Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * originally based on the dummy device. | |
3 | * | |
4 | * Copyright 1999, Thomas Davis, tadavis@lbl.gov. | |
5 | * Licensed under the GPL. Based on dummy.c, and eql.c devices. | |
6 | * | |
7 | * bonding.c: an Ethernet Bonding driver | |
8 | * | |
9 | * This is useful to talk to a Cisco EtherChannel compatible equipment: | |
10 | * Cisco 5500 | |
11 | * Sun Trunking (Solaris) | |
12 | * Alteon AceDirector Trunks | |
13 | * Linux Bonding | |
14 | * and probably many L2 switches ... | |
15 | * | |
16 | * How it works: | |
17 | * ifconfig bond0 ipaddress netmask up | |
18 | * will setup a network device, with an ip address. No mac address | |
19 | * will be assigned at this time. The hw mac address will come from | |
20 | * the first slave bonded to the channel. All slaves will then use | |
21 | * this hw mac address. | |
22 | * | |
23 | * ifconfig bond0 down | |
24 | * will release all slaves, marking them as down. | |
25 | * | |
26 | * ifenslave bond0 eth0 | |
27 | * will attach eth0 to bond0 as a slave. eth0 hw mac address will either | |
28 | * a: be used as initial mac address | |
29 | * b: if a hw mac address already is there, eth0's hw mac address | |
30 | * will then be set from bond0. | |
31 | * | |
1da177e4 LT |
32 | */ |
33 | ||
a4aee5c8 JP |
34 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
35 | ||
1da177e4 LT |
36 | #include <linux/kernel.h> |
37 | #include <linux/module.h> | |
1da177e4 LT |
38 | #include <linux/types.h> |
39 | #include <linux/fcntl.h> | |
40 | #include <linux/interrupt.h> | |
41 | #include <linux/ptrace.h> | |
42 | #include <linux/ioport.h> | |
43 | #include <linux/in.h> | |
169a3e66 | 44 | #include <net/ip.h> |
1da177e4 | 45 | #include <linux/ip.h> |
169a3e66 JV |
46 | #include <linux/tcp.h> |
47 | #include <linux/udp.h> | |
1da177e4 LT |
48 | #include <linux/slab.h> |
49 | #include <linux/string.h> | |
50 | #include <linux/init.h> | |
51 | #include <linux/timer.h> | |
52 | #include <linux/socket.h> | |
53 | #include <linux/ctype.h> | |
54 | #include <linux/inet.h> | |
55 | #include <linux/bitops.h> | |
3d632c3f | 56 | #include <linux/io.h> |
1da177e4 | 57 | #include <asm/system.h> |
1da177e4 | 58 | #include <asm/dma.h> |
3d632c3f | 59 | #include <linux/uaccess.h> |
1da177e4 LT |
60 | #include <linux/errno.h> |
61 | #include <linux/netdevice.h> | |
62 | #include <linux/inetdevice.h> | |
a816c7c7 | 63 | #include <linux/igmp.h> |
1da177e4 LT |
64 | #include <linux/etherdevice.h> |
65 | #include <linux/skbuff.h> | |
66 | #include <net/sock.h> | |
67 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
68 | #include <linux/smp.h> |
69 | #include <linux/if_ether.h> | |
70 | #include <net/arp.h> | |
71 | #include <linux/mii.h> | |
72 | #include <linux/ethtool.h> | |
73 | #include <linux/if_vlan.h> | |
74 | #include <linux/if_bonding.h> | |
b63bb739 | 75 | #include <linux/jiffies.h> |
e843fa50 | 76 | #include <linux/preempt.h> |
c3ade5ca | 77 | #include <net/route.h> |
457c4cbc | 78 | #include <net/net_namespace.h> |
ec87fd3b | 79 | #include <net/netns/generic.h> |
1da177e4 LT |
80 | #include "bonding.h" |
81 | #include "bond_3ad.h" | |
82 | #include "bond_alb.h" | |
83 | ||
84 | /*---------------------------- Module parameters ----------------------------*/ | |
85 | ||
86 | /* monitor all links that often (in milliseconds). <=0 disables monitoring */ | |
87 | #define BOND_LINK_MON_INTERV 0 | |
88 | #define BOND_LINK_ARP_INTERV 0 | |
89 | ||
90 | static int max_bonds = BOND_DEFAULT_MAX_BONDS; | |
bb1d9123 | 91 | static int tx_queues = BOND_DEFAULT_TX_QUEUES; |
7893b249 | 92 | static int num_grat_arp = 1; |
305d552a | 93 | static int num_unsol_na = 1; |
1da177e4 | 94 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
95 | static int updelay; |
96 | static int downdelay; | |
1da177e4 | 97 | static int use_carrier = 1; |
3d632c3f SH |
98 | static char *mode; |
99 | static char *primary; | |
a549952a | 100 | static char *primary_reselect; |
3d632c3f SH |
101 | static char *lacp_rate; |
102 | static char *ad_select; | |
103 | static char *xmit_hash_policy; | |
1da177e4 | 104 | static int arp_interval = BOND_LINK_ARP_INTERV; |
3d632c3f SH |
105 | static char *arp_ip_target[BOND_MAX_ARP_TARGETS]; |
106 | static char *arp_validate; | |
107 | static char *fail_over_mac; | |
ebd8e497 | 108 | static int all_slaves_active = 0; |
d2991f75 | 109 | static struct bond_params bonding_defaults; |
c2952c31 | 110 | static int resend_igmp = BOND_DEFAULT_RESEND_IGMP; |
1da177e4 LT |
111 | |
112 | module_param(max_bonds, int, 0); | |
113 | MODULE_PARM_DESC(max_bonds, "Max number of bonded devices"); | |
bb1d9123 AG |
114 | module_param(tx_queues, int, 0); |
115 | MODULE_PARM_DESC(tx_queues, "Max number of transmit queues (default = 16)"); | |
7893b249 MS |
116 | module_param(num_grat_arp, int, 0644); |
117 | MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packets to send on failover event"); | |
305d552a BH |
118 | module_param(num_unsol_na, int, 0644); |
119 | MODULE_PARM_DESC(num_unsol_na, "Number of unsolicited IPv6 Neighbor Advertisements packets to send on failover event"); | |
1da177e4 LT |
120 | module_param(miimon, int, 0); |
121 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
122 | module_param(updelay, int, 0); | |
123 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
124 | module_param(downdelay, int, 0); | |
2ac47660 MW |
125 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
126 | "in milliseconds"); | |
1da177e4 | 127 | module_param(use_carrier, int, 0); |
2ac47660 MW |
128 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
129 | "0 for off, 1 for on (default)"); | |
1da177e4 | 130 | module_param(mode, charp, 0); |
2ac47660 MW |
131 | MODULE_PARM_DESC(mode, "Mode of operation : 0 for balance-rr, " |
132 | "1 for active-backup, 2 for balance-xor, " | |
133 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
134 | "6 for balance-alb"); | |
1da177e4 LT |
135 | module_param(primary, charp, 0); |
136 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
137 | module_param(primary_reselect, charp, 0); |
138 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
139 | "once it comes up; " | |
140 | "0 for always (default), " | |
141 | "1 for only if speed of primary is " | |
142 | "better, " | |
143 | "2 for only on active slave " | |
144 | "failure"); | |
1da177e4 | 145 | module_param(lacp_rate, charp, 0); |
2ac47660 MW |
146 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner " |
147 | "(slow/fast)"); | |
fd989c83 JV |
148 | module_param(ad_select, charp, 0); |
149 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic: stable (0, default), bandwidth (1), count (2)"); | |
169a3e66 | 150 | module_param(xmit_hash_policy, charp, 0); |
2ac47660 MW |
151 | MODULE_PARM_DESC(xmit_hash_policy, "XOR hashing method: 0 for layer 2 (default)" |
152 | ", 1 for layer 3+4"); | |
1da177e4 LT |
153 | module_param(arp_interval, int, 0); |
154 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
155 | module_param_array(arp_ip_target, charp, NULL, 0); | |
156 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 JV |
157 | module_param(arp_validate, charp, 0); |
158 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes: none (default), active, backup or all"); | |
3915c1e8 JV |
159 | module_param(fail_over_mac, charp, 0); |
160 | 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 |
161 | module_param(all_slaves_active, int, 0); |
162 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
163 | "by setting active flag for all slaves. " | |
164 | "0 for never (default), 1 for always."); | |
c2952c31 FL |
165 | module_param(resend_igmp, int, 0); |
166 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on link failure"); | |
1da177e4 LT |
167 | |
168 | /*----------------------------- Global variables ----------------------------*/ | |
169 | ||
e843fa50 | 170 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 171 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
172 | #endif |
173 | ||
f99189b1 | 174 | int bond_net_id __read_mostly; |
1da177e4 | 175 | |
3d632c3f SH |
176 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
177 | static int arp_ip_count; | |
1da177e4 | 178 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
179 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
180 | static int lacp_fast; | |
1da177e4 | 181 | |
e97fd7c6 | 182 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
183 | { "slow", AD_LACP_SLOW}, |
184 | { "fast", AD_LACP_FAST}, | |
185 | { NULL, -1}, | |
186 | }; | |
187 | ||
e97fd7c6 | 188 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
189 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
190 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
191 | { "balance-xor", BOND_MODE_XOR}, | |
192 | { "broadcast", BOND_MODE_BROADCAST}, | |
193 | { "802.3ad", BOND_MODE_8023AD}, | |
194 | { "balance-tlb", BOND_MODE_TLB}, | |
195 | { "balance-alb", BOND_MODE_ALB}, | |
196 | { NULL, -1}, | |
197 | }; | |
198 | ||
e97fd7c6 | 199 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
200 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
201 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 202 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
203 | { NULL, -1}, |
204 | }; | |
205 | ||
e97fd7c6 | 206 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
207 | { "none", BOND_ARP_VALIDATE_NONE}, |
208 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
209 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
210 | { "all", BOND_ARP_VALIDATE_ALL}, | |
211 | { NULL, -1}, | |
212 | }; | |
213 | ||
e97fd7c6 | 214 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
215 | { "none", BOND_FOM_NONE}, |
216 | { "active", BOND_FOM_ACTIVE}, | |
217 | { "follow", BOND_FOM_FOLLOW}, | |
218 | { NULL, -1}, | |
219 | }; | |
220 | ||
a549952a JP |
221 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
222 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
223 | { "better", BOND_PRI_RESELECT_BETTER}, | |
224 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
225 | { NULL, -1}, | |
226 | }; | |
227 | ||
fd989c83 JV |
228 | struct bond_parm_tbl ad_select_tbl[] = { |
229 | { "stable", BOND_AD_STABLE}, | |
230 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
231 | { "count", BOND_AD_COUNT}, | |
232 | { NULL, -1}, | |
233 | }; | |
234 | ||
1da177e4 LT |
235 | /*-------------------------- Forward declarations ---------------------------*/ |
236 | ||
c3ade5ca | 237 | static void bond_send_gratuitous_arp(struct bonding *bond); |
181470fc | 238 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 239 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
240 | |
241 | /*---------------------------- General routines -----------------------------*/ | |
242 | ||
bd33acc3 | 243 | const char *bond_mode_name(int mode) |
1da177e4 | 244 | { |
77afc92b HE |
245 | static const char *names[] = { |
246 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
247 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
248 | [BOND_MODE_XOR] = "load balancing (xor)", | |
249 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 250 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
251 | [BOND_MODE_TLB] = "transmit load balancing", |
252 | [BOND_MODE_ALB] = "adaptive load balancing", | |
253 | }; | |
254 | ||
255 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 256 | return "unknown"; |
77afc92b HE |
257 | |
258 | return names[mode]; | |
1da177e4 LT |
259 | } |
260 | ||
261 | /*---------------------------------- VLAN -----------------------------------*/ | |
262 | ||
263 | /** | |
264 | * bond_add_vlan - add a new vlan id on bond | |
265 | * @bond: bond that got the notification | |
266 | * @vlan_id: the vlan id to add | |
267 | * | |
268 | * Returns -ENOMEM if allocation failed. | |
269 | */ | |
270 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
271 | { | |
272 | struct vlan_entry *vlan; | |
273 | ||
5a03cdb7 | 274 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 275 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 276 | |
305d552a | 277 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 278 | if (!vlan) |
1da177e4 | 279 | return -ENOMEM; |
1da177e4 LT |
280 | |
281 | INIT_LIST_HEAD(&vlan->vlan_list); | |
282 | vlan->vlan_id = vlan_id; | |
283 | ||
284 | write_lock_bh(&bond->lock); | |
285 | ||
286 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
287 | ||
288 | write_unlock_bh(&bond->lock); | |
289 | ||
5a03cdb7 | 290 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
291 | |
292 | return 0; | |
293 | } | |
294 | ||
295 | /** | |
296 | * bond_del_vlan - delete a vlan id from bond | |
297 | * @bond: bond that got the notification | |
298 | * @vlan_id: the vlan id to delete | |
299 | * | |
300 | * returns -ENODEV if @vlan_id was not found in @bond. | |
301 | */ | |
302 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
303 | { | |
0883beca | 304 | struct vlan_entry *vlan; |
1da177e4 LT |
305 | int res = -ENODEV; |
306 | ||
5a03cdb7 | 307 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 | 308 | |
e843fa50 | 309 | block_netpoll_tx(); |
1da177e4 LT |
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); | |
e843fa50 | 344 | unblock_netpoll_tx(); |
1da177e4 LT |
345 | return res; |
346 | } | |
347 | ||
348 | /** | |
349 | * bond_has_challenged_slaves | |
350 | * @bond: the bond we're working on | |
351 | * | |
352 | * Searches the slave list. Returns 1 if a vlan challenged slave | |
353 | * was found, 0 otherwise. | |
354 | * | |
355 | * Assumes bond->lock is held. | |
356 | */ | |
357 | static int bond_has_challenged_slaves(struct bonding *bond) | |
358 | { | |
359 | struct slave *slave; | |
360 | int i; | |
361 | ||
362 | bond_for_each_slave(bond, slave, i) { | |
363 | if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 364 | pr_debug("found VLAN challenged slave - %s\n", |
a4aee5c8 | 365 | slave->dev->name); |
1da177e4 LT |
366 | return 1; |
367 | } | |
368 | } | |
369 | ||
5a03cdb7 | 370 | pr_debug("no VLAN challenged slaves found\n"); |
1da177e4 LT |
371 | return 0; |
372 | } | |
373 | ||
374 | /** | |
375 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
376 | * @bond: the bond we're working on | |
377 | * @curr: item we're advancing from | |
378 | * | |
379 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
380 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 381 | * |
1da177e4 LT |
382 | * Caller must hold bond->lock |
383 | */ | |
384 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
385 | { | |
386 | struct vlan_entry *next, *last; | |
387 | ||
3d632c3f | 388 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 389 | return NULL; |
1da177e4 LT |
390 | |
391 | if (!curr) { | |
392 | next = list_entry(bond->vlan_list.next, | |
393 | struct vlan_entry, vlan_list); | |
394 | } else { | |
395 | last = list_entry(bond->vlan_list.prev, | |
396 | struct vlan_entry, vlan_list); | |
397 | if (last == curr) { | |
398 | next = list_entry(bond->vlan_list.next, | |
399 | struct vlan_entry, vlan_list); | |
400 | } else { | |
401 | next = list_entry(curr->vlan_list.next, | |
402 | struct vlan_entry, vlan_list); | |
403 | } | |
404 | } | |
405 | ||
406 | return next; | |
407 | } | |
408 | ||
409 | /** | |
410 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 411 | * |
1da177e4 LT |
412 | * @bond: bond device that got this skb for tx. |
413 | * @skb: hw accel VLAN tagged skb to transmit | |
414 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 415 | */ |
3d632c3f SH |
416 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
417 | struct net_device *slave_dev) | |
1da177e4 | 418 | { |
8387451e | 419 | skb->dev = slave_dev; |
1da177e4 | 420 | skb->priority = 1; |
080e4130 | 421 | if (unlikely(netpoll_tx_running(slave_dev))) |
8a8efa22 | 422 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 423 | else |
f6dc31a8 | 424 | dev_queue_xmit(skb); |
1da177e4 LT |
425 | |
426 | return 0; | |
427 | } | |
428 | ||
429 | /* | |
430 | * In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid | |
431 | * and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a | |
432 | * lock because: | |
433 | * a. This operation is performed in IOCTL context, | |
434 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
435 | * c. Holding a lock with BH disabled while directly calling a base driver | |
436 | * entry point is generally a BAD idea. | |
3d632c3f | 437 | * |
1da177e4 LT |
438 | * The design of synchronization/protection for this operation in the 8021q |
439 | * module is good for one or more VLAN devices over a single physical device | |
440 | * and cannot be extended for a teaming solution like bonding, so there is a | |
441 | * potential race condition here where a net device from the vlan group might | |
442 | * be referenced (either by a base driver or the 8021q code) while it is being | |
443 | * removed from the system. However, it turns out we're not making matters | |
444 | * worse, and if it works for regular VLAN usage it will work here too. | |
445 | */ | |
446 | ||
447 | /** | |
448 | * bond_vlan_rx_register - Propagates registration to slaves | |
449 | * @bond_dev: bonding net device that got called | |
450 | * @grp: vlan group being registered | |
451 | */ | |
3d632c3f SH |
452 | static void bond_vlan_rx_register(struct net_device *bond_dev, |
453 | struct vlan_group *grp) | |
1da177e4 | 454 | { |
454d7c9b | 455 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
456 | struct slave *slave; |
457 | int i; | |
458 | ||
a71fb881 | 459 | write_lock_bh(&bond->lock); |
1da177e4 | 460 | bond->vlgrp = grp; |
a71fb881 | 461 | write_unlock_bh(&bond->lock); |
1da177e4 LT |
462 | |
463 | bond_for_each_slave(bond, slave, i) { | |
464 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 465 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
466 | |
467 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
468 | slave_ops->ndo_vlan_rx_register) { |
469 | slave_ops->ndo_vlan_rx_register(slave_dev, grp); | |
1da177e4 LT |
470 | } |
471 | } | |
472 | } | |
473 | ||
474 | /** | |
475 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
476 | * @bond_dev: bonding net device that got called | |
477 | * @vid: vlan id being added | |
478 | */ | |
479 | static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) | |
480 | { | |
454d7c9b | 481 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
482 | struct slave *slave; |
483 | int i, res; | |
484 | ||
485 | bond_for_each_slave(bond, slave, i) { | |
486 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 487 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
488 | |
489 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a SH |
490 | slave_ops->ndo_vlan_rx_add_vid) { |
491 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vid); | |
1da177e4 LT |
492 | } |
493 | } | |
494 | ||
495 | res = bond_add_vlan(bond, vid); | |
496 | if (res) { | |
a4aee5c8 | 497 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 LT |
498 | bond_dev->name, vid); |
499 | } | |
500 | } | |
501 | ||
502 | /** | |
503 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
504 | * @bond_dev: bonding net device that got called | |
505 | * @vid: vlan id being removed | |
506 | */ | |
507 | static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) | |
508 | { | |
454d7c9b | 509 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
510 | struct slave *slave; |
511 | struct net_device *vlan_dev; | |
512 | int i, res; | |
513 | ||
514 | bond_for_each_slave(bond, slave, i) { | |
515 | struct net_device *slave_dev = slave->dev; | |
eb7cc59a | 516 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
517 | |
518 | if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) && | |
eb7cc59a | 519 | slave_ops->ndo_vlan_rx_kill_vid) { |
1da177e4 LT |
520 | /* Save and then restore vlan_dev in the grp array, |
521 | * since the slave's driver might clear it. | |
522 | */ | |
5c15bdec | 523 | vlan_dev = vlan_group_get_device(bond->vlgrp, vid); |
eb7cc59a | 524 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vid); |
5c15bdec | 525 | vlan_group_set_device(bond->vlgrp, vid, vlan_dev); |
1da177e4 LT |
526 | } |
527 | } | |
528 | ||
529 | res = bond_del_vlan(bond, vid); | |
530 | if (res) { | |
a4aee5c8 | 531 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 LT |
532 | bond_dev->name, vid); |
533 | } | |
534 | } | |
535 | ||
536 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
537 | { | |
538 | struct vlan_entry *vlan; | |
eb7cc59a | 539 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 540 | |
f35188fa | 541 | if (!bond->vlgrp) |
03dc2f4c | 542 | return; |
1da177e4 LT |
543 | |
544 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
545 | slave_ops->ndo_vlan_rx_register) |
546 | slave_ops->ndo_vlan_rx_register(slave_dev, bond->vlgrp); | |
1da177e4 LT |
547 | |
548 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 549 | !(slave_ops->ndo_vlan_rx_add_vid)) |
03dc2f4c | 550 | return; |
1da177e4 | 551 | |
eb7cc59a SH |
552 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) |
553 | slave_ops->ndo_vlan_rx_add_vid(slave_dev, vlan->vlan_id); | |
1da177e4 LT |
554 | } |
555 | ||
3d632c3f SH |
556 | static void bond_del_vlans_from_slave(struct bonding *bond, |
557 | struct net_device *slave_dev) | |
1da177e4 | 558 | { |
eb7cc59a | 559 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 LT |
560 | struct vlan_entry *vlan; |
561 | struct net_device *vlan_dev; | |
562 | ||
f35188fa | 563 | if (!bond->vlgrp) |
03dc2f4c | 564 | return; |
1da177e4 LT |
565 | |
566 | if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) || | |
eb7cc59a | 567 | !(slave_ops->ndo_vlan_rx_kill_vid)) |
1da177e4 | 568 | goto unreg; |
1da177e4 LT |
569 | |
570 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
f35188fa JV |
571 | if (!vlan->vlan_id) |
572 | continue; | |
1da177e4 LT |
573 | /* Save and then restore vlan_dev in the grp array, |
574 | * since the slave's driver might clear it. | |
575 | */ | |
5c15bdec | 576 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
eb7cc59a | 577 | slave_ops->ndo_vlan_rx_kill_vid(slave_dev, vlan->vlan_id); |
5c15bdec | 578 | vlan_group_set_device(bond->vlgrp, vlan->vlan_id, vlan_dev); |
1da177e4 LT |
579 | } |
580 | ||
581 | unreg: | |
582 | if ((slave_dev->features & NETIF_F_HW_VLAN_RX) && | |
eb7cc59a SH |
583 | slave_ops->ndo_vlan_rx_register) |
584 | slave_ops->ndo_vlan_rx_register(slave_dev, NULL); | |
1da177e4 LT |
585 | } |
586 | ||
587 | /*------------------------------- Link status -------------------------------*/ | |
588 | ||
ff59c456 JV |
589 | /* |
590 | * Set the carrier state for the master according to the state of its | |
591 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
592 | * do special 802.3ad magic. | |
593 | * | |
594 | * Returns zero if carrier state does not change, nonzero if it does. | |
595 | */ | |
596 | static int bond_set_carrier(struct bonding *bond) | |
597 | { | |
598 | struct slave *slave; | |
599 | int i; | |
600 | ||
601 | if (bond->slave_cnt == 0) | |
602 | goto down; | |
603 | ||
604 | if (bond->params.mode == BOND_MODE_8023AD) | |
605 | return bond_3ad_set_carrier(bond); | |
606 | ||
607 | bond_for_each_slave(bond, slave, i) { | |
608 | if (slave->link == BOND_LINK_UP) { | |
609 | if (!netif_carrier_ok(bond->dev)) { | |
610 | netif_carrier_on(bond->dev); | |
611 | return 1; | |
612 | } | |
613 | return 0; | |
614 | } | |
615 | } | |
616 | ||
617 | down: | |
618 | if (netif_carrier_ok(bond->dev)) { | |
619 | netif_carrier_off(bond->dev); | |
620 | return 1; | |
621 | } | |
622 | return 0; | |
623 | } | |
624 | ||
1da177e4 LT |
625 | /* |
626 | * Get link speed and duplex from the slave's base driver | |
627 | * using ethtool. If for some reason the call fails or the | |
628 | * values are invalid, fake speed and duplex to 100/Full | |
629 | * and return error. | |
630 | */ | |
631 | static int bond_update_speed_duplex(struct slave *slave) | |
632 | { | |
633 | struct net_device *slave_dev = slave->dev; | |
5d30530e DD |
634 | struct ethtool_cmd etool = { .cmd = ETHTOOL_GSET }; |
635 | u32 slave_speed; | |
61a44b9c | 636 | int res; |
1da177e4 LT |
637 | |
638 | /* Fake speed and duplex */ | |
639 | slave->speed = SPEED_100; | |
640 | slave->duplex = DUPLEX_FULL; | |
641 | ||
61a44b9c MW |
642 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
643 | return -1; | |
1da177e4 | 644 | |
61a44b9c MW |
645 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
646 | if (res < 0) | |
1da177e4 | 647 | return -1; |
1da177e4 | 648 | |
5d30530e DD |
649 | slave_speed = ethtool_cmd_speed(&etool); |
650 | switch (slave_speed) { | |
1da177e4 LT |
651 | case SPEED_10: |
652 | case SPEED_100: | |
653 | case SPEED_1000: | |
94dbffd5 | 654 | case SPEED_10000: |
1da177e4 LT |
655 | break; |
656 | default: | |
657 | return -1; | |
658 | } | |
659 | ||
660 | switch (etool.duplex) { | |
661 | case DUPLEX_FULL: | |
662 | case DUPLEX_HALF: | |
663 | break; | |
664 | default: | |
665 | return -1; | |
666 | } | |
667 | ||
5d30530e | 668 | slave->speed = slave_speed; |
1da177e4 LT |
669 | slave->duplex = etool.duplex; |
670 | ||
671 | return 0; | |
672 | } | |
673 | ||
674 | /* | |
675 | * if <dev> supports MII link status reporting, check its link status. | |
676 | * | |
677 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 678 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
679 | * |
680 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
681 | * can't tell and just pretend it is), or 0, meaning that the link is | |
682 | * down. | |
683 | * | |
684 | * If reporting is non-zero, instead of faking link up, return -1 if | |
685 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
686 | * support them). If use_carrier is set, return whatever it says. | |
687 | * It'd be nice if there was a good way to tell if a driver supports | |
688 | * netif_carrier, but there really isn't. | |
689 | */ | |
3d632c3f SH |
690 | static int bond_check_dev_link(struct bonding *bond, |
691 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 692 | { |
eb7cc59a | 693 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 694 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
695 | struct ifreq ifr; |
696 | struct mii_ioctl_data *mii; | |
1da177e4 | 697 | |
6c988853 PG |
698 | if (!reporting && !netif_running(slave_dev)) |
699 | return 0; | |
700 | ||
eb7cc59a | 701 | if (bond->params.use_carrier) |
1da177e4 | 702 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 703 | |
29112f4e JP |
704 | /* Try to get link status using Ethtool first. */ |
705 | if (slave_dev->ethtool_ops) { | |
706 | if (slave_dev->ethtool_ops->get_link) { | |
707 | u32 link; | |
708 | ||
709 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
710 | ||
711 | return link ? BMSR_LSTATUS : 0; | |
712 | } | |
713 | } | |
714 | ||
3d632c3f | 715 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 716 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
717 | if (ioctl) { |
718 | /* TODO: set pointer to correct ioctl on a per team member */ | |
719 | /* bases to make this more efficient. that is, once */ | |
720 | /* we determine the correct ioctl, we will always */ | |
721 | /* call it and not the others for that team */ | |
722 | /* member. */ | |
723 | ||
724 | /* | |
725 | * We cannot assume that SIOCGMIIPHY will also read a | |
726 | * register; not all network drivers (e.g., e100) | |
727 | * support that. | |
728 | */ | |
729 | ||
730 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
731 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
732 | mii = if_mii(&ifr); | |
733 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
734 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
735 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
736 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
737 | } |
738 | } | |
739 | ||
1da177e4 LT |
740 | /* |
741 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 742 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
743 | * cannot report link status). If not reporting, pretend |
744 | * we're ok. | |
745 | */ | |
3d632c3f | 746 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
747 | } |
748 | ||
749 | /*----------------------------- Multicast list ------------------------------*/ | |
750 | ||
1da177e4 LT |
751 | /* |
752 | * Push the promiscuity flag down to appropriate slaves | |
753 | */ | |
7e1a1ac1 | 754 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 755 | { |
7e1a1ac1 | 756 | int err = 0; |
1da177e4 LT |
757 | if (USES_PRIMARY(bond->params.mode)) { |
758 | /* write lock already acquired */ | |
759 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
760 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
761 | inc); | |
1da177e4 LT |
762 | } |
763 | } else { | |
764 | struct slave *slave; | |
765 | int i; | |
766 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
767 | err = dev_set_promiscuity(slave->dev, inc); |
768 | if (err) | |
769 | return err; | |
1da177e4 LT |
770 | } |
771 | } | |
7e1a1ac1 | 772 | return err; |
1da177e4 LT |
773 | } |
774 | ||
775 | /* | |
776 | * Push the allmulti flag down to all slaves | |
777 | */ | |
7e1a1ac1 | 778 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 779 | { |
7e1a1ac1 | 780 | int err = 0; |
1da177e4 LT |
781 | if (USES_PRIMARY(bond->params.mode)) { |
782 | /* write lock already acquired */ | |
783 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
784 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
785 | inc); | |
1da177e4 LT |
786 | } |
787 | } else { | |
788 | struct slave *slave; | |
789 | int i; | |
790 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
791 | err = dev_set_allmulti(slave->dev, inc); |
792 | if (err) | |
793 | return err; | |
1da177e4 LT |
794 | } |
795 | } | |
7e1a1ac1 | 796 | return err; |
1da177e4 LT |
797 | } |
798 | ||
799 | /* | |
800 | * Add a Multicast address to slaves | |
801 | * according to mode | |
802 | */ | |
22bedad3 | 803 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
804 | { |
805 | if (USES_PRIMARY(bond->params.mode)) { | |
806 | /* write lock already acquired */ | |
3d632c3f | 807 | if (bond->curr_active_slave) |
22bedad3 | 808 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
809 | } else { |
810 | struct slave *slave; | |
811 | int i; | |
3d632c3f SH |
812 | |
813 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 814 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
815 | } |
816 | } | |
817 | ||
818 | /* | |
819 | * Remove a multicast address from slave | |
820 | * according to mode | |
821 | */ | |
22bedad3 | 822 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
823 | { |
824 | if (USES_PRIMARY(bond->params.mode)) { | |
825 | /* write lock already acquired */ | |
3d632c3f | 826 | if (bond->curr_active_slave) |
22bedad3 | 827 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
828 | } else { |
829 | struct slave *slave; | |
830 | int i; | |
831 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 832 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
833 | } |
834 | } | |
835 | } | |
836 | ||
a816c7c7 | 837 | |
5a37e8ca | 838 | static void __bond_resend_igmp_join_requests(struct net_device *dev) |
a816c7c7 JV |
839 | { |
840 | struct in_device *in_dev; | |
a816c7c7 JV |
841 | |
842 | rcu_read_lock(); | |
5a37e8ca | 843 | in_dev = __in_dev_get_rcu(dev); |
866f3b25 | 844 | if (in_dev) |
24912420 | 845 | ip_mc_rejoin_groups(in_dev); |
a816c7c7 JV |
846 | rcu_read_unlock(); |
847 | } | |
848 | ||
5a37e8ca FL |
849 | /* |
850 | * Retrieve the list of registered multicast addresses for the bonding | |
851 | * device and retransmit an IGMP JOIN request to the current active | |
852 | * slave. | |
853 | */ | |
854 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
855 | { | |
856 | struct net_device *vlan_dev; | |
857 | struct vlan_entry *vlan; | |
858 | ||
859 | read_lock(&bond->lock); | |
860 | ||
861 | /* rejoin all groups on bond device */ | |
862 | __bond_resend_igmp_join_requests(bond->dev); | |
863 | ||
864 | /* rejoin all groups on vlan devices */ | |
865 | if (bond->vlgrp) { | |
866 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
867 | vlan_dev = vlan_group_get_device(bond->vlgrp, | |
868 | vlan->vlan_id); | |
869 | if (vlan_dev) | |
870 | __bond_resend_igmp_join_requests(vlan_dev); | |
871 | } | |
872 | } | |
873 | ||
c2952c31 FL |
874 | if (--bond->igmp_retrans > 0) |
875 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); | |
876 | ||
5a37e8ca FL |
877 | read_unlock(&bond->lock); |
878 | } | |
879 | ||
379b7383 | 880 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
881 | { |
882 | struct bonding *bond = container_of(work, struct bonding, | |
883 | mcast_work.work); | |
884 | bond_resend_igmp_join_requests(bond); | |
885 | } | |
886 | ||
1da177e4 LT |
887 | /* |
888 | * flush all members of flush->mc_list from device dev->mc_list | |
889 | */ | |
3d632c3f SH |
890 | static void bond_mc_list_flush(struct net_device *bond_dev, |
891 | struct net_device *slave_dev) | |
1da177e4 | 892 | { |
454d7c9b | 893 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 894 | struct netdev_hw_addr *ha; |
1da177e4 | 895 | |
22bedad3 JP |
896 | netdev_for_each_mc_addr(ha, bond_dev) |
897 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
898 | |
899 | if (bond->params.mode == BOND_MODE_8023AD) { | |
900 | /* del lacpdu mc addr from mc list */ | |
901 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
902 | ||
22bedad3 | 903 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
904 | } |
905 | } | |
906 | ||
907 | /*--------------------------- Active slave change ---------------------------*/ | |
908 | ||
909 | /* | |
910 | * Update the mc list and multicast-related flags for the new and | |
911 | * old active slaves (if any) according to the multicast mode, and | |
912 | * promiscuous flags unconditionally. | |
913 | */ | |
3d632c3f SH |
914 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
915 | struct slave *old_active) | |
1da177e4 | 916 | { |
22bedad3 | 917 | struct netdev_hw_addr *ha; |
1da177e4 | 918 | |
3d632c3f | 919 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
920 | /* nothing to do - mc list is already up-to-date on |
921 | * all slaves | |
922 | */ | |
923 | return; | |
1da177e4 LT |
924 | |
925 | if (old_active) { | |
3d632c3f | 926 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 927 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 928 | |
3d632c3f | 929 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 930 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 931 | |
22bedad3 JP |
932 | netdev_for_each_mc_addr(ha, bond->dev) |
933 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
934 | } |
935 | ||
936 | if (new_active) { | |
7e1a1ac1 | 937 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 938 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 939 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 940 | |
3d632c3f | 941 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 942 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 943 | |
22bedad3 JP |
944 | netdev_for_each_mc_addr(ha, bond->dev) |
945 | dev_mc_add(new_active->dev, ha->addr); | |
1da177e4 LT |
946 | } |
947 | } | |
948 | ||
3915c1e8 JV |
949 | /* |
950 | * bond_do_fail_over_mac | |
951 | * | |
952 | * Perform special MAC address swapping for fail_over_mac settings | |
953 | * | |
954 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
955 | */ | |
956 | static void bond_do_fail_over_mac(struct bonding *bond, | |
957 | struct slave *new_active, | |
958 | struct slave *old_active) | |
1f78d9f9 HE |
959 | __releases(&bond->curr_slave_lock) |
960 | __releases(&bond->lock) | |
961 | __acquires(&bond->lock) | |
962 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
963 | { |
964 | u8 tmp_mac[ETH_ALEN]; | |
965 | struct sockaddr saddr; | |
966 | int rv; | |
967 | ||
968 | switch (bond->params.fail_over_mac) { | |
969 | case BOND_FOM_ACTIVE: | |
970 | if (new_active) | |
971 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
972 | new_active->dev->addr_len); | |
973 | break; | |
974 | case BOND_FOM_FOLLOW: | |
975 | /* | |
976 | * if new_active && old_active, swap them | |
977 | * if just old_active, do nothing (going to no active slave) | |
978 | * if just new_active, set new_active to bond's MAC | |
979 | */ | |
980 | if (!new_active) | |
981 | return; | |
982 | ||
983 | write_unlock_bh(&bond->curr_slave_lock); | |
984 | read_unlock(&bond->lock); | |
985 | ||
986 | if (old_active) { | |
987 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
988 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
989 | ETH_ALEN); | |
990 | saddr.sa_family = new_active->dev->type; | |
991 | } else { | |
992 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
993 | saddr.sa_family = bond->dev->type; | |
994 | } | |
995 | ||
996 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
997 | if (rv) { | |
a4aee5c8 | 998 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
999 | bond->dev->name, -rv, new_active->dev->name); |
1000 | goto out; | |
1001 | } | |
1002 | ||
1003 | if (!old_active) | |
1004 | goto out; | |
1005 | ||
1006 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1007 | saddr.sa_family = old_active->dev->type; | |
1008 | ||
1009 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1010 | if (rv) | |
a4aee5c8 | 1011 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1012 | bond->dev->name, -rv, new_active->dev->name); |
1013 | out: | |
1014 | read_lock(&bond->lock); | |
1015 | write_lock_bh(&bond->curr_slave_lock); | |
1016 | break; | |
1017 | default: | |
a4aee5c8 | 1018 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
1019 | bond->dev->name, bond->params.fail_over_mac); |
1020 | break; | |
1021 | } | |
1022 | ||
1023 | } | |
1024 | ||
a549952a JP |
1025 | static bool bond_should_change_active(struct bonding *bond) |
1026 | { | |
1027 | struct slave *prim = bond->primary_slave; | |
1028 | struct slave *curr = bond->curr_active_slave; | |
1029 | ||
1030 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1031 | return true; | |
1032 | if (bond->force_primary) { | |
1033 | bond->force_primary = false; | |
1034 | return true; | |
1035 | } | |
1036 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1037 | (prim->speed < curr->speed || | |
1038 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1039 | return false; | |
1040 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1041 | return false; | |
1042 | return true; | |
1043 | } | |
3915c1e8 | 1044 | |
1da177e4 LT |
1045 | /** |
1046 | * find_best_interface - select the best available slave to be the active one | |
1047 | * @bond: our bonding struct | |
1048 | * | |
1049 | * Warning: Caller must hold curr_slave_lock for writing. | |
1050 | */ | |
1051 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1052 | { | |
1053 | struct slave *new_active, *old_active; | |
1054 | struct slave *bestslave = NULL; | |
1055 | int mintime = bond->params.updelay; | |
1056 | int i; | |
1057 | ||
49b4ad92 | 1058 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1059 | |
1060 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1061 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1062 | new_active = bond->first_slave; |
3d632c3f | 1063 | else |
1da177e4 | 1064 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1065 | } |
1066 | ||
1da177e4 | 1067 | if ((bond->primary_slave) && |
a549952a JP |
1068 | bond->primary_slave->link == BOND_LINK_UP && |
1069 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1070 | new_active = bond->primary_slave; |
1071 | } | |
1072 | ||
1073 | /* remember where to stop iterating over the slaves */ | |
1074 | old_active = new_active; | |
1075 | ||
1076 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1077 | if (new_active->link == BOND_LINK_UP) { |
1078 | return new_active; | |
1079 | } else if (new_active->link == BOND_LINK_BACK && | |
1080 | IS_UP(new_active->dev)) { | |
1081 | /* link up, but waiting for stabilization */ | |
1082 | if (new_active->delay < mintime) { | |
1083 | mintime = new_active->delay; | |
1084 | bestslave = new_active; | |
1da177e4 LT |
1085 | } |
1086 | } | |
1087 | } | |
1088 | ||
1089 | return bestslave; | |
1090 | } | |
1091 | ||
1092 | /** | |
1093 | * change_active_interface - change the active slave into the specified one | |
1094 | * @bond: our bonding struct | |
1095 | * @new: the new slave to make the active one | |
1096 | * | |
1097 | * Set the new slave to the bond's settings and unset them on the old | |
1098 | * curr_active_slave. | |
1099 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1100 | * | |
1101 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1102 | * because it is apparently the best available slave we have, even though its | |
1103 | * updelay hasn't timed out yet. | |
1104 | * | |
3915c1e8 JV |
1105 | * If new_active is not NULL, caller must hold bond->lock for read and |
1106 | * curr_slave_lock for write_bh. | |
1da177e4 | 1107 | */ |
a77b5325 | 1108 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1109 | { |
1110 | struct slave *old_active = bond->curr_active_slave; | |
1111 | ||
3d632c3f | 1112 | if (old_active == new_active) |
1da177e4 | 1113 | return; |
1da177e4 LT |
1114 | |
1115 | if (new_active) { | |
b2220cad JV |
1116 | new_active->jiffies = jiffies; |
1117 | ||
1da177e4 LT |
1118 | if (new_active->link == BOND_LINK_BACK) { |
1119 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1120 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1121 | bond->dev->name, new_active->dev->name, | |
1122 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1123 | } |
1124 | ||
1125 | new_active->delay = 0; | |
1126 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1127 | |
3d632c3f | 1128 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1129 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1130 | |
58402054 | 1131 | if (bond_is_lb(bond)) |
1da177e4 | 1132 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1133 | } else { |
1134 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1135 | pr_info("%s: making interface %s the new active one.\n", |
1136 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1137 | } |
1138 | } | |
1139 | } | |
1140 | ||
3d632c3f | 1141 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1142 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1143 | |
58402054 | 1144 | if (bond_is_lb(bond)) { |
1da177e4 | 1145 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1146 | if (old_active) |
1147 | bond_set_slave_inactive_flags(old_active); | |
1148 | if (new_active) | |
1149 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1150 | } else { |
1151 | bond->curr_active_slave = new_active; | |
1152 | } | |
c3ade5ca JV |
1153 | |
1154 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1155 | if (old_active) |
c3ade5ca | 1156 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1157 | |
1158 | if (new_active) { | |
1159 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1160 | |
709f8a45 OG |
1161 | if (bond->params.fail_over_mac) |
1162 | bond_do_fail_over_mac(bond, new_active, | |
1163 | old_active); | |
3915c1e8 | 1164 | |
ffa95ed5 BH |
1165 | if (netif_running(bond->dev)) { |
1166 | bond->send_grat_arp = bond->params.num_grat_arp; | |
1167 | bond_send_gratuitous_arp(bond); | |
01f3109d | 1168 | |
ffa95ed5 BH |
1169 | bond->send_unsol_na = bond->params.num_unsol_na; |
1170 | bond_send_unsolicited_na(bond); | |
1171 | } | |
305d552a | 1172 | |
01f3109d OG |
1173 | write_unlock_bh(&bond->curr_slave_lock); |
1174 | read_unlock(&bond->lock); | |
1175 | ||
75c78500 | 1176 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1177 | |
1178 | read_lock(&bond->lock); | |
1179 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1180 | } |
c3ade5ca | 1181 | } |
a2fd940f | 1182 | |
5a37e8ca FL |
1183 | /* resend IGMP joins since active slave has changed or |
1184 | * all were sent on curr_active_slave */ | |
ffa95ed5 BH |
1185 | if (((USES_PRIMARY(bond->params.mode) && new_active) || |
1186 | bond->params.mode == BOND_MODE_ROUNDROBIN) && | |
1187 | netif_running(bond->dev)) { | |
c2952c31 | 1188 | bond->igmp_retrans = bond->params.resend_igmp; |
5a37e8ca | 1189 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); |
a2fd940f | 1190 | } |
1da177e4 LT |
1191 | } |
1192 | ||
1193 | /** | |
1194 | * bond_select_active_slave - select a new active slave, if needed | |
1195 | * @bond: our bonding struct | |
1196 | * | |
3d632c3f | 1197 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1198 | * - The old curr_active_slave has been released or lost its link. |
1199 | * - The primary_slave has got its link back. | |
1200 | * - A slave has got its link back and there's no old curr_active_slave. | |
1201 | * | |
3915c1e8 | 1202 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1203 | */ |
a77b5325 | 1204 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1205 | { |
1206 | struct slave *best_slave; | |
ff59c456 | 1207 | int rv; |
1da177e4 LT |
1208 | |
1209 | best_slave = bond_find_best_slave(bond); | |
1210 | if (best_slave != bond->curr_active_slave) { | |
1211 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1212 | rv = bond_set_carrier(bond); |
1213 | if (!rv) | |
1214 | return; | |
1215 | ||
1216 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1217 | pr_info("%s: first active interface up!\n", |
1218 | bond->dev->name); | |
ff59c456 | 1219 | } else { |
a4aee5c8 JP |
1220 | pr_info("%s: now running without any active interface !\n", |
1221 | bond->dev->name); | |
ff59c456 | 1222 | } |
1da177e4 LT |
1223 | } |
1224 | } | |
1225 | ||
1226 | /*--------------------------- slave list handling ---------------------------*/ | |
1227 | ||
1228 | /* | |
1229 | * This function attaches the slave to the end of list. | |
1230 | * | |
1231 | * bond->lock held for writing by caller. | |
1232 | */ | |
1233 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1234 | { | |
1235 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1236 | new_slave->next = new_slave; | |
1237 | new_slave->prev = new_slave; | |
1238 | bond->first_slave = new_slave; | |
1239 | } else { | |
1240 | new_slave->next = bond->first_slave; | |
1241 | new_slave->prev = bond->first_slave->prev; | |
1242 | new_slave->next->prev = new_slave; | |
1243 | new_slave->prev->next = new_slave; | |
1244 | } | |
1245 | ||
1246 | bond->slave_cnt++; | |
1247 | } | |
1248 | ||
1249 | /* | |
1250 | * This function detaches the slave from the list. | |
1251 | * WARNING: no check is made to verify if the slave effectively | |
1252 | * belongs to <bond>. | |
1253 | * Nothing is freed on return, structures are just unchained. | |
1254 | * If any slave pointer in bond was pointing to <slave>, | |
1255 | * it should be changed by the calling function. | |
1256 | * | |
1257 | * bond->lock held for writing by caller. | |
1258 | */ | |
1259 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1260 | { | |
3d632c3f | 1261 | if (slave->next) |
1da177e4 | 1262 | slave->next->prev = slave->prev; |
1da177e4 | 1263 | |
3d632c3f | 1264 | if (slave->prev) |
1da177e4 | 1265 | slave->prev->next = slave->next; |
1da177e4 LT |
1266 | |
1267 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1268 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1269 | bond->first_slave = slave->next; | |
1270 | } else { | |
1271 | bond->first_slave = NULL; /* slave was the last one */ | |
1272 | } | |
1273 | } | |
1274 | ||
1275 | slave->next = NULL; | |
1276 | slave->prev = NULL; | |
1277 | bond->slave_cnt--; | |
1278 | } | |
1279 | ||
f6dc31a8 | 1280 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1281 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1282 | { |
8a8efa22 AW |
1283 | struct netpoll *np; |
1284 | int err = 0; | |
f6dc31a8 | 1285 | |
8a8efa22 AW |
1286 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
1287 | err = -ENOMEM; | |
1288 | if (!np) | |
1289 | goto out; | |
1290 | ||
1291 | np->dev = slave->dev; | |
1292 | err = __netpoll_setup(np); | |
1293 | if (err) { | |
1294 | kfree(np); | |
1295 | goto out; | |
f6dc31a8 | 1296 | } |
8a8efa22 AW |
1297 | slave->np = np; |
1298 | out: | |
1299 | return err; | |
1300 | } | |
1301 | static inline void slave_disable_netpoll(struct slave *slave) | |
1302 | { | |
1303 | struct netpoll *np = slave->np; | |
1304 | ||
1305 | if (!np) | |
1306 | return; | |
1307 | ||
1308 | slave->np = NULL; | |
1309 | synchronize_rcu_bh(); | |
1310 | __netpoll_cleanup(np); | |
1311 | kfree(np); | |
1312 | } | |
1313 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1314 | { | |
1315 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1316 | return false; | |
1317 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1318 | return false; | |
1319 | return true; | |
f6dc31a8 WC |
1320 | } |
1321 | ||
1322 | static void bond_poll_controller(struct net_device *bond_dev) | |
1323 | { | |
8a8efa22 AW |
1324 | } |
1325 | ||
1326 | static void __bond_netpoll_cleanup(struct bonding *bond) | |
1327 | { | |
c2355e1a NH |
1328 | struct slave *slave; |
1329 | int i; | |
1330 | ||
8a8efa22 AW |
1331 | bond_for_each_slave(bond, slave, i) |
1332 | if (IS_UP(slave->dev)) | |
1333 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1334 | } |
f6dc31a8 WC |
1335 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1336 | { | |
1337 | struct bonding *bond = netdev_priv(bond_dev); | |
8a8efa22 AW |
1338 | |
1339 | read_lock(&bond->lock); | |
1340 | __bond_netpoll_cleanup(bond); | |
1341 | read_unlock(&bond->lock); | |
1342 | } | |
1343 | ||
1344 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) | |
1345 | { | |
1346 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1347 | struct slave *slave; |
8a8efa22 | 1348 | int i, err = 0; |
f6dc31a8 WC |
1349 | |
1350 | read_lock(&bond->lock); | |
f6dc31a8 | 1351 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1352 | err = slave_enable_netpoll(slave); |
1353 | if (err) { | |
1354 | __bond_netpoll_cleanup(bond); | |
1355 | break; | |
f6dc31a8 WC |
1356 | } |
1357 | } | |
1358 | read_unlock(&bond->lock); | |
8a8efa22 | 1359 | return err; |
f6dc31a8 WC |
1360 | } |
1361 | ||
8a8efa22 AW |
1362 | static struct netpoll_info *bond_netpoll_info(struct bonding *bond) |
1363 | { | |
1364 | return bond->dev->npinfo; | |
1365 | } | |
f6dc31a8 | 1366 | |
8a8efa22 AW |
1367 | #else |
1368 | static inline int slave_enable_netpoll(struct slave *slave) | |
1369 | { | |
1370 | return 0; | |
1371 | } | |
1372 | static inline void slave_disable_netpoll(struct slave *slave) | |
1373 | { | |
1374 | } | |
f6dc31a8 WC |
1375 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1376 | { | |
1377 | } | |
f6dc31a8 WC |
1378 | #endif |
1379 | ||
1da177e4 LT |
1380 | /*---------------------------------- IOCTL ----------------------------------*/ |
1381 | ||
4ad072c9 AB |
1382 | static int bond_sethwaddr(struct net_device *bond_dev, |
1383 | struct net_device *slave_dev) | |
1da177e4 | 1384 | { |
5a03cdb7 HE |
1385 | pr_debug("bond_dev=%p\n", bond_dev); |
1386 | pr_debug("slave_dev=%p\n", slave_dev); | |
1387 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1388 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1389 | return 0; | |
1390 | } | |
1391 | ||
7f353bf2 HX |
1392 | #define BOND_VLAN_FEATURES \ |
1393 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1394 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1395 | |
3d632c3f | 1396 | /* |
8e3babcd | 1397 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1398 | * feature bits are managed elsewhere, so preserve those feature bits |
1399 | * on the master device. | |
8531c5ff AK |
1400 | */ |
1401 | static int bond_compute_features(struct bonding *bond) | |
1402 | { | |
8531c5ff AK |
1403 | struct slave *slave; |
1404 | struct net_device *bond_dev = bond->dev; | |
04ed3e74 MM |
1405 | u32 features = bond_dev->features; |
1406 | u32 vlan_features = 0; | |
3158bf7d MS |
1407 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1408 | bond_dev->hard_header_len); | |
8e3babcd | 1409 | int i; |
8531c5ff | 1410 | |
7f353bf2 | 1411 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
c6e1a0d1 | 1412 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM | NETIF_F_NOCACHE_COPY; |
b63365a2 HX |
1413 | |
1414 | if (!bond->first_slave) | |
1415 | goto done; | |
1416 | ||
1417 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1418 | |
278339a4 | 1419 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1420 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1421 | features = netdev_increment_features(features, |
1422 | slave->dev->features, | |
1423 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1424 | vlan_features = netdev_increment_features(vlan_features, |
1425 | slave->dev->vlan_features, | |
1426 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1427 | if (slave->dev->hard_header_len > max_hard_header_len) |
1428 | max_hard_header_len = slave->dev->hard_header_len; | |
1429 | } | |
8531c5ff | 1430 | |
b63365a2 | 1431 | done: |
7f353bf2 | 1432 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
acd1130e MM |
1433 | bond_dev->features = netdev_fix_features(bond_dev, features); |
1434 | bond_dev->vlan_features = netdev_fix_features(bond_dev, vlan_features); | |
54ef3137 | 1435 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1436 | |
1437 | return 0; | |
1438 | } | |
1439 | ||
872254dd MS |
1440 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1441 | struct net_device *slave_dev) | |
1442 | { | |
454d7c9b | 1443 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1444 | |
00829823 | 1445 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1446 | |
1447 | bond_dev->type = slave_dev->type; | |
1448 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1449 | bond_dev->addr_len = slave_dev->addr_len; | |
1450 | ||
1451 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1452 | slave_dev->addr_len); | |
d90a162a | 1453 | bond->setup_by_slave = 1; |
872254dd MS |
1454 | } |
1455 | ||
5b2c4dd2 JP |
1456 | /* On bonding slaves other than the currently active slave, suppress |
1457 | * duplicates except for 802.3ad ETH_P_SLOW, alb non-mcast/bcast, and | |
1458 | * ARP on active-backup slaves with arp_validate enabled. | |
1459 | */ | |
1460 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1461 | struct slave *slave, |
1462 | struct bonding *bond) | |
5b2c4dd2 | 1463 | { |
2d7011ca | 1464 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1465 | if (slave_do_arp_validate(bond, slave) && |
5b2c4dd2 JP |
1466 | skb->protocol == __cpu_to_be16(ETH_P_ARP)) |
1467 | return false; | |
1468 | ||
0bd80dad | 1469 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1470 | skb->pkt_type != PACKET_BROADCAST && |
1471 | skb->pkt_type != PACKET_MULTICAST) | |
1472 | return false; | |
1473 | ||
0bd80dad | 1474 | if (bond->params.mode == BOND_MODE_8023AD && |
5b2c4dd2 JP |
1475 | skb->protocol == __cpu_to_be16(ETH_P_SLOW)) |
1476 | return false; | |
1477 | ||
1478 | return true; | |
1479 | } | |
1480 | return false; | |
1481 | } | |
1482 | ||
8a4eb573 | 1483 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1484 | { |
8a4eb573 | 1485 | struct sk_buff *skb = *pskb; |
f1c1775a | 1486 | struct slave *slave; |
0bd80dad | 1487 | struct bonding *bond; |
5b2c4dd2 | 1488 | |
8a4eb573 JP |
1489 | skb = skb_share_check(skb, GFP_ATOMIC); |
1490 | if (unlikely(!skb)) | |
1491 | return RX_HANDLER_CONSUMED; | |
1492 | ||
1493 | *pskb = skb; | |
5b2c4dd2 | 1494 | |
35d48903 JP |
1495 | slave = bond_slave_get_rcu(skb->dev); |
1496 | bond = slave->bond; | |
0bd80dad JP |
1497 | |
1498 | if (bond->params.arp_interval) | |
f1c1775a | 1499 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1500 | |
0bd80dad | 1501 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1502 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1503 | } |
1504 | ||
35d48903 | 1505 | skb->dev = bond->dev; |
5b2c4dd2 | 1506 | |
0bd80dad | 1507 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1508 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1509 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1510 | |
541ac7c9 CG |
1511 | if (unlikely(skb_cow_head(skb, |
1512 | skb->data - skb_mac_header(skb)))) { | |
1513 | kfree_skb(skb); | |
8a4eb573 | 1514 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1515 | } |
35d48903 | 1516 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1517 | } |
1518 | ||
8a4eb573 | 1519 | return RX_HANDLER_ANOTHER; |
5b2c4dd2 JP |
1520 | } |
1521 | ||
1da177e4 | 1522 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1523 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1524 | { |
454d7c9b | 1525 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1526 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1527 | struct slave *new_slave = NULL; |
22bedad3 | 1528 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1529 | struct sockaddr addr; |
1530 | int link_reporting; | |
1531 | int old_features = bond_dev->features; | |
1532 | int res = 0; | |
1533 | ||
552709d5 | 1534 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1535 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1536 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1537 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1538 | } |
1539 | ||
1540 | /* bond must be initialized by bond_open() before enslaving */ | |
1541 | if (!(bond_dev->flags & IFF_UP)) { | |
a4aee5c8 JP |
1542 | pr_warning("%s: master_dev is not up in bond_enslave\n", |
1543 | bond_dev->name); | |
1da177e4 LT |
1544 | } |
1545 | ||
1546 | /* already enslaved */ | |
1547 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1548 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1549 | return -EBUSY; |
1550 | } | |
1551 | ||
1552 | /* vlan challenged mutual exclusion */ | |
1553 | /* no need to lock since we're protected by rtnl_lock */ | |
1554 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1555 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
f35188fa | 1556 | if (bond->vlgrp) { |
a4aee5c8 JP |
1557 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1558 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1559 | return -EPERM; |
1560 | } else { | |
a4aee5c8 JP |
1561 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1562 | bond_dev->name, slave_dev->name, | |
1563 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1564 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1565 | } | |
1566 | } else { | |
5a03cdb7 | 1567 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1568 | if (bond->slave_cnt == 0) { |
1569 | /* First slave, and it is not VLAN challenged, | |
1570 | * so remove the block of adding VLANs over the bond. | |
1571 | */ | |
1572 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1573 | } | |
1574 | } | |
1575 | ||
217df670 JV |
1576 | /* |
1577 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1578 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1579 | * the current ifenslave will set the interface down prior to |
1580 | * enslaving it; the old ifenslave will not. | |
1581 | */ | |
1582 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1583 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1584 | slave_dev->name); |
1585 | res = -EPERM; | |
1586 | goto err_undo_flags; | |
1587 | } | |
1da177e4 | 1588 | |
872254dd MS |
1589 | /* set bonding device ether type by slave - bonding netdevices are |
1590 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1591 | * there is a need to override some of the type dependent attribs/funcs. | |
1592 | * | |
1593 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1594 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1595 | */ | |
1596 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1597 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1598 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1599 | bond_dev->name, |
1600 | bond_dev->type, slave_dev->type); | |
75c78500 | 1601 | |
3ca5b404 JP |
1602 | res = netdev_bonding_change(bond_dev, |
1603 | NETDEV_PRE_TYPE_CHANGE); | |
1604 | res = notifier_to_errno(res); | |
1605 | if (res) { | |
1606 | pr_err("%s: refused to change device type\n", | |
1607 | bond_dev->name); | |
1608 | res = -EBUSY; | |
1609 | goto err_undo_flags; | |
1610 | } | |
75c78500 | 1611 | |
32a806c1 | 1612 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1613 | dev_uc_flush(bond_dev); |
22bedad3 | 1614 | dev_mc_flush(bond_dev); |
32a806c1 | 1615 | |
e36b9d16 MS |
1616 | if (slave_dev->type != ARPHRD_ETHER) |
1617 | bond_setup_by_slave(bond_dev, slave_dev); | |
1618 | else | |
1619 | ether_setup(bond_dev); | |
75c78500 | 1620 | |
93d9b7d7 JP |
1621 | netdev_bonding_change(bond_dev, |
1622 | NETDEV_POST_TYPE_CHANGE); | |
e36b9d16 | 1623 | } |
872254dd | 1624 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1625 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1626 | slave_dev->name, | |
1627 | slave_dev->type, bond_dev->type); | |
1628 | res = -EINVAL; | |
1629 | goto err_undo_flags; | |
872254dd MS |
1630 | } |
1631 | ||
eb7cc59a | 1632 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1633 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1634 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1635 | bond_dev->name); | |
3915c1e8 JV |
1636 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1637 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1638 | 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", |
1639 | bond_dev->name); | |
2ab82852 MS |
1640 | res = -EOPNOTSUPP; |
1641 | goto err_undo_flags; | |
1642 | } | |
1da177e4 LT |
1643 | } |
1644 | ||
c20811a7 JP |
1645 | /* If this is the first slave, then we need to set the master's hardware |
1646 | * address to be the same as the slave's. */ | |
d13a2cb6 | 1647 | if (is_zero_ether_addr(bond->dev->dev_addr)) |
c20811a7 JP |
1648 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, |
1649 | slave_dev->addr_len); | |
1650 | ||
1651 | ||
243cb4e5 | 1652 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1653 | if (!new_slave) { |
1654 | res = -ENOMEM; | |
1655 | goto err_undo_flags; | |
1656 | } | |
1657 | ||
bb1d9123 AG |
1658 | /* |
1659 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1660 | * is set via sysfs or module option if desired. | |
1661 | */ | |
1662 | new_slave->queue_id = 0; | |
1663 | ||
b15ba0fb JP |
1664 | /* Save slave's original mtu and then set it to match the bond */ |
1665 | new_slave->original_mtu = slave_dev->mtu; | |
1666 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1667 | if (res) { | |
1668 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1669 | goto err_free; | |
1670 | } | |
1671 | ||
217df670 JV |
1672 | /* |
1673 | * Save slave's original ("permanent") mac address for modes | |
1674 | * that need it, and for restoring it upon release, and then | |
1675 | * set it to the master's address | |
1676 | */ | |
1677 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1678 | |
dd957c57 | 1679 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1680 | /* |
1681 | * Set slave to master's mac address. The application already | |
1682 | * set the master's mac address to that of the first slave | |
1683 | */ | |
1684 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1685 | addr.sa_family = slave_dev->type; | |
1686 | res = dev_set_mac_address(slave_dev, &addr); | |
1687 | if (res) { | |
5a03cdb7 | 1688 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1689 | goto err_restore_mtu; |
2ab82852 | 1690 | } |
217df670 | 1691 | } |
1da177e4 | 1692 | |
1765a575 | 1693 | res = netdev_set_bond_master(slave_dev, bond_dev); |
c2edacf8 | 1694 | if (res) { |
1765a575 | 1695 | pr_debug("Error %d calling netdev_set_bond_master\n", res); |
569f0c4d | 1696 | goto err_restore_mac; |
c2edacf8 | 1697 | } |
5b2c4dd2 | 1698 | |
217df670 JV |
1699 | /* open the slave since the application closed it */ |
1700 | res = dev_open(slave_dev); | |
1701 | if (res) { | |
3d632c3f | 1702 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1703 | goto err_unset_master; |
1da177e4 LT |
1704 | } |
1705 | ||
35d48903 | 1706 | new_slave->bond = bond; |
1da177e4 | 1707 | new_slave->dev = slave_dev; |
0b680e75 | 1708 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1709 | |
58402054 | 1710 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1711 | /* bond_alb_init_slave() must be called before all other stages since |
1712 | * it might fail and we do not want to have to undo everything | |
1713 | */ | |
1714 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1715 | if (res) |
569f0c4d | 1716 | goto err_close; |
1da177e4 LT |
1717 | } |
1718 | ||
1719 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1720 | * master's promisc (and mc) settings only if it becomes the | |
1721 | * curr_active_slave, and that is taken care of later when calling | |
1722 | * bond_change_active() | |
1723 | */ | |
1724 | if (!USES_PRIMARY(bond->params.mode)) { | |
1725 | /* set promiscuity level to new slave */ | |
1726 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1727 | res = dev_set_promiscuity(slave_dev, 1); |
1728 | if (res) | |
1729 | goto err_close; | |
1da177e4 LT |
1730 | } |
1731 | ||
1732 | /* set allmulti level to new slave */ | |
1733 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1734 | res = dev_set_allmulti(slave_dev, 1); |
1735 | if (res) | |
1736 | goto err_close; | |
1da177e4 LT |
1737 | } |
1738 | ||
b9e40857 | 1739 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1740 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1741 | netdev_for_each_mc_addr(ha, bond_dev) |
1742 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1743 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1744 | } |
1745 | ||
1746 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1747 | /* add lacpdu mc addr to mc list */ | |
1748 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1749 | ||
22bedad3 | 1750 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1751 | } |
1752 | ||
1753 | bond_add_vlans_on_slave(bond, slave_dev); | |
1754 | ||
1755 | write_lock_bh(&bond->lock); | |
1756 | ||
1757 | bond_attach_slave(bond, new_slave); | |
1758 | ||
1759 | new_slave->delay = 0; | |
1760 | new_slave->link_failure_count = 0; | |
1761 | ||
8531c5ff AK |
1762 | bond_compute_features(bond); |
1763 | ||
3915c1e8 JV |
1764 | write_unlock_bh(&bond->lock); |
1765 | ||
1766 | read_lock(&bond->lock); | |
1767 | ||
f5b2b966 JV |
1768 | new_slave->last_arp_rx = jiffies; |
1769 | ||
1da177e4 LT |
1770 | if (bond->params.miimon && !bond->params.use_carrier) { |
1771 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1772 | ||
1773 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1774 | /* | |
1775 | * miimon is set but a bonded network driver | |
1776 | * does not support ETHTOOL/MII and | |
1777 | * arp_interval is not set. Note: if | |
1778 | * use_carrier is enabled, we will never go | |
1779 | * here (because netif_carrier is always | |
1780 | * supported); thus, we don't need to change | |
1781 | * the messages for netif_carrier. | |
1782 | */ | |
a4aee5c8 | 1783 | 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 | 1784 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1785 | } else if (link_reporting == -1) { |
1786 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1787 | 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", |
1788 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1789 | } |
1790 | } | |
1791 | ||
1792 | /* check for initial state */ | |
1793 | if (!bond->params.miimon || | |
1794 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1795 | if (bond->params.updelay) { | |
a4aee5c8 | 1796 | pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n"); |
1da177e4 LT |
1797 | new_slave->link = BOND_LINK_BACK; |
1798 | new_slave->delay = bond->params.updelay; | |
1799 | } else { | |
a4aee5c8 | 1800 | pr_debug("Initial state of slave_dev is BOND_LINK_UP\n"); |
1da177e4 LT |
1801 | new_slave->link = BOND_LINK_UP; |
1802 | } | |
1803 | new_slave->jiffies = jiffies; | |
1804 | } else { | |
a4aee5c8 | 1805 | pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n"); |
1da177e4 LT |
1806 | new_slave->link = BOND_LINK_DOWN; |
1807 | } | |
1808 | ||
1809 | if (bond_update_speed_duplex(new_slave) && | |
1810 | (new_slave->link != BOND_LINK_DOWN)) { | |
a4aee5c8 JP |
1811 | pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n", |
1812 | bond_dev->name, new_slave->dev->name); | |
1da177e4 LT |
1813 | |
1814 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a4aee5c8 JP |
1815 | pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n", |
1816 | bond_dev->name); | |
1da177e4 LT |
1817 | } |
1818 | } | |
1819 | ||
1820 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1821 | /* if there is a primary slave, remember it */ | |
a549952a | 1822 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1823 | bond->primary_slave = new_slave; |
a549952a JP |
1824 | bond->force_primary = true; |
1825 | } | |
1da177e4 LT |
1826 | } |
1827 | ||
3915c1e8 JV |
1828 | write_lock_bh(&bond->curr_slave_lock); |
1829 | ||
1da177e4 LT |
1830 | switch (bond->params.mode) { |
1831 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1832 | bond_set_slave_inactive_flags(new_slave); |
1833 | bond_select_active_slave(bond); | |
1da177e4 LT |
1834 | break; |
1835 | case BOND_MODE_8023AD: | |
1836 | /* in 802.3ad mode, the internal mechanism | |
1837 | * will activate the slaves in the selected | |
1838 | * aggregator | |
1839 | */ | |
1840 | bond_set_slave_inactive_flags(new_slave); | |
1841 | /* if this is the first slave */ | |
1842 | if (bond->slave_cnt == 1) { | |
1843 | SLAVE_AD_INFO(new_slave).id = 1; | |
1844 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1845 | * can be called only after the mac address of the bond is set | |
1846 | */ | |
1847 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1848 | bond->params.lacp_fast); | |
1849 | } else { | |
1850 | SLAVE_AD_INFO(new_slave).id = | |
1851 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1852 | } | |
1853 | ||
1854 | bond_3ad_bind_slave(new_slave); | |
1855 | break; | |
1856 | case BOND_MODE_TLB: | |
1857 | case BOND_MODE_ALB: | |
e30bc066 | 1858 | bond_set_active_slave(new_slave); |
059fe7a5 | 1859 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1860 | bond_select_active_slave(bond); |
1da177e4 LT |
1861 | break; |
1862 | default: | |
5a03cdb7 | 1863 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1864 | |
1865 | /* always active in trunk mode */ | |
e30bc066 | 1866 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1867 | |
1868 | /* In trunking mode there is little meaning to curr_active_slave | |
1869 | * anyway (it holds no special properties of the bond device), | |
1870 | * so we can change it without calling change_active_interface() | |
1871 | */ | |
3d632c3f | 1872 | if (!bond->curr_active_slave) |
1da177e4 | 1873 | bond->curr_active_slave = new_slave; |
3d632c3f | 1874 | |
1da177e4 LT |
1875 | break; |
1876 | } /* switch(bond_mode) */ | |
1877 | ||
3915c1e8 JV |
1878 | write_unlock_bh(&bond->curr_slave_lock); |
1879 | ||
ff59c456 JV |
1880 | bond_set_carrier(bond); |
1881 | ||
f6dc31a8 | 1882 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 AW |
1883 | slave_dev->npinfo = bond_netpoll_info(bond); |
1884 | if (slave_dev->npinfo) { | |
1885 | if (slave_enable_netpoll(new_slave)) { | |
1886 | read_unlock(&bond->lock); | |
1887 | pr_info("Error, %s: master_dev is using netpoll, " | |
1888 | "but new slave device does not support netpoll.\n", | |
1889 | bond_dev->name); | |
1890 | res = -EBUSY; | |
1891 | goto err_close; | |
1892 | } | |
f6dc31a8 WC |
1893 | } |
1894 | #endif | |
8a8efa22 | 1895 | |
3915c1e8 | 1896 | read_unlock(&bond->lock); |
1da177e4 | 1897 | |
b76cdba9 MW |
1898 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1899 | if (res) | |
569f0c4d | 1900 | goto err_close; |
b76cdba9 | 1901 | |
35d48903 JP |
1902 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1903 | new_slave); | |
1904 | if (res) { | |
1905 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1906 | goto err_dest_symlinks; | |
1907 | } | |
1908 | ||
a4aee5c8 JP |
1909 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1910 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1911 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1912 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1913 | |
1914 | /* enslave is successful */ | |
1915 | return 0; | |
1916 | ||
1917 | /* Undo stages on error */ | |
35d48903 JP |
1918 | err_dest_symlinks: |
1919 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1920 | ||
1da177e4 LT |
1921 | err_close: |
1922 | dev_close(slave_dev); | |
1923 | ||
569f0c4d | 1924 | err_unset_master: |
1765a575 | 1925 | netdev_set_bond_master(slave_dev, NULL); |
569f0c4d | 1926 | |
1da177e4 | 1927 | err_restore_mac: |
dd957c57 | 1928 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1929 | /* XXX TODO - fom follow mode needs to change master's |
1930 | * MAC if this slave's MAC is in use by the bond, or at | |
1931 | * least print a warning. | |
1932 | */ | |
2ab82852 MS |
1933 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1934 | addr.sa_family = slave_dev->type; | |
1935 | dev_set_mac_address(slave_dev, &addr); | |
1936 | } | |
1da177e4 | 1937 | |
b15ba0fb JP |
1938 | err_restore_mtu: |
1939 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1940 | ||
1da177e4 LT |
1941 | err_free: |
1942 | kfree(new_slave); | |
1943 | ||
1944 | err_undo_flags: | |
1945 | bond_dev->features = old_features; | |
3d632c3f | 1946 | |
1da177e4 LT |
1947 | return res; |
1948 | } | |
1949 | ||
1950 | /* | |
1951 | * Try to release the slave device <slave> from the bond device <master> | |
1952 | * It is legal to access curr_active_slave without a lock because all the function | |
1953 | * is write-locked. | |
1954 | * | |
1955 | * The rules for slave state should be: | |
1956 | * for Active/Backup: | |
1957 | * Active stays on all backups go down | |
1958 | * for Bonded connections: | |
1959 | * The first up interface should be left on and all others downed. | |
1960 | */ | |
a77b5325 | 1961 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1962 | { |
454d7c9b | 1963 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1964 | struct slave *slave, *oldcurrent; |
1965 | struct sockaddr addr; | |
1da177e4 LT |
1966 | |
1967 | /* slave is not a slave or master is not master of this slave */ | |
1968 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1969 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1970 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1971 | bond_dev->name, slave_dev->name); |
1972 | return -EINVAL; | |
1973 | } | |
1974 | ||
e843fa50 | 1975 | block_netpoll_tx(); |
f6dc31a8 | 1976 | netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE); |
1da177e4 LT |
1977 | write_lock_bh(&bond->lock); |
1978 | ||
1979 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1980 | if (!slave) { | |
1981 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1982 | pr_info("%s: %s not enslaved\n", |
1983 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1984 | write_unlock_bh(&bond->lock); |
e843fa50 | 1985 | unblock_netpoll_tx(); |
1da177e4 LT |
1986 | return -EINVAL; |
1987 | } | |
1988 | ||
35d48903 JP |
1989 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1990 | * for this slave anymore. | |
1991 | */ | |
1992 | netdev_rx_handler_unregister(slave_dev); | |
1993 | write_unlock_bh(&bond->lock); | |
1994 | synchronize_net(); | |
1995 | write_lock_bh(&bond->lock); | |
1996 | ||
3915c1e8 | 1997 | if (!bond->params.fail_over_mac) { |
8e95a202 JP |
1998 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) && |
1999 | bond->slave_cnt > 1) | |
a4aee5c8 JP |
2000 | 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", |
2001 | bond_dev->name, slave_dev->name, | |
2002 | slave->perm_hwaddr, | |
2003 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
2004 | } |
2005 | ||
2006 | /* Inform AD package of unbinding of slave. */ | |
2007 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2008 | /* must be called before the slave is | |
2009 | * detached from the list | |
2010 | */ | |
2011 | bond_3ad_unbind_slave(slave); | |
2012 | } | |
2013 | ||
a4aee5c8 JP |
2014 | pr_info("%s: releasing %s interface %s\n", |
2015 | bond_dev->name, | |
e30bc066 | 2016 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 2017 | slave_dev->name); |
1da177e4 LT |
2018 | |
2019 | oldcurrent = bond->curr_active_slave; | |
2020 | ||
2021 | bond->current_arp_slave = NULL; | |
2022 | ||
2023 | /* release the slave from its bond */ | |
2024 | bond_detach_slave(bond, slave); | |
2025 | ||
8531c5ff AK |
2026 | bond_compute_features(bond); |
2027 | ||
3d632c3f | 2028 | if (bond->primary_slave == slave) |
1da177e4 | 2029 | bond->primary_slave = NULL; |
1da177e4 | 2030 | |
3d632c3f | 2031 | if (oldcurrent == slave) |
1da177e4 | 2032 | bond_change_active_slave(bond, NULL); |
1da177e4 | 2033 | |
58402054 | 2034 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2035 | /* Must be called only after the slave has been |
2036 | * detached from the list and the curr_active_slave | |
2037 | * has been cleared (if our_slave == old_current), | |
2038 | * but before a new active slave is selected. | |
2039 | */ | |
2543331d | 2040 | write_unlock_bh(&bond->lock); |
1da177e4 | 2041 | bond_alb_deinit_slave(bond, slave); |
2543331d | 2042 | write_lock_bh(&bond->lock); |
1da177e4 LT |
2043 | } |
2044 | ||
059fe7a5 JV |
2045 | if (oldcurrent == slave) { |
2046 | /* | |
2047 | * Note that we hold RTNL over this sequence, so there | |
2048 | * is no concern that another slave add/remove event | |
2049 | * will interfere. | |
2050 | */ | |
2051 | write_unlock_bh(&bond->lock); | |
2052 | read_lock(&bond->lock); | |
2053 | write_lock_bh(&bond->curr_slave_lock); | |
2054 | ||
1da177e4 LT |
2055 | bond_select_active_slave(bond); |
2056 | ||
059fe7a5 JV |
2057 | write_unlock_bh(&bond->curr_slave_lock); |
2058 | read_unlock(&bond->lock); | |
2059 | write_lock_bh(&bond->lock); | |
2060 | } | |
2061 | ||
1da177e4 | 2062 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
2063 | bond_set_carrier(bond); |
2064 | ||
1da177e4 LT |
2065 | /* if the last slave was removed, zero the mac address |
2066 | * of the master so it will be set by the application | |
2067 | * to the mac address of the first slave | |
2068 | */ | |
2069 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2070 | ||
f35188fa | 2071 | if (!bond->vlgrp) { |
1da177e4 LT |
2072 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
2073 | } else { | |
a4aee5c8 JP |
2074 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2075 | bond_dev->name, bond_dev->name); | |
2076 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2077 | bond_dev->name); | |
1da177e4 LT |
2078 | } |
2079 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2080 | !bond_has_challenged_slaves(bond)) { | |
a4aee5c8 JP |
2081 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", |
2082 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
2083 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
2084 | } | |
2085 | ||
2086 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2087 | unblock_netpoll_tx(); |
1da177e4 | 2088 | |
b76cdba9 MW |
2089 | /* must do this from outside any spinlocks */ |
2090 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2091 | ||
1da177e4 LT |
2092 | bond_del_vlans_from_slave(bond, slave_dev); |
2093 | ||
2094 | /* If the mode USES_PRIMARY, then we should only remove its | |
2095 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2096 | * already taken care of above when we detached the slave | |
2097 | */ | |
2098 | if (!USES_PRIMARY(bond->params.mode)) { | |
2099 | /* unset promiscuity level from slave */ | |
3d632c3f | 2100 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2101 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2102 | |
2103 | /* unset allmulti level from slave */ | |
3d632c3f | 2104 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2105 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2106 | |
2107 | /* flush master's mc_list from slave */ | |
b9e40857 | 2108 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2109 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2110 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2111 | } |
2112 | ||
1765a575 | 2113 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2114 | |
8a8efa22 | 2115 | slave_disable_netpoll(slave); |
f6dc31a8 | 2116 | |
1da177e4 LT |
2117 | /* close slave before restoring its mac address */ |
2118 | dev_close(slave_dev); | |
2119 | ||
3915c1e8 | 2120 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
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 | |
b15ba0fb JP |
2127 | dev_set_mtu(slave_dev, slave->original_mtu); |
2128 | ||
2d7011ca | 2129 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2130 | |
2131 | kfree(slave); | |
2132 | ||
2133 | return 0; /* deletion OK */ | |
2134 | } | |
2135 | ||
d90a162a | 2136 | /* |
dadaa10b | 2137 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2138 | * Must be under rtnl_lock when this function is called. |
2139 | */ | |
26d8ee75 | 2140 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2141 | struct net_device *slave_dev) | |
d90a162a | 2142 | { |
454d7c9b | 2143 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2144 | int ret; |
2145 | ||
2146 | ret = bond_release(bond_dev, slave_dev); | |
2147 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2148 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2149 | pr_info("%s: destroying bond %s.\n", |
2150 | bond_dev->name, bond_dev->name); | |
9e71626c | 2151 | unregister_netdevice(bond_dev); |
d90a162a MS |
2152 | } |
2153 | return ret; | |
2154 | } | |
2155 | ||
1da177e4 LT |
2156 | /* |
2157 | * This function releases all slaves. | |
2158 | */ | |
2159 | static int bond_release_all(struct net_device *bond_dev) | |
2160 | { | |
454d7c9b | 2161 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2162 | struct slave *slave; |
2163 | struct net_device *slave_dev; | |
2164 | struct sockaddr addr; | |
2165 | ||
2166 | write_lock_bh(&bond->lock); | |
2167 | ||
ff59c456 JV |
2168 | netif_carrier_off(bond_dev); |
2169 | ||
3d632c3f | 2170 | if (bond->slave_cnt == 0) |
1da177e4 | 2171 | goto out; |
1da177e4 LT |
2172 | |
2173 | bond->current_arp_slave = NULL; | |
2174 | bond->primary_slave = NULL; | |
2175 | bond_change_active_slave(bond, NULL); | |
2176 | ||
2177 | while ((slave = bond->first_slave) != NULL) { | |
2178 | /* Inform AD package of unbinding of slave | |
2179 | * before slave is detached from the list. | |
2180 | */ | |
3d632c3f | 2181 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2182 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2183 | |
2184 | slave_dev = slave->dev; | |
2185 | bond_detach_slave(bond, slave); | |
2186 | ||
2543331d JV |
2187 | /* now that the slave is detached, unlock and perform |
2188 | * all the undo steps that should not be called from | |
2189 | * within a lock. | |
2190 | */ | |
2191 | write_unlock_bh(&bond->lock); | |
2192 | ||
35d48903 JP |
2193 | /* unregister rx_handler early so bond_handle_frame wouldn't |
2194 | * be called for this slave anymore. | |
2195 | */ | |
2196 | netdev_rx_handler_unregister(slave_dev); | |
2197 | synchronize_net(); | |
2198 | ||
58402054 | 2199 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2200 | /* must be called only after the slave |
2201 | * has been detached from the list | |
2202 | */ | |
2203 | bond_alb_deinit_slave(bond, slave); | |
2204 | } | |
2205 | ||
8531c5ff AK |
2206 | bond_compute_features(bond); |
2207 | ||
b76cdba9 | 2208 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2209 | bond_del_vlans_from_slave(bond, slave_dev); |
2210 | ||
2211 | /* If the mode USES_PRIMARY, then we should only remove its | |
2212 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2213 | * already taken care of above when we detached the slave | |
2214 | */ | |
2215 | if (!USES_PRIMARY(bond->params.mode)) { | |
2216 | /* unset promiscuity level from slave */ | |
3d632c3f | 2217 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2218 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2219 | |
2220 | /* unset allmulti level from slave */ | |
3d632c3f | 2221 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2222 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2223 | |
2224 | /* flush master's mc_list from slave */ | |
b9e40857 | 2225 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2226 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2227 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2228 | } |
2229 | ||
1765a575 | 2230 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2231 | |
8a8efa22 AW |
2232 | slave_disable_netpoll(slave); |
2233 | ||
1da177e4 LT |
2234 | /* close slave before restoring its mac address */ |
2235 | dev_close(slave_dev); | |
2236 | ||
dd957c57 | 2237 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2238 | /* restore original ("permanent") mac address*/ |
2239 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2240 | addr.sa_family = slave_dev->type; | |
2241 | dev_set_mac_address(slave_dev, &addr); | |
2242 | } | |
1da177e4 | 2243 | |
1da177e4 LT |
2244 | kfree(slave); |
2245 | ||
2246 | /* re-acquire the lock before getting the next slave */ | |
2247 | write_lock_bh(&bond->lock); | |
2248 | } | |
2249 | ||
2250 | /* zero the mac address of the master so it will be | |
2251 | * set by the application to the mac address of the | |
2252 | * first slave | |
2253 | */ | |
2254 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2255 | ||
f35188fa | 2256 | if (!bond->vlgrp) { |
1da177e4 | 2257 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
f35188fa | 2258 | } else { |
a4aee5c8 JP |
2259 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2260 | bond_dev->name, bond_dev->name); | |
2261 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2262 | bond_dev->name); | |
1da177e4 LT |
2263 | } |
2264 | ||
a4aee5c8 | 2265 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2266 | |
2267 | out: | |
2268 | write_unlock_bh(&bond->lock); | |
1da177e4 LT |
2269 | return 0; |
2270 | } | |
2271 | ||
2272 | /* | |
2273 | * This function changes the active slave to slave <slave_dev>. | |
2274 | * It returns -EINVAL in the following cases. | |
2275 | * - <slave_dev> is not found in the list. | |
2276 | * - There is not active slave now. | |
2277 | * - <slave_dev> is already active. | |
2278 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2279 | * - <slave_dev> is not running. | |
3d632c3f SH |
2280 | * In these cases, this function does nothing. |
2281 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2282 | */ |
2283 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2284 | { | |
454d7c9b | 2285 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2286 | struct slave *old_active = NULL; |
2287 | struct slave *new_active = NULL; | |
2288 | int res = 0; | |
2289 | ||
3d632c3f | 2290 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2291 | return -EINVAL; |
1da177e4 LT |
2292 | |
2293 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2294 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2295 | return -EINVAL; |
1da177e4 | 2296 | |
059fe7a5 | 2297 | read_lock(&bond->lock); |
1da177e4 | 2298 | |
059fe7a5 | 2299 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2300 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2301 | read_unlock(&bond->curr_slave_lock); |
2302 | ||
1da177e4 LT |
2303 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2304 | ||
2305 | /* | |
2306 | * Changing to the current active: do nothing; return success. | |
2307 | */ | |
2308 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2309 | read_unlock(&bond->lock); |
1da177e4 LT |
2310 | return 0; |
2311 | } | |
2312 | ||
2313 | if ((new_active) && | |
2314 | (old_active) && | |
2315 | (new_active->link == BOND_LINK_UP) && | |
2316 | IS_UP(new_active->dev)) { | |
e843fa50 | 2317 | block_netpoll_tx(); |
059fe7a5 | 2318 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2319 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2320 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2321 | unblock_netpoll_tx(); |
3d632c3f | 2322 | } else |
1da177e4 | 2323 | res = -EINVAL; |
1da177e4 | 2324 | |
059fe7a5 | 2325 | read_unlock(&bond->lock); |
1da177e4 LT |
2326 | |
2327 | return res; | |
2328 | } | |
2329 | ||
1da177e4 LT |
2330 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2331 | { | |
454d7c9b | 2332 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2333 | |
2334 | info->bond_mode = bond->params.mode; | |
2335 | info->miimon = bond->params.miimon; | |
2336 | ||
6603a6f2 | 2337 | read_lock(&bond->lock); |
1da177e4 | 2338 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2339 | read_unlock(&bond->lock); |
1da177e4 LT |
2340 | |
2341 | return 0; | |
2342 | } | |
2343 | ||
2344 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2345 | { | |
454d7c9b | 2346 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2347 | struct slave *slave; |
689c96cc | 2348 | int i, res = -ENODEV; |
1da177e4 | 2349 | |
6603a6f2 | 2350 | read_lock(&bond->lock); |
1da177e4 LT |
2351 | |
2352 | bond_for_each_slave(bond, slave, i) { | |
2353 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2354 | res = 0; |
2355 | strcpy(info->slave_name, slave->dev->name); | |
2356 | info->link = slave->link; | |
e30bc066 | 2357 | info->state = bond_slave_state(slave); |
689c96cc | 2358 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2359 | break; |
2360 | } | |
2361 | } | |
2362 | ||
6603a6f2 | 2363 | read_unlock(&bond->lock); |
1da177e4 | 2364 | |
689c96cc | 2365 | return res; |
1da177e4 LT |
2366 | } |
2367 | ||
2368 | /*-------------------------------- Monitoring -------------------------------*/ | |
2369 | ||
1da177e4 | 2370 | |
f0c76d61 JV |
2371 | static int bond_miimon_inspect(struct bonding *bond) |
2372 | { | |
2373 | struct slave *slave; | |
2374 | int i, link_state, commit = 0; | |
41f89100 JP |
2375 | bool ignore_updelay; |
2376 | ||
2377 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2378 | |
2379 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2380 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2381 | |
f0c76d61 | 2382 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2383 | |
2384 | switch (slave->link) { | |
f0c76d61 JV |
2385 | case BOND_LINK_UP: |
2386 | if (link_state) | |
2387 | continue; | |
1da177e4 | 2388 | |
f0c76d61 JV |
2389 | slave->link = BOND_LINK_FAIL; |
2390 | slave->delay = bond->params.downdelay; | |
2391 | if (slave->delay) { | |
a4aee5c8 JP |
2392 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2393 | bond->dev->name, | |
2394 | (bond->params.mode == | |
2395 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2396 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2397 | "active " : "backup ") : "", |
2398 | slave->dev->name, | |
2399 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2400 | } |
f0c76d61 JV |
2401 | /*FALLTHRU*/ |
2402 | case BOND_LINK_FAIL: | |
2403 | if (link_state) { | |
2404 | /* | |
2405 | * recovered before downdelay expired | |
2406 | */ | |
2407 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2408 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2409 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2410 | bond->dev->name, | |
2411 | (bond->params.downdelay - slave->delay) * | |
2412 | bond->params.miimon, | |
2413 | slave->dev->name); | |
f0c76d61 | 2414 | continue; |
1da177e4 | 2415 | } |
f0c76d61 JV |
2416 | |
2417 | if (slave->delay <= 0) { | |
2418 | slave->new_link = BOND_LINK_DOWN; | |
2419 | commit++; | |
2420 | continue; | |
1da177e4 | 2421 | } |
1da177e4 | 2422 | |
f0c76d61 JV |
2423 | slave->delay--; |
2424 | break; | |
2425 | ||
2426 | case BOND_LINK_DOWN: | |
2427 | if (!link_state) | |
2428 | continue; | |
2429 | ||
2430 | slave->link = BOND_LINK_BACK; | |
2431 | slave->delay = bond->params.updelay; | |
2432 | ||
2433 | if (slave->delay) { | |
a4aee5c8 JP |
2434 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2435 | bond->dev->name, slave->dev->name, | |
2436 | ignore_updelay ? 0 : | |
2437 | bond->params.updelay * | |
2438 | bond->params.miimon); | |
f0c76d61 JV |
2439 | } |
2440 | /*FALLTHRU*/ | |
2441 | case BOND_LINK_BACK: | |
2442 | if (!link_state) { | |
2443 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2444 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2445 | bond->dev->name, | |
2446 | (bond->params.updelay - slave->delay) * | |
2447 | bond->params.miimon, | |
2448 | slave->dev->name); | |
f0c76d61 JV |
2449 | |
2450 | continue; | |
2451 | } | |
2452 | ||
41f89100 JP |
2453 | if (ignore_updelay) |
2454 | slave->delay = 0; | |
2455 | ||
f0c76d61 JV |
2456 | if (slave->delay <= 0) { |
2457 | slave->new_link = BOND_LINK_UP; | |
2458 | commit++; | |
41f89100 | 2459 | ignore_updelay = false; |
f0c76d61 | 2460 | continue; |
1da177e4 | 2461 | } |
f0c76d61 JV |
2462 | |
2463 | slave->delay--; | |
1da177e4 | 2464 | break; |
f0c76d61 JV |
2465 | } |
2466 | } | |
1da177e4 | 2467 | |
f0c76d61 JV |
2468 | return commit; |
2469 | } | |
1da177e4 | 2470 | |
f0c76d61 JV |
2471 | static void bond_miimon_commit(struct bonding *bond) |
2472 | { | |
2473 | struct slave *slave; | |
2474 | int i; | |
2475 | ||
2476 | bond_for_each_slave(bond, slave, i) { | |
2477 | switch (slave->new_link) { | |
2478 | case BOND_LINK_NOCHANGE: | |
2479 | continue; | |
1da177e4 | 2480 | |
f0c76d61 JV |
2481 | case BOND_LINK_UP: |
2482 | slave->link = BOND_LINK_UP; | |
2483 | slave->jiffies = jiffies; | |
2484 | ||
2485 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2486 | /* prevent it from being the active one */ | |
e30bc066 | 2487 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2488 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2489 | /* make it immediately active */ | |
e30bc066 | 2490 | bond_set_active_slave(slave); |
f0c76d61 JV |
2491 | } else if (slave != bond->primary_slave) { |
2492 | /* prevent it from being the active one */ | |
e30bc066 | 2493 | bond_set_backup_slave(slave); |
1da177e4 | 2494 | } |
1da177e4 | 2495 | |
546add79 KPO |
2496 | bond_update_speed_duplex(slave); |
2497 | ||
5d30530e | 2498 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 KPO |
2499 | bond->dev->name, slave->dev->name, |
2500 | slave->speed, slave->duplex ? "full" : "half"); | |
1da177e4 | 2501 | |
f0c76d61 JV |
2502 | /* notify ad that the link status has changed */ |
2503 | if (bond->params.mode == BOND_MODE_8023AD) | |
2504 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2505 | |
58402054 | 2506 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2507 | bond_alb_handle_link_change(bond, slave, |
2508 | BOND_LINK_UP); | |
1da177e4 | 2509 | |
f0c76d61 JV |
2510 | if (!bond->curr_active_slave || |
2511 | (slave == bond->primary_slave)) | |
2512 | goto do_failover; | |
1da177e4 | 2513 | |
f0c76d61 | 2514 | continue; |
059fe7a5 | 2515 | |
f0c76d61 | 2516 | case BOND_LINK_DOWN: |
fba4acda JV |
2517 | if (slave->link_failure_count < UINT_MAX) |
2518 | slave->link_failure_count++; | |
2519 | ||
f0c76d61 | 2520 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2521 | |
f0c76d61 JV |
2522 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2523 | bond->params.mode == BOND_MODE_8023AD) | |
2524 | bond_set_slave_inactive_flags(slave); | |
2525 | ||
a4aee5c8 JP |
2526 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2527 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2528 | |
2529 | if (bond->params.mode == BOND_MODE_8023AD) | |
2530 | bond_3ad_handle_link_change(slave, | |
2531 | BOND_LINK_DOWN); | |
2532 | ||
ae63e808 | 2533 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2534 | bond_alb_handle_link_change(bond, slave, |
2535 | BOND_LINK_DOWN); | |
2536 | ||
2537 | if (slave == bond->curr_active_slave) | |
2538 | goto do_failover; | |
2539 | ||
2540 | continue; | |
2541 | ||
2542 | default: | |
a4aee5c8 | 2543 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2544 | bond->dev->name, slave->new_link, |
2545 | slave->dev->name); | |
2546 | slave->new_link = BOND_LINK_NOCHANGE; | |
2547 | ||
2548 | continue; | |
2549 | } | |
2550 | ||
2551 | do_failover: | |
2552 | ASSERT_RTNL(); | |
e843fa50 | 2553 | block_netpoll_tx(); |
f0c76d61 JV |
2554 | write_lock_bh(&bond->curr_slave_lock); |
2555 | bond_select_active_slave(bond); | |
2556 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2557 | unblock_netpoll_tx(); |
f0c76d61 JV |
2558 | } |
2559 | ||
2560 | bond_set_carrier(bond); | |
1da177e4 LT |
2561 | } |
2562 | ||
0b0eef66 JV |
2563 | /* |
2564 | * bond_mii_monitor | |
2565 | * | |
2566 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2567 | * inspection, then (if inspection indicates something needs to be done) |
2568 | * an acquisition of appropriate locks followed by a commit phase to | |
2569 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2570 | */ |
2571 | void bond_mii_monitor(struct work_struct *work) | |
2572 | { | |
2573 | struct bonding *bond = container_of(work, struct bonding, | |
2574 | mii_work.work); | |
0b0eef66 JV |
2575 | |
2576 | read_lock(&bond->lock); | |
f0c76d61 JV |
2577 | if (bond->kill_timers) |
2578 | goto out; | |
2579 | ||
2580 | if (bond->slave_cnt == 0) | |
2581 | goto re_arm; | |
b59f9f74 JV |
2582 | |
2583 | if (bond->send_grat_arp) { | |
2584 | read_lock(&bond->curr_slave_lock); | |
2585 | bond_send_gratuitous_arp(bond); | |
2586 | read_unlock(&bond->curr_slave_lock); | |
2587 | } | |
2588 | ||
305d552a BH |
2589 | if (bond->send_unsol_na) { |
2590 | read_lock(&bond->curr_slave_lock); | |
2591 | bond_send_unsolicited_na(bond); | |
2592 | read_unlock(&bond->curr_slave_lock); | |
2593 | } | |
2594 | ||
f0c76d61 | 2595 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2596 | read_unlock(&bond->lock); |
2597 | rtnl_lock(); | |
2598 | read_lock(&bond->lock); | |
f0c76d61 JV |
2599 | |
2600 | bond_miimon_commit(bond); | |
2601 | ||
5655662d JV |
2602 | read_unlock(&bond->lock); |
2603 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2604 | read_lock(&bond->lock); | |
0b0eef66 JV |
2605 | } |
2606 | ||
f0c76d61 JV |
2607 | re_arm: |
2608 | if (bond->params.miimon) | |
2609 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2610 | msecs_to_jiffies(bond->params.miimon)); | |
2611 | out: | |
0b0eef66 | 2612 | read_unlock(&bond->lock); |
0b0eef66 | 2613 | } |
c3ade5ca | 2614 | |
d3bb52b0 | 2615 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2616 | { |
2617 | struct in_device *idev; | |
2618 | struct in_ifaddr *ifa; | |
a144ea4b | 2619 | __be32 addr = 0; |
c3ade5ca JV |
2620 | |
2621 | if (!dev) | |
2622 | return 0; | |
2623 | ||
2624 | rcu_read_lock(); | |
e5ed6399 | 2625 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2626 | if (!idev) |
2627 | goto out; | |
2628 | ||
2629 | ifa = idev->ifa_list; | |
2630 | if (!ifa) | |
2631 | goto out; | |
2632 | ||
2633 | addr = ifa->ifa_local; | |
2634 | out: | |
2635 | rcu_read_unlock(); | |
2636 | return addr; | |
2637 | } | |
2638 | ||
d3bb52b0 | 2639 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2640 | { |
0883beca | 2641 | struct vlan_entry *vlan; |
f5b2b966 JV |
2642 | |
2643 | if (ip == bond->master_ip) | |
2644 | return 1; | |
2645 | ||
0883beca | 2646 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2647 | if (ip == vlan->vlan_ip) |
2648 | return 1; | |
2649 | } | |
2650 | ||
2651 | return 0; | |
2652 | } | |
2653 | ||
c3ade5ca JV |
2654 | /* |
2655 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2656 | * switches in VLAN mode (especially if ports are configured as | |
2657 | * "native" to a VLAN) might not pass non-tagged frames. | |
2658 | */ | |
d3bb52b0 | 2659 | 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 |
2660 | { |
2661 | struct sk_buff *skb; | |
2662 | ||
5a03cdb7 | 2663 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2664 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2665 | |
c3ade5ca JV |
2666 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2667 | NULL, slave_dev->dev_addr, NULL); | |
2668 | ||
2669 | if (!skb) { | |
a4aee5c8 | 2670 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2671 | return; |
2672 | } | |
2673 | if (vlan_id) { | |
2674 | skb = vlan_put_tag(skb, vlan_id); | |
2675 | if (!skb) { | |
a4aee5c8 | 2676 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2677 | return; |
2678 | } | |
2679 | } | |
2680 | arp_xmit(skb); | |
2681 | } | |
2682 | ||
2683 | ||
1da177e4 LT |
2684 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2685 | { | |
b23dd4fe | 2686 | int i, vlan_id; |
d3bb52b0 | 2687 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2688 | struct vlan_entry *vlan; |
c3ade5ca | 2689 | struct net_device *vlan_dev; |
c3ade5ca | 2690 | struct rtable *rt; |
1da177e4 | 2691 | |
6b780567 MW |
2692 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2693 | if (!targets[i]) | |
5a31bec0 | 2694 | break; |
5a03cdb7 | 2695 | pr_debug("basa: target %x\n", targets[i]); |
f35188fa | 2696 | if (!bond->vlgrp) { |
5a03cdb7 | 2697 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2698 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2699 | bond->master_ip, 0); | |
2700 | continue; | |
2701 | } | |
2702 | ||
2703 | /* | |
2704 | * If VLANs are configured, we do a route lookup to | |
2705 | * determine which VLAN interface would be used, so we | |
2706 | * can tag the ARP with the proper VLAN tag. | |
2707 | */ | |
78fbfd8a DM |
2708 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2709 | RTO_ONLINK, 0); | |
b23dd4fe | 2710 | if (IS_ERR(rt)) { |
c3ade5ca | 2711 | if (net_ratelimit()) { |
a4aee5c8 | 2712 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2713 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2714 | } |
2715 | continue; | |
2716 | } | |
2717 | ||
2718 | /* | |
2719 | * This target is not on a VLAN | |
2720 | */ | |
d8d1f30b | 2721 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2722 | ip_rt_put(rt); |
5a03cdb7 | 2723 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2724 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2725 | bond->master_ip, 0); | |
2726 | continue; | |
2727 | } | |
2728 | ||
2729 | vlan_id = 0; | |
0883beca | 2730 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2731 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
d8d1f30b | 2732 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2733 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2734 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2735 | vlan_dev->name, vlan_id); |
2736 | break; | |
2737 | } | |
2738 | } | |
2739 | ||
2740 | if (vlan_id) { | |
ed4b9f80 | 2741 | ip_rt_put(rt); |
c3ade5ca JV |
2742 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2743 | vlan->vlan_ip, vlan_id); | |
2744 | continue; | |
2745 | } | |
2746 | ||
2747 | if (net_ratelimit()) { | |
a4aee5c8 | 2748 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2749 | bond->dev->name, &targets[i], |
d8d1f30b | 2750 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2751 | } |
ed4b9f80 | 2752 | ip_rt_put(rt); |
c3ade5ca JV |
2753 | } |
2754 | } | |
2755 | ||
2756 | /* | |
2757 | * Kick out a gratuitous ARP for an IP on the bonding master plus one | |
2758 | * for each VLAN above us. | |
b59f9f74 JV |
2759 | * |
2760 | * Caller must hold curr_slave_lock for read or better | |
c3ade5ca JV |
2761 | */ |
2762 | static void bond_send_gratuitous_arp(struct bonding *bond) | |
2763 | { | |
2764 | struct slave *slave = bond->curr_active_slave; | |
2765 | struct vlan_entry *vlan; | |
2766 | struct net_device *vlan_dev; | |
2767 | ||
a4aee5c8 JP |
2768 | pr_debug("bond_send_grat_arp: bond %s slave %s\n", |
2769 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
b59f9f74 JV |
2770 | |
2771 | if (!slave || !bond->send_grat_arp || | |
2772 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
c3ade5ca JV |
2773 | return; |
2774 | ||
b59f9f74 JV |
2775 | bond->send_grat_arp--; |
2776 | ||
c3ade5ca JV |
2777 | if (bond->master_ip) { |
2778 | bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip, | |
1053f62c | 2779 | bond->master_ip, 0); |
c3ade5ca JV |
2780 | } |
2781 | ||
f35188fa JV |
2782 | if (!bond->vlgrp) |
2783 | return; | |
2784 | ||
c3ade5ca | 2785 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2786 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2787 | if (vlan->vlan_ip) { |
2788 | bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip, | |
2789 | vlan->vlan_ip, vlan->vlan_id); | |
2790 | } | |
1da177e4 LT |
2791 | } |
2792 | } | |
2793 | ||
d3bb52b0 | 2794 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2795 | { |
2796 | int i; | |
d3bb52b0 | 2797 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2798 | |
f5b2b966 | 2799 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2800 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2801 | &sip, &tip, i, &targets[i], |
2802 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2803 | if (sip == targets[i]) { |
2804 | if (bond_has_this_ip(bond, tip)) | |
2805 | slave->last_arp_rx = jiffies; | |
2806 | return; | |
2807 | } | |
2808 | } | |
2809 | } | |
2810 | ||
2811 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2812 | { | |
2813 | struct arphdr *arp; | |
2814 | struct slave *slave; | |
2815 | struct bonding *bond; | |
2816 | unsigned char *arp_ptr; | |
d3bb52b0 | 2817 | __be32 sip, tip; |
f5b2b966 | 2818 | |
1f3c8804 AG |
2819 | if (dev->priv_flags & IFF_802_1Q_VLAN) { |
2820 | /* | |
2821 | * When using VLANS and bonding, dev and oriv_dev may be | |
2822 | * incorrect if the physical interface supports VLAN | |
2823 | * acceleration. With this change ARP validation now | |
2824 | * works for hosts only reachable on the VLAN interface. | |
2825 | */ | |
2826 | dev = vlan_dev_real_dev(dev); | |
2827 | orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif); | |
2828 | } | |
2829 | ||
f5b2b966 JV |
2830 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2831 | goto out; | |
2832 | ||
454d7c9b | 2833 | bond = netdev_priv(dev); |
f5b2b966 JV |
2834 | read_lock(&bond->lock); |
2835 | ||
5a03cdb7 | 2836 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
a4aee5c8 JP |
2837 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2838 | orig_dev ? orig_dev->name : "NULL"); | |
f5b2b966 JV |
2839 | |
2840 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2841 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2842 | goto out_unlock; | |
2843 | ||
b3053251 NH |
2844 | skb = skb_share_check(skb, GFP_ATOMIC); |
2845 | if (!skb) | |
2846 | goto out_unlock; | |
2847 | ||
988b7050 | 2848 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2849 | goto out_unlock; |
2850 | ||
d0a92be0 | 2851 | arp = arp_hdr(skb); |
f5b2b966 JV |
2852 | if (arp->ar_hln != dev->addr_len || |
2853 | skb->pkt_type == PACKET_OTHERHOST || | |
2854 | skb->pkt_type == PACKET_LOOPBACK || | |
2855 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2856 | arp->ar_pro != htons(ETH_P_IP) || | |
2857 | arp->ar_pln != 4) | |
2858 | goto out_unlock; | |
2859 | ||
2860 | arp_ptr = (unsigned char *)(arp + 1); | |
2861 | arp_ptr += dev->addr_len; | |
2862 | memcpy(&sip, arp_ptr, 4); | |
2863 | arp_ptr += 4 + dev->addr_len; | |
2864 | memcpy(&tip, arp_ptr, 4); | |
2865 | ||
5a03cdb7 | 2866 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2867 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2868 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2869 | &sip, &tip); | |
f5b2b966 JV |
2870 | |
2871 | /* | |
2872 | * Backup slaves won't see the ARP reply, but do come through | |
2873 | * here for each ARP probe (so we swap the sip/tip to validate | |
2874 | * the probe). In a "redundant switch, common router" type of | |
2875 | * configuration, the ARP probe will (hopefully) travel from | |
2876 | * the active, through one switch, the router, then the other | |
2877 | * switch before reaching the backup. | |
2878 | */ | |
e30bc066 | 2879 | if (bond_is_active_slave(slave)) |
f5b2b966 JV |
2880 | bond_validate_arp(bond, slave, sip, tip); |
2881 | else | |
2882 | bond_validate_arp(bond, slave, tip, sip); | |
2883 | ||
2884 | out_unlock: | |
2885 | read_unlock(&bond->lock); | |
2886 | out: | |
2887 | dev_kfree_skb(skb); | |
2888 | return NET_RX_SUCCESS; | |
2889 | } | |
2890 | ||
1da177e4 LT |
2891 | /* |
2892 | * this function is called regularly to monitor each slave's link | |
2893 | * ensuring that traffic is being sent and received when arp monitoring | |
2894 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2895 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2896 | * arp monitoring in active backup mode. | |
2897 | */ | |
1b76b316 | 2898 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2899 | { |
1b76b316 JV |
2900 | struct bonding *bond = container_of(work, struct bonding, |
2901 | arp_work.work); | |
1da177e4 LT |
2902 | struct slave *slave, *oldcurrent; |
2903 | int do_failover = 0; | |
2904 | int delta_in_ticks; | |
2905 | int i; | |
2906 | ||
2907 | read_lock(&bond->lock); | |
2908 | ||
5ce0da8f | 2909 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2910 | |
3d632c3f | 2911 | if (bond->kill_timers) |
1da177e4 | 2912 | goto out; |
1da177e4 | 2913 | |
3d632c3f | 2914 | if (bond->slave_cnt == 0) |
1da177e4 | 2915 | goto re_arm; |
1da177e4 LT |
2916 | |
2917 | read_lock(&bond->curr_slave_lock); | |
2918 | oldcurrent = bond->curr_active_slave; | |
2919 | read_unlock(&bond->curr_slave_lock); | |
2920 | ||
2921 | /* see if any of the previous devices are up now (i.e. they have | |
2922 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2923 | * the picture unless it is null. also, slave->jiffies is not needed | |
2924 | * here because we send an arp on each slave and give a slave as | |
2925 | * long as it needs to get the tx/rx within the delta. | |
2926 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2927 | * so it can wait | |
2928 | */ | |
2929 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2930 | unsigned long trans_start = dev_trans_start(slave->dev); |
2931 | ||
1da177e4 | 2932 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2933 | if (time_in_range(jiffies, |
2934 | trans_start - delta_in_ticks, | |
2935 | trans_start + delta_in_ticks) && | |
2936 | time_in_range(jiffies, | |
2937 | slave->dev->last_rx - delta_in_ticks, | |
2938 | slave->dev->last_rx + delta_in_ticks)) { | |
1da177e4 LT |
2939 | |
2940 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2941 | bond_set_active_slave(slave); |
1da177e4 LT |
2942 | |
2943 | /* primary_slave has no meaning in round-robin | |
2944 | * mode. the window of a slave being up and | |
2945 | * curr_active_slave being null after enslaving | |
2946 | * is closed. | |
2947 | */ | |
2948 | if (!oldcurrent) { | |
a4aee5c8 JP |
2949 | pr_info("%s: link status definitely up for interface %s, ", |
2950 | bond->dev->name, | |
2951 | slave->dev->name); | |
1da177e4 LT |
2952 | do_failover = 1; |
2953 | } else { | |
a4aee5c8 JP |
2954 | pr_info("%s: interface %s is now up\n", |
2955 | bond->dev->name, | |
2956 | slave->dev->name); | |
1da177e4 LT |
2957 | } |
2958 | } | |
2959 | } else { | |
2960 | /* slave->link == BOND_LINK_UP */ | |
2961 | ||
2962 | /* not all switches will respond to an arp request | |
2963 | * when the source ip is 0, so don't take the link down | |
2964 | * if we don't know our ip yet | |
2965 | */ | |
cb32f2a0 JB |
2966 | if (!time_in_range(jiffies, |
2967 | trans_start - delta_in_ticks, | |
2968 | trans_start + 2 * delta_in_ticks) || | |
2969 | !time_in_range(jiffies, | |
2970 | slave->dev->last_rx - delta_in_ticks, | |
2971 | slave->dev->last_rx + 2 * delta_in_ticks)) { | |
1da177e4 LT |
2972 | |
2973 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2974 | bond_set_backup_slave(slave); |
1da177e4 | 2975 | |
3d632c3f | 2976 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2977 | slave->link_failure_count++; |
1da177e4 | 2978 | |
a4aee5c8 JP |
2979 | pr_info("%s: interface %s is now down.\n", |
2980 | bond->dev->name, | |
2981 | slave->dev->name); | |
1da177e4 | 2982 | |
3d632c3f | 2983 | if (slave == oldcurrent) |
1da177e4 | 2984 | do_failover = 1; |
1da177e4 LT |
2985 | } |
2986 | } | |
2987 | ||
2988 | /* note: if switch is in round-robin mode, all links | |
2989 | * must tx arp to ensure all links rx an arp - otherwise | |
2990 | * links may oscillate or not come up at all; if switch is | |
2991 | * in something like xor mode, there is nothing we can | |
2992 | * do - all replies will be rx'ed on same link causing slaves | |
2993 | * to be unstable during low/no traffic periods | |
2994 | */ | |
3d632c3f | 2995 | if (IS_UP(slave->dev)) |
1da177e4 | 2996 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2997 | } |
2998 | ||
2999 | if (do_failover) { | |
e843fa50 | 3000 | block_netpoll_tx(); |
059fe7a5 | 3001 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
3002 | |
3003 | bond_select_active_slave(bond); | |
3004 | ||
059fe7a5 | 3005 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3006 | unblock_netpoll_tx(); |
1da177e4 LT |
3007 | } |
3008 | ||
3009 | re_arm: | |
1b76b316 JV |
3010 | if (bond->params.arp_interval) |
3011 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
3012 | out: |
3013 | read_unlock(&bond->lock); | |
3014 | } | |
3015 | ||
3016 | /* | |
b2220cad JV |
3017 | * Called to inspect slaves for active-backup mode ARP monitor link state |
3018 | * changes. Sets new_link in slaves to specify what action should take | |
3019 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
3020 | * to link states must be committed. | |
3021 | * | |
3022 | * Called with bond->lock held for read. | |
1da177e4 | 3023 | */ |
b2220cad | 3024 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 3025 | { |
1da177e4 | 3026 | struct slave *slave; |
b2220cad | 3027 | int i, commit = 0; |
cb32f2a0 | 3028 | unsigned long trans_start; |
1da177e4 | 3029 | |
b2220cad JV |
3030 | bond_for_each_slave(bond, slave, i) { |
3031 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 3032 | |
b2220cad | 3033 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
3034 | if (time_in_range(jiffies, |
3035 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3036 | slave_last_rx(bond, slave) + delta_in_ticks)) { | |
3037 | ||
b2220cad JV |
3038 | slave->new_link = BOND_LINK_UP; |
3039 | commit++; | |
3040 | } | |
1da177e4 | 3041 | |
b2220cad JV |
3042 | continue; |
3043 | } | |
1da177e4 | 3044 | |
b2220cad JV |
3045 | /* |
3046 | * Give slaves 2*delta after being enslaved or made | |
3047 | * active. This avoids bouncing, as the last receive | |
3048 | * times need a full ARP monitor cycle to be updated. | |
3049 | */ | |
cb32f2a0 JB |
3050 | if (time_in_range(jiffies, |
3051 | slave->jiffies - delta_in_ticks, | |
3052 | slave->jiffies + 2 * delta_in_ticks)) | |
b2220cad JV |
3053 | continue; |
3054 | ||
3055 | /* | |
3056 | * Backup slave is down if: | |
3057 | * - No current_arp_slave AND | |
3058 | * - more than 3*delta since last receive AND | |
3059 | * - the bond has an IP address | |
3060 | * | |
3061 | * Note: a non-null current_arp_slave indicates | |
3062 | * the curr_active_slave went down and we are | |
3063 | * searching for a new one; under this condition | |
3064 | * we only take the curr_active_slave down - this | |
3065 | * gives each slave a chance to tx/rx traffic | |
3066 | * before being taken out | |
3067 | */ | |
e30bc066 | 3068 | if (!bond_is_active_slave(slave) && |
b2220cad | 3069 | !bond->current_arp_slave && |
cb32f2a0 JB |
3070 | !time_in_range(jiffies, |
3071 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3072 | slave_last_rx(bond, slave) + 3 * delta_in_ticks)) { | |
3073 | ||
b2220cad JV |
3074 | slave->new_link = BOND_LINK_DOWN; |
3075 | commit++; | |
3076 | } | |
3077 | ||
3078 | /* | |
3079 | * Active slave is down if: | |
3080 | * - more than 2*delta since transmitting OR | |
3081 | * - (more than 2*delta since receive AND | |
3082 | * the bond has an IP address) | |
3083 | */ | |
cb32f2a0 | 3084 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 3085 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
3086 | (!time_in_range(jiffies, |
3087 | trans_start - delta_in_ticks, | |
3088 | trans_start + 2 * delta_in_ticks) || | |
3089 | !time_in_range(jiffies, | |
3090 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3091 | slave_last_rx(bond, slave) + 2 * delta_in_ticks))) { | |
3092 | ||
b2220cad JV |
3093 | slave->new_link = BOND_LINK_DOWN; |
3094 | commit++; | |
3095 | } | |
1da177e4 LT |
3096 | } |
3097 | ||
b2220cad JV |
3098 | return commit; |
3099 | } | |
1da177e4 | 3100 | |
b2220cad JV |
3101 | /* |
3102 | * Called to commit link state changes noted by inspection step of | |
3103 | * active-backup mode ARP monitor. | |
3104 | * | |
3105 | * Called with RTNL and bond->lock for read. | |
3106 | */ | |
3107 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
3108 | { | |
3109 | struct slave *slave; | |
3110 | int i; | |
cb32f2a0 | 3111 | unsigned long trans_start; |
1da177e4 | 3112 | |
b2220cad JV |
3113 | bond_for_each_slave(bond, slave, i) { |
3114 | switch (slave->new_link) { | |
3115 | case BOND_LINK_NOCHANGE: | |
3116 | continue; | |
ff59c456 | 3117 | |
b2220cad | 3118 | case BOND_LINK_UP: |
cb32f2a0 | 3119 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 3120 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
3121 | time_in_range(jiffies, |
3122 | trans_start - delta_in_ticks, | |
3123 | trans_start + delta_in_ticks)) || | |
b9f60253 | 3124 | bond->curr_active_slave != slave) { |
b2220cad | 3125 | slave->link = BOND_LINK_UP; |
b2220cad JV |
3126 | bond->current_arp_slave = NULL; |
3127 | ||
a4aee5c8 | 3128 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 3129 | bond->dev->name, slave->dev->name); |
b2220cad | 3130 | |
b9f60253 JP |
3131 | if (!bond->curr_active_slave || |
3132 | (slave == bond->primary_slave)) | |
3133 | goto do_failover; | |
1da177e4 | 3134 | |
b9f60253 | 3135 | } |
1da177e4 | 3136 | |
b9f60253 | 3137 | continue; |
1da177e4 | 3138 | |
b2220cad JV |
3139 | case BOND_LINK_DOWN: |
3140 | if (slave->link_failure_count < UINT_MAX) | |
3141 | slave->link_failure_count++; | |
3142 | ||
3143 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 3144 | bond_set_slave_inactive_flags(slave); |
b2220cad | 3145 | |
a4aee5c8 | 3146 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 3147 | bond->dev->name, slave->dev->name); |
b2220cad | 3148 | |
b9f60253 | 3149 | if (slave == bond->curr_active_slave) { |
b2220cad | 3150 | bond->current_arp_slave = NULL; |
b9f60253 | 3151 | goto do_failover; |
1da177e4 | 3152 | } |
b9f60253 JP |
3153 | |
3154 | continue; | |
b2220cad JV |
3155 | |
3156 | default: | |
a4aee5c8 | 3157 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
3158 | bond->dev->name, slave->new_link, |
3159 | slave->dev->name); | |
b9f60253 | 3160 | continue; |
1da177e4 | 3161 | } |
1da177e4 | 3162 | |
b9f60253 JP |
3163 | do_failover: |
3164 | ASSERT_RTNL(); | |
e843fa50 | 3165 | block_netpoll_tx(); |
b2220cad | 3166 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3167 | bond_select_active_slave(bond); |
b2220cad | 3168 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3169 | unblock_netpoll_tx(); |
b2220cad | 3170 | } |
1da177e4 | 3171 | |
b2220cad JV |
3172 | bond_set_carrier(bond); |
3173 | } | |
1da177e4 | 3174 | |
b2220cad JV |
3175 | /* |
3176 | * Send ARP probes for active-backup mode ARP monitor. | |
3177 | * | |
3178 | * Called with bond->lock held for read. | |
3179 | */ | |
3180 | static void bond_ab_arp_probe(struct bonding *bond) | |
3181 | { | |
3182 | struct slave *slave; | |
3183 | int i; | |
1da177e4 | 3184 | |
b2220cad | 3185 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3186 | |
b2220cad | 3187 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3188 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3189 | bond->current_arp_slave->dev->name, | |
3190 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3191 | |
b2220cad JV |
3192 | if (bond->curr_active_slave) { |
3193 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3194 | read_unlock(&bond->curr_slave_lock); | |
3195 | return; | |
3196 | } | |
1da177e4 | 3197 | |
b2220cad | 3198 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3199 | |
b2220cad JV |
3200 | /* if we don't have a curr_active_slave, search for the next available |
3201 | * backup slave from the current_arp_slave and make it the candidate | |
3202 | * for becoming the curr_active_slave | |
3203 | */ | |
1da177e4 | 3204 | |
b2220cad JV |
3205 | if (!bond->current_arp_slave) { |
3206 | bond->current_arp_slave = bond->first_slave; | |
3207 | if (!bond->current_arp_slave) | |
3208 | return; | |
3209 | } | |
1da177e4 | 3210 | |
b2220cad | 3211 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3212 | |
b2220cad JV |
3213 | /* search for next candidate */ |
3214 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3215 | if (IS_UP(slave->dev)) { | |
3216 | slave->link = BOND_LINK_BACK; | |
3217 | bond_set_slave_active_flags(slave); | |
3218 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3219 | slave->jiffies = jiffies; |
b2220cad JV |
3220 | bond->current_arp_slave = slave; |
3221 | break; | |
1da177e4 LT |
3222 | } |
3223 | ||
b2220cad JV |
3224 | /* if the link state is up at this point, we |
3225 | * mark it down - this can happen if we have | |
3226 | * simultaneous link failures and | |
3227 | * reselect_active_interface doesn't make this | |
3228 | * one the current slave so it is still marked | |
3229 | * up when it is actually down | |
1da177e4 | 3230 | */ |
b2220cad JV |
3231 | if (slave->link == BOND_LINK_UP) { |
3232 | slave->link = BOND_LINK_DOWN; | |
3233 | if (slave->link_failure_count < UINT_MAX) | |
3234 | slave->link_failure_count++; | |
1da177e4 | 3235 | |
b2220cad JV |
3236 | bond_set_slave_inactive_flags(slave); |
3237 | ||
a4aee5c8 JP |
3238 | pr_info("%s: backup interface %s is now down.\n", |
3239 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3240 | } |
b2220cad JV |
3241 | } |
3242 | } | |
1da177e4 | 3243 | |
b2220cad JV |
3244 | void bond_activebackup_arp_mon(struct work_struct *work) |
3245 | { | |
3246 | struct bonding *bond = container_of(work, struct bonding, | |
3247 | arp_work.work); | |
3248 | int delta_in_ticks; | |
1da177e4 | 3249 | |
b2220cad | 3250 | read_lock(&bond->lock); |
1da177e4 | 3251 | |
b2220cad JV |
3252 | if (bond->kill_timers) |
3253 | goto out; | |
1da177e4 | 3254 | |
b2220cad | 3255 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3256 | |
b2220cad JV |
3257 | if (bond->slave_cnt == 0) |
3258 | goto re_arm; | |
3259 | ||
b59f9f74 JV |
3260 | if (bond->send_grat_arp) { |
3261 | read_lock(&bond->curr_slave_lock); | |
3262 | bond_send_gratuitous_arp(bond); | |
3263 | read_unlock(&bond->curr_slave_lock); | |
3264 | } | |
3265 | ||
305d552a BH |
3266 | if (bond->send_unsol_na) { |
3267 | read_lock(&bond->curr_slave_lock); | |
3268 | bond_send_unsolicited_na(bond); | |
3269 | read_unlock(&bond->curr_slave_lock); | |
3270 | } | |
3271 | ||
b2220cad JV |
3272 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3273 | read_unlock(&bond->lock); | |
3274 | rtnl_lock(); | |
3275 | read_lock(&bond->lock); | |
3276 | ||
3277 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3278 | ||
3279 | read_unlock(&bond->lock); | |
3280 | rtnl_unlock(); | |
3281 | read_lock(&bond->lock); | |
1da177e4 LT |
3282 | } |
3283 | ||
b2220cad JV |
3284 | bond_ab_arp_probe(bond); |
3285 | ||
1da177e4 | 3286 | re_arm: |
3d632c3f | 3287 | if (bond->params.arp_interval) |
1b76b316 | 3288 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3289 | out: |
3290 | read_unlock(&bond->lock); | |
3291 | } | |
3292 | ||
1da177e4 LT |
3293 | /*-------------------------- netdev event handling --------------------------*/ |
3294 | ||
3295 | /* | |
3296 | * Change device name | |
3297 | */ | |
3298 | static int bond_event_changename(struct bonding *bond) | |
3299 | { | |
1da177e4 LT |
3300 | bond_remove_proc_entry(bond); |
3301 | bond_create_proc_entry(bond); | |
7e083840 | 3302 | |
f073c7ca TI |
3303 | bond_debug_reregister(bond); |
3304 | ||
1da177e4 LT |
3305 | return NOTIFY_DONE; |
3306 | } | |
3307 | ||
3d632c3f SH |
3308 | static int bond_master_netdev_event(unsigned long event, |
3309 | struct net_device *bond_dev) | |
1da177e4 | 3310 | { |
454d7c9b | 3311 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3312 | |
3313 | switch (event) { | |
3314 | case NETDEV_CHANGENAME: | |
3315 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3316 | default: |
3317 | break; | |
3318 | } | |
3319 | ||
3320 | return NOTIFY_DONE; | |
3321 | } | |
3322 | ||
3d632c3f SH |
3323 | static int bond_slave_netdev_event(unsigned long event, |
3324 | struct net_device *slave_dev) | |
1da177e4 LT |
3325 | { |
3326 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3327 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3328 | |
3329 | switch (event) { | |
3330 | case NETDEV_UNREGISTER: | |
3331 | if (bond_dev) { | |
1284cd3a JV |
3332 | if (bond->setup_by_slave) |
3333 | bond_release_and_destroy(bond_dev, slave_dev); | |
3334 | else | |
3335 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3336 | } |
3337 | break; | |
3338 | case NETDEV_CHANGE: | |
17d04500 JV |
3339 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3340 | struct slave *slave; | |
3341 | ||
3342 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3343 | if (slave) { | |
5d30530e | 3344 | u32 old_speed = slave->speed; |
65cce19c | 3345 | u8 old_duplex = slave->duplex; |
17d04500 JV |
3346 | |
3347 | bond_update_speed_duplex(slave); | |
3348 | ||
3349 | if (bond_is_lb(bond)) | |
3350 | break; | |
3351 | ||
3352 | if (old_speed != slave->speed) | |
3353 | bond_3ad_adapter_speed_changed(slave); | |
3354 | if (old_duplex != slave->duplex) | |
3355 | bond_3ad_adapter_duplex_changed(slave); | |
3356 | } | |
3357 | } | |
3358 | ||
1da177e4 LT |
3359 | break; |
3360 | case NETDEV_DOWN: | |
3361 | /* | |
3362 | * ... Or is it this? | |
3363 | */ | |
3364 | break; | |
3365 | case NETDEV_CHANGEMTU: | |
3366 | /* | |
3367 | * TODO: Should slaves be allowed to | |
3368 | * independently alter their MTU? For | |
3369 | * an active-backup bond, slaves need | |
3370 | * not be the same type of device, so | |
3371 | * MTUs may vary. For other modes, | |
3372 | * slaves arguably should have the | |
3373 | * same MTUs. To do this, we'd need to | |
3374 | * take over the slave's change_mtu | |
3375 | * function for the duration of their | |
3376 | * servitude. | |
3377 | */ | |
3378 | break; | |
3379 | case NETDEV_CHANGENAME: | |
3380 | /* | |
3381 | * TODO: handle changing the primary's name | |
3382 | */ | |
3383 | break; | |
8531c5ff AK |
3384 | case NETDEV_FEAT_CHANGE: |
3385 | bond_compute_features(bond); | |
3386 | break; | |
1da177e4 LT |
3387 | default: |
3388 | break; | |
3389 | } | |
3390 | ||
3391 | return NOTIFY_DONE; | |
3392 | } | |
3393 | ||
3394 | /* | |
3395 | * bond_netdev_event: handle netdev notifier chain events. | |
3396 | * | |
3397 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3398 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3399 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3400 | * dev_probe_lock). | |
3401 | */ | |
3d632c3f SH |
3402 | static int bond_netdev_event(struct notifier_block *this, |
3403 | unsigned long event, void *ptr) | |
1da177e4 LT |
3404 | { |
3405 | struct net_device *event_dev = (struct net_device *)ptr; | |
3406 | ||
5a03cdb7 | 3407 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3408 | event_dev ? event_dev->name : "None", |
3409 | event); | |
1da177e4 | 3410 | |
0b680e75 JV |
3411 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3412 | return NOTIFY_DONE; | |
3413 | ||
1da177e4 | 3414 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3415 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3416 | return bond_master_netdev_event(event, event_dev); |
3417 | } | |
3418 | ||
3419 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3420 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3421 | return bond_slave_netdev_event(event, event_dev); |
3422 | } | |
3423 | ||
3424 | return NOTIFY_DONE; | |
3425 | } | |
3426 | ||
c3ade5ca JV |
3427 | /* |
3428 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3429 | * | |
3430 | * We keep track of device IPs primarily to use as source addresses in | |
3431 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3432 | * | |
3433 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3434 | */ | |
3435 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3436 | { | |
3437 | struct in_ifaddr *ifa = ptr; | |
3438 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
ec87fd3b | 3439 | struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id); |
0883beca PE |
3440 | struct bonding *bond; |
3441 | struct vlan_entry *vlan; | |
c3ade5ca | 3442 | |
ec87fd3b | 3443 | list_for_each_entry(bond, &bn->dev_list, bond_list) { |
c3ade5ca JV |
3444 | if (bond->dev == event_dev) { |
3445 | switch (event) { | |
3446 | case NETDEV_UP: | |
3447 | bond->master_ip = ifa->ifa_local; | |
3448 | return NOTIFY_OK; | |
3449 | case NETDEV_DOWN: | |
3450 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3451 | return NOTIFY_OK; | |
3452 | default: | |
3453 | return NOTIFY_DONE; | |
3454 | } | |
3455 | } | |
3456 | ||
0883beca | 3457 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
3458 | if (!bond->vlgrp) |
3459 | continue; | |
5c15bdec | 3460 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3461 | if (vlan_dev == event_dev) { |
3462 | switch (event) { | |
3463 | case NETDEV_UP: | |
3464 | vlan->vlan_ip = ifa->ifa_local; | |
3465 | return NOTIFY_OK; | |
3466 | case NETDEV_DOWN: | |
3467 | vlan->vlan_ip = | |
3468 | bond_glean_dev_ip(vlan_dev); | |
3469 | return NOTIFY_OK; | |
3470 | default: | |
3471 | return NOTIFY_DONE; | |
3472 | } | |
3473 | } | |
3474 | } | |
3475 | } | |
3476 | return NOTIFY_DONE; | |
3477 | } | |
3478 | ||
1da177e4 LT |
3479 | static struct notifier_block bond_netdev_notifier = { |
3480 | .notifier_call = bond_netdev_event, | |
3481 | }; | |
3482 | ||
c3ade5ca JV |
3483 | static struct notifier_block bond_inetaddr_notifier = { |
3484 | .notifier_call = bond_inetaddr_event, | |
3485 | }; | |
3486 | ||
1da177e4 LT |
3487 | /*-------------------------- Packet type handling ---------------------------*/ |
3488 | ||
3489 | /* register to receive lacpdus on a bond */ | |
3490 | static void bond_register_lacpdu(struct bonding *bond) | |
3491 | { | |
3492 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3493 | ||
3494 | /* initialize packet type */ | |
3495 | pk_type->type = PKT_TYPE_LACPDU; | |
3496 | pk_type->dev = bond->dev; | |
3497 | pk_type->func = bond_3ad_lacpdu_recv; | |
3498 | ||
3499 | dev_add_pack(pk_type); | |
3500 | } | |
3501 | ||
3502 | /* unregister to receive lacpdus on a bond */ | |
3503 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3504 | { | |
3505 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3506 | } | |
3507 | ||
f5b2b966 JV |
3508 | void bond_register_arp(struct bonding *bond) |
3509 | { | |
3510 | struct packet_type *pt = &bond->arp_mon_pt; | |
3511 | ||
c4f283b1 JV |
3512 | if (pt->type) |
3513 | return; | |
3514 | ||
f5b2b966 | 3515 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3516 | pt->dev = bond->dev; |
f5b2b966 JV |
3517 | pt->func = bond_arp_rcv; |
3518 | dev_add_pack(pt); | |
3519 | } | |
3520 | ||
3521 | void bond_unregister_arp(struct bonding *bond) | |
3522 | { | |
c4f283b1 JV |
3523 | struct packet_type *pt = &bond->arp_mon_pt; |
3524 | ||
3525 | dev_remove_pack(pt); | |
3526 | pt->type = 0; | |
f5b2b966 JV |
3527 | } |
3528 | ||
169a3e66 JV |
3529 | /*---------------------------- Hashing Policies -----------------------------*/ |
3530 | ||
6f6652be JV |
3531 | /* |
3532 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3533 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3534 | */ | |
a361c83c | 3535 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3536 | { |
3537 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3538 | struct iphdr *iph = ip_hdr(skb); | |
3539 | ||
f14c4e4e | 3540 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3541 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3542 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3543 | } |
3544 | ||
d3da6831 | 3545 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3546 | } |
3547 | ||
169a3e66 | 3548 | /* |
59c51591 | 3549 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3550 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3551 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3552 | */ | |
a361c83c | 3553 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3554 | { |
3555 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3556 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3557 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3558 | int layer4_xor = 0; |
3559 | ||
f14c4e4e BH |
3560 | if (skb->protocol == htons(ETH_P_IP)) { |
3561 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3562 | (iph->protocol == IPPROTO_TCP || |
3563 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3564 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3565 | } |
3566 | return (layer4_xor ^ | |
3567 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3568 | ||
3569 | } | |
3570 | ||
d3da6831 | 3571 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3572 | } |
3573 | ||
3574 | /* | |
3575 | * Hash for the output device based upon layer 2 data | |
3576 | */ | |
a361c83c | 3577 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3578 | { |
3579 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3580 | ||
d3da6831 | 3581 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3582 | } |
3583 | ||
1da177e4 LT |
3584 | /*-------------------------- Device entry points ----------------------------*/ |
3585 | ||
3586 | static int bond_open(struct net_device *bond_dev) | |
3587 | { | |
454d7c9b | 3588 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3589 | |
3590 | bond->kill_timers = 0; | |
3591 | ||
5a37e8ca FL |
3592 | INIT_DELAYED_WORK(&bond->mcast_work, bond_resend_igmp_join_requests_delayed); |
3593 | ||
58402054 | 3594 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3595 | /* bond_alb_initialize must be called before the timer |
3596 | * is started. | |
3597 | */ | |
3598 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3599 | /* something went wrong - fail the open operation */ | |
b473946a | 3600 | return -ENOMEM; |
1da177e4 LT |
3601 | } |
3602 | ||
1b76b316 JV |
3603 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3604 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3605 | } |
3606 | ||
3607 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3608 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3609 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3610 | } |
3611 | ||
3612 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3613 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3614 | INIT_DELAYED_WORK(&bond->arp_work, | |
3615 | bond_activebackup_arp_mon); | |
3616 | else | |
3617 | INIT_DELAYED_WORK(&bond->arp_work, | |
3618 | bond_loadbalance_arp_mon); | |
3619 | ||
3620 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3621 | if (bond->params.arp_validate) |
3622 | bond_register_arp(bond); | |
1da177e4 LT |
3623 | } |
3624 | ||
3625 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3626 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3627 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3628 | /* register to receive LACPDUs */ |
3629 | bond_register_lacpdu(bond); | |
fd989c83 | 3630 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3631 | } |
3632 | ||
3633 | return 0; | |
3634 | } | |
3635 | ||
3636 | static int bond_close(struct net_device *bond_dev) | |
3637 | { | |
454d7c9b | 3638 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3639 | |
3640 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3641 | /* Unregister the receive of LACPDUs */ | |
3642 | bond_unregister_lacpdu(bond); | |
3643 | } | |
3644 | ||
f5b2b966 JV |
3645 | if (bond->params.arp_validate) |
3646 | bond_unregister_arp(bond); | |
3647 | ||
1da177e4 LT |
3648 | write_lock_bh(&bond->lock); |
3649 | ||
b59f9f74 | 3650 | bond->send_grat_arp = 0; |
305d552a | 3651 | bond->send_unsol_na = 0; |
1da177e4 LT |
3652 | |
3653 | /* signal timers not to re-arm */ | |
3654 | bond->kill_timers = 1; | |
3655 | ||
3656 | write_unlock_bh(&bond->lock); | |
3657 | ||
1da177e4 | 3658 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3659 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3660 | } |
3661 | ||
3662 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3663 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3664 | } |
3665 | ||
3666 | switch (bond->params.mode) { | |
3667 | case BOND_MODE_8023AD: | |
1b76b316 | 3668 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3669 | break; |
3670 | case BOND_MODE_TLB: | |
3671 | case BOND_MODE_ALB: | |
1b76b316 | 3672 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3673 | break; |
3674 | default: | |
3675 | break; | |
3676 | } | |
3677 | ||
5a37e8ca FL |
3678 | if (delayed_work_pending(&bond->mcast_work)) |
3679 | cancel_delayed_work(&bond->mcast_work); | |
1da177e4 | 3680 | |
58402054 | 3681 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3682 | /* Must be called only after all |
3683 | * slaves have been released | |
3684 | */ | |
3685 | bond_alb_deinitialize(bond); | |
3686 | } | |
3687 | ||
3688 | return 0; | |
3689 | } | |
3690 | ||
28172739 ED |
3691 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3692 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3693 | { |
454d7c9b | 3694 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3695 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3696 | struct slave *slave; |
3697 | int i; | |
3698 | ||
28172739 | 3699 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3700 | |
3701 | read_lock_bh(&bond->lock); | |
3702 | ||
3703 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3704 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3705 | dev_get_stats(slave->dev, &temp); |
3706 | ||
3707 | stats->rx_packets += sstats->rx_packets; | |
3708 | stats->rx_bytes += sstats->rx_bytes; | |
3709 | stats->rx_errors += sstats->rx_errors; | |
3710 | stats->rx_dropped += sstats->rx_dropped; | |
3711 | ||
3712 | stats->tx_packets += sstats->tx_packets; | |
3713 | stats->tx_bytes += sstats->tx_bytes; | |
3714 | stats->tx_errors += sstats->tx_errors; | |
3715 | stats->tx_dropped += sstats->tx_dropped; | |
3716 | ||
3717 | stats->multicast += sstats->multicast; | |
3718 | stats->collisions += sstats->collisions; | |
3719 | ||
3720 | stats->rx_length_errors += sstats->rx_length_errors; | |
3721 | stats->rx_over_errors += sstats->rx_over_errors; | |
3722 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3723 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3724 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3725 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3726 | ||
3727 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3728 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3729 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3730 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3731 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3732 | } |
3733 | ||
1da177e4 LT |
3734 | read_unlock_bh(&bond->lock); |
3735 | ||
3736 | return stats; | |
3737 | } | |
3738 | ||
3739 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3740 | { | |
3741 | struct net_device *slave_dev = NULL; | |
3742 | struct ifbond k_binfo; | |
3743 | struct ifbond __user *u_binfo = NULL; | |
3744 | struct ifslave k_sinfo; | |
3745 | struct ifslave __user *u_sinfo = NULL; | |
3746 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3747 | int res = 0; |
3748 | ||
a4aee5c8 | 3749 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3750 | |
3751 | switch (cmd) { | |
1da177e4 LT |
3752 | case SIOCGMIIPHY: |
3753 | mii = if_mii(ifr); | |
3d632c3f | 3754 | if (!mii) |
1da177e4 | 3755 | return -EINVAL; |
3d632c3f | 3756 | |
1da177e4 LT |
3757 | mii->phy_id = 0; |
3758 | /* Fall Through */ | |
3759 | case SIOCGMIIREG: | |
3760 | /* | |
3761 | * We do this again just in case we were called by SIOCGMIIREG | |
3762 | * instead of SIOCGMIIPHY. | |
3763 | */ | |
3764 | mii = if_mii(ifr); | |
3d632c3f | 3765 | if (!mii) |
1da177e4 | 3766 | return -EINVAL; |
3d632c3f | 3767 | |
1da177e4 LT |
3768 | |
3769 | if (mii->reg_num == 1) { | |
454d7c9b | 3770 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3771 | mii->val_out = 0; |
6603a6f2 | 3772 | read_lock(&bond->lock); |
1da177e4 | 3773 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3774 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3775 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3776 | |
1da177e4 | 3777 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3778 | read_unlock(&bond->lock); |
1da177e4 LT |
3779 | } |
3780 | ||
3781 | return 0; | |
3782 | case BOND_INFO_QUERY_OLD: | |
3783 | case SIOCBONDINFOQUERY: | |
3784 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3785 | ||
3d632c3f | 3786 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3787 | return -EFAULT; |
1da177e4 LT |
3788 | |
3789 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3790 | if (res == 0 && |
3791 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3792 | return -EFAULT; | |
1da177e4 LT |
3793 | |
3794 | return res; | |
3795 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3796 | case SIOCBONDSLAVEINFOQUERY: | |
3797 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3798 | ||
3d632c3f | 3799 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3800 | return -EFAULT; |
1da177e4 LT |
3801 | |
3802 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3803 | if (res == 0 && |
3804 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3805 | return -EFAULT; | |
1da177e4 LT |
3806 | |
3807 | return res; | |
3808 | default: | |
3809 | /* Go on */ | |
3810 | break; | |
3811 | } | |
3812 | ||
3d632c3f | 3813 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3814 | return -EPERM; |
1da177e4 | 3815 | |
ec87fd3b | 3816 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3817 | |
a4aee5c8 | 3818 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3819 | |
3d632c3f | 3820 | if (!slave_dev) |
1da177e4 | 3821 | res = -ENODEV; |
3d632c3f | 3822 | else { |
a4aee5c8 | 3823 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3824 | switch (cmd) { |
3825 | case BOND_ENSLAVE_OLD: | |
3826 | case SIOCBONDENSLAVE: | |
3827 | res = bond_enslave(bond_dev, slave_dev); | |
3828 | break; | |
3829 | case BOND_RELEASE_OLD: | |
3830 | case SIOCBONDRELEASE: | |
3831 | res = bond_release(bond_dev, slave_dev); | |
3832 | break; | |
3833 | case BOND_SETHWADDR_OLD: | |
3834 | case SIOCBONDSETHWADDR: | |
3835 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3836 | break; | |
3837 | case BOND_CHANGE_ACTIVE_OLD: | |
3838 | case SIOCBONDCHANGEACTIVE: | |
3839 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3840 | break; | |
3841 | default: | |
3842 | res = -EOPNOTSUPP; | |
3843 | } | |
3844 | ||
3845 | dev_put(slave_dev); | |
3846 | } | |
3847 | ||
1da177e4 LT |
3848 | return res; |
3849 | } | |
3850 | ||
22bedad3 JP |
3851 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3852 | struct netdev_hw_addr_list *list, | |
3853 | int addrlen) | |
3854 | { | |
3855 | struct netdev_hw_addr *ha; | |
3856 | ||
3857 | netdev_hw_addr_list_for_each(ha, list) | |
3858 | if (!memcmp(ha->addr, addr, addrlen)) | |
3859 | return true; | |
3860 | ||
3861 | return false; | |
3862 | } | |
3863 | ||
1da177e4 LT |
3864 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3865 | { | |
454d7c9b | 3866 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 JP |
3867 | struct netdev_hw_addr *ha; |
3868 | bool found; | |
1da177e4 | 3869 | |
1da177e4 LT |
3870 | /* |
3871 | * Do promisc before checking multicast_mode | |
3872 | */ | |
3d632c3f | 3873 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
3874 | /* |
3875 | * FIXME: Need to handle the error when one of the multi-slaves | |
3876 | * encounters error. | |
3877 | */ | |
1da177e4 | 3878 | bond_set_promiscuity(bond, 1); |
1da177e4 | 3879 | |
3d632c3f SH |
3880 | |
3881 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 3882 | bond_set_promiscuity(bond, -1); |
3d632c3f | 3883 | |
1da177e4 LT |
3884 | |
3885 | /* set allmulti flag to slaves */ | |
3d632c3f | 3886 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
3887 | /* |
3888 | * FIXME: Need to handle the error when one of the multi-slaves | |
3889 | * encounters error. | |
3890 | */ | |
1da177e4 | 3891 | bond_set_allmulti(bond, 1); |
1da177e4 | 3892 | |
3d632c3f SH |
3893 | |
3894 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 3895 | bond_set_allmulti(bond, -1); |
3d632c3f | 3896 | |
1da177e4 | 3897 | |
80ee5ad2 JV |
3898 | read_lock(&bond->lock); |
3899 | ||
1da177e4 LT |
3900 | bond->flags = bond_dev->flags; |
3901 | ||
3902 | /* looking for addresses to add to slaves' mc list */ | |
22bedad3 JP |
3903 | netdev_for_each_mc_addr(ha, bond_dev) { |
3904 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
3905 | bond_dev->addr_len); | |
3906 | if (!found) | |
3907 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
3908 | } |
3909 | ||
3910 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
3911 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
3912 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
3913 | bond_dev->addr_len); | |
3914 | if (!found) | |
3915 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
3916 | } |
3917 | ||
3918 | /* save master's multicast list */ | |
22bedad3 JP |
3919 | __hw_addr_flush(&bond->mc_list); |
3920 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
3921 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 3922 | |
80ee5ad2 | 3923 | read_unlock(&bond->lock); |
1da177e4 LT |
3924 | } |
3925 | ||
00829823 SH |
3926 | static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms) |
3927 | { | |
3928 | struct bonding *bond = netdev_priv(dev); | |
3929 | struct slave *slave = bond->first_slave; | |
3930 | ||
3931 | if (slave) { | |
3932 | const struct net_device_ops *slave_ops | |
3933 | = slave->dev->netdev_ops; | |
3934 | if (slave_ops->ndo_neigh_setup) | |
72e2240f | 3935 | return slave_ops->ndo_neigh_setup(slave->dev, parms); |
00829823 SH |
3936 | } |
3937 | return 0; | |
3938 | } | |
3939 | ||
1da177e4 LT |
3940 | /* |
3941 | * Change the MTU of all of a master's slaves to match the master | |
3942 | */ | |
3943 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3944 | { | |
454d7c9b | 3945 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3946 | struct slave *slave, *stop_at; |
3947 | int res = 0; | |
3948 | int i; | |
3949 | ||
5a03cdb7 | 3950 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3951 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3952 | |
3953 | /* Can't hold bond->lock with bh disabled here since | |
3954 | * some base drivers panic. On the other hand we can't | |
3955 | * hold bond->lock without bh disabled because we'll | |
3956 | * deadlock. The only solution is to rely on the fact | |
3957 | * that we're under rtnl_lock here, and the slaves | |
3958 | * list won't change. This doesn't solve the problem | |
3959 | * of setting the slave's MTU while it is | |
3960 | * transmitting, but the assumption is that the base | |
3961 | * driver can handle that. | |
3962 | * | |
3963 | * TODO: figure out a way to safely iterate the slaves | |
3964 | * list, but without holding a lock around the actual | |
3965 | * call to the base driver. | |
3966 | */ | |
3967 | ||
3968 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3969 | pr_debug("s %p s->p %p c_m %p\n", |
3970 | slave, | |
3971 | slave->prev, | |
3972 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3973 | |
1da177e4 LT |
3974 | res = dev_set_mtu(slave->dev, new_mtu); |
3975 | ||
3976 | if (res) { | |
3977 | /* If we failed to set the slave's mtu to the new value | |
3978 | * we must abort the operation even in ACTIVE_BACKUP | |
3979 | * mode, because if we allow the backup slaves to have | |
3980 | * different mtu values than the active slave we'll | |
3981 | * need to change their mtu when doing a failover. That | |
3982 | * means changing their mtu from timer context, which | |
3983 | * is probably not a good idea. | |
3984 | */ | |
5a03cdb7 | 3985 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3986 | goto unwind; |
3987 | } | |
3988 | } | |
3989 | ||
3990 | bond_dev->mtu = new_mtu; | |
3991 | ||
3992 | return 0; | |
3993 | ||
3994 | unwind: | |
3995 | /* unwind from head to the slave that failed */ | |
3996 | stop_at = slave; | |
3997 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3998 | int tmp_res; | |
3999 | ||
4000 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
4001 | if (tmp_res) { | |
a4aee5c8 JP |
4002 | pr_debug("unwind err %d dev %s\n", |
4003 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4004 | } |
4005 | } | |
4006 | ||
4007 | return res; | |
4008 | } | |
4009 | ||
4010 | /* | |
4011 | * Change HW address | |
4012 | * | |
4013 | * Note that many devices must be down to change the HW address, and | |
4014 | * downing the master releases all slaves. We can make bonds full of | |
4015 | * bonding devices to test this, however. | |
4016 | */ | |
4017 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4018 | { | |
454d7c9b | 4019 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4020 | struct sockaddr *sa = addr, tmp_sa; |
4021 | struct slave *slave, *stop_at; | |
4022 | int res = 0; | |
4023 | int i; | |
4024 | ||
eb7cc59a SH |
4025 | if (bond->params.mode == BOND_MODE_ALB) |
4026 | return bond_alb_set_mac_address(bond_dev, addr); | |
4027 | ||
4028 | ||
a4aee5c8 JP |
4029 | pr_debug("bond=%p, name=%s\n", |
4030 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 4031 | |
dd957c57 | 4032 | /* |
3915c1e8 JV |
4033 | * If fail_over_mac is set to active, do nothing and return |
4034 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 4035 | */ |
3915c1e8 | 4036 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 4037 | return 0; |
2ab82852 | 4038 | |
3d632c3f | 4039 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 4040 | return -EADDRNOTAVAIL; |
1da177e4 LT |
4041 | |
4042 | /* Can't hold bond->lock with bh disabled here since | |
4043 | * some base drivers panic. On the other hand we can't | |
4044 | * hold bond->lock without bh disabled because we'll | |
4045 | * deadlock. The only solution is to rely on the fact | |
4046 | * that we're under rtnl_lock here, and the slaves | |
4047 | * list won't change. This doesn't solve the problem | |
4048 | * of setting the slave's hw address while it is | |
4049 | * transmitting, but the assumption is that the base | |
4050 | * driver can handle that. | |
4051 | * | |
4052 | * TODO: figure out a way to safely iterate the slaves | |
4053 | * list, but without holding a lock around the actual | |
4054 | * call to the base driver. | |
4055 | */ | |
4056 | ||
4057 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 4058 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 4059 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 4060 | |
eb7cc59a | 4061 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 4062 | res = -EOPNOTSUPP; |
5a03cdb7 | 4063 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
4064 | goto unwind; |
4065 | } | |
4066 | ||
4067 | res = dev_set_mac_address(slave->dev, addr); | |
4068 | if (res) { | |
4069 | /* TODO: consider downing the slave | |
4070 | * and retry ? | |
4071 | * User should expect communications | |
4072 | * breakage anyway until ARP finish | |
4073 | * updating, so... | |
4074 | */ | |
5a03cdb7 | 4075 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4076 | goto unwind; |
4077 | } | |
4078 | } | |
4079 | ||
4080 | /* success */ | |
4081 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4082 | return 0; | |
4083 | ||
4084 | unwind: | |
4085 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4086 | tmp_sa.sa_family = bond_dev->type; | |
4087 | ||
4088 | /* unwind from head to the slave that failed */ | |
4089 | stop_at = slave; | |
4090 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4091 | int tmp_res; | |
4092 | ||
4093 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4094 | if (tmp_res) { | |
a4aee5c8 JP |
4095 | pr_debug("unwind err %d dev %s\n", |
4096 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4097 | } |
4098 | } | |
4099 | ||
4100 | return res; | |
4101 | } | |
4102 | ||
4103 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4104 | { | |
454d7c9b | 4105 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4106 | struct slave *slave, *start_at; |
cf5f9044 | 4107 | int i, slave_no, res = 1; |
a2fd940f | 4108 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
4109 | |
4110 | read_lock(&bond->lock); | |
4111 | ||
3d632c3f | 4112 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4113 | goto out; |
cf5f9044 | 4114 | /* |
a2fd940f AG |
4115 | * Start with the curr_active_slave that joined the bond as the |
4116 | * default for sending IGMP traffic. For failover purposes one | |
4117 | * needs to maintain some consistency for the interface that will | |
4118 | * send the join/membership reports. The curr_active_slave found | |
4119 | * will send all of this type of traffic. | |
cf5f9044 | 4120 | */ |
00ae7028 | 4121 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4122 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4123 | |
a2fd940f AG |
4124 | read_lock(&bond->curr_slave_lock); |
4125 | slave = bond->curr_active_slave; | |
4126 | read_unlock(&bond->curr_slave_lock); | |
4127 | ||
4128 | if (!slave) | |
4129 | goto out; | |
4130 | } else { | |
4131 | /* | |
4132 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4133 | * that as being rare enough not to justify using an | |
4134 | * atomic op here. | |
4135 | */ | |
4136 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4137 | ||
4138 | bond_for_each_slave(bond, slave, i) { | |
4139 | slave_no--; | |
4140 | if (slave_no < 0) | |
4141 | break; | |
4142 | } | |
1da177e4 LT |
4143 | } |
4144 | ||
cf5f9044 | 4145 | start_at = slave; |
1da177e4 LT |
4146 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4147 | if (IS_UP(slave->dev) && | |
4148 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4149 | bond_is_active_slave(slave)) { |
1da177e4 | 4150 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
4151 | break; |
4152 | } | |
4153 | } | |
4154 | ||
1da177e4 LT |
4155 | out: |
4156 | if (res) { | |
4157 | /* no suitable interface, frame not sent */ | |
4158 | dev_kfree_skb(skb); | |
4159 | } | |
4160 | read_unlock(&bond->lock); | |
ec634fe3 | 4161 | return NETDEV_TX_OK; |
1da177e4 LT |
4162 | } |
4163 | ||
075897ce | 4164 | |
1da177e4 LT |
4165 | /* |
4166 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4167 | * the bond has a usable interface. | |
4168 | */ | |
4169 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4170 | { | |
454d7c9b | 4171 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4172 | int res = 1; |
4173 | ||
1da177e4 LT |
4174 | read_lock(&bond->lock); |
4175 | read_lock(&bond->curr_slave_lock); | |
4176 | ||
3d632c3f | 4177 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4178 | goto out; |
1da177e4 | 4179 | |
075897ce JL |
4180 | if (!bond->curr_active_slave) |
4181 | goto out; | |
4182 | ||
075897ce JL |
4183 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4184 | ||
1da177e4 | 4185 | out: |
3d632c3f | 4186 | if (res) |
1da177e4 LT |
4187 | /* no suitable interface, frame not sent */ |
4188 | dev_kfree_skb(skb); | |
3d632c3f | 4189 | |
1da177e4 LT |
4190 | read_unlock(&bond->curr_slave_lock); |
4191 | read_unlock(&bond->lock); | |
ec634fe3 | 4192 | return NETDEV_TX_OK; |
1da177e4 LT |
4193 | } |
4194 | ||
4195 | /* | |
169a3e66 JV |
4196 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4197 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4198 | * find the next active slave. | |
1da177e4 LT |
4199 | */ |
4200 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4201 | { | |
454d7c9b | 4202 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4203 | struct slave *slave, *start_at; |
4204 | int slave_no; | |
4205 | int i; | |
4206 | int res = 1; | |
4207 | ||
4208 | read_lock(&bond->lock); | |
4209 | ||
3d632c3f | 4210 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4211 | goto out; |
1da177e4 | 4212 | |
a361c83c | 4213 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4214 | |
4215 | bond_for_each_slave(bond, slave, i) { | |
4216 | slave_no--; | |
3d632c3f | 4217 | if (slave_no < 0) |
1da177e4 | 4218 | break; |
1da177e4 LT |
4219 | } |
4220 | ||
4221 | start_at = slave; | |
4222 | ||
4223 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4224 | if (IS_UP(slave->dev) && | |
4225 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4226 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4227 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
4228 | break; | |
4229 | } | |
4230 | } | |
4231 | ||
4232 | out: | |
4233 | if (res) { | |
4234 | /* no suitable interface, frame not sent */ | |
4235 | dev_kfree_skb(skb); | |
4236 | } | |
4237 | read_unlock(&bond->lock); | |
ec634fe3 | 4238 | return NETDEV_TX_OK; |
1da177e4 LT |
4239 | } |
4240 | ||
4241 | /* | |
4242 | * in broadcast mode, we send everything to all usable interfaces. | |
4243 | */ | |
4244 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4245 | { | |
454d7c9b | 4246 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4247 | struct slave *slave, *start_at; |
4248 | struct net_device *tx_dev = NULL; | |
4249 | int i; | |
4250 | int res = 1; | |
4251 | ||
4252 | read_lock(&bond->lock); | |
4253 | ||
3d632c3f | 4254 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4255 | goto out; |
1da177e4 LT |
4256 | |
4257 | read_lock(&bond->curr_slave_lock); | |
4258 | start_at = bond->curr_active_slave; | |
4259 | read_unlock(&bond->curr_slave_lock); | |
4260 | ||
3d632c3f | 4261 | if (!start_at) |
1da177e4 | 4262 | goto out; |
1da177e4 LT |
4263 | |
4264 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4265 | if (IS_UP(slave->dev) && | |
4266 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4267 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4268 | if (tx_dev) { |
4269 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4270 | if (!skb2) { | |
a4aee5c8 | 4271 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4272 | bond_dev->name); |
1da177e4 LT |
4273 | continue; |
4274 | } | |
4275 | ||
4276 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4277 | if (res) { | |
4278 | dev_kfree_skb(skb2); | |
4279 | continue; | |
4280 | } | |
4281 | } | |
4282 | tx_dev = slave->dev; | |
4283 | } | |
4284 | } | |
4285 | ||
3d632c3f | 4286 | if (tx_dev) |
1da177e4 | 4287 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4288 | |
4289 | out: | |
3d632c3f | 4290 | if (res) |
1da177e4 LT |
4291 | /* no suitable interface, frame not sent */ |
4292 | dev_kfree_skb(skb); | |
3d632c3f | 4293 | |
1da177e4 LT |
4294 | /* frame sent to all suitable interfaces */ |
4295 | read_unlock(&bond->lock); | |
ec634fe3 | 4296 | return NETDEV_TX_OK; |
1da177e4 LT |
4297 | } |
4298 | ||
4299 | /*------------------------- Device initialization ---------------------------*/ | |
4300 | ||
6f6652be JV |
4301 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4302 | { | |
4303 | switch (bond->params.xmit_policy) { | |
4304 | case BOND_XMIT_POLICY_LAYER23: | |
4305 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4306 | break; | |
4307 | case BOND_XMIT_POLICY_LAYER34: | |
4308 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4309 | break; | |
4310 | case BOND_XMIT_POLICY_LAYER2: | |
4311 | default: | |
4312 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4313 | break; | |
4314 | } | |
4315 | } | |
4316 | ||
bb1d9123 AG |
4317 | /* |
4318 | * Lookup the slave that corresponds to a qid | |
4319 | */ | |
4320 | static inline int bond_slave_override(struct bonding *bond, | |
4321 | struct sk_buff *skb) | |
4322 | { | |
4323 | int i, res = 1; | |
4324 | struct slave *slave = NULL; | |
4325 | struct slave *check_slave; | |
4326 | ||
4327 | read_lock(&bond->lock); | |
4328 | ||
4329 | if (!BOND_IS_OK(bond) || !skb->queue_mapping) | |
4330 | goto out; | |
4331 | ||
4332 | /* Find out if any slaves have the same mapping as this skb. */ | |
4333 | bond_for_each_slave(bond, check_slave, i) { | |
4334 | if (check_slave->queue_id == skb->queue_mapping) { | |
4335 | slave = check_slave; | |
4336 | break; | |
4337 | } | |
4338 | } | |
4339 | ||
4340 | /* If the slave isn't UP, use default transmit policy. */ | |
4341 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4342 | (slave->link == BOND_LINK_UP)) { | |
4343 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4344 | } | |
4345 | ||
4346 | out: | |
4347 | read_unlock(&bond->lock); | |
4348 | return res; | |
4349 | } | |
4350 | ||
4351 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) | |
4352 | { | |
4353 | /* | |
4354 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4355 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4356 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4357 | * way down to the bonding driver. | |
4358 | */ | |
fd0e435b PO |
4359 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4360 | ||
4361 | if (unlikely(txq >= dev->real_num_tx_queues)) { | |
d30ee670 | 4362 | do { |
fd0e435b | 4363 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4364 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4365 | } |
4366 | return txq; | |
bb1d9123 AG |
4367 | } |
4368 | ||
424efe9c | 4369 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4370 | { |
bb1d9123 AG |
4371 | struct bonding *bond = netdev_priv(dev); |
4372 | ||
e843fa50 NH |
4373 | /* |
4374 | * If we risk deadlock from transmitting this in the | |
4375 | * netpoll path, tell netpoll to queue the frame for later tx | |
4376 | */ | |
4377 | if (is_netpoll_tx_blocked(dev)) | |
4378 | return NETDEV_TX_BUSY; | |
4379 | ||
bb1d9123 AG |
4380 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { |
4381 | if (!bond_slave_override(bond, skb)) | |
4382 | return NETDEV_TX_OK; | |
4383 | } | |
00829823 SH |
4384 | |
4385 | switch (bond->params.mode) { | |
4386 | case BOND_MODE_ROUNDROBIN: | |
4387 | return bond_xmit_roundrobin(skb, dev); | |
4388 | case BOND_MODE_ACTIVEBACKUP: | |
4389 | return bond_xmit_activebackup(skb, dev); | |
4390 | case BOND_MODE_XOR: | |
4391 | return bond_xmit_xor(skb, dev); | |
4392 | case BOND_MODE_BROADCAST: | |
4393 | return bond_xmit_broadcast(skb, dev); | |
4394 | case BOND_MODE_8023AD: | |
4395 | return bond_3ad_xmit_xor(skb, dev); | |
4396 | case BOND_MODE_ALB: | |
4397 | case BOND_MODE_TLB: | |
4398 | return bond_alb_xmit(skb, dev); | |
4399 | default: | |
4400 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4401 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4402 | dev->name, bond->params.mode); | |
00829823 SH |
4403 | WARN_ON_ONCE(1); |
4404 | dev_kfree_skb(skb); | |
4405 | return NETDEV_TX_OK; | |
4406 | } | |
4407 | } | |
4408 | ||
4409 | ||
1da177e4 LT |
4410 | /* |
4411 | * set bond mode specific net device operations | |
4412 | */ | |
a77b5325 | 4413 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4414 | { |
169a3e66 JV |
4415 | struct net_device *bond_dev = bond->dev; |
4416 | ||
1da177e4 LT |
4417 | switch (mode) { |
4418 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4419 | break; |
4420 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4421 | break; |
4422 | case BOND_MODE_XOR: | |
6f6652be | 4423 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4424 | break; |
4425 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4426 | break; |
4427 | case BOND_MODE_8023AD: | |
6f6652be | 4428 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4429 | break; |
1da177e4 | 4430 | case BOND_MODE_ALB: |
8f903c70 JV |
4431 | /* FALLTHRU */ |
4432 | case BOND_MODE_TLB: | |
1da177e4 LT |
4433 | break; |
4434 | default: | |
4435 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4436 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4437 | bond_dev->name, mode); | |
1da177e4 LT |
4438 | break; |
4439 | } | |
4440 | } | |
4441 | ||
217df670 JV |
4442 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4443 | struct ethtool_drvinfo *drvinfo) | |
4444 | { | |
4445 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4446 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4447 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4448 | } | |
4449 | ||
7282d491 | 4450 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4451 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4452 | .get_link = ethtool_op_get_link, |
4453 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4454 | .get_sg = ethtool_op_get_sg, | |
4455 | .get_tso = ethtool_op_get_tso, | |
4456 | .get_ufo = ethtool_op_get_ufo, | |
4457 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4458 | }; |
4459 | ||
eb7cc59a | 4460 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4461 | .ndo_init = bond_init, |
9e71626c | 4462 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4463 | .ndo_open = bond_open, |
4464 | .ndo_stop = bond_close, | |
00829823 | 4465 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4466 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4467 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a SH |
4468 | .ndo_do_ioctl = bond_do_ioctl, |
4469 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4470 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4471 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4472 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4473 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4474 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4475 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
f6dc31a8 | 4476 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4477 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4478 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4479 | .ndo_poll_controller = bond_poll_controller, | |
4480 | #endif | |
9232ecca JP |
4481 | .ndo_add_slave = bond_enslave, |
4482 | .ndo_del_slave = bond_release, | |
eb7cc59a SH |
4483 | }; |
4484 | ||
9e2e61fb AW |
4485 | static void bond_destructor(struct net_device *bond_dev) |
4486 | { | |
4487 | struct bonding *bond = netdev_priv(bond_dev); | |
4488 | if (bond->wq) | |
4489 | destroy_workqueue(bond->wq); | |
4490 | free_netdev(bond_dev); | |
4491 | } | |
4492 | ||
181470fc | 4493 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4494 | { |
454d7c9b | 4495 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4496 | |
1da177e4 LT |
4497 | /* initialize rwlocks */ |
4498 | rwlock_init(&bond->lock); | |
4499 | rwlock_init(&bond->curr_slave_lock); | |
4500 | ||
d2991f75 | 4501 | bond->params = bonding_defaults; |
1da177e4 LT |
4502 | |
4503 | /* Initialize pointers */ | |
1da177e4 LT |
4504 | bond->dev = bond_dev; |
4505 | INIT_LIST_HEAD(&bond->vlan_list); | |
4506 | ||
4507 | /* Initialize the device entry points */ | |
181470fc | 4508 | ether_setup(bond_dev); |
eb7cc59a | 4509 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4510 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4511 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4512 | |
9e2e61fb | 4513 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4514 | |
4515 | /* Initialize the device options */ | |
4516 | bond_dev->tx_queue_len = 0; | |
4517 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4518 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4519 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4520 | ||
1da177e4 LT |
4521 | /* At first, we block adding VLANs. That's the only way to |
4522 | * prevent problems that occur when adding VLANs over an | |
4523 | * empty bond. The block will be removed once non-challenged | |
4524 | * slaves are enslaved. | |
4525 | */ | |
4526 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4527 | ||
932ff279 | 4528 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4529 | * transmitting */ |
4530 | bond_dev->features |= NETIF_F_LLTX; | |
4531 | ||
4532 | /* By default, we declare the bond to be fully | |
4533 | * VLAN hardware accelerated capable. Special | |
4534 | * care is taken in the various xmit functions | |
4535 | * when there are slaves that are not hw accel | |
4536 | * capable | |
4537 | */ | |
1da177e4 LT |
4538 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4539 | NETIF_F_HW_VLAN_RX | | |
4540 | NETIF_F_HW_VLAN_FILTER); | |
4541 | ||
e6599c2e ED |
4542 | /* By default, we enable GRO on bonding devices. |
4543 | * Actual support requires lowlevel drivers are GRO ready. | |
4544 | */ | |
4545 | bond_dev->features |= NETIF_F_GRO; | |
1da177e4 LT |
4546 | } |
4547 | ||
fdaea7a9 JV |
4548 | static void bond_work_cancel_all(struct bonding *bond) |
4549 | { | |
4550 | write_lock_bh(&bond->lock); | |
4551 | bond->kill_timers = 1; | |
4552 | write_unlock_bh(&bond->lock); | |
4553 | ||
4554 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4555 | cancel_delayed_work(&bond->mii_work); | |
4556 | ||
4557 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4558 | cancel_delayed_work(&bond->arp_work); | |
4559 | ||
4560 | if (bond->params.mode == BOND_MODE_ALB && | |
4561 | delayed_work_pending(&bond->alb_work)) | |
4562 | cancel_delayed_work(&bond->alb_work); | |
4563 | ||
4564 | if (bond->params.mode == BOND_MODE_8023AD && | |
4565 | delayed_work_pending(&bond->ad_work)) | |
4566 | cancel_delayed_work(&bond->ad_work); | |
5a37e8ca FL |
4567 | |
4568 | if (delayed_work_pending(&bond->mcast_work)) | |
4569 | cancel_delayed_work(&bond->mcast_work); | |
fdaea7a9 JV |
4570 | } |
4571 | ||
c67dfb29 EB |
4572 | /* |
4573 | * Destroy a bonding device. | |
4574 | * Must be under rtnl_lock when this function is called. | |
4575 | */ | |
4576 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4577 | { |
454d7c9b | 4578 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4579 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4580 | |
f6dc31a8 WC |
4581 | bond_netpoll_cleanup(bond_dev); |
4582 | ||
c67dfb29 EB |
4583 | /* Release the bonded slaves */ |
4584 | bond_release_all(bond_dev); | |
4585 | ||
a434e43f JV |
4586 | list_del(&bond->bond_list); |
4587 | ||
4588 | bond_work_cancel_all(bond); | |
4589 | ||
a434e43f | 4590 | bond_remove_proc_entry(bond); |
c67dfb29 | 4591 | |
f073c7ca TI |
4592 | bond_debug_unregister(bond); |
4593 | ||
22bedad3 | 4594 | __hw_addr_flush(&bond->mc_list); |
f35188fa JV |
4595 | |
4596 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { | |
4597 | list_del(&vlan->vlan_list); | |
4598 | kfree(vlan); | |
4599 | } | |
a434e43f JV |
4600 | } |
4601 | ||
1da177e4 LT |
4602 | /*------------------------- Module initialization ---------------------------*/ |
4603 | ||
4604 | /* | |
4605 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4606 | * number of the mode or its string name. A bit complicated because |
4607 | * some mode names are substrings of other names, and calls from sysfs | |
4608 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4609 | */ |
325dcf7a | 4610 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4611 | { |
54b87323 | 4612 | int modeint = -1, i, rv; |
a42e534f | 4613 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4614 | |
a42e534f JV |
4615 | for (p = (char *)buf; *p; p++) |
4616 | if (!(isdigit(*p) || isspace(*p))) | |
4617 | break; | |
4618 | ||
4619 | if (*p) | |
ece95f7f | 4620 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4621 | else |
54b87323 | 4622 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4623 | |
4624 | if (!rv) | |
4625 | return -1; | |
1da177e4 LT |
4626 | |
4627 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4628 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4629 | return tbl[i].mode; |
4630 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4631 | return tbl[i].mode; |
1da177e4 LT |
4632 | } |
4633 | ||
4634 | return -1; | |
4635 | } | |
4636 | ||
4637 | static int bond_check_params(struct bond_params *params) | |
4638 | { | |
a549952a | 4639 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4640 | |
1da177e4 LT |
4641 | /* |
4642 | * Convert string parameters. | |
4643 | */ | |
4644 | if (mode) { | |
4645 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4646 | if (bond_mode == -1) { | |
a4aee5c8 | 4647 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4648 | mode == NULL ? "NULL" : mode); |
4649 | return -EINVAL; | |
4650 | } | |
4651 | } | |
4652 | ||
169a3e66 JV |
4653 | if (xmit_hash_policy) { |
4654 | if ((bond_mode != BOND_MODE_XOR) && | |
4655 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4656 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4657 | bond_mode_name(bond_mode)); |
4658 | } else { | |
4659 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4660 | xmit_hashtype_tbl); | |
4661 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4662 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4663 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4664 | xmit_hash_policy); |
4665 | return -EINVAL; | |
4666 | } | |
4667 | } | |
4668 | } | |
4669 | ||
1da177e4 LT |
4670 | if (lacp_rate) { |
4671 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4672 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4673 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4674 | } else { |
4675 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4676 | if (lacp_fast == -1) { | |
a4aee5c8 | 4677 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4678 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4679 | return -EINVAL; | |
4680 | } | |
4681 | } | |
4682 | } | |
4683 | ||
fd989c83 JV |
4684 | if (ad_select) { |
4685 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4686 | if (params->ad_select == -1) { | |
a4aee5c8 | 4687 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4688 | ad_select == NULL ? "NULL" : ad_select); |
4689 | return -EINVAL; | |
4690 | } | |
4691 | ||
4692 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4693 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4694 | } |
4695 | } else { | |
4696 | params->ad_select = BOND_AD_STABLE; | |
4697 | } | |
4698 | ||
f5841306 | 4699 | if (max_bonds < 0) { |
a4aee5c8 JP |
4700 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4701 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4702 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4703 | } | |
4704 | ||
4705 | if (miimon < 0) { | |
a4aee5c8 JP |
4706 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4707 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4708 | miimon = BOND_LINK_MON_INTERV; |
4709 | } | |
4710 | ||
4711 | if (updelay < 0) { | |
a4aee5c8 JP |
4712 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4713 | updelay, INT_MAX); | |
1da177e4 LT |
4714 | updelay = 0; |
4715 | } | |
4716 | ||
4717 | if (downdelay < 0) { | |
a4aee5c8 JP |
4718 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4719 | downdelay, INT_MAX); | |
1da177e4 LT |
4720 | downdelay = 0; |
4721 | } | |
4722 | ||
4723 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4724 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4725 | use_carrier); | |
1da177e4 LT |
4726 | use_carrier = 1; |
4727 | } | |
4728 | ||
7893b249 | 4729 | if (num_grat_arp < 0 || num_grat_arp > 255) { |
2381a55c | 4730 | pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4731 | num_grat_arp); |
7893b249 MS |
4732 | num_grat_arp = 1; |
4733 | } | |
4734 | ||
305d552a | 4735 | if (num_unsol_na < 0 || num_unsol_na > 255) { |
2381a55c | 4736 | pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4737 | num_unsol_na); |
305d552a BH |
4738 | num_unsol_na = 1; |
4739 | } | |
4740 | ||
1da177e4 LT |
4741 | /* reset values for 802.3ad */ |
4742 | if (bond_mode == BOND_MODE_8023AD) { | |
4743 | if (!miimon) { | |
a4aee5c8 | 4744 | 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 | 4745 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4746 | miimon = 100; |
4747 | } | |
4748 | } | |
4749 | ||
bb1d9123 AG |
4750 | if (tx_queues < 1 || tx_queues > 255) { |
4751 | pr_warning("Warning: tx_queues (%d) should be between " | |
4752 | "1 and 255, resetting to %d\n", | |
4753 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4754 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4755 | } | |
4756 | ||
ebd8e497 AG |
4757 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4758 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4759 | "not of valid value (0/1), so it was set to " | |
4760 | "0\n", all_slaves_active); | |
4761 | all_slaves_active = 0; | |
4762 | } | |
4763 | ||
c2952c31 FL |
4764 | if (resend_igmp < 0 || resend_igmp > 255) { |
4765 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4766 | "0 and 255, resetting to %d\n", | |
4767 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4768 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4769 | } | |
4770 | ||
1da177e4 LT |
4771 | /* reset values for TLB/ALB */ |
4772 | if ((bond_mode == BOND_MODE_TLB) || | |
4773 | (bond_mode == BOND_MODE_ALB)) { | |
4774 | if (!miimon) { | |
a4aee5c8 | 4775 | 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 | 4776 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4777 | miimon = 100; |
4778 | } | |
4779 | } | |
4780 | ||
4781 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4782 | 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", |
4783 | updelay); | |
1da177e4 LT |
4784 | } |
4785 | ||
4786 | if (!miimon) { | |
4787 | if (updelay || downdelay) { | |
4788 | /* just warn the user the up/down delay will have | |
4789 | * no effect since miimon is zero... | |
4790 | */ | |
a4aee5c8 JP |
4791 | 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", |
4792 | updelay, downdelay); | |
1da177e4 LT |
4793 | } |
4794 | } else { | |
4795 | /* don't allow arp monitoring */ | |
4796 | if (arp_interval) { | |
a4aee5c8 JP |
4797 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4798 | miimon, arp_interval); | |
1da177e4 LT |
4799 | arp_interval = 0; |
4800 | } | |
4801 | ||
4802 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4803 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4804 | updelay, miimon, | |
4805 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4806 | } |
4807 | ||
4808 | updelay /= miimon; | |
4809 | ||
4810 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4811 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4812 | downdelay, miimon, | |
4813 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4814 | } |
4815 | ||
4816 | downdelay /= miimon; | |
4817 | } | |
4818 | ||
4819 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4820 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4821 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4822 | arp_interval = BOND_LINK_ARP_INTERV; |
4823 | } | |
4824 | ||
4825 | for (arp_ip_count = 0; | |
4826 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4827 | arp_ip_count++) { | |
4828 | /* not complete check, but should be good enough to | |
4829 | catch mistakes */ | |
4830 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4831 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4832 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4833 | arp_interval = 0; |
4834 | } else { | |
d3bb52b0 | 4835 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4836 | arp_target[arp_ip_count] = ip; |
4837 | } | |
4838 | } | |
4839 | ||
4840 | if (arp_interval && !arp_ip_count) { | |
4841 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4842 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4843 | arp_interval); | |
1da177e4 LT |
4844 | arp_interval = 0; |
4845 | } | |
4846 | ||
f5b2b966 JV |
4847 | if (arp_validate) { |
4848 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4849 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4850 | return -EINVAL; |
4851 | } | |
4852 | if (!arp_interval) { | |
a4aee5c8 | 4853 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4854 | return -EINVAL; |
4855 | } | |
4856 | ||
4857 | arp_validate_value = bond_parse_parm(arp_validate, | |
4858 | arp_validate_tbl); | |
4859 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4860 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4861 | arp_validate == NULL ? "NULL" : arp_validate); |
4862 | return -EINVAL; | |
4863 | } | |
4864 | } else | |
4865 | arp_validate_value = 0; | |
4866 | ||
1da177e4 | 4867 | if (miimon) { |
a4aee5c8 | 4868 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4869 | } else if (arp_interval) { |
4870 | int i; | |
4871 | ||
a4aee5c8 JP |
4872 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4873 | arp_interval, | |
4874 | arp_validate_tbl[arp_validate_value].modename, | |
4875 | arp_ip_count); | |
1da177e4 LT |
4876 | |
4877 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4878 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4879 | |
e5e2a8fd | 4880 | pr_info("\n"); |
1da177e4 | 4881 | |
b8a9787e | 4882 | } else if (max_bonds) { |
1da177e4 LT |
4883 | /* miimon and arp_interval not set, we need one so things |
4884 | * work as expected, see bonding.txt for details | |
4885 | */ | |
a4aee5c8 | 4886 | 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 |
4887 | } |
4888 | ||
4889 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4890 | /* currently, using a primary only makes sense | |
4891 | * in active backup, TLB or ALB modes | |
4892 | */ | |
a4aee5c8 JP |
4893 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4894 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4895 | primary = NULL; |
4896 | } | |
4897 | ||
a549952a JP |
4898 | if (primary && primary_reselect) { |
4899 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4900 | pri_reselect_tbl); | |
4901 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4902 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4903 | primary_reselect == |
4904 | NULL ? "NULL" : primary_reselect); | |
4905 | return -EINVAL; | |
4906 | } | |
4907 | } else { | |
4908 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4909 | } | |
4910 | ||
3915c1e8 JV |
4911 | if (fail_over_mac) { |
4912 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4913 | fail_over_mac_tbl); | |
4914 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4915 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4916 | arp_validate == NULL ? "NULL" : arp_validate); |
4917 | return -EINVAL; | |
4918 | } | |
4919 | ||
4920 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4921 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4922 | } else { |
4923 | fail_over_mac_value = BOND_FOM_NONE; | |
4924 | } | |
dd957c57 | 4925 | |
1da177e4 LT |
4926 | /* fill params struct with the proper values */ |
4927 | params->mode = bond_mode; | |
169a3e66 | 4928 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4929 | params->miimon = miimon; |
7893b249 | 4930 | params->num_grat_arp = num_grat_arp; |
305d552a | 4931 | params->num_unsol_na = num_unsol_na; |
1da177e4 | 4932 | params->arp_interval = arp_interval; |
f5b2b966 | 4933 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4934 | params->updelay = updelay; |
4935 | params->downdelay = downdelay; | |
4936 | params->use_carrier = use_carrier; | |
4937 | params->lacp_fast = lacp_fast; | |
4938 | params->primary[0] = 0; | |
a549952a | 4939 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4940 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4941 | params->tx_queues = tx_queues; |
ebd8e497 | 4942 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4943 | params->resend_igmp = resend_igmp; |
1da177e4 LT |
4944 | |
4945 | if (primary) { | |
4946 | strncpy(params->primary, primary, IFNAMSIZ); | |
4947 | params->primary[IFNAMSIZ - 1] = 0; | |
4948 | } | |
4949 | ||
4950 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4951 | ||
4952 | return 0; | |
4953 | } | |
4954 | ||
0daa2303 | 4955 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4956 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 4957 | |
e8a0464c DM |
4958 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4959 | struct netdev_queue *txq, | |
4960 | void *_unused) | |
c773e847 DM |
4961 | { |
4962 | lockdep_set_class(&txq->_xmit_lock, | |
4963 | &bonding_netdev_xmit_lock_key); | |
4964 | } | |
4965 | ||
4966 | static void bond_set_lockdep_class(struct net_device *dev) | |
4967 | { | |
cf508b12 DM |
4968 | lockdep_set_class(&dev->addr_list_lock, |
4969 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4970 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
4971 | } |
4972 | ||
181470fc SH |
4973 | /* |
4974 | * Called from registration process | |
4975 | */ | |
4976 | static int bond_init(struct net_device *bond_dev) | |
4977 | { | |
4978 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4979 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc SH |
4980 | |
4981 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4982 | ||
4983 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4984 | if (!bond->wq) | |
4985 | return -ENOMEM; | |
4986 | ||
4987 | bond_set_lockdep_class(bond_dev); | |
4988 | ||
181470fc | 4989 | bond_create_proc_entry(bond); |
ec87fd3b | 4990 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4991 | |
6151b3d4 | 4992 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4993 | |
f073c7ca TI |
4994 | bond_debug_register(bond); |
4995 | ||
22bedad3 | 4996 | __hw_addr_init(&bond->mc_list); |
181470fc SH |
4997 | return 0; |
4998 | } | |
4999 | ||
88ead977 EB |
5000 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
5001 | { | |
5002 | if (tb[IFLA_ADDRESS]) { | |
5003 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
5004 | return -EINVAL; | |
5005 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
5006 | return -EADDRNOTAVAIL; | |
5007 | } | |
5008 | return 0; | |
5009 | } | |
5010 | ||
5011 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
5012 | .kind = "bond", | |
6639104b | 5013 | .priv_size = sizeof(struct bonding), |
88ead977 EB |
5014 | .setup = bond_setup, |
5015 | .validate = bond_validate, | |
5016 | }; | |
5017 | ||
dfe60397 | 5018 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5019 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5020 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5021 | * set up our sysfs entries. | |
5022 | */ | |
ec87fd3b | 5023 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
5024 | { |
5025 | struct net_device *bond_dev; | |
5026 | int res; | |
5027 | ||
5028 | rtnl_lock(); | |
027ea041 | 5029 | |
bb1d9123 AG |
5030 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "", |
5031 | bond_setup, tx_queues); | |
dfe60397 | 5032 | if (!bond_dev) { |
a4aee5c8 | 5033 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
5034 | rtnl_unlock(); |
5035 | return -ENOMEM; | |
dfe60397 MW |
5036 | } |
5037 | ||
ec87fd3b | 5038 | dev_net_set(bond_dev, net); |
88ead977 EB |
5039 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5040 | ||
e4b91c48 JV |
5041 | if (!name) { |
5042 | res = dev_alloc_name(bond_dev, "bond%d"); | |
5043 | if (res < 0) | |
9e2e61fb | 5044 | goto out; |
27e6f065 NH |
5045 | } else { |
5046 | /* | |
5047 | * If we're given a name to register | |
5048 | * we need to ensure that its not already | |
5049 | * registered | |
5050 | */ | |
5051 | res = -EEXIST; | |
5052 | if (__dev_get_by_name(net, name) != NULL) | |
5053 | goto out; | |
e4b91c48 JV |
5054 | } |
5055 | ||
dfe60397 | 5056 | res = register_netdevice(bond_dev); |
0daa2303 | 5057 | |
e826eafa PO |
5058 | netif_carrier_off(bond_dev); |
5059 | ||
30c15ba9 | 5060 | out: |
7e083840 | 5061 | rtnl_unlock(); |
9e2e61fb AW |
5062 | if (res < 0) |
5063 | bond_destructor(bond_dev); | |
30c15ba9 | 5064 | return res; |
dfe60397 MW |
5065 | } |
5066 | ||
2c8c1e72 | 5067 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 5068 | { |
15449745 | 5069 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5070 | |
5071 | bn->net = net; | |
5072 | INIT_LIST_HEAD(&bn->dev_list); | |
5073 | ||
ec87fd3b | 5074 | bond_create_proc_dir(bn); |
15449745 EB |
5075 | |
5076 | return 0; | |
ec87fd3b EB |
5077 | } |
5078 | ||
2c8c1e72 | 5079 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 5080 | { |
15449745 | 5081 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5082 | |
5083 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
5084 | } |
5085 | ||
5086 | static struct pernet_operations bond_net_ops = { | |
5087 | .init = bond_net_init, | |
5088 | .exit = bond_net_exit, | |
15449745 EB |
5089 | .id = &bond_net_id, |
5090 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
5091 | }; |
5092 | ||
1da177e4 LT |
5093 | static int __init bonding_init(void) |
5094 | { | |
1da177e4 LT |
5095 | int i; |
5096 | int res; | |
5097 | ||
bd33acc3 | 5098 | pr_info("%s", bond_version); |
1da177e4 | 5099 | |
dfe60397 | 5100 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5101 | if (res) |
dfe60397 | 5102 | goto out; |
1da177e4 | 5103 | |
15449745 | 5104 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
5105 | if (res) |
5106 | goto out; | |
027ea041 | 5107 | |
88ead977 EB |
5108 | res = rtnl_link_register(&bond_link_ops); |
5109 | if (res) | |
6639104b | 5110 | goto err_link; |
88ead977 | 5111 | |
f073c7ca TI |
5112 | bond_create_debugfs(); |
5113 | ||
1da177e4 | 5114 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5115 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5116 | if (res) |
5117 | goto err; | |
1da177e4 LT |
5118 | } |
5119 | ||
b76cdba9 MW |
5120 | res = bond_create_sysfs(); |
5121 | if (res) | |
5122 | goto err; | |
5123 | ||
1da177e4 | 5124 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5125 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5126 | bond_register_ipv6_notifier(); |
dfe60397 | 5127 | out: |
1da177e4 | 5128 | return res; |
88ead977 EB |
5129 | err: |
5130 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5131 | err_link: |
15449745 | 5132 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5133 | goto out; |
dfe60397 | 5134 | |
1da177e4 LT |
5135 | } |
5136 | ||
5137 | static void __exit bonding_exit(void) | |
5138 | { | |
5139 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5140 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5141 | bond_unregister_ipv6_notifier(); |
1da177e4 | 5142 | |
ae68c398 | 5143 | bond_destroy_sysfs(); |
f073c7ca | 5144 | bond_destroy_debugfs(); |
ae68c398 | 5145 | |
88ead977 | 5146 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5147 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
5148 | |
5149 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
5150 | /* |
5151 | * Make sure we don't have an imbalance on our netpoll blocking | |
5152 | */ | |
5153 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 5154 | #endif |
1da177e4 LT |
5155 | } |
5156 | ||
5157 | module_init(bonding_init); | |
5158 | module_exit(bonding_exit); | |
5159 | MODULE_LICENSE("GPL"); | |
5160 | MODULE_VERSION(DRV_VERSION); | |
5161 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5162 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5163 | MODULE_ALIAS_RTNL_LINK("bond"); |