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/dma.h> |
3d632c3f | 58 | #include <linux/uaccess.h> |
1da177e4 LT |
59 | #include <linux/errno.h> |
60 | #include <linux/netdevice.h> | |
61 | #include <linux/inetdevice.h> | |
a816c7c7 | 62 | #include <linux/igmp.h> |
1da177e4 LT |
63 | #include <linux/etherdevice.h> |
64 | #include <linux/skbuff.h> | |
65 | #include <net/sock.h> | |
66 | #include <linux/rtnetlink.h> | |
1da177e4 LT |
67 | #include <linux/smp.h> |
68 | #include <linux/if_ether.h> | |
69 | #include <net/arp.h> | |
70 | #include <linux/mii.h> | |
71 | #include <linux/ethtool.h> | |
72 | #include <linux/if_vlan.h> | |
73 | #include <linux/if_bonding.h> | |
b63bb739 | 74 | #include <linux/jiffies.h> |
e843fa50 | 75 | #include <linux/preempt.h> |
c3ade5ca | 76 | #include <net/route.h> |
457c4cbc | 77 | #include <net/net_namespace.h> |
ec87fd3b | 78 | #include <net/netns/generic.h> |
5ee31c68 | 79 | #include <net/pkt_sched.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; |
ad246c99 | 92 | static int num_peer_notif = 1; |
1da177e4 | 93 | static int miimon = BOND_LINK_MON_INTERV; |
3d632c3f SH |
94 | static int updelay; |
95 | static int downdelay; | |
1da177e4 | 96 | static int use_carrier = 1; |
3d632c3f SH |
97 | static char *mode; |
98 | static char *primary; | |
a549952a | 99 | static char *primary_reselect; |
3d632c3f | 100 | static char *lacp_rate; |
655f8919 | 101 | static int min_links; |
3d632c3f SH |
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)"); | |
ad246c99 | 116 | module_param_named(num_grat_arp, num_peer_notif, int, 0644); |
90e62474 AG |
117 | MODULE_PARM_DESC(num_grat_arp, "Number of peer notifications to send on " |
118 | "failover event (alias of num_unsol_na)"); | |
ad246c99 | 119 | module_param_named(num_unsol_na, num_peer_notif, int, 0644); |
90e62474 AG |
120 | MODULE_PARM_DESC(num_unsol_na, "Number of peer notifications to send on " |
121 | "failover event (alias of num_grat_arp)"); | |
1da177e4 LT |
122 | module_param(miimon, int, 0); |
123 | MODULE_PARM_DESC(miimon, "Link check interval in milliseconds"); | |
124 | module_param(updelay, int, 0); | |
125 | MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds"); | |
126 | module_param(downdelay, int, 0); | |
2ac47660 MW |
127 | MODULE_PARM_DESC(downdelay, "Delay before considering link down, " |
128 | "in milliseconds"); | |
1da177e4 | 129 | module_param(use_carrier, int, 0); |
2ac47660 MW |
130 | MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; " |
131 | "0 for off, 1 for on (default)"); | |
1da177e4 | 132 | module_param(mode, charp, 0); |
90e62474 | 133 | MODULE_PARM_DESC(mode, "Mode of operation; 0 for balance-rr, " |
2ac47660 MW |
134 | "1 for active-backup, 2 for balance-xor, " |
135 | "3 for broadcast, 4 for 802.3ad, 5 for balance-tlb, " | |
136 | "6 for balance-alb"); | |
1da177e4 LT |
137 | module_param(primary, charp, 0); |
138 | MODULE_PARM_DESC(primary, "Primary network device to use"); | |
a549952a JP |
139 | module_param(primary_reselect, charp, 0); |
140 | MODULE_PARM_DESC(primary_reselect, "Reselect primary slave " | |
141 | "once it comes up; " | |
142 | "0 for always (default), " | |
143 | "1 for only if speed of primary is " | |
144 | "better, " | |
145 | "2 for only on active slave " | |
146 | "failure"); | |
1da177e4 | 147 | module_param(lacp_rate, charp, 0); |
90e62474 AG |
148 | MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner; " |
149 | "0 for slow, 1 for fast"); | |
fd989c83 | 150 | module_param(ad_select, charp, 0); |
90e62474 AG |
151 | MODULE_PARM_DESC(ad_select, "803.ad aggregation selection logic; " |
152 | "0 for stable (default), 1 for bandwidth, " | |
153 | "2 for count"); | |
655f8919 | 154 | module_param(min_links, int, 0); |
155 | MODULE_PARM_DESC(min_links, "Minimum number of available links before turning on carrier"); | |
156 | ||
169a3e66 | 157 | module_param(xmit_hash_policy, charp, 0); |
90e62474 AG |
158 | MODULE_PARM_DESC(xmit_hash_policy, "balance-xor and 802.3ad hashing method; " |
159 | "0 for layer 2 (default), 1 for layer 3+4, " | |
160 | "2 for layer 2+3"); | |
1da177e4 LT |
161 | module_param(arp_interval, int, 0); |
162 | MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds"); | |
163 | module_param_array(arp_ip_target, charp, NULL, 0); | |
164 | MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form"); | |
f5b2b966 | 165 | module_param(arp_validate, charp, 0); |
90e62474 AG |
166 | MODULE_PARM_DESC(arp_validate, "validate src/dst of ARP probes; " |
167 | "0 for none (default), 1 for active, " | |
168 | "2 for backup, 3 for all"); | |
3915c1e8 | 169 | module_param(fail_over_mac, charp, 0); |
90e62474 AG |
170 | MODULE_PARM_DESC(fail_over_mac, "For active-backup, do not set all slaves to " |
171 | "the same MAC; 0 for none (default), " | |
172 | "1 for active, 2 for follow"); | |
ebd8e497 AG |
173 | module_param(all_slaves_active, int, 0); |
174 | MODULE_PARM_DESC(all_slaves_active, "Keep all frames received on an interface" | |
90e62474 | 175 | "by setting active flag for all slaves; " |
ebd8e497 | 176 | "0 for never (default), 1 for always."); |
c2952c31 | 177 | module_param(resend_igmp, int, 0); |
90e62474 AG |
178 | MODULE_PARM_DESC(resend_igmp, "Number of IGMP membership reports to send on " |
179 | "link failure"); | |
1da177e4 LT |
180 | |
181 | /*----------------------------- Global variables ----------------------------*/ | |
182 | ||
e843fa50 | 183 | #ifdef CONFIG_NET_POLL_CONTROLLER |
fb4fa76a | 184 | atomic_t netpoll_block_tx = ATOMIC_INIT(0); |
e843fa50 NH |
185 | #endif |
186 | ||
f99189b1 | 187 | int bond_net_id __read_mostly; |
1da177e4 | 188 | |
3d632c3f SH |
189 | static __be32 arp_target[BOND_MAX_ARP_TARGETS]; |
190 | static int arp_ip_count; | |
1da177e4 | 191 | static int bond_mode = BOND_MODE_ROUNDROBIN; |
3d632c3f SH |
192 | static int xmit_hashtype = BOND_XMIT_POLICY_LAYER2; |
193 | static int lacp_fast; | |
1da177e4 | 194 | |
e97fd7c6 | 195 | const struct bond_parm_tbl bond_lacp_tbl[] = { |
1da177e4 LT |
196 | { "slow", AD_LACP_SLOW}, |
197 | { "fast", AD_LACP_FAST}, | |
198 | { NULL, -1}, | |
199 | }; | |
200 | ||
e97fd7c6 | 201 | const struct bond_parm_tbl bond_mode_tbl[] = { |
1da177e4 LT |
202 | { "balance-rr", BOND_MODE_ROUNDROBIN}, |
203 | { "active-backup", BOND_MODE_ACTIVEBACKUP}, | |
204 | { "balance-xor", BOND_MODE_XOR}, | |
205 | { "broadcast", BOND_MODE_BROADCAST}, | |
206 | { "802.3ad", BOND_MODE_8023AD}, | |
207 | { "balance-tlb", BOND_MODE_TLB}, | |
208 | { "balance-alb", BOND_MODE_ALB}, | |
209 | { NULL, -1}, | |
210 | }; | |
211 | ||
e97fd7c6 | 212 | const struct bond_parm_tbl xmit_hashtype_tbl[] = { |
169a3e66 JV |
213 | { "layer2", BOND_XMIT_POLICY_LAYER2}, |
214 | { "layer3+4", BOND_XMIT_POLICY_LAYER34}, | |
6f6652be | 215 | { "layer2+3", BOND_XMIT_POLICY_LAYER23}, |
169a3e66 JV |
216 | { NULL, -1}, |
217 | }; | |
218 | ||
e97fd7c6 | 219 | const struct bond_parm_tbl arp_validate_tbl[] = { |
f5b2b966 JV |
220 | { "none", BOND_ARP_VALIDATE_NONE}, |
221 | { "active", BOND_ARP_VALIDATE_ACTIVE}, | |
222 | { "backup", BOND_ARP_VALIDATE_BACKUP}, | |
223 | { "all", BOND_ARP_VALIDATE_ALL}, | |
224 | { NULL, -1}, | |
225 | }; | |
226 | ||
e97fd7c6 | 227 | const struct bond_parm_tbl fail_over_mac_tbl[] = { |
3915c1e8 JV |
228 | { "none", BOND_FOM_NONE}, |
229 | { "active", BOND_FOM_ACTIVE}, | |
230 | { "follow", BOND_FOM_FOLLOW}, | |
231 | { NULL, -1}, | |
232 | }; | |
233 | ||
a549952a JP |
234 | const struct bond_parm_tbl pri_reselect_tbl[] = { |
235 | { "always", BOND_PRI_RESELECT_ALWAYS}, | |
236 | { "better", BOND_PRI_RESELECT_BETTER}, | |
237 | { "failure", BOND_PRI_RESELECT_FAILURE}, | |
238 | { NULL, -1}, | |
239 | }; | |
240 | ||
fd989c83 JV |
241 | struct bond_parm_tbl ad_select_tbl[] = { |
242 | { "stable", BOND_AD_STABLE}, | |
243 | { "bandwidth", BOND_AD_BANDWIDTH}, | |
244 | { "count", BOND_AD_COUNT}, | |
245 | { NULL, -1}, | |
246 | }; | |
247 | ||
1da177e4 LT |
248 | /*-------------------------- Forward declarations ---------------------------*/ |
249 | ||
181470fc | 250 | static int bond_init(struct net_device *bond_dev); |
c67dfb29 | 251 | static void bond_uninit(struct net_device *bond_dev); |
1da177e4 LT |
252 | |
253 | /*---------------------------- General routines -----------------------------*/ | |
254 | ||
bd33acc3 | 255 | const char *bond_mode_name(int mode) |
1da177e4 | 256 | { |
77afc92b HE |
257 | static const char *names[] = { |
258 | [BOND_MODE_ROUNDROBIN] = "load balancing (round-robin)", | |
259 | [BOND_MODE_ACTIVEBACKUP] = "fault-tolerance (active-backup)", | |
260 | [BOND_MODE_XOR] = "load balancing (xor)", | |
261 | [BOND_MODE_BROADCAST] = "fault-tolerance (broadcast)", | |
3d632c3f | 262 | [BOND_MODE_8023AD] = "IEEE 802.3ad Dynamic link aggregation", |
77afc92b HE |
263 | [BOND_MODE_TLB] = "transmit load balancing", |
264 | [BOND_MODE_ALB] = "adaptive load balancing", | |
265 | }; | |
266 | ||
267 | if (mode < 0 || mode > BOND_MODE_ALB) | |
1da177e4 | 268 | return "unknown"; |
77afc92b HE |
269 | |
270 | return names[mode]; | |
1da177e4 LT |
271 | } |
272 | ||
273 | /*---------------------------------- VLAN -----------------------------------*/ | |
274 | ||
275 | /** | |
276 | * bond_add_vlan - add a new vlan id on bond | |
277 | * @bond: bond that got the notification | |
278 | * @vlan_id: the vlan id to add | |
279 | * | |
280 | * Returns -ENOMEM if allocation failed. | |
281 | */ | |
282 | static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id) | |
283 | { | |
284 | struct vlan_entry *vlan; | |
285 | ||
5a03cdb7 | 286 | pr_debug("bond: %s, vlan id %d\n", |
a4aee5c8 | 287 | (bond ? bond->dev->name : "None"), vlan_id); |
1da177e4 | 288 | |
305d552a | 289 | vlan = kzalloc(sizeof(struct vlan_entry), GFP_KERNEL); |
3d632c3f | 290 | if (!vlan) |
1da177e4 | 291 | return -ENOMEM; |
1da177e4 LT |
292 | |
293 | INIT_LIST_HEAD(&vlan->vlan_list); | |
294 | vlan->vlan_id = vlan_id; | |
295 | ||
296 | write_lock_bh(&bond->lock); | |
297 | ||
298 | list_add_tail(&vlan->vlan_list, &bond->vlan_list); | |
299 | ||
300 | write_unlock_bh(&bond->lock); | |
301 | ||
5a03cdb7 | 302 | pr_debug("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name); |
1da177e4 LT |
303 | |
304 | return 0; | |
305 | } | |
306 | ||
307 | /** | |
308 | * bond_del_vlan - delete a vlan id from bond | |
309 | * @bond: bond that got the notification | |
310 | * @vlan_id: the vlan id to delete | |
311 | * | |
312 | * returns -ENODEV if @vlan_id was not found in @bond. | |
313 | */ | |
314 | static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id) | |
315 | { | |
0883beca | 316 | struct vlan_entry *vlan; |
1da177e4 LT |
317 | int res = -ENODEV; |
318 | ||
5a03cdb7 | 319 | pr_debug("bond: %s, vlan id %d\n", bond->dev->name, vlan_id); |
1da177e4 | 320 | |
e843fa50 | 321 | block_netpoll_tx(); |
1da177e4 LT |
322 | write_lock_bh(&bond->lock); |
323 | ||
0883beca | 324 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1da177e4 LT |
325 | if (vlan->vlan_id == vlan_id) { |
326 | list_del(&vlan->vlan_list); | |
327 | ||
58402054 | 328 | if (bond_is_lb(bond)) |
1da177e4 | 329 | bond_alb_clear_vlan(bond, vlan_id); |
1da177e4 | 330 | |
a4aee5c8 JP |
331 | pr_debug("removed VLAN ID %d from bond %s\n", |
332 | vlan_id, bond->dev->name); | |
1da177e4 LT |
333 | |
334 | kfree(vlan); | |
335 | ||
1da177e4 LT |
336 | res = 0; |
337 | goto out; | |
338 | } | |
339 | } | |
340 | ||
a4aee5c8 JP |
341 | pr_debug("couldn't find VLAN ID %d in bond %s\n", |
342 | vlan_id, bond->dev->name); | |
1da177e4 LT |
343 | |
344 | out: | |
345 | write_unlock_bh(&bond->lock); | |
e843fa50 | 346 | unblock_netpoll_tx(); |
1da177e4 LT |
347 | return res; |
348 | } | |
349 | ||
1da177e4 LT |
350 | /** |
351 | * bond_next_vlan - safely skip to the next item in the vlans list. | |
352 | * @bond: the bond we're working on | |
353 | * @curr: item we're advancing from | |
354 | * | |
355 | * Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL, | |
356 | * or @curr->next otherwise (even if it is @curr itself again). | |
3d632c3f | 357 | * |
1da177e4 LT |
358 | * Caller must hold bond->lock |
359 | */ | |
360 | struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr) | |
361 | { | |
362 | struct vlan_entry *next, *last; | |
363 | ||
3d632c3f | 364 | if (list_empty(&bond->vlan_list)) |
1da177e4 | 365 | return NULL; |
1da177e4 LT |
366 | |
367 | if (!curr) { | |
368 | next = list_entry(bond->vlan_list.next, | |
369 | struct vlan_entry, vlan_list); | |
370 | } else { | |
371 | last = list_entry(bond->vlan_list.prev, | |
372 | struct vlan_entry, vlan_list); | |
373 | if (last == curr) { | |
374 | next = list_entry(bond->vlan_list.next, | |
375 | struct vlan_entry, vlan_list); | |
376 | } else { | |
377 | next = list_entry(curr->vlan_list.next, | |
378 | struct vlan_entry, vlan_list); | |
379 | } | |
380 | } | |
381 | ||
382 | return next; | |
383 | } | |
384 | ||
385 | /** | |
386 | * bond_dev_queue_xmit - Prepare skb for xmit. | |
3d632c3f | 387 | * |
1da177e4 LT |
388 | * @bond: bond device that got this skb for tx. |
389 | * @skb: hw accel VLAN tagged skb to transmit | |
390 | * @slave_dev: slave that is supposed to xmit this skbuff | |
1da177e4 | 391 | */ |
3d632c3f SH |
392 | int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, |
393 | struct net_device *slave_dev) | |
1da177e4 | 394 | { |
8387451e | 395 | skb->dev = slave_dev; |
374eeb5a | 396 | |
5ee31c68 | 397 | BUILD_BUG_ON(sizeof(skb->queue_mapping) != |
df4ab5b3 JP |
398 | sizeof(qdisc_skb_cb(skb)->slave_dev_queue_mapping)); |
399 | skb->queue_mapping = qdisc_skb_cb(skb)->slave_dev_queue_mapping; | |
374eeb5a | 400 | |
e15c3c22 | 401 | if (unlikely(netpoll_tx_running(bond->dev))) |
8a8efa22 | 402 | bond_netpoll_send_skb(bond_get_slave_by_dev(bond, slave_dev), skb); |
080e4130 | 403 | else |
f6dc31a8 | 404 | dev_queue_xmit(skb); |
1da177e4 LT |
405 | |
406 | return 0; | |
407 | } | |
408 | ||
409 | /* | |
cc0e4070 JP |
410 | * In the following 2 functions, bond_vlan_rx_add_vid and bond_vlan_rx_kill_vid, |
411 | * We don't protect the slave list iteration with a lock because: | |
1da177e4 LT |
412 | * a. This operation is performed in IOCTL context, |
413 | * b. The operation is protected by the RTNL semaphore in the 8021q code, | |
414 | * c. Holding a lock with BH disabled while directly calling a base driver | |
415 | * entry point is generally a BAD idea. | |
3d632c3f | 416 | * |
1da177e4 LT |
417 | * The design of synchronization/protection for this operation in the 8021q |
418 | * module is good for one or more VLAN devices over a single physical device | |
419 | * and cannot be extended for a teaming solution like bonding, so there is a | |
420 | * potential race condition here where a net device from the vlan group might | |
421 | * be referenced (either by a base driver or the 8021q code) while it is being | |
422 | * removed from the system. However, it turns out we're not making matters | |
423 | * worse, and if it works for regular VLAN usage it will work here too. | |
424 | */ | |
425 | ||
1da177e4 LT |
426 | /** |
427 | * bond_vlan_rx_add_vid - Propagates adding an id to slaves | |
428 | * @bond_dev: bonding net device that got called | |
429 | * @vid: vlan id being added | |
430 | */ | |
8e586137 | 431 | static int bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid) |
1da177e4 | 432 | { |
454d7c9b | 433 | struct bonding *bond = netdev_priv(bond_dev); |
87002b03 | 434 | struct slave *slave, *stop_at; |
1da177e4 LT |
435 | int i, res; |
436 | ||
437 | bond_for_each_slave(bond, slave, i) { | |
87002b03 JP |
438 | res = vlan_vid_add(slave->dev, vid); |
439 | if (res) | |
440 | goto unwind; | |
1da177e4 LT |
441 | } |
442 | ||
443 | res = bond_add_vlan(bond, vid); | |
444 | if (res) { | |
a4aee5c8 | 445 | pr_err("%s: Error: Failed to add vlan id %d\n", |
1da177e4 | 446 | bond_dev->name, vid); |
8e586137 | 447 | return res; |
1da177e4 | 448 | } |
8e586137 JP |
449 | |
450 | return 0; | |
87002b03 JP |
451 | |
452 | unwind: | |
453 | /* unwind from head to the slave that failed */ | |
454 | stop_at = slave; | |
455 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) | |
456 | vlan_vid_del(slave->dev, vid); | |
457 | ||
458 | return res; | |
1da177e4 LT |
459 | } |
460 | ||
461 | /** | |
462 | * bond_vlan_rx_kill_vid - Propagates deleting an id to slaves | |
463 | * @bond_dev: bonding net device that got called | |
464 | * @vid: vlan id being removed | |
465 | */ | |
8e586137 | 466 | static int bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid) |
1da177e4 | 467 | { |
454d7c9b | 468 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 469 | struct slave *slave; |
1da177e4 LT |
470 | int i, res; |
471 | ||
87002b03 JP |
472 | bond_for_each_slave(bond, slave, i) |
473 | vlan_vid_del(slave->dev, vid); | |
1da177e4 LT |
474 | |
475 | res = bond_del_vlan(bond, vid); | |
476 | if (res) { | |
a4aee5c8 | 477 | pr_err("%s: Error: Failed to remove vlan id %d\n", |
1da177e4 | 478 | bond_dev->name, vid); |
8e586137 | 479 | return res; |
1da177e4 | 480 | } |
8e586137 JP |
481 | |
482 | return 0; | |
1da177e4 LT |
483 | } |
484 | ||
485 | static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev) | |
486 | { | |
487 | struct vlan_entry *vlan; | |
87002b03 | 488 | int res; |
1da177e4 | 489 | |
87002b03 JP |
490 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
491 | res = vlan_vid_add(slave_dev, vlan->vlan_id); | |
492 | if (res) | |
493 | pr_warning("%s: Failed to add vlan id %d to device %s\n", | |
494 | bond->dev->name, vlan->vlan_id, | |
495 | slave_dev->name); | |
496 | } | |
1da177e4 LT |
497 | } |
498 | ||
3d632c3f SH |
499 | static void bond_del_vlans_from_slave(struct bonding *bond, |
500 | struct net_device *slave_dev) | |
1da177e4 LT |
501 | { |
502 | struct vlan_entry *vlan; | |
1da177e4 | 503 | |
1da177e4 | 504 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
f35188fa JV |
505 | if (!vlan->vlan_id) |
506 | continue; | |
87002b03 | 507 | vlan_vid_del(slave_dev, vlan->vlan_id); |
1da177e4 | 508 | } |
1da177e4 LT |
509 | } |
510 | ||
511 | /*------------------------------- Link status -------------------------------*/ | |
512 | ||
ff59c456 JV |
513 | /* |
514 | * Set the carrier state for the master according to the state of its | |
515 | * slaves. If any slaves are up, the master is up. In 802.3ad mode, | |
516 | * do special 802.3ad magic. | |
517 | * | |
518 | * Returns zero if carrier state does not change, nonzero if it does. | |
519 | */ | |
520 | static int bond_set_carrier(struct bonding *bond) | |
521 | { | |
522 | struct slave *slave; | |
523 | int i; | |
524 | ||
525 | if (bond->slave_cnt == 0) | |
526 | goto down; | |
527 | ||
528 | if (bond->params.mode == BOND_MODE_8023AD) | |
529 | return bond_3ad_set_carrier(bond); | |
530 | ||
531 | bond_for_each_slave(bond, slave, i) { | |
532 | if (slave->link == BOND_LINK_UP) { | |
533 | if (!netif_carrier_ok(bond->dev)) { | |
534 | netif_carrier_on(bond->dev); | |
535 | return 1; | |
536 | } | |
537 | return 0; | |
538 | } | |
539 | } | |
540 | ||
541 | down: | |
542 | if (netif_carrier_ok(bond->dev)) { | |
543 | netif_carrier_off(bond->dev); | |
544 | return 1; | |
545 | } | |
546 | return 0; | |
547 | } | |
548 | ||
1da177e4 LT |
549 | /* |
550 | * Get link speed and duplex from the slave's base driver | |
551 | * using ethtool. If for some reason the call fails or the | |
98f41f69 | 552 | * values are invalid, set speed and duplex to -1, |
13b95fb7 | 553 | * and return. |
1da177e4 | 554 | */ |
13b95fb7 | 555 | static void bond_update_speed_duplex(struct slave *slave) |
1da177e4 LT |
556 | { |
557 | struct net_device *slave_dev = slave->dev; | |
4bc71cb9 | 558 | struct ethtool_cmd ecmd; |
5d30530e | 559 | u32 slave_speed; |
61a44b9c | 560 | int res; |
1da177e4 | 561 | |
589665f5 DC |
562 | slave->speed = SPEED_UNKNOWN; |
563 | slave->duplex = DUPLEX_UNKNOWN; | |
1da177e4 | 564 | |
4bc71cb9 | 565 | res = __ethtool_get_settings(slave_dev, &ecmd); |
61a44b9c | 566 | if (res < 0) |
13b95fb7 | 567 | return; |
1da177e4 | 568 | |
4bc71cb9 | 569 | slave_speed = ethtool_cmd_speed(&ecmd); |
6f92c66f | 570 | if (slave_speed == 0 || slave_speed == ((__u32) -1)) |
13b95fb7 | 571 | return; |
1da177e4 | 572 | |
4bc71cb9 | 573 | switch (ecmd.duplex) { |
1da177e4 LT |
574 | case DUPLEX_FULL: |
575 | case DUPLEX_HALF: | |
576 | break; | |
577 | default: | |
13b95fb7 | 578 | return; |
1da177e4 LT |
579 | } |
580 | ||
5d30530e | 581 | slave->speed = slave_speed; |
4bc71cb9 | 582 | slave->duplex = ecmd.duplex; |
1da177e4 | 583 | |
13b95fb7 | 584 | return; |
1da177e4 LT |
585 | } |
586 | ||
587 | /* | |
588 | * if <dev> supports MII link status reporting, check its link status. | |
589 | * | |
590 | * We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(), | |
3d632c3f | 591 | * depending upon the setting of the use_carrier parameter. |
1da177e4 LT |
592 | * |
593 | * Return either BMSR_LSTATUS, meaning that the link is up (or we | |
594 | * can't tell and just pretend it is), or 0, meaning that the link is | |
595 | * down. | |
596 | * | |
597 | * If reporting is non-zero, instead of faking link up, return -1 if | |
598 | * both ETHTOOL and MII ioctls fail (meaning the device does not | |
599 | * support them). If use_carrier is set, return whatever it says. | |
600 | * It'd be nice if there was a good way to tell if a driver supports | |
601 | * netif_carrier, but there really isn't. | |
602 | */ | |
3d632c3f SH |
603 | static int bond_check_dev_link(struct bonding *bond, |
604 | struct net_device *slave_dev, int reporting) | |
1da177e4 | 605 | { |
eb7cc59a | 606 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
d9d52832 | 607 | int (*ioctl)(struct net_device *, struct ifreq *, int); |
1da177e4 LT |
608 | struct ifreq ifr; |
609 | struct mii_ioctl_data *mii; | |
1da177e4 | 610 | |
6c988853 PG |
611 | if (!reporting && !netif_running(slave_dev)) |
612 | return 0; | |
613 | ||
eb7cc59a | 614 | if (bond->params.use_carrier) |
1da177e4 | 615 | return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0; |
1da177e4 | 616 | |
29112f4e | 617 | /* Try to get link status using Ethtool first. */ |
c772dde3 BH |
618 | if (slave_dev->ethtool_ops->get_link) |
619 | return slave_dev->ethtool_ops->get_link(slave_dev) ? | |
620 | BMSR_LSTATUS : 0; | |
29112f4e | 621 | |
3d632c3f | 622 | /* Ethtool can't be used, fallback to MII ioctls. */ |
eb7cc59a | 623 | ioctl = slave_ops->ndo_do_ioctl; |
1da177e4 LT |
624 | if (ioctl) { |
625 | /* TODO: set pointer to correct ioctl on a per team member */ | |
626 | /* bases to make this more efficient. that is, once */ | |
627 | /* we determine the correct ioctl, we will always */ | |
628 | /* call it and not the others for that team */ | |
629 | /* member. */ | |
630 | ||
631 | /* | |
632 | * We cannot assume that SIOCGMIIPHY will also read a | |
633 | * register; not all network drivers (e.g., e100) | |
634 | * support that. | |
635 | */ | |
636 | ||
637 | /* Yes, the mii is overlaid on the ifreq.ifr_ifru */ | |
638 | strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ); | |
639 | mii = if_mii(&ifr); | |
640 | if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) { | |
641 | mii->reg_num = MII_BMSR; | |
3d632c3f SH |
642 | if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) |
643 | return mii->val_out & BMSR_LSTATUS; | |
1da177e4 LT |
644 | } |
645 | } | |
646 | ||
1da177e4 LT |
647 | /* |
648 | * If reporting, report that either there's no dev->do_ioctl, | |
61a44b9c | 649 | * or both SIOCGMIIREG and get_link failed (meaning that we |
1da177e4 LT |
650 | * cannot report link status). If not reporting, pretend |
651 | * we're ok. | |
652 | */ | |
3d632c3f | 653 | return reporting ? -1 : BMSR_LSTATUS; |
1da177e4 LT |
654 | } |
655 | ||
656 | /*----------------------------- Multicast list ------------------------------*/ | |
657 | ||
1da177e4 LT |
658 | /* |
659 | * Push the promiscuity flag down to appropriate slaves | |
660 | */ | |
7e1a1ac1 | 661 | static int bond_set_promiscuity(struct bonding *bond, int inc) |
1da177e4 | 662 | { |
7e1a1ac1 | 663 | int err = 0; |
1da177e4 LT |
664 | if (USES_PRIMARY(bond->params.mode)) { |
665 | /* write lock already acquired */ | |
666 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
667 | err = dev_set_promiscuity(bond->curr_active_slave->dev, |
668 | inc); | |
1da177e4 LT |
669 | } |
670 | } else { | |
671 | struct slave *slave; | |
672 | int i; | |
673 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
674 | err = dev_set_promiscuity(slave->dev, inc); |
675 | if (err) | |
676 | return err; | |
1da177e4 LT |
677 | } |
678 | } | |
7e1a1ac1 | 679 | return err; |
1da177e4 LT |
680 | } |
681 | ||
682 | /* | |
683 | * Push the allmulti flag down to all slaves | |
684 | */ | |
7e1a1ac1 | 685 | static int bond_set_allmulti(struct bonding *bond, int inc) |
1da177e4 | 686 | { |
7e1a1ac1 | 687 | int err = 0; |
1da177e4 LT |
688 | if (USES_PRIMARY(bond->params.mode)) { |
689 | /* write lock already acquired */ | |
690 | if (bond->curr_active_slave) { | |
7e1a1ac1 WC |
691 | err = dev_set_allmulti(bond->curr_active_slave->dev, |
692 | inc); | |
1da177e4 LT |
693 | } |
694 | } else { | |
695 | struct slave *slave; | |
696 | int i; | |
697 | bond_for_each_slave(bond, slave, i) { | |
7e1a1ac1 WC |
698 | err = dev_set_allmulti(slave->dev, inc); |
699 | if (err) | |
700 | return err; | |
1da177e4 LT |
701 | } |
702 | } | |
7e1a1ac1 | 703 | return err; |
1da177e4 LT |
704 | } |
705 | ||
706 | /* | |
707 | * Add a Multicast address to slaves | |
708 | * according to mode | |
709 | */ | |
22bedad3 | 710 | static void bond_mc_add(struct bonding *bond, void *addr) |
1da177e4 LT |
711 | { |
712 | if (USES_PRIMARY(bond->params.mode)) { | |
713 | /* write lock already acquired */ | |
3d632c3f | 714 | if (bond->curr_active_slave) |
22bedad3 | 715 | dev_mc_add(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
716 | } else { |
717 | struct slave *slave; | |
718 | int i; | |
3d632c3f SH |
719 | |
720 | bond_for_each_slave(bond, slave, i) | |
22bedad3 | 721 | dev_mc_add(slave->dev, addr); |
1da177e4 LT |
722 | } |
723 | } | |
724 | ||
725 | /* | |
726 | * Remove a multicast address from slave | |
727 | * according to mode | |
728 | */ | |
22bedad3 | 729 | static void bond_mc_del(struct bonding *bond, void *addr) |
1da177e4 LT |
730 | { |
731 | if (USES_PRIMARY(bond->params.mode)) { | |
732 | /* write lock already acquired */ | |
3d632c3f | 733 | if (bond->curr_active_slave) |
22bedad3 | 734 | dev_mc_del(bond->curr_active_slave->dev, addr); |
1da177e4 LT |
735 | } else { |
736 | struct slave *slave; | |
737 | int i; | |
738 | bond_for_each_slave(bond, slave, i) { | |
22bedad3 | 739 | dev_mc_del(slave->dev, addr); |
1da177e4 LT |
740 | } |
741 | } | |
742 | } | |
743 | ||
a816c7c7 | 744 | |
5a37e8ca | 745 | static void __bond_resend_igmp_join_requests(struct net_device *dev) |
a816c7c7 JV |
746 | { |
747 | struct in_device *in_dev; | |
a816c7c7 | 748 | |
5a37e8ca | 749 | in_dev = __in_dev_get_rcu(dev); |
866f3b25 | 750 | if (in_dev) |
24912420 | 751 | ip_mc_rejoin_groups(in_dev); |
a816c7c7 JV |
752 | } |
753 | ||
5a37e8ca FL |
754 | /* |
755 | * Retrieve the list of registered multicast addresses for the bonding | |
756 | * device and retransmit an IGMP JOIN request to the current active | |
757 | * slave. | |
758 | */ | |
759 | static void bond_resend_igmp_join_requests(struct bonding *bond) | |
760 | { | |
471cb5a3 | 761 | struct net_device *bond_dev, *vlan_dev, *upper_dev; |
5a37e8ca FL |
762 | struct vlan_entry *vlan; |
763 | ||
471cb5a3 | 764 | rcu_read_lock(); |
5a37e8ca FL |
765 | read_lock(&bond->lock); |
766 | ||
1c3ac428 PP |
767 | bond_dev = bond->dev; |
768 | ||
5a37e8ca | 769 | /* rejoin all groups on bond device */ |
1c3ac428 PP |
770 | __bond_resend_igmp_join_requests(bond_dev); |
771 | ||
772 | /* | |
773 | * if bond is enslaved to a bridge, | |
774 | * then rejoin all groups on its master | |
775 | */ | |
471cb5a3 JP |
776 | upper_dev = netdev_master_upper_dev_get_rcu(bond_dev); |
777 | if (upper_dev && upper_dev->priv_flags & IFF_EBRIDGE) | |
778 | __bond_resend_igmp_join_requests(upper_dev); | |
5a37e8ca FL |
779 | |
780 | /* rejoin all groups on vlan devices */ | |
cc0e4070 | 781 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
1c3ac428 | 782 | vlan_dev = __vlan_find_dev_deep(bond_dev, |
cc0e4070 | 783 | vlan->vlan_id); |
cc0e4070 JP |
784 | if (vlan_dev) |
785 | __bond_resend_igmp_join_requests(vlan_dev); | |
5a37e8ca FL |
786 | } |
787 | ||
e6d265e8 | 788 | if (--bond->igmp_retrans > 0) |
c2952c31 | 789 | queue_delayed_work(bond->wq, &bond->mcast_work, HZ/5); |
e6d265e8 | 790 | |
5a37e8ca | 791 | read_unlock(&bond->lock); |
471cb5a3 | 792 | rcu_read_unlock(); |
5a37e8ca FL |
793 | } |
794 | ||
379b7383 | 795 | static void bond_resend_igmp_join_requests_delayed(struct work_struct *work) |
5a37e8ca FL |
796 | { |
797 | struct bonding *bond = container_of(work, struct bonding, | |
94265cf5 | 798 | mcast_work.work); |
471cb5a3 | 799 | rcu_read_lock(); |
5a37e8ca | 800 | bond_resend_igmp_join_requests(bond); |
471cb5a3 | 801 | rcu_read_unlock(); |
5a37e8ca FL |
802 | } |
803 | ||
1da177e4 LT |
804 | /* |
805 | * flush all members of flush->mc_list from device dev->mc_list | |
806 | */ | |
3d632c3f SH |
807 | static void bond_mc_list_flush(struct net_device *bond_dev, |
808 | struct net_device *slave_dev) | |
1da177e4 | 809 | { |
454d7c9b | 810 | struct bonding *bond = netdev_priv(bond_dev); |
22bedad3 | 811 | struct netdev_hw_addr *ha; |
1da177e4 | 812 | |
22bedad3 JP |
813 | netdev_for_each_mc_addr(ha, bond_dev) |
814 | dev_mc_del(slave_dev, ha->addr); | |
1da177e4 LT |
815 | |
816 | if (bond->params.mode == BOND_MODE_8023AD) { | |
817 | /* del lacpdu mc addr from mc list */ | |
818 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
819 | ||
22bedad3 | 820 | dev_mc_del(slave_dev, lacpdu_multicast); |
1da177e4 LT |
821 | } |
822 | } | |
823 | ||
824 | /*--------------------------- Active slave change ---------------------------*/ | |
825 | ||
826 | /* | |
827 | * Update the mc list and multicast-related flags for the new and | |
828 | * old active slaves (if any) according to the multicast mode, and | |
829 | * promiscuous flags unconditionally. | |
830 | */ | |
3d632c3f SH |
831 | static void bond_mc_swap(struct bonding *bond, struct slave *new_active, |
832 | struct slave *old_active) | |
1da177e4 | 833 | { |
22bedad3 | 834 | struct netdev_hw_addr *ha; |
1da177e4 | 835 | |
3d632c3f | 836 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 LT |
837 | /* nothing to do - mc list is already up-to-date on |
838 | * all slaves | |
839 | */ | |
840 | return; | |
1da177e4 LT |
841 | |
842 | if (old_active) { | |
3d632c3f | 843 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 844 | dev_set_promiscuity(old_active->dev, -1); |
1da177e4 | 845 | |
3d632c3f | 846 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 847 | dev_set_allmulti(old_active->dev, -1); |
1da177e4 | 848 | |
22bedad3 JP |
849 | netdev_for_each_mc_addr(ha, bond->dev) |
850 | dev_mc_del(old_active->dev, ha->addr); | |
1da177e4 LT |
851 | } |
852 | ||
853 | if (new_active) { | |
7e1a1ac1 | 854 | /* FIXME: Signal errors upstream. */ |
3d632c3f | 855 | if (bond->dev->flags & IFF_PROMISC) |
1da177e4 | 856 | dev_set_promiscuity(new_active->dev, 1); |
1da177e4 | 857 | |
3d632c3f | 858 | if (bond->dev->flags & IFF_ALLMULTI) |
1da177e4 | 859 | dev_set_allmulti(new_active->dev, 1); |
1da177e4 | 860 | |
22bedad3 JP |
861 | netdev_for_each_mc_addr(ha, bond->dev) |
862 | dev_mc_add(new_active->dev, ha->addr); | |
1da177e4 LT |
863 | } |
864 | } | |
865 | ||
3915c1e8 JV |
866 | /* |
867 | * bond_do_fail_over_mac | |
868 | * | |
869 | * Perform special MAC address swapping for fail_over_mac settings | |
870 | * | |
871 | * Called with RTNL, bond->lock for read, curr_slave_lock for write_bh. | |
872 | */ | |
873 | static void bond_do_fail_over_mac(struct bonding *bond, | |
874 | struct slave *new_active, | |
875 | struct slave *old_active) | |
1f78d9f9 HE |
876 | __releases(&bond->curr_slave_lock) |
877 | __releases(&bond->lock) | |
878 | __acquires(&bond->lock) | |
879 | __acquires(&bond->curr_slave_lock) | |
3915c1e8 JV |
880 | { |
881 | u8 tmp_mac[ETH_ALEN]; | |
882 | struct sockaddr saddr; | |
883 | int rv; | |
884 | ||
885 | switch (bond->params.fail_over_mac) { | |
886 | case BOND_FOM_ACTIVE: | |
7d26bb10 | 887 | if (new_active) { |
3915c1e8 JV |
888 | memcpy(bond->dev->dev_addr, new_active->dev->dev_addr, |
889 | new_active->dev->addr_len); | |
7d26bb10 WP |
890 | write_unlock_bh(&bond->curr_slave_lock); |
891 | read_unlock(&bond->lock); | |
892 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); | |
893 | read_lock(&bond->lock); | |
894 | write_lock_bh(&bond->curr_slave_lock); | |
895 | } | |
3915c1e8 JV |
896 | break; |
897 | case BOND_FOM_FOLLOW: | |
898 | /* | |
899 | * if new_active && old_active, swap them | |
900 | * if just old_active, do nothing (going to no active slave) | |
901 | * if just new_active, set new_active to bond's MAC | |
902 | */ | |
903 | if (!new_active) | |
904 | return; | |
905 | ||
906 | write_unlock_bh(&bond->curr_slave_lock); | |
907 | read_unlock(&bond->lock); | |
908 | ||
909 | if (old_active) { | |
910 | memcpy(tmp_mac, new_active->dev->dev_addr, ETH_ALEN); | |
911 | memcpy(saddr.sa_data, old_active->dev->dev_addr, | |
912 | ETH_ALEN); | |
913 | saddr.sa_family = new_active->dev->type; | |
914 | } else { | |
915 | memcpy(saddr.sa_data, bond->dev->dev_addr, ETH_ALEN); | |
916 | saddr.sa_family = bond->dev->type; | |
917 | } | |
918 | ||
919 | rv = dev_set_mac_address(new_active->dev, &saddr); | |
920 | if (rv) { | |
a4aee5c8 | 921 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
922 | bond->dev->name, -rv, new_active->dev->name); |
923 | goto out; | |
924 | } | |
925 | ||
926 | if (!old_active) | |
927 | goto out; | |
928 | ||
929 | memcpy(saddr.sa_data, tmp_mac, ETH_ALEN); | |
930 | saddr.sa_family = old_active->dev->type; | |
931 | ||
932 | rv = dev_set_mac_address(old_active->dev, &saddr); | |
933 | if (rv) | |
a4aee5c8 | 934 | pr_err("%s: Error %d setting MAC of slave %s\n", |
3915c1e8 JV |
935 | bond->dev->name, -rv, new_active->dev->name); |
936 | out: | |
937 | read_lock(&bond->lock); | |
938 | write_lock_bh(&bond->curr_slave_lock); | |
939 | break; | |
940 | default: | |
a4aee5c8 | 941 | pr_err("%s: bond_do_fail_over_mac impossible: bad policy %d\n", |
3915c1e8 JV |
942 | bond->dev->name, bond->params.fail_over_mac); |
943 | break; | |
944 | } | |
945 | ||
946 | } | |
947 | ||
a549952a JP |
948 | static bool bond_should_change_active(struct bonding *bond) |
949 | { | |
950 | struct slave *prim = bond->primary_slave; | |
951 | struct slave *curr = bond->curr_active_slave; | |
952 | ||
953 | if (!prim || !curr || curr->link != BOND_LINK_UP) | |
954 | return true; | |
955 | if (bond->force_primary) { | |
956 | bond->force_primary = false; | |
957 | return true; | |
958 | } | |
959 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_BETTER && | |
960 | (prim->speed < curr->speed || | |
961 | (prim->speed == curr->speed && prim->duplex <= curr->duplex))) | |
962 | return false; | |
963 | if (bond->params.primary_reselect == BOND_PRI_RESELECT_FAILURE) | |
964 | return false; | |
965 | return true; | |
966 | } | |
3915c1e8 | 967 | |
1da177e4 LT |
968 | /** |
969 | * find_best_interface - select the best available slave to be the active one | |
970 | * @bond: our bonding struct | |
971 | * | |
972 | * Warning: Caller must hold curr_slave_lock for writing. | |
973 | */ | |
974 | static struct slave *bond_find_best_slave(struct bonding *bond) | |
975 | { | |
976 | struct slave *new_active, *old_active; | |
977 | struct slave *bestslave = NULL; | |
978 | int mintime = bond->params.updelay; | |
979 | int i; | |
980 | ||
49b4ad92 | 981 | new_active = bond->curr_active_slave; |
1da177e4 LT |
982 | |
983 | if (!new_active) { /* there were no active slaves left */ | |
3d632c3f | 984 | if (bond->slave_cnt > 0) /* found one slave */ |
1da177e4 | 985 | new_active = bond->first_slave; |
3d632c3f | 986 | else |
1da177e4 | 987 | return NULL; /* still no slave, return NULL */ |
1da177e4 LT |
988 | } |
989 | ||
1da177e4 | 990 | if ((bond->primary_slave) && |
a549952a JP |
991 | bond->primary_slave->link == BOND_LINK_UP && |
992 | bond_should_change_active(bond)) { | |
1da177e4 LT |
993 | new_active = bond->primary_slave; |
994 | } | |
995 | ||
996 | /* remember where to stop iterating over the slaves */ | |
997 | old_active = new_active; | |
998 | ||
999 | bond_for_each_slave_from(bond, new_active, i, old_active) { | |
b9f60253 JP |
1000 | if (new_active->link == BOND_LINK_UP) { |
1001 | return new_active; | |
1002 | } else if (new_active->link == BOND_LINK_BACK && | |
1003 | IS_UP(new_active->dev)) { | |
1004 | /* link up, but waiting for stabilization */ | |
1005 | if (new_active->delay < mintime) { | |
1006 | mintime = new_active->delay; | |
1007 | bestslave = new_active; | |
1da177e4 LT |
1008 | } |
1009 | } | |
1010 | } | |
1011 | ||
1012 | return bestslave; | |
1013 | } | |
1014 | ||
ad246c99 BH |
1015 | static bool bond_should_notify_peers(struct bonding *bond) |
1016 | { | |
1017 | struct slave *slave = bond->curr_active_slave; | |
1018 | ||
1019 | pr_debug("bond_should_notify_peers: bond %s slave %s\n", | |
1020 | bond->dev->name, slave ? slave->dev->name : "NULL"); | |
1021 | ||
1022 | if (!slave || !bond->send_peer_notif || | |
1023 | test_bit(__LINK_STATE_LINKWATCH_PENDING, &slave->dev->state)) | |
1024 | return false; | |
1025 | ||
1026 | bond->send_peer_notif--; | |
1027 | return true; | |
1028 | } | |
1029 | ||
1da177e4 LT |
1030 | /** |
1031 | * change_active_interface - change the active slave into the specified one | |
1032 | * @bond: our bonding struct | |
1033 | * @new: the new slave to make the active one | |
1034 | * | |
1035 | * Set the new slave to the bond's settings and unset them on the old | |
1036 | * curr_active_slave. | |
1037 | * Setting include flags, mc-list, promiscuity, allmulti, etc. | |
1038 | * | |
1039 | * If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP, | |
1040 | * because it is apparently the best available slave we have, even though its | |
1041 | * updelay hasn't timed out yet. | |
1042 | * | |
3915c1e8 JV |
1043 | * If new_active is not NULL, caller must hold bond->lock for read and |
1044 | * curr_slave_lock for write_bh. | |
1da177e4 | 1045 | */ |
a77b5325 | 1046 | void bond_change_active_slave(struct bonding *bond, struct slave *new_active) |
1da177e4 LT |
1047 | { |
1048 | struct slave *old_active = bond->curr_active_slave; | |
1049 | ||
3d632c3f | 1050 | if (old_active == new_active) |
1da177e4 | 1051 | return; |
1da177e4 LT |
1052 | |
1053 | if (new_active) { | |
b2220cad JV |
1054 | new_active->jiffies = jiffies; |
1055 | ||
1da177e4 LT |
1056 | if (new_active->link == BOND_LINK_BACK) { |
1057 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1058 | pr_info("%s: making interface %s the new active one %d ms earlier.\n", |
1059 | bond->dev->name, new_active->dev->name, | |
1060 | (bond->params.updelay - new_active->delay) * bond->params.miimon); | |
1da177e4 LT |
1061 | } |
1062 | ||
1063 | new_active->delay = 0; | |
1064 | new_active->link = BOND_LINK_UP; | |
1da177e4 | 1065 | |
3d632c3f | 1066 | if (bond->params.mode == BOND_MODE_8023AD) |
1da177e4 | 1067 | bond_3ad_handle_link_change(new_active, BOND_LINK_UP); |
1da177e4 | 1068 | |
58402054 | 1069 | if (bond_is_lb(bond)) |
1da177e4 | 1070 | bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP); |
1da177e4 LT |
1071 | } else { |
1072 | if (USES_PRIMARY(bond->params.mode)) { | |
a4aee5c8 JP |
1073 | pr_info("%s: making interface %s the new active one.\n", |
1074 | bond->dev->name, new_active->dev->name); | |
1da177e4 LT |
1075 | } |
1076 | } | |
1077 | } | |
1078 | ||
3d632c3f | 1079 | if (USES_PRIMARY(bond->params.mode)) |
1da177e4 | 1080 | bond_mc_swap(bond, new_active, old_active); |
1da177e4 | 1081 | |
58402054 | 1082 | if (bond_is_lb(bond)) { |
1da177e4 | 1083 | bond_alb_handle_active_change(bond, new_active); |
8f903c70 JV |
1084 | if (old_active) |
1085 | bond_set_slave_inactive_flags(old_active); | |
1086 | if (new_active) | |
1087 | bond_set_slave_active_flags(new_active); | |
1da177e4 LT |
1088 | } else { |
1089 | bond->curr_active_slave = new_active; | |
1090 | } | |
c3ade5ca JV |
1091 | |
1092 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) { | |
3d632c3f | 1093 | if (old_active) |
c3ade5ca | 1094 | bond_set_slave_inactive_flags(old_active); |
c3ade5ca JV |
1095 | |
1096 | if (new_active) { | |
ad246c99 BH |
1097 | bool should_notify_peers = false; |
1098 | ||
c3ade5ca | 1099 | bond_set_slave_active_flags(new_active); |
2ab82852 | 1100 | |
709f8a45 OG |
1101 | if (bond->params.fail_over_mac) |
1102 | bond_do_fail_over_mac(bond, new_active, | |
1103 | old_active); | |
3915c1e8 | 1104 | |
ad246c99 BH |
1105 | if (netif_running(bond->dev)) { |
1106 | bond->send_peer_notif = | |
1107 | bond->params.num_peer_notif; | |
1108 | should_notify_peers = | |
1109 | bond_should_notify_peers(bond); | |
1110 | } | |
1111 | ||
01f3109d OG |
1112 | write_unlock_bh(&bond->curr_slave_lock); |
1113 | read_unlock(&bond->lock); | |
1114 | ||
b7bc2a5b | 1115 | call_netdevice_notifiers(NETDEV_BONDING_FAILOVER, bond->dev); |
ad246c99 | 1116 | if (should_notify_peers) |
b7bc2a5b AW |
1117 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, |
1118 | bond->dev); | |
01f3109d OG |
1119 | |
1120 | read_lock(&bond->lock); | |
1121 | write_lock_bh(&bond->curr_slave_lock); | |
7893b249 | 1122 | } |
c3ade5ca | 1123 | } |
a2fd940f | 1124 | |
5a37e8ca | 1125 | /* resend IGMP joins since active slave has changed or |
94265cf5 FL |
1126 | * all were sent on curr_active_slave. |
1127 | * resend only if bond is brought up with the affected | |
1128 | * bonding modes and the retransmission is enabled */ | |
1129 | if (netif_running(bond->dev) && (bond->params.resend_igmp > 0) && | |
1130 | ((USES_PRIMARY(bond->params.mode) && new_active) || | |
1131 | bond->params.mode == BOND_MODE_ROUNDROBIN)) { | |
c2952c31 | 1132 | bond->igmp_retrans = bond->params.resend_igmp; |
5a37e8ca | 1133 | queue_delayed_work(bond->wq, &bond->mcast_work, 0); |
a2fd940f | 1134 | } |
1da177e4 LT |
1135 | } |
1136 | ||
1137 | /** | |
1138 | * bond_select_active_slave - select a new active slave, if needed | |
1139 | * @bond: our bonding struct | |
1140 | * | |
3d632c3f | 1141 | * This functions should be called when one of the following occurs: |
1da177e4 LT |
1142 | * - The old curr_active_slave has been released or lost its link. |
1143 | * - The primary_slave has got its link back. | |
1144 | * - A slave has got its link back and there's no old curr_active_slave. | |
1145 | * | |
3915c1e8 | 1146 | * Caller must hold bond->lock for read and curr_slave_lock for write_bh. |
1da177e4 | 1147 | */ |
a77b5325 | 1148 | void bond_select_active_slave(struct bonding *bond) |
1da177e4 LT |
1149 | { |
1150 | struct slave *best_slave; | |
ff59c456 | 1151 | int rv; |
1da177e4 LT |
1152 | |
1153 | best_slave = bond_find_best_slave(bond); | |
1154 | if (best_slave != bond->curr_active_slave) { | |
1155 | bond_change_active_slave(bond, best_slave); | |
ff59c456 JV |
1156 | rv = bond_set_carrier(bond); |
1157 | if (!rv) | |
1158 | return; | |
1159 | ||
1160 | if (netif_carrier_ok(bond->dev)) { | |
a4aee5c8 JP |
1161 | pr_info("%s: first active interface up!\n", |
1162 | bond->dev->name); | |
ff59c456 | 1163 | } else { |
a4aee5c8 JP |
1164 | pr_info("%s: now running without any active interface !\n", |
1165 | bond->dev->name); | |
ff59c456 | 1166 | } |
1da177e4 LT |
1167 | } |
1168 | } | |
1169 | ||
1170 | /*--------------------------- slave list handling ---------------------------*/ | |
1171 | ||
1172 | /* | |
1173 | * This function attaches the slave to the end of list. | |
1174 | * | |
1175 | * bond->lock held for writing by caller. | |
1176 | */ | |
1177 | static void bond_attach_slave(struct bonding *bond, struct slave *new_slave) | |
1178 | { | |
1179 | if (bond->first_slave == NULL) { /* attaching the first slave */ | |
1180 | new_slave->next = new_slave; | |
1181 | new_slave->prev = new_slave; | |
1182 | bond->first_slave = new_slave; | |
1183 | } else { | |
1184 | new_slave->next = bond->first_slave; | |
1185 | new_slave->prev = bond->first_slave->prev; | |
1186 | new_slave->next->prev = new_slave; | |
1187 | new_slave->prev->next = new_slave; | |
1188 | } | |
1189 | ||
1190 | bond->slave_cnt++; | |
1191 | } | |
1192 | ||
1193 | /* | |
1194 | * This function detaches the slave from the list. | |
1195 | * WARNING: no check is made to verify if the slave effectively | |
1196 | * belongs to <bond>. | |
1197 | * Nothing is freed on return, structures are just unchained. | |
1198 | * If any slave pointer in bond was pointing to <slave>, | |
1199 | * it should be changed by the calling function. | |
1200 | * | |
1201 | * bond->lock held for writing by caller. | |
1202 | */ | |
1203 | static void bond_detach_slave(struct bonding *bond, struct slave *slave) | |
1204 | { | |
3d632c3f | 1205 | if (slave->next) |
1da177e4 | 1206 | slave->next->prev = slave->prev; |
1da177e4 | 1207 | |
3d632c3f | 1208 | if (slave->prev) |
1da177e4 | 1209 | slave->prev->next = slave->next; |
1da177e4 LT |
1210 | |
1211 | if (bond->first_slave == slave) { /* slave is the first slave */ | |
1212 | if (bond->slave_cnt > 1) { /* there are more slave */ | |
1213 | bond->first_slave = slave->next; | |
1214 | } else { | |
1215 | bond->first_slave = NULL; /* slave was the last one */ | |
1216 | } | |
1217 | } | |
1218 | ||
1219 | slave->next = NULL; | |
1220 | slave->prev = NULL; | |
1221 | bond->slave_cnt--; | |
1222 | } | |
1223 | ||
f6dc31a8 | 1224 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 1225 | static inline int slave_enable_netpoll(struct slave *slave) |
f6dc31a8 | 1226 | { |
8a8efa22 AW |
1227 | struct netpoll *np; |
1228 | int err = 0; | |
f6dc31a8 | 1229 | |
47be03a2 | 1230 | np = kzalloc(sizeof(*np), GFP_ATOMIC); |
8a8efa22 AW |
1231 | err = -ENOMEM; |
1232 | if (!np) | |
1233 | goto out; | |
1234 | ||
47be03a2 | 1235 | err = __netpoll_setup(np, slave->dev, GFP_ATOMIC); |
8a8efa22 AW |
1236 | if (err) { |
1237 | kfree(np); | |
1238 | goto out; | |
f6dc31a8 | 1239 | } |
8a8efa22 AW |
1240 | slave->np = np; |
1241 | out: | |
1242 | return err; | |
1243 | } | |
1244 | static inline void slave_disable_netpoll(struct slave *slave) | |
1245 | { | |
1246 | struct netpoll *np = slave->np; | |
1247 | ||
1248 | if (!np) | |
1249 | return; | |
1250 | ||
1251 | slave->np = NULL; | |
2cde6acd | 1252 | __netpoll_free_async(np); |
8a8efa22 AW |
1253 | } |
1254 | static inline bool slave_dev_support_netpoll(struct net_device *slave_dev) | |
1255 | { | |
1256 | if (slave_dev->priv_flags & IFF_DISABLE_NETPOLL) | |
1257 | return false; | |
1258 | if (!slave_dev->netdev_ops->ndo_poll_controller) | |
1259 | return false; | |
1260 | return true; | |
f6dc31a8 WC |
1261 | } |
1262 | ||
1263 | static void bond_poll_controller(struct net_device *bond_dev) | |
1264 | { | |
8a8efa22 AW |
1265 | } |
1266 | ||
1267 | static void __bond_netpoll_cleanup(struct bonding *bond) | |
1268 | { | |
c2355e1a NH |
1269 | struct slave *slave; |
1270 | int i; | |
1271 | ||
8a8efa22 AW |
1272 | bond_for_each_slave(bond, slave, i) |
1273 | if (IS_UP(slave->dev)) | |
1274 | slave_disable_netpoll(slave); | |
f6dc31a8 | 1275 | } |
f6dc31a8 WC |
1276 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1277 | { | |
1278 | struct bonding *bond = netdev_priv(bond_dev); | |
8a8efa22 AW |
1279 | |
1280 | read_lock(&bond->lock); | |
1281 | __bond_netpoll_cleanup(bond); | |
1282 | read_unlock(&bond->lock); | |
1283 | } | |
1284 | ||
47be03a2 | 1285 | static int bond_netpoll_setup(struct net_device *dev, struct netpoll_info *ni, gfp_t gfp) |
8a8efa22 AW |
1286 | { |
1287 | struct bonding *bond = netdev_priv(dev); | |
f6dc31a8 | 1288 | struct slave *slave; |
8a8efa22 | 1289 | int i, err = 0; |
f6dc31a8 WC |
1290 | |
1291 | read_lock(&bond->lock); | |
f6dc31a8 | 1292 | bond_for_each_slave(bond, slave, i) { |
8a8efa22 AW |
1293 | err = slave_enable_netpoll(slave); |
1294 | if (err) { | |
1295 | __bond_netpoll_cleanup(bond); | |
1296 | break; | |
f6dc31a8 WC |
1297 | } |
1298 | } | |
1299 | read_unlock(&bond->lock); | |
8a8efa22 | 1300 | return err; |
f6dc31a8 WC |
1301 | } |
1302 | ||
8a8efa22 AW |
1303 | static struct netpoll_info *bond_netpoll_info(struct bonding *bond) |
1304 | { | |
1305 | return bond->dev->npinfo; | |
1306 | } | |
f6dc31a8 | 1307 | |
8a8efa22 AW |
1308 | #else |
1309 | static inline int slave_enable_netpoll(struct slave *slave) | |
1310 | { | |
1311 | return 0; | |
1312 | } | |
1313 | static inline void slave_disable_netpoll(struct slave *slave) | |
1314 | { | |
1315 | } | |
f6dc31a8 WC |
1316 | static void bond_netpoll_cleanup(struct net_device *bond_dev) |
1317 | { | |
1318 | } | |
f6dc31a8 WC |
1319 | #endif |
1320 | ||
1da177e4 LT |
1321 | /*---------------------------------- IOCTL ----------------------------------*/ |
1322 | ||
409cc1f8 JP |
1323 | static void bond_set_dev_addr(struct net_device *bond_dev, |
1324 | struct net_device *slave_dev) | |
1da177e4 | 1325 | { |
5a03cdb7 HE |
1326 | pr_debug("bond_dev=%p\n", bond_dev); |
1327 | pr_debug("slave_dev=%p\n", slave_dev); | |
1328 | pr_debug("slave_dev->addr_len=%d\n", slave_dev->addr_len); | |
1da177e4 | 1329 | memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len); |
409cc1f8 JP |
1330 | bond_dev->addr_assign_type = NET_ADDR_SET; |
1331 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond_dev); | |
1da177e4 LT |
1332 | } |
1333 | ||
c8f44aff MM |
1334 | static netdev_features_t bond_fix_features(struct net_device *dev, |
1335 | netdev_features_t features) | |
8531c5ff | 1336 | { |
8531c5ff | 1337 | struct slave *slave; |
b2a103e6 | 1338 | struct bonding *bond = netdev_priv(dev); |
c8f44aff | 1339 | netdev_features_t mask; |
8e3babcd | 1340 | int i; |
8531c5ff | 1341 | |
b2a103e6 | 1342 | read_lock(&bond->lock); |
b63365a2 | 1343 | |
b2a103e6 MM |
1344 | if (!bond->first_slave) { |
1345 | /* Disable adding VLANs to empty bond. But why? --mq */ | |
1346 | features |= NETIF_F_VLAN_CHALLENGED; | |
1347 | goto out; | |
1348 | } | |
b63365a2 | 1349 | |
b2a103e6 | 1350 | mask = features; |
b63365a2 | 1351 | features &= ~NETIF_F_ONE_FOR_ALL; |
b2a103e6 | 1352 | features |= NETIF_F_ALL_FOR_ALL; |
7f353bf2 | 1353 | |
54ef3137 | 1354 | bond_for_each_slave(bond, slave, i) { |
b63365a2 HX |
1355 | features = netdev_increment_features(features, |
1356 | slave->dev->features, | |
b2a103e6 MM |
1357 | mask); |
1358 | } | |
1359 | ||
1360 | out: | |
1361 | read_unlock(&bond->lock); | |
1362 | return features; | |
1363 | } | |
1364 | ||
62f2a3a4 MM |
1365 | #define BOND_VLAN_FEATURES (NETIF_F_ALL_CSUM | NETIF_F_SG | \ |
1366 | NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \ | |
1367 | NETIF_F_HIGHDMA | NETIF_F_LRO) | |
b2a103e6 MM |
1368 | |
1369 | static void bond_compute_features(struct bonding *bond) | |
1370 | { | |
1371 | struct slave *slave; | |
1372 | struct net_device *bond_dev = bond->dev; | |
c8f44aff | 1373 | netdev_features_t vlan_features = BOND_VLAN_FEATURES; |
b2a103e6 | 1374 | unsigned short max_hard_header_len = ETH_HLEN; |
0e376bd0 SB |
1375 | unsigned int gso_max_size = GSO_MAX_SIZE; |
1376 | u16 gso_max_segs = GSO_MAX_SEGS; | |
b2a103e6 | 1377 | int i; |
b6fe83e9 | 1378 | unsigned int flags, dst_release_flag = IFF_XMIT_DST_RELEASE; |
b2a103e6 MM |
1379 | |
1380 | read_lock(&bond->lock); | |
1381 | ||
1382 | if (!bond->first_slave) | |
1383 | goto done; | |
1384 | ||
1385 | bond_for_each_slave(bond, slave, i) { | |
278339a4 | 1386 | vlan_features = netdev_increment_features(vlan_features, |
b2a103e6 MM |
1387 | slave->dev->vlan_features, BOND_VLAN_FEATURES); |
1388 | ||
b6fe83e9 | 1389 | dst_release_flag &= slave->dev->priv_flags; |
54ef3137 JV |
1390 | if (slave->dev->hard_header_len > max_hard_header_len) |
1391 | max_hard_header_len = slave->dev->hard_header_len; | |
0e376bd0 SB |
1392 | |
1393 | gso_max_size = min(gso_max_size, slave->dev->gso_max_size); | |
1394 | gso_max_segs = min(gso_max_segs, slave->dev->gso_max_segs); | |
54ef3137 | 1395 | } |
8531c5ff | 1396 | |
b63365a2 | 1397 | done: |
b2a103e6 | 1398 | bond_dev->vlan_features = vlan_features; |
54ef3137 | 1399 | bond_dev->hard_header_len = max_hard_header_len; |
0e376bd0 SB |
1400 | bond_dev->gso_max_segs = gso_max_segs; |
1401 | netif_set_gso_max_size(bond_dev, gso_max_size); | |
8531c5ff | 1402 | |
b6fe83e9 ED |
1403 | flags = bond_dev->priv_flags & ~IFF_XMIT_DST_RELEASE; |
1404 | bond_dev->priv_flags = flags | dst_release_flag; | |
1405 | ||
b2a103e6 MM |
1406 | read_unlock(&bond->lock); |
1407 | ||
1408 | netdev_change_features(bond_dev); | |
8531c5ff AK |
1409 | } |
1410 | ||
872254dd MS |
1411 | static void bond_setup_by_slave(struct net_device *bond_dev, |
1412 | struct net_device *slave_dev) | |
1413 | { | |
454d7c9b | 1414 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a | 1415 | |
00829823 | 1416 | bond_dev->header_ops = slave_dev->header_ops; |
872254dd MS |
1417 | |
1418 | bond_dev->type = slave_dev->type; | |
1419 | bond_dev->hard_header_len = slave_dev->hard_header_len; | |
1420 | bond_dev->addr_len = slave_dev->addr_len; | |
1421 | ||
1422 | memcpy(bond_dev->broadcast, slave_dev->broadcast, | |
1423 | slave_dev->addr_len); | |
d90a162a | 1424 | bond->setup_by_slave = 1; |
872254dd MS |
1425 | } |
1426 | ||
5b2c4dd2 | 1427 | /* On bonding slaves other than the currently active slave, suppress |
3aba891d | 1428 | * duplicates except for alb non-mcast/bcast. |
5b2c4dd2 JP |
1429 | */ |
1430 | static bool bond_should_deliver_exact_match(struct sk_buff *skb, | |
0bd80dad JP |
1431 | struct slave *slave, |
1432 | struct bonding *bond) | |
5b2c4dd2 | 1433 | { |
2d7011ca | 1434 | if (bond_is_slave_inactive(slave)) { |
0bd80dad | 1435 | if (bond->params.mode == BOND_MODE_ALB && |
5b2c4dd2 JP |
1436 | skb->pkt_type != PACKET_BROADCAST && |
1437 | skb->pkt_type != PACKET_MULTICAST) | |
5b2c4dd2 | 1438 | return false; |
5b2c4dd2 JP |
1439 | return true; |
1440 | } | |
1441 | return false; | |
1442 | } | |
1443 | ||
8a4eb573 | 1444 | static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) |
5b2c4dd2 | 1445 | { |
8a4eb573 | 1446 | struct sk_buff *skb = *pskb; |
f1c1775a | 1447 | struct slave *slave; |
0bd80dad | 1448 | struct bonding *bond; |
de063b70 ED |
1449 | int (*recv_probe)(const struct sk_buff *, struct bonding *, |
1450 | struct slave *); | |
13a8e0c8 | 1451 | int ret = RX_HANDLER_ANOTHER; |
5b2c4dd2 | 1452 | |
8a4eb573 JP |
1453 | skb = skb_share_check(skb, GFP_ATOMIC); |
1454 | if (unlikely(!skb)) | |
1455 | return RX_HANDLER_CONSUMED; | |
1456 | ||
1457 | *pskb = skb; | |
5b2c4dd2 | 1458 | |
35d48903 JP |
1459 | slave = bond_slave_get_rcu(skb->dev); |
1460 | bond = slave->bond; | |
0bd80dad JP |
1461 | |
1462 | if (bond->params.arp_interval) | |
f1c1775a | 1463 | slave->dev->last_rx = jiffies; |
5b2c4dd2 | 1464 | |
4d97480b MH |
1465 | recv_probe = ACCESS_ONCE(bond->recv_probe); |
1466 | if (recv_probe) { | |
de063b70 ED |
1467 | ret = recv_probe(skb, bond, slave); |
1468 | if (ret == RX_HANDLER_CONSUMED) { | |
1469 | consume_skb(skb); | |
1470 | return ret; | |
3aba891d JP |
1471 | } |
1472 | } | |
1473 | ||
0bd80dad | 1474 | if (bond_should_deliver_exact_match(skb, slave, bond)) { |
8a4eb573 | 1475 | return RX_HANDLER_EXACT; |
5b2c4dd2 JP |
1476 | } |
1477 | ||
35d48903 | 1478 | skb->dev = bond->dev; |
5b2c4dd2 | 1479 | |
0bd80dad | 1480 | if (bond->params.mode == BOND_MODE_ALB && |
35d48903 | 1481 | bond->dev->priv_flags & IFF_BRIDGE_PORT && |
5b2c4dd2 | 1482 | skb->pkt_type == PACKET_HOST) { |
5b2c4dd2 | 1483 | |
541ac7c9 CG |
1484 | if (unlikely(skb_cow_head(skb, |
1485 | skb->data - skb_mac_header(skb)))) { | |
1486 | kfree_skb(skb); | |
8a4eb573 | 1487 | return RX_HANDLER_CONSUMED; |
541ac7c9 | 1488 | } |
35d48903 | 1489 | memcpy(eth_hdr(skb)->h_dest, bond->dev->dev_addr, ETH_ALEN); |
5b2c4dd2 JP |
1490 | } |
1491 | ||
13a8e0c8 | 1492 | return ret; |
5b2c4dd2 JP |
1493 | } |
1494 | ||
471cb5a3 JP |
1495 | static int bond_master_upper_dev_link(struct net_device *bond_dev, |
1496 | struct net_device *slave_dev) | |
1497 | { | |
1498 | int err; | |
1499 | ||
1500 | err = netdev_master_upper_dev_link(slave_dev, bond_dev); | |
1501 | if (err) | |
1502 | return err; | |
1503 | slave_dev->flags |= IFF_SLAVE; | |
1504 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1505 | return 0; | |
1506 | } | |
1507 | ||
1508 | static void bond_upper_dev_unlink(struct net_device *bond_dev, | |
1509 | struct net_device *slave_dev) | |
1510 | { | |
1511 | netdev_upper_dev_unlink(slave_dev, bond_dev); | |
1512 | slave_dev->flags &= ~IFF_SLAVE; | |
1513 | rtmsg_ifinfo(RTM_NEWLINK, slave_dev, IFF_SLAVE); | |
1514 | } | |
1515 | ||
1da177e4 | 1516 | /* enslave device <slave> to bond device <master> */ |
a77b5325 | 1517 | int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) |
1da177e4 | 1518 | { |
454d7c9b | 1519 | struct bonding *bond = netdev_priv(bond_dev); |
eb7cc59a | 1520 | const struct net_device_ops *slave_ops = slave_dev->netdev_ops; |
1da177e4 | 1521 | struct slave *new_slave = NULL; |
22bedad3 | 1522 | struct netdev_hw_addr *ha; |
1da177e4 LT |
1523 | struct sockaddr addr; |
1524 | int link_reporting; | |
1da177e4 LT |
1525 | int res = 0; |
1526 | ||
c772dde3 BH |
1527 | if (!bond->params.use_carrier && |
1528 | slave_dev->ethtool_ops->get_link == NULL && | |
1529 | slave_ops->ndo_do_ioctl == NULL) { | |
a4aee5c8 JP |
1530 | pr_warning("%s: Warning: no link monitoring support for %s\n", |
1531 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1532 | } |
1533 | ||
1da177e4 LT |
1534 | /* already enslaved */ |
1535 | if (slave_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 1536 | pr_debug("Error, Device was already enslaved\n"); |
1da177e4 LT |
1537 | return -EBUSY; |
1538 | } | |
1539 | ||
1540 | /* vlan challenged mutual exclusion */ | |
1541 | /* no need to lock since we're protected by rtnl_lock */ | |
1542 | if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) { | |
5a03cdb7 | 1543 | pr_debug("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
55462cf3 | 1544 | if (vlan_uses_dev(bond_dev)) { |
a4aee5c8 JP |
1545 | pr_err("%s: Error: cannot enslave VLAN challenged slave %s on VLAN enabled bond %s\n", |
1546 | bond_dev->name, slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1547 | return -EPERM; |
1548 | } else { | |
a4aee5c8 JP |
1549 | pr_warning("%s: Warning: enslaved VLAN challenged slave %s. Adding VLANs will be blocked as long as %s is part of bond %s\n", |
1550 | bond_dev->name, slave_dev->name, | |
1551 | slave_dev->name, bond_dev->name); | |
1da177e4 LT |
1552 | } |
1553 | } else { | |
5a03cdb7 | 1554 | pr_debug("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name); |
1da177e4 LT |
1555 | } |
1556 | ||
217df670 JV |
1557 | /* |
1558 | * Old ifenslave binaries are no longer supported. These can | |
3d632c3f | 1559 | * be identified with moderate accuracy by the state of the slave: |
217df670 JV |
1560 | * the current ifenslave will set the interface down prior to |
1561 | * enslaving it; the old ifenslave will not. | |
1562 | */ | |
1563 | if ((slave_dev->flags & IFF_UP)) { | |
a4aee5c8 | 1564 | pr_err("%s is up. This may be due to an out of date ifenslave.\n", |
217df670 JV |
1565 | slave_dev->name); |
1566 | res = -EPERM; | |
1567 | goto err_undo_flags; | |
1568 | } | |
1da177e4 | 1569 | |
872254dd MS |
1570 | /* set bonding device ether type by slave - bonding netdevices are |
1571 | * created with ether_setup, so when the slave type is not ARPHRD_ETHER | |
1572 | * there is a need to override some of the type dependent attribs/funcs. | |
1573 | * | |
1574 | * bond ether type mutual exclusion - don't allow slaves of dissimilar | |
1575 | * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond | |
1576 | */ | |
1577 | if (bond->slave_cnt == 0) { | |
e36b9d16 | 1578 | if (bond_dev->type != slave_dev->type) { |
e36b9d16 | 1579 | pr_debug("%s: change device type from %d to %d\n", |
a4aee5c8 JP |
1580 | bond_dev->name, |
1581 | bond_dev->type, slave_dev->type); | |
75c78500 | 1582 | |
b7bc2a5b AW |
1583 | res = call_netdevice_notifiers(NETDEV_PRE_TYPE_CHANGE, |
1584 | bond_dev); | |
3ca5b404 JP |
1585 | res = notifier_to_errno(res); |
1586 | if (res) { | |
1587 | pr_err("%s: refused to change device type\n", | |
1588 | bond_dev->name); | |
1589 | res = -EBUSY; | |
1590 | goto err_undo_flags; | |
1591 | } | |
75c78500 | 1592 | |
32a806c1 | 1593 | /* Flush unicast and multicast addresses */ |
a748ee24 | 1594 | dev_uc_flush(bond_dev); |
22bedad3 | 1595 | dev_mc_flush(bond_dev); |
32a806c1 | 1596 | |
e36b9d16 MS |
1597 | if (slave_dev->type != ARPHRD_ETHER) |
1598 | bond_setup_by_slave(bond_dev, slave_dev); | |
550fd08c | 1599 | else { |
e36b9d16 | 1600 | ether_setup(bond_dev); |
550fd08c NH |
1601 | bond_dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
1602 | } | |
75c78500 | 1603 | |
b7bc2a5b AW |
1604 | call_netdevice_notifiers(NETDEV_POST_TYPE_CHANGE, |
1605 | bond_dev); | |
e36b9d16 | 1606 | } |
872254dd | 1607 | } else if (bond_dev->type != slave_dev->type) { |
a4aee5c8 JP |
1608 | pr_err("%s ether type (%d) is different from other slaves (%d), can not enslave it.\n", |
1609 | slave_dev->name, | |
1610 | slave_dev->type, bond_dev->type); | |
1611 | res = -EINVAL; | |
1612 | goto err_undo_flags; | |
872254dd MS |
1613 | } |
1614 | ||
eb7cc59a | 1615 | if (slave_ops->ndo_set_mac_address == NULL) { |
2ab82852 | 1616 | if (bond->slave_cnt == 0) { |
a4aee5c8 JP |
1617 | pr_warning("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active.", |
1618 | bond_dev->name); | |
3915c1e8 JV |
1619 | bond->params.fail_over_mac = BOND_FOM_ACTIVE; |
1620 | } else if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { | |
a4aee5c8 JP |
1621 | 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", |
1622 | bond_dev->name); | |
2ab82852 MS |
1623 | res = -EOPNOTSUPP; |
1624 | goto err_undo_flags; | |
1625 | } | |
1da177e4 LT |
1626 | } |
1627 | ||
8d8fc29d AW |
1628 | call_netdevice_notifiers(NETDEV_JOIN, slave_dev); |
1629 | ||
c20811a7 JP |
1630 | /* If this is the first slave, then we need to set the master's hardware |
1631 | * address to be the same as the slave's. */ | |
6c8c4e4c | 1632 | if (bond->slave_cnt == 0 && bond->dev_addr_from_first) |
409cc1f8 | 1633 | bond_set_dev_addr(bond->dev, slave_dev); |
c20811a7 | 1634 | |
243cb4e5 | 1635 | new_slave = kzalloc(sizeof(struct slave), GFP_KERNEL); |
1da177e4 LT |
1636 | if (!new_slave) { |
1637 | res = -ENOMEM; | |
1638 | goto err_undo_flags; | |
1639 | } | |
1640 | ||
bb1d9123 AG |
1641 | /* |
1642 | * Set the new_slave's queue_id to be zero. Queue ID mapping | |
1643 | * is set via sysfs or module option if desired. | |
1644 | */ | |
1645 | new_slave->queue_id = 0; | |
1646 | ||
b15ba0fb JP |
1647 | /* Save slave's original mtu and then set it to match the bond */ |
1648 | new_slave->original_mtu = slave_dev->mtu; | |
1649 | res = dev_set_mtu(slave_dev, bond->dev->mtu); | |
1650 | if (res) { | |
1651 | pr_debug("Error %d calling dev_set_mtu\n", res); | |
1652 | goto err_free; | |
1653 | } | |
1654 | ||
217df670 JV |
1655 | /* |
1656 | * Save slave's original ("permanent") mac address for modes | |
1657 | * that need it, and for restoring it upon release, and then | |
1658 | * set it to the master's address | |
1659 | */ | |
1660 | memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN); | |
1da177e4 | 1661 | |
dd957c57 | 1662 | if (!bond->params.fail_over_mac) { |
2ab82852 MS |
1663 | /* |
1664 | * Set slave to master's mac address. The application already | |
1665 | * set the master's mac address to that of the first slave | |
1666 | */ | |
1667 | memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
1668 | addr.sa_family = slave_dev->type; | |
1669 | res = dev_set_mac_address(slave_dev, &addr); | |
1670 | if (res) { | |
5a03cdb7 | 1671 | pr_debug("Error %d calling set_mac_address\n", res); |
b15ba0fb | 1672 | goto err_restore_mtu; |
2ab82852 | 1673 | } |
217df670 | 1674 | } |
1da177e4 | 1675 | |
471cb5a3 | 1676 | res = bond_master_upper_dev_link(bond_dev, slave_dev); |
c2edacf8 | 1677 | if (res) { |
471cb5a3 | 1678 | pr_debug("Error %d calling bond_master_upper_dev_link\n", res); |
569f0c4d | 1679 | goto err_restore_mac; |
c2edacf8 | 1680 | } |
5b2c4dd2 | 1681 | |
217df670 JV |
1682 | /* open the slave since the application closed it */ |
1683 | res = dev_open(slave_dev); | |
1684 | if (res) { | |
3d632c3f | 1685 | pr_debug("Opening slave %s failed\n", slave_dev->name); |
35d48903 | 1686 | goto err_unset_master; |
1da177e4 LT |
1687 | } |
1688 | ||
35d48903 | 1689 | new_slave->bond = bond; |
1da177e4 | 1690 | new_slave->dev = slave_dev; |
0b680e75 | 1691 | slave_dev->priv_flags |= IFF_BONDING; |
1da177e4 | 1692 | |
58402054 | 1693 | if (bond_is_lb(bond)) { |
1da177e4 LT |
1694 | /* bond_alb_init_slave() must be called before all other stages since |
1695 | * it might fail and we do not want to have to undo everything | |
1696 | */ | |
1697 | res = bond_alb_init_slave(bond, new_slave); | |
3d632c3f | 1698 | if (res) |
569f0c4d | 1699 | goto err_close; |
1da177e4 LT |
1700 | } |
1701 | ||
1702 | /* If the mode USES_PRIMARY, then the new slave gets the | |
1703 | * master's promisc (and mc) settings only if it becomes the | |
1704 | * curr_active_slave, and that is taken care of later when calling | |
1705 | * bond_change_active() | |
1706 | */ | |
1707 | if (!USES_PRIMARY(bond->params.mode)) { | |
1708 | /* set promiscuity level to new slave */ | |
1709 | if (bond_dev->flags & IFF_PROMISC) { | |
7e1a1ac1 WC |
1710 | res = dev_set_promiscuity(slave_dev, 1); |
1711 | if (res) | |
1712 | goto err_close; | |
1da177e4 LT |
1713 | } |
1714 | ||
1715 | /* set allmulti level to new slave */ | |
1716 | if (bond_dev->flags & IFF_ALLMULTI) { | |
7e1a1ac1 WC |
1717 | res = dev_set_allmulti(slave_dev, 1); |
1718 | if (res) | |
1719 | goto err_close; | |
1da177e4 LT |
1720 | } |
1721 | ||
b9e40857 | 1722 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 1723 | /* upload master's mc_list to new slave */ |
22bedad3 JP |
1724 | netdev_for_each_mc_addr(ha, bond_dev) |
1725 | dev_mc_add(slave_dev, ha->addr); | |
b9e40857 | 1726 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
1727 | } |
1728 | ||
1729 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1730 | /* add lacpdu mc addr to mc list */ | |
1731 | u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR; | |
1732 | ||
22bedad3 | 1733 | dev_mc_add(slave_dev, lacpdu_multicast); |
1da177e4 LT |
1734 | } |
1735 | ||
1736 | bond_add_vlans_on_slave(bond, slave_dev); | |
1737 | ||
1738 | write_lock_bh(&bond->lock); | |
1739 | ||
1740 | bond_attach_slave(bond, new_slave); | |
1741 | ||
1742 | new_slave->delay = 0; | |
1743 | new_slave->link_failure_count = 0; | |
1744 | ||
3915c1e8 JV |
1745 | write_unlock_bh(&bond->lock); |
1746 | ||
b2a103e6 MM |
1747 | bond_compute_features(bond); |
1748 | ||
876254ae VF |
1749 | bond_update_speed_duplex(new_slave); |
1750 | ||
3915c1e8 JV |
1751 | read_lock(&bond->lock); |
1752 | ||
f31c7937 MK |
1753 | new_slave->last_arp_rx = jiffies - |
1754 | (msecs_to_jiffies(bond->params.arp_interval) + 1); | |
f5b2b966 | 1755 | |
1da177e4 LT |
1756 | if (bond->params.miimon && !bond->params.use_carrier) { |
1757 | link_reporting = bond_check_dev_link(bond, slave_dev, 1); | |
1758 | ||
1759 | if ((link_reporting == -1) && !bond->params.arp_interval) { | |
1760 | /* | |
1761 | * miimon is set but a bonded network driver | |
1762 | * does not support ETHTOOL/MII and | |
1763 | * arp_interval is not set. Note: if | |
1764 | * use_carrier is enabled, we will never go | |
1765 | * here (because netif_carrier is always | |
1766 | * supported); thus, we don't need to change | |
1767 | * the messages for netif_carrier. | |
1768 | */ | |
a4aee5c8 | 1769 | 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 | 1770 | bond_dev->name, slave_dev->name); |
1da177e4 LT |
1771 | } else if (link_reporting == -1) { |
1772 | /* unable get link status using mii/ethtool */ | |
a4aee5c8 JP |
1773 | 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", |
1774 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1775 | } |
1776 | } | |
1777 | ||
1778 | /* check for initial state */ | |
f31c7937 MK |
1779 | if (bond->params.miimon) { |
1780 | if (bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS) { | |
1781 | if (bond->params.updelay) { | |
1782 | new_slave->link = BOND_LINK_BACK; | |
1783 | new_slave->delay = bond->params.updelay; | |
1784 | } else { | |
1785 | new_slave->link = BOND_LINK_UP; | |
1786 | } | |
1da177e4 | 1787 | } else { |
f31c7937 | 1788 | new_slave->link = BOND_LINK_DOWN; |
1da177e4 | 1789 | } |
f31c7937 MK |
1790 | } else if (bond->params.arp_interval) { |
1791 | new_slave->link = (netif_carrier_ok(slave_dev) ? | |
1792 | BOND_LINK_UP : BOND_LINK_DOWN); | |
1da177e4 | 1793 | } else { |
f31c7937 | 1794 | new_slave->link = BOND_LINK_UP; |
1da177e4 LT |
1795 | } |
1796 | ||
f31c7937 MK |
1797 | if (new_slave->link != BOND_LINK_DOWN) |
1798 | new_slave->jiffies = jiffies; | |
1799 | pr_debug("Initial state of slave_dev is BOND_LINK_%s\n", | |
1800 | new_slave->link == BOND_LINK_DOWN ? "DOWN" : | |
1801 | (new_slave->link == BOND_LINK_UP ? "UP" : "BACK")); | |
1802 | ||
1da177e4 LT |
1803 | if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) { |
1804 | /* if there is a primary slave, remember it */ | |
a549952a | 1805 | if (strcmp(bond->params.primary, new_slave->dev->name) == 0) { |
1da177e4 | 1806 | bond->primary_slave = new_slave; |
a549952a JP |
1807 | bond->force_primary = true; |
1808 | } | |
1da177e4 LT |
1809 | } |
1810 | ||
3915c1e8 JV |
1811 | write_lock_bh(&bond->curr_slave_lock); |
1812 | ||
1da177e4 LT |
1813 | switch (bond->params.mode) { |
1814 | case BOND_MODE_ACTIVEBACKUP: | |
8a8e447b JV |
1815 | bond_set_slave_inactive_flags(new_slave); |
1816 | bond_select_active_slave(bond); | |
1da177e4 LT |
1817 | break; |
1818 | case BOND_MODE_8023AD: | |
1819 | /* in 802.3ad mode, the internal mechanism | |
1820 | * will activate the slaves in the selected | |
1821 | * aggregator | |
1822 | */ | |
1823 | bond_set_slave_inactive_flags(new_slave); | |
1824 | /* if this is the first slave */ | |
1825 | if (bond->slave_cnt == 1) { | |
1826 | SLAVE_AD_INFO(new_slave).id = 1; | |
1827 | /* Initialize AD with the number of times that the AD timer is called in 1 second | |
1828 | * can be called only after the mac address of the bond is set | |
1829 | */ | |
56d00c67 | 1830 | bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL); |
1da177e4 LT |
1831 | } else { |
1832 | SLAVE_AD_INFO(new_slave).id = | |
1833 | SLAVE_AD_INFO(new_slave->prev).id + 1; | |
1834 | } | |
1835 | ||
1836 | bond_3ad_bind_slave(new_slave); | |
1837 | break; | |
1838 | case BOND_MODE_TLB: | |
1839 | case BOND_MODE_ALB: | |
e30bc066 | 1840 | bond_set_active_slave(new_slave); |
059fe7a5 | 1841 | bond_set_slave_inactive_flags(new_slave); |
5a29f789 | 1842 | bond_select_active_slave(bond); |
1da177e4 LT |
1843 | break; |
1844 | default: | |
5a03cdb7 | 1845 | pr_debug("This slave is always active in trunk mode\n"); |
1da177e4 LT |
1846 | |
1847 | /* always active in trunk mode */ | |
e30bc066 | 1848 | bond_set_active_slave(new_slave); |
1da177e4 LT |
1849 | |
1850 | /* In trunking mode there is little meaning to curr_active_slave | |
1851 | * anyway (it holds no special properties of the bond device), | |
1852 | * so we can change it without calling change_active_interface() | |
1853 | */ | |
4e591b93 | 1854 | if (!bond->curr_active_slave && new_slave->link == BOND_LINK_UP) |
1da177e4 | 1855 | bond->curr_active_slave = new_slave; |
3d632c3f | 1856 | |
1da177e4 LT |
1857 | break; |
1858 | } /* switch(bond_mode) */ | |
1859 | ||
3915c1e8 JV |
1860 | write_unlock_bh(&bond->curr_slave_lock); |
1861 | ||
ff59c456 JV |
1862 | bond_set_carrier(bond); |
1863 | ||
f6dc31a8 | 1864 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 AW |
1865 | slave_dev->npinfo = bond_netpoll_info(bond); |
1866 | if (slave_dev->npinfo) { | |
1867 | if (slave_enable_netpoll(new_slave)) { | |
1868 | read_unlock(&bond->lock); | |
1869 | pr_info("Error, %s: master_dev is using netpoll, " | |
1870 | "but new slave device does not support netpoll.\n", | |
1871 | bond_dev->name); | |
1872 | res = -EBUSY; | |
f7d9821a | 1873 | goto err_detach; |
8a8efa22 | 1874 | } |
f6dc31a8 WC |
1875 | } |
1876 | #endif | |
8a8efa22 | 1877 | |
3915c1e8 | 1878 | read_unlock(&bond->lock); |
1da177e4 | 1879 | |
b76cdba9 MW |
1880 | res = bond_create_slave_symlinks(bond_dev, slave_dev); |
1881 | if (res) | |
f7d9821a | 1882 | goto err_detach; |
b76cdba9 | 1883 | |
35d48903 JP |
1884 | res = netdev_rx_handler_register(slave_dev, bond_handle_frame, |
1885 | new_slave); | |
1886 | if (res) { | |
1887 | pr_debug("Error %d calling netdev_rx_handler_register\n", res); | |
1888 | goto err_dest_symlinks; | |
1889 | } | |
1890 | ||
a4aee5c8 JP |
1891 | pr_info("%s: enslaving %s as a%s interface with a%s link.\n", |
1892 | bond_dev->name, slave_dev->name, | |
e30bc066 | 1893 | bond_is_active_slave(new_slave) ? "n active" : " backup", |
a4aee5c8 | 1894 | new_slave->link != BOND_LINK_DOWN ? "n up" : " down"); |
1da177e4 LT |
1895 | |
1896 | /* enslave is successful */ | |
1897 | return 0; | |
1898 | ||
1899 | /* Undo stages on error */ | |
35d48903 JP |
1900 | err_dest_symlinks: |
1901 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
1902 | ||
f7d9821a | 1903 | err_detach: |
1904 | write_lock_bh(&bond->lock); | |
1905 | bond_detach_slave(bond, new_slave); | |
1906 | write_unlock_bh(&bond->lock); | |
1907 | ||
1da177e4 LT |
1908 | err_close: |
1909 | dev_close(slave_dev); | |
1910 | ||
569f0c4d | 1911 | err_unset_master: |
471cb5a3 | 1912 | bond_upper_dev_unlink(bond_dev, slave_dev); |
569f0c4d | 1913 | |
1da177e4 | 1914 | err_restore_mac: |
dd957c57 | 1915 | if (!bond->params.fail_over_mac) { |
3915c1e8 JV |
1916 | /* XXX TODO - fom follow mode needs to change master's |
1917 | * MAC if this slave's MAC is in use by the bond, or at | |
1918 | * least print a warning. | |
1919 | */ | |
2ab82852 MS |
1920 | memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN); |
1921 | addr.sa_family = slave_dev->type; | |
1922 | dev_set_mac_address(slave_dev, &addr); | |
1923 | } | |
1da177e4 | 1924 | |
b15ba0fb JP |
1925 | err_restore_mtu: |
1926 | dev_set_mtu(slave_dev, new_slave->original_mtu); | |
1927 | ||
1da177e4 LT |
1928 | err_free: |
1929 | kfree(new_slave); | |
1930 | ||
1931 | err_undo_flags: | |
b2a103e6 | 1932 | bond_compute_features(bond); |
3d632c3f | 1933 | |
1da177e4 LT |
1934 | return res; |
1935 | } | |
1936 | ||
1937 | /* | |
1938 | * Try to release the slave device <slave> from the bond device <master> | |
1939 | * It is legal to access curr_active_slave without a lock because all the function | |
0896341a | 1940 | * is write-locked. If "all" is true it means that the function is being called |
1941 | * while destroying a bond interface and all slaves are being released. | |
1da177e4 LT |
1942 | * |
1943 | * The rules for slave state should be: | |
1944 | * for Active/Backup: | |
1945 | * Active stays on all backups go down | |
1946 | * for Bonded connections: | |
1947 | * The first up interface should be left on and all others downed. | |
1948 | */ | |
0896341a | 1949 | static int __bond_release_one(struct net_device *bond_dev, |
1950 | struct net_device *slave_dev, | |
1951 | bool all) | |
1da177e4 | 1952 | { |
454d7c9b | 1953 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
1954 | struct slave *slave, *oldcurrent; |
1955 | struct sockaddr addr; | |
c8f44aff | 1956 | netdev_features_t old_features = bond_dev->features; |
1da177e4 LT |
1957 | |
1958 | /* slave is not a slave or master is not master of this slave */ | |
1959 | if (!(slave_dev->flags & IFF_SLAVE) || | |
471cb5a3 | 1960 | !netdev_has_upper_dev(slave_dev, bond_dev)) { |
a4aee5c8 | 1961 | pr_err("%s: Error: cannot release %s.\n", |
1da177e4 LT |
1962 | bond_dev->name, slave_dev->name); |
1963 | return -EINVAL; | |
1964 | } | |
1965 | ||
e843fa50 | 1966 | block_netpoll_tx(); |
1da177e4 LT |
1967 | write_lock_bh(&bond->lock); |
1968 | ||
1969 | slave = bond_get_slave_by_dev(bond, slave_dev); | |
1970 | if (!slave) { | |
1971 | /* not a slave of this bond */ | |
a4aee5c8 JP |
1972 | pr_info("%s: %s not enslaved\n", |
1973 | bond_dev->name, slave_dev->name); | |
f5e2a7b2 | 1974 | write_unlock_bh(&bond->lock); |
e843fa50 | 1975 | unblock_netpoll_tx(); |
1da177e4 LT |
1976 | return -EINVAL; |
1977 | } | |
1978 | ||
fcd99434 | 1979 | write_unlock_bh(&bond->lock); |
35d48903 JP |
1980 | /* unregister rx_handler early so bond_handle_frame wouldn't be called |
1981 | * for this slave anymore. | |
1982 | */ | |
1983 | netdev_rx_handler_unregister(slave_dev); | |
35d48903 JP |
1984 | write_lock_bh(&bond->lock); |
1985 | ||
0896341a | 1986 | if (!all && !bond->params.fail_over_mac) { |
2e42e474 | 1987 | if (ether_addr_equal(bond_dev->dev_addr, slave->perm_hwaddr) && |
8e95a202 | 1988 | bond->slave_cnt > 1) |
a4aee5c8 JP |
1989 | 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", |
1990 | bond_dev->name, slave_dev->name, | |
1991 | slave->perm_hwaddr, | |
1992 | bond_dev->name, slave_dev->name); | |
1da177e4 LT |
1993 | } |
1994 | ||
1995 | /* Inform AD package of unbinding of slave. */ | |
1996 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1997 | /* must be called before the slave is | |
1998 | * detached from the list | |
1999 | */ | |
2000 | bond_3ad_unbind_slave(slave); | |
2001 | } | |
2002 | ||
a4aee5c8 JP |
2003 | pr_info("%s: releasing %s interface %s\n", |
2004 | bond_dev->name, | |
e30bc066 | 2005 | bond_is_active_slave(slave) ? "active" : "backup", |
a4aee5c8 | 2006 | slave_dev->name); |
1da177e4 LT |
2007 | |
2008 | oldcurrent = bond->curr_active_slave; | |
2009 | ||
2010 | bond->current_arp_slave = NULL; | |
2011 | ||
2012 | /* release the slave from its bond */ | |
2013 | bond_detach_slave(bond, slave); | |
2014 | ||
3d632c3f | 2015 | if (bond->primary_slave == slave) |
1da177e4 | 2016 | bond->primary_slave = NULL; |
1da177e4 | 2017 | |
3d632c3f | 2018 | if (oldcurrent == slave) |
1da177e4 | 2019 | bond_change_active_slave(bond, NULL); |
1da177e4 | 2020 | |
58402054 | 2021 | if (bond_is_lb(bond)) { |
1da177e4 LT |
2022 | /* Must be called only after the slave has been |
2023 | * detached from the list and the curr_active_slave | |
2024 | * has been cleared (if our_slave == old_current), | |
2025 | * but before a new active slave is selected. | |
2026 | */ | |
2543331d | 2027 | write_unlock_bh(&bond->lock); |
1da177e4 | 2028 | bond_alb_deinit_slave(bond, slave); |
2543331d | 2029 | write_lock_bh(&bond->lock); |
1da177e4 LT |
2030 | } |
2031 | ||
0896341a | 2032 | if (all) { |
2033 | bond->curr_active_slave = NULL; | |
2034 | } else if (oldcurrent == slave) { | |
059fe7a5 JV |
2035 | /* |
2036 | * Note that we hold RTNL over this sequence, so there | |
2037 | * is no concern that another slave add/remove event | |
2038 | * will interfere. | |
2039 | */ | |
2040 | write_unlock_bh(&bond->lock); | |
2041 | read_lock(&bond->lock); | |
2042 | write_lock_bh(&bond->curr_slave_lock); | |
2043 | ||
1da177e4 LT |
2044 | bond_select_active_slave(bond); |
2045 | ||
059fe7a5 JV |
2046 | write_unlock_bh(&bond->curr_slave_lock); |
2047 | read_unlock(&bond->lock); | |
2048 | write_lock_bh(&bond->lock); | |
2049 | } | |
2050 | ||
1da177e4 | 2051 | if (bond->slave_cnt == 0) { |
ff59c456 | 2052 | bond_set_carrier(bond); |
409cc1f8 JP |
2053 | eth_hw_addr_random(bond_dev); |
2054 | bond->dev_addr_from_first = true; | |
1da177e4 | 2055 | |
cc0e4070 | 2056 | if (bond_vlan_used(bond)) { |
a4aee5c8 JP |
2057 | pr_warning("%s: Warning: clearing HW address of %s while it still has VLANs.\n", |
2058 | bond_dev->name, bond_dev->name); | |
2059 | pr_warning("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n", | |
2060 | bond_dev->name); | |
1da177e4 | 2061 | } |
1da177e4 LT |
2062 | } |
2063 | ||
2064 | write_unlock_bh(&bond->lock); | |
e843fa50 | 2065 | unblock_netpoll_tx(); |
1da177e4 | 2066 | |
80028ea1 | 2067 | if (bond->slave_cnt == 0) { |
2af73d4b | 2068 | call_netdevice_notifiers(NETDEV_CHANGEADDR, bond->dev); |
80028ea1 VF |
2069 | call_netdevice_notifiers(NETDEV_RELEASE, bond->dev); |
2070 | } | |
2af73d4b | 2071 | |
b2a103e6 MM |
2072 | bond_compute_features(bond); |
2073 | if (!(bond_dev->features & NETIF_F_VLAN_CHALLENGED) && | |
2074 | (old_features & NETIF_F_VLAN_CHALLENGED)) | |
2075 | pr_info("%s: last VLAN challenged slave %s left bond %s. VLAN blocking is removed\n", | |
2076 | bond_dev->name, slave_dev->name, bond_dev->name); | |
2077 | ||
b76cdba9 MW |
2078 | /* must do this from outside any spinlocks */ |
2079 | bond_destroy_slave_symlinks(bond_dev, slave_dev); | |
2080 | ||
1da177e4 LT |
2081 | bond_del_vlans_from_slave(bond, slave_dev); |
2082 | ||
2083 | /* If the mode USES_PRIMARY, then we should only remove its | |
2084 | * promisc and mc settings if it was the curr_active_slave, but that was | |
2085 | * already taken care of above when we detached the slave | |
2086 | */ | |
2087 | if (!USES_PRIMARY(bond->params.mode)) { | |
2088 | /* unset promiscuity level from slave */ | |
3d632c3f | 2089 | if (bond_dev->flags & IFF_PROMISC) |
1da177e4 | 2090 | dev_set_promiscuity(slave_dev, -1); |
1da177e4 LT |
2091 | |
2092 | /* unset allmulti level from slave */ | |
3d632c3f | 2093 | if (bond_dev->flags & IFF_ALLMULTI) |
1da177e4 | 2094 | dev_set_allmulti(slave_dev, -1); |
1da177e4 LT |
2095 | |
2096 | /* flush master's mc_list from slave */ | |
b9e40857 | 2097 | netif_addr_lock_bh(bond_dev); |
1da177e4 | 2098 | bond_mc_list_flush(bond_dev, slave_dev); |
b9e40857 | 2099 | netif_addr_unlock_bh(bond_dev); |
1da177e4 LT |
2100 | } |
2101 | ||
471cb5a3 | 2102 | bond_upper_dev_unlink(bond_dev, slave_dev); |
1da177e4 | 2103 | |
8a8efa22 | 2104 | slave_disable_netpoll(slave); |
f6dc31a8 | 2105 | |
1da177e4 LT |
2106 | /* close slave before restoring its mac address */ |
2107 | dev_close(slave_dev); | |
2108 | ||
3915c1e8 | 2109 | if (bond->params.fail_over_mac != BOND_FOM_ACTIVE) { |
2ab82852 MS |
2110 | /* restore original ("permanent") mac address */ |
2111 | memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN); | |
2112 | addr.sa_family = slave_dev->type; | |
2113 | dev_set_mac_address(slave_dev, &addr); | |
2114 | } | |
1da177e4 | 2115 | |
b15ba0fb JP |
2116 | dev_set_mtu(slave_dev, slave->original_mtu); |
2117 | ||
2d7011ca | 2118 | slave_dev->priv_flags &= ~IFF_BONDING; |
1da177e4 LT |
2119 | |
2120 | kfree(slave); | |
2121 | ||
2122 | return 0; /* deletion OK */ | |
2123 | } | |
2124 | ||
0896341a | 2125 | /* A wrapper used because of ndo_del_link */ |
2126 | int bond_release(struct net_device *bond_dev, struct net_device *slave_dev) | |
2127 | { | |
2128 | return __bond_release_one(bond_dev, slave_dev, false); | |
2129 | } | |
2130 | ||
d90a162a | 2131 | /* |
dadaa10b | 2132 | * First release a slave and then destroy the bond if no more slaves are left. |
d90a162a MS |
2133 | * Must be under rtnl_lock when this function is called. |
2134 | */ | |
26d8ee75 | 2135 | static int bond_release_and_destroy(struct net_device *bond_dev, |
2136 | struct net_device *slave_dev) | |
d90a162a | 2137 | { |
454d7c9b | 2138 | struct bonding *bond = netdev_priv(bond_dev); |
d90a162a MS |
2139 | int ret; |
2140 | ||
2141 | ret = bond_release(bond_dev, slave_dev); | |
2142 | if ((ret == 0) && (bond->slave_cnt == 0)) { | |
8a8efa22 | 2143 | bond_dev->priv_flags |= IFF_DISABLE_NETPOLL; |
a4aee5c8 JP |
2144 | pr_info("%s: destroying bond %s.\n", |
2145 | bond_dev->name, bond_dev->name); | |
9e71626c | 2146 | unregister_netdevice(bond_dev); |
d90a162a MS |
2147 | } |
2148 | return ret; | |
2149 | } | |
2150 | ||
1da177e4 LT |
2151 | /* |
2152 | * This function changes the active slave to slave <slave_dev>. | |
2153 | * It returns -EINVAL in the following cases. | |
2154 | * - <slave_dev> is not found in the list. | |
2155 | * - There is not active slave now. | |
2156 | * - <slave_dev> is already active. | |
2157 | * - The link state of <slave_dev> is not BOND_LINK_UP. | |
2158 | * - <slave_dev> is not running. | |
3d632c3f SH |
2159 | * In these cases, this function does nothing. |
2160 | * In the other cases, current_slave pointer is changed and 0 is returned. | |
1da177e4 LT |
2161 | */ |
2162 | static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev) | |
2163 | { | |
454d7c9b | 2164 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2165 | struct slave *old_active = NULL; |
2166 | struct slave *new_active = NULL; | |
2167 | int res = 0; | |
2168 | ||
3d632c3f | 2169 | if (!USES_PRIMARY(bond->params.mode)) |
1da177e4 | 2170 | return -EINVAL; |
1da177e4 | 2171 | |
471cb5a3 JP |
2172 | /* Verify that bond_dev is indeed the master of slave_dev */ |
2173 | if (!(slave_dev->flags & IFF_SLAVE) || | |
2174 | !netdev_has_upper_dev(slave_dev, bond_dev)) | |
1da177e4 | 2175 | return -EINVAL; |
1da177e4 | 2176 | |
059fe7a5 | 2177 | read_lock(&bond->lock); |
1da177e4 | 2178 | |
059fe7a5 | 2179 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 2180 | old_active = bond->curr_active_slave; |
059fe7a5 JV |
2181 | read_unlock(&bond->curr_slave_lock); |
2182 | ||
1da177e4 LT |
2183 | new_active = bond_get_slave_by_dev(bond, slave_dev); |
2184 | ||
2185 | /* | |
2186 | * Changing to the current active: do nothing; return success. | |
2187 | */ | |
2188 | if (new_active && (new_active == old_active)) { | |
059fe7a5 | 2189 | read_unlock(&bond->lock); |
1da177e4 LT |
2190 | return 0; |
2191 | } | |
2192 | ||
2193 | if ((new_active) && | |
2194 | (old_active) && | |
2195 | (new_active->link == BOND_LINK_UP) && | |
2196 | IS_UP(new_active->dev)) { | |
e843fa50 | 2197 | block_netpoll_tx(); |
059fe7a5 | 2198 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 | 2199 | bond_change_active_slave(bond, new_active); |
059fe7a5 | 2200 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2201 | unblock_netpoll_tx(); |
3d632c3f | 2202 | } else |
1da177e4 | 2203 | res = -EINVAL; |
1da177e4 | 2204 | |
059fe7a5 | 2205 | read_unlock(&bond->lock); |
1da177e4 LT |
2206 | |
2207 | return res; | |
2208 | } | |
2209 | ||
1da177e4 LT |
2210 | static int bond_info_query(struct net_device *bond_dev, struct ifbond *info) |
2211 | { | |
454d7c9b | 2212 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
2213 | |
2214 | info->bond_mode = bond->params.mode; | |
2215 | info->miimon = bond->params.miimon; | |
2216 | ||
6603a6f2 | 2217 | read_lock(&bond->lock); |
1da177e4 | 2218 | info->num_slaves = bond->slave_cnt; |
6603a6f2 | 2219 | read_unlock(&bond->lock); |
1da177e4 LT |
2220 | |
2221 | return 0; | |
2222 | } | |
2223 | ||
2224 | static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info) | |
2225 | { | |
454d7c9b | 2226 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 2227 | struct slave *slave; |
689c96cc | 2228 | int i, res = -ENODEV; |
1da177e4 | 2229 | |
6603a6f2 | 2230 | read_lock(&bond->lock); |
1da177e4 LT |
2231 | |
2232 | bond_for_each_slave(bond, slave, i) { | |
2233 | if (i == (int)info->slave_id) { | |
689c96cc ED |
2234 | res = 0; |
2235 | strcpy(info->slave_name, slave->dev->name); | |
2236 | info->link = slave->link; | |
e30bc066 | 2237 | info->state = bond_slave_state(slave); |
689c96cc | 2238 | info->link_failure_count = slave->link_failure_count; |
1da177e4 LT |
2239 | break; |
2240 | } | |
2241 | } | |
2242 | ||
6603a6f2 | 2243 | read_unlock(&bond->lock); |
1da177e4 | 2244 | |
689c96cc | 2245 | return res; |
1da177e4 LT |
2246 | } |
2247 | ||
2248 | /*-------------------------------- Monitoring -------------------------------*/ | |
2249 | ||
1da177e4 | 2250 | |
f0c76d61 JV |
2251 | static int bond_miimon_inspect(struct bonding *bond) |
2252 | { | |
2253 | struct slave *slave; | |
2254 | int i, link_state, commit = 0; | |
41f89100 JP |
2255 | bool ignore_updelay; |
2256 | ||
2257 | ignore_updelay = !bond->curr_active_slave ? true : false; | |
1da177e4 LT |
2258 | |
2259 | bond_for_each_slave(bond, slave, i) { | |
f0c76d61 | 2260 | slave->new_link = BOND_LINK_NOCHANGE; |
1da177e4 | 2261 | |
f0c76d61 | 2262 | link_state = bond_check_dev_link(bond, slave->dev, 0); |
1da177e4 LT |
2263 | |
2264 | switch (slave->link) { | |
f0c76d61 JV |
2265 | case BOND_LINK_UP: |
2266 | if (link_state) | |
2267 | continue; | |
1da177e4 | 2268 | |
f0c76d61 JV |
2269 | slave->link = BOND_LINK_FAIL; |
2270 | slave->delay = bond->params.downdelay; | |
2271 | if (slave->delay) { | |
a4aee5c8 JP |
2272 | pr_info("%s: link status down for %sinterface %s, disabling it in %d ms.\n", |
2273 | bond->dev->name, | |
2274 | (bond->params.mode == | |
2275 | BOND_MODE_ACTIVEBACKUP) ? | |
e30bc066 | 2276 | (bond_is_active_slave(slave) ? |
a4aee5c8 JP |
2277 | "active " : "backup ") : "", |
2278 | slave->dev->name, | |
2279 | bond->params.downdelay * bond->params.miimon); | |
1da177e4 | 2280 | } |
f0c76d61 JV |
2281 | /*FALLTHRU*/ |
2282 | case BOND_LINK_FAIL: | |
2283 | if (link_state) { | |
2284 | /* | |
2285 | * recovered before downdelay expired | |
2286 | */ | |
2287 | slave->link = BOND_LINK_UP; | |
1da177e4 | 2288 | slave->jiffies = jiffies; |
a4aee5c8 JP |
2289 | pr_info("%s: link status up again after %d ms for interface %s.\n", |
2290 | bond->dev->name, | |
2291 | (bond->params.downdelay - slave->delay) * | |
2292 | bond->params.miimon, | |
2293 | slave->dev->name); | |
f0c76d61 | 2294 | continue; |
1da177e4 | 2295 | } |
f0c76d61 JV |
2296 | |
2297 | if (slave->delay <= 0) { | |
2298 | slave->new_link = BOND_LINK_DOWN; | |
2299 | commit++; | |
2300 | continue; | |
1da177e4 | 2301 | } |
1da177e4 | 2302 | |
f0c76d61 JV |
2303 | slave->delay--; |
2304 | break; | |
2305 | ||
2306 | case BOND_LINK_DOWN: | |
2307 | if (!link_state) | |
2308 | continue; | |
2309 | ||
2310 | slave->link = BOND_LINK_BACK; | |
2311 | slave->delay = bond->params.updelay; | |
2312 | ||
2313 | if (slave->delay) { | |
a4aee5c8 JP |
2314 | pr_info("%s: link status up for interface %s, enabling it in %d ms.\n", |
2315 | bond->dev->name, slave->dev->name, | |
2316 | ignore_updelay ? 0 : | |
2317 | bond->params.updelay * | |
2318 | bond->params.miimon); | |
f0c76d61 JV |
2319 | } |
2320 | /*FALLTHRU*/ | |
2321 | case BOND_LINK_BACK: | |
2322 | if (!link_state) { | |
2323 | slave->link = BOND_LINK_DOWN; | |
a4aee5c8 JP |
2324 | pr_info("%s: link status down again after %d ms for interface %s.\n", |
2325 | bond->dev->name, | |
2326 | (bond->params.updelay - slave->delay) * | |
2327 | bond->params.miimon, | |
2328 | slave->dev->name); | |
f0c76d61 JV |
2329 | |
2330 | continue; | |
2331 | } | |
2332 | ||
41f89100 JP |
2333 | if (ignore_updelay) |
2334 | slave->delay = 0; | |
2335 | ||
f0c76d61 JV |
2336 | if (slave->delay <= 0) { |
2337 | slave->new_link = BOND_LINK_UP; | |
2338 | commit++; | |
41f89100 | 2339 | ignore_updelay = false; |
f0c76d61 | 2340 | continue; |
1da177e4 | 2341 | } |
f0c76d61 JV |
2342 | |
2343 | slave->delay--; | |
1da177e4 | 2344 | break; |
f0c76d61 JV |
2345 | } |
2346 | } | |
1da177e4 | 2347 | |
f0c76d61 JV |
2348 | return commit; |
2349 | } | |
1da177e4 | 2350 | |
f0c76d61 JV |
2351 | static void bond_miimon_commit(struct bonding *bond) |
2352 | { | |
2353 | struct slave *slave; | |
2354 | int i; | |
2355 | ||
2356 | bond_for_each_slave(bond, slave, i) { | |
2357 | switch (slave->new_link) { | |
2358 | case BOND_LINK_NOCHANGE: | |
2359 | continue; | |
1da177e4 | 2360 | |
f0c76d61 JV |
2361 | case BOND_LINK_UP: |
2362 | slave->link = BOND_LINK_UP; | |
2363 | slave->jiffies = jiffies; | |
2364 | ||
2365 | if (bond->params.mode == BOND_MODE_8023AD) { | |
2366 | /* prevent it from being the active one */ | |
e30bc066 | 2367 | bond_set_backup_slave(slave); |
f0c76d61 JV |
2368 | } else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) { |
2369 | /* make it immediately active */ | |
e30bc066 | 2370 | bond_set_active_slave(slave); |
f0c76d61 JV |
2371 | } else if (slave != bond->primary_slave) { |
2372 | /* prevent it from being the active one */ | |
e30bc066 | 2373 | bond_set_backup_slave(slave); |
1da177e4 | 2374 | } |
1da177e4 | 2375 | |
5d30530e | 2376 | pr_info("%s: link status definitely up for interface %s, %u Mbps %s duplex.\n", |
700c2a77 KPO |
2377 | bond->dev->name, slave->dev->name, |
2378 | slave->speed, slave->duplex ? "full" : "half"); | |
1da177e4 | 2379 | |
f0c76d61 JV |
2380 | /* notify ad that the link status has changed */ |
2381 | if (bond->params.mode == BOND_MODE_8023AD) | |
2382 | bond_3ad_handle_link_change(slave, BOND_LINK_UP); | |
059fe7a5 | 2383 | |
58402054 | 2384 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2385 | bond_alb_handle_link_change(bond, slave, |
2386 | BOND_LINK_UP); | |
1da177e4 | 2387 | |
f0c76d61 JV |
2388 | if (!bond->curr_active_slave || |
2389 | (slave == bond->primary_slave)) | |
2390 | goto do_failover; | |
1da177e4 | 2391 | |
f0c76d61 | 2392 | continue; |
059fe7a5 | 2393 | |
f0c76d61 | 2394 | case BOND_LINK_DOWN: |
fba4acda JV |
2395 | if (slave->link_failure_count < UINT_MAX) |
2396 | slave->link_failure_count++; | |
2397 | ||
f0c76d61 | 2398 | slave->link = BOND_LINK_DOWN; |
1da177e4 | 2399 | |
f0c76d61 JV |
2400 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP || |
2401 | bond->params.mode == BOND_MODE_8023AD) | |
2402 | bond_set_slave_inactive_flags(slave); | |
2403 | ||
a4aee5c8 JP |
2404 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
2405 | bond->dev->name, slave->dev->name); | |
f0c76d61 JV |
2406 | |
2407 | if (bond->params.mode == BOND_MODE_8023AD) | |
2408 | bond_3ad_handle_link_change(slave, | |
2409 | BOND_LINK_DOWN); | |
2410 | ||
ae63e808 | 2411 | if (bond_is_lb(bond)) |
f0c76d61 JV |
2412 | bond_alb_handle_link_change(bond, slave, |
2413 | BOND_LINK_DOWN); | |
2414 | ||
2415 | if (slave == bond->curr_active_slave) | |
2416 | goto do_failover; | |
2417 | ||
2418 | continue; | |
2419 | ||
2420 | default: | |
a4aee5c8 | 2421 | pr_err("%s: invalid new link %d on slave %s\n", |
f0c76d61 JV |
2422 | bond->dev->name, slave->new_link, |
2423 | slave->dev->name); | |
2424 | slave->new_link = BOND_LINK_NOCHANGE; | |
2425 | ||
2426 | continue; | |
2427 | } | |
2428 | ||
2429 | do_failover: | |
2430 | ASSERT_RTNL(); | |
e843fa50 | 2431 | block_netpoll_tx(); |
f0c76d61 JV |
2432 | write_lock_bh(&bond->curr_slave_lock); |
2433 | bond_select_active_slave(bond); | |
2434 | write_unlock_bh(&bond->curr_slave_lock); | |
e843fa50 | 2435 | unblock_netpoll_tx(); |
f0c76d61 JV |
2436 | } |
2437 | ||
2438 | bond_set_carrier(bond); | |
1da177e4 LT |
2439 | } |
2440 | ||
0b0eef66 JV |
2441 | /* |
2442 | * bond_mii_monitor | |
2443 | * | |
2444 | * Really a wrapper that splits the mii monitor into two phases: an | |
f0c76d61 JV |
2445 | * inspection, then (if inspection indicates something needs to be done) |
2446 | * an acquisition of appropriate locks followed by a commit phase to | |
2447 | * implement whatever link state changes are indicated. | |
0b0eef66 JV |
2448 | */ |
2449 | void bond_mii_monitor(struct work_struct *work) | |
2450 | { | |
2451 | struct bonding *bond = container_of(work, struct bonding, | |
2452 | mii_work.work); | |
ad246c99 | 2453 | bool should_notify_peers = false; |
e6d265e8 | 2454 | unsigned long delay; |
0b0eef66 JV |
2455 | |
2456 | read_lock(&bond->lock); | |
e6d265e8 JV |
2457 | |
2458 | delay = msecs_to_jiffies(bond->params.miimon); | |
f0c76d61 JV |
2459 | |
2460 | if (bond->slave_cnt == 0) | |
2461 | goto re_arm; | |
b59f9f74 | 2462 | |
ad246c99 BH |
2463 | should_notify_peers = bond_should_notify_peers(bond); |
2464 | ||
f0c76d61 | 2465 | if (bond_miimon_inspect(bond)) { |
0b0eef66 | 2466 | read_unlock(&bond->lock); |
e6d265e8 JV |
2467 | |
2468 | /* Race avoidance with bond_close cancel of workqueue */ | |
2469 | if (!rtnl_trylock()) { | |
2470 | read_lock(&bond->lock); | |
2471 | delay = 1; | |
2472 | should_notify_peers = false; | |
2473 | goto re_arm; | |
2474 | } | |
2475 | ||
0b0eef66 | 2476 | read_lock(&bond->lock); |
f0c76d61 JV |
2477 | |
2478 | bond_miimon_commit(bond); | |
2479 | ||
5655662d JV |
2480 | read_unlock(&bond->lock); |
2481 | rtnl_unlock(); /* might sleep, hold no other locks */ | |
2482 | read_lock(&bond->lock); | |
0b0eef66 JV |
2483 | } |
2484 | ||
f0c76d61 | 2485 | re_arm: |
e6d265e8 JV |
2486 | if (bond->params.miimon) |
2487 | queue_delayed_work(bond->wq, &bond->mii_work, delay); | |
2488 | ||
0b0eef66 | 2489 | read_unlock(&bond->lock); |
ad246c99 BH |
2490 | |
2491 | if (should_notify_peers) { | |
e6d265e8 JV |
2492 | if (!rtnl_trylock()) { |
2493 | read_lock(&bond->lock); | |
2494 | bond->send_peer_notif++; | |
2495 | read_unlock(&bond->lock); | |
2496 | return; | |
2497 | } | |
b7bc2a5b | 2498 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
2499 | rtnl_unlock(); |
2500 | } | |
0b0eef66 | 2501 | } |
c3ade5ca | 2502 | |
d3bb52b0 | 2503 | static int bond_has_this_ip(struct bonding *bond, __be32 ip) |
f5b2b966 | 2504 | { |
0883beca | 2505 | struct vlan_entry *vlan; |
eaddcd76 | 2506 | struct net_device *vlan_dev; |
f5b2b966 | 2507 | |
eaddcd76 | 2508 | if (ip == bond_confirm_addr(bond->dev, 0, ip)) |
f5b2b966 JV |
2509 | return 1; |
2510 | ||
0883beca | 2511 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
eaddcd76 AG |
2512 | rcu_read_lock(); |
2513 | vlan_dev = __vlan_find_dev_deep(bond->dev, vlan->vlan_id); | |
2514 | rcu_read_unlock(); | |
2515 | if (vlan_dev && ip == bond_confirm_addr(vlan_dev, 0, ip)) | |
f5b2b966 JV |
2516 | return 1; |
2517 | } | |
2518 | ||
2519 | return 0; | |
2520 | } | |
2521 | ||
c3ade5ca JV |
2522 | /* |
2523 | * We go to the (large) trouble of VLAN tagging ARP frames because | |
2524 | * switches in VLAN mode (especially if ports are configured as | |
2525 | * "native" to a VLAN) might not pass non-tagged frames. | |
2526 | */ | |
d3bb52b0 | 2527 | 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 |
2528 | { |
2529 | struct sk_buff *skb; | |
2530 | ||
5a03cdb7 | 2531 | pr_debug("arp %d on slave %s: dst %x src %x vid %d\n", arp_op, |
a4aee5c8 | 2532 | slave_dev->name, dest_ip, src_ip, vlan_id); |
3d632c3f | 2533 | |
c3ade5ca JV |
2534 | skb = arp_create(arp_op, ETH_P_ARP, dest_ip, slave_dev, src_ip, |
2535 | NULL, slave_dev->dev_addr, NULL); | |
2536 | ||
2537 | if (!skb) { | |
a4aee5c8 | 2538 | pr_err("ARP packet allocation failed\n"); |
c3ade5ca JV |
2539 | return; |
2540 | } | |
2541 | if (vlan_id) { | |
2542 | skb = vlan_put_tag(skb, vlan_id); | |
2543 | if (!skb) { | |
a4aee5c8 | 2544 | pr_err("failed to insert VLAN tag\n"); |
c3ade5ca JV |
2545 | return; |
2546 | } | |
2547 | } | |
2548 | arp_xmit(skb); | |
2549 | } | |
2550 | ||
2551 | ||
1da177e4 LT |
2552 | static void bond_arp_send_all(struct bonding *bond, struct slave *slave) |
2553 | { | |
b23dd4fe | 2554 | int i, vlan_id; |
d3bb52b0 | 2555 | __be32 *targets = bond->params.arp_targets; |
0883beca | 2556 | struct vlan_entry *vlan; |
eaddcd76 | 2557 | struct net_device *vlan_dev = NULL; |
c3ade5ca | 2558 | struct rtable *rt; |
1da177e4 | 2559 | |
6b780567 | 2560 | for (i = 0; (i < BOND_MAX_ARP_TARGETS); i++) { |
eaddcd76 | 2561 | __be32 addr; |
6b780567 | 2562 | if (!targets[i]) |
5a31bec0 | 2563 | break; |
5a03cdb7 | 2564 | pr_debug("basa: target %x\n", targets[i]); |
cc0e4070 | 2565 | if (!bond_vlan_used(bond)) { |
5a03cdb7 | 2566 | pr_debug("basa: empty vlan: arp_send\n"); |
eaddcd76 | 2567 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2568 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2569 | addr, 0); |
c3ade5ca JV |
2570 | continue; |
2571 | } | |
2572 | ||
2573 | /* | |
2574 | * If VLANs are configured, we do a route lookup to | |
2575 | * determine which VLAN interface would be used, so we | |
2576 | * can tag the ARP with the proper VLAN tag. | |
2577 | */ | |
78fbfd8a DM |
2578 | rt = ip_route_output(dev_net(bond->dev), targets[i], 0, |
2579 | RTO_ONLINK, 0); | |
b23dd4fe | 2580 | if (IS_ERR(rt)) { |
c3ade5ca | 2581 | if (net_ratelimit()) { |
a4aee5c8 | 2582 | pr_warning("%s: no route to arp_ip_target %pI4\n", |
78fbfd8a | 2583 | bond->dev->name, &targets[i]); |
c3ade5ca JV |
2584 | } |
2585 | continue; | |
2586 | } | |
2587 | ||
2588 | /* | |
2589 | * This target is not on a VLAN | |
2590 | */ | |
d8d1f30b | 2591 | if (rt->dst.dev == bond->dev) { |
ed4b9f80 | 2592 | ip_rt_put(rt); |
5a03cdb7 | 2593 | pr_debug("basa: rtdev == bond->dev: arp_send\n"); |
eaddcd76 | 2594 | addr = bond_confirm_addr(bond->dev, targets[i], 0); |
c3ade5ca | 2595 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2596 | addr, 0); |
c3ade5ca JV |
2597 | continue; |
2598 | } | |
2599 | ||
2600 | vlan_id = 0; | |
0883beca | 2601 | list_for_each_entry(vlan, &bond->vlan_list, vlan_list) { |
cc0e4070 JP |
2602 | rcu_read_lock(); |
2603 | vlan_dev = __vlan_find_dev_deep(bond->dev, | |
2604 | vlan->vlan_id); | |
2605 | rcu_read_unlock(); | |
d8d1f30b | 2606 | if (vlan_dev == rt->dst.dev) { |
c3ade5ca | 2607 | vlan_id = vlan->vlan_id; |
5a03cdb7 | 2608 | pr_debug("basa: vlan match on %s %d\n", |
c3ade5ca JV |
2609 | vlan_dev->name, vlan_id); |
2610 | break; | |
2611 | } | |
2612 | } | |
2613 | ||
eaddcd76 | 2614 | if (vlan_id && vlan_dev) { |
ed4b9f80 | 2615 | ip_rt_put(rt); |
eaddcd76 | 2616 | addr = bond_confirm_addr(vlan_dev, targets[i], 0); |
c3ade5ca | 2617 | bond_arp_send(slave->dev, ARPOP_REQUEST, targets[i], |
eaddcd76 | 2618 | addr, vlan_id); |
c3ade5ca JV |
2619 | continue; |
2620 | } | |
2621 | ||
2622 | if (net_ratelimit()) { | |
a4aee5c8 | 2623 | pr_warning("%s: no path to arp_ip_target %pI4 via rt.dev %s\n", |
78fbfd8a | 2624 | bond->dev->name, &targets[i], |
d8d1f30b | 2625 | rt->dst.dev ? rt->dst.dev->name : "NULL"); |
c3ade5ca | 2626 | } |
ed4b9f80 | 2627 | ip_rt_put(rt); |
c3ade5ca JV |
2628 | } |
2629 | } | |
2630 | ||
d3bb52b0 | 2631 | static void bond_validate_arp(struct bonding *bond, struct slave *slave, __be32 sip, __be32 tip) |
f5b2b966 JV |
2632 | { |
2633 | int i; | |
d3bb52b0 | 2634 | __be32 *targets = bond->params.arp_targets; |
f5b2b966 | 2635 | |
f5b2b966 | 2636 | for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) { |
5a03cdb7 | 2637 | pr_debug("bva: sip %pI4 tip %pI4 t[%d] %pI4 bhti(tip) %d\n", |
a4aee5c8 JP |
2638 | &sip, &tip, i, &targets[i], |
2639 | bond_has_this_ip(bond, tip)); | |
f5b2b966 JV |
2640 | if (sip == targets[i]) { |
2641 | if (bond_has_this_ip(bond, tip)) | |
2642 | slave->last_arp_rx = jiffies; | |
2643 | return; | |
2644 | } | |
2645 | } | |
2646 | } | |
2647 | ||
de063b70 ED |
2648 | static int bond_arp_rcv(const struct sk_buff *skb, struct bonding *bond, |
2649 | struct slave *slave) | |
f5b2b966 | 2650 | { |
de063b70 | 2651 | struct arphdr *arp = (struct arphdr *)skb->data; |
f5b2b966 | 2652 | unsigned char *arp_ptr; |
d3bb52b0 | 2653 | __be32 sip, tip; |
de063b70 | 2654 | int alen; |
f5b2b966 | 2655 | |
3aba891d | 2656 | if (skb->protocol != __cpu_to_be16(ETH_P_ARP)) |
13a8e0c8 | 2657 | return RX_HANDLER_ANOTHER; |
f5b2b966 | 2658 | |
f5b2b966 | 2659 | read_lock(&bond->lock); |
de063b70 | 2660 | alen = arp_hdr_len(bond->dev); |
f5b2b966 | 2661 | |
3aba891d JP |
2662 | pr_debug("bond_arp_rcv: bond %s skb->dev %s\n", |
2663 | bond->dev->name, skb->dev->name); | |
f5b2b966 | 2664 | |
de063b70 ED |
2665 | if (alen > skb_headlen(skb)) { |
2666 | arp = kmalloc(alen, GFP_ATOMIC); | |
2667 | if (!arp) | |
2668 | goto out_unlock; | |
2669 | if (skb_copy_bits(skb, 0, arp, alen) < 0) | |
2670 | goto out_unlock; | |
2671 | } | |
f5b2b966 | 2672 | |
3aba891d | 2673 | if (arp->ar_hln != bond->dev->addr_len || |
f5b2b966 JV |
2674 | skb->pkt_type == PACKET_OTHERHOST || |
2675 | skb->pkt_type == PACKET_LOOPBACK || | |
2676 | arp->ar_hrd != htons(ARPHRD_ETHER) || | |
2677 | arp->ar_pro != htons(ETH_P_IP) || | |
2678 | arp->ar_pln != 4) | |
2679 | goto out_unlock; | |
2680 | ||
2681 | arp_ptr = (unsigned char *)(arp + 1); | |
3aba891d | 2682 | arp_ptr += bond->dev->addr_len; |
f5b2b966 | 2683 | memcpy(&sip, arp_ptr, 4); |
3aba891d | 2684 | arp_ptr += 4 + bond->dev->addr_len; |
f5b2b966 JV |
2685 | memcpy(&tip, arp_ptr, 4); |
2686 | ||
5a03cdb7 | 2687 | pr_debug("bond_arp_rcv: %s %s/%d av %d sv %d sip %pI4 tip %pI4\n", |
e30bc066 | 2688 | bond->dev->name, slave->dev->name, bond_slave_state(slave), |
a4aee5c8 JP |
2689 | bond->params.arp_validate, slave_do_arp_validate(bond, slave), |
2690 | &sip, &tip); | |
f5b2b966 JV |
2691 | |
2692 | /* | |
2693 | * Backup slaves won't see the ARP reply, but do come through | |
2694 | * here for each ARP probe (so we swap the sip/tip to validate | |
2695 | * the probe). In a "redundant switch, common router" type of | |
2696 | * configuration, the ARP probe will (hopefully) travel from | |
2697 | * the active, through one switch, the router, then the other | |
2698 | * switch before reaching the backup. | |
2699 | */ | |
e30bc066 | 2700 | if (bond_is_active_slave(slave)) |
f5b2b966 JV |
2701 | bond_validate_arp(bond, slave, sip, tip); |
2702 | else | |
2703 | bond_validate_arp(bond, slave, tip, sip); | |
2704 | ||
2705 | out_unlock: | |
2706 | read_unlock(&bond->lock); | |
de063b70 ED |
2707 | if (arp != (struct arphdr *)skb->data) |
2708 | kfree(arp); | |
13a8e0c8 | 2709 | return RX_HANDLER_ANOTHER; |
f5b2b966 JV |
2710 | } |
2711 | ||
1da177e4 LT |
2712 | /* |
2713 | * this function is called regularly to monitor each slave's link | |
2714 | * ensuring that traffic is being sent and received when arp monitoring | |
2715 | * is used in load-balancing mode. if the adapter has been dormant, then an | |
2716 | * arp is transmitted to generate traffic. see activebackup_arp_monitor for | |
2717 | * arp monitoring in active backup mode. | |
2718 | */ | |
1b76b316 | 2719 | void bond_loadbalance_arp_mon(struct work_struct *work) |
1da177e4 | 2720 | { |
1b76b316 JV |
2721 | struct bonding *bond = container_of(work, struct bonding, |
2722 | arp_work.work); | |
1da177e4 LT |
2723 | struct slave *slave, *oldcurrent; |
2724 | int do_failover = 0; | |
da210f55 | 2725 | int delta_in_ticks, extra_ticks; |
1da177e4 LT |
2726 | int i; |
2727 | ||
2728 | read_lock(&bond->lock); | |
2729 | ||
5ce0da8f | 2730 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
da210f55 | 2731 | extra_ticks = delta_in_ticks / 2; |
1da177e4 | 2732 | |
3d632c3f | 2733 | if (bond->slave_cnt == 0) |
1da177e4 | 2734 | goto re_arm; |
1da177e4 LT |
2735 | |
2736 | read_lock(&bond->curr_slave_lock); | |
2737 | oldcurrent = bond->curr_active_slave; | |
2738 | read_unlock(&bond->curr_slave_lock); | |
2739 | ||
2740 | /* see if any of the previous devices are up now (i.e. they have | |
2741 | * xmt and rcv traffic). the curr_active_slave does not come into | |
2742 | * the picture unless it is null. also, slave->jiffies is not needed | |
2743 | * here because we send an arp on each slave and give a slave as | |
2744 | * long as it needs to get the tx/rx within the delta. | |
2745 | * TODO: what about up/down delay in arp mode? it wasn't here before | |
2746 | * so it can wait | |
2747 | */ | |
2748 | bond_for_each_slave(bond, slave, i) { | |
cb32f2a0 JB |
2749 | unsigned long trans_start = dev_trans_start(slave->dev); |
2750 | ||
1da177e4 | 2751 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2752 | if (time_in_range(jiffies, |
2753 | trans_start - delta_in_ticks, | |
da210f55 | 2754 | trans_start + delta_in_ticks + extra_ticks) && |
cb32f2a0 JB |
2755 | time_in_range(jiffies, |
2756 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2757 | slave->dev->last_rx + delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2758 | |
2759 | slave->link = BOND_LINK_UP; | |
e30bc066 | 2760 | bond_set_active_slave(slave); |
1da177e4 LT |
2761 | |
2762 | /* primary_slave has no meaning in round-robin | |
2763 | * mode. the window of a slave being up and | |
2764 | * curr_active_slave being null after enslaving | |
2765 | * is closed. | |
2766 | */ | |
2767 | if (!oldcurrent) { | |
a4aee5c8 JP |
2768 | pr_info("%s: link status definitely up for interface %s, ", |
2769 | bond->dev->name, | |
2770 | slave->dev->name); | |
1da177e4 LT |
2771 | do_failover = 1; |
2772 | } else { | |
a4aee5c8 JP |
2773 | pr_info("%s: interface %s is now up\n", |
2774 | bond->dev->name, | |
2775 | slave->dev->name); | |
1da177e4 LT |
2776 | } |
2777 | } | |
2778 | } else { | |
2779 | /* slave->link == BOND_LINK_UP */ | |
2780 | ||
2781 | /* not all switches will respond to an arp request | |
2782 | * when the source ip is 0, so don't take the link down | |
2783 | * if we don't know our ip yet | |
2784 | */ | |
cb32f2a0 JB |
2785 | if (!time_in_range(jiffies, |
2786 | trans_start - delta_in_ticks, | |
da210f55 | 2787 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2788 | !time_in_range(jiffies, |
2789 | slave->dev->last_rx - delta_in_ticks, | |
da210f55 | 2790 | slave->dev->last_rx + 2 * delta_in_ticks + extra_ticks)) { |
1da177e4 LT |
2791 | |
2792 | slave->link = BOND_LINK_DOWN; | |
e30bc066 | 2793 | bond_set_backup_slave(slave); |
1da177e4 | 2794 | |
3d632c3f | 2795 | if (slave->link_failure_count < UINT_MAX) |
1da177e4 | 2796 | slave->link_failure_count++; |
1da177e4 | 2797 | |
a4aee5c8 JP |
2798 | pr_info("%s: interface %s is now down.\n", |
2799 | bond->dev->name, | |
2800 | slave->dev->name); | |
1da177e4 | 2801 | |
3d632c3f | 2802 | if (slave == oldcurrent) |
1da177e4 | 2803 | do_failover = 1; |
1da177e4 LT |
2804 | } |
2805 | } | |
2806 | ||
2807 | /* note: if switch is in round-robin mode, all links | |
2808 | * must tx arp to ensure all links rx an arp - otherwise | |
2809 | * links may oscillate or not come up at all; if switch is | |
2810 | * in something like xor mode, there is nothing we can | |
2811 | * do - all replies will be rx'ed on same link causing slaves | |
2812 | * to be unstable during low/no traffic periods | |
2813 | */ | |
3d632c3f | 2814 | if (IS_UP(slave->dev)) |
1da177e4 | 2815 | bond_arp_send_all(bond, slave); |
1da177e4 LT |
2816 | } |
2817 | ||
2818 | if (do_failover) { | |
e843fa50 | 2819 | block_netpoll_tx(); |
059fe7a5 | 2820 | write_lock_bh(&bond->curr_slave_lock); |
1da177e4 LT |
2821 | |
2822 | bond_select_active_slave(bond); | |
2823 | ||
059fe7a5 | 2824 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 2825 | unblock_netpoll_tx(); |
1da177e4 LT |
2826 | } |
2827 | ||
2828 | re_arm: | |
e6d265e8 | 2829 | if (bond->params.arp_interval) |
1b76b316 | 2830 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 2831 | |
1da177e4 LT |
2832 | read_unlock(&bond->lock); |
2833 | } | |
2834 | ||
2835 | /* | |
b2220cad JV |
2836 | * Called to inspect slaves for active-backup mode ARP monitor link state |
2837 | * changes. Sets new_link in slaves to specify what action should take | |
2838 | * place for the slave. Returns 0 if no changes are found, >0 if changes | |
2839 | * to link states must be committed. | |
2840 | * | |
2841 | * Called with bond->lock held for read. | |
1da177e4 | 2842 | */ |
b2220cad | 2843 | static int bond_ab_arp_inspect(struct bonding *bond, int delta_in_ticks) |
1da177e4 | 2844 | { |
1da177e4 | 2845 | struct slave *slave; |
b2220cad | 2846 | int i, commit = 0; |
cb32f2a0 | 2847 | unsigned long trans_start; |
da210f55 JB |
2848 | int extra_ticks; |
2849 | ||
2850 | /* All the time comparisons below need some extra time. Otherwise, on | |
2851 | * fast networks the ARP probe/reply may arrive within the same jiffy | |
2852 | * as it was sent. Then, the next time the ARP monitor is run, one | |
2853 | * arp_interval will already have passed in the comparisons. | |
2854 | */ | |
2855 | extra_ticks = delta_in_ticks / 2; | |
1da177e4 | 2856 | |
b2220cad JV |
2857 | bond_for_each_slave(bond, slave, i) { |
2858 | slave->new_link = BOND_LINK_NOCHANGE; | |
1da177e4 | 2859 | |
b2220cad | 2860 | if (slave->link != BOND_LINK_UP) { |
cb32f2a0 JB |
2861 | if (time_in_range(jiffies, |
2862 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2863 | slave_last_rx(bond, slave) + delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2864 | |
b2220cad JV |
2865 | slave->new_link = BOND_LINK_UP; |
2866 | commit++; | |
2867 | } | |
1da177e4 | 2868 | |
b2220cad JV |
2869 | continue; |
2870 | } | |
1da177e4 | 2871 | |
b2220cad JV |
2872 | /* |
2873 | * Give slaves 2*delta after being enslaved or made | |
2874 | * active. This avoids bouncing, as the last receive | |
2875 | * times need a full ARP monitor cycle to be updated. | |
2876 | */ | |
cb32f2a0 JB |
2877 | if (time_in_range(jiffies, |
2878 | slave->jiffies - delta_in_ticks, | |
da210f55 | 2879 | slave->jiffies + 2 * delta_in_ticks + extra_ticks)) |
b2220cad JV |
2880 | continue; |
2881 | ||
2882 | /* | |
2883 | * Backup slave is down if: | |
2884 | * - No current_arp_slave AND | |
2885 | * - more than 3*delta since last receive AND | |
2886 | * - the bond has an IP address | |
2887 | * | |
2888 | * Note: a non-null current_arp_slave indicates | |
2889 | * the curr_active_slave went down and we are | |
2890 | * searching for a new one; under this condition | |
2891 | * we only take the curr_active_slave down - this | |
2892 | * gives each slave a chance to tx/rx traffic | |
2893 | * before being taken out | |
2894 | */ | |
e30bc066 | 2895 | if (!bond_is_active_slave(slave) && |
b2220cad | 2896 | !bond->current_arp_slave && |
cb32f2a0 JB |
2897 | !time_in_range(jiffies, |
2898 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2899 | slave_last_rx(bond, slave) + 3 * delta_in_ticks + extra_ticks)) { |
cb32f2a0 | 2900 | |
b2220cad JV |
2901 | slave->new_link = BOND_LINK_DOWN; |
2902 | commit++; | |
2903 | } | |
2904 | ||
2905 | /* | |
2906 | * Active slave is down if: | |
2907 | * - more than 2*delta since transmitting OR | |
2908 | * - (more than 2*delta since receive AND | |
2909 | * the bond has an IP address) | |
2910 | */ | |
cb32f2a0 | 2911 | trans_start = dev_trans_start(slave->dev); |
e30bc066 | 2912 | if (bond_is_active_slave(slave) && |
cb32f2a0 JB |
2913 | (!time_in_range(jiffies, |
2914 | trans_start - delta_in_ticks, | |
da210f55 | 2915 | trans_start + 2 * delta_in_ticks + extra_ticks) || |
cb32f2a0 JB |
2916 | !time_in_range(jiffies, |
2917 | slave_last_rx(bond, slave) - delta_in_ticks, | |
da210f55 | 2918 | slave_last_rx(bond, slave) + 2 * delta_in_ticks + extra_ticks))) { |
cb32f2a0 | 2919 | |
b2220cad JV |
2920 | slave->new_link = BOND_LINK_DOWN; |
2921 | commit++; | |
2922 | } | |
1da177e4 LT |
2923 | } |
2924 | ||
b2220cad JV |
2925 | return commit; |
2926 | } | |
1da177e4 | 2927 | |
b2220cad JV |
2928 | /* |
2929 | * Called to commit link state changes noted by inspection step of | |
2930 | * active-backup mode ARP monitor. | |
2931 | * | |
2932 | * Called with RTNL and bond->lock for read. | |
2933 | */ | |
2934 | static void bond_ab_arp_commit(struct bonding *bond, int delta_in_ticks) | |
2935 | { | |
2936 | struct slave *slave; | |
2937 | int i; | |
cb32f2a0 | 2938 | unsigned long trans_start; |
1da177e4 | 2939 | |
b2220cad JV |
2940 | bond_for_each_slave(bond, slave, i) { |
2941 | switch (slave->new_link) { | |
2942 | case BOND_LINK_NOCHANGE: | |
2943 | continue; | |
ff59c456 | 2944 | |
b2220cad | 2945 | case BOND_LINK_UP: |
cb32f2a0 | 2946 | trans_start = dev_trans_start(slave->dev); |
b9f60253 | 2947 | if ((!bond->curr_active_slave && |
cb32f2a0 JB |
2948 | time_in_range(jiffies, |
2949 | trans_start - delta_in_ticks, | |
da210f55 | 2950 | trans_start + delta_in_ticks + delta_in_ticks / 2)) || |
b9f60253 | 2951 | bond->curr_active_slave != slave) { |
b2220cad | 2952 | slave->link = BOND_LINK_UP; |
5a430974 VF |
2953 | if (bond->current_arp_slave) { |
2954 | bond_set_slave_inactive_flags( | |
2955 | bond->current_arp_slave); | |
2956 | bond->current_arp_slave = NULL; | |
2957 | } | |
b2220cad | 2958 | |
a4aee5c8 | 2959 | pr_info("%s: link status definitely up for interface %s.\n", |
b9f60253 | 2960 | bond->dev->name, slave->dev->name); |
b2220cad | 2961 | |
b9f60253 JP |
2962 | if (!bond->curr_active_slave || |
2963 | (slave == bond->primary_slave)) | |
2964 | goto do_failover; | |
1da177e4 | 2965 | |
b9f60253 | 2966 | } |
1da177e4 | 2967 | |
b9f60253 | 2968 | continue; |
1da177e4 | 2969 | |
b2220cad JV |
2970 | case BOND_LINK_DOWN: |
2971 | if (slave->link_failure_count < UINT_MAX) | |
2972 | slave->link_failure_count++; | |
2973 | ||
2974 | slave->link = BOND_LINK_DOWN; | |
b9f60253 | 2975 | bond_set_slave_inactive_flags(slave); |
b2220cad | 2976 | |
a4aee5c8 | 2977 | pr_info("%s: link status definitely down for interface %s, disabling it\n", |
b9f60253 | 2978 | bond->dev->name, slave->dev->name); |
b2220cad | 2979 | |
b9f60253 | 2980 | if (slave == bond->curr_active_slave) { |
b2220cad | 2981 | bond->current_arp_slave = NULL; |
b9f60253 | 2982 | goto do_failover; |
1da177e4 | 2983 | } |
b9f60253 JP |
2984 | |
2985 | continue; | |
b2220cad JV |
2986 | |
2987 | default: | |
a4aee5c8 | 2988 | pr_err("%s: impossible: new_link %d on slave %s\n", |
b2220cad JV |
2989 | bond->dev->name, slave->new_link, |
2990 | slave->dev->name); | |
b9f60253 | 2991 | continue; |
1da177e4 | 2992 | } |
1da177e4 | 2993 | |
b9f60253 JP |
2994 | do_failover: |
2995 | ASSERT_RTNL(); | |
e843fa50 | 2996 | block_netpoll_tx(); |
b2220cad | 2997 | write_lock_bh(&bond->curr_slave_lock); |
b9f60253 | 2998 | bond_select_active_slave(bond); |
b2220cad | 2999 | write_unlock_bh(&bond->curr_slave_lock); |
e843fa50 | 3000 | unblock_netpoll_tx(); |
b2220cad | 3001 | } |
1da177e4 | 3002 | |
b2220cad JV |
3003 | bond_set_carrier(bond); |
3004 | } | |
1da177e4 | 3005 | |
b2220cad JV |
3006 | /* |
3007 | * Send ARP probes for active-backup mode ARP monitor. | |
3008 | * | |
3009 | * Called with bond->lock held for read. | |
3010 | */ | |
3011 | static void bond_ab_arp_probe(struct bonding *bond) | |
3012 | { | |
3013 | struct slave *slave; | |
3014 | int i; | |
1da177e4 | 3015 | |
b2220cad | 3016 | read_lock(&bond->curr_slave_lock); |
1da177e4 | 3017 | |
b2220cad | 3018 | if (bond->current_arp_slave && bond->curr_active_slave) |
a4aee5c8 JP |
3019 | pr_info("PROBE: c_arp %s && cas %s BAD\n", |
3020 | bond->current_arp_slave->dev->name, | |
3021 | bond->curr_active_slave->dev->name); | |
1da177e4 | 3022 | |
b2220cad JV |
3023 | if (bond->curr_active_slave) { |
3024 | bond_arp_send_all(bond, bond->curr_active_slave); | |
3025 | read_unlock(&bond->curr_slave_lock); | |
3026 | return; | |
3027 | } | |
1da177e4 | 3028 | |
b2220cad | 3029 | read_unlock(&bond->curr_slave_lock); |
059fe7a5 | 3030 | |
b2220cad JV |
3031 | /* if we don't have a curr_active_slave, search for the next available |
3032 | * backup slave from the current_arp_slave and make it the candidate | |
3033 | * for becoming the curr_active_slave | |
3034 | */ | |
1da177e4 | 3035 | |
b2220cad JV |
3036 | if (!bond->current_arp_slave) { |
3037 | bond->current_arp_slave = bond->first_slave; | |
3038 | if (!bond->current_arp_slave) | |
3039 | return; | |
3040 | } | |
1da177e4 | 3041 | |
b2220cad | 3042 | bond_set_slave_inactive_flags(bond->current_arp_slave); |
059fe7a5 | 3043 | |
b2220cad JV |
3044 | /* search for next candidate */ |
3045 | bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave->next) { | |
3046 | if (IS_UP(slave->dev)) { | |
3047 | slave->link = BOND_LINK_BACK; | |
3048 | bond_set_slave_active_flags(slave); | |
3049 | bond_arp_send_all(bond, slave); | |
1da177e4 | 3050 | slave->jiffies = jiffies; |
b2220cad JV |
3051 | bond->current_arp_slave = slave; |
3052 | break; | |
1da177e4 LT |
3053 | } |
3054 | ||
b2220cad JV |
3055 | /* if the link state is up at this point, we |
3056 | * mark it down - this can happen if we have | |
3057 | * simultaneous link failures and | |
3058 | * reselect_active_interface doesn't make this | |
3059 | * one the current slave so it is still marked | |
3060 | * up when it is actually down | |
1da177e4 | 3061 | */ |
b2220cad JV |
3062 | if (slave->link == BOND_LINK_UP) { |
3063 | slave->link = BOND_LINK_DOWN; | |
3064 | if (slave->link_failure_count < UINT_MAX) | |
3065 | slave->link_failure_count++; | |
1da177e4 | 3066 | |
b2220cad JV |
3067 | bond_set_slave_inactive_flags(slave); |
3068 | ||
a4aee5c8 JP |
3069 | pr_info("%s: backup interface %s is now down.\n", |
3070 | bond->dev->name, slave->dev->name); | |
1da177e4 | 3071 | } |
b2220cad JV |
3072 | } |
3073 | } | |
1da177e4 | 3074 | |
b2220cad JV |
3075 | void bond_activebackup_arp_mon(struct work_struct *work) |
3076 | { | |
3077 | struct bonding *bond = container_of(work, struct bonding, | |
3078 | arp_work.work); | |
ad246c99 | 3079 | bool should_notify_peers = false; |
b2220cad | 3080 | int delta_in_ticks; |
1da177e4 | 3081 | |
b2220cad | 3082 | read_lock(&bond->lock); |
1da177e4 | 3083 | |
b2220cad | 3084 | delta_in_ticks = msecs_to_jiffies(bond->params.arp_interval); |
1da177e4 | 3085 | |
b2220cad JV |
3086 | if (bond->slave_cnt == 0) |
3087 | goto re_arm; | |
3088 | ||
ad246c99 BH |
3089 | should_notify_peers = bond_should_notify_peers(bond); |
3090 | ||
b2220cad JV |
3091 | if (bond_ab_arp_inspect(bond, delta_in_ticks)) { |
3092 | read_unlock(&bond->lock); | |
e6d265e8 JV |
3093 | |
3094 | /* Race avoidance with bond_close flush of workqueue */ | |
3095 | if (!rtnl_trylock()) { | |
3096 | read_lock(&bond->lock); | |
3097 | delta_in_ticks = 1; | |
3098 | should_notify_peers = false; | |
3099 | goto re_arm; | |
3100 | } | |
3101 | ||
b2220cad JV |
3102 | read_lock(&bond->lock); |
3103 | ||
3104 | bond_ab_arp_commit(bond, delta_in_ticks); | |
3105 | ||
3106 | read_unlock(&bond->lock); | |
3107 | rtnl_unlock(); | |
3108 | read_lock(&bond->lock); | |
1da177e4 LT |
3109 | } |
3110 | ||
b2220cad JV |
3111 | bond_ab_arp_probe(bond); |
3112 | ||
1da177e4 | 3113 | re_arm: |
e6d265e8 | 3114 | if (bond->params.arp_interval) |
1b76b316 | 3115 | queue_delayed_work(bond->wq, &bond->arp_work, delta_in_ticks); |
e6d265e8 | 3116 | |
1da177e4 | 3117 | read_unlock(&bond->lock); |
ad246c99 BH |
3118 | |
3119 | if (should_notify_peers) { | |
e6d265e8 JV |
3120 | if (!rtnl_trylock()) { |
3121 | read_lock(&bond->lock); | |
3122 | bond->send_peer_notif++; | |
3123 | read_unlock(&bond->lock); | |
3124 | return; | |
3125 | } | |
b7bc2a5b | 3126 | call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, bond->dev); |
ad246c99 BH |
3127 | rtnl_unlock(); |
3128 | } | |
1da177e4 LT |
3129 | } |
3130 | ||
1da177e4 LT |
3131 | /*-------------------------- netdev event handling --------------------------*/ |
3132 | ||
3133 | /* | |
3134 | * Change device name | |
3135 | */ | |
3136 | static int bond_event_changename(struct bonding *bond) | |
3137 | { | |
1da177e4 LT |
3138 | bond_remove_proc_entry(bond); |
3139 | bond_create_proc_entry(bond); | |
7e083840 | 3140 | |
f073c7ca TI |
3141 | bond_debug_reregister(bond); |
3142 | ||
1da177e4 LT |
3143 | return NOTIFY_DONE; |
3144 | } | |
3145 | ||
3d632c3f SH |
3146 | static int bond_master_netdev_event(unsigned long event, |
3147 | struct net_device *bond_dev) | |
1da177e4 | 3148 | { |
454d7c9b | 3149 | struct bonding *event_bond = netdev_priv(bond_dev); |
1da177e4 LT |
3150 | |
3151 | switch (event) { | |
3152 | case NETDEV_CHANGENAME: | |
3153 | return bond_event_changename(event_bond); | |
a64d49c3 EB |
3154 | case NETDEV_UNREGISTER: |
3155 | bond_remove_proc_entry(event_bond); | |
3156 | break; | |
3157 | case NETDEV_REGISTER: | |
3158 | bond_create_proc_entry(event_bond); | |
3159 | break; | |
1da177e4 LT |
3160 | default: |
3161 | break; | |
3162 | } | |
3163 | ||
3164 | return NOTIFY_DONE; | |
3165 | } | |
3166 | ||
3d632c3f SH |
3167 | static int bond_slave_netdev_event(unsigned long event, |
3168 | struct net_device *slave_dev) | |
1da177e4 | 3169 | { |
471cb5a3 JP |
3170 | struct slave *slave = bond_slave_get_rtnl(slave_dev); |
3171 | struct bonding *bond = slave->bond; | |
3172 | struct net_device *bond_dev = slave->bond->dev; | |
3173 | u32 old_speed; | |
3174 | u8 old_duplex; | |
1da177e4 LT |
3175 | |
3176 | switch (event) { | |
3177 | case NETDEV_UNREGISTER: | |
471cb5a3 JP |
3178 | if (bond->setup_by_slave) |
3179 | bond_release_and_destroy(bond_dev, slave_dev); | |
3180 | else | |
3181 | bond_release(bond_dev, slave_dev); | |
1da177e4 | 3182 | break; |
98f41f69 | 3183 | case NETDEV_UP: |
1da177e4 | 3184 | case NETDEV_CHANGE: |
471cb5a3 JP |
3185 | old_speed = slave->speed; |
3186 | old_duplex = slave->duplex; | |
17d04500 | 3187 | |
471cb5a3 | 3188 | bond_update_speed_duplex(slave); |
17d04500 | 3189 | |
471cb5a3 JP |
3190 | if (bond->params.mode == BOND_MODE_8023AD) { |
3191 | if (old_speed != slave->speed) | |
3192 | bond_3ad_adapter_speed_changed(slave); | |
3193 | if (old_duplex != slave->duplex) | |
3194 | bond_3ad_adapter_duplex_changed(slave); | |
17d04500 | 3195 | } |
1da177e4 LT |
3196 | break; |
3197 | case NETDEV_DOWN: | |
3198 | /* | |
3199 | * ... Or is it this? | |
3200 | */ | |
3201 | break; | |
3202 | case NETDEV_CHANGEMTU: | |
3203 | /* | |
3204 | * TODO: Should slaves be allowed to | |
3205 | * independently alter their MTU? For | |
3206 | * an active-backup bond, slaves need | |
3207 | * not be the same type of device, so | |
3208 | * MTUs may vary. For other modes, | |
3209 | * slaves arguably should have the | |
3210 | * same MTUs. To do this, we'd need to | |
3211 | * take over the slave's change_mtu | |
3212 | * function for the duration of their | |
3213 | * servitude. | |
3214 | */ | |
3215 | break; | |
3216 | case NETDEV_CHANGENAME: | |
3217 | /* | |
3218 | * TODO: handle changing the primary's name | |
3219 | */ | |
3220 | break; | |
8531c5ff AK |
3221 | case NETDEV_FEAT_CHANGE: |
3222 | bond_compute_features(bond); | |
3223 | break; | |
1da177e4 LT |
3224 | default: |
3225 | break; | |
3226 | } | |
3227 | ||
3228 | return NOTIFY_DONE; | |
3229 | } | |
3230 | ||
3231 | /* | |
3232 | * bond_netdev_event: handle netdev notifier chain events. | |
3233 | * | |
3234 | * This function receives events for the netdev chain. The caller (an | |
e041c683 | 3235 | * ioctl handler calling blocking_notifier_call_chain) holds the necessary |
1da177e4 LT |
3236 | * locks for us to safely manipulate the slave devices (RTNL lock, |
3237 | * dev_probe_lock). | |
3238 | */ | |
3d632c3f SH |
3239 | static int bond_netdev_event(struct notifier_block *this, |
3240 | unsigned long event, void *ptr) | |
1da177e4 LT |
3241 | { |
3242 | struct net_device *event_dev = (struct net_device *)ptr; | |
3243 | ||
5a03cdb7 | 3244 | pr_debug("event_dev: %s, event: %lx\n", |
a4aee5c8 JP |
3245 | event_dev ? event_dev->name : "None", |
3246 | event); | |
1da177e4 | 3247 | |
0b680e75 JV |
3248 | if (!(event_dev->priv_flags & IFF_BONDING)) |
3249 | return NOTIFY_DONE; | |
3250 | ||
1da177e4 | 3251 | if (event_dev->flags & IFF_MASTER) { |
5a03cdb7 | 3252 | pr_debug("IFF_MASTER\n"); |
1da177e4 LT |
3253 | return bond_master_netdev_event(event, event_dev); |
3254 | } | |
3255 | ||
3256 | if (event_dev->flags & IFF_SLAVE) { | |
5a03cdb7 | 3257 | pr_debug("IFF_SLAVE\n"); |
1da177e4 LT |
3258 | return bond_slave_netdev_event(event, event_dev); |
3259 | } | |
3260 | ||
3261 | return NOTIFY_DONE; | |
3262 | } | |
3263 | ||
3264 | static struct notifier_block bond_netdev_notifier = { | |
3265 | .notifier_call = bond_netdev_event, | |
3266 | }; | |
3267 | ||
169a3e66 JV |
3268 | /*---------------------------- Hashing Policies -----------------------------*/ |
3269 | ||
6b923cb7 JE |
3270 | /* |
3271 | * Hash for the output device based upon layer 2 data | |
3272 | */ | |
3273 | static int bond_xmit_hash_policy_l2(struct sk_buff *skb, int count) | |
3274 | { | |
3275 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
3276 | ||
3277 | if (skb_headlen(skb) >= offsetof(struct ethhdr, h_proto)) | |
3278 | return (data->h_dest[5] ^ data->h_source[5]) % count; | |
3279 | ||
3280 | return 0; | |
3281 | } | |
3282 | ||
6f6652be JV |
3283 | /* |
3284 | * Hash for the output device based upon layer 2 and layer 3 data. If | |
6b923cb7 | 3285 | * the packet is not IP, fall back on bond_xmit_hash_policy_l2() |
6f6652be | 3286 | */ |
a361c83c | 3287 | static int bond_xmit_hash_policy_l23(struct sk_buff *skb, int count) |
6f6652be JV |
3288 | { |
3289 | struct ethhdr *data = (struct ethhdr *)skb->data; | |
6b923cb7 JE |
3290 | struct iphdr *iph; |
3291 | struct ipv6hdr *ipv6h; | |
3292 | u32 v6hash; | |
3293 | __be32 *s, *d; | |
3294 | ||
3295 | if (skb->protocol == htons(ETH_P_IP) && | |
3296 | skb_network_header_len(skb) >= sizeof(*iph)) { | |
3297 | iph = ip_hdr(skb); | |
6f6652be | 3298 | return ((ntohl(iph->saddr ^ iph->daddr) & 0xffff) ^ |
d3da6831 | 3299 | (data->h_dest[5] ^ data->h_source[5])) % count; |
6b923cb7 JE |
3300 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
3301 | skb_network_header_len(skb) >= sizeof(*ipv6h)) { | |
3302 | ipv6h = ipv6_hdr(skb); | |
3303 | s = &ipv6h->saddr.s6_addr32[0]; | |
3304 | d = &ipv6h->daddr.s6_addr32[0]; | |
3305 | v6hash = (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3306 | v6hash ^= (v6hash >> 24) ^ (v6hash >> 16) ^ (v6hash >> 8); | |
3307 | return (v6hash ^ data->h_dest[5] ^ data->h_source[5]) % count; | |
6f6652be JV |
3308 | } |
3309 | ||
6b923cb7 | 3310 | return bond_xmit_hash_policy_l2(skb, count); |
6f6652be JV |
3311 | } |
3312 | ||
169a3e66 | 3313 | /* |
59c51591 | 3314 | * Hash for the output device based upon layer 3 and layer 4 data. If |
169a3e66 | 3315 | * the packet is a frag or not TCP or UDP, just use layer 3 data. If it is |
6b923cb7 | 3316 | * altogether not IP, fall back on bond_xmit_hash_policy_l2() |
169a3e66 | 3317 | */ |
a361c83c | 3318 | static int bond_xmit_hash_policy_l34(struct sk_buff *skb, int count) |
169a3e66 | 3319 | { |
6b923cb7 JE |
3320 | u32 layer4_xor = 0; |
3321 | struct iphdr *iph; | |
3322 | struct ipv6hdr *ipv6h; | |
3323 | __be32 *s, *d; | |
3324 | __be16 *layer4hdr; | |
169a3e66 | 3325 | |
6b923cb7 JE |
3326 | if (skb->protocol == htons(ETH_P_IP) && |
3327 | skb_network_header_len(skb) >= sizeof(*iph)) { | |
3328 | iph = ip_hdr(skb); | |
56f8a75c | 3329 | if (!ip_is_fragment(iph) && |
169a3e66 | 3330 | (iph->protocol == IPPROTO_TCP || |
6b923cb7 JE |
3331 | iph->protocol == IPPROTO_UDP) && |
3332 | (skb_headlen(skb) - skb_network_offset(skb) >= | |
3333 | iph->ihl * sizeof(u32) + sizeof(*layer4hdr) * 2)) { | |
3334 | layer4hdr = (__be16 *)((u32 *)iph + iph->ihl); | |
3335 | layer4_xor = ntohs(*layer4hdr ^ *(layer4hdr + 1)); | |
169a3e66 JV |
3336 | } |
3337 | return (layer4_xor ^ | |
3338 | ((ntohl(iph->saddr ^ iph->daddr)) & 0xffff)) % count; | |
6b923cb7 JE |
3339 | } else if (skb->protocol == htons(ETH_P_IPV6) && |
3340 | skb_network_header_len(skb) >= sizeof(*ipv6h)) { | |
3341 | ipv6h = ipv6_hdr(skb); | |
3342 | if ((ipv6h->nexthdr == IPPROTO_TCP || | |
3343 | ipv6h->nexthdr == IPPROTO_UDP) && | |
3344 | (skb_headlen(skb) - skb_network_offset(skb) >= | |
3345 | sizeof(*ipv6h) + sizeof(*layer4hdr) * 2)) { | |
3346 | layer4hdr = (__be16 *)(ipv6h + 1); | |
3347 | layer4_xor = ntohs(*layer4hdr ^ *(layer4hdr + 1)); | |
3348 | } | |
3349 | s = &ipv6h->saddr.s6_addr32[0]; | |
3350 | d = &ipv6h->daddr.s6_addr32[0]; | |
3351 | layer4_xor ^= (s[1] ^ d[1]) ^ (s[2] ^ d[2]) ^ (s[3] ^ d[3]); | |
3352 | layer4_xor ^= (layer4_xor >> 24) ^ (layer4_xor >> 16) ^ | |
3353 | (layer4_xor >> 8); | |
3354 | return layer4_xor % count; | |
169a3e66 JV |
3355 | } |
3356 | ||
6b923cb7 | 3357 | return bond_xmit_hash_policy_l2(skb, count); |
169a3e66 JV |
3358 | } |
3359 | ||
1da177e4 LT |
3360 | /*-------------------------- Device entry points ----------------------------*/ |
3361 | ||
fbb0c41b | 3362 | static void bond_work_init_all(struct bonding *bond) |
3363 | { | |
3364 | INIT_DELAYED_WORK(&bond->mcast_work, | |
3365 | bond_resend_igmp_join_requests_delayed); | |
3366 | INIT_DELAYED_WORK(&bond->alb_work, bond_alb_monitor); | |
3367 | INIT_DELAYED_WORK(&bond->mii_work, bond_mii_monitor); | |
3368 | if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3369 | INIT_DELAYED_WORK(&bond->arp_work, bond_activebackup_arp_mon); | |
3370 | else | |
3371 | INIT_DELAYED_WORK(&bond->arp_work, bond_loadbalance_arp_mon); | |
3372 | INIT_DELAYED_WORK(&bond->ad_work, bond_3ad_state_machine_handler); | |
3373 | } | |
3374 | ||
3375 | static void bond_work_cancel_all(struct bonding *bond) | |
3376 | { | |
3377 | cancel_delayed_work_sync(&bond->mii_work); | |
3378 | cancel_delayed_work_sync(&bond->arp_work); | |
3379 | cancel_delayed_work_sync(&bond->alb_work); | |
3380 | cancel_delayed_work_sync(&bond->ad_work); | |
3381 | cancel_delayed_work_sync(&bond->mcast_work); | |
3382 | } | |
3383 | ||
1da177e4 LT |
3384 | static int bond_open(struct net_device *bond_dev) |
3385 | { | |
454d7c9b | 3386 | struct bonding *bond = netdev_priv(bond_dev); |
ba3211cc PP |
3387 | struct slave *slave; |
3388 | int i; | |
1da177e4 | 3389 | |
ba3211cc PP |
3390 | /* reset slave->backup and slave->inactive */ |
3391 | read_lock(&bond->lock); | |
3392 | if (bond->slave_cnt > 0) { | |
3393 | read_lock(&bond->curr_slave_lock); | |
3394 | bond_for_each_slave(bond, slave, i) { | |
3395 | if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) | |
3396 | && (slave != bond->curr_active_slave)) { | |
3397 | bond_set_slave_inactive_flags(slave); | |
3398 | } else { | |
3399 | bond_set_slave_active_flags(slave); | |
3400 | } | |
3401 | } | |
3402 | read_unlock(&bond->curr_slave_lock); | |
3403 | } | |
3404 | read_unlock(&bond->lock); | |
3405 | ||
fbb0c41b | 3406 | bond_work_init_all(bond); |
5a37e8ca | 3407 | |
58402054 | 3408 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3409 | /* bond_alb_initialize must be called before the timer |
3410 | * is started. | |
3411 | */ | |
fbb0c41b | 3412 | if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) |
b473946a | 3413 | return -ENOMEM; |
1b76b316 | 3414 | queue_delayed_work(bond->wq, &bond->alb_work, 0); |
1da177e4 LT |
3415 | } |
3416 | ||
fbb0c41b | 3417 | if (bond->params.miimon) /* link check interval, in milliseconds. */ |
1b76b316 | 3418 | queue_delayed_work(bond->wq, &bond->mii_work, 0); |
1da177e4 LT |
3419 | |
3420 | if (bond->params.arp_interval) { /* arp interval, in milliseconds. */ | |
1b76b316 | 3421 | queue_delayed_work(bond->wq, &bond->arp_work, 0); |
f5b2b966 | 3422 | if (bond->params.arp_validate) |
3aba891d | 3423 | bond->recv_probe = bond_arp_rcv; |
1da177e4 LT |
3424 | } |
3425 | ||
3426 | if (bond->params.mode == BOND_MODE_8023AD) { | |
1b76b316 | 3427 | queue_delayed_work(bond->wq, &bond->ad_work, 0); |
1da177e4 | 3428 | /* register to receive LACPDUs */ |
3aba891d | 3429 | bond->recv_probe = bond_3ad_lacpdu_recv; |
fd989c83 | 3430 | bond_3ad_initiate_agg_selection(bond, 1); |
1da177e4 LT |
3431 | } |
3432 | ||
3433 | return 0; | |
3434 | } | |
3435 | ||
3436 | static int bond_close(struct net_device *bond_dev) | |
3437 | { | |
454d7c9b | 3438 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3439 | |
1da177e4 | 3440 | write_lock_bh(&bond->lock); |
ad246c99 | 3441 | bond->send_peer_notif = 0; |
1da177e4 LT |
3442 | write_unlock_bh(&bond->lock); |
3443 | ||
fbb0c41b | 3444 | bond_work_cancel_all(bond); |
58402054 | 3445 | if (bond_is_lb(bond)) { |
1da177e4 LT |
3446 | /* Must be called only after all |
3447 | * slaves have been released | |
3448 | */ | |
3449 | bond_alb_deinitialize(bond); | |
3450 | } | |
3aba891d | 3451 | bond->recv_probe = NULL; |
1da177e4 LT |
3452 | |
3453 | return 0; | |
3454 | } | |
3455 | ||
28172739 ED |
3456 | static struct rtnl_link_stats64 *bond_get_stats(struct net_device *bond_dev, |
3457 | struct rtnl_link_stats64 *stats) | |
1da177e4 | 3458 | { |
454d7c9b | 3459 | struct bonding *bond = netdev_priv(bond_dev); |
28172739 | 3460 | struct rtnl_link_stats64 temp; |
1da177e4 LT |
3461 | struct slave *slave; |
3462 | int i; | |
3463 | ||
28172739 | 3464 | memset(stats, 0, sizeof(*stats)); |
1da177e4 LT |
3465 | |
3466 | read_lock_bh(&bond->lock); | |
3467 | ||
3468 | bond_for_each_slave(bond, slave, i) { | |
be1f3c2c | 3469 | const struct rtnl_link_stats64 *sstats = |
28172739 ED |
3470 | dev_get_stats(slave->dev, &temp); |
3471 | ||
3472 | stats->rx_packets += sstats->rx_packets; | |
3473 | stats->rx_bytes += sstats->rx_bytes; | |
3474 | stats->rx_errors += sstats->rx_errors; | |
3475 | stats->rx_dropped += sstats->rx_dropped; | |
3476 | ||
3477 | stats->tx_packets += sstats->tx_packets; | |
3478 | stats->tx_bytes += sstats->tx_bytes; | |
3479 | stats->tx_errors += sstats->tx_errors; | |
3480 | stats->tx_dropped += sstats->tx_dropped; | |
3481 | ||
3482 | stats->multicast += sstats->multicast; | |
3483 | stats->collisions += sstats->collisions; | |
3484 | ||
3485 | stats->rx_length_errors += sstats->rx_length_errors; | |
3486 | stats->rx_over_errors += sstats->rx_over_errors; | |
3487 | stats->rx_crc_errors += sstats->rx_crc_errors; | |
3488 | stats->rx_frame_errors += sstats->rx_frame_errors; | |
3489 | stats->rx_fifo_errors += sstats->rx_fifo_errors; | |
3490 | stats->rx_missed_errors += sstats->rx_missed_errors; | |
3491 | ||
3492 | stats->tx_aborted_errors += sstats->tx_aborted_errors; | |
3493 | stats->tx_carrier_errors += sstats->tx_carrier_errors; | |
3494 | stats->tx_fifo_errors += sstats->tx_fifo_errors; | |
3495 | stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors; | |
3496 | stats->tx_window_errors += sstats->tx_window_errors; | |
2439f9eb AG |
3497 | } |
3498 | ||
1da177e4 LT |
3499 | read_unlock_bh(&bond->lock); |
3500 | ||
3501 | return stats; | |
3502 | } | |
3503 | ||
3504 | static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd) | |
3505 | { | |
3506 | struct net_device *slave_dev = NULL; | |
3507 | struct ifbond k_binfo; | |
3508 | struct ifbond __user *u_binfo = NULL; | |
3509 | struct ifslave k_sinfo; | |
3510 | struct ifslave __user *u_sinfo = NULL; | |
3511 | struct mii_ioctl_data *mii = NULL; | |
387ff911 | 3512 | struct net *net; |
1da177e4 LT |
3513 | int res = 0; |
3514 | ||
a4aee5c8 | 3515 | pr_debug("bond_ioctl: master=%s, cmd=%d\n", bond_dev->name, cmd); |
1da177e4 LT |
3516 | |
3517 | switch (cmd) { | |
1da177e4 LT |
3518 | case SIOCGMIIPHY: |
3519 | mii = if_mii(ifr); | |
3d632c3f | 3520 | if (!mii) |
1da177e4 | 3521 | return -EINVAL; |
3d632c3f | 3522 | |
1da177e4 LT |
3523 | mii->phy_id = 0; |
3524 | /* Fall Through */ | |
3525 | case SIOCGMIIREG: | |
3526 | /* | |
3527 | * We do this again just in case we were called by SIOCGMIIREG | |
3528 | * instead of SIOCGMIIPHY. | |
3529 | */ | |
3530 | mii = if_mii(ifr); | |
3d632c3f | 3531 | if (!mii) |
1da177e4 | 3532 | return -EINVAL; |
3d632c3f | 3533 | |
1da177e4 LT |
3534 | |
3535 | if (mii->reg_num == 1) { | |
454d7c9b | 3536 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3537 | mii->val_out = 0; |
6603a6f2 | 3538 | read_lock(&bond->lock); |
1da177e4 | 3539 | read_lock(&bond->curr_slave_lock); |
3d632c3f | 3540 | if (netif_carrier_ok(bond->dev)) |
1da177e4 | 3541 | mii->val_out = BMSR_LSTATUS; |
3d632c3f | 3542 | |
1da177e4 | 3543 | read_unlock(&bond->curr_slave_lock); |
6603a6f2 | 3544 | read_unlock(&bond->lock); |
1da177e4 LT |
3545 | } |
3546 | ||
3547 | return 0; | |
3548 | case BOND_INFO_QUERY_OLD: | |
3549 | case SIOCBONDINFOQUERY: | |
3550 | u_binfo = (struct ifbond __user *)ifr->ifr_data; | |
3551 | ||
3d632c3f | 3552 | if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) |
1da177e4 | 3553 | return -EFAULT; |
1da177e4 LT |
3554 | |
3555 | res = bond_info_query(bond_dev, &k_binfo); | |
3d632c3f SH |
3556 | if (res == 0 && |
3557 | copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) | |
3558 | return -EFAULT; | |
1da177e4 LT |
3559 | |
3560 | return res; | |
3561 | case BOND_SLAVE_INFO_QUERY_OLD: | |
3562 | case SIOCBONDSLAVEINFOQUERY: | |
3563 | u_sinfo = (struct ifslave __user *)ifr->ifr_data; | |
3564 | ||
3d632c3f | 3565 | if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) |
1da177e4 | 3566 | return -EFAULT; |
1da177e4 LT |
3567 | |
3568 | res = bond_slave_info_query(bond_dev, &k_sinfo); | |
3d632c3f SH |
3569 | if (res == 0 && |
3570 | copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) | |
3571 | return -EFAULT; | |
1da177e4 LT |
3572 | |
3573 | return res; | |
3574 | default: | |
3575 | /* Go on */ | |
3576 | break; | |
3577 | } | |
3578 | ||
387ff911 G |
3579 | net = dev_net(bond_dev); |
3580 | ||
3581 | if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) | |
1da177e4 | 3582 | return -EPERM; |
1da177e4 | 3583 | |
387ff911 | 3584 | slave_dev = dev_get_by_name(net, ifr->ifr_slave); |
1da177e4 | 3585 | |
a4aee5c8 | 3586 | pr_debug("slave_dev=%p:\n", slave_dev); |
1da177e4 | 3587 | |
3d632c3f | 3588 | if (!slave_dev) |
1da177e4 | 3589 | res = -ENODEV; |
3d632c3f | 3590 | else { |
a4aee5c8 | 3591 | pr_debug("slave_dev->name=%s:\n", slave_dev->name); |
1da177e4 LT |
3592 | switch (cmd) { |
3593 | case BOND_ENSLAVE_OLD: | |
3594 | case SIOCBONDENSLAVE: | |
3595 | res = bond_enslave(bond_dev, slave_dev); | |
3596 | break; | |
3597 | case BOND_RELEASE_OLD: | |
3598 | case SIOCBONDRELEASE: | |
3599 | res = bond_release(bond_dev, slave_dev); | |
3600 | break; | |
3601 | case BOND_SETHWADDR_OLD: | |
3602 | case SIOCBONDSETHWADDR: | |
409cc1f8 JP |
3603 | bond_set_dev_addr(bond_dev, slave_dev); |
3604 | res = 0; | |
1da177e4 LT |
3605 | break; |
3606 | case BOND_CHANGE_ACTIVE_OLD: | |
3607 | case SIOCBONDCHANGEACTIVE: | |
3608 | res = bond_ioctl_change_active(bond_dev, slave_dev); | |
3609 | break; | |
3610 | default: | |
3611 | res = -EOPNOTSUPP; | |
3612 | } | |
3613 | ||
3614 | dev_put(slave_dev); | |
3615 | } | |
3616 | ||
1da177e4 LT |
3617 | return res; |
3618 | } | |
3619 | ||
22bedad3 JP |
3620 | static bool bond_addr_in_mc_list(unsigned char *addr, |
3621 | struct netdev_hw_addr_list *list, | |
3622 | int addrlen) | |
3623 | { | |
3624 | struct netdev_hw_addr *ha; | |
3625 | ||
3626 | netdev_hw_addr_list_for_each(ha, list) | |
3627 | if (!memcmp(ha->addr, addr, addrlen)) | |
3628 | return true; | |
3629 | ||
3630 | return false; | |
3631 | } | |
3632 | ||
d03462b9 | 3633 | static void bond_change_rx_flags(struct net_device *bond_dev, int change) |
1da177e4 | 3634 | { |
454d7c9b | 3635 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3636 | |
d03462b9 JP |
3637 | if (change & IFF_PROMISC) |
3638 | bond_set_promiscuity(bond, | |
3639 | bond_dev->flags & IFF_PROMISC ? 1 : -1); | |
3d632c3f | 3640 | |
d03462b9 JP |
3641 | if (change & IFF_ALLMULTI) |
3642 | bond_set_allmulti(bond, | |
3643 | bond_dev->flags & IFF_ALLMULTI ? 1 : -1); | |
3644 | } | |
1da177e4 | 3645 | |
d03462b9 JP |
3646 | static void bond_set_multicast_list(struct net_device *bond_dev) |
3647 | { | |
3648 | struct bonding *bond = netdev_priv(bond_dev); | |
3649 | struct netdev_hw_addr *ha; | |
3650 | bool found; | |
1da177e4 | 3651 | |
80ee5ad2 JV |
3652 | read_lock(&bond->lock); |
3653 | ||
1da177e4 | 3654 | /* looking for addresses to add to slaves' mc list */ |
22bedad3 JP |
3655 | netdev_for_each_mc_addr(ha, bond_dev) { |
3656 | found = bond_addr_in_mc_list(ha->addr, &bond->mc_list, | |
3657 | bond_dev->addr_len); | |
3658 | if (!found) | |
3659 | bond_mc_add(bond, ha->addr); | |
1da177e4 LT |
3660 | } |
3661 | ||
3662 | /* looking for addresses to delete from slaves' list */ | |
22bedad3 JP |
3663 | netdev_hw_addr_list_for_each(ha, &bond->mc_list) { |
3664 | found = bond_addr_in_mc_list(ha->addr, &bond_dev->mc, | |
3665 | bond_dev->addr_len); | |
3666 | if (!found) | |
3667 | bond_mc_del(bond, ha->addr); | |
1da177e4 LT |
3668 | } |
3669 | ||
3670 | /* save master's multicast list */ | |
22bedad3 JP |
3671 | __hw_addr_flush(&bond->mc_list); |
3672 | __hw_addr_add_multiple(&bond->mc_list, &bond_dev->mc, | |
3673 | bond_dev->addr_len, NETDEV_HW_ADDR_T_MULTICAST); | |
1da177e4 | 3674 | |
80ee5ad2 | 3675 | read_unlock(&bond->lock); |
1da177e4 LT |
3676 | } |
3677 | ||
234bcf8a | 3678 | static int bond_neigh_init(struct neighbour *n) |
00829823 | 3679 | { |
234bcf8a | 3680 | struct bonding *bond = netdev_priv(n->dev); |
00829823 | 3681 | struct slave *slave = bond->first_slave; |
234bcf8a SP |
3682 | const struct net_device_ops *slave_ops; |
3683 | struct neigh_parms parms; | |
3684 | int ret; | |
3685 | ||
3686 | if (!slave) | |
3687 | return 0; | |
3688 | ||
3689 | slave_ops = slave->dev->netdev_ops; | |
3690 | ||
3691 | if (!slave_ops->ndo_neigh_setup) | |
3692 | return 0; | |
3693 | ||
3694 | parms.neigh_setup = NULL; | |
3695 | parms.neigh_cleanup = NULL; | |
3696 | ret = slave_ops->ndo_neigh_setup(slave->dev, &parms); | |
3697 | if (ret) | |
3698 | return ret; | |
3699 | ||
3700 | /* | |
3701 | * Assign slave's neigh_cleanup to neighbour in case cleanup is called | |
3702 | * after the last slave has been detached. Assumes that all slaves | |
3703 | * utilize the same neigh_cleanup (true at this writing as only user | |
3704 | * is ipoib). | |
3705 | */ | |
3706 | n->parms->neigh_cleanup = parms.neigh_cleanup; | |
3707 | ||
3708 | if (!parms.neigh_setup) | |
3709 | return 0; | |
3710 | ||
3711 | return parms.neigh_setup(n); | |
3712 | } | |
3713 | ||
3714 | /* | |
3715 | * The bonding ndo_neigh_setup is called at init time beofre any | |
3716 | * slave exists. So we must declare proxy setup function which will | |
3717 | * be used at run time to resolve the actual slave neigh param setup. | |
3718 | */ | |
3719 | static int bond_neigh_setup(struct net_device *dev, | |
3720 | struct neigh_parms *parms) | |
3721 | { | |
3722 | parms->neigh_setup = bond_neigh_init; | |
00829823 | 3723 | |
00829823 SH |
3724 | return 0; |
3725 | } | |
3726 | ||
1da177e4 LT |
3727 | /* |
3728 | * Change the MTU of all of a master's slaves to match the master | |
3729 | */ | |
3730 | static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) | |
3731 | { | |
454d7c9b | 3732 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3733 | struct slave *slave, *stop_at; |
3734 | int res = 0; | |
3735 | int i; | |
3736 | ||
5a03cdb7 | 3737 | pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond, |
a4aee5c8 | 3738 | (bond_dev ? bond_dev->name : "None"), new_mtu); |
1da177e4 LT |
3739 | |
3740 | /* Can't hold bond->lock with bh disabled here since | |
3741 | * some base drivers panic. On the other hand we can't | |
3742 | * hold bond->lock without bh disabled because we'll | |
3743 | * deadlock. The only solution is to rely on the fact | |
3744 | * that we're under rtnl_lock here, and the slaves | |
3745 | * list won't change. This doesn't solve the problem | |
3746 | * of setting the slave's MTU while it is | |
3747 | * transmitting, but the assumption is that the base | |
3748 | * driver can handle that. | |
3749 | * | |
3750 | * TODO: figure out a way to safely iterate the slaves | |
3751 | * list, but without holding a lock around the actual | |
3752 | * call to the base driver. | |
3753 | */ | |
3754 | ||
3755 | bond_for_each_slave(bond, slave, i) { | |
a4aee5c8 JP |
3756 | pr_debug("s %p s->p %p c_m %p\n", |
3757 | slave, | |
3758 | slave->prev, | |
3759 | slave->dev->netdev_ops->ndo_change_mtu); | |
e944ef79 | 3760 | |
1da177e4 LT |
3761 | res = dev_set_mtu(slave->dev, new_mtu); |
3762 | ||
3763 | if (res) { | |
3764 | /* If we failed to set the slave's mtu to the new value | |
3765 | * we must abort the operation even in ACTIVE_BACKUP | |
3766 | * mode, because if we allow the backup slaves to have | |
3767 | * different mtu values than the active slave we'll | |
3768 | * need to change their mtu when doing a failover. That | |
3769 | * means changing their mtu from timer context, which | |
3770 | * is probably not a good idea. | |
3771 | */ | |
5a03cdb7 | 3772 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3773 | goto unwind; |
3774 | } | |
3775 | } | |
3776 | ||
3777 | bond_dev->mtu = new_mtu; | |
3778 | ||
3779 | return 0; | |
3780 | ||
3781 | unwind: | |
3782 | /* unwind from head to the slave that failed */ | |
3783 | stop_at = slave; | |
3784 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3785 | int tmp_res; | |
3786 | ||
3787 | tmp_res = dev_set_mtu(slave->dev, bond_dev->mtu); | |
3788 | if (tmp_res) { | |
a4aee5c8 JP |
3789 | pr_debug("unwind err %d dev %s\n", |
3790 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3791 | } |
3792 | } | |
3793 | ||
3794 | return res; | |
3795 | } | |
3796 | ||
3797 | /* | |
3798 | * Change HW address | |
3799 | * | |
3800 | * Note that many devices must be down to change the HW address, and | |
3801 | * downing the master releases all slaves. We can make bonds full of | |
3802 | * bonding devices to test this, however. | |
3803 | */ | |
3804 | static int bond_set_mac_address(struct net_device *bond_dev, void *addr) | |
3805 | { | |
454d7c9b | 3806 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3807 | struct sockaddr *sa = addr, tmp_sa; |
3808 | struct slave *slave, *stop_at; | |
3809 | int res = 0; | |
3810 | int i; | |
3811 | ||
eb7cc59a SH |
3812 | if (bond->params.mode == BOND_MODE_ALB) |
3813 | return bond_alb_set_mac_address(bond_dev, addr); | |
3814 | ||
3815 | ||
a4aee5c8 JP |
3816 | pr_debug("bond=%p, name=%s\n", |
3817 | bond, bond_dev ? bond_dev->name : "None"); | |
1da177e4 | 3818 | |
dd957c57 | 3819 | /* |
3915c1e8 JV |
3820 | * If fail_over_mac is set to active, do nothing and return |
3821 | * success. Returning an error causes ifenslave to fail. | |
dd957c57 | 3822 | */ |
3915c1e8 | 3823 | if (bond->params.fail_over_mac == BOND_FOM_ACTIVE) |
dd957c57 | 3824 | return 0; |
2ab82852 | 3825 | |
3d632c3f | 3826 | if (!is_valid_ether_addr(sa->sa_data)) |
1da177e4 | 3827 | return -EADDRNOTAVAIL; |
1da177e4 LT |
3828 | |
3829 | /* Can't hold bond->lock with bh disabled here since | |
3830 | * some base drivers panic. On the other hand we can't | |
3831 | * hold bond->lock without bh disabled because we'll | |
3832 | * deadlock. The only solution is to rely on the fact | |
3833 | * that we're under rtnl_lock here, and the slaves | |
3834 | * list won't change. This doesn't solve the problem | |
3835 | * of setting the slave's hw address while it is | |
3836 | * transmitting, but the assumption is that the base | |
3837 | * driver can handle that. | |
3838 | * | |
3839 | * TODO: figure out a way to safely iterate the slaves | |
3840 | * list, but without holding a lock around the actual | |
3841 | * call to the base driver. | |
3842 | */ | |
3843 | ||
3844 | bond_for_each_slave(bond, slave, i) { | |
eb7cc59a | 3845 | const struct net_device_ops *slave_ops = slave->dev->netdev_ops; |
5a03cdb7 | 3846 | pr_debug("slave %p %s\n", slave, slave->dev->name); |
1da177e4 | 3847 | |
eb7cc59a | 3848 | if (slave_ops->ndo_set_mac_address == NULL) { |
1da177e4 | 3849 | res = -EOPNOTSUPP; |
5a03cdb7 | 3850 | pr_debug("EOPNOTSUPP %s\n", slave->dev->name); |
1da177e4 LT |
3851 | goto unwind; |
3852 | } | |
3853 | ||
3854 | res = dev_set_mac_address(slave->dev, addr); | |
3855 | if (res) { | |
3856 | /* TODO: consider downing the slave | |
3857 | * and retry ? | |
3858 | * User should expect communications | |
3859 | * breakage anyway until ARP finish | |
3860 | * updating, so... | |
3861 | */ | |
5a03cdb7 | 3862 | pr_debug("err %d %s\n", res, slave->dev->name); |
1da177e4 LT |
3863 | goto unwind; |
3864 | } | |
3865 | } | |
3866 | ||
3867 | /* success */ | |
3868 | memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len); | |
3869 | return 0; | |
3870 | ||
3871 | unwind: | |
3872 | memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len); | |
3873 | tmp_sa.sa_family = bond_dev->type; | |
3874 | ||
3875 | /* unwind from head to the slave that failed */ | |
3876 | stop_at = slave; | |
3877 | bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) { | |
3878 | int tmp_res; | |
3879 | ||
3880 | tmp_res = dev_set_mac_address(slave->dev, &tmp_sa); | |
3881 | if (tmp_res) { | |
a4aee5c8 JP |
3882 | pr_debug("unwind err %d dev %s\n", |
3883 | tmp_res, slave->dev->name); | |
1da177e4 LT |
3884 | } |
3885 | } | |
3886 | ||
3887 | return res; | |
3888 | } | |
3889 | ||
3890 | static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev) | |
3891 | { | |
454d7c9b | 3892 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 3893 | struct slave *slave, *start_at; |
cf5f9044 | 3894 | int i, slave_no, res = 1; |
a2fd940f | 3895 | struct iphdr *iph = ip_hdr(skb); |
1da177e4 | 3896 | |
cf5f9044 | 3897 | /* |
a2fd940f AG |
3898 | * Start with the curr_active_slave that joined the bond as the |
3899 | * default for sending IGMP traffic. For failover purposes one | |
3900 | * needs to maintain some consistency for the interface that will | |
3901 | * send the join/membership reports. The curr_active_slave found | |
3902 | * will send all of this type of traffic. | |
cf5f9044 | 3903 | */ |
00ae7028 | 3904 | if ((iph->protocol == IPPROTO_IGMP) && |
a2fd940f | 3905 | (skb->protocol == htons(ETH_P_IP))) { |
1da177e4 | 3906 | |
a2fd940f AG |
3907 | read_lock(&bond->curr_slave_lock); |
3908 | slave = bond->curr_active_slave; | |
3909 | read_unlock(&bond->curr_slave_lock); | |
3910 | ||
3911 | if (!slave) | |
3912 | goto out; | |
3913 | } else { | |
3914 | /* | |
3915 | * Concurrent TX may collide on rr_tx_counter; we accept | |
3916 | * that as being rare enough not to justify using an | |
3917 | * atomic op here. | |
3918 | */ | |
3919 | slave_no = bond->rr_tx_counter++ % bond->slave_cnt; | |
3920 | ||
3921 | bond_for_each_slave(bond, slave, i) { | |
3922 | slave_no--; | |
3923 | if (slave_no < 0) | |
3924 | break; | |
3925 | } | |
1da177e4 LT |
3926 | } |
3927 | ||
cf5f9044 | 3928 | start_at = slave; |
1da177e4 LT |
3929 | bond_for_each_slave_from(bond, slave, i, start_at) { |
3930 | if (IS_UP(slave->dev) && | |
3931 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3932 | bond_is_active_slave(slave)) { |
1da177e4 | 3933 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
1da177e4 LT |
3934 | break; |
3935 | } | |
3936 | } | |
3937 | ||
1da177e4 LT |
3938 | out: |
3939 | if (res) { | |
3940 | /* no suitable interface, frame not sent */ | |
04502430 | 3941 | kfree_skb(skb); |
1da177e4 | 3942 | } |
0693e88e | 3943 | |
ec634fe3 | 3944 | return NETDEV_TX_OK; |
1da177e4 LT |
3945 | } |
3946 | ||
075897ce | 3947 | |
1da177e4 LT |
3948 | /* |
3949 | * in active-backup mode, we know that bond->curr_active_slave is always valid if | |
3950 | * the bond has a usable interface. | |
3951 | */ | |
3952 | static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev) | |
3953 | { | |
454d7c9b | 3954 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3955 | int res = 1; |
3956 | ||
1da177e4 LT |
3957 | read_lock(&bond->curr_slave_lock); |
3958 | ||
0693e88e MM |
3959 | if (bond->curr_active_slave) |
3960 | res = bond_dev_queue_xmit(bond, skb, | |
3961 | bond->curr_active_slave->dev); | |
1da177e4 | 3962 | |
04502430 ED |
3963 | read_unlock(&bond->curr_slave_lock); |
3964 | ||
3d632c3f | 3965 | if (res) |
1da177e4 | 3966 | /* no suitable interface, frame not sent */ |
04502430 | 3967 | kfree_skb(skb); |
0693e88e | 3968 | |
ec634fe3 | 3969 | return NETDEV_TX_OK; |
1da177e4 LT |
3970 | } |
3971 | ||
3972 | /* | |
169a3e66 JV |
3973 | * In bond_xmit_xor() , we determine the output device by using a pre- |
3974 | * determined xmit_hash_policy(), If the selected device is not enabled, | |
3975 | * find the next active slave. | |
1da177e4 LT |
3976 | */ |
3977 | static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev) | |
3978 | { | |
454d7c9b | 3979 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
3980 | struct slave *slave, *start_at; |
3981 | int slave_no; | |
3982 | int i; | |
3983 | int res = 1; | |
3984 | ||
a361c83c | 3985 | slave_no = bond->xmit_hash_policy(skb, bond->slave_cnt); |
1da177e4 LT |
3986 | |
3987 | bond_for_each_slave(bond, slave, i) { | |
3988 | slave_no--; | |
3d632c3f | 3989 | if (slave_no < 0) |
1da177e4 | 3990 | break; |
1da177e4 LT |
3991 | } |
3992 | ||
3993 | start_at = slave; | |
3994 | ||
3995 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
3996 | if (IS_UP(slave->dev) && | |
3997 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 3998 | bond_is_active_slave(slave)) { |
1da177e4 LT |
3999 | res = bond_dev_queue_xmit(bond, skb, slave->dev); |
4000 | break; | |
4001 | } | |
4002 | } | |
4003 | ||
1da177e4 LT |
4004 | if (res) { |
4005 | /* no suitable interface, frame not sent */ | |
04502430 | 4006 | kfree_skb(skb); |
1da177e4 | 4007 | } |
0693e88e | 4008 | |
ec634fe3 | 4009 | return NETDEV_TX_OK; |
1da177e4 LT |
4010 | } |
4011 | ||
4012 | /* | |
4013 | * in broadcast mode, we send everything to all usable interfaces. | |
4014 | */ | |
4015 | static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev) | |
4016 | { | |
454d7c9b | 4017 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 LT |
4018 | struct slave *slave, *start_at; |
4019 | struct net_device *tx_dev = NULL; | |
4020 | int i; | |
4021 | int res = 1; | |
4022 | ||
1da177e4 LT |
4023 | read_lock(&bond->curr_slave_lock); |
4024 | start_at = bond->curr_active_slave; | |
4025 | read_unlock(&bond->curr_slave_lock); | |
4026 | ||
3d632c3f | 4027 | if (!start_at) |
1da177e4 | 4028 | goto out; |
1da177e4 LT |
4029 | |
4030 | bond_for_each_slave_from(bond, slave, i, start_at) { | |
4031 | if (IS_UP(slave->dev) && | |
4032 | (slave->link == BOND_LINK_UP) && | |
e30bc066 | 4033 | bond_is_active_slave(slave)) { |
1da177e4 LT |
4034 | if (tx_dev) { |
4035 | struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC); | |
4036 | if (!skb2) { | |
a4aee5c8 | 4037 | pr_err("%s: Error: bond_xmit_broadcast(): skb_clone() failed\n", |
4e0952c7 | 4038 | bond_dev->name); |
1da177e4 LT |
4039 | continue; |
4040 | } | |
4041 | ||
4042 | res = bond_dev_queue_xmit(bond, skb2, tx_dev); | |
4043 | if (res) { | |
04502430 | 4044 | kfree_skb(skb2); |
1da177e4 LT |
4045 | continue; |
4046 | } | |
4047 | } | |
4048 | tx_dev = slave->dev; | |
4049 | } | |
4050 | } | |
4051 | ||
3d632c3f | 4052 | if (tx_dev) |
1da177e4 | 4053 | res = bond_dev_queue_xmit(bond, skb, tx_dev); |
1da177e4 LT |
4054 | |
4055 | out: | |
3d632c3f | 4056 | if (res) |
1da177e4 | 4057 | /* no suitable interface, frame not sent */ |
04502430 | 4058 | kfree_skb(skb); |
3d632c3f | 4059 | |
1da177e4 | 4060 | /* frame sent to all suitable interfaces */ |
ec634fe3 | 4061 | return NETDEV_TX_OK; |
1da177e4 LT |
4062 | } |
4063 | ||
4064 | /*------------------------- Device initialization ---------------------------*/ | |
4065 | ||
6f6652be JV |
4066 | static void bond_set_xmit_hash_policy(struct bonding *bond) |
4067 | { | |
4068 | switch (bond->params.xmit_policy) { | |
4069 | case BOND_XMIT_POLICY_LAYER23: | |
4070 | bond->xmit_hash_policy = bond_xmit_hash_policy_l23; | |
4071 | break; | |
4072 | case BOND_XMIT_POLICY_LAYER34: | |
4073 | bond->xmit_hash_policy = bond_xmit_hash_policy_l34; | |
4074 | break; | |
4075 | case BOND_XMIT_POLICY_LAYER2: | |
4076 | default: | |
4077 | bond->xmit_hash_policy = bond_xmit_hash_policy_l2; | |
4078 | break; | |
4079 | } | |
4080 | } | |
4081 | ||
bb1d9123 AG |
4082 | /* |
4083 | * Lookup the slave that corresponds to a qid | |
4084 | */ | |
4085 | static inline int bond_slave_override(struct bonding *bond, | |
4086 | struct sk_buff *skb) | |
4087 | { | |
4088 | int i, res = 1; | |
4089 | struct slave *slave = NULL; | |
4090 | struct slave *check_slave; | |
4091 | ||
0693e88e MM |
4092 | if (!skb->queue_mapping) |
4093 | return 1; | |
bb1d9123 AG |
4094 | |
4095 | /* Find out if any slaves have the same mapping as this skb. */ | |
4096 | bond_for_each_slave(bond, check_slave, i) { | |
4097 | if (check_slave->queue_id == skb->queue_mapping) { | |
4098 | slave = check_slave; | |
4099 | break; | |
4100 | } | |
4101 | } | |
4102 | ||
4103 | /* If the slave isn't UP, use default transmit policy. */ | |
4104 | if (slave && slave->queue_id && IS_UP(slave->dev) && | |
4105 | (slave->link == BOND_LINK_UP)) { | |
4106 | res = bond_dev_queue_xmit(bond, skb, slave->dev); | |
4107 | } | |
4108 | ||
bb1d9123 AG |
4109 | return res; |
4110 | } | |
4111 | ||
374eeb5a | 4112 | |
bb1d9123 AG |
4113 | static u16 bond_select_queue(struct net_device *dev, struct sk_buff *skb) |
4114 | { | |
4115 | /* | |
4116 | * This helper function exists to help dev_pick_tx get the correct | |
fd0e435b | 4117 | * destination queue. Using a helper function skips a call to |
bb1d9123 AG |
4118 | * skb_tx_hash and will put the skbs in the queue we expect on their |
4119 | * way down to the bonding driver. | |
4120 | */ | |
fd0e435b PO |
4121 | u16 txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : 0; |
4122 | ||
374eeb5a NH |
4123 | /* |
4124 | * Save the original txq to restore before passing to the driver | |
4125 | */ | |
df4ab5b3 | 4126 | qdisc_skb_cb(skb)->slave_dev_queue_mapping = skb->queue_mapping; |
374eeb5a | 4127 | |
fd0e435b | 4128 | if (unlikely(txq >= dev->real_num_tx_queues)) { |
d30ee670 | 4129 | do { |
fd0e435b | 4130 | txq -= dev->real_num_tx_queues; |
d30ee670 | 4131 | } while (txq >= dev->real_num_tx_queues); |
fd0e435b PO |
4132 | } |
4133 | return txq; | |
bb1d9123 AG |
4134 | } |
4135 | ||
0693e88e | 4136 | static netdev_tx_t __bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
00829823 | 4137 | { |
bb1d9123 AG |
4138 | struct bonding *bond = netdev_priv(dev); |
4139 | ||
4140 | if (TX_QUEUE_OVERRIDE(bond->params.mode)) { | |
4141 | if (!bond_slave_override(bond, skb)) | |
4142 | return NETDEV_TX_OK; | |
4143 | } | |
00829823 SH |
4144 | |
4145 | switch (bond->params.mode) { | |
4146 | case BOND_MODE_ROUNDROBIN: | |
4147 | return bond_xmit_roundrobin(skb, dev); | |
4148 | case BOND_MODE_ACTIVEBACKUP: | |
4149 | return bond_xmit_activebackup(skb, dev); | |
4150 | case BOND_MODE_XOR: | |
4151 | return bond_xmit_xor(skb, dev); | |
4152 | case BOND_MODE_BROADCAST: | |
4153 | return bond_xmit_broadcast(skb, dev); | |
4154 | case BOND_MODE_8023AD: | |
4155 | return bond_3ad_xmit_xor(skb, dev); | |
4156 | case BOND_MODE_ALB: | |
4157 | case BOND_MODE_TLB: | |
4158 | return bond_alb_xmit(skb, dev); | |
4159 | default: | |
4160 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4161 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4162 | dev->name, bond->params.mode); | |
00829823 | 4163 | WARN_ON_ONCE(1); |
04502430 | 4164 | kfree_skb(skb); |
00829823 SH |
4165 | return NETDEV_TX_OK; |
4166 | } | |
4167 | } | |
4168 | ||
0693e88e MM |
4169 | static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev) |
4170 | { | |
4171 | struct bonding *bond = netdev_priv(dev); | |
4172 | netdev_tx_t ret = NETDEV_TX_OK; | |
4173 | ||
4174 | /* | |
4175 | * If we risk deadlock from transmitting this in the | |
4176 | * netpoll path, tell netpoll to queue the frame for later tx | |
4177 | */ | |
4178 | if (is_netpoll_tx_blocked(dev)) | |
4179 | return NETDEV_TX_BUSY; | |
4180 | ||
4181 | read_lock(&bond->lock); | |
4182 | ||
4183 | if (bond->slave_cnt) | |
4184 | ret = __bond_start_xmit(skb, dev); | |
4185 | else | |
04502430 | 4186 | kfree_skb(skb); |
0693e88e MM |
4187 | |
4188 | read_unlock(&bond->lock); | |
4189 | ||
4190 | return ret; | |
4191 | } | |
00829823 | 4192 | |
1da177e4 LT |
4193 | /* |
4194 | * set bond mode specific net device operations | |
4195 | */ | |
a77b5325 | 4196 | void bond_set_mode_ops(struct bonding *bond, int mode) |
1da177e4 | 4197 | { |
169a3e66 JV |
4198 | struct net_device *bond_dev = bond->dev; |
4199 | ||
1da177e4 LT |
4200 | switch (mode) { |
4201 | case BOND_MODE_ROUNDROBIN: | |
1da177e4 LT |
4202 | break; |
4203 | case BOND_MODE_ACTIVEBACKUP: | |
1da177e4 LT |
4204 | break; |
4205 | case BOND_MODE_XOR: | |
6f6652be | 4206 | bond_set_xmit_hash_policy(bond); |
1da177e4 LT |
4207 | break; |
4208 | case BOND_MODE_BROADCAST: | |
1da177e4 LT |
4209 | break; |
4210 | case BOND_MODE_8023AD: | |
6f6652be | 4211 | bond_set_xmit_hash_policy(bond); |
1da177e4 | 4212 | break; |
1da177e4 | 4213 | case BOND_MODE_ALB: |
8f903c70 JV |
4214 | /* FALLTHRU */ |
4215 | case BOND_MODE_TLB: | |
1da177e4 LT |
4216 | break; |
4217 | default: | |
4218 | /* Should never happen, mode already checked */ | |
a4aee5c8 JP |
4219 | pr_err("%s: Error: Unknown bonding mode %d\n", |
4220 | bond_dev->name, mode); | |
1da177e4 LT |
4221 | break; |
4222 | } | |
4223 | } | |
4224 | ||
217df670 | 4225 | static void bond_ethtool_get_drvinfo(struct net_device *bond_dev, |
7826d43f | 4226 | struct ethtool_drvinfo *drvinfo) |
217df670 | 4227 | { |
7826d43f JP |
4228 | strlcpy(drvinfo->driver, DRV_NAME, sizeof(drvinfo->driver)); |
4229 | strlcpy(drvinfo->version, DRV_VERSION, sizeof(drvinfo->version)); | |
4230 | snprintf(drvinfo->fw_version, sizeof(drvinfo->fw_version), "%d", | |
4231 | BOND_ABI_VERSION); | |
217df670 JV |
4232 | } |
4233 | ||
7282d491 | 4234 | static const struct ethtool_ops bond_ethtool_ops = { |
217df670 | 4235 | .get_drvinfo = bond_ethtool_get_drvinfo, |
fa53ebac | 4236 | .get_link = ethtool_op_get_link, |
8531c5ff AK |
4237 | }; |
4238 | ||
eb7cc59a | 4239 | static const struct net_device_ops bond_netdev_ops = { |
181470fc | 4240 | .ndo_init = bond_init, |
9e71626c | 4241 | .ndo_uninit = bond_uninit, |
eb7cc59a SH |
4242 | .ndo_open = bond_open, |
4243 | .ndo_stop = bond_close, | |
00829823 | 4244 | .ndo_start_xmit = bond_start_xmit, |
bb1d9123 | 4245 | .ndo_select_queue = bond_select_queue, |
be1f3c2c | 4246 | .ndo_get_stats64 = bond_get_stats, |
eb7cc59a | 4247 | .ndo_do_ioctl = bond_do_ioctl, |
d03462b9 | 4248 | .ndo_change_rx_flags = bond_change_rx_flags, |
afc4b13d | 4249 | .ndo_set_rx_mode = bond_set_multicast_list, |
eb7cc59a | 4250 | .ndo_change_mtu = bond_change_mtu, |
cc0e4070 | 4251 | .ndo_set_mac_address = bond_set_mac_address, |
00829823 | 4252 | .ndo_neigh_setup = bond_neigh_setup, |
cc0e4070 | 4253 | .ndo_vlan_rx_add_vid = bond_vlan_rx_add_vid, |
eb7cc59a | 4254 | .ndo_vlan_rx_kill_vid = bond_vlan_rx_kill_vid, |
f6dc31a8 | 4255 | #ifdef CONFIG_NET_POLL_CONTROLLER |
8a8efa22 | 4256 | .ndo_netpoll_setup = bond_netpoll_setup, |
f6dc31a8 WC |
4257 | .ndo_netpoll_cleanup = bond_netpoll_cleanup, |
4258 | .ndo_poll_controller = bond_poll_controller, | |
4259 | #endif | |
9232ecca JP |
4260 | .ndo_add_slave = bond_enslave, |
4261 | .ndo_del_slave = bond_release, | |
b2a103e6 | 4262 | .ndo_fix_features = bond_fix_features, |
eb7cc59a SH |
4263 | }; |
4264 | ||
b3f92b63 DG |
4265 | static const struct device_type bond_type = { |
4266 | .name = "bond", | |
4267 | }; | |
4268 | ||
9e2e61fb AW |
4269 | static void bond_destructor(struct net_device *bond_dev) |
4270 | { | |
4271 | struct bonding *bond = netdev_priv(bond_dev); | |
4272 | if (bond->wq) | |
4273 | destroy_workqueue(bond->wq); | |
4274 | free_netdev(bond_dev); | |
4275 | } | |
4276 | ||
181470fc | 4277 | static void bond_setup(struct net_device *bond_dev) |
1da177e4 | 4278 | { |
454d7c9b | 4279 | struct bonding *bond = netdev_priv(bond_dev); |
1da177e4 | 4280 | |
1da177e4 LT |
4281 | /* initialize rwlocks */ |
4282 | rwlock_init(&bond->lock); | |
4283 | rwlock_init(&bond->curr_slave_lock); | |
4284 | ||
d2991f75 | 4285 | bond->params = bonding_defaults; |
1da177e4 LT |
4286 | |
4287 | /* Initialize pointers */ | |
1da177e4 LT |
4288 | bond->dev = bond_dev; |
4289 | INIT_LIST_HEAD(&bond->vlan_list); | |
4290 | ||
4291 | /* Initialize the device entry points */ | |
181470fc | 4292 | ether_setup(bond_dev); |
eb7cc59a | 4293 | bond_dev->netdev_ops = &bond_netdev_ops; |
8531c5ff | 4294 | bond_dev->ethtool_ops = &bond_ethtool_ops; |
169a3e66 | 4295 | bond_set_mode_ops(bond, bond->params.mode); |
1da177e4 | 4296 | |
9e2e61fb | 4297 | bond_dev->destructor = bond_destructor; |
1da177e4 | 4298 | |
b3f92b63 DG |
4299 | SET_NETDEV_DEVTYPE(bond_dev, &bond_type); |
4300 | ||
1da177e4 LT |
4301 | /* Initialize the device options */ |
4302 | bond_dev->tx_queue_len = 0; | |
4303 | bond_dev->flags |= IFF_MASTER|IFF_MULTICAST; | |
0b680e75 | 4304 | bond_dev->priv_flags |= IFF_BONDING; |
550fd08c | 4305 | bond_dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING); |
181470fc | 4306 | |
1da177e4 LT |
4307 | /* At first, we block adding VLANs. That's the only way to |
4308 | * prevent problems that occur when adding VLANs over an | |
4309 | * empty bond. The block will be removed once non-challenged | |
4310 | * slaves are enslaved. | |
4311 | */ | |
4312 | bond_dev->features |= NETIF_F_VLAN_CHALLENGED; | |
4313 | ||
932ff279 | 4314 | /* don't acquire bond device's netif_tx_lock when |
1da177e4 LT |
4315 | * transmitting */ |
4316 | bond_dev->features |= NETIF_F_LLTX; | |
4317 | ||
4318 | /* By default, we declare the bond to be fully | |
4319 | * VLAN hardware accelerated capable. Special | |
4320 | * care is taken in the various xmit functions | |
4321 | * when there are slaves that are not hw accel | |
4322 | * capable | |
4323 | */ | |
1da177e4 | 4324 | |
b2a103e6 MM |
4325 | bond_dev->hw_features = BOND_VLAN_FEATURES | |
4326 | NETIF_F_HW_VLAN_TX | | |
4327 | NETIF_F_HW_VLAN_RX | | |
4328 | NETIF_F_HW_VLAN_FILTER; | |
4329 | ||
34324dc2 | 4330 | bond_dev->hw_features &= ~(NETIF_F_ALL_CSUM & ~NETIF_F_HW_CSUM); |
b2a103e6 | 4331 | bond_dev->features |= bond_dev->hw_features; |
1da177e4 LT |
4332 | } |
4333 | ||
c67dfb29 EB |
4334 | /* |
4335 | * Destroy a bonding device. | |
4336 | * Must be under rtnl_lock when this function is called. | |
4337 | */ | |
4338 | static void bond_uninit(struct net_device *bond_dev) | |
a434e43f | 4339 | { |
454d7c9b | 4340 | struct bonding *bond = netdev_priv(bond_dev); |
f35188fa | 4341 | struct vlan_entry *vlan, *tmp; |
a434e43f | 4342 | |
f6dc31a8 WC |
4343 | bond_netpoll_cleanup(bond_dev); |
4344 | ||
c67dfb29 | 4345 | /* Release the bonded slaves */ |
0896341a | 4346 | while (bond->first_slave != NULL) |
4347 | __bond_release_one(bond_dev, bond->first_slave->dev, true); | |
4348 | pr_info("%s: released all slaves\n", bond_dev->name); | |
c67dfb29 | 4349 | |
a434e43f JV |
4350 | list_del(&bond->bond_list); |
4351 | ||
f073c7ca TI |
4352 | bond_debug_unregister(bond); |
4353 | ||
22bedad3 | 4354 | __hw_addr_flush(&bond->mc_list); |
f35188fa JV |
4355 | |
4356 | list_for_each_entry_safe(vlan, tmp, &bond->vlan_list, vlan_list) { | |
4357 | list_del(&vlan->vlan_list); | |
4358 | kfree(vlan); | |
4359 | } | |
a434e43f JV |
4360 | } |
4361 | ||
1da177e4 LT |
4362 | /*------------------------- Module initialization ---------------------------*/ |
4363 | ||
4364 | /* | |
4365 | * Convert string input module parms. Accept either the | |
ece95f7f JV |
4366 | * number of the mode or its string name. A bit complicated because |
4367 | * some mode names are substrings of other names, and calls from sysfs | |
4368 | * may have whitespace in the name (trailing newlines, for example). | |
1da177e4 | 4369 | */ |
325dcf7a | 4370 | int bond_parse_parm(const char *buf, const struct bond_parm_tbl *tbl) |
1da177e4 | 4371 | { |
54b87323 | 4372 | int modeint = -1, i, rv; |
a42e534f | 4373 | char *p, modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, }; |
ece95f7f | 4374 | |
a42e534f JV |
4375 | for (p = (char *)buf; *p; p++) |
4376 | if (!(isdigit(*p) || isspace(*p))) | |
4377 | break; | |
4378 | ||
4379 | if (*p) | |
ece95f7f | 4380 | rv = sscanf(buf, "%20s", modestr); |
a42e534f | 4381 | else |
54b87323 | 4382 | rv = sscanf(buf, "%d", &modeint); |
a42e534f JV |
4383 | |
4384 | if (!rv) | |
4385 | return -1; | |
1da177e4 LT |
4386 | |
4387 | for (i = 0; tbl[i].modename; i++) { | |
54b87323 | 4388 | if (modeint == tbl[i].mode) |
ece95f7f JV |
4389 | return tbl[i].mode; |
4390 | if (strcmp(modestr, tbl[i].modename) == 0) | |
1da177e4 | 4391 | return tbl[i].mode; |
1da177e4 LT |
4392 | } |
4393 | ||
4394 | return -1; | |
4395 | } | |
4396 | ||
4397 | static int bond_check_params(struct bond_params *params) | |
4398 | { | |
a549952a | 4399 | int arp_validate_value, fail_over_mac_value, primary_reselect_value; |
f5b2b966 | 4400 | |
1da177e4 LT |
4401 | /* |
4402 | * Convert string parameters. | |
4403 | */ | |
4404 | if (mode) { | |
4405 | bond_mode = bond_parse_parm(mode, bond_mode_tbl); | |
4406 | if (bond_mode == -1) { | |
a4aee5c8 | 4407 | pr_err("Error: Invalid bonding mode \"%s\"\n", |
1da177e4 LT |
4408 | mode == NULL ? "NULL" : mode); |
4409 | return -EINVAL; | |
4410 | } | |
4411 | } | |
4412 | ||
169a3e66 JV |
4413 | if (xmit_hash_policy) { |
4414 | if ((bond_mode != BOND_MODE_XOR) && | |
4415 | (bond_mode != BOND_MODE_8023AD)) { | |
a4aee5c8 | 4416 | pr_info("xmit_hash_policy param is irrelevant in mode %s\n", |
169a3e66 JV |
4417 | bond_mode_name(bond_mode)); |
4418 | } else { | |
4419 | xmit_hashtype = bond_parse_parm(xmit_hash_policy, | |
4420 | xmit_hashtype_tbl); | |
4421 | if (xmit_hashtype == -1) { | |
a4aee5c8 | 4422 | pr_err("Error: Invalid xmit_hash_policy \"%s\"\n", |
3d632c3f | 4423 | xmit_hash_policy == NULL ? "NULL" : |
169a3e66 JV |
4424 | xmit_hash_policy); |
4425 | return -EINVAL; | |
4426 | } | |
4427 | } | |
4428 | } | |
4429 | ||
1da177e4 LT |
4430 | if (lacp_rate) { |
4431 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 JP |
4432 | pr_info("lacp_rate param is irrelevant in mode %s\n", |
4433 | bond_mode_name(bond_mode)); | |
1da177e4 LT |
4434 | } else { |
4435 | lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl); | |
4436 | if (lacp_fast == -1) { | |
a4aee5c8 | 4437 | pr_err("Error: Invalid lacp rate \"%s\"\n", |
1da177e4 LT |
4438 | lacp_rate == NULL ? "NULL" : lacp_rate); |
4439 | return -EINVAL; | |
4440 | } | |
4441 | } | |
4442 | } | |
4443 | ||
fd989c83 JV |
4444 | if (ad_select) { |
4445 | params->ad_select = bond_parse_parm(ad_select, ad_select_tbl); | |
4446 | if (params->ad_select == -1) { | |
a4aee5c8 | 4447 | pr_err("Error: Invalid ad_select \"%s\"\n", |
fd989c83 JV |
4448 | ad_select == NULL ? "NULL" : ad_select); |
4449 | return -EINVAL; | |
4450 | } | |
4451 | ||
4452 | if (bond_mode != BOND_MODE_8023AD) { | |
a4aee5c8 | 4453 | pr_warning("ad_select param only affects 802.3ad mode\n"); |
fd989c83 JV |
4454 | } |
4455 | } else { | |
4456 | params->ad_select = BOND_AD_STABLE; | |
4457 | } | |
4458 | ||
f5841306 | 4459 | if (max_bonds < 0) { |
a4aee5c8 JP |
4460 | pr_warning("Warning: max_bonds (%d) not in range %d-%d, so it was reset to BOND_DEFAULT_MAX_BONDS (%d)\n", |
4461 | max_bonds, 0, INT_MAX, BOND_DEFAULT_MAX_BONDS); | |
1da177e4 LT |
4462 | max_bonds = BOND_DEFAULT_MAX_BONDS; |
4463 | } | |
4464 | ||
4465 | if (miimon < 0) { | |
a4aee5c8 JP |
4466 | pr_warning("Warning: miimon module parameter (%d), not in range 0-%d, so it was reset to %d\n", |
4467 | miimon, INT_MAX, BOND_LINK_MON_INTERV); | |
1da177e4 LT |
4468 | miimon = BOND_LINK_MON_INTERV; |
4469 | } | |
4470 | ||
4471 | if (updelay < 0) { | |
a4aee5c8 JP |
4472 | pr_warning("Warning: updelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4473 | updelay, INT_MAX); | |
1da177e4 LT |
4474 | updelay = 0; |
4475 | } | |
4476 | ||
4477 | if (downdelay < 0) { | |
a4aee5c8 JP |
4478 | pr_warning("Warning: downdelay module parameter (%d), not in range 0-%d, so it was reset to 0\n", |
4479 | downdelay, INT_MAX); | |
1da177e4 LT |
4480 | downdelay = 0; |
4481 | } | |
4482 | ||
4483 | if ((use_carrier != 0) && (use_carrier != 1)) { | |
a4aee5c8 JP |
4484 | pr_warning("Warning: use_carrier module parameter (%d), not of valid value (0/1), so it was set to 1\n", |
4485 | use_carrier); | |
1da177e4 LT |
4486 | use_carrier = 1; |
4487 | } | |
4488 | ||
ad246c99 BH |
4489 | if (num_peer_notif < 0 || num_peer_notif > 255) { |
4490 | pr_warning("Warning: num_grat_arp/num_unsol_na (%d) not in range 0-255 so it was reset to 1\n", | |
4491 | num_peer_notif); | |
4492 | num_peer_notif = 1; | |
4493 | } | |
4494 | ||
1da177e4 LT |
4495 | /* reset values for 802.3ad */ |
4496 | if (bond_mode == BOND_MODE_8023AD) { | |
4497 | if (!miimon) { | |
a4aee5c8 | 4498 | 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 | 4499 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4500 | miimon = 100; |
4501 | } | |
4502 | } | |
4503 | ||
bb1d9123 AG |
4504 | if (tx_queues < 1 || tx_queues > 255) { |
4505 | pr_warning("Warning: tx_queues (%d) should be between " | |
4506 | "1 and 255, resetting to %d\n", | |
4507 | tx_queues, BOND_DEFAULT_TX_QUEUES); | |
4508 | tx_queues = BOND_DEFAULT_TX_QUEUES; | |
4509 | } | |
4510 | ||
ebd8e497 AG |
4511 | if ((all_slaves_active != 0) && (all_slaves_active != 1)) { |
4512 | pr_warning("Warning: all_slaves_active module parameter (%d), " | |
4513 | "not of valid value (0/1), so it was set to " | |
4514 | "0\n", all_slaves_active); | |
4515 | all_slaves_active = 0; | |
4516 | } | |
4517 | ||
c2952c31 FL |
4518 | if (resend_igmp < 0 || resend_igmp > 255) { |
4519 | pr_warning("Warning: resend_igmp (%d) should be between " | |
4520 | "0 and 255, resetting to %d\n", | |
4521 | resend_igmp, BOND_DEFAULT_RESEND_IGMP); | |
4522 | resend_igmp = BOND_DEFAULT_RESEND_IGMP; | |
4523 | } | |
4524 | ||
1da177e4 LT |
4525 | /* reset values for TLB/ALB */ |
4526 | if ((bond_mode == BOND_MODE_TLB) || | |
4527 | (bond_mode == BOND_MODE_ALB)) { | |
4528 | if (!miimon) { | |
a4aee5c8 | 4529 | 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 | 4530 | pr_warning("Forcing miimon to 100msec\n"); |
1da177e4 LT |
4531 | miimon = 100; |
4532 | } | |
4533 | } | |
4534 | ||
4535 | if (bond_mode == BOND_MODE_ALB) { | |
a4aee5c8 JP |
4536 | 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", |
4537 | updelay); | |
1da177e4 LT |
4538 | } |
4539 | ||
4540 | if (!miimon) { | |
4541 | if (updelay || downdelay) { | |
4542 | /* just warn the user the up/down delay will have | |
4543 | * no effect since miimon is zero... | |
4544 | */ | |
a4aee5c8 JP |
4545 | 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", |
4546 | updelay, downdelay); | |
1da177e4 LT |
4547 | } |
4548 | } else { | |
4549 | /* don't allow arp monitoring */ | |
4550 | if (arp_interval) { | |
a4aee5c8 JP |
4551 | pr_warning("Warning: miimon (%d) and arp_interval (%d) can't be used simultaneously, disabling ARP monitoring\n", |
4552 | miimon, arp_interval); | |
1da177e4 LT |
4553 | arp_interval = 0; |
4554 | } | |
4555 | ||
4556 | if ((updelay % miimon) != 0) { | |
a4aee5c8 JP |
4557 | pr_warning("Warning: updelay (%d) is not a multiple of miimon (%d), updelay rounded to %d ms\n", |
4558 | updelay, miimon, | |
4559 | (updelay / miimon) * miimon); | |
1da177e4 LT |
4560 | } |
4561 | ||
4562 | updelay /= miimon; | |
4563 | ||
4564 | if ((downdelay % miimon) != 0) { | |
a4aee5c8 JP |
4565 | pr_warning("Warning: downdelay (%d) is not a multiple of miimon (%d), downdelay rounded to %d ms\n", |
4566 | downdelay, miimon, | |
4567 | (downdelay / miimon) * miimon); | |
1da177e4 LT |
4568 | } |
4569 | ||
4570 | downdelay /= miimon; | |
4571 | } | |
4572 | ||
4573 | if (arp_interval < 0) { | |
a4aee5c8 JP |
4574 | pr_warning("Warning: arp_interval module parameter (%d) , not in range 0-%d, so it was reset to %d\n", |
4575 | arp_interval, INT_MAX, BOND_LINK_ARP_INTERV); | |
1da177e4 LT |
4576 | arp_interval = BOND_LINK_ARP_INTERV; |
4577 | } | |
4578 | ||
4579 | for (arp_ip_count = 0; | |
4580 | (arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count]; | |
4581 | arp_ip_count++) { | |
4582 | /* not complete check, but should be good enough to | |
4583 | catch mistakes */ | |
90fb6250 | 4584 | __be32 ip = in_aton(arp_ip_target[arp_ip_count]); |
4585 | if (!isdigit(arp_ip_target[arp_ip_count][0]) || | |
4586 | ip == 0 || ip == htonl(INADDR_BROADCAST)) { | |
a4aee5c8 JP |
4587 | pr_warning("Warning: bad arp_ip_target module parameter (%s), ARP monitoring will not be performed\n", |
4588 | arp_ip_target[arp_ip_count]); | |
1da177e4 LT |
4589 | arp_interval = 0; |
4590 | } else { | |
1da177e4 LT |
4591 | arp_target[arp_ip_count] = ip; |
4592 | } | |
4593 | } | |
4594 | ||
4595 | if (arp_interval && !arp_ip_count) { | |
4596 | /* don't allow arping if no arp_ip_target given... */ | |
a4aee5c8 JP |
4597 | pr_warning("Warning: arp_interval module parameter (%d) specified without providing an arp_ip_target parameter, arp_interval was reset to 0\n", |
4598 | arp_interval); | |
1da177e4 LT |
4599 | arp_interval = 0; |
4600 | } | |
4601 | ||
f5b2b966 JV |
4602 | if (arp_validate) { |
4603 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) { | |
a4aee5c8 | 4604 | pr_err("arp_validate only supported in active-backup mode\n"); |
f5b2b966 JV |
4605 | return -EINVAL; |
4606 | } | |
4607 | if (!arp_interval) { | |
a4aee5c8 | 4608 | pr_err("arp_validate requires arp_interval\n"); |
f5b2b966 JV |
4609 | return -EINVAL; |
4610 | } | |
4611 | ||
4612 | arp_validate_value = bond_parse_parm(arp_validate, | |
4613 | arp_validate_tbl); | |
4614 | if (arp_validate_value == -1) { | |
a4aee5c8 | 4615 | pr_err("Error: invalid arp_validate \"%s\"\n", |
f5b2b966 JV |
4616 | arp_validate == NULL ? "NULL" : arp_validate); |
4617 | return -EINVAL; | |
4618 | } | |
4619 | } else | |
4620 | arp_validate_value = 0; | |
4621 | ||
1da177e4 | 4622 | if (miimon) { |
a4aee5c8 | 4623 | pr_info("MII link monitoring set to %d ms\n", miimon); |
1da177e4 LT |
4624 | } else if (arp_interval) { |
4625 | int i; | |
4626 | ||
a4aee5c8 JP |
4627 | pr_info("ARP monitoring set to %d ms, validate %s, with %d target(s):", |
4628 | arp_interval, | |
4629 | arp_validate_tbl[arp_validate_value].modename, | |
4630 | arp_ip_count); | |
1da177e4 LT |
4631 | |
4632 | for (i = 0; i < arp_ip_count; i++) | |
e5e2a8fd | 4633 | pr_info(" %s", arp_ip_target[i]); |
1da177e4 | 4634 | |
e5e2a8fd | 4635 | pr_info("\n"); |
1da177e4 | 4636 | |
b8a9787e | 4637 | } else if (max_bonds) { |
1da177e4 LT |
4638 | /* miimon and arp_interval not set, we need one so things |
4639 | * work as expected, see bonding.txt for details | |
4640 | */ | |
b2730f4f | 4641 | pr_debug("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 |
4642 | } |
4643 | ||
4644 | if (primary && !USES_PRIMARY(bond_mode)) { | |
4645 | /* currently, using a primary only makes sense | |
4646 | * in active backup, TLB or ALB modes | |
4647 | */ | |
a4aee5c8 JP |
4648 | pr_warning("Warning: %s primary device specified but has no effect in %s mode\n", |
4649 | primary, bond_mode_name(bond_mode)); | |
1da177e4 LT |
4650 | primary = NULL; |
4651 | } | |
4652 | ||
a549952a JP |
4653 | if (primary && primary_reselect) { |
4654 | primary_reselect_value = bond_parse_parm(primary_reselect, | |
4655 | pri_reselect_tbl); | |
4656 | if (primary_reselect_value == -1) { | |
a4aee5c8 | 4657 | pr_err("Error: Invalid primary_reselect \"%s\"\n", |
a549952a JP |
4658 | primary_reselect == |
4659 | NULL ? "NULL" : primary_reselect); | |
4660 | return -EINVAL; | |
4661 | } | |
4662 | } else { | |
4663 | primary_reselect_value = BOND_PRI_RESELECT_ALWAYS; | |
4664 | } | |
4665 | ||
3915c1e8 JV |
4666 | if (fail_over_mac) { |
4667 | fail_over_mac_value = bond_parse_parm(fail_over_mac, | |
4668 | fail_over_mac_tbl); | |
4669 | if (fail_over_mac_value == -1) { | |
a4aee5c8 | 4670 | pr_err("Error: invalid fail_over_mac \"%s\"\n", |
3915c1e8 JV |
4671 | arp_validate == NULL ? "NULL" : arp_validate); |
4672 | return -EINVAL; | |
4673 | } | |
4674 | ||
4675 | if (bond_mode != BOND_MODE_ACTIVEBACKUP) | |
a4aee5c8 | 4676 | pr_warning("Warning: fail_over_mac only affects active-backup mode.\n"); |
3915c1e8 JV |
4677 | } else { |
4678 | fail_over_mac_value = BOND_FOM_NONE; | |
4679 | } | |
dd957c57 | 4680 | |
1da177e4 LT |
4681 | /* fill params struct with the proper values */ |
4682 | params->mode = bond_mode; | |
169a3e66 | 4683 | params->xmit_policy = xmit_hashtype; |
1da177e4 | 4684 | params->miimon = miimon; |
ad246c99 | 4685 | params->num_peer_notif = num_peer_notif; |
1da177e4 | 4686 | params->arp_interval = arp_interval; |
f5b2b966 | 4687 | params->arp_validate = arp_validate_value; |
1da177e4 LT |
4688 | params->updelay = updelay; |
4689 | params->downdelay = downdelay; | |
4690 | params->use_carrier = use_carrier; | |
4691 | params->lacp_fast = lacp_fast; | |
4692 | params->primary[0] = 0; | |
a549952a | 4693 | params->primary_reselect = primary_reselect_value; |
3915c1e8 | 4694 | params->fail_over_mac = fail_over_mac_value; |
bb1d9123 | 4695 | params->tx_queues = tx_queues; |
ebd8e497 | 4696 | params->all_slaves_active = all_slaves_active; |
c2952c31 | 4697 | params->resend_igmp = resend_igmp; |
655f8919 | 4698 | params->min_links = min_links; |
1da177e4 LT |
4699 | |
4700 | if (primary) { | |
4701 | strncpy(params->primary, primary, IFNAMSIZ); | |
4702 | params->primary[IFNAMSIZ - 1] = 0; | |
4703 | } | |
4704 | ||
4705 | memcpy(params->arp_targets, arp_target, sizeof(arp_target)); | |
4706 | ||
4707 | return 0; | |
4708 | } | |
4709 | ||
0daa2303 | 4710 | static struct lock_class_key bonding_netdev_xmit_lock_key; |
cf508b12 | 4711 | static struct lock_class_key bonding_netdev_addr_lock_key; |
49ee4920 | 4712 | static struct lock_class_key bonding_tx_busylock_key; |
0daa2303 | 4713 | |
e8a0464c DM |
4714 | static void bond_set_lockdep_class_one(struct net_device *dev, |
4715 | struct netdev_queue *txq, | |
4716 | void *_unused) | |
c773e847 DM |
4717 | { |
4718 | lockdep_set_class(&txq->_xmit_lock, | |
4719 | &bonding_netdev_xmit_lock_key); | |
4720 | } | |
4721 | ||
4722 | static void bond_set_lockdep_class(struct net_device *dev) | |
4723 | { | |
cf508b12 DM |
4724 | lockdep_set_class(&dev->addr_list_lock, |
4725 | &bonding_netdev_addr_lock_key); | |
e8a0464c | 4726 | netdev_for_each_tx_queue(dev, bond_set_lockdep_class_one, NULL); |
49ee4920 | 4727 | dev->qdisc_tx_busylock = &bonding_tx_busylock_key; |
c773e847 DM |
4728 | } |
4729 | ||
181470fc SH |
4730 | /* |
4731 | * Called from registration process | |
4732 | */ | |
4733 | static int bond_init(struct net_device *bond_dev) | |
4734 | { | |
4735 | struct bonding *bond = netdev_priv(bond_dev); | |
ec87fd3b | 4736 | struct bond_net *bn = net_generic(dev_net(bond_dev), bond_net_id); |
9fe0617d | 4737 | struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); |
181470fc SH |
4738 | |
4739 | pr_debug("Begin bond_init for %s\n", bond_dev->name); | |
4740 | ||
9fe0617d NH |
4741 | /* |
4742 | * Initialize locks that may be required during | |
4743 | * en/deslave operations. All of the bond_open work | |
4744 | * (of which this is part) should really be moved to | |
4745 | * a phase prior to dev_open | |
4746 | */ | |
4747 | spin_lock_init(&(bond_info->tx_hashtbl_lock)); | |
4748 | spin_lock_init(&(bond_info->rx_hashtbl_lock)); | |
4749 | ||
181470fc SH |
4750 | bond->wq = create_singlethread_workqueue(bond_dev->name); |
4751 | if (!bond->wq) | |
4752 | return -ENOMEM; | |
4753 | ||
4754 | bond_set_lockdep_class(bond_dev); | |
4755 | ||
ec87fd3b | 4756 | list_add_tail(&bond->bond_list, &bn->dev_list); |
181470fc | 4757 | |
6151b3d4 | 4758 | bond_prepare_sysfs_group(bond); |
22bedad3 | 4759 | |
f073c7ca TI |
4760 | bond_debug_register(bond); |
4761 | ||
409cc1f8 JP |
4762 | /* Ensure valid dev_addr */ |
4763 | if (is_zero_ether_addr(bond_dev->dev_addr) && | |
4764 | bond_dev->addr_assign_type == NET_ADDR_PERM) { | |
4765 | eth_hw_addr_random(bond_dev); | |
4766 | bond->dev_addr_from_first = true; | |
4767 | } | |
4768 | ||
22bedad3 | 4769 | __hw_addr_init(&bond->mc_list); |
181470fc SH |
4770 | return 0; |
4771 | } | |
4772 | ||
88ead977 EB |
4773 | static int bond_validate(struct nlattr *tb[], struct nlattr *data[]) |
4774 | { | |
4775 | if (tb[IFLA_ADDRESS]) { | |
4776 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) | |
4777 | return -EINVAL; | |
4778 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) | |
4779 | return -EADDRNOTAVAIL; | |
4780 | } | |
4781 | return 0; | |
4782 | } | |
4783 | ||
d40156aa | 4784 | static unsigned int bond_get_num_tx_queues(void) |
d5da4510 | 4785 | { |
efacb309 | 4786 | return tx_queues; |
d5da4510 JP |
4787 | } |
4788 | ||
88ead977 | 4789 | static struct rtnl_link_ops bond_link_ops __read_mostly = { |
d40156aa JP |
4790 | .kind = "bond", |
4791 | .priv_size = sizeof(struct bonding), | |
4792 | .setup = bond_setup, | |
4793 | .validate = bond_validate, | |
4794 | .get_num_tx_queues = bond_get_num_tx_queues, | |
4795 | .get_num_rx_queues = bond_get_num_tx_queues, /* Use the same number | |
4796 | as for TX queues */ | |
88ead977 EB |
4797 | }; |
4798 | ||
dfe60397 | 4799 | /* Create a new bond based on the specified name and bonding parameters. |
e4b91c48 | 4800 | * If name is NULL, obtain a suitable "bond%d" name for us. |
dfe60397 MW |
4801 | * Caller must NOT hold rtnl_lock; we need to release it here before we |
4802 | * set up our sysfs entries. | |
4803 | */ | |
ec87fd3b | 4804 | int bond_create(struct net *net, const char *name) |
dfe60397 MW |
4805 | { |
4806 | struct net_device *bond_dev; | |
4807 | int res; | |
4808 | ||
4809 | rtnl_lock(); | |
027ea041 | 4810 | |
1c5cae81 JP |
4811 | bond_dev = alloc_netdev_mq(sizeof(struct bonding), |
4812 | name ? name : "bond%d", | |
4813 | bond_setup, tx_queues); | |
dfe60397 | 4814 | if (!bond_dev) { |
a4aee5c8 | 4815 | pr_err("%s: eek! can't alloc netdev!\n", name); |
9e2e61fb AW |
4816 | rtnl_unlock(); |
4817 | return -ENOMEM; | |
dfe60397 MW |
4818 | } |
4819 | ||
ec87fd3b | 4820 | dev_net_set(bond_dev, net); |
88ead977 EB |
4821 | bond_dev->rtnl_link_ops = &bond_link_ops; |
4822 | ||
dfe60397 | 4823 | res = register_netdevice(bond_dev); |
0daa2303 | 4824 | |
e826eafa PO |
4825 | netif_carrier_off(bond_dev); |
4826 | ||
7e083840 | 4827 | rtnl_unlock(); |
9e2e61fb AW |
4828 | if (res < 0) |
4829 | bond_destructor(bond_dev); | |
30c15ba9 | 4830 | return res; |
dfe60397 MW |
4831 | } |
4832 | ||
2c8c1e72 | 4833 | static int __net_init bond_net_init(struct net *net) |
ec87fd3b | 4834 | { |
15449745 | 4835 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b EB |
4836 | |
4837 | bn->net = net; | |
4838 | INIT_LIST_HEAD(&bn->dev_list); | |
4839 | ||
ec87fd3b | 4840 | bond_create_proc_dir(bn); |
4c22400a | 4841 | bond_create_sysfs(bn); |
15449745 EB |
4842 | |
4843 | return 0; | |
ec87fd3b EB |
4844 | } |
4845 | ||
2c8c1e72 | 4846 | static void __net_exit bond_net_exit(struct net *net) |
ec87fd3b | 4847 | { |
15449745 | 4848 | struct bond_net *bn = net_generic(net, bond_net_id); |
ec87fd3b | 4849 | |
4c22400a | 4850 | bond_destroy_sysfs(bn); |
ec87fd3b | 4851 | bond_destroy_proc_dir(bn); |
ec87fd3b EB |
4852 | } |
4853 | ||
4854 | static struct pernet_operations bond_net_ops = { | |
4855 | .init = bond_net_init, | |
4856 | .exit = bond_net_exit, | |
15449745 EB |
4857 | .id = &bond_net_id, |
4858 | .size = sizeof(struct bond_net), | |
ec87fd3b EB |
4859 | }; |
4860 | ||
1da177e4 LT |
4861 | static int __init bonding_init(void) |
4862 | { | |
1da177e4 LT |
4863 | int i; |
4864 | int res; | |
4865 | ||
bd33acc3 | 4866 | pr_info("%s", bond_version); |
1da177e4 | 4867 | |
dfe60397 | 4868 | res = bond_check_params(&bonding_defaults); |
3d632c3f | 4869 | if (res) |
dfe60397 | 4870 | goto out; |
1da177e4 | 4871 | |
15449745 | 4872 | res = register_pernet_subsys(&bond_net_ops); |
ec87fd3b EB |
4873 | if (res) |
4874 | goto out; | |
027ea041 | 4875 | |
88ead977 EB |
4876 | res = rtnl_link_register(&bond_link_ops); |
4877 | if (res) | |
6639104b | 4878 | goto err_link; |
88ead977 | 4879 | |
f073c7ca TI |
4880 | bond_create_debugfs(); |
4881 | ||
1da177e4 | 4882 | for (i = 0; i < max_bonds; i++) { |
ec87fd3b | 4883 | res = bond_create(&init_net, NULL); |
dfe60397 MW |
4884 | if (res) |
4885 | goto err; | |
1da177e4 LT |
4886 | } |
4887 | ||
1da177e4 | 4888 | register_netdevice_notifier(&bond_netdev_notifier); |
dfe60397 | 4889 | out: |
1da177e4 | 4890 | return res; |
88ead977 EB |
4891 | err: |
4892 | rtnl_link_unregister(&bond_link_ops); | |
6639104b | 4893 | err_link: |
15449745 | 4894 | unregister_pernet_subsys(&bond_net_ops); |
88ead977 | 4895 | goto out; |
dfe60397 | 4896 | |
1da177e4 LT |
4897 | } |
4898 | ||
4899 | static void __exit bonding_exit(void) | |
4900 | { | |
4901 | unregister_netdevice_notifier(&bond_netdev_notifier); | |
4902 | ||
f073c7ca | 4903 | bond_destroy_debugfs(); |
ae68c398 | 4904 | |
15449745 | 4905 | unregister_pernet_subsys(&bond_net_ops); |
4de79c73 | 4906 | rtnl_link_unregister(&bond_link_ops); |
e843fa50 NH |
4907 | |
4908 | #ifdef CONFIG_NET_POLL_CONTROLLER | |
fb4fa76a NH |
4909 | /* |
4910 | * Make sure we don't have an imbalance on our netpoll blocking | |
4911 | */ | |
4912 | WARN_ON(atomic_read(&netpoll_block_tx)); | |
e843fa50 | 4913 | #endif |
1da177e4 LT |
4914 | } |
4915 | ||
4916 | module_init(bonding_init); | |
4917 | module_exit(bonding_exit); | |
4918 | MODULE_LICENSE("GPL"); | |
4919 | MODULE_VERSION(DRV_VERSION); | |
4920 | MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION); | |
4921 | MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others"); | |
88ead977 | 4922 | MODULE_ALIAS_RTNL_LINK("bond"); |