436 | } |
4437 | ||
4438 | /*------------------------- Device initialization ---------------------------*/ | |
4439 | ||
6f6652be JV |
4440 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4441 | { | |
4442 | switch (bond->params.xmit_policy) { | |
4443 | case BOND_XMIT_POLICY_LAYER23: | |
4444 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4445 | break; | |
4446 | case BOND_XMIT_POLICY_LAYER34: | |
4447 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4448 | break; | |
4449 | case BOND_XMIT_POLICY_LAYER2: | |
4450 | default: | |
4451 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4452 | break; | |
4453 | } | |
4454 | } | |
4455 | ||
bb1d9123 AG |
4456 | /* |
4457 | * Lookup the slave that corresponds to a qid | |
4458 | */ | |
4459 | static inline int bond_slave_override(struct bonding *bond, | |
4460 | struct sk_buff *skb) | |
4461 | { | |
4462 | int i, res = 1; | |
4463 | struct slave *slave = NULL; | |
4464 | struct slave *check_slave; | |
4465 | ||
4466 | read_lock(&bond->lock); | |
4467 | ||
4468 | if (!BOND_IS_OK(bond) || !skb->queue_mapping) | |
4469 | goto out; | |
4470 | ||
4471 | /* Find out if any slaves have the same mapping as this skb. */ | |
4472 | bond_for_each_slave(bond, check_slave, i) { | |
4473 | if (check_slave->queue_id == skb->queue_mapping) { | |
4474 | slave = check_slave; | |
4475 | break; | |
4476 | } | |
4477 | } | |
4478 | ||
4479 | /* If the slave isn't UP, use default transmit policy. */ | |
4480 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4481 | (slave->link == BOND_LINK_UP)) { | |
4482 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4483 | } | |
4484 | ||
4485 | out: | |
4486 | read_unlock(&bond->lock); | |
4487 | return res; | |
4488 | } | |
4489 | ||
4490 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) | |
4491 | { | |
4492 | /* | |
4493 | * This helper function exists to help dev_pick_tx get the correct | |
4494 | * destination queue. Using a helper function skips the a call to | |
4495 | * skb_tx_hash and will put the skbs in the queue we expect on their | |
4496 | * way down to the bonding driver. | |
4497 | */ | |
4498 | return skb->queue_mapping; | |
4499 | } | |
4500 | ||
424efe9c | 4501 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4502 | { |
bb1d9123 AG |
4503 | struct bonding *bond = netdev_priv(dev); |
4504 | ||
4505 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
4506 | if (!bond_slave_override(bond, skb)) | |
4507 | return NETDEV_TX_OK; | |
4508 | } | |
00829823 SH |
4509 | |
4510 | switch (bond->params.mode) { | |
4511 | case BOND_MODE_ROUNDROBIN: | |
4512 | return bond_xmit_roundrobin(skb, dev); | |
4513 | case BOND_MODE_ACTIVEBACKUP: | |
4514 | return bond_xmit_activebackup(skb, dev); | |
4515 | case BOND_MODE_XOR: | |
4516 | return bond_xmit_xor(skb, dev); | |
4517 | case BOND_MODE_BROADCAST: | |
4518 | return bond_xmit_broadcast(skb, dev); | |
4519 | case BOND_MODE_8023AD: | |
4520 | return bond_3ad_xmit_xor(skb, dev); | |
4521 | case BOND_MODE_ALB: | |
4522 | case BOND_MODE_TLB: | |
4523 | return bond_alb_xmit(skb, dev); | |
4524 | default: | |
4525 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4526 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4527 | dev->name, bond->params.mode); | |
00829823 SH |
4528 | WARN_ON_ONCE(1); |
4529 | dev_kfree_skb(skb); | |
4530 | return NETDEV_TX_OK; | |
4531 | } | |
4532 | } | |
4533 | ||
4534 | ||
1da177e4 LT |
4535 | /* |
4536 | * set bond mode specific net device operations | |
4537 | */ | |
a77b5325 | 4538 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4539 | { |
169a3e66 JV |
4540 | struct net_device *bond_dev = bond->dev; |
4541 | ||
1da177e4 LT |
4542 | switch (mode) { |
4543 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4544 | break; |
4545 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4546 | break; |
4547 | case BOND_MODE_XOR: | |
6f6652be | 4548 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4549 | break; |
4550 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4551 | break; |
4552 | case BOND_MODE_8023AD: | |
8f903c70 | 4553 | bond_set_master_3ad_flags(bond); |
6f6652be | 4554 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4555 | break; |
1da177e4 | 4556 | case BOND_MODE_ALB: |
8f903c70 JV |
4557 | bond_set_master_alb_flags(bond); |
4558 | /* FALLTHRU */ | |
4559 | case BOND_MODE_TLB: | |
1da177e4 LT |
4560 | break; |
4561 | default: | |
4562 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4563 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4564 | bond_dev->name, mode); | |
1da177e4 LT |
4565 | break; |
4566 | } | |
4567 | } | |
4568 | ||
217df670 JV |
4569 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4570 | struct ethtool_drvinfo *drvinfo) | |
4571 | { | |
4572 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4573 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4574 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4575 | } | |
4576 | ||
7282d491 | 4577 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4578 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4579 | .get_link = ethtool_op_get_link, |
4580 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4581 | .get_sg = ethtool_op_get_sg, | |
4582 | .get_tso = ethtool_op_get_tso, | |
4583 | .get_ufo = ethtool_op_get_ufo, | |
4584 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4585 | }; |
4586 | ||
eb7cc59a | 4587 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4588 | .ndo_init = bond_init, |
9e71626c | 4589 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4590 | .ndo_open = bond_open, |
4591 | .ndo_stop = bond_close, | |
00829823 | 4592 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4593 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4594 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a SH |
4595 | .ndo_do_ioctl = bond_do_ioctl, |
4596 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4597 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4598 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4599 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4600 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4601 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4602 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
f6dc31a8 WC |
4603 | #ifdef CONFIG_NET_POLL_CONTROLLER |
4604 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, | |
4605 | .ndo_poll_controller = bond_poll_controller, | |
4606 | #endif | |
eb7cc59a SH |
4607 | }; |
4608 | ||
9e2e61fb AW |
4609 | static void bond_destructor(struct net_device *bond_dev) |
4610 | { | |
4611 | struct bonding *bond = netdev_priv(bond_dev); | |
4612 | if (bond->wq) | |
4613 | destroy_workqueue(bond->wq); | |
4614 | free_netdev(bond_dev); | |
4615 | } | |
4616 | ||
181470fc | 4617 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4618 | { |
454d7c9b | 4619 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4620 | |
1da177e4 LT |
4621 | /* initialize rwlocks */ |
4622 | rwlock_init(&bond->lock); | |
4623 | rwlock_init(&bond->curr_slave_lock); | |
4624 | ||
d2991f75 | 4625 | bond->params = bonding_defaults; |
1da177e4 LT |
4626 | |
4627 | /* Initialize pointers */ | |
1da177e4 LT |
4628 | bond->dev = bond_dev; |
4629 | INIT_LIST_HEAD(&bond->vlan_list); | |
4630 | ||
4631 | /* Initialize the device entry points */ | |
181470fc | 4632 | ether_setup(bond_dev); |
eb7cc59a | 4633 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4634 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4635 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4636 | |
9e2e61fb | 4637 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4638 | |
4639 | /* Initialize the device options */ | |
4640 | bond_dev->tx_queue_len = 0; | |
4641 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4642 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4643 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4644 | ||
6cf3f41e JV |
4645 | if (bond->params.arp_interval) |
4646 | bond_dev->priv_flags |= IFF_MASTER_ARPMON; | |
1da177e4 LT |
4647 | |
4648 | /* At first, we block adding VLANs. That's the only way to | |
4649 | * prevent problems that occur when adding VLANs over an | |
4650 | * empty bond. The block will be removed once non-challenged | |
4651 | * slaves are enslaved. | |
4652 | */ | |
4653 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4654 | ||
932ff279 | 4655 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4656 | * transmitting */ |
4657 | bond_dev->features |= NETIF_F_LLTX; | |
4658 | ||
4659 | /* By default, we declare the bond to be fully | |
4660 | * VLAN hardware accelerated capable. Special | |
4661 | * care is taken in the various xmit functions | |
4662 | * when there are slaves that are not hw accel | |
4663 | * capable | |
4664 | */ | |
1da177e4 LT |
4665 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4666 | NETIF_F_HW_VLAN_RX | | |
4667 | NETIF_F_HW_VLAN_FILTER); | |
4668 | ||
1da177e4 LT |
4669 | } |
4670 | ||
fdaea7a9 JV |
4671 | static void bond_work_cancel_all(struct bonding *bond) |
4672 | { | |
4673 | write_lock_bh(&bond->lock); | |
4674 | bond->kill_timers = 1; | |
4675 | write_unlock_bh(&bond->lock); | |
4676 | ||
4677 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4678 | cancel_delayed_work(&bond->mii_work); | |
4679 | ||
4680 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4681 | cancel_delayed_work(&bond->arp_work); | |
4682 | ||
4683 | if (bond->params.mode == BOND_MODE_ALB && | |
4684 | delayed_work_pending(&bond->alb_work)) | |
4685 | cancel_delayed_work(&bond->alb_work); | |
4686 | ||
4687 | if (bond->params.mode == BOND_MODE_8023AD && | |
4688 | delayed_work_pending(&bond->ad_work)) | |
4689 | cancel_delayed_work(&bond->ad_work); | |
4690 | } | |
4691 | ||
c67dfb29 EB |
4692 | /* |
4693 | * Destroy a bonding device. | |
4694 | * Must be under rtnl_lock when this function is called. | |
4695 | */ | |
4696 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4697 | { |
454d7c9b | 4698 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4699 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4700 | |
f6dc31a8 WC |
4701 | bond_netpoll_cleanup(bond_dev); |
4702 | ||
c67dfb29 EB |
4703 | /* Release the bonded slaves */ |
4704 | bond_release_all(bond_dev); | |
4705 | ||
a434e43f JV |
4706 | list_del(&bond->bond_list); |
4707 | ||
4708 | bond_work_cancel_all(bond); | |
4709 | ||
a434e43f | 4710 | bond_remove_proc_entry(bond); |
c67dfb29 | 4711 | |
22bedad3 | 4712 | __hw_addr_flush(&bond->mc_list); |
f35188fa JV |
4713 | |
4714 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { | |
4715 | list_del(&vlan->vlan_list); | |
4716 | kfree(vlan); | |
4717 | } | |
a434e43f JV |
4718 | } |
4719 | ||
1da177e4 LT |
4720 | /*------------------------- Module initialization ---------------------------*/ |
4721 | ||
4722 | /* | |
4723 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4724 | * number of the mode or its string name. A bit complicated because |
4725 | * some mode names are substrings of other names, and calls from sysfs | |
4726 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4727 | */ |
325dcf7a | 4728 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4729 | { |
54b87323 | 4730 | int modeint = -1, i, rv; |
a42e534f | 4731 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4732 | |
a42e534f JV |
4733 | for (p = (char *)buf; *p; p++) |
4734 | if (!(isdigit(*p) || isspace(*p))) | |
4735 | break; | |
4736 | ||
4737 | if (*p) | |
ece95f7f | 4738 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4739 | else |
54b87323 | 4740 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4741 | |
4742 | if (!rv) | |
4743 | return -1; | |
1da177e4 LT |
4744 | |
4745 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4746 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4747 | return tbl[i].mode; |
4748 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4749 | return tbl[i].mode; |
1da177e4 LT |
4750 | } |
4751 | ||
4752 | return -1; | |
4753 | } | |
4754 | ||
4755 | static int bond_check_params(struct bond_params *params) | |
4756 | { | |
a549952a | 4757 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4758 | |
1da177e4 LT |
4759 | /* |
4760 | * Convert string parameters. | |
4761 | */ | |
4762 | if (mode) { | |
4763 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4764 | if (bond_mode == -1) { | |
a4aee5c8 | 4765 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4766 | mode == NULL ? "NULL" : mode); |
4767 | return -EINVAL; | |
4768 | } | |
4769 | } | |
4770 | ||
169a3e66 JV |
4771 | if (xmit_hash_policy) { |
4772 | if ((bond_mode != BOND_MODE_XOR) && | |
4773 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4774 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4775 | bond_mode_name(bond_mode)); |
4776 | } else { | |
4777 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4778 | xmit_hashtype_tbl); | |
4779 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4780 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4781 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4782 | xmit_hash_policy); |
4783 | return -EINVAL; | |
4784 | } | |
4785 | } | |
4786 | } | |
4787 | ||
1da177e4 LT |
4788 | if (lacp_rate) { |
4789 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4790 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4791 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4792 | } else { |
4793 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4794 | if (lacp_fast == -1) { | |
a4aee5c8 | 4795 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4796 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4797 | return -EINVAL; | |
4798 | } | |
4799 | } | |
4800 | } | |
4801 | ||
fd989c83 JV |
4802 | if (ad_select) { |
4803 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4804 | if (params->ad_select == -1) { | |
a4aee5c8 | 4805 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4806 | ad_select == NULL ? "NULL" : ad_select); |
4807 | return -EINVAL; | |
4808 | } | |
4809 | ||
4810 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4811 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4812 | } |
4813 | } else { | |
4814 | params->ad_select = BOND_AD_STABLE; | |
4815 | } | |
4816 | ||
f5841306 | 4817 | if (max_bonds < 0) { |
a4aee5c8 JP |
4818 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4819 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4820 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4821 | } | |
4822 | ||
4823 | if (miimon < 0) { | |
a4aee5c8 JP |
4824 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4825 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4826 | miimon = BOND_LINK_MON_INTERV; |
4827 | } | |
4828 | ||
4829 | if (updelay < 0) { | |
a4aee5c8 JP |
4830 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4831 | updelay, INT_MAX); | |
1da177e4 LT |
4832 | updelay = 0; |
4833 | } | |
4834 | ||
4835 | if (downdelay < 0) { | |
a4aee5c8 JP |
4836 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4837 | downdelay, INT_MAX); | |
1da177e4 LT |
4838 | downdelay = 0; |
4839 | } | |
4840 | ||
4841 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4842 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4843 | use_carrier); | |
1da177e4 LT |
4844 | use_carrier = 1; |
4845 | } | |
4846 | ||
7893b249 | 4847 | if (num_grat_arp < 0 || num_grat_arp > 255) { |
2381a55c | 4848 | pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4849 | num_grat_arp); |
7893b249 MS |
4850 | num_grat_arp = 1; |
4851 | } | |
4852 | ||
305d552a | 4853 | if (num_unsol_na < 0 || num_unsol_na > 255) { |
2381a55c | 4854 | pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4855 | num_unsol_na); |
305d552a BH |
4856 | num_unsol_na = 1; |
4857 | } | |
4858 | ||
1da177e4 LT |
4859 | /* reset values for 802.3ad */ |
4860 | if (bond_mode == BOND_MODE_8023AD) { | |
4861 | if (!miimon) { | |
a4aee5c8 | 4862 | 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 | 4863 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4864 | miimon = 100; |
4865 | } | |
4866 | } | |
4867 | ||
bb1d9123 AG |
4868 | if (tx_queues < 1 || tx_queues > 255) { |
4869 | pr_warning("Warning: tx_queues (%d) should be between " | |
4870 | "1 and 255, resetting to %d\n", | |
4871 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4872 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4873 | } | |
4874 | ||
ebd8e497 AG |
4875 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4876 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4877 | "not of valid value (0/1), so it was set to " | |
4878 | "0\n", all_slaves_active); | |
4879 | all_slaves_active = 0; | |
4880 | } | |
4881 | ||
1da177e4 LT |
4882 | /* reset values for TLB/ALB */ |
4883 | if ((bond_mode == BOND_MODE_TLB) || | |
4884 | (bond_mode == BOND_MODE_ALB)) { | |
4885 | if (!miimon) { | |
a4aee5c8 | 4886 | 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 | 4887 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4888 | miimon = 100; |
4889 | } | |
4890 | } | |
4891 | ||
4892 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4893 | 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", |
4894 | updelay); | |
1da177e4 LT |
4895 | } |
4896 | ||
4897 | if (!miimon) { | |
4898 | if (updelay || downdelay) { | |
4899 | /* just warn the user the up/down delay will have | |
4900 | * no effect since miimon is zero... | |
4901 | */ | |
a4aee5c8 JP |
4902 | 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", |
4903 | updelay, downdelay); | |
1da177e4 LT |
4904 | } |
4905 | } else { | |
4906 | /* don't allow arp monitoring */ | |
4907 | if (arp_interval) { | |
a4aee5c8 JP |
4908 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4909 | miimon, arp_interval); | |
1da177e4 LT |
4910 | arp_interval = 0; |
4911 | } | |
4912 | ||
4913 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4914 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4915 | updelay, miimon, | |
4916 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4917 | } |
4918 | ||
4919 | updelay /= miimon; | |
4920 | ||
4921 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4922 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4923 | downdelay, miimon, | |
4924 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4925 | } |
4926 | ||
4927 | downdelay /= miimon; | |
4928 | } | |
4929 | ||
4930 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4931 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4932 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4933 | arp_interval = BOND_LINK_ARP_INTERV; |
4934 | } | |
4935 | ||
4936 | for (arp_ip_count = 0; | |
4937 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4938 | arp_ip_count++) { | |
4939 | /* not complete check, but should be good enough to | |
4940 | catch mistakes */ | |
4941 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4942 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4943 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4944 | arp_interval = 0; |
4945 | } else { | |
d3bb52b0 | 4946 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4947 | arp_target[arp_ip_count] = ip; |
4948 | } | |
4949 | } | |
4950 | ||
4951 | if (arp_interval && !arp_ip_count) { | |
4952 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4953 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4954 | arp_interval); | |
1da177e4 LT |
4955 | arp_interval = 0; |
4956 | } | |
4957 | ||
f5b2b966 JV |
4958 | if (arp_validate) { |
4959 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4960 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4961 | return -EINVAL; |
4962 | } | |
4963 | if (!arp_interval) { | |
a4aee5c8 | 4964 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4965 | return -EINVAL; |
4966 | } | |
4967 | ||
4968 | arp_validate_value = bond_parse_parm(arp_validate, | |
4969 | arp_validate_tbl); | |
4970 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4971 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4972 | arp_validate == NULL ? "NULL" : arp_validate); |
4973 | return -EINVAL; | |
4974 | } | |
4975 | } else | |
4976 | arp_validate_value = 0; | |
4977 | ||
1da177e4 | 4978 | if (miimon) { |
a4aee5c8 | 4979 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4980 | } else if (arp_interval) { |
4981 | int i; | |
4982 | ||
a4aee5c8 JP |
4983 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4984 | arp_interval, | |
4985 | arp_validate_tbl[arp_validate_value].modename, | |
4986 | arp_ip_count); | |
1da177e4 LT |
4987 | |
4988 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4989 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4990 | |
e5e2a8fd | 4991 | pr_info("\n"); |
1da177e4 | 4992 | |
b8a9787e | 4993 | } else if (max_bonds) { |
1da177e4 LT |
4994 | /* miimon and arp_interval not set, we need one so things |
4995 | * work as expected, see bonding.txt for details | |
4996 | */ | |
a4aee5c8 | 4997 | 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 |
4998 | } |
4999 | ||
5000 | if (primary && !USES_PRIMARY(bond_mode)) { | |
5001 | /* currently, using a primary only makes sense | |
5002 | * in active backup, TLB or ALB modes | |
5003 | */ | |
a4aee5c8 JP |
5004 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
5005 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
5006 | primary = NULL; |
5007 | } | |
5008 | ||
a549952a JP |
5009 | if (primary && primary_reselect) { |
5010 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
5011 | pri_reselect_tbl); | |
5012 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 5013 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
5014 | primary_reselect == |
5015 | NULL ? "NULL" : primary_reselect); | |
5016 | return -EINVAL; | |
5017 | } | |
5018 | } else { | |
5019 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
5020 | } | |
5021 | ||
3915c1e8 JV |
5022 | if (fail_over_mac) { |
5023 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
5024 | fail_over_mac_tbl); | |
5025 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 5026 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
5027 | arp_validate == NULL ? "NULL" : arp_validate); |
5028 | return -EINVAL; | |
5029 | } | |
5030 | ||
5031 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 5032 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
5033 | } else { |
5034 | fail_over_mac_value = BOND_FOM_NONE; | |
5035 | } | |
dd957c57 | 5036 | |
1da177e4 LT |
5037 | /* fill params struct with the proper values */ |
5038 | params->mode = bond_mode; | |
169a3e66 | 5039 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 5040 | params->miimon = miimon; |
7893b249 | 5041 | params->num_grat_arp = num_grat_arp; |
305d552a | 5042 | params->num_unsol_na = num_unsol_na; |
1da177e4 | 5043 | params->arp_interval = arp_interval; |
f5b2b966 | 5044 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
5045 | params->updelay = updelay; |
5046 | params->downdelay = downdelay; | |
5047 | params->use_carrier = use_carrier; | |
5048 | params->lacp_fast = lacp_fast; | |
5049 | params->primary[0] = 0; | |
a549952a | 5050 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 5051 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 5052 | params->tx_queues = tx_queues; |
ebd8e497 | 5053 | params->all_slaves_active = all_slaves_active; |
1da177e4 LT |
5054 | |
5055 | if (primary) { | |
5056 | strncpy(params->primary, primary, IFNAMSIZ); | |
5057 | params->primary[IFNAMSIZ - 1] = 0; | |
5058 | } | |
5059 | ||
5060 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
5061 | ||
5062 | return 0; | |
5063 | } | |
5064 | ||
0daa2303 | 5065 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 5066 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 5067 | |
e8a0464c DM |
5068 | static void bond_set_lockdep_class_one(struct net_device *dev, |
5069 | struct netdev_queue *txq, | |
5070 | void *_unused) | |
c773e847 DM |
5071 | { |
5072 | lockdep_set_class(&txq->_xmit_lock, | |
5073 | &bonding_netdev_xmit_lock_key); | |
5074 | } | |
5075 | ||
5076 | static void bond_set_lockdep_class(struct net_device *dev) | |
5077 | { | |
cf508b12 DM |
5078 | lockdep_set_class(&dev->addr_list_lock, |
5079 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 5080 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
5081 | } |
5082 | ||
181470fc SH |
5083 | /* |
5084 | * Called from registration process | |
5085 | */ | |
5086 | static int bond_init(struct net_device *bond_dev) | |
5087 | { | |
5088 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 5089 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc SH |
5090 | |
5091 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
5092 | ||
5093 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
5094 | if (!bond->wq) | |
5095 | return -ENOMEM; | |
5096 | ||
5097 | bond_set_lockdep_class(bond_dev); | |
5098 | ||
5099 | netif_carrier_off(bond_dev); | |
5100 | ||
5101 | bond_create_proc_entry(bond); | |
ec87fd3b | 5102 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 5103 | |
6151b3d4 | 5104 | bond_prepare_sysfs_group(bond); |
22bedad3 JP |
5105 | |
5106 | __hw_addr_init(&bond->mc_list); | |
181470fc SH |
5107 | return 0; |
5108 | } | |
5109 | ||
88ead977 EB |
5110 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
5111 | { | |
5112 | if (tb[IFLA_ADDRESS]) { | |
5113 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
5114 | return -EINVAL; | |
5115 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
5116 | return -EADDRNOTAVAIL; | |
5117 | } | |
5118 | return 0; | |
5119 | } | |
5120 | ||
5121 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
5122 | .kind = "bond", | |
6639104b | 5123 | .priv_size = sizeof(struct bonding), |
88ead977 EB |
5124 | .setup = bond_setup, |
5125 | .validate = bond_validate, | |
5126 | }; | |
5127 | ||
dfe60397 | 5128 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5129 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5130 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5131 | * set up our sysfs entries. | |
5132 | */ | |
ec87fd3b | 5133 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
5134 | { |
5135 | struct net_device *bond_dev; | |
5136 | int res; | |
5137 | ||
5138 | rtnl_lock(); | |
027ea041 | 5139 | |
bb1d9123 AG |
5140 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "", |
5141 | bond_setup, tx_queues); | |
dfe60397 | 5142 | if (!bond_dev) { |
a4aee5c8 | 5143 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
5144 | rtnl_unlock(); |
5145 | return -ENOMEM; | |
dfe60397 MW |
5146 | } |
5147 | ||
ec87fd3b | 5148 | dev_net_set(bond_dev, net); |
88ead977 EB |
5149 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5150 | ||
e4b91c48 JV |
5151 | if (!name) { |
5152 | res = dev_alloc_name(bond_dev, "bond%d"); | |
5153 | if (res < 0) | |
9e2e61fb | 5154 | goto out; |
e4b91c48 JV |
5155 | } |
5156 | ||
dfe60397 | 5157 | res = register_netdevice(bond_dev); |
0daa2303 | 5158 | |
30c15ba9 | 5159 | out: |
7e083840 | 5160 | rtnl_unlock(); |
9e2e61fb AW |
5161 | if (res < 0) |
5162 | bond_destructor(bond_dev); | |
30c15ba9 | 5163 | return res; |
dfe60397 MW |
5164 | } |
5165 | ||
2c8c1e72 | 5166 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 5167 | { |
15449745 | 5168 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5169 | |
5170 | bn->net = net; | |
5171 | INIT_LIST_HEAD(&bn->dev_list); | |
5172 | ||
ec87fd3b | 5173 | bond_create_proc_dir(bn); |
15449745 EB |
5174 | |
5175 | return 0; | |
ec87fd3b EB |
5176 | } |
5177 | ||
2c8c1e72 | 5178 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 5179 | { |
15449745 | 5180 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5181 | |
5182 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
5183 | } |
5184 | ||
5185 | static struct pernet_operations bond_net_ops = { | |
5186 | .init = bond_net_init, | |
5187 | .exit = bond_net_exit, | |
15449745 EB |
5188 | .id = &bond_net_id, |
5189 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
5190 | }; |
5191 | ||
1da177e4 LT |
5192 | static int __init bonding_init(void) |
5193 | { | |
1da177e4 LT |
5194 | int i; |
5195 | int res; | |
5196 | ||
3d632c3f | 5197 | pr_info("%s", version); |
1da177e4 | 5198 | |
dfe60397 | 5199 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5200 | if (res) |
dfe60397 | 5201 | goto out; |
1da177e4 | 5202 | |
15449745 | 5203 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
5204 | if (res) |
5205 | goto out; | |
027ea041 | 5206 | |
88ead977 EB |
5207 | res = rtnl_link_register(&bond_link_ops); |
5208 | if (res) | |
6639104b | 5209 | goto err_link; |
88ead977 | 5210 | |
1da177e4 | 5211 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5212 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5213 | if (res) |
5214 | goto err; | |
1da177e4 LT |
5215 | } |
5216 | ||
b76cdba9 MW |
5217 | res = bond_create_sysfs(); |
5218 | if (res) | |
5219 | goto err; | |
5220 | ||
1da177e4 | 5221 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5222 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5223 | bond_register_ipv6_notifier(); |
dfe60397 | 5224 | out: |
1da177e4 | 5225 | return res; |
88ead977 EB |
5226 | err: |
5227 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5228 | err_link: |
15449745 | 5229 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5230 | goto out; |
dfe60397 | 5231 | |
1da177e4 LT |
5232 | } |
5233 | ||
5234 | static void __exit bonding_exit(void) | |
5235 | { | |
5236 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5237 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5238 | bond_unregister_ipv6_notifier(); |
1da177e4 | 5239 | |
ae68c398 PE |
5240 | bond_destroy_sysfs(); |
5241 | ||
88ead977 | 5242 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5243 | unregister_pernet_subsys(&bond_net_ops); |
1da177e4 LT |
5244 | } |
5245 | ||
5246 | module_init(bonding_init); | |
5247 | module_exit(bonding_exit); | |
5248 | MODULE_LICENSE("GPL"); | |
5249 | MODULE_VERSION(DRV_VERSION); | |
5250 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5251 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5252 | MODULE_ALIAS_RTNL_LINK("bond"); |