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; | |
1da177e4 | 634 | struct ethtool_cmd etool; |
61a44b9c | 635 | int res; |
1da177e4 LT |
636 | |
637 | /* Fake speed and duplex */ | |
638 | slave->speed = SPEED_100; | |
639 | slave->duplex = DUPLEX_FULL; | |
640 | ||
61a44b9c MW |
641 | if (!slave_dev->ethtool_ops || !slave_dev->ethtool_ops->get_settings) |
642 | return -1; | |
1da177e4 | 643 | |
61a44b9c MW |
644 | res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool); |
645 | if (res < 0) | |
1da177e4 | 646 | return -1; |
1da177e4 | 647 | |
1da177e4 LT |
648 | switch (etool.speed) { |
649 | case SPEED_10: | |
650 | case SPEED_100: | |
651 | case SPEED_1000: | |
94dbffd5 | 652 | case SPEED_10000: |
1da177e4 LT |
653 | break; |
654 | default: | |
655 | return -1; | |
656 | } | |
657 | ||
658 | switch (etool.duplex) { | |
659 | case DUPLEX_FULL: | |
660 | case DUPLEX_HALF: | |
661 | break; | |
662 | default: | |
663 | return -1; | |
664 | } | |
665 | ||
666 | slave->speed = etool.speed; | |
667 | slave->duplex = etool.duplex; | |
668 | ||
669 | return 0; | |
670 | } | |
671 | ||
672 | /* | |
673 | * if <dev> supports MII link status reporting, check its link status. | |
674 | * | |
675 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 676 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
677 | * |
678 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
679 | * can't tell and just pretend it is), or 0, meaning that the link is | |
680 | * down. | |
681 | * | |
682 | * If reporting is non-zero, instead of faking link up, return -1 if | |
683 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
684 | * support them). If use_carrier is set, return whatever it says. | |
685 | * It'd be nice if there was a good way to tell if a driver supports | |
686 | * netif_carrier, but there really isn't. | |
687 | */ | |
3d632c3f SH |
688 | static int bond_check_dev_link(struct bonding *bond, |
689 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 690 | { |
eb7cc59a | 691 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 692 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
693 | struct ifreq ifr; |
694 | struct mii_ioctl_data *mii; | |
1da177e4 | 695 | |
6c988853 PG |
696 | if (!reporting && !netif_running(slave_dev)) |
697 | return 0; | |
698 | ||
eb7cc59a | 699 | if (bond->params.use_carrier) |
1da177e4 | 700 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 701 | |
29112f4e JP |
702 | /* Try to get link status using Ethtool first. */ |
703 | if (slave_dev->ethtool_ops) { | |
704 | if (slave_dev->ethtool_ops->get_link) { | |
705 | u32 link; | |
706 | ||
707 | link = slave_dev->ethtool_ops->get_link(slave_dev); | |
708 | ||
709 | return link ? BMSR_LSTATUS : 0; | |
710 | } | |
711 | } | |
712 | ||
3d632c3f | 713 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 714 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
715 | if (ioctl) { |
716 | /* TODO: set pointer to correct ioctl on a per team member */ | |
717 | /* bases to make this more efficient. that is, once */ | |
718 | /* we determine the correct ioctl, we will always */ | |
719 | /* call it and not the others for that team */ | |
720 | /* member. */ | |
721 | ||
722 | /* | |
723 | * We cannot assume that SIOCGMIIPHY will also read a | |
724 | * register; not all network drivers (e.g., e100) | |
725 | * support that. | |
726 | */ | |
727 | ||
728 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
729 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
730 | mii = if_mii(&ifr); | |
731 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
732 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
733 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
734 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
735 | } |
736 | } | |
737 | ||
1da177e4 LT |
738 | /* |
739 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 740 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
741 | * cannot report link status). If not reporting, pretend |
742 | * we're ok. | |
743 | */ | |
3d632c3f | 744 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
745 | } |
746 | ||
747 | /*----------------------------- Multicast list ------------------------------*/ | |
748 | ||
1da177e4 LT |
749 | /* |
750 | * Push the promiscuity flag down to appropriate slaves | |
751 | */ | |
7e1a1ac1 | 752 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 753 | { |
7e1a1ac1 | 754 | int err = 0; |
1da177e4 LT |
755 | if (USES_PRIMARY(bond->params.mode)) { |
756 | /* write lock already acquired */ | |
757 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
758 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
759 | inc); | |
1da177e4 LT |
760 | } |
761 | } else { | |
762 | struct slave *slave; | |
763 | int i; | |
764 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
765 | err = dev_set_promiscuity(slave->dev, inc); |
766 | if (err) | |
767 | return err; | |
1da177e4 LT |
768 | } |
769 | } | |
7e1a1ac1 | 770 | return err; |
1da177e4 LT |
771 | } |
772 | ||
773 | /* | |
774 | * Push the allmulti flag down to all slaves | |
775 | */ | |
7e1a1ac1 | 776 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 777 | { |
7e1a1ac1 | 778 | int err = 0; |
1da177e4 LT |
779 | if (USES_PRIMARY(bond->params.mode)) { |
780 | /* write lock already acquired */ | |
781 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
782 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
783 | inc); | |
1da177e4 LT |
784 | } |
785 | } else { | |
786 | struct slave *slave; | |
787 | int i; | |
788 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
789 | err = dev_set_allmulti(slave->dev, inc); |
790 | if (err) | |
791 | return err; | |
1da177e4 LT |
792 | } |
793 | } | |
7e1a1ac1 | 794 | return err; |
1da177e4 LT |
795 | } |
796 | ||
797 | /* | |
798 | * Add a Multicast address to slaves | |
799 | * according to mode | |
800 | */ | |
22bedad3 | 801 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
802 | { |
803 | if (USES_PRIMARY(bond->params.mode)) { | |
804 | /* write lock already acquired */ | |
3d632c3f | 805 | if (bond->curr_active_slave) |
22bedad3 | 806 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
807 | } else { |
808 | struct slave *slave; | |
809 | int i; | |
3d632c3f SH |
810 | |
811 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 812 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
813 | } |
814 | } | |
815 | ||
816 | /* | |
817 | * Remove a multicast address from slave | |
818 | * according to mode | |
819 | */ | |
22bedad3 | 820 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
821 | { |
822 | if (USES_PRIMARY(bond->params.mode)) { | |
823 | /* write lock already acquired */ | |
3d632c3f | 824 | if (bond->curr_active_slave) |
22bedad3 | 825 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
826 | } else { |
827 | struct slave *slave; | |
828 | int i; | |
829 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 830 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
831 | } |
832 | } | |
833 | } | |
834 | ||
a816c7c7 | 835 | |
5a37e8ca | 836 | static void __bond_resend_igmp_join_requests(struct net_device *dev) |
a816c7c7 JV |
837 | { |
838 | struct in_device *in_dev; | |
a816c7c7 JV |
839 | |
840 | rcu_read_lock(); | |
5a37e8ca | 841 | in_dev = __in_dev_get_rcu(dev); |
866f3b25 | 842 | if (in_dev) |
24912420 | 843 | ip_mc_rejoin_groups(in_dev); |
a816c7c7 JV |
844 | rcu_read_unlock(); |
845 | } | |
846 | ||
5a37e8ca FL |
847 | /* |
848 | * Retrieve the list of registered multicast addresses for the bonding | |
849 | * device and retransmit an IGMP JOIN request to the current active | |
850 | * slave. | |
851 | */ | |
852 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
853 | { | |
854 | struct net_device *vlan_dev; | |
855 | struct vlan_entry *vlan; | |
856 | ||
857 | read_lock(&bond->lock); | |
858 | ||
859 | /* rejoin all groups on bond device */ | |
860 | __bond_resend_igmp_join_requests(bond->dev); | |
861 | ||
862 | /* rejoin all groups on vlan devices */ | |
863 | if (bond->vlgrp) { | |
864 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { | |
865 | vlan_dev = vlan_group_get_device(bond->vlgrp, | |
866 | vlan->vlan_id); | |
867 | if (vlan_dev) | |
868 | __bond_resend_igmp_join_requests(vlan_dev); | |
869 | } | |
870 | } | |
871 | ||
c2952c31 FL |
872 | if (--bond->igmp_retrans > 0) |
873 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); | |
874 | ||
5a37e8ca FL |
875 | read_unlock(&bond->lock); |
876 | } | |
877 | ||
379b7383 | 878 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
879 | { |
880 | struct bonding *bond = container_of(work, struct bonding, | |
881 | mcast_work.work); | |
882 | bond_resend_igmp_join_requests(bond); | |
883 | } | |
884 | ||
1da177e4 LT |
885 | /* |
886 | * flush all members of flush->mc_list from device dev->mc_list | |
887 | */ | |
3d632c3f SH |
888 | static void bond_mc_list_flush(struct net_device *bond_dev, |
889 | struct net_device *slave_dev) | |
1da177e4 | 890 | { |
454d7c9b | 891 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 892 | struct netdev_hw_addr *ha; |
1da177e4 | 893 | |
22bedad3 JP |
894 | netdev_for_each_mc_addr(ha, bond_dev) |
895 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
896 | |
897 | if (bond->params.mode == BOND_MODE_8023AD) { | |
898 | /* del lacpdu mc addr from mc list */ | |
899 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
900 | ||
22bedad3 | 901 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
902 | } |
903 | } | |
904 | ||
905 | /*--------------------------- Active slave change ---------------------------*/ | |
906 | ||
907 | /* | |
908 | * Update the mc list and multicast-related flags for the new and | |
909 | * old active slaves (if any) according to the multicast mode, and | |
910 | * promiscuous flags unconditionally. | |
911 | */ | |
3d632c3f SH |
912 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
913 | struct slave *old_active) | |
1da177e4 | 914 | { |
22bedad3 | 915 | struct netdev_hw_addr *ha; |
1da177e4 | 916 | |
3d632c3f | 917 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
918 | /* nothing to do - mc list is already up-to-date on |
919 | * all slaves | |
920 | */ | |
921 | return; | |
1da177e4 LT |
922 | |
923 | if (old_active) { | |
3d632c3f | 924 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 925 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 926 | |
3d632c3f | 927 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 928 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 929 | |
22bedad3 JP |
930 | netdev_for_each_mc_addr(ha, bond->dev) |
931 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
932 | } |
933 | ||
934 | if (new_active) { | |
7e1a1ac1 | 935 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 936 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 937 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 938 | |
3d632c3f | 939 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 940 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 941 | |
22bedad3 JP |
942 | netdev_for_each_mc_addr(ha, bond->dev) |
943 | dev_mc_add(new_active->dev, ha->addr); | |
1da177e4 LT |
944 | } |
945 | } | |
946 | ||
3915c1e8 JV |
947 | /* |
948 | * bond_do_fail_over_mac | |
949 | * | |
950 | * Perform special MAC address swapping for fail_over_mac settings | |
951 | * | |
952 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
953 | */ | |
954 | static void bond_do_fail_over_mac(struct bonding *bond, | |
955 | struct slave *new_active, | |
956 | struct slave *old_active) | |
1f78d9f9 HE |
957 | __releases(&bond->curr_slave_lock) |
958 | __releases(&bond->lock) | |
959 | __acquires(&bond->lock) | |
960 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
961 | { |
962 | u8 tmp_mac[ETH_ALEN]; | |
963 | struct sockaddr saddr; | |
964 | int rv; | |
965 | ||
966 | switch (bond->params.fail_over_mac) { | |
967 | case BOND_FOM_ACTIVE: | |
968 | if (new_active) | |
969 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, | |
970 | new_active->dev->addr_len); | |
971 | break; | |
972 | case BOND_FOM_FOLLOW: | |
973 | /* | |
974 | * if new_active && old_active, swap them | |
975 | * if just old_active, do nothing (going to no active slave) | |
976 | * if just new_active, set new_active to bond's MAC | |
977 | */ | |
978 | if (!new_active) | |
979 | return; | |
980 | ||
981 | write_unlock_bh(&bond->curr_slave_lock); | |
982 | read_unlock(&bond->lock); | |
983 | ||
984 | if (old_active) { | |
985 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
986 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
987 | ETH_ALEN); | |
988 | saddr.sa_family = new_active->dev->type; | |
989 | } else { | |
990 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
991 | saddr.sa_family = bond->dev->type; | |
992 | } | |
993 | ||
994 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
995 | if (rv) { | |
a4aee5c8 | 996 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
997 | bond->dev->name, -rv, new_active->dev->name); |
998 | goto out; | |
999 | } | |
1000 | ||
1001 | if (!old_active) | |
1002 | goto out; | |
1003 | ||
1004 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
1005 | saddr.sa_family = old_active->dev->type; | |
1006 | ||
1007 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
1008 | if (rv) | |
a4aee5c8 | 1009 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
1010 | bond->dev->name, -rv, new_active->dev->name); |
1011 | out: | |
1012 | read_lock(&bond->lock); | |
1013 | write_lock_bh(&bond->curr_slave_lock); | |
1014 | break; | |
1015 | default: | |
a4aee5c8 | 1016 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
1017 | bond->dev->name, bond->params.fail_over_mac); |
1018 | break; | |
1019 | } | |
1020 | ||
1021 | } | |
1022 | ||
a549952a JP |
1023 | static bool bond_should_change_active(struct bonding *bond) |
1024 | { | |
1025 | struct slave *prim = bond->primary_slave; | |
1026 | struct slave *curr = bond->curr_active_slave; | |
1027 | ||
1028 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
1029 | return true; | |
1030 | if (bond->force_primary) { | |
1031 | bond->force_primary = false; | |
1032 | return true; | |
1033 | } | |
1034 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
1035 | (prim->speed < curr->speed || | |
1036 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
1037 | return false; | |
1038 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
1039 | return false; | |
1040 | return true; | |
1041 | } | |
3915c1e8 | 1042 | |
1da177e4 LT |
1043 | /** |
1044 | * find_best_interface - select the best available slave to be the active one | |
1045 | * @bond: our bonding struct | |
1046 | * | |
1047 | * Warning: Caller must hold curr_slave_lock for writing. | |
1048 | */ | |
1049 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
1050 | { | |
1051 | struct slave *new_active, *old_active; | |
1052 | struct slave *bestslave = NULL; | |
1053 | int mintime = bond->params.updelay; | |
1054 | int i; | |
1055 | ||
49b4ad92 | 1056 | new_active = bond->curr_active_slave; |
1da177e4 LT |
1057 | |
1058 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 1059 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 1060 | new_active = bond->first_slave; |
3d632c3f | 1061 | else |
1da177e4 | 1062 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
1063 | } |
1064 | ||
1da177e4 | 1065 | if ((bond->primary_slave) && |
a549952a JP |
1066 | bond->primary_slave->link == BOND_LINK_UP && |
1067 | bond_should_change_active(bond)) { | |
1da177e4 LT |
1068 | new_active = bond->primary_slave; |
1069 | } | |
1070 | ||
1071 | /* remember where to stop iterating over the slaves */ | |
1072 | old_active = new_active; | |
1073 | ||
1074 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1075 | if (new_active->link == BOND_LINK_UP) { |
1076 | return new_active; | |
1077 | } else if (new_active->link == BOND_LINK_BACK && | |
1078 | IS_UP(new_active->dev)) { | |
1079 | /* link up, but waiting for stabilization */ | |
1080 | if (new_active->delay < mintime) { | |
1081 | mintime = new_active->delay; | |
1082 | bestslave = new_active; | |
1da177e4 LT |
1083 | } |
1084 | } | |
1085 | } | |
1086 | ||
1087 | return bestslave; | |
1088 | } | |
1089 | ||
1090 | /** | |
1091 | * change_active_interface - change the active slave into the specified one | |
1092 | * @bond: our bonding struct | |
1093 | * @new: the new slave to make the active one | |
1094 | * | |
1095 | * Set the new slave to the bond's settings and unset them on the old | |
1096 | * curr_active_slave. | |
1097 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1098 | * | |
1099 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1100 | * because it is apparently the best available slave we have, even though its | |
1101 | * updelay hasn't timed out yet. | |
1102 | * | |
3915c1e8 JV |
1103 | * If new_active is not NULL, caller must hold bond->lock for read and |
1104 | * curr_slave_lock for write_bh. | |
1da177e4 | 1105 | */ |
a77b5325 | 1106 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1107 | { |
1108 | struct slave *old_active = bond->curr_active_slave; | |
1109 | ||
3d632c3f | 1110 | if (old_active == new_active) |
1da177e4 | 1111 | return; |
1da177e4 LT |
1112 | |
1113 | if (new_active) { | |
b2220cad JV |
1114 | new_active->jiffies = jiffies; |
1115 | ||
1da177e4 LT |
1116 | if (new_active->link == BOND_LINK_BACK) { |
1117 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1118 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1119 | bond->dev->name, new_active->dev->name, | |
1120 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1121 | } |
1122 | ||
1123 | new_active->delay = 0; | |
1124 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1125 | |
3d632c3f | 1126 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1127 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1128 | |
58402054 | 1129 | if (bond_is_lb(bond)) |
1da177e4 | 1130 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1131 | } else { |
1132 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1133 | pr_info("%s: making interface %s the new active one.\n", |
1134 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1135 | } |
1136 | } | |
1137 | } | |
1138 | ||
3d632c3f | 1139 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1140 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1141 | |
58402054 | 1142 | if (bond_is_lb(bond)) { |
1da177e4 | 1143 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1144 | if (old_active) |
1145 | bond_set_slave_inactive_flags(old_active); | |
1146 | if (new_active) | |
1147 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1148 | } else { |
1149 | bond->curr_active_slave = new_active; | |
1150 | } | |
c3ade5ca JV |
1151 | |
1152 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1153 | if (old_active) |
c3ade5ca | 1154 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1155 | |
1156 | if (new_active) { | |
1157 | bond_set_slave_active_flags(new_active); | |
2ab82852 | 1158 | |
709f8a45 OG |
1159 | if (bond->params.fail_over_mac) |
1160 | bond_do_fail_over_mac(bond, new_active, | |
1161 | old_active); | |
3915c1e8 | 1162 | |
ffa95ed5 BH |
1163 | if (netif_running(bond->dev)) { |
1164 | bond->send_grat_arp = bond->params.num_grat_arp; | |
1165 | bond_send_gratuitous_arp(bond); | |
01f3109d | 1166 | |
ffa95ed5 BH |
1167 | bond->send_unsol_na = bond->params.num_unsol_na; |
1168 | bond_send_unsolicited_na(bond); | |
1169 | } | |
305d552a | 1170 | |
01f3109d OG |
1171 | write_unlock_bh(&bond->curr_slave_lock); |
1172 | read_unlock(&bond->lock); | |
1173 | ||
75c78500 | 1174 | netdev_bonding_change(bond->dev, NETDEV_BONDING_FAILOVER); |
01f3109d OG |
1175 | |
1176 | read_lock(&bond->lock); | |
1177 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1178 | } |
c3ade5ca | 1179 | } |
a2fd940f | 1180 | |
5a37e8ca FL |
1181 | /* resend IGMP joins since active slave has changed or |
1182 | * all were sent on curr_active_slave */ | |
ffa95ed5 BH |
1183 | if (((USES_PRIMARY(bond->params.mode) && new_active) || |
1184 | bond->params.mode == BOND_MODE_ROUNDROBIN) && | |
1185 | netif_running(bond->dev)) { | |
c2952c31 | 1186 | bond->igmp_retrans = bond->params.resend_igmp; |
5a37e8ca | 1187 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); |
a2fd940f | 1188 | } |
1da177e4 LT |
1189 | } |
1190 | ||
1191 | /** | |
1192 | * bond_select_active_slave - select a new active slave, if needed | |
1193 | * @bond: our bonding struct | |
1194 | * | |
3d632c3f | 1195 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1196 | * - The old curr_active_slave has been released or lost its link. |
1197 | * - The primary_slave has got its link back. | |
1198 | * - A slave has got its link back and there's no old curr_active_slave. | |
1199 | * | |
3915c1e8 | 1200 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1201 | */ |
a77b5325 | 1202 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1203 | { |
1204 | struct slave *best_slave; | |
ff59c456 | 1205 | int rv; |
1da177e4 LT |
1206 | |
1207 | best_slave = bond_find_best_slave(bond); | |
1208 | if (best_slave != bond->curr_active_slave) { | |
1209 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1210 | rv = bond_set_carrier(bond); |
1211 | if (!rv) | |
1212 | return; | |
1213 | ||
1214 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1215 | pr_info("%s: first active interface up!\n", |
1216 | bond->dev->name); | |
ff59c456 | 1217 | } else { |
a4aee5c8 JP |
1218 | pr_info("%s: now running without any active interface !\n", |
1219 | bond->dev->name); | |
ff59c456 | 1220 | } |
1da177e4 LT |
1221 | } |
1222 | } | |
1223 | ||
1224 | /*--------------------------- slave list handling ---------------------------*/ | |
1225 | ||
1226 | /* | |
1227 | * This function attaches the slave to the end of list. | |
1228 | * | |
1229 | * bond->lock held for writing by caller. | |
1230 | */ | |
1231 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1232 | { | |
1233 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1234 | new_slave->next = new_slave; | |
1235 | new_slave->prev = new_slave; | |
1236 | bond->first_slave = new_slave; | |
1237 | } else { | |
1238 | new_slave->next = bond->first_slave; | |
1239 | new_slave->prev = bond->first_slave->prev; | |
1240 | new_slave->next->prev = new_slave; | |
1241 | new_slave->prev->next = new_slave; | |
1242 | } | |
1243 | ||
1244 | bond->slave_cnt++; | |
1245 | } | |
1246 | ||
1247 | /* | |
1248 | * This function detaches the slave from the list. | |
1249 | * WARNING: no check is made to verify if the slave effectively | |
1250 | * belongs to <bond>. | |
1251 | * Nothing is freed on return, structures are just unchained. | |
1252 | * If any slave pointer in bond was pointing to <slave>, | |
1253 | * it should be changed by the calling function. | |
1254 | * | |
1255 | * bond->lock held for writing by caller. | |
1256 | */ | |
1257 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1258 | { | |
3d632c3f | 1259 | if (slave->next) |
1da177e4 | 1260 | slave->next->prev = slave->prev; |
1da177e4 | 1261 | |
3d632c3f | 1262 | if (slave->prev) |
1da177e4 | 1263 | slave->prev->next = slave->next; |
1da177e4 LT |
1264 | |
1265 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1266 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1267 | bond->first_slave = slave->next; | |
1268 | } else { | |
1269 | bond->first_slave = NULL; /* slave was the last one */ | |
1270 | } | |
1271 | } | |
1272 | ||
1273 | slave->next = NULL; | |
1274 | slave->prev = NULL; | |
1275 | bond->slave_cnt--; | |
1276 | } | |
1277 | ||
f6dc31a8 | 1278 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1279 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1280 | { |
8a8efa22 AW |
1281 | struct netpoll *np; |
1282 | int err = 0; | |
f6dc31a8 | 1283 | |
8a8efa22 AW |
1284 | np = kzalloc(sizeof(*np), GFP_KERNEL); |
1285 | err = -ENOMEM; | |
1286 | if (!np) | |
1287 | goto out; | |
1288 | ||
1289 | np->dev = slave->dev; | |
1290 | err = __netpoll_setup(np); | |
1291 | if (err) { | |
1292 | kfree(np); | |
1293 | goto out; | |
f6dc31a8 | 1294 | } |
8a8efa22 AW |
1295 | slave->np = np; |
1296 | out: | |
1297 | return err; | |
1298 | } | |
1299 | static inline void slave_disable_netpoll(struct slave *slave) | |
1300 | { | |
1301 | struct netpoll *np = slave->np; | |
1302 | ||
1303 | if (!np) | |
1304 | return; | |
1305 | ||
1306 | slave->np = NULL; | |
1307 | synchronize_rcu_bh(); | |
1308 | __netpoll_cleanup(np); | |
1309 | kfree(np); | |
1310 | } | |
1311 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1312 | { | |
1313 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1314 | return false; | |
1315 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1316 | return false; | |
1317 | return true; | |
f6dc31a8 WC |
1318 | } |
1319 | ||
1320 | static void bond_poll_controller(struct net_device *bond_dev) | |
1321 | { | |
8a8efa22 AW |
1322 | } |
1323 | ||
1324 | static void __bond_netpoll_cleanup(struct bonding *bond) | |
1325 | { | |
c2355e1a NH |
1326 | struct slave *slave; |
1327 | int i; | |
1328 | ||
8a8efa22 AW |
1329 | bond_for_each_slave(bond, slave, i) |
1330 | if (IS_UP(slave->dev)) | |
1331 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1332 | } |
f6dc31a8 WC |
1333 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1334 | { | |
1335 | struct bonding *bond = netdev_priv(bond_dev); | |
8a8efa22 AW |
1336 | |
1337 | read_lock(&bond->lock); | |
1338 | __bond_netpoll_cleanup(bond); | |
1339 | read_unlock(&bond->lock); | |
1340 | } | |
1341 | ||
1342 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni) | |
1343 | { | |
1344 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1345 | struct slave *slave; |
8a8efa22 | 1346 | int i, err = 0; |
f6dc31a8 WC |
1347 | |
1348 | read_lock(&bond->lock); | |
f6dc31a8 | 1349 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1350 | err = slave_enable_netpoll(slave); |
1351 | if (err) { | |
1352 | __bond_netpoll_cleanup(bond); | |
1353 | break; | |
f6dc31a8 WC |
1354 | } |
1355 | } | |
1356 | read_unlock(&bond->lock); | |
8a8efa22 | 1357 | return err; |
f6dc31a8 WC |
1358 | } |
1359 | ||
8a8efa22 AW |
1360 | static struct netpoll_info *bond_netpoll_info(struct bonding *bond) |
1361 | { | |
1362 | return bond->dev->npinfo; | |
1363 | } | |
f6dc31a8 | 1364 | |
8a8efa22 AW |
1365 | #else |
1366 | static inline int slave_enable_netpoll(struct slave *slave) | |
1367 | { | |
1368 | return 0; | |
1369 | } | |
1370 | static inline void slave_disable_netpoll(struct slave *slave) | |
1371 | { | |
1372 | } | |
f6dc31a8 WC |
1373 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1374 | { | |
1375 | } | |
f6dc31a8 WC |
1376 | #endif |
1377 | ||
1da177e4 LT |
1378 | /*---------------------------------- IOCTL ----------------------------------*/ |
1379 | ||
4ad072c9 AB |
1380 | static int bond_sethwaddr(struct net_device *bond_dev, |
1381 | struct net_device *slave_dev) | |
1da177e4 | 1382 | { |
5a03cdb7 HE |
1383 | pr_debug("bond_dev=%p\n", bond_dev); |
1384 | pr_debug("slave_dev=%p\n", slave_dev); | |
1385 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 LT |
1386 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
1387 | return 0; | |
1388 | } | |
1389 | ||
7f353bf2 HX |
1390 | #define BOND_VLAN_FEATURES \ |
1391 | (NETIF_F_VLAN_CHALLENGED | NETIF_F_HW_VLAN_RX | NETIF_F_HW_VLAN_TX | \ | |
1392 | NETIF_F_HW_VLAN_FILTER) | |
8531c5ff | 1393 | |
3d632c3f | 1394 | /* |
8e3babcd | 1395 | * Compute the common dev->feature set available to all slaves. Some |
7f353bf2 HX |
1396 | * feature bits are managed elsewhere, so preserve those feature bits |
1397 | * on the master device. | |
8531c5ff AK |
1398 | */ |
1399 | static int bond_compute_features(struct bonding *bond) | |
1400 | { | |
8531c5ff AK |
1401 | struct slave *slave; |
1402 | struct net_device *bond_dev = bond->dev; | |
04ed3e74 MM |
1403 | u32 features = bond_dev->features; |
1404 | u32 vlan_features = 0; | |
3158bf7d MS |
1405 | unsigned short max_hard_header_len = max((u16)ETH_HLEN, |
1406 | bond_dev->hard_header_len); | |
8e3babcd | 1407 | int i; |
8531c5ff | 1408 | |
7f353bf2 | 1409 | features &= ~(NETIF_F_ALL_CSUM | BOND_VLAN_FEATURES); |
c6e1a0d1 | 1410 | features |= NETIF_F_GSO_MASK | NETIF_F_NO_CSUM | NETIF_F_NOCACHE_COPY; |
b63365a2 HX |
1411 | |
1412 | if (!bond->first_slave) | |
1413 | goto done; | |
1414 | ||
1415 | features &= ~NETIF_F_ONE_FOR_ALL; | |
7f353bf2 | 1416 | |
278339a4 | 1417 | vlan_features = bond->first_slave->dev->vlan_features; |
54ef3137 | 1418 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1419 | features = netdev_increment_features(features, |
1420 | slave->dev->features, | |
1421 | NETIF_F_ONE_FOR_ALL); | |
278339a4 JV |
1422 | vlan_features = netdev_increment_features(vlan_features, |
1423 | slave->dev->vlan_features, | |
1424 | NETIF_F_ONE_FOR_ALL); | |
54ef3137 JV |
1425 | if (slave->dev->hard_header_len > max_hard_header_len) |
1426 | max_hard_header_len = slave->dev->hard_header_len; | |
1427 | } | |
8531c5ff | 1428 | |
b63365a2 | 1429 | done: |
7f353bf2 | 1430 | features |= (bond_dev->features & BOND_VLAN_FEATURES); |
acd1130e MM |
1431 | bond_dev->features = netdev_fix_features(bond_dev, features); |
1432 | bond_dev->vlan_features = netdev_fix_features(bond_dev, vlan_features); | |
54ef3137 | 1433 | bond_dev->hard_header_len = max_hard_header_len; |
8531c5ff AK |
1434 | |
1435 | return 0; | |
1436 | } | |
1437 | ||
872254dd MS |
1438 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1439 | struct net_device *slave_dev) | |
1440 | { | |
454d7c9b | 1441 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1442 | |
00829823 | 1443 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1444 | |
1445 | bond_dev->type = slave_dev->type; | |
1446 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1447 | bond_dev->addr_len = slave_dev->addr_len; | |
1448 | ||
1449 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1450 | slave_dev->addr_len); | |
d90a162a | 1451 | bond->setup_by_slave = 1; |
872254dd MS |
1452 | } |
1453 | ||
5b2c4dd2 JP |
1454 | /* On bonding slaves other than the currently active slave, suppress |
1455 | * duplicates except for 802.3ad ETH_P_SLOW, alb non-mcast/bcast, and | |
1456 | * ARP on active-backup slaves with arp_validate enabled. | |
1457 | */ | |
1458 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1459 | struct slave *slave, |
1460 | struct bonding *bond) | |
5b2c4dd2 | 1461 | { |
2d7011ca | 1462 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1463 | if (slave_do_arp_validate(bond, slave) && |
5b2c4dd2 JP |
1464 | skb->protocol == __cpu_to_be16(ETH_P_ARP)) |
1465 | return false; | |
1466 | ||
0bd80dad | 1467 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1468 | skb->pkt_type != PACKET_BROADCAST && |
1469 | skb->pkt_type != PACKET_MULTICAST) | |
1470 | return false; | |
1471 | ||
0bd80dad | 1472 | if (bond->params.mode == BOND_MODE_8023AD && |
5b2c4dd2 JP |
1473 | skb->protocol == __cpu_to_be16(ETH_P_SLOW)) |
1474 | return false; | |
1475 | ||
1476 | return true; | |
1477 | } | |
1478 | return false; | |
1479 | } | |
1480 | ||
8a4eb573 | 1481 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1482 | { |
8a4eb573 | 1483 | struct sk_buff *skb = *pskb; |
f1c1775a | 1484 | struct slave *slave; |
0bd80dad | 1485 | struct bonding *bond; |
5b2c4dd2 | 1486 | |
8a4eb573 JP |
1487 | skb = skb_share_check(skb, GFP_ATOMIC); |
1488 | if (unlikely(!skb)) | |
1489 | return RX_HANDLER_CONSUMED; | |
1490 | ||
1491 | *pskb = skb; | |
5b2c4dd2 | 1492 | |
35d48903 JP |
1493 | slave = bond_slave_get_rcu(skb->dev); |
1494 | bond = slave->bond; | |
0bd80dad JP |
1495 | |
1496 | if (bond->params.arp_interval) | |
f1c1775a | 1497 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1498 | |
0bd80dad | 1499 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1500 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1501 | } |
1502 | ||
35d48903 | 1503 | skb->dev = bond->dev; |
5b2c4dd2 | 1504 | |
0bd80dad | 1505 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1506 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1507 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1508 | |
541ac7c9 CG |
1509 | if (unlikely(skb_cow_head(skb, |
1510 | skb->data - skb_mac_header(skb)))) { | |
1511 | kfree_skb(skb); | |
8a4eb573 | 1512 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1513 | } |
35d48903 | 1514 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1515 | } |
1516 | ||
8a4eb573 | 1517 | return RX_HANDLER_ANOTHER; |
5b2c4dd2 JP |
1518 | } |
1519 | ||
1da177e4 | 1520 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1521 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1522 | { |
454d7c9b | 1523 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1524 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1525 | struct slave *new_slave = NULL; |
22bedad3 | 1526 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1527 | struct sockaddr addr; |
1528 | int link_reporting; | |
1529 | int old_features = bond_dev->features; | |
1530 | int res = 0; | |
1531 | ||
552709d5 | 1532 | if (!bond->params.use_carrier && slave_dev->ethtool_ops == NULL && |
eb7cc59a | 1533 | slave_ops->ndo_do_ioctl == NULL) { |
a4aee5c8 JP |
1534 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1535 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1536 | } |
1537 | ||
1538 | /* bond must be initialized by bond_open() before enslaving */ | |
1539 | if (!(bond_dev->flags & IFF_UP)) { | |
a4aee5c8 JP |
1540 | pr_warning("%s: master_dev is not up in bond_enslave\n", |
1541 | bond_dev->name); | |
1da177e4 LT |
1542 | } |
1543 | ||
1544 | /* already enslaved */ | |
1545 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1546 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1547 | return -EBUSY; |
1548 | } | |
1549 | ||
1550 | /* vlan challenged mutual exclusion */ | |
1551 | /* no need to lock since we're protected by rtnl_lock */ | |
1552 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1553 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
f35188fa | 1554 | if (bond->vlgrp) { |
a4aee5c8 JP |
1555 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1556 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1557 | return -EPERM; |
1558 | } else { | |
a4aee5c8 JP |
1559 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1560 | bond_dev->name, slave_dev->name, | |
1561 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1562 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
1563 | } | |
1564 | } else { | |
5a03cdb7 | 1565 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1566 | if (bond->slave_cnt == 0) { |
1567 | /* First slave, and it is not VLAN challenged, | |
1568 | * so remove the block of adding VLANs over the bond. | |
1569 | */ | |
1570 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; | |
1571 | } | |
1572 | } | |
1573 | ||
217df670 JV |
1574 | /* |
1575 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1576 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1577 | * the current ifenslave will set the interface down prior to |
1578 | * enslaving it; the old ifenslave will not. | |
1579 | */ | |
1580 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1581 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1582 | slave_dev->name); |
1583 | res = -EPERM; | |
1584 | goto err_undo_flags; | |
1585 | } | |
1da177e4 | 1586 | |
872254dd MS |
1587 | /* set bonding device ether type by slave - bonding netdevices are |
1588 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1589 | * there is a need to override some of the type dependent attribs/funcs. | |
1590 | * | |
1591 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1592 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1593 | */ | |
1594 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1595 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1596 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1597 | bond_dev->name, |
1598 | bond_dev->type, slave_dev->type); | |
75c78500 | 1599 | |
3ca5b404 JP |
1600 | res = netdev_bonding_change(bond_dev, |
1601 | NETDEV_PRE_TYPE_CHANGE); | |
1602 | res = notifier_to_errno(res); | |
1603 | if (res) { | |
1604 | pr_err("%s: refused to change device type\n", | |
1605 | bond_dev->name); | |
1606 | res = -EBUSY; | |
1607 | goto err_undo_flags; | |
1608 | } | |
75c78500 | 1609 | |
32a806c1 | 1610 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1611 | dev_uc_flush(bond_dev); |
22bedad3 | 1612 | dev_mc_flush(bond_dev); |
32a806c1 | 1613 | |
e36b9d16 MS |
1614 | if (slave_dev->type != ARPHRD_ETHER) |
1615 | bond_setup_by_slave(bond_dev, slave_dev); | |
1616 | else | |
1617 | ether_setup(bond_dev); | |
75c78500 | 1618 | |
93d9b7d7 JP |
1619 | netdev_bonding_change(bond_dev, |
1620 | NETDEV_POST_TYPE_CHANGE); | |
e36b9d16 | 1621 | } |
872254dd | 1622 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1623 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1624 | slave_dev->name, | |
1625 | slave_dev->type, bond_dev->type); | |
1626 | res = -EINVAL; | |
1627 | goto err_undo_flags; | |
872254dd MS |
1628 | } |
1629 | ||
eb7cc59a | 1630 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1631 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1632 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1633 | bond_dev->name); | |
3915c1e8 JV |
1634 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1635 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1636 | 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", |
1637 | bond_dev->name); | |
2ab82852 MS |
1638 | res = -EOPNOTSUPP; |
1639 | goto err_undo_flags; | |
1640 | } | |
1da177e4 LT |
1641 | } |
1642 | ||
c20811a7 JP |
1643 | /* If this is the first slave, then we need to set the master's hardware |
1644 | * address to be the same as the slave's. */ | |
d13a2cb6 | 1645 | if (is_zero_ether_addr(bond->dev->dev_addr)) |
c20811a7 JP |
1646 | memcpy(bond->dev->dev_addr, slave_dev->dev_addr, |
1647 | slave_dev->addr_len); | |
1648 | ||
1649 | ||
243cb4e5 | 1650 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1651 | if (!new_slave) { |
1652 | res = -ENOMEM; | |
1653 | goto err_undo_flags; | |
1654 | } | |
1655 | ||
bb1d9123 AG |
1656 | /* |
1657 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1658 | * is set via sysfs or module option if desired. | |
1659 | */ | |
1660 | new_slave->queue_id = 0; | |
1661 | ||
b15ba0fb JP |
1662 | /* Save slave's original mtu and then set it to match the bond */ |
1663 | new_slave->original_mtu = slave_dev->mtu; | |
1664 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1665 | if (res) { | |
1666 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1667 | goto err_free; | |
1668 | } | |
1669 | ||
217df670 JV |
1670 | /* |
1671 | * Save slave's original ("permanent") mac address for modes | |
1672 | * that need it, and for restoring it upon release, and then | |
1673 | * set it to the master's address | |
1674 | */ | |
1675 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1676 | |
dd957c57 | 1677 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1678 | /* |
1679 | * Set slave to master's mac address. The application already | |
1680 | * set the master's mac address to that of the first slave | |
1681 | */ | |
1682 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1683 | addr.sa_family = slave_dev->type; | |
1684 | res = dev_set_mac_address(slave_dev, &addr); | |
1685 | if (res) { | |
5a03cdb7 | 1686 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1687 | goto err_restore_mtu; |
2ab82852 | 1688 | } |
217df670 | 1689 | } |
1da177e4 | 1690 | |
1765a575 | 1691 | res = netdev_set_bond_master(slave_dev, bond_dev); |
c2edacf8 | 1692 | if (res) { |
1765a575 | 1693 | pr_debug("Error %d calling netdev_set_bond_master\n", res); |
569f0c4d | 1694 | goto err_restore_mac; |
c2edacf8 | 1695 | } |
5b2c4dd2 | 1696 | |
217df670 JV |
1697 | /* open the slave since the application closed it */ |
1698 | res = dev_open(slave_dev); | |
1699 | if (res) { | |
3d632c3f | 1700 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1701 | goto err_unset_master; |
1da177e4 LT |
1702 | } |
1703 | ||
35d48903 | 1704 | new_slave->bond = bond; |
1da177e4 | 1705 | new_slave->dev = slave_dev; |
0b680e75 | 1706 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1707 | |
58402054 | 1708 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1709 | /* bond_alb_init_slave() must be called before all other stages since |
1710 | * it might fail and we do not want to have to undo everything | |
1711 | */ | |
1712 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1713 | if (res) |
569f0c4d | 1714 | goto err_close; |
1da177e4 LT |
1715 | } |
1716 | ||
1717 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1718 | * master's promisc (and mc) settings only if it becomes the | |
1719 | * curr_active_slave, and that is taken care of later when calling | |
1720 | * bond_change_active() | |
1721 | */ | |
1722 | if (!USES_PRIMARY(bond->params.mode)) { | |
1723 | /* set promiscuity level to new slave */ | |
1724 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1725 | res = dev_set_promiscuity(slave_dev, 1); |
1726 | if (res) | |
1727 | goto err_close; | |
1da177e4 LT |
1728 | } |
1729 | ||
1730 | /* set allmulti level to new slave */ | |
1731 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1732 | res = dev_set_allmulti(slave_dev, 1); |
1733 | if (res) | |
1734 | goto err_close; | |
1da177e4 LT |
1735 | } |
1736 | ||
b9e40857 | 1737 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1738 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1739 | netdev_for_each_mc_addr(ha, bond_dev) |
1740 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1741 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1742 | } |
1743 | ||
1744 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1745 | /* add lacpdu mc addr to mc list */ | |
1746 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1747 | ||
22bedad3 | 1748 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1749 | } |
1750 | ||
1751 | bond_add_vlans_on_slave(bond, slave_dev); | |
1752 | ||
1753 | write_lock_bh(&bond->lock); | |
1754 | ||
1755 | bond_attach_slave(bond, new_slave); | |
1756 | ||
1757 | new_slave->delay = 0; | |
1758 | new_slave->link_failure_count = 0; | |
1759 | ||
8531c5ff AK |
1760 | bond_compute_features(bond); |
1761 | ||
3915c1e8 JV |
1762 | write_unlock_bh(&bond->lock); |
1763 | ||
1764 | read_lock(&bond->lock); | |
1765 | ||
f5b2b966 JV |
1766 | new_slave->last_arp_rx = jiffies; |
1767 | ||
1da177e4 LT |
1768 | if (bond->params.miimon && !bond->params.use_carrier) { |
1769 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1770 | ||
1771 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1772 | /* | |
1773 | * miimon is set but a bonded network driver | |
1774 | * does not support ETHTOOL/MII and | |
1775 | * arp_interval is not set. Note: if | |
1776 | * use_carrier is enabled, we will never go | |
1777 | * here (because netif_carrier is always | |
1778 | * supported); thus, we don't need to change | |
1779 | * the messages for netif_carrier. | |
1780 | */ | |
a4aee5c8 | 1781 | 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 | 1782 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1783 | } else if (link_reporting == -1) { |
1784 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1785 | 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", |
1786 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1787 | } |
1788 | } | |
1789 | ||
1790 | /* check for initial state */ | |
1791 | if (!bond->params.miimon || | |
1792 | (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) { | |
1793 | if (bond->params.updelay) { | |
a4aee5c8 | 1794 | pr_debug("Initial state of slave_dev is BOND_LINK_BACK\n"); |
1da177e4 LT |
1795 | new_slave->link = BOND_LINK_BACK; |
1796 | new_slave->delay = bond->params.updelay; | |
1797 | } else { | |
a4aee5c8 | 1798 | pr_debug("Initial state of slave_dev is BOND_LINK_UP\n"); |
1da177e4 LT |
1799 | new_slave->link = BOND_LINK_UP; |
1800 | } | |
1801 | new_slave->jiffies = jiffies; | |
1802 | } else { | |
a4aee5c8 | 1803 | pr_debug("Initial state of slave_dev is BOND_LINK_DOWN\n"); |
1da177e4 LT |
1804 | new_slave->link = BOND_LINK_DOWN; |
1805 | } | |
1806 | ||
1807 | if (bond_update_speed_duplex(new_slave) && | |
1808 | (new_slave->link != BOND_LINK_DOWN)) { | |
a4aee5c8 JP |
1809 | pr_warning("%s: Warning: failed to get speed and duplex from %s, assumed to be 100Mb/sec and Full.\n", |
1810 | bond_dev->name, new_slave->dev->name); | |
1da177e4 LT |
1811 | |
1812 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a4aee5c8 JP |
1813 | pr_warning("%s: Warning: Operation of 802.3ad mode requires ETHTOOL support in base driver for proper aggregator selection.\n", |
1814 | bond_dev->name); | |
1da177e4 LT |
1815 | } |
1816 | } | |
1817 | ||
1818 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { | |
1819 | /* if there is a primary slave, remember it */ | |
a549952a | 1820 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1821 | bond->primary_slave = new_slave; |
a549952a JP |
1822 | bond->force_primary = true; |
1823 | } | |
1da177e4 LT |
1824 | } |
1825 | ||
3915c1e8 JV |
1826 | write_lock_bh(&bond->curr_slave_lock); |
1827 | ||
1da177e4 LT |
1828 | switch (bond->params.mode) { |
1829 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1830 | bond_set_slave_inactive_flags(new_slave); |
1831 | bond_select_active_slave(bond); | |
1da177e4 LT |
1832 | break; |
1833 | case BOND_MODE_8023AD: | |
1834 | /* in 802.3ad mode, the internal mechanism | |
1835 | * will activate the slaves in the selected | |
1836 | * aggregator | |
1837 | */ | |
1838 | bond_set_slave_inactive_flags(new_slave); | |
1839 | /* if this is the first slave */ | |
1840 | if (bond->slave_cnt == 1) { | |
1841 | SLAVE_AD_INFO(new_slave).id = 1; | |
1842 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1843 | * can be called only after the mac address of the bond is set | |
1844 | */ | |
1845 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL, | |
1846 | bond->params.lacp_fast); | |
1847 | } else { | |
1848 | SLAVE_AD_INFO(new_slave).id = | |
1849 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1850 | } | |
1851 | ||
1852 | bond_3ad_bind_slave(new_slave); | |
1853 | break; | |
1854 | case BOND_MODE_TLB: | |
1855 | case BOND_MODE_ALB: | |
e30bc066 | 1856 | bond_set_active_slave(new_slave); |
059fe7a5 | 1857 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1858 | bond_select_active_slave(bond); |
1da177e4 LT |
1859 | break; |
1860 | default: | |
5a03cdb7 | 1861 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1862 | |
1863 | /* always active in trunk mode */ | |
e30bc066 | 1864 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1865 | |
1866 | /* In trunking mode there is little meaning to curr_active_slave | |
1867 | * anyway (it holds no special properties of the bond device), | |
1868 | * so we can change it without calling change_active_interface() | |
1869 | */ | |
3d632c3f | 1870 | if (!bond->curr_active_slave) |
1da177e4 | 1871 | bond->curr_active_slave = new_slave; |
3d632c3f | 1872 | |
1da177e4 LT |
1873 | break; |
1874 | } /* switch(bond_mode) */ | |
1875 | ||
3915c1e8 JV |
1876 | write_unlock_bh(&bond->curr_slave_lock); |
1877 | ||
ff59c456 JV |
1878 | bond_set_carrier(bond); |
1879 | ||
f6dc31a8 | 1880 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 AW |
1881 | slave_dev->npinfo = bond_netpoll_info(bond); |
1882 | if (slave_dev->npinfo) { | |
1883 | if (slave_enable_netpoll(new_slave)) { | |
1884 | read_unlock(&bond->lock); | |
1885 | pr_info("Error, %s: master_dev is using netpoll, " | |
1886 | "but new slave device does not support netpoll.\n", | |
1887 | bond_dev->name); | |
1888 | res = -EBUSY; | |
1889 | goto err_close; | |
1890 | } | |
f6dc31a8 WC |
1891 | } |
1892 | #endif | |
8a8efa22 | 1893 | |
3915c1e8 | 1894 | read_unlock(&bond->lock); |
1da177e4 | 1895 | |
b76cdba9 MW |
1896 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1897 | if (res) | |
569f0c4d | 1898 | goto err_close; |
b76cdba9 | 1899 | |
35d48903 JP |
1900 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1901 | new_slave); | |
1902 | if (res) { | |
1903 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1904 | goto err_dest_symlinks; | |
1905 | } | |
1906 | ||
a4aee5c8 JP |
1907 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1908 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1909 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1910 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1911 | |
1912 | /* enslave is successful */ | |
1913 | return 0; | |
1914 | ||
1915 | /* Undo stages on error */ | |
35d48903 JP |
1916 | err_dest_symlinks: |
1917 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1918 | ||
1da177e4 LT |
1919 | err_close: |
1920 | dev_close(slave_dev); | |
1921 | ||
569f0c4d | 1922 | err_unset_master: |
1765a575 | 1923 | netdev_set_bond_master(slave_dev, NULL); |
569f0c4d | 1924 | |
1da177e4 | 1925 | err_restore_mac: |
dd957c57 | 1926 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1927 | /* XXX TODO - fom follow mode needs to change master's |
1928 | * MAC if this slave's MAC is in use by the bond, or at | |
1929 | * least print a warning. | |
1930 | */ | |
2ab82852 MS |
1931 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1932 | addr.sa_family = slave_dev->type; | |
1933 | dev_set_mac_address(slave_dev, &addr); | |
1934 | } | |
1da177e4 | 1935 | |
b15ba0fb JP |
1936 | err_restore_mtu: |
1937 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1938 | ||
1da177e4 LT |
1939 | err_free: |
1940 | kfree(new_slave); | |
1941 | ||
1942 | err_undo_flags: | |
1943 | bond_dev->features = old_features; | |
3d632c3f | 1944 | |
1da177e4 LT |
1945 | return res; |
1946 | } | |
1947 | ||
1948 | /* | |
1949 | * Try to release the slave device <slave> from the bond device <master> | |
1950 | * It is legal to access curr_active_slave without a lock because all the function | |
1951 | * is write-locked. | |
1952 | * | |
1953 | * The rules for slave state should be: | |
1954 | * for Active/Backup: | |
1955 | * Active stays on all backups go down | |
1956 | * for Bonded connections: | |
1957 | * The first up interface should be left on and all others downed. | |
1958 | */ | |
a77b5325 | 1959 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1960 | { |
454d7c9b | 1961 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1962 | struct slave *slave, *oldcurrent; |
1963 | struct sockaddr addr; | |
1da177e4 LT |
1964 | |
1965 | /* slave is not a slave or master is not master of this slave */ | |
1966 | if (!(slave_dev->flags & IFF_SLAVE) || | |
1967 | (slave_dev->master != bond_dev)) { | |
a4aee5c8 | 1968 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1969 | bond_dev->name, slave_dev->name); |
1970 | return -EINVAL; | |
1971 | } | |
1972 | ||
e843fa50 | 1973 | block_netpoll_tx(); |
f6dc31a8 | 1974 | netdev_bonding_change(bond_dev, NETDEV_BONDING_DESLAVE); |
1da177e4 LT |
1975 | write_lock_bh(&bond->lock); |
1976 | ||
1977 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1978 | if (!slave) { | |
1979 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1980 | pr_info("%s: %s not enslaved\n", |
1981 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1982 | write_unlock_bh(&bond->lock); |
e843fa50 | 1983 | unblock_netpoll_tx(); |
1da177e4 LT |
1984 | return -EINVAL; |
1985 | } | |
1986 | ||
35d48903 JP |
1987 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1988 | * for this slave anymore. | |
1989 | */ | |
1990 | netdev_rx_handler_unregister(slave_dev); | |
1991 | write_unlock_bh(&bond->lock); | |
1992 | synchronize_net(); | |
1993 | write_lock_bh(&bond->lock); | |
1994 | ||
3915c1e8 | 1995 | if (!bond->params.fail_over_mac) { |
8e95a202 JP |
1996 | if (!compare_ether_addr(bond_dev->dev_addr, slave->perm_hwaddr) && |
1997 | bond->slave_cnt > 1) | |
a4aee5c8 JP |
1998 | 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", |
1999 | bond_dev->name, slave_dev->name, | |
2000 | slave->perm_hwaddr, | |
2001 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
2002 | } |
2003 | ||
2004 | /* Inform AD package of unbinding of slave. */ | |
2005 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2006 | /* must be called before the slave is | |
2007 | * detached from the list | |
2008 | */ | |
2009 | bond_3ad_unbind_slave(slave); | |
2010 | } | |
2011 | ||
a4aee5c8 JP |
2012 | pr_info("%s: releasing %s interface %s\n", |
2013 | bond_dev->name, | |
e30bc066 | 2014 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 2015 | slave_dev->name); |
1da177e4 LT |
2016 | |
2017 | oldcurrent = bond->curr_active_slave; | |
2018 | ||
2019 | bond->current_arp_slave = NULL; | |
2020 | ||
2021 | /* release the slave from its bond */ | |
2022 | bond_detach_slave(bond, slave); | |
2023 | ||
8531c5ff AK |
2024 | bond_compute_features(bond); |
2025 | ||
3d632c3f | 2026 | if (bond->primary_slave == slave) |
1da177e4 | 2027 | bond->primary_slave = NULL; |
1da177e4 | 2028 | |
3d632c3f | 2029 | if (oldcurrent == slave) |
1da177e4 | 2030 | bond_change_active_slave(bond, NULL); |
1da177e4 | 2031 | |
58402054 | 2032 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2033 | /* Must be called only after the slave has been |
2034 | * detached from the list and the curr_active_slave | |
2035 | * has been cleared (if our_slave == old_current), | |
2036 | * but before a new active slave is selected. | |
2037 | */ | |
2543331d | 2038 | write_unlock_bh(&bond->lock); |
1da177e4 | 2039 | bond_alb_deinit_slave(bond, slave); |
2543331d | 2040 | write_lock_bh(&bond->lock); |
1da177e4 LT |
2041 | } |
2042 | ||
059fe7a5 JV |
2043 | if (oldcurrent == slave) { |
2044 | /* | |
2045 | * Note that we hold RTNL over this sequence, so there | |
2046 | * is no concern that another slave add/remove event | |
2047 | * will interfere. | |
2048 | */ | |
2049 | write_unlock_bh(&bond->lock); | |
2050 | read_lock(&bond->lock); | |
2051 | write_lock_bh(&bond->curr_slave_lock); | |
2052 | ||
1da177e4 LT |
2053 | bond_select_active_slave(bond); |
2054 | ||
059fe7a5 JV |
2055 | write_unlock_bh(&bond->curr_slave_lock); |
2056 | read_unlock(&bond->lock); | |
2057 | write_lock_bh(&bond->lock); | |
2058 | } | |
2059 | ||
1da177e4 | 2060 | if (bond->slave_cnt == 0) { |
ff59c456 JV |
2061 | bond_set_carrier(bond); |
2062 | ||
1da177e4 LT |
2063 | /* if the last slave was removed, zero the mac address |
2064 | * of the master so it will be set by the application | |
2065 | * to the mac address of the first slave | |
2066 | */ | |
2067 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2068 | ||
f35188fa | 2069 | if (!bond->vlgrp) { |
1da177e4 LT |
2070 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
2071 | } else { | |
a4aee5c8 JP |
2072 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2073 | bond_dev->name, bond_dev->name); | |
2074 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2075 | bond_dev->name); | |
1da177e4 LT |
2076 | } |
2077 | } else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2078 | !bond_has_challenged_slaves(bond)) { | |
a4aee5c8 JP |
2079 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", |
2080 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
2081 | bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED; |
2082 | } | |
2083 | ||
2084 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2085 | unblock_netpoll_tx(); |
1da177e4 | 2086 | |
b76cdba9 MW |
2087 | /* must do this from outside any spinlocks */ |
2088 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2089 | ||
1da177e4 LT |
2090 | bond_del_vlans_from_slave(bond, slave_dev); |
2091 | ||
2092 | /* If the mode USES_PRIMARY, then we should only remove its | |
2093 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2094 | * already taken care of above when we detached the slave | |
2095 | */ | |
2096 | if (!USES_PRIMARY(bond->params.mode)) { | |
2097 | /* unset promiscuity level from slave */ | |
3d632c3f | 2098 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2099 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2100 | |
2101 | /* unset allmulti level from slave */ | |
3d632c3f | 2102 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2103 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2104 | |
2105 | /* flush master's mc_list from slave */ | |
b9e40857 | 2106 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2107 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2108 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2109 | } |
2110 | ||
1765a575 | 2111 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2112 | |
8a8efa22 | 2113 | slave_disable_netpoll(slave); |
f6dc31a8 | 2114 | |
1da177e4 LT |
2115 | /* close slave before restoring its mac address */ |
2116 | dev_close(slave_dev); | |
2117 | ||
3915c1e8 | 2118 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2119 | /* restore original ("permanent") mac address */ |
2120 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2121 | addr.sa_family = slave_dev->type; | |
2122 | dev_set_mac_address(slave_dev, &addr); | |
2123 | } | |
1da177e4 | 2124 | |
b15ba0fb JP |
2125 | dev_set_mtu(slave_dev, slave->original_mtu); |
2126 | ||
2d7011ca | 2127 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2128 | |
2129 | kfree(slave); | |
2130 | ||
2131 | return 0; /* deletion OK */ | |
2132 | } | |
2133 | ||
d90a162a | 2134 | /* |
dadaa10b | 2135 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2136 | * Must be under rtnl_lock when this function is called. |
2137 | */ | |
26d8ee75 | 2138 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2139 | struct net_device *slave_dev) | |
d90a162a | 2140 | { |
454d7c9b | 2141 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2142 | int ret; |
2143 | ||
2144 | ret = bond_release(bond_dev, slave_dev); | |
2145 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2146 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2147 | pr_info("%s: destroying bond %s.\n", |
2148 | bond_dev->name, bond_dev->name); | |
9e71626c | 2149 | unregister_netdevice(bond_dev); |
d90a162a MS |
2150 | } |
2151 | return ret; | |
2152 | } | |
2153 | ||
1da177e4 LT |
2154 | /* |
2155 | * This function releases all slaves. | |
2156 | */ | |
2157 | static int bond_release_all(struct net_device *bond_dev) | |
2158 | { | |
454d7c9b | 2159 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2160 | struct slave *slave; |
2161 | struct net_device *slave_dev; | |
2162 | struct sockaddr addr; | |
2163 | ||
2164 | write_lock_bh(&bond->lock); | |
2165 | ||
ff59c456 JV |
2166 | netif_carrier_off(bond_dev); |
2167 | ||
3d632c3f | 2168 | if (bond->slave_cnt == 0) |
1da177e4 | 2169 | goto out; |
1da177e4 LT |
2170 | |
2171 | bond->current_arp_slave = NULL; | |
2172 | bond->primary_slave = NULL; | |
2173 | bond_change_active_slave(bond, NULL); | |
2174 | ||
2175 | while ((slave = bond->first_slave) != NULL) { | |
2176 | /* Inform AD package of unbinding of slave | |
2177 | * before slave is detached from the list. | |
2178 | */ | |
3d632c3f | 2179 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 2180 | bond_3ad_unbind_slave(slave); |
1da177e4 LT |
2181 | |
2182 | slave_dev = slave->dev; | |
2183 | bond_detach_slave(bond, slave); | |
2184 | ||
2543331d JV |
2185 | /* now that the slave is detached, unlock and perform |
2186 | * all the undo steps that should not be called from | |
2187 | * within a lock. | |
2188 | */ | |
2189 | write_unlock_bh(&bond->lock); | |
2190 | ||
35d48903 JP |
2191 | /* unregister rx_handler early so bond_handle_frame wouldn't |
2192 | * be called for this slave anymore. | |
2193 | */ | |
2194 | netdev_rx_handler_unregister(slave_dev); | |
2195 | synchronize_net(); | |
2196 | ||
58402054 | 2197 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2198 | /* must be called only after the slave |
2199 | * has been detached from the list | |
2200 | */ | |
2201 | bond_alb_deinit_slave(bond, slave); | |
2202 | } | |
2203 | ||
8531c5ff AK |
2204 | bond_compute_features(bond); |
2205 | ||
b76cdba9 | 2206 | bond_destroy_slave_symlinks(bond_dev, slave_dev); |
1da177e4 LT |
2207 | bond_del_vlans_from_slave(bond, slave_dev); |
2208 | ||
2209 | /* If the mode USES_PRIMARY, then we should only remove its | |
2210 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2211 | * already taken care of above when we detached the slave | |
2212 | */ | |
2213 | if (!USES_PRIMARY(bond->params.mode)) { | |
2214 | /* unset promiscuity level from slave */ | |
3d632c3f | 2215 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2216 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2217 | |
2218 | /* unset allmulti level from slave */ | |
3d632c3f | 2219 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2220 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2221 | |
2222 | /* flush master's mc_list from slave */ | |
b9e40857 | 2223 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2224 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2225 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2226 | } |
2227 | ||
1765a575 | 2228 | netdev_set_bond_master(slave_dev, NULL); |
1da177e4 | 2229 | |
8a8efa22 AW |
2230 | slave_disable_netpoll(slave); |
2231 | ||
1da177e4 LT |
2232 | /* close slave before restoring its mac address */ |
2233 | dev_close(slave_dev); | |
2234 | ||
dd957c57 | 2235 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
2236 | /* restore original ("permanent") mac address*/ |
2237 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2238 | addr.sa_family = slave_dev->type; | |
2239 | dev_set_mac_address(slave_dev, &addr); | |
2240 | } | |
1da177e4 | 2241 | |
1da177e4 LT |
2242 | kfree(slave); |
2243 | ||
2244 | /* re-acquire the lock before getting the next slave */ | |
2245 | write_lock_bh(&bond->lock); | |
2246 | } | |
2247 | ||
2248 | /* zero the mac address of the master so it will be | |
2249 | * set by the application to the mac address of the | |
2250 | * first slave | |
2251 | */ | |
2252 | memset(bond_dev->dev_addr, 0, bond_dev->addr_len); | |
2253 | ||
f35188fa | 2254 | if (!bond->vlgrp) { |
1da177e4 | 2255 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; |
f35188fa | 2256 | } else { |
a4aee5c8 JP |
2257 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2258 | bond_dev->name, bond_dev->name); | |
2259 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2260 | bond_dev->name); | |
1da177e4 LT |
2261 | } |
2262 | ||
a4aee5c8 | 2263 | pr_info("%s: released all slaves\n", bond_dev->name); |
1da177e4 LT |
2264 | |
2265 | out: | |
2266 | write_unlock_bh(&bond->lock); | |
1da177e4 LT |
2267 | return 0; |
2268 | } | |
2269 | ||
2270 | /* | |
2271 | * This function changes the active slave to slave <slave_dev>. | |
2272 | * It returns -EINVAL in the following cases. | |
2273 | * - <slave_dev> is not found in the list. | |
2274 | * - There is not active slave now. | |
2275 | * - <slave_dev> is already active. | |
2276 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2277 | * - <slave_dev> is not running. | |
3d632c3f SH |
2278 | * In these cases, this function does nothing. |
2279 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2280 | */ |
2281 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2282 | { | |
454d7c9b | 2283 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2284 | struct slave *old_active = NULL; |
2285 | struct slave *new_active = NULL; | |
2286 | int res = 0; | |
2287 | ||
3d632c3f | 2288 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2289 | return -EINVAL; |
1da177e4 LT |
2290 | |
2291 | /* Verify that master_dev is indeed the master of slave_dev */ | |
3d632c3f | 2292 | if (!(slave_dev->flags & IFF_SLAVE) || (slave_dev->master != bond_dev)) |
1da177e4 | 2293 | return -EINVAL; |
1da177e4 | 2294 | |
059fe7a5 | 2295 | read_lock(&bond->lock); |
1da177e4 | 2296 | |
059fe7a5 | 2297 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2298 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2299 | read_unlock(&bond->curr_slave_lock); |
2300 | ||
1da177e4 LT |
2301 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2302 | ||
2303 | /* | |
2304 | * Changing to the current active: do nothing; return success. | |
2305 | */ | |
2306 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2307 | read_unlock(&bond->lock); |
1da177e4 LT |
2308 | return 0; |
2309 | } | |
2310 | ||
2311 | if ((new_active) && | |
2312 | (old_active) && | |
2313 | (new_active->link == BOND_LINK_UP) && | |
2314 | IS_UP(new_active->dev)) { | |
e843fa50 | 2315 | block_netpoll_tx(); |
059fe7a5 | 2316 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2317 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2318 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2319 | unblock_netpoll_tx(); |
3d632c3f | 2320 | } else |
1da177e4 | 2321 | res = -EINVAL; |
1da177e4 | 2322 | |
059fe7a5 | 2323 | read_unlock(&bond->lock); |
1da177e4 LT |
2324 | |
2325 | return res; | |
2326 | } | |
2327 | ||
1da177e4 LT |
2328 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2329 | { | |
454d7c9b | 2330 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2331 | |
2332 | info->bond_mode = bond->params.mode; | |
2333 | info->miimon = bond->params.miimon; | |
2334 | ||
6603a6f2 | 2335 | read_lock(&bond->lock); |
1da177e4 | 2336 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2337 | read_unlock(&bond->lock); |
1da177e4 LT |
2338 | |
2339 | return 0; | |
2340 | } | |
2341 | ||
2342 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2343 | { | |
454d7c9b | 2344 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2345 | struct slave *slave; |
689c96cc | 2346 | int i, res = -ENODEV; |
1da177e4 | 2347 | |
6603a6f2 | 2348 | read_lock(&bond->lock); |
1da177e4 LT |
2349 | |
2350 | bond_for_each_slave(bond, slave, i) { | |
2351 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2352 | res = 0; |
2353 | strcpy(info->slave_name, slave->dev->name); | |
2354 | info->link = slave->link; | |
e30bc066 | 2355 | info->state = bond_slave_state(slave); |
689c96cc | 2356 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2357 | break; |
2358 | } | |
2359 | } | |
2360 | ||
6603a6f2 | 2361 | read_unlock(&bond->lock); |
1da177e4 | 2362 | |
689c96cc | 2363 | return res; |
1da177e4 LT |
2364 | } |
2365 | ||
2366 | /*-------------------------------- Monitoring -------------------------------*/ | |
2367 | ||
1da177e4 | 2368 | |
f0c76d61 JV |
2369 | static int bond_miimon_inspect(struct bonding *bond) |
2370 | { | |
2371 | struct slave *slave; | |
2372 | int i, link_state, commit = 0; | |
41f89100 JP |
2373 | bool ignore_updelay; |
2374 | ||
2375 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2376 | |
2377 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2378 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2379 | |
f0c76d61 | 2380 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2381 | |
2382 | switch (slave->link) { | |
f0c76d61 JV |
2383 | case BOND_LINK_UP: |
2384 | if (link_state) | |
2385 | continue; | |
1da177e4 | 2386 | |
f0c76d61 JV |
2387 | slave->link = BOND_LINK_FAIL; |
2388 | slave->delay = bond->params.downdelay; | |
2389 | if (slave->delay) { | |
a4aee5c8 JP |
2390 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2391 | bond->dev->name, | |
2392 | (bond->params.mode == | |
2393 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2394 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2395 | "active " : "backup ") : "", |
2396 | slave->dev->name, | |
2397 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2398 | } |
f0c76d61 JV |
2399 | /*FALLTHRU*/ |
2400 | case BOND_LINK_FAIL: | |
2401 | if (link_state) { | |
2402 | /* | |
2403 | * recovered before downdelay expired | |
2404 | */ | |
2405 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2406 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2407 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2408 | bond->dev->name, | |
2409 | (bond->params.downdelay - slave->delay) * | |
2410 | bond->params.miimon, | |
2411 | slave->dev->name); | |
f0c76d61 | 2412 | continue; |
1da177e4 | 2413 | } |
f0c76d61 JV |
2414 | |
2415 | if (slave->delay <= 0) { | |
2416 | slave->new_link = BOND_LINK_DOWN; | |
2417 | commit++; | |
2418 | continue; | |
1da177e4 | 2419 | } |
1da177e4 | 2420 | |
f0c76d61 JV |
2421 | slave->delay--; |
2422 | break; | |
2423 | ||
2424 | case BOND_LINK_DOWN: | |
2425 | if (!link_state) | |
2426 | continue; | |
2427 | ||
2428 | slave->link = BOND_LINK_BACK; | |
2429 | slave->delay = bond->params.updelay; | |
2430 | ||
2431 | if (slave->delay) { | |
a4aee5c8 JP |
2432 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2433 | bond->dev->name, slave->dev->name, | |
2434 | ignore_updelay ? 0 : | |
2435 | bond->params.updelay * | |
2436 | bond->params.miimon); | |
f0c76d61 JV |
2437 | } |
2438 | /*FALLTHRU*/ | |
2439 | case BOND_LINK_BACK: | |
2440 | if (!link_state) { | |
2441 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2442 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2443 | bond->dev->name, | |
2444 | (bond->params.updelay - slave->delay) * | |
2445 | bond->params.miimon, | |
2446 | slave->dev->name); | |
f0c76d61 JV |
2447 | |
2448 | continue; | |
2449 | } | |
2450 | ||
41f89100 JP |
2451 | if (ignore_updelay) |
2452 | slave->delay = 0; | |
2453 | ||
f0c76d61 JV |
2454 | if (slave->delay <= 0) { |
2455 | slave->new_link = BOND_LINK_UP; | |
2456 | commit++; | |
41f89100 | 2457 | ignore_updelay = false; |
f0c76d61 | 2458 | continue; |
1da177e4 | 2459 | } |
f0c76d61 JV |
2460 | |
2461 | slave->delay--; | |
1da177e4 | 2462 | break; |
f0c76d61 JV |
2463 | } |
2464 | } | |
1da177e4 | 2465 | |
f0c76d61 JV |
2466 | return commit; |
2467 | } | |
1da177e4 | 2468 | |
f0c76d61 JV |
2469 | static void bond_miimon_commit(struct bonding *bond) |
2470 | { | |
2471 | struct slave *slave; | |
2472 | int i; | |
2473 | ||
2474 | bond_for_each_slave(bond, slave, i) { | |
2475 | switch (slave->new_link) { | |
2476 | case BOND_LINK_NOCHANGE: | |
2477 | continue; | |
1da177e4 | 2478 | |
f0c76d61 JV |
2479 | case BOND_LINK_UP: |
2480 | slave->link = BOND_LINK_UP; | |
2481 | slave->jiffies = jiffies; | |
2482 | ||
2483 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2484 | /* prevent it from being the active one */ | |
e30bc066 | 2485 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2486 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2487 | /* make it immediately active */ | |
e30bc066 | 2488 | bond_set_active_slave(slave); |
f0c76d61 JV |
2489 | } else if (slave != bond->primary_slave) { |
2490 | /* prevent it from being the active one */ | |
e30bc066 | 2491 | bond_set_backup_slave(slave); |
1da177e4 | 2492 | } |
1da177e4 | 2493 | |
546add79 KPO |
2494 | bond_update_speed_duplex(slave); |
2495 | ||
700c2a77 KPO |
2496 | pr_info("%s: link status definitely up for interface %s, %d Mbps %s duplex.\n", |
2497 | bond->dev->name, slave->dev->name, | |
2498 | slave->speed, slave->duplex ? "full" : "half"); | |
1da177e4 | 2499 | |
f0c76d61 JV |
2500 | /* notify ad that the link status has changed */ |
2501 | if (bond->params.mode == BOND_MODE_8023AD) | |
2502 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2503 | |
58402054 | 2504 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2505 | bond_alb_handle_link_change(bond, slave, |
2506 | BOND_LINK_UP); | |
1da177e4 | 2507 | |
f0c76d61 JV |
2508 | if (!bond->curr_active_slave || |
2509 | (slave == bond->primary_slave)) | |
2510 | goto do_failover; | |
1da177e4 | 2511 | |
f0c76d61 | 2512 | continue; |
059fe7a5 | 2513 | |
f0c76d61 | 2514 | case BOND_LINK_DOWN: |
fba4acda JV |
2515 | if (slave->link_failure_count < UINT_MAX) |
2516 | slave->link_failure_count++; | |
2517 | ||
f0c76d61 | 2518 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2519 | |
f0c76d61 JV |
2520 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2521 | bond->params.mode == BOND_MODE_8023AD) | |
2522 | bond_set_slave_inactive_flags(slave); | |
2523 | ||
a4aee5c8 JP |
2524 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2525 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2526 | |
2527 | if (bond->params.mode == BOND_MODE_8023AD) | |
2528 | bond_3ad_handle_link_change(slave, | |
2529 | BOND_LINK_DOWN); | |
2530 | ||
ae63e808 | 2531 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2532 | bond_alb_handle_link_change(bond, slave, |
2533 | BOND_LINK_DOWN); | |
2534 | ||
2535 | if (slave == bond->curr_active_slave) | |
2536 | goto do_failover; | |
2537 | ||
2538 | continue; | |
2539 | ||
2540 | default: | |
a4aee5c8 | 2541 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2542 | bond->dev->name, slave->new_link, |
2543 | slave->dev->name); | |
2544 | slave->new_link = BOND_LINK_NOCHANGE; | |
2545 | ||
2546 | continue; | |
2547 | } | |
2548 | ||
2549 | do_failover: | |
2550 | ASSERT_RTNL(); | |
e843fa50 | 2551 | block_netpoll_tx(); |
f0c76d61 JV |
2552 | write_lock_bh(&bond->curr_slave_lock); |
2553 | bond_select_active_slave(bond); | |
2554 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2555 | unblock_netpoll_tx(); |
f0c76d61 JV |
2556 | } |
2557 | ||
2558 | bond_set_carrier(bond); | |
1da177e4 LT |
2559 | } |
2560 | ||
0b0eef66 JV |
2561 | /* |
2562 | * bond_mii_monitor | |
2563 | * | |
2564 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2565 | * inspection, then (if inspection indicates something needs to be done) |
2566 | * an acquisition of appropriate locks followed by a commit phase to | |
2567 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2568 | */ |
2569 | void bond_mii_monitor(struct work_struct *work) | |
2570 | { | |
2571 | struct bonding *bond = container_of(work, struct bonding, | |
2572 | mii_work.work); | |
0b0eef66 JV |
2573 | |
2574 | read_lock(&bond->lock); | |
f0c76d61 JV |
2575 | if (bond->kill_timers) |
2576 | goto out; | |
2577 | ||
2578 | if (bond->slave_cnt == 0) | |
2579 | goto re_arm; | |
b59f9f74 JV |
2580 | |
2581 | if (bond->send_grat_arp) { | |
2582 | read_lock(&bond->curr_slave_lock); | |
2583 | bond_send_gratuitous_arp(bond); | |
2584 | read_unlock(&bond->curr_slave_lock); | |
2585 | } | |
2586 | ||
305d552a BH |
2587 | if (bond->send_unsol_na) { |
2588 | read_lock(&bond->curr_slave_lock); | |
2589 | bond_send_unsolicited_na(bond); | |
2590 | read_unlock(&bond->curr_slave_lock); | |
2591 | } | |
2592 | ||
f0c76d61 | 2593 | if (bond_miimon_inspect(bond)) { |
0b0eef66 JV |
2594 | read_unlock(&bond->lock); |
2595 | rtnl_lock(); | |
2596 | read_lock(&bond->lock); | |
f0c76d61 JV |
2597 | |
2598 | bond_miimon_commit(bond); | |
2599 | ||
5655662d JV |
2600 | read_unlock(&bond->lock); |
2601 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2602 | read_lock(&bond->lock); | |
0b0eef66 JV |
2603 | } |
2604 | ||
f0c76d61 JV |
2605 | re_arm: |
2606 | if (bond->params.miimon) | |
2607 | queue_delayed_work(bond->wq, &bond->mii_work, | |
2608 | msecs_to_jiffies(bond->params.miimon)); | |
2609 | out: | |
0b0eef66 | 2610 | read_unlock(&bond->lock); |
0b0eef66 | 2611 | } |
c3ade5ca | 2612 | |
d3bb52b0 | 2613 | static __be32 bond_glean_dev_ip(struct net_device *dev) |
c3ade5ca JV |
2614 | { |
2615 | struct in_device *idev; | |
2616 | struct in_ifaddr *ifa; | |
a144ea4b | 2617 | __be32 addr = 0; |
c3ade5ca JV |
2618 | |
2619 | if (!dev) | |
2620 | return 0; | |
2621 | ||
2622 | rcu_read_lock(); | |
e5ed6399 | 2623 | idev = __in_dev_get_rcu(dev); |
c3ade5ca JV |
2624 | if (!idev) |
2625 | goto out; | |
2626 | ||
2627 | ifa = idev->ifa_list; | |
2628 | if (!ifa) | |
2629 | goto out; | |
2630 | ||
2631 | addr = ifa->ifa_local; | |
2632 | out: | |
2633 | rcu_read_unlock(); | |
2634 | return addr; | |
2635 | } | |
2636 | ||
d3bb52b0 | 2637 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2638 | { |
0883beca | 2639 | struct vlan_entry *vlan; |
f5b2b966 JV |
2640 | |
2641 | if (ip == bond->master_ip) | |
2642 | return 1; | |
2643 | ||
0883beca | 2644 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f5b2b966 JV |
2645 | if (ip == vlan->vlan_ip) |
2646 | return 1; | |
2647 | } | |
2648 | ||
2649 | return 0; | |
2650 | } | |
2651 | ||
c3ade5ca JV |
2652 | /* |
2653 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2654 | * switches in VLAN mode (especially if ports are configured as | |
2655 | * "native" to a VLAN) might not pass non-tagged frames. | |
2656 | */ | |
d3bb52b0 | 2657 | 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 |
2658 | { |
2659 | struct sk_buff *skb; | |
2660 | ||
5a03cdb7 | 2661 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2662 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2663 | |
c3ade5ca JV |
2664 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2665 | NULL, slave_dev->dev_addr, NULL); | |
2666 | ||
2667 | if (!skb) { | |
a4aee5c8 | 2668 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2669 | return; |
2670 | } | |
2671 | if (vlan_id) { | |
2672 | skb = vlan_put_tag(skb, vlan_id); | |
2673 | if (!skb) { | |
a4aee5c8 | 2674 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2675 | return; |
2676 | } | |
2677 | } | |
2678 | arp_xmit(skb); | |
2679 | } | |
2680 | ||
2681 | ||
1da177e4 LT |
2682 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2683 | { | |
b23dd4fe | 2684 | int i, vlan_id; |
d3bb52b0 | 2685 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2686 | struct vlan_entry *vlan; |
c3ade5ca | 2687 | struct net_device *vlan_dev; |
c3ade5ca | 2688 | struct rtable *rt; |
1da177e4 | 2689 | |
6b780567 MW |
2690 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
2691 | if (!targets[i]) | |
5a31bec0 | 2692 | break; |
5a03cdb7 | 2693 | pr_debug("basa: target %x\n", targets[i]); |
f35188fa | 2694 | if (!bond->vlgrp) { |
5a03cdb7 | 2695 | pr_debug("basa: empty vlan: arp_send\n"); |
c3ade5ca JV |
2696 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2697 | bond->master_ip, 0); | |
2698 | continue; | |
2699 | } | |
2700 | ||
2701 | /* | |
2702 | * If VLANs are configured, we do a route lookup to | |
2703 | * determine which VLAN interface would be used, so we | |
2704 | * can tag the ARP with the proper VLAN tag. | |
2705 | */ | |
78fbfd8a DM |
2706 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2707 | RTO_ONLINK, 0); | |
b23dd4fe | 2708 | if (IS_ERR(rt)) { |
c3ade5ca | 2709 | if (net_ratelimit()) { |
a4aee5c8 | 2710 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2711 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2712 | } |
2713 | continue; | |
2714 | } | |
2715 | ||
2716 | /* | |
2717 | * This target is not on a VLAN | |
2718 | */ | |
d8d1f30b | 2719 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2720 | ip_rt_put(rt); |
5a03cdb7 | 2721 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
c3ade5ca JV |
2722 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2723 | bond->master_ip, 0); | |
2724 | continue; | |
2725 | } | |
2726 | ||
2727 | vlan_id = 0; | |
0883beca | 2728 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2729 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
d8d1f30b | 2730 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2731 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2732 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2733 | vlan_dev->name, vlan_id); |
2734 | break; | |
2735 | } | |
2736 | } | |
2737 | ||
2738 | if (vlan_id) { | |
ed4b9f80 | 2739 | ip_rt_put(rt); |
c3ade5ca JV |
2740 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
2741 | vlan->vlan_ip, vlan_id); | |
2742 | continue; | |
2743 | } | |
2744 | ||
2745 | if (net_ratelimit()) { | |
a4aee5c8 | 2746 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2747 | bond->dev->name, &targets[i], |
d8d1f30b | 2748 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2749 | } |
ed4b9f80 | 2750 | ip_rt_put(rt); |
c3ade5ca JV |
2751 | } |
2752 | } | |
2753 | ||
2754 | /* | |
2755 | * Kick out a gratuitous ARP for an IP on the bonding master plus one | |
2756 | * for each VLAN above us. | |
b59f9f74 JV |
2757 | * |
2758 | * Caller must hold curr_slave_lock for read or better | |
c3ade5ca JV |
2759 | */ |
2760 | static void bond_send_gratuitous_arp(struct bonding *bond) | |
2761 | { | |
2762 | struct slave *slave = bond->curr_active_slave; | |
2763 | struct vlan_entry *vlan; | |
2764 | struct net_device *vlan_dev; | |
2765 | ||
a4aee5c8 JP |
2766 | pr_debug("bond_send_grat_arp: bond %s slave %s\n", |
2767 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
b59f9f74 JV |
2768 | |
2769 | if (!slave || !bond->send_grat_arp || | |
2770 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
c3ade5ca JV |
2771 | return; |
2772 | ||
b59f9f74 JV |
2773 | bond->send_grat_arp--; |
2774 | ||
c3ade5ca JV |
2775 | if (bond->master_ip) { |
2776 | bond_arp_send(slave->dev, ARPOP_REPLY, bond->master_ip, | |
1053f62c | 2777 | bond->master_ip, 0); |
c3ade5ca JV |
2778 | } |
2779 | ||
f35188fa JV |
2780 | if (!bond->vlgrp) |
2781 | return; | |
2782 | ||
c3ade5ca | 2783 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
5c15bdec | 2784 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
2785 | if (vlan->vlan_ip) { |
2786 | bond_arp_send(slave->dev, ARPOP_REPLY, vlan->vlan_ip, | |
2787 | vlan->vlan_ip, vlan->vlan_id); | |
2788 | } | |
1da177e4 LT |
2789 | } |
2790 | } | |
2791 | ||
d3bb52b0 | 2792 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2793 | { |
2794 | int i; | |
d3bb52b0 | 2795 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2796 | |
f5b2b966 | 2797 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2798 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2799 | &sip, &tip, i, &targets[i], |
2800 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2801 | if (sip == targets[i]) { |
2802 | if (bond_has_this_ip(bond, tip)) | |
2803 | slave->last_arp_rx = jiffies; | |
2804 | return; | |
2805 | } | |
2806 | } | |
2807 | } | |
2808 | ||
2809 | static int bond_arp_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) | |
2810 | { | |
2811 | struct arphdr *arp; | |
2812 | struct slave *slave; | |
2813 | struct bonding *bond; | |
2814 | unsigned char *arp_ptr; | |
d3bb52b0 | 2815 | __be32 sip, tip; |
f5b2b966 | 2816 | |
1f3c8804 AG |
2817 | if (dev->priv_flags & IFF_802_1Q_VLAN) { |
2818 | /* | |
2819 | * When using VLANS and bonding, dev and oriv_dev may be | |
2820 | * incorrect if the physical interface supports VLAN | |
2821 | * acceleration. With this change ARP validation now | |
2822 | * works for hosts only reachable on the VLAN interface. | |
2823 | */ | |
2824 | dev = vlan_dev_real_dev(dev); | |
2825 | orig_dev = dev_get_by_index_rcu(dev_net(skb->dev),skb->skb_iif); | |
2826 | } | |
2827 | ||
f5b2b966 JV |
2828 | if (!(dev->priv_flags & IFF_BONDING) || !(dev->flags & IFF_MASTER)) |
2829 | goto out; | |
2830 | ||
454d7c9b | 2831 | bond = netdev_priv(dev); |
f5b2b966 JV |
2832 | read_lock(&bond->lock); |
2833 | ||
5a03cdb7 | 2834 | pr_debug("bond_arp_rcv: bond %s skb->dev %s orig_dev %s\n", |
a4aee5c8 JP |
2835 | bond->dev->name, skb->dev ? skb->dev->name : "NULL", |
2836 | orig_dev ? orig_dev->name : "NULL"); | |
f5b2b966 JV |
2837 | |
2838 | slave = bond_get_slave_by_dev(bond, orig_dev); | |
2839 | if (!slave || !slave_do_arp_validate(bond, slave)) | |
2840 | goto out_unlock; | |
2841 | ||
b3053251 NH |
2842 | skb = skb_share_check(skb, GFP_ATOMIC); |
2843 | if (!skb) | |
2844 | goto out_unlock; | |
2845 | ||
988b7050 | 2846 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) |
f5b2b966 JV |
2847 | goto out_unlock; |
2848 | ||
d0a92be0 | 2849 | arp = arp_hdr(skb); |
f5b2b966 JV |
2850 | if (arp->ar_hln != dev->addr_len || |
2851 | skb->pkt_type == PACKET_OTHERHOST || | |
2852 | skb->pkt_type == PACKET_LOOPBACK || | |
2853 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2854 | arp->ar_pro != htons(ETH_P_IP) || | |
2855 | arp->ar_pln != 4) | |
2856 | goto out_unlock; | |
2857 | ||
2858 | arp_ptr = (unsigned char *)(arp + 1); | |
2859 | arp_ptr += dev->addr_len; | |
2860 | memcpy(&sip, arp_ptr, 4); | |
2861 | arp_ptr += 4 + dev->addr_len; | |
2862 | memcpy(&tip, arp_ptr, 4); | |
2863 | ||
5a03cdb7 | 2864 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2865 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2866 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2867 | &sip, &tip); | |
f5b2b966 JV |
2868 | |
2869 | /* | |
2870 | * Backup slaves won't see the ARP reply, but do come through | |
2871 | * here for each ARP probe (so we swap the sip/tip to validate | |
2872 | * the probe). In a "redundant switch, common router" type of | |
2873 | * configuration, the ARP probe will (hopefully) travel from | |
2874 | * the active, through one switch, the router, then the other | |
2875 | * switch before reaching the backup. | |
2876 | */ | |
e30bc066 | 2877 | if (bond_is_active_slave(slave)) |
f5b2b966 JV |
2878 | bond_validate_arp(bond, slave, sip, tip); |
2879 | else | |
2880 | bond_validate_arp(bond, slave, tip, sip); | |
2881 | ||
2882 | out_unlock: | |
2883 | read_unlock(&bond->lock); | |
2884 | out: | |
2885 | dev_kfree_skb(skb); | |
2886 | return NET_RX_SUCCESS; | |
2887 | } | |
2888 | ||
1da177e4 LT |
2889 | /* |
2890 | * this function is called regularly to monitor each slave's link | |
2891 | * ensuring that traffic is being sent and received when arp monitoring | |
2892 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2893 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2894 | * arp monitoring in active backup mode. | |
2895 | */ | |
1b76b316 | 2896 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2897 | { |
1b76b316 JV |
2898 | struct bonding *bond = container_of(work, struct bonding, |
2899 | arp_work.work); | |
1da177e4 LT |
2900 | struct slave *slave, *oldcurrent; |
2901 | int do_failover = 0; | |
2902 | int delta_in_ticks; | |
2903 | int i; | |
2904 | ||
2905 | read_lock(&bond->lock); | |
2906 | ||
5ce0da8f | 2907 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 2908 | |
3d632c3f | 2909 | if (bond->kill_timers) |
1da177e4 | 2910 | goto out; |
1da177e4 | 2911 | |
3d632c3f | 2912 | if (bond->slave_cnt == 0) |
1da177e4 | 2913 | goto re_arm; |
1da177e4 LT |
2914 | |
2915 | read_lock(&bond->curr_slave_lock); | |
2916 | oldcurrent = bond->curr_active_slave; | |
2917 | read_unlock(&bond->curr_slave_lock); | |
2918 | ||
2919 | /* see if any of the previous devices are up now (i.e. they have | |
2920 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2921 | * the picture unless it is null. also, slave->jiffies is not needed | |
2922 | * here because we send an arp on each slave and give a slave as | |
2923 | * long as it needs to get the tx/rx within the delta. | |
2924 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2925 | * so it can wait | |
2926 | */ | |
2927 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2928 | unsigned long trans_start = dev_trans_start(slave->dev); |
2929 | ||
1da177e4 | 2930 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2931 | if (time_in_range(jiffies, |
2932 | trans_start - delta_in_ticks, | |
2933 | trans_start + delta_in_ticks) && | |
2934 | time_in_range(jiffies, | |
2935 | slave->dev->last_rx - delta_in_ticks, | |
2936 | slave->dev->last_rx + delta_in_ticks)) { | |
1da177e4 LT |
2937 | |
2938 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2939 | bond_set_active_slave(slave); |
1da177e4 LT |
2940 | |
2941 | /* primary_slave has no meaning in round-robin | |
2942 | * mode. the window of a slave being up and | |
2943 | * curr_active_slave being null after enslaving | |
2944 | * is closed. | |
2945 | */ | |
2946 | if (!oldcurrent) { | |
a4aee5c8 JP |
2947 | pr_info("%s: link status definitely up for interface %s, ", |
2948 | bond->dev->name, | |
2949 | slave->dev->name); | |
1da177e4 LT |
2950 | do_failover = 1; |
2951 | } else { | |
a4aee5c8 JP |
2952 | pr_info("%s: interface %s is now up\n", |
2953 | bond->dev->name, | |
2954 | slave->dev->name); | |
1da177e4 LT |
2955 | } |
2956 | } | |
2957 | } else { | |
2958 | /* slave->link == BOND_LINK_UP */ | |
2959 | ||
2960 | /* not all switches will respond to an arp request | |
2961 | * when the source ip is 0, so don't take the link down | |
2962 | * if we don't know our ip yet | |
2963 | */ | |
cb32f2a0 JB |
2964 | if (!time_in_range(jiffies, |
2965 | trans_start - delta_in_ticks, | |
2966 | trans_start + 2 * delta_in_ticks) || | |
2967 | !time_in_range(jiffies, | |
2968 | slave->dev->last_rx - delta_in_ticks, | |
2969 | slave->dev->last_rx + 2 * delta_in_ticks)) { | |
1da177e4 LT |
2970 | |
2971 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2972 | bond_set_backup_slave(slave); |
1da177e4 | 2973 | |
3d632c3f | 2974 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2975 | slave->link_failure_count++; |
1da177e4 | 2976 | |
a4aee5c8 JP |
2977 | pr_info("%s: interface %s is now down.\n", |
2978 | bond->dev->name, | |
2979 | slave->dev->name); | |
1da177e4 | 2980 | |
3d632c3f | 2981 | if (slave == oldcurrent) |
1da177e4 | 2982 | do_failover = 1; |
1da177e4 LT |
2983 | } |
2984 | } | |
2985 | ||
2986 | /* note: if switch is in round-robin mode, all links | |
2987 | * must tx arp to ensure all links rx an arp - otherwise | |
2988 | * links may oscillate or not come up at all; if switch is | |
2989 | * in something like xor mode, there is nothing we can | |
2990 | * do - all replies will be rx'ed on same link causing slaves | |
2991 | * to be unstable during low/no traffic periods | |
2992 | */ | |
3d632c3f | 2993 | if (IS_UP(slave->dev)) |
1da177e4 | 2994 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2995 | } |
2996 | ||
2997 | if (do_failover) { | |
e843fa50 | 2998 | block_netpoll_tx(); |
059fe7a5 | 2999 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
3000 | |
3001 | bond_select_active_slave(bond); | |
3002 | ||
059fe7a5 | 3003 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3004 | unblock_netpoll_tx(); |
1da177e4 LT |
3005 | } |
3006 | ||
3007 | re_arm: | |
1b76b316 JV |
3008 | if (bond->params.arp_interval) |
3009 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); | |
1da177e4 LT |
3010 | out: |
3011 | read_unlock(&bond->lock); | |
3012 | } | |
3013 | ||
3014 | /* | |
b2220cad JV |
3015 | * Called to inspect slaves for active-backup mode ARP monitor link state |
3016 | * changes. Sets new_link in slaves to specify what action should take | |
3017 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
3018 | * to link states must be committed. | |
3019 | * | |
3020 | * Called with bond->lock held for read. | |
1da177e4 | 3021 | */ |
b2220cad | 3022 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 3023 | { |
1da177e4 | 3024 | struct slave *slave; |
b2220cad | 3025 | int i, commit = 0; |
cb32f2a0 | 3026 | unsigned long trans_start; |
1da177e4 | 3027 | |
b2220cad JV |
3028 | bond_for_each_slave(bond, slave, i) { |
3029 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 3030 | |
b2220cad | 3031 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
3032 | if (time_in_range(jiffies, |
3033 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3034 | slave_last_rx(bond, slave) + delta_in_ticks)) { | |
3035 | ||
b2220cad JV |
3036 | slave->new_link = BOND_LINK_UP; |
3037 | commit++; | |
3038 | } | |
1da177e4 | 3039 | |
b2220cad JV |
3040 | continue; |
3041 | } | |
1da177e4 | 3042 | |
b2220cad JV |
3043 | /* |
3044 | * Give slaves 2*delta after being enslaved or made | |
3045 | * active. This avoids bouncing, as the last receive | |
3046 | * times need a full ARP monitor cycle to be updated. | |
3047 | */ | |
cb32f2a0 JB |
3048 | if (time_in_range(jiffies, |
3049 | slave->jiffies - delta_in_ticks, | |
3050 | slave->jiffies + 2 * delta_in_ticks)) | |
b2220cad JV |
3051 | continue; |
3052 | ||
3053 | /* | |
3054 | * Backup slave is down if: | |
3055 | * - No current_arp_slave AND | |
3056 | * - more than 3*delta since last receive AND | |
3057 | * - the bond has an IP address | |
3058 | * | |
3059 | * Note: a non-null current_arp_slave indicates | |
3060 | * the curr_active_slave went down and we are | |
3061 | * searching for a new one; under this condition | |
3062 | * we only take the curr_active_slave down - this | |
3063 | * gives each slave a chance to tx/rx traffic | |
3064 | * before being taken out | |
3065 | */ | |
e30bc066 | 3066 | if (!bond_is_active_slave(slave) && |
b2220cad | 3067 | !bond->current_arp_slave && |
cb32f2a0 JB |
3068 | !time_in_range(jiffies, |
3069 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3070 | slave_last_rx(bond, slave) + 3 * delta_in_ticks)) { | |
3071 | ||
b2220cad JV |
3072 | slave->new_link = BOND_LINK_DOWN; |
3073 | commit++; | |
3074 | } | |
3075 | ||
3076 | /* | |
3077 | * Active slave is down if: | |
3078 | * - more than 2*delta since transmitting OR | |
3079 | * - (more than 2*delta since receive AND | |
3080 | * the bond has an IP address) | |
3081 | */ | |
cb32f2a0 | 3082 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 3083 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
3084 | (!time_in_range(jiffies, |
3085 | trans_start - delta_in_ticks, | |
3086 | trans_start + 2 * delta_in_ticks) || | |
3087 | !time_in_range(jiffies, | |
3088 | slave_last_rx(bond, slave) - delta_in_ticks, | |
3089 | slave_last_rx(bond, slave) + 2 * delta_in_ticks))) { | |
3090 | ||
b2220cad JV |
3091 | slave->new_link = BOND_LINK_DOWN; |
3092 | commit++; | |
3093 | } | |
1da177e4 LT |
3094 | } |
3095 | ||
b2220cad JV |
3096 | return commit; |
3097 | } | |
1da177e4 | 3098 | |
b2220cad JV |
3099 | /* |
3100 | * Called to commit link state changes noted by inspection step of | |
3101 | * active-backup mode ARP monitor. | |
3102 | * | |
3103 | * Called with RTNL and bond->lock for read. | |
3104 | */ | |
3105 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
3106 | { | |
3107 | struct slave *slave; | |
3108 | int i; | |
cb32f2a0 | 3109 | unsigned long trans_start; |
1da177e4 | 3110 | |
b2220cad JV |
3111 | bond_for_each_slave(bond, slave, i) { |
3112 | switch (slave->new_link) { | |
3113 | case BOND_LINK_NOCHANGE: | |
3114 | continue; | |
ff59c456 | 3115 | |
b2220cad | 3116 | case BOND_LINK_UP: |
cb32f2a0 | 3117 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 3118 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
3119 | time_in_range(jiffies, |
3120 | trans_start - delta_in_ticks, | |
3121 | trans_start + delta_in_ticks)) || | |
b9f60253 | 3122 | bond->curr_active_slave != slave) { |
b2220cad | 3123 | slave->link = BOND_LINK_UP; |
b2220cad JV |
3124 | bond->current_arp_slave = NULL; |
3125 | ||
a4aee5c8 | 3126 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 3127 | bond->dev->name, slave->dev->name); |
b2220cad | 3128 | |
b9f60253 JP |
3129 | if (!bond->curr_active_slave || |
3130 | (slave == bond->primary_slave)) | |
3131 | goto do_failover; | |
1da177e4 | 3132 | |
b9f60253 | 3133 | } |
1da177e4 | 3134 | |
b9f60253 | 3135 | continue; |
1da177e4 | 3136 | |
b2220cad JV |
3137 | case BOND_LINK_DOWN: |
3138 | if (slave->link_failure_count < UINT_MAX) | |
3139 | slave->link_failure_count++; | |
3140 | ||
3141 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 3142 | bond_set_slave_inactive_flags(slave); |
b2220cad | 3143 | |
a4aee5c8 | 3144 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 3145 | bond->dev->name, slave->dev->name); |
b2220cad | 3146 | |
b9f60253 | 3147 | if (slave == bond->curr_active_slave) { |
b2220cad | 3148 | bond->current_arp_slave = NULL; |
b9f60253 | 3149 | goto do_failover; |
1da177e4 | 3150 | } |
b9f60253 JP |
3151 | |
3152 | continue; | |
b2220cad JV |
3153 | |
3154 | default: | |
a4aee5c8 | 3155 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
3156 | bond->dev->name, slave->new_link, |
3157 | slave->dev->name); | |
b9f60253 | 3158 | continue; |
1da177e4 | 3159 | } |
1da177e4 | 3160 | |
b9f60253 JP |
3161 | do_failover: |
3162 | ASSERT_RTNL(); | |
e843fa50 | 3163 | block_netpoll_tx(); |
b2220cad | 3164 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 3165 | bond_select_active_slave(bond); |
b2220cad | 3166 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3167 | unblock_netpoll_tx(); |
b2220cad | 3168 | } |
1da177e4 | 3169 | |
b2220cad JV |
3170 | bond_set_carrier(bond); |
3171 | } | |
1da177e4 | 3172 | |
b2220cad JV |
3173 | /* |
3174 | * Send ARP probes for active-backup mode ARP monitor. | |
3175 | * | |
3176 | * Called with bond->lock held for read. | |
3177 | */ | |
3178 | static void bond_ab_arp_probe(struct bonding *bond) | |
3179 | { | |
3180 | struct slave *slave; | |
3181 | int i; | |
1da177e4 | 3182 | |
b2220cad | 3183 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3184 | |
b2220cad | 3185 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3186 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3187 | bond->current_arp_slave->dev->name, | |
3188 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3189 | |
b2220cad JV |
3190 | if (bond->curr_active_slave) { |
3191 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3192 | read_unlock(&bond->curr_slave_lock); | |
3193 | return; | |
3194 | } | |
1da177e4 | 3195 | |
b2220cad | 3196 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3197 | |
b2220cad JV |
3198 | /* if we don't have a curr_active_slave, search for the next available |
3199 | * backup slave from the current_arp_slave and make it the candidate | |
3200 | * for becoming the curr_active_slave | |
3201 | */ | |
1da177e4 | 3202 | |
b2220cad JV |
3203 | if (!bond->current_arp_slave) { |
3204 | bond->current_arp_slave = bond->first_slave; | |
3205 | if (!bond->current_arp_slave) | |
3206 | return; | |
3207 | } | |
1da177e4 | 3208 | |
b2220cad | 3209 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3210 | |
b2220cad JV |
3211 | /* search for next candidate */ |
3212 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3213 | if (IS_UP(slave->dev)) { | |
3214 | slave->link = BOND_LINK_BACK; | |
3215 | bond_set_slave_active_flags(slave); | |
3216 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3217 | slave->jiffies = jiffies; |
b2220cad JV |
3218 | bond->current_arp_slave = slave; |
3219 | break; | |
1da177e4 LT |
3220 | } |
3221 | ||
b2220cad JV |
3222 | /* if the link state is up at this point, we |
3223 | * mark it down - this can happen if we have | |
3224 | * simultaneous link failures and | |
3225 | * reselect_active_interface doesn't make this | |
3226 | * one the current slave so it is still marked | |
3227 | * up when it is actually down | |
1da177e4 | 3228 | */ |
b2220cad JV |
3229 | if (slave->link == BOND_LINK_UP) { |
3230 | slave->link = BOND_LINK_DOWN; | |
3231 | if (slave->link_failure_count < UINT_MAX) | |
3232 | slave->link_failure_count++; | |
1da177e4 | 3233 | |
b2220cad JV |
3234 | bond_set_slave_inactive_flags(slave); |
3235 | ||
a4aee5c8 JP |
3236 | pr_info("%s: backup interface %s is now down.\n", |
3237 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3238 | } |
b2220cad JV |
3239 | } |
3240 | } | |
1da177e4 | 3241 | |
b2220cad JV |
3242 | void bond_activebackup_arp_mon(struct work_struct *work) |
3243 | { | |
3244 | struct bonding *bond = container_of(work, struct bonding, | |
3245 | arp_work.work); | |
3246 | int delta_in_ticks; | |
1da177e4 | 3247 | |
b2220cad | 3248 | read_lock(&bond->lock); |
1da177e4 | 3249 | |
b2220cad JV |
3250 | if (bond->kill_timers) |
3251 | goto out; | |
1da177e4 | 3252 | |
b2220cad | 3253 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3254 | |
b2220cad JV |
3255 | if (bond->slave_cnt == 0) |
3256 | goto re_arm; | |
3257 | ||
b59f9f74 JV |
3258 | if (bond->send_grat_arp) { |
3259 | read_lock(&bond->curr_slave_lock); | |
3260 | bond_send_gratuitous_arp(bond); | |
3261 | read_unlock(&bond->curr_slave_lock); | |
3262 | } | |
3263 | ||
305d552a BH |
3264 | if (bond->send_unsol_na) { |
3265 | read_lock(&bond->curr_slave_lock); | |
3266 | bond_send_unsolicited_na(bond); | |
3267 | read_unlock(&bond->curr_slave_lock); | |
3268 | } | |
3269 | ||
b2220cad JV |
3270 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3271 | read_unlock(&bond->lock); | |
3272 | rtnl_lock(); | |
3273 | read_lock(&bond->lock); | |
3274 | ||
3275 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3276 | ||
3277 | read_unlock(&bond->lock); | |
3278 | rtnl_unlock(); | |
3279 | read_lock(&bond->lock); | |
1da177e4 LT |
3280 | } |
3281 | ||
b2220cad JV |
3282 | bond_ab_arp_probe(bond); |
3283 | ||
1da177e4 | 3284 | re_arm: |
3d632c3f | 3285 | if (bond->params.arp_interval) |
1b76b316 | 3286 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
1da177e4 LT |
3287 | out: |
3288 | read_unlock(&bond->lock); | |
3289 | } | |
3290 | ||
1da177e4 LT |
3291 | /*-------------------------- netdev event handling --------------------------*/ |
3292 | ||
3293 | /* | |
3294 | * Change device name | |
3295 | */ | |
3296 | static int bond_event_changename(struct bonding *bond) | |
3297 | { | |
1da177e4 LT |
3298 | bond_remove_proc_entry(bond); |
3299 | bond_create_proc_entry(bond); | |
7e083840 | 3300 | |
f073c7ca TI |
3301 | bond_debug_reregister(bond); |
3302 | ||
1da177e4 LT |
3303 | return NOTIFY_DONE; |
3304 | } | |
3305 | ||
3d632c3f SH |
3306 | static int bond_master_netdev_event(unsigned long event, |
3307 | struct net_device *bond_dev) | |
1da177e4 | 3308 | { |
454d7c9b | 3309 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3310 | |
3311 | switch (event) { | |
3312 | case NETDEV_CHANGENAME: | |
3313 | return bond_event_changename(event_bond); | |
1da177e4 LT |
3314 | default: |
3315 | break; | |
3316 | } | |
3317 | ||
3318 | return NOTIFY_DONE; | |
3319 | } | |
3320 | ||
3d632c3f SH |
3321 | static int bond_slave_netdev_event(unsigned long event, |
3322 | struct net_device *slave_dev) | |
1da177e4 LT |
3323 | { |
3324 | struct net_device *bond_dev = slave_dev->master; | |
454d7c9b | 3325 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3326 | |
3327 | switch (event) { | |
3328 | case NETDEV_UNREGISTER: | |
3329 | if (bond_dev) { | |
1284cd3a JV |
3330 | if (bond->setup_by_slave) |
3331 | bond_release_and_destroy(bond_dev, slave_dev); | |
3332 | else | |
3333 | bond_release(bond_dev, slave_dev); | |
1da177e4 LT |
3334 | } |
3335 | break; | |
3336 | case NETDEV_CHANGE: | |
17d04500 JV |
3337 | if (bond->params.mode == BOND_MODE_8023AD || bond_is_lb(bond)) { |
3338 | struct slave *slave; | |
3339 | ||
3340 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
3341 | if (slave) { | |
3342 | u16 old_speed = slave->speed; | |
65cce19c | 3343 | u8 old_duplex = slave->duplex; |
17d04500 JV |
3344 | |
3345 | bond_update_speed_duplex(slave); | |
3346 | ||
3347 | if (bond_is_lb(bond)) | |
3348 | break; | |
3349 | ||
3350 | if (old_speed != slave->speed) | |
3351 | bond_3ad_adapter_speed_changed(slave); | |
3352 | if (old_duplex != slave->duplex) | |
3353 | bond_3ad_adapter_duplex_changed(slave); | |
3354 | } | |
3355 | } | |
3356 | ||
1da177e4 LT |
3357 | break; |
3358 | case NETDEV_DOWN: | |
3359 | /* | |
3360 | * ... Or is it this? | |
3361 | */ | |
3362 | break; | |
3363 | case NETDEV_CHANGEMTU: | |
3364 | /* | |
3365 | * TODO: Should slaves be allowed to | |
3366 | * independently alter their MTU? For | |
3367 | * an active-backup bond, slaves need | |
3368 | * not be the same type of device, so | |
3369 | * MTUs may vary. For other modes, | |
3370 | * slaves arguably should have the | |
3371 | * same MTUs. To do this, we'd need to | |
3372 | * take over the slave's change_mtu | |
3373 | * function for the duration of their | |
3374 | * servitude. | |
3375 | */ | |
3376 | break; | |
3377 | case NETDEV_CHANGENAME: | |
3378 | /* | |
3379 | * TODO: handle changing the primary's name | |
3380 | */ | |
3381 | break; | |
8531c5ff AK |
3382 | case NETDEV_FEAT_CHANGE: |
3383 | bond_compute_features(bond); | |
3384 | break; | |
1da177e4 LT |
3385 | default: |
3386 | break; | |
3387 | } | |
3388 | ||
3389 | return NOTIFY_DONE; | |
3390 | } | |
3391 | ||
3392 | /* | |
3393 | * bond_netdev_event: handle netdev notifier chain events. | |
3394 | * | |
3395 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3396 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3397 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3398 | * dev_probe_lock). | |
3399 | */ | |
3d632c3f SH |
3400 | static int bond_netdev_event(struct notifier_block *this, |
3401 | unsigned long event, void *ptr) | |
1da177e4 LT |
3402 | { |
3403 | struct net_device *event_dev = (struct net_device *)ptr; | |
3404 | ||
5a03cdb7 | 3405 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3406 | event_dev ? event_dev->name : "None", |
3407 | event); | |
1da177e4 | 3408 | |
0b680e75 JV |
3409 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3410 | return NOTIFY_DONE; | |
3411 | ||
1da177e4 | 3412 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3413 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3414 | return bond_master_netdev_event(event, event_dev); |
3415 | } | |
3416 | ||
3417 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3418 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3419 | return bond_slave_netdev_event(event, event_dev); |
3420 | } | |
3421 | ||
3422 | return NOTIFY_DONE; | |
3423 | } | |
3424 | ||
c3ade5ca JV |
3425 | /* |
3426 | * bond_inetaddr_event: handle inetaddr notifier chain events. | |
3427 | * | |
3428 | * We keep track of device IPs primarily to use as source addresses in | |
3429 | * ARP monitor probes (rather than spewing out broadcasts all the time). | |
3430 | * | |
3431 | * We track one IP for the main device (if it has one), plus one per VLAN. | |
3432 | */ | |
3433 | static int bond_inetaddr_event(struct notifier_block *this, unsigned long event, void *ptr) | |
3434 | { | |
3435 | struct in_ifaddr *ifa = ptr; | |
3436 | struct net_device *vlan_dev, *event_dev = ifa->ifa_dev->dev; | |
ec87fd3b | 3437 | struct bond_net *bn = net_generic(dev_net(event_dev), bond_net_id); |
0883beca PE |
3438 | struct bonding *bond; |
3439 | struct vlan_entry *vlan; | |
c3ade5ca | 3440 | |
ec87fd3b | 3441 | list_for_each_entry(bond, &bn->dev_list, bond_list) { |
c3ade5ca JV |
3442 | if (bond->dev == event_dev) { |
3443 | switch (event) { | |
3444 | case NETDEV_UP: | |
3445 | bond->master_ip = ifa->ifa_local; | |
3446 | return NOTIFY_OK; | |
3447 | case NETDEV_DOWN: | |
3448 | bond->master_ip = bond_glean_dev_ip(bond->dev); | |
3449 | return NOTIFY_OK; | |
3450 | default: | |
3451 | return NOTIFY_DONE; | |
3452 | } | |
3453 | } | |
3454 | ||
0883beca | 3455 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
3456 | if (!bond->vlgrp) |
3457 | continue; | |
5c15bdec | 3458 | vlan_dev = vlan_group_get_device(bond->vlgrp, vlan->vlan_id); |
c3ade5ca JV |
3459 | if (vlan_dev == event_dev) { |
3460 | switch (event) { | |
3461 | case NETDEV_UP: | |
3462 | vlan->vlan_ip = ifa->ifa_local; | |
3463 | return NOTIFY_OK; | |
3464 | case NETDEV_DOWN: | |
3465 | vlan->vlan_ip = | |
3466 | bond_glean_dev_ip(vlan_dev); | |
3467 | return NOTIFY_OK; | |
3468 | default: | |
3469 | return NOTIFY_DONE; | |
3470 | } | |
3471 | } | |
3472 | } | |
3473 | } | |
3474 | return NOTIFY_DONE; | |
3475 | } | |
3476 | ||
1da177e4 LT |
3477 | static struct notifier_block bond_netdev_notifier = { |
3478 | .notifier_call = bond_netdev_event, | |
3479 | }; | |
3480 | ||
c3ade5ca JV |
3481 | static struct notifier_block bond_inetaddr_notifier = { |
3482 | .notifier_call = bond_inetaddr_event, | |
3483 | }; | |
3484 | ||
1da177e4 LT |
3485 | /*-------------------------- Packet type handling ---------------------------*/ |
3486 | ||
3487 | /* register to receive lacpdus on a bond */ | |
3488 | static void bond_register_lacpdu(struct bonding *bond) | |
3489 | { | |
3490 | struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type); | |
3491 | ||
3492 | /* initialize packet type */ | |
3493 | pk_type->type = PKT_TYPE_LACPDU; | |
3494 | pk_type->dev = bond->dev; | |
3495 | pk_type->func = bond_3ad_lacpdu_recv; | |
3496 | ||
3497 | dev_add_pack(pk_type); | |
3498 | } | |
3499 | ||
3500 | /* unregister to receive lacpdus on a bond */ | |
3501 | static void bond_unregister_lacpdu(struct bonding *bond) | |
3502 | { | |
3503 | dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type)); | |
3504 | } | |
3505 | ||
f5b2b966 JV |
3506 | void bond_register_arp(struct bonding *bond) |
3507 | { | |
3508 | struct packet_type *pt = &bond->arp_mon_pt; | |
3509 | ||
c4f283b1 JV |
3510 | if (pt->type) |
3511 | return; | |
3512 | ||
f5b2b966 | 3513 | pt->type = htons(ETH_P_ARP); |
e245cb71 | 3514 | pt->dev = bond->dev; |
f5b2b966 JV |
3515 | pt->func = bond_arp_rcv; |
3516 | dev_add_pack(pt); | |
3517 | } | |
3518 | ||
3519 | void bond_unregister_arp(struct bonding *bond) | |
3520 | { | |
c4f283b1 JV |
3521 | struct packet_type *pt = &bond->arp_mon_pt; |
3522 | ||
3523 | dev_remove_pack(pt); | |
3524 | pt->type = 0; | |
f5b2b966 JV |
3525 | } |
3526 | ||
169a3e66 JV |
3527 | /*---------------------------- Hashing Policies -----------------------------*/ |
3528 | ||
6f6652be JV |
3529 | /* |
3530 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
3531 | * the packet is not IP mimic bond_xmit_hash_policy_l2() | |
3532 | */ | |
a361c83c | 3533 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3534 | { |
3535 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3536 | struct iphdr *iph = ip_hdr(skb); | |
3537 | ||
f14c4e4e | 3538 | if (skb->protocol == htons(ETH_P_IP)) { |
6f6652be | 3539 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3540 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6f6652be JV |
3541 | } |
3542 | ||
d3da6831 | 3543 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
6f6652be JV |
3544 | } |
3545 | ||
169a3e66 | 3546 | /* |
59c51591 | 3547 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 JV |
3548 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
3549 | * altogether not IP, mimic bond_xmit_hash_policy_l2() | |
3550 | */ | |
a361c83c | 3551 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 JV |
3552 | { |
3553 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
eddc9ec5 | 3554 | struct iphdr *iph = ip_hdr(skb); |
d3bb52b0 | 3555 | __be16 *layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); |
169a3e66 JV |
3556 | int layer4_xor = 0; |
3557 | ||
f14c4e4e BH |
3558 | if (skb->protocol == htons(ETH_P_IP)) { |
3559 | if (!(iph->frag_off & htons(IP_MF|IP_OFFSET)) && | |
169a3e66 JV |
3560 | (iph->protocol == IPPROTO_TCP || |
3561 | iph->protocol == IPPROTO_UDP)) { | |
d3bb52b0 | 3562 | layer4_xor = ntohs((*layer4hdr ^ *(layer4hdr + 1))); |
169a3e66 JV |
3563 | } |
3564 | return (layer4_xor ^ | |
3565 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
3566 | ||
3567 | } | |
3568 | ||
d3da6831 | 3569 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3570 | } |
3571 | ||
3572 | /* | |
3573 | * Hash for the output device based upon layer 2 data | |
3574 | */ | |
a361c83c | 3575 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) |
169a3e66 JV |
3576 | { |
3577 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3578 | ||
d3da6831 | 3579 | return (data->h_dest[5] ^ data->h_source[5]) % count; |
169a3e66 JV |
3580 | } |
3581 | ||
1da177e4 LT |
3582 | /*-------------------------- Device entry points ----------------------------*/ |
3583 | ||
3584 | static int bond_open(struct net_device *bond_dev) | |
3585 | { | |
454d7c9b | 3586 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3587 | |
3588 | bond->kill_timers = 0; | |
3589 | ||
5a37e8ca FL |
3590 | INIT_DELAYED_WORK(&bond->mcast_work, bond_resend_igmp_join_requests_delayed); |
3591 | ||
58402054 | 3592 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3593 | /* bond_alb_initialize must be called before the timer |
3594 | * is started. | |
3595 | */ | |
3596 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) { | |
3597 | /* something went wrong - fail the open operation */ | |
b473946a | 3598 | return -ENOMEM; |
1da177e4 LT |
3599 | } |
3600 | ||
1b76b316 JV |
3601 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); |
3602 | queue_delayed_work(bond->wq, &bond->alb_work, 0); | |
1da177e4 LT |
3603 | } |
3604 | ||
3605 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ | |
1b76b316 JV |
3606 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); |
3607 | queue_delayed_work(bond->wq, &bond->mii_work, 0); | |
1da177e4 LT |
3608 | } |
3609 | ||
3610 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 JV |
3611 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) |
3612 | INIT_DELAYED_WORK(&bond->arp_work, | |
3613 | bond_activebackup_arp_mon); | |
3614 | else | |
3615 | INIT_DELAYED_WORK(&bond->arp_work, | |
3616 | bond_loadbalance_arp_mon); | |
3617 | ||
3618 | queue_delayed_work(bond->wq, &bond->arp_work, 0); | |
f5b2b966 JV |
3619 | if (bond->params.arp_validate) |
3620 | bond_register_arp(bond); | |
1da177e4 LT |
3621 | } |
3622 | ||
3623 | if (bond->params.mode == BOND_MODE_8023AD) { | |
a40745f5 | 3624 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); |
1b76b316 | 3625 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 LT |
3626 | /* register to receive LACPDUs */ |
3627 | bond_register_lacpdu(bond); | |
fd989c83 | 3628 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3629 | } |
3630 | ||
3631 | return 0; | |
3632 | } | |
3633 | ||
3634 | static int bond_close(struct net_device *bond_dev) | |
3635 | { | |
454d7c9b | 3636 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3637 | |
3638 | if (bond->params.mode == BOND_MODE_8023AD) { | |
3639 | /* Unregister the receive of LACPDUs */ | |
3640 | bond_unregister_lacpdu(bond); | |
3641 | } | |
3642 | ||
f5b2b966 JV |
3643 | if (bond->params.arp_validate) |
3644 | bond_unregister_arp(bond); | |
3645 | ||
1da177e4 LT |
3646 | write_lock_bh(&bond->lock); |
3647 | ||
b59f9f74 | 3648 | bond->send_grat_arp = 0; |
305d552a | 3649 | bond->send_unsol_na = 0; |
1da177e4 LT |
3650 | |
3651 | /* signal timers not to re-arm */ | |
3652 | bond->kill_timers = 1; | |
3653 | ||
3654 | write_unlock_bh(&bond->lock); | |
3655 | ||
1da177e4 | 3656 | if (bond->params.miimon) { /* link check interval, in milliseconds. */ |
1b76b316 | 3657 | cancel_delayed_work(&bond->mii_work); |
1da177e4 LT |
3658 | } |
3659 | ||
3660 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3661 | cancel_delayed_work(&bond->arp_work); |
1da177e4 LT |
3662 | } |
3663 | ||
3664 | switch (bond->params.mode) { | |
3665 | case BOND_MODE_8023AD: | |
1b76b316 | 3666 | cancel_delayed_work(&bond->ad_work); |
1da177e4 LT |
3667 | break; |
3668 | case BOND_MODE_TLB: | |
3669 | case BOND_MODE_ALB: | |
1b76b316 | 3670 | cancel_delayed_work(&bond->alb_work); |
1da177e4 LT |
3671 | break; |
3672 | default: | |
3673 | break; | |
3674 | } | |
3675 | ||
5a37e8ca FL |
3676 | if (delayed_work_pending(&bond->mcast_work)) |
3677 | cancel_delayed_work(&bond->mcast_work); | |
1da177e4 | 3678 | |
58402054 | 3679 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3680 | /* Must be called only after all |
3681 | * slaves have been released | |
3682 | */ | |
3683 | bond_alb_deinitialize(bond); | |
3684 | } | |
3685 | ||
3686 | return 0; | |
3687 | } | |
3688 | ||
28172739 ED |
3689 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3690 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3691 | { |
454d7c9b | 3692 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3693 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3694 | struct slave *slave; |
3695 | int i; | |
3696 | ||
28172739 | 3697 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3698 | |
3699 | read_lock_bh(&bond->lock); | |
3700 | ||
3701 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3702 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3703 | dev_get_stats(slave->dev, &temp); |
3704 | ||
3705 | stats->rx_packets += sstats->rx_packets; | |
3706 | stats->rx_bytes += sstats->rx_bytes; | |
3707 | stats->rx_errors += sstats->rx_errors; | |
3708 | stats->rx_dropped += sstats->rx_dropped; | |
3709 | ||
3710 | stats->tx_packets += sstats->tx_packets; | |
3711 | stats->tx_bytes += sstats->tx_bytes; | |
3712 | stats->tx_errors += sstats->tx_errors; | |
3713 | stats->tx_dropped += sstats->tx_dropped; | |
3714 | ||
3715 | stats->multicast += sstats->multicast; | |
3716 | stats->collisions += sstats->collisions; | |
3717 | ||
3718 | stats->rx_length_errors += sstats->rx_length_errors; | |
3719 | stats->rx_over_errors += sstats->rx_over_errors; | |
3720 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3721 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3722 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3723 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3724 | ||
3725 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3726 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3727 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3728 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3729 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3730 | } |
3731 | ||
1da177e4 LT |
3732 | read_unlock_bh(&bond->lock); |
3733 | ||
3734 | return stats; | |
3735 | } | |
3736 | ||
3737 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3738 | { | |
3739 | struct net_device *slave_dev = NULL; | |
3740 | struct ifbond k_binfo; | |
3741 | struct ifbond __user *u_binfo = NULL; | |
3742 | struct ifslave k_sinfo; | |
3743 | struct ifslave __user *u_sinfo = NULL; | |
3744 | struct mii_ioctl_data *mii = NULL; | |
1da177e4 LT |
3745 | int res = 0; |
3746 | ||
a4aee5c8 | 3747 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3748 | |
3749 | switch (cmd) { | |
1da177e4 LT |
3750 | case SIOCGMIIPHY: |
3751 | mii = if_mii(ifr); | |
3d632c3f | 3752 | if (!mii) |
1da177e4 | 3753 | return -EINVAL; |
3d632c3f | 3754 | |
1da177e4 LT |
3755 | mii->phy_id = 0; |
3756 | /* Fall Through */ | |
3757 | case SIOCGMIIREG: | |
3758 | /* | |
3759 | * We do this again just in case we were called by SIOCGMIIREG | |
3760 | * instead of SIOCGMIIPHY. | |
3761 | */ | |
3762 | mii = if_mii(ifr); | |
3d632c3f | 3763 | if (!mii) |
1da177e4 | 3764 | return -EINVAL; |
3d632c3f | 3765 | |
1da177e4 LT |
3766 | |
3767 | if (mii->reg_num == 1) { | |
454d7c9b | 3768 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3769 | mii->val_out = 0; |
6603a6f2 | 3770 | read_lock(&bond->lock); |
1da177e4 | 3771 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3772 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3773 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3774 | |
1da177e4 | 3775 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3776 | read_unlock(&bond->lock); |
1da177e4 LT |
3777 | } |
3778 | ||
3779 | return 0; | |
3780 | case BOND_INFO_QUERY_OLD: | |
3781 | case SIOCBONDINFOQUERY: | |
3782 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3783 | ||
3d632c3f | 3784 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3785 | return -EFAULT; |
1da177e4 LT |
3786 | |
3787 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3788 | if (res == 0 && |
3789 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3790 | return -EFAULT; | |
1da177e4 LT |
3791 | |
3792 | return res; | |
3793 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3794 | case SIOCBONDSLAVEINFOQUERY: | |
3795 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3796 | ||
3d632c3f | 3797 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3798 | return -EFAULT; |
1da177e4 LT |
3799 | |
3800 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3801 | if (res == 0 && |
3802 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3803 | return -EFAULT; | |
1da177e4 LT |
3804 | |
3805 | return res; | |
3806 | default: | |
3807 | /* Go on */ | |
3808 | break; | |
3809 | } | |
3810 | ||
3d632c3f | 3811 | if (!capable(CAP_NET_ADMIN)) |
1da177e4 | 3812 | return -EPERM; |
1da177e4 | 3813 | |
ec87fd3b | 3814 | slave_dev = dev_get_by_name(dev_net(bond_dev), ifr->ifr_slave); |
1da177e4 | 3815 | |
a4aee5c8 | 3816 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3817 | |
3d632c3f | 3818 | if (!slave_dev) |
1da177e4 | 3819 | res = -ENODEV; |
3d632c3f | 3820 | else { |
a4aee5c8 | 3821 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3822 | switch (cmd) { |
3823 | case BOND_ENSLAVE_OLD: | |
3824 | case SIOCBONDENSLAVE: | |
3825 | res = bond_enslave(bond_dev, slave_dev); | |
3826 | break; | |
3827 | case BOND_RELEASE_OLD: | |
3828 | case SIOCBONDRELEASE: | |
3829 | res = bond_release(bond_dev, slave_dev); | |
3830 | break; | |
3831 | case BOND_SETHWADDR_OLD: | |
3832 | case SIOCBONDSETHWADDR: | |
3833 | res = bond_sethwaddr(bond_dev, slave_dev); | |
3834 | break; | |
3835 | case BOND_CHANGE_ACTIVE_OLD: | |
3836 | case SIOCBONDCHANGEACTIVE: | |
3837 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3838 | break; | |
3839 | default: | |
3840 | res = -EOPNOTSUPP; | |
3841 | } | |
3842 | ||
3843 | dev_put(slave_dev); | |
3844 | } | |
3845 | ||
1da177e4 LT |
3846 | return res; |
3847 | } | |
3848 | ||
22bedad3 JP |
3849 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3850 | struct netdev_hw_addr_list *list, | |
3851 | int addrlen) | |
3852 | { | |
3853 | struct netdev_hw_addr *ha; | |
3854 | ||
3855 | netdev_hw_addr_list_for_each(ha, list) | |
3856 | if (!memcmp(ha->addr, addr, addrlen)) | |
3857 | return true; | |
3858 | ||
3859 | return false; | |
3860 | } | |
3861 | ||
1da177e4 LT |
3862 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3863 | { | |
454d7c9b | 3864 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 JP |
3865 | struct netdev_hw_addr *ha; |
3866 | bool found; | |
1da177e4 | 3867 | |
1da177e4 LT |
3868 | /* |
3869 | * Do promisc before checking multicast_mode | |
3870 | */ | |
3d632c3f | 3871 | if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) |
7e1a1ac1 WC |
3872 | /* |
3873 | * FIXME: Need to handle the error when one of the multi-slaves | |
3874 | * encounters error. | |
3875 | */ | |
1da177e4 | 3876 | bond_set_promiscuity(bond, 1); |
1da177e4 | 3877 | |
3d632c3f SH |
3878 | |
3879 | if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) | |
1da177e4 | 3880 | bond_set_promiscuity(bond, -1); |
3d632c3f | 3881 | |
1da177e4 LT |
3882 | |
3883 | /* set allmulti flag to slaves */ | |
3d632c3f | 3884 | if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) |
7e1a1ac1 WC |
3885 | /* |
3886 | * FIXME: Need to handle the error when one of the multi-slaves | |
3887 | * encounters error. | |
3888 | */ | |
1da177e4 | 3889 | bond_set_allmulti(bond, 1); |
1da177e4 | 3890 | |
3d632c3f SH |
3891 | |
3892 | if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) | |
1da177e4 | 3893 | bond_set_allmulti(bond, -1); |
3d632c3f | 3894 | |
1da177e4 | 3895 | |
80ee5ad2 JV |
3896 | read_lock(&bond->lock); |
3897 | ||
1da177e4 LT |
3898 | bond->flags = bond_dev->flags; |
3899 | ||
3900 | /* looking for addresses to add to slaves' mc list */ | |
22bedad3 JP |
3901 | netdev_for_each_mc_addr(ha, bond_dev) { |
3902 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
3903 | bond_dev->addr_len); | |
3904 | if (!found) | |
3905 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
3906 | } |
3907 | ||
3908 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
3909 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
3910 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
3911 | bond_dev->addr_len); | |
3912 | if (!found) | |
3913 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
3914 | } |
3915 | ||
3916 | /* save master's multicast list */ | |
22bedad3 JP |
3917 | __hw_addr_flush(&bond->mc_list); |
3918 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
3919 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 3920 | |
80ee5ad2 | 3921 | read_unlock(&bond->lock); |
1da177e4 LT |
3922 | } |
3923 | ||
00829823 SH |
3924 | static int bond_neigh_setup(struct net_device *dev, struct neigh_parms *parms) |
3925 | { | |
3926 | struct bonding *bond = netdev_priv(dev); | |
3927 | struct slave *slave = bond->first_slave; | |
3928 | ||
3929 | if (slave) { | |
3930 | const struct net_device_ops *slave_ops | |
3931 | = slave->dev->netdev_ops; | |
3932 | if (slave_ops->ndo_neigh_setup) | |
72e2240f | 3933 | return slave_ops->ndo_neigh_setup(slave->dev, parms); |
00829823 SH |
3934 | } |
3935 | return 0; | |
3936 | } | |
3937 | ||
1da177e4 LT |
3938 | /* |
3939 | * Change the MTU of all of a master's slaves to match the master | |
3940 | */ | |
3941 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3942 | { | |
454d7c9b | 3943 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3944 | struct slave *slave, *stop_at; |
3945 | int res = 0; | |
3946 | int i; | |
3947 | ||
5a03cdb7 | 3948 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3949 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3950 | |
3951 | /* Can't hold bond->lock with bh disabled here since | |
3952 | * some base drivers panic. On the other hand we can't | |
3953 | * hold bond->lock without bh disabled because we'll | |
3954 | * deadlock. The only solution is to rely on the fact | |
3955 | * that we're under rtnl_lock here, and the slaves | |
3956 | * list won't change. This doesn't solve the problem | |
3957 | * of setting the slave's MTU while it is | |
3958 | * transmitting, but the assumption is that the base | |
3959 | * driver can handle that. | |
3960 | * | |
3961 | * TODO: figure out a way to safely iterate the slaves | |
3962 | * list, but without holding a lock around the actual | |
3963 | * call to the base driver. | |
3964 | */ | |
3965 | ||
3966 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3967 | pr_debug("s %p s->p %p c_m %p\n", |
3968 | slave, | |
3969 | slave->prev, | |
3970 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3971 | |
1da177e4 LT |
3972 | res = dev_set_mtu(slave->dev, new_mtu); |
3973 | ||
3974 | if (res) { | |
3975 | /* If we failed to set the slave's mtu to the new value | |
3976 | * we must abort the operation even in ACTIVE_BACKUP | |
3977 | * mode, because if we allow the backup slaves to have | |
3978 | * different mtu values than the active slave we'll | |
3979 | * need to change their mtu when doing a failover. That | |
3980 | * means changing their mtu from timer context, which | |
3981 | * is probably not a good idea. | |
3982 | */ | |
5a03cdb7 | 3983 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3984 | goto unwind; |
3985 | } | |
3986 | } | |
3987 | ||
3988 | bond_dev->mtu = new_mtu; | |
3989 | ||
3990 | return 0; | |
3991 | ||
3992 | unwind: | |
3993 | /* unwind from head to the slave that failed */ | |
3994 | stop_at = slave; | |
3995 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3996 | int tmp_res; | |
3997 | ||
3998 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3999 | if (tmp_res) { | |
a4aee5c8 JP |
4000 | pr_debug("unwind err %d dev %s\n", |
4001 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4002 | } |
4003 | } | |
4004 | ||
4005 | return res; | |
4006 | } | |
4007 | ||
4008 | /* | |
4009 | * Change HW address | |
4010 | * | |
4011 | * Note that many devices must be down to change the HW address, and | |
4012 | * downing the master releases all slaves. We can make bonds full of | |
4013 | * bonding devices to test this, however. | |
4014 | */ | |
4015 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
4016 | { | |
454d7c9b | 4017 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4018 | struct sockaddr *sa = addr, tmp_sa; |
4019 | struct slave *slave, *stop_at; | |
4020 | int res = 0; | |
4021 | int i; | |
4022 | ||
eb7cc59a SH |
4023 | if (bond->params.mode == BOND_MODE_ALB) |
4024 | return bond_alb_set_mac_address(bond_dev, addr); | |
4025 | ||
4026 | ||
a4aee5c8 JP |
4027 | pr_debug("bond=%p, name=%s\n", |
4028 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 4029 | |
dd957c57 | 4030 | /* |
3915c1e8 JV |
4031 | * If fail_over_mac is set to active, do nothing and return |
4032 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 4033 | */ |
3915c1e8 | 4034 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 4035 | return 0; |
2ab82852 | 4036 | |
3d632c3f | 4037 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 4038 | return -EADDRNOTAVAIL; |
1da177e4 LT |
4039 | |
4040 | /* Can't hold bond->lock with bh disabled here since | |
4041 | * some base drivers panic. On the other hand we can't | |
4042 | * hold bond->lock without bh disabled because we'll | |
4043 | * deadlock. The only solution is to rely on the fact | |
4044 | * that we're under rtnl_lock here, and the slaves | |
4045 | * list won't change. This doesn't solve the problem | |
4046 | * of setting the slave's hw address while it is | |
4047 | * transmitting, but the assumption is that the base | |
4048 | * driver can handle that. | |
4049 | * | |
4050 | * TODO: figure out a way to safely iterate the slaves | |
4051 | * list, but without holding a lock around the actual | |
4052 | * call to the base driver. | |
4053 | */ | |
4054 | ||
4055 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 4056 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 4057 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 4058 | |
eb7cc59a | 4059 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 4060 | res = -EOPNOTSUPP; |
5a03cdb7 | 4061 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
4062 | goto unwind; |
4063 | } | |
4064 | ||
4065 | res = dev_set_mac_address(slave->dev, addr); | |
4066 | if (res) { | |
4067 | /* TODO: consider downing the slave | |
4068 | * and retry ? | |
4069 | * User should expect communications | |
4070 | * breakage anyway until ARP finish | |
4071 | * updating, so... | |
4072 | */ | |
5a03cdb7 | 4073 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
4074 | goto unwind; |
4075 | } | |
4076 | } | |
4077 | ||
4078 | /* success */ | |
4079 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
4080 | return 0; | |
4081 | ||
4082 | unwind: | |
4083 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
4084 | tmp_sa.sa_family = bond_dev->type; | |
4085 | ||
4086 | /* unwind from head to the slave that failed */ | |
4087 | stop_at = slave; | |
4088 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
4089 | int tmp_res; | |
4090 | ||
4091 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
4092 | if (tmp_res) { | |
a4aee5c8 JP |
4093 | pr_debug("unwind err %d dev %s\n", |
4094 | tmp_res, slave->dev->name); | |
1da177e4 LT |
4095 | } |
4096 | } | |
4097 | ||
4098 | return res; | |
4099 | } | |
4100 | ||
4101 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
4102 | { | |
454d7c9b | 4103 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4104 | struct slave *slave, *start_at; |
cf5f9044 | 4105 | int i, slave_no, res = 1; |
a2fd940f | 4106 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 LT |
4107 | |
4108 | read_lock(&bond->lock); | |
4109 | ||
3d632c3f | 4110 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4111 | goto out; |
cf5f9044 | 4112 | /* |
a2fd940f AG |
4113 | * Start with the curr_active_slave that joined the bond as the |
4114 | * default for sending IGMP traffic. For failover purposes one | |
4115 | * needs to maintain some consistency for the interface that will | |
4116 | * send the join/membership reports. The curr_active_slave found | |
4117 | * will send all of this type of traffic. | |
cf5f9044 | 4118 | */ |
00ae7028 | 4119 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 4120 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 4121 | |
a2fd940f AG |
4122 | read_lock(&bond->curr_slave_lock); |
4123 | slave = bond->curr_active_slave; | |
4124 | read_unlock(&bond->curr_slave_lock); | |
4125 | ||
4126 | if (!slave) | |
4127 | goto out; | |
4128 | } else { | |
4129 | /* | |
4130 | * Concurrent TX may collide on rr_tx_counter; we accept | |
4131 | * that as being rare enough not to justify using an | |
4132 | * atomic op here. | |
4133 | */ | |
4134 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
4135 | ||
4136 | bond_for_each_slave(bond, slave, i) { | |
4137 | slave_no--; | |
4138 | if (slave_no < 0) | |
4139 | break; | |
4140 | } | |
1da177e4 LT |
4141 | } |
4142 | ||
cf5f9044 | 4143 | start_at = slave; |
1da177e4 LT |
4144 | bond_for_each_slave_from(bond, slave, i, start_at) { |
4145 | if (IS_UP(slave->dev) && | |
4146 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4147 | bond_is_active_slave(slave)) { |
1da177e4 | 4148 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
4149 | break; |
4150 | } | |
4151 | } | |
4152 | ||
1da177e4 LT |
4153 | out: |
4154 | if (res) { | |
4155 | /* no suitable interface, frame not sent */ | |
4156 | dev_kfree_skb(skb); | |
4157 | } | |
4158 | read_unlock(&bond->lock); | |
ec634fe3 | 4159 | return NETDEV_TX_OK; |
1da177e4 LT |
4160 | } |
4161 | ||
075897ce | 4162 | |
1da177e4 LT |
4163 | /* |
4164 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
4165 | * the bond has a usable interface. | |
4166 | */ | |
4167 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
4168 | { | |
454d7c9b | 4169 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4170 | int res = 1; |
4171 | ||
1da177e4 LT |
4172 | read_lock(&bond->lock); |
4173 | read_lock(&bond->curr_slave_lock); | |
4174 | ||
3d632c3f | 4175 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4176 | goto out; |
1da177e4 | 4177 | |
075897ce JL |
4178 | if (!bond->curr_active_slave) |
4179 | goto out; | |
4180 | ||
075897ce JL |
4181 | res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev); |
4182 | ||
1da177e4 | 4183 | out: |
3d632c3f | 4184 | if (res) |
1da177e4 LT |
4185 | /* no suitable interface, frame not sent */ |
4186 | dev_kfree_skb(skb); | |
3d632c3f | 4187 | |
1da177e4 LT |
4188 | read_unlock(&bond->curr_slave_lock); |
4189 | read_unlock(&bond->lock); | |
ec634fe3 | 4190 | return NETDEV_TX_OK; |
1da177e4 LT |
4191 | } |
4192 | ||
4193 | /* | |
169a3e66 JV |
4194 | * In bond_xmit_xor() , we determine the output device by using a pre- |
4195 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
4196 | * find the next active slave. | |
1da177e4 LT |
4197 | */ |
4198 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
4199 | { | |
454d7c9b | 4200 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4201 | struct slave *slave, *start_at; |
4202 | int slave_no; | |
4203 | int i; | |
4204 | int res = 1; | |
4205 | ||
4206 | read_lock(&bond->lock); | |
4207 | ||
3d632c3f | 4208 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4209 | goto out; |
1da177e4 | 4210 | |
a361c83c | 4211 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
4212 | |
4213 | bond_for_each_slave(bond, slave, i) { | |
4214 | slave_no--; | |
3d632c3f | 4215 | if (slave_no < 0) |
1da177e4 | 4216 | break; |
1da177e4 LT |
4217 | } |
4218 | ||
4219 | start_at = slave; | |
4220 | ||
4221 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4222 | if (IS_UP(slave->dev) && | |
4223 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4224 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4225 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
4226 | break; | |
4227 | } | |
4228 | } | |
4229 | ||
4230 | out: | |
4231 | if (res) { | |
4232 | /* no suitable interface, frame not sent */ | |
4233 | dev_kfree_skb(skb); | |
4234 | } | |
4235 | read_unlock(&bond->lock); | |
ec634fe3 | 4236 | return NETDEV_TX_OK; |
1da177e4 LT |
4237 | } |
4238 | ||
4239 | /* | |
4240 | * in broadcast mode, we send everything to all usable interfaces. | |
4241 | */ | |
4242 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4243 | { | |
454d7c9b | 4244 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4245 | struct slave *slave, *start_at; |
4246 | struct net_device *tx_dev = NULL; | |
4247 | int i; | |
4248 | int res = 1; | |
4249 | ||
4250 | read_lock(&bond->lock); | |
4251 | ||
3d632c3f | 4252 | if (!BOND_IS_OK(bond)) |
1da177e4 | 4253 | goto out; |
1da177e4 LT |
4254 | |
4255 | read_lock(&bond->curr_slave_lock); | |
4256 | start_at = bond->curr_active_slave; | |
4257 | read_unlock(&bond->curr_slave_lock); | |
4258 | ||
3d632c3f | 4259 | if (!start_at) |
1da177e4 | 4260 | goto out; |
1da177e4 LT |
4261 | |
4262 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4263 | if (IS_UP(slave->dev) && | |
4264 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4265 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4266 | if (tx_dev) { |
4267 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4268 | if (!skb2) { | |
a4aee5c8 | 4269 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4270 | bond_dev->name); |
1da177e4 LT |
4271 | continue; |
4272 | } | |
4273 | ||
4274 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4275 | if (res) { | |
4276 | dev_kfree_skb(skb2); | |
4277 | continue; | |
4278 | } | |
4279 | } | |
4280 | tx_dev = slave->dev; | |
4281 | } | |
4282 | } | |
4283 | ||
3d632c3f | 4284 | if (tx_dev) |
1da177e4 | 4285 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4286 | |
4287 | out: | |
3d632c3f | 4288 | if (res) |
1da177e4 LT |
4289 | /* no suitable interface, frame not sent */ |
4290 | dev_kfree_skb(skb); | |
3d632c3f | 4291 | |
1da177e4 LT |
4292 | /* frame sent to all suitable interfaces */ |
4293 | read_unlock(&bond->lock); | |
ec634fe3 | 4294 | return NETDEV_TX_OK; |
1da177e4 LT |
4295 | } |
4296 | ||
4297 | /*------------------------- Device initialization ---------------------------*/ | |
4298 | ||
6f6652be JV |
4299 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4300 | { | |
4301 | switch (bond->params.xmit_policy) { | |
4302 | case BOND_XMIT_POLICY_LAYER23: | |
4303 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4304 | break; | |
4305 | case BOND_XMIT_POLICY_LAYER34: | |
4306 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4307 | break; | |
4308 | case BOND_XMIT_POLICY_LAYER2: | |
4309 | default: | |
4310 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4311 | break; | |
4312 | } | |
4313 | } | |
4314 | ||
bb1d9123 AG |
4315 | /* |
4316 | * Lookup the slave that corresponds to a qid | |
4317 | */ | |
4318 | static inline int bond_slave_override(struct bonding *bond, | |
4319 | struct sk_buff *skb) | |
4320 | { | |
4321 | int i, res = 1; | |
4322 | struct slave *slave = NULL; | |
4323 | struct slave *check_slave; | |
4324 | ||
4325 | read_lock(&bond->lock); | |
4326 | ||
4327 | if (!BOND_IS_OK(bond) || !skb->queue_mapping) | |
4328 | goto out; | |
4329 | ||
4330 | /* Find out if any slaves have the same mapping as this skb. */ | |
4331 | bond_for_each_slave(bond, check_slave, i) { | |
4332 | if (check_slave->queue_id == skb->queue_mapping) { | |
4333 | slave = check_slave; | |
4334 | break; | |
4335 | } | |
4336 | } | |
4337 | ||
4338 | /* If the slave isn't UP, use default transmit policy. */ | |
4339 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4340 | (slave->link == BOND_LINK_UP)) { | |
4341 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4342 | } | |
4343 | ||
4344 | out: | |
4345 | read_unlock(&bond->lock); | |
4346 | return res; | |
4347 | } | |
4348 | ||
4349 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) | |
4350 | { | |
4351 | /* | |
4352 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4353 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4354 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4355 | * way down to the bonding driver. | |
4356 | */ | |
fd0e435b PO |
4357 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4358 | ||
4359 | if (unlikely(txq >= dev->real_num_tx_queues)) { | |
d30ee670 | 4360 | do { |
fd0e435b | 4361 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4362 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4363 | } |
4364 | return txq; | |
bb1d9123 AG |
4365 | } |
4366 | ||
424efe9c | 4367 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4368 | { |
bb1d9123 AG |
4369 | struct bonding *bond = netdev_priv(dev); |
4370 | ||
e843fa50 NH |
4371 | /* |
4372 | * If we risk deadlock from transmitting this in the | |
4373 | * netpoll path, tell netpoll to queue the frame for later tx | |
4374 | */ | |
4375 | if (is_netpoll_tx_blocked(dev)) | |
4376 | return NETDEV_TX_BUSY; | |
4377 | ||
bb1d9123 AG |
4378 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { |
4379 | if (!bond_slave_override(bond, skb)) | |
4380 | return NETDEV_TX_OK; | |
4381 | } | |
00829823 SH |
4382 | |
4383 | switch (bond->params.mode) { | |
4384 | case BOND_MODE_ROUNDROBIN: | |
4385 | return bond_xmit_roundrobin(skb, dev); | |
4386 | case BOND_MODE_ACTIVEBACKUP: | |
4387 | return bond_xmit_activebackup(skb, dev); | |
4388 | case BOND_MODE_XOR: | |
4389 | return bond_xmit_xor(skb, dev); | |
4390 | case BOND_MODE_BROADCAST: | |
4391 | return bond_xmit_broadcast(skb, dev); | |
4392 | case BOND_MODE_8023AD: | |
4393 | return bond_3ad_xmit_xor(skb, dev); | |
4394 | case BOND_MODE_ALB: | |
4395 | case BOND_MODE_TLB: | |
4396 | return bond_alb_xmit(skb, dev); | |
4397 | default: | |
4398 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4399 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4400 | dev->name, bond->params.mode); | |
00829823 SH |
4401 | WARN_ON_ONCE(1); |
4402 | dev_kfree_skb(skb); | |
4403 | return NETDEV_TX_OK; | |
4404 | } | |
4405 | } | |
4406 | ||
4407 | ||
1da177e4 LT |
4408 | /* |
4409 | * set bond mode specific net device operations | |
4410 | */ | |
a77b5325 | 4411 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4412 | { |
169a3e66 JV |
4413 | struct net_device *bond_dev = bond->dev; |
4414 | ||
1da177e4 LT |
4415 | switch (mode) { |
4416 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4417 | break; |
4418 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4419 | break; |
4420 | case BOND_MODE_XOR: | |
6f6652be | 4421 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4422 | break; |
4423 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4424 | break; |
4425 | case BOND_MODE_8023AD: | |
6f6652be | 4426 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4427 | break; |
1da177e4 | 4428 | case BOND_MODE_ALB: |
8f903c70 JV |
4429 | /* FALLTHRU */ |
4430 | case BOND_MODE_TLB: | |
1da177e4 LT |
4431 | break; |
4432 | default: | |
4433 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4434 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4435 | bond_dev->name, mode); | |
1da177e4 LT |
4436 | break; |
4437 | } | |
4438 | } | |
4439 | ||
217df670 JV |
4440 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
4441 | struct ethtool_drvinfo *drvinfo) | |
4442 | { | |
4443 | strncpy(drvinfo->driver, DRV_NAME, 32); | |
4444 | strncpy(drvinfo->version, DRV_VERSION, 32); | |
4445 | snprintf(drvinfo->fw_version, 32, "%d", BOND_ABI_VERSION); | |
4446 | } | |
4447 | ||
7282d491 | 4448 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4449 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac SH |
4450 | .get_link = ethtool_op_get_link, |
4451 | .get_tx_csum = ethtool_op_get_tx_csum, | |
4452 | .get_sg = ethtool_op_get_sg, | |
4453 | .get_tso = ethtool_op_get_tso, | |
4454 | .get_ufo = ethtool_op_get_ufo, | |
4455 | .get_flags = ethtool_op_get_flags, | |
8531c5ff AK |
4456 | }; |
4457 | ||
eb7cc59a | 4458 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4459 | .ndo_init = bond_init, |
9e71626c | 4460 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4461 | .ndo_open = bond_open, |
4462 | .ndo_stop = bond_close, | |
00829823 | 4463 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4464 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4465 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a SH |
4466 | .ndo_do_ioctl = bond_do_ioctl, |
4467 | .ndo_set_multicast_list = bond_set_multicast_list, | |
4468 | .ndo_change_mtu = bond_change_mtu, | |
eb7cc59a | 4469 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4470 | .ndo_neigh_setup = bond_neigh_setup, |
eb7cc59a SH |
4471 | .ndo_vlan_rx_register = bond_vlan_rx_register, |
4472 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, | |
4473 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, | |
f6dc31a8 | 4474 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4475 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4476 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4477 | .ndo_poll_controller = bond_poll_controller, | |
4478 | #endif | |
9232ecca JP |
4479 | .ndo_add_slave = bond_enslave, |
4480 | .ndo_del_slave = bond_release, | |
eb7cc59a SH |
4481 | }; |
4482 | ||
9e2e61fb AW |
4483 | static void bond_destructor(struct net_device *bond_dev) |
4484 | { | |
4485 | struct bonding *bond = netdev_priv(bond_dev); | |
4486 | if (bond->wq) | |
4487 | destroy_workqueue(bond->wq); | |
4488 | free_netdev(bond_dev); | |
4489 | } | |
4490 | ||
181470fc | 4491 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4492 | { |
454d7c9b | 4493 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4494 | |
1da177e4 LT |
4495 | /* initialize rwlocks */ |
4496 | rwlock_init(&bond->lock); | |
4497 | rwlock_init(&bond->curr_slave_lock); | |
4498 | ||
d2991f75 | 4499 | bond->params = bonding_defaults; |
1da177e4 LT |
4500 | |
4501 | /* Initialize pointers */ | |
1da177e4 LT |
4502 | bond->dev = bond_dev; |
4503 | INIT_LIST_HEAD(&bond->vlan_list); | |
4504 | ||
4505 | /* Initialize the device entry points */ | |
181470fc | 4506 | ether_setup(bond_dev); |
eb7cc59a | 4507 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4508 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4509 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4510 | |
9e2e61fb | 4511 | bond_dev->destructor = bond_destructor; |
1da177e4 LT |
4512 | |
4513 | /* Initialize the device options */ | |
4514 | bond_dev->tx_queue_len = 0; | |
4515 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4516 | bond_dev->priv_flags |= IFF_BONDING; |
181470fc SH |
4517 | bond_dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
4518 | ||
1da177e4 LT |
4519 | /* At first, we block adding VLANs. That's the only way to |
4520 | * prevent problems that occur when adding VLANs over an | |
4521 | * empty bond. The block will be removed once non-challenged | |
4522 | * slaves are enslaved. | |
4523 | */ | |
4524 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4525 | ||
932ff279 | 4526 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4527 | * transmitting */ |
4528 | bond_dev->features |= NETIF_F_LLTX; | |
4529 | ||
4530 | /* By default, we declare the bond to be fully | |
4531 | * VLAN hardware accelerated capable. Special | |
4532 | * care is taken in the various xmit functions | |
4533 | * when there are slaves that are not hw accel | |
4534 | * capable | |
4535 | */ | |
1da177e4 LT |
4536 | bond_dev->features |= (NETIF_F_HW_VLAN_TX | |
4537 | NETIF_F_HW_VLAN_RX | | |
4538 | NETIF_F_HW_VLAN_FILTER); | |
4539 | ||
e6599c2e ED |
4540 | /* By default, we enable GRO on bonding devices. |
4541 | * Actual support requires lowlevel drivers are GRO ready. | |
4542 | */ | |
4543 | bond_dev->features |= NETIF_F_GRO; | |
1da177e4 LT |
4544 | } |
4545 | ||
fdaea7a9 JV |
4546 | static void bond_work_cancel_all(struct bonding *bond) |
4547 | { | |
4548 | write_lock_bh(&bond->lock); | |
4549 | bond->kill_timers = 1; | |
4550 | write_unlock_bh(&bond->lock); | |
4551 | ||
4552 | if (bond->params.miimon && delayed_work_pending(&bond->mii_work)) | |
4553 | cancel_delayed_work(&bond->mii_work); | |
4554 | ||
4555 | if (bond->params.arp_interval && delayed_work_pending(&bond->arp_work)) | |
4556 | cancel_delayed_work(&bond->arp_work); | |
4557 | ||
4558 | if (bond->params.mode == BOND_MODE_ALB && | |
4559 | delayed_work_pending(&bond->alb_work)) | |
4560 | cancel_delayed_work(&bond->alb_work); | |
4561 | ||
4562 | if (bond->params.mode == BOND_MODE_8023AD && | |
4563 | delayed_work_pending(&bond->ad_work)) | |
4564 | cancel_delayed_work(&bond->ad_work); | |
5a37e8ca FL |
4565 | |
4566 | if (delayed_work_pending(&bond->mcast_work)) | |
4567 | cancel_delayed_work(&bond->mcast_work); | |
fdaea7a9 JV |
4568 | } |
4569 | ||
c67dfb29 EB |
4570 | /* |
4571 | * Destroy a bonding device. | |
4572 | * Must be under rtnl_lock when this function is called. | |
4573 | */ | |
4574 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4575 | { |
454d7c9b | 4576 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4577 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4578 | |
f6dc31a8 WC |
4579 | bond_netpoll_cleanup(bond_dev); |
4580 | ||
c67dfb29 EB |
4581 | /* Release the bonded slaves */ |
4582 | bond_release_all(bond_dev); | |
4583 | ||
a434e43f JV |
4584 | list_del(&bond->bond_list); |
4585 | ||
4586 | bond_work_cancel_all(bond); | |
4587 | ||
a434e43f | 4588 | bond_remove_proc_entry(bond); |
c67dfb29 | 4589 | |
f073c7ca TI |
4590 | bond_debug_unregister(bond); |
4591 | ||
22bedad3 | 4592 | __hw_addr_flush(&bond->mc_list); |
f35188fa JV |
4593 | |
4594 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { | |
4595 | list_del(&vlan->vlan_list); | |
4596 | kfree(vlan); | |
4597 | } | |
a434e43f JV |
4598 | } |
4599 | ||
1da177e4 LT |
4600 | /*------------------------- Module initialization ---------------------------*/ |
4601 | ||
4602 | /* | |
4603 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4604 | * number of the mode or its string name. A bit complicated because |
4605 | * some mode names are substrings of other names, and calls from sysfs | |
4606 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4607 | */ |
325dcf7a | 4608 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4609 | { |
54b87323 | 4610 | int modeint = -1, i, rv; |
a42e534f | 4611 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4612 | |
a42e534f JV |
4613 | for (p = (char *)buf; *p; p++) |
4614 | if (!(isdigit(*p) || isspace(*p))) | |
4615 | break; | |
4616 | ||
4617 | if (*p) | |
ece95f7f | 4618 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4619 | else |
54b87323 | 4620 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4621 | |
4622 | if (!rv) | |
4623 | return -1; | |
1da177e4 LT |
4624 | |
4625 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4626 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4627 | return tbl[i].mode; |
4628 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4629 | return tbl[i].mode; |
1da177e4 LT |
4630 | } |
4631 | ||
4632 | return -1; | |
4633 | } | |
4634 | ||
4635 | static int bond_check_params(struct bond_params *params) | |
4636 | { | |
a549952a | 4637 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4638 | |
1da177e4 LT |
4639 | /* |
4640 | * Convert string parameters. | |
4641 | */ | |
4642 | if (mode) { | |
4643 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4644 | if (bond_mode == -1) { | |
a4aee5c8 | 4645 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4646 | mode == NULL ? "NULL" : mode); |
4647 | return -EINVAL; | |
4648 | } | |
4649 | } | |
4650 | ||
169a3e66 JV |
4651 | if (xmit_hash_policy) { |
4652 | if ((bond_mode != BOND_MODE_XOR) && | |
4653 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4654 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4655 | bond_mode_name(bond_mode)); |
4656 | } else { | |
4657 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4658 | xmit_hashtype_tbl); | |
4659 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4660 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4661 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4662 | xmit_hash_policy); |
4663 | return -EINVAL; | |
4664 | } | |
4665 | } | |
4666 | } | |
4667 | ||
1da177e4 LT |
4668 | if (lacp_rate) { |
4669 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4670 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4671 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4672 | } else { |
4673 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4674 | if (lacp_fast == -1) { | |
a4aee5c8 | 4675 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4676 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4677 | return -EINVAL; | |
4678 | } | |
4679 | } | |
4680 | } | |
4681 | ||
fd989c83 JV |
4682 | if (ad_select) { |
4683 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4684 | if (params->ad_select == -1) { | |
a4aee5c8 | 4685 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4686 | ad_select == NULL ? "NULL" : ad_select); |
4687 | return -EINVAL; | |
4688 | } | |
4689 | ||
4690 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4691 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4692 | } |
4693 | } else { | |
4694 | params->ad_select = BOND_AD_STABLE; | |
4695 | } | |
4696 | ||
f5841306 | 4697 | if (max_bonds < 0) { |
a4aee5c8 JP |
4698 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4699 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4700 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4701 | } | |
4702 | ||
4703 | if (miimon < 0) { | |
a4aee5c8 JP |
4704 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4705 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4706 | miimon = BOND_LINK_MON_INTERV; |
4707 | } | |
4708 | ||
4709 | if (updelay < 0) { | |
a4aee5c8 JP |
4710 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4711 | updelay, INT_MAX); | |
1da177e4 LT |
4712 | updelay = 0; |
4713 | } | |
4714 | ||
4715 | if (downdelay < 0) { | |
a4aee5c8 JP |
4716 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4717 | downdelay, INT_MAX); | |
1da177e4 LT |
4718 | downdelay = 0; |
4719 | } | |
4720 | ||
4721 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4722 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4723 | use_carrier); | |
1da177e4 LT |
4724 | use_carrier = 1; |
4725 | } | |
4726 | ||
7893b249 | 4727 | if (num_grat_arp < 0 || num_grat_arp > 255) { |
2381a55c | 4728 | pr_warning("Warning: num_grat_arp (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4729 | num_grat_arp); |
7893b249 MS |
4730 | num_grat_arp = 1; |
4731 | } | |
4732 | ||
305d552a | 4733 | if (num_unsol_na < 0 || num_unsol_na > 255) { |
2381a55c | 4734 | pr_warning("Warning: num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", |
a4aee5c8 | 4735 | num_unsol_na); |
305d552a BH |
4736 | num_unsol_na = 1; |
4737 | } | |
4738 | ||
1da177e4 LT |
4739 | /* reset values for 802.3ad */ |
4740 | if (bond_mode == BOND_MODE_8023AD) { | |
4741 | if (!miimon) { | |
a4aee5c8 | 4742 | 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 | 4743 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4744 | miimon = 100; |
4745 | } | |
4746 | } | |
4747 | ||
bb1d9123 AG |
4748 | if (tx_queues < 1 || tx_queues > 255) { |
4749 | pr_warning("Warning: tx_queues (%d) should be between " | |
4750 | "1 and 255, resetting to %d\n", | |
4751 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4752 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4753 | } | |
4754 | ||
ebd8e497 AG |
4755 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4756 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4757 | "not of valid value (0/1), so it was set to " | |
4758 | "0\n", all_slaves_active); | |
4759 | all_slaves_active = 0; | |
4760 | } | |
4761 | ||
c2952c31 FL |
4762 | if (resend_igmp < 0 || resend_igmp > 255) { |
4763 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4764 | "0 and 255, resetting to %d\n", | |
4765 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4766 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4767 | } | |
4768 | ||
1da177e4 LT |
4769 | /* reset values for TLB/ALB */ |
4770 | if ((bond_mode == BOND_MODE_TLB) || | |
4771 | (bond_mode == BOND_MODE_ALB)) { | |
4772 | if (!miimon) { | |
a4aee5c8 | 4773 | 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 | 4774 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4775 | miimon = 100; |
4776 | } | |
4777 | } | |
4778 | ||
4779 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4780 | 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", |
4781 | updelay); | |
1da177e4 LT |
4782 | } |
4783 | ||
4784 | if (!miimon) { | |
4785 | if (updelay || downdelay) { | |
4786 | /* just warn the user the up/down delay will have | |
4787 | * no effect since miimon is zero... | |
4788 | */ | |
a4aee5c8 JP |
4789 | 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", |
4790 | updelay, downdelay); | |
1da177e4 LT |
4791 | } |
4792 | } else { | |
4793 | /* don't allow arp monitoring */ | |
4794 | if (arp_interval) { | |
a4aee5c8 JP |
4795 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4796 | miimon, arp_interval); | |
1da177e4 LT |
4797 | arp_interval = 0; |
4798 | } | |
4799 | ||
4800 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4801 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4802 | updelay, miimon, | |
4803 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4804 | } |
4805 | ||
4806 | updelay /= miimon; | |
4807 | ||
4808 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4809 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4810 | downdelay, miimon, | |
4811 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4812 | } |
4813 | ||
4814 | downdelay /= miimon; | |
4815 | } | |
4816 | ||
4817 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4818 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4819 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4820 | arp_interval = BOND_LINK_ARP_INTERV; |
4821 | } | |
4822 | ||
4823 | for (arp_ip_count = 0; | |
4824 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4825 | arp_ip_count++) { | |
4826 | /* not complete check, but should be good enough to | |
4827 | catch mistakes */ | |
4828 | if (!isdigit(arp_ip_target[arp_ip_count][0])) { | |
a4aee5c8 JP |
4829 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4830 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4831 | arp_interval = 0; |
4832 | } else { | |
d3bb52b0 | 4833 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
1da177e4 LT |
4834 | arp_target[arp_ip_count] = ip; |
4835 | } | |
4836 | } | |
4837 | ||
4838 | if (arp_interval && !arp_ip_count) { | |
4839 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4840 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4841 | arp_interval); | |
1da177e4 LT |
4842 | arp_interval = 0; |
4843 | } | |
4844 | ||
f5b2b966 JV |
4845 | if (arp_validate) { |
4846 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4847 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4848 | return -EINVAL; |
4849 | } | |
4850 | if (!arp_interval) { | |
a4aee5c8 | 4851 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4852 | return -EINVAL; |
4853 | } | |
4854 | ||
4855 | arp_validate_value = bond_parse_parm(arp_validate, | |
4856 | arp_validate_tbl); | |
4857 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4858 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4859 | arp_validate == NULL ? "NULL" : arp_validate); |
4860 | return -EINVAL; | |
4861 | } | |
4862 | } else | |
4863 | arp_validate_value = 0; | |
4864 | ||
1da177e4 | 4865 | if (miimon) { |
a4aee5c8 | 4866 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4867 | } else if (arp_interval) { |
4868 | int i; | |
4869 | ||
a4aee5c8 JP |
4870 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4871 | arp_interval, | |
4872 | arp_validate_tbl[arp_validate_value].modename, | |
4873 | arp_ip_count); | |
1da177e4 LT |
4874 | |
4875 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4876 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4877 | |
e5e2a8fd | 4878 | pr_info("\n"); |
1da177e4 | 4879 | |
b8a9787e | 4880 | } else if (max_bonds) { |
1da177e4 LT |
4881 | /* miimon and arp_interval not set, we need one so things |
4882 | * work as expected, see bonding.txt for details | |
4883 | */ | |
a4aee5c8 | 4884 | 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 |
4885 | } |
4886 | ||
4887 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4888 | /* currently, using a primary only makes sense | |
4889 | * in active backup, TLB or ALB modes | |
4890 | */ | |
a4aee5c8 JP |
4891 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4892 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4893 | primary = NULL; |
4894 | } | |
4895 | ||
a549952a JP |
4896 | if (primary && primary_reselect) { |
4897 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4898 | pri_reselect_tbl); | |
4899 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4900 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4901 | primary_reselect == |
4902 | NULL ? "NULL" : primary_reselect); | |
4903 | return -EINVAL; | |
4904 | } | |
4905 | } else { | |
4906 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4907 | } | |
4908 | ||
3915c1e8 JV |
4909 | if (fail_over_mac) { |
4910 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4911 | fail_over_mac_tbl); | |
4912 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4913 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4914 | arp_validate == NULL ? "NULL" : arp_validate); |
4915 | return -EINVAL; | |
4916 | } | |
4917 | ||
4918 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4919 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4920 | } else { |
4921 | fail_over_mac_value = BOND_FOM_NONE; | |
4922 | } | |
dd957c57 | 4923 | |
1da177e4 LT |
4924 | /* fill params struct with the proper values */ |
4925 | params->mode = bond_mode; | |
169a3e66 | 4926 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4927 | params->miimon = miimon; |
7893b249 | 4928 | params->num_grat_arp = num_grat_arp; |
305d552a | 4929 | params->num_unsol_na = num_unsol_na; |
1da177e4 | 4930 | params->arp_interval = arp_interval; |
f5b2b966 | 4931 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4932 | params->updelay = updelay; |
4933 | params->downdelay = downdelay; | |
4934 | params->use_carrier = use_carrier; | |
4935 | params->lacp_fast = lacp_fast; | |
4936 | params->primary[0] = 0; | |
a549952a | 4937 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4938 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4939 | params->tx_queues = tx_queues; |
ebd8e497 | 4940 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4941 | params->resend_igmp = resend_igmp; |
1da177e4 LT |
4942 | |
4943 | if (primary) { | |
4944 | strncpy(params->primary, primary, IFNAMSIZ); | |
4945 | params->primary[IFNAMSIZ - 1] = 0; | |
4946 | } | |
4947 | ||
4948 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4949 | ||
4950 | return 0; | |
4951 | } | |
4952 | ||
0daa2303 | 4953 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4954 | static struct lock_class_key bonding_netdev_addr_lock_key; |
0daa2303 | 4955 | |
e8a0464c DM |
4956 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4957 | struct netdev_queue *txq, | |
4958 | void *_unused) | |
c773e847 DM |
4959 | { |
4960 | lockdep_set_class(&txq->_xmit_lock, | |
4961 | &bonding_netdev_xmit_lock_key); | |
4962 | } | |
4963 | ||
4964 | static void bond_set_lockdep_class(struct net_device *dev) | |
4965 | { | |
cf508b12 DM |
4966 | lockdep_set_class(&dev->addr_list_lock, |
4967 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4968 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
c773e847 DM |
4969 | } |
4970 | ||
181470fc SH |
4971 | /* |
4972 | * Called from registration process | |
4973 | */ | |
4974 | static int bond_init(struct net_device *bond_dev) | |
4975 | { | |
4976 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4977 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
181470fc SH |
4978 | |
4979 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4980 | ||
4981 | bond->wq = create_singlethread_workqueue(bond_dev->name); | |
4982 | if (!bond->wq) | |
4983 | return -ENOMEM; | |
4984 | ||
4985 | bond_set_lockdep_class(bond_dev); | |
4986 | ||
181470fc | 4987 | bond_create_proc_entry(bond); |
ec87fd3b | 4988 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4989 | |
6151b3d4 | 4990 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4991 | |
f073c7ca TI |
4992 | bond_debug_register(bond); |
4993 | ||
22bedad3 | 4994 | __hw_addr_init(&bond->mc_list); |
181470fc SH |
4995 | return 0; |
4996 | } | |
4997 | ||
88ead977 EB |
4998 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4999 | { | |
5000 | if (tb[IFLA_ADDRESS]) { | |
5001 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
5002 | return -EINVAL; | |
5003 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
5004 | return -EADDRNOTAVAIL; | |
5005 | } | |
5006 | return 0; | |
5007 | } | |
5008 | ||
5009 | static struct rtnl_link_ops bond_link_ops __read_mostly = { | |
5010 | .kind = "bond", | |
6639104b | 5011 | .priv_size = sizeof(struct bonding), |
88ead977 EB |
5012 | .setup = bond_setup, |
5013 | .validate = bond_validate, | |
5014 | }; | |
5015 | ||
dfe60397 | 5016 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 5017 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
5018 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
5019 | * set up our sysfs entries. | |
5020 | */ | |
ec87fd3b | 5021 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
5022 | { |
5023 | struct net_device *bond_dev; | |
5024 | int res; | |
5025 | ||
5026 | rtnl_lock(); | |
027ea041 | 5027 | |
bb1d9123 AG |
5028 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), name ? name : "", |
5029 | bond_setup, tx_queues); | |
dfe60397 | 5030 | if (!bond_dev) { |
a4aee5c8 | 5031 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
5032 | rtnl_unlock(); |
5033 | return -ENOMEM; | |
dfe60397 MW |
5034 | } |
5035 | ||
ec87fd3b | 5036 | dev_net_set(bond_dev, net); |
88ead977 EB |
5037 | bond_dev->rtnl_link_ops = &bond_link_ops; |
5038 | ||
e4b91c48 JV |
5039 | if (!name) { |
5040 | res = dev_alloc_name(bond_dev, "bond%d"); | |
5041 | if (res < 0) | |
9e2e61fb | 5042 | goto out; |
27e6f065 NH |
5043 | } else { |
5044 | /* | |
5045 | * If we're given a name to register | |
5046 | * we need to ensure that its not already | |
5047 | * registered | |
5048 | */ | |
5049 | res = -EEXIST; | |
5050 | if (__dev_get_by_name(net, name) != NULL) | |
5051 | goto out; | |
e4b91c48 JV |
5052 | } |
5053 | ||
dfe60397 | 5054 | res = register_netdevice(bond_dev); |
0daa2303 | 5055 | |
e826eafa PO |
5056 | netif_carrier_off(bond_dev); |
5057 | ||
30c15ba9 | 5058 | out: |
7e083840 | 5059 | rtnl_unlock(); |
9e2e61fb AW |
5060 | if (res < 0) |
5061 | bond_destructor(bond_dev); | |
30c15ba9 | 5062 | return res; |
dfe60397 MW |
5063 | } |
5064 | ||
2c8c1e72 | 5065 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 5066 | { |
15449745 | 5067 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5068 | |
5069 | bn->net = net; | |
5070 | INIT_LIST_HEAD(&bn->dev_list); | |
5071 | ||
ec87fd3b | 5072 | bond_create_proc_dir(bn); |
15449745 EB |
5073 | |
5074 | return 0; | |
ec87fd3b EB |
5075 | } |
5076 | ||
2c8c1e72 | 5077 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 5078 | { |
15449745 | 5079 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
5080 | |
5081 | bond_destroy_proc_dir(bn); | |
ec87fd3b EB |
5082 | } |
5083 | ||
5084 | static struct pernet_operations bond_net_ops = { | |
5085 | .init = bond_net_init, | |
5086 | .exit = bond_net_exit, | |
15449745 EB |
5087 | .id = &bond_net_id, |
5088 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
5089 | }; |
5090 | ||
1da177e4 LT |
5091 | static int __init bonding_init(void) |
5092 | { | |
1da177e4 LT |
5093 | int i; |
5094 | int res; | |
5095 | ||
bd33acc3 | 5096 | pr_info("%s", bond_version); |
1da177e4 | 5097 | |
dfe60397 | 5098 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 5099 | if (res) |
dfe60397 | 5100 | goto out; |
1da177e4 | 5101 | |
15449745 | 5102 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
5103 | if (res) |
5104 | goto out; | |
027ea041 | 5105 | |
88ead977 EB |
5106 | res = rtnl_link_register(&bond_link_ops); |
5107 | if (res) | |
6639104b | 5108 | goto err_link; |
88ead977 | 5109 | |
f073c7ca TI |
5110 | bond_create_debugfs(); |
5111 | ||
1da177e4 | 5112 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 5113 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
5114 | if (res) |
5115 | goto err; | |
1da177e4 LT |
5116 | } |
5117 | ||
b76cdba9 MW |
5118 | res = bond_create_sysfs(); |
5119 | if (res) | |
5120 | goto err; | |
5121 | ||
1da177e4 | 5122 | register_netdevice_notifier(&bond_netdev_notifier); |
c3ade5ca | 5123 | register_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5124 | bond_register_ipv6_notifier(); |
dfe60397 | 5125 | out: |
1da177e4 | 5126 | return res; |
88ead977 EB |
5127 | err: |
5128 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 5129 | err_link: |
15449745 | 5130 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 5131 | goto out; |
dfe60397 | 5132 | |
1da177e4 LT |
5133 | } |
5134 | ||
5135 | static void __exit bonding_exit(void) | |
5136 | { | |
5137 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
c3ade5ca | 5138 | unregister_inetaddr_notifier(&bond_inetaddr_notifier); |
305d552a | 5139 | bond_unregister_ipv6_notifier(); |
1da177e4 | 5140 | |
ae68c398 | 5141 | bond_destroy_sysfs(); |
f073c7ca | 5142 | bond_destroy_debugfs(); |
ae68c398 | 5143 | |
88ead977 | 5144 | rtnl_link_unregister(&bond_link_ops); |
15449745 | 5145 | unregister_pernet_subsys(&bond_net_ops); |
e843fa50 NH |
5146 | |
5147 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
5148 | /* |
5149 | * Make sure we don't have an imbalance on our netpoll blocking | |
5150 | */ | |
5151 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 5152 | #endif |
1da177e4 LT |
5153 | } |
5154 | ||
5155 | module_init(bonding_init); | |
5156 | module_exit(bonding_exit); | |
5157 | MODULE_LICENSE("GPL"); | |
5158 | MODULE_VERSION(DRV_VERSION); | |
5159 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
5160 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 5161 | MODULE_ALIAS_RTNL_LINK("bond"); |